(Reviews in Mineralogy and Geochemistry 69) Keith Putirka, Frank Tepley - Minerals, Inclusions, and Volcanic Processess (2008, Mineralogical Society of America)
Uploaded by
Akbar Susanto(Reviews in Mineralogy and Geochemistry 69) Keith Putirka, Frank Tepley - Minerals, Inclusions, and Volcanic Processess (2008, Mineralogical Society of America)
Uploaded by
Akbar SusantoMinerals, Inclusions and
Volcanic Processes
69 Reviews in Mineralogy and Geochemistry 69
FrOM the SerIeS edItOr
The chapters in this volume represent an extensive review of the material
presented by the invited speakers at a two-day short course on Minerals,
Inclusions and Volcanic Processes held prior to the annual fall American
Geophysical Union meeting in San Francisco, California (December 13-14,
2008). This meeting was sponsored by the Mineralogical Society of America and
the Department of Energy. The short course was also held in conjunction with an
AGU-sponsored topical session at the subsequent annual fall AGU meeting.
Any supplemental material and errata (if any) can be found at the MSA
website www.minsocam.org.
Jodi J. Rosso, Series Editor
West Richland, Washington
October 2008
ShOrt COurSe SerIeS dedICatIOn
Dr. William C. Luth has had a long and distinguished career in research,
education and in the government. He was a leader in experimental petrology and
in training graduate students at Stanford University. His efforts at Sandia National
Laboratory and at the Department of Energy’s headquarters resulted in the initiation
and long-term support of many of the cutting edge research projects whose results
form the foundations of these short courses. Bill’s broad interest in understanding
fundamental geochemical processes and their applications to national problems
is a continuous thread through both his university and government career. He
retired in 1996, but his efforts to foster excellent basic research, and to promote the
development of advanced analytical capabilities gave a unique focus to the basic
research portfolio in Geosciences at the Department of Energy. He has been, and
continues to be, a friend and mentor to many of us. It is appropriate to celebrate his
career in education and government service with this series of courses.
1529-6466/08/0069-0000$05.00 DOI: 10.2138/rmg.2008.69.0
Minerals, Inclusions and
Volcanic Processes
69 Reviews in Mineralogy and Geochemistry 69
TABLE OF CONTENTS
1 Introduction to Minerals, Inclusions andVolcanic Processes
Keith D. Putirka
OVERVIEW OF THE VOLUME .............................................................................................1
SOME IMMEDIATE AND SOME MORE DISTANT DIRECTIONS ...................................4
ACKNOWLEDGMENTS.........................................................................................................5
REFERENCES CITED.............................................................................................................5
2 experimental Studies of the Kinetics and
energetics of Magma Crystallization
Julia E. Hammer
INTRODUCTION ....................................................................................................................9
PRE-ERUPTIVE MAGMATIC ENVIRONMENTS ..............................................................10
General approach .........................................................................................................10
Apparatus ....................................................................................................................10
Evaluating experimental results ..................................................................................11
Phase equilibria in volcano-petrology .........................................................................12
CRYSTALLIzATION KINETICS IN SIMPLE SYSTEMS ..................................................12
Fundamental concepts of crystal nucleation................................................................14
Theory development through experimentation............................................................14
Crystal growth ............................................................................................................22
CRYSTALLIzATION KINETICS IN MAGMAS: METHODS ............................................23
Starting materials ........................................................................................................24
Thermal history ..........................................................................................................24
Static vs. dynamic .......................................................................................................25
Constant-rate vs. variable-rate cooling .......................................................................25
BASALT COOLING ExPERIMENTS ..................................................................................26
Nucleation controls texture..........................................................................................26
Growth rates decay with time ......................................................................................27
Undercooling and the approach to equilibrium ...........................................................29
Crystallization in dynamic environments ...................................................................30
v
Minerals, Inclusions & Volcanic Processes ‒ Table of Contents
FELSIC MAGMA DECOMPRESSION ExPERIMENTS ....................................................33
Effective undercooling.................................................................................................33
Decompression history ................................................................................................34
TExTURAL qUANTIFICATION OF ExPERIMENTAL RUN PRODUCTS .....................37
Crystal size distribution analysis ................................................................................38
Crystal aspect ratio .....................................................................................................41
PHASE BOUNDARIES AND INTERFACES .......................................................................44
Heterogeneous nucleation ..........................................................................................44
Solid-solid interfacial energy.......................................................................................44
Reaction relationships .................................................................................................47
MELT INCLUSION ENTRAPMENT....................................................................................47
Major element behavior ..............................................................................................47
Trace element behavior................................................................................................49
CONCLUSIONS.....................................................................................................................51
ACKNOWLEDGMENTS.......................................................................................................52
REFERENCES CITED...........................................................................................................52
3 thermometers and Barometers for Volcanic Systems
Keith D. Putirka
INTRODUCTION ..................................................................................................................61
THEORETICAL ASPECTS ...................................................................................................61
REGRESSION ANALYSIS & STATISTICAL CONSIDERATIONS ...................................63
Experimental data, regression analysis & statistical considerations ..........................63
IGNEOUS THERMOMETERS AND BAROMETERS ........................................................67
History and background ..............................................................................................67
Calculating “liquid” and mineral components, and activities .....................................68
“Glass” (or liquid) thermometers ................................................................................71
Olivine-liquid equilibria ..............................................................................................73
Plagioclase- and alkali feldspar-liquid thermobarometers ..........................................79
Two-feldspar thermometers .........................................................................................83
Orthopyroxene and orthopyroxene-liquid thermobarometers .....................................84
Clinopyroxene and clinopyroxene-liquid thermobarometers ......................................88
Two-pyroxene thermometers and barometers .............................................................94
The silica activity barometer .......................................................................................96
APPLICATIONS.....................................................................................................................98
Magma transport at Hawaii .........................................................................................98
The conditions of mantle partial melting.....................................................................98
Temperature estimates in felsic volcanic systems .....................................................101
SOURCES OF ERROR RELATIVE TO P AND T ..............................................................104
Model error ................................................................................................................104
Error of treatment ......................................................................................................107
CONCLUSIONS...................................................................................................................110
ACKNOWLEDGMENTS.....................................................................................................110
REFERENCES .....................................................................................................................111
vi
Minerals, Inclusions, & Volcanic Processes ‒ Table of Contents
4 thermometers and thermobarometers in Granitic Systems
J. Lawford Anderson, Andrew P. Barth,
Joseph L. Wooden, Frank Mazdab
INTRODUCTION ................................................................................................................121
Plutonic barometry of Cordilleran orogenic terranes - a historical perspective .......122
SINGLE MINERAL TRACE ELEMENT THERMOMETERS
IN IGNEOUS SYSTEMS ...............................................................................................122
Ti-in-quartz thermometer ..........................................................................................123
Ti-in-zircon and zr-in-rutile thermometers ...............................................................124
zr-in-sphene thermometer .........................................................................................128
ACCESSORY PHASE SATURATION THERMOMETRY.................................................129
THERMOBAROMETERS APPLICABLE TO GRANITIC ROCKS..................................130
Al-in-hornblende barometry ......................................................................................130
Thermobarometry for peraluminous granites ............................................................132
Oxybarometers and other mineral reactions involving oxygen .................................133
Fe-Ti oxides, pyroxenes, olivine, and quartz (qUILF) and sphene (titanite) ............134
Magnetite and ilmenite series of granitic plutons and oxygen fugacity ....................135
SUMMARY ..........................................................................................................................137
ACKNOWLEDGMENT .......................................................................................................138
REFERENCES .....................................................................................................................138
5 Fluid Inclusion thermobarometry
as a tracer for Magmatic Processes
Thor H. Hansteen, Andreas Klügel
INTRODUCTION ................................................................................................................143
OCCURRENCE AND PETROGRAPHY OF FLUID INCLUSIONS.................................144
INCLUSION CLASSIFICATION ........................................................................................144
FLUID INCLUSION FORMATION AND RE-EqUILIBRATION ....................................147
Trapping during crystal growth .................................................................................147
Formation of secondary inclusions ............................................................................147
Post-entrapment re-equilibration ...............................................................................147
MICROTHERMOMETRY ...................................................................................................149
FLUID SYSTEMS ................................................................................................................150
The CO2 system .........................................................................................................150
The H2O-NaCl system ...............................................................................................152
TRACKING VOLCANIC PLUMBING SYSTEMS USING CO2 INCLUSIONS ..............156
Rationale ....................................................................................................................158
Determining inclusion density, pressure and depth ...................................................160
The role of possibly missing H2O..............................................................................165
Volumetric re-equilibration during magma ascent: a case study ...............................166
vii
Minerals, Inclusions & Volcanic Processes ‒ Table of Contents
PRESSURE INFORMATION FROM AqUEOUS FLUIDS IN SHALLOW MAGMA
CHAMBERS AND PLUTONIC ROCKS ......................................................................168
Background................................................................................................................168
Fluids in shallow magma chambers ..........................................................................168
Fluid inclusions in plutonic rocks .............................................................................169
SUMMARY ..........................................................................................................................171
OUTLOOK ...........................................................................................................................171
ACKNOWLEDGMENTS.....................................................................................................172
REFERENCES .....................................................................................................................172
6 Petrologic reconstruction of Magmatic
System Variables and Processes
Jon Blundy, Kathy Cashman
INTRODUCTION ................................................................................................................179
MAGMATIC VARIABLES ..................................................................................................180
Petrologically-determined intensive variables...........................................................181
Linking intensive variables to the monitoring record ................................................181
METHODOLOGY ...............................................................................................................181
Scanning electron microscopy (SEM) .......................................................................181
Electron microprobe analysis (EMPA) ......................................................................184
Secondary ion mass spectrometry (SIMS) ................................................................186
Analysis of textures ..................................................................................................188
GEOTHERMOMETRY .......................................................................................................193
Iron-titanium oxide thermometry and oxybarometry ................................................194
Amphibole-Plagioclase thermometry ........................................................................197
Two-pyroxene thermometry ......................................................................................199
GEOBAROMETRY ..............................................................................................................200
Experimental reconstruction......................................................................................201
Volatile saturation pressures ......................................................................................203
Haplogranite projection .............................................................................................207
Crystal textures ..........................................................................................................209
COMPARISON OF PETROLOGICAL DATA WITH MONITORING SIGNALS.............214
Gas chemistry ............................................................................................................214
MAGMA ASCENT AND CRYSTALLIzATION ................................................................220
Phenocryst abundance ...............................................................................................220
Melt inclusion compositions .....................................................................................221
Modeling volatile elements ......................................................................................222
Combining degassing and crystallization ..................................................................224
Interpreting magmatic systems .................................................................................225
SUMMARY ..........................................................................................................................230
ACKNOWLEDGMENTS.....................................................................................................231
REFERENCES .....................................................................................................................231
viii
Minerals, Inclusions, & Volcanic Processes ‒ Table of Contents
7 Magma ascent rates
Malcolm J. Rutherford
INTRODUCTION ...............................................................................................................241
MAGMA SOURCE AND MAGMA RESERVOIRS ..........................................................242
BUBBLE FORMATION AND EVOLUTION DURING MAGMA ASCENT....................243
GROUNDMASS CRYSTALLIzATION DURING MAGMA ASCENT ............................246
PHENOCRYST-MELT REACTIONS DURING MAGMA ASCENT ................................250
Reactions between volatile-bearing minerals and melt during magma ascent ..........250
Phenocryst-melt reactions driven by pre-eruption magma mixing ...........................255
MAGMA ASCENT RATES FROM MASS ERUPTION RATES .......................................257
MAGMA ASCENT FROM SEISMIC DATA.......................................................................258
MAGMA ASCENT RATES BASED ON xENOLITH-MELT REACTIONS ....................260
xENOLITH TRANSPORT, FLOW MODELING AND MAGMA ASCENT RATES .......264
SUMMARY OF MAGMA ASCENT RATE ESTIMATES..................................................265
REFERENCES ....................................................................................................................267
8 Melt Inclusions in Basaltic and related Volcanic rocks
Adam J.R. Kent
INTRODUCTION ................................................................................................................273
Previous reviews of melt inclusions in basaltic and related rocks ............................276
TECHNIqUES USED FOR STUDY OF MELT INCLUSIONS .........................................276
Petrographic examination of melt inclusions ............................................................276
Preparation of melt inclusions for analysis ...............................................................278
Chemical analysis of melt inclusions ........................................................................281
FORMATION AND EVOLUTION OF MELT INCLUSIONS ...........................................292
Melt inclusion host minerals .....................................................................................292
Formation of melt inclusions .....................................................................................292
Evolution of melt inclusions after trapping ...............................................................299
Re-equilibration with host and or external melt ........................................................302
ORIGIN OF COMPOSITIONAL VARIATIONS IN MELT INCLUSIONS ......................311
How representative are trapped melts? ......................................................................316
MELT INCLUSIONS IN BASALTIC AND RELATED ROCKS........................................318
Mid-ocean ridges .......................................................................................................318
Subduction zones .......................................................................................................319
Oceanic islands ..........................................................................................................320
Continental flood basalts ...........................................................................................320
CONCLUDING REMARKS ................................................................................................321
ACKNOWLEDGMENTS.....................................................................................................322
REFERENCES .....................................................................................................................322
ix
Minerals, Inclusions & Volcanic Processes ‒ Table of Contents
9 Interpreting h2O and CO2 Contents in Melt Inclusions:
Constraints from Solubility experiments and Modeling
Gordon Moore
INTRODUCTION ................................................................................................................333
ExPERIMENTAL VOLATILE SOLUBILITY STUDIES IN NATURAL MELTS ............335
The “good” solubility experiment .............................................................................335
The solubility of pure H2O and CO2 in natural melts ................................................337
The solubility of mixed fluids (H2O + CO2) in natural melts ....................................343
VOLATILE SOLUBILITY MODELS FOR NATURAL SILICATE MELTS .....................345
H2O solubility models ...............................................................................................346
CO2 solubility models................................................................................................349
Mixed (H2O + CO2) solubility models ......................................................................350
The compositional variation in melt inclusions and its implications for
interpreting volatile contents ...........................................................................352
ACKNOWLEDGMENTS.....................................................................................................355
REFERENCES .....................................................................................................................358
10 Volatile abundances in Basaltic Magmas and
their degassing Paths tracked by Melt Inclusions
Nicole Métrich, Paul J. Wallace
INTRODUCTION ................................................................................................................363
WHAT ARE MELT INCLUSIONS AND HOW DO THEY FORM?..................................364
POST-ENTRAPMENT MODIFICATION OF MELT INCLUSIONS.................................369
ANALYTICAL TECHNIqUES FOR VOLATILES ............................................................372
VOLATILE SOLUBILITIES AND VAPOR MELT PARTITIONING .................................373
H2O AND CO2 ABUNDANCES IN BASALTIC MAGMAS: CRYSTALLIzATION
DEPTHS, MAGMA ASCENT AND DEGASSING PROCESSES ................................377
H2O-CO2 abundances in basaltic magmas .................................................................377
Pressures of crystallization and magma degassing ...................................................379
Sulfur and halogen degassing ...................................................................................384
ERUPTION STYLES AND PROCESSES – HOW DO THEY RELATE TO MELT
INCLUSION DATA? ......................................................................................................385
VOLCANIC DEGASSING: INTEGRATION OF VOLCANIC GAS COMPOSITIONS
AND GAS FLUxES WITH MELT INCLUSIONS .......................................................387
Syn-eruptive magma degassing and volatile degassing budgets ..............................387
Excessive degassing and the role of degassed but unerupted magma ......................388
Differential gas bubble transfer ................................................................................388
GEOCHEMICAL SYSTEMATICS OF THE PRIMARY VOLATILE CONTENTS
OF MAFIC MAGMAS ...................................................................................................388
CONCLUDING REMARKS ................................................................................................393
ACKNOWLEGMENTS........................................................................................................396
REFERENCES .....................................................................................................................396
x
Minerals, Inclusions, & Volcanic Processes ‒ Table of Contents
11 Inter- and Intracrystalline Isotopic disequilibria:
techniques and applications
Frank C. Ramos, Frank J. Tepley III
INTRODUCTION ................................................................................................................403
MICROSPATIAL ISOTOPE ANALYSIS .............................................................................404
Micromilling ..............................................................................................................406
Secondary ionization mass spectrometry (SIMS) .....................................................417
Laser ablation sampling.............................................................................................418
MICROANALYTE ANALYSIS ...........................................................................................432
Single crystal analyses...............................................................................................432
SUMMARY ..........................................................................................................................439
ACKNOWLEDGMENTS.....................................................................................................439
REFERENCES .....................................................................................................................440
12 Oxygen Isotopes in Mantle and Crustal Magmas
as revealed by Single Crystal analysis
Ilya Bindeman
PART I: INTRODUCTION ..................................................................................................445
Basics of oxygen isotope variations in nature and their causes.................................446
Oxygen isotopes in mantle-derived rocks, normal-δ18O, high-δ18O, and
low-δ18O magmas ............................................................................................451
PART II: SINGLE PHENOCRYST ISOTOPE STUDIES ..................................................453
Modern methods of oxygen isotope analysis of crystals ...........................................453
δ18O heterogeneity and ∆18Ocrystal-melt disequilibria ......................................................455
PART III: CASE STUDIES OF OxYGEN ISOTOPE DISEqUILIBRIA IN IGNEOUS
ROCKS BASED ON SINGLE CRYSTAL ISOTOPE ANALYSIS ................................457
Basaltic igneous systems: olivine in basalts ..............................................................458
Silicic igneous systems ..............................................................................................463
Case studies of large silicic magma systems .............................................................464
Phenocryst heterogeneity in plutonic rocks...............................................................467
PART IV: ISOTOPE DISEqUILIBRIA: WHAT HAVE WE LEARNED ABOUT
MAGMA GENESIS? ......................................................................................................469
Mineral-diffusive timescales and magma residence times ........................................469
Rapid magma genesis and high magma production rates .........................................469
Very shallow petrogenesis and the role of hydrothermal carapaces ..........................470
Accretion of large silicic magma bodies ..................................................................471
The origin of “phenocrysts” and interpretation of their melt inclusions ..................472
ACKNOWLEDGMENTS.....................................................................................................472
REFERENCES CITED.........................................................................................................473
xi
Minerals, Inclusions & Volcanic Processes ‒ Table of Contents
13 uranium-series Crystal ages
Kari M. Cooper, Mary R. Reid
INTRODUCTION ................................................................................................................479
Principles of U-series dating .....................................................................................479
U-series dating of crystals .........................................................................................482
ANALYTICAL AND PRACTICAL CONSIDERATIONS..................................................482
Introduction ...............................................................................................................482
Practical limitations: sample preparation and analysis..............................................483
Age calculations ........................................................................................................490
Effect of diffusion on crystal ages .............................................................................497
WHAT DO CRYSTAL AGES TELL US? ............................................................................499
Introduction ...............................................................................................................499
Using U-Pb and decay series ages to date eruptions .................................................500
Pre-eruptive crystal growth—‘simple’ scenarios ......................................................501
Crystal recycling—autocrysts vs. xenocrysts. vs. antecrysts ....................................503
Insights from dating of subvolcanic plutons..............................................................516
RESOLVING AGES FROM COMPLEx CRYSTAL POPULATIONS:
INCORPORATING OTHER OBSERVATIONS ............................................................519
Introduction ...............................................................................................................519
Combining U-Th and U-Pb SIMS dating ..................................................................519
U-Th/U-Pb spots versus mineral separates................................................................520
Multiple size fractions ...............................................................................................523
Textural perspectives .................................................................................................524
Crystal-scale major element/trace element variations ...............................................527
Crystal ages and thermometry ...................................................................................529
Single-crystal or sub-crystal oxygen isotopic data ....................................................532
Summary ...................................................................................................................534
KEY qUESTIONS AND FUTURE DIRECTIONS ............................................................535
What can crystal ages tell us about magmatic processes?.........................................535
What do we get from the different scales of analyses? .............................................536
Future directions ........................................................................................................537
Final thoughts ............................................................................................................538
ACKNOWLEDGMENTS.....................................................................................................538
REFERENCES .....................................................................................................................538
14 time Scales of Magmatic Processes from
Modeling the Zoning Patterns of Crystals
Fidel Costa, Ralf Dohmen, Sumit Chakraborty
INTRODUCTION ................................................................................................................545
EqUATIONS TO DESCRIBE DIFFUSION PROCESSES ................................................548
Diffusion, flux and constitutive laws .........................................................................548
Time dependence: the continuity equation ................................................................550
MODELING NATURAL CRYSTALS .................................................................................552
Initial and boundary conditions .................................................................................553
xii
Minerals, Inclusions, & Volcanic Processes ‒ Table of Contents
Non-isothermal diffusion...........................................................................................557
Variable boundary conditions – a numerical simulation using MELTS ....................558
Diffusion and crystal growth – the moving boundary problem.................................563
DIFFUSION COEFFICIENTS .............................................................................................565
Types of diffusion coefficients...................................................................................567
Multi-component diffusion in silicates ......................................................................567
Parameters that determine tracer diffusion coefficients.............................................570
DIFFUSION MODELING – POTENTIAL PITFALLS AND ERRORS .............................574
Multiple dimensions, sectioning effects and anisotropy ...........................................574
Uncertainties in the calculated times from the diffusion models ..............................576
A CHECKLIST FOR OBTAINING TIME SCALES OF MAGMATIC PROCESSES
BY DIFFUSION MODELING .......................................................................................577
TIME SCALES OF MAGMATIC PROCESSES THAT HAVE BEEN OBTAINED
FROM MODELING THE zONING OF CRYSTALS AND THEIR RELATION
TO OTHER INFORMATION ON TIME SCALES OF PROCESSES ..........................578
Time scales from diffusion modeling .......................................................................579
SUMMARY AND PROSPECTS ..........................................................................................581
ACKNOWLEDGMENTS.....................................................................................................582
REFERENCES .....................................................................................................................582
APPENDIx I ........................................................................................................................588
APPENDIx II .......................................................................................................................591
15 Mineral textures and Zoning as
evidence for Open System Processes
Martin J. Streck
INTRODUCTION ................................................................................................................595
METHODS TO IMAGE TExTURES AND qUANTIFY COMPOSITION .......................595
BRIEF REVIEW OF OPEN SYSTEM PROCESSES .........................................................596
Mixing .......................................................................................................................597
Contamination (assimilation) ....................................................................................597
MINERAL RECORD FOR MIxING OF MAGMAS AND COUNTRY ROCK –
PHENOCYRST, ANTECRYST, xENOCRYST ............................................................598
SUMMARY OF zONING AND TExTURES .....................................................................599
Crystal zoning............................................................................................................599
Crystal textures ..........................................................................................................601
MINERAL SPECIFIC TExTURES AND zONING ...........................................................604
Plagioclase .................................................................................................................604
Pyroxene ....................................................................................................................607
Olivine .......................................................................................................................611
Amphibole .................................................................................................................611
Other major minerals .................................................................................................613
Accessory minerals....................................................................................................613
MINERAL POPULATIONS AS EVIDENCE FOR OPEN SYSTEM BEHAVIOR ...........615
OUTLOOK AND CONCLUSIONS .....................................................................................618
ACKNOWLEDGMENTS.....................................................................................................618
REFERENCES .....................................................................................................................619
xiii
Minerals, Inclusions & Volcanic Processes ‒ Table of Contents
16 decryption of Igneous rock textures:
Crystal Size distribution tools
Pietro Armienti
CRYSTAL SIzE DISTRIBUTIONS ....................................................................................623
DETERMINATION OF THE CURVES OF POPULATION DENSITY .............................624
Analysis of distributions of areas ..............................................................................625
Stereologic reconstruction of CSD: monodispersed distributions.............................625
Stereologic reconstruction of CSD: poly-dispersed distributions .............................627
The “best” CSD for a set of unfolded planar data .....................................................631
Truncation effects ......................................................................................................633
CSDs AND THE PETROGENETIC CONTExT .................................................................635
Continuously taped and refilled magma chamber: the Stromboli example ...............636
Cooling of a magma batch: the case of Mount Etna .................................................639
Estimate of mineral growth rates for Mount Etna .....................................................643
TEMPORAL EVOLUTION OF CSDs: SIzE AS FUNCTION OF TIME ..........................644
CONCLUDING REMARKS ................................................................................................648
REFERENCES .....................................................................................................................648
17 deciphering Magma Chamber dynamics from
Styles of Compositional Zoning in Large Silicic ash Flow Sheets
Olivier Bachmann, George W. Bergantz
INTRODUCTION ................................................................................................................651
Types of gradients in ignimbrites ..............................................................................652
Abrupt gradients ........................................................................................................653
Linear gradients .........................................................................................................653
Lack of gradients .......................................................................................................655
MECHANISMS FOR GENERATING GRADIENTS IN SILICIC
MAGMA CHAMBERS ..................................................................................................656
The role of convection ...............................................................................................656
Origin of heterogeneity in silicic magmas: convective stirring vs.
phase separation ..............................................................................................661
Mechanisms to generate abrupt gradients .................................................................662
THE CASE OF THE ABRUPTLY zONED CRATER LAKE ERUPTION ........................662
Mechanism to generate linear gradients ....................................................................663
THE CASE OF THE LINEARLY zONED BISHOP TUFF................................................664
The generation of homogeneity .................................................................................665
THE CASE OF THE FISH CANYON MAGMA BODY ....................................................666
THE CASE OF THE HOMOGENEOUS GRANITOIDS ...................................................667
CONCLUSIONS...................................................................................................................667
ACKNOWLEDGMENTS.....................................................................................................668
REFERENCES .....................................................................................................................668
xiv
Reviews in Mineralogy & Geochemistry
Vol. 69 pp. 1-8, 2008 1
Copyright © Mineralogical Society of America
Introduction to Minerals, Inclusions and
Volcanic Processes
Keith d. Putirka
Department of Earth and Environmental Sciences
California State University, Fresno
2576 E. San Ramon Ave., MS/ST24
Fresno, California, 93740-8039, U.S.A.
[email protected]
Minerals are intrinsically resistant to the processes that homogenize silicate liquids—their
compositions thus yield an archive of volcanic and magmatic processes that are invisible at
the whole rock scale. Minerals and their inclusions record diverse magma compositions, the
depths and temperatures of magma storage, the nature of open system processes, and the rates
at which magmas ascend. The potential for understanding volcanic systems through minerals
and their inclusions has long been recognized (Sorby 1858). Sorby’s (1863) study of James
Hall’s reversal experiments helped resolve the “basalt controversy” in favor of a volcanic
origin, while zirkel’s (1863) discovery of quartz within a volcanic rock helped tip the balance
in favor of a magmatic origin for granite (Young 2003). Studies of phenocrysts have also long
illustrated the importance of wall rock assimilation and magma mixing (e.g., Fenner 1926;
Finch and Anderson 1930; Larson et al. 1938), and the potential for geothermometry (Barth
1934). Darwin’s (1844) mineralogical field-studies in the Galapagos archipelago, followed by
King’s (1878) studies at Hawaii, also inaugurated the establishment of fractional crystallization
as an important evolutionary process (Becker 1897; Bowen 1915).
Recent advances in micro-analytical techniques open a new realm of detail, building upon
a long history of mineralogical research; this volume summarizes some of this progress. Our
summary focuses on volcanologic and magmatic processes, but the methods reviewed here
extend well beyond terrestrial applications. Samples from the Stardust return mission, for
example, show that olivine, plagioclase and pyroxene pervade the solar system (Brownlee et
al. 2006)—while the topics covered here surely apply to all terrestrial-like planetary bodies,
relevance may extend to a cosmic scale. Our more modest hope is that this volume will aid the
study of disparate fields of terrestrial igneous systems, and perhaps provide a catalyst for new
collaborations and integrated studies.
OVerVIeW OF the VOLuMe
Our review begins by tracing the origins of mineral grains, and methods to estimate
pressures (P) and temperatures (T) of crystallization. Key to such attempts is an understanding
of textures, and in her review, Hammer (2008) shows how “dynamic” experiments (conducted
with varying P or T), yield important insights into crystal growth. Early dynamic experiments
(e.g., Lofgren et al. 1974; Walker et al. 1978) have shown that porphyritic textures can result
from a single episode of cooling. More recent experiments demonstrate that crystals can
form during ascent due to loss of volatiles (Hammer and Rutherford 2002). Hammer (2008)
describes these and other advances, and additional challenges that require new experimental
1529-6466/08/0069-0001$05.00 DOI: 10.2138/rmg.2008.69.1
2 Putirka
research. Her review also sets the stage for the next four chapters, where the hope above all
hopes is that not all is disequilibrium and discord.
Putirka (2008), Anderson et al. (2008) and Blundy and Cashman (2008) review various
igneous geothermometers and geobarometers. The Anderson et al. (2008) review includes
accessory phase trace-element thermometers, which can be used to estimate T for the same
grains for which U-series equilibria provide age dates (Cooper and Reid 2008). Also covered
among these chapters are many familiar models involving olivine, amphibole, feldspar,
pyroxene, and spinel, with many new calibrations. Anderson et al. (2008) also examine the
applications of thermobarometry to granitic systems—potentially the un-erupted residue
of high SiO2 volcanism (e.g., Bachman and Bergantz 2004). Hansteen and Klügel (2008)
review another method for P estimation, based on densities of entrapped fluids and appropriate
equations of state. Fluid inclusions appear to provide the single most precise means for P
estimation in geologic systems, and so should prove useful to test other methods. As a matter
of good fortune, Klügel and Klein (2005) show that P estimates obtained from fluid inclusions
match those from mineral-melt equilibria in at least one instance. Both barometers thus appear
to be valid—and indicate transport depths well below the base of the crust. The importance
of this kind of work is well summarized by Blundy and Cashman (2008), where they note
the correlation between petrologic and fumarole/seismic monitoring data at Mt. St. Helens,
underscoring the applicability of such studies to forensic volcanology.
Rutherford’s (2008) chapter returns us to the issue of disequilibrium, with a review of
magma ascent rates. This review shows that consistent values for ascent rates are obtained
from a range of textural and petrologic methods. For example, ascent rates can be ascertained
from amphibole reaction rims formed in slowly ascending silicic eruptions (e.g., Browne
and Gardner 2006) or H-loss profiles in melt inclusions and crystals (Peslier and Luhr 2006;
Humphreys et al. 2008). A most exciting proposition is that ascent rates positively correlate
with volcanic explosivity—an important forensic issue (e.g., Scandone and Malone 1985).
Our volume then moves to a review of melt inclusions, which have received much attention
in the past decade. The interest lies in the fact that melt inclusions expose compositional
variability far in excess of that exhibited by their host magmas (e.g., Sobolev and Shimizu 1993;
Nielsen et al. 1995; Kent et al. 2002; McLennan et al. 2003). Kent (2008) shows that pre-mixed
magma compositions can be preserved as inclusions, thus providing a window into pre-eruptive
conditions (although such interpretations require care, e.g., Danyushevsky et al. 2002; Kent
2008). Métrich and Wallace (2008) review the volatile contents in basaltic melt inclusions and
“magma degassing paths”. They conclude that most phenocrysts (and hence melt inclusions)
form during ascent and degassing, and so provide an invaluable record of volcanic volatile
budgets, from crustal depths to the surface. Métrich et al. (2005) provide a dramatic illustration
of such methods in their forensic analysis of an explosive 2003 “paroxysm” at Stromboli, as do
Roggensack et al. (1997) in a forward-looking study at Cerro Negro. Such methods rely upon
vapor saturation pressures, which are derived from experimentally calibrated models. Moore
(2008) and Blundy and Cashman (2008) test two of the most important models, by Newman
and Lowenstern (2002) (VolatileCalc) and Papale et al. (2006). Moore’s review in particular
shows that CO2 and H2O solubilities are highly sensitive to melt composition, and provides a
guide to the appropriate use of these models, and their errors.
The next four chapters document insights obtained from isotopic studies and diffusion
profiles. Ramos and Tepley (2008) review developments of micro-analytical isotope
measurements, which now have the potential to elucidate even the most cryptic of open system
behaviors. 87Sr/87Sr ratios, for example, can be matched to dissolution surfaces to identify magma
recharge events (Tepley et al. 2000). And 87Sr/86Sr-contrasts within and between phenocrysts
allow differentiation between the roles of wall rock assimilation and enriched mantle sources
to explain elevated 87Sr/86Sr (Ramos and Reid 2005). In the next chapter, Cooper and Reid
Introduction 3
(2008) examine the time scales for such processes through U-series age dating of minerals.
They show that crystals can be recycled over time scales of tens to hundreds of thousands of
years before being erupted. Perhaps even more interesting are discordant 230Th-238U and 226Ra-
230
Th age dates at Mt. St. Helens (Cooper and Reid 2003), which imply that young magmas
can mix with magmas that are tens of thousands of years older (Cooper and Reid 2008).
Such mixing is not entirely unanticipated. Murata and Richter (1966) show that high-energy
eruptions can dredge up and entrain masses of pre-existing crystals. And in what may be an
emerging paradigm, Ryan (1988) and Kuntz (1992) emphasize that volcanic plumbing systems
are best thought of as a plexus of dikes and sills, and that these in turn, as shown by Sinton
and Detrick (1992) and Marsh (1995), are not filled just with liquid, but are comprised of a
mush of crystals and magma (and perhaps bubbles too). Mineral age dates have spurred the
introduction of a new term, “antecryst” (by Wes Hildreth; Cooper and Reid 2008), to represent
minerals that are older than the magmas they inhabit (so are not “phenocrysts”), but are part of
an ancestral magmatic (mush) system (so are not “xenocrysts”). Oxygen isotopes (Bindeman
2008) may be particularly useful for addressing such issues. For example, Bindeman et al.
(2007) use mineral age dates and O isotopes to fingerprint the recycling of surface materials
into magmas; their work reveals a volcanic cycle at the Snake River Plain that begins with
partial melting of pre-existing crust, and is completed by re-melting of caldera-fill, altered by
meteoric waters. Bindeman’s (2008) review also reveals that at Iceland and Yellowstone, most
crystals are out of isotopic equilibrium with their host magmas and are largely inherited from
preexisting rocks; he suggests that true phenocrysts are rare. Costa et al. (2008) review yet
another means to estimate the rates of magmatic processes, using mineral diffusion profiles.
This highly versatile method has led to some very interesting insights. Age dates by Costa and
Dungan (2005), for example, show that wall rock assimilation (at Tatara-San Pedro) can occur
on time scales of months to years. In contrast, Klügel (2001) calculates time scales intervening
between magma aggregation and eruption (at La Palma) on the order of tens to hundreds of
years. Apparently, plenty of time is available for the assimilation and mixing processes that
Bowen (1928) thought were unimportant.
A recurring theme in this volume is the prevalence of open system processes and magma
hybridization. Though appreciated for more than a century (e.g., Bunsen 1851; Daly 1914),
the open-system view fell from favor following Bowen’s (1928) compelling arguments in
favor of fractional crystallization, and against assimilation. Eventually, studies of mineral
zoning by Macdonald and Katsura (1965), Sato (1975), and especially Eichelberger (1975),
Anderson (1976) and Dungan and Rhodes (1978), precipitated the modern view that mixing
and assimilation are ubiquitous. The review by Streck (2008) summarizes how zoning styles
in minerals continue to be important, not just for identifying, but for quantifying open system
processes. The study of Arenal in Costa Rica is illustrative, where Streck et al. (2005) show how
minerals contained within “monotonous” erupted units reveal at least four magmatic episodes.
In his review, Streck (2008) suggests that individual crystals can be composites of such events,
so are not easily classified as phenocrysts, antecrysts or xenocrysts. Perhaps we require a return
to the strictly observational origin of “phenocryst,” as defined by Iddings (1889), to denote
crystals that “stand out conspicuously from the surrounding crystals or glass composing the
groundmass.”
The volume returns to issues of kinetics with Armienti’s (2008) new look at the analysis of
crystal size distributions (CSD). Since Marsh (1988) introduced CSD theory to petrology, such
measurements have been crucial to understanding crystal textures. Where closed systems can
be identified, Armienti (2008) provides the methods by which self-consistent kinetic parameters
can be derived from CSD. The CSD approach can also be extremely important for understanding
open systems, as shown by Armienti et al. (2007) who use CSD, thermobarometry, and isotope-
ratios in minerals at Mt. Etna, to illustrate the time scales at which magmas flush through the
4 Putirka
plumbing system. Our volume concludes with a chapter by Bachmann and Bergantz (2008)
summarizing compositional zonations—not in minerals—but in erupted volcanic strata. They
also review the thermal and compositional forces that drive open system behavior. A seeming
departure from the first two thirds of the volume theme, this review is fundamental—our
microscopic observations are of greatest value when they inform and enlarge our views of
macro-level volcanic phenomena, including those manifest at the outcrop scale.
Many of the contributions in this volume emphasize the close link between technical
capabilities in analytical methods and advances in understanding magmatic and volcanic
systems. Descriptions of many of the most common analytical approaches are reviewed
within these chapters. Such topics include: secondary ion mass spectrometry (Blundy and
Cashman 2008; Kent 2008); electron microprobe (Blundy and Cashman 2008; Kent 2008;
Métrich and Wallace 2008); laser ablation ICP-MS (Kent 2008); Fourier transform infrared
spectroscopy (Moore 2008; Métrich and Wallace 2008); microsampling and isotope mass
spectrometry (Ramos and Tepley 2008); U-series measurement techniques (Cooper and Reid
2008); Nomarski differential interference contrasts (Streck 2008); micro-Raman spectroscopy
(Métrich and Wallace 2008); back-scattered electron microscopy, and cathodoluminescence
(Blundy and Cashman 2008).
SOMe IMMedIate and SOMe MOre dIStant dIreCtIOnS
The complementary nature of all the topics in this volume is not difficult to recognize. As
an example, Blundy et al. (2006) track heating-degassing events during ascent at Mt. St. Helens,
illustrating a marriage of thermometry to the study of melt inclusions. Works by Armienti et
al. (2007) and Roggensack (2001) similarly illustrate how important insights can be obtained
at the confluence of various approaches. But new allied studies are badly needed. Métrich and
Wallace (2008), for example, show that vapor-saturated melt inclusions often record pressures
<2 kbar and rarely above 5 kbar. In some cases, such P estimates are consistent with phase
equilibria (Johnson et al. 2008), or correspond to aseismic zones (Wallace and Gerlach 1994;
Spilliaert et al. 2006), suggesting a genuine link to shallow regions of magma storage. But
fluid inclusions and mineral-melt equilibria often record much higher pressures, reaching 8-10
kbar (Putirka 2008; Hansteen and Klügel 2008). Hansteen and Klügel (2008) suggest that fluid
inclusions can re-equilibrate rapidly, and so should have a natural bias to late-stage volcanic and
magmatic processes; the same may be even more true of melt inclusions, if they are captured at
(relatively) high crystallization rates, caused by ascent and degassing. In this interpretation, no
methods are by necessity wrong, but different methods yield information about different parts
of a volcanic plumbing system. Of course, another possibility is that some melt inclusions are
not vapor saturated (see Blundy and Cashman 2008), in which case P estimates from solubility
models are minimum values; however, variable CO2 contents among many melt inclusion
suites appear to require vapor saturation in certain cases (Metrich and Wallace 2008). Moore
(2008) furthermore notes that VolatileCalc is not calibrated above 5 kbar.
Costa et al. (2008) and Cooper and Reid (2008) discuss an issue of time scale estimates
that is parallel to that for geobarometry. They note that U-series methods provide age dates
on the order of tens to hundreds of thousands of years, while diffusion profiles often yield
times scales of months to years. One locality, the Bishop Tuff of CA, has been studied by both
methods, and as luck would have it, U-series (Simon and Reid 2005) and diffusion profiles
(Morgan and Blake 2006) yield similar time scales (≈100 k.y.). But as Costa et al. (2008)
note, diffusion profiles may access time scales shorter than those accessible to radioisotope
methods, so as with geobarometry, different methods may inform us about different aspects of
volcanic processes. Diffusion dating, for example, should inform us about the time a particular
mineral has spent at high temperature, which in a thermally complex magmatic system may
Introduction 5
be substantially less than the radioisotope record of the time at which a crystal first-formed. In
any case, these issues can undoubtedly be resolved by additional allied efforts.
The alliance of the sub-fields reviewed here bear upon fundamental issues of volcanology:
At what depths are eruptions triggered, and over what time scales? Where and why do
magmas coalesce before ascent? And if magmas stagnate for thousands of years, what forces
are responsible for initiating final ascent, or the degassing processes that accelerate upward
motion? To the extent that we can answer these questions, we move towards formulating tests
of mechanistic models of volcanic eruptions (e.g., Wilson 1980; Slezin 2003; Scandone et
al. 2007), and hypotheses of the tectonic controls on magma transport (e.g., ten Brink and
Brocher 1987; Takada 1994; Putirka and Busby 2007). We undoubtedly have made some
progress since Sorby (1863), Fenner (1926), Barth (1934), and many others—but as yet there
is still much work to do.
aCKnOWLedGMentS
This volume and the associated short course were supported by a generous grant from the
Department of Energy; their contribution was especially key in supporting student attendance.
Debts of thanks are owed to Alex Speer for his efforts to help organize the volume and short
course, and to Jodi Rosso, whose tireless efforts in assembling and editing our volume cannot
be overstated—the Mineralogical Society of America is very fortunate indeed to have their
services. Thanks also to my co-editor, Frank Tepley and to the co-authors of this volume for
their dedication to this project.
reFerenCeS CIted
Anderson AT (1976) Magma mixing: petrological process and volcanological tool. J Volcanol Geotherm Res
1:3-33
Anderson JL, Barth AP, Wooden JL, Mazdab F (2008) Thermometers and thermobarometers in granitic
systems. Rev Mineral Geochem 69:121-142
Armienti P (2008) Decryption of igneous rock textures: crystal size distribution tools. Rev Mineral Geochem
69:623-649
Armienti P, Tonarini S, Innocenti F, D’Orazio M (2007) Mount Etna pyroxene as tracer of petrogenetic
processes and dynamics of the feeding system. Geol Soc Am Spec Paper 418:265-276
Bachman O, Bergantz G (2004) On the origin of crystal-poor rhyolites: extracted from batholithic crystal
mushes. J Petrol 45:1565-1582
Bachmann O, Bergantz GW (2008) Deciphering magma chamber dynamics from styles of compositional
zoning in large silicic ash flow sheets. Rev Mineral Geochem 69:651-674
Barth TW (1934) Temperatures in lavas and magmas and a new geologic thermometer. Naturen 6:187-192
Becker GF (1897) Fractional crystallization of rocks. Am J Sci 4:257-261
Bindeman I (2008) Oxygen isotopes in mantle and crustal magmas as revealed by single crystal analysis. Rev
Mineral Geochem 69:445-478
Bindeman IN, Watts KE, Schmitt AK, Morgan LA, Shanks PWC (2007) Voluminous low δ18O magmas in
the late Miocene Heise volcanic field, Idaho: implications for the fate of Yellowstone hotspot calderas.
Geology 35:1019-1022
Blundy J, Cashman K (2008) Petrologic reconstruction of magmatic system variables and processes. Rev
Mineral Geochem 69:179-239
Blundy J, Cashman K, Humphreys M (2006) Magma heating by decompression-driven crystallization beneath
andesite volcanoes. Nature 443:76-80
Bowen NL (1915) The later stages of the evolution of the igneous rocks. J Geol Supp tono. 8 23:1-91
Bowen NL (1928) The Evolution of the Igneous Rocks. Dover Pub, NY
Browne BL, Gardner JE (2006) The influence of magma ascent pat on the texture, mineralogy and formation
of hornblende reaction rims. Earth Planet Sci Lett 246:161-140
Brownlee D, and others (2006) Comet 81P/Wild 2 Under a microscope. Science 314:1711-1716
Bunsen R (1851) Uber die processe der vulkanischen Gesteinsbildung Islands. Annal Phys Chemie 83:197-
272
6 Putirka
Cooper KM, Reid MR (2008) Uranium-series crystal ages. Rev Mineral Geochem 69:479-554
Cooper KM, Reid MR (2003) Re-examination of crystal ages in recent Mt. St. Helens lavas: implications for
magma reservoir processes. Earth Planet Sci Lett 213:149
Costa F, Dohmen R, Chakraborty S (2008) Time scales of magmatic processes from modeling the zoning
patterns of crystals. Rev Mineral Geochem 69:545-594
Costa F, Dungan M (2005) Short time scales of magmatic assimilation from diffusion modeling of multiple
elements in olivine. Geology 33:837-840
Daly RA (1914) Igneous Rocks and their Origin. McGraw Hill, London
Danyushevsky LV, McNeill LV, Sobolev AV (2002) Experimental and petrological studies of melt inclusions
in phenocrysts from mantle-derived magmas: an overview of techniques, advantages and complications.
Chem Geol 183:5-24
Darwin C (1844) Geolgoical Observations on the Volcanic Islands. London, Smith-Elder
Davis BTC, Boyd DR (1966) The join Mg2Si2O6-CaMgSi2O6 at 30 kilobars pressure and its application to
pyroxenes from kimberlites. J Geophys Res 71:3567-3576
Dungan MA, Rhodes JM (1978) Residual glasses and melt inclusions in basalts from DSDP legs 45 and 46:
evidence for magma mixing. Contrib Mineral Petrol 67:417-431
Eichelberger JC (1975) Origin of andesite and dacite: evidence of missing at Glass Mountain in California and
at other circum-Pacific volcanoes. Geol Soc Am Bull 86:1381-1391
Fenner RH (1926) The Katmai magmatic province. J Geol 34:673-772
Finch RH and Anderson CA (1930) The quartz basalt eruptions of Cinder Cone, Lassen Volcanic National
Park, California. Cal Univ Pubs Geol Sci 19:245-273
Hammer JE (2008) Experimental studies of the kinetics and energetic of magma crystallization. Rev Mineral
Geochem 69:9-59
Hammer JE, Rutherford MJ (2002) An experimental study of the kinetics of decompression-induced
crystallization in silicic melt. J Geophys Res 107
Hansteen TH, Klügel A (2008) Fluid inclusion thermobarometry as a tracer for magmatic processes. Rev
Mineral Geochem 69:143-177
Humphreys MCS, Menand T, Blludy JD, Klimm K (2008) Magma ascent rates in explosive eruptions:
constraints form H2O diffusion in melt inclusions. Earth Planet Sci Lett 270:25-40
Iddings JP (1889) On the crystallization of igneous rocks. Bull Phil Soc Wash 11:65-113
Johnson EJ, Wallace PJ, Cashman KV, Delgado Granados, H, Kent, A (2008) Magmatic volatile contents
and degassing-induced crystallization at Volcan Jorullo, Mexico: Implications for melt evolution and the
plumbing systems of monogenetic volcanoes. Earth Planet Sci Lett, 269:477-486
Kent AJR (2008) Melt inclusions in basaltic and related volcanic rocks. Rev Mineral Geochem 69:273-331
Kent AJR, Baker JA, Wiedenbeck M (2002) Contamination and melt aggregation processes in continental
flood basalts: constraints from melt inclusions in Oligocene basalts from Yemen. Earth Planet Sci Lett
202:577-594
King C (1878) Systematic Geology. U.S. Govt Print Office, Washington D.C.
Klügel A (2001) Prolonged reaction between harzburgite xenoliths and silica-undersaturated melt: implications
for dissolution and Fe-Mg interdiffusion rates of orthopyroxene. Contrib Mineral Petrol 141:1-14
Klügel A, Klein F (2005) Complex storage and ascent at embryonic submarine volcanoes from the Madeira
Archipelago. Geology 34:337-340
Kuntz MA (1992) A model-based perspective of basaltic volcanism, eastern Snake River Plain, Idaho. Geol
Soc Am Mem 179:289-304
Larson ES, Irving J, Gonyer FA, Larson ES III (1938) Petrologic results of a study of the minerals from the
tertiary volcanic rocks of the San Juan region, Colorado. Am Mineral 23:227-257
Lofgren GE, Donaldson CH, Williams RJ, Mullins O, Usselman TM (1974) Experimentally produced textures
and mineral chemistry of Apollo 15 quartz normative basalts. Lunar Planet Sci Conf 5th, p. 549-568
Macdonald GA, Katsura T (1965) Eruption of Lassen Peak, Cascade Range, California, in 1915: example of
mixed magmas. Geol Soc Am Bull 76:475-482
Maclennan J, McKenzie D, Hilton F, Gronvold K, Shimizu, N (2003) Geochemical variability in a single flow
from northern Iceland. J Geophys Res 108: doi 10.1029/2000JB000142
Marsh BD (1988) Crystal size distribution (CSD) in rocks and the kinetics and dynamics of crystallization. I
Theory. Contrib Mineral Petrol 99:277-291
Marsh BD (1995) Solidification fronts and magmatic evolution. Mineral Mag 60:5-40
Métrich N, Wallace PJ (2008) Volatile abundances in basaltic magmas and their degassing paths tracked by
melt inclusions. Rev Mineral Geochem 69:363-402
Métrich N, Bertagnini A, Landi P, Rosi M, Belhadj O (2005) Triggering mechanism at the origin of
paroxysms at Stromboli (Aeolian archipelago, Italy): the 5 April 2003 eruption. Geophys Res Lett
32:10.1029/2004GL022257
Introduction 7
Moore G (2008) Interpreting H2O and CO2 contents in melt inclusions: constraints from solubility experiments
and modeling. Rev Mineral Geochem 69:333-361
Morgan DJ, Blake S (2006) Magmatic residence times of zone phenocrysts: introduction and application of the
binary element diffusion modeling (BEDM) technique. Contrib Mineral Petrol 151:58-70
Murata KJ, Richter DH (1966) The settling of olivine in Kilauean magma as shown by lavas of the 1959
eruption. Am J Sci 264:194-203
Newman S, Lowenstern JB (2002) VolatileCalc: a silicate melt-H2O-CO2 solution model written in Visual
Basic for Excel. Comp Geosci 28:597-604
Nielsen RL, Crum J, Bourgeois R, Hascall K, Forsythe LM, Fisk MR, Christie DM (1995) Melt inclusions in
high-An plagioclase form the Gorda Ridge: an example of the local diversity of MORB parent magmas.
Contrib Mineral Petrol 122:34-50
Papale P, Moretti R, Barbato D (2006) The compositional dependence of the saturation surface of H2O+CO2
fluids in silicate melts. Chem Geol 229:78-95
Peslier AH, Luhr JF (2006) Hydrogen loss from olivines in mantle zenoliths from Sincoe (USA) and Mexico:
mafic alkalic magma ascent rates and water budget of the sub-continental lithosphere. Earth Planet Sci
Lett 242:302-319
Putirka KD (2008) Thermometers and barometers for volcanic systems. Rev Mineral Geochem 69:61-120
Putirka KD, Busby CJ (2007) The tectonic significance of high-K2O volcanism in the Sierra Nevada, California.
Geology 35:923-926
Ramos FC Reid MR (2005) Distinguishing melting of heterogeneous mantle sources from crustal contamination:
insights from Sr isotopes at the phenocrysts scale, Pisgah Crater, California. J Petrol 46:999-1012
Ramos FC, Tepley FJ III (2008) Inter- and intracrystalline isotopic disequilibria: techniques and applications.
Rev Mineral Geochem 69:403-443
Roedder E (1979) Origin and significance of magmatic inclusions. Bull Mineral 102:487-210
Roggensack K, Hervig RL, McKnight SB, Williams SN (1997) Explosive basaltic volcanism from Cerro
Negro volcano: influence of volatiles on eruptive style. Science 277:1639-1642
Roggensack K (2001) Sizing up crystals and their melt inclusions: a new approach to crystallization studies.
Earth Planet Sci Lett 187:221-237
Rutherford MJ (2008) Magma ascent rates. Rev Mineral Geochem 69:241-271
Ryan MP (1988) The mechanics and three-dimensional internal structure of active magmatic systems: Kilauea
volcano, Hawaii. J Geophys Res 93:4213-4248
Sato H (1975) Diffusion coronas around quartz xenocrysts in andesite and basalt from Tertiary volcanic region
in northeastern Shikoku, Japan. Contrib Mineral Petrol 50:49-64
Scandone R, Cashman KV, Malone SD (2007) Magma supply, magma ascent and the style of volcanic
eruptions. Earth Planet Sci Lett 253:513-529
Scandone R, Malone SD (1985) Magma supply, magma discharge and readjustment of the feeding system of
Mount St. Helens during 1980. J Volcanol Geotherm Res 23: 239-262
Simon JI, Reid MR (2005) The pace of rhyolite differentiation and storage in an ‘archetypical’ silicic magma
system, Long Valley, California. Earth Planet Sci Lett 235:123-140
Sinton JM, Detrick RS (1992) Mid-ocean ridge magma chambers. J Geophys Res 97:197-216
Slezin YB (2003) The mechanism of volcanic eruptions (a steady state approach). J Volcanol Geotherm Res
122:7-50
Sobolev AV, Shimizu N (1993) Ultra-depleted primary melt included in an olivine from the Mid-Atlantic ridge.
Nature 363:151-154
Sorby HC (1858) On the microscopic structures of crystals, indicating the origin of minerals and rocks. Geol
Soc London q J 14:453-500
Sorby HC (1863) On the microscopical structure of Mount Sorrel Syenite, artificially fused and slowly cooled.
Proc Geol Polytech Soc W Yorkshire 4:301-304
Spilliaert N, Allard, P, Métrich N, Sobolev AV (2006) Melt inclusion record of the conditions of ascent,
degassing, and extrusion of volatile-rich alkali basalt during the powerful 2002 flank eruption of Mount
Etna (Italy). J Geophys Res doi:10.1029/2005JB003934
Streck MJ (2008) Mineral textures and zoning as evidence for open system processes. Rev Mineral Geochem
69:595-622
Streck MJ, Dungan MA, Bussy F, Malvassi E (2005) Mineral inventory of continuously erupting basaltic
andesites at Arenal volcano, Costa Rica: implications for interpreting monotonous, crystal-rich, mafic arc
stratigraphies. J Volcanol Geotherm Res 140:133-155
Takada A (1994) The influence of regional stress and magmatic input on styles of monogenetic and polygenetic
volcanism. J Geophys Res 99:13563-13573
ten Brink US, Brocher TM (1987) Multichannel seismic evidence for a subcrustal intrusive complex under
Oahu and a model of Hawaiian volcanism. J Geophys Res 92:13687-13707
8 Putirka
Tepley FJ III, Davidson JP, Tilling RI, Arth JG (2000) Magma mixing, recharge and eruption histories recorded
in plagioclase phenocrysts from El Chicón volcano, Mexico. J Petrol 41:1397-1411
Walker D, Powell MA, Lofgren GE, Hays JF (1978) Dynamic crystallization of a eucrite basalt. Lunar Planet
Sci Conf 9: 1369-1391
Wallace PJ, Gerlach TM (1994) Magmatic vapor source for sulfur dioxide released during volcanic eruptions:
Evidence from Mount Pinatubo. Science 265:497-499
Wilson L (1980) Relationships between pressure, volatile content and ejecta velocity in three types of volcanic
explosion. J Volcanol Geotherm Res 8:297-313
Young DA (2003) Mind Over Magma: The Story of Igneous Petrology. Princeton Univ Press, Princeton, NJ
zirkel F (1863) Mikroskopische Gesteinsstudien. Sitz Akad Wissen Wien, Math-Natur Klasse 47:226-270
Reviews in Mineralogy & Geochemistry
Vol. 69 pp. 9-59, 2008 2
Copyright © Mineralogical Society of America
experimental Studies of the
Kinetics and energetics of Magma Crystallization
Julia e. hammer
Department of Geology and Geophysics
University of Hawaii
Honolulu, Hawaii, 96822, U.S.A.
[email protected]
IntrOduCtIOn
A quantitative understanding of crystallization and solidification is central to appreciating
a variety of phenomena at the interface between volcanology and petrology. Interpretation of
magmatic processes from the textures of erupted lavas and pyroclasts in turn depends upon
understanding how textures evolve during cooling, decompression, and devolatilization. Crystal
size, shape, and compositional distribution depend upon the underlying phase equilibria, mass
transport processes, and kinetics of reactions at interfaces. Applications of kinetics in physical
volcanology include estimating intra-eruptive magma residence times using crystal growth
rates (e.g., Mangan 1990); interpreting perturbations in the melt composition or thermal state
from microtextures (e.g., Sharp et al. 1996); assessing degassing as a driver of crystal growth
(e.g., Blundy and Cashman 2001) and microlite nucleation (e.g., Hammer et al. 1999) during
volcanic eruptions; and interpreting magmatic conditions as they change through time using
crystal population trends (e.g., zieg and Marsh 2002).
This contribution is chiefly concerned with magmas out of chemical and textural equilibrium
evolving due to changes in temperature, pressure, or composition. However, studies concerned
with dynamic aspects of syn-eruptive magma evolution are predicated on an understanding
of the equilibrium steady-state, with experiments typically incorporating an initial stage of
equilibration, imposing magma reservoir conditions as a prelude to controlled cooling or
decompression. This review likewise includes a brief overview of the phase equilibria approach
for determining the “initial conditions” for volcanic eruptions. The remainder of the chapter
considers kinetic aspects of crystallization, which are essential for interpreting the information
encoded in the compositions and textures of volcanic materials. A section each is allotted
to (1) presenting concepts and controversies in the theoretical understanding of kinetics of
crystallization since the last RIMG review on this topic (Kirkpatrick 1981), (2) summarizing
generic aspects of experimental methods and experiment design in geological studies, (2)
reviewing established trends in basalt crystallization, highlighting recent studies that build
upon these relationships, and examining pioneering research in olivine growth kinetics and
melt inclusion entrapment, (3) considering constraints on kinetics of crystallization in hydrous
felsic magmas due to decompression and devolatilization, (4) discussing conventional and
innovative methods of characterizing crystal textures of experimental run products, (5) assessing
the effects of phase interactions including epitaxial growth and heterogeneous nucleation, and
(6) describing frontier research on compositional aspects of the melt inclusion entrapment
process incorporating energetic and kinetic issues presented in the previous sections.
1529-6466/08/0069-0002$10.00 DOI: 10.2138/rmg.2008.69.2
10 Hammer
Pre-eruPtIVe MaGMatIC enVIrOnMentS
General approach
While magmas often contain evidence of disequilibrium (zoned phenocrysts; e.g.,
Bachmann and Bergantz 2008; Streck 2008), phenocryst rims and interstitial melt (glass)
typically achieve equilibrium in pre-eruption magmas (e.g., Rutherford et al. 1985; Rutherford
and Devine 1996). Thus, determination of pre-eruptive magma storage conditions is predicated
on analyses of natural samples and reproduction of natural phase compositions in experimental
charges (e.g., Rutherford et al. 1985; Webster 1987; and many others cited below). A typical
approach is to constrain as many environmental variables as possible using analytical techniques
applied to natural samples. Experiments are then run to establish the stability fields of the major
phenocryst phases at nodes in pressure- temperature space. For example, fO2 and temperature may
be determined from co-existing Fe-Ti oxide minerals in natural samples (Andersen et al. 1993);
H2O and CO2 contents may be ascertained from analyses of melt inclusions in phenocrysts (e.g.,
Wallace and Gerlach 1994; Wallace 2005; Kent 2008; Moore 2008; Metrich and Wallace 2008);
the assumption of H2O-saturation may follow from comparison of dissolved H2O contents and
solubility of H2O at reasonable estimates of chamber depth. Prior knowledge of such variables
drastically reduces the number of equilibrium experiments necessary to establish a compelling
match between the phase assemblage in an experiment and that in the natural magma. Matrix
glasses provide an important means of locating the appropriate equilibrium conditions, because
melts respond to the changing proportions and abundances of crystallizing phases. Thus, a
vitreous groundmass is an extremely helpful attribute in the natural materials.
Starting materials may be synthesized from reagent oxides and carbonates (less commonly
gels) to match the natural matrix glass. Application of heat treatments exceeding liquidus tem-
peratures is common practice to homogenize starting materials for phase equilibrium experi-
ments (Costa et al. 2004). This ensures that if particles or microscopic heterogeneities remain,
they are uniformly distributed. However, if natural samples are used as starting materials, care
should be taken to prevent exposure of phenocryst cores during preparation, because mass se-
questered inside a crystal is not necessarily in equilibrium with the matrix melt, and its exposure
to melt during an experiment could measurably draw the system away from the desired pre-
eruptive conditions. Selection of appropriate system boundaries to include only the “reactive
magma” is discussed in depth by Pichavant et al. (2007).
Experimental investigation of processes occurring at elevated temperature and pressure
usually requires assuming that the state of the system at run conditions is preserved when the
material is brought quickly to low temperature, i.e., quenched. Most diffusive processes and in-
terphase reactions of interest are sufficiently sluggish that the assumption is valid for the quench
rates commonly applied (100’s to 1,000’s of °C h−1).
apparatus
Three apparatus have produced the vast majority of phase equilibrium studies at pressures
corresponding to the Earth’s surface and extending to mid-crustal depths: the 1-atmosphere
gas-mixing furnace, the cold-seal (externally-heated) pressure vessel, CSPV, and the internally-
heated pressure vessel, IHPV. The operating techniques are described exhaustively in several
excellent texts (Ulmer 1971; Edgar 1973; Ulmer and Barnes 1987), and papers referenced in
this section. Only the basic definitions, key attributes, and limitations are summarized here to
provide a context for the non-experimentalist.
The gas-mixing furnace uses mixtures of H2 and CO2 or CO and CO2 to reduce the
oxygen content of the furnace atmosphere from Earth’s surface environment (~0.2 bars O2) to
conditions prevailing in planetary interiors (10−7 to 10−17 bars O2). For convenience, oxygen
concentrations are referenced to the oxygen concentrations fixed by temperature-sensitive
Experimental Studies of Magma Crystallization 11
oxygen-buffering reactions such as Ni + O2 = NiO (NNO); fayalite + O2 = quartz + magnetite
(FMq, alternately “qFM”); and iron + O2 = wustite (IW), using log-unit notation. For example,
qFM +1 corresponds to the oxygen fugacity one order of magnitude higher than that fixed by the
qFM buffer at a specified temperature. The gas-mixing system in conjunction with a resistance
furnace is a workhorse in experimental petrology for (a) its ability to generate the continuous
spectrum of relevant oxygen fugacities, not just discrete buffer values, (b) its ability to generate
temperatures required to melt basalts and highly refractory simple silicate compositions, and
modulate temperature using programmable controllers, and (c) for rapid quenching capabilities.
Of course, evaluation of the effects of volatiles in concentrations exceeding nominal values
necessitates study at elevated confining pressure.
As fluid-medium pressure apparatus, the CSPV and IHPV offer pressure control advantages
over solid-media apparatus such as the piston-cylinder and diamond anvil cell. The stress on
a sample capsule is hydrostatic and precisely measurable (commonly less than ± 1 MPa).
Attainable pressure ranges are relevant to volcanic processes in the Earth’s crust (5-600 MPa
for the CSPV, depending on temperature, and up to 1000 MPa for the IHPV), and may be varied
during a run to simulate eruptive decompression. In the CSPV setup, a vessel of high tensile
strength Mo or Ni alloy is joined to a fluid line containing either water or inert gas, and then
placed inside a tube furnace. The pressure seal is made outside the furnace, giving the apparatus
its name (although the junction is far from cold). quench rates are much lower than for 1-atm
gas-mixing apparatus, but various strategies (e.g., Gardner et al. 1999; Gaillard et al. 2003) have
improved quench rates dramatically. The sample fO2 may be controlled by the vessel itself, or
in combination with metal filler rods. Alternatively, a solid-medium buffering assemblage may
be included inside the sealed outer capsule along with the sample to monitor any changes in
the intrinsic sample fO2. Hydrous experiments in vessels pressurized by a gas are particularly
prone to oxidation.
The furnace-vessel relationship is reversed for the IHPV, with the electrical resistance fur-
nace encircling the sample being located inside a much larger pressure vessel. The large volume
in the IHPV allows multiple charges to be run simultaneously (e.g., Martel et al. 1999), or
simply for larger volumes of starting material to be used. The IHPV technique is also used in
seismic velocity determinations (e.g., Kern 1982), electrical resistivity measurements (Gaillard
2004), and falling-sphere viscometry (e.g., Shaw 1963; Vetere et al. 2006). IHPV technology
includes apparatus capable of deforming samples to large strains in a geometry resembling
simple shear (Arbaret et al. 2007; Champallier et al. 2008), with promising potential for inves-
tigating hydrodynamic effects in crystal-rich ascending magmas. Rapid quenching rates and
accurate oxygen fugacity control are achieved by incorporating drop-quenching capabilities and
H2 membranes, respectively (Holloway et al. 1992; Roux and Lefevre 1992; Scaillet et al. 1992;
Roux et al. 1994; Berndt et al. 2002). A sapphire-windowed IHPV recently developed by Gonde
et al. (2006) offers the possibility of observing high temperature/ high-pressure processes in situ.
Due to the technical challenges of maintaining and safely operating large-volume high-pressure
gas apparatus (Lofgren 1987), the IHPV is used in fewer laboratories worldwide than CSPVs.
evaluating experimental results
Criteria for establishing a match with natural pre-eruptive conditions include agreement
between experimental run products and natural samples in terms of the phase assemblage,
compositions of matrix melt (glass) and major silicate minerals, and mineral mode. Chemical
equilibrium and phase stability is ascertained by evaluating crystal texture (i.e., verifying faceted
surfaces), ensuring internal consistency with experiments at bracketing conditions, testing for
run time-independence of outcomes, and by approaching run conditions from above and below
the equilibrium crystal content. Mass balance (based on elemental compositions and abundance
of all phases in the sample) may also be employed to check for sample-container interactions,
and ensure the system was closed with respect to mass during the run.
12 Hammer
Unfortunately, not all erupted magmas are suitable for a study aimed at determining pre-
eruptive conditions. Holocrystalline samples pose the challenge of lacking a glass composition
to compare against experimental glasses. Features such as crystal resorbtion, mingled glasses,
and texturally or mineralogically distinct enclaves suggest that magmas are not in chemical
equilibrium prior to eruption. The experimental technique described above is unlikely to yield
meaningful results with such starting materials.
Phase equilibria in volcano-petrology
A listing of phase equilibrium studies that determine conditions prior to a specific volcanic
eruption or eruptive period is presented in Table 1. The table does not include experimental
studies of intrusive rocks (Clemens and Wall 1981; Clemens et al. 1986; Scaillet et al. 1995;
Dall’Agnol et al. 1999; Bogaerts et al. 2006), studies that explore the effects of environmental
parameters on phase equilibria (e.g., Sisson and Grove 1993; Martel et al. 1999; Brugger et
al. 2003; Feig et al. 2006; Pichavant and Macdonald 2007) or other magma differentiation
processes (e.g., Grove and Juster 1989; Berndt et al. 2005).
Several themes are apparent in the dataset. With the exception of the Laacher See study, all
magmas are from volcanoes at convergent plate boundaries. The majority cover intermediate
magma compositions (andesite to dacite bulk rock compositions) that contain highly evolved
matrix melts (rhyodacite to rhyolite). The majority of studies are performed at H2O -saturation
(PH2O = Ptotal, i.e., the mole fraction of H2O in the fluid phase =1), and at relatively oxidizing
conditions (∆NNO= 0 to +3). Studies of crystal-poor mafic magmas, magmas rich in volatiles
besides H2O, volatile-poor magmas, and magmas equilibrated at moderately reducing conditions
are scarce. Early studies of this type (Holloway and Burnham 1972; Eggler and Burnham
1973) yielded new insights of general petrological importance, but experienced oxidation-
state changes and container-sample reaction. These problems remain today for experiments at
oxidation states < qFM and temperatures above 1100 °C.
Table 1 allows comparison of the ranges of intrinsic conditions (temperature, pressure, oxy-
gen fugacity, H2O content) obtained as equilibrium conditions against the ranges that have been
actually tested as experimental variables. The data sets have merit beyond fulfilling the specific
volcanological objectives that motivated the research. As internally consistent probes of phase
equilibria for selected starting materials, the compositions and modes are typically incorporated
into the MELTS thermodynamic database (Ghiorso and Sack 1995). Apart from exceptionally
cool dacite (Mt. Pinatubo, 1991) and phonolite (Laacher See volcano), the experimental results
describe the expected negative correlation between bulk silica content and equilibrium tem-
perature. Basaltic andesite (Mt. Pelée) and rhyolite (Novarupta) define the high and low PH2O
endmembers, respectively, from among the H2O-saturated magmas. However, the pre-eruption
depths of equilibration are only weakly correlated with bulk composition, and entirely uncor-
related with matrix melt composition. For example, mafic andesite (SW Trident) and rhyolite
(Novarupta) eruptions are staged from similarly shallow depths (50-75 MPa), while dacites
erupt from both shallow (Novarupta, ≤ 50 MPa) and deep (Black Butte, 300 MPa) levels.
CryStaLLIZatIOn KInetICS In SIMPLe SySteMS
The vast majority of nucleation studies have been conducted in compositionally simple
systems and reported in the materials science literature (reviewed recently in Fokin et al. 2006),
with application to the synthesis of industrial glasses and ceramics. Materials such as Li2O·SiO2
(lithium disilicate) and Na2O-2CaO-3SiO2 are considered analogs for geologically relevant
melts (e.g., Davis and Ihinger 2002). The present treatment is limited to aspects of crystal
nucleation phenomena recently evaluated with experiments (steady-state nucleation, transient-
nucleation, the crystal-melt interfacial free energy) and adjustments to the classical nucleation
theory (CNT). The section concludes with an overview of aspects of crystal growth theory.
table 1. Determinations of pre-eruption magmatic conditions using the method of experimental phase equilibria.
Magma investigated experimental conditions Pre-eruption equilibrium conditions ref.
Volcano Eruption Bulk Bulk SiO2 Matrix Glass S.M.1 Pressure Temp. DNNO xH2O Pressure Temp. DNNO xH2O
Composition SiO2 (MPa) (°C) (MPa) (°C)
St. Helens 1980 dacite 62.8 72.7 1A 100-320 847-1190 −1 to +4 0.5-1 220 ± 30 930 ± 10 0-4 0.5-0.7 1
St. Helens 1980 dacite 63.0-63.8 72.7 1A 220, 320 890-970 0-4 0.4-1 220 920 1.5 0.67 2
St. Helens 1480 A.D. dacite 67.2 74.8 1A 100-350 825-875 0.67-1.17 0-1 250-350 848 ± 5 0.8 0.4-0.5 3
Mt. Pinatubo June 15, 1991 dacite 64.76 76.44 1B 160-390 760-925 2-4 0.5-1 220 ±50 780 ±10 3 1 4
Soufriere Hills January, 1996 andesite 59.15-59.46 n.d. 1A 50-210 800-935 1 0.5-1 115-130 820-840--> 1 1 5
840-880
W. Mexico n.d. andesite ~62 n.d. 1A 44.1-294.4 950-1100 1.1-2.8 1 70-150 950-975 2 1 6
W. Mexico n.d. basaltic andesite ~55 n.d. 1A 51.7-302.7 950-1150 1.6-5.6 1 <250 1000-1150 2 1 6
Santorini ~1645 B.C. rhyodacite1 69.07 unknown 1A 50-250 800-920 0.5-1 1 210-240 825±25 0.5-1 1 7
Santorini ~1645 B.C. rhyodacite2 69.07 72.1-73.9 1A 50-250 800-920 0.5-1 1 50 885 0.5-1 1 7
Santorini ~1645 B.C. andesite 56.33 n.d. 1A 50-150 865-910 0.5-1 1 >75 885 0.5-1 1 7
Mt. Pinatubo June 15, 1991 dacite 64.76 77.75 1C 220-390 750-900 0-2.7 0.3-1 220 760±20 1.5-1.7 >0.88 8
SW Trident 1953-1974 A.D. andesite 55.8-58.9 ≈63 1B 1000-1100 50-200 0 <1 - 1 90 1000±20 1 1 9
Novarupta June 6-9, 1912 rhyolite 77.23 77.64 1A 25-200 760-900 0.5 1 40-100 800-850 0.5 1 10
Soufriere Hills January, 1996 andesite 59.15 78.8 2 25-210 825-1100 1 1 >100 900-1000 1 1 11
Mt. Pelée 650 y.b.p., 1929 andesite 60-61 62-63 1C 200-400 850-1040 0-3 0.7-1 200 ±50 875-900 0.4-0.8 1 12
Mt. Pelée 40-19.5 ky.b.p. basaltic andesite 53 74.0-76.5 1C 398.8-427 949-1025 0.2-3.8 0.58-1 400 950-1025 0-4 1 13
Mt. Pinatubo June 15, 1991 dacite 64.76 78.2 1B 150-220 780 2 1 220-155 780 2 0.94-1 14
Novarupta June 6-9, 1912 andesite 59 67.6 1B 50-225 850-1050 1 0.34-1 100-75 930-960 1 0.7-1 15
Novarupta June 6-9, 1912 dacite 65.4 79.1 1B 15-200 820-950 0-2 0.27-1 50-25 850-880 0-2 0.9-1 15
Laacher See 12,900 y.b.p. phonolite 58.41 57.37 1A 75-175 725-800 1 1 115-145 750-760 1 1 16
Unzen 1992 hybrid3 64.6 68.2 1C 100-300 750-875 0 0.4-1 300 930 hybrid3 0 <1 17
Experimental Studies of Magma Crystallization
Stromboli 800-1600 A.D. basalt 49.4 49.8 1C 48-417 1050-1175 0-2 0.06-0.25 100-270 1140-1160 0-2 0.15-0.65 18
Aniakchak 3430 ybp rhyodacite 70.57 71.5 1B 50-200 820-950 0.5-1 1 125-150 870-880 0.5-1 0.8 19
Arenal ~2006 basaltic andesite 52.0-54.0 n.d. 2 500 1100-1300 −3 to −2.5 1 500 1150 −1 to 0 1 20
Black Butte 10,000 y.b.p. dacite 64.63 67.73 1C 100-450 800-950 0-1 1 300 890-910 1 1 21
1
Starting materials: [1A] powdered natural material, [1B] crushed natural material, [1C] fused natural material, [2] synthetic material
2
[1] Rutherford et al. 1985, [2] Rutherford and Devine 1988, [3] Gardner et al. 1995, [4] Rutherford and Devine 1996, [5] Barclay et al. 1998, [6] Moore and Carmichael 1998, [7] Cottrell et al. 1999, [8] Scaillet and Evans 1999,
[9] Coombs et al. 2000, [10] Coombs and Gardner 2001, [11] Couch et al. 2001, [12] Martel et al. 1999; Pichavant et al. 2002, [13] Pichavant et al. 2002, [14] Hammer and Rutherford 2003, [15] Hammer et al. 2002, [16] Harms
et al. 2004, [17] Holtz et al. 2005, [18] Di Carlo et al. 2006, [19] Larsen 2006, [20] Petermann and Lundstrom 2006, [21] McCanta et al. 2007
13
14 Hammer
Fundamental concepts of crystal nucleation
Nucleation exerts the primary control on the grain size of rocks (Swanson 1977), and thus
the texture and many of their mechanical properties (Kesson et al. 2001). Reviews of the first
several decades of nucleation research in igneous rocks include those by Kirkpatrick (1981),
Cashman (1990), and Lasaga (1998). Despite a long and distinguished history in metallurgy,
ceramics, and igneous petrology, nucleation continues to lie at the frontier in materials
science because it is a difficult process to study directly. Energy barriers to nucleation are
strongly dependent on melt structure (Toshiya et al. 1991) and thus the thermal pre-treatment,
initial volatile content, and major element composition (Davis et al. 1997) of the melt. In
many cases, the preparation of the experimental “starting material” is reported here, as these
steps may be as important to controlling eventual texture as subliquidus thermal history.
Supercritical clusters of atoms (nuclei) are exceedingly small (10-1000 atoms, or ~1-100 nm)
and ephemeral according to the classical theory of crystal nucleation (Kelton and Greer 1988).
In hydrous magmatic systems relevant to crustal-level reservoirs and conduits, the process is
seemingly even more inaccessible to the experimentalist because of the need to modulate the
H2O concentration in the melt. This prevents in situ study of nucleation at ambient pressure, as
in scanning calorimetric (Davis and Ihinger 1999) and electrical resistivity approaches (Wang
and Lu 2000). Nevertheless, advances in specimen analysis are allowing phenomenological
studies to push theory development (zhang et al. 2003).
theory development through experimentation
The traditional view of nucleation is that crystal-like clusters of atoms possessing the
physical and chemical properties of macroscopic crystalline solids form by random fluctuations
in a melt when below its liquidus temperature (Volmer and Weber 1926; Becker and Doring
1935). The excess free energy associated with maintaining the cluster-liquid interface causes
clusters below a critical size to shrink and those above it to grow into nuclei. The critical
cluster size itself varies with undercooling, becoming smaller as undercooling increases.
Homogeneous nucleation is rapid at the temperature where a large fraction of clusters achieve
the critical cluster size. The temperature dependence of the cluster size distribution is given
by a Boltzmann function: nr = noexp(−∆Gr/kT), where nr is the number of spherical clusters of
radius r, no is the total number of atoms in the system, ∆Gr is the excess free energy associated
with the cluster, T is temperature in Kelvin, and k is Boltzmann’s constant. Porter and Easterling
(1997) provide a plausible explanation for the extreme sensitivity of subliquidus nucleation
behavior to duration and temperature of superliquidus heat treatments, without calling upon
ephemeral heterogeneities to serve as nucleation sites: increasing temperature causes melt
depolymerization, lessening the quantity of large clusters poised to become nuclei should
the temperature drop below the equilibrium liquidus. If cooling after protracted superheating
is rapid, the effect of unrelaxed (i.e., disequilibrium) melt structure is to inhibit nucleation
at subliquidus temperatures. This deleterious effect on nucleation increases as the degree of
superheating increases.
Most experimental work in nucleation employs two-stage experiments, in which crystals
nucleate at low temperature, Tn, and grow to observable size in a separate “development”
step at a higher (but still subliquidus) temperature, Td. Recall that the critical size decreases
with decreasing temperature. Thus, the critical size at Tn is smaller than the critical size at
Td; only nuclei that have achieved critical size at the development temperature (Td) grow to
become observable crystals. Smaller clusters have a high probability of dissolving during the
development stage. Thus, not all nuclei formed in the nucleation stage survive to be counted,
which could lead to underestimation of the nucleation rate at Tn. However, nucleation rates can be
found through extrapolation by determining the functional dependence of the surviving crystal
quantity on Td (e.g., Fokin et al. 2006). A second important feature of nucleation phenomenon
is that the steady-state (time independent) nucleation rate is not attained immediately at a
Experimental Studies of Magma Crystallization 15
given temperature. Rather, a relaxation time is required for melt to structurally adjust and to
form the equilibrium distribution of clusters. An apparent lag time prior to nucleation, during
which nucleation rates are much lower than at steady state, is known as the period of transient
nucleation or induction time.
Steady state nucleation. According to classical nucleation theory (CNT), the steady
state rate equation for the formation of spherical critical nuclei (I, m−3 s−1) as a function of
temperature has the form (James 1985):
−( ∆G * + ∆GD ) 16 πσ3
I = A exp ; ∆G* = S ( θ) (1)
kBT 3∆GV 2
The pre-exponential factor A includes the frequency of attachment attempts and the specific
number density of reactant atoms in the melt. A = kB Tnv/h; where kB is the Boltzmann constant,
T is temperature, nv is the volumetric concentration of reactant atoms, and h is Planck’s
constant. ∆G* is the free energy required to form a spherical critical nucleus having properties
of the macroscopic solid. σ is the free energy associated with the crystal-liquid interface. ∆GV
is the free energy change per volume of the transformation, defined ∆GV = ∆G/VM, where VM
is the molar volume of the crystallizing phase and ∆G is the bulk free energy decrease driving
solidification. (Several processes can cause ∆G to decrease in natural magmas, as will be
explored below. Common mechanisms include decreasing the temperature of a melt below its
liquidus and changing the melt’s liquidus temperature through a change in melt composition.)
∆GD is the activation energy required for the attachment of atoms to a cluster, i.e., a kinetic
barrier that includes transport of components through the fluid. θ is the wetting angle between
the nucleus-wall and nucleus-liquid interfaces, and S(θ) = (2 + cosθ)(1 − cosθ)2/4.
The balance between the thermodynamic driving force (∆G*), which propels solidification
for all temperatures below the equilibrium melting point, and the kinetic barrier (∆GD), which
grows as temperature decreases, give rise to a nucleation function with a maximum. In the
event that nucleation is homogeneous, S(θ) = 1 and the thermodynamic barrier is simply a
function of the volumetric free energy difference between liquid and solid (James 1985).
Heterogeneous crystal nucleation is rare in metals and alloys, but common in silicate systems
(James 1974; Kelton and Greer 1988).
Several simplifying assumptions are employed to compare experimental data with the
CNT. The first is that activation energy of atomic jumps across the liquid-nucleus interface
is the same as that of shear relaxation of the liquid, i.e., the Stokes-Einstein approximation
(Ree and Eyring 1958; Dingwell and Webb 1989), allowing ∆GD to be expressed in terms
of viscosity (η). Second, the heat capacity of the liquid is assumed equal to the heat capacity
of the solid, which reduces the thermodynamic driving force ∆G for crystallization to ∆G =
∆H∆T/TL. Combined with the pre-exponential terms, the rate equation for homogenous crystal
nucleation becomes:
Ac −16 πσ3TL 2 nV kB
I= T exp ; Ac = (2 )
η 3∆H ∆T kBT
2 2
3πλ 3
(although alternate formulations are proposed for the pre-exponential term, Ac; e.g., Gránásy
et al. 2002).
Steady-state nucleation rate data as a function of reduced (homologous) temperature (T/
TL) are shown for several one-component systems (Fig. 1). The height and position of the
maxima are parameters that determine whether given materials are likely to produce crystal-
rich or glassy materials upon rapid cooling. If the kinetic barrier is decreased relative to the
thermodynamic driving force, the maximum increases and shifts toward lower temperatures.
The addition of H2O to silicate melts produces this effect by lowering melt viscosity (Fig. 2).
16 Hammer
Figure 1. Steady-state nucleation
rate versus reduced temperature
(TL= liquidus temperature) for
crystals forming in glasses of
the same composition: 3MgO
Al2O3·3SiO2 (curve 1), Li2O·2SiO2
(curve 2), Na2O·2CaO·3SiO2
(curve 3), 2Na2O·CaO·3SiO2
(curve 4). [Used by permission of
Elsevier, from Fokin et al. (2006),
J Non-Cryst Solids, Vol. 352, Fig.
11, p. 2681-2714.]
T/TL
Figure 1.
Figure 2. Temperature dependence
of the steady-state nucleation rate
in Na2O·2CaO·3SiO2 containing
different amounts of H2O: 0.01 mol%
(curve 1), 0.20 mol% (curve 2). [Used
by permission of Elsevier, from Fokin
et al. (2006), J Non-Cryst Solids, Vol.
352, Fig. 13b, pp. 2681-2714.]
Figure 2.
Conversely, if the free energy of forming a critical cluster is reduced by decreasing interfacial
energy, e.g., Muller et al. (2000), the maximum nucleation rate is also increased, but shifted
toward higher temperatures.
Transient nucleation. The transient rate of nucleation I’ at time t during the period
of structural relaxation and formation of the steady-state cluster distribution is given by a
stochastic equation (Kashchiev 1969):
∞
−n2t
I = I 1 + 2∑ (−1) exp
′ n
(3)
n =1 τ
Experimental Studies of Magma Crystallization 17
where τ is induction time and n is an integer. The nuclei density N(t) is related to time according
to the integrated form of Eqn. (3) obtained by Kashchiev (1969):
(
∞ −1
)
n
N (t ) t π2 t
= − − 2∑ exp − n2 ( 4)
Iτ τ 6 n =1 n
2
τ
Experimental data for 2Na2O·CaO·3SiO2 and typical N(t) curves are shown in Figure 3, indicat-
ing the transient stage of nucleation, the induction time, and the steady state nucleation region.
Davis et al. (1997) report nucle-
ation kinetics in H2O-bearing lithium
disilicate melt. Reagents and H2O are
fused at 1350 °C, quenched to glass,
reheated at 435 °C for periods from
100-10,000 minutes in ambient atmo-
sphere, and transferred to 600 °C for
growth of nuclei to observable size.
Small amounts (~10-1000 ppm) of
N, m-3
H2O in lithium disilicate glass increase
nucleation rate and decrease the in-
duction time (Fig. 4). The effects are
greater than can be attributed to reduc-
tion of the kinetic barrier through de-
creasing viscosity alone. However, a
3% decrease in σ can account for the
nucleation behavior, and is consistent
with a negative correspondence be-
tween σ and dissolved H2O content that
is inferred for natural silicate melts by
Figure 3. Typical curve of the number density of crystals
Figure 3.
Hammer (2004). forming in glass of the same composition. T = nucleation
n
Interfacial energy. Values of the temperature, Td=development (growth) temperature.
[Used by permission of Elsevier, from Fokin et al. (2006),
crystal-melt interfacial free energy J Non-Cryst Solids, Vol. 352, Fig. 3, pp. 2681-2714.]
(σ) defined in the classical nucleation
theory (Eqn. 1) are exceedingly difficult
to obtain independently from nucleation
data. The few such values in existence were derived from measurements of macroscopic crystal-
melt dihedral angles (Cooper and Kohlstedt 1982; Rose and Brenan 2001; Ikeda et al. 2002)
and numerical studies of hard sphere molecular interaction potentials (Battezzati 2001; Granasy
et al. 2002; Davidchack and Laird 2003). These data sets do not include many geologically-
important crystal-melt combinations. One method of obtaining σ for the case of homogeneous
nucleation is to assume the CNT is valid and determine its value from the slope of a plot of (1/
(T∆GV2)) versus ln(Iη/T) (Gonzalez-Oliver and James 1980). Interestingly, this method rarely
yields constant values of σ for all experiments in a given system. As will be discussed further
below, another approach is to assume that the form of the CNT is correct but that the interfacial
energy is not single-valued (James 1974; Kelton and Greer 1988). It is possible to rearrange
Equation (2) in order to solve for σ using nucleation rate data, thus generating a separate value
for each experiment. Note that in the event of heterogeneous nucleation, it is impossible to solve
for σ in this way without knowledge of θ.
Temperature and curvature dependence of σ. Recent studies critically examine the
assumption that σ remains constant as clusters grow into crystal nuclei and nuclei become
macroscopic crystals. One approach for determining σ that is consistent with the steady-state
18 Hammer
1x10-3
representative error
1x10-4
number density (um-3)
1x10-5
Figure 4. Number density of
lithium disilicate crystals forming
1x10-6 17 ppm H2O
in variably hydrated melt of the
same (anhydrous) composition
113 ppm at 435 °C. Symbols designate
1x10 -7
181 ppm
water contents of starting glasses
determined by FTIR (Davis et
328 ppm al. 1997). [Used by permission
1x10 -8
670 ppm
of Elsevier, from Davies et al.
(1997), J Non-Cryst Solids, Vol.
957 ppm 219, Fig. 1, pp. 62-69.]
1x10 -9
10 100 1000 10,000 100,000
time (min)
Figure 4. CNT but not explicitly dependent upon it, involves fitting the transient nucleation expression
(Shneidman 1988) to transient nucleation rates (Deubener and Weinberg 1998; Sycheva
1998). Three data sets obtained in this manner in the lithium disilicate system over a range of
development temperatures produce positive correlation between σ and nucleation temperature
(Deubener and Weinberg 1998); (Fig. 5) This result seemingly conflicts with the thermodynamic
point of view in which the surface energy of a crystal having infinite radius of curvature (σinf;
150
100
σ (mJ m-2)
Td=560 oC
50
Td=570 oC
Td=623 oC
0
690 710 730 750 770
Tn (K)
Figure 5. Crystal-liquid surface energies in Li2O·2SiO2 glass as a function of nucleation temperature (Tn)
obtained at development temperatures of 560 °C (squares), 570 °C (circles), and 623 °C (x’s); various
sourcesFigure 5.
in Deubner and Weinberg (1998). All sets include the same value at 500 °C (773 K), obtained from
transient nucleation data Deubner and Weinberg (1998). Straight lines are least squares fits to the data.
Experimental Studies of Magma Crystallization 19
i.e., a flat surface) decreases with increasing temperature (Gutzow et al. 1985; Skripov and
Faizullin 2005). However, Fokin and zonotto (2000) draw upon Tolman’s equation
σ inf
σ( R) = (5)
2δ
1 + R
(Tolman 1949) to address the apparent discrepancy and separate temperature and curvature
controls on σ. Tolman’s parameter, δ, characterizes the width of the interface and is on the
order of atomic dimensions; R is the radius of the critical cluster. Equation (5) shows how the
magnitude of the curvature correction for surface energy increases as the nucleus decreases in
size. Embedded in the equation is a temperature dependence of the critical radius: because the
bulk free energy driving force increases with decreasing subliquidus temperature, R decreases
with decreasing temperature. Near the liquidus temperature, σ closely approaches σinf because
R is a maximum value. As temperature decreases, both the numerator and denominator of the
ratio defining σ(R) decrease, with the value of σ dependent upon the relative magnitudes of δ
and R. Incorporating Tolman’s equation to the expression for the work of forming a nucleus,
Fokin and zanotto (2000) fix δ and find the range of values of σinf for which the calculated
nucleation rates equal published experimental values in the Li2O-2SiO2 and Na2O-2CaO-
3SiO2 systems. A plausible range in δ values (e.g., 2.8 × 10−10 m < δ < 3.4 × 10−10 m for
Li2O-2SiO2) produces σinf values that increase with decreasing temperature (in accord with
theory) and explains the correlation between σ and temperature observed by Deubener and
Weinberg (1998). The analysis shows that the observed increase in σ with temperature (Fig.
5) may actually arise from the dependence of the critical radius on temperature.
Non-CNT assessment of σ. The wetting, or dihedral, angle (θ) at the contact between two
crystals and melt is related to the ratio of interfacial free energies (Smith 1964):
σsolid-solid θ
= 2 cos (6)
σsolid-liquid 2
However, neither σsolid-solid nor σsolid-liquid is truly single-valued for a given system at equilibrium.
Euhedral crystals are nonspherical precisely because σsolid-liquid is directionally anisotropic.
Sessile drop techniques, in which a drop of liquid is placed on a solid surface and allowed to
mechanically equilibrate (e.g., Bagdassarov et al. 2000), confirm that different faces are usually
characterized by different values of σsolid-liquid, with some minerals (e.g., olivine and pyroxene)
more strongly anisotropic than others (e.g., spinel structured Fe-Ti oxides; Shafer and Foley
2002). The determination of σsolid-solid interfacial free energy is complicated by the fact that
solid-solid contacts include varying crystallographic misorientations among crystals, giving
rise to a spectrum of equilibrium dihedral angles. Nonetheless, assessment of characteristic
σ ratios through wetting angle studies is typically conducted in crystal-rich natural rocks and
near-solidus melting experiments containing abundant mineral contacts (Waff and Faul 1992;
Laporte and Watson 1995; Holness 2006), with obvious implications for mantle melt storage
and migration. Ikeda et al. (2002) measured θ among sparse crystals in relatively melt-rich
magmas in the Fe-free diopside-anorthite system to estimate the ratio of interfacial energies
between diopside and melt. They observe that σsolid-liquid is ~57% of σsolid-solid. Conceptually, as
the crystallographic misorientation angle approaches zero and lattices of adjacent crystals come
into alignment, the σsolid-solid value should decline to a minimum allowing optimal investigation
of σsolid-liquid through wetting angle measurements. If variation in σsolid-solid with crystallographic
misorientation is large compared with directional anisotropy of σsolid-liquid, i.e., if σsolid-liquid may
be assumed constant in a given crystal-melt system, then θ should increase with decreasing
misorientation. Evaluating this hypothesis and determining the maximum value of σsolid-liquid as
a fraction of σsolid-solid would require electron backscattered diffraction (EBSD), transmission
20 Hammer
electron microscopy (TEM) or universal stage petrography as well as accurate measurement of
θ. Bearing in mind the curvature (size) effects prescribed by Equation (5), such determinations
of macroscopic σsolid-liquid may yet fail to reconcile nucleation data with theory.
A novel approach to determining σ (i.e., σsolid-liquid) utilizing nucleation behavior
but independent from the CNT capitalizes on the observation from two-stage nucleation
experiments (Fokin et al. 2000) that higher development temperatures are associated with
longer induction times. The difference in induction time (∆t) scales with the difference in
development temperatures because the size of the critical cluster increases, while cluster
growth rate remains relatively constant. Using an analytical expression for ∆t, Kalinina et al.
(1976) and Fokin et al. (2000) incorporate the definition of critical size, R = 2σ/∆Gv, to obtain
σ from measurements of ∆t. The values of σ obtained are nearly an order of magnitude greater
than values obtained using nucleation data in conjunction with the CNT. If the independently-
derived values are used to forward calculate nucleation rate using the CNT, the prediction
underestimates observed nucleation rates by many orders of magnitude. The explanation for
this discrepancy put forth by Fokin et al. (2000) is that macroscopic values of thermodynamic
driving force do not match the driving force for cluster formation and growth. This shift in
blame from uncertainties in σ to uncertainties in ∆G is expanded in the section Beyond the
CNT, below.
Diffuse Interface Theory. An alternative to the thermodynamic description of heterogeneous
systems in which phases are separated by razor-sharp boundaries (Gibbs 1906) posits a spatial
distribution of thermodynamic properties in the vicinity of clusters (Turnbull 1964). The diffuse
interface theory, DIT (Spaepen 1994; Granasy et al. 1996), takes into account gradations
between bulk solid and bulk liquid values of enthalpy and entropy within an interface region.
This region is not a third phase in the thermodynamic sense, but rather a transitional interval that
exists solely because solid and liquid surround it. The model states that relative to the bulk melt
and cluster center (which possesses properties of a bulk solid), the crystal-melt interfaces are
regions in which excess free energy (σ) increases to a maximum at some radial distance from
the center of a subcritical cluster. In contrast to crystal-crystal boundaries, which are sharply
defined regions of large enthalpy due to a high density of broken bonds, cluster-liquid interfacial
regions contain a lower density of broken bonds. Bonding arrangements in this region may be
somewhat similar to those in the neighboring solid (Spaepen 1994). Because undercooling a
melt to produce glass is impossible without positive interfacial free energy (otherwise there
would be no thermodynamic barrier to crystallization), the interfacial free energy (σint), defined
as Hint − TSint must be greater than zero. Thus, the contribution of the enthalpy term (Hint) must
increase more steeply than the entropy term (TSint) with radial distance from the nucleus in
order for σ to remain positive.
The central concept of the DIT is employed to interpret the observed correspondence
of σ with melt H2O content found by Hammer (2004) upon fitting the CNT with plagioclase
nucleation data in hydrous silicate melt (Hammer and Rutherford 2002). The Hint and TSint
gradients are presumed sensitive to H2O content through a control over bonding and atomic
configuration in the diffuse interfacial region (Fig. 6a,c). Solubility mechanisms (Burnham
1975; Silver and Stolper 1989; McMillan 1994) and accepted views of melt structure suggest
that the gradients in Hint and TSint both shift toward the cluster as dissolved H2O increases,
while the observed σ - H2O relationship (Fig. 6b) suggests that entropy shifts more. That is,
the magnitude of the difference between Hint and TSint decreases with the addition of H2O to
the system (Fig. 6c).
Beyond the CNT. Despite predictive strengths in some areas, the CNT fails to quantitatively
describe observation for all cases in which σ is not used as a fit parameter (Fokin et al. 2005,
2006). A likely flaw is the assumption that critical nuclei possess the thermodynamic and
surface properties of the macroscopic phase (Fokin et al. 2006, 2007). Inverting the CNT and
Experimental Studies of Magma Crystallization 21
a 0.10 b
σ ' [J m-2]
Interfacial Free Energy
0.05
σ =Hint-TSint
0
0 1 2 3 4 5
H2O [wt. %]
decreasing
H2O
0 solid
(cluster interfacial region liquid
center)
radial distance from cluster center
decreasing H2O
Energy
Hint
TSint
low H2O
c high H2O
Figure 6. Application of the diffuse interface theory to experimental data (Hammer and Rutherford 2002).
Figure 6.
(a) Schematic cross section of the interfacial region between a subcritical cluster and the surrounding melt
showing thermodynamic properties of a diffuse interface. Materials in “solid” and “liquid” regions possess
properties of the bulk (i.e., macroscopic) phases. The region between bulk phases has interfacial free
energy, σ = Hint − TSint. Increasing σ with decreasing dissolved H2O is inferred from (b), values obtained
by back solving the CNT equation for nucleation rate using nucleation data from Hammer and Rutherford
(2002). (c) Gradients in enthalpy and entropy with radial distance from a subcritical cluster for H2O-rich
(dashed curve) and H2O-poor (unbroken curve) melts. A shift upon devolatilization of greater magnitude in
the entropy gradient is conceptually consistent with the σ-H2O relationship in (b). From Hammer (2004).
an accepted crystal growth theory (Gutzow and Schmelzer 1995), Fokin et al. (2007) solve for
the diffusion coefficient for the interface region (D) two ways: (1) using experimental crystal
growth data, (2) using experimentally determined nucleation induction times, literature values
for σ, and an expression for the time lag for formation of critical clusters:
16 kBT σ
τ= (7)
3 ∆Gv2 a 2 D
where a is a size parameter on the order of ionic radius and other variables are as defined
previously. Differing values of D, where the growth-based values greatly exceed the nucleation-
based values, suggest that either the definition of interfacial free energy is flawed (as explained
above), or the diffusion coefficient controlling nucleation of critical clusters differs from that
governing growth of macroscopic crystals. Further evidence in favor of the latter, and against
22 Hammer
the assumption of bulk properties in clusters, includes dozens of simple silicate systems
studied since the 1960’s in which nucleation occurs at lower thermodynamic driving force than
predicted by the CNT (Granasy and James 1998). The interfacial free energy characterizing
clusters may be considerably smaller than that estimated for the macrophase because clusters
may more closely resemble melt in composition and/or structure (Fokin et al. 2007). Clearly,
much experimental and analytical work remains in elucidating the thermodynamics and
transport properties in the interfacial regions of critical clusters.
Crystal growth
The present treatment presents only the fundamentals necessary to appreciate and
evaluate recent experimental literature in igneous petrology. The interested reader is pointed to
excellent descriptions of crystal growth theory, Avrami phase transformation theory, numerical
treatments, and related concepts by Christian (1965), Dowty (1980) and Kirkpatrick (1981).
We focus on crystal growth regimes that influence crystal morphology and are defined based on
the rate-limiting step of the crystallization process. If the attachment of atoms to the growing
crystal face is slow compared to the migration of compatible atoms through the surrounding
medium toward the surface and migration of rejected atoms away from the surface, then growth
is said to be “interface-controlled.” The equation describing steady-state growth (Turnbull and
Cohen 1960) in either regime, with simplifying assumptions (Kirkpatrick 1981) is:
−∆G′
G = faoν exp (8)
RT
where G is growth rate, f is the fraction of site on the crystal surface available for attachment,
ao is the thickness of a growth layer, ν is the frequency of atom attachment attempts, R is the
gas constant, T is temperature, ∆G′ is the activation energy for attachment (an energy barrier),
and again ∆G is the bulk chemical driving force. Growth rate is zero at the liquidus temperature
because the chemical driving force is zero, increases to a maximum as undercooling increases,
and then decreases as activation energy for attachment become prohibitively large. A unary
(one-component) system subjected to undercooling always crystallizes in the interface-limited
regime, because atomic mobility in the melt is not required for growth layers to advance.
Crystallization in multicomponent systems in which the crystal and melt compositions
differ (incongruent crystallization) may depart from steady state growth with increasing
undercooling. If the component mobility is very slow compared to the attachment rate, then
growth is “diffusion-controlled,” and the growth rate evolves through time as:
12
D
G = k (9)
t
where k is a constant involving correction terms, and D is the diffusion coefficient for the rate-
controlling species in the melt. Each regime is associated with a relative undercooling, and
gives rise to a characteristic set of growth morphologies.
Interface control. Interface-control is the prevailing case at low degrees of undercooling,
where high temperatures (and/or H2O contents) promote rapid diffusion in the melt. Crystals
growing near chemical equilibrium (i.e., in response to small ∆Gv) with the melt always undergo
interface-controlled growth. A simple inverse relationship between nucleation and growth rates
exists if a fixed volume of solid is distributed on nuclei or pre-existing crystals per unit of time.
If the area of growth substrate is small (e.g., a crystal-poor melt at low nucleation rate), distri-
bution of a fixed volume of solid produces a greater thickness of new growth than if the same
volume is distributed on a large area of substrate (Cashman 1993). Thus, for the case of crystal-
lization at low undercooling, sluggish nucleation encourages rapid growth. The corollary is that
rapid nucleation may suppress growth rates. This is essentially a mass balance argument that
predicts a separation of the growth and nucleation rate maxima as functions of undercooling.
Experimental Studies of Magma Crystallization 23
Interface-controlled growth produces faceted to euhedral crystals according to one of
three mechanisms: continuous growth, advance on screw dislocations, and layer-by-layer
growth. The relevant mechanism is dominantly controlled by interface roughness (Sunagawa
1981). Crystal surfaces are atomically rough if the entropy of fusion is small, and atomically
smooth if the entropy of fusion is large (Jackson 1967) as is typical of silicates (with the
exception of quartz). Continuous growth prevails when surfaces are atomically rough, because
this signifies the attachment energy is similar for all sites on the surface to which growth
units may attach. Screw dislocation and layer-by-layer growth occur on atomically smooth
surfaces. Attachment energies are high and thus greatly facilitated by ledges formed where
screw dislocations emerge at the crystal surface. In the absence of such defects, a new layer
must be nucleated. Addition of the first growth unit to a smooth face requires substantially
more energy than the addition of subsequent units, which attach at two or more sides. Thus, the
advance rate of a face depends on the frequency of successful attachments of isolated growth
units. Although more common at the high growth rates associated with large undercoolings,
anomalously high concentrations of incompatible elements (i.e., solute trapping) may also
occur in the interface-controlled regime. Elements that are incompatible in the bulk solid may
be stable at the surface if the surface region contains site types not present in the lattice interior
(Watson 1996). If atoms of these solute elements are slow to diffuse away from the advancing
front, they may be trapped. Anisotropy of surface interface site types on different crystal faces,
coupled with slow lattice diffusion is a cause of sector zoning (Watson and Liang 1995).
Through clever selection of melt compositions in the CaO+Al2O3+SiO2 (CAS) system,
Roskosz et al. (2005) investigate the importance of structural and compositional similarity
between melt and crystal in assessing controls on growth rate over a range of undercoolings,
all within the interface-controlled regime. They find that when the crystal and melt have the
same Al/Si ratio, crystal growth rate scales inversely with melt viscosity (given by temperature
as well as composition) regardless of other compositional differences between the phases. This
is interpreted as a reflection that both congruent and incongruent crystallization are controlled
by the rate of bond rearrangements between oxygen and network-forming cations. For crystals
growing from melts of a given viscosity, growth rate varies more strongly as a function of
thermodynamic undercooling than melt-crystal structural similarity.
Diffusion control. Diffusion-controlled crystal growth prevails at large undercoolings,
where the chemical potential driving solidification is large yet lower temperatures and/or lower
H2O contents reduce component mobility. Morphologies of crystals grown in this regime
are anhedral, progressing through a menagerie of forms including tablets, needles, hoppers,
dendrites, and skeletons (Sunagawa 1981) as undercooling increases. These morphologies
develop because protuberances on the corners and edges of crystals penetrate through the
boundary layer that is polluted with rejected components and into an environment slightly
richer in crystal-forming components.
CryStaLLIZatIOn KInetICS In MaGMaS: MethOdS
“Volcano petrology” studies are broadly characterized by their practical objectives. Works
that strive to elucidate the processes of crystallization, test available theories, and define
relationships among variables are emphasized in this section. These investigations aim to
generate reproducible relationships and internally consistent data sets. By contrast, applied
studies attempt to recreate conditions preceding a specific event, and then impose changes
meant to simulate volcanic processes. Selected site-specific research that tests eruption models,
quantifies ascent rates, and/or reproduces natural crystallization conditions during a particular
eruption is not emphasized here, although selected references to such studies are provided in
the sections below.
24 Hammer
Much of the present understanding of crystallization in silicate melts builds upon
experimental work associated with lunar petrology studies, which mapped out the relationships
between cooling rates, growth regimes, crystal sizes, and morphologies in lunar basalt (Lofgren
et al. 1974; Walker et al. 1976; Grove and Bence 1979). Elevated pressure experiments
addressed growth of plagioclase from simple hydrated plagioclase melts (Muncill and Lasaga
1988). Investigations of volatile-rich, granite-forming magmas (Fenn 1977; Swanson 1977;
Swanson and Fenn 1986) concurrently filled out the silicic end of the compositional spectrum.
These formative studies, reviewed thoroughly elsewhere (Dowty 1980; Kirkpatrick 1981;
Cashman 1990) paved the procedural and interpretive ground guiding the recent resurgence in
quantifying crystallization kinetics.
Chemical interactions in complex systems influence nucleation kinetics and growth
rate in ways that are of broad volcanological interest. For example, crystal nucleation rate
and density, crystal size and morphology, and the nature of multicrystalline and multiphase
interactions evolve through time in response to declining undercooling. At present these issues
must be considered through empirical observation by examining controlled natural samples
and performing laboratory experiments.
Starting materials
A variety of starting materials are employed according to the specific study objectives.
Synthetic materials emulate a natural multicomponent melt composition of interest (e.g.,
Martel and Schmidt 2003), or fall between end members of a compositionally simple system
(e.g., Faure et al. 2003a). Lavas are commonly used as starting materials when the process of
interest occurs in a magma chamber, conduit, or planetary surface (e.g., Lesher et al. 1999).
Pyroclastic rocks provide the advantage that the groundmass needs only to be melted and
equilibrium with crystal rims restored in order to match the chemical environment prior to
eruption (e.g., Hammer and Rutherford 2002).
thermal history
Overwhelming evidence demonstrates that nucleation at subliquidus temperatures is
sensitive to the near- or superliquidus thermal history of the material (Walker et al. 1978;
Lofgren 1980, 1983; Davis 2000; Davis and Ihinger 2002). The importance of thermal history
is exemplified by the basalt cooling experiments of Sato (1995), in which a 20 °C difference
in initial melting temperature causes a difference in plagioclase number density upon dwelling
isothermally 100 °C below the liquidus of more than four orders of magnitude. Therefore,
the thermal treatment selected for a particular set of experiments is governed by specific
objectives. Destruction of pre-existing structure by superliquidus heating up to 100 degrees
above the liquidus may be desired in order to evaluate the contribution of heterogeneities and
microscopic substrates in an ensuing nucleation step (London et al. 1998). Alternatively, if
the objective is to correlate crystal growth mechanisms with morphologies, starting materials
may be homogenized at superliquidus temperatures and subsequently seeded with metallic
impurities or mineral phases to provide substrates for heterogeneous nucleation or crystal
growth (e.g., Simakin and Chevychelov 1995; Couch 2003; Faure et al. 2003a; Simakin and
Salova 2004; Larsen 2005). Seeds facilitate identification of new growth if the seed composition
differs appreciably from the mineral of interest. However, if the seed crystals are far from
equilibrium with the surrounding melt, a reaction or dissolution may obfuscate interpretation
and introduce kinetic artifacts not related to growth of the target phase. To achieve results
with utmost relevance to natural volcanic processes, it may be necessary to sacrifice some
of the certainty that comes from isolating the processes of crystal nucleation and growth. A
preferred strategy in this case is to match pre-eruption magma conditions as closely as possible
prior to imposing a decompression or cooling history, and therefore preserve the naturally-
occurring heterogeneities and structure (e.g., degree of melt polymerization and concentration
Experimental Studies of Magma Crystallization 25
of sub-critical clusters). This is the approach taken by Lesher et al. (1999) and Hammer and
Rutherford (2002), among others.
Static vs. dynamic
Experimental procedures follow one of two end members or some combination of both.
“Static” experiments are composed of an initial equilibration period followed by rapid decom-
pression or rapid cooling, executed in a single step or ramp that is short compared to the duration
of the subsequent dwell period. This approach is conceptually linked to irreversible processes
defined by classical thermodynamics; the response of the system against constant externally ap-
plied conditions is monitored. If these experiments are decompressions, they are called “isobar-
ic” and if cooling “isothermal,” because the response of the system to the disequilibrating event
(either the pressure or temperature change) occurs at constant externally-controlled conditions.
After the desired anneal period, material is quenched to preserve a snapshot of the system at
elevated run temperature. A series of snapshots reveals the changes through time of the rates
of crystal growth and nucleation, changes in crystal morphology, and the evolution of texture.
The functional relationship between degree of undercooling and crystallization rates can only
be determined from short-duration static experiments, ideally using the zero-time extrapolation
of time series data. The sample undercooling thus applied is generally easily reconciled with
theory, but is probably dissimilar from what occurs during most natural processes.
At the other end of the spectrum, “dynamic” experiments include continuous changes in
temperature and/ or pressure, such as multi-step decompression, constant-rate cooling, and
various combinations of dwell and ramp segments. Conceptually, these experiments may ap-
proach the hypothetical reversible thermodynamic process, although in practice (as described
below) near-equilibrium along the path is far from assured. Continuous changes in temperature
or pressure impose continually changing undercoolings, the magnitudes of which at any instant
depend on the rate of process variable change and the crystallization kinetics at all previous
steps as well as the kinetics at the given conditions. Thus, the undercooling through time is
likely to be a sawtooth function, with unknown (and possibly varying) tooth height and spacing.
The results of dynamic experiments are less easily compared with theoretical predictions than
those of static experiments. However, dynamic experiments can be designed to more closely
match the conditions during natural magma cooling or decompression, and are thus suited to
applied research.
Constant-rate vs. variable-rate cooling
Performing constant-rate cooling experiments is a trivial procedure with modern
programmable furnace power controllers. Because the cooling rate history does not vary
arbitrarily, the results are not tied to a specific magma reservoir geometry or heat flow
mechanism. Despite being comparatively general, the thermal profiles of constant-rate or
“linear” cooling experiments have the disadvantage of not simulating natural cooling profiles.
Linearly decreasing temperature may be a poor approximation of natural magma cooling, since
the rate of heat transfer between the magma and surroundings strongly decreases over time
and the release of latent heat is not constant through the crystallization interval. An implicit
assumption of the constant-rate cooling approach is that a suite of constant-rate cooling runs
spanning at least an order of magnitude in rate probably captures essential differences in the
progress of crystal nucleation, crystal growth, and melt differentiation.
Bowles et al. (2007) investigated the morphological effects of variable-rate cooling by
discretizing a thermal history representing cooling at arbitrary depth within a conductively
cooling lava flow, incorporating release of latent heat using the solution to the Stefan problem
given by Turcotte and Schubert (1982). Cooling rates thus progressively decreased from 53 °C
h−1 to 2.2 °C h−1, with a time-integrated average above 600 °C of 19 °C h−1. The morphology
and aspect ratio of pyroxene crystals grown in this run were compared with those in constant-
26 Hammer
rate experiments performed at 2.8, 5.7, 19, 72, and 230 °C h−1. The variable-rate experiment
qualitatively and quantitatively matched the constant-rate cooling experiment performed at
72 °C h−1. This provisional test suggests that the early rapidly-cooled portion of the thermal
profile exerts dominant control over pyroxene texture. Constant-rate cooling experiments
therefore probably mimic the initial stages of magma crystallization in natural settings.
BaSaLt COOLInG exPerIMentS
nucleation controls texture
Dynamic cooling experiments on lunar and basaltic meteorite compositions in the mid
to late 1970’s produced a series of observations with profound implications. Since then, these
results have been widely reproduced, quantified for other compositions and experimental
techniques, and are now (largely) rectified with theory. One important observation was that
continuous, slow cooling at a constant rate (or a declining rate), can give rise to phenocrystic
textures (Lofgren et al. 1974; Walker et al. 1976; Grove and Bence 1979), thus shattering the
assumption that large euhedral crystals form during slow cooling and that anhedral groundmass
crystals require a second stage of crystallization at faster cooling. These observations have
been replicated recently in a study of Mars analog basalt (Hammer 2006). Several factors are
suggested to control whether a given melt will crystallize to produce a porphyritic texture: the
degree of superheating above the liquidus (Lofgren and Lanier 1992), the quantity of phases
appearing at the liquidus temperature (Lofgren et al. 1974), and the differential suppression
of phase appearance temperatures among phases (Walker et al. 1978). If melts are structurally
unrelaxed upon cooling, the delay in vigorous nucleation allows supersaturation to increase so
that first-formed crystals grow very rapidly; conversely a granular texture results from cooling
from near-liquidus temperatures because many clusters quickly achieve critical size and begin
to grow. Thus, the tradeoff in nucleation and growth rate leading to qualitatively different
texture is a manifestation of the aforementioned mass balance relationship, with nucleation
rate determined by melt structure. A related observation is the occurrence of heterogeneous
nucleation leading to fasciculate textures (intergrowths) when superliquidus heat treatments
affect one phase more than another. Destruction of plagioclase clusters during superheating of
Stannern meteorite basalt examined by Walker et al. (1978), for example, may have led to the
development of prominent plagioclase-pyroxene intergrowths, whereas subliquidus starting
temperatures produce intergranular to subophitic textures because plagioclase nucleation
occurs in the absence of a pyroxene substrate. In summary, if near-critical clusters are not
available at the start of cooling, then the degree of undercooling at which crystals nucleate
determines growth rate and crystal morphology; if crystals nucleate from a population of near-
critical clusters, then the initial number density and spatial distribution of these clusters is
more important than cooling rate for determining ultimate morphology and size (Walker et
al. 1978).
A related tenet of basalt crystallization developed initially for olivine (by Donaldson
1976, and replicated for other phases: Walker et al. 1978; Lofgren and Lanier 1992) is that the
amount of super-liquidus heating is (a) inversely proportional to the nucleation induction time
upon undercooling by a given amount, and (b) proportional to the magnitude of appearance
temperature suppression at a given time following the application of supersaturation. In
Donaldson’s (1976) experiments, 60 °C of undercooling is required for olivine nucleation
following a sudden drop in temperature from 38 °C above the olivine liquidus, while olivine
doesn’t appear at any temperature within 2 hours following a drop from 118 °C above the
liquidus. The superheating effects diminish with longer subliquidus dwell times. Above ~4 h,
the appearance temperature is suppressed similarly (<40 °C) by a wide range of superheating,
18-118 °C.
Experimental Studies of Magma Crystallization 27
Phase appearance suppression is also observed in constant-rate cooling experiments, in
which the magnitude of suppression increases with increasing cooling rate (Grove 1978; Grove
and Bence 1979). The formative (1970’s) experiments suggested that olivine is least-affected
by superheating and rapid cooling rates. In contrast, Lesher et al. (1999) observe olivine appear-
ance temperature in MORB suppressed to a greater extent (190 °C at 100 °C h−1) than plagio-
clase (30 °C at 100 °C h−1). Moreover, the lag time between the predicted and actual appearance
time estimated for plagioclase (3600 s at 10 °C h−1 cooling rate) is substantially less than for
olivine (12,000 s at 10 °C h−1). The contrast with lunar compositions is interpreted (Lesher et
al. 1999) to represent different near-liquidus melt structures and therefore different nucleation
energy barriers to formation of plagioclase vs. olivine. That is, nucleation of plagioclase may
be facilitated by structural similarity between plagioclase and melt and perhaps σplagioclase-melt
< σolivine-melt in MORB. Presumably, the reverse mineral-melt structural relationship exists in
lunar basalts.
Growth rates decay with time
A feature shared by all crystallization kinetics studies is a power law relationship
between crystal size (assessed as the length or width of the largest crystals or the widths of
dendrites) and cooling rate. Similar behavior is observed among natural cooling environments
in which crystallization durations are well constrained (Cashman 1993). This trend is widely
reported in the basalt dynamic cooling literature (Walker et al. 1976; Lesher et al. 1999;
Kohut and Nielsen 2004; Conte et al. 2006), and continues to be interpreted as evidence that
cooling rate is a paramount factor (after superliquidus heat treatments) controlling crystal
growth rate. The relationship between crystal size and cooling rate may be recast as the net
growth rate (defined below) versus crystallization interval duration if sufficient experimental
information is reported. Data thus converted for mafic systems crystallizing dynamically and
isothermally, and hydrous felsic systems crystallizing in response to multi-step and single-step
decompressions, are plotted in Figures 7 and 8, respectively. Growth rates vary by at least two
orders of magnitude across the data sets for any given crystallization time within the mafic and
felsic groupings (probably due to differences in experimental or measurement technique as
well as intrinsic differences arising from crystal and melt compositions).
It is important to note that growth rates are typically computed as a characteristic size of
the observed population divided by run duration, rather than by observing progressive growth
of individual crystals; one calculation of characteristic size is to find the square root of the
quotient of area fraction and area number density (both obtained from analysis of backscattered
electron images; e.g., Hammer and Rutherford 2002). Another method is to take the average
width or length of the five largest crystals (e.g., Walker et al. 1976). Thus, an increase in the
quantity of crystals in the system could produce unvarying or declining characteristic size
through time. A monotonic (slope = −1) relationship on a log-log plot of growth rate and
crystallization time would be generated if size were to remain constant through time. Datasets
exhibiting slopes < −1 represent systems in which the characteristic size decreases with
increasing crystallization time, usually due to appearance of new crystals; slopes > −1 represent
systems in which the characteristic size increases with crystallization time (whether or not new
crystals have been added). The majority of datasets contain deviations from the slope = 1 trend
lines, revealing changes in size (and thus net growth rate) with time. Slopes describing the
changing growth rate among plagioclase forming in the mafic systems range from −0.5 (Grove
1978) to −0.8 (Kohut and Nielsen 2004). In most cases, long duration experiments reflect
slow cooling rates (< 10 °C h−1), and these experiments produce crystals that are larger than
in the quickly cooled experiments of short overall duration. For example, the large crystals
produced in the Kohut and Nielsen (2004) study are the result of protracted dwell times after
a cooling stage. The characteristic crystal size in these experiments is not sensitive to the rate
of initial cooling. In the felsic systems, the range in slopes of individual datasets is from -0.5
28 Hammer
An30
-3
An10
log plagioclase growth rate (mm/s)
20-25 oC
25 oC
-4
11 oC
-5
20-400 oC
40-350 oC
-6 6m
m
-7 5m
icr
on
s
-8
2 3 4 5 6 7 8
log crystallization time (s)
Figure 7. Crystal growth rates with time in constant-rate cooling (except as indicated) experiments.
Crystallization time depends on both cooling rate and quench temperature. Diamonds: plagioclase in basalt
(Kohut and Nielsen 2004). Squares: plagioclase in lunar ferrobasalt (Grove 1978). Circles: plagioclase in
MORB (Grove et al. 1990). Triangles: Ca-pyroxene in lunar picritic basalt (Walker et al. 1976). Vertical black
bars: plagioclase growing from H2O-saturated plagioclase melt of same (anhydrous) composition at 200 MPa
at the range of undercoolings indicated; time information not available, the x-axis location is arbitrary (Muncill
and Lasaga 1988). Gray shaded boxes: isothermal experiments, crystallization time is the dwell period and
∆T as indicated. White plus symbols on black background: Ca-pyroxene crystals in hydrous hawaiite melt
(Simakin et al. 2003). x’s: plagioclase growing from plagioclase melt of the same composition, 1-atm (Muncill
and Lasaga 1987). Dashed lines show trends produced by zero net growth rate (i.e., constant mean size through
time) for sizes bounding those observed in the experimental datasets.
-5.5
pla
gio
c las
-6 e in
ma
fic
ma
log plagioclase growth rate (mm/s)
-6.5 gm
a s
-7
70
m
-7.5 icr
on
s
-8
-8.5
0.5
m
icr
-9 on
s
-9.5
2 3 4 5 6 7
log crystallization time (s)
Figure 8. Plagioclase growth rates with time in single-step (filled symbols) and multi-step (open symbols)
decompression experiments. Crystallization time depends on decompression rate and quench pressure.
Circles: plagioclase in Pinatubo dacite (Hammer and Rutherford 2002). Diamonds: plagioclase in synthetic
haplogranite (Couch 2003). Triangles: plagioclase in synthetic Soufriere Hills matrix melt (Couch et al. 2003).
Squares: anorthite seeds in Aniakchak rhyodacite (Larsen 2005). Pluses: plagioclase internally nucleated in
Aniakchak rhyodacite (Brugger, unpubl.). Dashed lines show trends that would be produced if size remained
constant through time. Shaded region shows a portion of the plagioclase growth rate field for mafic systems.
Experimental Studies of Magma Crystallization 29
in seeded crystal growth experiments (Larsen 2005) and slow decompressions (Brugger and
Hammer, unpubl.) to < −0.9 in rapidly decompressed hydrous magma (Brugger and Hammer,
unpubl.; Couch 2003), probably reflecting the varying dominance of nucleation in the overall
crystallization mechanism. Of course, growth rate must eventually decline (given sufficient
time and permissible kinetics) because as the system approaches equilibrium and degree of
undercooling decreases, growth rates must (by definition) approach zero. Declining contrast
between compositional zones in experimentally grown pyroxene crystals suggests strongly
declining growth rates over time (Schwandt and McKay 2006).
Interestingly, the minimum growth rate observed in static experiments is >2 orders of
magnitude faster than minimum growth rate observed in dynamic experiments, and static
experiments at low to moderate undercoolings (<30 °C) exhibit growth rates that are large
compared to growth rates calculated after an extended period of slow (dynamic) cooling (Fig.
7). The observation that growth rates dramatically decay with time while appearing relatively
insensitive to the magnitude of initial undercooling—while theory dictates undercooling
exerts the primary control over growth rate—seems to present a contradiction. The issue
has implications for many volcanological applications, such as interpretation of crystal size
distributions (e.g., Armienti 2008), in which growth rate is commonly assumed constant. The
most likely resolution lies in the fact that growth rates have not been directly measured at very
small degrees of imposed undercooling. Small growth rates evident in the long-duration static
and slow rate-of-change dynamic studies (Figs. 7 and 8) suggest that natural and experimental
systems achieve and maintain very small undercoolings. Future experimental work at low
undercoolings will be challenging. As the magnitude of undercooling is reduced, it becomes
increasingly difficult to maintain sample temperature and/or pressure within the necessary
tolerances, rapid quenching becomes increasingly important, and decreases in growth layers
smaller than a few microns are imperceptible by analysis methods short of high resolution
TEM. Ostwald ripening experiments (e.g., Cabane et al. 2001, 2005) are a promising means of
addressing growth in this context, as are extremely experiments of extremely long duration.
undercooling and the approach to equilibrium
The compositional effects of rapid crystal growth at moderate to large undercoolings
have been examined by Grove and Bence (1979) and Hammer (2006), among others. While
details differ among phases, Fe-Mg partitioning between mafic phases and melt depart from
equilibrium as cooling rate increases, having major implications for departure from equilibrium
liquid lines of descent (Grove and Bence 1979). Partitioning of minor elements, e.g., Al and
Ti in clinopyroxene, are also dependent upon crystal growth rate as modulated by cooling
rate (Walker et al. 1978; Grove and Bence 1979), although the Al content of pyroxene is
also affected by the Al2O3 content of the melt and whether plagioclase appearance has been
suppressed due to rapid cooling. Further compositional effects are discussed below in the
context of melt inclusion entrapment.
A kinetic effect worthy of mention is that the time scales for solidification increase with
decreasing rates of cooling. Solidification of mid-ocean ridge basalt occurs ~50× faster in
magma cooled rapidly (1000 °C h−1) than in magma cooled slowly (10 °C h−1; Lesher et al.
1999). A qualitatively similar result is obtained in an investigation of plagioclase, olivine, and
pyroxene crystallization in response to fast and slow cooling of three natural alkalic lavas (Conte
et al. 2006). Rapidly cooled samples more closely approach equilibrium phase proportions
than the slowly cooled experiments. Within limits, higher undercoolings imposed by rapid
cooling enhance the rate of crystallization. Clearly, the trend does not continue to extremely
rapid cooling rates, as it would be impossible to quench a melt to glass. Finding the cooling
rate at which crystallization reaches optimum efficiency may be the dynamic equivalent to
mapping out individual nucleation and growth rate maxima using static experiments.
30 Hammer
Crystallization in dynamic environments
Shaking and shearing. Novel approaches to nucleation include the Bartels and Furman
(2002) examination of the effects of ultrasonic vibration frequencies on the nucleation of olivine
and plagioclase in natural high-alumina basalt, and the Kouchi (1986) study of plagioclase
and pyroxene nucleation in basalts undergoing shear deformation. In the Bartels and Furman
(2002) study, beads of pressed powder were placed into a gas mixing furnace at sub-liquidus
run temperatures then either held or vibrated at one of two frequencies (1.2 MHz, or 150
kHz) for 21-38 h. Samples were removed from the furnace and placed in a stream of air to
quench. Interestingly, although the phase proportions were similar in all experiments of a given
temperature, the sizes and morphologies of plagioclase differ dramatically depending on whether
they were subjected to vibration; olivine texture did not differ among samples. Plagioclase
crystals in the static charges were comparatively large and of high aspect ratio while the vibrated
charges contained smaller, and nearly equant plagioclase crystals of high spatial number density
(Fig. 9a,b). Crystal size distribution plots are linear in static experiments and concave up in the
-12
c. d.
Frequency
-14
-16
ln (population density, um-4)
e.
-18
Frequency
-20
-22 crystal width/ length
-24
-26
200 um
-28
0 50 100 150 200 250 300 350 400
size (um)
Figure 9. qualitative and quantitative results of static and dynamic (150 kHz) basalt crystallization
Figure 9.
experiments. BSE images (a and b, same scale) reveal qualitative differences in texture; plagioclase CSDs
from static (filled diamonds) experiments quantify slightly coarser grain size than in dynamic (open circles)
experiments; frequency histograms of plagioclase crystal aspect ratio (d and e) quantify morphological
differences. [From Bartels and Furman (2002), Am Mineral, Vol. 87, Figs. 4 and 5, pp. 217-226.]
vibrated experiments (Fig. 9c), reflecting the difference in number frequency at the smallest
size classes. Vibration may increase the density of nuclei by assisting structural relaxation.
Applicability of the results to natural volcanic systems depends on whether the observed effects
are similarly important in the lower frequency range of volcano tectonic seismicity.
Kouchi (1986) applied shear forces to a disc of basaltic melt held between cylinders
rotating either in the same direction at the same rate (static experiments), or in opposite
directions at 9.5-70 rpm (dynamic experiments). The nucleation process and resulting textures
are clearly affected by the induced flow in the dynamic experiments. At a given undercooling,
the period of transient nucleation decreases as rotation rate increases. At a given observation
time, the magnitude of phase appearance suppression decreases as rotation rate increases. Both
findings suggest that nucleation is enhanced in the sheared melts, an observation previously
noted in the industrial crystallization literature (Mullin 1980). In fact, at a given differential
Experimental Studies of Magma Crystallization 31
8
a. static
c.
7.5
log N (cm-3)
7
nucleation density
b. dynamic 6.5
5.5
0 50 100 150
supercooling (deg)
Figure 10. qualitative and quantitative results of basalt crystallization experiments of Kouchi et al. (1986).
Figure 10.
BSE images (a and b, same scale) show textural effects of magma shearing caused by rotation (infrared
furnace setup shown in c, inset) at 75 rpm. (c) Population density of plagioclase and pyroxene crystals
grown in static (filled diamonds) and dynamic (open circles; 48 rpm) experiments. Curves merely guide
the eye through the data points; after Kouchi et al. (1986).
rotation rate, nuclei density is greater in the dynamic experiments (Fig. 10). Kouchi (1986)
suggests that nucleation enhancement in the dynamic experiments may result from a reduction
in the kinetic barrier (∆GD of Eqn. 1) brought about by stirring and advective transport. At
moderately large undercoolings (≥45 °C), the morphologies of plagioclase and pyroxene in the
dynamic experiments are faceted and acicular, whereas crystals in the static experiments are
dendritic to spherulitic. The shift in morphology suggests that the rate limiting step of crystal
growth changes from diffusion-controlled to the interface-controlled. Differential motions of
crystals and melt in the dynamic experiments could disrupt compositional boundary layers,
promoting interface-controlled growth.
Olivine morphology and growth mechanisms. Examinations of growth and development
of anhedral olivine crystals are abundant in the 1-atm basalt cooling literature (Donaldson et al.
1975; Donaldson 1976) and have been the topic of recent intensive work in the CaO+ MgO+
Al2O3+ SiO2 (CMAS) system (Faure et al. 2003a,b, 2006, 2007; Faure and Schiano 2004, 2005).
Faure et al. (2003a) performed dynamic cooling experiments on simple basalt analogs (51.3 wt%
SiO2) at 1-1890 °C h−1 on crystal-free materials that had been fused and homogenized 50 °C
above the liquidus then seeded with Pt impurities and remelted at 22-124 °C above the liquidus.
The quench temperatures ranged from 24-356 °C below the liquidus, a difference parameter
defined as the nominal undercooling to distinguish it from the undercooling, sensu strictu (i.e.,
applied by an instantaneous drop in temperature). Olivine nucleates heterogeneously on the Pt
impurities, and grows into crystals defined as polyhedral, tabular, hopper, “baby swallowtail,” or
swallowtail (Fig. 11) depending on nominal undercooling and cooling rate. At low cooling rates,
crystal morphology is a function of both variables, while above 47 °C h−1 crystal morphology is
sensitive only to the degree of nominal undercooling.
An important result of the crystallographic investigation is that during formation of hopper
crystals, growth is slowest along [010] and fastest along [100]. Atom attachment preferentially
occurs on corners and edges of (100), giving rise to melt-filled embayments that characterize
the hopper morphology. These crystals are hourglass viewed along [010] and hexagonal along
32 Hammer
Polyhedral crystal
cooling rate (oC/h)
Tabular crystal
Hopper crystal
Baby Swallowtail shape
{Swallowtail
Rods:
shape
-Lattice oliivne
-Chain olivine
-Branching olivine
1 Virtual limit separating the tabular and hopper domains
2 Virtual limit separating the hopper and baby swallowtail domains
3 Virtual limit separating the baby swallowtail and swallowtail domains
G
Degree of undercooling (oC)
E
F
Figure 11. Figure 11. Morphology variations of olivine as a function of cooling rate and undercooling,
emphasizing evolution in the section parallel to (010), after Faure et al. (2003a).
[100]. As nominal undercooling increases, overgrowths appear at the mouths of hopper cavities
and continue to grow rapidly along [100] to create the “baby swallowtail” morphology. The
overgrowths that define the faces are fine dendrites growing rapidly along [101] and [−101].
The resulting morphology is characteristically swallowtailed when viewed along [010] and
rod-like when [010] is in the section plane. The elongation direction of the dendrite-bounding
rhomb shifts from [100] to [010] with increasing nominal undercooling, as the morphology
shifts from hourglass hopper, to hopper with dendritic overgrowths, to dendrites forming
directly on Pt impurities. The authors posit that all other olivine morphologies described in the
literature (e.g., Donaldson 1976; Sunagawa 1981) can be reconciled as various cuts through
these four shape types.
The mechanism of olivine growth, derived from compositional and textural data, is
considered by Faure and Schiano (2005) and Faure et al. (2007). Correspondence of euhedral
morphology and compositional uniformity in the vicinity of crystal-melt boundaries suggest
that olivine growth is controlled by interface attachment mechanisms. Furthermore, growth
by screw dislocation propagation proposed to dominate at low undercoolings (Cabane et
al. 2005), is consistent with the curvilinear outlines of melt inclusions observed by Faure
and Schiano (2005). Compositional gradients in the melt develop at higher undercoolings
when rapid crystallization is required to approach equilibrium, yet component mobility in the
Experimental Studies of Magma Crystallization 33
melt limits the growth rate of crystals. Growth is more rapid in regions of high surface area
such as crystal edges and corners, leading to growth by layer spreading (layer nucleation)
and development of hopper crystals. The transition to dendritic morphologies is less well
understood, but may reflect a transition in mechanism to that of continuous growth. The platy
spinifex olivine textures characteristic of komatiites are interpreted by Faure et al. (2006)
as a special case of growth during slow cooling within a steep thermal gradient (7-35 °C
cm−1). Elongate dendritic morphologies suggest diffusion-control, further substantiated by
the presence of compositional boundary layers between dendrites. Crystal growth occurs
dominantly perpendicular to isotherms because heat is most efficiently transported along the
fast-growing [100] axis.
Olivine growth processes during dynamic cooling experiments are used to interpret
melt inclusion entrapment conditions in natural sparsely phyric Mid-Atlantic Ridge basalt
(Faure and Schiano 2004). Cycles of cooling and mild (subliquidus) re-heating are needed
to form chains of melt inclusions: growth at high undercoolings produces hopper cavities,
and subsequent faceted growth at lower undercooling seals them off (Fig. 12). Of course, the
applicability of forsterite growing in CMAS to natural basalts is contingent upon ideal Mg-Fe
mixing in olivine and absence of interface effects arising from presence of Fe.
Figure 12. Formation of closed olivine hopper crystals in the section parallel to (010), with schematic il-
lustration of the thermal fluctuations necessary to trap a chain of inclusions. Photomicrograph B′ shows a
hopper crystal obtained at cooling rate of 422 °C h−1. Photomicrograph C′ shows a hopper crystal formed at
437 °C h−1, and inclusion entrapped after reheating for 5 minutes to subliquidus temperature. [Used by per-
Figure 12.mission of Elsevier, from Faure and Schiano (2004), Earth Planet Sci Lett, Vol. 220, Fig. 7b, p. 331-344.]
FeLSIC MaGMa deCOMPreSSIOn exPerIMentS
effective undercooling
Formative experimental work showed that magmatic H2O degassing causes crystallization
of anhydrous phases (Fenn 1977; Swanson 1977; Muncill and Lasaga 1988). Because H2O
solubility is a strong function of pressure, magma decompression provides a driving force for
solidification that is analogous to cooling of an isobaric system. An effective undercooling
34 Hammer
250
200 Figure 13. Plagioclase-in curve and
pressure (MPa)
estimated solidus curve for melt of Pi-
150 natubo matrix dacite used in decom-
pression experiments of Hammer and
pl l
solid Rutherford (2002). ∆Teff is defined at
iqu
100 any given pressure as the plagioclase
idu
us
liquidus temperature - experimental
s
50 ΔT eff temperature (780 °C). Circles repre-
sent experimental pressures examined
in single-step decompression experi-
0 ments.
600 700 800 900 1000 1100
o
T ( C)
Figure 13. (∆Teff) is defined as the difference between the actual system temperature and the liquidus
temperature (TL−T) using the equilibrium PH2O-T phase diagram for the system of interest (Fig.
13). Practical applications of site-specific crystal nucleation and growth kinetics experiments
(a) establish the relevance of syn-eruptive, devolatilization-driven crystallization in erupting
arc magmas (Geschwind and Rutherford 1995), (b) constrain magma ascent rates during
explosive and effusive eruptions (McCanta et al. 2007; Rutherford 2008), (c) determine time
scales of magma mixing from strongly zoned plagioclase crystals (Larsen 2005; Martel et al.
2006), (d) consider the implications of crystal nucleation on bubble coalescence in volcanic
bombs (Simakin et al. 1999), and (e) challenge the notion that magma flow rates feeding
explosive eruptions exceed those of effusive eruptions (Castro and Gardner 2008). Given
appropriately contextualized natural samples and targeted experiments, it is now possible to
infer detailed multi-stage magma ascent histories from groundmass crystal textures (Nicholis
and Rutherford 2004; Szramek et al. 2006; Suzuki et al. 2007; Blundy and Cashman 2008).
Establishing baseline crystallization kinetics of quartz, plagioclase, and alkali feldspar
was the objective of a triad of process-oriented studies (Hammer and Rutherford 2002;
Couch 2003; Martel and Schmidt 2003). The key experimental variables in the Hammer
and Rutherford (2002) study were (a) final pressure, (b) dwell time at final pressure, and (c)
decompression path. The pressure-time path consisted of either a rapid decompression in a
single step, or incremental pressure drops in multiple steps. The range of ∆Teff imposed by
the single-step decompressions was 34-241 °C. This series is used to determine the functional
dependence of feldspar nucleation and growth rates on final pressure (assessed following a
fixed dwell time), and to monitor changes in crystallinity and texture through time in response
to a discrete perturbation. Multi-step decompressions monitor the evolution of the system in
response to stepwise decreasing pressure according to a fixed time-integrated decompression
rate of 1.2 MPa h−1.
decompression history
Static experiments. Key results of the single-step experiments include: (1) quantification
of the bell-shaped crystal nucleation and growth curves as functions of Pf; shown in Figure
14. (2) Decay of growth rates (computed as the characteristic size divided by run duration)
over time by 3 orders of magnitude between 0.33 and 931 h (Fig. 8). (3) Textural maturation
that includes disappearance of microlites at long dwell times and initially low ∆Teff. (4)
Maintenance of equilibrium at low to moderate ∆Teff, and departures at high ∆Teff.
The general forms of the nucleation and growth rate dependence on ∆Teff obtained at
168 h dwell times (Fig. 14) provide an explanation for the variable approach to equilibrium.
Experimental Studies of Magma Crystallization 35
10.000
Hammer and Rutherford 2002
nucleation rate (mm-3 s-1)
Couch 2003
1.000
Figure 14. Plagioclase nucle-
ation rates (a) and growth (b)
0.100 8h rates following isothermal,
single-step decompression +
168 h dwell periods of 8 and 168 h
from Hammer and Rutherford
0.010
(2002), open symbols, and
168 h Couch (2003), filled symbols.
(a) For a given dwell time,
0.001 nucleation rates vary strongly
1.00E-06
8h with final pressure, display-
ing a maximum at low pres-
1h
sures. Nucleation rate decays
growth rate (mm s-1)
rapidly with increasing dwell
3h
1.00E-07 time (Couch 2003). Note mis-
match in nucleation rates at 168
27 h h dwell time. (b) Growth rates
display a maximum at interme-
168 h
diate final pressures and also
1.00E-08 decay with increasing dwell
time. Growth rates at 168 h are
168 h
similar in the two data sets.
931 h
1.00E-09
0 50 100 150 200
final pressure (MPa)
Equilibrium is achieved at small to moderate ∆Teff chiefly by growth of existing plagioclase and
Figure 14. quartz crystals and sparse nuclei. Departure from equilibrium crystallinity occurs in the ∆Teff
interval between maximum growth rate and maximum nucleation rate. At large ∆Teff (150-250
°C) nucleation rate increases and microlite formation becomes the dominant crystallization
process, but it is insufficient to maintain equilibrium in the absence of rapid crystal growth.
Dynamic experiments. Striking contrasts between quasi-continuous decompression and
single-step decompression run products underscore the importance of decompression history
in controlling crystal texture (Fig. 15). For example, samples decompressed to final pressures
>40 MPa in multiple steps were much less likely to reach equilibrium crystallinity than samples
decompressed in a single step. Because crystallization during each decompression step brings
the system closer to equilibrium, stepwise decompression produces smaller instantaneous
∆Teff than is typical in the single-step decompressions. That the slowly-decompressed samples
were less likely to achieve chemical equilibrium than their rapidly-decompressed counterparts
is interpreted to reflect this generally lower degree of undercooling in the multistep run
throughout the experiment. This observation runs contrary to conventional wisdom, which
would suggest slow decompression is more likely to promote chemical equilibration than
rapid decompression. However, these results and interpretations parallel those of Lesher et al.
(1999) in describing the approach to equilibrium of cooling MORB. Kinetic effects, such as
the functional dependence of crystal growth on ∆Teff and the presence of the nucleation energy
barrier, manifestly control progress toward chemical equilibrium.
Seeded experiments. Kindred static and dynamic experiments were performed in the
simple system SiO2-Al2O3-CaO-Na2O-K2O; haplogranite with 73 wt% SiO2) by Couch
36 Hammer
effective undercooling (oC)
6 6 5
26 18 12 93 70 48 34 19 -7
1
e
e q u i l ibr i u m c u r
0.8
mass fraction crystallized
single-
step
0.6
ve
0.4
0.2 multi-
step
0
0 50 100 150 200 250
final pressure (MPa)
Figure 15. Comparison of equilibrium crystal content with experimental samples. Single-step
decompressions followed by 168 h dwell times (closed circles) result in greater crystal contents than
Figure
multi-step 15.
decompressions at similar integrated decompression rates (open squares) for all final pressured
examined, although both paths fail to closely approach equilibrium near the solidus (i.e., below ~40 MPa);
after Hammer and Rutherford (2002).
(2003). These experiments also examined decompression-induced crystallization by single-
step and multi-step decompression paths, but differed in starting material type and preparation.
After high temperature fusion and homogenization, the starting materials were held at initial
conditions of H2O -saturation at 160 MPa and 875 °C for 16 h. Charges were seeded with
1 vol% andesine feldspar, which were not in equilibrium with the bulk composition or any
subsequent melt along the liquid line of descent. Single-step decompressions imposed ∆Teff
of 35-167 °C, while the incremental decompressions in the MSD set simulated a variety of
average decompression rates, from 0.1-110 MPa h−1.
Couch (2003) obtained a similar functional dependence of growth rate on Pf (Fig. 14b) to
that of Hammer and Rutherford (2002), and nucleation and plagioclase growth rates steeply
declined with increasing dwell time (Fig. 14b). However, several differences are noteworthy:
(a) a 4 h nucleation lag time and cessation of nucleation after 8 h, (b) nucleation rates up to 40×
lower in comparable-duration experiments, and (c) failure to reach equilibrium crystallinity
at any final pressure along any decompression path. By analogy with the basalt cooling
literature, the differences in nucleation rates obtained by Couch (2003) in comparison with
Hammer and Rutherford (2002) likely arise from the use of synthetic starting materials and a
homogenization time that may have been insufficient for melt relaxation and formation of the
equilibrium cluster size distribution. The absence of equilibrium crystal substrates may have
inhibited growth at low ∆Teff. This inference is supported by the inference that seed crystals
partially dissolved in the first 4-8 h. In combination, these factors affecting the time-dependent
response of nucleation and crystal growth to application of ∆Teff may explain why equilibrium
crystallinities were not obtained in the Couch (2003) experiments, despite similar locations of
nucleation and growth rate maxima with ∆Teff for a given snapshot in time (Fig. 14).
Synthetic starting materials. Martel and Schmidt (2003) studied crystallization in H2O-
saturated synthetic rhyolite, focusing on both melt devolatilization and crystallization due to
multi-step decompression. The starting material was patterned on melts trapped in Soufriere
Hills (Montserrat) andesite crystals, normalized to the nine most abundant oxide components.
Experimental Studies of Magma Crystallization 37
After high temperature fusion and homogenization, the starting melts were held at 860 °C and
either 150 MPa or 50 MPa for 7 days, after which they were decompressed in multiple steps
to final pressures of either 50 or 15 MPa (respectively). Thus, the high-pressure (150 - > 50
MPa) and low-pressure (50-15 MPa) portions of magma ascent were examined separately. The
integrated decompression rates imposed by incremental stepwise decompression ranged from
3.6 - 3.5 × 106 MPa h−1.
Consistent with the findings of Hammer and Rutherford (2002) and Couch (2003),
decompression-induced crystallization at high pressure is dominated by growth of exsiting
crystals, while nucleation is the dominant process at lower pressure (higher ∆Teff). The
compositions of plagioclase, orthopyroxene, and melt strongly depend upon both the pressure
range and decompression rate. For example, plagioclase crystals forming at high pressures
Equilibrium
Figure 16. Plagioclase composition (a) and volume percentage (b) in equilibrium (upper panels) and multi-
step decompression (lower panels) experiments. Experiment pressure (MPa) is indicated by the equilibrium
Figure
symbols. 16.
Crystallization intervals (i.e., the duration of decompression) are as indicated by each symbol
in the lower panels. Solid and dotted ovals enclose the equilibrium values for the high pressure (closed
circles) and low-pressure (x’s) experiments, respectively. After Martel and Schmidt (2003).
are more anorthitic than those growing at lower pressure, and composition was only slightly
affected by varying decompression rate (Fig. 16) Pressure and decompression rate were
equally important controls over the approach to the equilibrium crystal contents, with closest
approach occurring in high pressure runs decompressed <360 MPa h−1 (0.1 MPa min−1; Fig.
16). Low pressure runs and faster decompressions at high pressure trailed equilibrium due
to insufficient plagioclase crystallization. An additional insight includes the observation of
incomplete exsolution of H2O from the melt in samples decompressed >3.6 105 MPa h−1.
texturaL quantIFICatIOn OF exPerIMentaL run PrOduCtS
Early work laid the qualitative groundwork for comparative petrography, with
development of precise morphological terminology linked to a theoretical understanding of
growth (Donaldson 1976; Sunagawa 1981). In the absence of temporal information for the
vast majority of natural volcanic samples, the application of experimental studies is through
the comparison of textures, quantified typically by crystal size (Marsh 2007) and shape.
38 Hammer
Unfortunately, few methods and metrics have been validated experimentally to ensure that rate-
dependent parameters are interpreted correctly. Investigation of several methods, including
CSD analysis and 3D crystal aspect ratio, are described below.
Crystal size distribution analysis
Basalt cooling experiments. zieg and Lofgren (2006) performed dynamic cooling
experiments on synthetic chondrule compositions seeded with natural olivine, and report
CSDs for six temperatures along the constant-rate (92 °C h−1) cooling trajectory between
1545 and 1176 °C. Seeds were initially not in stable equilibrium with the melt at the onset of
cooling; in fact most were destroyed during the 12 minute homogenization interval. However,
overgrowths on the surviving seeds indicate stabilization of olivine within the first 29 minutes
of cooling. Vertical zonation of crystal texture within the sample beads is extreme. Equant
seed-cored crystals accumulate at the base, and skeletal overgrowths penetrate the cumulate
layer toward the top of the bead. Neither of two bulk characterization parameters, mean crystal
length and crystal number density, vary systematically with quench temperature. In fact,
subtraction of skeletal crystals from the CSDs revealed no apparent difference in the CSDs of
cumulate crystals among the runs (Fig. 17). Therefore, traditional analysis of evolving CSD
slope and intercept to the experimental charges was not applied to extract characteristic growth
and nucleation rates using the known run times. However, separating results from the cumulate
zone and the skeletal zone, subtle differences emerge from examination of the time series of
binned population density data (Fig. 18) that are modeled numerically, taking the 29-minute
experiment as a baseline and using two endmember sets of assumptions. Evolution of crystal
sizes within the cumulate zone is modeled assuming zero-nucleation rate and constant (i.e.,
size and time-independent) growth rates of 2 × 10−7, 6 × 10−7 and 2 × 10−6 mm s−1. The
intermediate rate, which produces a similar frequency of small crystals, is best-fitting in view
of the greater statistical certainty of population density in the smallest bins. Crystallization in
the skeletal zone is modeled assuming constant growth rate (2 × 10−5 mm s−1) and accelerating
nucleation rate [I(t) = 9 × 10−6 exp(4 × 10−4t)]. This model predicts the population densities
in the larger size classes, but underestimates the observed frequencies in smaller bins. The
mismatch is attributed to crystal-crystal impingement and the ambiguous distinction between
skeletal and non-skeletal morphologies at the top of the cumulate pile.
A recent study reporting plagioclase CSD analyses for ~40 constant-rate basalt cooling
experiments in which superliquidus heat treatment temperature and duration, cooling rate
12
1176 oC 1325 oC
10 1375 oC 1447 oC
8
ln (n) [mm-4]
Figure 17. Crystal size distributions
of non-skeletal olivine crystals in the
6 cumulate portions of basalt cooling
experimental charges cooled at a
constant rate of 92 °C h−1 to the indi-
4 cated quench temperature; after zieg
and Lofgren (2006).
2
0
0 0.1 0.2 0.3 0.4
L [mm]
Experimental Studies of Magma Crystallization 39
size (mm)
Figure 18. Comparison of modeled
and observed population densities
from zieg and Lofgren (2006).
(a) Calculated trajectories in the
cumulate crystal zone assuming
cumulate zone constant growth rates as indicated
in top box. (b) Calculated size
isopleths in the skeletal zone
assuming constant growth rate (2.0
× 10−5 mm s−1), and exponentially
skeletal zone increasing nucleation rate. The
model matches large size class data,
0.201 but underestimates the frequencies
in smaller classes. [Used by
0.319 permission of Elsevier, from zieg
size (mm) and Lofgren (2006), J Volcanol
0.506 Geotherm Res, Vol. 154, Fig. 8, p.
74-88.]
0.801 mm
Figure 18.
(0.2-3.0 °C h−1), and subliquidus dwell times were varied, is particularly illuminating (Pupier
et al. 2008). The resulting data set (partially presented in Fig. 19) provides by far the most
complete means to date of evaluating common assumptions applied in the interpretation of
CSDs. Highlights include (a) nucleation rate that strongly increases with increasing cooling
rate for a given experiment duration and quench temperature, in accord with expectation; (b)
time-evolution of nucleation rate that cannot be modeled as exponentially increasing (as in
Marsh 1988), but rather sharply decreasing; (c) fanning distributions that cannot be (in this
case) the result of decreasing cooling rate through time (e.g., zieg and Marsh 2002), because
cooling rate is held constant; (d) downturn in the CSD at small sizes and flattening at large
sizes along a given cooling trajectory, provisionally interpreted as synneusis (crystal-crystal
attachment along (010) faces); and finally, (e) crystal growth rate that cannot be modeled
as constant, either because synneusis is important, or because growth rate is crystal size-
dependent (Eberl et al. 2002; Eberl and Kile 2005) or time (i.e., undercooling)-dependent (as
discussed in a previous section).
Basalt heating experiments. Burkhard (2002) compares the CSDs of Fe-Ti oxides,
pyroxene, and plagioclase in billets of Kilauea basalt before and after laboratory heat treatments
at temperatures of 850-934 °C for 43-240 h. Representative CSDs from a fixed position within
40 Hammer
Figure 19. CSDs of crys-
tal widths from experi-
ments designed to evaluate
cooling rate as a variable
controlling plagioclase
texture. Cooling rates are
as indicated in each panel;
experiment design shown
in (b) inset. Vertical dashed
lines represent the loca-
tions of maximum number
densities. After Pupier et al.
(2008).
Figure 19.
the billet are shown in Figure 20. The y-intercepts of all phase CSDs increase after the heating
segment suggesting an increase in the quantity of crystals. The breadth of the size distribution
increases for all phases, with particularly dramatic increases for plagioclase and pyroxene,
perhaps due to size-dependent (proportionate) growth (Eberl et al. 2002). While the slope
of the Fe-Ti oxide CSD is unchanged by the heating step, the initially linear plagioclase and
pyroxene CSDs become kinked upwards. Oddly, conventional CSD interpretive guidelines
would suggest a different set of crystallization conditions and/ or magmatic processes for each
phase (Higgins 1996; Marsh 1998; Higgins and Roberge 2003). For example, the transition
from straight to curved CSDs evident in the plagioclase and pyroxene data imply magma
mixing, while the parallel and linear Fe-Ti oxide patterns may be interpreted to reflect steady
state growth and exponentially increasing nucleation. None of the minerals displays fanning
7
l
6
5
4
plagioclase
Experimental Studies of Magma
3 Crystallization 41
0 2 4 6 8 10 12
size (um)
12 10
11 a. 9 c.
10
8
9
ln (n) 7
ln (n)
8
7 6
6
5
5
4
4
pyroxene Fe-Ti oxides
3 3
0 1 2 3 4
11 b.
10
size (um)
9
Figurereheated
20. at 934 oC
lava
ln (n)
8
Figure 20. Crystal size distributions of pyroxene (a),
7
plagioclase (b), and Fe-Ti oxide (c) crystals located
6
18mm from Kilauea lava sample edge before heating
5
(filled symbols) and after heating (open symbols) to
4
plagioclase 934 °C for 43 h. After Burkhard (2002).
3
0 2 4 6 8 10 12
size (um)
10
9 c.
CSDs 8
that characterize suites of natural samples that are thought to represent evolution during
declining undercooling, as might be expected from the heat treatment schedule applied. The
7
ln (n)
result is not surprising, perhaps, since experiments do not mimic the thermal history of natural
6
magma cooling. Nonetheless, the ambiguous results highlight the need to further verify the
efficacy
5
of the CSD technique for interpreting crystallization processes and time scales.
4
Fe-Ti oxides
Ostwald ripening experiments. Park and Hanson (1999) and Cabane et al. (2005) studied
3
evolution
0 of olivine
1 crystal
2 textures3 in isothermal
4 subliquidus experiments on simple basalts
in the CMAS system. Starting materials in the Park and Hanson (1999) study were prepared
size (um)
by fusing reagents above the liquidus, then reducing temperature and crystallizing for 48 h
Figure 20. subliquidus (1275 °C). The ensuing ripening experiments included a series of isothermal runs
at the homogenization temperature for durations of ~10 minutes to ~10 days. Identical matrix
melt compositions in all experiments indicate that textural changes occur at constant crystal
volume fraction. The width of the frequency distribution of crystal sizes increase by nearly an
order of magnitude over the applied dwell times. This broadening distribution is interpreted
to result from continuous dissolution of crystals smaller than the average size and coarsening
of the largest crystals at any given instant according to an Ostwald ripening process. With
increasing dwell time, the maximum frequencies of the CSDs decrease and the slopes flatten
(Fig. 21), both trends indicating that large crystals grow at the expense of small crystals. On the
observation that all crystals smaller than 5 microns dissolve within 5 days, the concept of an
effective critical size is introduced for ripening in magmatic environments. Systems held near
equilibrium for long durations will evolve so that the critical size approaches the mean size, and
consequently all information pertaining to nucleation will be erased from the CSD. Cabane et al.
(2005) performed similar experiments that included a series focused on plagioclase coarsening
in synthetic (CMAS+Na2O) andesite magma. They observe crystal size increasing with t1/3 and
positive skewness in normalized size distributions (Fig. 22), both characteristics attributable to
growth according to a screw dislocation mechanism consistent with nominal undercooling.
Crystal aspect ratio
The surface area to volume ratio (3D aspect ratio) is the comparative metric selected for
quantitative textural analysis of pyroxene, olivine and Fe-Ti oxides in a study of the effects of
cooling rate and fO2 on textures in synthetic Fe-rich Mars analog basalt (Hammer 2006). This
analysis technique (Dehoff and Rhines 1968) was adopted because crystal morphologies are
highly anhedral. In summary, circular test lines are digitally overlain on BSE images; locations
42 Hammer
-5
7 min
30 min
ln (population density, um-4)
58 min
-10
180 min
1.9 days
4.0 days
-15 6.9 days
9.8 days
-20
-25
0 25 50 75 100 125
size (um)
Figure 21. Crystal size distributions of forsterite in ripening experiments.
Dwell times are as indicated; replotted from Park and Hanson (1999).
Figure 21.
24000
Olivine
Cube of mean grain size (μm3)
80 min
16000
43.2 hr
496.4 hr
8000
0
0 100 200 300 400 500
Time (hours)
Plagioclase 200
1 hr
Cube of mean grain size (μm3)
150
20 hr al to
time
rtion
336 hr ropo
100 d p
3
50
0
0 100 200 300 400
Time (hours)
Figure 22. Two dimensional crystal size distributions normalized by mean size for three experiments
Figure 22.focusing on olivine (a) and plagioclase (b). Insets show cube of the 2D mean grain sizes plotted against run
duration. Plagioclase data (a inset) for durations ≥ 20 h (solid circles) are fitted with a line (empty circles
correspond to short duration experiments ≤ 10 h). All olivine data (b inset) are fitted with a line; plagioclase
mean grain sizes are shown for comparison. After Cabane et al. (2005).
Experimental Studies of Magma Crystallization 43
where crystal-melt boundaries intersect
the test line are manually identified and a.
231 oC/h
marked, and the quantity of marked
boundary intersections per unit length of
test line (NL) is tabulated. The surface area 1
per unit volume of a pyroxene population,
SvP, is computed as a combination of the
boundary intersection density and the
volume fraction:
SvP = 2NL/φ (10)
100 um
where the area fraction of pyroxene
b.
18.7 oC/h
crystals, presumed equivalent to the
volume fraction φ, (Hilliard 1968) which is
measured by standard thresholding of the
characteristic range of grayscale values. 2 1
In accord with qualitative assessment
of the morphlogical dependence upon
cooling rate, the 3D crystal aspect ratio
(hereafter “aspect ratio”) is sensitive to
cooling rate. The aspect ratio of pyroxene
100 um
is particularly well correlated with mor-
phology. For example, all pyroxene crys- c.
2.8 oC/h
tals characterized as euhedral have aspect
ratios < 1000 mm−1, while all anhedral
morphologies have values exceeding 1000
1
mm−1. Interestingly, the aspect ratios of
separate pyroxene populations progres-
sively diverge as cooling rate decreases
(Fig. 23). This trend in textural dispersion
2
results in the slowest-cooled materials con-
taining populations having both the lowest 100 um
and highest observed aspect ratios, while 10000
moderately-cooled experiments contain d. surface area per unit volume
2
two morphologically similar populations,
and fastest-cooled experiments contain 1000
only one population. In a relationship b c
(mm-1)
reminiscent of the power law relationship a
between growth rate and cooling rate, here 100
the logarithm of the cooling rate is linearly
1
related to the logarithm of the population
population
1 2
NNO pop 1: log (y) = 0.677* log (x) + 1.391 R2= 0.93
aspect ratio. The coarse-grained popula- 10
QFM pop 2: log (y) = -0.411* log (x) + 3.850 R2= 0.82
1000 100 10 1
cooling rate (oC h-1)
Figure 23. (a-c). Binary images of pyroxene in dynamic cooling experiments obtained from BSE images;
circled numbers indicate morphologically distinct populations. (d). quantitative relationship between
cooling rate and the surface area to volume ratio of each pyroxene population. Gray arrows indicate
populations represented in images (a-c). Abscissa axis values increase toward the left so that the theoretical
approach to textural equilibrium increases to the right. Long dashed lines are least squares fits to 1 σ
about sample means. Extension of second population trend lines to cooling rates above 18.7 °C h−1 is for
illustrative purposes only; no experiment cooled faster than 18.7 °C h−1 contained more than one crystal
population. Adapted from Hammer (in press).
44 Hammer
tion decreases by two orders of magnitude and the fine-grained population doubles over the
cooling rate range examined. Modest initial undercooling in slow-cooled charges probably led
to efficient growth of faceted crystals by an interface-controlled mechanism, delaying buildup
of sufficient driving force for a secondary nucleation event until late in the experiment. The an-
hedral morphologies and high SvP of the second generation reflect diffusion-limited growth and
insufficient time for coarsening. The comparative uniformity of anhedral crystal morphologies
in the most rapidly cooled charges (>100 °C h−1) is probably due to rapid diffusion-controlled
growth of a single generation of late (low-temperature) nuclei.
PhaSe BOundarIeS and InterFaCeS
Interface phenomena are at the frontier of both nucleation studies (described above) and
multiphase dynamics. Important questions being addressed include whether interfacial free
energies between crystalline or crystals and vapor bubbles play an important role in defining
texture, whether crystalline phases nucleate heterogeneously on vapor surfaces, the possibility
of epitaxial growth as an important process in natural compositions, and assessment of crystal
clustering and synneusis in magmatic environments.
heterogeneous nucleation
Davis and Ihinger (1998) report strong proclivity of lithium disilicate crystals to nucleate
on vapor bubbles, but only following specific thermal histories (two stage fusion treatments
prior to a crystal development step, with the second stage 330 °C cooler than the first) and
exposure to certain vapor species (Ar and N2). In keeping with observations in natural magmas,
their data suggest that heterogeneous crystal nucleation does not occur on arbitrary interfaces.
Rather, chemical reactions occur at specific crystal-vapor interfaces, possibly including
production of a surfactant.
The opposite relationship—that of crystals providing sites for heterogeneous nucleation
of vapor bubbles—has been explored by Hurwitz and Navon (1994), Gardner (2007), and
Simakin et al. (2000). While both spinel-structured and rhombohedral Fe-Ti oxides apparently
reduce the energy barrier to nucleation relative to the homogeneous case, major rock-forming
minerals plagioclase and pyroxene apparently do not. Gardner (2007) compares the effects
of melt viscosity, temperature, and water diffusivity predicted by bubble nucleation theory
with the experimental data, concluding that none of these factors can account for the observed
variation in nucleation rate. Rather, changes in melt structure with varying H2O content are
inferred to exert dominant control over the nucleation-hypersensitive σsolid-liquid and σliquid-vapor.
Solid-solid interfacial energy
The connectivity of liquid in systems near the solidus depends upon the dihedral angle
θ defined in Equation (6) as a ratio between solid-solid and solid-liquid interfacial energies.
Ikeda et al. (2002) measured dihedral angles among diopside crystals coexisting with melts at
high melt fractions in the diopside-anorthite system, applying Equation (6) to evaluate changes
in σsolid-liquid/σsolid-solid with changing temperature and melt composition. Holocrystalline
starting materials were prepared by equilibrating homogenized synthetic glass at subsolidus
temperature. The crystalline materials were subsequently either (1) brought directly to a run
temperature between 1280 and 1350 °C and held for 5-24 h, (2) heated at a constant rate (0.5
°C min−1) from near the solidus (1280 °C) to the run temperature, or (3) cooled at constant
rate (0.05 or 0.5 oC min-1) from 40 °C below the liquidus (1350 °C) to the run temperature.
Although sets (1) and (2) are both melting experiments, the isothermal experiments of treatment
(1) are perceived to result in attainment of textural equilibrium. Set (3) represents a reversal in
that equilibrium is approached from above the run temperature. Values of the dihedral angle
are correlated with temperature and melt composition in all experiments (Fig. 24), although
Experimental Studies of Magma Crystallization 45
σsolid-liquid/ σsolid-solid
Figure 24. Variation of the dihedral angle with Al2O3 concentration in the melt, an indicator of crystallinity.
Uncertainties in dihedral angles are ±4°. Numerical values near symbols indicate run temperatures (in
isothermal experiments, solid symbols) or quench temperatures (in cooling and heating experiments, open
symbols). Dihedral angles in the isothermal experiments are considered close to equilibrium values. [Used
Figure 24. permission from Springer Science+Business Media from Ikeda et al. (2002) Contrib Mineral
with kind
Petrol, Vol. 142, Fig. 15b, p. 397-415.].
heating yields σ ratios that are lower than the isothermal values and cooling yields σ ratios that
are higher. One interpretation supported by the observed trend is that the σsolid-liquid increases
when the melt composition departs from the crystal composition. Formation of boundary layers
enriched in rejected components around growing crystals may increase σsolid-liquid, thereby
providing a driving force for crystal clustering as a means of reducing the overall solid-liquid
interfacial area (Fig. 25).
Crystal intergrowths are common in both experimental and natural samples. Two factors are
particularly important in co-crystallization: (1) diffusion-limited crystal growth leading to sta-
bilization of phases that incorporate rejected components, i.e., constitutional supercooling, and
(2) interfacial energy reductions accommodated by coherent or semi-coherent lattice boundaries
between two phases. In the latter case, nucleation may be heterogeneous, and growth epitaxial,
leading to a lattice preferred orientation among phases. Interfacial energy controls are not ex-
pected to be important among minerals crystallizing from silicate melt at high undercooling.
However, orientation mapping using electron backscatter diffraction (EBSD) reveals consis-
tent textural characteristics in samples produced in 1-atm basalt cooling experiments (Hammer
2005). Clinopyroxene crystals are dendritic, with high surface area to volume ratios (≥ 3 × 103
mm−1), indicating diffusion-controlled growth. Orientation maps of contiguous regions of in-
dividual pyroxene dendrites indicate incremental rotations during crystal growth—poles to the
(010) planes are strongly aligned, whereas the poles to (100) and (001) describe great circles
in their respective pole figures. Titanomagnetite crystals occur preferentially on the surfaces of
feathery clinopyroxene dendrites (Fig. 26a). Poles of the six symmetrically equivalent (110)
planes in titanomagnetite are weakly clustered. One of these poles is always coincident with the
strongly oriented (010) pole of nearby clinopyroxene crystals (Fig. 26b,c).
Dendritic pyroxene morphologies in these experiments (Fig 26a) indicate chemical po-
tential gradients developing in a regime of constitutional supercooling. The high proportion
of shared pyroxene-titanomagnetite boundaries is consistent with either constitutional super-
46 Hammer
Figure 25. Textural diversification model depending upon the interfacial energy ratio (dihedral angle). The
Figure 25.
system with a large dihedral angle produces crystal clustering in both plutonic and volcanic rocks. [Used
with kind permission from Springer Science+Business Media from Ikeda et al. (2002) Contrib Mineral
Petrol, Vol. 142, Fig. 5, p. 397-415.]
a. b.
timt
Figure 26. Titanomagnetite crys-
tals decorate the tips of dendritic
Ca-pyroxene crystals in rapidly
Ca-pyroxene {010} cooled Fe-rich/ Al-poor basalt.
The (010) pyroxene and (110) pole
c. figures demonstrate coincidence
of crystallographic axes suggest-
ing epitaxial growth. (Hammer
pyx unpubl.)
titanomagnetite {110}
cooling or an interfacial energy advantage to co-crystallization. However, the lattice preferred
orientation observed among crystals of pyroxene and titanomagnetite indicates a strong con-
trol of lattice interfacial energies in the formation of titanomagnetite. The identical lattice pre-
ferred orientation has been described in oxide-pyroxene intergrowths occurring by subsolidus
exsolution, presumably at low degrees of undercooling (Okamura et al. 1976) resulting from
similarities in the oxygen frameworks of these minerals; i.e., the presence of a semi-coherent
Experimental Studies of Magma Crystallization 47
lattice boundary. Experiments (Hammer 2005) show that epitaxial growth and heterogeneous
nucleation may also be important in highly undercooled sililcate melts in which constitutional
supercooling drives multiphase intergrowths.
reaction relationships
A final example of interphase reaction kinetics with direct application to volcanic
processes is determination of magma mixing events prior to eruption and quench (e.g., Baker
1991; Coombs et al. 2003). The exposure time of dissimilar magmas thrust into contact may
be estimated using experimental studies of mineral-melt reactions, given plausible estimates
of temperature, melt H2O content, and PH2O. Coombs and Gardner (2004) monitored the
development of forsteritic olivine reacting with hydrous high-silica melt to form either
orthopyroxene or amphibole+orthopyroxene. As may be expected, the reaction product is
strongly dependent upon dissolved H2O content. However, there is also a strong dependence
of reaction rate on dissolved H2O, presumably because of enhanced component mobility in
the melt. Rim growth rate increases by an order of magnitude as PH2O increased from 50 to150
MPa, despite 50 °C cooling.
MeLt InCLuSIOn entraPMent
Crystal growth mechanisms and regime are especially important in the context of
interpreting the compositions of melts trapped within growing crystals. It is extremely important
to determine whether diffusion-controlled growth conditions are required to entrap melts, and
whether the melts are boundary layers rather than representative of the far-field melt. As yet,
very few experimental studies report information sufficient to evaluate the issue, and the extant
data are intriguingly contradictory.
Major element behavior
Melt inclusion formation was studied by Kohut and Nielsen (2004) in a series of dynamic
crystallization experiments using natural starting materials. MORB was placed inside capsules
of An92, seeded with Fo92 crystals, held for two hours at the computed liquidus temperature (1300
°C), cooled to 1230 or 1210 °C at rates of 1, 5, or 10 °C min−1, held for dwell times ranging
from 0-24 h, and then quenched. Melt inclusions were evaluated in terms of major element
composition and size distribution. Evaluated in terms of Al2O3, CaO in olivine and Mg# and
TiO2 in plagioclase, melt inclusion compositions are not correlated with size. Importantly, melt
inclusion compositions matched matrix glass composition for all runs except those quenched
directly after cooling (apparently reflecting failure to equilibrate during ramped cooling). The
interpreted sequence of events (for both plagioclase and olivine) is that protuberances evolve
to form hopper crystals during the cooling stage, which subsequently overgrow and seal off
at reduced growth rate during the dwell period. In the overwhelming majority of cases, the
shift from diffusion-control to interface-controlled growth occurs before the melt inclusion is
isolated from the matrix melt, and therefore any boundary layer that existed during rapid growth
has been homogenized by diffusion.
Dramatically different results are obtained in a study of melt inclusion entrapment in for-
sterite crystallizing from synthetic basalt by Faure and Schiano (2005). In this study, reagents
are fused at 1400 °C for 24 h, quenched, powdered, pressed into pellets, and reheated to super-
liquidus temperatures (19-52 °C above the liquidus), held for an hour, then cooled along one of
four schedules. The majority of runs are either single-segment constant rate cooling ramps at
1-1890 °C h−1 to varying quench temperatures (representing 51-176 °C below the liquidus), or
dual-segment ramped cooling where the first segment is slow (2 °C h−1), the second segment is
fast (1644-1730 °C h−1), and quench temperatures are low (175-236 °C below the liquidus). As
the charges were not seeded, all observed olivine crystals nucleated during cooling. In contrast
48 Hammer
to the Kohut and Nielsen finding that inclusion entrapment requires growth at rapid cooling
followed by growth at isothermal conditions, faceted olivine crystals in the slowly cooled (1 °C
h−1) single-segment experiments occasionally develop embayments and trap inclusions, never
having undergone rapid cooling. The near-surface melt and melt inclusion compositions formed
during slow cooling are identical to far-field matrix melt in all elements. Faure and Schiano
(2005) suggest that interactions among screw dislocations could promote development of em-
bayments and ultimately seal off of melt inclusions. In contrast, skeletal and dendritic crystals
produced in rapidly cooled single-segment experiments, as well as dendrite-tipped polyhedral
crystals formed during the dual segment runs, are surrounded by a boundary layer of melt that
is depleted in MgO and enriched in Al2O3 and CaO compared to the far-field melt (Fig. 27).
As may be expected from such inclusions, the compositions within skeletal crystals deviate
substantially from the equilibrium liquid line of descent followed by the far-field melt. Remark-
ably, inclusions contained within dendritic-polyhedral crystals also deviate from the liquid line
of descent and far field melts, despite gross morphological similarity with slowly grown faceted
crystals. The telltale sign of the latter entrapment process is elongation of inclusions parallel to
(010) faces, indicating growth entrapment occurred while the crystal was skeletal (Fig. 28).
The Kohut and Nielsen (2004) and Faure and Schiano (2005) studies arrive at dramatically
different conclusions regarding the melt inclusion entrapment process as a process in which far-
field melts are faithfully recorded in crystals growing under a range of conditions. While it is
tempting to pick apart differences in cooling rates, dwell temperatures, durations, and undercooling,
the two studies differ in many respects, not the greatest of which is cooling schedule. The Kohut
a. b.
Figure 27. (a) Reflected light photomicrograph of a section parallel to the (010) plane of a skeletal crystal
of olivine, showing cavities filled by glass at each end of the crystal. (b) BSE image of a dendrite-tipped
polyhedral olivine crystal. Graphs show chemical gradients along the marked traverses of the olivine
crystal-liquid interface. [Used by permission of Elsevier, from Faure and Schiano (2005), Earth Planet Sci
Lett, Vol. 236, Figs. 7 and 10, pp. 882-898.]
Figure 27.
Experimental Studies of Magma Crystallization 49
Figure 28. Back-scattered electron image
of a dendritic polyhedral forsterite crystal
cut nearly parallel to the (100) plane. The
enclosed melt inclusions lie parallel to the
external (010) or (021) faces. [Used by
permission of Elsevier, from Faure and
Schiano (2005), Earth Planet Sci Lett,
Vol. 236, Fig. 9, pp. 882-898.]
and Nielsen experiments were run on natural materials with multicomponent compositions,
and were seeded with natural phenocrysts. The Faure and Schiano (2005) experiments were
Figure 28. performed unseeded in a compositionally simple system with synthetic materials. In view of
nucleation and growth studies in general, it would not be surprising if starting material types
and preparations were at least as important as cooling schedule in controlling melt inclusion
processes. Future work that applies a variety of approaches, emphasizing characterization of
the evolving melt structure and sensitivity to initial conditions, may be necessary to resolve the
apparent discrepancies.
trace element behavior
Small undercooling. Crystallization at small undercoolings results in characteristic
partitioning of compatible and incompatible elements into the crystal lattice. Kennedy et
al. (1993) evaluated the effect of cooling rate on partitioning of 32 elements in olivine and
orthopyroxene coexisting with melt of chondrule composition, finding no correlation between
the partition coefficients of incompatible elements and cooling rate, up to 100 °C h−1. Similar
results were obtained by Lofgren et al. (2006) in Ti-Al-clinopyroxene growing in near-
isochemical melt at small undercooling. Such relationships underpin much of trace element
geochemistry as a powerful interpretative tool in igneous petrology. An alternative described
by Watson and Liang (1995) explains sector zoning in slowly cooled igneous minerals as the
result of anisotropic enrichment at crystal surfaces coupled with sluggish lattice diffusion.
That is, partitioning is an equilibrium process, but the partition coefficient for the surface of
the crystals is different from (and often higher than) the bulk value.
Growing crystals can incorporate trace elements in excess of bulk partitioning equilibrium
if crystal surfaces provide greater flexibility in the diversity of type of atomic sites present; trace
element atoms are essentially engulfed by the advancing growth front (Watson 1996, 2004).
Watson and Liang (1995) define a dimensionless growth Peclét number, Pe = V × l/Di, where
V is the velocity of crystal growth, l is the half thickness of the surface layer (on the order of
atomic dimensions), and Di is the diffusivity of component i in the near surface region of the
crystal. Growth entrapment is not predicted to be important in crystals growing at moderate to
low undercoolings in natural magmas, as growth rates must be exceedingly rapid to overcome
rapid atomic mobility in melt at magmatic temperatures (Watson 1996). However, this effect
may be important at large undercoolings typical of syn-eruptive crystallization.
Large undercooling. Crystal growth at large undercoolings is more commonly associated
with non-equilibrium partitioning, with at least four manifestations described in the literature.
(1) Diffusion-limited crystal growth results in formation of a boundary layer enriched in
rejected components, so the concentrations incorporated into the growing crystal are higher
than the equilibrium values based on far-field concentrations in the melt. The observation
by Grove and Raudsepp (1978) that elements with pyroxene-melt distribution coefficients
50 Hammer
greater than 1 were depleted around pyroxene phenocrysts while incompatible elements were
enriched, is consistent with this process. Similarly, in their study of clinopyroxene growing
in response to varying degrees of undercooling, Lofgren et al. (2006) report trace element
distribution coefficients increasing to unity or above with increasing crystal growth rate. (2)
Rapid, highly anisotropic growth causes entrapment of submicroscopic melt inclusions, and
their boundary-layer enrichments of incompatible elements. Above 100 °C h−1 cooling rate,
Kennedy et al. (1993) observe extreme departures from equilibrium partitioning in olivine
and orthopyroxene. The element variation patterns are consistent with incorporation of tiny
melt inclusions within the ion probe analyses; very small amounts of melt contamination (~5
wt%) could account for trace element patterns. The authors suggest that cooling rates 100-
2000 °C h−1 lead to skeletal crystal morphologies for olivine and orthopyroxene, facilitating
incorporation of submicroscopic melt inclusions. (3) Rapid growth causes incompatible
elements to be incorporated into crystals at concentrations exceeding the bulk or surface
partitioning values, in extreme cases nearing the absolute concentration in the boundary layer
melt. Such partitionless growth or “solute trapping” is a disequilbrium phenomenon that affects
elements selectively, based on their effective mobilities in the near-surface region. (4) Sector
zoning results from differences in impurity uptake by faces advancing at different velocities
as well as a strong crystal-chemical control. Shwandt and McKay (2006) describe rapidly
growing pyroxene faces in the {001} sectors that incorporate fewer impurity cations than slow-
growing faces (Fig. 29). The slow-growing sectors {100} and {010} cope with a boundary
layer enriched in Al and impurity elements produced in the wake of the fast-growing faces by
substituting octahedral cations and octahedral cations as charge-compensation couples. These
slow-growing sectors are prone to incorporating differing impurity concentrations depending
on whether tetrahedral and octahedral layers are exposed simultaneously as edges on {100}
faces or alternating in sequence as planes parallel to the surface, on {010} faces.
Elevated P contents are observed in experimentally grown olivines, even as major
elements (Fe-Mg) are distributed between melt and crystals according to equilibrium
partitioning (Milman-Barris et al. 2008). The concentrations of P in the near-surface region
are grossly insufficient to represent equilibrium partitioning according to process (1), above.
Incorporation of tiny melt inclusions into analytical spots (2) is precluded by vastly differing
Al/P ratios in olivine and boundary layer melt. Rather, crystal morphology implicates rapid
Figure 29. Aluminum map showing
cooling steps and sector zoning in
synthetic enstatite (shades of gray),
viewed down the fast-growing [001]
axis, with basaltic glass (white).
The legend indicates the number of
Al x-ray counts per pixel. Within
one concurrent growth interval the
concentration of Al is higher in the
{010} sectors than the {100} sectors.
The distinctiveness of all sectors
decreases within each cooling step
as growth slowed, prior to the start
of each new cooling step, whereby
the intensity of sector development is
intensified.
Experimental Studies of Magma Crystallization 51
growth in disequilibrium P trapping (3), as P-rich zones define relict hopper and skeletal
morphologies within externally-faceted and euhedral crystals (Fig. 30). Given sufficient time
at high temperature, high-P olivine in contact with melt dissolves and subsequently precipitates
at low growth rates (controlled by slow replenishment through tortuous pathways to the crystal
surface) as low-P olivine. The result is low-P interior linings of skeletal crystals. A similar
dissolution/ precipitation process may produce halos of low-P olivine in the interiors of natural
melt inclusions (Ikeda 1998; Danyushevsky et al. 2004). This study underscores the need to
consider lattice and surface mobilities of trace element ions in the context of crystal growth
mechanisms in the interpretation of natural magmatic processes, and demonstrates the need
for further targeted experimental study.
Figure 30. Phosphorus x-ray map of experimen-
tal olivine grown from basalt cooled from 1,400
to 1,190 °C at 15 °C h−1. A central high-P skel-
eton [F] is zoned outward toward low-P rims [F]
and inward toward elongate melt pools enclosed
in low-P olivine [D]. Other features include fine-
scale oscillatory zoning [A] and [H], knobs of
high-P olivine close to interior melt pools [E]
and a region of oscillatory bands [B]. [Modified
after Milman-Barris et al. (2008), Fig. 7.]
100 um
COnCLuSIOnS
The formative applications of experimental petrology to problems in volcanology were
both simple and profound: to determine the depth, temperature, and chemical environments of
Figure 30. restless magma prior to eruption. Many of the most important breakthroughs in volcanology,
such as appreciation of magma interactions in eruption triggering and pre-eruptive gas saturation,
are the fruits of interdisciplinary approaches linking field observations with experimental phase
equilibria studies. Looking toward the future of volcanology, experimental studies promise to
provide insights into dynamic aspects of magma transport, as the focus shifts from defining
equilibrium states toward constraining the kinetics of magmatic processes. “Experimental
volcanology” encompasses a variety of complementary approaches, including the use of natural
silicate magma, synthetic magmas based on natural compositions, and analog materials. The
strength of traditional high-pressure/temperature experiments continues to be the ability to study
natural chemical processes at conditions that can be varied to emulate changing geological
environments. Innovation in experimental apparatus development and improvements in the
spatial and compositional resolution of analytical instruments hold the promise of increasing
the applicability of experimental research in volcano hazard assessment, as well as deepening
the theoretical understanding of underlying physical and chemical processes.
52 Hammer
aCKnOWLedGMentS
This contribution was conceived and nurtured by discussions with M. Rutherford. C.
Brugger, O.K.Neill, and L. Tatsumi are thanked for their assistance and enthusiasm. The
manuscript was improved by constructive reviews from F. Faure, Y. Liang, G. Lofgren, K.
Putirka, and B. Scaillet. This work was supported in part by NSF award EAR04-49888. It is
SOEST publication #7501.
reFerenCeS CIted
Andersen DJ, Lindsley DH, Davidson PM (1993) qUILF; a Pascal program to assess equilibria among Fe-Mg-
Mn-Ti oxides, pyroxenes, olivine, and quartz. Computers Geosci 19:1333-1350
Arbaret L, Bystricky M, Champallier R (2007) Microstructures and rheology of hydrous synthetic magmatic
suspensions deformed in torsion at high pressure. J Geophys Res doi:10.1029/2006JB004856
Armienti P (2008) Decryption of igneous rock textures: crystal size distribution tools. Rev Mineral Geochem
69:623-649
Bachmann O, Bergantz GW (2008) Deciphering magma chamber dynamics from styles of compositional zoning
in large silicic ash flow sheets. Rev Mineral Geochem 69:651-674
Bagdassarov N, Dorfman A, Dingwell DB (2000) Effect of alkalis, phosphorus, and water on the surface tension
of haplogranite melt. Am Mineral 85:33-40
Baker DR (1991) Interdiffusion of hydrous dacitic and rhyolitic melts and the efficacy of rhyolite contamination
of dacitic enclaves. Contrib Mineral Petrol 106:462-473
Barclay J, Rutherford MJ, Carroll MR, Murphy MD, Devine JD, Gardner J, Sparks RSJ (1998) Experimental
phase equilibria constraints on pre-eruptive storage conditions of the Soufriere Hills magma. Geophys Res
Lett 25:3437-3440
Bartels KS, Furman T (2002) Effect of sonic and ultrasonic frequencies on the crystallization of basalt. Am
Mineral 87:217-226
Battezzati L (2001) Thermodynamic quantities in nucleation. Mater Sci Eng A 304:103-107
Becker R, Doring W (1935) Kinetische behandlung der keim buildung in ubersattingten dampfen. Ann Physik
5:719-752
Berndt J, Koepke J, Holtz F (2005) An experimental investigation of the influence of water and oxygen fugacity
on differentiation of MORB at 200 MPa. J Petrol 46:135-167
Berndt J, Liebske C, Holtz F, Freise M, Nowak M, ziegenbein D, Hurkuck W, Koepke J (2002) A combined
rapid-quench and H2-membrane setup for internally heated pressure vessels: Description and application
for water solubility in basaltic melts. Am Mineral 87:1717-1726
Blundy J, Cashman K (2001) Ascent-driven crystallisation of dacite magmas at Mount St Helens, 1980-1986.
Contrib Mineral Petrol 140:631-650
Blundy J, Cashman K (2008) Petrologic reconstruction of magmatic system variables and processes. Rev
Mineral Geochem 69:179-239
Bogaerts M, Scaillet B, Vander Auwera J (2006) Phase equilibria of the Lyngdal Granodiorite (Norway);
implications for the origin of metaluminous ferroan granitoids. J Petrol 47:2405-2431
Bowles JA, Hammer JE, Brachfeld SA (2007) Magnetic and petrographic characterization of synthetic martian
basalts. 7th International Conference on Mars, p. 3255, Pasadena, CA
Brugger C, Johnston AD, Cashman KV (2003) Phase relations in silicic systems at one-atmosphere pressure.
Contrib Mineral Petrol 146:356-369
Burkhard DJM (2002) Kinetics of crystallization: example of micro-crystallization in basalt lava. Contrib
Mineral Petrol 142:724-737
Burnham CW (1975) Water and magmas: a mixing model. Geochim Cosmochim Acta 39:1077-1084
Cabane H, Laporte D, Provost A (2001) Experimental investigation of the kinetics of Ostwald ripening of quartz
in silicic melts. Contrib Mineral Petrol 142:361-373
Cabane H, Laporte D, Provost A (2005) An experimental study of Ostwald ripening of olivine and plagioclase in
silicate melts; implications for the growth and size of crystals in magmas. Contrib Mineral Petrol 150:37-
53
Cashman K (1990) Textural constraints on the kinetics of crystallization of igneous rocks. Rev Mineral Geochem
24:259-314
Cashman KV (1993) Relationship Between plagioclase crystallization and cooling rate in basaltic melts. Contrib
Mineral Petrol 113:126-142
Castro JM, Gardner J (2008) Did magma ascent rate control the explosive-effusive transition at the Inyo volcanic
chain, California? Geology 36:279-282
Experimental Studies of Magma Crystallization 53
Champallier R, Bystricky M, Arbaret L (2008) Experimental investigation of magma rheology at 300 MPa:
From pure hydrous melt to 76 vol.% of crystals. Earth Planet Sci Lett 267:571-583
Christian JW (1965) The Theory of Transformations in Metals and Alloys. Pergamon Press, New York
Clemens JD, Holloway JR, White AJR (1986) Origin of an A-type granite: Experimental constraints. Am
Mineral 71:317-324
Clemens JD, Wall VJ (1981) Origin and crystallization of some peraluminous (S-type) granitic magmas. Can
Mineral 19:111-131
Conte AM, Perinelli C, Trigila R (2006) Cooling kinetics experiments on different Stromboli lavas: Effects on
crystal morphologies and phases composition. J Volc Geotherm Res 155:179-200
Coombs ML, Eichelberger JC, Rutherford MJ (2000) Magma storage and mixing conditions for the 1953-1974
eruptions of Southwest Trident Volcano, Katmai National Park, Alaska. Contrib Mineral Petrol 140:99-
118
Coombs ML, Eichelberger JC, Rutherford MJ (2003) Experimental and textural constraints on mafic enclave
formation in volcanic rocks. J Volcanol Geotherm Res 119:125-144
Coombs ML, Gardner JE (2001) Shallow-storage conditions for the rhyolite of the 1912 eruption at Novarupta,
Alaska. Geology 29:775-778
Coombs ML, Gardner JE (2004) Reaction rim growth on olivine in silicic melts: Implications for magma
mixing. Am Mineral 89:748-758
Cooper RF, Kohlstedt DL (1982) Interfacial energies in the olivine-basalt system. In: High-Pressure Research in
Geophysics, 12. Akimoto S, Manghnani MH (eds) D. Reidel Publishing Co., Tokyo. p. 217-228
Costa F, Scaillet B, Pichavant M (2004) Petrological and experimental constraints on the pre-eruption conditions
of Holocene dacite from Volcan San Pedro (36 degrees S, Chilean Andes) and the importance of sulphur in
silicic subduction-related magmas. J Petrol 45:855-881
Cottrell E, Gardner JE, Rutherford MJ (1999) Petrologic and experimental evidence for the movement and
heating of the pre-eruptive Minoan rhyodacite (Santorini, Greece). Contrib Mineral Petrol 135:315-331
Couch S (2003) Experimental investigation of crystallization kinetics in a haplogranite system. Am Mineral
88:1471-1485
Couch S, Sparks RSJ, Carroll MR (2001) Mineral disequilibrium in lavas explained by convective self-mixing
in open magma chambers. Nature 411:1037-1039
Couch S, Sparks RSJ, Carroll MR (2003) The kinetics of degassing-induced crystallization at Soufriere Hills
volcano, Montserrat. J Petrol 44:1477-1502
Dall’Agnol R, Scaillet B, Pichavant M (1999) An experimental study of a lower Proterozoic A-type granite from
the eastern Amazonian Craton, Brazil. J Petrol 40:1673-1698
Danyushevsky LV, Leslie RAJ, Crawford AJ, Durance P, Niu Y, Herzberg C, Wilson M (2004) Melt inclusions
in primitive olivine phenocrysts; the role of localized reaction processes in the origin of anomalous
compositions. J Petrol 45:2531-2553
Davidchack RL, Laird BB (2003) Direct calculation of the crystal-melt interfacial free energies for continuous
potentials:Application to the Lennard- Jones system. J Chem Phys 118:7651-7657
Davis M (2000) Effect of the growth treatment on two-stage nucleation experiments: The “flushing” effect.
Glass Sci Tech 73:170-177
Davis MJ, Ihinger PD (1998) Heterogeneous crystal nucleation on bubbles in silicate melt. Am Mineral 83:1008-
1015
Davis MJ, Ihinger PD (1999) Influence of hydroxyl on glass transformation kinetics in lithium disilicate melt
and a re-evaluation of structural relaxation in NBS 710 and 711. J Non-Cryst Solids 244:1-15
Davis MJ, Ihinger PD (2002) Effects of thermal history on crystal nucleation in silicate melt: Numerical
simulations. J Geophys Res 107 doi:10.1029/2001JB000392
Davis MJ, Ihinger PD, Lasaga AC (1997) Influence of water on nucleation kinetics in silicate melt. J Non-Cryst
Solids 219:62-69
Dehoff RT, Rhines FN (1968) quantitative Microscopy. McGraw-Hill, New York
Deubener J, Weinberg MC (1998) Crystal-liquid surface energies from transient nucleation. J Non-Cryst Solids
231:143-151
Di Carlo I, Pichavant M, Rotolo SG, Scaillet B (2006) Experimental crystallization of a high-K arc basalt: The
golden pumice, Stromboli volcano (Italy). J Petrol 47:1317-1343
Dingwell DB, Webb SL (1989) Structural relaxation in silicate melts and non-Newtonian melt rheology in
geologic processes. Phys Chem Mineral 16:508-516
Donaldson CH (1976) An experimental investigation of olivine morphology. Contrib Mineral Petrol 57:187-
213
Donaldson CH, Usselman TM, Williams RJ, Lofgren GE (1975) Experimental modeling of the cooling history
of Apollo 12 olivine basalts. Proc Lunar Sci Conf 6th, E43-87
Dowty E (1980) Crystal growth and nucleation theory. In: Physics of Magmatic Processes. Hargraves R (ed)
Princeton University Press, Princeton, NJ. p 487-551
54 Hammer
Eberl DD, Kile DE (2005) Crystal growth rate law identified from changes in variance of crystal size distributions.
Geochim Cosmochim Acta 69:A402-A402
Eberl DD, Kile DE, Drits VA (2002) On geological interpretations of crystal size distributions: Constant vs.
proportionate growth. Am Mineral 87:1235-1241
Edgar AD (1973) Experimental Petrology- Basic Principles and Techniques. Oxford University Press, New
York
Eggler DH, Burnham CW (1973) Crystallization and fractionation trends in the system andesite-H2O-CO2-O2 at
pressures to 10 Kb. Geo Soc Am Bull 84:2517-2532
Faure F, Arndt N, Libourel G (2006) Formation of spinifex texture in komatiites: an experimental study. J Petrol
47:1591-1610
Faure F, Schiano P (2004) Crystal morphologies in pillow basalts: implications for mid-ocean ridge processes.
Earth Planet Sci Lett 220:331-344
Faure F, Schiano P (2005) Experimental investigation of equilibration conditions during forsterite growth and
melt inclusion formation. Earth Planet Sci Lett 236:882-898
Faure F, Schiano P, Trolliard G, Nicollet C, Soulestin B (2007) Textural evolution of polyhedral olivine
experiencing rapid cooling rates. Contrib Mineral Petrol 153:405-416
Faure F, Trolliard G, Nicollet C, Montel JM (2003a) A developmental model of olivine morphology as a function
of the cooling rate and the degree of undercooling. Contrib Mineral Petrol 145:251-263
Faure F, Trolliard G, Soulestin B (2003b) TEM investigation of forsterite dendrites. Am Mineral 88:1241-1250
Feig ST, Koepke J, Snow JE (2006) Effect of water on tholeiitic basalt phase equilibria: an experimental study
under oxidizing conditions. Contrib Mineral Petrol 152:611-638
Fenn PM (1977) The nucleation and growth of alkali feldspar from hydrous melts. Can Mineral 15:135-161
Fokin VM, Schmelzer JWP, Nascimento MLF, zanotto ED (2007) Diffusion coefficients for crystal nucleation
and growth in deeply undercooled glass-forming liquids. J Chem Phys 126 doi:10.1063/1.2746502
Fokin VM, zanotto ED (2000) Crystal nucleation in silicate glasses: the temperature and size dependence of
crystal/liquid surface energy. J Non-Cryst Solids 265:105-112
Fokin VM, zanotto ED, Schmelzer JWP (2000) Method to estimate crystal/liquid surface energy by dissolution
of subcritical nuclei. J Non-Cryst Solids 278:24-34
Fokin VM, zanotto ED, Schmelzer JWP, Potapov OV (2005) New insights on the thermodynamic barrier for
nucleation in glasses: The case of lithium disilicate. J Non-Cryst Solids 351:1491-1499
Fokin VM, zanotto ED, Yuritsyn NS, Schmelzer JWP (2006) Homogeneous crystal nucleation in silicate
glasses: A 40 years perspective. J Non-Cryst Solids 352:2681-2714
Gaillard F (2004) Laboratory measurements of electrical conductivity of hydrous and dry silicic melts under
pressure. Earth and Planet Sci Lett 218:215-228
Gaillard F, Pichavant M, Mackwell S, Champallier R, Scaillet B, McCammon C (2003) Chemical transfer
during redox exchanges between H2 and Fe-bearing silicate melts. Am Mineral 88:308-315
Gardner JE (2007) Heterogeneous bubble nucleation in highly viscous silicate melts during instantaneous
decompression from high pressure. Chem Geol 236:1-12
Gardner JE, Hilton M, Carroll MR (1999) Experimental constraints on degassing of magma: isothermal bubble
growth during continuous decompression from high pressure. Earth Planet Sci Lett 168:201-218
Gardner JE, Rutherford M, Carey S, Sigurdsson H (1995) Experimental constraints on pre-eruptive water
contents and changing magma storage prior to explosive eruptions of Mount St. Helens volcano. Bull
Volcanol 57:1-17
Geschwind C, Rutherford MJ (1995) Crystallization of microlites during magma ascent: the fluid mechanics of
recent eruptions at Mount St. Helens. Bull Volcanol 57:356-370
Ghiorso MS, Sack RO (1995) Chemical mass transfer in magmatic processes IV. A revised and internally
consistent thermodynamic model for the interpolation and extrapolation of liquid-solid equilibria in
magmatic systems at elevated temperatures and pressures. Contrib Mineral Petrol 119:197-212
Gibbs JW (1906) On the equilibrium of heterogeneous substances. In: The Scientific Papers of J. Willard Gibbs,
1. Ox Bow Press, Woodbridge. p 55-353
Gonde C, Massare D, Bureau H, Martel C, Pichavant M, Clocchiatti R (2006) In situ study of magmatic
processes: a new experimental approach. High Pressure Res 26:243-250
Gonzalez-Oliver CJR, James PF (1980) Crystal nucleation and growth in a Na2O-2CaO-3SiO2 glass. J Non-
Cryst Solids 38/39:699-704
Granasy L, James PF (1998) Nucleation in oxide glasses: comparison of theory and experiment. Proc R Soc
London Ser A-Math Phys Eng Sci 454:1745-1766
Granasy L, Pusztai T, Hartmann E (1996) Diffuse interface model of nucleation. J Crystal Growth 167:756-
765
Granasy L, Pusztai T, James PF (2002) Interfacial properties deduced from nucleation experiments: A Cahn-
Hilliard analysis. J Chem Phys 117:6157-6168
Grove TL (1978) Cooling histories of Luna 24 very low Ti(VLT) ferrobasalts; an experimental study. Proc Lunar
Planet Sci Conf 9:565-584
Experimental Studies of Magma Crystallization 55
Grove TL, Bence AE (1979) Crystallization kinetics in a multiply saturated basalt magma; an experimental
study of Luna 24 ferrobasalt. Lunar Planet Sci Conf x:439-478
Grove TL, Detrick RS, Honnorez J, Adamson AC, Brass GW, Gillis KM, Humphris SE, Mevel C, Meyer PS,
Petersen N, Rautenschlein M, Shibata T, Staudigel H, Wooldridge AL, Yamamoto K, Bryan WB, Juteau T,
Autio LK, Becker K, Bina MM, Eissen J-P, Fujii T, Grove TL, Hamano Y, Hebert R, Komor SC, Kopietz
J, Krammer K, Loubet M, Moos D, Richards HG (1990) Cooling histories of lavas from Serocki Volcano.
Proc Oceanic Drilling Prog 106/109:3-8
Grove TL, Juster TC (1989) Experimental investigations of low-Ca pyroxene stability and olivine pyroxene
liquid equilibria at 1-atm in natural basaltic and andesitic liquids. Contrib Mineral Petrol 103:287-305
Grove TL, Raudsepp M (1978) Effects of kinetics on the crystallization of quartz normative basalt 15597; an
experimental study. In: Petrogenic Studies; The Moon and Meteorites, vol. l. Merrill RB (ed) Pergamon,
p 585-599
Gutzow I, Kashchiev D, Avramov I (1985) Nucleation and crystallization in glass-forming melts: old problems
and new questions. J Non-Cryst Solids 73:477-499
Gutzow I, Schmelzer J (1995) The Vitreous State: Thermodynamics, Structure, Rheology, and Crystallization.
Springer, Berlin
Hammer JE (2004) Crystal nucleation in hydrous rhyolite: Experimental data applied to classical theory. Am
Mineral 89:1673-1679
Hammer JE (2005) Strange attractors: symbiosis in magma crystallization. Eos, Trans AGU, 86, p. V12A-07
Hammer JE (2006) Influence of fO2 and cooling rate on the kinetics and energetics of Fe-rich basalt crystallization.
Earth Planet Sci Lett 248:618-637
Hammer JE (2008) Application of a textural geospeedometer to late-stage magmatic history of MIL03346.
Meteoritics Planet Sci (in press)
Hammer JE, Cashman KV, Hoblitt RP, Newman S (1999) Degassing and microlite crystallization during pre-
climactic events of the 1991 eruption of Mt. Pinatubo, Philippines. Bull Volcanol 60:355-380
Hammer JE, Rutherford MJ (2002) An experimental study of the kinetics of decompression-induced
crystallization in silicic melt. J Geophys Res 107 doi:10.1029/2001JB000281
Hammer JE, Rutherford MJ (2003) Petrologic indicators of preeruption magma dynamics. Geology 31:79-82
Hammer JE, Rutherford MJ, Hildreth W (2002) Magma storage prior to the 1912 eruption at Novarupta, Alaska.
Contrib Mineral Petrol 144:144-162
Harms E, Gardner JE, Schmincke HU (2004) Phase equilibria of the Lower Laacher See Tephra (East Eifel,
Germany): constraints on pre-eruptive storage conditions of a phonolitic magma reservoir. J Volc Geotherm
Res 134:125-138
Higgins MD (1996) Magma dynamics beneath Kameni volcano, Thera, Greece, as revealed by crystal size and
shape measurements. J Volc Geotherm Res 70:37-48
Higgins MD, Roberge J (2003) Crystal size distribution of plagioclase and amphibole from Soufriëre Hills
Volcano, Montserrat: Evidence for dynamic crystallization-textural coarsening cycle. J Petrol 44:1401-
1411
Hilliard JE (1968) Measurement of volume in volume. In: quantitative Microscopy. DeHoff RT, Rhines FN,
(eds) McGraw-Hill, New York, p 45-76
Holloway JR, Burnham CW (1972) Melting relations of basalt with equilibrium water pressure less than total
pressure. J Petrol 13:1-29
Holloway JR, Dixon JE, Pawley AR (1992) An internally heated, rapid quench, high-pressure vessel. Am
Mineral 77:643-646
Holness MB (2006) Melt-Solid dihedral angles of common minerals in natural rocks. J Petrol 47:791-800
Holtz F, Sato H, Lewis J, Behrens H, Nakada S (2005) Experimental petrology of the 1991-1995 Unzen dacite,
Japan. Part I: Phase relations, phase composition and pre-eruptive conditions. J Petrol 46:319-337
Hurwitz S, Navon O (1994) Bubble nucleation in rhyolitic melts: Experiments at high pressure, temperature, and
water content. Earth Planet Sci Lett 122:267-280
Ikeda S, Toriumi M, Yoshida H, Shimizu I (2002) Experimental study of the textural development of igneous
rocks in the late stage of crystallization: the importance of interfacial energies under non-equilibrium
conditions. Contrib Mineral Petrol 142:397-415
Ikeda Y (1998) Petrology of magmatic silicate inclusions in the Allan Hills 77005 lherzolitic shergottite.
Meteoritics Planet Sci 33:803-812
Jackson KA (1967) Current concepts in crystal growth from the melt. In: Progress in Solid State Chemistry, 4.
Reiss H (ed) Pergamon, New York, p 53-80
James PF (1974) Kinetics of crystal nucleation in lithium silicate glasses. Phys Chem Glasses 15:95-105
James PF (1985) Kinetics of crystal nucleation in silicate glasses. J Non-Cryst Solids 73:517-540
Kalinina AM, Fokin VM, Filipovich VN (1976) Method for determining the parameters characterizing the
formation of crystals in glasses. Fizika i Khimiya Stekla 2:298-305
Kashchiev D (1969) Solution of the non-steady state problem in nucleation kinetics. Surface Science 14:209-
220
56 Hammer
Kelton KF, Greer AL (1988) Test of classical nucleation theory in a condensed system. Phys Rev B 38:10089-
10092
Kennedy AK, Lofgren GE, Wasserburg GJ (1993) An experimental study of trace element partitioning between
olivine, orthopyroxene and melt in chondrules; equilibrium values and kinetic effects. Earth Planet Sci
Lett 115:177-195
Kent AJR (2008) Melt inclusions in basaltic and related volcanic rocks. Rev Mineral Geochem 69:273-331
Kern H (1982) P- and S-wave velocities in crustal and mantle rocks under the simultaneous action of high
confining pressure and high temperature and the effect of the rock microstructure. In: High-Pressure
Researches in Geoscience. Schreyer W (ed) Schweizerbart, Stuttgart, p 15-45
Kesson U, Lindqvist JE, Gransson M, Stigh J (2001) Relationship between texture and mechanical properties of
granites, Central Sweden, by use of image-analysing techniques. Bull Eng Geol Environ 60:277-284
Kirkpatrick RJ (1981) Kinetics of crystallization of igneous rocks. Rev Mineral Geochem 8:321-397
Kohut E, Nielsen RL (2004) Melt inclusion formation mechanisms and compositional effects in high-An feldspar
and high-Fo olivine in anhydrous mafic silicate liquids. Contrib Mineral Petrol 147:684-704
Kouchi A (1986) Effect of stirring on crystallization kinetics of basalt texture and element partitioning. Contrib
Mineral Petrol 93:429-438
Laporte D, Watson EB (1995) Experimental and theoretical constraints on melt distribution in crustal sources—
the effect of crystalline anisotropy on melt interconnectivity. Chem Geol 124:161-184
Larsen JF (2005) Experimental study of plagioclase rim growth around anorthite seed crystals in rhyodacitic
melt. Am Mineral 90:417-427
Larsen JF (2006) Rhyodacite magma storage conditions prior to the 3430 yBP caldera-forming eruption of
Aniakchak volcano, Alaska. Contrib Mineral Petrol 152:523-540
Lasaga AC (1998) Kinetic Theory in the Earth Sciences. Princeton University Press, Princeton
Lesher CE, Cashman KV, Mayfield JD (1999) Kinetic controls on crystallization of Tertiary North Atlantic
basalt and implications for the emplacement and cooling history of lava at Site 989, Southeast Greenland
rifted margin. Proc Ocean Drilling Prog 163:135-148
Lofgren G (1983) Effect of heterogeneous nucleation on basaltic textures; a dynamic crystallization study. J
Petrol 24:229-255
Lofgren G (1987) Internally heated systems. In: Hydrothermal Experimental Techniques. Ulmer GC, Barnes HL
(eds) John Wiley & Sons, New York, p 325-332
Lofgren GE (1980) Experimental studies on the dynamic crystallization of silicate melts. In: Physics of
Magmatic Processes. Hargraves R (ed) Princeton University Press, p 487-551
Lofgren GE, Donaldson CH, Williams RJ, Mullins O, Usselman TM (1974) Experimentally produced textures
and mineral chemistry of Apollo 15 quartz normative basalts. Lunar Planet Sci Conf 5:549-568
Lofgren GE, Huss GR, Wasserburg GJ, Shearer C, Vaniman D, Labotka T (2006) An experimental study of
trace-element partitioning between Ti-Al-clinopyroxene and melt; equilibrium and kinetic effects including
sector zoning. Am Mineral 91:1596-1606
Lofgren GE, Lanier AB (1992) Dynamic crystallization experiments on the Angra-Dos-Reis achondritic
meteorite. Earth Planet Sci Lett 111:455-466
London D, Morgan GB VI (1998) Experimental crystal growth from undercooled granitic melts; nucleation
response, texture, and crystallization sequence. Eos, Trans AGU, 79:366
Mangan MT (1990) Crystal size distribution systematics and the determination of magma storage times; the
1959 eruption of Kilauea Volcano, Hawaii. J Volc Geotherm Res 44:295-302
Marsh BD (1988) Crystal size distributions (CSD) in rocks and the kinetics and dynamics of crystallization I.
Theory. Contrib Mineral Petrol 99:277-291
Marsh BD (1998) On the interpretation of crystal size distributions in magmatic systems. J Petrol 39:553-599
Marsh BD (2007) Crystallization of silicate magmas deciphered using crystal size distributions. J Am Ceramic
Soc 90:746-757
Martel C, Ali AR, Poussineau S, Gourgaud A, Pichavant M (2006) Basalt-inherited microlites in silicic magmas:
Evidence from Mount Pelee (Martinique, French West Indies). Geology 34:905-908
Martel C, Bourdier JL, Traineau H, Pichavant M, Holtz F, Scaillet B (1999) Effects of f(O2) and H2O on andesite
phase relations between 2 and 4 kbar. J Geophys Res 104:29,453-29,470
Martel C, Schmidt BC (2003) Decompression experiments as an insight into ascent rates of silicic magmas.
Contrib Mineral Petrol 144:397-415
McCanta MC, Rutherford MJ, Hammer JE (2007) Pre-eruptive and syn-eruptive conditions in the Black Butte,
California dacite: Insight into crystallization kinetics in a silicic magma system. J Volc Geotherm Res
160:263-284
McMillan PF (1994) Water solubility and speciation models. Rev Mineral 30:131-156
Métrich N, Wallace PJ (2008) Volatile abundances in basaltic magmas and their degassing paths tracked by melt
inclusions. Rev Mineral Geochem 69:363-402
Milman-Barris MS, Beckett JR, Baker MB, Hofmann EA, Morgan z, Crowley MR, Vielzeuf D, Stolper E
(2008) zoning of phosphorus in igneous olivine. Contrib Mineral Petrol 155:739-765
Experimental Studies of Magma Crystallization 57
Moore G (2008) Interpreting H2O and CO2 contents in melt inclusions: constraints from solubility experiments
and modeling. Rev Mineral Geochem 69:333-361
Moore G, Carmichael ISE (1998) The hydrous phase equilibria (to 3 kbar) of an andesite and basaltic andesite
from western Mexico: constraints on water content and conditions of phenocryst growth. Contrib Mineral
Petrol 130:304-319
Muller R, zanotto ED, Fokin VM (2000) Surface crystallization of silicate glasses: nucleation sites and kinetics.
J Non-Cryst Solids 274:208-231
Mullin JW (1980) Bulk crystallization. In: Crystal Growth. Pamplin BR (ed) Pergamon, Oxford, p 521-565
Muncill GE, Lasaga AC (1987) Crystal-growth kinetics of plagioclase in igneous systems: One atmosphere
experiments and application of a simplified growth model. Am Mineral 72:299-311
Muncill GE, Lasaga AC (1988) Crystal-growth kinetics of plagioclase in igneous systems: Isothermal H2O-
saturated experiments and extension of a growth model to complex silicate melts. Am Mineral 73:982-
992
Nicholis MG, Rutherford MJ (2004) Experimental constraints on magma ascent rate for the Crater Flat volcanic
zone hawaiite. Geology 32:489-492
Okamura FP, McCallum IS, Stroh JM, Ghose S (1976) Pyroxene-spinel intergrowths in lunar and terrestrial
pyroxenes. Proc 7th Lunar Planet Sci Conf 2:1889-1899
Park Y, Hanson B (1999) Experimental investigation of Ostwald-ripening rates of forsterite in the haplobasaltic
system. J Volc Geotherm Res 90:103-113
Pertermann M, Lundstrom CC (2006) Phase equilibrium experiments at 0.5 GPa and 1100-1300 degrees C on a
basaltic andesite from Arenal volcano, Costa Rica. J Volc Geotherm Res 157:222-235
Pichavant M, Costa F, Burgisser A, Scaillet B, Martel C, Poussineau S (2007) Equilibration scales in silicic to
intermediate magmas - Implications for experimental studies. J Petrol 48:1955-1972
Pichavant M, Macdonald R (2007) Crystallization of primitive basaltic magmas at crustal pressures and genesis
of the calc-alkaline igneous suite: experimental evidence from St Vincent, Lesser Antilles arc. Contrib
Mineral Petrol 154:535-558
Pichavant M, Martel C, Bourdier JL, Scaillet B (2002) Physical conditions, structure, and dynamics of
a zoned magma chamber: Mount Pelee (Martinique, Lesser Antilles Arc). J Geophys Res 107 doi:
10.1029/2001JB000315
Porter D, Easterling K (1997) Phase Transformations in Metals and Alloys. Chapman & Hall, London
Pupier E, Duchene S, Toplis MJ (2008) Experimental quantification of plagioclase crystal size distribution
during cooling of a basaltic liquid. Contrib Mineral Petrol 155:555-570
Ree T, Eyring H (1958) The relaxation theory of transport phenomena. In: Rheology: Theory and Applications,
2. Eirich FR (ed) Academic Press, New York
Rose LA, Brenan JM (2001) Wetting properties of Fe-Ni-Co-Cu-O-S melts against olivine: Implications for
sulfide melt mobility. Econ Geol Bull Soc Econ Geol 96:145-157
Roskosz M, Toplis MJ, Richet P (2005) Experimental determination of crystal growth rates in highly supercooled
aluminosilicate liquids; implications for rate-controlling processes. Am Mineral 90:1146-1156
Roux J, Holtz F, Lefevre A, Schulze F (1994) A reliable high-temperature setup for internally heated pressure
vessels; applications to silicate melt studies. Am Mineral 79:1145-1149
Roux J, Lefevre A (1992) A fast-quench device for internally heated pressure vessels. Eur J Mineral 4:279-
281
Rutherford MJ (2008) Magma ascent rates. Rev Mineral Geochem 69:241-271
Rutherford M, Devine J (1988) The May 18, 1980, eruption of Mount St. Helens 3. Stability and chemistry of
amphibole in the magma chamber. J Geophys Res 93:11,949-11,959
Rutherford MJ, Devine JD (1996) Pre-eruption pressure-temperature conditions and volatiles in the 1991 dacitic
magma of Mount Pinatubo. In: Fire and Mud: Eruptions and Lahars of Mount Pinatubo, Philippines.
Newhall C, Punongbayan R (eds) University of Washington Press, Seattle, p. 751-766
Rutherford MJ, Sigurdsson H, Carey S, Davis A (1985) The May 18, 1980, eruption of Mount St. Helens 1.
Melt composition and experimental phase equilibria. J Geophys Res 90:2929-2947
Sato H (1995) Textural difference between pahoehoe and A’a lavas of Izu-Oshima Volcano, Japan - an
experimental study on population density of plagioclase. J Volc Geotherm Res 66:101-113
Scaillet B, Evans BW (1999) The 15 June 1991 eruption of Mount Pinatubo. I. Phase equilibria and pre-
eruption P-T-fO2-fH2O conditions of the dacite magma. J Petrol 40:381-411
Scaillet B, Pichavant M, Roux J (1995) Experimental crystallization of leukogranite magmas. J Petrol 36:663-
705
Scaillet B, Pichavant M, Roux J, Humbert G, Lefevre A (1992) Improvements of the Shaw membrane technique
for measurement and control of fH2 at high temperatures and pressures. Am Mineral 77:647-655
Schwandt CS, McKay GA (2006) Minor- and trace-element sector zoning in synthetic enstatite. Am Mineral
91:1607-1615
Shafer FN, Foley SF (2002) The effect of crystal orientation on the wetting behavior of silicate melts on the
surfaces of spinel peridotite minerals. Contrib Mineral Petrol 143:254-261
58 Hammer
Sharp TG, Stevenson RJ, Dingwell DB (1996) Microlites and “nanolites” in rhyolitic glass; microstructural
and chemical characterization. Bull Volcanol 57:631-640
Shaw HR (1963) Obsidian-H2O viscosities at 1000 and 2000 bars in the temperature range 700-900 °C. J
Geophys Res 68:6337-6343
Shneidman VA (1988) Establishment of a steady-state nucleation regime. Theory and comparison with
experimental data for glasses. Soviet Phys-Tech Phys 33:1338-1342
Silver L, Stolper E (1989) Water in albitic glasses. J Petrol 30:667-709
Simakin AG, Armienti P, Epel’baum MB (1999) Coupled degassing and crystallization; experimental study at
continuous pressure drop, with application to volcanic bombs. Bull Volcanol 61:275-287
Simakin AG, Armienti P, Salova TP (2000) Joint degassing and crystallization: Experimental study with a
gradual pressure release. Geochem Intl 38:523-534
Simakin AG, Chevychelov VY (1995) Experimental studies of feldspar crystallization from the granitic melt
with various water content. Geokhimiya 216-238
Simakin AG, Salova TP (2004) Plagioclase crystallization from a hawaiitic melt in experiments and in a
volcanic conduit. Petrol 12:82-92
Simakin AG, Salova TP, Armienti P (2003) Kinetics of clinopyroxene growth from a hydrous hawaiite melt.
Geochem Intl 41:1165-1175
Sisson TW, Grove TL (1993) Experimental investigations of the role of H2O in calc-alkaline differentiation and
subduction zone magmatism. Contrib Mineral Petrol 113:143-166
Skripov VP, Faizullin Mz (2005) Solid-liquid and liquid-vapor phase transitions: similarities and differences.
In: Nucleation Theory and Applications. Schmelzer JWP (ed) Wiley. p. 4-38
Smith CS (1964) Some elementary principles of polycrystalline microstructure. Metal Rev 9:1-48
Spaepen F (1994) Homogeneous nucleation and the temperature dependence of the crystal-melt interfacial
tension. Solid State Phys Adv Res Appl 47:1-32
Streck MJ (2008) Mineral textures and zoning as evidence for open system processes. Rev Mineral Geochem
69:595-622
Sunagawa I (1981) Characteristics of crystal growth in nature as seen from the morphology of mineral crystals.
Bull Mineral 104:81-87
Suzuki Y, Gardner JE, Larsen JF (2007) Experimental constraints on syneruptive magma ascent related to the
phreatomagmatic phase of the 2000AD eruption of Usu volcano, Japan. Bull Volcanol 69:423-444
Swanson SE (1977) Relation of nucleation and crystal-growth rate to the development of granitic textures. Am
Mineral 62:966-978
Swanson SE, Fenn PM (1986) quartz crystallization in igneous rocks. Am Mineral 71:331-342
Sycheva GA (1998) Surface energy at the crystal nucleus glass interface in alkali silicate glasses. Glass Phys
Chem 24:342-347
Szramek L, Gardner JE, Larsen J (2006) Degassing and microlite crystallization of basaltic andesite magma
erupting at Arenal volcano, Costa Rica. J Volc Geotherm Res 157:182-201
Tolman RC (1949) The effect of droplet size on surface tension. J Chem Phys 17:333-337
Toshiya A, Katsuo T, Ichiro S (1991) Nucleation, growth and stability of CaAl2Si2O8 polymorphs. Phys Chem
Mineral 17:473-484
Turcotte DL, Schubert G (1982) Geodynamics: Applications of Continuum Physics to Geological Problems.
John Wiley & Sons, New York
Turnbull D (1964) Physics of non-crystalline solids. In: Physics of Non-crystalline Solids. Proc Int Conf. Prins
JA (ed) North-Holland, Deift, p 41-56
Turnbull D, Cohen M (1960) Crystallization kinetics and glass formation. In: Modern Aspects of the Vitreous
State. MacKenzie JD (ed) Butterworth and Co., London, p 38-62
Ulmer GC (1971) Research Techniques for High Pressure and High Temperature. Springer Verlag
Ulmer GC, Barnes HL (1987) Hydrothermal Experimental Techniques. John Wiley & Sons, New York
Vetere F, Behrens H, Holtz F, Neuville DR (2006) Viscosity of andesitic melts: new experimental data and a
revised calculation model. Chem Geol 228:233-245
Volmer M, Weber A (1926) Kimbildung in ubersattingten gebilden. z Phys Chem 119:277-301
Waff HS, Faul UH (1992) Effects of crystalline anisotropy on fluid distribution in ultramafic partial melts. J
Geophys Res 97:9003-9014
Walker D, Kirkpatrick RJ, Longhi J, Hays JF (1976) Crystallization history of lunar picritic basalt Sample
12002; phase-equilibria and cooling-rate studies. Geol Soc Am Bull 87:646-656
Walker D, Powell MA, Lofgren GE, Hays JF (1978) Dynamic crystallization of a eucrite basalt. Proc 9th Lunar
Planet Sci Conf 1:1369-1391
Wallace PJ (2005) Volatiles in subduction zone magmas: concentrations and fluxes based on melt inclusion and
volcanic gas data. J Volc Geotherm Res 140:217-240
Wallace PJ, Gerlach TM (1994) Magmatic vapor source for sulfur dioxide released during volcanic eruptions;
evidence from Mount Pinatubo. Science 265:497-499
Experimental Studies of Magma Crystallization 59
Wang YP, Lu K (2000) Crystallization kinetics of amorphous Ni80P20 alloy investigated by electrical resistance
measurements. z Metallk 91:285-290
Watson EB (1996) Surface enrichment and trace-element uptake during crystal growth. Geochim Cosmochim
Acta 60:5013-5020
Watson EB (2004) A conceptual model for near-surface kinetic controls on the trace-element and stable isotope
composition of abiogenic calcite crystals. Geochim Cosmochim Acta 68:1473-1488
Watson EB, Liang Y (1995) A simple model for sector zoning in slowly grown crystals: Implications for
growth rate and lattice diffusion, with emphasis on accessory minerals in crustal rocks. Am Mineral
80:1179-1187
Webster JD, Holloway JR, Hervig RL (1987) Phase equilibria of a Be, U and F-enriched vitrophyre from Spor
Mountain, Utah. Geochem Cosmochim Acta 51:389-402
zhang J, Wei YH, qiu Kq, zhang HF, quan Mx, Hu zq (2003) Crystallization kinetics and pressure effect on
crystallization of zr55Al10Ni5Cu30 bulk glass. Mater Sci Eng Struct Mater Prop Microstruct Proc 357:386-
392
zieg MJ, Lofgren GE (2006) An experimental investigation of texture evolution during continuous cooling. J
Volc Geotherm Res 154:74-88
zieg MJ, Marsh BD (2002) Crystal size distributions and scaling laws in the quantification of igneous textures.
J Petrol 43:85-101
Reviews in Mineralogy & Geochemistry
Vol. 69 pp. 61-120, 2008 3
Copyright © Mineralogical Society of America
Thermometers and Barometers for Volcanic Systems
Keith D. Putirka
Department of Earth and Environmental Sciences
California State University, Fresno
2576 E. San Ramon Ave., MS/ST24
Fresno, California, 93740-8039, U.S.A.
[email protected]
INTRODUCTION
Knowledge of temperature and pressure, however qualitative, has been central to our views
of geology since at least the early 19th century. In 1822, for example, Charles Daubeny presented
what may be the very first “Geological Thermometer,” comparing temperatures of various
geologic processes (Torrens 2006). Daubeny (1835) may even have been the first to measure
the temperature of a lava flow, by laying a thermometer on the top of a flow at Vesuvius—albeit
several months following the eruption, after intervening rain (his estimate was 390 °F). In any
case, pressure (P) and temperature (T) estimation lie at the heart of fundamental questions: How
hot is Earth, and at what rate has the planet cooled. Are volcanoes the products of thermally
driven mantle plumes? Where are magmas stored, and how are they transported to the surface—
and how do storage and transport relate to plate tectonics? Well-calibrated thermometers and
barometers are essential tools if we are to fully appreciate the driving forces and inner workings
of volcanic systems.
This chapter presents methods to estimate the P-T conditions of volcanic and other igneous
processes. The coverage includes a review of existing geothermometers and geobarometers, and
a presentation of approximately 30 new models, including a new plagioclase-liquid hygrometer.
Our emphasis is on experimentally calibrated “thermobarometers,” based on analytic expressions
using P or T as dependent variables. For numerical reasons (touched on below) such expressions
will always provide the most accurate means of P-T estimation, and are also most easily employed.
Analytical expressions also allow error to be ascertained; in the absence of estimates of error,
P-T estimates are nearly meaningless. This chapter is intended to complement the chapters by
Anderson et al. (2008), who cover granitic systems, and by Blundy and Cashman (2008) and
Hansteen and Klügel (2008), who consider additional methods for P estimation.
THEORETICAL ASPECTS
At the foundation of well-founded belief lies belief that is not founded
– Ludwig Wittgenstein
In the search for useful geothermometers or geobarometers, the goal is to find some
chemical equilibrium where there is a significant difference between the entropy (∆Sr) (for a
thermometer) or volume (∆Vr) (for a barometer) of products and reactants. Consider:
0.5 + AlO1.5 + 2SiO 2 = NaAlSi 2 O 6
NaOliq liq liq cpx
(1)
where superscripts denote phase (liq = liquid; cpx = clinopyroxene); NaAlSi2O6cpx is the jadeite
(Jd) content of clinopyroxene. To compare S and V, we use partial molar entropies ( S) and
1529-6466/08/0069-0003$10.00 DOI: 10.2138/rmg.2008.69.3
62 Putirka
partial molar volumes (V ), where S = S/n and V = V/n, and n = number of moles. In Equation
(1), ∆Sr = S Jd − S NaOliq0.5 − S AlO1.5liq − 2 S SiOliq2 , ∆Vr = V Jd − V NaOliq0.5 − V AlO1liq.5 − 2V SiOliq2 . The equilibrium
constant, Keq, is:
cpx
aJd
K eq = (2)
( )
2
liq
aNaO 0.5
⋅ aAl
liq
2 O3
⋅ aSiO
liq
2
cpx liq
where terms such as aJd and aNaO 0.5
represent the activities of Jd in clinopyroxene, and SiO2
in liquid respectively. Activity is something like an ‘effective concentration’, and in the ideal
case, is equivalent to mole fraction: aJd cpx
= X Jdcpx, where X Jdcpx is the mole fraction of Jd in a
clinopyroxene solid solution. In non-ideal solutions, the two quantities can be related to one
another by an activity coefficient, λ, so that aJd cpx
= λ cpx cpx
Jd X Jd .
A “good” thermometer has a large ∆Sr and a good barometer has a large ∆Vr. But what is
“good”? For metamorphic systems, Essene (1982) suggests that ∆Sr ≥ 4.0 J/mole⋅K and ∆Vr ≥
0.2 J/bar; these limits perhaps also apply to igneous systems. The presence of a liquid, though,
poses hurdles to the igneous petrologist. Consider the very large number of distinct equilibria
used for P-T estimation in metamorphic systems (Spear 1993). Far fewer are available to
igneous petrologists. Why? To begin with, igneous rocks carry fewer crystalline phases. And
the numbers of crystalline phases is perhaps in turn limited by the presence of the liquid itself. A
silicate liquid is a ‘high variance’ phase in that, compared to minerals, it is highly compressible
and has no compositional stoichiometric constraints. Liquids can thus absorb significant
changes in P, T and composition ( Xij , the mole fraction of i in some phase j), by expanding,
contracting or mixing, without resorting to the extreme measure of precipitating a new phase.
Metamorphic rocks have no such luxury. Carrying nothing but crystalline phases, with sharp
limits on their ability to expand, contract or mix, new phases readily nucleate in response to
changing P-T- Xij . This results in sharp changes in S and V in metamorphic systems, which
can be exploited to formulate geothermobarometers. By their versatility, silicate liquid-bearing
systems are intrinsically less amenable to thermobarometer formulation—but all is not lost.
Perhaps the single most useful equation for petrologists is:
− RT ln K eq = ∆Gr ° (3)
where R is the gas constant (R = 1.9872 cal/K·mole, or 8.3144 J/dK·mole) and ∆Gr° is the
Gibbs free energy change of “reaction” for a balanced equilibrium, such as Equation (1). In
∆Gr°, the superscript ° means that the Gibbs free energy is for a standard state (pure substances
at 1 bar), an important distinction from ∆Gr, which is the Gibbs free energy change between
products and reactants at any arbitrary P and T. At equilibrium, ∆Gr = 0, by definition. But
rarely is ∆Gr = 0 at a standard state. So when ∆Gr = 0, it is usual that ∆Gr° ≠ 0, which means
that usually Keq ≠ 1.
Despite its name, however, Keq is only constant at fixed P and T. Driven by differences in
∆Sr and ∆Vr , Keq can vary greatly as a function of P and T. So for example, in Equation (1), if
NaAlSi 2O6cpx has a smaller partial molar volume compared to the collective partial molar vol-
umes of the reactants (∆Vr > 0), then with increased P, Equation (1) will shift to the right and Keq
will increase. Similarly, higher T favors whichever side of an equilibrium that has higher partial
molar entropy. Equation (3) provides a quantitative footing to relate Keq to P and T and is at the
root of all thermometers and barometers based on ‘theoretical’ or thermodynamic models.
The differentiation of “empirical” and “theoretical” is not always clear. The laws of
thermodynamics are rooted in fundamentally empirical expressions derived by Robert Boyle,
who in 1662 determined that volume, V, is related to P by the expression: V = C/P, where
C is a constant. In the 19th century, Jacques Charles and Joseph-Louis Gay-Lussac, further
illuminated the phenomenon of thermal expansion in gases: V = αVoT, where α is the coefficient
Thermometers & Barometers for Volcanic Systems 63
of thermal expansion, and Vo is some reference gas volume (see Castellan 1971). These two
expressions, known today as Boyle’s Law and Charles’ (or Gay-Lussac’s) Law are combined,
with Avogadro’s law (that equal volumes of gases have equal numbers of molecules), to
yield the ideal gas law: PV = nRT (which recognizes that the constant C in Boyle’s Law is
proportional to n). The relevance is that the form of Equation (3) is not accidental: it derives
from the ideal gas law and an assumption that activities can be represented by fugacities or
partial pressures, where we integrate (∂G/∂P)T = V by substituting V = nRT/P (see Nordstrom
and Munoz 1986). In effect, our most sophisticated and complex activity models for silicate
liquids and minerals are based on the assumption that they act, in the phrase of David Walker,
as a “perverted ideal gas.” In the event, practice and experience indicate that models based on
−RTlnKeq = ∆Gr° are more accurate than those based on more arbitrary relationships and so
provide the greatest promise for extrapolation outside calibration boundaries.
To make use of lnKeq = −∆Gr°/RT, we use the relationship ∆Gr° = ∆Hr° − T∆S° (where ∆Hr°
and ∆Sr° are respectively the enthalpy and entropy differences for a chemical equilibrium at stan-
dard state), and the expressions (∂∆Gr/∂P)T = ∆Vr and (∂∆Gr/∂P)P = ∆Sr. If P is constant, and the
differences in heat capacities (Cp) between products and reactants are effectively equal (∆Cp = 0)
(meaning that ∆Hr and ∆Sr are independent of T), then substitution yields the following:
lnKeq = −∆Hr°/RT + ∆Sr°/R (4)
If we know ∆Hr° and ∆Sr°, then we can use Equations (2) and (4) to calculate T, and we have
a thermometer. If ∆Hr° and ∆Sr° are unknown, we can preserve the functional relationship
between Keq and T in Equation (4) and use as a regression model: lnKeq = a/T + b, where a =
−∆Hr°/R and b = ∆Sr°/R; we can then derive a and b from isobaric experimental data, where
composition and T are known (and an activity model is assumed). Of course, P need not be
constant, and ∆Hr and ∆Sr will vary with T if ∆Cp ≠ 0. Products and reactants might also
exhibit significant differences in compressibility (β) or thermal expansion (α) such that, over a
wide enough P interval, ∆βr ≠ 0 and ∆αr ≠ 0. The following regression equation accounts for
such changes (see Putirka 1998):
A T P P2
ln K eq = + B + C ln + D + E + FP + GP 2 (5)
T To T T
where A = −∆Hr°/R, B = ∆Sr°/R, C = ∆Cp/R, D = −∆Vr°/R, E = ∆βr ∆Vr°/2R, F = ∆αr∆Vr°/R, and
G = ∆αr∆βr∆Vr°/2R. Equation (5) contains all the terms in Equation (4), plus additional terms
that correct for the cases that P is not constant, ∆Cp ≠ 0, etc. More terms could be added if ∆αr,
∆βr, and ∆Cp are not constant, though for most geological purposes Equation (5) is sufficiently
complex. In Equation (5), To is a reference temperature; in Putirka (1998) To was taken as the
melting point of a pure mineral, Tm, in which case reference values for ∆Hr°, ∆Sr°, etc., are for
fusions of pure substances. Equation (5) is also simplified in that the reference pressure (Po,
needed to calculate the effects of a non-zero ∆βr and ∆αr) is taken as 1 atm, or effectively zero on
a kilobar or GPa scale. If some other reference pressure is required, or if 1 bar is large compared
to the pressure of interest, then the term (P-Po) replaces P in each instance in Equation (5).
REGRESSION ANALYSIS & STATISTICAL CONSIDERATIONS
The stars might lie but the numbers never do
– Mary Chapin Carpenter
Experimental data, regression analysis & statistical considerations
Although some geothermometers or geobarometers are developed using calorimetric and
volumetric data, most derive from regression analysis of experimental data, where Keq, P and
T are all known. Partial melting experiments reported in the Library of Experimental Phase
64 Putirka
Relations (LEPR) (Hirschmann et al. 2008), and elsewhere, are here used to test existing,
and calibrate new thermobarometers. Regression analysis of such data yields regression
coefficients that minimize error for the selected dependent variable. This approach assumes
that the experiments used for regression approach equilibrium. There are several possible
validity tests, but none guarantee success. One is to check whether the regression-derived
coefficients compare favorably with independently determined values for quantities such as
∆Hr°, ∆Sr°, and ∆Vr°. One can also test model results for their ability to predict P or T for
natural systems where P and T might be known (e.g., a lava lake). Issues of error are discussed
in detail in the penultimate section of this chapter.
Regression strategies. For the calibration of new models, no data-filters are applied at
the outset, but generally, experiments that yield P or T estimates that are >3σ outside standard
errors of estimate (SEE) for remaining data are considered suspect and not used for calibration
or testing. And if particular studies contain multiple aberrant data, such studies are excluded.
Most new calibrations utilize “global” regressions, which use all available data to determine
the minimum number of parameters that are required to explain observed variability. All
tests make use of “leverage plots” to ensure that a given model parameter is significant and
explained by more than just one or a few potentially aberrant data (see JMP Statistics and
Graphics Guide). We also employ regression models where P and T appear as dependent
variables—a numerical necessity if one is to maximize precision.
Regression models. William of Ockham, a 14th century philosopher, suggested that
“plurality must not be posited without necessity” (Ockham’s razor), a recommendation
supported by Bayesian analysis (Jefferys and Berger 1992). Jefferys and Berger (1992) reword
this principle: “if a law has many adjustable parameters then it will be significantly preferred
to the simpler law only if its predictions are considerably more accurate”. To illustrate, they
show how Galileo’s quadratic expression (relating distance (x) and time (t) to the acceleration
due to gravity (g), i.e., x = xo + vt + 1/2gt2) is highly superior to a 6th order polynomial fit to
the same data. The problem is not whether additional coefficients in the 6th order equation
have theoretical meaning: gravity changes with elevation; this fact could be used to develop
additional terms beyond the familiar quadratic expression. But Galileo’s original experiments
have nowhere near the level of precision required to extract such information.
Similar issues apply to petrology. Using Equation (2) as an example, Figure 1a shows two
regression models: 1) a two-parameter model where lnKeq = aP/T + b, and 2) a five-parameter
model, where lnK = aP/T + bln(T) + c(P2/T) + dP2 + e. All of the coefficients (a, b, c, d, e)
have thermodynamic meaning (Eqn. 5), and the 5-parameter model describes a greater fraction
(99.8%) of the variation of lnKeq for the calibration data (ankaramites from Putirka et al. 1996).
However, the 5-parameter model yields gross systematic error when predicting lnKeq for “test”
data (Fig. 1b; all remaining data in Putirka et al. 1996), with a slope and intercept through lnKeq
(measured) vs. lnKeq (calculated) of 0.5 and 1.5 respectively. In contrast, our lowly 2-parameter
Figure 1 (on facing page). In (a) and (b), predicted values for lnKeq based on Equation (2) are used to
illustrate how model complexity, no matter how valid from theoretical considerations, can yield incorrect
predictions if model parameters are not supported by data. In (a), calculated and measured values of lnKeq
are compared using ankaramite bulk compositions from Putirka et al. 1996). For both the 2- and 5-parameter
equations, all regression coefficients have thermodynamic significance (Eqn. 5), and the 5-parameter
model provides a better fit to the calibration data. However, in (b) the 5-parameter equation yields strong
systematic error when used to predict lnKeq for experimental data not used for regression (as illustrated by
the slope and intercept of the regression line lnKeq(calculated) vs. lnKeq(measured); ideally they should be
1 and 0 respectively). In contrast, the 2-parameter model is nearly free of systematic error, and describes
the ‘test’ data with higher precision; the 5-parameter equation is of little use beyond interpolating between
calibration points, while the 2-parameter equation is a useful predictive tool. Models cannot be judged
solely by their fit to calibration data, nor by their thermodynamic complexity.
Thermometers & Barometers for Volcanic Systems 65
5
Calibration data (Ankaramites from Putirka et al. 1996)
lnKeq = aP/T + bln(T) + cP2/T + dP2 + e
slope = 1.1
4.5 intercept = - 0.252
R2 = 0.998
lnK (Measured)
5-parameter model
4 2-parameter model
lnKeq = aP/T +b
3.5 slope = 1.0
intercept = - 4x10-7
R2 = 0.96
a
2.5
2.5 3 3.5 4 4.5 5
lnK (Calculated)
5
Test data (remaining data from Putirka et al. 1996)
5-parameter model
lnKeq = aP/T +b 2-parameter model
4.5 slope = 0.99
intercept = - 0.13
lnK (Measured)
R2 = 0.57
3.5
3 lnKeq = aP/T + bln(T) + cP2/T + dP2 + e
slope = 0.5
intercept = 1.5
R2 = 0.48
b
2.5
2.5 3 3.5 4 4.5 5 5.5
lnK (Calculated)
Figure 1. caption on facing page
66 Putirka
model yields a higher correlation coefficient (R) for the test data, and is nearly absent systematic
error, with slope and intercept values much closer to the ideal of one and zero respectively. Box
and Draper (1987, p. 74) summarize the issue rather succinctly: “Remember that all models are
wrong; the practical question is how wrong do they have to be to not be useful.” In geothermo-
barometry, the answer is that only the first 4 terms of Equation (5) are usually needed.
Linear regression methods can be used to avoid the pitfalls noted in Figure 1 because
unlike non-linear methods, validity tests are available for regression coefficients; parameters
that are not supported by the data (no matter their thermodynamic significance), can be culled
from a potential model. An occasional roadblock is that equations such as Equation (5) cannot
always be rearranged so that P or T can be isolated as dependent variables. One solution is to
perform a linear regression using some other quantity as the independent variable, and then
rearrange the expression. But in multiple linear regression, error is minimized only for the
dependent variable (regression analysis is not symmetric), so error will not be minimized for
the variable of interest, and systematic error is assured. Equation (4), for example, will yield a
very good model for predicting lnKeq, but will be less precise, and yield systematic error, if the
same coefficients are rearranged to calculate T. Regression models that use P or T as dependent
variables will thus always be more precise and accurate compared to models that minimize
error on such quantities as lnKeq, Gibbs Free Energy, etc., an unavoidable consequence of
numerical issues of regression analysis. If variations in P are not of concern, Equation (4)
should be rearranged to yield:
1
= a ln K eq + b (6)
T
where a = −R/∆Hr° and b = ∆Sr°/∆Hr°. If variations in P are important, remaining choices are
to add P as an “empirical” factor,
1
= a ln K eq + b + cP ( 7)
T
or apply non-linear regression methods, using:
∆Hr ° + P∆V a + bP
T= = (8)
R ln K eq − ∆Sr ° c ln K eq + d
Here, ∆Cp = 0, and Po = 0 (i.e., 1 bar); if an additional P term is required, the term (1/2) P2∆βr ∆Vr°
can be added to the numerator (a, b, c, and d have thermodynamic meaning as in Equation (8),
but not are not identical to the coefficients in Equations (6-7)). A disadvantage to non-linear
regression is that it does not yield an absolute minimum for error on the dependent variable,
and there are no intrinsic validity tests for model parameters.
Activity modeling. Mixing and interaction between chemical components in an equilibrium
may result in non-ideal behavior, complicating the calculation of Keq. To address just such
compositional dependencies, Thompson (1967) developed a polynomial model for activity
coefficients, λi = ƒ(Xi). The approach uses so-called Margules parameters (WG), which represent
the energy of interaction between chemical components. Thompson’s (1967) parameters can
be substituted into Equation (3). For example, using the definitions for a two-parameter model
(WG1, and WG2), describing interaction between two components, X1 and X2, the Margules
model is: RTlnλ1 = (2WG2 − WG1)(X2)2 + 2(WG1 − WG2)(X2)3 and RTlnλ1 = (2WG1 − WG2)(X1)2
+ 2(WG2 − WG1)(X1)3 (Nordstrom and Munoz 1986). Because of the extrinsic dependency of λ
on T, such substitutions require non-linear regression methods for thermometers (see below).
As an alternative approach, consider forsterite (Fo, Mg2SiO4) melting,
Mg2SiO4ol = 2MgOliq + SiO4liq (9)
Thermometers & Barometers for Volcanic Systems 67
In the ideal case, aFool = XFool, aMgOliq = XMgOliq and aSiO2liq = XSiO2liq, and so Keqol-liq = XFool/
(XMgOliq·XSiO2liq). If the system is non-ideal and we know the values for λMgOliq, λFool, etc., the
activity coefficients can be inserted to yield Keqol-liq = λFoolXFool/(λMgOliqXMgOliq· λSiO2liqXSiO2liq).
O’Nions and Powell (1977) extract such terms to yield linear coefficients in Equations such as
(4) and (8). Using Equation (9) as an example,
ln K eq = ln
(
MgO )
a liq 2 ⋅ a liq
SiO2 (
= ln MgO MgO )
λ liq X liq 2 ⋅ λ liq X liq
SiO2 SiO2
(10)
ol
aFo λ olFo X Fo
ol
λ( ) ⋅ λ SiO2 (
X liq 2 ⋅ X liq )
2
liq liq
= ln + ln MgO
MgO SiO2
λ olFo ol
X Fo
(
= 2 ln λ liq ) ( ) ( ) (
MgO + 2 ln X MgO + ln λ SiO2 + ln X SiO2 − ln λ Fo − ln X Fo
liq liq liq ol ol
) ( ) ( )
where a term such as Xij refers to the mole fraction of component i in phase j. A possible
regression model based on Equations (6), (9) and (10) is thus:
1 ( )
X liq 2 ⋅ X liq
T
= a ln
MgO
ol
X Fo
SiO2
(
MgO )
SiO2 (
+ b + c ln λ liq + d ln λ liq + e ln λ ol
Fo ) ( ) (11)
where the coefficients a and b are as in Equation (6). For ideal mixing, the λ ij are all 1, and c,
d, and e are zero; if λ ij are known, then c = d = e = 1. For non-ideal mixing, λ ij can be related
to mole fractions, Xij, using
1 ( )
X liq 2 ⋅ X liq
T
= a ln
MgO
ol
X Fo
SiO2
(
MgO )
SiO2 (
+ b + c X liq + d ln X liq + e ln X ol
Fo ) ( ) (12)
where a and b are as in Equation (6), and c, d, and e are “empirical” regression coefficients
liq
that can be related to the activity coefficients: in the case of X MgO , we presume that there is
( )
some number c such that λ MgO = exp cX MgO , where c is as in Equation (12). Stormer (1975)
liq liq
shows how coefficients to a polynomial expansion can be equated to Margules parameters.
In any case, activity models are important here, only to the extent that they provide increased
precision when estimating P or T.
IGNEOUS THERMOMETERS AND BAROMETERS
All models are wrong, some models are useful
– George Box
History and background
James Hall (1798) presented the first attempts to measure the melting temperatures of
volcanic rocks. However, at the time there were few methods for determining temperatures above
the boiling point of mercury, and his use of the Wedgwood pyrometer (based on the volume
change in clays when heated) yielded widely ranging estimates (1029-3954 °C; see Hall’s “Table
of Fusibilities,” where °W ≈ 130 °F). Later experiments had much better temperature control,
with Joly (1891) using his “meldometer” and later experimentalists eventually employing the
thermocouple (e.g., Doelter 1902; Day and Allen 1905; Shepherd et al. 1909).
Some of the earliest temperature-controlled experimental studies of melting temperatures
for minerals and rocks (Joly 1891; Doelter 1902; Day and Allen 1905; Shepherd et al. 1909)
were no doubt conducted in part to understand the thermal conditions of partial melting and
68 Putirka
crystallization. But the first study to establish, with some accuracy, the temperature interval for
the crystallization of natural basaltic magmas was conducted by Sosman and Merwin (1913).
They determined that the solidus and liquidus temperatures for the Palisades Sill in New York are
≈1150 °C and 1300 °C respectively. Contributions by Bowen (1913, 1915) have outshone such
work. In his 1913 paper, Bowen established with accuracy the continuous solid solution of the
plagioclase feldspars, and then in 1915 applied experimental results, thermodynamic theory and
phase diagrams to show that a wide range of igneous rocks can be derived from basaltic parental
liquids by the process of fractional crystallization (Bowen 1915). But Sosman and Merwin’s
estimates were nonetheless largely correct: their estimation of the 1 atm temperature interval for
basalt crystallization matched well with Perret’s (1913) and Jagger’s (1917) direct temperature
estimates of freshly formed lava at Hawaii (est. 1050-1200 °C), and were qualitatively verified
by experiments in simple systems (Bowen 1915); together, these experimental and field-based
studies represent the first accurate determinations of the temperatures at which basaltic rocks
form. More thorough experimental studies (e.g., Tuttle and Bowen 1958; Yoder and Tilley 1962;
Green and Ringwood 1967) would later establish temperatures of partial crystallization for both
granitic and basaltic systems, over a range of P-T conditions.
None of these early efforts, though, established a means to estimate T or P for any particular
igneous rock. One of the first attempts to establish a true geothermometer was by Wright and
Larson (1909), using the optical and morphological properties of quartz. The first analytic
geothermometer, though, was proposed by Barth (1934) and presented by Barth (1951), who
developed an approach that is still in use today: the two-feldspar thermometer. Barth (1951)
presented a theoretical model based on the alkali-feldspar solvus, which was retracted in proof
when he discovered that his theoretical solvus did not match the experimentally determined
curve of Bowen and Tuttle (1950). Barth’s (1951) effort was followed by a Ti-in-magnetite
thermometer by Buddington et al. (1955) and Buddington and Lindsley (1964), a two-
pyroxene thermometer by Davis and Boyd (1966), a thermometer based on Ni partitioning
between olivine and liquid by Hakli and Wright (1967), and Barth’s own recalibrations of a
two-feldspar thermometer (Barth 1962, 1968). The following decade then saw a substantial
increase in the methods of igneous thermobarometry, with new calibrations of a plagioclase
thermometer by Kudo and Weill (1970), a new olivine thermometer by Roeder and Emslie
(1970), new calibrations of the two-feldspar thermometer (e.g., Stormer 1975) and the
calibration of multiple P- and T-sensitive igneous equilibria by Nicholls et al. (1971), Nicholls
and Carmichael (1972) and Bacon and Carmichael (1973), among others. Since the 1970’s the
number of igneous thermometers and barometers has greatly expanded.
Selection criteria. Within the space of a single chapter, it is possible to review neither all
the various igneous thermometers that have been calibrated, nor to detail their thermodynamic
bases. Emphasis is thus placed upon cataloging those tools that can be used widely, especially
for volcanic systems. The thermometers and barometers presented here are thus based on
commonly found minerals, using, for the most part, elemental components that are easily
determined in most laboratories.
Calculating “liquid” and mineral components, and activities
It is absolutely crucial that all components, be they minerals or liquids, are calculated in
precisely the same way as originally handled for a particular thermometer or barometer. Here,
component calculations that are common to many models shown below are discussed, but care
should be taken to examine specific component calculation procedures when presented below.
Several Tables are presented to illustrate nearly all the calculations presented in this
chapter (Tables 1-7); readers attempting to reproduce these and other numbers should note
the following: Table 1 in Putirka (1997) contains small errors relative to P and T (using the
models in Putirka et al. 1996); Table A2 in Putirka et al. (2007) contains small errors relative
to melt fraction F, and derivative calculations. All related equations are valid as published;
Thermometers & Barometers for Volcanic Systems 69
the errors result from the use of earlier versions of certain equations, later abandoned for the
published (more accurate) models (the Tables were not updated). In addition, some P and T
calculations presented here require numerical solution methods (when T and P depend upon
one another and both are to be calculated). Here and in other works P-T estimates are based
on the numerical module found within “Excel – Preferences – Calculation – Iteration”. This
feature may have numerical instabilities, and the “Solver” or “Goal Seek” options of Excel, or
other numerical methods, should perhaps be preferred.
Liquid components. In the models that follow, various calculation schemes are used to
represent the chemical components of silicate phases. For the empirical “glass” thermometers,
most liquid components are entered as wt%, so no special calculations are needed, except for
terms such as Mg#, which are always represented on a mole fraction basis, and calculated
as Mg#(liq) = (MgOliq/40.3)/[((MgOliq/40.3) + (FeOliq/71.85))], where terms such as MgOliq
represent the wt% of MgO in the liquid. For all other models, liquid components are calculated
either as a mole fraction or a cation fraction (Table 1), always on an anhydrous basis (even if
H2O occurs in the equation or analysis, so as to avoid estimating molecular H2O vs. OH−). The
purpose of such calculations is to be able to compare numbers of molecules, as this is how we
calculate Keq. Mole and cation fractions are very similar. For a mole fraction, the wt% of each
oxide is divided by its molecular weight; for example, Al2O3 would be divided by 101.96 and
Na2O would be divided by 61.98; this calculation yields a “mole proportion” for the oxides,
which is converted to a mole fraction by renormalizing the sum of the molecular proportions
so that they sum to 1.0 (see Table 1). In a chemical equilibrium, the Al content of a liquid, for
liq
example, would be written as Al2O3liq, and in a Keq, the mole fraction would appear as X Al 2 O3
.
Liquid components using cation fractions. For a cation fraction, each oxide is normalized
by the molecular weight of the oxide having only one cation. So Al2O3 is divided by the
molecular weight of AlO1.5, or 50.98; similarly, the wt% of Na2O is divided by the weight of
NaO0.5, or 30.99. Compared to the usual mole fraction, this approach doubles the “molecular”
units for oxides that contain two cations per formula unit, such as Al2O3, Na2O, K2O, Cr2O3,
and P2O5; using Al2O3 again as an example, we effectively divide one molecular unit of Al2O3
into two molecular units of AlO1.5. For oxides such as SiO2, which already have only one cation
per formula unit, the wt% of SiO2 is divided by 60.08, as usual. These “cation proportions”
are converted to a “cation fraction” by renormalizing each oxide by the sum of the cation
proportions, so that the new sum is 1.0 (Table 1). In a chemical equilibrium, the Al content of
liq
a liquid would be written as AlO1.5liq, and in a Keq, would appear as X AlO 1.5
.
The advantage of the cation fraction is that the relative numbers of cations can be compared
directly. Ratios of Fe/Mg (or equivalently, FeO/MgO) are the same whether a “mole” or cation
fraction are considered; but the cation fraction allows a more straightforward comparison of,
for example, Ca/Al (CaO/AlO1.5) or Na/Al (NaO0.5/AlO1.5) ratios. With a mineral composition,
cation fractions can also be multiplied by the number of cations in a mineral formula so that
the sum, instead of 1, is equal to the number of cations per formula unit (e.g., 2 for olivine, 4
for pyroxene, 5 for feldpsar); this procedure yields the number of cations per formula unit.
Activity models take these calculations further, recognizing that the actual chemical
species in liquids and minerals probably do not occur in the form of isolated, non-interacting
oxides. Two notable approaches for calculating the activities of liquid components are the two-
lattice (Bottinga and Weill 1972; Nielsen and Drake 1979) and the regular solution (Ghiorso
1987) models. Beattie (1993), however, recognized that because of numerous interactions at
the molecular level, silicate liquids have many more chemical species than components, which
limits what activity models can reveal about liquid properties. Beattie (1993) developed an
activity model that increased the precision of his olivine-liquid thermometer, and his two-
lattice model appears to be more successful than any proposed thus far. Beattie’s (1993) two-
lattice model yields several terms (calculated from cation fractions) that appear in the models
Table 1. Sample calculations for liquid components.
70
SiO2 TiO2 Al2O3 FeO MnO MgO CaO Na2O K2O total
1) Weight % 47.05 0.89 16.11 7.96 0.15 12.78 11.13 2.18 0.04 98.29
2) Molecular weights 60.08 79.9 101.96 71.85 70.94 40.3 56.08 61.98 94.2
3) Mole proportions 0.7831 0.0111 0.1580 0.1108 0.0021 0.3171 0.1985 0.0352 0.0004 1.62
liq liq liq liq liq liq liq
X SiO 2
XTiO 2
X Alliq2O3 X FeO X MnO X MgO XCaO X Na 2O
X Kliq2O
4) Mole fractions 0.4845 0.0069 0.0978 0.0685 0.0013 0.1962 0.1228 0.0218 0.0003 1.00
SiO2 TiO2 AlO1.5 FeO MnO MgO CaO NaO0.5 KO0.5
5) Molecular weights 60.08 79.9 50.98 71.85 70.94 40.3 56.08 30.99 47.1
6) Cation proportions 0.7831 0.0111 0.3160 0.1108 0.0021 0.3171 0.1985 0.0703 0.0008 1.81
Putirka
liq liq liq liq liq liq liq liq liq
X SiO 2
XTiO 2
X AlO 1.5 X FeO X MnO X MgO XCaO X NaO 0.5
X KO 0.5
7) Cation fractions 0.4327 0.0062 0.1746 0.0612 0.0012 0.1752 0.1096 0.0389 0.0005 1.00
liq liq liq
CSiO 2
C NM C NF NF
8) Beattie (1993) components 0.4327 0.3472 0.4782 -0.7148
The mole proportions in row 3) are equal to row 1) divided by row 2); mole fractions, row 6) are equal to row 3) divided by its total, 1.62. Cation
proportions in row 6) are equal to row 1) divided by row 5); cation fractions; row 7), is row 6) divided by its total, 1.81. Beattie (1993) components
liq liq liq liq liq liq liq liq liq liq liq liq liq liq
are calculated from cation fractions as follows: CSiO 2
= X SiO 2
; C NM = X FeO + X MnO + X MgO + XCaO + XCoO + X NiO ; C NF = X SiO 2
+ X NaO 0.5
+ X KO 0.5
+ XTiO 2
;
liq liq
NF = 7 ln 1 − X AlO + 7 ln 1 − XTiO , where Xij is the cation fraction of i in phase j. Here, the values of Xij are equal to the contents of row 7).
2 1.5 2
( ) ( )
Thermometers & Barometers for Volcanic Systems 71
below (Table 1): NF = 27 ln(1 − X AlO
liq
1.5
) + 7 ln(1 − X TiO
liq
), (where NF refers to “network formers,”
2
i.e., chemical constituents that comprise a silicate network in the liquid); the concentration of
network modifying cations, C NMliq
= X MgO
liq
+ X MnO
liq
+ X FeOliq
+ X CaO
liq
+ X CoO
liq
+ X NiO
liq
; the concentration
of Si, CSiO2 = XSiO2 .
liq liq
Mineral components. Mineral components are typically calculated using the numbers
of “cations per formula unit,” which can be derived from the “cation fractions” (just as for
liquids), sometimes with consideration of charge balance constraints. The forsterite (Fo)
content of olivine, for example, is often presented as Mg/(Mg+Fe), where Mg and Fe represent
the number of cations per formula unit in olivine; because Fo is expressed as a ratio, it does not
matter whether the formula is expressed as (Mg,Fe)Si0.5O2 or (Mg,Fe)2SiO4. This calculation is
also numerically equivalent to XMgOol/(XMgOol+XFeOol), where a term such as XMgOol represents
the mole or cation fraction of MgO, just as is calculated for a liquid (Table 1). Similarly, the
anorthite (An), albite (Ab) and orthoclase (Or) contents of plagioclase, can be represented
as Ca/(Ca+N+K), Na/(Ca+Na+K) and K/(Ca+Na+K) respectively, and are numerically
equivalent to XCaOpl/(XCaOpl+XNaO0.5pl+XKO0.5pl), XNaO0.5pl/(XCaOpl+XNaO0.5pl+XKO0.5pl) and XKO0.5pl/
(XCaOpl+XNaO0.5pl+XKO0.5pl), where here, all Xi are cation fractions.
Pyroxene components can be calculated in a similar manner, but normative type methods
are also employed, to handle charge-couple substitutions. In addition, instead of using a fixed
number of cations, the cations in clinopyroxene are usually calculated on the basis of a fixed
number of oxygen’s (Table 2). The advantage here is that the number of cations can then be added
together to test whether the analysis is valid. For example, if clinopyroxene cations are calculated
on the basis of 6 oxygens per formula unit, then the sum of all cations should be 4. If the sum is
very different from 4, then either the given composition is not (at least wholly) a pyroxene, or the
analysis itself is in error. Because of the complexity of clinopyroxene compositions, component
calculations are handled differently in different models; for this reason clinopyroxene components
are discussed in conjunction with specific thermometers and barometers below.
Units and notation. It is now customary to report P in units of GPa, instead of kbar, while
T units are commonly reported in either Kelvins (K) or degrees centigrade (°C). Because this
review reproduces equations from earlier publications, and because unit conversions would
in some cases require recalibration of regression coefficients, with some risk of introducing
additional error, P-T units from original works are largely preserved; the conversions in any
case are simple (10 kbar = 1 GPa; K = °C + 273.15). Additionally, since in geobarometry most
pressures of interest occur within the highly convenient magnitude range of 0-10 kbar, new
geobarometers are calibrated on a kbar scale, custom be damned.
For ease of calculation, all equations are presented using the same compositional notation
throughout (see Tables 1-3). Terms such as SiO2liq represent the wt% of SiO2 (or some other
oxide) in the liquid; terms such as Xij are either the mole or (nearly always, here) the cation
fraction of i in phase j, the choice of which will be indicated as appropriate (and will be evident
from oxides of Al or Na, for example, where for a mole fraction we write XAl2O3 and XNa2O, and
for a cation fraction we write XAlO1.5 and XNaO1.5). Terms such as NF and CNMliq are calculated
from cation fractions and derive from the activity model of Beattie (1993) (see Table 1). All
mole and cation fractions are calculated on an anhydrous basis, from weight percent oxides that
are not renormalized to 100%, even when hydrous. Compositional terms involving water are
always in units of wt% H2O.
“Glass” (or liquid) thermometers
The Helz and Thornber (1987) thermometer is simplicity itself, depending only upon the
wt% of MgO in a liquid (MgOliq): T(°C) = 20.1MgOliq +1014 °C. Their model was updated by
Montierth et al. (1995; T(°C) = 23.0MgOliq + 1012 °C), and Thornber et al. (2003). In spite of
the simplicity and narrow calibration range, the Helz and Thornber (1987) and Montierth et
Table 2. Pyroxene calculations example part I. Calculation of pyroxenes on the basis of six oxygens, and orthopyroxene components.
72
SiO2 TiO2 Al2O3 FeO MnO MgO CaO Na2O K2O Cr2O3
1) Opx wt% 55.7 0.03 1.88 13.82 0 27.7 1.47 0.44 0 0
2) Mol wt 60.08 79.9 101.96 71.85 70.94 40.3 56.08 61.98 94.2 152.0
3) mole prop 0.9271 0.0004 0.0184 0.1923 0.0000 0.6873 0.0262 0.0071 0 0
4) # of oxygens 2(0.9271) 2(0.0004) 3(0.0184) 0.1923 0.00 0.6873 0.0262 0.0071 0.00 3(0.00)
5) # of oxygens 1.8542 0.0008 0.0553 0.1923 0.0000 0.6873 0.0262 0.0071 0.0000 0.0000
6) sum of row 5 2.8233
6/(oxy
7) ORF 2.1252
sum) =
Cations on the basis of 6 oxygens
opx opx opx opx opx opx
X Siopx XTiopx X Alopx X Fe X Mn X Mg XCa X Na X Kopx XCr
8) cat/6 O 1.9703 0.0008 0.0784 0.4088 0 1.4607 0.0557 0.0302 0 0
Putirka
9) cation sum 4.0048
Orthopyroxene Components
opx opx opx opx opx opx
X Alcpx( IV ) X Alcpx(VI ) X NaAlSi2O6
X FmTiAlSiO6
XCrAl2 SiO6
X FmAl2 SiO6
XCaFmSi2O6
X Fm 2 Si2O6
total
10) 0.0297 0.0486 0.0302 0.0008 0.0000 0.0185 0.0557 0.8973 1.0024
To obtain row 8), the basis for P-T calculations involving pyroxenes (opx=orthopyroxene) first divide row 1) by row 2) to obtain row 3). To get row 4), multiply
row 3) by the numbers of oxygens that occur in each formula unit, as indicated; results are shown in row 5). Row 6) shows the sum of row 5), and ORF, the
“Oxygen Renormalization Factor”, in row 7), is equal to the number of oxygens per mineral formula unit, divided by the sum shown in 6); in our case, we
will calculate pyroxenes on the basis of six oxygens, yielding ORF = 2.1252. To obtain the numbers of cations on a six oxygen basis, as in row 8), multiply
each entry in row 3) by the ORF and by the number of cations in each formula unit (so for SiO2, X Siopx = (0.9271)(2.1252), and for Al2O3, X Alopx = (2)(0.0184)
(2.1252)). Row 9) shows the sum of cations per six oxygens (per formula unit). For a “good” pyroxene analysis, this value should be, as in this case, close to
the ideal value of 4, thus providing a test of the quality of a pyroxene analysis. Orthopyroxene components, in 10) are calculated from the cations in 8), using
opx opx opx opx opx opx
the following algorithm: X Alopx( IV ) = 2 − X Siopx ; X Alopx(VI ) = X Alopx − X Alopx( IV ); X NaAlSi2O6
= X Alopx(VI ) or X Na , whichever is less; X FmTiAlSiO6
= XTiopx ; XCrAl2 SiO6
= XCr ; X FmAl2 SiO6
= X Alopx(VI )
opx opx opx opx opx opx opx opx opx opx opx
− X NaAlSi2O6
− X CrAl2 SiO6 ; X CaFmSi2O6 = X Ca ; X Fm2 Si2O6 = [(X Fe + X Mn + X Mg ) − X FmTiAlSiO6 − X FmAl2 SiO6 − X CaFmSi2O6 ]/2. The sum of the opx components in row 10) should
be close to 1.0.
Thermometers & Barometers for Volcanic Systems 73
al. (1995) thermometers work remarkably well. When predicting T for 1,536 olivine-saturated
(± other phases) experimental data (Fig. 2), the models yield systematic error, but very high
correlation coefficients (R = 0.91for both models), capturing 84% of experimental T variation
(R2 = 0.84). A simple linear correction to the Helz and Thornber (1987) model removes the
systematic error, yielding a new model:
T(°C) = 26.3MgOliq + 994.4 °C (13)
2
Equation (13) yields R = 83 (R = 0.91) and a standard error of estimate (SEE) of ±71 °C (Fig.
2). Additional compositional terms can be added to reduce error further, yielding the following
empirical expressions, one P-independent the other P-dependent,
T ( °C ) = 754 + 190.6 [ Mg # ] + 25.52 MgOliq + 9.585 FeOliq (14)
+14.87 ( Na 2O + K 2O ) − 9.176 H 2Oliq
liq
T ( °C ) = 815.3 + 265.5 Mg # liq + 15.37 MgOliq + 8.61 FeOliq (15)
+6.646 ( Na 2 O + K 2 O ) + 39.16 P ( GPa ) − 12.83 H 2 Oliq
liq
where Mg#liq is a molar ratio and the remaining terms are weight percent oxides in a liquid (or
glass). In Equation (14), R2 = 0.92 and SEE = 51 °C; in Equation (15) R2 = 0.89 and SEE = 60
°C. Equations (13)-(15) are applicable to any volcanic rock saturated with olivine and any other
collection of phases, over the following compositional and P-T range: P = 0.0001-14.4 GPa; T =
729-2000 °C; SiO2 = 31.5-73.64 wt%; Na2O+K2O = 0-14.3 wt%; H2O = 0-18.6 wt% (Fig. 2).
Yang et al. (1996) suggest an improvement to this type of geothermometer by requiring
that additional phases be in equilibrium with the liquid, placing further constraints on
compositional degrees of freedom. Their thermometer applies to liquids in equilibrium with
olivine + plagioclase + clinopyroxene; a new equation similar to their equation 4 is:
T (°C) = −583 + 3141 XSiO + 15779 X Al
2 O3 + 1338.6 X MgO
liq liq liq
2
(16)
−31440[ XSiO
liq
2
⋅ X Al
liq
2 O3
] + 77.67 [ P(GPa )]]
where terms such as XMgOliq represent the mole fraction of MgO in the liquid. Equation (16)
performs less well for liquids that are additionally saturated with other phases, such as spinel
or other oxides, but for experiments where only liquid + olivine + plagioclase + clinopyroxene
are reported, Equation (16) yields R2 = 0.92 and SEE = ±19 °C (n = 73) for the calibration data
and R2 = 0.75 and SEE = 26 °C for 119 test data (Fig. 2).
Olivine-liquid equilibria
Tests for equilibrium. When using thermometers or barometers based on equilibrium
constants, it is essential that equilibrium between the phases in question is approached,
otherwise calculated P-T conditions have no meaning. Roeder and Emslie (1970) produced the
first experiments designed to test whether olivine of a given forsterite (Fo) content might be in
equilibrium with a liquid composition. Their landmark study of Fe-Mg partitioning between
olivine and liquid showed that the equilibrium constant for the following reaction,
MgOol + FeOliq = MgOliq + FeOol (17)
otherwise known as the Fe-Mg exchange coefficient, or KD(Fe-Mg)ol-liq = [(XFeolXMgliq)/
(XMgolXFeliq)] varies little with T or composition, and so is nearly constant at 0.30±0.03.
Subsequent work has shown that KD(Fe-Mg)ol-liq decreases with decreasing SiO2 or increasing
alkalis (Gee and Sack 1988), and increases with increased P (Herzberg and O’Hara 1998;
Putirka 2005a) (Toplis (2005) calibrates all these effects). But the Roeder and Emslie (1970)
74 Putirka Figure 2
2300 1400
Experiments Saturated
with additional phases
2 1350 Calib. data R2=0.54
2100 R =0.84 SEE = ±60oC
SEE = ±72oC R2=0.92 n=154
1300 SEE = ±19oC
slope=1.0
1900 n=73
intercept = -3.9oC 1250
n=1535 Test data
T(oC) Measured
T(oC) Measured 1700 R2=0.75
1200
SEE = ±26oC
n=119
1500 1150
1100
1300
1050 ol+cpx+pl+liq
ol+cpx+pl+liq + il or mt
1100
ol+cpx+pl+liq + sp
1000
ol+cpx+pl+liq + opx
900
ol+cpx+pl+liq + sp + opx
950 ol+cpx+pl+liq+hyd ph
a d
700 900
700 900 1100 1300 1500 1700 1900 2100 2300 900 1000 1100 1200 1300 1400
T(oC) Calculated (Eqn. 16)
T(oC) Calculated (Eqn. 13)
2300
2100 R2=0.90
SEE = ±58oC
1900 slope=1.0
intercept = -0.4oC
n=1530
T(oC) Measured
1700
1500
1300
1100
900
b
700
700 900 1100 1300 1500 1700 1900 2100 2300
T(oC) Calculated (Eqn. 14)
2300
Figure 2. Calculated and measured temperatures are compared for the empirical glass geothermometers
using: (a) Equation (13),2100
(b) Equation (14), (c) Equation (15) and (d) Equation (16) from the text. In (a) –
R2=0.93
(c) all available data are used for calibration: Agee and Walker (1990), Agee and Draper (2004), Almeev
o
et al. (2007), Arndt (1977), Baker and SEE = ±46
Eggler C Baker et al. (1994), Barclay and Carmichael (2004),
(1987),
1900
Bartels et al. (1991), Bender slope=1.0
et al. (1978), Blatter and Carmichael (2001), Bulatov et al. (2002), Chen et al.
o
(1982), Dann et al. (2001), Delano intercept
(1977), =Di-0.2
CarloCet al. (2006), Draper and Green (1997, 1999), Dunn
T(oC) Measured
and Sen (1994), Elkins and n=1530
1700 Grove (2003), Elkins et al. (2000, 2003, 2007), Falloon and Danyushevsky
(2000), Falloon et al. (1997, 1999, 2001), Feig et al. (2006), Fram and Longhi (1992), Fuji and Bougault
(1983), Gaetani and Grove (1998), Gee and Sack (1988), Grove and Beaty (1980), Grove and Bryan (1983),
1500
Grove et al. (1982, 1992, 1997, 2003), Grove and Juster (1989), Hesse and Grove (2003), Holbig and Grove
(2008), Jurewicz et al. (1993),
1300
Juster et al. (1989), Kawamoto (1996), Kelemen et al. (1990), Kennedy et al.
(1990), Kinzler and Grove (1985), Kinzler and Grove (1992a), Kinzler (1997), Kogi et al. (2005), Kogiso
et al. (1998), Kogiso and Hirschmann (2001), Kushiro and Mysen (2002), Kushiro and Walter (1998),
1100
Laporte et al. (2004), Longhi (1995), Longhi (2002), Longhi and Pan (1988, 1989), Longhi et al. (1978),
Maaloe (2004), Mahood and Baker (1986), McCoy and Lofgren (1999), McDade et al. (2003), Medard
and Grove (2008), Medard 900et al. (2004), Meen (1987, 1990), Müntener et al. (2001), Mibe et al. (2005),
Moore and Carmichael (1998), Morse et al. (2004), Müntener, et al. (2001),c Musselwhite et al. (2006),
700 caption continued on facing page
700 900 1100 1300 1500 1700 1900 2100 2300
T(oC) Calculated (Eqn. 15)
1700
n=1530
T(oC) Measure
1500
1300
1100
900
b
Thermometers
700
700 900
&1100Barometers
1300 1500
for1700Volcanic
1900
Systems
2100 2300
75
T(oC) Calculated (Eqn. 14)
2300
2100
R2=0.93
SEE = ±46oC
1900 slope=1.0
intercept = -0.2oC
T(oC) Measured
1700 n=1530
1500
1300
1100
900
c
700
Figure 2 700 900 1100 1300 1500 1700 1900 2100 2300
T(oC) Calculated (Eqn. 15)
2300 1400
Experiments Saturated
with additional phases
Calib. data R2=0.54
2100 R2=0.84 1350
SEE = ±60oC
SEE = ±72oC R2=0.92 n=154
1300 SEE = ±19oC
slope=1.0
1900 n=73
intercept = -3.9oC 1250
n=1535 Test data
T(oC) Measured
1700 R2=0.75
T(oC) Measured
1200
SEE = ±26oC
n=119
1500 1150
1100
1300
1050 ol+cpx+pl+liq
ol+cpx+pl+liq + il or mt
1100
ol+cpx+pl+liq + sp
1000
ol+cpx+pl+liq + opx
900
ol+cpx+pl+liq + sp + opx
950 ol+cpx+pl+liq+hyd ph
a d
700 900
700 900 1100 1300 1500 1700 1900 2100 2300 900 1000 1100 1200 1300 1400
T(oC) Calculated (Eqn. 16)
T(oC) Calculated (Eqn. 13)
2300
Figure 2. (continued) Parman et al. (1997), Parman and Grove (2004), Pichavant et al. (2002a,b), Pickering-
Witter and Johnston (2000), Putirka (1998), Ratajeski et al. (2005), Rhodes et al. (1979b), Robinson et
2100 R2=0.90 al. (1998), Sack et al. (1987), Salters and Longhi (1999), Scaillet and MacDonald (2003), Schwab and
SEE = ±58oC Johnston (2001), Scoates et al. (2006), Sisson and Grove (1993a.b), Stolper (1977), Stolper (1980), Takagi
1900 slope=1.0 et al. (2005), Takahashi (1980), Taura et al. (1998), Thy et al. (2004), Tormey et al. (1987), Vander Auwera
intercept = -0.4oC and Longhi (1994), Vander Auwera et al. (1998), Villiger et al. (2004), Wagner and Grove (1997, 1998),
n=1530 Walter (1998), Wasylenki et al. (2003), Xirouchakis et al. (2001), Yang et al. (1996). In (d) the calibration
T(oC) Measured
1700
data are only the subset of experiments where liquid is in equilibrium with clinopyroxene (cpx), olivine
(ol) and plagioclase (pl). In applying models with such restrictions, not only must all the relevant phases
1500 be present, but additional phases perhaps should also be absent, except for the addition of ilmenite or
magnetite, Equation (16) yields poorer T estimates for liquids that in addition to cpx+liq+plag are also
1300 saturated with orthopyroxene (opx), spinel (sp) and a hydrous phase such as amphibole, or a fluid (fl).
Calibration data are liquids saturated with cpx + ol + plag, and are from: Baker and Eggler (1987), Grove
1100
et al. (1992), Yang et al. (1996), Grove et al. (1997), and Feig et al. 2006). Test data are also saturated with
cpx + ol + plag, and are from: Kennedy et al. (1990), Juster et al. (1989), Dunn and Sen (1994), Grove and
Juster (1989), Grove and Bryan (1983), Bender et al. (1978), Baker et al. (1994), Thy et al. (2004), Medard
900
et al. (2004), bMorse et al. (2004), Pichavant et al. (2002), Mahood and Baker (1986), Meen (1990), Grove
et al. (1982), Tormey et al. (1987), Almeev et al. (2007), Stolper (1980), Sisson and Grove (1993a,b), Di
700 Carlo et al. (2006).
700 900 1100 1300 1500 1700 1900 2100 2300
T(oC) Calculated (Eqn. 14)
2300
2100
R2=0.93
SEE = ±46oC
1900 slope=1.0
intercept = -0.2oC
Measured
1700 n=1530
76 Putirka
model, i.e., KD(Fe-Mg)ol-liq = 0.30, is still valid for basaltic systems generally, at P < 2-3 GPa
(1504 experiments yield KD(Fe-Mg)ol-liq = 0.299±0.053).
J. Michael Rhodes (Dungan et al. 1978; Rhodes et al. 1979a) developed a highly elegant
graphical means to apply such a test for equilibrium between olivine and a potential liquid
by comparing 100⋅Mg#liq vs. 100⋅Mg#ol (where ol = olivine, and the Mg#ol = Fo = Mgol/
(Mgol+Feol); though technically calculated on the basis of cations, it is numerically equivalent
to calculate Fo from XMgOol/(XMgol + XFeOol) as in Table 1). The Rhodes’s diagram (Fig. 3)
simultaneously tests for equilibrium while also displaying various forms of open system
behavior to explain disequilibrium.
Thermometers. Hakli and Wright (1967) were probably the first to calibrate an olivine-
liquid thermometer, in their case based on Ni partitioning. Their effort was updated by Leeman
and Lindstrom (1978) and subsequently expanded to other phases (e.g., Petry et al. 1997; Canil
1999; Righter 2006). Unfortunately, the numbers of experiments reporting NiO are few. Tests
using 67 experimental data with NiO analyses show that the model of Arndt (1977):
10430
T (°C) = − 273.15 (18)
ol -liq
ln DNiO + 4.79
yields the least systematic error, but a disappointing SEE of ±108 °C—and recalibration is of
little help.
Most recent interest in olivine thermometry has revolved about the partitioning of Mg
between olivine and liquid, the first such calibration of which was by Roeder and Emslie (1970).
Though apparently less appreciated than their calibration of KD(Fe-Mg)ol‑liq, their Figure 7
presents a highly elegant graphical thermometer that is still quite useful (see Putirka 2005a).
A recent review by Putirka et al. (2007) shows tests of olivine-liquid thermometers based
on Mg partitioning. Of published models, that by Beattie (1993) is by far the superior. His
equation, re-written here by combining terms, is:
T (°C) =
13603 + 4.943 × 10 −7 P(GPa ) × 109 − 10 −5( ) − 273.15 (19)
6.26 + 2 ln DMg
ol / liq
+ 2 ln[1.5(C NM
L
)] + 2 ln[3(CSiO
L
)] − NF
Figure 3 2
95
Figure 3. The Rhodes diagram yields
tests of olivine-liquid equilibrium.
90
If olivines are in equilibrium with
Fo (= 100 X Olivine Mg#)
coexisting whole rock or glass
compositions, then the putative liquid-
Olivine removal Olivine accumulation
olivine pair should lie along the solid
ng
line, within some established error
ixi
85
aM
bound, here given as KD(Fe-Mg)ol-liq
m
ag
M
= 0.30±0.03. An additional utility of
Differentiation
this diagram is that deviations from
80 equilibrium can be used to explain
disequilibrium, as shown by arrows.
For example, if the maximum forsterite
75
(Fo) content of a suite of olivines from
a given rock were in equilibrium with
KD(Fe-Mg)ol-liq = 0.30 ± 0.03 the whole rock, the vertical trend
would (as indicated) be consistent
70 with (though not uniquely attributable
40 50 60 70 80 90 to) closed system differentiation.
100xMg# Liquid
Thermometers & Barometers for Volcanic Systems 77
ol liq
where = X MgO / X MgO ; other components are calculated as in Table 1. Interestingly, the Beattie
ol / liq
(1993) model works just as well when DMg is calculated from the liquid composition, so like
the “glass” thermometers above, Equation (19) can be used absent an olivine composition
(where the thermometer retrieves the T at which a liquid would become saturated with olivine
ol / liq
at a given P). For this calculation, Beattie’s (1993) expression for DMg (ignoring terms for Ni
and Co content in his Table 1) must be substituted into Equation (19):
ol / liq
DMg =
(
0.666 − −0.049 X MnO
liq
+ 0.027 X FeO
liq
) (20)
liq
X MgO + 0.259 X MliqnO + 0.299 X FeO
liq
The combination of Equations (19) and (20) provide remarkably accurate predictions for olivine
equilibration temperatures, and if one is interested only in anhydrous systems, at T < 1650 °C,
no new calibrations are needed. The Beattie (1993) model yields systematic error at very high
temperatures and pressures, and Herzberg and O’Hara (2002) provide a correction to eliminate
B 93 93
such error: T(°C) = T (°C)1bar + 54P(GPa) − 2P(GPa)2, where T (°C)1Bbar is the temperature from
Beattie (1993), as in Equation (19), but calculated for the case that P = 1 bar.
The Beattie (1993) thermometer also overestimates T for experiments conducted on hy-
drous bulk compositions. To rectify this issue, and to integrate the pressure sensitivity noted by
Herzberg and O’Hara (2002) into the thermodynamic regression equations, Putirka et al. (2007)
presented several new equations, the best thermometers of which are Putirka et al.’s (2007)
Equations (2) and (4), as reproduced here:
P(GPa ) 4430
ol / liq
ln DMg = −2.158 + 55.09 − 6.213 × 10 −2 H 2 Oliq + (21)
T (°C) T (°C)
+5.115 × 10 −2 Na 2 Oliq + K 2 Oliq
T (°C) = {15294.6 + 1318.8P(GPa ) + 2.4834[ P (GPa )]2} / {8.048 (22)
+2.8352 ln D ol / liq
Mg + 2.097 ln[1.5(C L
NM )] + 2.575 ln[3(C liq
SiO2 )]
−1.41NF + 0.222H 2 O + 0.5P(GPa )}liq
In Equation (21) all compositional terms, such as Na2Oliq and H2Oliq are in wt%, except
ol / liq ol / liq ol liq
for DMg , which is a cation fraction ratio: DMg = X MgO / X MgO . All other components are
calculated from cation fractions on an anhydrous basis, using the notation of Beattie (1993)
(Table 1). Incidentally, tests conducted by Putirka et al. (2007) of the Sisson and Grove
(1993b) thermometers showed that they did not perform well overall, with systematic errors
for various systems, including those that are mafic or alkalic. However, their second equation:
log10 [ X Mgol liq
/ ( X MgO liq
( XSiO 2
)0.5 )] = 4129/T(K) − 2.082 + 0.0146[P(bar) – 1]/T(K), yields very good
predictions for T for the CO2-bearing experiments of Dasgupta et al. (2007) (regression of
Tmeas vs. Tpred yields R2 = 0.92; SEE = 27 °C; n = 21), greatly outperforming models by Beattie
(1993) and Putirka et al. (2007) for these special compositions. The Sisson and Grove (1993b)
expression should thus be considered for peridotitic systems containing 2-25 wt% CO2.
Generally, though, Equations (19) and (22) are by far the most precise (Fig. 4 and Putirka
et al. (2007)). Equation (19) provides the best estimates for anhydrous conditions, and (22)
the best estimates when water is present (H2O ranges from 0-18.6% for the test data set).
However, although these thermometers are calibrated using different experimental data, they
yield similar errors using test data. This suggests that differences in their respective estimates
reflect calibration error. In such a case, there is no disadvantage to using both equations and
averaging the results, perhaps including the Sisson and Grove (1993b) model in the average,
depending upon the extent to which the models are appropriate for the system in question.
78
Figure 4
2200
Anhydrous Eqn. (19) Anhydrous Eqn. (21) Anhydrous Eqn. (22)
2
R2=0.94 R =0.91 R2=0.94
2000
SEE=±44oC SEE=±53oC SEE=±45oC
slope=1.01 Hydrous slope=0.94 Hydrous slope=0.98 Hydrous
1800 o 2 o 2 o 2
intercept=-15 C R =0.83 intercept=62 C R =0.92 intercept=24 C R =0.95
n=1302 SEE=±55oC n=1280 SEE=±36oC n=1282 SEE=±29oC
1600 slope=0.82 slope=0.79 slope=0.94
o o o
intercept=141 C intercept=247 C intercept=71 C
n=233 n=223 n=223
1400
Overall Overall Overall
T(oC) Measured
1200 R2=0.91 R2=0.91 R2=0.94
SEE=±53oC SEE=±52oC SEE=±43oC
1000 slope=1.00 slope=0.91 slope=0.97
o o o
Putirka
intercept=-23 C intercept=100 C intercept=34 C
n=1535 n=1505 n=1505
800 Anyhydrous experiments
Hydrous experiments a b c
600
600 1100 1600 2100 2600 1100 1600 2100 2600 1100 1600 2100 2600
T(oC) Beattie (1993) with H&O(02) correction T(oC) Putirka et al.’s (2007) Eqn. (2) T(oC) Putirka et al.’s (2007) Eqn. (4)
Figure 4. Tests of olivine-liquid thermometers based on the partitioning of Mg. In (a), the Beattie (1993) thermometer (Eqn.
19), using the Herzberg and O’Hara (2002) pressure correction; in (b), Putirka et al.’s (2007) Equation (2) (Eqn. 21); in (c),
Putirka et al.’s (2007) Equation (4) (Eqn. 22). For anhydrous conditions, the Beattie (1993) model is best. If water is present,
Putirka et al.’s (2007) Equation (4) is best. Data are as in Figures 2a-c.
Thermometers & Barometers for Volcanic Systems 79
Plagioclase- and alkali feldspar-liquid thermobarometers
Kudo and Weill (1970) presented the first plagioclase-liquid geothermometer. Given the
common occurrence of plagioclase, it is unsurprising that plagioclase thermometry has since
received much attention (e.g., Mathez 1973; Drake 1976; Loomis 1979; Glazner 1984; Ariskin
and Barmina 1990; Marsh et al. 1990). Putirka’s (2005b) review of plagioclase thermometers
showed that the calibration of Sugawara (2001), and the MELTS/pMELTS models of Ghiorso
et al. (2002) provided the most accurate predictors of T for existing models, but that these
models fell short at low T (<1100 °C) and for hydrous systems. That review showed that
improvements in T estimation could be obtained from a model that is simpler:
10 4 pl
X An
= 6.12 + 0.257 ln − 3.166 X liq
( ) (23)
liq
( )( )
2 2 CaO
T (K ) X X liq liq
XSiO
CaO AlO1.5 2
liq
X AlO
( )
2
−3.137 liq 1.5
+ 1.216 X Ab
pl
X AlO + XSiliqO
1.5 2
−2.475 × 10 −2 ( P( kbar ) ) + 0.2166 H 2O liq ( )
Here, X = X / X + X
pl
An
pl
CaO
j
( pl
CaO X and X = X
pl
NaO0.5
pl
KO0.5 )
/ X + X NaOpl
pl
Ab 0.5
pl
X KO pl
(
, where mineral
NaO0.5
0.5
pl
CaO )
components (all Xi ) are calculated as cation fractions; liq = liquid and pl = plagioclase. All
liquid components except H2Oliq are cation fractions, calculated on an anhydrous basis without
renormalization of weight percent values (Table 1); H2Oliq is the wt% of H2O in the liquid
phase. Equation (23) has performed well in other tests (Blundy et al. 2006), and new tests
using data from the LEPR indicate high precision (Fig. 5a), but a global regression yields a ∼6
°C improvement in the SEE (Fig. 5b):
10 4 pl
= 6.4706 + 0.3128 ln
X An − 8.103 X liq
( ) (24a )
liq
( )( )
2 2 SiO2
T (K ) X X liq liq
XSiO
CaO AlO1.5 2
( ) ( ) ( )
2 2
+4.872 X KO
liq
0.5
+ 1.5346 X Aplb + 8.661 XSiO
liq
2
−3.341 × 10 −2 ( P( kbar ) ) + 0.18047 H 2 Oliq ( )
For alkali feldspars, T can be recovered from two new models based on albite-liquid
equilibrium:
10 4 afs
X Ab
= 17.3 − 1.03 ln ( ) − 200 X liq (24 b)
liq
( )
3 CaO
T (K ) X X liq X liq
NaO0.5 AlO1.5 SiO2
( ) ( ) + 13500 ( X − 0.0037)
2
−2.42 X NaO
liq
0.5
− 29.8 X KO
liq
.05
liq
CaO
−550 ( X liq
KO.05 − 0.056 ) ( X liq
NaO.05 − 0.089 ) − 0.078P( kbar )
10 4
T (K )
= 14.6 + 0.055 ( H 2O(wt%) ) − 0.06 P( kbar ) − 99.6 X NaO
liq
0.5
liq
X Al O1.5 ( ) (24c)
(
−2313 X CaO
liq liq
X AlO 1.5
+ 395 X KO
liq
)0.5
(
liq
X AlO 1.5
− 151 X KO
liq
0.5
liq
XSiO )
2
( )
( )
2
+15037 X CaO
liq
Figure 5. The plagioclase-liquid thermometer 1550
80
from (a) Putirka (2005b) (Eqn. 23), captures R2=0.93
much of the variation of data from the 1450 R2=0.90 SEE = ±36oC
Library of Experimental Phase Relations SEE = ±43oC slope=1.0
slope=0.92 intercept = -3.1oC
(LEPR); a global regression (b; Eqn. 1350
intercept = 83oC n=1186
24a) reduces error slightly; data are the n=1203
1250
subset of experiments in Figures 2a-c that
are plagioclase saturated, plus data from
1150
Blundy (1997), Gerke and Kilinc (1992),
Grove and Raudsepp (1978), Helz (1976),
1050
Holtz et al. (2005), Housh and Luhr (1991),
T(oC) Measured
Johnston (1986), Kawamoto et al. (1996), 950
Koester et al. (2002), Martel et al. (1999),
Liu et al. (1998, 2000), Minitti and Rutherford 850
(2000), Panjasawatwong et al. (1995), Patino-
Douce (1995), Patino-Douce and Beard 750
(1995), Petermann and Hirschmann (2000), a b
Petermann and Lundstrom (2006), Prouteau 650
and Scaillet (2003), Rutherford et al. (1985), 650 850 1050 1250 1450 1650 850 1050 1250 1450 1650
T(oC) Eqn. (23) T(oC) Eqn. (24a)
Scaillet and Evans (1999), Skjerlie and Patino-
1600 1550
Douce (2002), Springer and Seck (1997),
Putirka
Calibration Data R2=0.92
Whitaker et al. (2007). In (c), a thermometer 1500 SEE = ±37oC
R2=0.94 1450
based on alkali feldspar-liquid equilibrium SEE = ±23oC slope=1.0
(Eqn. 24b) reproduces T for experiments 1400 slope=0.99 intercept = 3.0oC
1350 n=1190
on natural compositions. This model was intercept = 8.8oC
1300 n=88 Tsuruta
calibrated using experiments performed at & Takahashi 1250
T < 1050 °C from Gerke and Kilinc (1992), (1998)
1200
Koester et al. (2002), Patino-Douce (2005), 1150
Patino-Douce and Beard (1995), Patino-Douce 1100 Holtz et al.
(2001)
and Harris (1998), Scaillet and MacDonald 1050
1000
T(oC) Measured
T(oC) Measured
(2003). Test data are from these same studies,
All Data
but performed at T > 1050 °C, with additional 950
900 (Natural Compositions)
data from Holtz et al. (2001) and Tsuruta and R2=0.96
Takahashi (1998). In (d), the temperature at SEE = ±25oC 850
800
slope=0.94
which a liquid should become saturated with intercept = 44oC
plagioclase (Eqn. (26)). If this plagioclase 700 n=95 750
c d
saturation temperature is in agreement with
600 650
that calculated using Equations (23) or (24a), 600 800 1000 1200 1400 1600 850 1050 1250 1450 1650
then an approach to equilibrium is indicated, T(oC) Eqn. (24b) T(oC) Eqn. (26)
and calculated temperatures may be valid. In 30
P > 1 atm P = 1 atm
(e), the barometer of Putirka (2005b) yields R2=0.37 µ = 3.55 kbar 14 Anhydrous Data
mixed results. The model can predict P to SEE = ±3.8 kbar σ = ±4.01 kbar µ=0.005 wt. % H2O
slope=0.64 n=334 σ=±1.03 wt. % H2O
within ± 3 kbar for just under half of all data, 25
intercept = 2.9 kbar n=825
12
but fails to capture P variations for the LEPR n=727
data set as a whole; the barometer appears to
20
be work well for a subset consisting of about
2/3 of data obtained at P > 1 atm, from: Vander
Auwera and Longhi (1994), Vander Auwera et 8
15
al. (1998), Baker and Eggler (1987), Bartels
Data subset
et al. (1991), Bender et al. (1978), Draper and (P> 1atm)
P(kbar) Measured
6
Johnston (1992), Falloon et al. (1997, 1999), R2=0.67
H2O (wt. %) Measured
Calibration Data
10 SEE = ±2.47 kbar
Fram and Longhi (1992), Grove et al. (1992), (all hydrous data)
slope=0.75 R2=0.87
4
Housh and Luhr (1991), Kinzler and Grove intercept = 1.3 kbar SEE=±1.1 wt. % H2O
(1992), Meen (1987), Panjasawatwong et al. n=497 slope=1.0
5
(1995), Blatter and Carmichael (2001), Feig 2 intercept=0 wt. % H2O
n=730
et al. (2006), Gardner et al. (1995), Holtz et e f
al. (2005), Martel et al. (1999), Moore and 0 0
Carmichael (1998), Patino-Douce and Beard 0 5 10 15 20 25 30 0 2 4 6 8 10 12 14
P(kbar) Eqn. (25a) H2O (wt. %) Eqn. (25b)
(1995), Rutherford et al. (1985), Scaillet and
Evans (1999), Sisson and Grove (1993a,b),
Takagi et al. (2005), Whitaker et al. (2007). 1050 Eqn. (27a) Calibration data
In (f), a new hygrometer (based on equation Eqn. (27a) Test data
(H) of Putirka (2005b)) may be useful for Eqn. (27b) (Global Calibration)
systems with H2O < 9 wt%; 1σ model error 950
is <±1 wt%. In (g), existing two-feldspar
models are tested using experimental data
not used for their calibration; newer, more 850
complex thermodynamic models, share the
same systematic error as with Barth (1962,
his model a). In (h), a new two-feldspar
Thermometers & Barometers for Volcanic Systems
T(oC) Measured
750 Eqn (27a) Eqn. (27b)
thermometer, calibrated and tested using the R2=0.89 R2=0.88
simpler approach of Barth (1951, 1962) with SEE = ±30oC SEE = ±30oC
Barth (1968) (model a)
slope=0.97 slope=1.0
empirical compositional corrections so as to Stormer (1975) intercept = 25oC intercept = -9.7oC
650 Elkins and Grove (1990) n=42 n=42
reduce model error.
Benisek et al. (2004)
g h
550
550 650 750 850 950 1050 1150 550 650 750 850 950 1050
T(oC) From Prior Published Models T(oC) Eqns. (27a, b)
81
82 Putirka
In Equation (24b), all components are calculated as for (24a); afs = alkali feldspar. The ther-
mometer is calibrated at T < 1050 °C, with a calibration error of ±23 °C. The model does not
perform well for experimental data for the synthetic system of Holtz et al. (2005) and efforts to
include these data result in strong systematic error at T > 1050 °C. Although nearly all experi-
ments are hydrous, inclusion of an H2Oliq term similarly yields problems when extrapolating to
T > 1050 °C, and incorporation of the high T data does not solve this problem. Equation (24c)
yields the saturation T for alkali feldspar and is independent of an alkali feldspar composition.
afs liq liq
As a possible test for equilibrium, the Ab-An exchange (KD(An-Ab)afs-liq = X Ab X AlO1.5 X CaO /
afs liq liq
X An X NaO0.5 XSiO 2
) is nearly invariant with respect to T-P-H 2 O, (but with a large standard deviation)
yielding a mean of 0.27±0.18 for 60 experimental data. (The equilibrium constant for Or-Ab
exchange is too dependent on T-P-H2O to be useful in this regard).
Test results for the plagioclase barometer of Putirka (2005b),
X Ab
pl liq
X AlO liq
X CaO
P( kbar ) = −42.2 + 4.94 × 10 −2 T (K ) + 1.16 × 10 −2 T (K )ln pl liq 1.5 liq (25a )
X An X NaO XSiO
0.5 2
( ) + 514.2 ( X ) − 19.6 ln ( X ) − 139.8 ( X )
2 3
−382.3 XSiO
liq
2
liq
SiO2
pl
Ab
liq
CaO
+287.2 ( X ) + 163.9 ( X )
liq
NaO0.5
liq
KO0.5
are not encouraging (Fig. 5). In the original model, data from Sack et al. (1987) and Sugawara
(2001) were excluded due to the potential for Na loss from liquids when using an open furnace
design (see Tormey et al. 1987). However, Na loss should not be an issue for experiments con-
ducted in sealed glass tubes, internally heated pressure vessels or in piston-cylinder apparatus.
Equation (25) predicts P to within <3 kbar for new data (not used by Putirka 2005b) from Falloon
et al. (1997), Scaillet and Evans (1999), Feig et al. (2006) (1-2 kbar experiments), Holtz et al.
(2005), Takagi et al. (2005), Almeev et al. (2007), and Whitaker et al. (2007). However, Equa-
tion (25) yields poor results for experiments from Meen (1990), Jurewicz et al. (1993), Patino-
Douce and Beard (1995), Kawamoto et al. (1996), Patino-Douce and Harris (1998), Koester et
al. (2002), Pichavant et al. (2002), Skjerlie et al. (2002), Prouteau et al. (2001), Medard et al.
(2004), and Villiger et al. (2004). Worse still, no global correlation appears capable of rectifying
these deficiencies, nor explaining why the model can perform well for some data subsets but
not for others. Indeed, with some irony, the equilibrium constant for An-Ab exchange, might
be more useful as a test for equilibrium, because although it clearly varies with T, P and H2O
(e.g., Hamada and Fuji 2007), the values are normally distributed, and when divided over two
pl liq liq pl liq liq
temperature intervals yield KD(An-Ab)pl-liq = X Ab X AlO 1.5
X CaO / X An X NaO 0.5
XSiO 2
= 0.10±0.05 at T <
1050 °C (n = 390; mostly hydrous), and 0.27±0.11 for T ≥ 1050 °C (n = 908). New experiments,
designed for the purpose of testing or developing a plagioclase-liquid barometer are needed. At
best, the plagioclase barometer cannot be recommended for any other than natural compositions
very similar to experimentally studied compositions where the model performs well.
While the status of plagioclase-liquid barometry is firmly in doubt, plagioclase-liquid
equilibrium might be useful as a hygrometer, at least if T is well known. Equation (H) of
Putirka (2005b) tends to over-predict water contents for the LEPR-derived data set, but a global
calibration yields:
pl
X An
H 2O(wt%) = 25.95 − 0.0032T (°C) ln (25b)
liq
( )( )
2 2
X CaO X AlO1.5 XSiO2
liq liq
( ) ( ) − 40.3 ( X ) + 5.7 ( X )
2
−18.9 X KO
liq
0.5
+ 14.5 X MgO
liq liq
CaO
pl
An + 0.108 P( kbar )
Thermometers & Barometers for Volcanic Systems 83
Equation (25b) yields water contents to within ±1.1 wt% H2O for 730 hydrous experimental
data, and recovers a mean H2O content of 0.04±1.0 wt% for 825 anhydrous compositions not
used for regression analysis (Fig. 5). Of course, Equation (25b) is something of a rearrangement
of Equation (24), so as expected, estimates of water content are highly sensitive to T: just a ±38
°C error in T (~1.3σ) translates to a ±1.0 wt% error in H2O.
As a test for plagioclase-liquid equilibrium, T from Equations (23) or (24a) can be
compared to the T required for a liquid to reach plagioclase saturation. The Putirka (2005b)
model predicts T to within ±48 °C for the global LEPR data set; a global regression of this data
yields a 10 °C improvement in precision (Fig. 5c):
10 4
T (K )
(
= 10.86 − 9.7654 XSiO
liq
2
)
+ 4.241 X CaO
liq
(
− 55.56 X CaO
liq
) liq
X AlO (
1.5
) (26)
( ) ( )
3
+37.50 X KO
liq
0.5
liq
X AlO 1.5
+ 11.206 XSiO
liq
2
(
−3.151 × 10 −2 ( P( kbar ) ) + 0.1709 H 2 Oliq )
Equation (26) yields the temperature at which plagioclase should crystallize from a silicate
liquid at a given P.
Two-feldspar thermometers
Thomas W. Barth, with some fairness, could be said to be the father of modern
geothermobarometry. Barth (1934) first suggested that the relative solution of albite
(Ab; NaAlSi3O8) into plagioclase and alkali feldspar could be formulated as an analytic
geothermometer, and he followed his own suggestion with several calibrations (Barth 1951,
1962, 1968). Stormer (1975) presented the first major revision of the two-feldspar thermometer.
In Barth’s (1951) original formulation, he assumed that albite solutions followed Henry’s Law,
i.e., that the activity of a component in a dilute solution varies linearly with composition.
Stormer (1975) illustrated some “serious deficiencies” in the Barth (1962) model, at least
insofar as when one desires to predict the activity of albite. Intervening years have seen the
development of additional thermodynamic models, exhibiting increasing complexity (e.g.,
Whitney and Stormer 1977; Powell and Powell 1977; Fuhrman and Lindsley 1988; Elkins
and Grove 1990; Benisek et al. 2004). These efforts, though, have not demonstrated that more
complex expressions yield better thermometers. Ever since 1934, calibration efforts have
been hampered by a lack of experimental data, and there are probably more published pages
devoted to feldspar thermodynamic modeling than there are experimental observations; in the
latest model (Benisek et al. 2004), the number of adjustable parameters (21) is more than half
the total of modern experimental observations.
Here, tests and new calibrations are performed using 41 experimental data from Elkins
and Grove (1990), Gerke and Kilinc (1992), Patino-Douce and Beard (1995), Patino-Douce
and Harris (1998), Koester et al. (2002), Patino-Douce (2005) and Auzanneau et al. (2006).
As a departure, compositions of experiments are not “adjusted” as was done by Furhman and
Lindsley (1988) and Elkins and Grove (1990). In those studies, experimental input data were
modified so as to minimize the residuals on parameter estimates in least squares analysis. A
problem, though, is that there is no clear path by which to “adjust” the compositions of natural
samples, or other experimental data that might be used for test purposes. The issue is not trivial
because T estimates are highly sensitive to even small changes in XAb, XAn and XOr. In the event,
it is unclear that such adjustments are required to produce a reliable thermometer.
In spite of nearly four decades of thermodynamic modeling, our most complex efforts
barely improve upon Barth’s (1962) simple partitioning model (his Equation a). Indeed, Barth’s
(1951) original approach, though seemingly lost among more recent thermodynamic models,
84 Putirka
provides a remarkably precise (new) thermometer, when just a few empirical corrections are
added:
10 4 X Ab
afs
T (°C)
= 9.8 − 0.098P( kbar ) − 2.46 ln plag
X
( )
− 14.2 XSi + 423 X Ca
afs afs
( ) (27a )
Ab
(
−2.42 ln X An
afs
)
− 11.4 X An
plag
(pl
X Ab )
−442 − 3.72 P( kbar )
T (°C) = (27b)
X afs
−0.11 + 0.1 ln Ab
pl ( ) ( ) (
− 3.27 X An + 0.098 ln X An + 0.52 X An X Ab
afs afs plag pl
)
X Ab
The term XSiafs is the cation fraction of Si in alkali feldspar, which is calculated in the same
way as XSiO2liq. Equation (27a) was calibrated from 30 of the 41 experimental data, using
all data from Elkins and Grove (1990) except A″1, the sole observation from Patino-Douce
(2005), all data from Patino-Douce and Harris (1998) except APD-624, samples APD570,
APD571 and APD547 from Patino-Douce and Beard (1995), and sample 13-89-3 from Gerke
and Kilinc (1992). Remaining data and data from Koester et al. (2002) and Auzanneau et al.
(2006) were reserved for test purposes. Equation (27a) recovers T for the calibration data to
±23 °C, and ±44 °C for the test data. Equation (27b) represents a global calibration from all 41
experimental observations and recovers T to ±30 °C; these models, though much simpler than
recent models, also lack the systematic error of recent efforts.
Orthopyroxene and orthopyroxene-liquid thermobarometers
Tests for equilibrium. As with olivine, the Rhodes’ diagram can be used to test for
equilibrium for orthopyroxene: 785 experimental data yield KD(Fe-Mg)opx-liq = 0.29±0.06
(where opx = orthopyroxene). The value is independent of P or T, but decreases slightly with
increased silica contents, where KD(Fe-Mg)opx-liq = 0.4805 − 0.3733XSiliq (XSiliq is the cation
fraction of SiO2 in the equilibrium liquid).
Calculation of orthopyroxene components. All liquid components are based on cation
fractions, as illustrated in Table 1. All orthopyroxene components are based on the numbers
of cations calculated on a 6-oxygen basis (Table 2). The usefulness of this procedure is that
it leads easily to stoichiometric components, and the sum of such cations (ideally 4.0) yields
a test for the quality of orthopyroxene analyses. Before investigating geothermobarometers,
it is important to understand how certain components are partitioned into orthopyroxene.
For example, how is Al charge balanced in orthopyroxene? As a Ca-Tschermak component,
CaAl2SiO6? Or as FmAl2SiO6 (where Fm = Fe+Mg)? Various such components likely exist, but
to write equilibria involving orthopyroxene there is some advantage to understanding which, if
any, are dominant. Longhi (1976) and Thompson (1982) show how mixing trajectories can be
used to make such identifications. Experimental orthopyroxenes are shown in Figure 6, with
end member components calculated using Thompson’s (1982) linear algebra approach and the
end members: Fm6O6, Ca2Si2O6, Na4Si2O6, Ti3O6, Si3O6, Al4O6 and Cr4O6, which are projected
onto the triangle Si3O6 - Al4O6 - Fm6O6 (Fig. 6). This projection shows that the dominant
mechanism by which Al is incorporated into orthopyroxene is as FmAl2SiO6. The algorithm
used to calculate orthopyroxene for thermobarometry is based on a normative scheme similar
to that use for clinopyroxene (Table 2).
Thermometers and barometers. Beattie (1993) appears to present the only thermometer
based on orthopyroxene-liquid equilibria. This model works well for some compositions, but
over-predicts T for hydrous and low-T, nominally anhydrous compositions (Fig. 7). Two new
thermometers rectify these problems:
Figure 6 for Volcanic Systems
Thermometers & Barometers 85
Al4O6
4+
FmTi Al2O6
CrAl2SiO6
FmAl2SiO6
NaAlSi2O6
Si3O6 Fm6O6
CrAl2SiO6 FmAl2SiO6
4+
FmTi Al2O6
NaAlSi2O6
Si3O6 Fm6O6
Figure 6. Experimental orthopyroxene compositions are re-cast as various oxide end members using
Thompson’s (1982) algebraic approach: Fm6O6, Ca2Si2O6, Na4Si2O6, Ti3O6, Si3O6, Al4O6 and Cr4O6. These
end-member components are projected onto the triangle Si3O6 - Al4O6 - Fm6O6 (upper figure) (Fm = Fe+Mg);
the data are concentrated at the base of this triangle, of which a magnified region is shown (lower figure).
The projection shows that the dominant mixing vector for Al is toward the component FmAl2SiO6.
10 4
= 4.07 − 0.329[ P(GPa )] + 0.12[H 2 Oliq ] (28a )
T (°C)
opx
X Fm
+0.567 ln 2 Si 2 O 6
X liq
( )( )
2 2
SiO2
liq
X FeO + X MnO
liq
+ X MgO
liq
−3.006[ X MgO ] − 6.17[ X KO0.5 ] + 1.89[ Mg # ] + 2.57[ X Fe
liq liq liq opx
]
5573.8 + 587.9 P (GPa ) − 61[ P (GPa )]
2
T (°C) = (28b)
( ) − 3.97 (C ) + 0.06 NF + 24.7 ( X ) + 0.081H O
2
5.3 − 0.633 ln Mg # + 0.156 P (GPa )
liq L liq liq
NM CaO 2
opx
For Equation (28a), all components are calculated as in Tables 1 and 2. The term X Fm 2 Si2 O6
is
the mole fraction of Fm2Si2O6 (or enstatite + ferrosilite, EnFs) as calculated in Table 2, where
opx
Fm = Fe + Mn + Mg; terms such as X Fe are the number of cations of the indicated element
(in this case Fe) in opx, when calculated on a 6 oxygen basis (Table 2). Equation (28b) uses
only liquid components to calculate the T at which a liquid should become saturated with
orthopyroxene. Equation (28a) was calibrated and tested using only those experimental data
86 Putirka
Figure 7
1900 Anhydrous Anhydrous & Hydrous
2
R =0.82 R2=0.95
SEE=±61oC SEE=±39oC
slope=1.2 slope=0.98
1700 o
intercept=-278 C o
intercept=23 C
n=525 n=793
T(oC) Measured
1500
1300
Calibration Data
R2=0.97
SEE=±28oC
1100 slope=1.0
o
Anyhydrous experiments intercept=8.2 C
Hydrous experiments n=177
900
a b
700
700 900 1100 1300 1500 1700 1900 2100 900 1100 1300 1500 1700 1900 2100
T(oC) Beattie (1993) T(oC) Eqn. (28a)
1900 Calibration Data
R2=0.97
SEE=±48oC
1700 slope=1.0
o
intercept=-9.5 C
n=814
T(oC) Measured
1500
1300
1100
900
c
700
700 900 1100 1300 1500 1700 1900 2100
T(oC) Eqn. (28b)
35
Anhydrous & Hydrous 75 Anhydrous
R2 = 0.87 R2 = 0.89
30
SEE = ±2.6 kbar SEE = ±3.0 kbar
65
slope = 1.0 slope = 0.90
25 intercept = 0.04 kbar intercept = -0.04kbar
n=592 55 n=412
P(kbar) Measured
20
45
15
35
Hydrous
10
R2 = 0.91 25 Hydrous
SEE = ±2.1 kbar R2 = 0.67
5 slope = 0.92 15
SEE = ±4.1 kbar
intercept = 0.58 kbar slope = 1.01
n=212 intercept = 1.82 kbar
0 5 n=144
d e
-5 -5
-1 0 10 20 30 -15 5 25 45 65
P(kbar) Eqn. (29a) P(kbar) Eqn. (29c)
Figure 7. caption on facing page
Thermometers & Barometers for Volcanic Systems 87
whose orthopyroxenes had cations sums in the range 3.97-4.03. The expression recovers T for
the calibration data to ±26 °C (slope and intercept are 0.996 and 5.17 °C; R2 = 0.98; n = 148),
and to ±41 °C for the test data (slope and intercept are 0.985 and 15.3 °C; R2 = 0.95; n = 770).
The range of applicability for (28a) is: T = 750-1600 °C; P = 0.0001-11.0 GPa; SiO2 = 33-77
wt%; H2O = 0-14.2 wt%. Equation (28b) has a similar range of applicability, and is the result
of a global non-linear regression.
Wood (1974) was one of the first to calibrate the Al content in orthopyroxene as a barometer.
His Al-in-orthopyroxene model was restricted to rocks in equilibrium with garnet. However,
orthopyroxene barometry need not be restricted to such equilibria. The FmAl2SiO6 component
increases slightly with increased P between 1 atm and 0.5 GPa. And as with clinopyroxene,
equilibria involving a jadeite-like component NaAlSi2O6 (or Jd) shows a steep slope when
plotted against P/T. Barometers calibrated using NaAlSi2O6 or FmAl2SiO6 yield errors of ca.
±2.5-3.0 GPa, and potentially useful expressions include:
opx
X NaAlSi
P( kbar ) = −13.97 + 0.0129T (°C) + 0.001416T (°C) ln 2 O6 (29a )
X liq X liq
( )
2
X liq
NaO0.5 AlO1.5 SiO2
−19.64( XSiO
liq
2
) + 47.49( X MgO
liq
) + 6.99( X Fe
opx
)
(
+37.37 X FmAl
opx
2 SiO6
)
+ 0.748(H 2 Oliq ) + 79.67( X NaO
liq
0.5
+ X KO
liq
0.5
)
opx
X FmAl
P( kbar ) = 1.788 + 0.0375T (°C) + 1.295 × 10 −3 T (°C)ln 2 SiO6 (29b)
FmO AlO1.5 (
X liq X liq 2 X liq
)
SiO2
(
−33.42 X AlO
liq
1.5
) ( )
+ 9.795 Mg # liq + 36.08 X NaO
liq
(
0.5
+ X KO
liq
0.5
)
+ 0.784 H 2Oliq ( )
−26.2 X( opx
Si ) + 14.21( X )
opx
Fe
P( kbar ) = 2064 + 0.321T (°C) − 343.4 ln T (°C) + 31.52 X Al
opx
( )
− 12.28 X Copx
a ( ) (29c)
( ) ( ) ( − 0.1715 )
2
−290 X Cr
opx
+ 1.54 ln X Cr
opx
− 177.2 X Al
opx
−372 ( X − 0.1715 ) ( X − 0.0736 )
opx
Al
opx
Ca
The term XAlopx is the total number of Al atoms in orthopyroxene when cations are calculated on
a 6 oxygen basis (equal to Al(IV) + Al(VI)). Equations (29a,b) derive from global calibrations
of 592 experiments. Equation (29a) yields R2 = 0.83 and SEE = ±2.6 kbar; both models under-
estimate P at P > 30 kbar. However, P for 1 atm data, even though included in the regression,
yield a mean of 2.2±2.3 kbar, and their exclusion from the regression does not improve accu-
racy at P > 0.001 kbar. Remarkably, though, P is more accurately recovered for hydrous data,
Figure 7. (on facing page) Orthopyroxene-liquid thermometers of (a) Beattie (1993) and (b) and (c) this study
are tested using orthopyroxene saturated experiments from data in Figures 2a-c, and from Barnes (1986),
Keshav et al. (2004), Schmidt et al. (2004), Hammer et al. (2002), Holtz et al. (2005), Martel et al. (1999),
Naney (1983), Prouteau and Scaillet (2003), Springer and Seck (1997). In (d) and (e) are tests of potential
orthopyroxene barometers, based on global regressions of experimental data. For (d) the new barometer is
based on Jd-liq equilibrium and was calibrated using data from Baker and Eggler (1987), Bartels et al. (1991),
Elkins-Tanton et al. (2000, 2003), Falloon and Danyushevsky (2000), Falloon et al. (2001), Gaetani and
Grove (1998), Grove and Juster (1989), Holbig and Grove (2007), Kinzler and Grove (1992), Kinzler (1997),
Moore and Carmichael (1998), Ratajeski et al. (2005), and Wagner and Grove (1997, 1998). Equation (29c)
(e) was derived from a global regression, using only orthopyroxene components as independent variables.
88 Putirka
yielding a respectable R2 = 0.91 and SEE ±2.1 kbar—perhaps reflecting more rapid orthopy-
roxene equilibration when H2O is present. Because it is sometimes useful to calculate P when
an equilibrium liquid is unavailable (see Nimis 1995), Equation (29c) was calibrated, using data
from Longhi and Pan (1988), Grove and Juster (1989), Bartels et al. (1991), Kinzler and Grove
(1992) Kinzler (1997), Wagner and Grove (1998), Walter (1998), Falloon et al. (1997, 1999),
Blatter and Carmichael (2001), Gaetani and Grove (1998). The model recovers pressure for the
calibration data with a R2 = 0.93 and an SEE of ±2.5 kbar. Equation (29c) exhibits much less
systematic error to 70 kbar for anhydrous data, but at the cost of systematic error for hydrous
data; predicted pressures for a global dataset (n = 556) yield R2 = 0.84 and an SEE of ±3.6 kbar;
statistics for hydrous and anhydrous data are shown in Figure 7.
Clinopyroxene and clinopyroxene-liquid thermobarometers
Davis and Boyd (1966) introduced the first pyroxene thermometer, a formulation that
requires equilibrium between clino- and orthopyroxene. The two-pyroxene approach has since
received much attention, but few volcanic rocks precipitate both ortho- and clino-pyroxene
phenocrysts, which limits its usefulness to volcanic systems. For this reason, thermometers
and barometers were eventually developed using clinopyroxene compositions alone (Nimis
1995; Nimis and Ulmer 1998; Nimis and Taylor 2000), or clinopyroxene-liquid equilibria
(Putirka et al. 1996, 2003).
Calculation of clinopyroxene components. For the models that follow, all liquid
components are cation fractions, calculated on an anhydrous basis, without renormalization of
wt% values (Table 1). For both the Nimis and Putirka et al. models, clinopyroxene components
are based on the numbers of cations on a 6 oxygen basis. As with orthopyroxene, a “good”
clinopyroxene analysis has a cation sum very close to 4. Indeed, Nimis (1995) uses cation
fractions that are normalized to 4 (which can be achieved by simply multiplying cation fractions
by 4). In Putirka et al. (1996, 2003, and below), cation fractions are never renormalized, and
the sum is used to check for the validity of a given clinopyroxene analysis. Clinopyroxene
components, such as jadeite (Jd), diopside + hedenbergite, (DiHd), and enstatite + ferrosilite
(EnFs) are calculated using a normative procedure of Putirka et al. (2003); Nimis (1995) makes
direct use of clinopyroxene cations calculated on a 6-oxygen basis (Table 3).
Thermometers and barometers. The highest precision and least systematic error for
existing clinopyroxene barometers is Putirka et al. (1996) for anhydrous systems and Putirka
et al. (2003) for those that are hydrous (Fig. 8). The Nimis (1995) model contains stronger
systematic error, for both anhydrous and hydrous systems (see below). However, the Nimis
(1995) model is better than Nimis (1999) (even though Nimis (1999) included corrections for
thermal expansion) and Putirka et al. (1996, 2003) in recovering P for 1 atm data. For Putirka
et al. (1996, 2003), except for 1 atm experiments from Grove and Juster (1989), most all other
1 atm experiments yield high P estimates, ca. 2-3 kbar (Fig. 8). And, as with other equilibria,
the problem is not solved by adding 1 atm data to regression models. The problems of open
furnace experiments have been discussed relative to plagioclase and are relevant here: volatile
losses of Na (Tormey et al. 1987) can lead to systematically high P estimates (Putirka et
al. 1996, 2003). However, although the problem is apparently obviated by the Nimis (1995)
approach, which depends only upon clinopyroxene compositions, the Nimis (1995) model
also yields anomalously low P estimates for high-P experiments. For example, the Putirka
et al. (2003) barometer yields a mean value µ of 9.5 kbar for 184 experiments equilibrated
at 10 kbar, compared to µ = 6.4 kbar for Nimis (1995); for 99 experiments equilibrated at 15
kbar, Putirka et al. (2003) yields µ = 14.5 kbar, compared to µ = 11.7 kbar for Nimis (1995).
This is not to say that the Nimis (1995) model is not useful, but rather that all silicate-based
barometers have inherent error and should be used with this understanding.
To better describe hydrous samples, new barometers based on Equation (1) were calibrated
using H2Oliq (in wt%) as a variable:
Table 3. Pyroxene calculations example part II. Calculation of clinopyroxene components.
SiO2 TiO2 Al2O3 FeO MnO MgO CaO Na2O K 2O Cr2O3
1) Cpx, wt% 52.5 0.33 1.58 9.89 0.24 13.29 22.5 0.37 0 0.05
2) mol. wt 60.08 79.9 101.96 71.85 70.94 40.3 56.08 61.98 94.2 152.0
3) mol. prop. 0.8738 0.0041 0.0155 0.1376 0.0034 0.3298 0.4012 0.0060 0.0000 0.0003
4) # of O 1.7477 0.0083 0.0465 0.1376 0.0034 0.3298 0.4012 0.0060 0.0000 0.0010
5) oxy sum 2.6814
6) ORF 6/(oxy sum) = 2.2376
Cations on the basis of 6 oxygens
cpx cpx cpx cpx cpx cpx
X Sicpx XTicpx X Alcpx X Fe X Mn X Mg XCa X Na X Kcpx XCr
7) cat/6 Oxy 1.9553 0.0092 0.0694 0.3080 0.0076 0.7379 0.8978 0.0267 0.0000 0.0015
8) cation sum 4.0134
Clinopyroxene Components
cpx cpx cpx cpx cpx cpx
X Alcpx( IV ) X Alcpx(VI ) X Fe ( 3+ ) X Jdcpx XCaTs XCaTi XCrCaTs X DiHd X EnFs total
9) 0.0447 0.0247 0.0268 0.0247 0.0000 0.0223 0.0007 0.8747 0.0856 1.0081
To obtain row 7), the basis for calculations by Nimis (1995), first divide row 1) by row 2) to obtain row 3). To get row 4), multiply row 3) by the numbers
of oxygens that occur in each formula unit. Row 5) shows the sum of row 4), and the Oxygen Renormalization Factor (ORF) in row 6), is equal to the
number of oxygens per mineral formula unit, divided by the sum shown in 5); in our case, we calculate pyroxenes on the basis of six oxygens, yielding
ORF = 2.2376. To obtain the numbers of cations on a six oxygen basis, as in row 7), multiply each entry in row 3) by the ORF and the number of cations
Thermometers & Barometers for Volcanic Systems
in each formula unit (see Table 2). Row 8) shows the sum of cations per six oxygens (per formula unit). For a “good” pyroxene analysis, this value should
be, as for orthopyroxenes, close to 4, thus providing a test of the quality of a pyroxene analysis. Clinopyroxene components in 9) are calculated from the
cpx cpx cpx
cations in 7), using the following algorithm: X Alcpx( IV ) = 2 − X Sicpx ; X Alcpx(VI ) = X Alcpx − X Alcpx( IV ); X Fe ( 3+ )
= X Na + X Alcpx( IV ) − X Alcpx(VI ) − 2 XTicpx − XCr (Papike te al. 1974); X Jdcpx
cpx cpx cpx cpx cpx
= X Al (VI ) or X Na , whichever is less; if excess X Alcpx(VI ) remains after forming X Jdcpx , XCaTs (CaAlIVAlVISiO6) = X Alcpx(VI ) − X Jdcpx ; if X Alcpx( IV ) > XCaTs , XCaTi (CaTiAl2O6)
cpx cpx cpx cpx cpx cpx cpx cpx cpx cpx cpx cpx cpx
= [X Al ( IV ) − XCaTs]/2; XCrCaTs (CaCr2SiO6) = XCr /2; X DiHd (CaFmSi2O6) = XCa − XCaTi − XCaTs − XCrCaTs; X EnFs (Fm2Si2O6) = [X Fe + Mg − X DiHd ]/2. The sum of X
the cpx components in 9) should be close to 1.0.
89
90 Putirka
Figure 8
2400
Anhydrous
Anhydrous Hydrous Anhydrous
2200
T=950-1650oC T=950-1650oC
R2=0.82 R2=0.75
2000
SEE = ±52oC SEE = ±61oC
T(oC) Putirka et al. (1996)
slope=1.0 slope=0.84
1800 intercept = -13.6oC intercept = 197oC
n=818 n=818
1600
1400
Hydrous Anhydrous
1200 T=679-1435oC T=679-1435oC
R2=0.65 R2=0.74
SEE = ±69oC SEE = ±59oC
1000
a slope=0.95 b slope=0.70
intercept = -72oC intercept = 271oC
800
n=324 n=324
600
600 800 1000 1200 1400 1600 1800 2000 2200 2400 800 1000 1200 1400 1600 1800 2000 2200 2400
T(oC) Putirka et al. (1996) T(oC) Putirka et al. (2003)
Anhydrous 1 atm Anhydrous
65
P=0-40 kbar µ = -0.61 kbar P=0-40 kbar
R2=0.68 1σ = ±2.93 kbar R2=0.73
55 SEE = ±5.0 kbar n=306 SEE = ±4.6 kbar
slope=0.89 slope=0.82
intercept = 3.8 kbar intercept = 2.2 kbar
45 n=928 n=848
P(kbar) Measured
35 1 atm
µ = 1.63 kbar
1σ = ±3.88 kbar
25 Measured
P(kbar)
n=254
Hydrous Hydrous
15
P=0-32 kbar P=0-32 kbar
R2=0.74 R2=0.52
SEE = ±4.0 kbar SEE = ±5.4 kbar
5
c slope=0.62 d slope=0.80
intercept = 3.7 kbar intercept = 6.6 kbar
n=329 n=324
-5
-25 -15 -5 5 15 25 35 45 55 65 -15 -5 5 15 25 35 45 55 65
P(kbar) Nimis (1995) P(kbar) Putirka et al. (1996)
Anhydrous
Figure 8. In (a), the clinopyroxene-liquid 65
P=0-40 kbar
thermometer from Putirka et al. (1996). In R2=0.71
(b), the clinopyroxene-liquid thermometer of 55
SEE = ±4.8 kbar
slope=0.89
Putirka et al. (2003). In (c), the clinopyroxene intercept = 0.90 kbar
barometer of Nimis (1995). Also shown: n=848
P(kbar) Measured
45
clinopyroxene-liquid barometers from (d)
Putirka et al. (1996) and (e) Putirka et al. 1 atm
35
(2003). Data are clinopyroxene-saturated µ = 2.21 kbar
experiments conducted at P < 40kbar, from 1σ = ±3.63 kbar
n=254
Figures 2a-c, 5a-b, and 7, plus data from 25
Hydrous
Berndt et al. (2001), Hirschmann et al. P=0-32 kbar
(2003), Johnson (1998), Johnston (1986), 15 R2=0.58
Putirka et al. (1996, 2003), Takahashi et SEE = ±5.0 kbar
al. (1989), Tsuruta and Takahashi (1998), slope=0.89
5 e intercept = 3.6 kbar
Ulmer and Sweeny (2002), Wood and Triglia n=324
(2001), Yasuda et al. (1994), Berndt et al. -5
(2001), Walter and Presnall (1994). -25 -15 -5 5 15 25 35 45 55 65
P(kbar) Putirka et al. (2003)
Thermometers & Barometers for Volcanic Systems 91
cpx
X NaAlSi
T (K ) T (K ) 2 O6
P( kbar ) = −48.7 + 271 + 32 ln (30)
) (
4 4 2
10 10 liq liq liq
X NaO0.5 X AlO1.5 XSiO2
−8.2 ln( X FeO
liq
) + 4.6 ln(( X MgO
liq
) − 0.96 ln( X ) − 2.2 ln ( X
liq
KO0.5 ) cpx
DiHd
−31( Mg # liq ) + 56(( X NaO
liq
0.5
+X ) + 0.76 ( H O )
liq
KO0.5 2
liq
cpx
X NaAlSi
T (K ) T (K ) 2 O6
P( kbar ) = −40.73 + 358 + 21. 69 ln (31)
X liq X liq
( )
2
10 4 10 4 liq
NaO0.5 AlO1.5 XSiO2
−105.7( X CaO
liq
) − 165.5( X NaO
liq
0.5
+ X KO
liq
0.5
)2 − 50.15 XSiO
liq
2
( )( X liq
FeO + X MgO
liq
)
−3.178 ln X ( cpx
DiHd ) − 2.205 ln( X cpx
EnFs ) + 0.864 ln( X cpx
Al ) + 0.3962 H 2O ( liq
)
The term XAlcpx is the total number of Al atoms in clinopyroxene when the formula is calculated
on a 6 oxygen basis (equal to XAl(IV)cpx + XAl(VI)cpx). Equation (30) was calibrated using Grove
and Juster (1989), Kinzler and Grove (1992), Walter and Presnall (1994), Sisson and Grove
(1993a,b), Putirka et al. (1996), Scaillet and MacDonald (2003), and Patino-Douce (2005).
Equation (31) was calibrated from a global regression of clinopyroxene-saturated experiments,
excluding experiments performed at 1 atm (except Grove and Juster 1989) and P > 40 kbar (Fig.
9). A new model based on the Nimis (1995) approach was also recalibrated using experiments
performed from 0.001-80 kbar:
P( kbar ) = 3205 + 0.384T (K ) − 518 ln T (K ) − 5.62 X Mg
cpx
+ 83.2 X Na
cpx
( ) ( ) (32a )
( ) ( ) ( ) ( )
2 2
+68.2 X DiHd
cpx
+ 2.52 ln X Al(VI)
cpx
− 51.1 X DiHd
cpx
+ 34.8 X EnFs
cpx
P( kbar ) = 1458 + 0.197T (K ) − 241 ln T (K ) + 0.453 H 2Oliq ) ( (32 b)
(
+55.5 X Acpxl(VI) ) + 8.05 ( X ) − 277 ( X ) + 18 ( X ) + 44.1( X )
cpx
Fe
cpx
K
cpx
Jd
cpx
DiHd
( ) − 17.7 ( X ) + 97.3 ( X )
2 2
+2.2 ln X Jdcpx cpx
Al
cpx
Fe(M2)
( ) − 27.6 ( X )
2 2
+30.7 X Mg(
cpx
(M2)
cpx
DiHd
Equation (32a) depends only upon the composition of clinopyroxene, and improves the
precision of the Nimis (1995) model, but it preserves the systematic error with respect to
hydrous experiments. Equation (32b) removes this systematic error, but at the cost of requiring
an estimate of the H2O content of the liquid in equilibrium with the clinopyroxene; it also
requires use of additional calculations for the fractions of Fe and Mg occupying M1 and M2
sites, which can be found in Nimis (1995).
Equation (32c) represents a new barometer based on the partitioning of Al between
clinopyroxene and liquid.
P( kbar ) = −57.9 + 0.0475T (K ) − 40.6( X FeO
liq
) − 47.7( XCaTs
cpx
) (32c)
X cpx
+0.676(H 2Oliq ) − 153( X CaO
liq liq
XSiO ) + 6.89 liAl
0.5 2 X AlO
q
1.5
Here, the last term represents the ratio of the total number of Al atoms in clinopyroxene when
calculated on a 6 oxygen basis (XAlcpx = XAl(IV)cpx + XAl(VI)cpx; see Table 3), and the cation
Figure 9
2400
92
All Data P<70 kbar Anhydrous All Data P<70 kbar Anhydrous
R2=0.92 R2=0.90 R2=0.93 R2=0.91
2200 SEE = ±45oC SEE = ±46oC SEE = ±45oC SEE = ±45oC
slope=1.0 slope=0.98 slope=0.99 slope=0.97
intercept = 6.6oC intercept = 26oC intercept = 17oC intercept = 35oC
2000
n=1174 n=854 n=1186 n=861
1800
1600
1400
T (oC) Measured
1200 Hydrous Hydrous
R2=0.86 R2=0.85
SEE = ±42oC SEE = ±45oC
1000 slope=1.0 slope=0.93
intercept = -0.9oC intercept = 65oC
n=320 n=325
800
a b
600
600 800 1000 1200 1400 1600 1800 2000 2200 2400 800 1000 1200 1400 1600 1800 2000 2200 2400
T (oC) Eqn. (33) T (oC) Eqn. (34)
Figure 9. New clinopyroxene-liquid
thermometers and barometers based on
Putirka
65 All Data P<40 kbar All Data P<40 kbar
global calibrations. In (a), clinopyroxene- R2=0.83 R2=0.89
SEE = ±3.6 kbar SEE = ±2.9 kbar
liquid thermometer based on Jd-DiHd slope=1.1 slope=1.0
55
exchange (Eqn. 33). In (b), clinopyroxene intercept = -0.4 kbar intercept = -0.8 kbar
n=1066 n=1176
saturation thermometer (Eqn. 34). In (c),
a barometer based on Jd-liq equilibrium, 45
1 atm 1 atm
using calibration data as noted in text µ = 2.8 kbar µ = 2.7 kbar
1σ = ±3.0 kbar 1σ = ±3.2 kbar
(Eqn. 30). In (d), a barometer as in (c) but 35 n=217 n=228
based on a global calibration (Eqn. 31).
P (kbar) Measured
In (e), a clinopyroxene-only thermometer 25
of Nimis and Taylor (2000) and in (f) a
new calibration for a clinopyroxene-only Calibration Data
15 R2=0.97
thermometer, using the Nimis and Taylor
SEE = ±1.6 kbar
(2000) activity model (Eqn. 32d). In (g), slope=1.0
a clinopyroxene-only barometer (Eqn. 5 intercept = 0 kbar
n=117
(32a)) and in (h) a similar calibration c d
(32b), but incorporating an H2Oliq term to -5
rectify systematic error for hydrous data. -25 -15 -5 5 15 25 35 45 55 65 -25 -15 -5 5 15 25 35 45 55 65
Data are as in Figure 8. P (kbar) Eqn. (30) P (kbar) Eqn. (31)
2400
Anhydrous Anhydrous
R2=0.40 R2=0.82
2200
SEE = ±107oC SEE = ±58oC
slope=0.49 slope=1.0
2000 intercept = 822oC intercept = 1.5oC
n=910 n=910
1800
1600
1400
T (oC) Measured
1200 Hydrous
R2=0.22 Hydrous
SEE = ±95oC R2=0.36
1000 SEE = ±87oC
slope=0.30
intercept = 754oC slope=0.82
n=314 intercept = 61oC
800
n=314
e f
600
100 300 500 700 900 1100 1300 1500 1700 1900 2100 2300 800 1000 1200 1400 1600 1800 2000 2200 2400
T (oC) Nimis & Taylor (2000) T (oC) Eqn. 32d
75 All Data P<40 kbar
Anhydrous
R2=0.91
R2=0.92 SEE = ±2.6 kbar
65 SEE = ±3.1 kbar slope=1.0
slope=1.0 intercept = 0 kbar
intercept = 0 kbar n=1173
55 n=854
45
35
Thermometers & Barometers for Volcanic Systems
25
P (kbar) Measured
1 atm 1 atm
15 µ = 1.7 kbar µ = 1.6 kbar
1σ = ±2.2 kbar 1σ = ±2.0 kbar
n=226 n=226
5
g h
-5
-15 -5 5 15 25 35 45 55 65 75 -15 -5 5 15 25 35 45 55 65 75
P (kbar) Eqn. 32b
93
P (kbar) Eqn. 32a
94 Putirka
fraction of Al in liquid (see Table 1). This barometer was calibrated using data from Putirka
et al. (1996), Kinzler and Grove (1992), Sisson and Grove (1993a,b) and Walter and Presnall
(1994); regression statistics are: R2 = 0.95, SEE = ±1.5 kbar, n = 99. A test using all available
data (n = 1303) yield R2 = 0.82 and SEE = ±5 kbar.
Nimis and Taylor (2000) have calibrated a clinopyroxene-only thermometer from a limited
data set (Fig. 9e). Using their activity model, a new and more precise thermometer is (Fig. 9f):
93100 + 544 P( kbar )
T (K ) = (32d )
( ) +10.9 ( X ) − 0.95 ( X ) ( )
2
61.1 + 36.6 X cpx
Ti
cpx
Fe
cpx
Al + X Cr
cpx
− X Na
cpx
− X Kcpx + 0.395 ln aEn
cpx
Here, all cation fractions are on the basis of 6 oxygens (as in Eqns. 32a,b), the term XAlcpx refers
cpx
to total Al (XAlcpx = XAl(IV)cpx + XAl(VI)cpx) and the activity of enstatite in clinopyroxene, aEn , is as
in Nimis and Taylor (2000): aEn = (1 − X Ca − X Na − X K ) ⋅ (1 − 0.5( X Al + X Cr + X Na + X Kcpx )).
cpx cpx cpx cpx cpx cpx cpx
The Jd-DiHd exchange thermometers (Putirka et al. 1996, 2003) reproduce T to ±52-60
°C. Global calibrations using experiments conducted at P < 70 kbar reduce these uncertainties
by about 10-20 °C (Fig. 9):
10 4 X cpx X liq X Fm
liq
T (K )
= 7.53 − 0.14 ln Jdcpx CaO
X X liq liq
X
(
+ 0.07 H 2 Oliq − 14.9 X CaO
liq
)
liq
XSiO 2
( ) (33)
DiHd Na Al
( )
−0.08 ln X TiO
liq
2
(
− 3.62 X NaO
liq
0.5
+ X KO
liq
0.5
)
− 1.1 Mg # liq ( )
−0.18 ln ( X ) − 0.027 P( kbar )
cpx
EnFs
10 4
T (K )
( )
= 6.39 + 0.076 H 2Oliq − 5.55 X CaO
liq liq
XSiO 2
( )
− 0.386 ln X MliqgO ( ) (34)
−0.046 P( kbar ) + 2.2 × 10 −4 [ P( kbar )]
2
Equation (34) yields the temperature at which a liquid will become saturated with clinopyroxene
at a given pressure, and so can be used as a test of temperatures derived from expressions such
as Equation (33).
Tests for equilibrium. Some petrologists (e.g., Klügel and Klein 2005) test for
equilibrium on the basis of Fe-Mg exchange coefficients. The danger is that Fe-Mg exchange
equilibrium does not necessitate, for example, Na-Al, or Ca-Na exchange equilibrium, etc.,
and so Rhodes et al. (1979a) and Putirka (1999a, 2005b) suggest that equilibrium values for
mineral compositions also be calculated, and compared to observed mineral compositions.
Among experimental and natural data, however, deviations in observed and calculated values
for KD(Fe-Mg)cpx-liq are highly correlated with deviations between observed and calculated
values for DiHd, EnFs, etc., so tests based on Fe-Mg exchange may be sufficient. The Fe-Mg
exchange coefficient derived from 1,245 experimental observations yields KD(Fe-Mg)cpx-liq =
0.28±0.08 (with a range from 0.04-0.68, with a roughly normal distribution). This value is not
perfectly invariant, and the equation
1719
ln K D ( Fe − Mg )cpx − liq = −0.107 − (35)
T (K )
accounts for a slight temperature dependency, where KD(Fe-Mg)cpx-liq is recovered with a not
so stunning R2 = 0.12 and an SEE of 0.08.
Two-pyroxene thermometers and barometers
Sen et al. (2005) illustrate how sub-solidus equilibria can be crucial to understanding
volcanic systems. Thus, since the introduction of the two-pyroxene geothermometer by Davis
Thermometers & Barometers for Volcanic Systems 95
and Boyd (1966), much attention has been given to estimating T utilizing some form of enstatite-
diopside partitioning (e.g., Wood and Banno 1973; Herzberg and Chapman 1976; Wells 1977;
Lindsley 1983; Lindsley and Anderson 1983; Mercier et al. 1984; Sen 1985; Sen and Jones
1989; Brey and Köhler 1990).
As with other systems, it is important to test for equilibrium. Using Fe-Mg exchange,
311 experiments yield K D ( Fe − Mg)cpx − opx = ( X Fe
cpx cpx
/ X Mg opx
) / ( X Fe opx
/ X Mg ) = 1.09 ± 0.14. For
calibration and test purposes we exclude data that lie at the outskirts of 3σ.
Of existing thermometers, the Brey and Köhler (1990) model T(BKN) is the most precise
(Fig. 10a). A new global regression, based on the partitioning of enstatite + ferrosilite (=
Fm2Si2O6 = EnFs; FmO = FeO + MgO + MnO) between clinopyroxene and orthopyroxene
increases precision for the available experimental data (Fig. 10b):
Figure 10
1800
Cpx with Mg#>0.75
Cpx with Mg#>0.75
R2=0.88
R2=0.91
1600 SEE = ±50oC
SEE = ±45oC
slope=1.07
slope=1.00
intercept = 72oC
intercept = -1.6oC
n=390
n=390
T(oC) Measured
1400
1200
All Data
All Data
R2=0.81
R2=0.88
SEE = ±70oC
1000 SEE = ±56oC
slope=0.96
slope=0.97
intercept = 58oC
intercept = 43oC
n=487
n=487
800 Experimental Bulk Compositions
Natural compositions
SiO2+CaO+Al2O3+MgO (SCAM)
a b
600
600 800 1000 1200 1400 1600 1800 800 1000 1200 1400 1600 1800
T(oC) Brey and Köhler (1990) (TBKN) T(oC) Eqn. (36)
1800 50
45 Cpx with Mg#>0.75
Cpx with Mg#>0.75
R2=0.82
1600 R2=0.93
40 SEE = ±2.8 kbar
SEE = ±38oC
slope=1.0
slope=0.98
intercept = 0.1 kbar
P(kbar) Measured
intercept = 16oC 35
1400 n=393
T(oC) Measured
n=390
30
1200 25
20
All Data
1000 All Data
R2=0.86 15 R2=0.77
SEE = ±60oC
SEE = ±3.2 kbar
slope=0.98
10 slope=0.96
800 intercept = 59oC
intercept = 0.5 kbar
n=487
5 n=483
c d
600 0
600 800 1000 1200 1400 1600 1800 0 10 20 30 40 50
T(oC) Eqn. (37) P(kbar) Eqn. (39)
Figure 10. The Brey and Kohler (1990) two-pyroxene thermometer (a) performs reasonably well when
predicting T for experimental data saturated with two pyroxenes from Figure 2a-c, plus data from Green
and Ringwood (1967), Klemme and O’Neill (2000), Gudfinnsson and Presnall (1996), Milholland and
Presnall (1998), and Sen (1985). A global regression (Eqn. 36) (b) provides somewhat greater precision,
but precision is improved further if the calibration data base is restricted to include only Mg-rich systems,
in this case defined as those cpx-opx pairs where Mg#cpx>0.75. In (d), a barometer, based on the Mercier et
al. (1984) approach is also calibrated using only high Mg# compositions.
96 Putirka
10 4 X cpx
T (°C)
= 11.2 − 1.96 ln EnFs
X opx
− 3.3 X Ca
cpx
(
− 25.8 X CrCaTs
cpx
) ( ) (36)
EnFs
(
+33.2 X Mn
opx
)
− 23.6 X Na
opx
(
− 2.08 X En
opx
)
− 8.33 X Di(
opx
− 0.05P( kbar )) ( )
where the EnFs components in clino- and orthopyroxene are calculated as in Tables 2 and 3 (EnFs
= Fm2Si2O6 for orthopyroxene), and where X En opx
= [( X Mg
opx opx
) / ( X Mg + X Fe
opx
+ X Mn
opx opx
)]( X Fm 2 Si 2 O6
), and
X Di = [( X Mg ) / ( X Mg + X Fe + X Mn )]( X CaFmSi2 O6 ) (with cation fractions calculated on a 6 oxygen
opx opx opx opx opx opx
basis). The models perform best for mafic systems where Mg#cpx >0.75. A separate regression
using only experiments with Mg#cpx>0.75 (Fig. 10c) yields the following expression:
10 4 X cpx
T (°C)
= 13.4 − 3.4 ln EnFs
X opx
+ 5.59 ln X Mg
cpx
(
− 8.8 Mg # cpx ) ( ) (37)
EnFs
(
+23.85 X Mn
opx
)
+ 6.48 X FmAl
opx
(
2 SiO6
− 2.38 X Di
cpx
)
− 0.044 P( kbar ) ( )
With a similar restriction on Mg#, a new barometer (Fig. 10d) was calibrated from a
global regression, utilizing the strategy of Mercier et al. (1984):
P( kbar ) = −279.8 + 293 X Al(VI)
opx
(
+ 455 X Na
opx
) ( )
+ 229 X Cr
opx
( ) (38)
(
+519 X Fm
opx
2 Si 2 O6
) − 563 ( X ) + 371( X ) + 327 ( a ) + 1K.119
opx
En
opx
Di
opx
En
f
Equation (38) is T-independent and the term Kf = X / (1 − X ) is as in Mercier et al. (1984);
opx
Ca
cpx
Ca
opx
X En = ( X Fm
opx
2 Si 2 O6
opx
)( X Mg opx
/ [ X Mg + X Mn
opx
+ X Fe
opx
]) and X = ( X CaFmSi
opx
opx
Di 2 O6
opx
)( X Mg opx
/ [ X Mg + X Mn
opx
+ X Fe
opx
]);
opx
the activity of enstatite in orthopyroxene, aEn , is similar to Wood and Banno (1973) (with all
cations calculated on the basis of 6 oxygens):
opx
0.5 X Mg 0.5 X Mg
opx
aEn =
opx
X opx + 0.5 X opx + 0.5 X opx + X opx + X opx
0.5 X Fe + X Fe + X Al(VI) + X Ti + X Cr + 0.5 X Mg
opx opx opx opx opx opx
Ca Mg Fe
2+ Mn Na 2+ 3+
2 + = X Fe − X
opx opx opx
Here, X Fe Fe3+
where Fe3+ is calculated as in Papike et al. (1974): Fe3+ = Al(IV) +
Na − Al(VI) − Cr − 2Ti. Equation (38) recovers P to ±3.7 kbar with R2 = 0.68 (n = 389; Fig.
11). Precision is greatly increased when T is used as input:
P( kbar ) = −94.25 + 0.045T (°C) + 187.7 X Al(VI)
opx
+ 246.8 X Fm
opx
(
2 Si 2 O6
) ( ) (39)
1.66
−212.5 X En(
opx
)
+ 127.5 aEn
opx
− ( ) Kf
− 69.4 X EnFs
cpx
( ) − 133.9 ( a )
cpx
Di
cpx
where terms are as in Equation (38), and the activity of diopside in clinopyroxene, aDi , is
aDi = ( X Ca ) / ( X Ca + 0.5 X Mg + 0.5 X Fe2+ + X Mn + X Na ). Equation (39) recovers P to ±2.8 kbar,
cpx cpx cpx cpx cpx cpx cpx
with R2 = 0.82 (n = 393; Fig. 11).
The silica activity barometer
Carmichael et al. (1970) and Nicholls et al. (1971) were the first to show that because the
liq
activity of SiO2 (aSiO 2
) may be buffered at constant P and T by reactions such as
Mg2SiO4ol + SiO2liq = Mg2Si2O6opx (40)
liq
that a could be used as a thermometer or barometer for igneous processes. Because most if
SiO2
not all basalts equilibrate with olivine and orthopyroxene in their mantle source regions, there
liq
has been much interest in using aSiO 2
as a barometer for mantle partial melting (though nothing
precludes application to other olivine + orthopyroxene saturated liquids). For example, the
Thermometers & Barometers for Volcanic Systems 97
Figure 11
100
90 R2=0.91
SEE = ±2.87 kbar
slope=1.0
80
intercept = 0 kbar
n=510
70
P(kbar) Measured
Figure 11. A silica activity barometer,
60 using the activity for SiO2 in silicate
liquids devised by Beattie (1993). The
50 data set is restricted to those liquids
that are in equilibrium with both olivine
40 and orthopyroxene (as in Nicholls et al.
1971), which include data from Figure
30 6, and data from Falloon et al. (1988),
and Kushiro (1996).
20
10
0
0 10 20 30 40 50 60 70 80 90 100
P(kbar) Eqn. (42)
work of Nicholls et al. (1971) is implicit in the SiO2-based barometers of Klein and Langmuir
(1987), Albarede (1992) and Haase (1996). Langmuir and Hanson (1980) later showed how
FeO can be used as a barometer (supported by experiments by Stolper 1980), and Putirka
(1999b) used Na/Ti ratios to estimate partial melting depths. But both FeO and Na/Ti are
sufficiently sensitive to variations in source region composition and degree of partial melting
that they do not readily lend themselves to a simple analytic expression.
The quantitative models of Albarede (1992) and Haase (1996) were calibrated using small
liq
data sets, and use wt% SiO2liq as input, not aSiO 2
; these models capture <35% of the variation
liq
in P for the larger experimental data set studied here. However, Beattie’s (1993) aSiO 2
is highly
correlated with P:
( ) (1 − X ) (1 − X )
−2 7/2 7
liq
aSiO 2
= 3 XSiO
liq
2
liq
AlO1.5
liq
TiO2 (41)
liq
When calculated in this way, aSiO 2
alone describes 80% of the variation in P for experimental
liq
data when Mg#liq >0.75, and for 510 experiments with no restrictions on Mg#, aSiO 2
captures
>60% of variation (over the P-T range 0.001–70 kbar and 825-2000 °C; SiO2 ranges from
liq
31.5-70%; data include hydrous compositions). Using the Beattie (1993) aSiO 2
a new model
loosely based on Equations (5) and (41) is:
( ) ( )
1/ 2
P( kbar ) = 231.5 + 0.186T (°C) + 0.1244T (°C)ln aSiO
liq
2
− 528.5 aSiO
liq
2
( 42)
X liq
(
+103.3 X TiO
liq
2
)
+ 69.9 X NaO
liq
0.5
(
+ X KO
liq
0.5
+ 77.3 liq )
X AlO + XSiO
liq
AlO1.5
1.5 2
where all compositional terms are cation fractions (Table 1). Equation (42) was calibrated
using a “global” database of partial melting experiments (iteratively removing data outside
±3σ) where liquids are in equilibrium with olivine and orthopyroxene (± other phases).
Interestingly, calibrations on smaller data sets yield little improvement in the SEE for P, and
fared poorly when predicting P for data not used in the calibration. Equation (42) recovers P
to ±2.9 kbar, with R2 = 0.91 (Fig. 11).
98 Putirka
APPLICATIONS
Magma transport at Hawaii
Thermobarometers can be of immense use for understanding the mechanical controls of
magma transport. For example, although it is reasonably well understood that density contrasts,
such as at the Moho, may control magma transport (Stolper and Walker 1980), other transport
models emphasize the roles of stress states within the lithosphere (ten-Brink and Brocher
1987; Parsons and Thompson 1993) or fracture toughness (Putirka 1997). These views yield
contrasting predictions regarding where magmas are stored prior to eruption. Presuming that
magmas undergo an episode of cooling and crystallization during storage, thermobarometers
provide a test.
To illustrate, Putirka (1997) posited that magmas are stored over a wide depth range at
Hawaii, and that the Moho does not provide a necessary barrier to magma transport. New data
from the Hawaii Scientific Drilling project, and a re-analysis of some existing data, modify this
view. Table 4 shows P-T calculations for a Pu’u O’o epsisode-3 flow (Garcia et al. 1992) using
several new and existing models. The thermobarometers are in agreement yielding mean P-T
estimates of 4.8±0.8 kbar and 1469±16 °C; standard deviations are well within model error. For
Pu’u O’o episodes 1-10, depth-time estimates are shown for individual crystals whose mineral-
whole rock pairs yield KD(Fe-Mg)cpx-liq within ±0.08 of values predicted using (35) (Fig. 12a).
As in Putirka (1997), crystallization depths become shallower with time. These magmas appear
to have been stored simultaneously over a range of depths, with a top-to-bottom emptying of a
rather lengthy and semi-continuous and connected conduit.
However, existing data from Mauna Kea (Frey et al. 1990, 1991) and 291 new clinopyroxene-
whole rock pairs from the HSDP-2 (Rhodes and Vollinger 2004; new mineral compositions
from this study) alter the story line. Of these data, 120 clinopyroxenes yield KD(Fe-Mg)cpx-liq
within ±0.08 of values determined from Equation (35). Seventy-five (73%) fall within the depth
range for “high density cumulates” determined from seismic studies (Hill and Zucca 1987).
This clustering, and a trend to lower T just above the Moho, was not evident in the smaller data
set examined by Putirka (1997). Moreover, P-T estimates from ultramafic xenoliths (Fodor and
Galar 1997; Eqns. 38, 39) (Table 5) also exhibit clustering near the Moho, and a trend towards
higher (more volcanic-like) T. The volcanic phenocryst, seismic, and xenolith data thus all
support the conclusions of Fodor and Galar (1997, p.1) that ultramafic xenoliths at Mauna Kea
are “gravity settled and in situ cumulates from reservoir bottoms.” However, both the volcanic
phenocryst and xenolith suites also yield depth ranges as great as 35 km. A hybrid model is
supported, involving a magma mush column (e.g., Ryan 1988; Marsh 1995) that extends below
the Moho, possibly with the mechanical controls advocated by Putirka (1997), with transport
inhibited by density contrasts near the Moho (Garcia et al. 1995; 8-14 km).
The conditions of mantle partial melting
The glass and olivine-liquid thermometers, and the silica activity barometer are calibrated
to 60-70 kbar, and can be applied to understand mantle melting. For example, Morgan (1971)
suggested that volcanism at Hawaii or Yellowstone is driven by plumes—thermal upwellings,
which rise though the mantle. Like Hess (1962), Morgan (1971) argued that heat from the core
drives mantle convection.
The mantle plume model thus predicts that “hot spots” should be hotter than ambient
mantle (which feeds mid-ocean ridge basalt (MORB) volcanism). In other words, they should
exhibit an “excess temperature,” reflecting their derivation from a source with excess heat.
Recent skepticism of the mantle plume model (e.g., Foulger and Natland 2003) has prompted
efforts to estimate mantle potential temperatures (Tp) at ocean islands, to test whether hot
spots are truly hot (e.g., Putirka 2005a; Putirka et al. 2007; Herzberg et al. 2007; Putirka
2008). Tp represents the T the mantle would have if it rose to the surface, adiabatically,
Table 4. Example calculations for clinopyroxene thermobarometers.
SiO2 TiO2 Al2O3 FeO MnO MgO CaO Na2O K2O Cr2O3
1
Whl-rk (wt%) 50.2 2.83 13.74 11.07 0.17 6.6 10.48 2.49 0.57 0.202
Whl-Rk (cat fr) 0.4795 0.0203 0.1547 0.0884 0.0014 0.0940 0.1072 0.0461 0.0069 0.0015
Cpx (wt%) 52.9 0.7 2.1 7.6 0 17.9 18.7 0.27 0 0.2
Cat per 6 O2 1.9337 0.0192 0.0905 0.2323 0.0000 0.9755 0.7323 0.0191 0.0000 0.0058
cpx cpx cpx cpx cpx
X AlcpxIV−6 ox X AlcpxVI−6 ox Fe3+ KD(Fe-Mg)cpx-liq X Jdcpx XCaTs XCaTi XCrCaTs X DiHd X EnFs
0.0663 0.0242 0.0169 0.253 0.0191 0.0051 0.0306 0.0029 0.6938 0.2570
Putirka et al. (1996) Putirka et al. (2003)
Eqn. T2 Eqn. P1
T(°C) P(kbar)3 T(°C) P(kbar)
1194 5.3 1214 5.1
New Models (this study)
Eqn. 30 Eqn. 31 Eqn. 32a Eqn. 32b
P(kbar) P(kbar) P(kbar) P(kbar)
4.9 6.3 4.0 4.1 Mean estimates
Eqn. 32d Eqn. 33 Eqn. 34 Eqn. (35) P(kbar) T(°C)
T(°C ) T(°C ) T(°C ) KD(Fe-Mg)cpx-liq Mean 4.95 1201
Thermometers & Barometers for Volcanic Systems
1199 1198 1198 0.283 std dev 0.85 7.7
1
“Whl-Rk” is a whole rock composition, for sample 3-88, Garcia et al. (1992).
2
Clinopyroxene (cpx) cations (on the basis of 6 oxygens) are calculated as in Table 2.
3
P calculated from Putirka et al. (1996) solved iteratively using Equation (T2) of Putirka et al. (1996); all other T-sensitive P estimates use T(K) from Putirka et al.
(2003).
99
100 Putirka
Figure 12 Figure 12
Days after eruption Days after eruption T(oC) Putirka et al. (1996) T(oC) Putirka et al. (1996)
0 50 100
0 150
50 200
100 900 9501501000 2001100
1050 900 1150
950 1200
1000 1250
1050 1300 1100 1150 1200 1250 1300
0 0 0 0
a a b b Model error (1σ) Model error (1σ)
Pu’u O’o Eruption Pu’u O’o Eruption
Depth (km) (P from Putirka et al., 1996)
Depth (km) (P from Putirka et al., 1996)
Episodes 1-10 Episodes
5 1-10 5
5 5 Individual crystals Individual crystals
Episode mean Episode mean
10 10
Mafic Mafic
cumulates cumulates
(at Mauna (at Mauna
15 15 Loa) Loa)
10 10
Depth interval Depth interval
for mafic 20 for mafic 20
cumulates at cumulates
Elasticat Elastic
Kilauea Kilauea
Thickness Thickness
15 15
25 25
Model error (1σ) Model error (1σ) 30 Mauna Kea 30 Mauna Kea
20 20 Volcanic rocks with cpx phenocrysts
Volcanic rocks with cpx phenocrysts
Xenoliths with cpx + opx Xenoliths with cpx + opx
35 35 cpx-opx
Xenoliths where KD(Fe-Mg) Xenoliths where KD(Fe-Mg)cpx-opx=1.09±0.28
=1.09±0.28
25 25 40 40
4 4
c c d d
3.5 3.5
P(GPa)= 3.3 - 0.63(Na2O/TiO2P(GPa)=
) 3.3 - 0.63(Na2O/TiO2)
Nominal Equilibration Depths Nominal Equilibration Depths
P(GPa) Eqn. (42)
P(GPa) Eqn. (42)
for Parental Ocean Island for Parental Ocean Island
3 & MORB Magmas 3 & MORB Magmas
2.5 2.5
2 2
1.5 1.5 P(GPa)= 1 P(GPa)= 1
MORB 1.18 - 0.07FeOt
MORB 1.18 - 0.07FeOt MORB MORB
1 1
7 8 9 10 7 11 8 12 9 13 10 14 11 112 131.5 14 2 1 2.5 1.5 3 2 2.5 3
FeOt FeOt Na2O/TiO2 Na2O/TiO2
900 900
e Two-feldspar e Two-feldspar
880 Plag-liq 880 Plag-liq
T(oC) Plag- Afs-sat, 2-Fspar (Eqn 27a)
T(oC) Plag- Afs-sat, 2-Fspar (Eqn 27a)
Afs-liq Afs-liq
860 860
840 840
820 820
800 800
780 780
760 760
740 740
720 720
Model error (1σ) Model error (1σ)
700 700
700 750 800
700 850
750 900
800 850 900
T(oC) Plag-, Afs-liq, Benisek
T(oC)
etPlag-,
al. 2-Fspar
Afs-liq, Benisek et al. 2-Fspar
Figure 12. caption on facing page
Thermometers & Barometers for Volcanic Systems 101
without melting, and so provides a convenient reference for inter-volcano comparisons (see
McKenzie and Bickle 1988). Table 6 shows calculations using parental liquid compositions for
MORB and Hawaii (from Putirka 2008). For these calculations, Equations (22) and (42) are
solved simultaneously to derive P and T, where T represents the temperature of olivine-liquid
equilibration and P the pressure of olivine+orthopyroxene+liquid equilibration. Additional T
estimates (Table 6) are derived using this P (from Eqns. 22, 42) as input. Mantle equilibration
temperatures are converted to Tp by correcting for the heat of fusion, and adiabatic upwelling
(Table 6). The thermometers yield highly consistent results.
The difference between Tp at Hawaii and Tp at MORB is the excess temperature, Tex, which
reflects some fraction of the excess heat derived from the plume source. The Tex at Hawaii is ~260
°C, and the nominal depth of equilibration there is more than 50% greater than beneath mid-
ocean ridges (Table 6). The calculated pressures of equilibration are called “nominal” because
the reconstructed “parental liquids” used as input (Putirka 2008) are likely to represent some
averaging of a range of melt compositions that exist within a melting column of considerable
depth extent (Klein and Langmuir 1987). For 27 other ocean islands (Putirka 2008), average Tex
= 160±52 °C and average equilibration pressures and depths are 2.6±0.4 GPa and 85±11 km.
liq
Computed inter-island variations in P may be real. Pressures calculated from aSiO 2
are
correlated with other P-sensitive components. For example, a positive correlation between
liq
P (based on aSiO 2
) and FeOt is expected (Fig. 12c) because increased P means that greater
amounts of olivine (and FeOt) enter the liquid (Langmuir and Hanson 1980; Stolper 1980;
Klein and Langmuir 1987). And a negative correlation is expected for Na2O/TiO2 (Fig. 12d),
because the distribution coefficient for Na in clinopyroxene increases relative to that for Ti,
with increased P (Putirka 1999b).
Temperature estimates in felsic volcanic systems
Unfortunately, few published studies on felsic volcanic rocks report both alkali feldspar,
plagioclase feldspar and glass compositions. But such compositions from the 75 ka Toba tuffs
(Beddoe-Stephens et al. 1983) indicate how some of the models presented here might be used
to better understand the thermal record of such systems.
Figure 12. (on facing page) In (a) and (b) Hawaiian rock and mineral compositions are used to illustrate
the use of several of the thermometers and barometers given in this study. Whole-rock-cpx pairs from the
Pu’u O’o eruption of Kilauea (Garcia et al. 1989, 1992), are used in (a) to compare depth and time; the
relationship indicates a top-to-bottom emptying of a continuous conduit that extends to >20 km below
the Kilauea summit. In (b), depth (d)-T estimates at Mauna Kea from xenoliths (Fodor and Galar 1997)
are compared to d-T estimates from volcanic rocks with clinopyroxene (cpx) phenocrysts (Frey et al.
1990, 1991; Yang et al. 1996; Rhodes and Vollinger 2004 with mineral compositions from this study). The
volcanic and xenolith suites both indicate transport of magmas from as great at 35 km, but also storage and
cooling of magma near the base of the volcano, in the depth range of “mafic cumulates” (Hill and Zucca
1987). In (c), depth estimates are calculated using calculated parental magma compositions for MORB
and 27 ocean islands from Putirka (2008), and are compared to FeOt and Na2O/TiO2 contents for these
magmas. There is some circularity for (c) because FeOt contents are used to estimate P. But the cross
correlations are as expected if variations in SiO2-FeOt-Na2O/TiO2 reflect partial melting depths, driven
by differences in partial melting temperatures (see Klein and Langmuir 1987 and Putirka 1999b). In (e),
temperatures are calculated for Toba Tuff ignimbrites (Beddoe-Stephens et al. 1983) using the two feldspar
(2-Fspar) and plagioclase- and alkali feldspar-liquid equilibria. Glass compositions are melt inclusions
in plagioclase and alkali feldspar (Beddoe-Stephens et al. 1983). The match between the two plagioclase
models indicates that plagioclase most closely approaches equilibrium with included glass. Alkali feldspar
T estimates are within 2σ of one another, but only one sample yields temperatures within 1σ. Though the
plagioclase (plag) and alkali feldspar (afs) temperatures are within model error, plagioclase may be closer
to equilibrium, and perhaps precipitated at a higher T than alkali feldspar, and temperatures derived from
two-feldspar thermometry may be too low.
Table 5. Examples of two-pyroxene P-T calculations.
102
SiO2 TiO2 Al2O3 FeO MnO MgO CaO Na2O K2O Cr2O3
Cpx (wt%) 52.3 0.7 3 5.1 0.11 16.6 21.5 0.33 0 0.58
Cat per 6 O1 1.9122 0.0192 0.1293 0.1559 0.0034 0.9048 0.8422 0.0234 0 0.0168
cpx cpx cpx cpx cpx
X AlcpxIV−6 ox X AlcpxVI−6 ox Fe3+ X Jdcpx XCaTs XCaTi XCrCaTs X DiHd X EnFs
0.0878 0.0414 0.0146 0.0234 0.0180 0.0349 0.0084 0.7809 0.1399
cpx cpx cpx
X En X Di aDi
0.1189 0.6640 0.6050
SiO2 TiO2 Al2O3 FeO MnO MgO CaO Na2O K2O Cr2O3
Opx (wt%) 55 0.34 1.5 11.3 0.24 30.7 0.9 0.01 0 0.19
Cat per 6 O1 1.944 0.0090 0.0625 0.3340 0.0072 1.617 0.0341 0.0007 0 0.0053
Putirka
opx opx opx opx opx
X AlcpxIV−6 ox X AlcpxVI−6 ox Fe3+ X Jdopx X FmTiAlSiO6
XCrCaTs X FmAl2 SiO6
X DiHd X EnFs
0.0563 0.00614 0.0275 0.0007 0.0090 0.0053 0.0001 0.0341 0.9576
opx opx opx opx
X En X Di aEn aDi
0.7908 0.0281 0.6451 0.0340
New Models (this study)2
Eqn. 38 Eqn. 39/36 Eqn. 36/39 Eqn. 37/38
KD(Fe-Mg)cpx-opx P(kbar) P(kbar) T(°C) T(°C)
0.835 2.2 3.4 985 964
1
Clinopyroxene (cpx)-orthopyroxene (opx) pair is from sample C2 of Fodor and Galar (1997) with compositions shown in wt% and as cations on the basis of 6 oxygens.
2
Equations (39) and (36) are solved simultaneously; Equation (38) is T-independent and Equation (36) uses 3.4 kbar as input. All clinopyroxene and orthopyroxene
calculations are as in Tables 2-4 and as noted in the text.
Table 6. Examples of liquid, olivine-liquid, and mantle potential temperature calculations.
SiO2 TiO2 Al2O3 FeO MnO MgO CaO Na2O K2O H2O
1
MORB 48.5 0.9 15.9 8 0.1 13.2 11.1 2.3 0 0.11
Hawaii1 47.2 1.6 9 11.6 0.2 21.4 7.3 1.4 0.2 0.65
DMgol-liq Eqn. 19 Mean Std. dev.
Eqn. 13 Eqn. 14 Eqn. 15 Eqn. 16 Eqn. 19 Eqn. 22
(Eqn. 20) w/H&O3 T(°C) T(°C)
MORB2 3.39 1342 1343 1359 1375 1390 1381 1374 1366 19
2
Hawaii 2.12 1557 1575 1547 1513 1614 1579 1557 1563 31
Eqn. 42 Depth5
Tpcorr Tp Tex
P(kbar) (km)
MORB4 15.6 52.5 33 1399 -
4
Hawaii 27.3 88.8 97 1660 261
1
Mid-ocean ridge (MORB) and Hawaiian compositions are “parental” magmas, as calculated in Putirka (2008).
2
All temperatures are calculated in oC.
3
“Equation (19) w/H&O” uses Equation (19) and the correction for pressure from Herzberg and O’Hara (2002); see text.
Thermometers & Barometers for Volcanic Systems
4
Tp is mantle potential temperature (see text); Tpcorr is the correction required to convert an olivine equilibration T (i.e., Eqn. 22) to Tp and is given by: Tp = Tol-liq
+ 667F – 1.33P(kbar), where F (melt fraction) is 0.08 for MORB and 0.20 at Hawaii (see Putirka et al. 2007) and Equation (22) is used as input.
5
P from Equation (42) is converted to depth (d) using a regression fit based on PREM (Anderson, 1989): d(km)=14.6 + 29P(GPa) − 0.157[P(GPa) − 8.4]2.
103
104 Putirka
T estimates at Toba based on plagioclase-liquid equilibrium (Eqn. 23) and plagioclase-
saturation (Eqn. 26) are nearly identical (Fig. 12e) (Table 7). Although this might reflect the
higher precision of the plagioclase-based models, it might also indicate that plagioclase-hosted
melt inclusions have more closely approached equilibrium with their host minerals, since all but
one alkali feldspar-glass inclusion pair yield mineral-melt and mineral saturation temperatures
that are >1σ apart. The one alkali feldspar-glass pair that most closely approaches equilibrium
has a T estimate slightly less than that for plagioclase, perhaps indicating that plagioclase
precipitated at a slightly higher T than alkali feldspar. If this is the case, temperatures from
two-feldspar thermometry might not be valid, since if alkali-feldspars are not in equilibrium
with their included glass, they would seem unlikely to be in equilibrium with neighboring
plagioclase feldspars. In that case, T estimates from the two-feldspar thermometers may be too
low by 40-70 °C. A greater number of samples are clearly needed to interpret these data with
confidence, but the Toba examples illustrate how multiple thermometers can be used to better
understand the thermal history of silicic systems.
SOURCES OF ERROR RELATIVE TO P AND T
Geothermometry and geobarometry involve two sources of error. 1) Model error measures
the precision of a model, the magnitude of which can be identified from regression statistics.
This error results from experimental error, and includes the error in reproducing P or T in
the laboratory, lack of attainment of equilibrium during an experiment, and errors related to
compositional measurements; the latter two are likely the most important (Putirka et al. 1996).
2) The error of treatment is related to the application of a model, and will trend towards
greater values as natural systems deviate from experimental conditions; the most serious issue
here is probably the pairing of a given mineral with its equilibrium liquid—less an issue for
experiments, but problematic for natural magmas when they behave as open systems. It is also
important that natural magmas and minerals have compositions and equilibration conditions
similar to those studied experimentally. The best guard against error is to compare P-T
conditions from different equilibria, which need not match precisely (olivine may precipitate
before clinopyroxene) but should be correlated.
Model error
Estimates for model error can be derived from regression statistics. Usually, a “standard
error of estimate” or SEE (equivalent to the root mean square error, RMSE), is reported for T
or P. Like a standard deviation (1σ) in a normal distribution, the SEE (or RMSE) represents
the interval ±E such that there is a 68.26% probability that the true value lies between −E and
+E. If the calibration data are small in number and of high quality, the SEE will probably
underestimate true model error, since by the nature of regression, coefficients are optimized for
a highly specific set of circumstances, which natural samples are unlikely to mimic perfectly.
If the calibration data derive from a single laboratory, it is also possible that there is systematic
error related to experimental design. A truer test of model error can be derived from “test data,”
i.e., experimental data not used for regression.
How does one decide which experimental data should be used for test or calibration
purposes? Hirschmann et al. (2008) note that many experiments are conducted at too short
a time scale to allow diffusional exchange between minerals and liquid. Long-duration
experiments, however, provide no assurances of equilibrium since thermal gradients in such
can induce compositional gradients (Lesher and Walker 1988). In any case, errors on estimates
of P (Eqn. 30) and T (Eqn. 33) are nearly independent of experimental run duration, as are
values for KD(Fe-Mg)ol-liq, an often-used test of equilibrium (Fig. 13).
Estimates of error. Various thermometers tend to yield comparable calibration errors, on
the order of ±25-30 °C, when more than a few hundred experiments are used for regression.
Table 7. Examples of feldspar calculations.
SiO2 TiO2 Al2O3 FeO MnO MgO CaO Na2O K2O H2O
1
Toba, R301 P 73.78 0.08 12.56 1 0.04 0.02 0.89 3.01 4.47 4.0
Two Feldspar Temperatures
An Ab Or Eqn 27a Eqn 27b B(2004)
1 T(°C) T(°C) T(°C)2
Plagioclase 0.372 0.586 0.042
Alkali Feldspar1 0.012 0.276 0.712 743 818 778
Feldspar-liquid P-T and H2O Calculations
Eqn. 23 Eqn. 24a Eqn. 24b Eqn. 24c Eqn. 26 Eqn. 25 Eqn. 25b/23 Eqn. 25b/24b
T(°C) T(°C) T(°C) T(°C) T(°C) P(kbar) H2O3 H2O3
846 815 794 815 843 4.03 3.2 4.05
1
Glass and feldspar data are from Beddoe-Stephens et al. (1983).
2
Thermometers & Barometers for Volcanic Systems
Two feldspar thermometer of Benisek et al. (2004).
3
H2O calculations are in wt%, calculated at a pressure of 1 kbar, and at a T calculated from Equations (23) and (24b); when calculated using a pressure of
4 kbar, wt% values for H2O are 0.5% less.
105
106 Putirka
Figure 13
6 8
R2(anhydrous) = 0.001 R2(anhydrous) = 0.000
R2(hydrous) = 0.000 R2(hydrous) = 0.020
Error on P, ln[abs(Pmeas - Pcalc)]
Error on T, ln[abs(Tmeas - Tcalc)]
4
6
2
Figure 13
4
2 8
6
-2 R2(anhydrous) = 0.001 R2(anhydrous) = 0.000
R2(hydrous) = 0.000 R2(hydrous) = 0.020
Error on P, ln[abs(Pmeas - Pcalc)]
Error on T, ln[abs(Tmeas - Tcalc)]
4 0 6
-4
2
-2 4
-6
a b
0
-8 -4 2
R2(anhydrous) = 0.013 -4 -2 0 2 4 6 8 10
R2(hydrous) = 0.030
-2
Duration of Experiment, ln(hrs)
0.5 0
-4
KD(Fe-Mg)ol-liq
0.4 -2
-6
a b
0.3 -8 -4
R2(anhydrous) = 0.013 -4 -2 0 2 4 6 8 10
R2(hydrous) = 0.030 Duration of Experiment, ln(hrs)
0.20.5
KD(Fe-Mg)ol-liq
0.10.4
c
00.3
-4 -2 0 2 4 6 8 10
Duration of Experiment, ln(hrs)
0.2
Figure 13. In (a) and (b) model errors are compared to experimental “run” duration, for P (Eqn. 30) and T
(Eqn. 0.133) respectively. Model error is calculated as the natural log of the absolute value of the difference
between predictedc and experimentally reported P or T. In (c), the Fe-Mg exchange coefficient between olivine
and liquid,
0 KD(Fe-Mg)ol-liq, is compared to experimental run duration. R2 values are shown for hydrous (open
symbols) -4
and anhydrous
-2 0
(closed
2
symbols)
4 6
data. 8Except10for some anhydrous experiments conducted at less than
a few hours, increasedDuration of Experiment,
experimental ln(hrs) do not necessarily yield improvement in model precision, or
run durations
apparently even a closer approach to Fe-Mg exchange equilibrium.
Thus ±25-30 °C probably represents the smallest magnitude for model error that one can
expect for any individual T estimate derived from a generalized silicate-based thermometer
(regardless of whether smaller errors are reported). Similarly, the highest precision to be
expected from silicate-based barometers is in the range ±1.5-2.5 kbar.
As noted in prior sections, there is hope that if experimental error is random, then more
precise estimates can be derived by averaging individual P-T estimates. Because many
experiments are conducted at 8, 10, 12 kbar, etc., it is easy to test this idea by calculating P for a
number of experiments performed at a given pressure (say 10 kbar), then compare the mean of
those estimates, and the standard error on that mean, to the measured value. Such calculations
indicate that mineral-melt barometers may be as precise as ±0.5-1.0 kbar (a “standard error on
the mean”) when estimates are averaged for isobaric systems (e.g., Putirka et al. 1996, 2003).
For thermometers the increase in precision is not so clear. For 35 experiments on dry
basalts conducted at 1200 °C, Equation (19) (Beattie 1990) yields a mean of 1225±52 °C (the
±52 °C is the standard deviation) with a standard error on the mean of ±9 °C; the mean value
thus deviates from the measured value by more than twice the standard error on the mean and
Thermometers & Barometers for Volcanic Systems 107
so the latter does not reflect greater accuracy. For the same data, Equation (22) yields the same
standard error on the mean (±9 °C) with a mean value of 1213±50 °C; the mean estimate is
still more than one standard error from the measured value. Although these standard errors
on the means do not represent increased accuracy, the absolute differences between the mean
and measured T estimates might, in which case averaging several T estimates (for isothermal
systems), yield error as low as ±13-25 °C.
P-T slopes and correlated error. Mordick and Glazner (2006) note that because the
clinopyroxene-liquid barometer is highly dependent on T, P-T estimates are necessarily
correlated. This principle holds for any T-sensitive barometer or P-sensitive thermometer.
Mordick and Glazner (2006) use a Monte Carlo technique to simulate random error and
generate a P-T error array for clinopyroxene-liquid pairs to illustrate the correlation. Putirka
and Condit (2003) adopt the method of Kohn and Spear (1991), using multiple analyses on a
single grain to generate a similar array. However determined, a key issue is that the coefficients
of any T-sensitive barometer or P-sensitive thermometer lead to a natural slope in P-T space. In
the clinopyroxene-liquid example, the slope is not without meaning: the P-T arrays generated
from Equations (30) or (31) yield the T at which a given liquid will become saturated with
clinopyroxene at a given P—a clinopyroxene saturation surface.
Of potentially great interest is when such P-T slopes deviate from natural saturation
conditions. For example, Putirka and Condit (2003) interpret near-isobaric extensions from
the saturation P-T trend to represent episodes of magma stagnation, as magmas isobarically
migrate down-T toward the ambient geotherm. A near-isothermal slope, in contrast, might
indicate partial crystallization during a period of rapid vertical transport (Armienti et al.
2007). To make such interpretations, though, the observed P-T slope must deviate significantly
from the intrinsic slope of mineral saturation. To make such a test, one could select a single
equilibrium mineral-liquid pair to use as input, select an arbitrary range of temperatures, also
to be used as input, and then use Equations (30) or (31) to calculate P for such conditions. The
P-T array thus generated will represent the clinopyroxene saturation curve for that particular
liquid. The difference between the slopes derived from different input compositions (as in
Mordick and Glazner 2006) may thus also be used to derive model error on the P-T slope.
Error of treatment
Error of treatment is much more difficult to define than model error, since it depends upon
how closely a natural system mimics experimental compositions and conditions. Because of the
wide array of compositions that have been experimentally studied, over a very wide range of
P and T, few natural conditions probably require model extrapolation. However, nearly all the
models discussed in this chapter presume equilibration between a given mineral-melt or min-
eral-mineral pair. In natural systems, the pairing of equilibrium phases is not always clear. For
example, in a lava flow, what is the liquid? Is it the whole rock? The glass? Some combination
thereof? One might expect that the whole rock is the most appropriate “liquid” when consider-
ing phenocrysts. But this further presumes that the whole rock acted as a closed system.
The methods of Roeder and Emslie (1970) and Rhodes et al. (1979a) are key to testing
for equilibrium. These tests are based on another assumption: two or more naturally occurring
phases are equilibrated if their compositions yield exchange coefficients (e.g., KD(Fe-Mg)ol-liq,
KD(Fe-Mg)ol-liq, KD(Fe-Mg)cpx-opx), or predicted mineral compositions (Rhodes et al. 1979a;
Putirka 1999), that mimic values determined in “equilibrium” (isothermal and isobaric)
experiments. Fe-Mg exchange coefficients are commonly used since they appear to be largely
independent of P or T. One strategy to estimate such error, then, is to estimate the standard
deviation for P-T estimates for crystals that pass some such equilibration filter. If the error is
less than model error then it is plausible that the crystals either precipitated from the liquid
used as input, or a liquid much like it, which in any case should provide a useful P-T estimate.
The hope then is that model error approaches true error.
108 Putirka
Compositional corrections. If mineral compositions from a single lava flow yield a
wide range of P-T estimates, then either the conditions of crystallization were not static (and
differences in P and T are real and should not be averaged), or the minerals crystallized from
different melts. In the later case, mineral-melt pairs might not yield “equilibrium” values for a
particular exchange coefficient, and one could make a correction for open system processing.
This correction could involve using the whole rock as a starting composition, then using mass
balance equations to add or subtract observed phenocryst phases until a “calculated liquid”
composition is derived that yields an equilibrium value for Fe-Mg exchange (between either
olivine-liquid or clinopyroxene-liquid; Putirka 1997; Putirka and Condit 2003), or a better
match between observed and calculated mineral components (Putirka and Condit 2003). These
corrections can result in a very large shift in estimates of P and T compared to estimates based
on observed (i.e., disequilibrium) mineral-whole rock or mineral-glass pairs. The assumption
is that error is decreased by such corrections, since the calculated liquid composition is, by
design, consistent with tests for equilibrium. But assuredly, error is attached to the calculation.
The error must be in proportion to whether the calculated liquid represents an actual liquid
composition. To estimate error, one could examine the variance of P-T estimates using observed
magma compositions that are similar to calculated compositions. And temperatures derived from
different equilibria should be better correlated with one another following such corrections.
The effects of disequilibrium. Since our thermometers and barometers are based on an
assumption of equilibrium, any form of disequilibrium (such as pairing a given mineral with
the wrong liquid) will yield errors. But how large are such errors? And how reliable are our
tests for equilibrium? Cooling rate experiments (see review by Hammer 2008), conducted at
constant P, but with a constantly changing T, provide an example of forced disequilibrium.
Here, Equation (32c) is tested using cooling rate experiments on a lunar ferro-basalt by Grove
and Bence (1979); the experiments were conducted at 1 atm (0.001 kbar) at cooling rates of 0.5
°C/hr to >600 °C/hr.
For these tests, P was calculated using the initial experimental T, and observed mineral
and melt compositions as input. For the 12 experiments conducted at cooling rates of 0.5-
10 °C/hr, Equation (32c) yields a P range of −1.82 to 1.86 kbar and a mean of −0.5±1.4
kbar—well within model error of 0.001 kbar (Fig. 14a), and with a standard deviation similar
to the model error. One lesson is that not all negative values of P should be ignored—they
may sometimes reflect the error that results when one attempts to predict a value that is
effectively zero. Another is that error is apparently random (and so the mean estimate is better
than any individual estimate). Still another is that at cooling rates ≤10 °C/hr, near-equilibrium
conditions are apparently achieved, since the model recovers the equilibrium P. Still another
lesson is relative to the use of the initial T as input into Equation (32c). If mean T or final
experimental T is used as input, the slow cooling rate experiments do not yield a mean of 1
atm. Apparently, slowly diffusing species, such as Al, record and retain initial P-T conditions,
and will not rapidly re-equilibrate, at least at lower temperatures. This means that, in practice,
clinopyroxene-based barometers might be expected to yield information about the deeper parts
of a magmatic system, and will not readily yield shallow P-T conditions unless precipitated at
such conditions.
Also interesting is that at cooling rates ≥40 °C/hr, the mean of P estimates using Equation
(32c) increases monotonically, to a value of 4.6 kbar at cooling rates >550 °C/hr (Fig. 14a).
Equation (32c) makes use of the clinopyroxene-melt partition coefficient for Al, DAlcpx-liq, and
this coefficient also increases with cooling rate, which provides the source for this 4.6 kbar
error (Fig. 14b). The interpretation: at low cooling rates, DAlcpx-liq retains its equilibrium value
and so provides the correct P (on average); at higher cooling rates, higher values of DAlcpx-liq
are disequilibrium values (see Lofgren et al. 2006). (One intriguing implication is that if P and
T are known, partition coefficients might be used to estimate cooling rates.)
Thermometers & Barometers for Volcanic Systems 109
Figure 14
3
Cooling Rate & Mean Calculated P
9
0.5-10 oC/hr (-0.5 kbar)
P(kbar)- Calculated (Eqn. 29d) 40 oC/hr (2.8 kbar)
150 oC/hr (4.0 kbar) 2.5
7 >550 oC/hr (4.6 kbar)
5 2
DAlcpx-liq
3
1.5
-1 P is calculated using initial
T (1473 K) as input
0.5
-3
a b
-5 0
0 100 200 300 400 500 600 700 800 0 100 200 300 400 500 600 700 800
Cooling Rate (oC/hr) Cooling Rate (oC/hr)
0.8 0.6
Cooling Rate & Mean Observed KD
0.5-10 oC/hr (0.18)
0.7
40 oC/hr (0.26)
DiHd
Cpx components - calculated
0.5
150 oC/hr (0.28)
0.6
>550 oC/hr (0.37)
KD(Fe-Mg)cpx-liq
0.4
0.5
0.4 0.3 KD calculated at 1200oC (initial T)
EnFs
0.3
0.2 KD calculated at 997oC (lowest final T)
0.2
0.1
0.1
c d
CaTs 0
0
0 100 200 300 400 500 600 700 800
0 0.2 0.4 0.6 0.8
Cpx components - measured Cooling Rate (oC/hr)
Figure 14. In (a) P is calculated using Equation (32c) for 1 atm cooling rate experiments (Grove and
Bence 1979); initial experimental T (1200 °C) is used as input. Experiments conducted at cooling rates
≤10 °C/hr yield a mean value of −0.5 kbar, effectively equivalent to the actual value of 0.001 (or ca.
0) kbar. (b) As crystallization rates exceed 40 °C/hr, the error on P increases, due to the increase in the
partition coefficient for Al. (c) The models of Putirka (1999) are used to predict the equilibrium values for
the clinopyroxene components DiHd, EnFs and CaTs, using experimental P, T and liquid compositions as
input. Only experiments conducted at cooling rates ≤10 °C/hr cluster near calculated equilibrium values.
The graph is effectively identical if calculated rather than observed P (0.001 kbar for all) is used as input.
(d) Observed values for KD(Fe-Mg)cpx-liq are compared to cooling rate. The highest cooling rate experiments
most strongly deviate from expected equilibrium values, but at lower cooling rates, Fe-Mg is a less effective
discriminator for equilibrium as observed values for KD closely match expected equilibrium values. The
gray box shows the range of KD values obtained from Equation (35) when using the highest T (initial T =
1200 °C) and the lowest of the final experimental T (975 °C) as input.
Can our tests for equilibrium effectively filter these disequilibrium cases (i.e., Roeder
and Emslie 1970; Putirka 1999)? Figure 14c shows results for the models of Putirka (1999)
when used to predict equilibrium clinopyroxene compositions for DiHd, EnFs, and CaTs (Fig.
14c), using initial T and observed P (0.001 kbar) as input. At the lower cooling rates, the
measured components match closely the predicted equilibrium values, and at higher cooling
rates, measured values deviate from equilibrium. So the test appears to be effective (Figure
14c is effectively identical when calculated P, instead of the observed 0.001 kbar, is used as
input).
110 Putirka
Values for KD(Fe-Mg)cpx-liq are also affected by cooling rate (see Hammer 2008), and P
derived from very high cooling rates could be rejected on the basis of such values. But at
cooling rates as high as 150 °C/hr, observed values are close to the equilibrium value (0.28,
when using initial T as input) and, curiously, low cooling rate experiments yield values that are
too low on average, even if the final experimental T is used as input into Equation (35). Thus Fe-
Mg exchange appears to be somewhat less effective a discriminator than DiHd, EnFs or CaTs,
which may reflect the more rapid diffusion of Fe and Mg compared to Al and Ca. Various tests
for equilibrium, at least in combination, should thus be successful at filtering erroneous results
(Fig. 14), but further tests of this sort should help greatly to better delimit error.
CONCLUSIONS
A great number of igneous thermometers and barometers have been calibrated since T.
W. Barth’s early efforts to calibrate a two-feldspar thermometer. In light of a rather expansive
experimental database (e.g., Hirschmann et al. 2008) we can now confidently calculate T, and in
some cases P, for nearly any volcanic or other igneous system, using relatively simple models.
The geologic limits of such methods are few. Nearly all the models presented here apply to
rocks equilibrated between 700-1700 °C, 1 bar- to 40 or 70 kbar, with water contents ranging
from 0-10 wt%. Liquid compositions range from basalt to rhyolite, from picrite to phonolite,
from tephrite to trachyte, and everything in between. Thermometers and barometers can be used
to calculate T and P to a precision of ±30 °C and ±1.5 kbar, respectively, for individual mineral-
melt or mineral-mineral pairs. Cooling rate (disequilibrium) experiments further indicate that
existing tests for equilibrium (i.e., Roeder and Emslie 1970; Rhodes et al. 1979b; Putirka 1999)
are useful as filters against non-equilibrium mineral-melt pairs.
A natural question is: Can precision be increased? More clever activity models offer a
seemingly slim hope; only the activity models of Wood and Banno (1973), Beattie (1993) and
Nimis and Taylor (2000) prove their worth in this regard, yielding thermometric errors of ~25-
30 °C. New experiments would aid the cases of the two-feldspar and two-pyroxene models,
and clarify the potential for plagioclase barometry. But absent the development of experimental
techniques that circumvent or subdue the obstacle of diffusion, increased precision might
instead be accomplished by the calibration of models targeted to a narrow range of rock types,
or geologic problems (using a narrow set of P-T-Xij conditions). This should obviate the need
for some of the compositional corrections that are required for thermometers and barometers
intended for broad application.
To increase accuracy, model tests could also be conducted using natural samples, where
P-T conditions are known. For example, Rosalind Helz (pers. comm.) reports that at Hawaii,
sequential skylight sampling shows the growth of new silicate phases, as lavas flow and cool
downstream. Geobarometers tend to yield 1 bar estimates in such cases (Putirka 2005b), even
when they do not yield 1 atm estimates for all 1atm experimental studies; additional tests of this
sort can provide compelling tests of accuracy and error for experimentally derived calibrations.
Undoubtedly, similar or more clever tests could be devised using well characterized systems,
such as at Mt. Etna, Mt. St. Helens, or Long Valley, etc., especially where new isotopic
methods (Cooper and Reid 2008) indicate the presence of long-lived systems, yielding “long
duration” (near-equilibrium?) experiments, mitigated, of course, by the potential problems of
open system behavior.
ACKNOWLEDGMENTS
Thanks to Alex Speer for all is his aid in organizing various aspects of the short course
and volume, and to Jodi Rosso for her many editorial efforts and endless patience in answering
Thermometers & Barometers for Volcanic Systems 111
questions. Thanks to my co-authors in this volume for their willingness to contribute. Finally,
this chapter greatly benefited by thoughtful reviews and comments by Lawford Anderson, Cin-
Ty Lee and Frank Tepley. This work was supported by NSF-EAR grants 0337345, 0421272
and 0313688.
REFERENCES
Agee CB, Walker D (1990) Aluminum partitioning between olivine and ultrabasic silicate liquid to 6 GPa.
Contrib Mineral Petrol 105:243-254
Agee CB, Draper DS (2004) Experimental constraints on the origin of Martian meteorites and the composition
of the Martian mantle. Earth Planet Sci Lett 224:415-429
Albarede F (1992) How deep to common basaltic magmas form and differentiate? J Geophys Res 97:10997-
11009
Almeev RA, Holtz F, Koepke J, Parat F, Botcharnikov RE (2007) The effect of H2O on olivine crystallization in
MORB: experimental calibration at 200 MPa. Am Mineral 92:670-674
Anderson DL (1989) Theory of the Earth. Blackwell, Brookline Village, MA
Anderson JL, Barth AP, Wooden JL, Mazdab F (2008) Thermometers and thermobarometers in granitic systems.
Rev Mineral Geochem 69:121-142
Ariskin AA (1999) Phase equilibria modeling in igneous petrology: use of COMAGMAT model for simulation
fractionation of ferro-basaltic magmas and the genesis of high-alumina basalt. J Volcanol Geotherm Res
90:115-162
Ariskin AA, Barmina GS (1990) Equilibria thermometry between plagioclases and basalt or andesite magmas.
Geochem Int 27:129-134
Armienti P, Tonarini S, Innocenti F, D’Orazio M (2007) Mount Etna pyroxene as tracer of petrogenetic processes
and dynamics of the feeding system. In: Cenozoic Volcanism in the Mediterranean Area. Beccaluva L,
Bianchini G, Wilson M (eds), Geol Soc Am Spec Paper 418:265-276
Arndt NT (1977) Partitioning of nickel between olivine and ultrabasic komatiite liquids. Carn Inst Wash Year
Book 76:553-557
Auzanneau E, Vielzeuf D, Schmidt MW (2006) Experimental evidence of decompression melting during
exhumation of subducted continental crust. Contrib Mineral Petrol 152:125-148
Bacon C, Carmichael ISE (1973) Stages in the P-T path of ascending basalt magma: an example from San
Quintin, Baja California. Contrib Mineral Petrol 41:1-22
Baker RB, Eggler DH (1987) Compositions of anhydrous and hydrous melts coexisting with plagioclase, augite,
and olivine or low-Ca pyroxene from 1 atm to 8 kbar: Application to the Aleutian volcanic center of Atka.
Am Mineral 72:12-28
Baker MB, Grove TL, Price R (1994) Primitive basalts and andesites from the Mt. Shasta region, N. California:
products of varying melt fraction and water content. Contrib Mineral Petrol 118:111-129
Barclay J, Carmichael ISE (2004) A hornblende basalt from western Mexico: water saturated phase relations
constrain a pressure-temperature window of eruptibility. J Petrol 45:485-506
Barnes SJ (1986) The distribution of chromium among orthopyroxene, spinel and silicateliquid at atmospheric
pressure. Geochim Cosmochim Acta 50:1889-1909
Bartels KS, Kinzler RJ, Grove TL (1991) High pressure phase relations of primitive high-alumina basalts from
Medicine Lake volcano, northern California. Contrib Mineral Petrol 108:253-270.
Barth TW (1934) Temperatures in lavas and magmas and a new geologic thermometer. Naturen 6:187-192
Barth TW (1951) The feldspar geologic thermometers. Neues Jahrb Mineral 82:143-154
Barth TW (1962) The feldspar geologic thermometers. Norsk Geol Tidsskr 42:330-339
Barth TW (1968) Additional data for the two-feldspar geothermometer. Lithos1:21-22
Beattie P (1993) Olivine-melt and orthopyroxene-melt equilibria. Contrib Mineral Petrol 115:103-111
Becquerel AC (1826) Procédé a l’aide duquel on peut mesurer d’intensité d’un courant électrique. Ann Chim
Phys 31:371
Beddoe-Stephens B, Aspden JA, Shepherd TJ (1983) Glass inclusions and melt compositions of the Toba tuffs,
northern Sumatra. Contrib Mineral Petrol 83:278-287
Bender JF, Langmuir CH, Hanson GN (1984) Petrogenesis of basalt glasses from the Tamayo Region, East
Pacific Rise. J Petrol 25:213-254
Benisek A, Kroll H, Cemic L (2004) New developments in two-feldspar thermometry. Am Mineral 89:1496-
1504
Berndt J, Holtz F, Koepke J (2001) Experimental constraints on storage conditions in the chemically zoned
phonolitic magma chamber of the Laacher See volcano. Contrib Mineral Petrol 140:469-486
Blatter DL, Carmichael ISE (2001) Hydrous phase equilibria of a Mexican high-silica andesite: A candidate for
a mantle origin? Geochim Cosmochim Acta 65:4043-4065
112 Putirka
Blundy J, Cashman K (2008) Petrologic reconstruction of magmatic system variables and processes. Rev
Mineral Geochem 69:179-239
Blundy J, Cashman K, Humphreys M (2006) Magma heating by decompression-driven crystallization beneath
andesite volcanoes. Nature 443:76-80
Bottinga Y, Weill DF (1972) The viscosity of magmatic silicate liquid, a model for calculation. Am J Sci
272:438-475
Bowen NL (1913) The melting phenomena of the plagioclase feldspars. Am J Sci 35:577-599
Bowen NL (1914) The ternary system: diopside-forsterite-silica. Am J Sci 38:270-264
Bowen NL (1915) The later stages of the evolution of the igneous rocks. J Geology Supp 8, 23:1-91
Bowen NL, Tuttle OF (1950) The system NaAlSi3O8-KAlSi3O8-H2O. J Geol 489-511
Box GEP, Draper, NR (1987) Empirical Model-Building and Response Surfaces. Wiley, New York
Brey GP, Kohler T (1990) Geothermobarometry in four-phase lherzolites II. New thermobarometers, and
practical assessment of existing thermobarometers. J Petrol 31:1353-1378
Buddington AF, Fahey J, Vlisdis A (1955) Thermometric and petrogenetic significance of titaniferous magnetite.
Am J Sci 253:497-532
Buddington AF, Lindsley DH (1964) Iron-titanium oxide minerals and synthetic equivalents. J Petrol 5:210-
357
Bulatov VK, Girnis AV, Brey GP (2002) Experimental melting of a modally heterogeneous mantle. Mineral
Petrol 75:131-152
Canil D (1999) The Ni-in-garnet geothermometer: calibration at natural abundances. Contrib Mineral Petrol
136:240-246
Carmichael ISE, Nicholls J, Smith AL (1970) Silica activity in igneous rocks. Am Mineral 55:246-263
Carmichael I.S.E., Turner FJ, Verhoogen J (1974) Igneous Petrology. McGraw-Hill Book Company, New York
Castellan GW (1971) Physical Chemistry, 2nd ed. Addison-Wesley, Reading, Massachusetts
Chen HK, Delano JW, Lindsley DH (1982) Chemistry and phase relations of VLT volcanic glasses from Apollo
14 and Apollo 17. Proc Lunar Planet Sci Conf 13:A171-A181
Cooper KM, Reid MR (2008) Uranium-series crystal ages. Rev Mineral Geochem 69:479-554
Dann JC, Holzheid AH, Grove TL, McSween HY (2001) Phase equilibria of the Shergotty meteorite: Constraints
on pre-eruptive water contents of Martian magmas and fractional crystallization under hydrous conditions.
Meteorit Planet Sci 36:793-806
Dasgupta R, Hirschmann MM, Smith N (2007) Partial melting experiments of peridotite + CO2 at 3 GPa and
genesis of alkalic ocean island basalts. J Petrol 48:2093-2124
Daubeny C (1835) Some account of the eruption of Vesuvius, which occurred in the month of August 1834,
extracted from the manuscript notes of the Cavaliere Monticelli, foreign member of the Geological Society;
together with a statement of the products of the eruption and of the condition of the volcano subsequently
to it. Philos Trans R Soc London 125:153-159
Davis BTC, Boyd DR (1966) The join Mg2Si2O6-CaMgSi2O6 at 30 kilobars pressure and its application to
pyroxenes from kimberlites. J Geophys Res 71:3567-3576
Day AL, Allen ET (1905) The isomorphism and thermal properties of the feldspars. Am J Sci 14:93-131
Delano JW (1977) Experimental melting relations of 63545, 76015, and 76055. Proc Lunar Planet Sci Conf
8:2097-2123
Di Carlo I, Pichavant M, Rotolo SG, Scaillet B (2006) Experimental crystallization of a high-K arc basalt: the
golden pumice, Stromboli Volcano (Italy). J Petrol 47:1317-1343
Doelter C (1902) Die chemische Zusammensetzung und die Genesis der Monzonigesteine. Tscher Mineral
Petrograph Mitteilung 21:191-225
Drake MJ (1976) Plagioclase-melt equilibria. Geochim Cosmochim Acta 40:457-465
Draper DS, Green TH (1997) P-T phase relations of silicic, alkaline, aluminous mantle-xenolith glasses under
anhydrous and C-O-H fluid-saturated conditions. J Petrol 38:1187-1224
Draper DS, Green TH (1999) P-T phase relations of silicic, alkaline, aluminous liquids: new results and
applications to mantle melting and metasomatism. Earth Planet Sci Lett 170:255-268
Draper DS, Johnston AD (1992) Anhydrous PT phase relations of an Aleutian high-MgO basalt; an investigation
of the role of olivine-liquid reaction in the generation of arc high-alumina basalts. Contrib Mineral Petrol
112:501-519
Dungan MA, Long PE, Rhodes JM (1978) Magma mixing at mid-ocean ridges: evidence from legs 45 and 45-
DSDP. Geophys Res Lett 5:423-425
Dunn T, Sen C (1994) Mineral/matrix partition coefficient for orthopyroxene, plagioclase, and olivine in basaltic
to andesitic systems: a combined analytical and experimental study. Geochim Cosmochim Acta 58:717-
733
Elkins-Tanton LT, Draper DS, Agee CB, Jewell J, Thorpe A, Hess PC (2007) The last lavas erupted during the
main phase of the Siberian flood volcanic province: results from experimental petrology. Contrib Mineral
Petrol 153:191-209
Thermometers & Barometers for Volcanic Systems 113
Elkins LT, Fernandes VA, Delano JW, Grove TL (2000) Origin of lunar ultramafic green glasses: Constraints
from phase equilibrium studies. Geochim Cosmochim Acta 64:2339-2350
Elkins LT, Grove TL (1990) Ternary feldspar experiments and thermodynamic models. Am Mineral 75:544-
559
Elkins-Tanton, LT, Chatterjee N, Grove TL (2003) Experimental and petrological constraints on lunar
differentiation from the Apollo 15 green picritic glasses. Meteorit Planet Sci 38:515-527
Essene EJ (1982) Geologic thermometry and barometry. Rev Mineral 10:153-206
Elthon D, Scarfe CM (1984) High-pressure phase equilibria of a high-magnesia basalt and the genesis of primary
oceanic basalts. Am Mineral 69:1-15
Esperança S, Holloway JR (1987) On the origin of some mica-lamprophyres; experimental evidence from a
mafic minette. Contrib Mineral Petrol 95:207-216
Falloon, TJ, Danyushevsky LV (2000) Melting of refractory mantle at 1.5, 2 and 2.5 GPa under anhydrous and
H20-undersaturated conditions: implications for the petrogenesis of high-Ca boninites and the influence of
subduction components on mantle melting. J Petrol 41:257-283
Falloon TJ, Danyushevsky LV, Green DH (2001) Peridotite melting at 1 GPa; reversal experiments on partial
melt compositions produced by peridotite-basalt sandwich experiments. J Petrol 42:2363-2390
Falloon TJ, Green DH (1987) Anhydrous partial melting of MORB pyrolite and other peridotite compositions at
10 kbar; implications for the origin of primitive MORB glasses. Mineral Petrol 37:181-219
Falloon TJ, Green DH, Danyushevsky LV, Faul UH (1999) Peridotite melting at 1.0 and 1.5 GPa: an experimental
evaluation of techniques using diamond aggregates and mineral mixes for determination of near-solidus
melts. J Petrol 40:1343-1375
Falloon TJ, Green DH, Hatton CJ, Harris KL (1988) Anhydrous partial melting of a fertile and depleted peridotite
from 2 to 30 Kb and application to basalt petrogenesis. J Petrol 29:1257-1282
Falloon TJ, Green DH, O-Neill H-St, Hibberson WO (1997) Experimental tests of low degree peridotite partial
melt compositions; implications for the nature of anhydrous near-solidus peridotite melts at 1 GPa. Earth
Planet Sci Lett 152:149-162
Feig ST, Koepke J, Snow JE (2006) Effect of water on tholeiitic basalt phase equilibria: an experimental study
under oxidizing conditions. Contrib Mineral Petrol 152:611-638
Fodor RV, Galar P (1997) A view into the subsurface of Mauna Kea Volcano, Hawaii: crystallization processes
interpreted through the petrology and petrography of gabbroic and ultramafic xenoliths. J Petrol 38:581-
624
Foulger GR, Natland JH (2003) Is “hotspot” volcanism a consequence of plate tectonics? Science 300:921-922
Fram MS, Longhi J (1992) Phase equilibria of dikes associated with Proterozoic anorthosite complexes. Am
Mineral 77:605-616
Frey FA, Garcia MO, Wise WS, Kennedy A, Gurriet P, Albarede F (1991) The evolution of Mauna Kea volcano,
Hawaii: petrogenesis of tholeiitic and alkalic basalts. J Geophys Res 96:14347-14375
Frey FA, Wise WS, Garcia MO, West H, Kwon S-T, Kennedy A (1990) Evolution of Mauna Kea volcano,
Hawaii: petrologic and geochemical constraints on postshield volcanism. J Geophys Res 95:1271-1300
Fuhrman ML, Lindsley DH (1988) Ternary-feldspar modeling and thermometry. Am Mineral 73:201-215
Fujii T, Bougault H (1983) Melting relations of a magnesian abyssal tholeiite and the origin of MORBs. Earth
Planet Sci Lett 62:283-295
Gaetani GA, Grove T (1998) The influence of melting of mantle peridotite. Contrib Mineral Petrol 131:323-
346
Garcia MO, Ho RA, Rhodes JM, Wolfe EW (1989) Petrologic constraints on rift-zone processes: results from
episode 1 of the Pu’u O’o eruption of Kilauea volcano, Hawaii. Bull Volcanol 52:81-96
Garcia MO, Hulsebosch TP, Rhodes JM (1995) Olivine-rich submarine basalts from the southwest rift zone
of Mauna Loa Volcano: implications for magmatic processes and geochemical evolution. In: Mauna
Loa Revealed: Structure, Composition, History, and Hazards. Rhodes JM, Lockwood JP (eds) AGU
Geophysical Monograph 92:219-239
Garcia MO, Rhodes JM, Wolfe EW, Ulrich GE, Ho RA (1992) Petrology of lavas from episodes 2-47 of the Pu’u
O’o eruption of Kilauea Volcano, Hawaii: evolution of magmatic processes. Bull Volcanol 55:1-16
Gardner JE, Rutherford M, Carey S, Sigurdsson H (1995) Experimental constraints on pre-eruptive water
contents and changing magma storage prior to explosive eruptions of Mount St Helens volcano. Bull
Volcanol 57:1-17
Gee LL, Sack RO (1988) Experimental petrology of melilite nephelinites. J Petrol 29:1233-1255
Gerke TL, Kilinc AI (1992) Enrichment of SiO2 in rhyolites by fractional crystallization: an experimental study
of peraluminous granitic rocks from the St. Francois Mountains, Missouri, USA. Lithos 29:273-283
Ghiorso MS (1987) Modeling magmatic systems: thermodynamic relations. Rev Mineral 17:443-465
Ghiorso MS, Hirschmann MM, Reiners PW, Kress VC III (2002) The pMELTS: a revision of MELTS for
improved calculation of phase relations and major element partitioning related to partial melting of the
mantle to 3 GPa. Geochem Geophys Geosyst 3:1-36, 10.1029/2001GC000217
114 Putirka
Glazner AF (1984) Activities of olivine and plagioclase components in silicate melts and their application to
geothermometry. Contrib Mineral Petrol 88:260-268
Green DH, Ringwood AE (1967) The genesis of basaltic magmas. Contrib Mineral Petrol 15:103-190
Grove TL, Beaty DW (1980) Classification, experimental petrology and possible volcanic histories of the Apollo
11 high-K basalts. Proc Lunar Planet Sci Conf 11:149-177
Grove TL, Bence AE (1979) Crystallization kinetics in a multiply saturated basalt magma: an experimental
study of Luna 24 ferrobasalt. Proc Lunar Planet Sci Conf 10:439-478
Grove TL, Bryan WB (1983) Fractionation of pyroxene-phyric MORB at low pressure: an experimental study.
Contrib Mineral Petrol 84:293-309
Grove TL, Donnelly-Nolan, JM Housh T (1997) Magmatic processes that generated the rhyolite of Glass
Mountain, Medicine lake volcano, N. California. Contrib Mineral Petrol 127:205-223
Grove TL, Elkins-Tanton LT, Parman SW, Chatterjee N, Müntener O, Gaetani GA (2003) Fractional crystallization
and mantle-melting controls on calk-alkaline differentiation trends. Contrib Mineral Petrol 145:515-533
Grove TL, Gerlach DC, Sando TW (1982) Origin of calc-alkaline series lavas at Medicine Lake Volcano by
fractionation, assimilation and mixing. Contrib Mineral Petrol 80:160-182
Grove TL, Juster TC (1989) Experimental investigations of low-Ca pyroxene stability and olivine-pyroxene-
liquid equilibria at 1-atm in natural basaltic and andesitic liquids. Contrib Mineral Petrol 103:287-305
Grove TL, Kinzler RJ, Bryan WB (1990) Natural and experimental phase relations of lavas from Serocki
Volcano. Proc Ocean Drill Prog 106/109:9-17
Grove TL, Kinzler RJ, Bryan WB (1992) Fractionation of mid-ocean ridge basalt (MORB). In: Mantle Flow
and Melt Generation at Mid-Ocean Ridges. Phipps Morgan J, Blackman, DK, Sinton JM (eds) AGU
Geophysical Monograph Series 7:281-310
Grove TL, Raudsepp M (1978) Effects of kinetics on the crystallization of quartz normative basalt 15597: an
experimental study. Proc Lunar Planet Sci Conf 9:585-599
Gudfinnsson GH, Presnall DC (1996) Melting relations of model lherzolite in the system CaO-MgO-Al2O3-SiO2
at 2.4-3.4 GPa and the generation of komatiites. J Geophys Res 101:27701-27709
Haase KM (1996) The relationships between the age of the lithosphere and the composition of oceanic magmas:
constraints on partial melting, mantle sources and the thermal structure of the plates. Earth Planet Sci Lett
144:75-92
Hakli TA, Wright TL (1967) The fractionation of nickel between olivine and augite as a geothermometer.
Geochim Cosmochim Acta 31:877-884
Hamada M, Fuji T (2007) H2O-rich island arc low-K tholeiite magma inferred from Ca-rich plagioclase-melt
inclusion equilibria. Geochem J 41:437-461
Hall J (1798) Experiments on whinstone and lava. Proc R Soc Edinburgh 5:43-74
Hammer JE (2008) Experimental studies of the kinetics and energetic of magma crystallization. Rev Mineral
Geochem 69:9-59
Hammer JE, Rutherford MJ, Hildreth W (2002) Magma storage prior to the 1912 eruption at Novarupta, Alaska.
Contrib Mineral Petrol 144:144-162
Hansteen TH, Klügel A (2008) Fluid inclusion thermobarometry as a tracer for magmatic processes. Rev
Mineral Geochem 69:143-177
Helz RT (1976) Phase relations of basalts in their melting ranges at PH2O = 5kb. Part II. Melt compositions. J
Petrol 17:139-193
Helz RT, Thornber CR (1987) Geothermometry of Kilauea Iki lava lake, Hawaii. Bull Volcanol 49:651-668
Herzberg C, Asimow PD, Arndt N, Niu Y, Lesher CM, Fitton JG, Cheadle MJ, Saunders AD (2007) Temperature
in ambient mantle and plumes: constraints form basalts, picrites and komatiites. Geochem Geoophys
Geosyst 8:10.1029/2006GC001390
Herzberg CT, Chapman NA (1976) Clinopyroxene geothermometry of spinel-lherzolites. Am Mineral 61:626-
637
Herzberg C, O’Hara MJ (1998) Phase equilibrium constraints on the origin of basalts, picrites and komatiites.
Earth Sci Rev 44:39-79
Herzberg C, O’Hara MJ (2002) Plume-associated ultramafic magmas off Phanerozoic age. J Petrol 43:1857-
1883
Hess H (1962) History of the ocean basins. In: Petrologic Studies: A Volume in Honor of A.F. Buddington.
Engle AE, James HL, Leonard BF (eds) GSA Special Volume,p 599-620
Hesse M, Grove TL (2003) Absarokites from the western Mexican Volcanic Belt: constraints on mantle wedge
conditions. Contrib Mineral Petrol 146:10-27
Hill DP, Zucca JJ (1987) Geophysical constraints on the structure of Kilauea and Mauna Loa volcanoes and
some implications for seismomagmatic processes. USGS Surv Prof Paper 1350:903-917
Hirschmann MM, Ghiorso MS, Davis FA, Gordon SM, Mukherjee S, Grove TL, Krawczynski M, Medard
E, Till CB (2008) Library of experimental phase relations: a database and web portal for experimental
magmatic phase equilibria data. Geochem Geophys Geosyst 9, Q03011, doi:10.1029/2007GC001894
Thermometers & Barometers for Volcanic Systems 115
Holbig ES, Grove TL (2008) Mantle melting beneath the Tibetan Plateau: Experimental constraints on
ultrapotassic magmatism. J Geophys Res doi:10.1029/2007JB005149
Holtz F, Becker A, Freise M, Johannes W (2001) The water-undersaturated and dry Qz-Ab-Or system revisited.
Experimental results at very low water activities and geological implications. Contrib Mineral Petrol
141:347-357
Holtz F, Sato H, Lewis J, Behrens H, Nakada S (2005) Experimental Petrology of the 1991-19995 Unzen Dacite,
Japan. Part I: Phase relations, phase composition, and pre-eruptive conditions. J Petrol 46:319-337
Housh TB, Luhr JF (1991) Plagioclase-melt equilibria in hydrous systems. Am Mineral 76:477-492
Jaggar TA (1917) Thermal gradient of Kilauea lava lake. Wash Acad Sci 7:397-405
Jefferys WH, Berger JO (1992) Ockham’s razor and Bayesian analysis. Am J Sci 80:64-72
Johnson KTM (1998) Experimental determination of partition coefficients for rare earth and high-field-strength
elements between clinopyroxene, garnet, and basaltic melt at high pressures. Contrib Mineral Petrol
133:60-68
Johnston AD (1986) Anhydrous P-T phase relations of near-primary high-alumina basalt from the South
Sandwich Islands. Contrib Mineral Petrol 92:368-382
JMP, Statistics and Graphics Guide (2002), Version 5, SAS Institute, Cary, NC, USA
Joly J (1891) On the determination of the melting points of minerals, Part 1. Uses of the meldometer. Proc R
Irish Acad 2:38
Jurewicz AJG, Mittlefehldt DW, Jones JH (1993) Experimental partial melting of the Allende (CV) and Murchison
(CM) chondrites and the origin of asteroidal basalts. Geochim Cosmochim Acta 57:2123-2139
Juster CT, Grove TL, Perfit MR (1989) Experimental constraints on the generation of FeTi basalts, andesite, and
rhyodacites at the Galapagos Spreading Center, 85°W and 95°W. J Geophys Res 94:9251-9274
Kawamoto T, Hervig RL, Holloway JR (1996) Experimental evidence for a hydrous transition zone in the early
Earth’s mantle. Earth Planet Sci Lett 142:587-592
Kelemen PB, Joyce DB, Webster JD, Holloway JR (1990) Reaction between ultramafic rock and fractionating
basaltic magma II. Experimental investigation of reaction between olivine tholeiite and harzburgite at
1150-1050 °C and 5kb. J Petrol 31:99-134
Kennedy AK, Grove TL, Johnson RW (1990) Experimental and major element constraints on the evolution of
lavas from Lihir Island, Papua New Guinea. Contrib Mineral Petrol 104:722-734.
Keshav S, Gudfinnsson GH, Sen G, Fei Y (2004) High-pressure melting experiments on garnet clinopyroxenite
and the alkalic to tholeiitic transition in ocean-island basalts. Earth Planet Sci Lett 223:365-379
Kinzler RJ, Grove TL (1985) Crystallization and differentiation of Archean komatiite lavas from northeast
Ontario: phase equilibrium and kinetic studies. Am Mineral 70:40-51
Kinzler RJ (1997) Melting of mantle peridotite at pressures approaching the spinel to garnet transition:
Application to mid-ocean ridge basalt petrogenesis. J Geophys Res 102:853-874
Kinzler RJ, Grove TL (1992) Primary magmas of mid-ocean ridge basalts 1. Experiments and methods. J
Geophys Res 97:6885-6906
Kjarsgaard BA (1998) Phase relations of a carbonated high-CaO nephelinite at 0.2 and 0.5 GPa. J Petrol 39:061-
2075
Klein EM, Langmuir CH (1987) Global correlations of ocean ridge basalt chemistry with axial depth and crustal
thickness. J Geophys Res 92:8089-8115
Klemme S, O’Neil HSC (2000) The near-solidus transition from garnet lherzolite to spinel lherzolite. Contrib
Mineral Petrol 138:237-248
Klügel A, Klein F (2005) Complex storage and ascent at embryonic submarine volcanoes from the Madeira
Archipelago. Geology 34:337-340
Koester E, Pawley AR, Fernandes LA, Porcher CC, Soliani Jr E ( 2002) Experimental melting of cordierite
gneiss and the petrogenesis of syntranscurrent peraluminous granites in southern Brazil. J Petrol 43:1595-
1616
Kogi R, Müntener O, Ulmer P, Ottolini L (2005) Piston-cylinder experiments on H2O undersaturated Fe-bearing
systems: an experimental setup approaching fO2 conditions of natural calc-alkaline magmas. Am Mineral
90:708-717
Kogiso T, Hirose K, Takahashi E (1998) Melting experiments on homogeneous mixtures of peridotite and basalt;
application to the genesis of ocean island basalts. Earth Planet Sci Lett 162:45-61
Kogiso T, Hirschmann MM (2001) Experimental study of clinopyroxenite partial melting and the origin of ultra-
calcic melt inclusions. Contrib Mineral Petrol 142:347-360
Kohn MT, Spear FJ (1991) Error propagation for barometers: 2. Applications to rocks. Am Mineral 76:138-
147
Kudo AM, Weill DF (1970) An igneous plagioclase thermometer. Contrib Mineral Petrol 25:52-65
Kushiro I (1996) Partial melting of a fertile mantle peridotite at high pressures: an experimental study using
aggregates of diamond. In: Earth Processes: Reading the Isotopic Code. Baso A, Hart S (eds) AGU
Geophysical Monograph 95:109-122
116 Putirka
Kushiro I, Mysen B (2002) A possible effect of melt structure on the Mg-Fe2+ partitioning between olivine and
melt. Geochim Cosmochim Acta 66:2267-2272
Kushiro I, Walter MJ (1998) Mg-Fe partitioning between olivine and mafic-ultramafic Melts. Geophys Res Lett
25:2337-2340
Langmuir CH, Hanson GN (1980) An evaluation of major element heterogeneity in the mantle sources of
basalts. Phil Trans R Soc Lond A 297:383-407
Laporte D, Toplis MJ, Seyler M, Devidal J-L (2004) A new experimental technique for extracting liquids from
peridotite at very low degrees of melting: application to partial melting of depleted peridotite. Contrib
Mineral Petrol 146:463-484
Leeman WP, Lindstrom DJ (1978) Partitioning of Ni2+ between basaltic and synthetic melts and olivines – an
experimental study. Geochim Cosmochim Acta 42:801-816
Lesher CE, Walker D (1988) Cumulate maturation and melt migration in a temperature gradient. J Geophys Res
93:10295-10311
Lindsley DH (1983) Pyroxene thermometry. Am Mineral 68:477-493
Lindsley DH, Anderson DJ (1983) A two-pyroxene thermometer. J Geophys Res 88:A887-A906
Liu TC, Chen BR, Chen CH (1998) Anhydrous melting experiment of a Wannienta basalt in the Kuanyinshan
area, northern Taiwan, at atmospheric pressure. Terr Atm Ocean Sci 9:165-182
Liu TC, Chen BR, Chen, CH (2000) Melting experiment of a Wannienta basalt in the Kuanyinshan area, northern
Taiwan, at pressures up to 2.0 GPa. J Asian Earth Sci 18:519-531
Lofgren GE, Huss GR, Wasserburg GJ (1006) An experimental study of trace-element partitioning between Ti-
Al-clinopyroxene and melt: equilibrium and kinetic effects including sector zoning. Am Mineral 91:1596-
1606
Longhi J (1976) Iron, Magnesium, and Silica in Plagioclase. Ph.D. Thesis, Harvard University, 296 pp
Longhi J (1995) Liquidus equilibria of some primary lunar and terrestrial melts in the garnet stability field.
Geochim Cosmochim Acta 59:2375-2386
Longhi J (2002) Some phase equilibrium systematics of lherzolite melting: I. Geochem Geophys Geosys
3:doi:10.1029/2001GC000204
Longhi J, Pan V (1988) A reconnaissance study of phase boundaries in low-alkali basaltic liquids. J Petrol
29:115-147
Longhi J, Pan V (1989) The parent magmas of the SNC meteorites. Proc Lunar Planet Sci Conf 19:451-464
Loomis TP (1979) An empirical model for plagioclase equilibrium in hydrous melts. Geochim Cosmochim Acta
43:1753-1759
Maaloe S (2004) The PT-phase relations of an MgO-rich Hawaiian tholeiite: the compositions of primary
Hawaiian tholeiites. Contrib Mineral Petrol 148:236-246
Mahood GA, Baker DR (1986) Experimental constraints on depths of fractionation of mildly alkalic basalts and
associated felsic rocks: Pantelleria, Strait of Sicily. Contrib Mineral Petrol 93:251-264
Marsh BD (1995) Solidification fronts and magmatic evolution. Mineral Mag 60:5-40
Marsh BD, Fournelle J, Myers JD, Chou I-M. (1990) On plagioclase thermometry in island-arc rocks:
experiments and theory. In: Fluid-Mineral interactions: A Tribute to H.P. Eugster. Spencer RJ, Chou I-M
(eds) Geochemical Society Special Publication 2:65-83
Martel C, Pichavant M, Holtz F, Scaillet B, Bourdier J, Traineau H (1999) Effects of fO2 and H2O on andesite
phase relations between 2 and 4 kbar. J Geophys Res 104:29453 29470
Mathez EA (1973) Refinement of the Kudo-Weill plagioclase thermometer and its applications to basaltic rocks.
Contrib Mineral Petrol 41:61-72
McKenzie D, Bickle MJ (1988) The volume and composition of melt generated by extension of the lithosphere.
J Petrol 29:625-679
McCoy TJ, Lofgren GE (1999) Crystallization of the Zagami shergottite: an experimental study. Earth Planet
Sci Lett 173:397-411
McDade P, Blundy JD, Wood BJ (2003) Trace element partitioning on the Tinaquillo lherzolite solidus at 1.5
GPa. Phys Earth Planet Interior 139:129-147
Medard E, Grove TL (2008) The effect of H2O on the olivine liquidus of basaltic melts: experiments and
thermodynamic models. Contrib Mineral Petrol 155:417-432
Medard E, Schmidt MW, Schiano P (2004) Liquidus surfaces of ultracalcic primitive melts: formation conditions
and sources. Contrib Mineral Petrol 148:201-215
Meen JK (1987) Formation of shoshonites from calcalkaline basalt magmas: geochemical and experimental
constraints from the type locality. Contrib Mineral Petrol 97:333-351
Meen JK (1990) Elevation of potassium content of basaltic magma by fractional crystallization: the effect of
pressure. Contrib Mineral Petrol 104:309-331
Mercier J-C, Beoit V, Girardeau J (1984) Equilibrium state of diopside-bearing harzburgites from ophiolites:
geobarometric and geodynamic implications. Contrib Mineral Petrol 85:391-403
Mibe K, Fuji T, Yasuda A, Ono S (2005) Mg-Fe partitioning between olivine and ultramafic melts at high
pressure. Geochim Cosmochim Acta 70:757-766
Thermometers & Barometers for Volcanic Systems 117
Milholland CS, Presnall DC (1998) Liquidus phase relations in the CaO-MgO-Al2O3-SiO2 system at 3.0 GPa:
the aluminous pyroxene thermal divide and high pressure fractionation of picritic and komatiitic magmas.
J Petrol 39:3-27
Minitti ME, Rutherford MJ (2000) Genesis of the Mars Pathfinder sulfur-free rock from SNC parental liquids.
Geochim Cosmochim Acta 64:2535-2547
Montierth C, Johnston AD, Cashman KV (1995) An empirical glass-composition-based geothermometer for
Mauna Loa lavas. In: Mauna Loa Revealed: Structure, Composition, History, and Hazards. Rhodes JM,
Lockwood JP (eds) AGU Geophysical Monograph 92:207-217
Moore G, Carmichael ISE (1998) The hydrous phase equilibria (to 3 kbar) of an andesite and basaltic andesite
from western Mexico: constraints on water content and conditions of phenocryst growth. Contrib Mineral
Petrol 130:304-319
Mordick BE, Glazner AF (2006) Clinopyroxene thermobarometry of basalts from the Coso and Big Pine
volcanic fields, California. Contrib Mineral Petrol doi 10.1007s00410-006-0097-0
Morgan WJ (1971) Convection plumes in the lower mantle. Nature 230:42-43
Morse SA, Brady JB, Sporleder BA (2004) Experimental petrology of the Kiglapait Intrusion; cotectic trace for
the lower zone at 5 kbar in graphite. J Petrol 45:2225-2259
Müntener O, Kelemen PB, Grove TL (2001) The role of H2O during crystallization of primitive arc magmas
under uppermost mantle conditions and genesis of igneous pyroxenites: and experimental study. Contrib
Mineral Petrol 141:643-658
Musselwhite DS, Dalton HA, Kiefer WS, Treiman AH (2006) Experimental petrology of the basaltic shergottite
Yamato-980459: Implications for the thermal structure of the Martian mantle. Meteorit Planet Sci 41:1271-
1290
Naney MT (1983) Phase equilibria of rock-forming ferromagnesian silicates in granitic Systems. Am J Sci
283:993-1033
Nicholls J, Carmichael ISE (1972) The equilibration temperature and pressure of various lava types with spine-
and garnet-peridotite. Am Mineral 57:941-959
Nicholls J, Carmichael ISE, Stormer JC (1971) Silica activity and Ptotal in igneous rocks. Contrib Mineral Petrol
33:1-20
Nielsen RL, Drake MJ (1979) Pyroxene-melt equilibria. Geochim Cosmochim Acta 43:1259-1272
Nimis P (1995) A clinopyroxene geobarometer for basaltic systems based on crystals-structure modeling.
Contrib Mineral Petrol 121:115-125
Nimis P (1999) Clinopyroxene geobarometry of magmatic rocks. Part 2. Structural geobarometers for basic to
acid, tholeiitic and mildly alkaline systems. Contrib Mineral Petrol 135:62-74
Nimis P, Taylor WR (2000) Single clinopyroxene thermobarometry for garnet peridotites. Part 1 Calibration and
testing of a Cr-in-cpx barometer and an enstatite-in-cpx thermometer. Contrib Mineral Petrol 139:541-
554
Nimis P, Ulmer P (1998) Clinopyroxene geobarometry of magmatic rocks. Part 1: an expanded structural
geobarometer for anhydrous and hydrous basic and ultrabasic systems. Contrib Mineral Petrol 133:122-
135
Nordstrom DK, Munoz JL (1986) Geochemical Thermodynamics. Blackwell, Brookline, Massachusetts
O’Nions RK, Powell R (1977) The thermodynamics of trace element distribution. In: Thermodynamics in
Geology. Fraser DG (ed) Reidel, Dordrecht-Holland, p 349-363
Panjasawatwong Y, Danyushevsky LV, Crawford AJ, Harris KL (1995) An experimental study of the effects of
melt composition on plagioclase-melt equilibria at 5 and 10 kbar: implications for the origin of magmatic
high-An plagioclase. Contrib Mineral Petrol 118:420-432.
Papike JJ, Cameron KL, Baldwin K (1974) Amphiboles and pyroxenes: characterization of other than
quadrilateral components and estimates of ferric iron from microprobe data. Geol Soc Am Abst Prog
6:1053-1054
Parman SW, Dann JC, Grove TL, deWit MJ (1997) Emplacement conditions of komatiite magmas from the 3.49
Ga Komati Formation, Barberton Greenstone Belt, South Africa. Earth Planet Sci Lett 150:303-323
Parman SW, Grove TL (2004) Harzburgite melting with and without H2O: experimental data and predictive
modeling. J Geophys Res 109:doi:10.1029/2003JB002566
Parsons T, Thompson, GA (1993) Does magmatism influence low-angle normal faulting? Geology 21:247-250
Patino-Douce AE (1995) Experimental generation of hybrid silicic melts by reaction of high-Al basalt with
metamorphic rocks. J Geophys Res 100:15623-15639
Patino-Douce AE (2005) Vapor-absent melting of tonalite at 15-32 kbar. J Petrol 46:275-290.
Patino-Douce AE, Beard JS (1995) Dehydration-melting of biotite gneiss and quartz amphibolite from 3 to 15
kbar. J Petrol 36:707-738.
Patino-Douce AE, Harris N (1998) Experimental constraints on Himalayan anatexis. J Petrol 39:689-710
Perret RA (1913) Volcanic research at Kilauea in the summer of 1911. Am J Sci 4:475-483.
118 Putirka
Petermann M, Hirschmann MM (2000) Anhydrous partial melting experiments on MORB-like eclogite:
Phase relations, phase compositions and mineral-melt partitioning of major elements at 2-3 GPa. J Petrol
44:2173-2201
Petermann M, Lundstrom CC (2006) Phase equilibrium experiments at 0.5 GPa and 1100-1300 °C on a basaltic
andesite from Arenal volcano, Costa Rica. J Volcanol 157:222-235
Petry C, Chakraborty S, Palme H (1997) Olivine-melt nickel-iron exchange Thermometer: cosmochemical
significance and preliminary experimental results. Meteor Planet Sci 32:106-107
Pichavant M, Martel C, Bourdier J-L, Scaillet B (2002a) Physical conditions, structure, and dynamics of a zoned
magma chamber: Mount Pele (Martinique, Lesser Antilles Arc). J Geophys Res 107:1-25
Pichavant M, Mysen BO, MacDonald R (2002b) Source and H2O content of high-MgO magmas in island arc
settings: an experimental study of a primitive calc-alkaline basalt from St. Vincent, Lesser Antilles arc.
Geochim Cosmochim Acta 66:2193-2209
Pickering-Witter J, Johnston AD (2000) The effects of variable bulk composition on the melting systematics of
fertile peridotitic assemblages. Contrib Mineral Petrol 140:190-211
Powell M, Powell R (1977) Plagioclase-alkali feldspar geothermometry revisited. Mineral Mag 41:253-256
Prouteau G, Scaillet B, Pichavant M, Maury R (2001) Evidence for mantle metasomatism by hydrous silicic
melts derived from subducted oceanic crust. Nature 410:197-200
Putirka K (1997) Magma transport at Hawaii: inferences from igneous thermobarometry. Geology 25:69-72
Putirka K (1998) Garnet + liquid equilibrium. Contrib Mineral Petrol 131:273-288
Putirka K (1999a) Clinopyroxene+liquid equilibrium to 100 kbar and 2450 K. Contrib Mineral Petrol 135:151-
163
Putirka K (1999b) Melting depths and mantle heterogeneity beneath Hawaii and the East Pacific Rise: Constraints
from Na/Ti and REE ratios. J Geophys Res 104:2817-2829.
Putirka K (2005a) Mantle potential temperatures at Hawaii, Iceland, and the mid-ocean ridge system, as inferred
from olivine phenocrysts: Evidence for thermally–driven mantle plumes. Geochem Geophys Geosys
doi:10.1029/2005GC000915
Putirka K (2005b) Igneous thermometers and barometers based on plagioclase + liquid equilibria: tests of some
existing models and new calibrations. Am Mineral 90:336-346
Putirka K (2008) Excess temperatures at ocean islands: implications for mantle layering and convection.
Geology 36:283-286
Putirka K, Condit C (2003) A cross section of a magma conduit system at the margins of the Colorado Plateau.
Geology 31:701-704
Putirka K, Johnson M, Kinzler R, Walker D (1996) Thermobarometry of mafic igneous rocks based on
clinopyroxene-liquid equilibria, 0-30 kbar. Contrib Mineral Petrol 123:92-108
Putirka K, Perfit M, Ryerson FJ, Jackson MG (2007) Ambient and excess mantle temperatures, olivine
thermometry, and active vs. passive upwelling. Chem Geol 241:177-206.
Putirka K, Ryerson FJ, Mikaelian H (2003) New igneous thermobarometers for mafic and evolved lava
compositions, based on clinopyroxene + liquid equilibria. Am Mineral 88:1542-1554
Ratajeski K, Sisson TW, Glazner AF (2005) Experimental and geochemical evidence for derivation of the El
Capitan Granite, California, by partial melting of hydrous gabbroic lower crust. Contrib Mineral Petrol
149:713-734
Rhodes JM, Dungan MA, Blanchard DP, Long PE (1979a) Magma mixing at mid-ocean ridges: evidence form
basalts drilled near 22°N on the mid-Atlantic ridge. Tectonophys 55:35-61
Rhodes JM, Lofgren GE, Smith DP (1979b) One atmosphere melting experiments on ilmenite basalt 12008.
Proc Lunar Planet Sci Conf 10:407-422
Rhodes JM, Vollinger MJ (2004) Composition of basaltic lavas sampled by phase-2 of the Hawaii Scientific
Drilling Project: geochemical stratigraphy and magma series types. Geochem Geophys Geosys
5:doi:10.1029/2002GC00434
Righter K, Leeman WP, Hervig RL (2006) Partitioning of Ni, Co, and V between spinel-structured oxides and
silicate melts; importance of spinel composition. Chem Geol 227:1-25
Robinson JAC, Wood BJ, Blundy JD (1998) The beginning of melting of fertile and depleted peridotite at 1.5
GPa. Earth Planet Sci Lett 155:97-111
Roeder PL, Emslie RF (1970) Olivine-liquid equilibrium. Contrib Mineral Petrol 29:275-289
Rutherford MJ, Sigurdsson H, Carey S, Andrew. (1985) The May 18, 1980, eruption of Mount St. Helens; 1,
Melt composition and experimental phase equilibria. J Geophys Res 90:2929-2947
Ryan MP (1988) The mechanics and three-dimensional internal structure of active magmatic systems: Kilauea
volcano, Hawaii. J Geophys Res 93:4213-4248
Ryan TP (1997) Modern Regression Methods. Wiley, New York
Sack RO, Walker D, Carmichael ISE (1987) Experimental petrology of alkalic lavas: constraints on cotectics of
multiple saturation in natural basic liquids. Contrib Mineral Petrol 96:1-23
Salters VJM, Longhi J (1999) Trace element partitioning during the initial stages of melting beneath mid-ocean
ridges. Earth Planet Sci Lett 166:15-30
Thermometers & Barometers for Volcanic Systems 119
Scaillet B, Evans BW (1999) The 15 June 1991 Eruption of Mount Pinatubo. I. Phase Equilibria and pre-
eruption P-T-fO2-fH2O conditions of the dacite magma. J Petrol 40:381-411
Scaillet B, MacDonald R (2003) Experimental constraints on the relationships between peralkaline rhyolites of
the Kenya rift valley. J Petrol 44:1867-1894
Schmidt MW, Green DH, Hibberson WO (2004) Ultra-calcic magmas generated from Ca-depleted mantle: an
experimental study on the origin of ankaramites. J Petrol 45:531-554.
Schwab BE, Johnston AD (2001) Melting systematics of modally variable, compositionally intermediate
peridotites and the effects of mineral fertility. J Petrol 42:1789-1811
Scoates JS, Lo Cascio M, Weis, D, Lindsley DH (2006) Experimental constraints on the origin and evolution
of mildly alkalic basalts from the Kerguelen Archipelago, Southeast Indian Ocean. Contrib Mineral Petrol
151:582-599
Sen G (1985) Experimental determination of pyroxene compositions in the system CaO-MgO-Al2O3-SiO2 at
900-1200 °C and 10-15 kbar using PbO and H2O fluxes. Am Mineral 70:678-695
Sen G, Jones R (1989) Experimental equilibration of multicomponent pyroxenes in the spinel peridotite field:
implications for practical thermometers and a possible barometer. J Geophys Res 94:17871-17880
Sen G, Keshav S, Bizimis M (2005) Hawaiian mantle xenoliths and magmas: composition and thermal character
of the lithosphere. Am Mineral 90:871-887
Shepherd FS, Rankin GA, Wright FE (1909) The binary systems of alumina with silica, lime and magnesia. Am
J Sci 28:293-315
Sisson TW, Grove TL (1993a) Experimental investigations of the role of H2O in calc-alkaline differentiation and
subduction zone magmatism. Contrib Mineral Petrol 113:143-166
Sisson TW, Grove TL (1993b) Temperatures and H2O contents of low-MgO high-alumina basalts. Contrib
Mineral Petrol 113:167-184
Skjerlie KP, Patino-Douce AE (2002) The fluid-absent partial melting of a zoisite-bearing quartz eclogite from
1.0 to 3.2 GPa; implications for melting in thickened continental crust and for subduction-zone processes.
J Petrol 43:291-314
Sosman RB, Merwin HE (1913) Data on the intrusion temperature of the Palisade diabase. J Wash Acad 3:389-
395
Spear FS (1993) Metamorphic Phase Equilibria and Pressure-Temperature-Time Paths. Mineralogical Society
of America Monograph, Washington D.C.
Springer W, Seck HA (1997) Partial fusion of basic granulites at 5 to 15 kbar: implications for the origin of TTG
magmas. Contrib Mineral Petrol 127:30-45
Stolper E (1977) Experimental petrology of eucritic meteorites Geochim Cosmochim Acta 41:587-611
Stolper E (1980) A Phase Diagram for Mid-Ocean Ridge Basalts: Preliminary Results and Implications for
Petrogenesis. Contrib Mineral Petrol 74:13-27
Stolper E, Walker D (1980) Melt density and the average composition of basalt. Contrib Mineral Petrol 74:7-
12
Stormer JC (1975) A practical two-feldspar geothermometer. Am Mineral 60:667-674
Sugawara T (2001) Ferric iron partitioning between plagioclase and silicate liquid: thermodynamics and
petrological applications. Contrib Mineral Petrol 141:659-686
Takagi D, Sato H, Nakagawa N (2005) Experimental study of a low-alkali tholeiite at 1-5kbar:optimal condition
for the crystallization of high-An plagioclase in hydrous arc tholeiite. Contrib Mineral Petrol 149:527-
540
Takahashi E (1980) Melting relations of an alkali-olivine basalt to 30 kbar, and their bearing on the origin of
alkali basalt magmas. Carneige Institute Washington Year Book 79:271-276
Takahashi E, Nakajima K, Wright TL (1998) Origin of the Columbia River basalts: melting model of a
heterogeneous plume head. Earth Planet Sci Lett 162:63-80
Taura H, Yurimoto H, Kurita K, Sueno S (1998) Pressure dependence on partition coefficients for trace elements
between olivine and the coexisting melts. Phys Chem Mineral 25:469-484
ten Brink US, Brocher TM (1987) Multichannel seismic evidence for a subcrustal intrusive complex under Oahu
and a model for Hawaiian volcanism. J Geophys Res 92:13687-13707
Thompson JB (1967) Thermodynamic properties of simple solutions. In: Researches in Geochemistry, Vol 2.
Abelson PH (ed) Wiley, New York, p 340-361
Thompson JB (1982) Composition space: an algebraic and geometric approach. Rev Mineral 10:1-31
Thornber CR, Heliker C, Sherrod DR, Kauahikaua JP, Miklius A, Okubo PG, Trusdell FA, Budahn JR, Ridley
WI, Meeker GP (2003) Kilauea East Rift Zone magmatism: an episode 54 perspective. J Petrol 44:1525-
1559
Thy P (1991) High and low pressure phase equilibria of a mildly alkalic lava from the 965 Surtsey eruption:
experimental results. Lithos 26:223-243
Thy P, Lesher CE, Fram MS (2004) Low Pressure experimental constraints on the evolution of basaltic lavas
from site 917, southeast Greenland continental margin. Proc Ocean Drill Prog 152:359-372
120 Putirka
Toplis MJ (2005) The thermodynamics of iron and magnesium partitioning between olivine and liquid: criteria
for assessing and predicting equilibrium in natural and experimental systems. Contrib Mineral Petrol
149:22-39
Tormey DR, Grove TL, Bryan WB (1987) Experimental petrology of normal MORB near the Kane Fracture
Zone: 22°-25° N, mid-Atlantic Ridge. Contrib Mineral Petrol 96:121-139
Torrens HS (2006) The geological work of Gregory Watt, his travels with William Maclure in Italy (1801-1802)
and Watt’s “proto-geological” map of Italy (1804). In: The Origins of Geology in Italy. Vai GB, Caldwell
WGE (eds) p 179-198
Tsuruta K, Takahashi E (1998) Melting study of the an alkali basalt JB-1 up to 12.5 GPa: behavior of potassium
in the deep mantle. Phys Earth Planet Int 107:119-130
Tuttle OF, Bowen NL (1958) The origin of granite in the light of experimental studies in the system NaAlSi3O8-
KAlSi3O8-SiO2-H2O. Geol Soc Am Mem 174
Vander Auwera J, Longhi J (1994) Experimental study of jotunite (hypersthene monzodiorite): constraints on
the parent magma composition and crystallization conditions (P, T, fO2) of the Bjerkreim-Sokndal layered
intrusion (Norway). Contrib Mineral Petrol 118:60-78
Vander Auwera J, Longhi J, Duchesne J-C (1998) A liquid line of descent of the jotunite (hypersthene
monzodiorite) suite. J Petrol 39:439-468
Villiger S, Ulmer P, Müntener O, Thompson AB (2004) The liquid line of descent of anhydrous, mantle-derived,
tholeiitic liquids by fractional and equilibrium crystallization - an experimental study at 1.0 GPa. J Geol
45:2369-2388
Wagner TP, Grove TL (1997) Experimental constraints on the origin of lunar high-Ti ultramafic glasses. Geochim
Cosmochim Acta 61:1315-1327
Wagner TP, Grove TL (1998) Melt/harzburgite reaction in the petrogenesis of tholeiitic magma from Kilauea
volcano, Hawaii. Contrib Mineral Petrol 131:1-12
Walter MJ (1998) Melting of garnet peridotite and the origin of komatiite and depleted lithosphere of komatiite
and depleted lithosphere. J Petrol 39:29-60
Walter MJ, Presnall DC (1994) Melting behavior of simplified lherzolite in the system CaO-MgO-Al2O3-SiO2-
Na2O from 7 to 35 kbar. J Petrol 35:329-359
Wasylenki LE, Baker MB, Kent AJR, Stolper EM (2003) Near-solidus melting of the shallow upper mantle:
partial melting experiments on depleted peridotite. J Petrol 44:1163-1191
Wells PRA (1977) Pyroxene thermometry in simple and complex systems. Contrib Mineral Petrol 62:129-139
Whitaker ML, Nekvasil, H, Lindsley DH, Difrancesco NJ (2007) The role of pressure in producing compositional
diversity in intraplate basaltic magmas. J Petrol 48:365-393
Whitney JA, Stormer JC (1977) The distribution of NaAlSi3O8 between coexisting microcline and plagioclase
and its effect on geothermometric calculations. Am Mineral 62:687-691
Wood BJ (1974) The solubility of alumina in orthopyroxene coexisting with garnet. Contrib Mineral Petrol
46:1-15
Wood BJ, Banno S (1973) Garnet-orthopyroxene and orthopyroxene-clinopyroxene relationships in simple and
complex systems. Contrib Mineral Petrol 42:109-124
Wood BJ, Trigila R (2001) Experimental determination of aluminous clinopyroxene-melt partition coefficients
for potassic liquids, with application to the evolution of the Roman province potassic magmas. Chem Geol
172:213-223
Wright FE, Larsen ES (1909) Quartz as a geologic thermometer. Am J Sci 27:421-447
Xirouchakis D, Hirschmann MM, Simpson J (2001) The effect of titanium on the silica content of mantle-
derived melts. Geochem Cosmochim Acta 65:2029-2045
Yang HJ, Frey FA, Clague DA, Garcia MO (1996) Mineral chemistry of submarine lavas from Hilo Ridge,
Hawaii: implications for magmatic processes within Hawaiian rift zones. Contrib Mineral Petrol 135:355-
372
Yasuda A, Fujii T, Kurita K (1994) Melting phase relations of an anhydrous mid-ocean ridge basalt from 3 to
20 GPa: Implications for the behavior of subducted oceanic crust in the mantle. J Geophys Res 99:9401-
9414
Yoder HS, Tilley CE (1962) Origin of basalt magmas: an experimental study of natural and synthetic rock
systems. J Petrol 3:342-532
Reviews in Mineralogy & Geochemistry
Vol. 69 pp. 121-142, 2008 4
Copyright © Mineralogical Society of America
thermometers and thermobarometers
in Granitic Systems
J. Lawford anderson
Department of Earth Sciences, University of Southern California
Los Angeles, California, 90089-0740, U.S.A.
[email protected]
andrew P. Barth
Department of Earth Sciences, Indiana University-Purdue University
Indianapolis, Indiana, 46202, U.S.A.
Joseph L. Wooden, Frank Mazdab
U.S. Geological Survey, 345 Middlefield Road
Menlo Park, California, 94205, U.S.A.
IntrOduCtIOn
Recent advances in our ability to resolve the P-T-t (pressure-temperature-time) paths of
orogenic terranes have continued to re-emphasize the common marriage between petrology
and tectonics. The importance of determining the P-T history of metamorphic rocks, as through
chemical analysis of mineral zoning profiles and inclusions initially paved the way into this
area of our science (e.g., Essene 1982, 1989; Hoisch 1991; Speer 1993). Parallel success has
been demonstrated for igneous systems (see review by Anderson 1996), with the derived
information being primarily related to ascent and/or cooling of the crystallizing magma. And, in
multiply-intruded terranes, where the ages of intrusion span a significant portion of the orogenic
deformation, dated plutons act as important crustal “nails,” which offer first-order insight into
the descent and/or ascent of deformed crust during the orogenic process.
Emplacement barometry can be retrieved from contact metamorphic rocks where the
enclosing host assemblages contain appropriate compositions for barometric calculations.
However, in pervasively intruded terranes where a range of ages of magmatic activity exists,
multiple and extensive overprinting makes the use of host metamorphic mineral assemblages
and mineral compositions challenging. In contrast, several igneous minerals are less prone to
the loss of primary compositions, in part due to the inherent insensitivity of the rock to high-
grade subsolidus reequilibration and the paucity of post-crystallization fluids. However, there
are exceptions to this generalization and below we discuss an example where high-T, post-
emplacement fluids have altered the entire margin of the Mt. Stuart batholith of Washington.
In general, many igneous rocks lack mineral assemblages suitable for thermobarometric
analysis. Either the thermodynamic variance is high or, in specific cases, the phases are
characterized by component exchanges easily reset during slow cooling or later thermal events.
Two feldspar and iron-titanium oxide thermometry are examples of the latter case; only in
shallow-emplaced plutons does feldspar or iron-titanium oxide thermometry yield consistent
hypersolidus results. Nevertheless, as reviewed below, many granitic intrusions provide the
means to determine temperature and pressure attending crystallization through mineral equilibria
less prone to subsequent readjustment. For example, recent calibrations of Ti solubility in
1529-6466/08/0069-0004$05.00 DOI: 10.2138/rmg.2008.69.4
122 Anderson, Barth, Wooden & Mazdab
quartz and zircon and zr solubility in sphene offer new means to determine crystallization
temperatures in most igneous rocks. These can be also be used in concert with apatite, zircon,
and allanite saturation temperatures.
For metaluminous granites, the Al-in-hornblende thermobarometer continues to be impor-
tant, particularly where used in conjunction with the above thermometers or that of hornblende-
plagioclase, pyroxene-hornblende, or two pyroxenes, with the caveat that the temperature-
sensitive equilibria record the same conditions as that used for barometry. For peraluminous
granites, the recently calibrated pressure-sensitive equilibrium of garnet-plagioclase-biotite-
muscovite can be used in concert with the garnet-biotite thermometer. Other potential barometers
are based on the pressure dependent Si-content of muscovite in equilibrium with biotite, alkali
feldspar, quartz, and a hydrous fluid.
A basic question in igneous emplacement thermobarometry is whether or not the mineral
phase compositions represent near-solidus conditions, as opposed to entrained solids grown or
acquired during ascent or retrogressed during subsolidus cooling. Thorough petrography is an
ever important prerequisite for the identification of solid phases in equilibrium with melt at the
crustal level of final crystallization and emplacement as is the full elemental characterization
of zoned minerals. Clearly, not all magmatic phases record the same history (e.g., Claiborne et
al. 2006).
Plutonic barometry of Cordilleran orogenic terranes - a historical perspective
In their now classic papers, E-An zen and Jane Hammarstrom (zen 1985, 1989; zen and
Hammarstrom 1984; Hammarstrom and zen 1986) demonstrated that many post-accretionary
plutons within the exotic terranes of northwest North America (Idaho to British Columbia)
are deep-seated with estimated emplacement depths ranging from 18 to 30 km. They founded
the basis of magmatic epidote as having pressure significance, and pioneered the ever popular
Al-in-hornblende barometer. For several northwest Cordilleran terranes, their work not only
revealed deep levels of pluton emplacement but also demonstrated that plutonism was followed
by tectonic ascent of these terranes at rates greater than 1 mm/yr. At about the same time,
Anderson et al. (1988) first documented the mid-crustal origin of the lower plate of several
metamorphic core complexes and, based on thermobarometry of successive plutonic events,
determined tectonic ascent at rates of 2 mm/yr during core complex extension. Similarly,
Barth (1990) documented the mid-crustal origin of a Mesozoic complex in the San Gabriel
(Tujunga) terrane of southern California. Barth (1990) further showed that portions of the San
Gabriel terrane remained at depths of approximately 22 km throughout all phases of Triassic
to Late Cretaceous plutonism, followed by rapid tectonic uplift during the early Tertiary. These
papers were part of a wave of renewed and continued interest in plutonic thermobarometry in
response to zen and Hammarstrom’s contributions. In 1996, Anderson provided an overview of
the status of thermobarometry in granitic batholiths. Now, over a decade later, this contribution
serves to build on that earlier work, along with companion papers by Putirka (2008) and
Blundy and Cashman (2008).
SInGLe MIneraL traCe eLeMent therMOMeterS
In IGneOuS SySteMS
Understanding the mechanisms that cause diversity in plutonic rocks and how these
mechanisms are linked to the processes controlling explosive volcanic eruptions requires
improved understanding of the thermal evolution of plutonic and volcanic rocks. However,
information about this thermal evolution is most often limited to the superstructure of the
youngest systems, because the evolutionary history of the older and the deeper parts of volcanic
systems, and their plutonic roots is often obscured or erased by hydrothermal alteration or later
metamorphism. Recent experimental calibrations, with support from empirical observations
Thermometers & Thermobarometers in Granitic Systems 123
in natural systems, have shown the value of using the trace element content of zircon, quartz,
and sphene for estimating temperatures and evaluating the thermal evolution of intermediate
to silicic igneous rocks (Wark and Watson 2006; Watson et al. 2006; Ferry and Watson 2007;
Hayden et al. 2007; Ferriss et al. 2008). These experiments are particularly important because
of the near ubiquitous occurrence of quartz in intermediate and silicic plutonic rock and as
phenocrysts in silicic volcanic rocks, the widespread occurrence of trace amounts of zircon and
sphene in these same rocks, and the potential resistance of these minerals to later metamorphic
and hydrothermal re-equilibration. New insights into deeper volcanic and plutonic processes,
and variation in these processes through greater spans of geologic time, are contained in
compositional variations within and among these major and trace phases.
The observation of the linear temperature dependence of trace element contents of zircon,
rutile, and sphene developed in these studies is particularly powerful because these minerals are
well-known hosts of the radiogenic parent isotopes of U, Th and Lu. Thus, temperature-time
histories may now be constructed from combined multi-element and isotopic analyses of these
phases. This is particularly relevant given recent findings confirming that many batholiths can
form over several million years through a protracted plutonic history (Matzel et al. 2006; Miller
et al. 2007; Walker et al. 2007). Furthermore, because zircon saturation is not always achieved
during partial melting reactions and/or during the ascent of intermediate and silicic magmas,
pre-magmatic zircon crystals are often incompletely dissolved, and thus may retain thermal,
petrologic and geochronologic information about the magma source region. The chemistry of
zircon, sphene, and quartz can, therefore, track elements from the full evolutionary path of
intermediate and felsic magma formation, evolution and solidification.
ti-in-quartz thermometer
Wark and Watson (2006) developed the following experimental calibration describing the
temperature dependence of the titanium content of quartz:
log Ti(ppm)qtz = 5.69 − [3765/T(K)] (1)
which expressed in T (°C) becomes
3765
T ( °C ) = − − 273 (2)
log Ti ( ppm ) − 5.69
qtz
For rutile-undersaturated rocks (a common condition in igneous rocks), Wark and Watson
(2006) offer the following:
3765 − 273
T ( °C ) = − (3)
(
log Ti ( ppm ) a
qtz )
TiO2 − 5.69
Ti concentrations greater than about 15-20 ppm can be measured with a conventional electron
microprobe, but quartz formed at lower temperatures requires ion microprobe or ICP-MS
analyses. Although all of the experiments on which (1) is based were run at 10 kbar, comparison
to limited 1 atm experiments and empirical observations (Wark and Watson 2006; Wark et al.
2007; Wiebe et al. 2007) suggests the effect of pressure on this relation is negligible for crustal
rocks formed at 10 kbar or less. Analytical uncertainties are quite small in the application of
this thermometer, but temperature can be seriously underestimated if aTiO2<1, as much as 50-
75 °C over the typical temperature range of quartz growth during solidification of intermediate
to silicic melts. If aTiO2 is unconstrained, Equation (1) yields a minimum temperature of
crystallization assuming aTiO2 = 1. For example at aTiO2 = 1, quartz Ti concentration of 40
ppm yields a temperature of 648 °C, and 100 ppm derives the estimate of 747 °C. At aTiO2
= 0.6, these estimates rise to 700 and 813 °C, respectively. Reduced titanium activity can
124 Anderson, Barth, Wooden & Mazdab
be estimated using the rutile solubility relation of Hayden and Watson (2007), or calculated
independently from Equation (1) if another mineral geothermometer is available, or through
sphene-based equilbria (xirouchakis et al. 2001a, b), as described further below.
An early application of this thermometer in igneous systems is the study of quartz
phenocrysts in the Bishop Tuff, east central California (Wark et al. 2007). Previous studies have
estimated the temperature and compositional zonation of this unit using other equilibria (e.g.
Hildreth 1981). Wark et al. (2007) showed that the quartz Ti thermometer yields temperatures
in good agreement with temperatures calculated from the compositions of coexisting Fe-Ti
oxides assuming an aTiO2 of 0.63±0.03. Furthermore, well-preserved compositional zonation
in quartz that was imaged by cathodoluminescence microscopy recorded an abrupt increase in
temperature of ca. 60 °C in the later erupted (hence deeper) parts of the tuff. Wark et al. (2007)
related these high-T quartz rims to renewed growth induced by an increase in T and decrease
in aH2O caused by mafic magma injection immediately preceding the climactic eruption of the
tuff. These results indicate that Equation (1) constitutes a robust thermometer that can recover
thermal variations in evolving, compositionally-zoned silicic volcanic systems. The work also
presents the issue of accurately assessing crystal growth history, given that each mineral can be
stable and grow over a range of conditions and matching such to that of other minerals remains
a serious challenge.
To date, this thermometer has not been tested on many granitic plutons and it may not
resistant to subsolidus re-equilibration given the ease that quartz undergoes recrystallization.
However, the study of Wiebe et al. (2007) on the Vinalhaven granite of Maine shows clear
promise. Assuming an aTiO2 of 0.6, based on the study of Wark et al. (2007) for the Bishop Tuff,
their work resulted in Ti-in-quartz temperatures ranging 700-770 °C for a fine grained phase
of the pluton, with quartz cores yielding a T near 810 °C. Likewise, quartz in a coarse-grained
granite yielded temperatures ranging 700-840 °C, the higher range of which clearly reflect
hypersolidus conditions.
ti-in-zircon and Zr-in-rutile thermometers
zircon in igneous systems is usually very near to the zrSiO4 end member, with 0.5 to
2.5% HfSiO4 solid solution (Speer 1982). Among the many trace elements found in zircon
(Hoskin and Schaltegger 2003), quadravalent Ti can substitute for Si, as indicated by the quartz
experiments of Wark and Watson (2006) described above. Watson et al. (2006) developed a
combined empirical and experimental calibration that describes the following temperature-
dependent variation in the abundance of titanium in zircon:
log Ti(ppm)zircon = 6.01 − [5080/T(K)] (4)
Rutile grown in the presence of zircon by Watson et al. (2006) preserved zr abundances that
were linearly dependent on temperature. Similarly, they offered the calibration
log zr(ppm)rutile = 7.36 – [4470/T(K)] (5)
Equation (5) is of limited utility as a thermometer in igneous rocks because of the rarity of
rutile-saturated magmas (Hayden and Watson 2007), but this thermometer will find utility in
metamorphic systems, particularly eclogites.
Ti concentrations can be measured in zircon formed at high temperature using a
conventional electron microprobe, but ICP-MS or ion microprobe analyses are necessary for
zircon crystallized at moderate to low temperatures characteristic of many intermediate and
felsic igneous systems (Wooden et al. 2006).
In a study of the Spirit Mountain batholith of Nevada, Claiborne et al. (2006) adjusted
Equation (4) to reflect an aTiO2 of 0.7, based on conditions appropriate for sphene and titano-
magnetite saturation, as the following
Thermometers & Thermobarometers in Granitic Systems 125
5080
T ( °C ) = − 273 (6)
( 6.01 − log ( 0.7 ⋅ ppmTizircon ) )
Their results for zircons having Ti concentrations ranging 3.3 to 33 ppm range 680-899 °C,
and are comparable to their reported zircon saturation temperatures (see below) which range
from 696-878 °C.
Ferry and Watson (2007) presented an improved thermodynamic analysis of Ti substitution
in zircon, demonstrating the dependence of Ti substitution on both the activities of TiO2 and
SiO2 (using rutile and α-quartz as standard states). Their revised calibration of the thermometer
emphasizes the effects of reduced activities on Ti abundance in zircon:
( ) (
logTi ( ppm )zircon = 5.711 − 4800 / T ( K ) − log aSiO2 + log aTiO2 ) ( 7)
The differences between Equations (6) and (7) derive from additional experiments used in
the later calibration, and the incorporation of coefficients to account for reduced activities of
SiO2 and/or TiO2 in natural systems that crystallized zircon but were not quartz- and/or rutile-
saturated. Ferry and Watson (2007) suggested that temperatures calculated using Equation (4)
will be underestimated by ≤70 °C if aTiO2 < 1 and overestimated by approximately the same
amount if aSiO2 < 1; these effects thus may partly compensate for each other, but care must be
taken to evaluate effects of reduced activities (Heiss et al. 2008)
Equation (7) is particularly powerful for relative thermometry from mineral compositional
zoning profiles and for accurate thermometry from systems where the activities are well-
known. As an example, consider the complexly zoned zircons preserved in typical calc-alkalic
plutonic rocks in the U.S Cordillera. Trace element analyses of zircons from a Late Cretaceous
granodiorite and granite from the eastern Transverse Ranges of California illustrate the
application of the zircon thermometer to gain insight into the petrogenesis of igneous rocks.
Analyses of the zircons were completed by ion microprobe on the USGS SHRIMP-RG and
representative analyses are given in Table 1. The ion microprobe has the advantage of lower
detection limits in comparison to the electron microprobe, but it necessitates using a 20-30
µm spot size, about 10 times the diameter of a typical electron microprobe spot. The value
and limitation of this technique are shown in Figure 1, where it is apparent that although the
ion microprobe yields both crucial age and extensive compositional data for each spot within
these complexly-zoned zircons, it clearly averages compositionally variable domains that are
visually resolvable through an SEM image and would be resolvable (in part, for Hf, Y, U and
perhaps some heavy rare earth elements) with an electron microprobe.
zircon thermometry results for a sphene, hornblende, biotite granodiorite indicate
crystallization of zircon over a wide temperature range, with 9000-16000 ppm Hf in solid
solution (Fig. 2). Hf-rich zircon rims yield an average temperature of ~713±30 °C (aSiO2 = 1), in
good agreement with the quartz- and water-saturated granodiorite solidus, which is appropriate
for this bulk composition, and in agreement with the temperature of 708±17 °C calculated from
hornblende - plagioclase thermobarometry in another sample of this granodiorite (Needy et al.,
in press). zircon interiors are typically Ti-enriched relative to rims and yield similar to higher
temperatures at progressively lower Hf concentrations, up to an average temperature about
60 °C above the estimated temperature of zircon saturation (see also Clairborne et al. 2006,
for correlation of whole-rock and zircon Hf content and fractionation). However, all of these
calculated temperatures are probably a few tens of degrees too high, because phase equilibria
suggest the granodiorite was not silica-saturated at these temperatures. Temperatures of zircon
interiors recalculated at aSiO2 = 0.55 are in reasonable agreement with phase equilibria and
with the estimated zircon saturation temperature. Comparing core and rim temperatures and
zoning profiles indicates that the zircons record solidification of a relatively cool, crystal-rich
126 Anderson, Barth, Wooden & Mazdab
table 1. Representative SHRIMP-RG ion microprobe
analyses of zircons from Blue granodiorite.
Spot Sample hf (ppm) ti (ppm) t(W) t(FW)1 t(FW)2
JW340-5.1 rim 15610 4.1 668 698 na
JW340-6.1 core 10945 6.8 708 743 690
JW340-6.2 rim 13530 5.4 690 722 na
JW340-6.3 intermediate 9481 24.9 828 881 813
Notes:
T(W) = temperature by calibration of Watson et al (2006)
T(FW)1 = temperature by calibration of Ferry and Watson (2007), aSiO2 = 1 and aTiO2 = 0.7
T(FW)2 = temperature by calibration of Ferry and Watson (2007), aSiO2 = 0.55 and aTiO2 = 0.7
722
743
100 microns
813
Precambrian 775
736
100 microns
Precambrian
Jurassic
Blue granodiorite
Figure 1. Scanning electron microscope (SEM) cathodoluminescence image of zircons in the Blue
granodiorite. Circles are the locations of ion microprobe spot analyses of pre-magmatic cores of
Precambrian and Jurassic age, and squares are locations of spot analyses of magmatic rims. Rims yield
an average temperature
Figure of 713±39
1. Scanning °Cmicroscope
electron with the Ferry and Watson
(SEM) (2007) calibration
cathodoluminescence (aSiOof
image =1, aTiO2 =0.7).
2 zircons in
Inset shows an oscillatory-zoned, magmatic zircon crystal preserving large temperature variations during
the Blue granodiorite. Circles are the locations of ion microprobe spot analyses of pre-
growth, within the interval between zircon saturation and the wet solidus.
magmatic cores of Precambrian and Jurassic age, and squares are locations of spot
analyses of magmatic rims. Rims yield an average temperature of 713±39°C with the
Ferry and Watson (2007) calibration (aSiO2=1, aTiO2=0.7). Inset shows an oscillatory-
granodiorite
zoned,magma thatzircon
magmatic may have never
crystal been above
preserving largeits saturation variations
temperature temperature, consistent
during growth,with
extensive preservation
within of premagmatic
the interval betweenzircon. Analyses
zircon of discrete
saturation andcompositional
the wet bands within
solidus.
the magmatic zircon indicate dynamic crystallization of the granodiorite within the temperature
range bounded by the zircon saturation temperature and the water-saturated solidus.
zircon thermometry results for a nearby pluton, the Palms biotite granite (Fig. 3), allow
us to compare the crystallization histories of two calc-alkalic plutons. The Palms granite has
a much more limited range in temperature between the estimated water-saturated liquidus and
solidus and earlier silica saturation, illustrating the well-known near eutectic-like composition
of hydrous granitic bulk compositions. Hf-rich zircon rims yield temperatures in reasonable
agreement with the wet solidus (aSiO2 = 1). Hf-poorer zircon interiors yield temperatures well
above the nominal liquidus temperature and many yield temperatures up to 100 °C above
zircon saturation. However, all of these calculated temperatures are too high because phase
equilibria suggest that the granite was probably not quartz saturated over the temperature
range. Recalculated temperatures of Hf-poor zircon interiors at aSiO2 = 0.5 average ~800 °C,
near to the estimated zircon saturation temperature; these values are more reasonable, because
the granite also preserves premagmatic zircon in the cores of many grains.
Thermometers & Thermobarometers in Granitic Systems 127
Figure 2. Hf vs. temperature plot for all analyzed magmatic zircons in the Blue granodiorite, using the
thermometer of Ferry and Watson (2007). Open circles for all spot analyses (aSiO2 = 1, aTiO2 = 0.7), and closed
circles for recalculated grain interiors (aSiO2 = 0.55, aTiO2 = 0.7). Representative analytical precision (box,
central left) based on replicate analyses of standard zircon Cz3. Water-saturated granodiorite liquidus, qz-in,
and solidus at 4 kb from Piwinskii (1968) and Stern et al. (1975). zircon saturation temperature from Watson
and Harrison (1983) and apatite saturation temperature from Harrison and Watson (1984). Solidus temperature
of 708±17 °C estimated from hornblende - plagioclase thermobarometry (Needy et al., in press). Dashed
arrows connect analyses within a single zircon crystal (see Fig. 1), illustrating large temperature variations
without significant zircon dissolution during solidification of the granodiorite.
Figure 3. Hf vs. temperature plot for all analyzed magmatic zircons in Palms granite, using the thermometer
of Ferry and Watson (2007). Open circles for all spot analyses (aSiO2 = 1, aTiO2 = 0.7), and closed circles for
recalculated grain interiors (aSiO2 = 0.5, aTiO2 = 0.7). Water-saturated granite liquidus, qz-in, and solidus at 4
kbar from Piwinskii (1968) and Stern et al. (1975). zircon saturation temperature from Watson and Harrison
(1983), and apatite saturation temperature from Harrison and Watson (1984).
128 Anderson, Barth, Wooden & Mazdab
The core to rim temperature estimates of zircon indicate that the granite and granodiorite
magmas crystallized over similar temperature ranges, from near to or below zircon saturation
down to the wet solidii. Most of the Palms granite zircons indicate crystallization above
the water-saturated liquidus, indicating that early crystallization was probably under water-
undersaturated conditions and that the magma may have been largely liquid early in its
crystallization history. In contrast, even for water-saturated conditions, the granodiorite records
no evidence for crystallization near its liquidus, and thus was probably a comparatively crystal-
rich magma. These results, along with crystal zoning profiles, provide important limitations
for models of chemical evolution and physical processes of magma transport and interaction
in this calc-alkalic arc setting.
Two other recent studies (Page et al. 2007; Fu et al. 2008) have found apparent low
temperatures derived from the Ti-in-zircon geothermometer in a range of settings and have
suggested that parameters other than temperature and aTiO2 and aSiO2 may be important in
controlling the Ti content of zircon. One issue may be the effect of pressure, which needs to
be further evaluated. Although most of the experiments from which Equation (4) was derived
were run at 10 kbar, limited experimental data combined with empirical observations from
rocks crystallized at higher pressure initially suggested that the effect of pressure on Ti-in-
zircon thermometry was on the order of 50 °C per 1 GPa (10 kbar) of uncertainty (Ferry and
Watson 2007). In contrast, the more recent study of Ferriss et al. (2008) indicates a larger
pressure dependence for the geothermometer, on the order 100 °C per 1 GPa. Clearly, this
promising geothermometer needs to be used in awareness of these issues.
Zr-in-sphene thermometer
Sphene in igneous systems is usually near to the CaTiSiO5 end member, exhibiting some
substitution of Al and Fe+3 charge-balanced by coupled substitution of OH or F (Ribbe 1982).
Among the many trace elements found in sphene, quadravalent zr can substitute for Ti, as
can be inferred from the rutile experiments of Watson et al. (2006) described above. Hayden
et al. (2007) developed a combined empirical and experimental calibration that describes the
temperature and pressure dependence of the abundance in sphene, which is as follows:
7708 P ( GPa )
logzr ( ppm )sphene = 10.52 − − 960 ×
T ( K ) T ( K )
( ) (
− log aTiO2 − log aSiO2 ) (8)
The experiments upon which Equation (8) are based were run at 800-1000 °C and 10-24
kbars; addition of some natural, well-characterized sphene samples can extend the calibration
to lower temperatures and pressures (Mazdab et al. 2007). zr concentrations greater than
about 150 ppm (~800 °C) can be measured with a conventional electron microprobe. Care
must be taken to avoid interferences caused by overlapping peaks from Nb and Y. Although
most problematic for background location at low zr concentrations, tails from some of these
interferences on the zr peak may also be important; the significance of this is a function
of the relative concentrations of zr, Nb and Y, as well as the quality of the electron probe
and analyzing crystal. Sphene formed at lower temperatures requires ion microprobe analyses
(Mazdab et al. 2007). All the zr isotopes are interfered by dimers that are largely not resolvable
by any mass spectrometer, limiting the detection limit to approximately 10 ppm.
Analytical uncertainties are quite small in the application of this thermometer, but temperature
can be seriously underestimated if aTiO2 < 1, which would be expected to be common in igneous
rocks as almost all lack rutile (see Hayden and Watson 2007) and/or if aSiO2 <1. Our findings
are that underestimates for the above reasons can be as much as 50-75 °C over the typical
temperature range for solidification of intermediate to silicic melts. An additional important
complication is that sphene typically displays sector zoning, with anomalously zr-enriched
sectors that should be recognized and avoided in using Equation (8) to calculate temperatures
Thermometers & Thermobarometers in Granitic Systems 129
of crystallization (Hayden et al. 2007; Mazdab et al. 2007). The origin of zr-enriched sectors
in sphene are not known to us and offer new areas of study. However, Watson and Liang (1995)
suggest that sector zoning can occur during slow growth when lattice diffusion rates for the
relevant element are low, and when the diffusing element is highly charged. Their model was
developed for rare earth elements, but likely may also apply to zr in sphene.
For the typical case where rocks contain sphene and quartz, but not rutile, aSiO2 can be
assumed to be unity, at least in the later stages of crystallization. Finally, aTiO2 can be estimated
from the rutile solubility model of Hayden and Watson (2007), requiring knowledge of the
composition of the melt from which the sphene crystallized. This can also be derived from
electron microprobe analysis of glass coexisting with sphene or estimated from whole rock
analysis.
Precise and accurate application of the zr content of sphene as a thermometer for igneous
rocks requires an independent estimate or calculation of temperature (and pressure), for example
from the hornblende-plagioclase thermobarometer (Anderson and Smith 1995; see below).
aCCeSSOry PhaSe SaturatIOn therMOMetry
Harrison and Watson (1984) and Watson and Harrison (1983) initially calibrated compo-
sitional and temperature constraints on the saturation conditions of zircon and apatite in silicic
melt systems. Subsequently, Green and Adam (2002) determined that apatite saturation is not
effected by pressure up to 25 kbar. More recently, Tollari et al. (2006) have offered a revision of
the apatite saturation thermometer that is particularly applicable to peraluminous granites.
Saturation thermometers, such as these presented here, offer a powerful means to assess
temperature of magmatic systems. Where used in conjunction with critical petrography,
they offer constraints on magmatic liquidus temperatures as well as an additional test of
solidus conditions. Should one of these phases, usually zircon, be inherited, antecrystic, or of
cumulate origin, derived temperatures may not have geologic significance and be in excess of
the liquidus. In contrast, should zircon saturation temperatures be near-solidus, then inherited
or antecrystic zircon can be expected to partially or fully dissolve. Recent applications include
Hoskin et al. (2000), Miller et al. (2003), Hoskin and Schaltegger (2003), Janousek et al.
(2004), Lopez-Moro and Lopez-Plaza (2004), Walker et al. (2007), and Miller et al. (2007) and
all concluded their findings consistent with other forms of thermometry (see also Clairborne
et al. 2006, mentioned above). However, Harrison et al. (2007) has presented evidence that
the zircon saturation thermometer may at times underestimate the onset temperature of zircon
crystallization, which will require further evaluation.
Figure 4 exhibits new apatite and zircon saturation temperatures for Cretaceous plutons
near Yosemite National Park, California, near Cinko Lake of the Sierra Nevada batholith (Foley
et al. 2007). Notably, petrography for these granitoids shows that apatite is an early phase in of
the crystallization sequence relative to that for zircon. As indicated below, these plutons, based
on hornblende-plagioclase thermobarometry (Anderson and Smith 1995; Holland and Blundy
1994 calibrations) were emplaced at 698-715±18 °C (error based on multiple analysis of
hornblende and plagioclase rims, usually in contact with quartz). In contrast, apatite saturation
temperatures for both plutons are near 900 °C, which offers an estimate of liquidus temperatures.
zircon saturation temperatures range 790±10 °C for the Harriet Lake pluton and 740±6 °C for
the Fremont Lake pluton, which are above but near the solidus and are comparable to zircon
saturation temperatures obtained elsewhere in the Yosemite region (Miller et al. 2007).
Klimm et al. (2008) have recently offered experimental data in support of temperature
constraints on allanite saturation. As one of the epidote group minerals, the LREE enrichment of
allanite expands its stability field. Unlike epidote, allanite appears as an accessory phase in many
130 Anderson, Barth, Wooden & Mazdab
Cinko Lake Apatite and Zircon Saturation Temperatures
950
Diorite
900
Quartz Diorite
apatite
Fremont Lake granodiorite
850
Harriet Lake granodiorite
T (°C)
Lecucogranite
800
Metavolcanics
750 zircon
700
650
50 55 60 65 70 75 80
SiO2 (wt%)
Figure 4. Apatite and zircon saturation temperatures for plutons and older metavolcanics
units in the Cinko Lake region of Yosemite National Park (Foley et al. 2007).
metaluminous granites with possibly no pressure significance. In magmatic epidote-bearing
plutons, allanite also can also form cores of these crystals. Thus, its onset of crystallization
provides additional information on plutonic melt conditions between the liquidus and solidus.
Klimm et al. (2008) have conducted melt experiments with a mid ocean ridge basalt at 750-900
°C on the Ni-NiO buffer to constrain the T-X (REE composition) control of allanite stability
in melts having a silica range above 50% SiO2. This accessory phase saturation thermometer
has not yet been tested for plutonic rocks but offers important additional information on
temperatures of magmas, both plutonic and volcanic. The pressure effect needs also to be
further investigated.
therMOBarOMeterS aPPLICaBLe tO GranItIC rOCKS
al-in-hornblende barometry
This popular barometer has yet to be fully experimentally evaluated but it has clearly
served well. Anderson and Smith (1995) recalibrated the barometer to account for the effects
of temperature using the experimental data of Johnson and Rutherford (1989) and Schmidt
(1992). Their calibration is
T °C − 675
P ( kbar ) = 4.76Al − 3.01 − × 0.530 Al + 0.005294 ( T °C − 675 ) (9)
85
where Al is the sum of Aliv + Alvi per 13 cations. Ague (1997) has offered a thermodynamic
model for the barometer and Ernst and Liu (1998) have presented new experimental data in
support of a more qualitative model that emphasizes both the Al and Ti contents of amphibole.
Earlier calibrations, both empirical and experimental (Hammarstrom and zen 1986; Hollister
et al. 1987; Johnson and Rutherford 1989; Schmidt 1992), which built the early formulation
of this thermobarometer, lacked correction for the effects of temperature, which can be
substantial (see Anderson and Smith 1995). Hence, early uses of the barometer that lacked
temperature corrections should be viewed with caution and subject to revision. An appropriate
thermometer is that based on plagioclase + hornblende (Holland and Blundy 1994) which
Thermometers & Thermobarometers in Granitic Systems 131
offers calibrations based on two equilibria:
edenite + 4 quartz = tremolite + albite (10a)
edenite + albite = richterite + anorthite (10b)
Anderson (1996) concluded that the calibration based on the second reaction (10b), which
yielded lower temperatures, more precisely produced temperatures from other thermometers. It
was thus the preferred calibration of the thermometer, including in conjunction with the Anderson
and Smith (1995) calibration of the Al-in-hornblende barometer, which in simultaneous solution
often yield P-T solutions at or above the appropriate granite to tonalite solidii. In their own
investigation of the Fish Canyon tuff, Bachman and Dungan (2002) made the same conclusion.
It should be noted that Holland and Blundy (1994) have an appropriate hornblende formula
scheme imbedded in their two thermometers that requires its use in their application.
Interest in the Al-in-hornblende thermobarometer is quite evident in recent publications
(e.g., Sial et al. 1999; Stone 2000; Barnes et al. 2001; Stein and Dietl 2001; Percival et al. 2002,
Asrat et al. 2004; Janousek et al. 2004; zhang et al. 2006; Rodriguez et al. 2007). Most of these
have used the P-T information to document the depth of batholith emplacement. In volcanic
systems, Manley and Bacon (2000), Lindsay et al. (2001), and Bachman et al. (2002), used
the thermobarometer to determine the magma chamber depth feeding into volcanic systems.
In contrast, Bachl et al. (2001), Miller and Miller, (2002), Needy et al. (2008), and Barth et al.
(2008) used the thermobarometer to document tilted crustal sections enclosing plutons.
Anderson et al. (2007) applied granitic pluton thermobarometry and host metavolcanic
units thermometry to present a model of upper crustal overturn attending pluton emplacement
in construction of the Sierra Nevada batholith in the vicinity of Yosemite Valley. Their results
depict shallow pluton emplacement at ~2.7±0.4 kbar at 698-715±18 °C (Fig. 5), leading to near
thermal metamorphic equilibration of steeply-dipping and lineated host metavolcanic units at
614-712 °C, which are inferred to have descended to pluton levels at ca. 9-10 km before or
during pluton emplacement. Their model of “upper crustal overturn” is based on the conclusion
that surficial volcanic deposits had, subsequent to eruption, had been overturned and descended
to plutonic depths of ~10 km, a topic further addressed by Paterson et al. (2008).
The Al-in-hornblende thermobarometer also had success in documenting cases of
subsolidus reequilibration on a batholith scale. Figure 6 offers comparison of P-T estimates
for emplacement of 14 Mesozoic plutons in the Mojave Desert region of southern California
(Anderson 1996) to the Cretaceous Mt. Stuart batholith of Washington (Anderson 1997;
Anderson et al. 2008). Both apply the same calibration of hornblende-plagioclase thermometer
(Holland and Blundy 1994) used in concert with the Anderson and Smith (1995) calibration
of the Al-in-hornblende barometer. The Mojave Desert region of California is rather unique
in the western U.S. in its variable exposures of upper and middle crustal exposures of the
Mesozoic arc (Anderson et al. 1992), and, it is notable in this context that all of these plutons
yielded P-T solutions at or above the granite to tonalite solidii. Clearly, sizeable portions of the
Mt. Stuart batholith yield erroneously low and subsolidus P-T solutions. The batholith is also
isotopically zoned with respect to margin enrichment in δ18O (ranging 7-8‰ upwards to over
10‰). Anderson et al. (2008) documented that where the batholith margin whole-rock oxygen
isotope values of δ18O are elevated, ∆Pl-qz (δ18O of plagioclase vs. that of quartz) values
are large (> 1.5; due largely to increased δ18O of plagioclase, but in detail the δ18O of both
minerals have increased), which relates to a lower temperature of oxygen isotope equilibration
of these minerals. They attributed this to subsolidus infiltration of high-temperature contact
aureole fluids into the batholith margin, which both elevated the oxygen isotope values of all
major minerals of the batholith margin and also reset hornblende and plagioclase elemental
compositions. Bucher and Frost (2006) reported similar influx of contact metamorphic fluids
into an anorogenic granite in Antarctica.
132 Anderson, Barth, Wooden & Mazdab
Cinko Lake Plutons
10
granite solidus
Metamorphism
8 tonalite solidus metavolcanic rocks
640 - 712 °C
Pressure (kb)
6
Plutons
Fremont: 2.7 ± 0.3 kb
4 @ 715 ± 18 °C
Harriet: 2.8 ± 0.4 kb
@ 698 ± 16 °C
2
0
600 650 700 750 800 850
Temperature °C
Figure 5. P-T estimates of pluton emplacement and contact metamorphism of host metavolcanic units,
based on hornblende-plagioclase thermobarometry for Cinko Lake area of Yosemite National Park,
based on the calibrations of Anderson and Smith (1995) and Holland and Blundy (1994). (Modified from
Anderson et al. 2007).
10 10
A granite solidus B
Fourteen14
Mojave Plutons
Mojave Plutons granite solidus
granite solidus Mount Stuart Batholith
8 tonalite solidus 8
tonalite solidus H olland & Blu nd y rxn B
H olland & Blundy rxn A
Pressure (kb)
Pressure (kb)
6 6
4 4
2 2
0 0
600 650 700 750 800 850 600 650 700 750 800 850
Temperature °C Temperature °C
Figure 6. Results of hornblende-plagioclase thermobarometry for A) fourteen deep- to shallow-emplaced
plutons exposed in the mountains of the of the Mojave Desert, California (modified from Anderson (1996;
symbols represent different plutons); and B) the Mount Stuart batholith (Anderson et al. 2008). Estimates
of pressure based on the Al-in-hornblende calibration of Anderson and Smith (1995) with temperature
from the hornblende-plagioclase calibration (reaction B) of Holland and Blundy (1994). For the Mount
Stuart, results of two calibrations of the Holland and Blundy (1994) hornblende-plagioclase thermometer
are depicted.
thermobarometry for peraluminous granites
The low molar volume of garnet introduces a significant pressure dependence to garnet-
based mineral equilibria in metamorphic and igneous systems. The garnet-hornblende-
plagioclase equilibrium, initially calibrated by Kohn and Spear (1990), has been revised
by Dale et al. (2000). Holdaway (2001) likewise offered a revision of the GASP (garnet-Al
silicate-silica-plagioclase) geobarometer in concert with an earlier revision of the garnet-biotite
thermometer (Holdaway 2000). Wu et al. (2004a) presented an empirical-based barometer for
garnet-biotite-plagioclase-quartz assemblages based on Holdaway’s (2000, 2001) revisions. It
offers a new avenue of pluton barometry in garnet, two-mica granites in comparison to prior
Thermometers & Thermobarometers in Granitic Systems 133
empirical calibrations by Ghent and Stout (1981) and Hodges and Crowley (1985) for the
assemblage garnet-biotite-plagioclase-muscovite.
The basic premise of these barometric reactions is that with increasing pressure, garnet
becomes more enriched in the grossular component at the expense the An (anorthite) component
of plagioclase. Regardless of calibration, petrographic evidence for early vs. late crystallization
is of first order importance. Figure 7 depicts P-T solutions for core to rim compositions of a
garnet, two mica granite using the Ghent and Stout (1981) and Ganguly and Saxena (1984)
calibrations. As a reflection of mineral zoning, the data reveal lower P-T solutions from core
to rim and, thus, are seen as indicative of crystallization during ascent. Some rims have also
yielded subsolidus results, which is a further indication of retrogression that is reported above
for this batholith.
Many peraluminous, garnet-bearing plutons also contain muscovite with a significant
celadonite component (K(Mg,Fe2+)(Fe3+,Al)[Si4O10](OH)2, but with important substitutions
such as that for Ti, F, and Cl). The pressure-dependent solution toward the celadonite end
member (Velde 1965) also provides useful pressure information. Wei and Powell (2004,
2006) have offered new thermodynamic datasets to evaluate pressure for muscovite-bearing
assemblages. Their work should be evaluated for pluton barometry along with that of Berman
et al (2007) on constraints on the stability of Al end members of biotite, but such is beyond the
scope of this contribution.
Oxybarometers and other mineral reactions involving oxygen
In contrast to the other intensive parameters (T, P, fH2O) that affect mineral crystallization
and composition in granitic magmas, oxygen fugacity stands out due to its remarkable ability
to range several tens of order of magnitude even in crustal-derived magmas. Such is not only
reflected in the relative abundance of magnetite (high fO2) versus ilmenite (low fO2), but also in
the modal percentages of mafic minerals and their Fe/Mg composition.
10
Kse91-2a (Gar-Bio-Pl-Mu)
8
6 cores
GASP
P (kbar)
6.4 ±0.9 kb
"Contact"
@ 738 ±41 (°C)
rims
4
4.6 ±1.3 kb
@ 672 ±26 (°C)
2 Wet tonalite
0
600 650 700 750 800 850
Temperature (°C)
Figure 7. Estimates of emplacement conditions of a garnet, two-mica granodiorite, Mt. Stuart batholith,
based on garnet-plagioclase-muscovite-biotite equilibria, using calibrations of Ghent and Stout (1981)
and Ganguly and Saxena (1984). The box labeled “GASP ‘Contact’” refers to estimate of the P-T of the
contact aureole based on garnet-aluminosilicate- silica-plagioclase equilibria. Modified from Paterson et
al. (1994).
134 Anderson, Barth, Wooden & Mazdab
Fe-ti oxides, pyroxenes, olivine, and quartz (quILF) and sphene (titanite)
The computational program and related data set termed “qUILF” for the reaction quartz
+ ulvospinel = ilmenite + fayalite (Andersen et al. 1993) continues to be utilized, e.g., Elliott
et al. (1998), Percival and Mortensen (2002), Ren et al. (2006), and Rodriguez et al. (2007).
It offers a range of T-fO2 estimates for oxide-, pyroxene-, and/or olivine-bearing assemblages,
even where not all included minerals are present, including olivine.
xirouchakis et al. (2001a) expanded qUILF to include sphene and in doing so
optimized standard-state properties for end member sphene (P21/a structure) consistent
with thermodynamic data in the qUILF program. Their work confirmed the Wones’ (1989)
calibration of the assemblage sphene + magnetite + quartz and his prediction that the assemblage
is a fundamental boundary between relatively reducing and oxidizing conditions. It is also
considered to improve the precision of calculations derived through qUILF and, in addition
to its use to determine T and fO2, it potentially can be used to compute the activities of silica
and titania with respect to quartz and rutile in other thermometers (see above section) where
these phases are lacking. Manon et al. (2008) have provided new entropy data for sphene,
which should provide means to further improve this application. The current version of qUILF
(qUILF95) can be downloaded from http://www.geosciences.stonybrook.edu/people/faculty/
lindsley/lindsley.html.
xirouchakis et al. (2001b) explored the use of sphene-based qUILF in several magmatic
systems including an ilmenite-series granitic complex in Japan, the Skaergaard Intrusion of
Greenland, and the Fish Canyon Tuff of Colorado. More recent applications include that of
Ryabchikov and Kogarko (2006) on an alkaline complex in the Kola Peninsula of Russia.
Figure 8 shows minimum T-fO2 estimates for the ilmenite-series Mt. Stuart batholith
(Anderson et al. 2008) of Washington using the qUILF program indicating crystallization
conditions near the quartz-fayalite-magnetite buffer. However, the more reduced determinations
derive from more silica-rich portions of the batholith and are considered to reflect subsolidus
-10
HM
SiO 2 = 53 - 54 w t. %
-15
SiO 2 = 59 - 66 w t. %
log fO 2
QFM
-20
SiO 2 = 64 - 69 w t. %
-25
600 700 800 900
Temperature °C
Figure 8. Minimum T-fO2, based on qUILF estimates of ilmenite solid solution stability, for the Mt. Stuart
batholith at temperatures obtained from thermometry. Each line, some of which are dashed for clarity,
represents individual samples. Low fO2 results for more silica rich rocks of the batholith, which mostly
occur near the pluton margin, are viewed as the result of reequilibration with high-T fluids derived from the
contact aureole (modified from Anderson et al. 2008).
Thermometers & Thermobarometers in Granitic Systems 135
infiltration of high-T fluids from the graphite-rich units of the contact aureole, which, as
indicated above, also lowered Al-in-hornblende values.
Magnetite and ilmenite series of granitic plutons and oxygen fugacity
Ishihara’s (1977) recognition of magnetite- and ilmenite-series granites was another
milestone that in its own right prompted renewed attention to the redox equilibria expressed
in the mineralogy of granitic batholiths, particularly that of the Fe-Ti oxides. It followed the
classic contributions of Buddington and Lindsley (1964) on Fe-Ti oxide oxybarometer and
that of Wones and Eugster (1965) on biotite-alkali feldspar-magnetite equilibria, which was
further advanced by the studies of Wones (1981, 1989), Carmichael and Ghiorso (1990), Frost
(1991), Ghiorso and Sack (1991), and Andersen et al. (1993).
Frost et al. (2001) concludes the magnetite/ilmenite-series classification to be particularly
applicable to continental margin-related batholiths. Ishihara’s (1977) proposal was that
ilmenite-series granitoids were derived and crystallized under oxygen reducing conditions in
contrast to those of magnetite-series. Gastil et al. (1990) used the basis of the classification
to divide the Peninsular Ranges batholith of southern and Baja California with more reduced
(ilmenite-series) granitic assemblages occurring on the continental side of the arc. In contrast,
Bateman et al. (1991) determined that the Sierra Nevada batholith, while mostly magnetite
series, has western margin plutons that are more reduced. Such implies that batholithic arcs
can have significant and variable portions that were derived from more oxidized versus more
reduced source materials which provide important evidence regarding the prevailing fO2 of
melt origin.
The utilization of this classification remains essential as evidenced by contributions of
Dall’Agnol et al. (1999), Sial et al (1999), Ishihara et al. (2000), Takagi (2004), Hart et al.
(2004), Anderson and Morrison (2005), Bogaerts et al. (2006), Bucher and Frost (2006), and
Dall’Agnol and Oliveria (2007).
It is the common experience that Fe-Ti oxides in plutonic rocks do not retain magmatic
compositions, but exceptions occur (Anderson and Thomas 1985; Anderson 1996) and the
qUILF program indicated above offers solutions to T-fO2 determinations. However, the biotite,
K-feldspar, and magnetite oxybarometer, although now dated (Wones 1981), is more robust
given that magmatic compositions of K-feldspar and magnetite can be calculated based of
compositions on plagioclase for K-feldspar and ilmenite for magnetite, with the assumption
that these other phases have more likely retained magmatic compositions.
The calibration of Wones (1981) is
4819 Pbar − 1
logfO 2 ( bar ) = −2 + 6.69 − logfH 2 O − 0.011 + log aAnnite − log aSan − logaMt ] (11)
TK T K
It is based on the equilibrium (termed ASM):
annite + 1/2 O2 = sanidine + magnetite + H2O (12)
which leads to a continuous reaction where, with increasing fO2, biotite becomes more Mg-rich
at the expense of components released to K-feldspar and magnetite.
Figure 9 depicts biotite compositions of Mesoproterozoic granites of the former Laurentia
supercontinent (data from Anderson 1978; Anderson and Thomas 1985; Anderson and Bender
1989). This remarkable magmatic epoch, considered to be anorogenic, has been subdivided
into transcontinental provinces of ilmenite-series granites (Baltic regions to Wyoming) and
magnetite-series granites (mid-continent to SW USA), the latter of which also includes a
peraluminous subprovince ranging from Colorado to central Arizona (Anderson and Morrison
2005). Whole-rock compositions of these A-type granites are elevated in K2O and other
136 Anderson, Barth, Wooden & Mazdab
Biotite + K-feldspar +/- Magnetite
Fe/ (Fe+Mg) ∼∆ QFM
1.0
Pikes Peak
Wiborg Wolf River
-1
Ilmenite -
series
0.8
Sorvik
Marble Montello
0
Revsund Parker, St. Vrain
Newberry Davis Dam
0.6 Ak Chin +1
Bowmans Silver Magnetite -
Wash Gold Plume
Butte Ruin series
Sierra Estrella
Ft. Huachuca Oracle +2
Holy
0.4 Moses
(two mica granites)
(biotite + hornblende)
Hualapai +3
0.2
Laurentia Mesoproterozoic Granites
0.0
2.0 2.4 2.8 3.2 3.6 4.0
Aliv + Alvi (22 oxygens)
Figure 9. Biotite compositions of Mesoproterozoic ilmenite-series and magnetite-series granites (data from
Anderson 1980; Anderson and Thomas 1985; Anderson and Bender 1999) and approximate fO2 relative to
the quartz-fayalite-magnetite (qFM) buffer (assuming PH2O = Ptotal) based on the calibration of Wones
(1981).
incompatible elements and have high Fe/Mg ratios. As noted by Anderson and Morrison
(2005), this generalization is particularly true for the Mesoproterozoic granites in the ilmenite-
series province; however, these compositional parameters range lower for these granites in the
magnetite-series province, particularly so for those peraluminous, presumably due to higher
fO2 and fH2O of the crustal source and related increased levels of partial melting.
The biotites from these Mesoproterozoic plutons show well the control of ASM equilibria.
Of these plutons, the magnetite-bearing granites have biotite with Fe/(Fe+Mg) ratios which
show no relation to whole-rock composition and instead reflect intensive parameter control
(T, fO2, fH2O). For example, the 1.4 Ga Holy Moses and Hualapai granites of western Arizona
have typical high whole-rock FeO/(FeO+MgO) wt% ratios averaging at or above 0.85, yet
their biotite Fe/(Fe+Mg) ratios average 0.45 and 0.30, respectively, and yield T-fO2 estimates,
based on ASM equilbria, at 2 to 3 log units above the qFM oxygen buffer. These two plutons
also are characterized by striking aeromagnetic anomalies due to their high magnetite content
(Anderson and Bender 1989). Hence, we suggest that oxygen fugacity is the principal factor
controlling the Fe/Mg composition of biotite and other mafic phases in magnetite-series
granites, as evidenced by calculated log units of fO2 (calibration of Wones 1981) in Figure 9,
which spans over four levels of magnitude.
For ilmenite-series granites, the ASM equilibria are incomplete (as it is a non-limiting
assemblage) where magnetite is absent; hence biotite Fe/Mg compositions lack intensive
Thermometers & Thermobarometers in Granitic Systems 137
parameter control and instead follow that of rock composition. Figure 10 depicts this for
biotites from the ilmenite-series Wolf River batholith (Anderson 1980), which compares Fe/
Mg biotite composition with that of rock composition and mineral assemblage in contrast
to the few samples of the batholith that have magnetite. Where magnetite is absent in this
batholith, biotite Fe/(Fe+Mg) ratios correlate well with that of whole-rock FeO/(FeO+MgO),
with appropriate shifts with mineral assemblage, including presence or absence of fayalitic
olivine, pyroxene, hornblende, and sphene. For the samples that contain magnetite, biotite Fe/
(Fe+Mg) ratios generally depart from any such correlation, although they remain high and,
thus, indicative of the oxygen-reduced nature of this ilmenite-series batholith.
SuMMary
The ability to determine the thermal and barometric history during crystallization and
emplacement of granitic plutons has been enhanced by several new calibrations applicable to
granitic mineral assemblages. Other existing calibrations for granitic plutons have continued
to be popular and fairly robust. Recent advances include the trace element thermometers Ti-
in-quartz, Ti-in-zircon, and zr-in-sphene (titanite), which need to be further evaluated on the
roles of reduced activities due to lack of a saturating phase, the effect of pressure dependence
(particularly for the Ti-in-zircon thermometer), and how resistive these thermometers are to
subsolidus reequilibration. As zircon and sphene are also hosts to radiogenic isotopes, these
minerals potentially also provide new insights into the temperature - time history of magmas.
When used in conjunction with pressure-sensitive mineral equilibria in the same rocks, a
complete assessment of the P-T-t (pressure-temperature-time) path is possible given that the
mineralogy of plutons can reflect crystallization over a range of pressure and temperature
during ascent and emplacement and that many intrusions are now seen as forming over several
millions of years during the protracted history of batholith construction. Accessory mineral
saturation thermometers, such as those for zircon, apatite, and allanite, provide a different
1.00
Ilmenite-Series Wolf River Batholith
FeO/(FeO +MgO) - Whole Rock
0.95
Mt-bearing
0.90
h
Sp
+
+O
l Hb
b +
+H o
0.85 Bio Bi b Bio
H
o+
Bi
+ ilmenite, Ksp, Plag, Qz
0.80
0.75
0.60 0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00
Fe/(Fe +Mg) - Biotite
Figure 10. Biotite composition
Figure (Fe/(Fe+Mg)
10. Biotite composition - atomic basis)
(Fe/(Fe+Mg) from
- atomic the from
basis) ilmenite-series Wolf River batholith
the ilmenite-series
as a function of mineral
Wolf River assemblage
batholith as aand whole
function of rock (FeO/(FeO+MgO),
mineral wt%)
assemblage and whole in comparison to samples
rock
(FeO/(FeO+MgO),
that contain magnetite (modifiedwt. %)Anderson
from in comparison to samples that contain magnetite (modified
1980).
from Anderson, 1980).
138 Anderson, Barth, Wooden & Mazdab
and powerful perspective, specifically that of the temperature of the onset of crystallization of
these minerals, which can allow an estimate of the range of temperature between the liquidus
and solidus of a given pluton.
In assessment of the depth of crystallization and emplacement of granitic plutons, the Al-
in-hornblende remains popular for metaluminous granites when appropriately corrected for
temperature. For peraluminous granites, potential new calibrations exist for the assemblages
bearing garnet, biotite, plagioclase, muscovite, and quartz.
Other thermometers, based on oxygen abundance, and including Fe-Ti oxides, pyroxene,
fayalitic olivine, quartz, sphene, and/or biotite, some of which have been recently revised,
can provide additional information on temperature and oxygen fugacity. Oxygen fugacity can
range over several orders of magnitude in different magmatic systems and can have profound
influence on the mineralogy and mineral compositions in granitic magmas. It also forms the
foundation of the popular magnetite- versus ilmenite-series granite classification.
aCKnOWLedGMent
The authors thank the very helpful reviews by Bernard Evans, Calvin Miller, Jennifer
Matzel, Keith Putirka, and Eric Essene who all made insightful comments to this contribution.
The senior author also wishes to thank Scott Paterson and our students, Rita Economos, Vali
Memeti, Claire Coyne, and Tao zhang for continuing to offer advice on issues of granite
thermobarometry.
reFerenCeS
Ague JJ (1997) Thermodynamic calculation of emplacement pressures for batholithic rocks, California:
implications for the aluminum-in-hornblende barometer. Geology 35:563-566
Andersen DJ, Lindsley DH, Davidson PM (1993) qUILF: A Pascal program to assess equilibria among Fe-
Mg-Mn-Ti oxides, pyroxenes, olivine, and quartz. Computers and Geosciences 19:1333-1350
Anderson JL (1980) Mineral equilibria and crystallization conditions in the Late Precambrian Wolf River
rapakivi massif, Wisconsin. Am J Sci 280:389-332
Anderson JL (1996) Status of thermobarometry in granitic batholiths. Trans R Soc Edinburgh 87:125-138
Anderson JL (1997) Regional tilt of the Mount Stuart batholith, Washington, determined using aluminum-in-
hornblende barometry, implications for northward translation of Baja British Columbia: Discussion. Geol
Soc Am Bull 109:1223-1225
Anderson JL, Barth AP, Young ED, Davis MJ, Farber D, Hayes EM, Johnson KA (1992). Plutonism across the
Tujunga-North American terrane boundary: A middle to upper crustal view of two juxtaposed arcs. In:
Characterization and Comparison of Ancient and Mesozoic Continental Margins - Proceedings of the 8th
International Conference on Basement Tectonics. Bartholomew MJ, Hyndman DW, Mogk DW, Mason R
(eds) Kluwer Academic Publishers, Dordrecht, p 205-230
Anderson JL, Barth, AP, Young ED (1988) Mid-crustal Cretaceous roots of Cordilleran core complexes.
Geology 16:366-369
Anderson JL, Bender EE (1989) Nature and origin of Proterozoic A-type granite magmatism in the southwestern
United States. Lithos 23:19-52
Anderson JL, Cullers RL (1978) Geochemistry and evolution of the Wolf River Batholith, a late Precambrian
rapakivi massif in North Wisconsin, U.S.A. Precambrian Res 7: 287-324
Anderson JL, Foley BJ, Ball EN, Paterson SR, Memeti V, Pignotta GS (2007) Upper crustal overturn during
magmatic surges – a potential Sierra-wide process. Geol Soc Am Abstr Programs 39(6):526 Paper #196-
6
Anderson JL, Morrison J (2005) Ilmenite, magnetite, and peraluminous Mesoproterozoic anorogenic granites
of Laurentia and Baltica. Lithos 80:45-60
Anderson JL, Morrison J, Paterson S, Francis J (2008) Post-emplacement fluids and pluton thermobarometry:
Mt. Stuart batholith, Washington Cascades. J Metamorph Geol (in revision)
Anderson JL, Smith DR (1995) The effect of temperature and oxygen fugacity on Al-in-hornblende barometry.
Am Mineral 80:549-559
Thermometers & Thermobarometers in Granitic Systems 139
Anderson JL, Thomas WM (1985) Proterozoic anorogenic two-mica granites: the Silver Plume and the St.
Vrain Batholiths of Colorado: Geology 13:177-180
Asrat A, Barbey P, Ludden N, Reisberg L, Gleies G, Ayalew D (2004) Petrology and isotope geochemistry of the
Pan-African Negash pluton, northern Ethiopia: mafic-felsic magma interactions during the construction
of shallow-level calc-alkaline plutons. J Petrol 45:1147-1179
Bachl CA, Miller CF, Miller JS, Faulds JE (2001) Construction of a pluton: Evidence from an exposed cross
section of the Searchlight pluton, Eldorado Mountains, Nevada. Geol Soc Am Bull 113:1213-1228
Bachmann O, Dungan MA (2002) Temperature-induced Al-zoning in hornblendes of the Fish Canyon magma,
Colorado. Am Mineral 87:1062-1076
Bachmann O, Dungan MA, Lipman PW (2002) The Fish Canyon magma body, San Juan volcanic field,
Colorado: Rejuvenation and eruption of an upper crustal batholith. J Petrol 43:1469-1503
Barnes CG, Burton BR, Burling TG, Wright JE, Karlsson HR (2001) Petrology and geochemistry of the late
Eocene Harrison Pass pluton, Ruby Mountains core complex, northeastern Nevada. J Petrol 42:901-929
Barth AP (1990) Mid-crustal emplacement of Mesozoic plutons, San Gabriel Mountains, California, and
implications for the geologic history of the San Gabriel Terrane. In: The Nature and Origin of Cordilleran
Magmatism. Anderson JL (ed) GSA Memoir 174:33-45
Barth AP, Anderson JL, Jacobsen CE, Paterson SR, Wooden JL (2008) Magmatism and tectonics in a tilted
crustal section through a continental arc, eastern Transverse Ranges and southern Mojave Desert. Geol
Soc Am Field Guide 11:1-18
Bateman PC, Dodge FCW, Kistler RW (1991) Magnetic susceptibility and relation to initial 87Sr/86Sr for
granitoids of the central Sierra Nevada, California. J Geophys Res 96:19155-19568
Berman RG, Aranovich LY, Rancourt DG, Mercier PHJ (2007) Reversed phase equilibrium constraints on the
stability of Mg-Fe-Al biotite. Am Mineral 92:139-150
Blundy J, Cashman K (2008) Petrologic reconstruction of magmatic system variables and processes. Rev
Mineral Geochem 69:179-239
Bogaerts M, Scaillet B, Auwera JV (2006) Phase equilibria of the Lyngdal granodiorite (Norway): Implications
for the origin of the metaluminous ferroan granitoids. J Petrol 47:2405-2431
Bucher K, Frost BR (2006) Fluid transfer in high-grade metamorphic terrains intruded by anorogenic granites:
The Thor Range, Antarctica. J Petrol 47:567-593
Buddington AL, Lindsley DH (1964) Iron-titanium oxide minerals and synthetic equivalents: J Petrol 5:310-
357
Carmichael ISE, Ghiorso MS (1990) The effect of oxygen fugacity on the redox state of natural liquids and
their crystallizing phases. Rev Mineral Geochem 24:191-212
Claiborne LE, Miller CF, Walker BA, Wooden JL, Mazdab FK, Bea F (2006) Tracking magmatic processes
through zr/Hf ratios in rocks and Hf and Ti zoning in zircons: An example from the Spirit Mountain
batholith, Nevada. Mineral Mag 70:517-543
Dale J, Holland T, Powell R (2000) Hornblende-garnet-plagioclase thermobarometry: a natural assemblage
calibration on the thermodynamics of hornblende. Contrib Mineral Petrol 140:353-362
Dall’Agnol R, Oliveira DC (2007) Oxidized, magnetite-series, rapakivi granites of Carajas, Brazil: Implications
for classification and petrogenesis of A-type granites. Lithos 93:215-233
Dall’Agnol R, Ramo OT, Magalhaes MS, Macambira MJB (1999) Petrology of the anorogenic, oxidised Jamon
and Musa granites, Amazonian Craton: implications for the genesis of Proterozoic A-type granites. Lithos
46:431-462
Elliott BA, Ramo OT, Nironen M (1998) Mineral chemistry constraints on the evolution of the 1.88-1.87 Ga
post-kinematic granite plutons in the Central Finland Granitoid Complex. Lithos 45:109-129
Ernst WG, Liu J (1998) Experimental phase-equlibrium study of Al- and Ti- contents of calcic amphibole in
MORB - a semiquantitative thermobarometer. Am Mineral 83:952-969
Essene EJ (1982) Geologic thermometry and barometry. Rev Mineral 10:153-206
Essene EJ (1989) The current status of thermobarometry in metamorphic rocks. In: Evolution of Metamorphic
Belts. Daly JS, Cliff RA, Yardley BWD (eds) Geological Society of London, p 1-44
Ferriss EDA, Essene EJ, Becker U (2008) Computational study of the effect of pressure on the Ti-in-zircon
geothermometer. Eur J Mineral (Werner Schreyer Special Issue), (in press)
Ferry JM, Watson EB (2007) New thermodynamic models and revised calibrations for the Ti-in-zircon and
zr-in-rutile thermometers. Contrib Mineral Petrol 154:429-437
Foley BJ, Ball EN, Fischer GC, Thompson JM, Memeti V, Pignotta GS, Anderson JL, Paterson SR, Matzel
J, Mundil R (2007) Downward ductile displacement of volcanic crust during pluton emplacement in
the central Sierra Nevada: Undergraduate Team Research at USC. GSA Cordilleran Section meeting,
Bellingham Paper #31-4
Frost BR (1991) Introduction to oxygen fugacity and its petrologic importance. Rev Mineral 25:1-9
Frost RB, Barnes CG, Collins WJ, Arculus RJ, Ellis DJ, Frost CR (2001) A geochemical classification of
granitic rocks. J Petrol 42:2033-2048
140 Anderson, Barth, Wooden & Mazdab
Fu B, Page z, Cavosie AJ, Fournelle J, Kita NT, Lackey JS, Wilde SA, Valley JW (2008) Ti-in-zircon
thermometry: applications and limitations. Contrib Mineral Petrol 156:197-215
Ganguly J, Saxena S (1984) Mixing properties of aluminosilicate garnets: constraints from natural and
experimental data, and applications to geothermo-barometry. Am Mineral 69:88-97
Gastil RG, Diamond J, Knaack C, Walawendar M, Marshall M, Boyles C, Chadwick B, Erskine B (1990) The
problem of the magnetite-ilmenite boundary in southern and Baja California. In: The Nature and Origin
of Cordilleran Magmatism. Anderson JL (ed) Geol Soc Am Memoir 174:19-32
Ghent ED, Stout Mz (1981) Geobarometry and geothermometry of plagioclase-biotite-garnet-muscovite
assemblages. Contrib Mineral Petrol 76:92-97
Ghiorso MS, Sack RO (1991) Fe-Ti oxide thermometry: thermodynamic formulation and estimation of
intensive variables in silicic magmas. Contrib Mineral Petrol 108:485-510
Green TH, Adam J (2002) Pressure effect on Ti- or P-rich accessory mineral saturation in evolved granitic
melts with differing K2O/Na2O ratios. Lithos 61:271-282
Hammarstrom JM, zen E (1986) Aluminum in hornblende, an empirical igneous geobarometer: Am Mineral
71:1297-1313
Harrison TM, Watson EB (1984) The behavior of apatite during crustal anatexis: equilibrium and kinetic
considerations: Geochim Cosmochim Acta 48:1467–1477
Harrison TM, Watson EB, Aikman AB (2007) Temperature spectra of zircon crystallization in plutonic rocks.
Geology 35:635-638
Hart CJR, Goldfarb RJ, Lewis LL, Mair JL (2004) The northern Cordilleran mid-Cretaceous plutonic province:
Ilmenite/magnetite-series granitoids and intrusion-related mineralization. Resour Geol 54:253-280
Hayden LA, Watson EB (2007) Rutile saturation in hydrous siliceous melts and its bearing on Ti-thermometry
of quartz and zircon. Earth Planet Sci Lett 258:561-568
Hayden LA, Watson EB, Wark DA (2007) A thermobarometer for sphene. Contrib Mineral Petrol Online 12 p.
http://dx.doi.org/10.1007/s00410-007-0256-y
Hildreth W (1981) Gradients in silicic magma chambers: implications for lithospheric magmatism. J Geophys
Res 86:10153-10192
Hodges KV, Crowley PD (1985) Error estimation and empirical geothermobarometry for pelitic systems. Am
Mineral 70:702-709
Hoisch TD (1991) Equilibria with the mineral assemblage quartz = muscovite + biotite + garnet + plagioclase.
Contrib Mineral Petrol 108:43-54
Holdaway MJ (2000) Application of new experimental and garnet Margules data to the garnet-biotite
geothermometer. Am Mineral 85:881-892
Holdaway MJ (2001) Recalibration of the GASP geobarometer in light of recent garnet and plagioclase models
and versions of the garnet-biotite geothermometer. Am Mineral 86:1117-1129
Holland T, Blundy J (1994) Non-ideal interactions in calcic amphiboles and their bearing on amphibole-
plagioclase thermometry. Contrib Mineral Petrol 116:433-447
Hollister LS, Grissom GC, Peters EK, Stowell HH, Sisson VB (1987) Confirmation of the empirical correlation
of Al in hornblende with pressure of solidification of calc-alkaline plutons. Am Mineral 72:231-239
Hoskin PWO, Kinney PD, Wyborn D, Chappell BW (2000) Identifying accessory mineral saturation during
differentiation in granitoid magmas: an integrated approach. J Petrol 41:1365-1396
Hoskin PWO, Schaltegger U (2003) The composition of zircon and igneous and metamorphic petrogenesis.
Rev Mineral Geochem 53:27-62
Ishihara S (1977) The magnetite-series and ilmenite-series granitic rocks. Mining Geol 27:293-305
Ishihara S, Hashimoto M, Machida M (2000) Magnetite/ilmenite-series classification and magnetic
susceptibility of the Mesozoic-Cenozoic batholiths of Peru. Resour Geol 50:123-129
Janousek V, Braithwaite CJR, Bowes DR, Gerdes A (2004) Magma mixing in the genesis of Hercynian calc-
alkaline granitoids: an integrated petographic and geochemical study of the Sazava intrusion, Central
Bohemian Pluton, Czech Republic. Lithos 78:67-99
Johnson MC, Rutherford M J (1989) Experimental calibration of an aluminum-in-hornblende geobarometer
with application to Long Valley caldera (California) volcanic rocks. Geology 17:837-841
Klimm K, Blundy JD, Green TH (2008) Trace element partitioning and accessory phase saturation during
H2O-saturated melting of basalt with implications for subduction zone chemical fluxes. J Petrol 49:523-
533
Kohn MJ, Spear FS (1990) Two new geobarometers for garnet amphibolites, with applications to southeastern
Vermont. Am Mineral 75:89-96
Lindsay JM, Schmidtt AK, Trumbull RB, de Silva SL, Siebel W, Emmermann R (2001) Magmatic evolution
of the La Pacana caldera system, central Andes, Chile: Compositional variation of two cogenetic, large
volume felsic ignimbrites. J Petrol 42:459-486
Lopez-Moro FJ, Lopez-Plaza M (2004) Monzonitic series from the Variscan Tormes dome (Central Iberian
zone): petrogenetic evolution from monzogabbro to granitic magmas. Lithos 72:19-44
Thermometers & Thermobarometers in Granitic Systems 141
Manley CR, Bacon CR (2000) Rhyolite thermobarometry and the shallowing of the magmatic reservoir, Coso
volcanic field, California. J Petrol 41:149-174
Manon MR, Essene EJ, Dachs E (2008) Low T heat capacity measurements and new entropy data for titanite:
implications for thermobarometry of high pressure rocks. Contrib Mineral Petrol, in press. (UPDATE)
Matzel J, Bowring SA, Miller RB (2006) Timescales of pluton construction at differing crustal levels: examples
from the Mount Stuart batholith and Ten Peak pluton, north Cascades, WA. Geol Soc Am Bull 118:1412-
1430
Mazdab F, Wooden JL, Barth AP (2007) Trace element variability in titanite from diverse geologic environments:
GSA Abstr Programs 39, no. 6, p. 406
Miller CF, McDowell SM, Mapes RW (2003) Hot and cold granites? Implications of zircon saturation
temperatures and preservation of inheritance. Geology 31:529-532
Miller CF, Miller JS (2002) Contrasting stratified plutons exposed in tilt blocks, El Dorado Mountains,
Colorado River rift, NV, USA. Lithos 61:209-224
Miller JS, Matzel JEP, Miller CF, Burgess SD, Miller RB (2007) zircon growth and recycling during the
assembly of large, composite plutons. J Volcanol Geotherm Res 167:282-299
Needy SK, Anderson JL, Wooden JL, Barth AP, Paterson SR, Memeti V, Pignotta GS (2008), Mesozoic
magmatism in an upper- to middle-crustal section through the Cordilleran continental margin arc, eastern
Transverse Ranges, California. In: Crustal cross-sections from the western North America Cordillera and
elsewhere: Implications for tectonic and petrologic processes. Miller RB, Snoke AW (eds) Geol Soc Am
Special Paper (accepted pending revisions)
Page F, zeb F, Bin K, Noriko T, Fournelle J, Spicuzza, MJ, Schulze DJ, Viljoen F, Basei MAS; Valley JW
(2007) zircons from kimberlite; new insights from oxygen isotopes, trace elements, and Ti in zircon
thermometry. Geochim Cosmochim Acta 71:3887-3903
Paterson SR, Memeti V, Anderson JL, Miller R, zak J, Jacobs R, Seyum S, Shimono S, Wenrong C (2008)
Transpression and downward crustal flow during the cretaceous high flux magmatic event in the Central
Sierra Nevada, California: Abstract, GSA Cordilleran Section meeting, Las Vegas, Paper #34-7
Paterson SR, Miller RB, Anderson JL, Lund S, Bendixen J, Taylor N, Fink T (1994) Emplacement and evolution
of the Mt. Stuart batholith. In: Geologic Field Trips in the Pacific Northwest Swanson DA, Haugerud RA
(eds) Department of Geological Sciences, University of Washington in conjunction with the Geological
Society of America, Seattle, 2:2F1-2F27
Percival JA, Mortensen JK (2002) Water deficient calc-alkaline plutonic rocks of northeastern Superior
Province, Canada: significance of charnockitic magmatism: J Petrol 43:1617-1650
Piwinskii AJ (1968) Experimental studies of rock series, central Sierra Nevada batholith, California: J Geol
76:548-570
Putirka KD (2008) Thermometers and barometers for volcanic systems. Rev Mineral Geochem 69:61-120
Ren M, Omenda PA, Anthony EY, White JC, Macdonald R, Bailey DK (2006) Application of the qUILF
thermbarometer to the peralkaline trachytes and pantellerites of the Eburru volcanic complex, East Africa
Rift, Kenya. Lithos 9:109-124
Ribbe PH (1982) Titanite (Sphene). Rev Mineral 5:137-154
Rodriguez C, Selles D, Dungan M, Langmuir C, Leemanm W (2007) Adakitic dacites formed by intracrustal
fractionation of water-rich parent magmas at Nevado de Longavi’ Volcano (36.2°S) Andean Southern
Volcanic zone, Central Chile. J Petrol 48:2033-2061
Ryabchikov ID, Kogarko LN (2006) Magnetite compositions and oxygen fugacities of the Khibina magmatic
system. Lithos 9:35-45
Schmidt MW (1992) Amphibole composition in tonalite as a function of pressure: an experimental calibration
of the Al-in-hornblende barometer. Contrib MineralPetrol 110:304-310
Sial AN, Toselli AJ, Saavedra J, Parada MA, Ferreira VP (1999) Emplacement, petrological and magnetic
susceptibility characteristics of diverse magmatic epidote-bearing granitoid rocks in Brazil, Argentina,
and Chile. Lithos 46:367-392
Speer JA (1982) zircon. Rev Mineral 5:67-112
Stein E, Dietl C (2001) Hornblende thermobarometry of granitoids from the Central Odenwald (Germany) and
their implications for the geotectonic development of the Odenwald. Mineral Petrol 72:185-207
Stern CR, Huang W, Wyllie PJ (1975) Basalt-andesite-rhyolite-H2O: Crystallization intervals with excess
H2O and H2O-undersaturated liquidus surface to 35 kilobars with implications for magma genesis. Earth
Planet Sci Lett 28:189-196
Stone D (2000) Temperature and pressure variations in suites of Archean felsic plutonic rocks, Berens River
area, northwest Superior Province, Ontario, Canada. Can Mineral 38:455-470
Takagi T (2004) Origin of magnetite- and ilmenite-series granitic rocks in the Japan arc. Am J Sci 304:169-
202
Tollari N, Toplis MJ, Barnes S-J (2006) Predicting phosphate saturation in silicic magmas: An experimental
study of the effects of melt composition and temperature. Geochim Cosmochim Acta 70:1518-1536
142 Anderson, Barth, Wooden & Mazdab
Velde B (1965) Phengite micas; synthesis, stability, and natural occurrence. Am J Sci 263:886-913
Walker BA Jr, Miller, CF, Claiborne, LE, Wooden JL, Miller JS (2007) Geology and geochronology of the
Spirit Mountain batholith, southern Nevada: implications for timescales and physical processes of
batholith construction. J Volcanol Geotherm Res 167:239-262
Wark DA, Hildreth W, Spear FS, Cherniak DJ, Watson EB (2007) Pre-eruption recharge of the Bishop magma
system. Geology 35:235-238
Wark DA, Watson EB (2006) Titaniq: a titanium-in-quartz geothermometer. Contrib Mineral Petrol 152:743-
754
Watson EB, Harrison TM (1983) zircon saturation revisited: temperature and composition effects in a variety
of crustal magma types. Earth Planet Sci Lett 64:295-304
Watson EB, Liang Y (1995) A simple model for sector zoning in slowly grown crystals: Implications for
growth rate and lattice diffusion, with emphasis on accessory minerals in crustal rocks. Am Mineral
80:1179-1187
Watson EB, Wark DA, Thomas JB (2006) Crystallization thermometers for zircon and rutile. Contrib Mineral
Petrol 151:413-433
Wei C, Powell R (2004) Calculated phase relations in high pressure metapelites in the system NKFMASH
(Na2O-K2O-FeO-MgO-Al2O3-SiO2-H2O). J Petrol 45:183-202
Wei C, Powell R (2006) Calculated phase relations in the system NCKFMASH (Na2O-CaO-K2O-FeO-MgO-
Al2O3-SiO2-H2O) for high pressure metapelites. J Petrol 47:385-408
Wiebe RA, Wark DA, Hawkins DP (2007) Insights from quartz cathodoluminescence zoning into crystallization
of the Vinalhaven granite, coastal Maine. Contrib Mineral Petrol 154:439-453
Wones DR (1981) Mafic silicates as indicators of intensive parameters in granitic magmas. Mining Geol
31:191-212
Wones DR (1989) Significance of the assemblage titanite + magnetite + quartz in granitic rocks. Am Mineral
74:744-49
Wones DR, Eugster HP (1965) Stability of biotite: experiment, theory, and application. Am Mineral 50:1228-
1272
Wooden JL, Mazdab F, Claiborne LEL , Miller C.F, Barth AP (2006) Elemental analysis of zircon by High
Mass Resolution USGS-Stanford SHRIMP-RG: Measuring and evaluating Ti-in-zircon temperatures and
compositional characteristics. EOS 31:F-04
Wu CM, zhang J, Ren L-D (2004a) Empirical garnet-biotite-plagioclase-quartz (GBPq) geobarometry in
medium- to high-grade metapelites. J Petrol 45:1907-1921
Wu CM, zhang J, Ren L-D (2004a) Empirical garnet-muscovite-plagioclase-quartz geobarometry in medium-
to high-grade metapelites. Lithos 78:319-332
xirouchakis D, Lindsley DH, Frost BR (2001a) Assemblages with titanite (CaTiOSiO4), Ca-Fe-Mg olivine and
pyroxenes, Fe-Mg-Ti oxides, and quartz: Part I. Theory. Am Mineral 86:247-253
xirouchakis D, Lindsley DH, Frost BR (2001b) Assemblages with titanite (CaTiOSiO4), Ca-Fe-Mg olivine and
pyroxenes, Fe-Mg-Ti oxides, and quartz: Part II. Application. Am Mineral 86:254-264
zen E (1985) Implications of magmatic epidote-bearing plutons on crustal evolution in the accreted terranes of
northwestern North America. Geology 13:266-269
zen E (1989) Plumbing the depths of batholiths. Am J Sci 289:1137-1157
zen E, Hammarstrom JM (1984) Magmatic epidote and its petrologic significance. Geology 12:515-518
zhang SH, zhao Y, Song B (2006) Hornblende thermobarometry of the Carboniferous granitoids from the
Inner Mongolia paleo-uplift: implications for the tectonic evolution of the northern margin of the North
China block. Mineral Petrol 87:123-141
Reviews in Mineralogy & Geochemistry
Vol. 69 pp. 143-177, 2008 5
Copyright © Mineralogical Society of America
Fluid Inclusion thermobarometry as a
tracer for Magmatic Processes
thor h. hansteen
IFM-GEOMAR, Leibniz-Institute for Marine Sciences
Dynamics of the Ocean Floor
D-24148 Kiel, Germany
[email protected]
andreas Klügel
Fachbereich Geowissenschaften
Universität Bremen
D-28334 Bremen, Germany
[email protected]
IntrOduCtIOn
Fluid inclusions in minerals may form in any type of volcanic or plutonic rock ranging
from mafic to silicic compositions. Because all igneous rocks reach fluid saturation at some
stage during their evolution, fluids trapped as inclusions in magmatic minerals belong to a
certain paragenesis or phase assemblage, which may include minerals, melts and one or more
fluid phases. These fluid inclusions reflect one or more stages during rock evolution, and can
be used to constrain multistage formation and evolution processes including ascent histories,
magma chamber processes and crystallization behavior.
Fluid inclusions can provide thermobarometric data on various timescales. During ascent
of mafic to intermediate magmas, fluid inclusions may form within hours to days and record
transient magma stagnation levels, whereas chemical mineral-melt thermobarometry requires
equilibrium mineral growth and thus typically reflects well-defined crystallization events
(Roedder and Bodnar 1980; Wanamaker et al. 1990; Hansteen et al. 1998; Klügel et al. 2000;
Frezzotti and Peccerillo 2004). Such data has only rarely been combined with melt inclusion
investigations in order to depict detailed magma ascent histories (e.g., Bureau et al. 1998).
During prolonged crystallization in magma chambers or in the plutonic environment, however,
fluid inclusions reflect the equilibrium situation, and data from various thermobarometric
methods should overlap.
This chapter focuses on the use of fluid inclusions as thermobarometers to constrain
magmatic processes and timescales. After an introduction of basic principles and explanation
of the most relevant fluid systems we show in a “cookbook” style how barometric data are
derived, and discuss error magnitudes and pitfalls. Thermobarometric information attained
from chemical mineral-melt equilibria are compared to that obtained from fluid inclusions in
order to provide detailed accounts for magma ascent and crystallization. Although the term
“fluid inclusion” is often used in a general way comprising all inclusions that were trapped
in the liquid state (Roedder 1984), this chapter is restricted to CO2- and H2O-dominated
inclusions and does not include sulfide- or silicate melt inclusions, which are covered in Kent
(2008) and Metrich and Wallace (2008).
1529-6466/08/0069-0005$05.00 DOI: 10.2138/rmg.2008.69.5
144 Hansteen & Klügel
OCCurrenCe and PetrOGraPhy OF FLuId InCLuSIOnS
Every igneous mineral may contain fluid inclusions. Among pheno- and xenocrysts from
volcanic rocks, a majority of inclusion investigations have been performed on the host minerals
quartz, plagioclase, amphibole, pyroxene and olivine. In mantle xenoliths and in plutonic
rocks the occurrence of thousands of fluid inclusions within a regular thin section is not
uncommon (Roedder 1965). Inclusion sizes are one to three orders of magnitude smaller than
the mineral in which they occur, typically <2 to 30 micrometers, the smaller ones being much
more abundant than the larger ones. Depending on composition and density, fluid inclusions at
room temperature contain one or more phases (Fig. 1). Because the lower refractive indices of
most fluid inclusions compared to their host minerals cause internal reflections in the inclusion
cavity, many inclusions appear dark in ordinary plane-polarized light under the petrographic
microscope. Applying conoscopic light enhances the visibility of the inclusion contents, and is
thus invaluable for fluid inclusion petrography.
Accurate observations of fluid inclusions rely on an optimal microscopic image. Three-
dimensional petrography using 50 to 1000 µm thick doubly polished sections is essential
for the successful determination of phase assemblages of inclusions and the chronology of
trapping (e.g., Roedder 1984). Standard petrographic thin sections are 25 to 30 µm thick
essentially providing a two-dimensional view only. As common inclusion sizes are comparable
to the thickness of thin sections, most fluid inclusions are destroyed upon preparation and will
simply not be discovered. The ideal slide thickness for petrographic inclusion investigations
depends on the actual sizes of the fluid inclusions, and properties of the host mineral including
color, optical anisotropy and cleavage. Because the visibility of inclusions within a mineral
deteriorates quickly with increasing depth below the surface, the ideal sample thickness is
often in the range between 60 and 120 µm. Minerals with a strong color require comparatively
thin sections, while those with a good cleavage, like pyroxene, amphibole and plagioclase,
may require comparatively thick sections for the inclusions to survive preparation.
InCLuSIOn CLaSSIFICatIOn
Fluid inclusions may be classified according to their phases visible at room temperature,
which may vary from single-phase (vapor, liquid or supercritical fluid), two-phase liquid-
vapor, three-phase liquid-liquid-vapor (typical for mixed-system CO2-H2O inclusions at room
temperature), and each of the above types may additionally contain one or more daughter
minerals (defined as a mineral formed after sealing off of the inclusion from its surroundings;
Roedder 1984) (Fig. 2). This descriptive classification may be expanded to classify complex
inclusions containing several solid and fluid/liquid phases. The drawback of this classification is
that inclusions with similar compositions and densities may be classified as distinct types. More
often, inclusions in minerals are chronologically classified according to their time of trapping
relative to host mineral growth (Sobolev and Kostyuk 1975; Roedder 1979, 1984). Generally,
fluid inclusions may be trapped both during and after growth of the host crystal (Fig. 3).
Primary inclusions occur as isolated, single inclusions or in small non-oriented groups.
They may also be aligned within and often parallel to mineral growth zones. Primary inclusions
have been formed during, and as a direct result of, growth of the host crystal. Secondary
inclusions occur as trails that extend to the surface of the host mineral and may crosscut several
neighboring grains. They were thus formed after termination of crystal growth and represent
fluids trapped during healing of fractures or microcracks. Pseudosecondary inclusions occur
along healed microcracks which were formed during host mineral growth, and are thus
senso stricto a subgroup of the primary inclusions. They are texturally distinguished by their
occurrence along straight or curved trails or planes which do not coincide with growth zones
Fluid Inclusion Thermobarometry: Tracer for Magmatic Processes 145
Figure 1. Photomicrographs of fluid inclusions in phenocrysts and xenoliths. Scale bars are 20 micrometers.
(a) Texturally early CO2-rich inclusions occurring as irregular groups in plagioclase from a magmatically
overprinted tholeiitic gabbro xenolith (La Palma, Canary Islands). (b) Trails of mature secondary fluid
inclusions cut by a young trail of immature inclusions with elongated shapes due to arrested necking-down
processes (dunite xenolith, La Palma, Canary Islands). (c) Texturally early, leaked CO2 inclusion with
decrepitation trails extending from two edges; olivine phenocryst from a basanite lava (Gran Canaria,
Canary Islands). (d) Decrepitated primary CO2 fluid inclusion in olivine from a dunite xenolith (La Palma,
Canary Islands). (e) Homogenization of two-phase CO2 inclusions into the liquid phase (plagioclase from
a tholeiitic gabbro xenolith). (f) Boiling assemblage comprising coexisting high-salinity liquid phase and
low-salinity vapor phase inclusions. All inclusions homogenize at 512 ± 40 ° C; the liquid phase has a
salinity of 63 ± 3 wt% NaCl equivalents. The daughter crystals are halite (hal), sylvite (syl), and ilmenite
(ilm). Miarolitic quartz in the Permian Eikeren alkali granite, Oslo Rift, Norway. (g) Boiling assemblage
with coexisting liquid phase and low-salinity vapor phase inclusions. All inclusions homogenize at 351 ±
25 °C; the liquid phase has a salinity of 20 ± 2 wt% NaCl equivalents. Same locality as in f).
(Fig. 3a). Pseudosecondary trails may crosscut growth zones, but do not reach the grain
boundaries of the host crystal.
The above nomenclature may be misleading if the primary mineralogy is not preserved,
like in many plutonic rocks (Van den Kerkhof and Hein 2001), where fluid inclusion
generations may be termed “early” and “late” only. In such cases, the relative chronology of the
inclusion groups and trails can be constrained from their spatial distribution and crosscutting
relationships (e.g., Touret 2001).
A fluid inclusion assemblage (FIA) describes a group of inclusions that were all trapped
at the same time or during the same event (Goldstein and Reynolds 1994); the term “group of
146 Hansteen & Klügel
Inclusion type Abbreviation Appearance Essential phases
Monophase liquid L L = 100%
Monophase vapor V V = 100%
Liquid-rich,
L+V L > 50%
two-phase
Vapor-rich,
V+L V = 50-80%
two-phase
Multiphase solid S+L±V L, S
Immiscible liquid L1 + L2 ± V L1, L2
Figure 2. Classification scheme for fluid inclusions based upon inclusion contents observed at room
temperature. L = liquid; V = vapor; S = solid. Modified after Shepherd et al. (1985).
Figure 2
a) b)
Figure 3. (a) Petrographic classification of primary (P), secondary (S) and pseudosecondary (PS) fluid
inclusions based on their three-dimensional distribution within single crystals (see text for details).
Modified after Shepherd et al. (1985). (b) Formation of secondary fluid inclusions through crack healing
and consecutive “necking-down” processes. Modified from Roedder (1984).
Fluid Inclusion Thermobarometry: Tracer for Magmatic Processes 147
synchronous inclusions” (GSI; Touret 2001) can be used as a synonym for FIA. Such a genetic,
chronology-based classification is useful for rocks with clear-cut petrographic relationships,
where in the ideal case each inclusion generation can be identified as part of a certain
paragenesis. In many cases, however, a classification based on the chronology of isolated
inclusions and healed inclusion-filled cracks may be ambiguous, and a classification scheme
based on inclusion contents will be more practical. Thus, in order to successfully classify
magmatic fluids, a combination of inclusion chronology, inclusion compositions and the
assignment to a certain paragenesis in the rock should be used whenever possible. Inclusions
trapped under liquidus conditions should be classified as parts of paragenesis including
coexisting fluid and melt, e.g., occurring along single cracks or within primary groups.
FLuId InCLuSIOn FOrMatIOn and re-equILIBratIOn
trapping during crystal growth
Most primary fluid inclusions represent fluid bubbles exsolved from a magma that adhered
to the surface of an actively growing crystal, possibly at sites of increased dislocation density.
Under favorable conditions such as periods of enhanced crystal growth these bubbles became
subsequently overgrown and trapped (Roedder 1963, 1984). In subvolcanic magma chambers
the rate of inclusion trapping can be strongly augmented by events that cause rapid degassing
and thus increased liquidus temperature and increased crystallization rates. Such events may
be reflected by increased occurrences of inclusions along growth zones of phenocrysts, and
may generally be caused by changes in pressure, temperature and melt composition during
magma movements or mixing (e.g., Pearce and Kolisnik 1990; Coombs and Gardner 1991;
Troll and Schmincke 2002; Blundy and Cashman 2005). Primary fluid inclusions may also
become trapped during recrystallization of minerals as a result of deformation or heating with
partial dissolution, in which case they occur as irregular groups rather than singly or along
growth zones (e.g., Hansteen et al. 1998; Fig. 1a).
Formation of secondary inclusions
Secondary and pseudosecondary inclusions originate as fluid-filled microcracks in the
host mineral (Figs. 1b and 3b). Such cracks form after stress buildup in minerals upon pressure
release, temperature change, and/or deformation. When an available fluid enters the crack
due to capillary forces the crack tends to heal rapidly at magmatic temperatures (Roedder
1979, 1984). The healing process begins with formation of elongated, often wedge-shaped
compartments enclosing portions of the fluid. By recrystallization processes the surface energy
in the crack-fluid system, which is the product of surface tension and area, becomes minimized.
If the host mineral has a finite solubility in the fluid, material of the host crystal dissolves
from high-tension surfaces (small curve diameters) and re-precipitates on low-tension surfaces
(greater curve diameters), a process termed necking-down (Fig. 3b). The minimum surface
area is that of a regular arrangement of spheres within the former crack. The minimum surface
energy is given by a partly or wholly faceted surface enclosing the same volume, referred to as
a negative crystal shape of an inclusion.
Post-entrapment re-equilibration
Roedder’s rules. An investigated primary or secondary fluid inclusion in principle represents
the P-T-X conditions at the time of entrapment only if it satisfies the following criteria, often
referred to as “Roedder’s rules” (Roedder 1984): 1) a single homogeneous fluid phase was
trapped, 2) the inclusion remained at a constant volume after trapping, and 3) nothing was
added to or removed from the inclusion after trapping. Criterion 1) cannot be simply claimed
but needs to be verified by careful microscopic investigations, although this is not easy for
small inclusions. Criteria 2) and 3) address the basic assumption for the interpretation of fluid
148 Hansteen & Klügel
inclusions, namely, that they have behaved as an isochoric system having constant volume and
mass under the conditions of interest. In most cases this assumption is a first approximation at
best because, after inclusion trapping, the P-T path of the host mineral carried by an ascending
magma differs strongly from the isochores of the trapped fluids. This results in the evolution
of a pressure gradient between inclusions and surroundings of the host crystal. If the pressure
differences become large, the host responds by plastic deformation (stretching) and/or brittle
deformation (decrepitation), which is termed volumetric re-equilibration of inclusions (e.g.,
Bodnar 2003). Fluid inclusions in igneous rocks usually become over-pressurized, but at other
P-T conditions an internal under-pressure may also develop.
Stretching. This type of volumetric re-equilibration describes permanent, plastic
deformation (creep) of the enclosing crystal without loss of fluid from the inclusion. Stretching
is not relevant at short time scales except for soft minerals (low Mohs hardness, e.g., fluorite
and calcite) but is the typical mode of re-equilibration at low strain rates, often associated with
long time scales. At high temperatures, however, stretching may also affect hard minerals and
significantly reduce inclusion density in a short time. Experiments with synthetic NaCl-H2O
inclusions in quartz having an internal overpressure of about 210 MPa showed significant
stretching after a few days at 625 °C and recovery to 150 MPa after 30 days (Vityk and Bodnar
1998). Re-equilibration experiments with olivine at 1400 °C and ambient pressure showed
stretching of natural CO2 inclusions and related density decrease from about 1.0 to 0.7 g/cm3
within a few days (Wanamaker and Evans 1989). Stretching is problematic insofar as it affects
some inclusions more than others, which cannot be recognized by optical microscopy (Vityk
and Bodnar 1998). In addition, stretching may promote diffusive loss of fluid components
through crystal defects involved in the deformation processes (Bakker and Jansen 1991).
Stretching may however be inferred from different phases or volume proportions in inclusions
belonging to the same fluid inclusion assemblage, and from the distribution of homogenization
temperatures (and hence densities) of a fluid inclusion assemblage as is discussed below.
Compared to stretching, expansion of the host crystal due to decompression is generally
of minor concern because most minerals in igneous rocks and xenoliths (olivine, pyroxene,
feldspar, amphibole, quartz) have small compressibilities. The inclusion volume will merely
increase by about 0.1% for every 100 MPa of pressure release, which can usually be ignored
(Roedder and Bodnar 1980). Likewise, the thermal expansion of silicate minerals is also rather
moderate (on the order of 0.2-0.5% per 100 °C; Robertson 1988).
Decrepitation. The typical mode of volumetric re-equilibration at high strain rates and for
relatively hard minerals is (partial) decrepitation, sometimes also referred to as leakage. The
host crystal fails by fracturing rather than creep and the inclusion leaks parts of its contents
by advection along the rapidly propagating microcracks. If the fluid is completely lost after
intense fracturing then the term total decrepitation is used. Decrepitation can be recognized
in petrographic investigations by halos of minute fluid inclusions adjacent to inclusions and
by microcracks (“whiskers”) often extending from sharp corners (Fig. 1c,d). Leakage due
to internal overpressure causes a drop in inclusion pressure and density to unpredictable
values. If the cracks penetrate grain boundaries and become connected to the surrounding, the
inclusion may adopt the actual ambient fluid pressure. In many cases the inclusion becomes
sealed again by microcrack healing, which occurs in a short time at magmatic temperatures;
e.g., in olivine within a few hours to days depending on crack size (Wanamaker et al. 1990).
If leakage and subsequent sealing occur at identical P-T conditions such as during ponding
of an ascending magma, then the inclusion may record the new ambient pressure. Thus the
geological interpretation of partially decrepitated fluid inclusions can still be very important,
as they often reflect a specific stage of host rock evolution.
Other re-equilibration processes. Necking-down during the evolution of secondary fluid
inclusions can also be considered as re-equilibration (Fig. 3b). If formation and sealing of
Fluid Inclusion Thermobarometry: Tracer for Magmatic Processes 149
the individual daughter inclusions involves a homogeneous fluid then each inclusion has the
same density as the entrapped fluid. However, if necking-down is accompanied by cooling and
related phase changes then the densities of the new inclusions may differ strongly from each
other (Roedder 1984). The composition and bulk density of inclusions may also change by
compositional re-equilibration involving 1) diffusion of components through the host mineral
and 2) reactions between fluid and inclusion walls that occur after trapping. These processes
are discussed below in the section of missing H2O. In general, experimental data suggest that
various mechanisms of re-equilibration operate together in a specific environment and that
extrapolation of data may be ambiguous. For a detailed treatment we refer to the review by
Bodnar (2003) and references therein.
MICrOtherMOMetry
Microthermometry is defined as the observation of phase changes occurring in a fluid
inclusion upon heating. For comparatively well-known unary and binary fluid systems (i.e.,
CO2, H2O, CO2-H2O, H2O-NaCl), such phase changes provide quantitative information on the
composition and density of single fluid inclusions. For more complex systems involving three
or more major components, microthermometry can provide semi-quantitative to qualitative
data depending on the chemical system involved. Thus additional microanalytical methods
are often needed to characterize the system, including non-destructive methods for major and
minor elements, such as Laser Raman spectroscopy and Fourier Transform Infrared (FTIR)
spectroscopy (e.g., Wopenka et al. 1990; Burke 2001). Raman and FTIR spectroscopy can
provide species ratios (e.g., of CO2, CH4, N2, CO, SO2) in single fluid inclusions, and allow
for identification of daughter minerals. Trace element analysis of single fluid inclusions can be
performed in-situ with particle induced x-ray emission (PIxE) and synchrotron xRF, and by
laser ablation ICP-MS by ablating the inclusion. A description of these methods is outside the
scope of this chapter and the interested reader is referred to Samson et al. (2003), Cauzid et al.
(2006), Simon et al. (2007) and references cited therein.
Microthermometry measurements are performed using a purpose-designed heating-
freezing stage for the use with a petrographic microscope. Two designs are commercially
available, namely the Reynolds gas flow stage (Fluid Inc., Denver, CO) (based on the USGS
gas flow stage; Werre et al. 1979; Woods et al. 1981), and the Linkam stage (Shepherd 1981),
which are built to operate in the temperature range between −196 °C (the temperature of the
liquid nitrogen cooling agent) and about 600 °C. The petrographic microscope has to be fitted
with long working distance lenses and condensor system. High-temperature heating stages for
measurements above 600 °C include the commercially available Linkam TH 1500 (Oskarsson
and Hansteen 1992) and the Vernadsky rapid-heating and quenching stage (designed at the
Vernadsky Institute, Moscow; zapunnyy et al. 1988). The latter attains heating and cooling
rates up to about 100 °C/s.
Because metastability is a common problem in such microscopic systems as fluid inclusions,
microthermometry measurements are always performed upon heating. The nucleation of new
phases like a mineral (ice, solid CO2, etc) or a vapor bubble requires significant undercooling,
where the nucleation temperature is not reproducible. Thus microthermometry measurements
comprise melting and dissolution reactions and homogenization of inclusions into liquid or
vapor upon heating. The most common microthermometry data include a) temperature of first
melting, denoted Tt (triple point temperature) for unary systems and Tm(eutectic) or Tm(initial)
for binary and higher systems, respectively; b) final melting temperature Tm(final); c)
homogenization temperature into the liquid (Th(LV→L)) or into the vapor (Th(LV→V)) phase,
respectively; and d) Ttrap (temperature of trapping) (cf. Roedder 1984). Partial homogenization,
involving homogenization into the liquid or vapor phase in the presence of either daughter
minerals or a second liquid phase, is still denoted Th(LV→L) or Th(LV→V), respectively.
150 Hansteen & Klügel
Temperature standards for microthermometry include compounds with known melting
points, and to an increasing extent comprise synthetic and natural fluid inclusions with known
compositions and densities (e.g., Sterner and Bodnar 1991). For detailed discussions on lab
techniques and interpretation of microthermometric data in general see Roedder (1984),
Shepherd et al. (1985), and Samson et al. (2003).
FLuId SySteMS
In the following discussion of fluid phase systems, we focus on the use of microthermometry
to obtain densities of magmatic fluids under isochoric conditions. When an isochoric system is
heated, its internal pressure changes as a complex function of composition and density. Thus
microthermometry observations are usually depicted in isochoric phase diagrams, rather than
their isobaric or isothermal siblings used in petrology.
The most common fluid inclusions in magmatic systems belong to the pseudo-system
H-O-C-N-S-NaCl, where NaCl represents highly-water-soluble salts including various halides
and sulfates. The model systems briefly introduced here are CO2 (±CH4), and H2O-NaCl,
because these are most relevant to barometry and are good approximations of commonly
occurring geological fluids. A more detailed description of these and other fluid systems is
outside the scope of this chapter, and the reader is referred to Bowers and Helgeson (1983a,b),
Van den Kerkhof (1990), Sterner and Bodnar (1991), Schmidt and Bodnar (2000), Diamond
(2003) and references cited therein.
the CO2 system
CO2 is a comparatively simple unary system that can be used to explain basic features
pertaining to most fluid systems. Depicted in a P-T projection, the system CO2 comprises
three primary phase fields solid (S), liquid (L) and vapor (V) (Fig. 4). Applying the Gibbs
phase rule, the primary phase fields have a variance of 2 and are separated by univariant curves
where S+L, S+V and, most commonly, L+V coexist. The L+V univariant curve ends at the
critical point at 30.98 °C and 7.38 MPa (Span and Wagner 1996). The three phases S+L+V
coexist only at the invariant triple point, which for pure CO2 is at −56.56 °C and 0.52 MPa.
The density ρ, which is essential for the interpretation of fluid phase systems, is not directly
visible in the P-T-projection but can be visualized in a three-dimensional P-T-ρ diagram (Fig.
4a). Because each CO2 inclusion can have only one unique density and hence one unique
univariant two-phase curve, the entire array of these curves covers all possible densities.
Microthermometry analysis involves cooling the inclusion towards the temperature of
liquid N2 until its contents comprise solid and vapor phase CO2. By heating up the inclusion
liquid CO2 appears at the invariant triple point at temperature Tt. Slight heating above Tt makes
solid CO2 melt completely, and the inclusion consists of liquid and vapor (Fig. 5). Depending
on the inclusion density, further heating along the L+V curve produces one of the following
reactions: (1) At densities below 0.468 g/cm3 (the critical density), the vapor bubble will expand
at the expense of the liquid phase, until the inclusion homogenizes into the vapor phase at
temperature Th(LV→V). (2) At densities above 0.468 g/cm3, the liquid phase will expand
at the expense of the vapor bubble, until the inclusion homogenizes into the liquid phase at
temperature Th(LV→L) (Figs. 1e and 5). (3) For inclusions having the critical density, the liquid
phase and vapor bubble will maintain their original sizes until the inclusion homogenizes by
fading and disappearance of the meniscus between both phases, termed critical homogenization.
Inclusions having near-critical densities may also show critical behavior because the two-phase
field is broad and flat near the critical temperature (Roedder 1965; Angus et al. 1976). Once
homogenized, each inclusion consists of a supercritical fluid and must follow an isochore where
further heating will not result in any additional phase changes.
Fluid Inclusion Thermobarometry: Tracer for Magmatic Processes 151
a)
b)
Figure 4. (a) Three-dimensional pressure (P) - temperature (T) - density (ρ) block diagram for the CO2
system schematically showing the relationship between P-T and T-ρ projections. L= liquid; V= vapor; c.p.
= critical point. The shaded section represents the density of an arbitrary inclusion homogenizing into the
liquid phase. Modified from Roedder (1984) and Goldstein and Reynolds (1994). (b) P-T projection of
the block diagram showing CO2 isochores in the low-temperature range based on data from Angus et al.
(1976); modified after Roedder (1984). Numbers on isochores are densities in g/cm3; Tt = triple point; c.p.
Figure 4
= critical point.
To sum up, temperature measurements of two phase transitions are necessary and
sufficient to characterize the properties of CO2 or any other unary inclusions. Triple point
melting is used to identify the chemical system, and homogenization into liquid or vapor is
used for an accurate density determination. The volumetric and microthermometric behavior
of fluids in the system CO2-N2-CH4 has been summarized by Van den Kerkhof (1990), Kooi et
al. (1998), and Van den Kerkhof and Thiéry (2001).
152 Hansteen & Klügel
Trapping
Ptrap
S L
e
Pressure
or
ch
Iso
critical point
triple
point V
L+V
S+V -56.6 °C +31.1 °C
Tm Th(LV→L) Ttrap
Temperature
Figure 5. P-T projection illustrating the phase transitions that occur in an arbitrary pure CO2 inclusion upon
heating. The inclusion has a density above the critical density and thus homogenizes into the liquid. The
homogenization temperature Th(LV→L) defines the isochore, from which the pressure and temperature of
trapping can be derived. Isochores of inclusions homogenizing into vapor run below the critical point as
indicated
Figureby the
5 dashed line. Modified after Van den Kerkhof (1988).
The H2O-NaCl system
Overview. The low pressure part of the P-T phase diagram of pure H2O has a topology
similar to that of CO2, but the absolute P and T values are different (Fig. 6). Phase transitions
relevant to fluid inclusion investigations occur at comparatively higher temperatures, including
the triple point at 0.15 °C and the critical point at 374 °C (representing a density of 0.32 g/
cm3). Like for the CO2 system, low-density inclusions homogenize into the vapor phase and
high-density ones into the liquid phase. Isochores for high-density water inclusions are notably
steeper than those for most CO2 densities (Crawford 1981).
Adding NaCl to the H2O system changes it from a unary to a binary, adding one degree of
variance to each of the phase elements. At low temperatures, the triple point of H2O is replaced
by a binary melting curve in the H2O-NaCl system (Fig. 7). The low-temperature part of the
system is a binary eutectic with a minimum at −20.8 °C, and a eutectic composition of 23.3
wt% NaCl. Below 0.1 °C, the salt hydrate hydrohalite (HH; NaCl·2H2O) is stable.
Phase changes observed during microthermometry. Upon heating of a low-salinity H2O-
NaCl inclusion, either hydrohalite or ice will melt last, where final ice melting reflects the lower
salinities. The final melting temperatures of either ice (Tm(Ice)) or hydrohalite (Tm(HH)) can be
used to estimate accurately the salinity and thus the composition of the inclusion. For inclusions
containing halite as a daughter mineral, salinities are measured using the temperature of halite
dissolution (see below).
Adding NaCl to the water system in amounts up to the room-temperature saturation level
(26.3 wt% NaCl) has three major effects (Fig. 6): a) the bulk densities increase because salt
dissolves readily in water with very little net volume expansion; b) the critical point moves to
progressively higher temperatures with the amounts of salt added, leading to an extended two-
phase L+V curve; and c) the isochores for a given liquid homogenization temperature become
less steep.
Fluid Inclusion Thermobarometry: Tracer for Magmatic Processes 153
0 .9 2
0.80
8
0.90
0.99
3
1.02
0.5
0.91
1.11
0.7
c.p.
120
100
Pressure (MPa)
Liquid
80 Vapor
60 c.p.
40 H2O
c.p. 10 wt% NaCl
20 Liquid 25 wt% NaCl
Vapor
100 300 400 500 600
Temperature (°C)
Figure 6. Pressure-temperature projection of the system H2O and the low-salinity part of the system H2O-
NaCl at elevated temperatures. In the system H2O, the L+V two-phase curve extends to the critical point
at 374 °C. Addition of salt to the water system gradually increases the extent of the L+V two phase curve
until nearly 700 °C for a salinity of 25 wt% NaCl. With increasing salinity the fluid densities increase and
the isochores become somewhat less steep. Isochores are marked with densities in g/cm3, and c.p. is the
critical point for each respective composition. Modified after Crawford (1981).
Figure 6 Y Solid (halite)
+
Liquid
Temperature (°C)
+0.10°
0 Liquid
Solid (hydrohalite)
+
Solid (ice) + Liquid Liquid
-20.8
E
Solid (ice + hydrohalite)
23.3 26.3
NaCl (wt%)
Figure 7. Melting properties for fluid inclusions in the low-temperature part of the system H2O-NaCl.
Salinities below the room-temperature saturation of 26.3 wt% NaCl can be estimated by either observing
the temperature of final ice melting (for salinities below the eutectic composition of 23.3 wt% NaCl), or
the temperature of final hydrohalite melting (for salinities between 23.3 and 26.3 wt%). Modified after
Crawford (1981) and Hall et al. (1988).
154 Hansteen & Klügel
High-salinity inclusions contain halite (H) as a daughter mineral at room temperature
(Fig. 1f). Thus the two-phase L+V curve for the low-salinity compositions is replaced by a
three-phase H+L+V curve (Fig. 8). In this case, the salinity of an inclusion can be accurately
measured by observing the temperature of final halite dissolution (Tm(Halite)) at the appropriate
halite liquidus (Bodnar et al. 1985; Chou 1987; Sterner et al. 1988, 1992). The inclusion density
is measured by observing its homogenization into the liquid (Th(LV→L)) or vapor (Th(LV→V))
phase. Depending on bulk density, four different sequences of high-temperature phase transitions
are possible for inclusions having the same salinity, i.e., for inclusions occurring along the
same isopleth in the H2O-NaCl system. This is illustrated by an example using four different
inclusions (A to D) with compositions of 50 wt% NaCl and 50 wt% H2O but four different
densities (Fig. 8). A) Upon heating, the highest-density inclusion containing the smallest vapor
bubble at room temperature will lose its bubble prior to halite dissolution. After homogenization
into the liquid phase at temperature T1 (Fig. 8), it will follow a path across the two-phase L+H
field until final homogenization at the 50 wt% halite liquidus (Tm(Halite) at T2). The resulting
isochore is comparatively steep. B) Inclusion B having a lower density and thus a larger bubble
at room temperature will homogenize into the liquid phase at T2 and simultaneously dissolve
its halite at the appropriate halite liquidus (Tm(Halite); Fig. 8). The isochore thus extends from
the intersection point of the halite liquidus with the H+L+V three phase curve. C) Inclusion
C has a still larger vapor bubble than inclusion B, and will homogenize into liquid after halite
dissolution. It will thus cross the two-phase L+V field until homogenization (Th(LV→L) at
temperature T3; Fig. 8). The vapor-rich inclusion D will also intersect the halite liquidus at T2.
Because it has a density lower than the critical density, it will homogenize into the vapor phase
and must thus follow a low-pressure path across the two-phase L+V field until homogenization
Tm(Halite) A B C
Room temperature
phase proportions L
H+L=L
B CP(50 wt% NaCl)
H+L
Pressure
D
C L+V
H+V H+L+V
T1 T2 T3 T4
Temperature
Figure 8. Possible modes of homogenization for fluid inclusions in the system H2O-NaCl containing
halite at room temperature. The example considers the 50 wt% NaCl isopleth, thus all inclusions have the
same salinity but different densities decreasing from inclusion “A” to inclusion “D”. The modes of final
homogenization are (inclusion designation in parenthesis): (A) Halite melting Tm(Halite); (B) simultaneous
halite melting and liquid phase homogenization Th(LV→L); (C) liquid homogenization, and (D) vapor
homogenization Th(LV→V), respectively. c.p. is the critical point and Tm(H) the final melting temperature
of halite. Modified from Konnerup-Madsen (1995).
Fluid Inclusion Thermobarometry: Tracer for Magmatic Processes 155
into vapor (Th(LV→V) at T4; Fig. 8). The resulting isochore will be much less steep than for the
higher-density inclusions with the same salinity.
Phase separation (boiling) in the H2O-NaCl system. A simplified P-T-X diagram of the
system H2O-NaCl is shown in Figure 9. A large immiscibility field exisits within this system at
elevated temperatures and low to moderate pressures (Sourirajan and Kennedy 1962; Bodnar
et al. 1985; Chou 1987; Driesner and Heinrich 2007; Simon et al. 2007). For clarity, the low-
temperature part of this field is shown as a pressure-composition projection in Figure 10.
All saline aqueous fluids entering the immiscibility region will instantaneously separate into
coexisting vapor and liquid phases. Taking the 800 °C isotherm at 100 MPa pressure as an
example, the two equilibrium phases are a high-density liquid (brine) containing 70 wt% NaCl,
and a low-density vapor containing about 2 wt% NaCl (points A and B in Figure 10). All fluids
with compositions between 2 and 70 wt% NaCl at 800 °C and 100 MPa must instantaneously
unmix to form such coexisting vapor and liquid phases. Generally, changes in each of the
parameters P, T and X can result in boiling at elevated temperatures. A common mechanism
for boiling of originally homogeneous fluids is pressure decrease, induced by fracturing of
roof rocks above shallow intrusions or magma chambers (cf. Roedder and Bodnar 1980). Thus
a rapid pressure drop corresponding to a change from lithostatic to hydrostatic conditions can
be facilitated. Boiling of an originally homogeneous fluid phase may lead to volume expansion
of the melt-fluid system and thus promote further wall-rock fracturing.
Other components. Most naturally occurring aqueous fluid inclusions contain major
components in addition to H2O and NaCl, the most important being CaCl2 and MgCl2
(Konnerup-Madsen 1979; Crawford 1981; zhang and Frantz 1987). In order to relate natural
samples to the H2O-NaCl system, microthermometric measurements of aqueous inclusions are
Halite liquidus Critical curve
200
150
P (MPa)
L+V 100
coexistence
H+L+V
coexistence:
liquid 50
100 H+L+V
90 coexistence:
80
70 vapor
60
50 Critical point
40 of H2O 1000
30 700 800 900
20 600
Wt. % NaCl 10 400 500
300
0 0 100 200
T (°C)
Figure 9. Topology of the system H2O-NaCl at elevated temperatures
and pressures. Modified after Driesner and Heinrich (2007).
Figure 9
156 Hansteen & Klügel
150
1000
100
900
0
80
80
0 °C
0
70
0
Vapor Liquid
100
Pressure (MPa)
A B
600
.
C.C
500
50
400 H+L+V
H+V
0
0 10 20 30 40 50 60 70 80 90 100
NaCl (wt.%)
Figure 10. Composition-pressure projection of immiscibility relations in the system H2O-NaCl, showing
the boundaries for the L+V two-phase fields for selected temperatures. The compositions “A” and “B”
connected with a horizontal tie line indicate the compositions of the coexisting vapor and liquid phases,
respectively, at 800 °C and 100 MPa. Modified from figures in Bodnar et al. (1985) and Chou (1987).
generally treated as10
Figure if they belong to this simplified system. Thus salinities measured in natural
fluid inclusions are given as wt% NaCl equivalents (Clynne and Potter 1977).
Descriptions of the geologically relevant but complex ternary system H2O-NaCl-CO2 is
beyond the scope of this paper. In general, however, adding CO2 to the system H2O-NaCl
extends the two-phase (immiscibility) conditions to higher pressures and temperatures
compared to either of the binary systems H2O-NaCl or H2O-CO2 (Takenouchi and Kennedy
1965; Bowers and Helgeson 1983a,b; Roedder 1984; Brown and Lamb 1989; Labotka 1991;
Schmidt and Bodnar 2000).
traCKInG VOLCanIC PLuMBInG SySteMS uSInG CO2 InCLuSIOnS
Pure CO2 or CO2-dominated fluid inclusions occur ubiquitously in xenoliths and
phenocrysts hosted by basaltic as well as evolved magmas (Roedder 1965, 1984). They
provide an excellent geobarometer even though the method is based on some assumptions and
has several sources of serious errors (Roedder and Bodnar 1980). This section is a “handbook”
to show the usage of CO2-dominated inclusions for barometry, to review the principles and
limitations of the method, and to constrain sources and magnitudes of possible errors.
Examples. In a study of CO2 inclusions in ultramafic xenoliths from Loihi seamount
(Hawaii), Roedder (1983) was probably the first who related the depths recorded by primary
and secondary inclusions to that of a magma chamber inferred from seismic data. Hansteen
(1991) used the density distribution of CO2-rich inclusions in olivine phenocrysts to show that
a picritic magma from Iceland was trapped at 7-10 km depth near the crust-mantle boundary
before eruption. Likewise, the density distributions of CO2 inclusions in phenocrysts and
xenoliths from the Canary Islands (Hansteen et al. 1991, 1998; Klügel et al. 1997, 2005;
Galipp et al. 2006) and Madeira Archipelago (Schwarz et al. 2004; Klügel and Klein 2006)
Fluid Inclusion Thermobarometry: Tracer for Magmatic Processes 157
combined with clinopyroxene-melt barometry indicate the presence of well-defined levels of
magma stagnation in both mantle (>15 km depth) and crust; a bimodal pressure distribution is
not uncommon (Fig. 11). For Réunion Island, in contrast, fluid inclusion densities of olivine
phenocrysts suggest that zones of magma crystallization and fluid trapping extend from near
the surface to 15 km depth at least (Bureau et al. 1998). Pure CO2 inclusions in quartz xenoliths
from Vulcano Island yield two distinct density intervals, which suggests ponding of magmas
first in a lower crustal reservoir (21-17 km) and then fractionation in a shallower reservoir
(14-8 km; Zanon et al. 2003; Frezzotti and Peccerillo 2004); an additional zone of magma
accumulation is indicated at 2-1.5 km depth (Clocchiatti et al. 1994) (Fig. 11). Belkin and De
Vivo (1993) used CO2 and H2O-CO2 inclusions in ultramafic cumulate xenoliths from plinian
eruptions of Vesuvius to infer a depth range of crystallization of 4-10 km. Fulignati et al.
(2004) interpreted CO2-rich fluid inclusions in skarn xenoliths of the 1944 K-phonotephritic
eruption of Vesuvius to indicate a pressure of <100 MPa for the shallow reservoir, which
implies a multistage evolution before eruption. Looking at continental intraplate volcanoes, a
change of magma ascent rates at Moho depths was inferred from the density distribution of two
generations of CO2 inclusions in mantle xenoliths from a volcanic field at the Hungary-Slovakia
border (Szabó and Bodnar 1996). For the East Eifel volcanic field (Germany) overlapping
pressures indicated by CO2-inclusions in both cumulate and metamorphic xenoliths were
taken as evidence for a main magma reservoir in the lower crust at 22-25 km depth, near the
brittle-ductile transition (Sachs and Hansteen 2000).
These examples show that multi-stage ascent of magmas with stagnation at Moho or
crustal depths is the rule rather than the exception, and that CO2-dominated fluid inclusions
played the major role in identifying these levels. Because the P-T-V properties of CO2 are
reasonably well known, the determination of inclusion densities and pressures is relatively
straightforward. Nevertheless, some pitfalls and error sources do exist and need to be considered
in any barometric study based on fluid inclusions.
La Palma Vulcano
East West
km
C C
Sediments
Oceanic 10 U.C. Rhyolite
crust B magmas
B
Mantle
L.C.
20 A
A Mafic
Mantle magmas
30
Figure 11. Models of the Recent magma plumbing systems beneath La Palma (Canary Islands) and
Vulcano (Aeolian arc) based on data and illustrations from Hansteen et al. (1998); Zanon et al. (2003);
Frezzotti and Peccerillo (2004); Klügel et al. (2005); and Peccerillo et al. (2006). For La Palma, the magma
Figure 11
storage systems in the lower oceanic crust (B) and uppermost crust (C) were inferred by fluid inclusion
barometry, whereas those in the mantle (A) reflect clinopyroxene-melt barometry data and subordinate
fluid inclusion data. The active part of a single eruption is shown in black. For Vulcano, three main levels
of magma storage have been identified by fluid inclusion barometry: (A) in the mafic granulite lower
crust (L.C.) to upper mantle; (B) in the felsic granulitic upper crust (U.C.) where mafic to intermediate
magmas rising from (A) fractionate to rhyolitic magmas; and (C) within the volcano-sedimentary terrain.
The barometry-based models for both islands are consistent with petrological and geophysical data on the
underlying lithological structures.
158 Hansteen & Klügel
rationale
In an ideal fluid inclusion pressure and temperature are related by an isochore described
by an equation of state for the respective fluid enclosed. If the fluid composition X and the
equation of state are known and T is determined by an independent method, then a pressure
represented by the inclusion can be derived. By investigating a sufficiently large number of
fluid inclusions, their density distributions can yield the depths of phenocryst crystallization
and/or xenolith entrainment (which in many cases reflect a magma chamber) and the depths of
temporary magma stagnation during ascent. This is highly valuable information on a volcanic
plumbing system not easily available from other methods. We point out that the study of both
primary and secondary inclusions is relevant and useful here because early as well as late
processes characterize the nature of a plumbing system.
Applicability. Compared to barometers based on mineral-melt equilibria that are suitable
for volcanic rocks (reviewed in Putirka 2008), fluid inclusion barometry offers two advantages:
1) it is not confined to certain mineral phases or phase assemblages but can be used for basically
any mineral occurring as phenocryst phase or in a xenolith, and 2) it has the potential to reveal
short-lived events during magma ascent since fluid inclusions can rapidly adapt to decreasing
ambient pressures. On the other hand, processes occurring after trapping bias the results and
limit the extent to which high pressures in the inclusions are preserved. In general, pressures
recorded by both primary and secondary inclusions considered here always represent a
minimum limit for their trapping pressure. Barometers using H2O and CO2 contents of silicate
melt inclusions offer similar advantages and disadvantages as compared to barometers based
on mineral-melt equilibria; see Metrich and Wallace (2008) for a discussion.
Fluid inclusion barometry is thus a tool complementary to mineral-melt barometry,
and we strongly suggest using both methods wherever possible, combined with careful
petrographic observations and additional petrological data, in order to obtain the most complete
information for a reconstruction of the ascent history of magmas (e.g., Peccerillo et al. 2006).
Xenoliths are particularly valuable in this respect as they are often derived from wall-rocks
of a magma reservoir where melt accumulation and degassing cause high deviatoric stresses
and fluid penetration. Many xenolith crystals become therefore fractured and rehealed under
the presence of CO2-rich fluids, which results in an extremely large number of secondary
inclusions (Roedder 1965, 1984; Andersen and Neumann 2001) that may reflect the P-T
conditions at the presumptive magma reservoir. Phenocrysts, in contrast, usually show far less
secondary inclusions because of the small magnitude of deviatoric stresses exerted on them
by the host melt.
Density distribution. The re-equilibration of primary and secondary fluid inclusions
obviously results in a biased density distribution. The extent of re-equilibration and whether
it occurs by either stretching or partial decrepitation depend on inclusion size, the rate at
which P-T conditions change, and the deformation properties of the host mineral and hence
its structure and composition (e.g., Bodnar 2003). The important point in fluid inclusion
barometry is that data from natural systems may provide ample information on the magmatic
plumbing system even if—or maybe because!—Roedder’s rules 2) or 3) are not satisfied. The
key to success is the nature of the density distribution of an investigated fluid inclusion suite
(e.g., Hansteen et al. 1998; Touret 2001; Andersen and Neumann 2001).
This is schematically shown in Figure 12 for fluid inclusions in phenocrysts and xenoliths
trapped at a deep level. If the host magma ascends from this level to the surface more or less
continuously then the density distribution will be skewed and show a broad “re-equilibration
tail” (Fig. 12a). Depending on factors such as the ascent velocity, the original density may
still be recognizable as a frequency maximum or merely as an upper limit of measured values.
In contrast, if the ascending magma stagnates temporarily at one or more levels such as the
mantle-crust boundary, then the extent of re-equilibration is larger at these levels and additional
Fluid Inclusion Thermobarometry: Tracer for Magmatic Processes 159
(a)
Entrapment or 100 Lanzarote
reequilib. level peridotite
Frequency
75 xenoliths
"Reequilibration tail"
50 Reequil. trail
25
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Inclusion density
Density (g/cm3)
(b)
Vulcano
Entrapment or Entrapment or 40
quartz
reequilib. level reequilib. level
Frequency
30
xenoliths
20
10
0.17 0.19 0.23 0.29 0.42 0.54 0.66 0.73 0.77 0.81 0.85
Inclusion density
Density (g/cm3) (non-linear scale)
(c)
Level of prolonged La Palma Moho
reequilibration 30 peridotite
Frequency
xenoliths
Inferred
20
Entrapment or level of
reequilib. level entrapment
10
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Inclusion density
Density (g/cm3)
Figure 12. Sketch showing idealized (left) and measured (right) density distributions of fluid inclusion
associations in phenocrysts or xenoliths hosted by magmas with different ascent histories. (a) Single-
stage ascent from the depth of entrapment causes random re-equilibration of inclusions en route to the
surface. This results in a single frequency maximum at the high-density end and a large spread and random
distribution of data towards lower densities. The example for a suite of peridotite xenoliths from Lanzarote
(Canary Islands) is based on data from Neumann et al. (1995). (b) Multi-stage ascent from depth with
temporal ponding at a shallower level. This results in a large spread of inclusion densities with two distinct
frequency maxima, as is exemplified by early and late fluid inclusions in quartz xenoliths from Vulcano
Island where the second maximum corresponds to a shallow level of trapping and re-equilibration (data
from first volcanic cycle; Zanon et al. 2003). (c) Multi-stage ascent from depth with prolonged ponding at
a shallower level. This results in almost complete re-equilibration of inclusions with a single pronounced
maximum and little spread of densities. The example for a suite of Recent peridotite xenoliths from La
Palma (Canary Islands) is based on Hansteen et al. (1998).
frequency maxima become recognizable in the density distribution; new inclusions may also
become trapped (Fig. 12b). An example is Vulcano Island where two distinct crustal reservoirs
could be identified by distinct density intervals of two generations of CO2-rich fluid inclusions
in quartz xenoliths (Zanon et al. 2003). It is highly unlikely that such distributions are fortuitous
if sufficient inclusions are analyzed. If magma ponds at a certain level for a long period of time
(months to years), then almost complete re-equilibration may occur. The resulting density
distribution shows a well-defined maximum at the shallower level and the memory of the
former deep level is almost completely lost (Fig. 12c). This situation is well exemplified by the
active Cumbre Vieja volcano on La Palma (Canary Islands) where fluid inclusions in xenoliths
and phenocrysts show an almost Gaussian density distribution reflecting pressures within the
lower crust but almost no memory of former mantle pressures (Klügel et al. 2005).
160 Hansteen & Klügel
Density distributions broadly resembling those in Figure 12 have also been derived by
decompression experiments with synthetic fluid inclusions in natural quartz (e.g., Bakker
and Jansen 1991, Vityk and Bodnar 1998). Based on systematic variations of parameters
such as decompression P-T path and duration of re-equilibration combined with statistical
analyses of the results, comprehensive relations between P-T-t parameters, nature and extent
of re-equilibration, and histogram properties could be shown. Although the experiments
were restricted to aqueous inclusions in quartz and cannot be rigorously compared to other
fluids and minerals, the results have general implications for the interpretation of density
distributions. As a particular result, it was found that a small percentage of inclusions, that
cannot be distinguished or recognized petrographically, maintained original densities even
after extensive re-equilibration.
determining inclusion density, pressure and depth
Composition of the inclusions. Most fluids enclosed in basalt-hosted xenoliths and
phenocrysts consist of essentially pure CO2 (Roedder 1965, 1984), but of course this needs
to be verified for any fluid inclusion investigated for pressure determinations since exceptions
do exist (e.g., Andersen and Neumann 2001, and references therein). Upon heating of pure
CO2 inclusions from ca. −190 °C, the first phase transition observed in microthermometric
investigations is instantaneous melting of frozen CO2 at −56.6 °C (Tm). If additional fluid
components are present then the initial melting temperature will be lowered and a temperature
interval will be noticeable between initial and final melting (Tm(initial) and Tm(final),
respectively). Isochoric T-X sections of the CO2-CH4 and CO2-N2 systems show that Tm(initial)
of CO2 is depressed by about 0.13 °C per mol% of CH4 or 0.24 °C per mol% of N2 in an
inclusion, and a melting interval is clearly observable for amounts >1 mol% of CH4 or a
few mol% of N2 (Van den Kerkhof 1990). Likewise, an amount of 1-2 mol% of SO2 lowers
Tm(initial) by about 1.4 °C and causes a melting interval of about 0.8-1.3 °C (Frezzotti et al.
2002). A combination of microthermometric and Raman microspectrometric measurements of
CO2 inclusions in natural samples indicates the amounts of CH4, N2, SO2, H2S or CO to be
less than 0.1 mol% if Tm(initial) is depressed by no more than 0.2 °C (zanon et al. 2003).
Although it is difficult to detect small amounts of additional components in CO2-dominated
inclusions by microthermometric investigations, it has been suggested that a few mol% do not
significantly affect the interpretation of the trapping conditions of inclusions (Van den Kerkhof
1990; Frezzotti et al. 2002).
The presence of H2O in CO2-rich inclusions causes no depression of Tm(initial) because of its
little solubility, but introduces additional phase transitions including formation of CO2 clathrate
hydrate at low T, its subsequent melting at T ≤ 10 °C, and final homogenization of coexisting
CO2 and H2O phases. Whereas amounts of <10 vol% of water are hard to detect in microscopic
investigations (but are detectable by Raman and FTIR spectroscopy), clathrate formation in
CO2-rich inclusions is readily recognized (Roedder 1984). According to Burruss (1981) the
bulk composition and density of such an inclusion can be obtained from the homogenization
temperature of CO2 liquid and vapor (yielding the density of the CO2 phase as shown below)
and from the relative volume proportions of both phases estimated at some temperature between
that of CO2 homogenization and 50 °C. The estimation of volume proportions, however, is not
trivial and is subject to considerable error (Roedder 1984; Bakker and Diamond 2006).
Determination of inclusion density and error magnitude. The density of CO2 inclusions
is accurately derived from the temperature and nature of final homogenization. Once
Th(LV→L) or Th(LV→V) have been established as shown above the bulk density can directly
be derived from the relation shown in Figure 13 using e.g., the auxiliary equations (3.14) and
(3.15) of Span and Wagner (1996). Since Th(LV→L) can be determined with an accuracy
and reproducibility of better than ±0.2 °C it follows from the slope of the saturated-liquid
curve in Figure 13 that the densities are accurate to within 0.001-0.01 g/cm3 (0.1-2% relative
Fluid Inclusion Thermobarometry: Tracer for Magmatic Processes 161
1.2
1
liquid
0.8
3)
Density (g/cm
liquid + vapor
0.6
Critical point
0.4 (31.98/0.468)
0.2 Triple
point
(-56.56) vapor
0
-60 -50 -40 -30 -20 -10 0 10 20 30
T hom (°C)
Figure 13. Isobaric section showing the relation between homogenization temperature and density of
saturated liquid and vapor, respectively, for pure CO2 inclusions calculated using the auxiliary equations
(3.14) and (3.15) of Span and Wagner (1996); cf. Figure 4a.
Figure 13
uncertainty), except for near the critical temperature where ∆ρ/∆T becomes very high (2-8%
relative uncertainty below 30.9 °C). The determination of Th(LV→V) is less accurate than
Th(LV→L) because the complete evaporation of a liquid film is more difficult to recognize than
elimination of a bubble that shows pseudo-Brownian motion. By assuming an accuracy of 1 °C
for the determination of Th(LV→V) the relative uncertainty for density estimates is 3-12% for
values <0.35 g/cm3 (Th < 30.2 °C). The microthermometric determination of densities of CO2
inclusions is thus surprisingly accurate, except for near-critical densities, and is not a major
error source for barometric investigations.
Which equation of state? A number of equations of state (EOS) for pure CO2 have been
published in the last decades based on different theoretical approaches, different P-T ranges
of experimental data, and different adjustable parameters (see discussions in Mäder and
Berman (1991); Seitz et al. (1994); Sterner and Pitzer (1994); and Span and Wagner (1996)).
Igneous petrologists face the problem that their range of interest, about 100 to >1000 MPa
at 800-1300 °C, is poorly represented by experiments and requires extrapolation beyond the
data used to adjust the EOS parameters. Most commonly the Modified Redlich-Kwong-type
EOS by Kerrick and Jacobs (1981) and Holloway (1981) (hereafter referred to as KJ81 and
H81, respectively) were used in the past, assuming that their extrapolation to conditions of
intratelluric magmas was adequate (e.g., Hansteen 1991; Szabó and Bodnar 1996; Hansteen et
al. 1998; zanon et al. 2003; Klügel et al. 2005).
Because of the paucity of experimental data for CO2 at temperatures >800 °C it is difficult
to judge the suitability of any EOS for magmatic systems. The calibration of the EOS of
Span and Wagner (1996; SW96) comprises some 2800 experimental ρ-P-T data in the range
from 0.1 MPa and −56 °C to 800 MPa and 830 °C with most P-T values falling within the
technically important region of <30 MPa and <100 °C. This EOS includes the latest state-of-
the-art data and is probably the most reliable equation for this range. In the high-T region the
reported uncertainty of calculated densities is ±2% at 400-800 MPa, ±1% at 100-400 MPa, and
±0.2% at 10-100 MPa. Span and Wagner (1996) suggest that their equation should also yield
reasonable data outside its validity range but the uncertainty cannot be estimated. Alternatively,
162 Hansteen & Klügel
for magmatic systems the EOS of Sterner and Pitzer (1994; SP94) is a good choice because
1) its calibration includes additional high P-T data and 2) with 28 parameters it is easier to
implement in spreadsheets than the sophisticated SW96 equation. Sterner and Pitzer (1994)
claim that their EOS provides valid isochores from 0.1 MPa and −57 °C to 10 GPa and >1700
°C, but because of lacking experimental data for the P-T conditions of magmatic systems the
respective uncertainties were not stated. Based on new high-pressure experiments, Abramson
and Brown (2004) suggest that pressures derived by the SW96 and SP94 equations need to be
decreased by a few percent at P-T conditions of the uppermost mantle.
Figure 14 compares pressures calculated by SP94, deliberately used here as a reference,
against those calculated by other common EOS using the software packages FLUIDS (Bakker
2003) and FLINCOR (Brown 1989) for a model temperature of 1200 °C. This comparison
covers almost the entire density range of CO2 inclusions reported so far. The frequently used
KJ81 equation yields negligible deviations from the reference pressures for densities <0.8 g/
cm3 but increasingly overestimates pressures at higher densities (up to 11% at 1.1 g/cm3). The
H81 and Bottinga and Richet (1981; BR81) equations underestimate pressures for low densities
by up to 10% and 7%, respectively, and overestimate for high densities by up to 10%. The
SW96 equation shows only minor deviations from SP94 over the entire density range, with a
maximum difference of −3% around the critical density of 0.466 g/cm3. An example illustrates
the effect of these differences: For an inclusion with a density of 1.1 g/cm3 the different EOS
yield a pressure range of 1032 to 1184 MPa at 1200 °C corresponding to a depth uncertainty of
ca. 4-5 km; for 0.6 g/cm3 the respective values are 269 to 303 MPa and ca. 1 km.
To summarize, the SP94 and SW96 equations are probably the best choice for calculating
CO2 isochores, with an advantage of SP94 being its easier implementation, but both equations
15 1400
10
T = 1200 °C 1200
1000
5
KJ81 800
P (MPa)
∆P (%)
0
SW96 600
-5
BR81 400
-10 H81
200
P (SP94)
-15 0
0 0.2 0.4 0.6 0.8 1
Density (g/cm3)
Figure Figure
14. Plot of14
inclusion density against pressures calculated using different equations of state (EOS)
for pure CO2 and a trapping temperature of 1200 °C. Dashed line (right scale) shows the pressure given by
the EOS of Sterner and Pitzer (1994) used here as a reference (SP94). Thick solid lines (left scale) show
the relative deviation from the reference (∆P in %) for pressures calculated using the following EOS: KJ81
(Kerrick and Jacobs 1981), H81 (Holloway 1981), BR81 (Bottinga and Richet 1981), and SW96 (Span
and Wagner 1996). The calculations used the software packages FLUIDS 1 (Bakker 2003) and FLINCOR
(Brown 1989).
Fluid Inclusion Thermobarometry: Tracer for Magmatic Processes 163
still need to be tested by experimental data for the P-T conditions of magmatic systems. The
KJ81 equation, often used in the past, retrieves pressures equally well for inclusions with
densities below 0.8 g/cm3 (Fig. 14). An electronic spreadsheet for the calculation of CO2
densities and isochores from homogenization temperatures using SP94 is available from the
authors upon request.
Trapping temperature. For a pressure to be calculated the temperature at the time
of inclusion trapping (Ttrap) must be known. For phenocrysts Ttrap can be obtained by an
independent method such as mineral-melt equilibrium (e.g., Putirka 2008 and references
therein), homogenization temperature of melt inclusions, or estimation based on melt
composition. Eruption temperatures of lavas are generally a good proxy for Ttrap since the magma
temperature is unlikely to have changed significantly during the final ascent. Since ∆P/∆Τ of
CO2 isochores is rather low at magmatic conditions and slightly increases with increasing fluid
density (Fig. 15), calculated pressures are relatively insensitive to uncertainties in temperature
determinations (Roedder 1965, 1983). For example, a choice of 1150 rather than 1200 °C
results in a calculated pressure error of 2 MPa (3.8%) for an inclusion density of 0.2 g/cm3, 12
MPa (4%) for 0.6 g/cm3, and 34 MPa (3.1%) for 1.1 g/cm3. Even a gross error in estimation of
Ttrap has a limited effect on calculated pressures as shown by a comparison of panels a) and c)
in Figure 16. It can be concluded that, at least for phenocrysts, the determination of Ttrap is not
a critical issue in fluid inclusion barometry.
For crustal and mantle xenoliths the determination of Ttrap is less straightforward because
a xenolith’s history before entrainment into the host magma, and the relation between xenolith
entrainment and fluid entrapment, cannot always be established. Fluid and melt inclusions
can be trapped long before xenolith entrainment, concomitant (and in some cases causally
related) to entrainment, or during subsequent xenolith transport to the surface. In many cases
the temperature of the host magma may be a good choice for Ttrap because heating of a xenolith
occurs rather rapidly: For a representative thermal diffusivity of 10−6 m2/sec (Robertson 1988)
the center of a spherical xenolith 10 cm in diameter reaches 90% of the surrounding melt
1.1
1000
1.0
800
Pressure (MPa)
0.9
Inclusion
600
0.8
Host magma
0.7
400
0.6
0.5
200 0.4
0.3
0.2
0
700 800 900 1000 1100 1200 1300
Temperature (°C)
Figure 15. Isochores for pure CO2 calculated using the EOS of Sterner and Pitzer (1994). Numbers at
isochores indicate fluid density in g/cm3. Dashed lines illustrate the strongly increasing pressure difference
between an adiabatically ascending basaltic magma and an isochoric fluid inclusion trapped at P-T
conditionsFigure
indicated15by the circle.
164 Hansteen & Klügel
temperature within 12 minutes (cf. Fig.
30 (a) 29 of Carslaw and Jaeger 1959). zanon
SP94 et al. (2003) show that the combination
25
(pure CO2)
of geothermometers for the host magmas
Frequency
20
and homogenization temperatures of melt
15
inclusions in xenolith minerals yield a
10
consistent and plausible range for Ttrap.
5
0
For some xenolith types such as
mantle peridotites the local ambient
30 (b) temperature can be determined by mineral
KJ81
25
(pure CO2) geothermometers, which, however, are
Frequency
20 slow in adapting to temperature changes
15 because chemical diffusivities in minerals
10 are orders of magnitude smaller than
5 thermal diffusivities. As a consequence,
0 Ttrap may actually be close to the host
magma temperature even though xenolith
30 (c)
25
SP94 minerals (and their cores in particular)
(950 °C, pure CO2) indicate much lower values. The same
Frequency
20
holds for a regional geotherm when claimed
15
to represent the P-T conditions of trapped
10
fluids (e.g., Andersen and Neumann
5 2001). The actual Ttrap can be anywhere
0 between the regional geotherm and magma
30 (d) temperatures because xenoliths represent
KJ81 the vicinity of conduits and magma holding
25
(0.9 CO2 + 0.1 H2O)
reservoirs and consequently were entrained
Frequency
20
15
at elevated temperatures. This view is based
on the frequently observed abundance of
10
deep-seated xenoliths in late rather than
5
early lavas of an eruption, which plausibly
0
0 200 400 600 800 1000 1200 reflects massive wall-rock collapse when
Pressure [MPa] magmatic overpressure declines (Cox et al.
1993; Klügel et al. 1999). The interpretation
Figure 16. Effect of changing calculation parameters of xenoliths as biased samples overprinted
Figure 16
on pressure distributions inferred for CO2 fluid
inclusions in peridotite xenoliths from Gran Canaria. by magmatism is also consistent with the
The calculations use measured densities from Hansteen penetration or infiltration of xenoliths by
et al. (1998), a model temperature of 1150 °C except basaltic melts often observed or with the
for panel c), and the EOS of Sterner and Pitzer (1994) association of crustal or mantle xenoliths
and Kerrick and Jacobs (1981) indicated as SP94 and with cumulates (e.g., Kempton 1987;
KJ81, respectively. (a) Pressure distribution for SP94.
(b) KJ81 yields a similar pressure distribution as SP94 Wulff-Pedersen et al. 1996; Klügel 1998;
except for higher pressures where KJ81 overestimates Sachs and Hansteen 2000; Shaw et al.
resulting in a larger data spread. (c) Decreasing the 2006). A regional geotherm is therefore
model temperature by as much as 200 °C causes a not an adequate measure for Ttrap of fluid
comparatively mild decrease of calculated pressures inclusions in a volcanic region where local
that increases with increasing inclusion density
(compare with panel a). (d) Data for an assumed temperatures vary strongly on a kilometer-
original fluid composition of 90 mol% CO2 + 10 or meter-scale. Such local variations
mol% H2O, which was accounted for by increasing can also be reflected in a wide range of
the measured density by 4.5% and by specifying temperature estimates for mantle xenoliths
the CO2-H2O system in KJ81 used to calculate the from a single locality (e.g., Frezzotti et al.
isochores. See text for details.
2002), which together with the host magma
Fluid Inclusion Thermobarometry: Tracer for Magmatic Processes 165
temperature provide brackets for Ttrap rather than a single value.
Calculation of depth. The derivation of depth from the calculated pressures appears to be
a trivial issue at first sight, but it requires some caution because it is not always clear whether
the derived pressure is hydrostatic or lithostatic. The local pressure at which a fluid is trapped
is in principle a fluid pressure and hence not lithostatic, thus its relation to the height of the
overlying rock mass is ambiguous. The fluid pressure during trapping can be equal to the
lithostatic pressure or it can be lower if an overlying column of magma rather than rock needs to
be considered. In this case further complications can arise because degassing and fragmentation
of an erupting magma affect the fluid pressure at depth. The fluid pressure can also exceed
the lithostatic pressure by up to a few tens of MPa, such as in the case of an overpressurized
magma chamber. Roedder and Bodnar (1980) discuss a variety of pressure conditions for fluid
entrapment and suggest that pressures are probably between lithostatic and hydrostatic in most
natural situations. As an example, assumption of a lithostatic column (lithostatic pressure) with
an average rock density ρ = 3.3 g/cm3 as in the upper mantle will yield a 18% lower depth than
assumption of a hydrostatic column (hydrostatic pressure) with a magma density ρmelt = 2.7 g/
cm3 (Roedder 1983); for continental crust this difference becomes smaller. Remarkably, at high
pressures, the uncertainty of the seemingly simple depth calculation may be larger than the
uncertainties in determining density and pressure as discussed above!
the role of possibly missing h2O
The predominance of pure CO2 fluid inclusions in basalt-hosted xenoliths and phenocrysts
does not necessarily preclude the earlier presence of H2O in the trapped fluid. To the contrary,
a significant amount of H2O in the fluid phase at depth is indicated e.g., by degassing basalts
(reviewed by e.g., Metrich and Wallace 2008) or by xenoliths with fluid inclusions being
associated with hydrous mineral phases (Sachs and Hansteen 2000). Much of the trapped H2O
can be lost, however, by compositional re-equilibration (Bodnar 2003) involving 1) reactions
and 2) preferred leakage and diffusion. First, H2O in a mixed CO2-H2O inclusion may become
removed through reaction between fluid and inclusion wall to produce hydrous minerals and
carbonate (Andersen et al. 1984; Kleinefeld and Bakker 2002); in this case the observed pure
CO2 is actually a residual fluid. Occurrences of such reaction products in small amounts are
often hard to recognize by visual observation. Second, and more significant, mixed CO2-
H2O inclusions show preferential leakage of H2O and thus a differentiation of supercritical
fluid upon changing ambient conditions. Synthetic mixed inclusions in quartz that were re-
equilibrated at reduced pressures and T between 560 and 825 °C showed a significant decrease
in XH2O within days to weeks (Bakker and Jansen 1991; Hall and Sterner 1993). Here, non-
decrepitative preferential H2O leakage is thought to occur mainly along crystal defects and
is probably enhanced by ductile deformation and recrystallization (Hollister 1990). Because
of different molecule sizes and wetting properties, H2O is more prone than CO2 to migration
along dislocations and planar defects whereas the latter is solely lost by hydrofractures
(Bakker and Jansen 1994). In addition, hydrogen can diffuse out of a quartz-hosted inclusion
within days (Mavrogenes and Bodnar 1994). In olivine, experiments revealed extremely fast
diffusion of water-derived hydrogen and hydroxyl, which may leave an inclusion devoid of
H2O within hours at magmatic temperatures (Mackwell et al. 1985; Mackwell and Kohlstedt
1990). The rapidity of diffusion indicates that it is governed by the inherent Fe3+/Fe2+ and
point defect abundance in olivine. Likewise, the very fast transfer of oxygen through olivine
as experimentally observed was explained by point-defect diffusion (Pasteris and Wanamaker
1988); therefore diffusion of water-derived species may be fast in other ferromagnesian
minerals as well. Diffusive loss and preferential H2O leakage are thus plausible explanations
for the absence of H2O in most fluid inclusions in mantle xenoliths.
For pressure determinations the problem of “missing H2O” in fluid inclusions boils down
to an estimation of their former H2O content and its adequate correction (cf. Sachs and Hansteen
166 Hansteen & Klügel
2000). If α is the molar H2O/CO2 ratio of a mixed CO2-H2O fluid trapped in an inclusion and
all H2O becomes subsequently lost, then for an isochoric system the original fluid density ρorig
is simply derived from the measured density of the H2O-free inclusion ρmeas by: ρorig = ρmeas ×
(1 + α × 18/44), i.e., corrected by the H2O/CO2 weight ratio. The isochore for the original fluid
is then calculated using the corrected density and an equation of state for the CO2-H2O system
(e.g., Kerrick and Jacobs 1981) with the respective H2O content given by α. The estimation
of the former H2O content, however, is notoriously difficult and in most cases confined to
an estimate of a plausible upper limit. Even for basalts where much data on CO2 and H2O
solubilities exist, estimation of the CO2/H2O ratio in the fluid phase (vapor) is hampered by
missing information on the H2O and CO2 contents in the melt at the time of fluid trapping and
on the conditions of degassing such as open vs. closed system. In general, the vapor phase
exsolved from basaltic melts at depth is strongly CO2-dominated until most of the CO2 is lost
and significant H2O exsolution occurs (Metrich and Wallace 2008). For MORB and intraplate
basalts at depths exceeding a few kilometers, a maximum H2O content in the fluid phase of 10
mol% is probably a conservative estimate in most cases (cf. Dixon and Stolper 1995); it may
be higher at subduction zones. For 10 mol% of H2O the correction factor for density is 1.045.
Using the Kerrick and Jacobs (1981) EOS for the CO2-H2O system the resulting corrections
for calculated pressures range from e.g., −7% for the 0.3 g/cm3 isochore to +23% for the 0.8 g/
cm3 isochore at 1150 °C. These pressure corrections show a strong non-linear dependency on
density and therefore strongly skewed calculated pressure distributions (Fig. 16d).
Volumetric re-equilibration during magma ascent: a case study
It has been discussed above that the healing of microcracks in basalt-hosted olivine occurs
within a few hours to days (Wanamaker et al. 1990). Therefore, if a magma stagnates during
its ascent for a period of more than about a day, even those inclusions showing petrographic
evidence for leakage can under favorable conditions yield the pressure of this stagnation. We
reiterate that it is not the density of a single decrepitated inclusion that provides the relevant
barometric information, but the clustering of many inclusions around a certain density.
However, the results can still be biased by stretching during subsequent magma ascent, as has
been calculated for olivine-hosted CO2 inclusions during isothermal decompression using a
power-law creep model (Wanamaker and Evans 1989). An extreme decompression by 3000
MPa within 6 hours (magma ascent from 100 km depth to the surface at 5 m/s) would increase
an inclusion’s radius by about 0.8% at 1100 °C and 9% at 1200 °C resulting in a density
decrease by 3% and 22%, respectively. Likewise, decompression by 1000 MPa within two days
(magma ascent from 33 km depth at 0.2 m/s) would result in a density decrease by 8% at 1100
°C and 30% at 1300 °C. The stretching rate generally increases as the overpressure of inclusions
increases, and the total amount of stretching for a fixed overpressure increases as the external
pressure decreases (Wanamaker and Evans 1989). Clearly, stretching may significantly reduce
inclusion densities in xenoliths and phenocrysts even during short ascent times, resulting in a
wide range of homogenization temperatures as is often observed.
A case in point for volumetric re-equilibration is shown by CO2 inclusions in basanite-hosted
xenoliths and phenocrysts from the 1949 eruption on La Palma, Canary Islands (Hansteen et al.
1998; Klügel et al. 2000). Inclusions in tholeiitic gabbro xenoliths that show ample evidence for
magmatic overprinting indicate a dominant pressure range of 270-330 MPa reflecting the lower
oceanic crust where the gabbros originated. This density range overlaps perfectly with the ranges
indicated by inclusions in ultramafic cumulates (90% of values between 190 and 340 MPa),
peridotite mantle xenoliths (90% between 190 and 400 MPa), and clinopyroxene phenocrysts
(Figs. 11 and 17). Remarkably, almost none of the investigated primary and secondary inclusions
in mantle xenoliths showed mantle pressures. This was attributed to storage of the xenoliths
in the lower crust for a long period of time, probably tens of years before eruption, where
almost complete re-equilibration of inclusions and interactions between xenoliths and mafic
magma occurred. Some re-equilibration did also occur during the final xenolith uplift, which
Fluid Inclusion Thermobarometry: Tracer for Magmatic Processes 167
100
80 (a)
60
Th(LV→L)
40
Frequency
20
0
4 Th(LV→V)
6
8
5 9 13 17 21 25 29 33
Th (°C)
100 (b) Gabbro xenoliths
Cumulate xenoliths
80 Peridotite xenoliths
Phenocrysts
Frequency
60 Moho Cpx-melt
barom.
40
20
0
0 200 400 600 800
Pressure (MPa)
Figure 17. Distribution of (a) homogenization temperatures and (b) calculated pressures indicated by fluid
inclusions in basanite-hosted xenoliths and phenocrysts from the 1949 eruption on La Palma (Canary
Islands). The diagrams are based on data of Hansteen et al. (1998) recalculated with the EOS of Sterner and
Pitzer (1994). A well-defined maximum between about 200 and 350 MPa is reflected in both crustal and
mantle xenoliths as well as clinopyroxene phenocrysts, and is consistent with the depth of the lower oceanic
crust where MORB Figure 17
gabbro xenoliths originate. Mantle pressures (above ca. 450 MPa) are indicated almost
exclusively by olivine phenocrysts and less so by peridotites, which is taken as evidence for prolonged
storage of mantle xenoliths in the lower crust. The pressure range indicated by fluid inclusions in olivine
phenocrysts overlaps with that indicated by mineral-melt barometry of clinopyroxene phenocrysts.
took only a few hours (Klügel 1998), but its effect was limited because the frequency maximum
of the pressure distribution shown in Figure 17 overlaps perfectly with the inferred pressure of
origin of the gabbro xenoliths. The basanite carrying the xenoliths to the surface ascended from
mantle depths and stagnated in the lower crust only briefly (Klügel et al. 2000) so that CO2
inclusions in olivine phenocrysts could retain their original mantle pressures, which fit well
with pressures indicated by clinopyroxene-melt barometry. CO2 inclusions in clinopyroxene
phenocrysts, however, became re-equilibrated in the lower crust because this mineral is more
prone to leakage than olivine due to better cleavage. The results of this case study illustrate the
usefulness of fluid inclusion barometry of xenoliths and phenocrysts especially when combined
with mineral-melt barometry and geospeedometry.
168 Hansteen & Klügel
PreSSure InFOrMatIOn FrOM aqueOuS FLuIdS In ShaLLOW
MaGMa ChaMBerS and PLutOnIC rOCKS
Background
Fluids coexisting with crystallizing magmas exhibit a wide range of chemical compositions,
covering major parts of the ternary system H2O-NaCl-CO2 (Burnham 1979; Holloway and
Blank 1994; Lowenstern 1994; Dixon et al. 1995; Schmidt and Bodnar 2000; Kamenetsky et
al. 2002). Both magma composition and pressure exert major influence on fluid composition.
Mafic intraplate magmas at depth typically contain comparable amounts of dissolved CO2 and
H2O, which are exsolved upon ascent and crystallization (e.g., Metrich and Wallace 2008).
Mafic subduction zone magmas are, however, H2O dominated. Due to the higher solubility of
H2O than CO2 in silicate melts at high pressures, the first fluid phase formed after initial fluid
saturation in the uppermost mantle or lower to middle crust is CO2-rich. If the magmatic fluids
are progressively lost to the surroundings during crystallization and degassing in a shallow
reservoir, the fluids continuously released from the magma will evolve towards H2O-richer
compositions. As discussed above, however, an ascending melt can produce CO2-dominated
fluids even in the upper crust, provided that either the initial water contents are low or that
crustal ascent has been rapid and stagnation times limited.
In consequence, felsic magmas usually contain comparatively high amounts of dissolved
water. At high temperatures and moderate to low pressures, both a high-salinity liquid and a
vapor phase may coexist with silicate melt (this phenomenon is described above for the system
H2O-NaCl). Such fluid phase separation is common in the plutonic environment because a
wide range of possible magmatic fluid compositions fall within the immiscibility fields for
the systems H2O-NaCl and H2O-NaCl-CO2 at low pressures (Bodnar et al. 1985; Chou 1987).
It should be stressed that the resulting major element compositions of the liquid and vapor
phases are consequently governed by the boundaries of the immiscibility field rather than by
the volatile contents of the melt.
In the following, we give examples of fluid inclusions representing the magmatic fluid
phase in shallow magma chambers prior to or during volcanic eruptions, and those occurring
in a variety of plutonic rocks. Both scenarios involve residual fluids exsolved from magmas at
shallow levels, where the combined parameters pressure, temperature and fluid composition
control whether fluid immiscibility (“boiling”) occurs. The degassing rate, which is important
for wall rock fracturing and eventually for the hazard potential of a shallow magma chamber,
is intimately linked with the pressure-dependent degree of fluid oversaturation.
Fluids in shallow magma chambers
The number of fluid inclusion studies from volcanic rocks seems to correlate with magma
chamber depth. The main reasons are probably the scarcity of fluid inclusions with adequate
sizes in phenocrysts formed at shallow levels, and the relative difficulty of observing phase
changes in low-density (i.e., vapor-rich) fluid inclusions.
Primary magmatic fluids coexisting with rhyolitic melt inclusions in quartz from igneous
complexes in New zealand and Chile have been interpreted as shallow-level magma chamber
paragenesis (Davidson and Kamenetsky 2007). The saline fluids coexisting with the melts are
comparatively enriched in halogens (chlorine and bromine). Lowenstern (1994) demonstrated
immiscibility among vapor, highly saline liquid, and silicate melt during pre-eruptive
crystallization of peralkaline rhyolites from Pantelleria, Italy. The Cl contents of Pantellerian
rhyolites indicate equilibration at pressures between 50 and 100 MPa. Lava samples erupted
from Mt. Unzen, Japan, in June 1991, contain abundant, complexly zoned plagioclase (Browne
et al. 2006, and references therein), which contain coexisting rhyolitic melt inclusions and
fluid inclusions (Hansteen, unpubl. data). Fluid inclusions are two-phase (about 3% liquid and
Fluid Inclusion Thermobarometry: Tracer for Magmatic Processes 169
97% vapor) at room temperature, and are well represented by the system H2O-NaCl-CO2. The
aqueous phase in the fluid inclusions has a salinity of 21.5 ± 1 wt% NaCl equivalents. The
density of the CO2-dominated phase was estimated at 0.02 g/cm3, giving a total fluid inclusion
density of about 0.05 g/cm3 when a density of 1.15 g/cm3 is assumed for the aqueous phase.
Using a post-mixing magma temperature of 900 °C, the magmatic fluid inclusions were formed
at a pressure of 35±15 MPa (1.5±0.5 km) (equation of state of Brown and Lamb 1989).
xenoliths interpreted as wall-rock fragments from shallow magma reservoirs have
successfully been used to estimate magma stagnation depths in both mafic (see above) and
silicic systems. From the last La Fossa eruption of Vulcano Island (1888-90), Clocchiatti et
al. (1994) described secondary H2O-NaCl-CO2 inclusions in metamorphic xenoliths giving
equilibration pressures of 30-60 MPa, which translates into a location of the higher portions
of the chamber at around 1500-2000 m depth (cf. Fig. 11). One cumulate xenolith from the
primitive basaltic 1999 eruption of Cerro Negro, Nicaragua, contains abundant water-rich fluid
inclusions with a salinity of 3.7 wt% NaCl equivalents (Hansteen and Weiss, unpubl. data). No
CO2 could be detected in the fluid inclusions, demonstrating the water-rich composition of the
residual magmatic fluid. Inclusion densities correspond to a magma chamber pressure of 100
±16 MPa. Compositional zoning on a cm-scale even within single glassy clasts demonstates
late-stage magma mingling during this eruption. The two compositional endmembers give
pressures of 275 ±25 MPa and 100 ± 20 MPa respectively, using the clinopyroxene-melt
thermobarometer of Putirka et al. (1996), which testifies to a two-level magma chamber
system at Cerro Negro. Based on volatile contents in melt inclusions, Roggensack et al. (1997)
similarly estimated fractionation pressures of about 300 MPa for the 1992 eruption products,
overlapping with our cpx-melt thermobarometry.
Summing up, fluid inclusions in phenocrysts and xenoliths have successfully been used
to demonstrate shallow-level magma stagnation between 30 and 100 MPa for both basaltic
and silicic volcanic systems. The composition of the fluids were found to be variable but well
represented by the system H2O-NaCl-CO2.
Fluid inclusions in plutonic rocks
Plutonic rocks typically contain abundant fluid inclusions representing various stages of
their evolution (e.g., Roedder and Coombs 1967; Konnerup-Madsen 1979; Weisbrod 1981).
Given the time available for differentiation processes, P-T variations, interaction with external
fluids of changing compositions and chemical re-equilibration of mineral paragenesis, fluid
inclusion records in plutonic rocks may be hard to decipher. Early inclusion assemblages may
also have partly or fully re-equilibrated at later stages (see above). Nevertheless, successful
fluid inclusion studies have been performed on both xenolith samples and in-situ samples from
a considerable number of plutonic rocks. Fluid inclusion studies are also an important tool in
pegmatite research, a topic which is beyond the scope of this paper; the interested reader is
referred to reviews by London (1996) and Thomas et al. (2006). Primary fluids exsolving from
felsic magma and eventually evolving into hydrothermal fluids is a topic of long-standing
interest, which has recently been highlighted by studies of subvolcanic and plutonic rocks
(Kamenetsky et al. 2004; Davidson and Kamenetsky 2007), showing a continuous evolution
from magmatic to hydrothermal fluids.
Fluid inclusions in deep-seated intrusions generally have a significant component of CO2,
probably related to its elevated solubility in silicate melts at high pressures. Konnerup-Madsen
(1979) interpreted CO2-dominated and CO2-H2O inclusions in quartz from the Precambrian
Farsund plutonic complex, south Norway, to represent magmatic fluids indicating an origin at
0.5 to 0.6 GPa. Later inclusions related to uplift were saline (2 to 60 wt% NaCl equivalents),
and contained significant amounts of the cations K+, Ca2+ and Mg2+. Frost and Touret (1989)
investigated fluid inclusions in quartz from the Sybille Monzosyenite, Laramie Anorthosite
170 Hansteen & Klügel
Complex, Wyoming, reporting that the primary magmatic fluid consisted predominantly of
CO2, which exsolved at 0.3 GPa and 950-1000 °C, confirming formation conditions inferred
from mineralogic data. Halite and sylvite crystals, which occur as part of the magmatic mineral
paragenesis trapped as inclusions in feldspar, were interpreted as remnants of immiscible
hydrosaline melts, that may have coexisted with the CO2-rich fluid. Frezzotti et al. (1994)
also reported possible brine-CO2 immiscibility at magmatic conditions of 0.3 GPa and 750 °C
for peraluminous graphite-bearing leucogranites from Deep Freeze Range (northern Victoria
Land, Antarctica).
In a classical study of peralkaline syenitic and granitic xenoliths from Ascencion Island,
Harris (1986) reported coexisting silicate melt inclusions and low-salinity aqueous inclusions,
and additionally silicate melt coexisting with both a high-salinity (> 40 wt% NaCl equivalents)
aqueous liquid phase and a CO2-rich vapor phase. Harris (1986) interpreted the homogeneous
low-salinity fluids as the original magmatic composition, and the coexisting brines and CO2-
rich vapor inclusions to represent the conditions after phase separation at 100 to 200 MPa.
An example of repeated fluid boiling in a subvolcanic granite is represented by the Permian
Eikeren alkali granite, Oslo rift (Hansteen and Burke 1990). Fluid evolution during cooling of
the granite can be summarized as four stages: a) boiling during late crystallization at about 700
°C; b) near-isobaric cooling and closed system fluid evolution down to about 450 °C; c) repeat-
ed boiling at 412 ± 51°C (Fig. 18), leading to opening of the fluid system to external fluids, and
d) rapid dilution by meteoric waters below 400 °C. Because both pressure and temperature can
be obtained by microthermometry of
boiling NaCl-H2O solutions (see above;
Figs. 1f and 1g), the two boiling events
24
at about 700 and 412 °C act as reference
points for the proposed cooling path. 20
Boiling under magmatic conditions in- Tm (Halite)
volved a liquid phase with a salinity of 16
48 ± 7 wt% NaCl equivalents, giving
a pressure of about 95 MPa at 700 °C. 12 n = 198
Boiling at 412 °C represents a liquid
8
phase with 47 ± 5 wt% NaCl equiva- Th (LV-L)
lents, giving a pressure of 25 ± 6 MPa, 4
interpreted as hydrostatic conditions.
Sub-solidus fluids in the temperature
interval between the two boiling events
did not coexist with a vapor phase, indi- 4
Th (LV-V)
cating pressures well above hydrostatic
8
conditions. Fluid immiscibility in mag-
matic systems has been described in de- 12
tail by Roedder (1992).
16
Highly peralkaline (agpaitic)
300 400 500
magmas have higher solubilities for
T (°C)
water and halogens than other magmas,
and thus exhibit different degassing Figure 18. Boiling event at 412 ± 51 °C for the subvolcanic
behavior. Konnerup-Madsen and Rose- Eikeren Granite, Permian Oslo Rift, Norway. Pressure
Hansen (1982) investigated magmatic and temperature can be simultaneously estimated since
H2O-CO2-CH4 fluid inclusions in the both a highly saline liquid and a low-salinity vapor phase
coexist as separate inclusions. Note that homogenization
agapaitic Ilímaussaq intrusion and Figure 18
into liquid (Th(LV→L)) and into vapor (Th(LV→V))
alkaline Gardar granitic complexes occur at the same temperature, a necessary observation
(south Greenland), and found that the for the identification of boiling.
Fluid Inclusion Thermobarometry: Tracer for Magmatic Processes 171
more alkalic magmas exsolved the more CH4-rich fluids. Krumrei et al. (2007) demonstrated
that hydrocarbons in fluid inclusions in the Ilímaussaq intrusion are of magmatic origin.
Summing up, fluid inclusions in a variety of mafic and silicic intrusive rocks indicate a
pressure-dependent influence on fluid compositions, where the deeper magmas have elevated
CO2 contents. Agpaitic magmas may additionally contain hydrocarbons of magmatic origin.
Fluid immiscibility resulting in the formation of a high-salinity liquid phase and a low-density
vapor phase is common in plutonites covering pressures up to 0.6 GPa.
SuMMary
Fluid inclusions can occur in any igneous mineral. In order to relate such inclusions to
magmatic processes, their time of formation relative to host mineral growth must be considered.
By petrographic investigations of doubly polished sections fluid inclusions assemblages are
determined, and in the ideal case related to mineral paragenesis. Early and late generations
of fluid inclusions can often be visually determined according to their occurrence in primary
growth zones or in later, secondary trails.
In order to obtain thermobarometric information from fluid inclusions, their chemical
composition and density have to be determined. Microthermometry is still a primary tool
to provide the densities and also qualitative to semi-quantitative chemical information of
single fluid inclusions, especially in the system H2O-CO2-NaCl. For more complex chemical
compositions, additional methods like Laser Raman microprobe and micro-FTIR should be
applied in order to characterize the systems.
Within a magma plumbing system of a volcanic rock, magma stagnation typically
occurs at discrete levels. During ascent of a magma, increasing pressure gradients between
fluid inclusions in phenocrysts and xenoliths and the host magma cause volumetric (leakage,
stretching) and compositional (e.g., loss of H2O) re-equilibration of the inclusions, and
additionally new inclusions may form. The resulting density distribution of fluid inclusions
depends on the ascent rates and stagnation periods, and may indicate the pressures of formation
and/or re-equilibration as discrete frequency maxima. These pressures are readily determined
from calculated isochores if the respective temperatures are known or can be estimated.
In shallow magma chambers and plutonic rocks, magmatic processes are often reflected by
several generations of fluid inclusions, allowing for detailed investigations of fluid evolution.
Phase separation involving immiscibility of a low-density vapor phase and a higher-density
liquid phase at moderate pressures may lead to the formation of coexisting vapor-rich and
water-rich fluid inclusions. By determining densities and compositions of both inclusion types,
pressure and temperature can be simultaneously determined.
OutLOOK
Integrated studies involving both silicate melt and fluid inclusions combined with additional
petrological and geophysical data will increasingly be the way to go for igneous petrologists
and volcanologists. For the former, a combination of melt and fluid inclusion geochemistry
will eventually provide unprecedented insights into deep-seated magmatic processes, although
the analytical effort may be high. Considering the limited number of microthermometric
studies of volcanic rocks published so far, a wealth of basic information is still to be obtained
on magma plumbing systems from all tectonic settings. For volcanologists, fluid inclusions are
the prime source for barometric information pertaining to shallow magma reservoirs of active
volcanoes and to short-duration stagnation of magma immediately prior to an eruption. Basic
information like CO2/H2O ratios in the degassing fluids, important for volcano dynamics and
172 Hansteen & Klügel
eruptive behavior, are an intergral part of the microthermometric data. Fluid inclusion data
are thus an invaluable tool to understand the past behavior of a volcano and, in conjunction
with geophysical and geochemical monitoring, contribute to the understanding of a volcano’s
plumbing system, which is a prerequisite for hazard mitigation research (cf. Peccerillo et
al. 2006). Thus the combination of efficient low-cost methods like detailed petrography and
microthermometry is still highly relevant in volcanology and petrology.
In the plutonic environment, fluid inclusions are typically larger and more abundant than
in their eruptive siblings, offering one explanation for the large number of microthermometric
studies available. Sizeable inclusions and the common occurrence of high-salinity aqueous
fluids with high trace element contents make such rocks ideal for microbeam analysis of single
inclusions using e.g., laser ablation ICP-MS. In combination with detailed petrography and
microthermometry, geochemical fluid evolution trends can be reconstructed.
aCKnOWLedGMentS
The research of the authors over the past years has been supported by the German Science
Foundation (DFG) and the German Ministry for Education, Science, Research and Technology
(BMBF). We gratefully acknowledge the constructive reviews from Ronald Bakker, Maria Luce
Frezzotti, Keith Putirka and Paul Wallace that helped to improve the paper. This publication
is contribution no. 165 of the Collaborative Research Center 574 “Volatiles and Hazards in
Subduction zones” at Kiel University.
reFerenCeS
Abramson EH, Brown JM (2004) The melting curve of carbon dioxide and implication for the fluid. EOS
Trans, Am Geophys Union 85(47):V41A-1358
Andersen T, Neumann ER (2001) Fluid inclusions in mantle xenoliths. Lithos 55:301-320
Andersen T, O’Reilly SY, Griffin WL (1984) The trapped fluid phase in upper mantle xenoliths from Victoria,
Australia: implications for mantle metasomatism. Contrib Mineral Petrol 88:72-85
Angus S, Armstrong B, de Reuck KM, Altunin VV, Gadetskii OG, Chapela GA, Rowlinson JS (1976)
International Tables of the Fluid State, Vol 3, Carbon dioxide. Pergamon Press, Oxford
Bakker RJ (2003) Package FLUIDS 1. Computer programs for analysis of fluid inclusion data and for modelling
bulk fluid properties. Chem Geol 194:3-23
Bakker RJ, Diamond LW (2006) Estimation of volume fractions of liquid and vapor phases in fluid inclusions,
and definition of inclusion shapes. Am Mineral 91:635-657
Bakker RJ, Jansen JBH (1991) Experimental post-entrapment water loss from synthetic CO2-H2O inclusions
in natural quartz. Geochim Cosmochim Acta 55:2215-2230
Bakker RJ, Jansen JBH (1994) A mechanism for preferential H2O leakage from fluid inclusions in quartz,
based on TEM observations. Contrib Mineral Petrol 116:7-20
Belkin HE, DeVivo B (1993) Fluid inclusion studies of ejected nodules from plinian eruptions of Mt. Somma-
Vesuvius. J Volcanol Geotherm Res 58:89-100
Blundy J, Cashman K (2005) Rapid decompression-driven crystallization recorded by melt inclusions from
Mount St Helens volcano. Geology 33:793-796
Bodnar RJ (2003) Re-equilibration of fluid inclusions. In: Fluid Inclusions: Analysis and Interpretation, Short
Course Series. Vol 32. Samson I, Anderson A, Marshall D (eds) Mineral Assoc Can, Vancouver, p 213-
231
Bodnar RJ, Burnham CW, Sterner SM (1985) Synthetic fluid inclusions in natural quartz. III. Determination of
phase equilibrium properties in the system H2O-NaCl to 1000 °C and 1500 bars. Geochim Cosmochim
Acta 49:1861-1873
Bottinga Y, Richet P (1981) High pressure and temperature equation of state and calculation of the
thermodynamic properties of gaseous carbon dioxide. Am J Sci 281:615-660
Bowers TS, Helgeson HC (1983a) Calculation of the thermodynamic and geochemical consequences of
nonideal mixing in the system H2O-CO2-NaCl on phase relations in geologic systems: Equation of state
for H2O-CO2-NaCl fluids at high pressures and temperatures. Geochim Cosmochim Acta 47:1247-1275
Fluid Inclusion Thermobarometry: Tracer for Magmatic Processes 173
Bowers TS, Helgeson HC (1983b) Calculation of the thermodynamic and geochemical consequences of
nonideal mixing in the system H2O-CO2-NaCl on phase relations in geologic systems: Metamorphic
equilibria at high pressures and temperatures. Am Mineral 68:1059-1075
Brown PE (1989) FLINCOR: A microcomputer program for the reduction and investigation of fluid-inclusion
data. Am Mineral 74:1390-1393
Brown PE, Lamb WM (1989) P-V-T properties of fluids in the system H2O±CO2±NaCl: New graphical
presentations and implications for fluid inclusion studies. Geochim Cosmochim Acta 53:1209-1221
Browne BL, Eichelberger JC, Patino LC, Vogel TA, Uto K, Hoshizumi H (2006) Magma mingling as indicated
by texture and Sr/Ba ratios of plagioclase phenocrysts from Unzen volcano, SW Japan. J Volcanol
Geotherm Res 154:103-116
Bureau H, Métrich N, Pineau F, Semet MP (1998) Magma-conduit interaction at Piton de la Fournaise volcano
(Réunion Island): a melt and fluid inclusion study. J Volcanol Geotherm Res 84:39-60
Burke EAJ (2001) Raman microspectrometry of fluid inclusions. Lithos 55:139-158
Burnham CW (1979) Magmas and hydrothermal fluids. In: Geochemistry of hydrothermal ore deposits. 2nd
ed. Barnes HL (ed) J. Wiley & Sons, New York, p. 71-136
Burruss RC (1981) Analysis of phase equilibria in C-O-H-S fluid inclusions. In: Short course in fluid inclusions:
applications to petrology. Vol 6. Hollister LS, Crawford ML (eds), Mineral Assoc Can, Vancouver, p 101-
137
Carslaw HS, Jaeger JC (1959) Conduction of Heat in Solids. Oxford University Press, Oxford
Cauzid J, Philippot P, Somogyi A, Ménez B, Simionovici A, Bleuet P (2006) Standardless quantification of
single fluid inclusions using synchrotron radiation induced x-ray fluorescence. Chem Geol 227:165-183
Chou I-M (1987) Phase relations in the system NaCl-KCl-H2O. III: Solubilities of halite in vapour saturated
liquids above 445 C and redetermination of phase equilibrium properties in the system NaCl-H2O to
1000 C and 1500 bars. Geochim Cosmochim Acta 51:1965-1975
Clocchiatti R, Del Moro A, Gioncada A, Joron JL, Mosbah M, Pinarelli L, Sbrana A (1994) Assessment of a
shallow magmatic system: the 1888-90 eruption, Vulcano Island, Italy. Bull Volcanol 56:466-486
Clynne MA, Potter RW II (1977) Freezing point depression of synthetic brines (abst). Geol Soc Am Abstr
Programs 9, 930
Coombs ML, Gardner JE (1991) Shallow-storage conditions for the rhyolite of the 1912 eruption at Novarupta,
Alaska. Geology 29:775-778
Cox KG, Charnley N, Gill RCO, Parish KA (1993) Alkali basalts from Shuqra, Yemen: magmas generated in
the crust-mantle transition zone? In: Magmatic Processes and Plate Tectonics, Geol Soc Spec Publ 76.
Prichard HM, Alabaster T, Harris NBW, Neary CR (eds), p 443-453
Crawford ML (1981) Phase equilibria in aqueous fluid inclusions. In: Short course in fluid inclusions:
applications to petrology. Vol 6. Hollister LS, Crawford ML (eds), Mineral Assoc Can, Vancouver, p
75-100
Davidson P, Kamenetsky VS (2007) Primary aqueous fluids in rhyolitic magmas: Melt inclusion evidence for
pre- and post-trapping exsolution. Chem Geol 237:372-383
Diamond LW (2003) Systematics of H2O inclusions. In: Fluid Inclusions: Analysis and Interpretation, Short
Course Series. Vol 32. Samson I, Anderson A, Marshall D (eds) Mineral Assoc Can, Vancouver, p 55-78
Dixon JE, Stolper EM (1995) An experimental study of water and carbon dioxide solubilities in mid-ocean
ridge basaltic liquids; Part II, Applications to degassing. J Petrol 36:1633-1646
Dixon JE, Stolper EM, Holloway JR (1995) An experimental study of water and carbon dioxide solubilities in
mid-ocean ridge basaltic liquids. Part I: calibration and solubility models. J Petrol 36:1607-1631
Driesner T, Heinrich CA (2007) The system H2O–NaCl. Part I: Correlation formulae for phase relations in
temperature–pressure–composition space from 0 to 1000 °C, 0 to 5000 bar, and 0 to 1 xNaCl. Geochim
Cosmochim Acta 71:4880-4901
Frezzotti ML, Andersen T, Neumann ER, Simonsen SL (2002) Carbonatite melt-CO2 fluid inclusions in mantle
xenoliths from Tenerife, Canary Islands: a story of trapping, immiscibility and fluid–rock interaction in
the upper mantle. Lithos 64:77-96
Frezzotti ML, Di Vincenzo G, Ghezzo C, Burke EAJ (1994) Evidence of magmatic CO2-rich fluids in
peraluminous graphite-bearing leucogranites from Deep Freeze Range (northern Victoria Land,
Antarctica). Contrib Mineral Petrol 117:111-123
Frezzotti ML, Peccerillo A (2004) Fluid inclusion and petrological studies elucidate reconstruction of magma
conduits. Eos Trans Am Geophys Union 85:157-163
Frost BR, Touret JLR (1989) Magmatic CO2 and saline melts from the Sybille Monzosyenite, Laramie
Anorthosite complex, Wyoming. Contrib Mineral Petrol 103:l78-186
Fulignati P, Marianelli P, Metrich N, Santacroce R, Sbrana A (2004) Towards a reconstruction of the magmatic
feeding system of the 1944 eruption of Mt Vesuvius. J Volcanol Geotherm Res 133:13-22
174 Hansteen & Klügel
Galipp K, Klügel A, Hansteen TH (2006) Changing depths of magma fractionation and stagnation during the
evolution of an oceanic island volcano: La Palma (Canary Islands). J Volcanol Geotherm Res 155:285-
306
Goldstein RH, Reynolds TJ (1994) Systematics of fluid inclusions in diagenetic minerals. SEPM Short Course
31, Tulsa, 199 pp
Hall DL, Sterner SM (1993) Preferential water loss from synthetic fluid inclusions. Contrib Mineral Petrol
114:489-500
Hall DL, Sterner SM, Bodnar RJ (1988) Freezing point depression of NaCl-KCl-H2O solutions. Econ Geol
83:197-202
Hansteen TH (1991) Multi-stage evolution of the picritic Mælifell rocks, SW Iceland: A mineralogical and
glass inclusion study. Contrib Mineral Petrol 109:225-239
Hansteen TH, Andersen T, Neumann ER, Jelsma H (1991) Fluid and silicate glass inclusions in ultramafic and
mafic xenoliths from Hierro, Canary Islands: implications for mantle metasomatism. Contrib Mineral
Petrol 107:242-254
Hansteen TH, Burke EAJ (1990) Melt-mineral-fluid interaction in peralkaline silicic intrusions in the Oslo
rift, Southeast Norway. II: High-temperature fluid inclusions in the Eikeren-Skrim complex. Nor Geol
Unders Bull 417:15-32
Hansteen TH, Klügel A, Schmincke HU (1998) Multi-stage magma ascent beneath the Canary Islands:
evidence from fluid inclusions. Contrib Mineral Petrol 132:48-64
Harris C (1986) A quantitative study of magmatic inclusions in the plutonic ejecta of Ascension Island. J Petrol
27:251-276
Hollister LS (1990) Enrichment of CO2 in fluid inclusions in quartz by removal of H2O during crystal-plastic
deformation. J Struct Geol 12:895-901
Holloway JR (1981) Compositions and volumes of supercritical fluids in the Earth”s crust. In: Short course in
fluid inclusions: applications to petrology. Vol 6. Hollister LS, Crawford ML (eds), Mineral Assoc Can,
Vancouver, p 13-38
Holloway JR, Blank JG (1994) Application of experimental results to C-O-H species in natural melts. Rev
Mineral Geochem 30:187-230
Kamenetsky VS, Naumov VB, Davidson P, van Achterbergh E, Ryan CG (2004) Immiscibility between silicate
magmas and aqueous fluids: a melt inclusion pursuit into the magmatic-hydrothermal transition in the
Omsukchan Granite (NE Russia). Chem Geol 210:73-90
Kamenetsky VS, van Achterbergh E, Ryan CG, Naumov VB, Menagh TP, Davidson P (2002) Extreme chemical
heterogeneity of granite-derived hydrothermal fluids: an example from inclusions in a single crystal of
miarolitic quartz. Geology 30:459-462
Kempton PD (1987) Mineralogic and geochemical evidence for differing styles of metasomatism in spinel
lherzolite xenoliths: Enriched mantle source regions of basalts? In: Mantle metasomatism. Vol 1. Menzies
MA, Hawkesworth CJ (eds) Academic Press, London, p 45-89
Kent AJR (2008) Melt inclusions in basaltic and related volcanic rocks. Rev Mineral Geochem 69:273-331
Kerrick DM, Jacobs GK (1981) A modified Redlich-Kwong equation for H2O, CO2 and H2O-CO2 mixtures at
elevated temperatures and pressures. Am J Sci 281:735-767
Kleinefeld B, Bakker RJ (2002) Fluid inclusions as microchemical systems: evidence and modelling of fluid-
host interactions in plagioclase. J Metamorphic Geol 20:845-858
Klügel A (1998) Reactions between mantle xenoliths and host magma beneath La Palma (Canary Islands):
constraints on magma ascent rates and crustal reservoirs. Contrib Mineral Petrol 131:237-257
Klügel A, Hansteen TH, Galipp K (2005) Magma storage and underplating beneath Cumbre Vieja volcano, La
Palma (Canary Islands). Earth Planet Sci Lett 236:211-226
Klügel A, Hansteen TH, Schmincke HU (1997) Rates of magma ascent and depths of magma reservoirs beneath
La Palma (Canary Islands). Terra Nova 9:117-121
Klügel A, Hoernle KA, Schmincke HU, White JDL (2000) The chemically zoned 1949 eruption on La Palma
(Canary Islands): Petrologic evolution and magma supply dynamics of a rift-zone eruption. J Geophys
Res 105(B3):5997-6016
Klügel A, Klein F (2006) Complex magma storage and ascent at embryonic submarine volcanoes from Madeira
Archipelago. Geology 34:337-340
Klügel A, Schmincke HU, White JDL, Hoernle KA (1999) Chronology and volcanology of the 1949 multi-vent
rift-zone eruption on La Palma (Canary Islands). J Volcanol Geotherm Res 94:267-282
Konnerup-Madsen J (1995) Basic microthermometric observations on fluid inclusions in minerals – a working
manual. NorFA Short Course Notes, Copenhagen, pp 100, ISBN: 87-89980-12-3
Konnerup-Madsen J, Rose-Hansen J (1982) Volatiles associated with alkaline igneous rift activity: Fluid
inclusions in the Ilímaussaq intrusion and the Gardar granitic complexes (south Greenland). Chem Geol
37:79-93
Fluid Inclusion Thermobarometry: Tracer for Magmatic Processes 175
Konnerup-Madsen J (1979) Fluid inclusions in quartz from deep-seated granitic intrusions, South Norway.
Lithos 12:13-23
Kooi ME, Schouten JA, Van den Kerkhof AM, Istrate G, Althaus E (1998) The system CO2-N2 at high pressure
and applications to fluid inclusions. Geochim Cosmochim Acta 62:2837-2843
Krumrei TV, Pernicka E, Kaliwoda M, Markl G (2007) Volatiles in a peralkaline system: abiogenic hydrocarbons
and F-Cl-Br systematics in the naujaite of the Ilímaussaq intrusion, South Greenland. Lithos 95:298-314
Labotka TC (1991) Chemical and physical properties of fluids. Rev Mineral 26:43-104
London D (1996) Granitic pegmatites. GSA Special Paper 315: The Third Hutton Symposium on the Origin of
Granites and Related Rocks 315:305-319
Lowenstern JB (1994) Chlorine, fluid immiscibility and degassing in peralkaline magmas from Pantelleria,
Italy. Am Mineral 79:353-370
Mackwell SJ, Kohlstedt DL (1990) Diffusion of hydrogen in olivine: Implications for water in the mantle. J
Geophys Res 95:5079-5088
Mackwell SJ, Kohlstedt DL, Paterson MS (1985) The role of water in the deformation of olivine single crystals.
J Geophys Res 90:11319-11333
Mäder UK, Berman RG (1991) An equation of state for carbon dioxide to high pressure and temperature. Am
Mineral 76:1547-1559
Mavrogenes JA, Bodnar RJ (1994) Hydrogen movement into and out of fluid inclusions in quartz: Experimental
evidence and geologic implications. Geochim Cosmochim Acta 58:141-148
Métrich N, Wallace PJ (2008) Volatile abundances in basaltic magmas and their degassing paths tracked by
melt inclusions. Rev Mineral Geochem 69:363-402
Neumann ER, Wulff-Pedersen E, Johnsen K, Andersen T, Krogh E (1995) Petrogenesis of spinel harzburgite
and dunite suite xenoliths from Lanzarote, eastern Canary Islands: Implications for the upper mantle.
Lithos 35:83-107
Oskarsson N, Hansteen TH (1992) The use of graphite for the removal of oxygen from nitrogen puge-gas in
high-temperature microthermometry using the Linkam TH1500 stage. Eur J Mineral 4:865-871
Pasteris JD, Wanamaker BJ (1988) Laser Raman microprobe analysis of experimentally re-equilibrated fluid
inclusions in olivine: Some implications for mantle fluids. Am Mineral 73:1074-1088
Pearce TH, Kolisnik AM (1990) Observations of plagioclase zoning using interference imaging. Earth Sci Rev
29:9-26
Peccerillo A, Frezzotti ML, De Astis G, Ventura G (2006) Modeling the magma plumbing system of Vulcano
(Aeolian Islands, Italy) by integrated fluid-inclusion geobarometry, petrology, and geophysics. Geology
34:17-20
Putirka KD (2008) Thermometers and barometers for volcanic systems. Rev Mineral Geochem 69:61-120
Robertson EC (1988) Thermal properties of rocks. Open file report 88-441, United States Geological Survey,
Reston, VA
Roedder E (1963) Studies of fluid inclusions II: Freezing data and their interpretation. Econ Geol 58:167-211
Roedder E (1965) Liquid CO2 inclusions in olivine-bearing nodules and phenocrysts from basalts. Am Mineral
50:1746-1782
Roedder E (1979) Origin and significance of magmatic inclusions. Bull Minéral 102:487-510
Roedder E (1983) Geobarometry of ultramafic xenoliths from Loihi Seamount, Hawaii, on the basis of CO2
inclusions in olivine. Earth Planet Sci Lett 66:369-379
Roedder E (1984) Fluid inclusions. Reviews in Mineralogy, volume 12. Mineral Soc Am, Washington
Roedder E (1992) Fluid inclusion evidence for immiscibility in magmatic differentiation. Geochim Cosmochim
Acta 56:5-20
Roedder E, Bodnar RJ (1980) geologic pressure determinations from fluid inclusion studies. Ann Rev Earth
Planet Sci 8:263-301
Roedder E, Coombs DS (1967) Immiscibility in granitic melts, indicated by fluid inclusions in ejected granitic
blocks from Ascension Island. J Petrol 8:417-451
Roggensack KL, Hervig RL, McKnight SB, Williams SN (1997) Explosive basaltic volcanism from Cerro
Negro volcano: Influence of volatiles on eruption style. Science 277:1639-1642
Sachs PM, Hansteen TH (2000) Pleistocene underplating and metasomatism of the lower continental crust: a
xenolith study. J Petrol 41:331-356
Samson I, Anderson A, Marshall D (2003) Fluid Inclusions: Analysis and Interpretation, Short Course Series,
vol 32. Mineralogical Association of Canada, Vancouver
Schmidt C, Bodnar RJ (2000) Synthetic fluid inclusions: xVI. PVTx properties in the system H2O-NaCl-CO2
at elevated temperatures, pressures, and salinities. Geochim Cosmochim Acta 64:3853-3869
Schwarz S, Klügel A, Wohlgemuth-Ueberwasser C (2004) Melt extraction pathways and stagnation depths
beneath the Madeira and Desertas rift zones (NE Atlantic) inferred from barometric studies. Contrib
Mineral Petrol 147:228-240
176 Hansteen & Klügel
Seitz JC, Blencoe JG, Joyce DB, Bodnar RJ (1994) Volumetric properties of CO2-CH4-N2 fluids at 200°C
and 1000 bars: A comparison of equations of state and experimental data. Geochim Cosmochim Acta
58:1065-1071
Shaw CSJ, Heidelbach F, Dingwell DB (2006) The origin of reaction textures in mantle peridotite xenoliths
from Sal Island, Cape Verde: the case for “metasomatism” by the host lava. Contrib Mineral Petrol
151:681-697
Shepherd TJ (1981) Temperature-programmable, heating-freezing stage for microthermometric analysis of
fluid inclusions. Econ Geol 76:1244-1247
Shepherd TJ, Rankin AH, Alderton DHM (1985) A practical guide to fluid inclusion studies. Chapman and
Hall, New York, pp 239
Simon AC, Frank MR, Pettke T, Candela PA, Piccoli PM, Heinrich CA, Glascock M (2007) An evaluation of
synthetic fluid inclusions for the purpose of trapping equilibrated, coexisting, immiscible fluid phases at
magmatic conditions. Am Mineral 92:124-138
Sobolev VS, Kostyuk VP (1975) Magmatic crystallization based on a study of melt inclusions. “Nauka”
Press, Novosibirsk, 182-253. (in Russian; Translated in part in Fluid Inclusion Research, Proc COFFI
9 (1976))
Sourirajan S, Kennedy GC (1962) The system H2O-NaCl at elevated temperatures and pressures. Am J Sci
260:115-141
Span R, Wagner W (1996) A new equation of state for carbon dioxide covering the fluid region from the triple
point temperature to 1100 K at pressures up to 800 MPa. J Phys Chem Ref Data 25:1509-1596
Sterner SM, Bodnar RJ (1991) Synthetic fluid inclusions. x: Experimental determination of P-V-T-x properties
in the CO2-H2O system to 6 kb and 700°C. Am J Sci 291:1-54
Sterner SM, Chou I-M, Downs RT, Pitzer KS (1992) Phase relations in the system NaCl-KCl-H2O. V:
Thermodynamic-PTx analysis of solid-liquid equilibria at high temperatures and pressures. Geochim
Cosmochim Acta 56: 2295-2309
Sterner SM, Hall DL, Bodnar RJ (1988) Synthetic fluid inclusions. V. Solubility relations in the system NaCl-
KCl-H2O under vapour-saturated conditions. Geochim Cosmochim Acta 52:989-1005
Sterner SM, Pitzer KS (1994) An equation of state for carbon dioxide valid from zero to extreme pressures.
Contrib Mineral Petrol 117:362-374
Szabó CS, Bodnar RJ (1996) Changing magma ascent rates in the Nógrád-Gömör volcanic field, Northern
Hungary / Southern Slovakia: Evidence from CO2-rich fluid inclusions in metasomatized mantle
xenoliths. Petrology 4:240-249
Takenouchi S, Kennedy GC (1965) The solubility of carbon dioxide in NaCl solutions at high temperatures and
pressures. Am J Sci 263:445-454
Thomas R, Webster JD, Davidson P (2006) Understanding pegmatite formation: The melt and fluid inclusion
approach. In: Melt inclusions in plutonic rocks. Webster JD (ed) Mineral Assoc Can, Vancouver, p 189-
210
Touret JLR (2001) Fluids in metamorphic rocks. Lithos 55:1-25
Troll VR, Schmincke H-U (2002) Magma mixing and crustal recycling recorded in ternary feldspar from
compositionally zoned peralkaline ignimbrite “A”, Gran Canaria, Canary Islands. J Petrol 43:243-270
Van den Kerkhof AM (1988) The system CO2-CH4-N2 in fluid inclusions: theoretical modelling and geological
applications. Ph.D. Dissertation, Free University, Amsterdam
Van den Kerkhof AM (1990) Isochoric phase diagrams in the systems CO2-CH4 and CO2-N2: Application to
fluid inclusions. Geochim Cosmochim Acta 54:621-629
Van den Kerkhof AM, Hein UF (2001) Fluid inclusion petrography. Lithos 55:27-47
Van den Kerkhof AM, Thiéry R (2001) Carbonic inclusions. Lithos 55:49-68
Vityk MO, Bodnar RJ (1998) Statistical microthermometry of synthetic fluid inclusions in quartz during
decompression re-equilibration. Contrib Mineral Petrol 132:149-162
Wanamaker BJ, Evans B (1989) Mechanical re-equilibration of fluid inclusions in San Carlos olivine by power-
law creep. Contrib Mineral Petrol 102:102-111
Wanamaker BJ, Wong TF, Evans B (1990) Decrepitation and crack healing of fluid inclusions in San Carlos
olivine. J Geophys Res 95:15623-15641
Weisbrod A (1981) Fluid inclusions in shallow intrusives. In: Short course in fluid inclusions: applications to
petrology. Vol 6. Hollister LS, Crawford ML (eds), Mineral Assoc Can, Vancouver, p 241-277
Werre RW Jr, Bodnar RJ, Bethke PM, Barton PB Jr (1979) A novel gas-flow fluid inclusion heating/freezing
stage. Geol Soc Am Abstr Prog 11:539 (abstr)
Woods TL, Bethke PM, Bodnar RJ, Werne RW Jr (1981) Supplementary components and operation of the U.S.
Geological Survey gas-flow heating/freezing stage. US Geol Survey Open File Report #81-954
Wopenka B, Pasteris JD, Freeman JJ (1990) Analysis of individual fluid inclusions by Fourier transform
infrared and Raman microspectroscopy. Geochim Cosmochim Acta 54:519-533
Fluid Inclusion Thermobarometry: Tracer for Magmatic Processes 177
Wulff-Pedersen E, Neumann ER, Jensen BJ (1996) The upper mantle under La Palma, Canary Islands: formation
of Si-K-Na-rich melt and its importance as a metasomatic agent. Contrib Mineral Petrol 125:113-139
zanon V, Frezzotti ML, Peccerillo A (2003) Magmatic feeding system and crustal magma accumulation
beneath Vulcano Island (Italy): Evidence from CO2 fluid inclusions in quartz xenoliths. J Geophys Res
108(B6, 2298): doi:10.1029/2002JB002140
zapunnyy SA, Sobolev VS, Bogdanov AA, Slutskiy LV, Dmitriyev LV, Kunin LL (1988) An apparatus for
hightemperature optical research with controlled fugacity. Geokhimiya 7:1044-1052
zhang YG, Frantz JD (1987) Determination of homogenization temperatures and densities of supercritical
fluids in the system NaCl-KCl-CaCl2-H2O using synthetic fluid inclusions. Chem Geol 64:335-350
Reviews in Mineralogy & Geochemistry
Vol. 69 pp. 179-239, 2008 6
Copyright © Mineralogical Society of America
Petrologic reconstruction of Magmatic System
Variables and Processes
Jon Blundy
Department of Earth Sciences
University of Bristol
Wills Memorial Building, Bristol BS8 1RJ, United Kingdom
[email protected]
Kathy Cashman
Department of Geological Sciences
University of Oregon
Eugene, Oregon 97403-1272, U.S.A.
[email protected]
IntrOduCtIOn
Explosive volcanic eruptions constitute a major class of natural hazard with potentially
profound economic and societal consequences. Although such eruptions cannot be prevented
and only rarely may be anticipated with any degree of accuracy, better understanding of how
explosive volcanoes work will lead to improved volcano monitoring and disaster mitigation.
A major goal of modern volcanology is linking of surface-monitored signals from active
volcanoes, such as seismicity, ground deformation and gas chemistry, to the subterranean
processes that generate them. Because sub-volcanic systems cannot be accessed directly, most
of what we know about these systems comes from studies of erupted products. Such studies
shed light on what happens underground prior to and during eruptions, thereby providing
an interpretative framework for post hoc evaluation of monitoring data. The aim of this
review is to present some of the current petrological techniques that can be used for studying
eruptive products and for constraining key magmatic variables such as pressure, temperature,
and volatile content. We first review analytical techniques, paying particular attention to
pitfalls and strategies for analyzing volcanic samples. We then examine commonly used
geothermometry schemes, evaluating each by comparison with experimental data not used
in the original geothermometer calibrations. As there are few mineral-based geobarometers
applicable to magma storage regions, we review other methods used to determine pre-
eruptive magma equilibration pressures. We then demonstrate how petrologically-constrained
parameters can be compared to the contemporaneous monitoring record. These examples are
drawn largely from Mount St. Helens volcano, for which there are abundant petrological and
monitoring data. However, we emphasize that our approaches can be applied to any number
of active volcanoes worldwide. Finally, we illustrate the application of these techniques to two
different types of magmatic systems—large silicic magma chambers and small intermediate-
composition magma storage regions—with particular focus on the combined evolution of melt
volatiles and magmatic crystal contents in these two different scenarios.
1529-6466/08/0069-0006$10.00 DOI: 10.2138/rmg.2008.69.6
180 Blundy & Cashman
MaGMatIC VarIaBLeS
Magmatic variables can be both intensive and extensive. Intensive variables are those
whose values do not depend on the amount of magma or gas in the system, namely composi-
tion, crystallinity, temperature, storage pressure and oxidation state. By combining knowledge
of these intensive variables with experimental measurements it is possible to constrain key addi-
tional variables that exert important controls on eruptive behavior, such as density and viscosity,
or that can refine seismic or potential-field studies of volcanic regions, such as seismic wave ve-
locity and electrical conductivity. Extensive magmatic variables are the counterpart of intensive
variables—their value is proportional to the size of the system. Examples include the volume
and depth of stored magma, the magmatic flux into and out of the sub-volcanic system, and
various forms of heat and stored energy. Extensive variables are much harder to determine from
petrology and require integration with other techniques such as seismic tomography, ground
deformation and physical volcanology. Some extensive variables have complex relationships
with intensive variables. For example, magma storage depth (an extensive variable) is related to
storage pressure (an intensive variable) through the nature of the overburden, the magma flow
velocity and the extent to which
the system is over-pressured. In
general, numerical modeling
of the dynamics of volcanic Edifice
systems requires knowledge
of both intensive and extensive
variables.
Conduit
Magmatic variables evolve
in space and time, reflecting
the complexity of underground
magmatic plumbing sys-
tems and processes by which Reservoir
those plumbing systems are
recharged and emptied (e.g.,
Marsh 2006). Although no two
volcanoes are alike, one can
envisage a generic magmatic 1
plumbing system that encapsu- 2
Different
magma inputs
lates the key variables (Fig. 1). 3
Feeder system
Our generic volcano comprises
a region of principal magma Crystal rich magma
storage, typically referred to
as a magma chamber or reser-
voir, and a plexus of fractures,
which may or may not be Source region
magma-filled to form dykes,
sills or conduits, which con-
nect the chamber to the surface
above and to the magma source Figure 1. Schematic of a sub-volcanic magma system showing
key features to be investigated using petrological techniques. The
region below. Intensive param- chamber is recharged via a feeder system from a deeper source
eters can vary spatially within region and discharged via a shallow conduit system connected to
this system due to heat loss to the surface. The magma reservoir is a zone of mixing and blending
(and physical exchange with) between different magma inputs from the source region. Crystal-
the wall rocks, heat input from lization and degassing occur in the conduit system and reservoir,
which may extend laterally in the form of sills. The conduit con-
new magma additions from be- necting the magma reservoir to the surface may be either magma-
low, post-eruptive changes in filled or a fracture network.
Reconstruction of Magmatic System Variables & Processes 181
the volcanic load, gas loss (or throughput) and associated redox reactions, and crystallization.
Both intensive and extensive variables may show significant temporal variability that can serve
to enhance or diminish the likelihood of eruption and that should be manifest as changes in the
monitoring record.
Petrologically-determined intensive variables
We focus on those variables that can be deduced from petrology alone, i.e., those that can
be determined from studying eruptive products in the laboratory. These include:
1. Temperature and pressure, determined by geothermometry and geobarometry using
the chemistry of coexisting minerals and/or glasses.
2. The concentration, distribution and migration of volatile species determined by
analysis of glassy melt inclusions.
3. The extent, distribution and timing of crystallization and degassing, as determined
using textural and chemical constraints.
4. Experimental simulation of sub-volcanic systems involving matching of the phase
assemblage and phase compositions of erupted rocks.
Linking intensive variables to the monitoring record
The chemistry and textural characteristics of erupted rocks contain a wealth of information
about processes occurring within the sub-volcanic system prior to and during eruption. Magma
that remained underground after one eruption may be discharged subsequently, so that some
textural or chemical features may record information about previous eruptive episodes (e.g.,
Hammer et al. 1999; Cashman and McConnell 2005). Unraveling the complex testimony of
volcanic rocks is a key objective of this review, and one that will be illustrated using selected ex-
amples. Linking this testimony to the contemporaneous monitoring record, however, is fraught
with difficulties, even for the world’s most extensively monitored volcanoes, such as Mount St.
Helens (USA), Soufrière Hills (Montserrat), Unzen (Japan) and Mt. Etna (Italy). It is increas-
ingly clear that only through a multidisciplinary effort involving volcano monitoring agencies,
seismologists, petrologists, gas geochemists and numerical modelers will we be able to create
an accurate picture of the underground geometry and dynamics of magmatic systems.
MethOdOLOGy
Petrology relies to varying extents on chemical and textural characterization of eruptive
products using microbeam techniques. Three commonly used analytical techniques that
we review here are back-scattered electron microscopy, electron-microprobe analysis, and
secondary ion mass spectrometry (or ion-microprobe analysis); other important techniques
are discussed elsewhere in this volume (e.g., FTIR and micro-Raman in Metrich and Wallace
2008; LA-ICP-MS in Kent 2008; isotopic analysis in Ramos and Tepley 2008). We also briefly
review new techniques for imaging and analysis (such as nanoSIMS and x-ray computer
tomography) as well as methods of textural analysis.
Scanning electron microscopy (SeM)
SEMs can be used in several different modes for different applications. Most useful for
petrologic applications is the acquisition of images from back-scattered electrons (BSEM)
that originate from elastic collisions between the incident electron beam and atomic nuclei at
or close to the sample surface. The intensity of the back-scattered signal depends primarily
on the mean atomic number (Z) of the target. Because the back-scattered signal derives
predominantly from the outermost few nanometers of the sample, BSEM does not suffer
from significant blurring of chemical contrast due to vertical chemical heterogeneity, thus
182 Blundy & Cashman
the technique provides an excellent map of Z variations at the sample surface. Lateral spatial
resolution is controlled by the beam diameter, which is a function of the incident beam current
used. Lower incident beam currents confer greater spatial resolution at the expense of reduced
signal. The electron beam accelerating voltage is an important term in the back-scatter cross-
section equation, however within the range commonly used for petrological work (5 to 20 kV),
the influence of accelerating voltage is subordinate to that of target Z and beam current.
BSEM is most useful for elucidating
fine-scale chemical zoning in minerals
showing solid solution between end-
members of contrasted Z, such as plagioclase
and ferromagnesian minerals (Figs. 2 and
3), and for imaging crystals that have low Z
contrast with the surrounding matrix glass
(e.g., plagioclase in high-silica glass). When
a sample has both low Z and high Z crystals
(say, plagioclase and hornblende), high
quality imaging usually requires separate
image acquisition for the two phases using
different analysis conditions (beam current
and contrast settings). BSEM images can
also be processed to generate mineralogical
modes by binning the image for certain
45
grayscale bands corresponding to particular
minerals and then using image-processing 40
software to calculate areas of the different
An mol%
35
bins (e.g., Hammer et al. 2000). However, care
must be taken when minerals have similar 30
and overlapping Z values, such as is the case
25
for amphiboles and clinopyroxenes.
20
BSEM can also be used to generate 90 80 70 60 50 40 30 20 10 0
qualitative diffusion profiles across Distance to rim (μm)
individual chemical zones. In those cases
where a single chemical component is Figure 2. BSEM image of fine-scale zoning at
giving rise to most or all of the backscatter the rim of a plagioclase crystal from Parinacota
contrast, quantitative diffusion profiles can volcano, Chile. Below the BSEM image is a graph
showing the compositional variation (as mol%
be generated provided that independent anorthite) represented by that zoning. Image and
analysis (e.g., by EPMA or SIMS) is used data courtesy of C. Ginibre; further details can be
to constrain the relationship between found in Ginibre et al. (2002).
backscatter intensity and composition (e.g.,
Ginibre et al. 2002; Morgan et al. 2004,
2006; Humphreys et al. 2008a). Examples include anorthite content of plagioclase (Fig. 2),
Mg# of olivine (Fig. 3), H2O content of matrix glass (Fig. 4) and Fe-Ti zonation in oxides
(Devine et al. 2003). quantitative diffusion profiles can be used to derive temporal information
provided that diffusivity of the key chemical component at the conditions of interest is known
(Costa et al. 2003; Costa and Chakraborty 2004; Humphreys et al. 2008a; Costa et al. 2008).
As diffusivity is often anisotropic, some additional information on the orientation of the crystal
across the diffusion interface is required. This can be obtained by conventional universal stage
microscope techniques or from electron back-scatter diffraction analysis (EBSD, e.g., Schwartz
et al. 2000; Randle and Engler 2000). Where multiple major components are diffusing, no
simple relationship between backscatter intensity and composition can be obtained.
Reconstruction of Magmatic System Variables & Processes 183
Figure 3. BSEM image of zoned
olivine crystal from Santorini, Greece.
The grayscale variation reflects
variation from Mg-rich (dark) core to
Fe-rich (light) rim. The gradation in
grayscale across the core-rim boundary
can be used to quantify the extent of
Mg-Fe interdiffusion. Image courtesy
of D. Morgan and V. Martin.
Figure 4. BSEM image of melt tube in plagioclase phenocryst from Mount St. Helens (left). The melt
tube connects the interior of the crystal with the matrix glass and shows a grayscale variation due to an
outwards decrease in dissolved H2O. Ion-microprobe analyzes of H2O in the melt tube and groundmass
glasses (4 white circles on left hand image) are used to calibrate the relationship between dissolved H2O
and grayscale intensity to produce a diffusion profile (right) that can be modeled to estimate the time taken
for H2O to diffuse out of the interior of the crystal and into the matrix. Fit parameters give diffusion times
(t) of 107-166 s for initial H2O contents of 4.6-6.4 wt% in melt in the crystal interior. From Humphreys et
al. (2008a).
Another type of SEM imaging uses cathodoluminescence (CL), an optical technique
that captures photons generated by interaction of the electron beam with the sample surface.
Although the origin of the CL signal is often unclear, as photons are emitted from defects
in the crystal structure created either by vacancies or by the presence of trace elements, CL
imaging can reveal complexities of crystal growth behavior that are not observable using other
methods. Absolute CL intensity is difficult to quantify, however, because it is dependent on
operating conditions; for this reason it is common to describe CL domains simply as “CL-
dark” and “CL-bright”. For petrologic applications, CL is a particularly effective technique
for imaging zoning patterns in quartz (e.g., Peppard et al. 2001; Rusk and Reed 2002; Rusk
et al. 2008) and zircon (e.g., Nemchin et al. 2001), where growth patterns revealed by CL
184 Blundy & Cashman
imaging provide insight into both the mechanism (e.g., Ostwald ripening) and location (within
the magma reservoir) of crystal growth. When the CL characteristics can be related to specific
trace elements, such as Ti in quartz, and quantified using the microbeam techniques described
above, spatial patterns of growth can be directly linked to intensive parameters responsible for
crystallization through use of mineral equilibria, such as the titanium-in-quartz geothermometer
of Wark and Watson (2006).
electron microprobe analysis (eMPa)
EMPA uses a focused electron beam to generate x-rays via electron transfer between
inner orbitals (K, L, M) in the target atoms. The energy (or wavelength) of the resulting
x-radiation is chemically diagnostic and the intensity of a particular x-ray is a function of
the concentration of the element that produced it within the target volume. x-rays can be
detected as a function of their energy (energy-dispersive EMPA, or EDS) or their wavelength
(wavelength-dispersive EMPA, or WDS). In EDS, the entire energy spectrum is collected
simultaneously, while in WDS a series of moveable spectrometers are tuned to characteristic
wavelengths for a particular length of time. In both cases, analysis time is several minutes,
although for low abundance elements analysis times can be significantly longer. WDS involves
more complex and expensive instrumentation but has lower detection limits than EDS, as well
as less problematic calibration and data reduction. WDS is therefore considered the technique
of choice for most petrological studies.
By scanning the beam across the sample it is possible to collect x-rays of a particular
wavelength or energy and so generate a chemical map (Fig. 5) to complement a BSEM image
Figure 5. Maps of polished dacite sample (SH184, erupted September 1984) from Mount St. Helens. (a)
BSEM map showing bright ferromagnesian mineral, mid-grey plagioclase and glass and black vesicles.
(b) Al map allowing clear distinction between plagioclase (light) and glass (dark). (c) Fe map showing
bright oxide minerals and distinguishes between pyroxene and hornblende. (d) Digitized binary image of
orthopyroxene in sample. This image can be used to quantify the volume fraction of orthopyroxene in the
sample and to derive crystal size distribution data. Similar maps can be generated for all other phases by
combining individual chemical maps. Image is 1 mm across. Images courtesy of D. Muir.
Reconstruction of Magmatic System Variables & Processes 185
of the same area. This offers the advantage of distinguishing between phases that cannot be
readily distinguished by BSEM alone because they have similar Z values. In our example,
while plagioclase crystals can be distinguished by BESM alone (confirmed by the Al map),
distinguishing between the mafic phases (hornblende, orthopyroxene and Fe-Ti oxides)
requires additional element maps, here illustrated by the identification of orthopyroxene using
the Fe map. Chemical mapping is also a valuable technique when looking at chemical zonation
of a component whose abundance is too low to affect BSEM intensity. A recent example is the
mapping of phosphorous in igneous olivine to elucidate fine-scale zoning that is less sensitive
to diffusive re-homogenization than Fe-Mg zoning (Milman-Barris et al. 2008).
As with BSEM, the intensity of the x-ray signal is a function of the input electron beam
energy and flux: higher beam currents excite x-rays from lower concentrations of target
elements at the expense of poorer spatial resolution, whereas higher accelerating voltages
excite higher energy x-rays from heavier elements at the expense of greater sample penetration.
Elements with Z greater than 4 (beryllium) can be analyzed by EMPA, although only elements
with Z > 8 are routinely analyzed. With suitably intense electron beams and long counting
times (several minutes) minimum detection limits of <100 ppm are attainable, although for
some minor and trace elements considerable care must be taken to deconvolve interferences
from minor x-ray lines of major elements, e.g., SiKβ on SrLα, by selecting suitably clean
background positions.
Analytical conditions are carefully selected for the type of materials being analyzed,
particularly those that are, to variable degrees, unstable under the electron beam, e.g., hydrous
silicate glasses, sodium-rich feldspars, carbonates, sulfates and phosphates. Most silicate
minerals can be analyzed using 15-20 keV electrons, with a beam current of 5-15 nA, focused
to a spot at the sample surface. Unlike backscatter electrons, x-rays are generated from a
significant interaction volume below the sample surface, which limits the effective spatial
resolution to 2-3 microns. Lower beam currents and accelerating voltages are recommended
for microlites whose small sizes make them susceptible to x-ray excitation from surrounding
and underlying glass. Microlite analyses should be carefully screened for glass contamination,
manifest as stoichiometrically excess Si and/or anomalously high K.
Hydrous silicate glasses pose a particular analytical challenge because of alkali-migration
during electron beam irradiation. The problem is most acute in rhyolites with >1 wt% dissolved
H2O and appears to be a consequence of electron implantation into the outer few microns of
the sample. Cations within the sample are attracted electrostatically away from the surface
towards the zone of highest electron density, moving at a rate dependent on their solid-state
diffusivity. Conversely, H2O concentrations increase towards the sample surface, possibly due
to migration of OH− away from the implanted electrons (Humphreys et al. 2006). The migration
process is thermally mediated by heating at the sample surface due to phonon excitation by
the electron beam. Those species with the highest diffusivities, i.e., Li, Na, K and H2O, are the
most strongly affected and H2O-rich glasses are the most susceptible. SIMS depth-profiling of
electron-beam irradiated regions (Humphreys et al. 2006) reveals that the process is a strong
function of net electron beam energy and is irreversible. Alkali-migration can lead to low Na
and correspondingly high Si and Al, as well as low analytical totals. Low totals are especially
problematic when the shortfall from 100% is assumed to be due to H2O as in the volatile-by-
difference (VBD) technique (e.g., Devine et al. 1995; King et al. 2002).
The simplest way to mitigate beam damage is to reduce beam currents to 2-4 nA and
defocus the electron spot over a 10-15 µm diameter. Na and K are counted first and for short
periods (e.g., 6s peak, 3s background) calibrated using similar beam conditions, and run using
a glass (or suite of glasses) of known H2O content as a secondary standard. VBDs within 0.5
to 0.7 wt% of the true H2O content should be attainable using this technique (Humphreys
et al. 2006), assuming negligible concentrations of other unanalyzed species (e.g., CO2, F,
186 Blundy & Cashman
etc.). A drawback of this approach is that it requires relatively large pools of glass, which are
not always available in highly vesicular pumices or where melt inclusions are small. Another
method is to use conventional electron beam conditions, but to extrapolate the Na signal to
zero time to compensate for the effects of migration. This approach has the advantage that
smaller glass pools can be analyzed, but requires software capable of taking counts at regular
intervals and requires some key assumptions about the shape of the Na decay curve with time.
Agreement between VBD and known H2O contents of rhyolite glasses to within 0.5 wt% is
reported using this approach, particularly when combined with iterative addition of H2O to the
matrix correction (e.g., Devine et al. 1995; Roman et al. 2006). The final method is to freeze
the sample within the analytical chamber to minimize thermal influences on alkali diffusion
(Kearns et al. 2002). At liquid nitrogen temperatures a cold stage can effectively eliminate alkali
migration. However, considerable technological modification to most electron-microprobes is
required for cold-stage analysis and contamination build-up is a major problem.
Secondary ion mass spectrometry (SIMS)
SIMS is a microbeam technique in which a beam of focused ions (positively or negatively
charged) is used to sputter particles, some of which are charged (secondary ions), from a sample
surface under high vacuum (Shimizu and Hart 1982; Hinton 1995). The secondary ions are then
accelerated into a double-focusing mass spectrometer where they are separated according to
their energy and mass/charge ratio before being detected using an electron multiplier or Faraday
cup. The most commonly used primary beams are O− and Cs+. SIMS analysis of geological
samples is conventionally carried out on gold-coated polished thin sections or grain mounts.
Unlike EMPA the technique is destructive in that the sputtering process leaves a small
hole at the sample surface. The size of the sputter crater depends on beam intensity and
analytical duration, but for most petrological work is ~2 microns deep. SIMS affords several
advantages over EMPA in that it boasts much greater sensitivity for many elements of interest
and can be used to analyze light elements, such as Li, B, H and C. The latter can be readily
converted to H2O and CO2. The ease with which elements can be analyzed depends primarily
on their secondary ion-yield under the analytical conditions. For a given set of primary beam
conditions, secondary ion-yields can vary by 4 orders of magnitude across the periodic table
(Hinton 1990). Elements with favorable positive or negative secondary ion-yields (e.g., Li+, B+,
Ti+, REE+, U+, Th+, Br−, I−) have detection limits of less than few 10s of ppb. Other elements,
for example Group xII metals (zn, Cd, Hg) and platinum group elements, have negligible
high-energy secondary ion yields and cannot be analyzed at very low concentrations.
A further complication is the presence of molecular ion interferences. Sputtered ions
include both monatomic and polyatomic species (molecular ions). Examples include the
isobaric interferences of 24Mg2+ on 12C+, 29Si16O+ on 45Sc+ and 137Ba16O+ on 153Eu+. Low mass
resolving power (~500) prevents direct resolution of these isobaric species on the basis of their
small mass differences; for this reason, molecular ion interferences are commonly addressed
using kinetic energy filtering, which relies on the narrower energy distribution of sputtered
molecular ions compared to monatomic ions. Selective sampling of the high-energy tail of
the energy distribution (by reducing the sample extraction voltage by 50-100 V) considerably
reduces transmission of molecular ions, although in some cases additional conventional peak
stripping is necessary. For example, to eliminate 29Si16O+ on 45Sc+, one must first analyze 42Ca+,
from which 44Ca+ can be calculated. The difference between the measured and calculated
count rate at mass 44 is ascribed to 28Si16O+, from which 29Si16O+ can be calculated (assuming
identical ion-yield for 28Si and 29Si) and subtracted from mass 45 to yield the corrected 45Sc+.
In contrast to x-ray generation by electrons, the physics of sputtering is not well understood
and SIMS relies heavily on calibration using standards of known composition whose matrix is
closely matched to that of the target materials. It is possible to calibrate on a single multi-
element standard, such as NIST glass SRM610, but careful consideration must be given to
Reconstruction of Magmatic System Variables & Processes 187
relative ion-yield difference between this glass and natural silicate glasses and minerals, which
may be up to 25% relative. quantitative SIMS analysis requires that the secondary ion count is
referenced to that of a species whose concentration is known: conventionally this is 30Si, whose
concentration is determined by EMPA of the same spot.
Analysis of H2O and CO2. An important petrological application of SIMS is the analysis
of H2O (as 1H+) and CO2 (as 12C+). Calibration for these volatile species requires use of a suite
of glasses, similar in composition to those of interest, with known H2O and CO2 contents.
As there is no empirical evidence that 1H ion yields are affected by the presence of CO2,
and vice versa, it is possible to use calibrant glasses that contain both species to generate
working curves that relate normalized 1H and 12C secondary-ion count rates to H2O and CO2,
respectively, in the standards. In principal the secondary-ion counts should be corrected for
variations in Si content (Shimizu 1997). In practice, the variation in relative ion yields is
correlated with SiO2 content, such that plotting 1H/30Si (or 12C/30Si) versus H2O (or CO2) yields
a single working curve for glasses of variable silica content (Fig. 6; Blundy and Cashman
2005). This (fortuitously) provides a convenient and robust means of determining H2O and
CO2 that requires relatively few calibrant glasses.
The quality of the analysis depends critically on the instrumental background. Elevated
background counts are a perennial problem for H2O and CO2 analysis because of their high
concentration in organic materials used in sample preparation and mounting (especially epoxy
resins used for grain mounts). The problem is greatest for CO2 because of its low concentration
in most geological materials and its inherently low positive secondary-ion yield. Surface
contamination with H or C, especially from organic solvents or diamond polishing materials, can
be minimized by allowing a few minutes “burn-in” of the beam before analysis and elimination
of the first few analytical cycles. Surface contamination can be minimized further by chemical
polishing with colloidal silica, by analyzing thin sections (rather than grain mounts), by using
low volatility mounting resins (where necessary), and by working at high vacuum. The latter
is facilitated on ion-microprobes fitted with multi-sample airlocks in which samples can be
Figure 6. Typical SIMS working curves for (a) H2O and (b) CO2 derived from rhyolite, dacite and andesite
glasses of known volatile content analyzed on a single day in March 2006 using the Edinburgh University
Cameca ims4f ion microprobe. Abscissa denotes ratios of counts per second for 1H+ and 12C+, normalized
to 30Si. 12C+ is corrected for 24Mg2+ interference. H2O based on 13 different glasses; CO2 based on 10
different glasses. The equations of the working curves are given. Fully propagated uncertainties (1 s.d.),
are approximately same size as symbol, or smaller.
188 Blundy & Cashman
pumped down to < 5×10−8 Torr prior to insertion into the analytical chamber, which is typically
at < 2×10−9 Torr. Backgrounds can be monitored during analysis using a small chip of quartz
mounted alongside the sample. Experience on the Cameca ims-4f at Edinburgh University
using a 10kV (nominal) 16O− primary beam, 4.5 kV secondary accelerating 75±20 V offset,
and 2-5 nA current at the sample surface, shows that 30Si-normalised backgrounds measured
on NIST28 quartz are typically 5×10−3 for 1H+ and 1×10−4 for 12C+ for samples pumped to 10−9
Torr or less in an airlock prior to analysis. This corresponds to minimum detection limits of the
order 150 ppm or less for H2O and 60 ppm or less for CO2 in Mg-free materials (see below).
Alternatively, individual crystals or glass chips can be pressed into indium, thereby
reducing the introduction of organics into the sample chamber. Using this technique coupled
with a Cs+ primary beam, high primary ion currents (200-300 nA) and an electron flood gun
to reduce sample charging, detections limits for negative secondary ions equate to ≤3 ppm for
C and ≤30 ppm for H2O (Hauri et al. 2002). The improved detectability, however, comes at the
expense of more time-consuming sample preparation, painstaking analytical set up, relatively
large (15-30 micron) beam diameters and potential damage to the glass by the electron flood
gun. For typical H2O contents of silica-rich volcanic glasses (≥0.5 wt%) such enhanced
detection limits are not warranted.
A further problem with CO2 analysis is the isobaric interference of 24Mg2+ on 12C+. This
interference is not removed by energy filtering as Mg2+/Mg ratios do not significantly change
with energy, and therefore must be resolved by peak stripping using 25Mg2+ measured on
mass 12.5. The small differences in ion-yield between 25Mg and 24Mg should be determined
from analysis of an Mg-rich silicate glasses under the same analytical conditions as the ratio
may differ slightly from the canonical value of 7.90. For Mg-poor rhyolite glasses with ~0.3
wt% MgO, the interference amounts to approximately 1000 ppm CO2. Thus it is essential
to determine the 24Mg+/25Mg+ ion-yield ratio very precisely, to employ long count times on
masses 12 and 12.5, and to measure background count rates for each sample analyzed to
optimize determination of CO2 concentrations of ≥100 ppm (e.g., Fig 6b). For more Mg-rich
glasses the contribution from 24Mg2+ is too large to be adequately removed by peak-stripping
and higher mass resolving power (≥1500) must be used, at the expense of poorer secondary
ion transmission.
NanoSIMS is a relatively new SIMS technique, increasingly used in cosmochemistry
for the isotopic analysis of tiny interplanetary dust particles (Hoppe 2006). The technique has
many similarities to conventional SIMS, although the coaxial ion optics of the primary and
secondary ion beams enable the incident ion beam (O− or Cs+) to be focused to a much smaller
spot at the sample surface (e.g., ~50 nm for Cs+, 150 nm for O−), while maintaining high
secondary ion collection efficiency. Like conventional SIMS, NanoSIMS is not standardless
and calibration against standards of known composition is required. There have been relatively
few petrological applications of NanoSIMS to date (e.g., Hellebrand et al. 2005), and the
considerable potential applications to trace element and volatile element variations in volcanic
materials have yet to be explored. However, the high spatial resolution, linked to analytical
sensitivity, makes NanoSIMS ideally suited to the study of fine-scale trace element zoning
in minerals for use in diffusion geospeedometry, or chemical analysis of very small melt
inclusions. We anticipate that NanoSIMS applications in volcanic petrology will increase
substantially in the coming years.
analysis of textures
The textures of volcanic samples provide fundamental information on conditions of
magma storage and ascent. So integral are textures to our thinking about igneous rocks that
they form the basis for much descriptive classification (e.g., obsidian vs. pumice; porphyritic
vs. aphanitic; basalt vs. diabase vs. gabbro). Here we briefly review two aspects of textural
analysis: data collection and data interpretation. More detailed discussions of crystal and vesicle
Reconstruction of Magmatic System Variables & Processes 189
size distribution analysis can be found elsewhere (e.g., Marsh 1988; Cashman and Marsh 1988;
Cashman 1990; Higgins 1994, 2006; Kile et al. 2000; Klug et al. 2002; Madras and McCoy
2002; Gurioli et al. 2005; Hersum and Marsh 2006, 2007; Resmini 2007; Armienti 2008).
Image acquisition and processing.
Modal analysis of phenocryst phases is
traditionally performed by point counting
using a mechanical stage mounted on a
petrographic microscope (e.g., Chayes
1956). The uncertainty on point-counted
modes can be estimated graphically using
Van der Plas and Tobi (1965). The full range
of textural features is best captured using
a high-resolution slide scanner to obtain
a digital image of a thin section (Fig. 7a),
while groundmass phases are best imaged
using BSEM or x-ray maps (see above).
The resulting digital images can be analyzed
for crystal (or bubble) size, shape, number
density and orientation using automated
image analysis programs available as free
downloads (e.g., NIH Image, http://rsb.info.
nih.gov/nih-image/index.html). Automated
thresholding to produce the binary image
required for analysis is easiest (and most
accurate) when the contrast between the
phase of interest and the background is
large (e.g., plagioclase phenocrysts in a fine-
grained matrix; Fig. 7a,b). When contrasts Figure 7. Example of a thin section scan of a 1956
are not sufficient for automated analysis, cryptodome sample from Bezymianny Volcano,
Kamchatka, showing white plagioclase crystals
individual phases may be distinguished
(a) and associated binary image of plagioclase (b)
using cross-polarized light, Z contrast from which textural data can be easily extracted.
variations in BSEM images or x-ray maps The total image area is 107 mm2, of which 21.9%
of selected elements (e.g., Fig. 5) and then is occupied by plagioclase phenocrysts. 1670
outlined by hand. individual plagioclase crystals are identified, for an
average of 15.57 crystals mm−2.
Measurements from thin sections are
two-dimensional. If bulk magma properties
are to be investigated, measurements must be converted from 2-D to 3-D (e.g., Jerram and
Higgins 2007); this conversion requires additional information about the crystal population
(such as phase proportions and the distribution of crystal shapes). For this reason, there have
been several techniques introduced to measure true 3-D crystal size data directly, including
extraction of whole crystals from a sample (e.g., Dunbar et al. 1994; Bindeman 2005), optical
focusing through a thick section with low crystallinity (e.g., Castro et al. 2003), statistical
methods (e.g., Morgan and Jerram 2006; Armienti 2008) and x-ray computed tomography
(xCT; e.g., Carlson 2006). Of these, xCT is potentially the most useful because images can be
acquired rapidly, several phases can be isolated from a single image, spatial relations between
individual bubbles or crystals are preserved, and images allow visualization of the phase
topology that is not possible in 2D images.
xCT uses the attenuation of an x-ray beam passing through a specimen to image phases of
different mass density or Z contrast. Measurements are typically made on a rotating specimen
that allows the sample to be viewed from multiple angles; individual tomographic slices map
190 Blundy & Cashman
attenuation within the plane of the x-ray beam. The resulting xCT image is a 3D array of
voxels (volume elements), each of which is assigned an attenuation coefficient. Processing of
these data requires smoothing and filtering algorithms that may limit resolution, particularly
of thin bubble walls. 3D images are commonly presented as movies (e.g., Wright et al. 2006)
or renderings of shaded relief (e.g., Gualda and Rivers 2006; Polacci et al. 2007). Current
limitations of the technique relate to image resolution, which depends on both the sample size
and average attenuation, and the need for large density contrasts between phases of interest. For
this reason, volcanological applications of xCT have focused primarily on imaging vesicles
(e.g., Lindquist and Venkatarangan 1999; Song et al. 2001; Shin et al. 2005; Wright et al. 2006;
Polacci et al. 2007), although phenocrysts can also be imaged provided that there is sufficient
contrast with matrix material (e.g., Gualda and Rivers 2006). Processed images can be used to
test models of permeable flow through actual pore pathways (e.g., Wright et al. 2006) or for
extraction of quantitative textural data (e.g., Song et al. 2001; Shin et al. 2005; Gualda 2006;
Polacci et al. 2007).
Textural analysis. Textural data acquired by the methods reviewed above can be used
to address both the pre-eruptive state of the magmatic system and the processes by which
magma ascends and erupts. Both the bulk crystallinity and the proportions of phenocryst phases
provide important information on magma storage conditions; these data typically provide
baseline information for assessment of phase equilibria experiments designed to constrain
intensive parameters of magma storage reservoirs (see below). In theory, measurement of the
size distribution of different crystal phases could also provide information on time scales of
crystallization, although this would require better constraints on crystal growth and nucleation
rates than are currently available. Alternatively, time constraints provided by other methods
such as short-lived isotopes (e.g., Gauthier and Condomines 1999) or diffusional profiles along
phenocryst margins (e.g., zellmer et al. 2003) can be used in conjunction with textural data to
measure growth rates directly.
The size distributions of individual phases (crystals and bubbles) reflect the integrated
crystallization and vesiculation history of a magma batch, and therefore provide critical
information not only on the physical state of the magma at the time of eruption but also on changes
in the physical state through space and time. Size analysis provides a particularly powerful tool
when the emplacement history of the samples is well constrained. For example, lava samples
collected along active flows through open channels (where cooling and crystallization are
rapid) and lava tubes (where insulation slows cooling and crystallization) provide constraints
on the syn-emplacement crystallization histories of these two lava types. Textural differences
are evident in a simple comparison of the number of crystals per unit area (Na) plotted as
a function of crystallinity (φ), as shown in Figure 8. Rapidly cooled channel samples show
a dramatic increase in crystal number (from 0 to 4000 mm−2) with increasing crystallinity,
indicating crystallization by the addition of crystals (nucleation-dominated crystallization) of
average size (d):
φ
d= (1)
Na
Envelopes enclosing the channel data show that the average crystal size in these samples
ranges from 6 to 9 mm. In contrast, slow-cooling lava tube and lava lake samples have lower,
and near-constant, crystal number densities (< 200-300 mm−2) but larger crystal sizes (20-100
µm), indicating that in these cases crystallization was growth-dominated.
Average Na data may be extrapolated to three dimensions by assuming that Na is related
to the volume-based number Nv as Na/d (based on intersection probabilities; e.g., Underwood
1970; Hammer et al. 1999). A more detailed description of crystal textures, however, is
provided by analysis of crystal size distributions (CSDs; e.g., Cashman and Marsh 1988). As
Reconstruction of Magmatic System Variables & Processes 191
5000
6 μm
KIL channel
Plagioclase number density (mm -2 )
KIL lava lake
d=
4000 ML channel
ML tube
μm
9
=
3000
d
d
te
ina
2000
om
-d
i on
1000 per area (mm2)
at
Number
cle
nu growth-dominated
0
0 0.1 0.2 0.3 0.4 0.5
Plagioclase volume fraction
Figure 8. Plagioclase crystal number (per area) plotted against plagioclase volume fraction for samples
collected both along active lava channels and from within lava tubes and lava lakes from Kilauea and Mauna
Loa volcanoes, Hawaii. Dashed lines illustrate samples that are related by the addition of crystals of similar
average size (labeled). We describe these crystallization trends as “nucleation-dominated”, as compared to
the “growth-dominated” trends where crystallinity increases are accomplished by changing the crystal size
instead of the crystal number. Data from Folley (1999), Riker (2005) and Cashman (unpublished).
CSD theory is reviewed in detail by Armienti (2008), here we summarize techniques that are
specifically relevant to the interpretation of volcanological processes. CSD theory utilizes the
slope of the cumulative number distribution (the density distribution n, with units of number
per volume per bin size) as a stable measure of the population distribution (e.g., Randolph and
Larson 1971). Under steady state conditions, n is related to crystal size (L) as
L
n = n° exp − (2 )
Ld
where n°, the nucleation density, is the intercept at L = 0, and Ld is the dominant size as
measured by the slope of the distribution (Ld = −1/slope). Figure 9a shows example CSDs for
the two (nucleation-dominated) Kilauea channel samples that are circled in Figure 8. These
samples have linear CSDs (i.e., they are well described by Eqn. 2) with near parallel slopes
(Ld = 3-4 mm). Note that Ld calculated from the CSDs is smaller than the average d of 9 µm
estimated from the Na − φ plot shown in Figure 8; this is because Ld is the mode of a distribution
that is positively skewed (hence the mean derived from Fig. 8 is larger than the mode derived
from the CSD). In fact, 3LD (the mode of the volume-based distribution; Cashman 1992) is
usually the closest approximation to the value d determined from Na − φ plots.
The extrapolated number density of nuclei-sized crystals, n°, indicates the importance
of nucleation in the crystallization process, while the area defined by the CSD curve provides
a measure of the total number of crystals per unit volume: NT = n°Ld. For example, the
nucleation-dominated channel samples shown in Figure 9a show an increase in NT over an
order of magnitude (from 3×104 mm−3 to 7×105 mm−3) as the plagioclase crystallinity increases
from 1 to 16%. In contrast, time-sequential CSDs for the growth-dominated Makaopuhi lava
192 Blundy & Cashman
20 14
(a) Kilauea channel (b) Makaopuhi lava lake
12
18
10
Ln n (/mm 4 )
16 8
φφ==
φφ=
=0
00..116 6 .44
φ φ=
14
6
=00.
φ=
φ=
1.1
4
00.0.01
12
1
10 0
0 0.01 0.02 0.03 0.04 0 0.1 0.2 0.3
L (mm) L (mm)
Figure 9. Plagioclase CSDs from time-sequential samples of basaltic lava. (a) Samples collected along
a channelized lava flow from Kilauea (circled points in Fig. 8) that show a nucleation-dominated trend
with similar slopes (crystal size) but different intercepts (crystal abundance). (b) Samples collected from
Makaopuhi lava lake that show a growth-dominated trend controlled by diminishing cooling rates with
depth in the lava lake (e.g., Cashman and Marsh 1988). Note scale differences between plots. Data from
Folley (1999) and unpublished data.
lake samples (Fig. 9b) have decreasing slopes (Ld increases from 17 to 36 µm) and decreasing
intercepts as plagioclase crystallinity increases from 10 to 41%. The net result is a slight
decrease in crystal number density (from 4000 to 1500 mm−3) as crystallization proceeds,
probably the result of ripening (e.g., Cashman and Marsh 1988).
Parameters determined from a CSD plot may be related to average rates of crystal nucleation
(J) and growth (G) by assuming that the dominant size (Ld) is a consequence of steady growth
over an appropriate duration of time (τ): Ld = Gτ. In industrial crystallization systems, τ is
the residence time of a system crystallizing at steady state. In geological applications, τ is the
effective crystallization time, as determined by some independent measure. Nucleation rate (J
= dNv /dt) is related to growth rate as J = n°G, and thus can be determined for any CSD for
which G is known. This relationship shows that n° is a direct measure of the ratio between the
rates of nucleation and growth (J/G) that are ultimately responsible for the final sample textures.
Reported values of n° vary over 11 orders of magnitude (from <101 to >1012 mm−4), illustrating
the wide range in crystallization conditions experienced in magmatic systems. In theory, this ratio
can be related directly to the undercooling, ∆T, if the appropriate thermodynamic parameters
are known (Armienti et al. 1994). As J is directly proportional to n°, however, estimates of J are
sensitive to the accuracy of crystal measurements at small sizes.
Finally, textural analysis also permits measurements of bubble and crystal orientations
and shapes. Orientation can be used to infer the dynamics of magma flow and emplacement
(e.g., Castro et al. 2002; Rust et al. 2003). Experimentally, it has been shown that crystal shape
is closely related to the conditions of crystallization, with higher degrees of supersaturation
leading to less equant shapes (more disparity between growth rates on different crystal faces,
e.g., Donaldson 1976; Lofgren 1980; Hammer and Rutherford 2002). As most measurements
are made in 2D, the biggest challenge to quantifying true crystal shapes lies in knowing the
intersection probabilities for crystals of different morphology (e.g., Higgins 1994). Frequency
distributions for different crystal shapes are presented by Morgan and Jerram (2006), who provide
a simple computer program for fitting 2D aspect ratio distributions to 3D crystal shapes that
Reconstruction of Magmatic System Variables & Processes 193
can be used with CSD Corrections (http://geologie.uqac.ca/~mhiggins/csdcorrections.html),
a computer program written by M. Higgins that calculates CSDs for different average particle
shapes. A volcanological application of crystal shape analysis is illustrated by a study of tephra
produced during early (pre-climactic) stages of the 1991 eruption of Mt. Pinatubo, Philippines
(Hammer et al. 1999). Crystallization in these samples was driven by rapid decompression
within the conduit followed by short (hours) periods of magma arrest at shallow levels prior to
eruption. Longer magma residence times at shallow levels within the conduit resulted in both
a change from nucleation-dominated to growth-dominated crystallization (as shown by CSDs)
and an accompanying change in plagioclase crystal habit from tabular to prismatic (Fig. 10).
In summary, textural analysis can be used to track the evolution of the crystal (or bubble)
populations in a volcanic sample. Measurement of crystal textures can then be combined
with micro-analytical techniques (above) and temperature and pressure constraints (below) to
determine not only the conditions of crystallization in magma storage regions but also phase
changes resulting from decompression and/or cooling during magma ascent and eruption. The
continued development of new imaging techniques (both physical – e.g., tomography – and
chemical – e.g., BSEM or nanoSIMS imaging) suggests that the power of combined physical
and chemical analysis of volcanic samples has only just begun to be realized.
CRYSTALLIZATION REGIME
nucleation - growth -
dominated dominated
1
25 tabular
Figure 10. Variations in plagioclase shape
0.8 (measured as I/L; ratio of intermediate:long
Ln n
CRYSTAL SHAPE
axis) with increasing crystallization
in pyroclasts from Pinatubo volcano,
0.6 20 Philippines. Inset shows CSD plots for
I/L
0 L (mm) 0.1 the samples identified by black symbols;
0.4
comparison of the two plots shows that
as crystallization becomes progressively
dominated by crystal growth, crystal
0.2 shapes become more equant. Data from
Hammer et al. (1999).
prismatic
0
0 0.10 0.20
volume fraction plagioclase
GeOtherMOMetry
Geothermometers and geobarometers use equilibrium chemical reaction(s) between two
or more phases (e.g., minerals, melt) in a rock. Because the extent of chemical equilibrium
is determined by kinetic factors, such as solid-state diffusion, reactions become inhibited
below a threshold (or “closure”) temperature. In volcanic rocks this is often considered to
be the eruption temperature, although care must be taken in cases where entrained wall rock
or cumulate material has not fully equilibrated with the melt prior to eruption or where the
component of interest, e.g., Al in clinopyroxene, has such low diffusivity that re-equilibration
at any time after initial crystallization is unlikely. In all cases, textural evidence for equilibrium
between the phases of interest must be examined before attempting thermobarometry.
The most widely used magmatic geothermometers involve exchange reactions between
two components in coexisting mineral-mineral or mineral-melt pairs, or net transfer reactions in
194 Blundy & Cashman
which one or more phases is consumed or produced. Thermobarometers require thermodynamic
calibration of the equilibrium constant for a particular reaction or set of reactions. In general,
geothermometers use reactions with large enthalpy changes, while barometers involve
reactions with large volume changes. Thermodynamic data used to calculate the equilibrium
constant may be derived calorimetrically or empirically. In either case a reliable description of
activity-composition (a-X) relationships is required for determining the equilibrium constant
from natural mineral or melt compositions.
Mineral-melt thermometry is reviewed by Putirka (2008) and a wide range of mineral-
mineral thermometers and barometers is reviewed by Anderson (2008); here we limit our
discussion to three commonly-used magmatic mineral-mineral thermometers:
1. Coexisting iron-titanium oxides (also used as an oxybarometer for the determination
of magmatic oxygen fugacity)
2. Coexisting amphibole and plagioclase, with or without a coexisting silica phase
3. Coexisting orthopyroxene and clinopyroxene
Our primary objective is not to describe the thermodynamics of the thermometers, which is
covered in detail elsewhere, but to assess their reliability in sub-volcanic magma systems by
applying them to experimental data not included in the original calibrations.
Iron-titanium oxide thermometry and oxybarometry
Many intermediate and acid volcanic rocks contain two discrete Fe-Ti oxide phases:
rhombohedral oxides close to the ilmenite (FeTiO3)-hematite (Fe2O3) binary and cubic oxides
close to the magnetite (Fe3O4)-ulvospinel (Fe2TiO4) binary. Minor components, notably
Mg, Mn, Al and zn, may substitute for Fe and Ti in these oxides. Minor components can
create deviations from the binary join that, in some cases, lead to significant modifications to
calculated temperatures. Bacon and Hirschmann (1988) use the Mg/Mn ratio of coexisting
oxides as a test for equilibrium:
Mg Mg
log = 0.9317+−00..0113
0104 × log − 0.0909+−00..0785
0787 (3)
Mn mt Mn ilm
where Mg and Mn are atomic fractions in magnetite (mt) and ilmenite (ilm) and the error
bounds denote the acceptable range for equilibrium.
The two key equilibria for iron-titanium oxide thermometry are:
FeTiO3 + Fe3O4 = Fe2TiO4 + Fe2O3 (4a)
4Fe3O4 + O2 = 6Fe2O3 (4b)
Reaction (4a) is an exchange reaction that can be used to estimate temperature; reaction (4b) is
a redox reaction that can be used to estimate fO2. In practice the two reactions are combined to
give a T- fO2 estimate (Fig. 11). The reaction volumes of both (4a) and (4b) are sufficiently small
that the effect of pressure on the T-fO2 estimate is negligible for most sub-volcanic pressures.
Development of geothermometers requires accurate calibration of a-X relationships, which
in the case of Fe-Ti oxides includes ordering in the spinel phase. Different formulations of a-X
relationships provided by Andersen and Lindsley (1988), Andersen et al. (1993), Ghiorso and
Sack (1991) and Ghiorso and Evans (2008) result in broadly similar T-fO2 estimates except
at very oxidizing conditions. A further complication is the calculation of oxide compositions
using EMPA data, because only rarely are Fe2+ and Fe3+ contents determined independently
(e.g., measuring oxygen by EMPA). Assumptions required about oxide stoichiometry, and
hence mole fractions of the various components, have led to several different recalculation
schemes. The simplest way to calculate T-fO2 from Fe-Ti oxides is to use freeware programs:
the Ghiorso and Evans (2008) formulation is available at http://ctserver.ofm-research.org/
Reconstruction of Magmatic System Variables & Processes 195
-3
-2
-1 -2
NNO
+1 -1
-0.5
0.8 NNO
1000°
+0.5
800°
600°
+1
0.6
+1.5
+2
XIl
+2.5
0.4
+3
0.2
0°
130
0.2 0.4 0.6 0.8
XUv
Figure 11. Plot of ilmenite mole fraction (Xil) in rhombohedral oxide versus ülvospinel mole fraction (xuv)
in coexisting cubic oxide, contoured for temperature (°C) and fO2 (log10 units relative to NNO buffer) in the
system Fe-Ti-O. The dashed lines denote the miscibility gap in the rhombohedral oxide phase; the presence
of minor components profoundly affects the size and shape of this gap. Figure from Ghiorso and Evans
(2008) kindly supplied by M. Ghiorso.
OxideGeothrm/OxideGeothrm.php; the formulation of Andersen and Lindsley is included in
the qUILF package (Andersen et al. 1993). The reader is also referred to ILMAT, a useful
downloadable Excel spreadsheet developed by LePage (2003) that calculates T-fO2 from pairs
of oxides using different thermometer formulations of a-X relationships and recalculations of
oxide formulae from EMPA data.
Fe-Ti oxides have been shown experimentally to re-equilibrate quickly in response to
changes in temperature and/or fO2 (Hammond and Taylor 1982; Venezky and Rutherford
1999). The rate at which re-equilibration occurs depends on the size of the grains and their
physical proximity. A common practice is to analyze several grains of titano-magnetite and
several of ilmenite (typically much scarcer in relatively oxidized volcanic rocks) and use the
averages in T-fO2 calculations. Equilibrium between the two averaged compositions is normally
checked using Equation (3). A problem with this approach is that, because the analyzed grains
may vary in size, it is never clear whether all of the ilmenites are in equilibrium with all of
the magnetites, particularly as their co-occurrence in a single volcanic sample is not in itself
196 Blundy & Cashman
evidence for textural equilibrium. For example, eruptive processes can mix crystals that grew
in different parts of a magma chamber.
A more reliable approach (Blundy et al. 2006) is to analyze only adjacent Fe-Ti oxide
pairs that either share a common boundary (i.e., are in contact) or occur as inclusions in the
same silicate phenocryst phase. In this way it is possible to ensure textural equilibrium, even
in the case of zoned crystals (e.g., Devine et al. 2003), and to distinguish between crystals with
different growth histories (e.g., phenocrysts, inclusions, microlites etc.). Each touching pair
can then be checked for equilibrium using Equation (3) and used to calculate an individual
T and fO2. This approach yields information on the temporal evolution of T and fO2 that is
lost through conventional averaging procedures. We assume that a homogeneous set of T-fO2
estimates from all pairs will result only if the entire oxide population, including inclusions
in phenocrysts, reached full chemical equilibrium prior to eruption. This is unlikely for most
volcanic rocks, despite the relatively rapid rates of diffusive re-equilibration in oxides.
An example of this approach is provided by Fe-Ti oxides from Mount St. Helens vol-
cano. Previous T- fO2 estimates for the 1980-86 dacite magmas, based upon averaged analyzes,
yielded temperatures that ranged from 860 to 1030 °C (Melson and Hopson 1981; Rutherford
et al. 1985; Rutherford and Devine 1988; Rutherford and Hill 1993). Blundy et al. (2006) used
touching (or co-included) pairs from samples spanning the 1980-81 phase of the eruption to
demonstrate systematic variations between individual crystal types, both in a single sample and
between different samples. Specifically, the lowest temperatures (880 °C) derived from inclu-
sions in phenocrysts from the May 18th Plinian deposit, whilst microlites and phenocrysts from
all subsequent eruptions showed elevated temperatures, which Blundy et al. (2006) attributed
to magma heating by latent heat release during decompression crystallization. Figure 12 shows
an expanded Mount St. Helens T-fO2 dataset, extending from 1980 to 1986 (Blundy et al. 2008).
Some of the lower temperatures recorded by cryptodome and May 1985 oxide pairs probably
reflect entrainment of cooler material from the walls of the conduit. Significantly, oxide pairs
from samples containing microlites formed during slow ascent and decompression crystalliza-
tion showed systematically lower fO2 than phenocrysts from the Plinian deposit (Fig. 12; see be-
low). These features would not have been apparent if averaged oxide compositions were used.
How well does the thermometer work? To test the reliability of the different Fe-Ti
oxide thermometers discussed above, we have applied them to the analyzed run products
of experiments that (1) crystallized both oxides and (2) were not included in any of the
calibrations. A total of 75 additional oxide pairs were taken from the following experimental
studies: Venezky and Rutherford (1999), Beard and Lofgren (1991), Blatter and Carmichael
(2001), Bogaerts et al. (2006), Dall’Agnol et al. (1999), Gardner et al. (1995), Martel et al.
(1998), Pichavant et al. (2002), Rutherford and Devine (1988), Scaillet and Evans (1999).
These data cover a wide range of pressure (50-690 MPa) and temperature (707-1000 °C). For
the fO2 calculations we have included only a subset (n = 30) of the data where the experimental
fO2 was buffered either by solid oxygen buffers (e.g., Ni-NiO or Re-ReO2) or by controlling
the partial pressure of H2 using a Shaw membrane. The resulting range in experimental fO2 is
from NNO-0.1 to NNO+2.8 log units.
Temperature and fO2 were calculated from the following formulations:
1. Andersen and Lindsley (1988) thermometer with Lindsley and Spencer (1982) oxide
formula recalculation
2. Andersen and Lindsley (1988) thermometer with Stormer (1983) oxide formula
recalculation
3. Ghiorso and Evans (2008)
4. qUILF (Andersen et al. 1993) using “FeTi” and “HM” reactions only.
Reconstruction of Magmatic System Variables & Processes 197
Figure 12. Temperature- fO2 plot of Fe-Ti oxides in various eruptive phases of the 1980-86 eruption of
Mount St. Helens. All data are taking from touching pairs of oxides or oxides co-included in the same
silicate phenocryst phase. The NNO buffer is shown for reference. Only the plinian eruption of May 18th,
1980 (filled circles) lacks microlites; all other samples (open symbols) are microlite-bearing due to slower
ascent and decompression crystallization. Oxide data are from Blundy et al. (2008).
For each formulation we calculated the average absolute deviation (aad) on T and fO2 for the
experimental dataset. For T, aad ranges from 33-60 °C (Fig. 13). The smallest aad comes from
the thermometer of Andersen and Lindsley (1988) although this formulation fails to capture
the full range in T, with systematic underestimates above 860°C and overestimates below.
In contrast, Ghiorso and Evans (2008) and qUILF are more scattered, but reproduce the full
range of T, although qUILF shows a systematic underestimate at all T. fO2 estimates from
Andersen and Lindsley (1988) and Ghiorso and Evans (2008) have similar aad of 0.25-0.34
log units; qUILF consistently underestimates fO2 for this relatively oxidizing experimental
dataset (aad = 1.09 log units).
amphibole-Plagioclase thermometry
Many intermediate and acid volcanic rocks contain coexisting amphibole and plagioclase.
The geothermometer of Holland and Blundy (1994) is based on two exchange equilibria
involving these phases:
edenite + 4 quartz = tremolite + albite (5)
edenite + albite = richterite + anorthite (6)
Equilibrium (5) is applicable to silica-saturated rocks, whereas (6) is applicable to silica-
undersaturated rocks. Holland and Blundy (1994) developed a-X models for both amphibole
and plagioclase and calibrated the thermometers (available at http://rock.esc.cam.ac.uk/astaff/
holland/hbplag.html) on coexisting amphibole-plagioclase pairs from a large number of
metamorphic rocks and experimental studies to cover a wide range of P-T and composition.
As both thermometers have modest (±72 °C GPa−1) pressure dependence at pressures of sub-
volcanic magma chambers, a pressure of 200 or 300 MPa can be assumed in the absence of
other constraints. Holland and Blundy (1994) report an aad of ±28 °C for thermometer (5) and
198 Blundy & Cashman
Figure 13. Comparison of calculated Fe-Ti temperatures for 75 experimental oxide pairs from studies
not used in original thermometer calibrations. Four thermometers are shown: (a) Andersen and Lindsley
(1988) thermometer with Lindsley and Spencer (1982) oxide formula recalculation; (b) Andersen and
Lindsley (1988) thermometer with Stormer (1983) oxide formula recalculation; (c) Ghiorso and Evans
(2008); and (d) qUILF-95 (Andersen et al. 1993). For each thermometer we show the number of points
(n) and the average absolute deviation (aad). Data references are provided in the text. Note that we use aad
in preference to Root Mean Square Error (RMSE) as aad gives a better impression of the average misfit of
calculated and experimental temperatures for relatively small datasets with outliers. The latter can strongly
skew the RMSE to high values.
±32 °C for thermometer (6) over a temperature range of 400-1000 °C and 100 to 1500 MPa
pressure. Anderson et al. (2008) further consider the application of reactions (5) and (6) to
granitic rocks and review the potential use of amphibole as a thermobarometer.
As with all thermometers, care must be taken to ensure that the amphibole-plagioclase
pairs are in textural equilibrium. Both phases are commonly zoned, so it is essential to use the
outermost rims of phenocrysts for thermometry, or touching pairs of crystals where possible.
Many volcanic rocks contain amphibole phenocrysts with plagioclase inclusions and vice
versa that can be used for thermometry provided that the immediately adjacent host mineral
is analyzed (e.g., Bachmann and Dungan 2002; Couch et al. 2003a). As in the case of Fe-Ti
oxides, we would caution against taking average compositions of amphibole and plagioclase
from an entire thin section; instead we recommend that temperatures be calculated from a
large number of (touching) pairs. The use of core-core pairs is only recommended where
Reconstruction of Magmatic System Variables & Processes 199
there is clear textural evidence that plagioclase cores and amphibole cores were precipitating
simultaneously from the same melt; this is rarely easy to establish.
How well does the thermometer work? We have assembled a dataset of 294 amphibole-
plagioclase pairs from experiments at magmatic temperatures published since 1994 and
therefore not included in the original thermometer calibration. The dataset covers a pressure
range of 42 to 1500 MPa and 640 to 1050 °C. The Holland and Blundy (1994) thermometers
were used to calculate T for each experiment at the experimental P. As few experiments
contain quartz in addition to amphibole and plagioclase, all temperatures were calculated
with thermometer (6). The average absolute deviation between calculated and experimental
temperature for the entire dataset is 61 °C (Fig. 14a). The deviation correlates negatively with
the Mg# of the amphibole (Fig. 14b), suggesting that the a-X relationships of Holland and
Blundy (1994) do not adequately address non-ideality between Fe2+ and Mg on the M1, M2
and M3 crystallographic sites. Evidently, in magmas with high Mg# amphibole-plagioclase
thermometry offers a less accurate means of estimating magmatic temperatures than Fe-Ti
oxide thermometry, but nonetheless a valuable one in rocks that lack two oxides.
two-pyroxene thermometry
Many intermediate and acid volcanic rocks contain coexisting ortho- and clinopyroxene.
The reader is referred to Putirka (2008) for a discussion of the various pyroxene equilibria upon
which thermometry is based and the techniques used for recalculating pyroxene structural
formulae.
Figure 14. Application of Thermometer B of Holland and Blundy (1994) to 298 experimental amphibole-
plagioclase pairs not included in original calibration dataset. (a) comparison of calculated and experimental
temperatures; (b) difference between calculated and experimental temperature as a function of amphibole
Mg#. The aad of the calculated temperature is 61 °C for the entire dataset. The negative correlation between
calculated temperature and Mg# in (b) suggests that there is additional non-ideality on the amphibole
M1-3 sites not fully addressed by the thermometer. Experimental data from Nakajima and Arima (1998),
Bogaerts et al. (2006), Pichavant et al. (2002), Grove et al. (1997), Klimm et al. (2003), Rutherford and
Devine (2008), Gardner et al. (1995), Donato et al. (2006), McCanta et al. (2007), Larsen (2006), Brooker
and Blundy (unpublished data), Poli (1993), Rapp and Watson (1995), Ernst and Liu (1998), Sisson et
al. (2005), Barclay and Carmichael (2004), Blatter and Carmichael (2001), Costa et al. (2004), Holtz et
al. (2005), Kawamoto (1996), Martel et al. (1999), Moore and Carmichael (1998), Prouteau and Scaillet
(2003), Sato et al. (1999), Scaillet and Evans (1999), Schmidt and Thompson (1996), Patino-Douce and
Beard (1995), Alonso-Perez et al. (2008), López and Castro (2001), Nicholis and Rutherford (2004),
Johannes and Koepke (2001), Molina and Poli 2000, Kaszuba and Wendlandt (2000), Couch (2002),
Gardien et al. (2000), Sen and Dunn (1994), Dall’Agnol et al. (1999).
200 Blundy & Cashman
Three formulations of the two-pyroxene thermometer are in widespread usage:
1. Wells (1977)
2. Brey and Kohler (1990)
3. qUILF (Andersen et al. 1993)
How well do the thermometers work? We have assembled a dataset of 142 experimental
two-pyroxene pairs from experiments at magmatic temperatures published since 1996. Data
span a range of 0.1 to 1500 MPa and 819 to 1230 °C (Figure 15). The Wells (1977) thermometer
(not shown), which has no pressure dependency, has an aad of 99 °C for the entire dataset.
The aad of the Brey and Kohler (1990) thermometer is 66 °C (Fig. 15a) and comparable to
that of the amphibole-plagioclase thermometer, although the errors are much larger than the
±30 °C (2 s.d.) quoted by Brey and Kohler (1990). The thermometer is much more reliable at
temperatures above 1000 °C, and is therefore better suited to andesitic and basaltic magmas.
qUILF (Fig 15b) systematically overestimates all experimental temperatures with an aad of
114 °C. However the uncertainty calculated qUILF is demonstrably more realistic and nearly
50% of the calculated temperatures fall within error of the experimental temperature.
Figure 15. Comparison of calculated two-pyroxene and experimental temperature for 142 experimental
pyroxene pairs using the thermometers of (a) Brey and Köhler (1990); and (b) qUILF-95 (Andersen
et al. 1993). The aad for each thermometer is given. Error bars are those provided by original authors.
Experimental data from Singh and Johannes (1996), Alonso-Perez et al. (2008), Moore and Carmichael
(1998), Larsen (2006), Lopez and Castro 2001, Johannes and Koepke (2001), Grove et al. (2003), Sisson
et al. (2005), Grove et al. (1997), Dall’agnol et al. (1999), Pichavant et al. (2002), Costa et al. (2004),
Kawamoto (1996), Martel et al. (1999), Prouteau and Scaillet (2003), Sato et al. (1999), Blatter and
Carmichael (1998, 2001), Pertermann and Lundstrom (2006), Pichavant and Macdonald (2007), Patino-
Douce and Beard (1995), Müntener et al. (2001).
GeOBarOMetry
In nature there are far fewer reactions suitable for magmatic geobarometry than for
geothermometry at conditions appropriate for sub-volcanic systems. The aluminum-in-
hornblende barometer—which relies on the presence of a low-variance mineral assemblage—
is discussed by Anderson (2008), while Putirka (2008) discusses the barometric potential
of some mineral-melt equilibria. Here we consider four alternative methods for estimating
pressure in sub-volcanic systems:
Reconstruction of Magmatic System Variables & Processes 201
1. Experimental reproduction of phase assemblages and compositions in natural
volcanic rocks
2. The saturation pressure of volatiles in melt inclusions
3. The chemical composition of silica-rich glasses in equilibrium with quartz and
feldspar
4. The proportion and number of plagioclase crystals.
experimental reconstruction
For multiply-saturated volcanic rocks it is possible to reconstruct the pre-eruptive magma
storage conditions using experiments designed to reproduce the phase assemblage and phase
compositions of natural samples. This approach requires two key assumptions. The first is that
the crystallizing assemblage represents equilibrium at the moment of eruption and the second
is that the assemblage equates to equilibrium at a unique set of pressure, temperature and
volatile conditions. The ubiquity of zoned crystals, especially plagioclase, in most volcanic
rocks renders both of these assumptions critical when considering whole-rock samples.
As crystallization over a range of pressures is probably the norm in many natural systems,
experiments should be ‘tuned’ to examine specific parts of the system (such as the uppermost
pre-eruptive storage region).
Pichavant et al. (2007) elegantly address experimental investigations of magma storage
by examining the equilibration length-scales that operate in natural magmas, and the resulting
effective composition, or “reactive magma,” that can be studied experimentally. If the reactive
magma at the pre-eruptive storage conditions is only a subset of the bulk rock volume, e.g.,
it excludes the cores of zoned crystals, then there is relatively little virtue in performing
experiments on the bulk sample. As an illustration, Pichavant et al. (2007) demonstrate that
starting experiments with a rock powder that contains xenocrystic calcic plagioclase will
prevent determination of a unique set of pre-eruptive conditions because the starting material
deviates substantially from the true reactive magma composition. “Partial equilibrium”
experiments can be used to circumvent this problem, whereby natural samples are lightly
crushed, thus chemically isolating crystal cores on the timescales of typical experimental runs.
An alternative approach is to attempt to equilibrate the matrix glass composition with crystals
of the same composition as the rims of the phenocrysts (e.g., Couch et al. 2003a). In practice
this can be done either by using a glass separate from the natural sample as the experimental
starting material, or by synthesizing a starting composition based on analysis of glass in the
natural sample. If there are nucleation problems, the glass starting material may be seeded
with tiny amounts (1-5%) of the desired crystallizing assemblage and checked for overgrowth
or dissolution textures in the resultant run products. Implicit in this approach is the assumption
that the reactive magma includes the groundmass glass and the phenocryst rims.
Typical pre-eruptive storage conditions of 0-500 MPa make externally-heated hydrothermal
(“cold-seal”) or gas-pressure (e.g., TzM) vessels, or internally-heated gas pressure (IHPV)
vessels, the ideal experimental techniques for these studies. Pressure and temperature can be
controlled precisely in both externally- and internally-heated apparatus. Temperature accuracy
is considerably greater in IHPV because the thermocouple sits adjacent to the sample, in
contrast to externally heated pressure vessels where the thermocouple is located inside the
walls of the pressure vessel.
As fO2 has a profound effect on phase relations and compositions, this intensive variable
must also be controlled. In IHPV a small amount of H2 can be added to the argon pressurizing
gas and the fugacity of H2 measured using a Shaw membrane (e.g., Martel et al. 1998, 1999;
Scaillet and Evans 1999). fO2 can then be calculated using the disproportionation reaction of
H2O,
202 Blundy & Cashman
2H2O = O2 + 2H2 (7)
log fO2 = log K w − 2 log fH2 + 2 log fH2 O (8)
Here Kw is the disproportionation constant for H2O (e.g., Robie et al. 1978) and fH2O is the
fugacity of H2O in the sample, which will be the fugacity of the pure gas in H2O-saturated
experiments. Provided that H2 can diffuse through the noble metal sample capsule on the
timescales of the experiment, fH2 inside the capsule equates to that in the pressurizing gas.
Although Au is relatively impermeable to H2 (Chou 1987), it can be used as a capsule material
in long duration experiments given a sustained source of H2, e.g., Ar-H2 gas mixtures.
In an externally-heated apparatus an (inner) sample capsule is embedded in an outer
capsule containing a solid buffer (e.g., Ni and NiO, plus a small amount of H2O; Chou 1987). fO2
is buffered via exchange of H2 between the sample and the buffer across the walls of the inner
capsule. The sample capsule must be sufficiently permeable to H2 (e.g., AgPd, AuPd, Pt) that
the external buffer and sample can attain the same fH2, whereas the outer capsule should then be
relatively impermeable to ensure that H2 is not lost from the system during the experiment.
Where the sample is undersaturated in H2O, the fO2 is buffered at a value that deviates
from that of the pure buffer by an amount controlled by the H2O activity in the sample capsule
(e.g., Klimm et al. 2003; Sisson et al. 2005). Specifically, as H2O activity in the undersaturated
sample is less than 1, the fO2 in the sample is buffered at a value below that of the buffer (i.e.,
more reducing). Except for highly H2O-undersaturated melt or H2O-poor fluids, the deviation
of the sample fO2 from that of the buffer is quite small, e.g., 0.3 log units lower than the buffer
for aHmelt
2O
= 0.7.
Starting conditions for equilibrium experiments typically use an initial temperature, fO2 and/
or pressure obtained from mineral thermobarometers to explore the neighbouring P-T-H2O-fO2
space for the best match to the natural assemblage. The extent to which this approach affords
a unique solution depends on the number of degrees of freedom of the natural assemblage,
the uncertainties in the petrologically-estimated P, T and fO2, and the variability in the natural
phase compositions. The greater the number of equilibrium crystal phases, the fewer the
degrees of freedom and the more unique the solution. Consider, for example, a volcanic rock
containing phenocrysts of amphibole, plagioclase, orthopyroxene, ilmenite and magnetite in
equilibrium with a vesicular (vapor-saturated) glass. The total number of major components
is likely to be 11 (SiO2, TiO2, Al2O3, FeO, Fe2O3, MgO, CaO, Na2O, K2O, H2O, CO2). The
total number of degrees of freedom in this 7-phase assemblage is 6. Thus even at fixed P, T,
fO2 and vapor composition there are still 2 additional degrees of freedom. However, if the rock
also contains clinopyroxene and alkali-feldspar, then the coexistence of these phases with glass
would constitute a unique solution. The degrees of freedom may also be limited by independent
compositional parameters, such as the anorthite content of the plagioclase, the Mg# of the
pyroxene or amphibole, the Al2O3 content of the glass, or the proportion of a particular phase.
To illustrate the principles of the experimental approach, Figure 16 shows the results of a
careful experimental investigation of pre-eruptive conditions for a Holocene dacite from Volcán
San Pedro, Chile (Costa et al. 2004). Oxide-thermometry indicates a pre-eruptive temperature
of 850 °C. For experiments conducted at this temperature and 200 MPa pressure we compare
7 independent parameters (crystallinity, modal plagioclase: amphibole ratio, SiO2, Al2O3 and
CaO in the glass, anorthite content (XAn) in plagioclase and Al2O3 content of amphibole) with
those in the natural sample with respect to the bulk H2O content of the magma. We note that
all 7 parameters show a consistent value of 5.1±0.4 wt% H2O, which overlaps with the H2O
content of glasses in experiments that yield the correct phase assemblage (Fig. 16, shaded
region). These results show a clear example of a reactive magma composition that closely
matches that of the bulk dacite used in the experiments. The small deviation of H2O content
estimated from glass CaO content, compared to other parameters, suggests that the natural
Reconstruction of Magmatic System Variables & Processes 203
Figure 16. Summary of experimental phase relations for Volcán San Pedro, Chile, dacite from Costa et
al. (2004). Results are expressed in terms of the bulk magmatic H2O content that best matches the seven
parameters from the natural sample listed on the ordinate axis: crystallinity; weight plagioclase:hornblende
(Plag/Hbl) ratio; wt% SiO2, Al2O3 and CaO in glass (gl); anorthite content of plagioclase (XAn) and Al2O3
content of hornblende. The data points are derived by linear regression of each parameter against H2O for
experiments performed at 850 °C and 200 MPa, the estimated pre-eruptive temperature and pressure, The
error bars are based on uncertainty in the regressions and variability in the natural sample. The shaded bar
denotes the range of H2O contents that yield the correct phenocryst assemblage (±biotite). The convergence
of all parameters on a pre-eruptive H2O content of 5.1±0.4 wt% suggest that the bulk rock approximates
the reactive magma at San Pedro, although the slight mismatch for CaO (gl) suggests the presence of small
amounts of calcic plagioclase cores.
sample may contain a small quantity of calcic plagioclase cores that are not reactive, a finding
consistent with the petrographic description of rare An-rich plagioclase cores in the San Pedro
dacite (Costa et al. 2004). Pichavant et al. (2007) draw a similar conclusion about the Mont
Pelée reactive magma composition based on the experiments of Martel et al. (1998, 1999).
How well does the barometer work? In Table 1 we list 40 published experimental studies
designed to constrain pre-eruptive magma storage conditions at 25 different volcanoes. All of
these studies provide a close match for many of the observed parameters in the natural samples.
However, in most cases the match is not as close as observed for Volcan San Pedro (Fig. 16)
and some parameters cannot be adequately matched at a single experimental condition.
A failure to match phase proportions and compositions at any reasonable permutation of
P-T-H2O-fO2 conditions most likely reflects a failure of the natural sample (or the experiments)
to attain equilibrium, omission of an important compositional variable such as CO2, S or F,
or violation of the reactive magma principle discussed above. For example, based on the ex-
periments of Rutherford et al. (1985) and Rutherford and Devine (1988), Blundy and Cashman
(2001) show that the experimentally determined P-T-H2O-fO2 conditions that best match the
phase compositions of Mount St. Helens dacite, differ from the conditions that best match the
phase proportions. One possible explanation for this discrepancy, as noted above, is that the ex-
perimental starting material contains calcic plagioclase cores that are relicts from earlier phases
of the crystallization history of the magma (e.g., Cooper and Reid 2003; Berlo et al. 2007) and
therefore not present in the reactive magma. Despite these caveats, however, careful experi-
mental exploration of phase equilibrium space has been a cornerstone of magma petrology and
continues to provide critical constraints on magma storage conditions (e.g., Hammer 2008).
Volatile saturation pressures
Another means of determining pressure is provided by the strong pressure dependence
of the solubility of H2O and CO2, the dominant magmatic volatile species, in silicate melts.
204 Blundy & Cashman
table 1. Experimental studies designed to reproduce pre-eruptive storage conditions beneath
andesite and dacite volcanoes. Table lists the volcano studied and the P-T range investigated.
P range
Source Volcano T range (°C)
(MPa)
Arce et al. (2006) Nevado de Toluca, Mexico 100-250 780-880
Barclay and Carmichael (2004) Jorullo, Mexico 40-300 950-1100
Barclay et al. (1998) Soufrière Hills, Montserrat 50-200 800-940
Blatter and Carmichael (1998) zitácuaro, Mexico 0.1-292 950-1150
Blatter and Carmichael (2001) Valle de Bravo, Mexico 52-301 909-1140
Browne and Gardner (2006) Redoubt, AK 30-250 775-900
Coombs and Gardner (2001) Novarupta, AK 20-200 760-900
Coombs et al (2000) Southwest Trident, AK 50-200 1000-1100
Coombs et al (2002) Southwest Trident, AK 90 890-1000
Costa et al. (2004) San Pedro, Chile 54-405 800-950
Cottrell et al (1999) Santorini, Greece 50-250 800-1050
Couch et al (2003a) Soufrière Hills, Montserrat 5-225 825-1100
Donato et al. (2006) Salina Island, Italy 200-300 755-825
Gardner et al. (1995) Mount St. Helens, WA 100-350 850
Geschwind and Rutherford (1992) Mount St. Helens, WA 100-390 760-930
Grove et al (2003) Shasta, CA 0.1-200 940-1250
Grove et al. (1997) Medicine Lake, CA 100-200 865-1050
Hammer et al. (2002) Novarupta, AK 15-225 850-1050
Holtz et al (2005) Unzen, Japan 100-300 775-875
Johnson and Rutherford (1989) Fish Canyon, CO 200-500 740-930
Larsen (2006) Aniakchak, AK 50-200 820-950
Luhr (1990) El Chichón, Mexico 100-400 800-1000
Martel et al (1998) Mont Pelée, Martinique 200-311 850-925
Martel et al. (1999) Mont Pelée, Martinique 200-400 850-1040
McCanta et al (2007) Black Butte, CA 100-450 800-950
Merzbacher and Eggler (1984) Mount St. Helens, WA 100-400 850-1000
Moore and Carmichael (1998) Colima, Mexico 0.1-303 900-1251
Nicholls et al. (1992) Taupo, New zealand 200-500 700-800
Parat et al (2008) Fish Canyon, CO 400 850-950
Pertermann and Lundstrom (2006) Arenal, Costa Rica 500 1100-1300
Pichavant et al. (2002) Mont Pelée, Martinique 400 950-1025
Prouteau and Scaillet (2003) Pinatubo, Philippines 400-980 750-995
Rutherford and Devine (1988) Mount St. Helens, WA 220-320 890-970
Rutherford and Devine (2003) Soufrière Hills, Montserrat 10-250 780-950
Rutherford and Devine (2008) Mount St. Helens, WA 70-260 840-940
Rutherford and Hill (1993) Mount St. Helens, WA 120-220 790-940
Rutherford et al (1985) Mount St. Helens, WA 0.1-320 847-1190
Sato et al (1999) Unzen, Japan 29-196 800-1032
Scaillet & Evans (1999) Piantubo, Philippines 220-390 750-900
Venezky and Rutherford (1997) Mount Rainer, WA 25-200 900-945
Venezky and Rutherford (1999) Unzen, Japan 40-200 750-1050
Reconstruction of Magmatic System Variables & Processes 205
H2O and CO2 solubilities are dependent on melt and vapor compositions, for which various
experimentally-calibrated solubility relationships are available. Perhaps the easiest relationship
to apply is that of Newman and Lowenstern (2002), available as an Excel spreadsheet,
VolatileCalc. This formulation has separate relationships for rhyolite and basalt liquids, but is
not calibrated for andesite liquids, which lack experimental constraints. However, as most melt
inclusions in andesites are rhyolite or dacite in composition, this is not a critical problem, Volatile
solubility is weakly dependent on temperature, therefore an estimate of magmatic temperature
is required to calculate the saturation pressure (see above). For a given concentration of H2O
and CO2, an uncertainty of ±50 °C in temperature translates to an uncertainty in saturation
pressure of approximately ±10 MPa at 200 MPa and ±5 MPa at 100 MPa. Alternative solubility
models for mixed H2O-CO2 volatile phase include those of Tamic et al. (2001), which is only
valid from 75-200 MPa, and Liu et al. (2005), valid from 0 to 500 MPa.
Magma storage pressures can be estimated from the volatile contents of melt inclusions (e.g.,
Wallace et al. 1999; Schmitt 2001; Liu et al. 2006) provided there is evidence that the melt was
volatile-saturated at the time the melt inclusions became sealed. The only unequivocal evidence
for volatile saturation in a magma containing H2O and CO2 is a decrease in CO2 content of melt
inclusions with increasing crystallization, as revealed, for example, by increasing contents of
incompatible trace elements (e.g., Wallace et al. 1999). This is because gas loss, whether driven
by isobaric volatile-saturated crystallization or by decompression of volatile-saturated magma,
will always drive CO2 preferentially into the vapor phase, leading to its preferential depletion,
relative to H2O, in the melt. In cases where the melt is not volatile-saturated, the estimated
pressure based on volatile saturation will underestimate the true storage pressure. Note that the
presence of gas bubbles in a melt inclusion does not necessarily indicate that the magma was
gas-saturated as bubbles can also form by shrinkage during cooling.
Melt inclusions can provide information not only on the pressures at which magma
was stored pre-eruptively, but also on the relationship between degassing and crystallization
(Moore 2008). The different paths described by melt inclusions generated under a variety of
different scenarios are described in detail in a later section. At active volcanoes, the calculated
pressures from melt inclusions can be converted into depth, using a suitable density model,
and used to corroborate independent constraints on magma storage depths, for example from
seismology or geodetic measurements (see below).
Different pressure constraints are provided by groundmass glasses, which are invariably
degassed such that saturation pressures, although accurate, pertain to the process of eruption
(specifically, the pressure at which the magma passed through the glass transition temperature)
rather than to conditions of magma storage. As kinetic factors may inhibit diffusion of H2O and
CO2 into bubbles during eruption (e.g., Gonnermann and Manga 2005), pressures recorded in
matrix glass may be over-estimates. Nonetheless, in the case of Mount St. Helens dacites, H2O
saturation pressures of 2 to 40 MPa (mean of 12 measurements = 16±12 MPa at an assumed
eruption temperature of 900 °C) in the groundmass glass from the May 18th, 1980 plinian phase
(Blundy and Cashman 2005; Humphreys et al. 2008a) agree well with the calculated pressure
range over which the Mount St. Helens magma crossed the glass transition temperature (e.g.,
Carey and Sigurdsson 1985; Papale et al. 1998), suggesting that, in this case at least, kinetic
effects played a minor role in limiting effective H2O escape to bubbles.
How well does the barometer work? The principal limitation of using volatile solubility
as a barometer is the obvious one that the magma must contain crystals with melt inclusions
sufficiently large to analyze, as well as the need to distinguish equilibrium volatile saturation
at the time of entrapment from post-entrapment processes, and the compositional dependence
of volatile solubility. Some degree of disequilibrium may be inevitable for rapidly erupted
volcanic liquids; where diffusion-limited disequilibrium has occurred, calculated saturation
206 Blundy & Cashman
pressures will be overestimates. The compositional dependence of volatile solubility is well
documented in simple systems (e.g., Johannes and Holtz 1996; Moore 2008), although it
is rarely taken into account in solubility models, which tend to be based on generic melt
compositions. Deviations in solubility from the models are likely to be greatest for peralkaline
and peraluminous melts.
To gauge the reliability of saturation pressures calculated from dissolved volatile
contents we have assembled a dataset of 118 published, experimentally-determined H2O±CO2
measurements (SIMS or FTIR analyzes only) on silicic to intermediate melt compositions,
none of which were used in the original solubility model calibrations. We have used both static
and decompression experiments to encompass the likely range of behaviors in natural volcanic
rocks. We have not, however, included experiments that demonstrate homogeneous nucleation,
where substantial supersaturations are possible (e.g., Mangan et al. 2004b). We have used
the experimental temperature for our calculations and test the models of both Newman and
Lowenstern (2002) and Liu et al. (2005); at all pressures below 300 MPa the two models agree
to within < ±5 MPa.
The agreement between calculated and experimental pressures is remarkably good for all
experiments at P < 400 MPa (Fig. 17). The large negative deviations from calculated pressures
apparent in the 400 MPa experiments of Scaillet et al. (1995) are almost certainly the result
of low quench rates in these IHPV experiments leading to some degassing during quenching.
Ignoring those data, we calculate an aad between calculated and experimental pressures of 22
MPa for Newman and Lowenstern (2002) and 23 MPa for Liu et al. (2005). The aad would
increase if we had to rely on alternative means of calculating temperature (e.g., from mineral
thermometers). To assess the importance of melt composition on solubility we have also
calculated saturation pressures for 17 glasses from the phonolite decompression experiments
of Larsen and Gardner (2004). Despite the obvious compositional differences to the calibrant
dataset, the calculated pressures are still good to ±35 MPa. Evidently small differences in melt
composition from the rhyolites used in the solubility models propagate to relatively small
uncertainties in calculated pressures.
Figure 17. Calculated volatile solubility pressures. (a) Model of Liu et al. (2005); (b) VolatileCalc from
Newman and Lowenstern (2002). Pressures calculated at experimental temperature. Data from Humphreys
et al. (2008a), Martel et al. (1999) Pichavant et al. (2002), Gardner et al. (1995) Martel and Schmidt (2003),
Mangan and Sisson (2000), Mangan et al. (2004a), Gardner (2007), Gardner et al. (1999), Larsen and
Gardner (2004), Scaillet et al. (1995).
Reconstruction of Magmatic System Variables & Processes 207
haplogranite projection
An alternative means of estimating pressure, especially for rocks that lack suitable melt
inclusions for volatile analysis, is to use the well known relationship between silica solubility
in melts and pressure. Since the seminal work of Tuttle and Bowen (1958) it has been known
that in the hydrous orthoclase-albite-silica (Or-Ab-Qz) “haplogranite” system, the silica content
of melts saturated in feldspar(s) and quartz increases appreciably with decreasing pH2O. Sub-
sequent experimental studies, summarized in Johannes and Holtz (1996), have confirmed and
refined Tuttle and Bowen’s conclusions. At pressures of ≥ 500 MPa, there is a miscibility gap
between K- and Na-feldspar and the ternary minimum becomes a eutectic connected to the Ab-
Or binary by a two-feldspar cotectic. At pressures below 500 MPa, the Qz-Ab-Or-H2O system
is characterized by two primary phase volumes (quartz and feldspar) separated by a curved
cotectic (Fig. 18). Because of complete solid solution between albite and orthoclase on the soli-
dus at low pressure, the cotectic contains a minimum rather than a eutectic. For H2O-saturated
conditions, the silica phase volume decreases (i.e., silica solubility increases) with decreasing
pressure, the minimum moves towards the Qz-Or join, and the temperature of the minimum
(or eutectic) increases from 645 °C at 500 MPa, to 680 at 200 MPa, to ~950 °C at atmospheric
pressure (Tuttle and Bowen 1958). At any given pressure, H2O-undersaturation has little effect
on the position of the cotectic, but shifts the minimum towards the Qz-Or join and increases
its temperature. For example, at 200 MPa the minimum temperature changes from 764 °C at
a water activity (aH2O) of 0.5 to 902 °C at aH2O = 0.1. The stable liquidus silica phase is high
quartz, except at P ≤20 MPa, where tridymite appears near the ternary minimum. At still lower
pressures tridymite becomes stable along the entire cotectic (Tuttle and Bowen 1958). There are
no reliable experimental data on the Qz-Ab-Or ternary at pressures below 50 MPa because of
difficulties in attaining equilibrium on experimental time scales (e.g., Schairer 1950; Brugger et
Qz
1
0.5
a
0.5 0.1 MPa 50 MP
100
200
50 40
+ 300
+
+ MPa
P = 220 MPa + 1000
5
1
Ab Or
1
0.5
Figure 18. Haplogranite ternary (as summarized by Blundy and Cashman 2001), showing changes in
experimental glass composition as a function of pressure, using projection scheme in equations (9a-c).
Staring material for experiments was lightly crushed dacite pumice from the 1991 eruption of Mt. Pinatubo.
Experimental charges were decompressed from 200 MPa to the final pressure and then allowed to equilibrate
for one week. At pressures ≥ 50 MPa, experimental data lie close to calculated cotectic. However, at P ≤
40 MPa, progression of melt compositions away from the Qz′ cotectic probably reflects disequilibrium
crystallization over the 7-day run duration. Redrafted from Hammer and Rutherford (2002).
208 Blundy & Cashman
al. 2003). A simple consequence of these observations is that silicate liquids with SiO2 contents
above ~75% can be generated only at low pressure, as saturation of the melt with a silica phase
(typically quartz) at higher pressure limits the maximum possible SiO2 content of the melt.
In natural silicic liquids, the relationship between SiO2 and pressure is more complicated
because of the presence of other components, notably Ca, which stabilizes plagioclase
feldspar, and Mg+Fe, which stabilize amphibole and biotite. In an attempt to account for the
effects of other components on SiO2, Blundy and Cashman (2001) developed a projection
scheme to convert natural glass compositions into the haplogranite ternary. Their projection
scheme considered the effect of An on water-saturated phase relations in the system Qz-Ab-Or
using experiments in the quinary system qz-Ab-Or-An-H2O and its constituent sub-systems
(e.g., Johannes and Holtz 1996). Fit parameters were obtained by trial and error to obtain the
following projection scheme:
Qz′ = Qzn × (1 − 0.03 An + 6 × 10 −5[Orn × An] + 10 −5 [ Abn × Orn × An]) (9a )
Or ′ = Orn × (1 − 0.07 An + 10 −3 [Qzn × An]) (9b)
Ab′ = 100 − Qz′ − Or ′ (9c)
where the prime denotes the revised ternary co-ordinates and Qz, Ab, Or and An are the CIPW
normative components. All units are expressed in weight percent, and the subscript n denotes
that the normative components Qz, Ab and Or are first summed to 100%. As no attempt was
made to account for the effect of normative corundum (Cor), the normalization scheme is not
applicable to liquids with >1% Cor. For glasses saturated in a silica phase, the projection can be
used to constrain the pressure of crystallization, while for glasses lacking a silica phase, mini-
mum pressures and a liquid line of descent can be determined (Blundy and Cashman 2001).
How well does the barometer work? A principal limitation of this method is the lack of
reliable experimental data on the haplogranite system at pressures below 50 MPa. Since Blundy
and Cashman’s (2001) original paper, a number of experimental studies have been published in
which H2O-saturated glasses containing feldspar and a silica phase (usually quartz) have been
performed over a range of pressures, temperatures, and decompression paths (e.g., Hammer
and Rutherford 2002; Couch et al. 2003b; Martel and Schmidt 2003). Only experiments
performed at temperatures below 800 °C (Hammer and Rutherford 2002) contain a silica phase
at pressures > 50 MPa; these experiments represent the most pertinent test of this technique. In
experiments that involved rapid decompression to pressures between 200 and 50 MPa followed
by equilibration times of one week, glass compositions provide a good estimate of relative
pressure from the plotting position on the Qz′-Ab′-Or′ plot, but only moderate accuracy in
absolute pressure (e.g., Hammer and Rutherford 2003). At lower pressures, glass compositions
diverge from the anticipated cotectic and move away from the Qz′ apex (Fig. 18). We suspect
that observed decreases in the Qz′ component of the melt reflect disequilibrium crystallization
because of the sluggish kinetics in these cool viscous melts (e.g., Cashman and Blundy 2000).
Disequilibrium is suggested texturally by the observed isolated clusters of quartz-feldspar
aggregates within apparently homogeneous matrix glass, a texture that is common in shallow
plugs and domes (e.g., Cashman 1992; Hammer et al. 1999). These disequilibrium effects
are probably transient, as suggested by broad beam analyzes of holocrystalline groundmass
produced by shallow crystallization of feeder dikes below Unzen volcano (Noguchi et al.
2008a). These groundmass compositions plot between the 50 MPa and (approximate) 0.1 MPa
cotectics in Figure 18, consistent with their crystallization of ~ 40 MPa.
We conclude that the haplogranite barometer can be used as a crude guide to pressure
in the absence of other constraints. Evidence for disequilibrium precipitation of silica phases
suggests that relative pressures are best established from the Ab′ component of the glass (Fig.
19; see also Hammer and Rutherford 2003). The scheme would benefit from more equilibrium
Reconstruction of Magmatic System Variables & Processes 209
200
Or' Qz' Ab'
150
Pressure (MPa)
100
P = 5.8 Ab' - 176
Pressure
50 (MPa)
0
0 20 40 60 80
Component
Figure 19. Individual haplogranite projection parameters (Qz′, Ab′, Or′) plotted against pressure derived
from decompression of H2O-saturated magma of bulk composition representative of the reactive magma
of the current eruption of Soufriere Hills, Montserrat (rhyolite). As seen in Figure 18, the Qz′ component
increases as P decreases to 50 MPa, then decreases at shallow pressure. Or′ shows little change until P
< 50 MPa, at which point it increases dramatically. Ab′ shows the most consistent change with pressure,
thereby exhibiting the greatest barometric potential for this bulk composition; regression equation shown
on Figure. Data from Martel and Schmidt (2003).
experimental studies of low-P phase relations in haplogranites and natural granitic compositions,
although there are substantial problems with obtaining equilibrium in such experiments (e.g.,
Schairer 1950; Brugger et al. 2003), problems that may also persist in nature (e.g., Cashman
1992; Cashman and Blundy 2000; Blundy and Cashman 2001). Simultaneous refinement of
thermodynamic models of melts in the haplogranite system (e.g., Holland and Powell 2001)
and their extrapolation to pressures below 100 MPa would greatly improve our understanding
of kinetic influences on crystallization at shallow pressures.
Crystal textures
One final method of estimating pressure is to use crystal textures, where those textures can
be reasonably assumed to result from decompression. In H2O-saturated magmas, plagioclase
± orthopyroxene tend to be the dominant crystallizing phases at low pressure, because
their stability is strongly dependent on pH2O. This pressure-dependence can generate high
supersaturations (effective undercooling > 200 °C) if H2O-saturated magma is decompressed
isothermally from ~ 200 MPa to near-surface conditions (Fig. 20). For this reason, phase
proportions, alone, carry important information about equilibration conditions. However, as
the kinetics of crystallization vary depending on decompression conditions, equilibrium may
not be fully achieved, particularly at shallow pressures (e.g., Geschwind and Rutherford 1995;
Hammer and Rutherford 2002; Couch et al. 2003b; Martel and Schmidt 2003; Larsen 2005;
Suzuki et al. 2006). To infer magma storage and ascent conditions from textural data, therefore,
we need to briefly review the kinetics of decompression-driven crystallization, as illuminated
by these recent experiments (see Hammer 2008 for a more complete review).
The phase diagram shown in Figure 20 shows that the equilibrium abundance of both
plagioclase and pyroxene increases dramatically with decreasing pressure, particularly at P <
50 MPa. This is demonstrated by experiments where H2O-saturated silicic magma (typically at
starting pressures of ~ 200 MPa) is decompressed rapidly and then allowed to equilibrate for
at least one week (Fig. 21). Here plagioclase crystallinity increases gradually from 0 to 20%
210 Blundy & Cashman
1200
Liquid
1100 Plagioclase
Temperature (˚C)
1000 Opx
Amphibole
900
800
Qz-Ab-O
700 r solidus
0 100 200 300 400
Pressure (MPa)
Figure 20. Phase diagram for H2O-saturated (Ptot = pH2O) Mount St. Helens dacite magma (based on
compilation of Blundy and Cashman 2001). Note the strong sensitivity of phase boundaries to pH2O. The
low temperature stability of alkali feldspar + quartz have not been determined experimentally and are
therefore not shown. Oxide phases are omitted for clarity; their composition and stability are sensitive to fO2.
50
M&S
H&R (SSD)
Volume % plagioclase
40
H&R (MSD)
30
20
10
Volume plagioclase (%)
0
0 50 100 150 200
Pf (MPa)
Figure 21. Variation in plagioclase abundance as a function of pressure for H2O-saturated rhyolitic
magmas. Filled symbols represent experiments where the magma was decompressed from 200 MPa to
the final pressure and then allowed to equilibrate prior to quenching (Single Step Decompressions, SSD);
open symbols represent decompression at a constant rate (1.2 MPa/hr) to the quench pressure (Multi-Step
Decompressions, MSD). Plagioclase abundance increases rapidly at P ≤ 50 MPa (also seen in natural
samples, e.g., Blundy et al. 2006; Blundy and Cashman 2008). Data from Hammer and Rutherford (2002),
Martel and Schmidt (2003).
between 200 and 50 MPa (0.13% MPa−1), then rapidly from 20 to 45% between 50 and 5 MPa
(0.56% MPa−1). A similar pattern of crystallinity increase with decreasing pressure can be
inferred from melt inclusion data obtained from pyroclasts erupted from Mount St. Helens in
1980 (Blundy et al. 2006) and from scoria and lapilli erupted from Mt. Etna in 2002 (Spilliaert
et al. 2006). In contrast, experiments that simulate steady (~1.2 MPa hr−1) decompression
over ≤ 180 hours do not attain equilibrium plagioclase crystallinities, particularly at pressures
between 5 and 50 MPa (Fig. 21). Thus while in theory plagioclase abundance should provide
information on equilibration pressure, in practice this method can be used only where
Reconstruction of Magmatic System Variables & Processes 211
decompression conditions allow an equilibrium phase assemblage to be attained (either
through temporary magma arrest at intermediate pressures for days to weeks or with magma
ascent at rates < ~ 1 MPa hr−1).
The kinetics of crystal nucleation 35
may also provide information on equil- (a)
ibration pressures, especially at low 30
Plagioclase (vol%)
pressures (high effective undercool- 50 MPa
25
ing). Specifically, recent experiments
suggest that rapid decompression to 20
shallow pressures causes plagioclase
15 100 MPa
crystallinity to increase rapidly by in-
ducing very high rates of nucleation 10
Plagioclase (vol%)
(Fig. 22). The plagioclase number den-
5
sity then stabilizes after 24-48 hours
(shaded areas in Fig. 22b), while the 0
abundance of plagioclase continues 0 200 400 600
to increase slowly (crystallization be- Time (hours)
comes growth-dominated; Fig. 22a).
These changes are illustrated in Figure Log plagioclase Nv (/mm 3 ) 7
23a, which shows that plateau values of (b)
both number density and crystallinity
6
increase with decreasing equilibration 50 MPa
pressure. Because the data indicate that
equilibrium crystal number densities 5
are achieved more rapidly than equilib- 125 MPa
rium crystallinities, and as crystal num-
4
ber density varies by several orders of
Log plagioclase Nv (/mm3)
magnitude, it appears that crystal num-
ber density may provide the more reli- 3
able geobarometer (Fig. 23b). 0 200 400 600
How well does the barometer Time (hours)
work? Unfortunately, there are not Figure 22. Temporal evolution of plagioclase (a) abundance
sufficient experimental data to test the and (b) number density resulting from instantaneous
correlation shown in Figure 23b. We isothermal decompression from 160 MPa, 875 °C to
can, however, make some comparisons different pressures (shown on curves). Time held at the
using data from natural samples. lower pressure is given on the abscissa. Gray bars illustrate
equilibrium plagioclase number achieved for each pressure.
Episodic vulcanian/sub-plinian Experimental data from Couch et al. (2003b).
eruptions of Mount St. Helens during
the summer of 1980 ejected magma
that had been stored for weeks at different depths within the conduit (Cashman and McConnell
2005; Scandone et al. 2007). In most of these pyroclasts, log plagioclase number densities lie
between 5.3 and 6.6 (Fig. 23c), which suggests equilibration pressures between 85 and 30 MPa
using the correlation shown in Figure 23b. As these pressures are in good agreement with those
estimated by Cashman and McConnell (2005) and Blundy et al. (2008) from the H2O content
of melt inclusions, in this example, at least, it appears that groundmass textures provide a fairly
robust indicator of equilibration pressure (see also Clarke et al. 2007).
This calibration does not work under all conditions, however. First, slow and steady
extrusion of dome lavas will cause crystal nucleation over a range of effective undercoolings,
thus resulting number densities will not be representative of any single pressure (e.g., Couch
et al. 2003b). Second, the calibration is for crystallization from melts of rhyolitic composition
212 Blundy & Cashman
Couch 125
9 Couch 100
(a) Couch 75
8 Couch 50
HR 175
Log Nv (mm )
3
7 HR 150
HR 100
6 HR 75
10 MPa HR 50
5 HR 40
Nv (/mm3) 50 MPa HR 25
4 HR 10
100 MPa
3
0 20 40 60 80
Crystallinity (%)
7
(b)
Log Nv (mm )
3
5
Log Nv (/mm3)
Log Nv = -2.8 Log P + 10.7
2
R = 0.96
4
0 1 2 3
Log P (MPa)
8 10 MPa
Log Nv (mm )
3
7 25 MPa
6 50 MPa
Nv (/mm3) Mount St. Helens
Nv (/mm3)
5
100 MPa Pinatubo surge
Spurr pumice
4
Merapi dome
(c) Montserrat
3
0 20 40 60 80
Crystallinity (%)
Figure 23. Number density (Nv) as a potential barometer. (a) Correlation of Nv with crystallinity for
experiments of Couch et al. (2003) and Hammer and Rutherford (2002) for different equilibration pressures.
Only plateau values are plotted, i.e., after Nv and crystallinity have attained equilibrium values (see Fig.
22). (b) Correlation of log number density with log equilibration pressure for experimental data in (a). (c)
Variation in Nv with crystallinity for pumiceous samples of dacite from Pinatubo (Hammer et al. 1999)
and Mount St. Helens (Cashman and McConnell 2005); andesite from Spurr (Gardner et al. 1998), and
andesitic dome lavas from Montserrat (Clarke et al. 2007) and Merapi (Hammer et al. 2000). Dashed lines
show pressures for rhyolitic compositions determined from the calibration shown in (b).
Reconstruction of Magmatic System Variables & Processes 213
and cannot be applied directly to melts of other compositions (e.g, andesitic magmas of Merapi
and Mt. Spurr, Fig. 23c). Third, silicic pyroclasts from the pre-climactic eruptions of Pinatubo
(Hammer et al. 1999) and from the May 18th lateral blast of Mount St. Helens (Cashman and
Hoblitt 2004) have extremely high plagioclase number densities (> 107 mm−3) that have not
been replicated in any decompression experiments (Fig. 23c). The volcanological context for
these samples argues that their textures must result from rapid decompression to shallow levels
(below ~15 MPa suggested by volatile contents of matrix glass and melt inclusions) followed
by growth times of hours to weeks. As these pressures are consistent with those predicted
from the calibration in Figure 23b, the inability to replicate these textures experimentally is
puzzling, but may reflect degassing differences between melts saturated with H2O alone and
those that contain multiple volatile phases (see below).
Textures as geospeedometer? Textural characteristics, particularly formation of
hornblende breakdown rims during decompression, have been used to monitor rates of magma
ascent (Rutherford and Hill 1993; Rutherford and Devine 2003, 2008; Browne and Gardner
2006; Rutherford 2008) when calibrated with experimental data. As distinguishing between
textures resulting from steady decompression versus temporary arrest may not be possible,
application of textural geospeedometers requires field evidence for steady magma discharge
(Q), such as continuous and long-lived effusion of lava domes.
Texturally, variations in effusion rate may be manifested as variations in both total crystal
abundance and crystal number density (e.g., Cashman 1992; Geschwind and Rutherford 1995).
Crystal abundance affects the magma rheology, which, 6 in turn, controls the morphology of
lava domes (e.g., Fink and Griffiths 1998). For example, slow effusion (< ~1-2 m3s−1) of (a)
125 MPa H2O-
saturated andesitic magma allows complete crystallization during magma ascent to the Earth’s
100 MPa
Log Nv (mm 3 )
surface and creates holocrystalline spines such as the 75 MPaspine of Mt. Pelee. In contrast,
5 famous
3 −1 50 MPa
higher effusion rates (> 5-10 m s ) limit the extent of syn-ascent crystallization and produce
flows with more fluid morphologies (e.g., Cashman et al. 2008). Crystal number density varies
over orders of magnitude as a function of decompression
4 rate (e.g., Couch et al. 2003) and
therefore offers a potentially sensitive measure of both decompression
Log Nv (mm3) rate (e.g., Fig. 24) and
the rate of H2O exsolution (Toramaru et al. 2008). Effusion rate control on plagioclase number
Log Nv = 0.66 Log dP/dT + 6.5
density has been observed during effusive eruptions 3 of Merapi (Hammer et al. 2000) and
Unzen volcanoes (Noguchi et al. 2008b; Fig. 24b). -5Unfortunately, -4 the natural-3 data cannot -2
Log dP/dt
6 7
125 MPa (a) (b)
Merapi
100 MPa Log Nv = 4e-5Q + 6.0
Log Nv (mm 3 )
Log Nv (mm 3 )
5 75 MPa
6
50 MPa
4 5
Log Nv (mm3) Unzen
Log Nv = 3e-6Q + 4.8
Log Nv = 0.66 Log dP/dT + 6.5
3 4
-5 -4 -3 -2 0 1 2 3 4 5
3 5
Log dP/dt Discharge rate (m /day) x 10
Figure
7
24. Textures as geospeedometers. (a) Variations in log number density (Nv) resulting from
decompressions performed at different rates to(b)
the final pressures indicated in the legend. Note that in these
Merapi
experiments, the rate of decompression (dP/dt, in MPa/s) is more important in determining Nv than the
Log Nv = 4e-5Q + 6.0
final pressure. Data from Couch et al. (2003). (b) Relationship between plagioclase number density (per
Log Nv (mm 3 )
area)6and magma flux (Q) at andesitic Merapi volcano, Indonesia (Hammer et al. 2000) and dacitic Unzen
volcano, Japan (Noguchi et al. 2008b).
5
Unzen
214 Blundy & Cashman
be directly compared with the experiments because the geometry of the Merapi and Unzen
conduits is unknown (and thus discharge rate cannot be directly compared with decompression
rate), although inferred ascent rates through the Unzen conduit of 0.008-0.05 ms−1 (~0.0002-
0.001 MPas−1; Noguchi et al. 2008b) yield an order of magnitude higher number density than
predicted by the calibration shown in Figure 24a. In summary, textural geobarometers and
speedometers require much more extensive calibration, but may provide extensive untapped
potential for probing conditions of magma storage and ascent.
COMParISOn OF PetrOLOGICaL data WIth MOnItOrInG SIGnaLS
A key objective of volcano petrology is to shed light on underground processes that precede
and accompany eruptions and that are reflected in monitored signals, such as seismicity, gas
chemistry, ground deformation and eruption dynamics. In this section we illustrate how the
petrological techniques discussed above can be used, via reference to our ongoing research at
Mount St. Helens. In particular, we show the utility of acquiring melt inclusion data sets that
are both complete (i.e., including analyzes of major, trace, and volatile elements) and well
constrained (i.e., are from samples for which both timing and eruption conditions are known).
Gas chemistry
Syn-eruptive and inter-eruptive fluxes of volcanic gas and its composition can be measured
by a variety of indirect spectroscopic methods and direct analyzes of fumaroles. Both types of
measurement are widely used as a volcano monitoring tool. The gases most widely measured
by spectroscopic techniques are SO2 and halogens. The more abundant volcanic gases, H2O and
CO2, can also be determined by IR spectroscopy. However, their abundance in the atmosphere
makes measurements of the volcanogenic flux subject to considerable uncertainty, although
recent technical developments with unmanned vehicles have made direct CO2 measurements
in the plume possible (McGonigle et al. 2008). Alternatively, H2O and CO2 can be measured in
enclosed volcanic craters where the beam path-length is short and atmospheric contamination
kept to a minimum. To date such measurements have been confined to persistently degassing
basaltic volcanoes (e.g., Burton et al. 2000; Aiuppa et al. 2007). At andesite and dacite volcanoes
the only reliable means of measuring H2O and CO2 comes from analyzes of sampled fumarole
gases. Fumarole gas chemistry can be used to determine the proportion and fugacity of the
various H-C-O species, including fO2. In this section we focus on two types of volcanic gas
measurements—fumarole gas determination of fO2 and spectroscopic measurement of SO2—and
compare these data with petrologic information to gain insights into the degassing process.
Oxygen fugacity. Fifty high-temperature (650-830 °C) fumarole gas analyzes from Mount
St. Helens in the period 1980-1982 are presented by Gerlach and Casadevall (1986). All gases
are H2O-rich with 91-99 mol% H2O, as also observed during the 1991-1995 eruption of Unzen
volcano (Ohba et al. 2008). CO2 is the next most abundant species (1-8 mol%) with minor H2,
H2S, CO and SO2. Field chromatograph data were restored to equilibrium high-temperature
values using thermodynamic data. The calculated fO2 of the fumarole gases lies within NNO±0.3
at temperatures >700 °C but decreases to NNO-1.6 at lower temperatures (Fig. 25).
The fumarole fO2 data can be compared to those calculated from coexisting Fe-Ti oxide
pairs from magma erupted over the same time period (Fig. 12). Oxides from the May 18th
Plinian eruption, which tapped magma stored at pressures of 160 to 320 MPa (Blundy and
Cashman 2005; see above), define an fO2 of NNO+0.8, which is significantly more oxidizing
than the highest temperature fumarole gases. In contrast, oxide pairs from the cryptodome and
post-May 18th eruptions show fO2 down to NNO and slightly below, thereby overlapping the
fumarole gases (Fig. 25). All of these samples contain microlites and show abundant evidence
of shallow-level crystallization. Their apparent reduction is a consequence of degassing-related
equilibria such as:
Reconstruction of Magmatic System Variables & Processes 215
Figure 25. Comparison of temperature
and oxygen fugacity (fO2) defined by
oxide pairs with values from high-
temperature fumaroles at Mount St.
Helens. Conditions in the deep-seated
magma chamber (P ≥ 160 MPa), as
sampled by the May 18th, 1980 plinian
eruption, display higher fO2 than the
fumarole gases. In contrast, fO2 from
the cryptodome and post-May 18th
eruptions, which show independent
evidence for shallow (P < 160 MPa)
crystallization, define lower fO2 that
extends down to the fumarole values.
This is a consequence of degassing-
related reduction, as demonstrated
thermodynamically by Burgisser and
Scaillet (2007). The NNO buffer is
shown for reference. Oxide data from
Blundy et al. (2008); fumarole data
from Gerlach and Casadevall (1986).
Fe3+ (melt) + S2− (melt) = Fe2+ (melt) + S4+(vapor)
Thermodynamic calculations by Burgisser and Scaillet (2007) demonstrate that such equilibria
in the rhyolite-H-S-O system cause reduction in magmatic fO2 by an amount related to the total
S content of the magma, its initial fO2, and exsolved gas content, and the pressure at which
gas and melt last equilibrate. Their calculations show that reduction in a rhyolitic melt with
200-250 ppm dissolved sulfur (a reasonable value for Mount St. Helens undegassed magma;
Fig. 26a) at 200 MPa and NNO+1 and containing 0.1 wt% exsolved gas, is of the order 0.7
log units if equilibrium is maintained down to 10 MPa. This is in excellent agreement with
the oxide data in Figure 25. The amount of reduction is only slightly greater for higher initial
exsolved gas contents. These data show clearly that the fO2 of fumarole gases corresponds not
to the fO2 of the deep-seated magma body, as tapped during the May 18th plinian phase, but to
the shallow sub-volcanic conduit (Blundy et al. 2008). Clearly, as emphasized by Burgisser
and Scaillet (2007), interpreting sub-volcanic magmatic processes in terms of fumarole gas
chemistry must take into account gas-melt equilibrium occurring within the conduit.
Sulfur dioxide. UV spectroscopic (COSPEC) measurements of SO2 emissions from
volcanoes provide valuable constraints on the volume of magma degassed both during and
between eruptions. The sulfur contents of melt inclusions can be combined with information
about eruptive volume and duration to calculate the sulfur released during eruption, assuming
complete degassing of the magma, to obtain the “petrological flux”, Fp:
V
Fp =
t
(
× SO2MI − SO2gm )
where MI and gm denote SO2 contents of melt inclusions and (degassed) groundmass glass,
respectively, V is the erupted volume (in dense rock equivalent) and t is the eruptive duration.
Fp can then be compared to the measured syn-eruptive flux. It has long been recognized that
there is a substantial (one to two orders of magnitude) mismatch between the calculated
petrological flux of SO2 and the observed syn-eruptive flux for large silicic eruptions (e.g.,
Wallace 2001, 2005). This has led to the conclusion that large-scale eruptions involve gas loss
from a much larger volume of magma than is actually erupted. Candidate processes include the
addition of sulfur-rich magma to the base of the chamber shortly prior to eruption, convective
216 Blundy & Cashman
Figure 26. Observed and petrological fluxes at Mount St. Helens. (a) Temporal evolution of SO2 in the
eruption plume, as measured by airborne COSPEC (dots; McGee and Casadevall 1994), and SO2 in melt
inclusions (crosses; Blundy et al. 2008). Both parameters show a similar temporal pattern with a marked
drop-off after May 18th 1980. The pre-eruptive SO2 content of the melt is of the order 1000-1500 ppm (or
300-500 ppm S). Note the logarithmic abscissa; major eruptive events are labeled. Note also the significant
inter-eruptive flux of SO2. (b) Comparison of observed and calculated petrological SO2 flux for post-May
18th dome-forming and dome-related eruptions. Calculated flux uses maximum SO2 melt inclusion contents
from (a) and the erupted volumes and durations of Swanson et al. (1987) and Swanson and Holcomb
(1990). Observed fluxes are taken from McGee and Casadevall (1994), averaged over the duration of the
eruptive period, with 1 s.d. error bars; it is not straightforward to estimate uncertainty on the petrological
flux. Note the strong 1:1 correlation between the two parameters.
Reconstruction of Magmatic System Variables & Processes 217
cycling of magma through the conduit system, or development of permeability throughout the
magma reservoir during eruption.
The post-plinian eruptions of Mount St. Helens afford an excellent opportunity to
compare petrological and observed fluxes because of the abundant gas flux data (McGee and
Casadevall 1994) and the substantial melt inclusion database of Blundy et al. (2008). At Mount
St. Helens, SO2 fluxes generally decrease with time. This decrease is mirrored by the trend
in melt inclusion SO2 contents (Fig. 26a). We can calculate the maximum petrological flux
based on the maximum SO2 content in melt inclusions from any single eruptive phase and
compare this with the observed syn-eruptive flux for 11 post-plinian episodes (Fig. 26b). The
clear correlation of these data suggest that, in contrast to large plinian eruptions, smaller,
dome-related events discharge a quantity of gas commensurate with the petrological data. This
interpretation cannot, however, account for the inter-eruptive flux, which at times exceeds the
subsequent syn-eruptive flux (Harris and Rose 1996; Fig. 26a). Either the inter-eruptive flux
extracts gas from outside the small magma volumes that are subsequently erupted, or the upper
reaches of the magma system are effectively replenished with gas on inter-eruptive timescales.
Both interpretations require the development of permeability in dacitic magma. We consider
that a better understanding of the physical mechanisms of gas transport in silicic magmas
should be a key research priority.
Seismicity. A full discussion of the types and causes of sub-volcanic seismicity is beyond
the scope of this review. There is, nonetheless, widespread consensus that the distribution of
sub-volcanic earthquakes contains information about the underground storage and movement
of magma and gas (see recent reviews of volcano seismicity by Neuberg 2000 and Chouet
2003). It is therefore useful to compare seismic information with melt inclusion data. For
well-monitored volcanoes there are abundant data on the temporal evolution of seismic events
that can be related to melt inclusion trapping pressures, as calculated from volatile solubility,
provided suitable conversion is made between pressure and depth.
Converting pressure to depth beneath active volcanoes is not straightforward as the exact
density structure beneath the volcano is rarely known. In the simplest case the system is
assumed to be either lithostatic or “magmastatic”, although it is likely that the dynamics of
the system and the effect of the volcanic edifice on the underlying stress distribution make
either assumption unreliable (e.g., Pinel and Jaupart 2000). Despite these limitations, useful
inferences can be made from simple models. The difference between magma density and rock
density is of the order 10%, if the presence of bubbles is ignored. We have already shown that
the typical accuracy of melt inclusion pressure calculated from H2O and CO2 is ±22 MPa. In
combination the likely uncertainty on calculated depth is of the order ±1.5 km for magma
storage depths of ≤15 km.
In the case of Mount St. Helens, a density model is available from Williams et al. (1987).
Assuming that the pressure in the system is lithostatic and taking into account the effect of
the volcanic edifice, Blundy et al. (2008) arrive at the following conversion for total volatile
pressure (Ptot) to depth (z) below sea-level:
z( km) = 0.029654 Ptot + 0.22704 Ptot 0.5 − 2.95 (10a )
This expression applies to the magma system prior to May 18th 1980. The removal of ~400 m
of the edifice on May 18th, leads to a slightly revised version for subsequent eruptions:
z( km) = 0.03074 Ptot + 0.18334 Ptot 0.5 − 2.55 (10 b)
Magma storage pressures can be calculated directly from saturation pressures for those melt
inclusions where both H2O and CO2 data are available; where we lack CO2 data, we calculate
pH2O from measured H2O and then assume that XH2O = 0.8 in the coexisting vapor phase,
leading to the correction Ptot = 1.287 pH2O. Figure 27 compares the calculated melt inclusion
218 Blundy & Cashman
Figure 27. Comparison of melt inclusion trapping depths (squares), calculated from volatile saturation pres-
sures (Newman and Lowenstern 2002) converted to depth (relative to sea-level) using equations (10a) and
(10b), and earthquake depths (open grey circles) from the PNSN catalogue for the period 1980-1986. Note
logarithmic abscissa. Selected eruptive episodes shown on top axis. The filled squares are for melt inclusions
for which both H2O and CO2 were analyzed; open squares are for melt inclusions in which only H2O was
analyzed and XH2O assumed to be 0.80, hence Ptot=1.287pH2O. Calculated depths assuming magmastatic
pressure rather than lithostatic pressure would place the melt inclusion approximately 10% deeper.
trapping depths to the depths of contemporaneous earthquakes for the period 1980-1986 taken
from the PNSN catalogue (http://www.geophys.washington.edu/SEIS/PNSN/). The close
match between absolute depths and their temporal evolution strongly supports a link between
seismicity and magma storage. Close inspection of Figure 27 reveals that in the months of 1980
following the plinian eruption of May 18th there were relatively few sub-volcanic earthquakes
except those that followed magma withdrawal from depths of at least 6 km during individual
explosive eruptions. After December 1980 shallow earthquakes appeared in the weeks prior
to individual effusive eruptions, coinciding with a considerable reduction in eruptive vigor.
The simplest interpretation of these data is that the magmatic system was relatively open to
magma ascent during 1980, but subsequently became effectively plugged by viscous magma
inhibiting further magma ascent (Scandone et al. 2007). Post-1980 melt inclusions testify to
relatively shallow pre-eruptive magma storage during this time period, suggesting that magmas
consistently paused, degassed, and crystallized at shallow levels prior to eruption.
Constraints on conduit dimensions. In the same way that melt inclusion SO2 contents can
be combined with eruptive volumes and durations to calculate a petrological sulfur flux, so the
trapping depths of melt inclusions can be used to estimate conduit and chamber dimensions,
particularly when melt inclusions are from relatively small, dome-related eruptions that derive
predominantly from the conduit and uppermost reaches of the magma system. For example, if
we assume that in a large population of melt inclusions the maximum and minimum trapping
depths correspond to the top and bottom of the erupted magma volume, we can calculate the
dimensions of a cylinder that encompasses this volume.
Reconstruction of Magmatic System Variables & Processes 219
We have made calculations for the seven post-plinian eruptive episodes from Mount St.
Helens for which we have more than 8 melt inclusions and for which the erupted volume is well
constrained (e.g., Swanson and Holcomb 1990). In each case we assume that the shallowest
erupted magma came from the very top of the conduit and use the melt inclusions to define the
lowermost extent of the erupted magma, which we assume also to be within the conduit. In
reality the shallowest erupted magma may derive from some distance below the edifice, while
the deepest erupted magma may have been stored within the uppermost chamber rather than
the conduit. Nonetheless, our calculations illustrate the potential of the approach.
We use both the deepest and second deepest melt inclusions for our calculations to safe-
guard against bias in the small datasets: for example, it is possible that some melt inclusions
are a legacy from un-erupted magma of the plinian phase. We use the lithostatic model (Eqn.
10b) to calculate depths and assume that the conduit is a cylinder of radius, r, which can be
obtained from:
V
r= (11)
πz
where V is the erupted volume (in dense rock equivalent; taken from Rutherford and Hill 1993)
and z is the depth of the deepest (or second deepest) melt inclusion. Note that using magma
density to calculate pressure will increase z and reduce r slightly; the presence of bubbles
will exacerbate this effect. However, because of the square root relationship of r to z, the
consequences of adopting different density models are muted.
The calculated conduit radii are shown in Figure 28. They are in good agreement with
those derived by Chadwick et al. (1988) from numerical modeling of crater-floor deformation
that preceded four separate dome extrusion events in late 1981 and1982 (Fig. 28, shaded area).
Figure 28. Temporal evolution of conduit radii for 7 post-May 18th 1980 eruptions of Mount St. Helens.
Radii are calculated using equation (11). The two filled circles for each eruption denote the radii calculated
from the greatest and second-greatest melt inclusion trapping depth. The shaded region is the range of
conduit radii estimated by Chadwick et al. (1988) from crater-floor deformation data in 1981 and 1982.
Note that changes in overburden density, for example due to magma vesiculation, cannot account for the
observed variation in conduit radius; for example, a 30% reduction in overburden density result in only a
14% relative reduction in radius.
220 Blundy & Cashman
Through 1980 there is a marked decrease in r. This decrease accompanied a transition from
explosive to effusive activity in late 1980 and has been interpreted to reflect plating of the
shallow conduit walls with partially solidified magma (e.g., Geschwind and Rutherford 1995)
or gradual collapse of the fully connected conduit (Scandone et al. 2007). After 1980, the
diameter of the shallow conduit apparently stabilized until an explosive eruption in March
1982, when our data suggest an increase of the effective conduit radius, probably as the result
of magma accumulation at intermediate depths combined with erosion of the walls of the
shallow system during the eruption.
Melt inclusion data for the current eruption at Mount St. Helens are more limited and
the maximum pressure recorded by any melt inclusion is 78 MPa (Blundy et al. 2008). Using
a density of 2000 kg m−3 for the extruded spines and a total erupted volume of 73×106 km3
we calculate r = 78 m for the current conduit. This is in good agreement with the value of
95 m independently derived by Iverson et al. (2006) on the basis of the extrusion flux, and
substantially larger than the calculated values for 1980-1982 (Fig. 28).
Our calculations suggest that there is virtue in trying to use melt inclusion data to constrain
conduit diameters for use in numerical modeling calculations, as conduit dimensions are
notoriously hard to estimate by other means (see geospeedometry discussion above) and yet
are a key input to any numerical model.
MaGMa aSCent and CryStaLLIZatIOn
The previous section showed ways in which petrologic data can be integrated with volcano
monitoring data to examine active systems. The power of petrologic studies, however, lies in
their application to the generation and storage of magma between eruptive events. Here we
illustrate ways in which intensive parameters derived from petrologic studies allow us to track
the degassing and crystallization history of different magmatic systems.
Phenocryst abundance
The phenocryst abundance in volcanic samples provides information on conditions of
magma storage, particularly in rapidly erupted pumice, where the rate of transport from the
magma storage region to the surface is sufficiently fast to inhibit additional crystallization. In
kinetic terms, the larger size and lower number density of phenocrysts reflects a dominance
of crystal growth over crystal nucleation, a consequence of the relatively small undercoolings
caused by either small degrees of cooling within magma storage regions or small reductions
in magmatic pH2O.
Although data on phenocryst phase proportions are surprisingly hard to find, some
generalizations may be made. First, the abundance of plagioclase phenocrysts in samples of
intermediate composition rocks erupted from stratovolcanoes is commonly 20-35%. This value
appears to reflect magma residence times of years to centuries in subvolcanic magma storage
regions. Explosively erupted samples with unusually high phenocryst contents (e.g., Pinatubo
1991; Pallister et al. 1996) typically have relatively low equilibration temperatures (< 800
°C) that suggest extensive cooling-related crystallization; basaltic magma inputs often trigger
(or at least aid) eruption of these magmas. Second, the plagioclase phenocryst content of
dome lavas is typically higher than that of pumice erupted explosively from the same storage
region (that is, during the same eruption). This difference has been documented at Mount
St. Helens (e.g., Cashman and Taggart 1983), Unzen (Nakada and Motomura 1999), Merapi
(e.g., Hammer et al. 2000) and Mt. Pelée (Martel et al. 2000), and suggests that phenocrysts
continue to grow during slow magma ascent from the storage region to the surface. Further
evidence for continuous crystal growth during ascent is provided both by zoning patterns and
compositional ranges in phenocryst rims (Cashman and Blundy 2000; Blundy and Cashman
Reconstruction of Magmatic System Variables & Processes 221
2001; Berlo et al. 2007) and by equilibration pressures recorded in plagioclase-hosted melt
inclusions (e.g., Blundy and Cashman 2005).
Melt inclusion compositions
The value of melt inclusion studies is well established for studying magmatic systems that
crystallize olivine or quartz (e.g., Wallace 2005; Metrich and Wallace 2008). Here we review
the use of plagioclase-hosted melt inclusions in studying conditions of magma storage and as-
cent. The ubiquity of plagioclase in many volcanic rocks, and the probability that plagioclase
may record the entire crystallization and ascent history of the magma, make plagioclase melt
inclusions an appropriate target of study. As it is unlikely that any single population of melt
inclusions can be used to infer pre-eruptive magma storage conditions, we recommend using
data from a large suite of melt inclusions with different textures and in different phenocryst
phases to reconstruct an overall picture of the sub-volcanic system. A large sample size also
allows the data to be screened for processes such as post-entrapment crystallization or diffu-
sive volatile loss, or to subdivide the data according to host mineral and/or texture (e.g., Liu
et al. 2006).
Once formed, plagioclase-hosted melt inclusions remain connected to the matrix melt via
thin melt tubes (e.g., Stewart and Pearce 2004; Blundy and Cashman 2005; Humphreys et al.
2008a). Diffusion of chemical components along the melt tubes, coupled with crystallization
of plagioclase around the walls of the melt inclusions, modifies the melt inclusion composition
until the inclusion becomes isolated, or occluded, from the matrix melt. The extent to which
diffusion maintains chemical equilibrium between inclusion and matrix melts depends on the
timescale, temperature and diffusivities of the components of interest. Volatile elements (e.g.,
H2O, CO2, Li) will diffuse rapidly, while network-forming cations such as Si and Al are likely
to be the rate-limiting step in controlling full chemical equilibration. Diffusive timescales
based on the experimentally-determined interdiffusivity of Si between hydrous dacitic and
rhyolite melts (Baker 1991) are ~7 hours for a (typical) 100 µm long melt tube at 900°C when
the melt has 6 wt% H2O and ~2 days with 3 wt% H2O. Thus melt inclusions are likely to
preserve the major element chemistry of the matrix melt on a timescale of days or less before
their eventual occlusion except for the most slowly diffusing trace species, i.e., large, highly
charged cations such as LREE, U and Th.
It is important to distinguish between this process and that of simple post-entrapment
crystallization, which occurs under closed conditions. When the system is closed, elements that
do not occur in the host crystal should increase with increasing post-entrapment crystallization,
for example, Mg and Fe in plagioclase-hosted melt inclusions. If, instead, Mg and Fe show
compatible behavior in both evolving plagioclase-hosted melt inclusions and in matrix glass,
a ferromagnesian phase such as orthopyroxene is implicated. If no such daughter minerals are
found in the melt inclusions then chemical exchange of Mg and Fe along the melt tubes must
have maintained chemical coherence between the inclusion and matrix melts (e.g., Blundy et
al. 2008).
The trace element chemistry of melt inclusions can be used to identify new inputs of
magma into the system, provided that account is taken of the possible consequences of
diffusive fractionation (Baker 2008). At Mount St. Helens, for example, REE patterns in melt
inclusions from the current (2004-2008) eruption are quite distinct from those observed during
the 1980-86 eruption, providing evidence for new melt being added to the sub-volcanic system
since 1986 (Blundy et al. 2008). Similarly, Humphreys et al. (2008b) have described “exotic”
melt inclusions from Shiveluch volcano, whose compositions suggest inputs of distinct melts
into the sub-volcanic system. Melt inclusions therefore provide information on open-system
processes for which there is abundant isotopic evidence from bulk rocks, single melt inclusions
and individual zoned phenocrysts (Ramos and Tepley 2008).
222 Blundy & Cashman
Modeling volatile elements
Melt inclusions are most commonly used to interpret degassing paths followed by
ascending and cooling magma. The systematics of H2O and CO2 in magmatic systems are
discussed more fully by Moore (2008). Here we explore H2O-CO2 trajectories that would be
followed by melts having different ascent and/or crystallization histories, as indicated by the
textural data described above. We consider 5 distinct scenarios (Fig. 29a):
1. Decompression without crystallization
2. Rapid decompression crystallization
3. Slow decompression crystallization
4. Isobaric vapor-saturated crystallization
5. Isobaric crystallization initially vapor under-saturated
Our calculations consider only closed systems, where vapor bubbles and melt maintain chemical
equilibrium and no vapor is lost from the system. Volatile solubility is taken from the model of
Liu et al. (2005). Our model melt is initially at 900 °C and 325 MPa. For scenarios 1-4 the melt
has initial contents of H2O and CO2 of 6.44 wt% and 500 ppm, respectively, consistent with an
equilibrium vapor phase composition of 80 mol% H2O. For the vapor-undersaturated scenario
(5) the initial values are 4.34 wt% and 400 ppm. Isobaric crystallization is driven by cooling
and occurs at 0.33 wt% crystals per °C for both vapor-saturated and undersaturated scenarios.
Decompression crystallization occurs isothermally at rates of 0.2 wt% crystals per MPa
(“rapid crystallization,” scenario 2) and 0.1 wt% MPa−1 (“slow crystallization,” scenario 3).
Note that these rates bracket the value of 0.13% crystals per MPa derived from decompression
experiments conducted between 200 and 50 MPa (e.g., Couch et al. 2003b). No account is
taken of the release of latent heat of crystallization, which is of the order 2.5 °C for each wt%
crystallized (Blundy et al. 2006).
Decompression without crystallization (scenario 1) leads to loss of both H2O and CO2
to the vapor phase (Fig. 29b). The greater solubility of H2O compared to CO2 means that the
vapor becomes increasingly H2O-rich as decompression occurs. The initial vapor comprises
80 mol% H2O, while the final vapor is over 99.5 mol% H2O, manifested as a sharp drop in CO2
in the melt at near constant H2O. Although not shown, open system decompression results in
an even sharper fall in CO2 at near constant H2O (e.g., Papale 2005).
Decompression crystallization follows similar trajectories to decompression alone
although there is a slight increase in H2O contents during the earliest stages of crystallization.
As both H2O and CO2 are considered perfectly incompatible in the crystallizing assemblage,
the effect of crystallization is to increase both components in the melt. Because CO2 is the
less soluble it is preferentially driven off into the vapor phase, enriching the melt slightly in
H2O. The enrichment is greater in the case of fast decompression crystallization (scenario
2) than slow decompression crystallization (scenario 3); in both scenarios, decompression
crystallization shows some similarities to open system behavior described above, although the
increase in H2O during the early stages of crystallization is quite distinct (Fig. 29d).
Isobaric vapor-saturated crystallization (scenario 4) increases H2O in the melt throughout
crystallization (see also Liu et al. 2006, Fig. 5b), due again to the preferential loss of CO2 to
the vapor and the (slight) increase in solubility with decreasing temperature; this is in contrast
to the case for CO2-free vapor-saturated systems, when H2O contents remain approximately
constant with crystallization (Blundy and Cashman 2005). The melt follows an isobar in
CO2-H2O space (Fig. 29c), albeit polythermally, resulting in a trend distinct from that of
decompression crystallization. CO2 decreases with increasing crystallization (Fig. 29d), hence
the previous assertion that the only unequivocal evidence for vapor saturation comes from a
negative correlation between CO2 and some index of crystallization (Wallace et al. 1999).
Reconstruction of Magmatic System Variables & Processes 223
Figure 29. Volatile evolution in melt inclusions
generated by different crystallization paths for
a generic andesite or dacite magma. (a) Model
crystallization scenarios used in calculations
in terms of wt% crystallized as a function of
pressure. Same symbols are used in subsequent
plots. (b) Molar composition of coexisting
vapor as a function of pressure. The initial
vapor composition is shown with the large
open circles: in the case of vapor saturated
crystallization this is fixed at 80 mol% H2O; for
undersaturated crystallization the first vapor to
evolve has 74 mol% H2O. (c) Melt inclusion
composition. Parental melt shown as large open
circles. (d) CO2 content of melt inclusions as a
function of crystallization. Parental melt shown
as large open circles. Note that only vapor-
undersaturated conditions lead to an increase in
CO2 with crystallization, up to the point of saturation. (e) H2O content of melt inclusions as a function of
crystallization. Parental melt shown as large open circles. Note that even vapor-saturated conditions can
lead to a small initial increase in H2O with crystallization, up to the point of saturation. Isobaric conditions
lead to a continual increase in H2O during crystallization.
224 Blundy & Cashman
Finally, isobaric vapor-undersaturated crystallization (scenario 5) follows a unique
trajectory on all diagrams in Figure 29. Initially crystallization increases both H2O and CO2 in
the melt until vapor saturation is attained at ~30% crystallinity. At that point the equilibrium
vapor phase is 74 mol% H2O. Once vapor saturation occurs, CO2 in the melt begins to fall
rapidly. H2O increases throughout crystallization. In Figure 29c, the vapor-undersaturated
period of crystallization is marked by a positive trajectory, which turns sharply to follow the
isobar (as per scenario 4) once saturation is reached.
The five scenarios above summarize simple, end-member degassing situations, as com-
monly invoked in the interpretation of melt inclusion H2O and CO2 data. However, because
real magma chambers are thermally and spatially variable, we would expect different portions
of the chamber to experience different degassing histories. For a large suite of melt inclusions
it is unlikely, therefore, that any single degassing trajectory will be defined. In fact, one could
envisage all five of the above scenarios occurring in different parts of a single large chamber.
Moreover, magma recharge, convection, sidewall crystallization, eruptions from an overly-
ing magma reservoir, volatile fluxing, and magma chamber over-pressurization can all serve
to complicate the degassing signal. In the following section we consider how more realistic
magma systems are likely to behave.
Combining degassing and crystallization
To illustrate the above complexities we have created a simple model magma chamber in
which there is a vertical, lithostatic pressure gradient from the floor (325 MPa) to the roof (125
MPa) and a lateral thermal gradient from the walls (750 °C) to the interior (925 °C) that drives
sidewall crystallization (Fig. 30). Magma entering the chamber at its base contains 5 wt%
H2O and 500 ppm CO2, but is volatile-undersaturated. At 925 °C it is also crystal-free. The
solubility of H2O and CO2 are taken from the expressions of Liu et al. (2005), and the weight
fraction crystallinity (Xc) of the magma is taken to be a simple function of both pressure and
temperature:
Figure 30. Model magma chamber
in terms of pressure and temperature,
contoured for crystal weight fraction (Xc)
and exsolved gas weight fraction (Xg).
The chamber is radially symmetrical
about the right-hand axis. The magma
initially enters the chamber at 900 °C
under vapor-undersaturated conditions
(lower right-hand corner) crystallizing in
response to both decompression (towards
the roof) and cooling (towards the wall).
The relationship between P, T and Xc
is given in Equation (10); the H2O and
CO2 solubility is from Liu et al. (2005).
The P-T conditions at which vapor-
saturation is attained are shown with the
bold dotted line. The magma is vapor
saturated throughout the shaded portion
of the figure.
Reconstruction of Magmatic System Variables & Processes 225
( P − 325) (900 − T )
Xc = + (12)
1000 300
where P is pressure in MPa and T is temperature in °C. This relationship is based crudely on
experimental studies on andesite and dacite magmas. We assume that the magma chamber
behaves as a closed system in which no gas is lost and we have not modeled the effects of
vapor composition on liquidus temperatures (and hence crystallinity), or of latent heat release
on magma temperature. Despite these simplifications, our model chamber shows many of
the characteristics to be expected in natural systems, such as vertical and lateral gradients in
crystallinity and bubble content (Fig. 30). Of course, magma chambers with different aspect
ratios will differ in detail from the model in Figure 30, although the overall spatial variation of
parameters will be similar. Note also that in the following discussion we ignore the effects of
magma chamber convection on the thermal structure.
As formulated, magma attains volatile saturation at 270 MPa within the interior of the
chamber, but at slightly higher pressure towards the cooler chamber walls. This divides
the chamber into a thin lower layer (or lens) that is vapor-undersaturated and a dominant
upper layer that is vapor-saturated (shaded region of Fig. 30). Decreasing pressure within the
chamber leads to crystallization, as does cooling of the magma towards the walls. We have
calculated the evolution of the melt composition along a series of isobaric sections (Fig. 31),
which covers a substantial portion of the CO2-H2O-Xc plots (Fig. 30). Importantly, the entire
population of equilibrium melts residing within the chamber (shown as shaded fields in Fig.
31) does not follow any of the simple end-member scenarios illustrated above.
Interpreting magmatic systems
We can use the insights gained from our model to interpret the behavior of the limited
number of sub-volcanic systems for which substantial datasets of melt inclusion CO2, H2O
and incompatible element analyzes exist. In particular, we compare the crystallization and
degassing behavior of large silicic caldera-forming systems (Bishop Tuff, California; Oruanui,
New zealand) with that of smaller, less evolved and more frequently active systems (Mt. Etna,
Italy; Popocatépetl, Mexico). For each example we have calculated the crystallinity (Xc) from
incompatible elements measured within the inclusions by assuming perfect incompatibility
(i.e., D = 0) and adopting the lowest incompatible element content of any melt inclusion as the
parental melt. Ideally, these calculated crystallinities would be compared with actual crystal
size distributions, but to date, no such complete study exists.
Large silicic systems. Although H2O and CO2 contents of rhyolitic ignimbrites produced
during caldera-forming eruptions of the Bishop Tuff and Oruanui record evidence of both
isobaric and polybaric crystallization (Fig. 32), the dominant signature is that of isobaric
volatile-saturated crystallization. Evidence for isobaric crystallization in the Oruanui system
includes the narrow (120-150 MPa) pressure range recorded by the melt inclusions (Fig. 32a)
and the remarkable constancy of H2O over a wide range of crystallinities (Fig. 32b). The low
CO2 contents of all Oruanui inclusions (Fig. 32c) suggest either some open system degassing
early in the evolution of the system or a low CO2 parent magma. Isobaric crystallization at
about 200 MPa is also indicated by the range of crystallinities at near-constant H2O shown by
the early and middle Bishop Tuff samples (Fig. 32b). Late Bishop Tuff samples, however, show
a pronounced decrease in CO2 with increasing crystallinity (Fig. 32c) that provides evidence
for vapor-saturation throughout the system (Wallace et al. 1999) and records crystallization
over a wide pressure range (100-300 MPa). The nature of crystallization in the late Bishop Tuff
has been the subject of detailed studies, some of which have suggested settling of shallow-
formed crystals followed by growth at deeper levels (Anderson et al. 2000; Peppard et al.
2001). Alternatively, explaining this trend by the isobaric sections illustrated in Figure 31
would require crystallization across a large temperature interval (which is not observed).
226 Blundy & Cashman
Figure 31. Calculated evolution of melt
inclusions in the model magma chamber
of Fig. 30 in terms of (a) CO2 and H2O
contents; (b) H2O as a function of wt%
crystallized and (c) CO2 as a function of
wt% crystallized. The solid lines denote
different isobaric section through the
chamber and the large filled circle is the
vapor-undersaturated parental magma.
The shaded region denotes the field
of melt inclusions likely to be tapped
during an eruption of the chamber. Note
that they do not match any of the simple
scenarios illustrated in Figure 29.
Reconstruction of Magmatic System Variables & Processes 227
1600 Mt. Etna. Melt inclusions in
(a) samples from the 2001-2002 eruptions
35
Oruanui (early)
1400
0
Oruanui (middle) of Mt. Etna preserve a very different
M
Pa
1200 Oruanui (late)
crystallization and degassing story to
25
Bishop (early)
0
1000 that seen in large rhyolitic systems.
M
Bishop (middle)
Pa
800 Bishop (late) As indicated by contemporaneous
15 monitoring data (e.g., Andronico et
600 0
M
CO2 (ppm) Pa al. 2005), the basaltic magma that fed
400
50 these eruptions ascended rapidly from
200 MP intermediate crustal levels, particularly
a
0 in 2001. This ascent path is nicely
0 1 2 3 4 5 6 7 recorded by the volatiles trapped in
melt inclusions and melt embayments,
H2O (wt%)
7 which indicate decompression and
(b) degassing from almost 500 MPa to
6 the surface (Fig. 33a). Crystallization
5 accompanying degassing is evinced
by strong depletions in both H2O
4 (Fig. 33b) and CO2 (Fig. 33c) with
3 increasing crystallinity. The data also
H2O (wt%)
retain interesting details about the
2
decompression path, as highlighted
1 by the spread of H2O and CO2 data
at pressures at and somewhat above
0
200 MPa, particularly in the 2002
0 20 40 60
samples (Fig. 33a). These data
Crystallinity (%) have been interpreted to indicate
1200
temporary magma storage at about 5
(c) km, which appears to coincide with
1000 a stratigraphic boundary between
underlying flysch and carbonate units
800 (Spilliaert et al. 2006). The range in
volatile compositions at this level is
600
inferred to reflect variable addition of
400(ppm) CO2-rich gas to the ponded magma.
CO2
An interesting consequence of this
200 fluxing would have been dehydration
of the magma (by addition of CO2).
0 Reduction of pH2O, in turn, would
0 20 40 60
promote crystallization, as seen in
Crystallinity (%) the ‘blurring’ of the CO2-crystallinity
data for the 2002 samples as compared
Figure 32. Volatile contents and crystallinity of early,
middle and late Oruanui (New zealand) and Bishop Tuff to the simpler story of rapid CO2
(USA) magmas as recorded by melt inclusions in terms depletion in 2001 (Fig. 33c).
of: (a) H2O-CO2, with isobars calculated at 760 °C from
Liu et al. (2005); (b) H2O versus wt% crystallized; (c) CO2 Small intermediate systems.
versus wt% crystallized. Oruanui data are taken from Liu There are only a few andesitic (and
et al. (2006); crystallinity is calculated from Rb content dacitic systems) for which complete
of melt inclusions relative to parent Rb content of 89 ppm (or near-complete) melt inclusion
assuming DRb = 0. Bishop data are taken from Wallace et data sets have been published. Here
al. (1999); crystallinity is calculated from Th content of
melt inclusions relative to parent Th content of 12.2 ppm we examine data from Popocaté-
assuming DTh = 0. petl (Mexico) and more limited data
228 Blundy & Cashman
5000 from Augustine (Alaska). Popocaté-
500
2001 MPa petl has been erupting intermittently
4000 2002 since 1994, during which time it has
emitted substantially more gas than
3000 300 MPa magma (e.g., Delgados-Granados et
al. 2001, 2008). The melt inclusion
2000
CO2 (ppm)
data are interesting in that they record
1000
of wide range of entrapment pres-
100 MPa
sures (50-300 MPa) at near-constant
0 H2O (~ 2 wt%; Fig. 34a). The result is
0 1 2 3 4 crystallization at constant H2O (Fig.
H2O (wt%) 34b) but a wide spread (and no ob-
vious pattern) of CO2 versus crystal-
4
linity (Atlas et al. 2006). These data
are particularly intriguing when com-
pared to those from Augustine, which
3
have low CO2 and well-developed
degassing-crystallization trends (Fig.
2 34a,b). They also contrast with the
strong degassing-crystallization trend
H2O (wt%) preserved in H2O-crystallinity data
1
from Mount St. Helens (Blundy and
0
Cashman 2005, 2008; Fig. 34b). We
0 20 40 60 suspect that the range in melt inclu-
Crystallinity (%)
sion compositions in Popocatépetl
magma result from the strong gas
fluxing observed in this system. In
5000 this way, Popocatépetl is similar to
Etna, except that the overall crystal-
4000
linity is higher (and average magma
3000
flux is lower). Like Etna, however,
both the degassing and gas fluxing
2000 trends shown in Figure 34 record
CO2 (ppm) crystallization over a wide range in
1000 pressure that seems characteristic of
small, rapidly replenished arc magma
0 systems.
0 20 40 60
Implications for magma accu-
Crystallinity (%) mulation and eruption. If the con-
Figure 33. Volatile contents and crystallinity of basaltic clusions about crystallization and de-
tephra erupted from Mt. Etna (Italy) in 2001 and 2002 gassing of magmatic systems drawn
as recorded by melt inclusions in terms of: (a) H2O-CO2, from the examples above can be gen-
with isobars from Metrich et al. (2004); (b) CO2 versus eralized, they might provide impor-
wt% crystallized; (c) H2O versus wt% crystallized. Data tant insights into conditions related
are taken from Metrich et al. (2004) and Spilliaert et al.
(2006); crystallinity calculated from K2O content of the to recharge and eruption patterns in
melt inclusions relative to the parent concentration (from the different tectonic environments.
appropriate bulk rock composition) assuming DK = 0.
Bishop Tuff and Oruanui are
large bodies of limited vertical extent
where crystallization appears to be dominantly isobaric, volatile-saturated, and driven by cool-
ing. They therefore resemble a version of our simple magma chamber (Fig. 30) with a high
width:height ratio. The large size of these magma bodies requires that they form over long time
Reconstruction of Magmatic System Variables & Processes 229
periods, probably by periodic
1600
(a) replenishment from below (e.g.,
1400 Annen et al. 2006). One can en-
1200 vision large laterally extensive
30 sheet-like bodies, similar to many
1000 0
M granite batholiths, whose tem-
800 Pa
20 perature wax and wane over time
0
600 M in response to episodic recharge
CO2 (ppm) Pa
400 10 events. Tectonic environments
0 that promote large, shallow, peri-
200 MP
a odically replenished magma bod-
0
ies tend to be extensional. Erup-
0 1 2 3 4 5 6 7
tions are infrequent and involve
H2O (wt%) wholesale roof collapse; it is not
clear whether eruptions are trig-
7 gered by external factors (e.g.,
(b) basaltic inputs, roof collapse) or
6
internal factors (e.g., overpres-
5 MSH
surization by gas accumulation
Popo
4 during crystallization).
Augustine
3 Etna, Augustine, Popoca-
H2O
2 (wt%) tépetl and Mount St. Helens
are stratovolcanoes with small,
1
vertically extensive magmatic
0 systems over which substantial
0 20 40 60 80 crystallization occurs due to the
Crystallinity (%) pressure drop alone. They there-
fore resemble our model magma
1600 chamber with high height:width
(c) ratio. While the pipe- or dyke-
1400
like shape of the magma reser-
1200
voirs argues for crystallization
1000 controlled primarily by decom-
800 pression, one might anticipate
600
additional crystallization by
CO2 (ppm) cooling along the vertically ex-
400
tensive walls. However, not only
200 does the steady fall in H2O with
0 increasing crystallinity argue
0 20 40 60 80 against much cooling during
crystallization, but combined
Crystallinity (%) melt inclusion, crystallinity and
Figure 34. Volatile contents and crystallinity of Augustine plagioclase-melt geothermom-
(Alaska)1986 and Popocatépetl (Mexico)1994-1998 magma etry data from Mount St. Helens
as recorded by melt inclusions in terms of: (a) H2O-CO2, with show that magma undergoing
isobars calculated at 1050 °C from Liu et al. (2005); (b) H2O decompression heats up due to
versus wt% crystallized; (c) CO2 versus wt% crystallized. Data the release of latent heat of crys-
are taken from Roman et al. (2006) and Atlas et al. (2006).
Crystallinity is calculated from K2O content of melt inclusions tallization (Blundy et al. 2006).
relative to parent K2O content assuming DK = 0. In (b), data from Stratovolcanoes are common in
Mount St. Helens are shown for comparison (from Blundy and arc environments. They tend to
Cashman 2005). be characterized by small fre-
230 Blundy & Cashman
quent eruptions that are often (although not always) triggered by mafic recharge events that mix
incompletely with reservoir magma, as evidenced by the common presence of mafic enclaves
(e.g., Montserrat, Unzen, Pinatubo). Absence of a cooling signature may derive from the rela-
tively short repose intervals that characterize such eruptions.
SuMMary
In summary, petrologic studies are so integral to the investigation of magmatic systems
that they are often taken for granted by those who apply new geophysical and geochemical
monitoring techniques to active volcanoes. Here we have shown ways in which the traditional
strengths of petrologic studies of magma generation and storage can be integrated with real-
time monitoring data to connect direct investigations of magmatic processes (particularly
volatile exsolution and crystallization) with manifestations of those processes that are used in
monitoring, such as seismicity and gas emissions.
“Traditional” petrologic methods used to study magma storage conditions focus on con-
straining intensive parameters such as temperature, pressure, oxygen fugacity and dissolved
volatile contents. Our review of commonly-used mineral-mineral geothermometers shows that
they are generally robust, with Fe-Ti oxides providing the best constraints. In contrast, geo-
barometry is notoriously difficult, particularly for upper crustal pressures. For this reason, we
review alternative methods of estimating equilibration pressures, including phase equilibria ex-
periments, volatile saturation models, matrix glass compositions and crystal textures. As no one
method is perfect, we recommend using as many different methods as permitted by the sample
set. Mineral-melt thermobarometry (Putirka 2008) offers significant advances in this regard.
We confine much of our comparison of petrologic and monitoring data to the well-
characterized eruption of Mount St. Helens in the 1980s. This is not to suggest that Mount
St. Helens should, in any way, be considered a reference standard, but simply because both
the nature of the eruption (with styles ranging from plinian to sub-plinian, vulcanian, and
effusive) and the wealth of monitoring and petrologic data allow us to compare these datasets
for different eruptive conditions. We find an impressive correlation between the petrology and
the monitoring data, which provides some interesting insights into the chemical and physical
processes occurring during the 1980-1986 eruption. Specifically, comparison of fO2 recorded
in Fe-Ti oxides and crater fumaroles demonstrates that shallow degassing causes a reduction
in magmatic fO2, as predicted by Burgisser and Scaillet (2007). Comparison of petrologic
and measured SO2 emissions shows these estimates to be similar for eruptive episodes, but
implicates non-erupted magma in generating inter-eruptive gas fluxes, thereby raising questions
about the physical mechanisms of gas transport in active, but non-erupting, magmatic systems.
Comparison of petrologic estimates of magma storage pressures with seismic data provides
important insights into the development of subvolcanic regions of magma storage and transport
and raises the tantalizing possibility that the geometry of subvolcanic regions may control the
style of ensuing eruptions. Where large numbers of melt inclusions from a single eruption are
analyzed for H2O and CO2, there exists the possibility of constraining conduit diameter, an
essential, but poorly-constrained, parameter in dynamical modeling. Improved geospeedometry
techniques (either chemical or textural) would also improve geometric constraints.
In stepping back to examine the evolution of magma storage regions between eruptions,
we demonstrate the importance of fully characterizing the phase proportions and composition
of individual samples, as well as the full range of melt inclusion compositions contained
within those phases. Simple models of degassing and crystallization illustrate the potential for
comprehensive melt inclusion studies to elucidate magma decompression and crystallization
histories, particularly when combined with trace element and isotopic analyzes of the same
inclusions. Application of our simple models to examples of both large silicic and small
Reconstruction of Magmatic System Variables & Processes 231
intermediate systems demonstrates the profound difference in these systems, from their geometry
and crystallization history to implied conditions of magma input and eruptive frequency.
Generalization of the conclusions drawn from examination of the few systems for which
complete data sets are available will require generation of similar (or improved, particularly by
addition of textural data) datasets for more magmatic systems. As the tools for such analysis are
now readily available, we hope this review will inspire collaborative efforts to fully characterize
many different types of magmatic systems, and to integrate those observations with studies in
physical volcanology, numerical modeling and volcano monitoring.
aCKnOWLedGMentS
JDB thanks NERC for the award of a Senior Research Fellowship (NER/K/S/2001/00771);
KVC thanks NSF for funding support through award EAR0510437. We are grateful to M.
Mangan and M. Humphreys for providing timely and constructive reviews of this manuscript,
to B. Scaillet, J. Beard, M. Ghiorso, K. Klimm, V. Martin, D. Morgan, D. Muir, C. Ginibre, E.
Gottesfeld, J. Riker, M. Folley and M. Humphreys for providing unpublished data or images,
J. Craven, R. Hinton and S, Kearns for critical reading of the microbeam analytical sections
and K. Putirka for his comments and editorial patience.
reFerenCeS
Aiuppa A, Moretti R, Federico C, Giudice G, Gurrieri S, Liuzzo M, Papale P, Shinohara H, Valenza M (2007)
Forecasting Etna eruptions by real-time observation of volcanic gas composition. Geology 35:1115-118
Alonso-Perez R, Müntener O, Ulmer P (2008) Igneous garnet and amphibole fractionation in the roots of island
arcs: experimental constraints on H2O undersaturated andesitic liquids. Contrib Mineral Petrol (in press)
Andersen DJ, Lindsley DH (1988) Internally-consistent solution models for Fe-Ti-Mg-Mn oxides – Fe-Ti
oxides. Am Mineral 73:714-726
Andersen DJ, Lindsley DH, Davidson PM (1993) qUILF: a Pascal program to assess equilibria among Fe-Mg-
Mn-Ti oxides, pyroxenes, olivine, and quartz. Comp Geosci 19:1333-1350
Anderson AT Jr, Davis AM, Lu Fq (2000) Evolution of Bishop Tuff rhyolitic magma based on melt and
magnetite inclusions and zoned phenocrysts. J Petrol 41:449-473
Anderson JL, Barth AP, Wooden JL, Mazdab F (2008) Thermometers and thermobarometers in granitic systems.
Rev Mineral Geochem 69:121-142
Andronico D, Branca S, Calvari S, Burton M, Caltabiano T, Corsaro RA, Del Carlo P Garfì G, Lodato L,
Miraglia L, Murè F, Neri M, Pecora E, Pompilio M, Salerno G, Spampinato L (2005) A multi-disciplinary
study of the 2002-03 Etna eruption: Insights into a complex plumbing system. Bull Volcanol 67:314-330
Annen C, Blundy JD, Sparks RSJ (2006) The genesis of intermediate and silicic magmas in deep crustal hot
zones. J Petrol 47:505-539
Arce JL, Macias JL, Gardner JE, Layer PW (2006) A 2.5 ka History of Dacitic Magmatism at Nevado de Toluca,
Mexico: Petrological, 40Ar/39Ar Dating, and Experimental Constraints on Petrogenesis. J Petrol 47:457-
479
Armienti P (2008) Decryption of igneous rock textures: crystal size distribution tools. Rev Mineral Geochem
69:623-649
Armienti P, Pareschi MT, Innocenti F, Pompilio M (1994) Effects of magma storage and ascent on the kinetics
of crystal growth. Contrib Mineral Petrol 115:402-414
Atlas zD, Dixon JE, Sen G, Finny M, Martin-Del Pozzo AL (2006) Melt inclusions from Volcan Popocate´petl
and Volca´n de Colima, Mexico: Melt evolution due to vapor-saturated crystallization during ascent. J
Volcanol Geotherm Res 153:221-240
Bachmann O, Dungan M (2002) Temperature-induced Al-zoning in hornblendes of the Fish Canyon magma,
Colorado. Am Mineral 87:1062-1076
Bacon CR, Hirschmann MM (1988) Mg/Mn partitioning as a test for equilibrium between coexisting Fe-Ti
oxides. Am Mineral 73:57-61
Baker DR (1991) Interdiffusion of hydrous dacitic and rhyolitic melts and the efficacy of rhyolite contamination
of dacitic enclaves. Contrib Mineral Petrol 106:462-473
Baker DR (2008) The fidelity of melt inclusions as records of melt composition. Contrib Mineral Petrol
doi10.1007/s00410-008-0291-3
232 Blundy & Cashman
Barclay J, Rutherford MJ, Carroll MR, Murphy MD, Devine JD, Gardner J, Sparks RSJ (1998) Experimental
phase equilibria constraints on pre-eruptive storage conditions of the Soufriere Hills magma. Geophys Res
Lett 25:3437-3440
Barclay J, Carmichael ISE (2004) A Hornblende Basalt from Western Mexico: Water-saturated Phase Relations
Constrain a Pressure-Temperature Window of Eruptibility. J Petrol 45:485-506
Beard JS, Lofgren GE (1991) Dehydration melting and water-saturated melting of basaltic and andesitic
greenstones and amphibolites at 1, 3 and 6.9 kb. J Petrol 32:365-401
Berlo K, Blundy J, Turner S, Hawkesworth C (2007) Textural and chemical variation in plagioclase phenocrysts
from the 1980 eruptions of Mount St. Helens, USA. Contrib Mineral Petrol 154:291-308
Bindeman IN (2005) Fragmentation phenomena in populations of magmatic crystals. Am Mineral 90:1801-
1815
Blatter DL, Carmichael ISE (1998) Plagioclase-free andesites from zitacuro (Michoacan), Mexico: petrology
and experimental constraints. Contrib Mineral Petrol 132: 121-138
Blatter DL, Carmichael ISE (2001) Hydrous phase equilibria of a Mexican high-silica andesite: A candidate for
a mantle origin? Geochim Cosmochim Acta 65:4043–4065
Blundy J, Cashman K (2001) Magma ascent and crystallization at Mount St. Helens, 1980-1986. Contrib
Mineral Petrol 140:631-650
Blundy J, Cashman KV (2005) Rapid decompression crystallization recorded by melt inclusions from Mount St.
Helens volcano. Geology 33:793-796
Blundy J, Cashman K, Humphreys M (2006) Magma heating by decompression-driven crystallization beneath
andesite volcanoes. Nature 443:76-80
Blundy J, Cashman K, Berlo K (2008) Evolving magma storage conditions beneath Mount St. Helens inferred
from chemical variations in melt inclusions from the 1980-1986 and current eruptions. Sherrod DR, Scott
WE, Stauffer PH eds, A volcano rekindled; the renewed eruption of Mount St. Helens, 2004–2006: U.S.
Geological Survey Professional Paper 1750 (in press)
Bogaerts M, Scaillet B, van der Auwera J (2006) Phase equilibria of the Lyngdal granodiorite (Norway):
Implications for the origin of metaluminous ferroan granitoids. J Petrol 47:2405-2431
Brey GP, Köhler T (1990) Geothermobarometry in Four-phase Lherzolites II. New Thermobarometers, and
Practical Assessment of Existing Thermobarometers. J Petrol 31:1353-1378
Browne BL, Gardner JE (2006) The influence of magma ascent path on the texture, mineralogy, and formation
of hornblende reaction rims. Earth Planet Sci Lett 246: 161-176
Brugger CR, Johnston AD, Cashman KV (2003) Phase equilibria in silicic systems at one-atmosphere pressure
Contrib Mineral Petrol 146:356-369
Burgisser A, Scaillet B (2007) Redox evolution of a degassing magma rising to the surface. Nature 445:194-
197
Burton MR, Oppenheimer C, Horrock LA, Francis PW (2000) Remote sensing of CO2 and H2O emission rates
from Masaya volcano, Nicaragua. Geology 28:915-918
Carey S, Sigurdsson H (1985) The May 18, 1980 eruption of Mount St, Helens. 2. Modeling of dynamics of the
plinian phase. J Geophys Res 90:2948-2958
Carlson WD (2006) Three-dimensional imaging of earth and planetary materials. Earth Planet Sci Lett 249:133-
147
Cashman KV (1990) Textural constraints on the kinetics of crystallization of igneous rocks. Rev Mineral
24:259-314
Cashman KV (1992) Groundmass crystallization of Mount St. Helens dacite, 1980-1986: A tool for interpreting
shallow magmatic processes. Contrib Mineral Petrol 109:431-449
Cashman K, Blundy J (2000) Degassing and crystallization of ascending andesite Phil Trans Roy Soc 358:1487-
1513
Cashman KV, Hoblitt RP (2004) Magmatic precursors to the May 18, 1980 eruption of Mount St. Helens, WA.
Geology 32:141-144
Cashman KV, Marsh BD (1988) Crystal size distribution (CSD) in rocks and the kinetics and dynamics of
crystallization II. Makaopuhi lava lake. Contrib Mineral Petrol 99:292-305
Cashman KV, McConnell S (2005) Transitions from explosive to effusive activity – the summer 1980 eruptions
of Mount St. Helens. Bull Volcanol 68:57-75
Cashman KV, Taggart JE (1983) Petrologic monitoring of 1981-1982 eruptive products from Mount St. Helens,
Washington. Science 221:1385-1387
Cashman KV, Thornber CR, Kauahikaua JP (1999) Cooling and crystallization of lava in open channels, and the
transition of pahoehoe lava to `a`a. Bull Volcanol 61:306-323
Cashman KV, Thornber CR, Pallister JS (2008) From dome to dust: shallow crystallization and fragmentation
of conduit magma during the 2004-2006 dome extrusion of Mount St. Helens. In Sherrod DR, Scott WE,
Stauffer PH, eds, A volcano rekindled: the renewed eruption of Mount St. Helens, 2004–2006. US Geol
Surv Prof Pap 1750 (in press)
Reconstruction of Magmatic System Variables & Processes 233
Castro JM, Cashman KV, Manga M (2003) A technique for measuring 3D crystal-size distributions of prismatic
microlites in obsidian. Am Mineral 88:1230-1240
Castro JM, Manga M, Cashman KV (2002) Dynamics of obsidian flows inferred from microstructures: insights
from microlite preferred orientations. Earth Planet Sci Lett 199:211-226
Chayes F (1956) Petrographic Modal Analysis. John Wiley & Sons, New York
Chou I M (1987) Oxygen buffer and hydrogen sensor techniques at elevated pressures and temperatures. In:
Hydrothermal Experimental Techniques. Barnes HL, Ulme GC (eds) John Wiley, New York, p 69-99
Chouet BA (2003) Volcano seismology. Pure Appl Geophys 160:739-788
Clarke AB, Stephens S, Teasdale R, Sparks RSJ, Diller K (2007) Petrologic constraints on the decompression
history of magma prior to Vulcanian explosions at the Souffriere Hills volcano, Montserrat. J Volc
Geotherm Res 161:261-274
Coombs ML, Gardner JE (2001) Shallow-storage conditions for the rhyolite of the 1912 eruption at Novarupta,
Alaska. Geology 29:775-778
Coombs ML, Eichelberger JC, Rutherford MJ (2000) Magma storage conditions for the 1953-1974 eruption of
Southwest Trident volcano, Katmai National Park, Alaska. Contrib Mineral Petrol 140:99-118
Coombs ML, Eichelberger JC, Rutherford MJ (2002) Experimental and textural constraints on mafic enclave
formation in volcanic rocks. J Volc Geotherm Res 119:125-144
Cooper KM, Reid MR (2003) Re-examination of crystal ages in recent Mount St. Helens lavas: Implications for
magma reservoir processes. Earth Planet Sci Lett 213:149-167
Costa F, Chakraborty S, Dohmen R (2003) Diffusion coupling between trace and major elements and a model
for calculation of magma residence times using plagioclase. Geochim Cosmochim Acta 67:2189-2200
Costa F, Chakraborty S (2004) Decadal time gaps between mafic intrusion and silicic eruption obtained from
chemical zoning patterns in olivine. Earth Planet Sci Lett 227:517-530
Costa F, Scaillet B, Pichavant M (2004) Petrological and experimental constraints on the pre-eruption conditions
of Holocene dacite from Volcán San Pedro (36°S, Chilean Andes) and the importance of sulfur in silicic
subduction-related magmas. J Petrol 45:855–881
Costa F, Dohmen R, Chakraborty S (2008) Time scales of magmatic processes from modeling the zoning
patterns of crystals. Rev Mineral Geochem 69:545-594
Cottrell E, Gardner JE, Rutherford MJ (1999) Petrologic and experimental evidence for the movement and
heating of the pre-eruptive Minoan rhyodacite (Santorini, Greece). Contrib Mineral Petrol 135: 315-331
Couch S (2003) Experimental investigation of crystallization kinetics in a haplogranite system. Am Mineral
88:1471-1485
Couch S, Harford CL, Sparks RSJ, Carroll MJ (2003a) Experimental constraints on the conditions of formation
of highly calcic plagioclase microlites at the Soufriere Hills Volcano, Montserrat. J Petrol 44:1455-1475
Couch S, Sparks RSJ, Carroll MR (2003b) The kinetics of degassing-induced crystallization at Soufriere Hills
volcano, Montserrat. J Petrol 44:1477-1502
Dall’Agnol R, Scaillet B, Pichavant M (1999) An experimental study of a Lower Proterozoic A-type granite
from the eastern Amazonian craton, Brazil. J Petrol 40:1673-1698
Davidson JP, Hora JM, Garrison J, Dungan MA (2005) Crustal forensics in arc magmas. J Volc Geotherm Res
140:157-170
Davidson JP, Morgan DJ, Charlier BLA, Harlou R, Hora JM (2007) Microsampling and isotopic analysis of
igneous rocks: Implications for the study of magmatic systems. Ann Rev Earth Planet Sci 35:273-311
Devine JD, Gardner JE, Brack HP, Layne GD, Rutherford MJ (1995) Comparison of analytical methods for
estimating H2O contents of silicic volcanic glasses. Am Mineral 80:319-328
Devine JD, Rutherford, MJ, Norton GE, Young SR (2003) Magma storage region processes inferred from
geochemistry of Fe-Ti oxides in andesitic magma, Soufriere Hills Volcano, Montserrat, WI. J Petrol
44:1375-1400
Donato P, Behrens H, De Rosa R, Holtz F, Parat F (2006) Crystallization conditions in the Upper Pollara magma
chamber, Salina Island, Southern Tyrrhenian Sea. Mineral Petrol 86: 89–108
Donaldson CH (1976) An experimental investigation of olivine morphology. Contrib Mineral Petrol 57:187-
213
Dunbar NW, Cashman KV, Dupre R, Kyle PR (1994) Crystallization processes of anorthoclase phenocrysts:
Evidence from crystal compositions, crystal size distributions and volatile contents of melt inclusions.
Antarctic Res Ser 66:129-146
Eberl DD, Kile DE, Drits VA (2002) On geological interpretations of crystal size distributions: Constant vs.
proportionate growth. Am Mineral 87:1235-1241
Ernst WG, Liu J (1998) Experimental phase-equilibrium study of Al- and Ti-contents of calcic amphibole in
MORB - A semiquantitative thermobarometer. Am Mineral 83:952-969
Folley M (1999) Crystallinity, rheology and surface morphology of basaltic lavas, Kilauea Volcano, Hawai`i.
Unpublished MS Thesis, Univ Oregon
234 Blundy & Cashman
Fink JH, Griffiths (1998) Morphology, eruption rates, and rheology of lava domes: Insights from laboratory
models. J Geophys Res 103: 527-545
Gardner JE (2007) Bubble coalescence in rhyolitic melts during decompression from high pressure. J Volcanol
Geotherm Res 166:161–176
Gardner CA, Cashman KV, Neal CA (1998) Tephra-fall deposits from the 1992 eruption of Crater Peak, Alaska:
implications of clast textures for eruptive processes Bull Volcanol 59: 537-555
Gardner JE, Hilton M, Carroll M (1999) Experimental constraints on degassing of magma: isothermal bubble
growth during continuous decompression from high pressure. Earth Planet Sci Lett 168:201–218
Gardner JE, Rutherford M, Carey S, Sigurdsson H (1995) Experimental constraints on pre-eruptive water
contens and changing magma storage prior to explosive eruptions of Mount St. Helens volcano. Bull Volc
57:1-17
Gauthier PJ, Condomines M (1999) 210Pb-226Ra radioactive disequilibria in recent lavas and radon degassing:
inferences on the magma chamber dynamics at Stromboli and Merapi volcanoes. Earth Planet Sci Lett
172: 111-126
Gerlach TM, Casadevall TJ (1986) Evaluation of gas data from high-temperature fumaroles at Mount St. Helens,
1980-1982. J Volcanol Geotherm Res 28:107-140
Geschwind CH, Rutherford MJ (1992) Cummingtonite and the evolution of the Mount St. Helens (Washington)
magma system: An experimental study. Geology 20:1011-1014
Ghiorso MS, Sack RO (1991) Fe-Ti oxide geothermometry – thermodynamic formulation and the estimation of
intensive variables in silicic magmas. Contrib Mineral Petrol108: 485-510
Ghiorso MS, Evans BW (2008) Thermodynamics of rhombohedral oxide solid solutions and a revision of the
Fe-Ti two-oxide geothermometer and oxygen-barometer. Am J Sci (in press)
Ginibre C, Kronz A, Woerner G (2002) High-resolution quantitative imaging of plagioclase composition using
accumulated backscattered electron images; new constraints on oscillatory zoning. Contrib Mineral Petrol
142: 436-44
Gonnerman HM, Manga M (2005) Nonequilibrium magma degassing: Results from modeling of the ca. 1340
A.D. eruption of Mono Craters, California. Earth Planet Sci Lett 238:1-16
Grove TL, Donnelly-Nolan JM, Housch T (1997) Magmatic processes that generated the rhyolite of Glass
Mountain, Medicine Lake volcano, N. California. Contrib Mineral Petrol 127: 205-223
Grove TL, Elkins-Tanton LT, Parman SW, Chatterjee N, Münetner O, Gaetani G (2003) Fractional crystallization
and mantle-melting controls on calc-alkaline differentiation trends. Contrib Mineral Petrol 145: 515–533
Gualda GAR, Rivers M (2006) quantitative 3D petrography using x-ray tomography: Application to Bishop
Tuff pumice clasts. J Volcanol Geotherm Res154: 48-62
Gualda GAR (2006) Crystal size distributions derived from 3D datasets: sample size versus uncertainties. J
Petrol 47: 1245-1254
Gardien V, Thompson AB, Ulmer P (2000) Melting of biotite + plagioclase + quartz gneisses: the role of H2O in
the stability of amphibole. J Petrol 41:651-666
Gurioli L, Houghton BF, Cashman KV, Cioni R (2005) Complex changes in eruption dynamics and the transition
between Plinian and phreatomagmatic activity during the 79 AD eruption of Vesuvius Bull Volcanol
67:144-159
Halter WE, Pettke T, Heinrich CA, Rother-Rutishauser B (2002) Major to trace element analysis of melt
inclusions by laser-ablation ICP-MS: methods of quantification. Chem Geol 183: 63-86
Hammer JE (2008) Experimental studies of the kinetics and energetic of magma crystallization. Rev Mineral
Geochem 69:9-59
Hammer JE, Cashman KV, Voight B. Magmatic processes revealed by textural and compositional trends in
Merapi dome lavas. J Volcanol Geotherm Res100:165–192
Hammer J, Rutherford M (2002) An experimental study of the kinetics of decompression-induced crystallization
in silicic melt. J Geophys Res 107, doi:10.1029/2001JB000281
Hammer JE, Rutherford MJ, Hildreth W (2002) Magma storage prior to the 1912 eruption at Novarupta, Alaska.
Contrib Mineral Petrol 144: 144-162
Hammer JE, Rutherford MJ (2003) Petrologic indicators of preeruption magma dynamics. Geology 31: 79-82
Hammer JE, Cashman KV, Voight B (2000) Magmatic processes revealed by textural and compositional trends
in Merapi dome lavas. J Volcanol Geotherm Res 100:165-192
Hammer JE, Cashman KV, Hoblitt RP, Newman S (1999) Degassing and microlite crystallization during pre-
climactic events of the 1991 eruption of Mt. Pinatubo, Phillipines. Bull Volcanol 60:355-380
Hammond PA, Taylor LA (1982) The ilmenite titano-magnetite assemblage – kinetics of reequilibration. Earth
Planet Sci Lett 61:143-150
Harris DM, Rose WI (1996) Dynamics of carbon dioxide emissions, crystallization and magma ascent:
hypotheses, theory, and applications to volcano monitoring at Mount St. Helens. Bull Volcanol 58:163-
174
Hauri E, Wang J, Dixon JE, King PL, Mandeville C, Newman S (2002) SIMS analysis of volatiles in silicate
glasses 1. Calibration, matrix effects and comparisons with FTIR. Chem Geol 183: 99-114
Reconstruction of Magmatic System Variables & Processes 235
Hellebrand E, Snow J, Mostefaoui S, Hoppe P (2005) Trace element distribution between orthopyroxene and
clinopyroxene in peridotites from the Gakkel Ridge: a SIMS and NanoSIMS study. Contrib Mineral Petrol
150: 486-504
Hersum TG, Marsh BD (2006) Igneous microstructures from kinetic models of crystallization. J Volcanol
Geotherm Res 154: 34-47
Hersum TG, Marsh BD (2007) Igneous textures: on the kinetics behind the words. Elements 3: 247-252
Higgins M (1994) Numerical modeling of crystal shapes in thin sections: estimation of crystal habit and true
size. Am Mineral 79: 113-119.
Higgins MD (2006) quantitative Textural Measurements in Igneous and Metamorphic Petrology. Cambridge
University Press, Cambridge, 276 pp
Hinton RW (1990) Ion microprobe trace-element analysis of silicates – measurement of multi-element glasses.
Chem Geol 83:11-25
Hinton RW (1995) Ion microprobe analysis in geology. In: Potts PJ, Bowles JFW, Reed SJB, Cave MR (eds)
Microprobe Techniques in the Earth Sciences, Chapman and Hall, London, pp 235-289
Holland T, Blundy J (1994) Non-ideal interactions in calcic amphiboles and their bearing on amphibole-
plagioclase thermometry. Contrib Mineral Petrol 116(4): 433-447
Holland T, Powell R (2001) Calculation of phase relations involving haplogranitic melts using an internally
consistent thermodynamic dataset. J Petrol 42:673-683
Holtz F, Sato H, Lewis J, Behrens H, Nakada S (2005) Experimental Petrology of the 1991–1995 Unzen Dacite,
Japan. Part I: Phase Relations, Phase Composition and Pre eruptive Conditions. J Petrol 46:319-337
Hoppe P (2006) NanoSIMS: A new tool in cosmochemistry. Appl Surf Sci 252:7102-7106
Humphreys MCS, Kearns SL, Blundy JD (2006) SIMS investigation of electron-beam damage to hydrous,
rhyolitic glasses: Implications for melt inclusion analysis. Am Mineral 91:667-679
Humphreys MCS, Menand T, Blundy JD, Klimm K (2008a) Magma ascent rates in explosive eruptions:
Constraints from H2O diffusion in melt inclusions. Earth Planet Sci Lett 270:25-40
Humphreys MCS, Blundy JD, Sparks RSJ (2008b) Shallow-level decompression crystallization and deep
magma supply at Shiveluch Volcano. Contrib Mineral Petrol 155:45-61
Ihinger PD, Hervig RL, Macmillan PF (1994). Analytical methods for volatiles in glasses. Rev Mineral Geochem
30:67-121
Iverson RM, Dzurisin D, Gardner CA, Gerlach TM, LaHusen R, Lisowski M, Major JJ, Malone SD, Messerich
JA, Moran SC, Pallister JS, qamar AI, Schilling SP, Vallance JW (2006) Dynamics of seismogenic volcanic
extrusion at Mount St. Helens in 2004–05. Nature 444:439-443
Jerram DA, Higgins MD (2007) 3D analysis of rock textures: quantifying igneous microstructures. Elements
3(4): 239-245
Johannes W, Holtz F (1996). Petrogenesis and Experimental Petrology of Granitic Rocks. Springer-Verlag,
Berlin, 335 pp
Johannes W, Koepke J (2001) Incomplete reaction of plagioclase in experimental dehydration melting of
amphibolite. Aust J Earth Sci 48:581-590
Johnson, M.C. and Rutherford, M.J. (1989) Experimentally determined conditions in the Fish Canyon Tuff,
Colorado, Magma Chamber. J Petrol 30:711-737
Jurewicz SR, Watson EB (1985) The distribution of partial melt in a granitic system: the application of liquid
phase sintering theory. Geochim Cosmochim Acta 49:1109-1121
Kaszuba JP, Wendlandt RK (2000) Effect of carbon dioxide on dehydration melting reactions and melt
compositions in the lower crust and the origin of alkaline rocks. J Petrol 41:363-386
Kawamoto T (1996) Experimental constraints on differentiation and H2O abundance of calc-alkaline magmas.
Earth Planet Sci Lett 144:577-589
Kearns SL, Steen N, Erlund E (2002) Electronprobe microanalysis of volcanic glass at cryogenic temperatures.
Microscop Microanal 8, Suppl 2:1562CD
Kent AJR (2008) Melt inclusions in basaltic and related volcanic rocks. Rev Mineral Geochem 69:273-331
Kile DE, Eberl DD, Hoch AR, Reddy MM (2000) An assessment of calcite crystal growth mechanisms based on
crystal size distributions. Geochim Cosmochim Acta 64:2937-2950
King PL, Venneman, TW, Holloway JR, Hervig RL, Lowenstern JE and Forneris, JF (2002) Analytical techniques
for volatiles: A case study using intermediate (andesitic) glasses. Am Mineral 87:1077-1089
Klimm K, Holtz F, Johannes W, King PL (2003) Fractionation of metaluminous A-type granites: an experimental
study of the Wangrah Suite, Lachlan Fold Belt, Australia. Precamb Res 124:327–341
Klug C, Cashman KV, Bacon CR (2002) Structure and physical characteristics of pumice from the climactic
eruption of Mt. Mazama (Crater Lake), Oregon. Bull Volcanol 64: 486-501
Larsen JF (2005) Experimental study of plagioclase rim growth around anorthite seed crystals in rhyodacitic
melt. Am Mineral 90:417-427
Larsen JF (2006) Rhyodacite magma storage conditions prior to the 3430 yBP caldera-forming eruption of
Aniakchak volcano, Alaska. Contrib Mineral Petrol 152:523–540
236 Blundy & Cashman
Larsen JF, Gardner JE (2004) Experimental study of water degassing from phonolite melts: implications for
volatile oversaturation during magmatic ascent. J Volc Geotherm Res 134:109-124
LePage L (2003) ILMAT: an excel worksheet for ilmenite--magnetite geothermometry and geobarometry.
Comput Geosci 29:673-678
Lindquist WB, Venkatarangan A (1999) Investigating 3D geometry of porous media from high resolution
images. Phys Chem Earth 25:593-599
Lindsley DH, Spencer KJ (1982) Fe–Ti oxide geothermometry: Reducing analyzes of coexisting Ti-magnetite
(Mt) and ilmenite (Ilm). Eos Trans, Am Geophys Union 63: 471
Liu Y, zhang Y, Behrens H (2005) Solubility of H2O in rhyolitic melts at low pressures and a new empirical
model for mixed H2O–CO2 solubility in rhyolitic melts. J Volcanol Geotherm Res 143:219-235
Liu Y, Anderson AT, Wilson CJN, Davis AM, Steele IM (2006) Mixing and differentiation in the Oruanui
rhyolitic magma, Taupo, New zealand: evidence from volatiles and trace elements in melt inclusions.
Contrib Mineral Petrol 151:71–87
Lofgren GE (1980) Experimental studies on the dynamic crystallization of silicate melts. In: Physics of Magmatic
Processes. Hargraves RB (ed) Princeton University Press, Princeton, p 487-551
López S, Castro A (2001) Determination of the fluid–absent solidus and supersolidus phase relationships of
MORB-derived amphibolites in the range 4–14 kbar. Am Mineral 86:1396-1403
Luhr JF (1990) Experimental phase relations of water- and sulfur-saturated arc magmas and the 1982 eruptions
of El Chichon Volcano. J Petrol 31:1071-1114
Madras G, McCoy BJ (2002) Ostwald ripening with size-dependent rates: Similarity and power-law solutions.
J Chem Phys 117: 8042-8049
Mangan MT (1990) Crystal size distribution systematics and the determination of magma storage times: The
1959 eruption of Kilauea volcano, Hawaii. J Volcanol Geotherm Res 44:295-302
Mangan M, Sisson T (2000) Delayed, disequilibrium degassing in rhyolite magma: decompression experiments
and implications for explosive volcanism. Earth Planet Sci Lett 183:441-455
Mangan M, Mastin L, Sisson T (2004a) Gas evolution in eruptive conduits: combining insights from high
temperature and pressure decompression experiments with steady-state flow modeling. J Volcanol
Geotherm Res 129:23-36
Mangan M, Sisson T, Hankins W (2004b) Decompression experiments identify kinetic controls on explosive
silicic Eruptions. Geophys Res Lett 31, L08605, doi:10.1029/2004GL019509
Marsh BD (1988) Crystal size distributions (CSD) in rocks and the kinetics and dynamics of crystallization I.
Theory. Contrib Mineral Petrol 99:277-291
Marsh BD (2006) Dynamics of magma chambers. Elements 2:287-292
Martel C, Pichavant M, Bourdier J-L, Traineau H, Holtz F, Scaillet B (1998). Magma storage conditions and
control of eruption regime in silicic volcanoes: experimental evidence from Mt. Pelée. Earth Planet Sci
Lett 156:89-99
Martel C, Pichavant M, Holtz F, Scaillet B (1999) Effects of fO2 and H2O on andesite phase relations between 2
and 4 kbar. J Geophys Res 104:29453-29470
Martel C, Bourdier JL, Pichavant M, Traineau H (2000) Textures, water content and degassing of silicic andesites
from recent plinian and dome-forming eruptions at Mount Pelee volcano (Martinique, Lesser Antilles arc).
J Volc Geotherm Res 96:191-206
Martel C, Schmidt B (2003) Decompression experiments as an insight into ascent rates of silicic magmas.
Contrib Mineral Petrol 144: 397–415
McCanta MC, Rutherford MJ, Hammer JE (2007) Pre-eruptive and syn-eruptive conditions in the Black Butte,
California dacite: Insight into crystallization kinetics in a silicic magma system. J Volcanol Geotherm Res
160:263–284
McGee KA, Casadevall TJ (1994) A Compilation of Sulfur Dioxide and Carbon Dioxide Emission-Rate Data
from Mount St. Helens during 1980-88. US Geol Surv Open-File Rep 94-212
McGonigle AJS, Aiuppa A, Giudice G, Tamburello G, Hodson AJ, Gurrieri S (2008) Unmanned aerial vehicle
measurements of volcanic carbon dioxide fluxes. Geophys Res Lett 35:L06303
Melson WG, Hopson CA (1981) Preeruption temperatures and oxygen fugacities in the 1980 eruptive sequence.
USGS Prof Pap 1250:641-648
Métrich N, Wallace PJ (2008) Volatile abundances in basaltic magmas and their degassing paths tracked by melt
inclusions. Rev Mineral Geochem 69:363-402
Metrich N, Allard P, Spilliaert N, Andronico D, Burton M (2004) 2001 flank eruption of the alkali- and volatile-
rich primitive basalt responsible for Mount Etna’s evolution in the last three decades. Earth Planet Sci Lett
228:1-17
Milman-Barris MS, Beckett JR, Baker MB, Hofmann AE, Morgan z, Crowley MR, Vielzeuf D, Stolper E
(2008) zoning of phosphorus in igneous olivine. Contrib Mineral Petrol 155:739-765
Molina JF, Poli S (2000) Carbonate stability and fluid composition in subducted oceanic crust: an experimental
study on H2O-CO2-bearing basalts. Earth Planet Sci Lett 176:295-310
Reconstruction of Magmatic System Variables & Processes 237
Moore G (2008) Interpreting H2O and CO2 contents in melt inclusions: constraints from solubility experiments
and modeling. Rev Mineral Geochem 69:333-361
Moore G, Carmichael ISE (1998) The hydrous phase equilibria (to 3 kbar) of an andesite and basaltic andesite
from western Mexico: constraints on water content and conditions of phenocryst growth. Contrib Mineral
Petrol 130:304-319
Morgan DJ, Blake SR, Rogers NW, DeVivo B, Rolandi G, Macdonald R, Hawkesworth CJ (2004) Time scales
of crystal residence and magma chamber volume from modeling of diffusion profiles in phenocrysts:
Vesuvius 1944. Earth Planet Sci Lett 222:933-946
Morgan DJ, Jerram DA (2006) On estimating crystal shape for crystal size distribution analysis. J Volcanol
Geotherm Res154:1-7
Morgan DJ, Blake SR, Rogers NW, DeVivo B, Rolandi G, Davidson JP (2006) Magma chamber recharge at
Vesuvius in the century prior to the eruption of A.D. 79. Geology 34: 845-848
Müntener O, Kelemen PB, Grove TL (2001) The role of H2O during crystallization of primitive arc magmas
under uppermost mantle conditions and genesis of igneous pyroxenites: an experimental study. Contrib
Mineral Petrol 141:643-658
Nakada S, Motomura Y (1999) Petrology of the 1991-1995 eruption at Unzen: effusion pulsation and groundmass
crystallization. J Volcanol Geotherm Res 89:173-196
Nakajima K, Arima M (1998) Melting experiments on hydrous low-K tholeiite: Implications for the genesis of
tonalitic crust in the Izu-Bonin-Mariana arc. Island Arc 7:359-373
Nemchin AA, Giannini LM, Bodorkos S, Oliver NHS (2001 Ostwald ripening as a possible mechanism for
zircon overgrowth formation during anatexis: theoretical constraints, a numerical model, and its application
to pelitic migmatites of the Tickalara Metamorphics, northwestern Australia. Geochim Cosmochim Acta
65:2771-1788
Neuberg J (2000) Characteristics and causes of shallow seismicity in andesite volcanoes. Phil Trans R Soc
358:1533-1546
Newman S, Lowenstern JB (2002) VolatileCalc: a silicate melt-H2O-CO2 solution model written in Visual Basic
for Excel. Comput Geosci 28:597-604
Nicholis MG, Rutherford MJ (2004) Experimental constraints on magma ascent rate for the Crater Flat volcanic
zone hawaiite. Geology 32:489-492
Nicholls IA, Oba T, Conrad WK (1992) The nature of primary rhyolitic magmas involved in crustal evolution:
Evidence from an experimental study of cummingtonite bearing rhyolites, Taupo Volcanic zone, New
zealand. Geochim Cosmochim Acta 56:955-962
Noguchi S, Toramaru A, Nakada S (2008a) Groundmass crystallization in dacite dykes taken in Unzen Scientific
Drilling Project (USDP-4) J Volcanol Geotherm Res doi:10.1016/j.jvolgeores.2008.03.037
Noguchi S, Toramaru A, Nakada S (2008b) Relation between microlite textures and discharge rate during the
1991-1995 eruptions at Unzen, Japan. J Volcanol Geotherm Res, doi:10.1016/j.jvolgeores.2008.03.025
Ohba T, Hirabayashi J-I, Nogami K, Kusakabe M, Yoshida M (2008) Magma degassing process during the
eruption of Mt. Unzen, Japan in 1991 to 1995: Modeling with the chemical composition of volcanic gas. J
Volcanol Geotherm Res doi:10.1016/j.jvolgeores.2008.03.040
Pallister JS, Hoblitt RP, Meeker GP, Knight RJ, Siems DF (1996) Magma mixing at Mount Pinatubo: Petrographic
and chemical evidence from the 1991 deposits. In: Fire and Mud: Eruptions and Lahars of Mount Pinatubo,
Philippines. Punongbayan RS, Newhall CG (eds) Univ Washington Press, Seattle, p 687-731
Papale P (2005) Determination of total H2O and CO2 budgets in evolving magmas from melt inclusion data. J
Geophys Res 110, doi:10.1029/2004JB003033
Papale P, Neri A, Macedonio G (1998) The role of magma composition and water content in explosive Eruptions
1. Conduit ascent dynamics. J Volcanol Geotherm Res 87:75-93
Parat F, Holtz F, Feig S (2008) Pre-eruptive conditions of the Huerto Andesite (Fish Canyon system, San Juan
volcanic field, Colorado): Influence of volatiles (COHS) on phase equilibria and mineral composition. J
Petrol 49:911-935
Patiño-Douce AE, Beard BS (1995) Dehydration-melting of Biotite Gneiss and quartz Amphibolite from 3 to
15 kbar. J Petrol 36:707-738
Peppard BT, Steele IM, Davis AM, Wallace PJ, Anderson AT (2001) zoned quartz phenocrysts from the rhyolitic
Bishop Tuff. Am Mineral 86:1034-1052
Pertermann M, Lundstrom C (2006) Phase equilibrium experiments at 0.5 GPa and 1100-1300 degrees C on a
basaltic andesite from Arenal volcano, Costa Rica. J Volcanol Geotherm Res 157:222-235
Pichavant M, Martel C, Boudier J-L, Scaillet B (2002) Physical conditions, structure, and dynamics
of a zoned magma chamber: Mount Pelée (Martinique, Lesser Antilles Arc). J Geophys Res 107,
doi:10.1029/2001JB000315
Pichavant M, Macdonald R (2007) Crystallization of primitive basaltic magmas at crustal pressures and genesis
of the calc-alkaline igneous suite: experimental evidence from St Vincent, Lesser Antilles arc. Contrib
Mineral Petrol 154:535–558
238 Blundy & Cashman
Pichavant M, Costa F, Burgisser A, Scaillet B, Martel C, Poussineau S (2007) Equilibration Scales in Silicic to
Intermediate Magmas Implications for Experimental Studies. J Petrol 48:1955-1972
Pinel V, Jaupart C (2000) The effect of edifice load on magma ascent beneath a volcano. Phil Trans Roy Soc
358:1515-1532
Polacci M, Baker DR, Bai L, Mancini L (2007) Large vesicles record pathways of degassing at basalt volcanoes.
Bull Volcanol doi 10.1007/s00445-007-0184-8
Poli S (1993) The amphibolite-eclogite transition: an experimental study on basalt. Am J Sci 293:1061-1107
Pompilio M (1998) Melting experiments on Mt. Etna lavas; I, The calibration of an empirical geothermometer
to estimate the eruptive temperature. Acta Vulcanol 10:1-9
Prouteau G, Scaillet B (2003) Experimental constraints on the origin of the 1991 Pinatubo dacite. J Petrol
44:2203-2241
Putirka KD (2008) Thermometers and barometers for volcanic systems. Rev Mineral Geochem 69:61-120
Ramos FC, Tepley FJ III (2008) Inter- and intracrystalline isotopic disequilibria: techniques and applications.
Rev Mineral Geochem 69:403-443
Randle V, Engler O (2000) Texture Analysis: Macrotexture, Microtexture, and Orientation Mapping. CRC
Press
Randolph AD, Larson MA (1971) Theory of Particulate Processes. Academic Press, New York, 251 pp
Rapp RP, Watson EB (1995) Dehydration melting of metabaslt at 8-32 kbar – implications for continental
growth and crust-mantle recycling. J Petrol 36:891-931
Resmini RG (2007) Modeling of crystal size distributions (CSDs) in sills. J Volcanol Geotherm Res161:118-
130
Robie RA, BS Hemingway, JR Fisher (1978) Thermodynamic properties of minerals and related substances at
298.15 K and 1 bar (105 pascals) pressure and at higher temperature. USGS Bull 1452
Roman DC, Cashman KV, Gardner CA, Wallace PJ, Donovan JJ (2006) Storage, degassing and pre-eruption
crystallization of compositionally heterogeneous magmas from the 1986 eruption of Augustine Volcano,
Alaska. Bull Volcanol 68:240-254
Rusk BG, Reed MH (2002) Scanning electron microscope-cathodoluminescence analysis of quartz reveals
complex growth histories in veins from the Butte porphyry copper deposit, Montana. Geology 30:727-
730
Rusk BG, Lowers HA, Reed MH (2008) Trace elements in hydrothermal quartz: Relationships to
cathodoluminescence textures and insights into vein formation. Geology 36:547-550
Rust A, Manga M, Cashman KV (2003) Determining flow type, shear rate and shear stress in magmas from
bubble shapes and orientations J Volcanol Geotherm Res 122:111-132
Rutherford MJ (2008) Magma ascent rates. Rev Mineral Geochem 69:241-271
Rutherford MJ, Devine JD (1988) The May 18, 1980 eruption of Mount St. Helens 3. Stability and chemistry of
amphibole in the magma chamber. J Geophys Res 93: 11949-11959
Rutherford MJ, Devine JD (2003) Magmatic conditions and magma ascent as indicated by hornblende phase
equilibria and reactions in the 1995-2002 Soufriere Hills magma. J Petrol 44:1433-1454
Rutherford MJ, Devine JD (2008) Magmatic Conditions and Processes in the Storage zone of the 2004–2006
Mount St. Helens Dacite. In: Sherrod DR, Scott WE, Stauffer PH (eds) A volcano rekindled: the renewed
eruption of Mount St. Helens, 2004–2006. US Geol Surv Prof Pap 1750 (in press)
Rutherford MJ, Hill PM (1993) Magma ascent rates from amphibole breakdown: An experimental study applied
to the 1980-1986 Mount St. Helens eruption. J Geophys Res 98:19667-19686
Rutherford MJ, Sigurdsson H, Carey S, Davis A (1985) The May 18, 1980 eruption of Mount St. Helens 1. Melt
composition and experimental phase equilibria. J Geophys Res 90: 2929-2947
Sato H, Nakada S, Fujii T, Nakamura M, Suzuki-Kamata K (1999) Groundmass pargasite in the 1991–1995
dacite of Unzen volcano: phase stability experiments and volcanological implications. J Volc Geotherm
Res 89:197-212
Scaillet B, Pichavant M, Roux J (1995) Experimental crystallization of leucogranite magmas. J Petrol 36:663-
705
Scaillet B, Evans BW (1999) The 15 June 1991 eruption of Mount Pinatubo. I. phase equilibria and pre-eruption
P-T- fO2-fH2O conditions of the dacite magma. J Petrol 40:381-411
Scandone R, Cashman KV, Malone SD (2007) Magma supply, magma ascent and the style of volcanic eruptions.
Earth Planet Sci Lett 253:513-529
Schairer JF (1950) The alkali feldspar join in the system NaAlSiO4-KAlSiO4-SiO2. J Geol 58: 512-517
Schmidt MW, Thompson AB (1996) Epidote in calc-alkaline magmas: An experimental study of stability, phase
relationships, and the role of epidote in magmatic evolution. Am Mineral 81:462-474
Schmitt AK (2001) Gas-saturated crystallization and degassing in large-volume, crystal-rich dacitic magmas
from the Altiplano-Puna, northern Chile. J Geophys Res 106:30,561-30,578
Schwartz AJ, Kumar M, Adams BL (2000) Electron Backscatter Diffraction in Materials Science. Kluwer
Academic/Plenum Publishers, New York
Reconstruction of Magmatic System Variables & Processes 239
Sen C, Dunn T (1994) Dehydration melting of a basaltic composition amphibolite at 1.5 and 2.0 GPa: implications
for the origin of adakites. Contrib Mineral Petrol 117:394-409
Shimizu N, Hart SR (1982) Applications of the ion-microprobe to geochemistry and cosmochemistry. Ann Rev
Earth Planet Sci 10:483-526
Shimizu N (1997) Principles of SIMS and modern ion microprobes. In Modern Analytical Geochemistry, Gill
J (ed) Longman, p 235-242
Shin H, Lindquist WB, Sahagian DL, Song S-R (2005) Analysis of the vesicular structure of basalts. Comput
Geosci 31:473-487
Silver LA, Ihinger PD, Stolper E (1990) The influence of bulk composition on the speciation of water in silicate
glasses. Contrib Mineral Petrol 104:142-162
Singh J, Johannes W (1996) Dehydration melting of tonalites 2. Composition of melts and solids. Contrib
Mineral Petrol 125:26-44
Sisson TW, Ratajeski K, Hankins WB, Glazner AF (2005) Voluminous granitic magmas from common basaltic
sources. Contrib Mineral Petrol 148:635-661
Song S-R, Jones KW, Lindquist WB, Dowd BA, Sahagian DL (2001) Synchotron x-ray computed
microtomography: studies on vesiculated basaltic rocks. Bull Volcanol 63:252-263
Spilliaert N, Allard P, Metrich N, Sobolev AV (2006) Melt inclusion record of the conditions of ascent, degassing,
and extrusion of volatile-rich alkali basalt during the powerful 2002 flank eruption of Mount Etna (Italy).
J Geophys Res 111 doi:10.1029/2005JB003934
Stewart M, Pearce TH (2004) Sieve-textured plagioclase in dacitic magma: Interference imaging results. Am
Mineral 89:348-351
Stormer JC (1983) The effects of recalculation on estimates of temperature and oxygen fugacity from analyzes
of multicomponent iron-titanium oxides. Am Mineral 66:586-594
Swanson DA, Holcomb RT (1990) Regularities in growth of the Mount St. Helens dacite dome, 1980–1986.
In Lava flows and domes, emplacement mechanisms and hazard implications. IAVCEI Proceedings in
Volcanology. Fink JH (ed) Springer-Verlag, Berlin 2:3–24
Tamic N, Behrens H, Holtz F (2001) The solubility of H O and CO in rhyolitic melts in equilibrium with a mixed
CO –H O fluid phase. Chem Geol 174:333–347
Thomas R (2000) Determination of water contents of granite melt inclusions by confocal laser Raman microprobe
spectroscopy. Am Mineral 85:868-872
Thomas R, Davidson P (2007) Progress in the determination of water in glasses and melt inclusions with Raman
spectroscopy: A short review. Acta Petrologica Sinica 23:15-20
Toramaru A, Noguchi S, Oyoshihara S, Tsune A (2008) MND (microlite number density)-water exsolution rate
meter. J Volc Geotherm Res doi:10.1016/j.jvolgeores.2008.03.035
Tuttle OF, Bowen NL (1958) Origin of granite in the light of experimental studies in the system NaAlSi3O8-
KAlSi3O8-SiO2-H2O. Geol Soc Am Mem 74: 153 pp
Underwood E (1970) quantitative Stereology. Addison-Wesley, Massachusetts, 274 pp
Van der Plas L, Tobi AC (1965) A chart for judging reliability of point-counting results. Am J Sci 263:87-89
Venezky DY, Rutherford MJ (1997). Preeruption conditions and timing of dacite-andesite magma mixing in the
2.2 ka eruption at Mount Rainier. J Geophys Res 109:20069-20086
Venezky DY, Rutherford MJ (1999) Petrology and Fe-Ti oxide reequilibration of the 1991 Mount Unzen mixed
magma. J Volcanol Geotherm Res 89:213-230
Wallace PJ (2001) Volcanic SO2 emissions and the abundance and distribution of exsolved gas in magmatic
bodies. J Volc Geotherm Res 108:85-106
Wallace PJ (2005) Volatiles in subduction zone magmas: concentrations and fluxes based on melt inclusion and
volcanic gas data. J Volcanol Geotherm Res 140:217-240
Wallace PJ, Anderson AT Jr, Davis AM (1999) Gradients in H2O, CO2 and exsolved gas in a large-volume silici
magma system: Interpreting the record preserved in melt inclusions from the Bishop Tuff. J Geophys Res
104:20097-20122
Wark D, Watson E (2006) Titaniq: a titanium-in-quartz geothermometer. Contrib Mineral Petrol 152:743-754
Watson EB (1982) Melt infiltration and magma evolution. Geology 10:236-240
Wells PRA (1977) Pyroxene thermometry in simple and complex systems. Contrib Mineral Petrol 62:129-139
Williams DL, Abrams G, Finn C, Dzurisin D, Johnson DJ, Denlinger R (1987) Evidence from gravity data for
an intrusive complex beneath Mount St. Helens. J Geophys Res 92:10207-10222
Wright HMN, Roberts J, Cashman KV (2006) Permeability of anisotropic tube pumice – model calculations and
measurements Geophys Res Lett 33, doi:10.1029/2006GL027224
zellmer GF, Sparks RSJ, Hawkesworth CJ, Wiedenbeck M (2003) Magma Emplacement and Remobilization
Timescales Beneath Montserrat: Insights from Sr and Ba zonation in Plagioclase Phenocrysts. J Petrol
44:1413-1431
Reviews in Mineralogy & Geochemistry
Vol. 69, pp. 241-271, 2008 7
Copyright © Mineralogical Society of America
Magma ascent rates
Malcolm J. rutherford
Department of Geological Sciences
Brown University
Providence, Rhode Island, 02912, U.S.A.
[email protected]
IntrOduCtIOn
Volcanoes can erupt explosively, generating high columns of ash and occasional pyroclastic
flows, or they can erupt slowly forming lava flows and domes. This variation reflects significant
differences in mass eruption rate (e.g., Wilson et al. 1980). Mass eruption rates in turn are
controlled by the rates of magma ascent through the volcanic conduit, and the conduit size. The
magma ascent rate itself is a function of the pressure in the magma storage region, the physical
properties of the magma, such as its density, viscosity and crystallinity, and the resistance to flow
in the conduit that connects the magma storage zone to the surface (Papale and Dobran 1994;
Mastin and Ghiorso 2001; Pinkerton et al. 2002; Sparks et al. 2006). A number of important
characteristics of volcanic eruptions are affected or controlled by the rate at which magma
ascends from depth. For example, the bubble content (i.e., vesicularity) and the degree of
crystallization that develop in the melt phase can be significantly different in rapidly vs. slowly
ascended magma. In fact, the inability of rapidly ascending magma to effectively lose exsolved
gas may be one factor causing an eruption to change from effusive to explosive behavior, as
recently documented for the Soufriere Hills eruption on Montserrat in the West Indies (Sparks
et al. 1998). Therefore, in order to better understand the processes involved and the changes
that occur in volcanic eruptions, it is important to quantify the rates at which different magmas
rise to the surface.
The fact that the rate of magma ascent controls a number of reactions that occur in
volatile-rich magmas suggests several ways to study magma ascent rates using reaction
data and theoretical flow models. One reaction that is a direct result of magma ascent is the
exsolution of volatiles from the melt to form gas bubbles (Sparks 1978). Water is the most
abundant bubble-forming gas in intermediate and silica-rich composition magmas, but CO2,
SO2, and Cl are also very important in some magma types. Available data suggest that these
gases diffuse out of the melt at different rates as pressure exerted on an ascending magma
decreases during ascent (Watson 1994; Baker and Rutherford 1996: zhang and Behrens 2000).
It is therefore theoretically possible to use the extent of degassing of each of these different
species as an indicator of the time involved in magma ascent from depth. Degassing during
ascent also causes the melt portion of the magma to crystallize. The rate and extent of this
crystallization have been determined for some magma types and can be used to estimate magma
ascent rates. Similarly, some crystals formed in magmas at depth can contain volatile species
such as OH, CO2 and S in their structure, e.g., biotite, hornblende, and Fe-sulfide, and these
phenocrysts tend to become unstable and react with the surrounding melt during magma ascent.
These reactions occur because the decrease in pressure during ascent causes a decrease in the
volatile species dissolved in the melt that destabilizes the phenocryst. The amount of reaction
between phenocryst and melt is controlled by the ascent rate and the kinetics of the mineral
breakdown. These degassing-related reactions are particularly important in studying volatile-
1529-6466/08/0069-0007$05.00 DOI: 10.2138/rmg.2008.69.7
242 Rutherford
rich, subduction zone volcanoes. These reactions are less useful in studies of basaltic volcanic
centers because of the generally lower volatile contents in these magmas, but examples have
been described where such reactions occur (Vaniman et al. 1982; Corsaro and Pompillio 2004).
Magma ascent rates can also be estimated using the rate of movement of seismic activity toward
the surface (Endo and Murray 1992) and using theoretical analyses of conduit flow together
with measurements of the mass eruption rate for a given event (Carey and Sigurdsson 1985).
Other methods of assessing magma ascent rates involve study of xenolith transport from
depth by the magma, the extent of chemical reaction between these xenoliths and their host liquid
in the magma, and evidence of U-Th-Ra isotopic disequilibrium in a magma (Condomines et
al. 1988). The latter approach involves the identification of isotopic disequilibrium of the short-
lived isotopes in the U and Th decay series; the disequilibrium indicates a maximum time of
magma existence (4-5 half lives for the decay involved, i.e.; ~8000 years for 230Th-226Ra and
30 years for 232Th-238Ra) since the disequilibrium was generated. (This topic will be treated in
other chapters in this volume, and therefore will not be considered further here.) The ability of
magma to physically transport xenoliths to the surface depends on the ascent velocity and the
rheology of the magma as well as the xenolith size and density (Spera 1984). Reactions between
xenoliths and magma may occur where a xenocryst or xenolith and magma are in contact but
not in equilibrium. In such cases, the crystal undergoes a time dependent diffusion exchange
of atoms with the melt, allowing an assessment of the time involved in magma ascent from the
point (depth) of crystal-melt contact which must be determined (Klugel 1998). The following
sections describe some of the interesting results on magma ascent rate determinations using
these different approaches, but first we consider the problem of the magma source, and where
ascent begins.
MaGMa SOurCe and MaGMa reSerVOIrS
One of the significant problems that must be addressed when attempting to determine
magma ascent rates for volcanic eruptions is the question of where ascent was initiated. For
many volcanoes there is good evidence for a magma storage region in the crust, a magma
reservoir from which the final ascent began. Magma storage zones in the upper crust have been
identified seismically for almost all well studied (recently active) volcanoes in subduction-zone
environments, including Mount St. Helens (Scandone and Malone 1985), Mount Pinatubo in
the Philippines (Pallister et al. 1996), Soufriere Hills on Montserrat (Aspinall et al. 1998),
Unzen in Japan, and Redoubt (Power et al. 1994) in the Aleutians. The presence of these near
surface (5 to 15 km depth) reservoirs is confirmed by petrological data (Fig. 1) that indicates the
phenocryst assemblage equilibrated in the reservoir (Hammer and Rutherford 2003) if they did
not form here initially (Rutherford and Devine 2008). Density contrasts between magma and
surrounding rocks (Ryan 1993) also indicate that there are likely both deep and shallow magma
reservoirs beneath many calc-alkaline volcanoes, and petrologic data indicates that two such
reservoirs did exist at 3-5 and >8 km under Mount Rainier at the time of its most recent (2,200
years ago) explosive eruption (Venezky and Rutherford 1997). Using the methods outlined
above, estimates of magma ascent rates for such volcanoes will generally refer to the transport
from the higher, crustal-level storage chamber to the surface.
Eruption from a crustal reservoir has also generally been the case for basaltic volcanic
centers such as Hawaii, Iceland, Etna, and the Canary Islands (Decker 1987; Klugel et al. 1997;
Corsaro and Pompilio 2004). For eruptions of basaltic magma at mid-ocean ridges, available
seismic evidence generally indicates a magma storage region in the crust or near the crust-
mantle boundary. However, some magma, specifically kimberlite and more mafic magma
erupting in mid-plate oceanic islands and in some island arcs (Carmichael 2002), appears to
ascend directly from the mantle without encountering a crustal storage region. In such cases,
Magma Ascent Rates 243
west east
0
A
-5
depth (km)
155 ± 10 MPa II previously designated
unerupted magma
220 ± 50 MPa I
-10
boundary of
magma storage
zone
-15
0 10 20
distance (km)
H2O-saturated
dacite
porphyry 90 km3
pluton
mixed volatile
phase
3
11 km
basalt
3
1.4 km
26 May - 8 June
Figure 1. A. Vertical E-W cross-section through the Mount Pinatubo volcanic system. Black dots are
earthquake foci mostly outside the magma volume. I and II show petrological constraints on phenocryst-
melt equilibration depths. B. Sketch cross sections illustrating interaction between basalt and pre-existing
dacite at Mount Pinatubo. [Used with permission of the Geological Society of America from Hammer and
Rutherford (2003) Geology, Vol. 31, Fig. 3, p. 81.]
the magma ascent rate would refer to the time between separation of magma from its source
in the Earth’s mantle and its arrival at the surface. However, most of the magma ascent rate
estimates are for volcanic centers where there is strong evidence of a crustal-level magma
storage zone, and the ascent rate is for magma transport from this reservoir to the surface.
BuBBLe FOrMatIOn and eVOLutIOn durInG MaGMa aSCent
As magma rises from deep in the crust, volatiles dissolved in the melt will eventually
tend to exsolve and form bubbles because of their lower solubility in melt at lower pressures.
The depth at which the gas phase is generated depends on the CO2 and H2O dissolved in the
magma, and the solubility of these species in the melt (e.g., Holloway and Blank 1994). Based
244 Rutherford
on the abundances of CO2 and H2O determined for melt inclusions trapped in early formed
phenocrysts (Fig. 2), the first gas phase generated in basalts and andesites is likely to be CO2-
rich (Roggensack et al. 1997; Metrich et al. 1999). The same is likely true for more evolved
magmas as well, but they tend to also contain higher (3-7 wt%) dissolved H2O (Cervantes and
Wallace 2003; Mangiacapra et al. 2008). What is the rate of this gas exsolution process, and
can measurements of the gas released be used to estimate the rates of magma ascent? Analyses
of matrix glasses in both explosively erupted and extrusive magmas illustrate that most of the
water originally in the melt at depth has been lost to the gas phase during ascent (Rutherford
et al. 1985). In other words, even for the relatively rapid ascent involved in explosive eruptions
of a volatile-rich magma, there appears to have been sufficient time during ascent for most of
the original H2O to escape from the melt. Other volatile species that have an affinity for the gas
phase include CO2, SO2, and Cl, based on their common occurrence in volcanic gas emissions.
The available diffusion data for silicate melts (Watson 1994) suggest that these species diffuse
at somewhat slower rates than water. Thus, analytical data on the abundance of the different
volatile species lost from the melt to the gas phase during an eruption may record the rate of
magma ascent, although the rapid diffusion of H2O may mean that it is nearly all transferred
to the gas phase in most eruptions.
Figure 2. Plot of dissolved CO2 vs. dissolved H2O for melt inclusions in K-trachybasalt and latite magmas
erupted at Campi Flegrei volcanic center, Italy. Paths 1, 2 and 3 show two possible decompression paths
relating high- (200 MPa) and low-pressure inclusions. Arrows indicate how crystallization or CO2 gas
fluxing would move H2O content at close to constant pressure. (Modified from Mangiocapra et al. 2008;
Isobars are calculated from Papale et al. 2006.)
Experiments indicate that the loss of water from melt into gas bubbles during decompression
(ascent) of a rhyolitic melt is generally very rapid (Hurwitz and Navon 1994; Gardner et al.
1999; Mangan and Sisson 2000; Gardner and Denis 2004; Baker et al. 2006; Gardner 2007).
At pressure drops of 0.025 MPa/s (1 bar per 4 seconds) water is able diffuse out of the melt
and into the bubbles rapidly enough to maintain equilibrium. At faster decompression rates of
0.25 and 1.0 MPa/s, water diffused too slowly to maintain equilibrium and the melts became
supersaturated with water. These experiments indicate that a transition occurs between 0.025
and 0.25 MPa/s decompression rates, beyond which gas bubble-melt equilibria cannot be
maintained. Given typical rhyolitic magma densities, these decompression rates equal ascent
Magma Ascent Rates 245
velocities of 0.7 and 5 m/s, rates much faster than are typically inferred for effusive lava
and dome eruptions. In the high-silica rhyolite (77 wt% SiO2) decompression experiments
referenced above, the 825 °C temperature produces a magma that is relatively viscous, so
water diffusion is relatively slow. In less viscous basaltic magmas, water diffusion is faster
at the higher (~1200 °C) temperatures (zhang and Stolper 1991). Numerical modeling of
bubble growth in basaltic magmas suggests that for reasonable ascent rates, water diffusion
and bubble growth are rapid enough to always maintain equilibrium between bubbles and melt
(Toramaru 1995). This includes ascent rates up to 100 m/s.
The above discussion focuses on degassing of water, which is generally the fastest diffusing
volatile species in melts. In contrast, the diffusivity of CO2, S and Cl in rhyolitic melts is 10’s
to100’s of times slower than water at a given temperature (Watson 1994: Baker and Rutherford
1996; Alletti et al. 2007). If magma ascends during explosive eruptions at rates controlled by
water exsolution (the maximum at which water and melt can maintain equilibrium), then it is
expected that slower diffusing gas species would not maintain equilibrium. Indeed, this appears
to be the case, as matrix glasses in volcanic pumices typically contain much if not all of the
pre-erupted contents of S and Cl. For example, the pre-eruptive concentrations of H2O and
Cl in the 1991 Pinatubo magma are about 6.4 wt% and 1200 ppm, respectively (Pallister et
al. 1996), whereas their final concentrations dissolved in matrix glasses are about 0.4 wt%
and 880 ppm. Degassing of Cl and S is complicated, however; since the partitioning of Cl
and S between the melt and a water-rich gas phase is likely to vary strongly as a function of
water content, pressure, melt composition, and temperature. Thus, in intermediate and mafic
composition magmas, where there is significant dissolved CO2 in the melt at depth, the initial
gas phase will be CO2-rich and only at much lower pressures (Fig. 2) will significant H2O
partition into the gas phase (Dixon and Stolper 1995; Wallace 2005). There is some preliminary
data on how S, Cl and F partition into the CO2-rich vs. the H2O-rich gas phase (Nicholis 2006),
but this is a relatively unstudied research area where additional work may be very productive
in understanding magma ascent during volcanic eruptions. An important implication of the
above findings is that water is able to rapidly diffuse out of silicate melts into bubbles during
magma ascent such that the melt-bubble equilibrium is almost always continuously maintained
(Gardner et al. 1999), and this is important because several of the reaction processes described
below are driven by water degassing of the interstitial melt during magma ascent. Hence, the
rates at which those reactions occur are not constrained by the exsolution of water from the
melt, but by their own reaction kinetics.
There is one group of volatile studies where the diffusion rates of volatile species such as
H2O and CO2 have been used with some success to determine the ascent rate of magmas during
volcanic eruptions. Anderson (1991) initially illustrated the potential of this approach using
gas bubbles in hourglass melt inclusions in quartz phenocrysts in the Bishop Tuff rhyolite.
He showed that the bubble-melt volume ratios suggest a decompression time of ~2 minutes
during the plinian phase of the eruption, i.e., an average ascent rate of 40 m/s for ascent from
180 MPa to 40 MPa where the magma was fragmented. Liu et al. (2007) refined this model by
measuring the concentrations of both H2O and CO2 in quartz-hosted, bubble-free cylindrical
melt inclusions that are connected to the surrounding bubble-rich melt. During magma ascent,
bubbles form and grow outside the cylindrical inclusion creating a concentration gradient in
the cylinder. Liu et al. (2007) used CO2 and H2O diffusion data to model the diffusion profiles
that would be produced for different starting conditions and different length inclusions as a
function of decompression rate. Fitting the measured melt diffusion profiles to the theoretical
data suggests that average magma ascent rates were 0.05-0.35 m/s for the large phreato-
magmatic Oruanui eruption of rhyolitic magma in New zealand. This estimate is somewhat
lower than the average velocity (~5-8 m/s) that is predicted for this magma by theoretical
models (Papale et al. 1998), but the approach is promising because it makes use of both CO2 and
H2O diffusive loss during decompression. Analyses of more complete profiles along a tubular
246 Rutherford
inclusion may be possible in the future using an instrument like the NANOSIMS (Saal et al.
2008). Humphreys et al. (2008) used H2O diffusion profiles in tubular glassy melt inclusions in
plagioclase phenocrysts to estimate the decompression time (and average velocity) of the May
18 Mount St Helens plinian eruption. They obtained an average velocity of 37-64 m/s for the
ascent of this magma from 7 km to the fragmentation depth (~500 m), significantly higher than
estimates from mass eruption constraints (Scandone and Malone 1985; ~3 m/s, Geschwind and
Rutherford 1995), and from numerical models (~4 m/s, Carey and Sigurdsson 1985; ~15 m/s,
Papale et al. 1998). There are several possible explanations discussed for the high estimate
from H2O diffusion data, including the fact that at these ascent rates the decompression is too
fast for H2O diffusive equilibrium to be maintained; the lack of equilibrium would tend to
give high ascent rate estimates as would the presence of CO2 in the gas phase. Of course, the
estimates from theoretical models may too low; the models continue to be revised as new data
on critical parameters such as volatile solubility and melt viscosity becomes available.
GrOundMaSS CryStaLLIZatIOn durInG MaGMa aSCent
The degassing of the melt in ascending water-rich magma discussed in the previous
section causes gas bubbles to form and/or grow, and also tends to cause the interstitial melt
phase to crystallize, ascent time permitting. This crystallization occurs because any near
adiabatic ascent path approaches and crosses the solidus for such magmas (Fig. 3). The process
has been documented by the descriptions of numerous volcanic complexes (e.g., Mount
St. Helens, etc.) where rapidly ascended magmas (explosive eruptions) erupt with a glassy
groundmass but effusive eruptions of the same magma have experienced extensive groundmass
3000
An70
An60
2500
32 8
2000 An50
PH2O
An42 Hbd Plag
+Ilm Opx
1500 An38 Cpx
+ Mte
51
1000
Magma
Ascent
500 Path
Phase Equilibria of 2005-06
Mount St. Helens Dacite
Xals
fO2 = NNO + 1 log units
0
840 860 880 900 920 940
O
T C
Figure 3. Phase equilibria of a typical intermediate composition, H2O-rich magma showing the onset of
crystallization of the major phenocryst phases plagioclse (Plag), orthopyroxene (Opx), Ca-pyroxene (Cpx),
Amphibole (Amph), magnetite (Mte) and ilmenite (Ilm). The Mount St. Helens 2004-06 dacite ascended
from a depth of approximately 4 km (120 MPa) where amphibole was stable and phenocryst content was
~50 vol%) at 860 °C in the final ascent. Numbers in boxes show the magma crystallinity at the P and T
of the box. Curve labeled xals is the approximate solidus for the dacite. (Modified after Rutherford and
Devine 2008).
Magma Ascent Rates 247
crystallization (Cashman 1992; Hammer et al. 1999; Couch et al. 2003). The process has also
been demonstrated experimentally beginning as far back as the Tuttle and Bowen (1958) work
on granite melts; these phase equilibrium studies show that melts with more than ~2 wt%
dissolved water have anhydrous phenocryst-in curves that have a negative slope on a P-T
diagram (Eggler and Burnham 1973). Although driven by the loss of volatiles from the melt with
decreasing pressure, as discussed in the previous section, the crystallization of the groundmass
melt does not necessarily occur at the same rate. For example, while the exsolution of water
from a decompressing melt follows equilibrium solubility for decompression rates of < 0.25
MPa/s (magma ascent rates less than 5 m/s), a rhyolitic glass crystallizes much more slowly
(Cashman 1988). The rate of crystallization will also differ for different temperatures and melt
compositions, being slower for more silica-rich melts at a given water concentration.
A study of groundmass glass crystallization in the eruption products of the 1980-86
Mount St. Helens eruption products (Geschwind and Rutherford 1995) illustrated what can be
done using the extent of groundmass crystallization to estimate magma ascent rates. Earlier
work (Rutherford and Devine 1988) established that all of the erupted magmas came from a
storage region at 8-11 km depth, and also established that the depth, temperature (880 °C)
and dissolved water content (4.6 wt%) of the melt in that magma storage zone did not change
over the time of these eruptions. In addition, the similarity of phenocryst and groundmass
mineral compositions (Fe-Ti oxides) in these samples suggests that there was essentially no
change in the temperature of any of these magmas during ascent. Geschwind and Rutherford
(1995) concluded that differences observed in the groundmass crystallinity of these samples
are the result of different ascent rates. Data on the extent of groundmass crystallization in
natural Mount St. Helens samples were compared with the results of experiments on May
18 dacite samples decompressed from 220 to 2 MPa over different times, i.e., at different
decompression rates (Fig. 4). These experiments demonstrate that nucleation and growth of
microlites in the melt phase is detectable after a one-day constant-rate decompression from 220
to 2 MPa, and changes continuously for decompressions taking up to 8 days, as measured by
Figure 4. Plot of melt composition as a function of time in an H2O-saturated dacite simulating adiabatic
magma ascending from 220 MPa (8 km) to the surface at 900 °C. The compositions produced are compared
with those in the natural dacite on the right. Ascent times of 100-200 hours are required to reproduce
the natural groundmass glass of the post-May 18 Mount St. Helens lava dome samples. (Modified after
Geschwind and Rutherford 1995.)
248 Rutherford
the change in melt (glass) composition. Using the composition of the remaining matrix glass in
the experimental and natural samples, average magma ascent rates have been obtained which
range from >0.2 m/s for the 1980 explosive plinian event to 0.01 to 0.02 m/s for the different
dome forming eruptions of 1980-86 (Fig. 4). Nucleation is unlikely to have been a factor
affecting groundmass crystallization in either the natural system or the experiments because
the Mount St. Helens magma was crystal-rich prior to ascent. A similar study of a crystal-poor
magma might show that growth of microlites from the groundmass melt takes much longer
when phenocrysts are not present to serve as nucleation sites. Westrich et al. (1988) describe
eruptions where the same phenocryst-poor, high-silica rhyolite erupted explosively and then
effusively without groundmass crystallization in either case, even though the effusive magma
must have lost the same amount of water over a much longer time frame.
Several recent studies have shown the difficulty in using groundmass crystallization to
estimate magma ascent rates. Hammer and Rutherford (2002) demonstrated experimentally,
that the crystal nucleation and growth rates in an ascending water-rich magma go through
maxima at large and small degrees of under-cooling respectively. Thus, both the texture and the
approach to equilibrium during crystallization depend on the ascent rate and also on whether
the rate is steady or whether there are pauses in the rate of ascent. Although an adiabatic ascent
is often assumed (e.g., Geschwind and Rutherford 1995), it is generally recognized that the
ascending magma could experience either heating or cooling. Cooling could occur by loss of
heat to the rocks surrounding the conduit or by gas bubble formation and expansion; heating
would tend to occur from the release of the latent heat of crystallization as it occurred. Blundy
et al. (2006) used plagioclase-melt geothermometry and pressures based on dissolved water
in melt inclusions to identify an increase in temperature for final M.I. entrapment conditions
during the 1980-82 Mount St. Helens eruptions (Fig. 5). The temperature increase, from 880 to
as high as 940 °C for the post-May 18 eruptions, appears to confirm that release of latent heat
of crystallization can raise the temperature of ascending magmas. A correlation of increased
250
May 18,1980 (Plin)
May 25, 1980
June 1980
July 1980
200 August 1980
October 1980
December 1980
Pressure (MPa)
150
Amph
100
50
0
820 840 860 880 900 920 940 960
o
Temperature ( C)
Figure 5. Plot of pressure of sample equilibration from volatiles in plagioclase melt inclusions vs. temperature
from plagioclase-melt geothermometry for 1980 Mount St. Helens samples. Trend of the data shows an
increase in temperature of more slowly ascended (and more crystalline) samples interpreted as the effect of
latent heat build-up. (Modified after Blundy et al. 2006 to exclude May 18, 1980 and 1982 data).
Magma Ascent Rates 249
crystallinity with increased M.I. sealing temperature was also determined in these same samples
(Blundy et al. 2006). However, these eruptions occurred in a time period where the Mount
St Helens eruption was alternating from periods of conduit-emptying explosions to effusive
dome building (Swanson et al. 1987); thus the ascent path was definitely not constant-rate
decompression, and some parts of each eruption may have spent days to weeks at a constant
pressure. Furthermore, each hand sample from these eruptions contains amphiboles with a
range of rim thicknesses that have been interpreted to indicate the samples are mixtures of
magma newly ascended from depth with samples of the same magma that spent longer periods
of time along the margins of the active conduit (Rutherford and Hill 1993). If the magma is a
mixture, the determination of ascent rates from crystallinity arguments becomes very difficult
unless the ascent path of the different magma components can be characterized.
The McCanta et al. (2007) study of the Black Butte dome near Mount Shasta (CA) indicates
that this eruption involved an intermediate-composition, water-rich magma and it contains no
evidence of magma mixing or interruptions in the magma ascent. It was determined experimen-
tally that the unzoned plagioclase (Fig. 6), amphibole and titaniferous-magnetite phenocryst
cores formed at a uniform set of conditions (P = 350 MPa, T = 890 °C). The melt interstitial to
these phenocrysts was almost completely crystallized in the ascent of this magma to the surface
as indicated by the flow alignment of the microphenocryts and microlites in the rock. Two pla-
gioclase crystallization processes were used to assess the time involved in the magma ascent
to the surface, and for comparison with the observed amphibole-melt reaction (next section).
Normally zoned overgrowths (17±3 µm thick) on the uniform composition (An77) cores range
from the core composition to An40, and represent essentially continuous crystallization during
the magma ascent (assumed to be adiabatic). The plagioclase microphenocrysts (and microlites
that make up much of the groundmass) have cores ≤ An65, and represent a parallel crystalliza-
tion process. According to the P-T ascent path the An60-65 plagioclase microphenocryst cores
would have become stable at ~150 MPa. Crystal size distribution (CSD) analysis of the plagio-
clase crystal population (Fig. 7) verify that there were two separate plagioclase growth events;
the phenocryst growth in the magma storage zone at ~300 MPa, and the microphenocrysts and
phenocrysts that grew during the ascent. Using an ascent rate based on amphibole rim thick-
nesses, a growth rate for the plagioclase phenocryst rims and microphenocrysts (G = 2.5-8.7 ×
10-8 mm/s) was obtained that is in the range ((1 × 10−7 to 2 × 10−11) of previous determinations
for experimental (Couch et al. 2003) and natural (Cashman 1992) samples; this agreement sup-
ports the model proposed for the Black Butte magma ascent.
A A B B
Figure 6. Photomicrographs of plagioclase phenocrysts with relatively homogenous cores and thin (15-20
µm), normally zone rims in a groundmass rich in plagioclase microphenocrysts (50-150 µm) and microlites
(<50 µm). Phenocryst cores are unzoned; thin (~20 µm) rims are normally zoned. (Modified after McCanta
et al. 2007.)
250 Rutherford
Figure 7. CSD plots of plagioclase for representative samples of the Black Butte (CA) dacite. Two discrete
populations are present as indicated by the two line segments required to fit the data, one for phenocrysts
(>150 µm) and microphenocrysts (50-150 µm) and one for the microlites (<50 µm). [Used with permission
of Elsevier from McCanta et al. (2007), J Volc Geotherm Res, Vol. 160, Fig. 9, p. 276. http://www.
sciencedirect.com/science/journal/03770273].
PhenOCrySt-MeLt reaCtIOnS durInG MaGMa aSCent
reactions between volatile-bearing minerals and melt during magma ascent
Another reaction that is caused by the loss of volatiles from the melt as magma rises
toward the surface is the breakdown of volatile-bearing phenocrysts, such as amphibole and
biotite. Sulfur is also partitioned from the melt to the gas phase during magma ascent, and thus
anhydrite and FeS, which occur in some volcanic rocks, would also tend to breakdown. Most of
these reactions will be controlled by the kinetics of the reaction between the phenocrysts and the
volatile-depleted melt. The rates of the phenocryst-melt reactions produced by decompression
have generally not been determined because they are complex functions of many variables,
including mineral and glass compositions, melt viscosity, and temperature. However, the rates
of reaction between amphibole phenocrysts and the coexisting rhyolitic melt have been studied
(Rutherford and Hill 1993; Browne and Gardner 2006), and these calibrations can be used to
estimate ascent rates for dacite and andesite composition magmas.
The rims produced by the amphibole-melt reaction have been widely observed in
volcanic rocks, particularly andesite to rhyolite composition effusive eruptions (MacGregor
1938; Garcia and Jacobson 1979). Textures suggest that the rim thickness can be accurately
measured at least where they occur in more viscous dacite and andesite magmas. The textures
of these reaction rims (Fig. 8) also indicate the reaction is between the amphibole and the
melt and will begin when the melt no longer contains sufficient dissolved water to stabilize
the amphibole (Rutherford and Hill 1993). Depending on the temperature of the ascending
magma, the breakdown tends to begin when the pressure decreases below about 100 MPa
(Fig. 3). The reaction of the amphibole with the water-deficient melt produces a fine-grained
assemblage of Ca-pyroxene, low Ca-pyroxene, plagioclase and magnetite (Rutherford and Hill
1993). Buckley et al. (2006) re-examined this reaction, analyzing amphibole rims in the recent
Soufriere Hills, Montserrat andesite; they emphasized that the reaction rim must be an open
system, losing alkalis and H2O to the surrounding groundmass.
Mount St. Helens sample studies show no detectable reaction between the amphibole
phenocrysts and the groundmass glass (melt) in the May 18, 1980, pumice samples, even
though the melt clearly lost all but about 0.3 wt% of the ~ 4.6 wt% dissolved water it contained
before erupting (Rutherford et al. 1985). In almost all subsequently erupted samples,
however, the amphibole phenocrysts have reaction rims at the contact of the crystal with the
groundmass melt. The dome-forming eruption of October 1986 was an exception, containing
only amphiboles without reaction rims. Each of the other post-May 18 samples had a main
Magma Ascent Rates 251
A B
Figure 8. Photomicrograph showing a typical amphibole decompression-induced reaction rim on a 2 mm
phenocryst with a hollow core (A) from dacite lava dome, and (B) a BSE close-up image of the same
phenocryst (rim) showing coarse to fine texture outward from the amphibole. The phases in the rim, from
white to black are magnetite, low- and high-Ca pyroxene, plagioclase, glass and vesicles. (Modified after
McCanta et al. 2007.)
population of amphibole phenocrysts (30 to 100 vol% of the crystals present) with uniform
thickness (0 to 50 µm range) reaction rims, and a second group with rims ranging to greater
thicknesses. These amphiboles are interpreted to represent, respectively, a main batch of
magma that ascended directly from the storage region at 8-11 km beneath the surface, and
magma from previous eruptions left temporarily in the conduit before being mixed into new
magma from depth. An experimental calibration of the amphibole breakdown reaction was
made for a constant-rate ascent of the 1980 Mount St. Helens dacite magma at 900 °C (Fig. 9).
Experiments that simulated ascent from 8 km depth to the surface in less than 4 days showed
no reaction rims on amphibole crystals; 2 µm wide rims were produced in 7-day ascents, and
32 µm rims formed in 20 days. Using this calibration, the magma ascent rates during the lava
50
845C 130-4 MPa MSH 9 0 0 o C
860C,130-4MPa 160-2 MPa
40 830C,130-4MPa
Rim Width (microns)
30
20
x
Anneal at 50 MPa, 8 6 0 o C
10 +
Anneal at 50 MPa, 8 3 0 o C
0 x
+
0 5 10 15 20 25 30
Run Duration (days)
Figure 9. Plot of experimentally determined amphibole reaction-rim widths as a function of a constant rate
ascent duration at temperatures of 830 to 900 °C. Two curves labeled “anneal” show rim development in
experiments run at the indicated conditions using a 160 MPa assemblage as the starting material. (Modified
after Rutherford and Devine 2003.)
252 Rutherford
dome phase of the Mount St. Helens eruption were determined to vary from as low as 0.004
m/s up to 0.015 m/s. This is in contrast with the 2 to 3 m/s ascent rate during the May 18, 1980
plinian phase of the eruption based on the mass eruption rate (Carey and Sigurdsson 1985;
Geschwind and Rutherford 1995).
Rutherford and Devine (2003) estimated magma ascent rates by measuring the width
of amphibole reaction rims for samples collected over time (1995-2002) as Soufriere Hills
volcano on Montserrat erupted a phenocryst-rich siliceous andesite. This dome-forming
lava experienced phenocryst-melt equilibration in a magma storage region at 5-6 km depth
(130 MPa) and 830 °C. Reaction rims on amphibole crystals adjacent to groundmass melt
showed significant changes in samples collected over the next several years. The first magma
erupted contains amphiboles with 120 µm thick rims. The magma erupted two months later, in
February 1996, has only 18 µm-thick rims, and the rims are only 10 µm wide in an April 1996
eruption. Most amphiboles in magma erupted in July through September 1996 contain no
reaction rims. The phenocryst assemblage in these magmas is the same throughout this period,
and thus the rims are not produced by changes in the magma storage region; they formed
during magma ascent. Using an experimental calibration of rim width versus time at 830 °C
(Fig. 9), absolute ascent rates and ascent rate changes can be calculated for the Montserrat
eruption. The magma ascent rate increased from 0.001 to 0.012 m/s during the first 300 days
of this eruption. Interestingly, there was an increase in the mass extrusion rate over this same
time period (Fig. 10). The amphibole rim data show that this increase in extrusion rate was
not produced simply by an increase in the size of the conduit; it required an increased magma
ascent rate. Another significant observation for this eruption is the association of numerous
explosive eruptions from the summit of the dome with periods of high magma ascent rate.
These explosions indicate periods of less efficient separation of gas and magma, a logical
consequence of higher magma ascent rates.
McCanta et al. (2007) saw evidence of a single continuous magma ascent at Black Butte
Ca. based on the texture and composition of plagioclase of the dacite samples (Fig. 6). Looking
Montserrat Magma Ascent Rate and Volume
Nov 1995 Mar 1998 Nov 1999 Sept 2002
0.025 500
Vulcanian Explosions
0.02 400
Ascent rate (m/s)
Volume (10 m )
3
Volume
6
0.015 300
0.01 Ascent rate 200
0.005 100
0 0
0 500 1000 1500 2000 2500
Time (days)
Figure 10. Ascent-rate and erupted volume vs. time for the 1995-2002 andesite eruptions at Soufriere Hills,
Montserrat. Ascent rate is based on amphibole rim-thickness development (Figure 11), which goes to zero
early and late in the time period shown. Note volcanian explosions associated with the high ascent rate
eruptions in 1996-97. (Modified after Rutherford and Devine 2003.)
Magma Ascent Rates 253
at the thickness of breakdown rims on amphiboles (Fig. 8) in the same samples, they found
a peak at 34±10 µm in all four lobes of the dome. The small spread in the rim thicknesses
observed confirms that the Black Butte eruption was a continuous event and there was no
mingling of magma batches with different ascent histories as at Mount St Helens. Using the
measured rim thickness range (24-44 µm) and the Rutherford and Hill (1993) calibration of
rim growth rate, a constant rate magma ascent rate of 0.004-0.006 m/s was determined for the
dacite magma ascent from ~200 MPa or ~8 km to the surface at 890 °C. This ascent rate is
supported by the fact it yields a plagioclase growth rate (G = 2.5-8.7 × 10−8 mm/s; previous
section) that is similar to other estimates of plagioclase growth in similar composition systems
(Cashman 1992; Izbekov et al. 2002).
Several lines of evidence suggest that many intermediate and silicic composition
eruptions involve magma ascent rates that are different in different parts of the conduit and
also over time. The post-May 18 (1980) eruptions at Mount St. Helens, regularly involved a
slow effusion of magma followed by an explosive eruption that rapidly removed significant
amounts of magma that was in the conduit (Pallister et al. 1992). The sequence of explosions
followed by additional slow lava effusion suggests that these eruptions involved magma
batches that ascended at different rates. Similar eruptions occurred at Mount Pinatubo (Hoblitt
et al. 1996), and at Soufriere Hills on Montserrat (Sparks et al. 1998). Browne and Gardner
(2006) suggested that variations in the ascent rate occurred during the 1989-90 eruptions at
Redoubt volcano. These variations are generally attributed to magma degassing which affects
both the buoyancy of a rising magma and the physical properties of the magma as gas is lost
to bubbles and bubbles may separate from the magma. By performing both single step and
multi-step decompressions to a given final pressure, Browne and Gardner (2006) showed that
rim growth on amphiboles in dacite magma is extremely slow just outside the stability field (at
840 °C) and also at very low pressures (Fig. 11). The highest rate of rim growth occurs at 30-40
MPa below the amphibole stability field in the single step decompressions, but the rim forms
preferentially at lower pressures (10-50 MPa) in multi-step (~ constant-rate) decompression
Figure 11. Reaction rim growth plotted against pressure and corresponding depth for single-step (a)
and multi-step (b) decompression experiments on a 1989 Redoubt dacite. Reaction rim growth rate is
equal to the reaction rim width/ total time below the amphibole stability boundary (from phase equilibria
experiments). Relative rim growth is the difference in rim thickness from sequential experiments quenched
at lower and lower pressures. [Published with permission of Elsevier from Browne and Gardner (2006),
Earth Planet Sci Lett, Vol. 246, Fig. 5. p. 170. http://www.sciencedirect.com/science/journal/0012821X].
254 Rutherford
experiments. Clearly, the possibility of variations in ascent rate must be considered when using
reaction rims on amphibole (or any other phenocryst-melt reaction driven by decompression)
to determine magma ascent rates.
The above point about variations in ascent rate is further illustrated by the Nicholis and
Rutherford (2004) study of amphibole rims developed in hawaiite (trachybasalt) that erupted
as tephra and lava flows near Yucca Mountain (Nevada, USA). Amphiboles do not occur in all
eruptions in the area, but when present, there are several phenocrysts in any sample collected,
and many have a thick (~30 µm), fine-grained reaction rim. Nicholis and Rutherford (2004)
determined the conditions where the amphibole would be stable (P > 180 MPa; T ~ 900 °C)
and determined experimentally that no constant-rate decompression would produce a reaction
rim like those in the natural sample. Amphiboles remaining in all but the very fast magma
decompressions were invariably rounded and reduced in modal abundance, suggesting dissolution
rather than crystallization at the rim (see Browne and Gardner 2006). The amphibole breaking
down was replaced by additions to preexisting anhydrous phenocrysts in all experiments that
produced groundmass crystallization comparable to that of the natural samples. It is speculated
that these observations are due to the relatively low viscosity of the H2O-rich trachybasalt
magma. Comparison of plagioclase microlites in experimental samples with those in natural
tephras (CSD analysis) indicated that magma ascent rates were >0.04 m/s for these hawaiites.
The rimmed amphiboles were only produced when the ascending magma was additionally held
for several days at ~ 20 MPa (~800 m depth), and Nicholis and Rutherford (2004) postulated
that the final eruption product was produced by mixing samples of magma stored at the shallow
depth into rapidly ascending magma of the same composition.
The 2004-08 eruption of Mount St. Helens consists of a series of domes or spines composed
of phenocryst-rich dacite almost identical to the magma erupted there in 1980-86, but with
some interesting differences. The recent eruption involves a degassed magma with low volatile
emissions (Gerlach et al. 2008), has an Fe-Ti oxide temperature of 860±15 °C, is phenocryst
rich (~40%), and has erupted with a largely crystalline groundmass (Pallister et al. 2008). The
2004-08 magma contains amphibole, sometimes rounded, with ~3 µm thick reaction rims in the
first-erupted material and 5 µm thick rims in all subsequent eruptions (Thornber et al. 2008),
somewhat less than the 10-20 µm thick rims in the 1980-85 lava domes. The thinner reaction
rims in the recent magma could be the result of the lower temperature of the recent eruptions,
but the amphiboles in magmas erupting in mid 2005 and beyond (when the mass eruption rate
had dropped from ~ 6 to < 2 m3/s (Schilling et al. 2008)) all have a thin outer zone of F-rich
amphibole (Rutherford and Devine 2008). Thin breakdown rims are present on this F- and Si-
rich amphibole. DeJesus and Rutherford (2007) investigated the origin and significance of the
new-growth zone of F-rich amphibole. They suggest the factors involved in its formation are
(1) increases in F and SiO2 activity in the melt produced by crystallization during relatively
slow magma ascent and (2) increased time at temperatures outside the stability field of OH-
rich amphibole. A mechanism such as dissolution or abrasion of pre-existing amphibole is also
required to get new amphibole growth in this magma since it has no modal or normative Ca-
pyroxene. As shown in Figure 12, the F-, Si- and Fe-rich amphibole is stable to pressures as
low as 50 MPa in dacitic magma at 860 °C. A decompression-induced reaction rim produced
on this amphibole during magma ascent would only form after the pressure dropped below 50
MPa. The rim would be much thinner than rims developed on amphiboles without this outer
F-rich layer and ascending at the same rate. Ascent rates of 0.005-0.015 m/s calculated for the
2004-06 Mount St. Helens eruption using the OH amphibole rim data (Thornber et al. 2008)
appear applicable to the early 2005 spine, but give a maximum estimate of magma ascent for
later 2005-06 eruptions. Although the presence of the F-, Fe-, and SiO2-rich rims on amphibole
phenocrysts is clearly important in assessing the significance of amphibole breakdown reactions,
such rims have not yet been described for other eruptions (DeJesus and Rutherford 2007), and
may be unique to the conditions and magma of this particular MSH eruption.
Magma Ascent Rates 255
200
An38 An42 F-Amph
An35
150
Melt + Xals
PH2O
(MPa) Xals
100 (0.10 F) Amph
Magma ascent
50 Xals (1.0 F) Path
Near-Solidus MSH-305
Dacite Phase Equilibria
fO2 = NNO + 1 log units
0
650 700 750 800 850 900 950 1000
o
T ( C)
Figure 12. PH2O-Temperatue diagram showing the stability of F-rich amphibole relative to OH-rich
amphibole from experiments on a 2004-06 Mount St. Helens dacite composition. F-Amph and xals (1.0 F)
show the change in amphibole stability and in the solidus with 1 wt% F in the melt; the light-weight lines
show similar curves for the dacite with low F (1000 ppm) in amphibole and melt (Rutherford and Devine
2008). Adiabatic magma ascent line (with latent heat) shows that the decomposition of a F-rich amphibole
will occur over a very short segment of the magma ascent path relative to the OH-rich phenocryst. (Modified
after DeJesus and Rutherford 2007.)
Phenocryst-melt reactions driven by pre-eruption magma mixing
Evidence of phenocryst reactions that are the result of mingling and mixing of two
contrasting magmas is almost ubiquitous in volcanic eruption products (Eichelberger 1978;
Pallister et al. 1992; Nakada and Fuji 1993). In fact, zoning in phenocrysts such as plagioclase
and amphibole and olivine often record the number of mixing events (Singer et al. 1995;
Rutherford and Devine 2003; Ginibre et al. 2007). In most cases, it is not possible to use
these disequilibria to assess magma ascent because the timing of the last mixing event relative
to the onset of ascent and eruption is unknown, or because the depth of the mixing event is
unknown. However, compositional zoning and disequilibrium in coexisting Ti-magnetite and
ilmenite in magmas has proven to be useful in determining ascent rates because the time scales
of re-equilibration are short (days to weeks). Nakamura (1995) extracted time information
from zoned titanomagnetite in the 1991-95 Mount Unzen magma where a high-temperature
basaltic andesite had mingled with a pre-existing dacite just prior to eruption. Enclaves of the
phenocryst-poor mafic magma occur in the dacite and some of the basaltic andesite was also
mixed into the dacite. Mingling is indicated by small blade-shaped Ca-rich plagioclase and
pargasitic amphibole crystals characteristic of the enclaves in the groundmass of the mixed
magma. A number of reactions began as a result of the magma mixing including embayment
of quartz and biotite phenocrysts, growth of pargasitic amphibole rims on biotite phenocrysts
and Ti-zoning in magnetite (Nakada and Fugii 1993). The thickness of the Ti-rich diffusion
profile at the outer margin of magnetite phenocrysts is typically 20-30 µm for samples erupted
over the 1991-93 time period. Nakamura (1995) measured and modeled the Ti-zoning in
magnetite (Fig. 13) by assuming the temperature of the dacite was 800 °C, and concluded
that the time from mixing to quench at the surface was a few months. Venezky and Rutherford
256 Rutherford
Figure 13. Ti x-ray image of zoned titanomagnetite and adjacent ilmenite phenocysts in the Mount Unzen
1991 dacite (left). Plot on the right shows the Ulvospinel component in the natural magnetite compared
to a diffusion model of rim to core zoning produced at 800 °C. Dt/a2 = 0.1 yields a time of 2-4 months at
800 °C. (Published with permission of the Geological Society of America from Nakamura (1995), Geology,
Vol. 23, Figs. 2, 3, p 808).
(1999) determined that the Mount Unzen mixed magma temperature was more likely 850-
870 °C based on the composition of coexisting magnetite and ilmenite in the phenocryst outer
margins and on coexisting pairs in the reaction rims around biotite. These temperatures were
confirmed by the P-T-XH2O conditions (P = 160 MPa; T1 = 790; T2 = 1030 °C, XH2O,1 = 5 wt%;
XH2O,2 = 3 wt%) determined from phase equilibria for the two end-member magmas involved in
the mixing and their proportions. Using the Freer and Hauptman (1978) data for Ti diffusion in
magnetite, Venezky and Rutherford (1999) determined that the 20-25 µm thick diffusion zone
in the 1991-93 Mount Unzen magnetite phenocrysts would be produced in 11 to 30 days at 900
and 850 °C respectively. If ascent began at the time of mixing, the average ascent from 7 km
to the surface would have been at the rate of 0.003 to 0.007 m/s. If the magma ascent began
somewhat after mixing, the rate obviously would have been faster. The fact that decompression-
induced amphibole reaction rims are not always present suggests that different batches of the
erupted Mount Unzen magma represent a short range of times between mixing and ascent.
Devine et al. (2003) also used Ti zoning in titanomagnetite to estimate the rate of magma
ascent for the andesite erupting at Soufriere Hills, Montserrat in 2001-02. As in the case of
Mount Unzen, the zoning reflects the mixing of hot basaltic andesite into a preexisting storage
zone. In this case the pre-mixing storage zone contained a siliceous andesite at ~830 °C at a
pressure of 130 MPa (Barclay et al. 1998). The temperature of the hotter magma is not well
constrained, but must have been >1050 °C because it contained no phenocrysts yet was able
to crystallize pargasite when quenched against the cooler siliceous andesite. The mixing in of
the mafic magma raised the temperature of the mixed magma to as high as 860 °C according
to the phase equilibria (Rutherford and Devine 2003). If the temperature of the mixed magma
were higher than 860 °C for as little as two days, a characteristic breakdown of the amphibole
phenocrysts would have been produced and is not observed. Based on diffusion calculations
and experiments at 860 °C, Devine et al. (2003) conclude that the Ti-rich rims on magnetite
(Fig. 14) required a mixing and ascent time of ~20 days at this temperature. However, many
batches of the magma that contain these zoned magnetite crystals also contain amphiboles with
no decompression-induced reaction rims (Fig. 10). According to decompression experiments
Magma Ascent Rates 257
14
Titanomagnetite Composition Profiles 960
13
940
12 MVO1229
Apparent T (oC)
920
TiO2 (wt %)
11 Expt M40 (860oC, 22 days)
900
10 880
9 860
8 840
7 820
MVO573
6 800
0 0.2 0.4 0.6 0.8 1
Normalized distance (rim to rim)
Figure 14. TiO2 concentrations in profiles across titanomagnetite crystals in 1995-2002 Soufriere Hills
volcano samples and one experimental sample. MVO573 (Sept. 1997) is from a rapidly ascended magma
sample (no reaction on amphibole phenocrysts) and was used in experiments such as M40. The profile
across the M40 titanomagnetite closely approximates those in natural samples like MVO1229 (Sept.
2001) and indicates the time required after mixing to create the zoning observed. Apparent temperature is
calculated for each Ti-magnetite composition paired with the adjacent unzoned ilmenite. (Modified after
Rutherford and Devine 2003.)
on samples of this material, the lack of rims on the amphibole requires ascent from 130 MPa
(5-6 km) to the surface in < 4 days, a steady state ascent rate of > 0.017 m/s. Rutherford and
Devine (2003) conclude that the < 4 day ascent began as many as 18 days after the heating and
mixing event that affected the erupting magma in the period Sept. 2001-(MVO-1228-9), but at
least some of the mixed magma reaching the surface in Sept. 2002 was heated for no more than
10 days (no Ti-zoning in magnetite) before ascending rapidly (< 4 days) to the surface.
MaGMa aSCent rateS FrOM MaSS eruPtIOn rateS
Measurements of the mass eruption rate of magma can be used to calculate conduit
diameter and ascent velocity, if certain assumptions are made about an eruption (Wilson et al.
1980). This technique is illustrated by studies of the May 18, 1980, plinian Mount St. Helens
eruption (Carey and Sigurdsson 1985; Geschwind and Rutherford 1995). Magma flowing
through a conduit below the depth where it is saturated with a gas phase can be described as a
one-phase, laminar flow of an essentially Newtonian fluid. For this case, magma ascent in the
conduit can be modeled using the Hagen-Poiseuille law
M = (∆P/∆L)·(Π ρFr4)/8η
where M is mass flux (in kg/s), ∆P/∆L is the pressure gradient driving the flow, ρF is the
fluid (magma) density, η is fluid viscosity, and r is conduit radius. Exsolution of a vapor
phase changes the bulk density and viscosity of the magma and thus affects flow conditions,
but exsolution is slow at first and does not significantly affect the flow until the magma is
within a kilometer or two of the surface (Wilson et al. 1980). Calculations are made for depths
greater than 1 km, assuming that the driving force for flow is due to the density difference
between the surrounding rocks and the magma. For quantitative applications of the Hagen-
Poiseuille law, numerical values for the magma mass flux, density, and viscosity of Mount
St. Helens magmas are needed. Using recently revised data for these parameters (Geschwind
and Rutherford 1995), and assuming the conduit radius remains constant at depth, the conduit
258 Rutherford
radius during the explosive 1980 eruption of Mount St. Helens is calculated to be 33 m. For
dome-forming eruptions in the period October 1980 to 1982, the calculated radius decreased
to 7 m and remained essentially constant (Fig. 15).
Ascent velocities for the Mount St. Helens eruptions were calculated from the relationship
expressing conservation of mass, M = Πr2ρFu, where u is the velocity. Given the conduit
radii calculated above, conservation of mass requires that ascent velocity decreased from
2-3 m/s on May 18, 1980, to 0.08 m/s for the October 1980 and 1981 eruptions (Fig. 15).
The former estimate is consistent with vent diameter estimates and the vent velocity is
required to explain plinian ash columns (Carey and Sigurdsson 1985). However, the latter
velocity leads to an ascent time of 30 hours from a depth of 8.5 km which is less than the
approximately 4-8 days indicated by the petrological methods for the post-May 18, dome-
forming eruptions (Rutherford and Hill 1993). The assumptions of constant magma viscosity,
and even constant conduit radius are almost certainly wrong for the dome-forming eruptions
because there were dome-destroying explosions followed by slow effusive dome growth
during this period (Swanson et al. 1987). Excessive crystallization of microlites during slow
ascent periods increases magma viscosity and decreases the magma velocity in the upper
parts of the conduit. This could explain the differences between the calculated velocity and
the velocity derived from hornblende breakdown for the 1980-82 dome lavas, and has also
been modeled as significant for explaining the cycles of effusive dome growth and volcanian
explosions that occurred at Soufriere Hills Montserrat in 1996-98 (Watts et al. 2002). Another
complicating factor in calculating magma ascent rates of intermediate composition eruptions
is the fact these magmas commonly contain evidence of a magma mixing event shortly before
the eruption. This observation, which suggests that overpressures may have developed in the
magma reservoir, is true for Soufriere Hills (Murphy et al. 2000; Rutherford and Devine 2003)
and for the 1991 Mount Pinatubo eruptions (Pallister et al. 1996).
Figure 15. Plots showing change in conduit radius (left) and magma ascent velocity (right) with eruption
date during the 1980-82 eruptions of Mount St, Helens as controlled by the observed Mass eruption rate
and the Hagen-Poiseuille law (see text). Note the rapid decrease in both conduit radius and ascent rate
during 1980. [Published with kind permission from Springer Science+Business Media from Geschwind
and Rutherford (1995) Bull Volc, Vol. 57, Figs. 8, 9, p. 367.]
MaGMa aSCent FrOM SeISMIC data
Earthquakes are commonly associated with volcanic eruptions and volcano unrest and
have been effectively used for monitoring and forecasting eruptions (e.g., Scandone and
Malone 1985; Decker 1987; Endo et al. 1996). Volcanic earthquakes generally occur in four
Magma Ascent Rates 259
main types, high- and low-frequency earthquakes, explosion earthquakes, and tremor. The
direct association of earthquakes with magma migration can be ambiguous, however, because
earthquake generation requires stressed and brittle rock, and so aseismic subvolcanic regions
may be unstressed and unfractured or hot and ductile. On the other hand, pressured magma
may generate earthquakes in brittle rock adjacent to magma-bearing conduits. Hence, magma
zones may be either seismic or aseismic.
One case of an association of earthquakes and possible magma movement occurred during
the 1977 deflation of Krafla Volcano in Iceland. At the same time that the volcano began to
deflate, earthquakes began under the central volcano and then migrated along the adjacent rift
zone (Tryggvason 1994). When the hypocenters of this migrating tremor passed a geothermal
drill-hole, fresh basaltic pumice erupted from the hole. The tremor hypocenters migrated along
the rift system at a rate of about 0.5 m/s, but it is not clear whether magma moved at this rate
or whether a pressure pulse moved through existing magma at this rate. Migration of tremor
hypocenters like that at Krafla Volcano has also been observed in the East and Southwest
Rift zones of Kilauea Volcano, Hawaii (Aki and Koyanagi 1981; Klein et al. 1987). These
swarms may define the motions of magma in the rift systems, as many of them are related to
the eruption of basaltic magma. The rates of down-rift swarm movement span a range of about
0.005 to 8.5 m/s. In a few cases, upward movement of earthquake swarms in the rift is also
identified, with rates on order of 0.03 to 1.7 m/s (Klein et al. 1987). Interestingly, this upward
migration tends to slow as the magma approaches the surface.
Although it is expected that earthquake hypocenters should migrate as magma migrates
beneath volcanoes, as in the above-described cases of Krafla and Hawaii, there are surprisingly
few documented cases of such trends. A magma ascent velocity of 0.6-0.7 m/s is estimated
for the explosive May 18, 1980 eruption of Mount St. Helens based on the arrival time of
new magma at the surface as detected initially by a change in the seismic signals adjacent to
the 8-11 km deep storage zone (Scandone and Malone 1985). Later in the 1980 phase of the
eruption, seismic signals at depth and the following onset of an eruption at the surface indicate
an average ascent velocity of 0.007-0.01 m/s, essentially identical to estimates from amphibole
reaction rims (Rutherford and Hill 1993). Kagiyama et al. (1999) estimated the rise rate of
magma at Mount Unzen (Japan) in 1991 as being 20 m/day (0.00023 m/s) using the depth
and time of seismic events. This estimate is significantly lower than the rate obtained from
Fe-Ti oxide zoning (above), possibly because it represents the average velocity of initial vent
opening as compared to an ascent rate once flow has been established.
There have been two areas of volcano seismic study that have yielded new insight into
magma ascent rates in the past 12 years. Endo et al. (1996) refined their modeling of pre-
eruption real time seismic amplitude measurements (RSAM) for several volcanic eruptions
including 1980-86 Mount St Helens, and 1991 Mount Pinatubo. They found that for many
dome-building eruptions an exponential curve was a good fit to the build of RSAM counts
prior to an eruption (Fig. 16). Using the deepest seismicity at the onset of the buildup, they
determined that the 1986 dome magma at Mount St. Helens ascended from 1.4 km to the
surface in 12 hours, i.e., magma ascended at an average rate of 0.032 m/s. This compares
reasonable well with a minimum estimate of <0.018 m/s for the average ascent rate from 8 km
depth to the surface based on the lack of amphibole reaction rims in this magma (Rutherford
and Hill 1993). Another estimate of magma ascent rate comes from the work of White (1996)
on the deep long period (DLP) earthquakes associated with the Mount Pinatubo eruptions of
1991. A swarm of DLP’s were located 28-35 km beneath the Pinatubo summit between May
26 and 28, 1991, about 11.5 days prior to the arrival of a dome-forming mixed magma at the
surface on June 7. The mixed magma consisted of 36 vol % basalt and 64% dacite (Pallister
et al. 1996), the dacite having been stored at 7-11 km depth according to the dacite petrology.
White (1996) interprets the DLP’s to have been produced by the flow of deep-seated basaltic
260 Rutherford
2000
RSAM (counts) or Tilt (microradians)
RSAM Data
SAUN Tilt
1500
Extrusion
Start Avalanche Phase
1000
500
0
124 126 128 130 132 134
Time (Julian days) 1986
Figure 16. Real Time Seismic amplitude (RSAM) counts build up at station YEL at Mount St. Helens in
May, 1986 compared to radial tilt at nearby station SAUN. RSAM for YEL peaked at 1400 UT, May 8
(Julian day 128), 1986. Tilt on the SAUN tiltmeter reversed direction several hours later; the extrusion of
lava is assumed to have occurred at the tilt reversal. (Modified after Endo and Murray 1992.)
magma upward from near the base of the crust, a flow that created vibrating fluid-filled cracks
adjacent to over-pressurized magma (Chouet et al. 1994). White (1996) speculates that the
batch of basaltic magma that produced the May 26 DLP’s may have been the same magma that
reached the surface on June 7 having passed by and partially mingled with dacite. The average
ascent rate from 35-40 km to the surface would be 0.031 to 0.035 m/s in such a case. Similar
basaltic magma ascent rates were determined over this depth range at Kilauea (Decker 1987).
MaGMa aSCent rateS BaSed On xenOLIth-MeLt reaCtIOnS
Many alkali basalt magmas erupt at the surface carrying xenoliths with clear textural and
compositional evidence to indicate that the inclusion was not in equilibrium with the host
magma. Among these xenoliths, spinel and garnet peridotite are clearly derived from the earth’s
mantle based on their high pressure phase assemblage, and must have been transported from a
source in the upper mantle (Carter 1970; Irving 1980). Reactions observed between the inclusion
and the host magma include phenocryst dissolution (Brearley and Scarfe 1986; Edwards and
Russell 1996), profiles of diffusive exchange (Klugel 1998), and zoning in inclusion glasses
(Frey and Green 1974). The extent of reaction that takes place at the magma-inclusion contact
can be used, along with available kinetic or diffusion data, to estimate the rate of magma ascent
if certain constraints can be placed on the ascent process. For example, it must be possible to
determine the depth at which the inclusion was incorporated into the magma, and it must also
be possible to assess any changes in temperature during ascent. This section describes three
examples where ascent-rate estimates have been made using such reactions.
Alkali basalt laden with inclusions of peridotite has erupted both historically (six eruptions
in the past 500 years) and prehistorically on La Palma in the Canary Islands (Klugel et al.
1997). The inclusions are spinel-bearing peridotites ranging up to 15 cm in diameter, and they
are commonly encased in a crystalline selvage (reaction) zone (< 1 mm thick) of titanium-
rich augite, amphibole, and magnetite. The selvage minerals are fine-grained at the inclusion
Magma Ascent Rates 261
contact, and coarsen toward the host magma. Fluid inclusions in recrystallized grains and in the
selvage minerals are essentially pure CO2, and their densities indicate that the selvage minerals
and recrystallized grains in the peridotite grew at pressures of 200 to 400 MPa (7-15 km depth).
Olivine crystals in the peridotite adjacent to the selvage zone have developed diffusion zones
that are 0.9 to 2.6 mm wide as a result of reaction with the magma (Fig. 17). This diffusion zone
is interpreted to have developed in olivine at the xenolith margins while the magma was stored in
a deep (7 to 15 km) crustal reservoir. Using diffusion data for Fe and Mg in olivine for a magma
temperature of 1200 °C, the measured diffusion zone thicknesses yield development times of 8
to 110 years. These times are interpreted as indicating the length of time the inclusions spent in
the magma storage zone and in ascent to the surface. Interestingly, there are some fractures and
surfaces in the xenoliths in the Canary Island samples that have no selvage zones. The diffusion
zones in olivine adjacent to the melt along these surfaces are 0 to 0.02 mm wide. The fractures
with no selvage zone are interpreted to have formed by a fracturing event at the time of the final
eruption (Klugel et al. 1997). The times calculated for development of the thin diffusion zones
range from 0 to 4 days. If it is correct to assume the surfaces formed when the magma began
to ascend, then average ascent rates of >0.06 m/s are calculated for the movement of this alkali
basalt from 11 km to the surface. This is in the same general range as the >0.2 m/s magma
ascent rates necessary to carry the xenoliths to the surface in this magma (Spera 1984). The lack
of diffusion rims of intermediate thickness on olivine in the xenoliths is interpreted to mean that
Figure 17. Drawing of a contact between an olivine-rich xenolith and host basalt with a thin selvage zone
at the boundary as is typical of such occurrences in La Palma lavas (Canary Islands). Middle panel shows
zoning profile (in olivine) along traverse in top view. Lowest panel shows diffusion-based solution for
times required to form large 1-2 mm diffusion path (6-83 years) and a thin reaction zone formed on cracks
in the olivine (<4 days). (Modified after Klugel 1998.)
262 Rutherford
the xenolith-bearing magma was stored in the lower crust and was not affected by a fracture-
producing event in the 8 to 110 year period before the eruption began.
A second petrologic indicator of magma ascent rate in alkali basalt and kimberlite magmas
is the amount of reaction that occurs between melt and garnet-bearing xenoliths; the garnet in
these lithologies is unstable in contact with melt at depths <40 km in the earth’s continental crust
for temperatures >850 °C (Green and Ringwood 1967). Based on the rate of garnet conversion
to spinel in experiments, O’Hara et al. (1971) concluded that well-preserved garnet lherzolite
could not have been exposed to temperatures as high as 850 °C at pressures <1.2 GPa for more
than a few hours. More recently, Canil and Fedortchouk (1999) have experimentally studied the
rate of garnet xenocryst dissolution in kimberlitic melt at pressures of 1.5 to 2.5 GPa and 1400 to
1500 °C, and compared the results to dissolution and reaction rims seen in natural kimberlites.
They found that the dissolution rate of garnet in H2O-saturated kimberlite is very rapid, linear
in time at a given T, and not significantly affected by differences in pressure. They also report
dissolution features on garnet xenocrysts in a natural kimberlite ranging up to 25 µm in width on
1 to 8 mm diameter spherical crystals. Based on these studies, Canil and Fedortchouck (1999)
conclude that the garnet xenocrysts spent <1 hour in 1200 °C magma once the melt and garnet
were in contact (Fig. 18). Two factors prevent us from using this dissolution data to calculate
other than minimum ascent rates. First, it is theorized that garnet xenocrysts are exposed to
the kimberlitic melt when xenoliths are fragmented, but when this happens is not known; it
may be at great depth, or it may be at the base of the diatreme zone at approximately 1-3 km
(Mitchell et al. 1980; Sparks et al. 2006). Secondly, the data and calculations are done for a
magma ascending adiabatically (Fig. 18), but if a gas phase is generated at significant depth, as
seems likely based on the evidence for abundant CO2 in kimberlite magmas, then a significant
cooling is calculated from the work done by the expanding gas phase, and the temperature
could drop from >1300 °C at 30 km (10 wt% CO2) to ~1000 °C at the surface (Sparks et al.
2006). Canil and Fedortchouck (1999) consider the two above factors, but note that the 25 µm
120
kimberlite
100
solidus
Fo
80 + Di
+ CO2
Depth (km)
Lower stability of
60
mantle carbonates
40
diatreme 1h
20
formation 10 h
. 10 d 1 min
0
200 400 600 800 1000 1200 1400
Temperature (oC)
Figure 18. Schematic plot of depth vs. temperature showing possible ascent paths for the Grizzley
(Canada) kimberlite magma based on garnet dissolution times. The arrows show simple adiabatic ascent
paths possible at different temperatures and the times required for garnet preservation on that path. The
kimberlite solidus is a minimum temperature constraint, and the lower pressure stability limit of carbonate
in the mantle is used as a possible depth constraint for the release of garnet xenocrysts into kimberlitic
magma. (Modified after Canil and Fedortchouk 1999.)
Magma Ascent Rates 263
thick dissolution features occurring on fractures in garnet are likely to have developed during
the explosive, late-stage emplacement of their kimberlite dike. They suggest that the explosive
emplacement probably occurred in only seconds to minutes. This estimate is consistent with the
Sparks et al. (2006) model for the final emplacement of kimberlite magmas.
A third example where reactions between xenoliths and the host magma have been used
to estimate magma ascent rates has developed from analytical studies of H2O in nominally
anhydrous minerals like olivine in the xenoliths. The diffusion of H2O in olivine varies with
crystallographic orientation (Mackwell and Kohlstedt 1990) and so the concentration of H2O
must be determined in a profile with a known crystallographic orientation. Diffusion rates
for H2O in olivine vary with temperature and hydrogen fugacity, but are sufficiently fast that
several mm diameter crystals will equilibrate in hours at >1000 °C (Mackell and Kohlstedt
1990). Recently, Peslier and Luhr (2006) described diffusion loss profiles in olivine crystals
from spinel peridotite inclusions carried to the surface in alkali basalt magma. They previously
found that associated pyroxenes contained significantly higher amounts of dissolved H2O but
there was no detectible H2O loss from the pyroxene rims. Demouchy et al. (2006) measured
similar H-loss profiles in olivine from a garnet lherzolite collected at a locality in Chile
(Fig. 19). In both of these studies it was assumed that the diffusive loss of water began when the
peridotite sample left the mantle source region at 40 and 60-70 km depth respectively, and the
Mackwell and Kohlstedt (1990) experimental data were used to estimate a time for the profiles
to develop. Peslier and Luhr (2006) derive times of 18-65 hours which correspond to magma
ascent rates of 0.2 to 0.5 m/s; Demouchy et al. (2006) determine times of 1.9 to 6.3 hours for
temperatures of 1200 to 1290 °C which is equivalent to a magma ascent rate of 6±3 m/s from
60 km. Both of these estimates are very similar to rates of alkali basalt ascent determined from
Figure 19. Water content profiles in a representative olivine crystal in a Garnet peridotite xenolith from Pali-
Aike, Chile (Demouchy et al. 2006). Individual data points are plotted for two crystallographic directions
in the crystal. Solid lines are calculated diffusion profiles based on previous experimental work for different
directions at 1200-1290 °C with an initial water content of 312 ppm H2O and a final 0 ppm at the rim.
Calculated profiles yield ascent rates of 1.9 h at 1290 °C and 6.3 h at 1200 °C for ascent from 70 km.
[Published with permission of the Geological Society of America from Demouchcy et al. (2006), Geology,
Vol. 34, Fig. 4, p. 431.]
264 Rutherford
dissolutions rates of garnet and from theoretical models of magma flow (Sparks et al. 2006),
but the 6 m/s determination is at high end of the range. The modeling of H-loss profiles (or
other fast-diffusing elements) shows significant promise for assessing magma ascent rates, but
again there is a significant need for additional work to determine if the diffusive loss of H from
olivine begins in the mantle as assumed. A related problem is the amount of water dissolved in
the surrounding alkali basalt, and the effect it has on H-loss from olivine xenocrysts.
xenOLIth tranSPOrt, FLOW MOdeLInG
and MaGMa aSCent rateS
The presence of dense mantle xenoliths in tephra and lava flow eruptions of alkali basalt
and kimberlite indicates that the magma carried them from depth (e.g., Irving 1980). An average
settling rate for xenoliths (minimum magma ascent rate) can be calculated by balancing the
different forces on the xenoliths during ascent, and by making certain assumptions about the
rheological properties of the magma (McGetchin and Ulrich 1973; Mitchell et al. 1980; Spera
1984; Sparks et al. 2006). Figure 20 shows the results of settling calculations for spherical
xenoliths, a density difference of 500 kg/m3 between the xenoliths and the melt (melt density
= 2,800 kg/m3), and an assumption of Bingham plastic rheology for the xenolith-laden
magma. As a Bingham fluid, the magma has a yield strength (σ°) that correlates well with
the apparent viscosity. Yield strength of 100 N/m2 corresponds to a phenocryst content of
~25 vol% in a basalt. The calculation indicates that a 20-cm-wide xenolith will settle at a
rate of 0.1 m/s. Obviously, the ascent rate of a magma carrying this xenolith would have to
be >0.1 m/s in order to carry the xenolith to the surface; the xenolith settling velocity gives
a minimum estimate of magma ascent rate. Published minimum estimates for alkali basalt
ascent rates range from 0.1 to 2 m/s based on a lack of chemical zonation in phenocrysts, or a
preservation of chemical zonation in melt pools (Brearley and Scarfe 1986). Two factors that
40
Xenolith Settling Rate (cm/s)
35
σo=20
30
25
σo=50
σo=0
20
σo=100
15 σo=150
10 σo=200
0
0 5 10 15 20 25
Xenolith Radius (cm)
Figure 20. Plot showing calculated xenolith settling velocity as a function of xenolith radius for different
magma yield strengths (σo). A density difference between the xenolith and melt of 500 kg/m3 is assumed;
the melt density is 2,800 kg/m3 and a Bingham rheology (viscosity of 350 poise) is assumed. Units of yield
strength are N/m2. A yield strength of 100 N/m2 corresponds approximately to a crystallinity of 25 vol%.
(Adapted from Spera 1984). The slight change in slope on the yield strength curves is due to the increased
importance of inertia relative to friction and buoyancy forces at high xenolith Reynolds numbers. The
calculations were made using Stokes law {U = 2gr2(dr)/9 η} modified to accommodate variations in the
properties of the melt and nodules (Modified after Spera 1984.)
Magma Ascent Rates 265
have not been considered in the settling calculation are the effect of nearby xenoliths and the
effect of bubbles in the magma. If significant, both of these would decrease the settling rate of
xenoliths, and therefore reduce the minimum ascent rate required to carry a xenolith of a given
size to the surface. Based on the relative abundance of vesicles and xenoliths in alkali basalt,
however, it is possible that neither has a significant effect on calculated ascent rates since there
is no obvious correlation between crystal content in the alkali basalt and xenolith abundance
(Spera 1984). Spera (1984) also argues that while the volatile abundance in such magmas is
likely to significantly affect the initial velocity of the xenolith-carrying magmas, increases in
ascent rate at depths >3-5 km will not result from volatile exsolution or expansion because the
solubility of the CO2-rich gas exsolved at depth changes slowly with pressure, and the gas is
relatively incompressible at >200 MPa.
Kimberlite, another mantle xenolith-bearing magma, is known to be very gas-rich and
also commonly has high phenocryst and/or xenolith contents (Dawson 1971). Spera (1984) did
not use xenolith settling rates to constrain magma ascent rate for kimberlite magmas, possibly
because the model (Fig. 20) depends strongly on good estimates for the magma viscosity (yield
strength). However, theoretical models of magma flow have been developed for evaluating as-
cent of magmas that range from basalt to kimbelite in composition (McGetchin and Ulrich
1973: Mitchell et al. 1980; Spera 1984; Sparks et al. 2006). Following McGetchin and Ulrich
(1973), Spera (1984) estimates that the ascent rate of kimberlitic magma is in the range of 10
to 30 m/s, higher than the ascent rate of alkali basalt by a factor of 10. However, Mitchell et al.
(1980) consider that the ascent of such gas-rich magmas is probably best considered as taking
place in three stages. Stage 1 (100-10 km) may be thought of as flow of a Newtonian fluid in
a pipe, although the flow will not be Newtonian if the magma contains a significant load of
xenoliths and xenocrysts. Additionally, there now seems to be general acceptance that kimber-
litic magma flow is in thin (0.3-1 m) dikes for stage 1 depths (Mitchell et al. 1980; Sparks et
al. 2006). Stage 2 (10-1 km) is where the gas phase becomes important in determining magma
properties and ascent rates; the time spent in this stage is 3-4 % of the total ~100 km ascent time
in most models. Stage 3 is the explosive diatreme stage (1-0 km) and lasts for an estimated 7 s
out of a total ascent time of 130 minutes (Mitchell et al. 1980). The velocity of the magma in
the final two stages is very high as required to explain the diatreme deposit characteristics. The
important point however is that the most recent theoretical models (Table 1) suggest that kim-
berlitic magma velocity in the dike stage (1) averages from 2.9-16.8 m/s (Sparks et al. 2006),
very much overlapping with the 10-30 m/s estimates from the earliest models.
table 1. Ascent rates for dike transported kimberlite magma
calculated using u = 7.7[w5/{µ(ρg∆r)3}]1/7g∆ρ (Sparks et al. 2006).
dike width (2w) Velocity Velocity
(m) ∆ρ = 100 kg/m3 ∆ρ = 300 kg/m3
0.2 2.9 5.4
0.3 3.8 7.2
0.6 6.3 11.7
1.0 9.1 16.8
SuMMary OF MaGMa aSCent rate eStIMateS
The estimates of average magma ascent rates for silicic to intermediate composition lava
extrusions are fairly well constrained in the range of 0.001 m/s for very slowly ascending
magmas, to 0.015 m/s for more rapidly ascending extrusive magmas (Table 2). These
266 Rutherford
table 2. Magma ascent rate estimates from different observations.*
explosive ascent extrusive ascent
Volcano Observation*
rate (m/s) rate (m/s)
Mount St. Helens groundmass crystallization >> 0.2 0.01 - 0.02
Mount St. Helens hornblende rims > 0.18 0.04 - 0.15
Mount St. Helens calculation from mass eruption 1-2 0.01 - 0.1
rate
Mount St. Helens seismicity 0.6 0.007 - 0.01
Soufriere Hills, Mont. amphibole rims > 0.2 0.001 - 0.012
Soufriere Hills, Mont. magnetite > 0.2 0.003 - 0.015
Mount Unzen magnetite zonation not present 0.002
Black Butte CA amphibole rims & plagioclase not present 0.004-0.006
growth
Hualali, HI alk basalt xenolith transport not present > 0.1
La Palma, Canary Is. olivine zonation not present > 0.06
xenocrysts in alk basalt hydrogen zoning in olv 0.2 to 0.5 m/s
Kimberlite theoretical modeling > 4 -16
Kimberlite garnet dissolution 1.1 to 30 m/s (final 2 km)
* Different types of observations are discussed in various sections of the paper.
estimates come from a variety of observations made at different volcanoes, and it is important
to remember that they are average rates. Ascent rates are likely to vary over the vertical extent
of the conduit carrying the magma, because the density and viscosity of the magma will vary
considerably over this length, particularly in water-rich magmas undergoing crystallization
and volatile exsolution. Additionally, the effective cross section area of the conduit at depth is
generally not known. One topic for future study is the nature of changes in flow within volcanic
conduits. It has also been suggested that convection may occur in conduits (Witter et al. 2005);
in such a case, the magma ascent would be very difficult to assess, but a record should be
preserved in some of the phenocryst and groundmass minerals such as the Fe-Ti oxides. It
is rather interesting that observations made for basaltic composition magma extrusions give
essentially the same range of average ascent rates as obtained for more silica-rich magmas.
The assessment of magma ascent rates during explosive eruptions is more complex
because there is almost certainly a much greater variation in the ascent rate from the base of
the conduit to the surface. This is particularly true for eruptions where there is evidence for
alternating effusive dome-forming eruptions and conduit-clearing explosive events (e.g., 1980
Mount St. Helens and 1996-2000 Soufriere Hills, Montserrat). The magma leaves the top of
the conduit at essentially sonic velocities for truly explosive eruptions (Papale et al. 1998;
Sparks et al. 2006). This high rate of ascent is achieved as a result of the low density and
rapid expansion of the gas phase in these gas-rich magmas. Deep in the conduit, very little of
the gas is exsolved, the magma is dense, and depending on the size and shape of the conduit,
flow rates are calculated to be much lower. In our view, some of the more profitable areas
for study of magma ascent lie with investigations of the behavior of the various magmatic
volatiles; the rates of their diffusion in melts, and the processes by which gas can be lost
from magmas during ascent. Additionally, since so many eruptions appear to be staged from
a magma storage zone in the earth’s crust, a better understanding of storage zone processes
(e.g., magma injection, convection and mixing) would greatly improve our understanding of
the magma ascent process.
Magma Ascent Rates 267
reFerenCeS
Alletti M, Baker DR, Feda C (2007) Halogen diffusion in a basaltic melt. Geochim Cosmochim Acta 71:3570-
3580
Aki K, Koyanagi RY, (1981) Deep volcanic tremor and magma ascent under Kilauea, Hawaii. J Geophys Res
86:7095-7109
Anderson AT (1991) Hourglass inclusions: Theory and application to the Bishop Rhyolitic Tuff. Am Mineral
76:530-547
Aspinall WP, Miller AD, Lynch LL, Latchman JL, Stewart RC, White RA, Power JA (1998) Soufriere Hills
eruption, Montserrat: Volcanic earthquake locations and fault plane solutions. Geophys Res Lett 25:3397-
3401
Baker, DR, Lang P, Robert G, Bergevin J-F, Allard E, Bai L (2006) Bubble growth in slightly supersaturated
albite melt at constant pressure. Geochim Cosmochim Acta 70:1821-1838
Baker LL, Rutherford M (1996) Sulfur diffusion in rhyolite melts. Contrib Mineral Petrol 123:335-344
Barclay J, Rutherford MJ, Carroll MR, Murphy MD, Devine JD, Gardner J, Sparks RSJ (1998) Experimental
phase equilibria constraints on pre-eruptive conditions of the Soufriere Hills magma. Geophys Res Lett
25:3437-3440
Blundy J, Cashman K, Humphreys M (2006) Magma heating by decompression-driven crystallization beneath
andesite volcanoes. Nature 44:76-80
Brearley M, Scarfe CM (1986) Dissolution rates of upper mantle minerals in an alkali basalt melt at high
pressure: an experimental study and implications for ultramafic xenolith survival. J Petrol 27:1157-1182
Browne BL, Gardner JE (2006) The influence of magma ascent path on the texture, mineralogy, and formation
of Hornblende reaction rims. Earth Planet Sci Lett 246:161-176
Buckley VJE, Sparks RSJ, Wood BJ (2006) Hornblende dehydration reactions during magma ascent at Soufriere
Hills Volcano, Montserrat. Contrib Mineral Petrol 151:121-140
Canil D, Fedortchouk Y (1999) Garnet dissolution and the emplacement of kimberlites. Earth Planet Sci Lett
167:227-237
Carey S, Sigurdsson H (1985) The May 18, 1980, eruption of Mount St. Helens 2: Modeling of dynamics of the
Plinian phase. J Geophys Res 90:2948-2958
Carmichael ISE (2002) The andesite aqueduct: perspectives on the evolution of intermediate magmatism in west
central (105-99°W) Mexico. Contrib Mineral Petrol 143:641-663
Carter JL (1970) Mineralogy and chemistry of the Earth’s upper mantle based on the partial fusion-partial
crystallization model. Bull Geol Soc Am 81:2021-2034
Cashman KV (1988) Crystallization of Mount St. Helens 1980-1986 dacite: a quantitative textural approach.
Bull Volcanol 50:194-209
Cashman KV (1992) Groundmass crystallization of Mount St. Helens dacite, 1980-1986: a tool for interpreting
shallow magmatic processes. Contrib Mineral Petrol 109:431-449
Cervantes P, Wallace P (2003) The role of water in subduction zone volcanism: New insights from melt inclusions
in high-Mg basalts from central Mexico. Geology 31:235-238.
Chouet BA, Page RA, Stephens CD, Lahr JC, Power JA (1994) Precursory swarms of long-period events at
Redoubt Volcano (1989-1990), Alaska: their origin and use as a forecasting tool. J Volc Geotherm Res
62:95-135
Condomines M, Hemond C, Allegre CJ (1988) U-Th-Ra radioactive disequilibria and magmatic processes.
Earth Planet Sci Lett 90:243-262
Corsaro R, Pompillio M (2004) Dynamics of magmas at Mount Etna. In: Mount Etna: Volcano Laboratory.
Bonaccorso A, Calvari S, Colteli M, Del Negro C, Falsaperla S (eds) AGU Geophysical Monograph
143:91-110
Couch S (2003) Experimental investigation of crystallization kinetics in a haplogranitic system. Am Mineral
88:1471-1485
Couch S, Sparks RSJ, Carroll MR (2003) The kinetics of degassing-induced crystallization at Soufrierre Hills
Volcano, Montserrat. J Petrol 44:1477-1502
Dawson JB (1971) Advances in kimberlite geology. Earth Sci Rev 7:182-214
Decker RW (1987) Dynamics of Hawaiian volcanoes: an overview. In: Volcanism in Hawaii. Decker RW,
Wright TL Staufer PH (eds) U.S. Geol Survey Prof Paper 1350:997-1018
Demouchy S, Jacobsen SD, Gaillard F, Stem CR (2006) Rapid magma ascent recorded by water diffusion
profiles in mantle olivine. Geology 34:429-432
Dejesus S, Rutherford MJ (2007) The origin and significance of F-rich amphibole in the 2004-6 Mount St.
Helens magma. EOS Trans, AGU 88(52) Fall Meet Suppl Abstract V41B-0603
Devine JD, Rutherford MJ, Norton GE, Young SR (2003) Magma storage zone processes inferred from
geochemistry of Fe-Ti oxides in andesitic magma, Soufriere Hills volcano, Montserrat. J Petrol 44:1375-
1400
268 Rutherford
Dixon JE, Stolper EM (1995) An experimental study of water and carbon dioxide solubilities in mid-ocean ridge
basaltic liquids: Part II: Applications to degassing. J Petrol 36:627-654
Edwards BR, Russell JK (1996) A review and analysis of silicate mineral dissolution experiments in natural
silicate melts. Chem Geol 130:233-245
Eichelberger JC (1978) Andesitic volcanism and crustal evolution. Science 275:21-27
Eggler DH, Burnham CW (1973) Crystallization and fractionation trends in the system andesite-H2O-CO2-O2 at
pressures to 10 kb. Bull Geol Soc Am 84:2517-2532
Endo ET, Murray T (1992) Real-time seismic amplitude measurement (RSAM): a volcano monitoring and
prediction tool. Bull Volc 53:533-545
Endo ET, Murray TL, Power JA (1996) A comparison of pre-eruption, real-time seismic amplitude measurements
for eruptions at Mount St. Helens, Redoubt Volcano, Mount Spurr, and Mount Pinatubo. In: Fire & Mud:
eruptions and lahars of Mount Pinatubo, Philippines. Newhall CG, Punongbayan RS (eds) University of
Washington Press, Seattle, p 233-246
Freer R, Hauptman z (1978) An experimental study of magnetite-titanomagnetite interdiffusion. Phys Earth
Planet Inter 16:223-231
Frey FA, Green DH (1974) The mineralogy, geochemistry and origin of lherzolitic inclusions in Victorian
basanites. Geochim Cosmochim Acta 38:1023-1059
Garcia MO, Jacobsen SS (1979) Crystal clots amphibole fraction and the evolution of calc-alkaline magmas.
Contrib Mineral Petrol 69:319-327
Gardner JE (2007) Heterogeneous bubble nucleation in highly viscous silicate melts during instantaneous
decompression from high pressure. Chem Geol 236:1-12
Gardner JE, Hilton M, Carroll MR (1999) experimental constraints on degassing of magma: Isothermal bubble
growth during continuous decompression from high pressure. Earth Planet Sci Lett 168:201-218
Gardner JE, Denis M-H (2004) Heterogeneous bubble nucleation on Fe-Ti oxide crystals in high-silica rhyolitic
melts. Geochim Cosmochim Acta 68:3587-3597
Gerlach TM, McGee KA, Doukas MP (2008) Emission rates of CO2, SO2, H2S, scrubbing, and pre-eruption
excess volatiles at Mount St. Helens, 2004-2005. In: A volcano rekindled: the first year of renewed
eruption at Mount St. Helens, 2004-2006. Sherrod, DR, Scott WE, Stauffer PH (eds) US Geol Survey Prof
Pap 1750, Chpt 26 (in press)
Geschwind C-H, Rutherford MJ (1995) Crystallization of microlites during magma ascent: the fluid mechanics
of 1980-86 eruptions at Mount St. Helens. Bull Volc 57:356-370
Ginibre C, Worner G, Kronz A (2007) Crystal zoning as an archive for magma evolution. Elements 3:261-266
Green DH, Ringwood AE (1967) An experimental investigation of the gabbro to eclogite transformation and its
petrological applications. Geochim Cosmochim Acta 81:767-833
Hammer JE, Cashman K, Hoblitt RP (1999) Degassing and microlite crystallization during pre-climatic events
of the 1991 eruption of Mount Pinatubo, Philippines. Bull Volc 60:355-380
Hammer JE, Rutherford MJ (2002) Kinetics of decompression-induced crystallization in silicic melt. J Geophys
Res 107, doi:10.1029/2001JB000281
Hammer, JE Rutherford MJ (2003) Glass Composition geobarometry: a petrologic indicator of pre-eruption
Pinatubo dacite magma dynamics. Geology 31:79-82
Hoblitt RP, Wolfe EW, Scott WE, Couchman MR, Pallister JS, Javier D (1996) The preclimatic eruptions of
Mount Pinatubo, June 1991. In: Fire & Mud: eruptions and lahars of Mount Pinatubo, Philippines. Newhall
CG, Punonngbayan RS (eds) University of Washington Press, Seattle, p 457-512
Holloway JR, Blank JG (1994) Application of experimental results to C-O-H species in natural melts. Rev
Mineral 30:187-230
Humphreys MCS, Menad T, Blundy JD, Klimm K (2008) Magma ascent rates in explosive eruptions: Constraints
from H2O diffusion in melt inclusions. Earth Planet Sci Lett 270:25-40
Hurwitz S, Navon O (1994) Bubble nucleation in rhyolitic melts: Experiments at high pressure, temperature and
water content. Earth Planet Sci Lett 122:267-280
Irving AJ (1980) Petrology and Geochemistry of composite ultramafic xenoliths in alkali basalts and implications
for magmatic processes in the mantle. Am J Sci 280A:389-426
Izbekov PE, Eichelberger JC, Patino LC, Vogel TA, Ivanov BA (2002) Calcic cores of plagioclase phenocrysts
in andesite from Karyminski volcano: evidence for rapid introduction by basaltic replenishment. Geology
80:799-802
Kagiyama T, Utada H, Yamamoto T (1999) Magma ascent beneath Unzen volcano, SW Japan, deduced from the
electrical resistivity structure. J Volc Geotherm Res 89:35-42
Klein FW, Koyanagi RY, Nakata JS, Tanigawa WR (1987) The seismicity of Kilauea’s magma system. In:
Volcanism in Hawaii Decker RW, Wright TL Staufer PH (eds) U.S. Geol Survey Prof Paper 1350:1019-
1185
Klugel A, Hansteen TH, Schminke H-U (1997) Rates of magma ascent and depths of magma reservoirs beneath
La Palma (Canary Islands). Terranova 9:117-121
Magma Ascent Rates 269
Klugel A (1998) Reactions between mantle xenoliths and host magma beneath La Palma (Canary Islands):
constraints on magma ascent rates and crustal reservoirs. Cont Mineral Petrol 131:237-257
Liu Y, Anderson AT, Wilson CJN (2007) Melt pockets in phenocrysts and decompression rates of silicic magmas
before fragmentation. J Geophys Res 112:B06204 doi:10.1029/2006JB004500
MacGregor AG (1938) The volcanic history and Petrology of Montserrat with observations on Mt. Pele in
Martinique: Royal Society expedition to Montserrat BWI. Phil Trans Roy Soc Ser B 229:1-90
Mackwell SJ, Kohlstedt DL (1990) Diffusion of hydrogen in olivine: implications for water in the mantle. J
Geophys Res 95:5079-5088
Mangan M, Sission T (2000) Delayed, disequilibrium degassing in rhyolite magma: decompression experiments
and implications for explosive volcanism. Earth Planet Sci Lett 183:441-455
Mangiacapra A, Moretti R, Rutherford MJ, Civetta L, Orsi G, Papale P (2008) The deep magma system at
Campi Flegrei caldera (Italy). Geophys Res Lett (in review)
Mastin LG, Ghiorso MS (2001) Adiabatic temperature changes of magma-gas mixtures during ascent and
eruption. Contrib Mineral Petrol 141:307-321
McCanta MC, Hammer JE, Rutherford MJ (2007) Pre-eruptive and syn-eruptive conditions in the Black Butte,
CA dacite: Insight into crystallization kinetics in a silicic magma system. J Volc Geotherm Res 160:263-
284
McGetchin TR, Ullrich GW (1973) xenoliths in Maars and diatremes with inferences for the Moon, Mars, and
Venus. J Geophys Res 78:1833-1853
Metrich N, Schiano P, Clocchiati R, Maury RC (1999) Transfer of Sulfur in subduction settings: an example
from Batan Island (Luzon volcanic arc, Philippines). Earth Planet Sci Lett 167:1-14
Mitchell RH, Carswell DA, Clarke DB (1980) Geological implications and validity of calculated equilibrium
conditions for ultramafic xenoliths from the pipe 200 kimberlite, Northern Lesotho. Contrib Mineral Petrol
72:205-217
Murphy MD, Sparks RSJ, Barclay J, Carroll MR, Brewer TS (2000) Remobilization of andesitic magma by
intrusion of mafic magma at the Soufriere Hills volcano, Montserrat, West Indies. J Petrol 41:21-42
Nakada S, Fujii T (1993) Preliminary report on the activity at Unzen volcano (Japan) November 1990- November
1991: dacite lava domes and pyroclastic flows. J Volc Geotherm Res 54:319-333
Nakamura M (1995) Continuous mixing of crystal mush and replenished magma in the ongoing Unzen eruption.
Geology 23:807-810
Nicholis MG, Rutherford MJ (2004) 32 Experimental constraints on magma ascent rate for the Crater Flat
volcanic zone hawaiite. Geology 32:489-492
Nicholis MG (2006) Degassing processes in basaltic magmas: terrestrial and lunar applications. PhD dissertation,
Brown University, Providence RI
O’Hara MJ, Richardson SW, Wilson G (1971) Garnet peridotite stability and occurrence in crust and mantle.
Contrib Mineral Petrol 32:48-68
Pallister JS, Hoblitt, RP, Crandell DTR, Mullineaux DR (1992) Mount St Helens a decade after the 1980
eruptions: Magmatic models, chemical cycles, and a revised hazards assessment. Bull Volc 54:126-146
Pallister JS, Hoblitt RP, Meeker GP, Knoght RJ, Siems DF (1996) Magma mixing at Mount Pinatubo: Petrographic
and chemical evidence from the 1991 deposits. In: Fire & Mud: eruptions and lahars of Mount Pinatubo,
Philippines. Newhall CG, Punonngbayan RS (eds) University of Washington Press, Seattle, p 687-732
Pallister JS, Thornber CR, Cashman KV, Clynne MA, Lowers HA, Brownfield IA, Meeker GP (2008) Petrology
of the 2004-2006 Mount St. Helens lava dome – implications for magmatic plumbing and eruption
triggering. In: A volcano rekindled: the first year of renewed eruption at Mount St. Helens, 2004-2006.
Sherrod, DR, Scott WE, Stauffer PH (eds) US Geol Survey Prof Paper 1750, chpt 19 (in press)
Papale P, Dobran, F (1994) Magma flow along the volcanic conduit during the plinian and pyroclastic flow
phases of the May 18, 1980, Mount St. Helens eruption. J Geophys Res 99:4355-4373
Papale P, Neri A, Macedonio G (1998) The role of magma composition and water content in explosive eruptions:
1. Conduit ascent dynamics. J Volc Geotherm Res 87:75-93
Papale P, Moretti R, Barbato D (2006) The compositional dependence of the saturation surface of H2O+CO2
fluids in silicate melts. Chem Geol 229:78-95
Peslier AH, Luhr JF (2006) Hydrogen loss from olivines in mantle xenoliths from Simcoe (USA) and Mexico:
mafic alkalic magma ascent rates and water budget of the sub-continental litosphere. Earth Planet Sci Lett
242:302-319
Pinkerton H, Wilson L, Macdonald R (2002) The transport and eruption of magma from volcanoes: a review.
Contemp Phys 43:197-210
Power JA, Lahr JC, Page RA, Chouet BA, Stephens CD, Harlow DH, Murray TL, Davies JN (1994) Seismic
evolution iof the 1989-90 eruption sequence of redoubt volcano, Alaska. J Volc Geotherm Res 62:69-94
Roggensack K, Hervig RL, McNiight SB, Williams SN (1997) Explosive basaltic volcanism from Cerro Negro
volcano: influence of volatiles on eruptive style. Science 277:1639-1642
Rutherford MJ, Sigurdsson H, Carey S (1985) The May 18, 1980 eruption of Mount St. Helens, 1: Melt
compositions and experimental phase equilibria. J Geophys Res 90:2929-2947
270 Rutherford
Rutherford MJ, Devine JD (1988) The May 18, 1980 eruption of Mount St. Helens Ill: Stability and chemistry
of amphibole in the magma chamber. J Geophys Res 93:11,949-11,959
Rutherford MJ, Hill PM (1993) Magma ascent rates from amphibole breakdown: Experiments and the 1980-
1986 Mount St. Helens eruptions. J Geophys Res 98:19667-19685
Rutherford MJ, Devine JD (2003) Magmatic conditions and magma ascent as indicated by hornblende phase
equilibria and reactions in the 1995-2001 Soufriere Hills magma. J Petrol 44:1433-1454
Rutherford MJ, Devine J (2008) Magmatic conditions and processes in the Storage zone of the 2004-06 Mount
St. Helens Eruption: The record in Amphibole and Plagioclase phenocrysts. In: A volcano rekindled: the
first year of renewed eruption at Mount St. Helens, 2004-2006. Sherrod, DR, Scott WE, Stauffer PH (eds)
US Geol Survey Prof Pap 1750, chpt 31 (in press)
Ryan M (1993) Neutral buoyancy and the structure of mid-ocean ridge magma reservoirs. J Geophys Res
98:22321-22338
Saal A, Hauri E, LoCasio M, Van Orman J, Rutherford M, Cooper R (2008). The volatile content of lunar
volcanic glasses: Evidence for the presence of water in the lunar interior. Nature 454:192-195
Scandone RS, Malone S (1985) Magma supply, magma discharge and readjustment of the feeding system of
Mount St. Helens during 1980. J Volc Geotherm Res 23:239-262
Schilling SP, Thompson RA, Messerich JA, Iwatsubo EY (2008) Mount St. Helens lava dome growth and
surface modeling, 2004-2005. In: A volcano rekindled: the first year of renewed eruption at Mount St.
Helens, 2004-2006. Sherrod, DR, Scott WE, Stauffer PH (eds) US Geol Survey Prof Pap 1750, chpt 8 (in
press)
Singer BS, Dungan MA, Layne GD (1995) Textures and Sr, Ba, Mg, Fe, K, and Ti compositional profiles in
volcanic plagioclase: Clues to the dynamics of calc-alkaline magma chambers. Am Mineral 80:776-798
Sparks RSJ (1978) the dynamics of bubble formation and growth in magmas: a review and analysis. J Volc
Geotherm Res 3:1-37
Sparks RSJ, Young SR, Barclay J, Calder ES, Cole P, Daroux B, Davies MA, Druitt TH, Harford C, Herd R,
James M, Lejeune AM, Loughlin S, Norton G, Skerrit G, Stasiuk MV, Stevens NS, Toothill J, Wage G,
Watts R (1998) Magma production and growth of the lava dome of the Soufriere Hills volcano, Montserrat:
November 1995 to December 1997. Geophys Res Lett 25:3421-3425
Sparks RSJ, Baker L, Brown RJ, Field M, Schumacher J, Stripp G, Walters A (2006) Dynamical constraints on
kimberlitic volcanism. J Volc Geotherm Res 155:18-48
Spera FJ (1984) Carbon dioxide in petrogenesis III: role of volatiles in the ascent of alkaline magma with special
reference to xenolith-bearing magmas. Contrib Mineral Petrol 88:217-232
Swanson DA, Dzurisin D, Holcomb TR, Iwatsubo EY, Chadwick WW Jr., Casadeval TJ, Ewert JW, Heliker CC
(1987) Growth of lava domes at Mount St. Helens, Washington. Geol Soc Am Spec Pap 212: 1-16
Thornber C, Pallister JS, Lowers H, Rowe M, Mandeville C, Meeker G (2008) The chemistry, mineralogy and
Petrology of amphibole in Mount St Helens 2004-2006 dacite. In: A volcano rekindled: the first year of
renewed eruption at Mount St. Helens, 2004-2006. Sherrod, DR, Scott WE, Stauffer PH (eds) US Geol
Survey Prof Pap 1750, chpt 32 (PAGE RANGE)
Toramaru A (1995) Numerical study of nucleation and growth of bubbles in viscous magma. J Geophys Res
104:20097-20122
Tryggvason E (1994) Surface deformation at the Krafla volcano, North Iceland, 1982-1992. Bull Volc 56:98-
107
Tuttle OF, Bowen NL (1958) Origin of granite in the light of experimental studies in the system NaAlSi3O8-
KalSi3O8-SiO2-H2O. Geol Soc Am Mem 74
Vaniman DT, Crowe BM, Gladney ES (1982) Petrology and geochemistry of Hawaiitic lavas from Crater Flat,
Nevada. Contrib Mineral Petrol 80:341-357
Venezky D, Rutherford MJ (1997) Pre-eruption conditions and timing of magma mixing in the 2.2 ka C-layer,
Mount Rainier. J Geophys Res 102:20,069-20,086
Venezky D, Rutherford MJ (1999) Petrology and Fe-Ti oxide reequilibration of the 1991 Mount Unzen mixed
magma. J Volc Geotherm Res 89:213-230
Wallace PJ (2005) Volatiles in subduction zone magmas: concentrations and fluxes based on melt inclusion and
volcanic gas data. J Volc Geotherm Res 140:217-240
Watson EB (1994) Diffusion in volatile-bearing magmas. Rev Mineral 30:371-411
Watts RB, Herd RA, Sparks RSJ, Young SR (2002) Growth patterns and emplacement of the andesitic lava dome
at Soufriere Hills Volcano, Montserrat. In: The Eruption of Soufriere Hills volcano Montserrat from 1995
to 1999. Druitt TH, Kokelar BP (eds) Geo Soc London Memoir 21:115-152
Westrich HR, Stockman HW, Eichelberger JC (1988) Degassing of rhyolitic magma during ascent and
emplacement. J Geophys Res. 93:6503-6511
White RA (1996) Precursory deep long-period earthquakes at Mount Pinatubo spatio-temporal link to a basalt
trigger. In: Fire & Mud: Eruptions and Lahars of Mount Pinatubo, Philippines. Newhall CG, Punonngbayan
RS (eds) University of Washington Press, Seattle, p 233-246
Magma Ascent Rates 271
Witter JB, Kress VG, Newhall CG (2005) Volcan Popocatepetl, Mexico. Petrology, magma mixing, and
intermediate sources for volatiles for the 1994-present eruption. J Petrol 46:2337-2366
Wilson L, Sparks RSJ, Walker GPL (1980) Explosive volcanic eruptions IV: the control of magma properties
and conduit geometry on eruption column behavior. Geophys J Roy Astron Soc 63:117-148
zhang Y Stolper EM (1991) Water diffusion in basaltic melt. Science 351:306-309
zhang Y Behrens H (2000) H2O diffusion in rhyolite glasses. Geochim Cosmochim Acta 55:441-456
Reviews in Mineralogy & Geochemistry
Vol. 69 pp. 273-331, 2008 8
Copyright © Mineralogical Society of America
Melt Inclusions in Basaltic and
related Volcanic rocks
adam J.r. Kent
Department of Geosciences
Oregon State University
Corvallis, Oregon, 97331, U.S.A.
[email protected]
IntrOduCtIOn
Melt inclusions are small parcels of melt trapped in crystals within magmatic systems,
and are analogous to fluid inclusions formed by trapping of hydrothermal and other fluids
during mineral growth in fluid-mineral systems (Sorby 1858; Roedder 1979, 1984). After
trapping, melt inclusions are potentially isolated from external melt and thus provide a way
to investigate melts trapped during magmatic evolution—driven by processes such as crystal-
liquid separation, vapor saturation and degassing, magma mixing and assimilation—which
can dramatically alter the compositions of the eventual erupted (or intruded) magmatic end
products. Melt inclusions are a powerful tool for the study of basaltic magma systems and
their mantle source regions, and are widely used to study the origin and evolution of mantle-
derived magmas. Melt inclusions have specific uses in the study of volatile elements (see
chapters by Metrich and Wallace 2008, Moore 2008, and Blundy and Cashman 2008), but
also provide unique information about the range of melt compositions present within basaltic
magmatic systems, and how these reflect mantle sources and the processes that occur during
melt generation, evolution, transport and eruption. This review outlines techniques used to
obtain chemical and other information from melt inclusions, discusses the processes which
lead to melt inclusion trapping in phenocryst minerals, examines the possible means by which
melt inclusion compositions might be fractionated during trapping or during subsequent re-
equilibration with the host mineral or external melt, and discusses some implications of melt
inclusion compositions for the nature of basaltic melt generation and transport systems.
This review is largely restricted in scope to studies of volcanic rocks of basaltic and related
composition. This refers to rocks erupted as lavas or tephra with broadly basaltic compositions:
SiO2 ~45-52 wt%, relatively high MgO and FeO, and typically containing one or more of
the following minerals as phenocrysts: forsterite-rich olivine, anorthite-rich plagioclase,
orthopyroxene, clinopyroxene and a spinel phase (aphyric basalts also occur but are less useful
for melt inclusion studies). These are magmas that ultimately derive from melting of the mantle,
but most have also experienced some modification during ascent and magma storage within
the oceanic or continental crust. Melt inclusions are found in all terrestrial environments in
which basaltic and related rocks occur, as well as in rocks of basaltic composition from lunar,
martian and other meteorite samples (e.g., Bombardieri et al. 2005; Nekvasil et al. 2007).
The ubiquity of melt inclusions in basaltic rocks emphasizes that the conditions amenable to
trapping melt inclusions occur during the evolution of almost all basaltic magmas, regardless
of environment of formation. It is also worth noting that there is a significant body of literature
devoted to the study of melt inclusions in rocks of other compositions—particularly those
related to silicic volcanic and plutonic rocks, but also carbonatites, kimberlites and other rarer
magma types (e.g., Nielsen and Veksler 2002; Kamenetsky et al. 2004), and inclusions also
1529-6466/08/0069-0008$10.00 DOI: 10.2138/rmg.2008.69.8
274 Kent
occur within mafic cumulates and peridotitic residues from mantle melting (e.g., Schiano and
Bourdon 1999; Schiano et al. 2004) and in layered mafic intrusions (e.g., Jakobsen et al. 2005).
These studies are largely beyond the scope of this review.
There are three primary advantages that the study of melt inclusions, in comparison to
studies based on whole rock or matrix glass samples of erupted basaltic rocks, can confer—of
which the first two are the most important in the literature. Firstly, melt inclusions trapped at
pressures that are higher than those of eventual eruption can preserve abundances of volatile
elements (H, C, Cl, S, F; see in this volume Metrich and Wallace 2008; Moore 2008, and Blundy
and Cashman 2008) and these may be directly measured within trapped glasses. In many cases
the abundances of volatile species can be inferred or demonstrated to be largely unaffected
by shallow-pressure degassing (e.g., Sisson and Bronto 1998; Newman et al. 2000; Saal et al.
2002; Cervantes and Wallace 2003a) and approach the compositions of undegassed magmas.
Melt inclusions also provide the opportunity to measure the isotopic composition (H, C, S, Cl)
and speciation (H, S) of volatile elements and to examine the relationship between volatiles and
other compositional or petrological features. This has been particularly beneficial for efforts to
understand the role of water in formation of primitive basalts in arc and back arc environments
(e.g., Sisson and Layne 1993; Sisson and Bronto 1998; Newman et al. 2000; Cervantes and
Wallace 2003a; Walker et al. 2003), studies of the interplay between volcanic eruption and
volatile degassing (e.g., Anderson 1975; Wallace and Anderson 1998; Cervantes and Wallace
2003b; Gurenko et al. 2005) and for efforts to study the volatile structure of the terrestrial
mantle (e.g., Sobolev and Chaussidon 1996; Lassiter et al. 2002; Saal et al. 2002). In many
cases this information cannot be gleaned from study of the host lavas, as these have degassed
or are altered. Even lavas erupted under significant confining pressures of water often have melt
inclusions that preserve volatile abundances that appear appreciably less degassed than those
measured in host matrix glasses (Sobolev 1996; Newman et al. 2000; Saal et al. 2002).
Secondly melt inclusions typically host a wider diversity of melt compositions than are
represented by the bulk or matrix glass compositions of the host and associated lavas or tephras
from a given lava suite or location. This is an almost ubiquitous observation in basaltic rocks,
and is commonly interpreted to indicate that magmatic diversity in most basaltic systems
is significantly greater in extent than the range of lava compositions that are erupted (e.g.,
Nielsen et al. 1995; Sobolev 1996; Kent et al. 1999a; Slater et al. 2001), although alternate
interpretations are possible (e.g., Danyushevsky et al. 2004). The best explanation for this
phenomenon is that inclusion-bearing magmas represent mixtures of smaller volumes of melts
of variable compositions that are present within basaltic melt generation and melt transport
systems. Mixing and blending to produce the erupted bulk compositions reduces compositional
diversity. Melts trapped within melt inclusions that formed prior to or during mixing preserve
a greater range of compositions (Sobolev and Shimizu 1993; Nielsen et al. 1995; Sobolev
1996; Saal et al. 1998, 2005; Kent et al. 1999a, 2002a; Sours-Page et al. 1999; Norman et al.
2002). This makes intuitive sense when considering the volumes that individual lavas (up to
many cubic kilometers) and melt inclusions (typically a few hundred cubic microns) represent.
Importantly, melt inclusions may also trap compositions that are more primitive than erupted
magma or liquid compositions, and in some cases may provide access to melts that approach
primary magma compositions.
Variations in magma compositions within and between comagmatic suites are a primary
source of information used to study the origin and evolution of basaltic and related rocks.
Thus the inherently greater diversity of compositions evident in melt inclusions often allow
the processes that drive magmatic diversity to be examined with greater resolution than studies
based on bulk rock compositions, and melt inclusions may also preserve melts present in
low volumetric proportions that are not evident amongst the erupted bulk compositions. Melt
inclusion suites have been described which exhibit compositional variations that reflect all of
Melt Inclusions in Basaltic and Related Volcanic Rocks 275
the major drivers of compositional change in basaltic and related rocks, including variations in
mantle source composition and mineralogy and differences in the style and degree of melting,
mantle-melt interaction, crystal fractionation, mixing and assimilation, and degassing.
A third advantage is that melt inclusions trapped within resistant phenocryst phases
in rocks that have been altered by weathering, hydrothermal alteration or low temperature
metamorphism, may provide the only means to directly establish the composition of magmatic
liquids (e.g., McDonough and Ireland 1993; Shimizu et al. 2001; Kamenetsky et al. 2002),
particularly for elements that are mobile during alteration (e.g., volatiles, alkalis and alkali
earths). This is particularly valuable for study of ancient and altered lavas.
Despite these clear advantages, caution must also be exercised when interpreting data from
melt inclusion suites. Melt inclusions are typically small, with a restricted volume available for
analysis, and specialized techniques are required to obtain chemical information. Measured
compositions, particularly trace element abundances and isotopic compositions, often have
higher uncertainties than conventional analyses based on relatively large amounts of bulk rock
powder or glass, although this may be offset to some degree by greater chemical and isotopic
diversity. Of additional major concern for studies of melt inclusions are processes that result
in chemical modification of the inclusion away from the original trapped composition. In
many cases this may produce measured compositions that were never present as a discrete
melt within the host magma system. In fact it is emphasized that melt inclusions very rarely,
if ever, preserve the exact composition of the trapped melt. This is largely due to the effects
of crystallization of the host mineral from the melt after trapping (e.g., Roedder 1979;
Danyushevsky et al. 2000; Kress and Ghiorso 2004), but also because the potential exists for
inclusions to interact with their host minerals and external melts through diffusive exchange
(e.g., qin et al. 1992; Danyushevsky et al. 2000; Gaetani and Watson 2000, 2002; Cottrell et al.
2002; Spandler et al. 2007; Portnyagin et al. 2008) and/or may reflect chemical fractionation
during the formation of the inclusion (e.g., Faure and Schiano 2005; Baker 2008). In some
instances the effects of post-entrapment modification may be identified and reversed using
laboratory heating or numerical treatments. In other cases use of specific chemical parameters,
such as incompatible element abundances and ratios can minimize the effects of modification.
However there is also the possibility that irreversible changes to composition may occur in a
manner that cannot be determined or subsequently reversed. Thus the challenge is determining
whether the compositional variations evident in a particular suite of melt inclusions reflect
primary petrological variations or inclusion-specific processes. In the latter case the danger is
that inclusion compositions may be interpreted as petrological in origin and provide misleading
insight into the processes of mama generation and evolution.
Another question to be carefully considered is the degree to which melt inclusions are
representative of the magmatic systems in which they occur. The use of data from melt
inclusions to draw conclusions about the host magma requires an extrapolation across spatial
scales of ~1010 or more. The processes which lead to inclusion formation may not produce
inclusion populations that directly reflect the ranges of melt compositions present at the time
of trapping, and some compositions may be trapped in preference to others. Melt inclusions
may also trap melts of anomalous compositions produced by boundary layer trapping, mineral
dissolution and reaction and other localized processes, and that never existed in significant
quantities within the magmatic system (e.g., Danyushevsky et al. 2003, 2004; Yaxley et al.
2004; Faure and Schiano 2005; Baker 2008). There is also the possibility that sample selection
and preparation itself may promote a bias, unconscious or otherwise, towards specific
compositions—the most obvious being selection of inclusions large enough for chemical
analysis. Thus careful consideration needs to be given to establishing petrological and chemical
relations between the rocks and minerals that host inclusions and the inclusions themselves.
An important source of information concerning the origin and significance of melt inclusions
276 Kent
is the comparison between inclusion populations and the bulk or matrix glass composition of
the lava or tephra which host inclusions, as well as other associated lavas.
Previous reviews of melt inclusions in basaltic and related rocks
A number of reviews concerning aspects of melt inclusion studies are already available in
the literature (Sobolev and Kostyuk 1975; Roedder et al. 1979, 1984; Lowenstern 1995, 2003;
Sobolev 1996; Frezzotti 2001; Danyushevsky et al. 2002a; Schiano 2003) and it is not the inten-
tion of this study to repeat in detail material available in these. In particular, the reader is referred
to these works for more detailed information on the use of melt inclusions to obtain estimates
of the conditions of melt inclusion formation via physical measurements and experimental re-
heating (Sobolev and Kostyuk 1975; Roedder 1979, 1984; Danyushevsky et al. 2002a; Schiano
2003). Discussions of the measurement and use of volatile abundances in melt inclusions and
applications of melt inclusions in silicic systems are available in companion articles by Metrich
and Wallace (2008), Moore (2008) and Blundy and Cashman (2008). Tepley and Ramos (2008)
also deal with measurement of radiogenic isotope compositions in melt inclusions.
teChnIqueS uSed FOr Study OF MeLt InCLuSIOnS
Petrographic examination of melt inclusions
Petrographic examination of melt inclusions can provide considerable information and is
required before commencing any subsequent analytical protocol. In many cases where the host
mineral is translucent and relatively clear melt inclusions may be observed via examination
of the host mineral grains under a petrographic microscope at medium to high magnification
(typically ≥ 10× objective) in transmitted light. Every effort should be made to examine melt
inclusions within the host mineral, although this is not always possible. Refractive index oil
or ethanol may be used to increase the amount of light passing through the host mineral.
Observation of inclusions at adequate magnification to distinguish key petrographic features,
and distinction between melt inclusions and small mineral inclusions, is generally not possible
using binocular microscopes, even ones of high quality. In cases where the surface of the
host mineral is covered by thin layers of glass, groundmass or alteration products then use of
double polished thick sections (similar to those used for fluid inclusion studies) or polished
grain mounts can improve viewing by allowing more light to pass into the crystal interior.
An alternate approach is to lightly leach the outside of crystals with acid to remove surface
coatings that impede transmitted light and several workers use cold leaching in a commercially
available ~50% aqueous solution of fluoroboric (HBF4) acid for this purpose.
The use of regular 30 µm thick petrographic thin sections is not encouraged for identification
and documentation of melt inclusions, as any inclusion that is large enough for quantitative
analysis (≥ 30 µm) will intersect one or more surfaces of the section and thus will not be fully
available for viewing in three dimensions. Features of importance to interpretation may be
removed by polishing, and in some cases features which appear to be melt inclusions in thin
section or in a polished surface may turn out instead to be truncated apophyses, regions of
crystal intersection, or zones of crystal resorbtion. In addition, fractures that indicate connection
between inclusion and the external environment on the outside of the host crystal may also have
been removed.
The common petrographic features of melt inclusions have been discussed in previous
reviews (e.g., Sobolev and Kostyuk 1975; Roedder 1979, 1984; Frezzotti 2001; Danyushevsky
et al. 2002a; Schiano 2003). The primary influence on the petrographic appearance of melt
inclusions is the thermal history experienced by the host minerals, together with the dissolved
volatile contents and the relative timing of trapping and vapor exsolution. Figure 1 shows
examples of common petrographic features of melt inclusions from basaltic and related
Melt Inclusions in Basaltic and Related Volcanic Rocks 277
Figure 1. Photographs of melt inclusions. (A) Transmitted light photo of glassy melt inclusion in olivine
without bubble. A small fracture has nucleated at one end. (B) Backscattered electron (BSE) image of olivine
hosted melt inclusion containing clinopyroxene, plagioclase and glass as well as a bubble. (C) BSE photo
showing two olivine-hosted melt inclusions. One (to the right) consists of glass and a bubble and shows a
well-developed Fe-rich rim on olivine surrounding the inclusion, the inclusion on the left shows alteration
of glass around the rim with remnant glass in the center. (D) Olivine hosted inclusion in transmitted light
containing a mixture of fine dendritic clinopyroxene and glass and a bubble. (E) Polyhedral olivine crystal
in transmitted light showing multiple ovoid and elongate melt inclusions containing glass and shrinkage
bubbles. (F) Polyhedral olivine crystal in transmitted light showing multiple glassy and bubble free elongate
to round melt inclusions. Craters remaining after LA-ICP-MS analyses are also evident (G) BSE photo of
an anorthite-rich MORB plagioclase phenocryst showing large numbers of light colored melt inclusions,
many occurring within inclusion rich bands.
278 Kent
volcanic rocks. The most common of these are glass (the result of quenching of trapped silicate
liquid), daughter minerals (that nucleate and grow after trapping), minerals that are co-trapped
with the inclusion (these do not nucleate within the trapped liquid but may grow further after
trapping), and vapor saturation or shrinkage bubbles.
In rapidly cooled rocks, such as tephras or glassy submarine pillow rims, melt inclusions
can consist entirely of glass ± vapor or shrinkage bubble ± co-trapped mineral grains, and
daughter minerals will be absent (Fig. 1; e.g., Roedder 1979). However it is important to note
that even completely glassy inclusions have almost always still experienced crystallization of
the host mineral on the wall of the inclusion (the interface between inclusion and host mineral).
As the new mineral growth occurs in thin shells on the inclusion wall, mineral growth may be
apparent on compositional profiles or backscattered electron images (Fig. 1).
Inclusions that have experienced relatively slow cooling rates may contain daughter
mineral phases. These minerals often have skeletal or dendritic textures indicative of rapid
mineral growth (Fig 1), indicating that significant undercooling or supersaturation is required
for nucleation. Formation of daughter crystals will result in closed system chemical fractionation
between mineral phases and residual liquid and may result in stabilization of mineral phases and
one or more residual liquid compositions that are not found elsewhere in the magmatic system
under investigation (Frezotti 2001). This phenomenon has been used to study crystallization
and immiscibility within trapped liquids (e.g., Clocchiatti and Massare 1985; Jambon et al.
1992; Nielsen and Veksler 2002). The formation of glassy or microcrystalline inclusions in any
given sample during cooling may in part also reflect inclusion size, composition and volatile
abundances (e.g., Lowenstern 2003) and thus may vary significantly between inclusions, even
within the same sample.
Bubbles, typically representing small fractions of the volume of the inclusion are also
often found within glassy and crystalline melt inclusions (Fig. 1). Although often referred to as
“shrinkage bubbles” bubbles represent a separate phase formed by vapor exsolution (Roedder
1984), although the greater thermal expansion of glass relative to crystalline host may assist
in bubble formation during cooling. Studies, largely of silicic melt inclusions, suggest that
significant underpressure may be required to nucleate a bubble phase (e.g., Mangan and Sisson
2000), and bubble formation and relative size is also a function of inclusion size and cooling
history: smaller inclusions and/or those forming at faster cooling rates (Tait 1992; Lowenstern
1994) may not nucleate bubbles. In basaltic and related rocks where high-pressure vapor
exsolution produces a CO2-rich vapor phase (e.g., Dixon et al. 1995) bubbles in inclusions
trapped at higher pressures may only contain CO2 in significant amounts (e.g., Hauri 2002;
Metrich and Wallace 2008), but might also be potentially important hosts for noble gases.
Preparation of melt inclusions for analysis
Petrological and geochemical information from melt inclusions comes from a number of
sources, including petrographic and experimental observations and micro chemical analyses
(e.g., Roedder 1979, 1984; Sobolev 1996; Danyushevsky et al. 2002a). Many earlier studies of
melt inclusions focused on studying melt inclusion crystallization or obtaining estimates of the
conditions of trapping through heating experiments, and although not the focus of this review
(see Sobolev and Kostyuk 1975; Roedder 1979, 1984; Danyushevsky et al. 2002a; Schiano
2003), this approach continues to provide valuable information. Recent work has increasingly
focused the use of microanalytical techniques to determine elemental abundances and isotopic
compositions of material trapped within inclusions.
Experimental rehomogenization of melt inclusions. For determination of the chemical
or isotopic composition of the trapped melt it is almost always preferable to analyze a
homogeneous glass rather than a mixture of various crystalline phases and residual glass, and
slowly cooled melt inclusions containing mixtures of crystals and glass require reheating and
quenching prior to analysis. Most applicable analytical techniques are based on calibration of
Melt Inclusions in Basaltic and Related Volcanic Rocks 279
a specific response, such as emission of characteristic x-rays, infrared absorption, or ion yields
produced by ion beam sputtering or laser ablation, and this response may vary between different
crystalline and glass phases. There may also be specific problems with analysis of individual
phases, such as the rapid Na loss reported by Nielsen et al. (1995) in albitic quench crystals
in plagioclase-hosted MORB melt inclusions. In addition, for surface analysis techniques, the
proportions of solid phases evident at the sample surface may not be representative of the three
dimensional volume of the inclusion, and the same applies to non-glassy inclusions where some
portion may have been removed by sample preparation prior to analysis.
Homogenization of melt inclusions is generally accompanied in one of two general
ways (Nielsen et al. 1998; Danyushevsky et al. 2002a). Inclusions can be heated individually
and observed using a microscope heating stage (Roedder 1979, 1984; Danyushevsky et al.
2002a), with specially designed “Vernadsky” style stages also allowing rapid quenching (e.g.,
Danyushevsky et al. 2002). In this manner the temperatures of disappearance of various mineral
phases and vapor bubbles can be directly measured. Alternatively, inclusion-bearing crystals
may be heated in a furnace—typically a 1-atmosphere redox controlled furnace (e.g., Nielsen
et al. 1995, 1998), although high-pressure apparatus has also been used—to a pre-determined
temperature and quenched. Nielsen et al. (1998) and Danyushevsky et al. (2002a) discuss the
advantages and disadvantages of each approach. The use of a heating stage provides a direct
estimate of the trapping temperature of an inclusion (although this may be difficult to interpret;
Danyushevsky et al. 2002a; Massare et al. 2002) but is hampered by slower throughput, greater
sample preparation and increased potential for fracturing of the host crystal (Nielsen et al.
1998). Homogenization within a gas-mixing furnace provides no direct measurement of trapping
temperatures although an iterative approach combined with calculation of phase equilibria may
allow these to be constrained (Nielsen et al. 1995, 1998). In many cases the period of time
that inclusions are held at the estimated trapping temperature is kept to a minimum to avoid
diffusive loss of water during heating (see below), and if this is significant it can also have
an effect on the estimated temperature of trapping and measured redox state (Danyushevsky
et al. 2002a; Rowe et al. 2007). Alternatively, Gaetani and Watson (2000) advocate a longer
period of isothermal heating (up to 72 hours) following homogenization and characterization
of compositional gradients adjacent to inclusions to assess the degree of inclusion-host
equilibration. Where compositional gradients are absent Gaetani and Watson (2000) estimate
that the trapping temperature is constrained to within ~±20 °C. Due to rapid loss of water via
diffusive exchange this approach would only work for relatively dry melts.
Both gas mixing furnace and heating stage homogenization techniques also allow for rapid
quenching of heated inclusions, which is required to reduce quench growth of the host mineral
during cooling. Nielsen et al. (1995) showed that quench crystallization with plagioclase-hosted
melt inclusions resulted in lower measured Ca and Al contents at quench times of 3s, compared
to the faster quenches (~0.5 s) provided by the 1-atm furnace. Overall both homogenization
approaches are widely used for study of melt inclusions, although direct comparisons between
the two are relatively rare. Norman et al. (2002) treated inclusions from Koolau, Hawaii, using
both methods and report essentially unchanged results.
For any rehomogenized inclusion it is important to ensure that analyzed melt inclusions
have not “breached” or vented through cracks during heating. Contraction during cooling often
results in the development of fractures at the margins of melt inclusions (e.g., Tait 1992) and
if these extend to the outside of the host crystal they can provide a pathway for loss of volatile
elements during reheating. In addition, fractures can also allow low temperature hydrous
fluids to devitrify and chemically alter inclusions (Fig 1). Prior to reheating affected inclusions
can be recognized by the presence of alteration within the inclusion—often concentrated
at the margins of the inclusion or adjacent to fractures (e.g., Fig. 1C, 2). However, where
inclusions have been homogenized experimentally, alteration minerals are also remelted and
incorporated into the inclusion composition resulting in an apparently glassy inclusion of
280 Kent
modified composition (Nielsen et al. 1998). This is particularly a problem in samples that have
experienced significant submarine or subaerial weathering, and one example is also shown in
Figure 2. In addition, for species that are undersaturated in the inclusion, rehomogenization
can also result in incorporation of material from within the fracture and outside the host
crystal into the inclusion via diffusive transport along melt-filled fractures. This phenomenon
has been documented in MORB melt inclusions where MnO contents in excess of 10 wt%
have been measured and related to Mn derived from Mn-rich coatings formed on fractures
and on the external grain boundaries of the host crystals (Nielsen et al. 1998). Homogenized
inclusions that have been modified by alteration prior to rehomogenization can sometimes be
recognized petrographically as they may show cuspate or irregular margins resulting from
reaction between inclusion and host olivine (e.g., Nielsen et al. 1998; Danyushevsky et al.
2002a). Nielsen et al. (1998) also emphasized that anomalously low S and/or Cl occur in
inclusions that have vented during heating due to the low solubilities of these elements in
basaltic melts at atmospheric pressure. These inclusions are often disregarded (Fig. 2; Nielsen
et al. 1998, 2000; Sour-Page et al. 2002), although it should be noted that this might also result
Figure 2. (A), (B) and (C) Selected oxide
compositions measured across an altered
and unhomogenized melt inclusion (see
inset BSE photo) from Baffin Island,
Canada (Kent, A.J.R. unpublished data).
The composition of matrix glass (sample
10021; Kent et al. 2004b) is also shown.
Melt inclusions show significant alteration
and compositional change within altered
material adjacent to inclusion margins,
but no apparent deviation from expected
composition in the visually unaltered
central portion of the inclusion. (D)
Composition of rehomogenized melt
inclusions from Baffin Island show
low CaO contents in some inclusions
with low S contents, suggesting they
incorporate low Ca alteration products.
The range of S contents above detection
limits is probably the result of shallow
pressure degassing.
Melt Inclusions in Basaltic and Related Volcanic Rocks 281
in removal of pristine melt inclusion compositions trapped at very shallow pressures (Fig. 2).
Sulfur is particularly useful as it is typically present at relatively high concentrations in basaltic
melts and in many cases the pre-degassing minimum S contents can be estimated based on
measured FeO contents (e.g., Mathez 1976, Wallace and Carmichael 1992).
Mounting and polishing melt inclusions. Preparation of melt inclusions for geochemical
analysis typically requires mounting inclusion-bearing mineral grains in epoxy or other mounting
media and then polishing to expose inclusions. Some techniques require more specialized
techniques, such as the preparation of double polished wafers for some spectroscopic analyses,
although measurement of major and trace elements by a combination of EMP, SIMS and/or
LA-ICP-MS requires only that the inclusion be polished on one side to expose the inclusion.
Two approaches can be taken. Firstly workers can prepare grain mounts containing a number
of inclusion-bearing crystals and polish to an arbitrary level. All inclusions exposed can then be
documented and analyzed, and the process repeated if necessary. This is a relatively inefficient
use of material, as many inclusions hosted by the mounted grains may not be exposed at the
level of polishing, but where crystalline material is relatively plentiful and inclusions abundant
it is often sufficient. This approach is also less labor intensive and has the beneficial effect of
providing an essentially random sample of the available inclusions. The alternate approach
is to select specific grains that contain melt inclusions prior to polishing through inspection
with the petrographic microscope, and then each crystal is polished individually to expose the
inclusion. This is considerably more labor intensive but is a more efficient use of material where
sample and or inclusions are scarce. However there is also the question of bias, unconscious or
otherwise, by selection of individual inclusions and this may lead to overemphasis on certain
trapped compositions within a dataset (e.g., Danyushevsky et al. 2004).
Several methods for mounting individual inclusions are available, including the published
techniques of Thomas and Bodnar (2002) and Hauri et al. (2002). In addition, polishing using
crystalbondTM or other thermoplastic cement or mounting of individual crystals using epoxy in
small tubes cut from ¼″ metal pipe are also used. More recently workers planning to use SIMS
for volatile and other analyses are mounting crystals using indium metal in preference to epoxy
to produce mounts with low vacuum backgrounds (e.g., Saal et al. 2002).
Chemical analysis of melt inclusions
A primary requirement for analysis of individual melt inclusions is that analytical
techniques are capable of providing information from small sample volumes. This largely
restricts analysis of melt inclusions to microbeam-based approaches using focused beams
of ions, photons, electrons, protons or x-rays, although dissolution of individual crystals has
also been used for measurement of radiogenic isotope compositions within melt inclusions
(Tepley and Ramos 2008). Because of these limitations, interest in geochemical studies of melt
inclusions has been strongly tied to technical advances in microchemical analysis techniques.
Development and widespread application of electron microprobe techniques enabled earlier
studies (e.g., Anderson and Wright 1972; Anderson 1974; Sobolev and Kostyuk 1975; Sobolev
et al. 1983a,b and others), and application of spectroscopic techniques for measurement of
water and CO2 allowed the first measurements of H2O and CO2 contents (e.g., Anderson 1974).
Subsequent development of the ion microprobe for measuring volatiles and trace elements, and
more recently, laser ablation ICP-MS for in situ trace element analyses, allowed application of
trace element techniques to the study of melt inclusions, and the integration of melt inclusion
data sets with the large amount of data already available for whole rock and glass samples.
These approaches also now offer the means to make in situ measurements of some stable and
radiogenic isotope systems.
In the following we briefly discuss the most common of these: electron microprobe
(EMP) analysis, Secondary Ion Mass Spectrometry (SIMS) and laser ablation inductively
282 Kent
coupled mass spectrometry (LA-ICP-MS). Fourier Transform Infrared spectroscopy (FTIR)
is also commonly used to determine H2O and CO2 abundance and speciation, but is discussed
elsewhere in this volume (Metrich and Wallace 2008; Moore 2008), and other spectroscopic
techniques have also been used to measure composition or speciation. The purpose is not to
exhaustively review these techniques but to focus on the specific application to melt inclusions.
In addition a number of other techniques are not yet in wide usage, but hold considerable
promise for future basaltic melt inclusions studies. Examples are synchrotron-based methods
to determine element speciation and volatile contents (e.g., Bonnin-Mosbah et al. 2002;
Sutton et al. 2005; Metrich et al. 2006; Seaman et al. 2006), and Laser Raman spectroscopy to
investigate the volatile contents of unexposed melt inclusions (Thomas et al. 2006).
An important point to note is that melt inclusions require analysis of a restricted volume
with limited number of atoms. For trace elements the total amount of material available for
analysis within individual inclusions is very limited. A single 100 µm diameter melt inclusion
contains just a few picograms of a trace element present at 1 µg/g concentration. Even with
sensitive analytical techniques the uncertainties associated with melt inclusion analyses are
often worse than that obtained from conventional analyses of bulk rock samples—where the
number of atoms available for analysis is typically much greater. However, in accord with the
general inverse relation between compositional diversity and analytical volume evident in many
natural materials, the poorer precision of measurements is often compensated for by the greater
range of measured element and isotopic compositions observed in melt inclusion suites.
Electron Microprobe Analysis (EMP). The most widely used technique for study of chem-
ical composition of melt inclusions is the electron microprobe, which provides measurements
of major and minor element compositions (typically Si, Ti, Al, Cr, Fe, Mn, Mg, Ca, Ni, Na, K,
P) as well as the volatile elements F, Cl and S (e.g., Nielsen et al. 1998; Davis et al. 2003; Straub
and Layne 2003; Blundy and Cashman 2008), and may also be used to measure the oxidation
state of S through the SKα peak scan technique (e.g., Wallace and Carmichael 1994; Rowe et al.
2007). The electron microprobe also provides the user with additional means to examine inclu-
sions using backscattered electron and compositional imaging and is indispensable as a means
to determine the composition of the mineral host using standard mineral analysis protocols.
The spatial resolution of most EMP analyses of basaltic glass is < 10-20 µm, and so
the relatively small size of most melt inclusions typically place no extra restrictions on
analytical protocols, and for many melt inclusions even allows for multiple measurements
of each inclusion to test homogeneity and analytical reproducibility (Nielsen et al. 1998).
Electron probe measurements of melt inclusions are thus directly comparable to routine
measurements of basaltic glass and precision and accuracy is limited by the quality of the
instrumental calibration, control of sample devolatilization, the duration of analysis, and the
abundance of each element. Significant redistribution of volatile elements (H, Li, Na) in glass
during EMP analysis using standard beam conditions has been shown to occur in felsic melt
inclusions (Humphreys et al. 2006). Sodium redistribution is also known to occur in basaltic
glasses although mobility of Li and H has not been demonstrated in mafic samples, which are
generally less prone to electron beam damage. Advances in techniques for direct measurement
of oxygen and correction for the effects of electron beam damage also promise improvements
in the ability to estimate H2O contents in melt inclusions and other glasses by difference
(Metrich and Wallace 2008).
Most elements present at concentrations of >~1 wt% can be measured with accuracy and
precision << 5%. Elements typically present in lower abundances (Cr, Mn, Ni, K, P, Cl, S, F)
may be measured less precisely, with uncertainties of up to ±10-50% depending on abundance.
Analysis of these minor species may also require longer count times, which slows analytical
throughout. Analysis of Cl and S, often present at abundances of << 0.1% may require total
analysis times as long as 10 minutes (Nielsen et al. 1998; Kent et al. 1999a; Danyushevsky
Melt Inclusions in Basaltic and Related Volcanic Rocks 283
et al. 2003; Davis et al. 2003) to obtain sufficient precision. Analysis of F may also require
corrections for spectral overlap from Fe (e.g., Davis et al. 2003; Straub and Layne 2003).
Several workers have also exploited the ability of EMP analysis to study the speciation
of sulfur trapped within melt inclusions from basaltic rocks. This technique has been widely
applied to volcanic glasses (Wallace and Carmichael 1994) but is also suited to melt inclusions
(Rowe et al. 2007). Further discussion of the technique may be found in the chapter by Metrich
and Wallace in this volume (2008). Fine shifts in the wavelength of the SKα x-ray are known
to correlate with S speciation, generally characterized as S6+/∑S, which in turn is a function
of the oxygen fugacity, composition and pressure and temperature of entrapment (Wallace
and Carmichael 1994). Rowe et al. (2007) demonstrated that melt inclusions could be used to
provide an estimate of oxidation state of basaltic melt trapped in inclusions, although the effects
of natural degassing, host mineral crystallization, overheating during re-homogenization and
natural Fe-loss (Danyushevsky et al. 2000) can also promote significant changes in oxygen
fugacity and sulfur speciation. Perhaps the most important potential change is that related to
diffusive loss (or gain) of H+ via equilibration between trapped melt and the host or external
ambient melt. Basaltic melts are potentially very sensitive to this process (see below).
Secondary Ion Mass Spectrometry (SIMS). Analyses of lithophile trace elements
(typically one or more of the following elements: Li, Be, B, Sc, Ti, V, Mn, Rb, Sr, Y, zr, Nb,
Cs, Ba, rare earth elements (REE), Hf, Ta, Pb, Th, U), present in concentrations of <1 to >
1000 µg/g, in melt inclusions from silicic and basaltic rocks were first conducted by secondary
ion mass spectrometry (SIMS) from the 1980’s and early-mid 1990’s (e.g., Dunbar et al. 1992;
McDonough and Ireland 1993; Sobolev and Shimizu 1993; Sobolev and Danyushevsky 1994;
Gurenko and Chaussidon 1995; Nielsen et al. 1995; Sobolev 1996; Shimizu 1998). The advent
of reliable techniques for measurement of trace element abundances using SIMS and other
techniques was an important driving force for expanded interest in the study of melt inclusions
during this time. Although laser ablation ICP-MS (LA-ICP-MS) has also become a widespread
method of trace element analyses of inclusions (see below) SIMS analysis is still routinely
used for melt inclusion studies and in addition to trace elements also offers the possibility
for measurements of volatile element (H, C, F, S, Cl), stable isotope (H, Li, B, O, C, S, Cl)
signatures and Pb isotope compositions.
General reviews and discussion of SIMS analysis, including applications of SIMS to
analysis of trace element abundances in silicate minerals and glasses, can be found in Shimizu
and Hart (1982) and Hinton (1995). SIMS analysis uses a focused primary ion beam, generally
16 −
O , for determination of lithophile trace element abundances as positive secondary ions or
133
Cs+ for analysis of volatile elements and some light stable isotopes as negative secondary
ions. The primary beam, focused to a spot size of 15-30 µm, is used to sputter material from
the melt inclusion or other region of analytical interest. A small proportion (generally < 1%)
of the sputtered atoms are ionized during the collision process and accelerated into a double-
focusing mass spectrometer where count rates at mass peaks corresponding to ions of interest
can be measured using electron multipliers or Faraday Cup detectors. The physics of the ion
sputtering and extraction process is not well understood and elemental concentrations are
typically calculated by reference to matrix matched standards of known composition analyzed
during the same analytical session (Shimizu and Hart 1982; Hinton 1995).
The concentration of unknown elements can be calculated from the following:
CiR
CR IU
CiU = CsU R S ij U (1)
I ij I Sk
IR
Sk
284 Kent
Where:
CUi = the concentration of trace element i in unknown material U
CUS = the concentration of the internal standard element S in the unknown material U.
CRi/ CRS = the ratio of the known concentrations of trace element i to the internal standard
element S in reference standard R.
IRij/IRSk = normalized ion yield (the ratio of measured ion beam intensities) for isotope j
of trace element i and isotope k of the internal standard element S in the reference
standard
IUij /IUSk = normalized ion yield for isotope j of trace element i and isotope k of the internal
standard element S in the unknown material
If multiple standards are used to generate a calibration then the term (CRi/CRS)/(IRij/IRSk) can be
replaced by the slope of the line defined by multiple standards on an x-y plot of IRij/IRSk vs. CRi/
CRS (as long as this line projects through the origin).
The internal standard element is typically a major element that is present in known quantity
(generally taken from electron microprobe measurements) and that has a minor isotope suitable
for analysis. Examples include 29Si, 30Si, 42Ca, 43Ca, and 47Ti. The use of an internal standard
corrects for changes in ion yield related to instrumental variations and also due to the geometry
of analyzed inclusions. Application of Equation (1) requires the following condition be met: (i)
that the variation between measured normalized ion intensities (IUij/IUSk) and CUi/CUS is linear
and that when CUi is zero that IUij/IUSk is zero (i.e., that there is no background or that this
has been corrected for prior to calculation of IUij/IUSk); (ii) a well-characterized homogenous
standard material is available for calibration; and (iii) that there are no differences in relative
ion yield between the isotopes of interest and the internal standard between the standard and
unknown materials. The latter phenomenon is generally termed a matrix effect. The adherence
to these requirements can be tested using secondary standards and fortunately for basaltic
melt inclusions there are a number of well-characterized basaltic composition glass standards
available (see below).
Secondary ions produced by the sputtering process often have a large range of initial kinetic
energies, and an electrostatic analyzer (ESA) and slit located between the ESA and magnet is
used to preclude entrance of ions with aberrant energies or trajectories into the magnet. In
addition, a small offset in the accelerating voltage is also used to select ions with a specified
energy range. Known as energy filtering, for trace element analysis this technique involves
selecting ions with an energy range centered on +60-100 eV with a slit width equivalent to
±10-30 eV. Although this reduces ion beam intensities the technique is critical for trace element
analyses using the smaller Cameca f-series ion probes, as selection of secondary ions with
significant initial kinetic energy both diminishes peak interferences from ionized molecular
species (particularly metal oxides M16O+) without resorting to higher mass resolving power
(and concomitant loss in transmission) and minimizes possible matrix effects. In some cases
energy filtering is not sufficient to reduce molecular interferences and this can limit the analysis
of certain elements or requires additional correction techniques such as peak stripping. For this
reason SIMS analysis of some transition metals (Sc, V, Ni, Cu, zn) and Rb can be problematic
(Hervig et al. 2006). Another important example is the requirement for a matrix-based correction
procedure for removing Ba and light REEO+ peaks from middle REE and middle REEO+
contributions from heavy REE (zinner and Crozaz 1986; Fahey et al. 1987).
One important feature of SIMS analysis is that ion sputtering rates are relatively low and
as a result, trace element measurements by SIMS typically takes 30-90 minutes per analysis.
However, in combination with the relatively high useful yield of the technique (defined as
the proportion of atoms of a given isotope removed from the analytical volume that produce
Melt Inclusions in Basaltic and Related Volcanic Rocks 285
detectable ions during analysis; Fig. 3; Hervig et al. 2006) slow sputter rates result in relatively
good preservation of inclusion material and for larger inclusions there is often sufficient
material remaining for additional analyses. This can be a tremendous advantage when multiple
analyses are planned for individual inclusions—such as measurements of trace elements,
volatile abundances and/or stable isotope compositions, and is one of the primary distinctions
between SIMS and LA-ICP-MS (see below). Detection limits for SIMS trace element analyses
depend on the instrument used, the size and location of the energy bandwidth chosen for
filtering (Fig. 3A) and the primary beam intensity, but are typically low, in the ng/g to µg/g
range and sufficient for trace element measurements even in incompatible element depleted
basalt compositions (Fig. 3A; Hervig et al. 2006). Analytical precision is largely a function
of signal intensity via the Poisson relationship (the standard deviation s = 1/√n, where n is the
number of counts obtained for a given isotope) and thus varies with elemental and isotopic
abundance. Accuracy is controlled by the quality of the calibration procedure, including the
degree to which standards are homogenous and well characterized. Overall, both precision and
accuracy is generally 5-10% relative for elements present at > 1-10 µg/g abundance and 10-
40% for elements present at lower abundances (e.g., Sobolev 1996; Shimizu 1998; Slater et al.
2001; Kent and Elliott 2002; Kent et al. 2002a; Portnyagin et al. 2007; Sadofsky et al. 2008).
SIMS can also be used for measuring abundance of volatile elements in melt inclusions and
other volcanic glasses (e.g., Sobolev 1996; Sobolev and Chaussidon 1996; Kent et al. 1999a;
Hauri et al. 2002; Lassiter et al. 2002; Gurenko et al. 2005; Benjamin et al. 2007; Portnyagin
et al. 2007; Sadofsky et al. 2008; Blundy and Cashman 2008). Recent developments of volatile
measurements using negative secondary ions have lowered detection limits and enabled
precise measurement of a full suite of volatile elements (H, C, F, S, Cl) in a single analysis
(Hauri 2002; Hauri et al. 2002; Saal et al. 2002). Further discussion of volatile measurement
techniques may be found in the chapter by Blundy and Cashman (2008).
Laser Ablation Inductively-Coupled Mass Spectrometry. Laser ablation has become a
powerful and widely used technique for in situ microsampling of solid materials in the Earth
and other sciences (Durrant 1999), and provides a sensitive means for measurement of trace
element abundances in small sample volumes in silicate minerals and glasses. Reviews and
discussion of relevant aspects of the LA-ICP-MS technique and ICP-MS mass spectrometry
may be found in Perkins and Pearce (1995), Niu and Houk (1996), Eggins et al. (1998), Russo
et al. (2002), Jackson and Gunther (2003); Hattendorf et al. (2003) and Heinrich et al. (2003).
Applications of LA-ICP-MS for analysis of silicate melt inclusion has also been discussed
by Taylor et al. (1997) and for fluid and melt inclusions by Halter et al. (2002), Heinrich et
al. (2003), and Pettke et al. (2004), and is now a widely applied technique (e.g., Kamenetsky
et al. 1997; Norman et al. 2002; Danyushevsky et al. 2003; Yaxley et al. 2004; Rowe et al.
2006; Kohut et al. 2006; Benjmain et al. 2007; Elburg et al. 2007; Wade et al. 2007; Johnson
et al. 2008; Vigouroux et al. 2008). LA-ICP-MS can provide information on a similar suite of
trace elements as SIMS, including all commonly used lithophile trace elements, but provides
increased access to transition metals and also, via specialized analyses, to select platinum
group and chalcophile elements (e.g., Sun et al. 2003). LA-ICP-MS has also been used to
measure Sr and Pb isotope compositions (see below).
Analyses of trace element abundances by LA-ICP-MS in melt inclusions and other
materials use a pulsed laser to ablate selected sample regions. Modern laser ablation systems of
266 nm, or more commonly, 213 or 193 nm wavelength allow focusing of the laser to a small
spot (generally from 10-100 µm) on a selected location with a homogenous energy profile
across the spot, enabling ablation of a flat bottomed circular crater (e.g., Eggins et al. 1998).
Interaction between each laser pulse and the sample surface results in localized destruction of
chemical bonds, either via thermal interaction or more advantageously via a “phase explosion”
and ejection of material as a jet of plasma and particulates above the sample surface (Russo et
286 Kent
Figure 3. Detection limits and useful yields (defined as the proportion of atoms within a given analyzed
volume that produce detectable ions during analysis) detected during analysis) for trace elements measured
by SIMS and LA-ICP-MS analyses. Detection limits and useful yields for SIMS are from Hervig et al.
(2006) and for LA-ICP-MS from Kent, A.J.R. unpublished data, calculated from the methods of Perkins
and Pearce (1995). (A) Detection limits for SIMS, showing data for both Cameca 6f and 3f instruments
using a 10 or 20 nA 16O- primary beam and secondary energy offset of 95 ± 20V (Hervig et al. 2006). (B)
Detection limits for LA-ICP-MS using three spot sizes (25, 40 and 80 µm diameter) and two pulse rates (2
and 4 Hz). The composition of depleted basalt standard glass BIR-1G is shown as a representative depleted
basaltic composition for comparison. In general both SIMS and LA-ICP-MS have sufficiently low detection
limits for trace element analyses in melt inclusions, even in depleted melt compositions, although specific
elements can be problematic (Pb for SIMS; Be, B, Rb for LA-ICP-MS). (C) Calculated useful yields for
selected trace elements for SIMS using energy filtering (for Cameca 3f and 6f) and LA-ICP-MS.
Melt Inclusions in Basaltic and Related Volcanic Rocks 287
al. 2002). UV laser systems, particularly those operating at 213 or, even better, at 193 nm are
preferred as they produce finer particulate, resulting in higher sensitivity and lesser elemental
fractionation associated with incomplete breakdown of ablated particulate within the plasma
furnace (e.g., Eggins et al. 1998; Gunther and Heinrich 1999; Jackson and Gunther 2003).
Ablation is carried out within a closed cell with continuous flow of inert gas (Ar, or more
commonly He; Eggins et al. 1998) sweeping ablated particulates into the ICP furnace where
they are rapidly atomized and ionized prior to entering the high vacuum mass analyzer (Niu and
Houk 1996; Gunther and Heinrich 1999). Rapidly scanning quadrupole or sector field ICP-MS
mass analyzers are used to serially monitor selected isotopes from the elements of interest with
short dwell times for individual masses (Longerich et al. 1996). ICP ion sources are capable of
efficiently ionizing the majority of elements in the periodic table (with the exception of some
halogens; e.g., Perkins and Pearce 1995) and thus a large number of elements are potentially
available for analysis.
As with SIMS, the physics of laser-solid interaction and particulate transport are incompletely
understood (e.g., Russo et al. 2002) and elemental abundances are calculated relative to analyses
of well characterized standard glasses, using internally-standardized techniques similar to those
described above for SIMS. The same Equation (1) is used to calculated unknown concentrations,
together with analysis of standard glasses for calibration, with similar assumptions as described
for SIMS. The most common choice of internal standard for LA-ICP-MS analysis of basaltic
melt inclusions is Ca, although Ti, Mg and Si have also been used. Calibration of LA-ICP-MS
analyses are best when the internal standard and analyzed trace elements have similar volatility,
and Ca, Mg, Ti and to a lesser extent Si, are relatively refractory, as are most common lithophile
trace elements (e.g., Li, Sr, Y, zr, Nb, Ba, REE, Hf, Th and U). More volatile elements calibrated
using these internal standards may be more problematic and include Pb, Cs, Rb, zn, P and B.
However it is also emphasized that use of He as the sweep gas, along with deep UV wavelength
ablation systems with homogenized energy distributions, minimizes elemental fractionation,
often to levels that are similar or less than overall measurement precision (<5%; e.g., Eggins
et al. 1998; Heinrich et al. 2003). Production of complex molecular species including metal
oxides is also more limited (e.g., Kent and Ungerer 2006) and can be largely controlled by
adjustment of plasma conditions. Background count rates are higher for LA-ICP-MS and are
generally corrected for before to calculation of normalized ion ratios by subtraction of count
rates measured before ablation starts.
Overall internal precision for LA-ICP-MS analyses of trace elements is again largely a
product of the number of counts recorded for each isotope during analysis via the Poisson
relationship and precision is inversely proportional to elemental and isotopic abundances. For
larger inclusions and analyses of matrix glasses it is often possible to use a larger spot size and
further improve precision. Ablation rates depend on the bulk composition of the material and
ablation parameters, but are generally in the order of 0.1-0.2 µm/pulse in most silicates (e.g.,
Eggins et al. 1998; Paul et al. 2005), allowing most inclusions of ≥30 µm diameter or more to
be analyzed in ~30-60 seconds of ablation at pulse rates of 2-5 Hz. At these conditions detection
limits are typically in the low µg/g to ng/g range and are typically sufficient for measurement of
depleted basalt compositions (Fig. 3A).
Most analyses reported in the literature are done on inclusions that are glassy (either through
natural quenching or rehomogenization) and polished to expose the inclusion, and where major
element compositions have been determined via prior electron microprobe analysis. In this way
the protocols used for analysis of inclusions are similar to those used for other silicate glasses
and minerals, except the size of the inclusion may limit the diameter of the laser spot used. The
smallest inclusions analyzed by LA-ICP-MS are typically in the 30-50 µm range. Analyses of
this type typically report precisions and accuracy that are better than 10-15% at the 2s limit (e.g.,
Fig. 4), with detection limits in the low µg/g to ng/g levels (e.g., Kent and Ungerer 2005).
288 Kent
Figure 4. Variations in observed standard deviations calculated from three repeat measurements of selected
trace element abundances in NIST 612 glass (measured using the isotopes shown in the legend) as a
function of laser spot size. Measurements used a laser pulse rate of 3 Hz.
Halter et al. (2002), Heinrich et al. (2003), and Pettke et al. (2004) describe an alternate
method, adapted from techniques developed for analysis of fluid inclusions, where the laser
is used to drill vertically through the host crystal to completely remove unexposed inclusions
in a single ablated volume. This approach involves ablating considerable amounts of the host
mineral in addition to the inclusion, and contributions from this must be accounted for (Halter et
al. 2002). Isotopes corresponding to all major and trace elements of interest are monitored and
LA-ICP-MS is thus used to determine both major element and trace element concentrations.
Although useful for reconnaissance, this approach is not widely used for petrological studies of
melt inclusions, as it results in complete destruction of the inclusion, and an alternate means are
required to estimate the composition of the internal standardizing element, which can introduce
additional uncertainty (Halter et al. 2002). There is also no provision for determination of the
abundances of volatile elements (Cl, S, H, C, F) in the destroyed inclusions, and although
measured incompatible trace element abundances by this technique can have precision and
accuracy that are similar to conventional LA-ICP-MS analyses (Pettke et al. 2003), analyses
of compatible elements will generally contain an additional component of uncertainty related
to the correction for host mineral within the ablated volume. Measurements reported to date of
major element abundances (Heinrich et al. 2003; Pettke et al. 2004) are also less precise than
those routinely obtained by electron microprobe and also limit petrological applications of the
obtained data.
Comparison between SIMS and LA-ICP-MS for trace element analysis. SIMS and LA-
ICP-MS are currently the two most commonly applied techniques for measuring trace element
abundances in melt inclusions. The two techniques are based on quite different physical principles
and instrumentation but rely on almost identical methods for calibration and calculation of
trace element abundances. The accuracy of both techniques is limited by the quality of this
calibration and adherence to the assumptions listed above. Precision is largely a factor of signal
size during analysis—which primarily reflects element and isotopic abundances. At present
the same standard glasses are widely used in SIMS and LA-ICP-MS laboratories (examples
include the NIST 610 and 612 glasses, MPI-DING glass set and the BCR-2G, BHVO-2G and
BIR-2G glasses from the USGS; e.g., Pearce et al. 1997; Rocholl 1998; Rocholl et al. 2000;
Raczek et al. 2001; Jochum et al. 2000, 2005a, 2006; Kent et al. 2004a). For this reason trace
element data produced by SIMS and LA-ICP-MS should be comparable, and this is borne
out by the few comparative studies available (e.g., Kent et al. 2004b; Pettke et al. 2005). An
additional comparison is shown in Figure 5 where we show the results of analysis of a suite of
melt inclusions hosted in olivine by SIMS (data from Kent et al. 2002a) and LA-ICP-MS. The
results are similar to other studies where direct comparisons have been made between SIMS
Melt Inclusions in Basaltic and Related Volcanic Rocks 289
and LA-ICP-MS trace element analysis of
volcanic glasses, and in general there is
good agreement within ± 10-15% across
a wide range of concentrations. In most
cases these differences are well within
the combined uncertainties of the two
techniques.
The primary advantage of SIMS over
LA-ICP-MS for trace element analyses
is that the higher useful yield of the
technique results in removal of relatively
small amounts of material per analysis
(Fig. 3; Hervig et al. 2006). SIMS useful
yields for lithophile trace elements range
between ~10-3-10-4 for f-series Cameca
ion probes using energy filtering (Hervig
et al. 2006). A typical analysis using a
10 nA primary 16O- beam focused to a
25 µm diameter spot will sputter a crater
from 2-5 µm deep during a typical 30-90
minute analysis. Except where inclusions
are at the smaller end of the analyzable
range (20-30 µm), this can leave material
remaining for further work—although
repolishing may be required. Useful yields
for LA-ICP-MS are typically 1-2 orders
of magnitude lower than SIMS (Fig. 4).
Coupled with faster sample removal rates
during ablation (Paul et al. 2005; Kent and
Ungerer 2006), this results in removal of
significantly more material during LA-
ICP-MS trace element analyses, often
limiting options for further analysis. For
example a 30-50 second ablation analysis
using a laser pulsed at 3-4 Hz will excavate
a crater of ~15-40 µm depth and can often
completely obliterate smaller inclusions
(< ~50 µm).
The primary advantage of LA-ICP-
MS over SIMS is that analysis times
are significantly shorter—typically 2
Figure 5. Comparison between trace element abun-
dances measured in the same olivine hosted melt in- minutes per individual analysis, allowing
clusions by SIMS and LA-ICP-MS. SIMS data are for significantly higher throughput
from Kent et al. (2002a) and subsequent LA-ICP-MS during laboratory analysis. Production of
analyses followed techniques outlined in Kent et al. molecular species are also much lower and
(2004b). Solid line shows 1:1 relation and dashed lines rates of ionization are higher during LA-
show ±10% deviation from this.
ICP-MS analysis (e.g., Kent and Ungerer
2005) and thus no special processing
techniques are required for REE analysis (c.f. zinner and Crozaz 1986; Fahey et al. 1987) and
many transition and chalcophile metals are simpler to measure.
290 Kent
It is also worth noting that analyses of incompatible elements in melt inclusions by SIMS
or LA-ICP-MS are both relatively insensitive to accidental incorporation of the host mineral
into the analyzed volume. This is particularly the case where the internal standardizing isotope
is also much higher in concentration in the inclusions than in the host mineral (e.g., 42Ca, 43Ca or
47
Ti in olivine hosted inclusions). As shown in Figure 6 incorporation of up to 60-80% or more
of the host mineral into the volume sputtered or abated for analysis will still produce less than
10-15% change in the measured concentration of moderately to highly incompatible elements
(KD ≤ 0.1-0.01). The effect is understandably the reverse for compatible elements, which are
very sensitive to incorporation of the host—addition of only a few percent will cause large
increases in the concentrations measured in the analyzed volume. Where contributions from the
host mineral are significant they can be subtracted from the measured total via comparison to
an analysis of the host mineral and calculation of relative ablated masses of host and inclusion
(e.g., Halter et al. 2002), although this may introduce additional uncertainties.
Isotopic compositions of melt inclusions. Both SIMS and LA-ICP-MS offer the ability
to determine some radiogenic isotope compositions of melt inclusions, as well as for SIMS
measurement of some stable isotope systems (H, Li, B, C, O, S, Cl). As with trace element
measurements in situ isotopic analyses generally have worse precision than for accepted
solution-based isotopic measurement techniques, but this is again is often compensated for
by relatively large isotopic variations observed within individual inclusion suites. Work on
radiogenic isotope systems to date has largely concentrated on measurements of Pb isotope
composition. For SIMS these measurements are made using large magnet instruments
capable of high transmission such as the Cameca 1270/1280 series. Saal et al. (1998, 2005)
reported measurements of 208Pb/206Pb and 207Pb/206Pb in relatively Pb-rich melt inclusions
from several OIB suites and other studies have also used a similar approach to measure
Pb isotope compositions from other localities (e.g., Kobayashi et al. 2004; Yurimoto et al.
2004). Uncertainties in 208Pb/206Pb and 207Pb/206Pb ratios are generally in the range of 0.2-
1.2%, depending on signal intensity. These studies have documented large variations in the
Pb isotope composition of melt inclusions from single lavas and lava suites, and have been
interpreted to represent highly heterogeneous Pb isotope compositions within the mantle
Figure 6. Calculated percent contributions to measured trace element concentrations in melt inclusions
resulting from incorporation of the mineral host into the analyzed volume. Contributions are shown for
a range of mineral-melt KD. Grey dashed box shows the range of 5-15% uncertainties typical for trace
element analyses by SIMS or LA-ICP-MS.
Melt Inclusions in Basaltic and Related Volcanic Rocks 291
source of individual lavas, or to record interaction between melts and mantle lithosphere
(Saal et al. 1998, 2005; Kobayashi et al. 2004; Yurimoto et al. 2004). Analyses of the low
abundance isotope, 204Pb may only be feasible in inclusions with sufficiently high Pb contents
(e.g., Saal et al. 2005), and so are rarely reported in the literature. Much of the variation in the
Pb isotope system is described by ratios of radiogenic isotopes to 204Pb, and thus this remains
a limitation. Application of multicollector SIMS instruments and improvements in sensitivity
promise improvements in measurement of 204Pb. LA-ICP-MS instruments using sector field
magnets have also been used to measure Pb isotope compositions of melt inclusions, either
using multicollector arrays (Paul et al. 2006) or via rapid electrostatic sequential scanning of
206
Pb-207Pb-208Pb abundances measured by electron multiplier to overcome the relatively short
signal duration during ablation of individual inclusions (Jochum et al. 2006). For the latter
approach the low abundance of 204Pb as well as dynamic range issues precludes measurement
of this isotope. Paul (2006) reports measurements made by laser ablation multicollector
ICP-MS of Pb isotope composition, including 204Pb, from melt inclusions from ocean island
basalt samples from Mangaia and Pitcairn islands as well as a subduction-related basalt from
Vanuatu. The range of reported Pb isotope compositions is significantly less than that reported
for Mangaia inclusions by Saal et al. (1998) using SIMS.
LA-ICP-MS has also been used to measure Sr isotope compositions of individual melt
inclusions, again using multicollector instruments (Jackson and Hart 2006). This measurement
requires large spot sizes, and correspondingly large melt inclusions in order to generate
sufficient signal during ablation. In addition, as melt inclusions typically include significant Rb
a correction is required to remove the interference of 87Rb from 87Sr, which cannot be resolved
using current instruments. The best precisions are generally achieved in inclusions with the
lowest Rb/Sr. As a result reported 87Sr/86Sr ratios have greater uncertainty than those made by
solution-based methods with Jackson and Hart (2006) reporting a typical uncertainty of ± 320
parts per million, although reported variations from Samoan olivine-hosted inclusions are
considerably larger than this. An alternative that is also promising is to use the microdrilling,
chemical processing and solution analysis techniques routinely used for study of isotopic
variations in phenocrysts (e.g., Ramos and Tepley 2008) to study individual melt inclusions
(e.g., Harlou et al. 2005).
SIMS instruments have also been used for measurements of stable isotope compositions
of melt inclusions. Measurements of δD, δ7Li, δ11B, δ13C, δ18O and δ34S, have been reported in
basaltic melt inclusions, with correction for instrumental mass fractionation typically made by
comparison to analyses of basaltic glass standards of known isotopic composition (e.g., Hauri
2002; Hauri et al. 2002; Kobayashi et al. 2004). Measurements of δ37Cl have also been made
by SIMS and are applicable to melt inclusions (e.g., Layne et al. 2004). Both large and small
format SIMS instruments have been used for stable isotope measurements in melt inclusions,
although the greater degree of transmission in the large format instruments is an inherent
advantage where high mass resolving power is required and/or overall precision is limited by
counting statistics on minor isotopes. Analyses of stable isotope compositions in melt inclusions
using SIMS are again more uncertain than those made by conventional mass spectrometric
techniques, with internal precisions of ~1-5 per mil reported for H, Li, B, C, S and Cl (e.g.,
Gurenko and Chaussidon 1997, 2001; Rose et al. 2001; Hauri et al. 2002; Kobayashi et al.
2004) and 0.1-1.5 per mil for O (Gurenko et al. 2001; Gurenko and Chaussidon 2002), with
precision dependent on abundance of individual elements in specific inclusions (e.g., Hauri
et al. 2002). Stable isotope analyses of melt inclusions often show variations well in excess
of these uncertainties, and consequently provide important information about the origin of
basaltic melt inclusions, including insight into the roles of the role of degassing, assimilation
and mantle source composition in basaltic magma systems (Gurenko and Chaussidon 1997,
2002; Gurenko et al. 2001; Rose et al. 2001; Hauri 2002; Kobayashi et al. 2004).
292 Kent
FOrMatIOn and eVOLutIOn OF MeLt InCLuSIOnS
Melt inclusion host minerals
The most common mineral hosts used for the study of melt inclusions, in approximate order
of importance in basaltic and related rocks, are olivine, plagioclase, spinel, clinopyroxene and
orthopyroxene. Olivine is often preferred as it is an early-formed mineral, together with spinel
in many basaltic and related magmas, preserving inclusions from the relatively early stages of
magmatic evolution, and readily traps melt inclusions in many volcanic rocks. Olivine also has
the advantage of being translucent and relatively incompatible with respect to most commonly
used lithophile trace elements (the exception being transition metals such as Ni, Mn, Co, Cr, V).
This is advantageous for three reasons (i) incorporation of a small amount of the host mineral in
the volume of inclusion analyzed for trace elements has little effect on measured trace element
abundances (Fig. 6); (ii) the rate at which diffusive equilibration occurs between a trapped
inclusion and the mineral host or external melt is lower for elements with lower mineral-melt
distribution coefficients (qin et al. 1992); and (iii) continued crystallization of the host mineral
on the interior of the inclusion does not fractionate incompatible element abundances and has
little overall effect on measured incompatible trace element abundances beyond the effect of
concentration due to olivine removal. Olivine-hosted inclusions can also be simpler to analyze
(as analysis of a small proportion of the host mineral does not significantly affect measured
incompatible trace element abundances; Fig. 6). Plagioclase is also a commonly used mineral
for melt inclusions, as it can contain very large numbers of melt inclusions (e.g., Fig. 1). In
many basaltic rocks, including MORB and related rocks, it is often the dominant phenocryst
phase (Sinton et al. 1993; Nielsen et al. 1995, 2001; Hansen et al. 2000; Sours-Page et al.
1999, 2002; Font et al. 2007). questions remain about the ability of plagioclase to preserve
trace element signatures over the timescales equivalent to magma ascent and storage within the
crust (Cottrell et al. 2002; Danyushevsky et al. 2002a), although variations which match those
of the host lavas are evident in a number of MORB suites (Sinton et al. 1993; Nielsen et al.
1995, 1998; Sours-Page et al. 1999, 2002). Melt inclusions also commonly occur in pyroxene,
although studies of melt inclusions hosted in pyroxenes are less common in basaltic and related
rocks, and are often made in concert with studies of inclusions in co-existing olivine (e.g.,
Sobolev and Danyushevsky 1994; Kent et al. 2002a).
Melt inclusions also occur within spinel, including Cr-rich spinel formed during the early
stages of crystallization of primitive basaltic magmas (Kamenetsky 1996, 2002; Schiano et al.
1997; Shimizu et al. 2001) providing an alternate mineral host for studies of melt inclusions
trapped in primitive basalts. However studies of spinel are limited in many cases by the small size
of spinel phenocrysts and contained inclusions in most basaltic rocks (e.g., Shimizu et al. 2001).
In addition, inclusions within spinel can only be examined in reflected light, placing limitations
on the petrographic examination of inclusions. Nevertheless several studies demonstrate that
studies of melt inclusions in Cr-spinel can provide useful information (Kamenetsky 1995,
1996; Shimizu et al. 2001; Kamenetsky et al. 2002). One specific advantage of spinel may be
its resistance to subsequent weathering or hydrothermal alteration, allowing melt inclusions
to provide a means to estimate the compositions of trapped melts in rocks where the whole
rock composition and other phenocryst phases have been heavily modified by weathering or
alteration (e.g., Shimizu et al. 2001; Kamenetsky et al. 2002).
Formation of melt inclusions
The common, near ubiquitous, presence of melt inclusions in phenocrysts phases in basaltic
rocks demonstrates that the processes that lead to trapping and formation of inclusions occur
during the evolution of almost all basaltic magmas, although abundance and size of inclusions
can vary between individual samples. In general terms melt inclusions form at the interface
between melt and crystalline phases in regions of relatively slow crystal growth and/or mineral
Melt Inclusions in Basaltic and Related Volcanic Rocks 293
dissolution or in regions isolated by the growth of intersecting crystal faces. Reviews, such as
those by Roedder (1979, 1984) and Sobolev and Kostyuk (1975), on the basis of petrographic
observations and analogies to fluid inclusions, suggest a number of general mechanisms of melt
inclusion formation, summarized in Figure 7. Of these, crystallization and textural equilibration
following rapid skeletal, dendritic or hopper growth (Fig. 7D,E) or differential dissolution (Fig.
7A), and localized slow crystal growth rates associated with defects or the presence of other
phases at the crystal-liquid interface (Fig. 7B,C) are probably the most relevant for basaltic
magmas (Kohut and Nielsen 2004; Faure and Schiano 2005; Goldstein and Luth 2006; Baker
2008). Small secondary inclusions may also form in healed fractures (Fig. 7F), but are often too
small for analysis. Growth following rapid differential dissolution may also be important for
plagioclase (Nakamura and Shimikata 1998; Michael et al. 2002).
Experimental studies aimed at understanding melt inclusion formation in basaltic and
related rocks are relatively few (Nakamura and Shimikata 1998; Kohut and Nielsen 2004; Faure
and Schiano 2005; Goldstein and Luth 2006; Baker 2008) but informative, and emphasize the
importance of the thermal history and relations between formation mechanisms and inclusion
composition. Experiments to date have largely been aimed at olivine, plagioclase and to a
lesser degree clinopyroxene, but inclusions in other minerals probably form through similar
processes. Faure and Schiano (2005) identified two different inclusion trapping mechanisms
in forsterite in the CMAS system. In rapidly growing skeletal or dendritic crystals, where
crystal growth is controlled by diffusion at the crystal-liquid interface, inclusions were trapped
by intersection or encirclement of melt by hopper, swallowtail or rod crystal morphologies.
Inclusions trapped in this way were typically small, irregular and showed orientations and
Figure 7. Schematic illustrations of common melt inclusion formation mechanisms relevant to basaltic
rocks. Top and bottom images in each panel represent early and later phases in crystal growth. (A) crystal
growth following a period of rapid mineral dissolution. (B) Inclusions forming where other minerals abut
the host-melt interface, inclusions are associated with mineral inclusions that did not crystallize from the
trapped inclusion. (C) Defects at the crystal interface limit localized growth rates. Inclusions are randomly
distributed within the host. (D) Textural equilibration and overgrowth following rapid dendritic growth,
inclusions are controlled by crystallographic orientations. (E) Overgrowth of skeletal or hopper crystals,
inclusions form in symmetric locations. (F.) Healing of melt-filled fractures, inclusions are typically small
and define a surface. These may be referred to as secondary inclusions, following common fluid inclusion
nomenclature. Modified from Roedder (1979).
294 Kent
distributions influenced by crystallographic orientation. In polyhedral crystals, where growth
rates are slower and growth is controlled by interface attachment, inclusions form at the site of
defects at the crystal-liquid interface (in natural systems this probably includes the presence
of other mineral phases, although Faure and Schiano [2005] did not investigate this directly).
Inclusions formed in this way in polyhedral olivine were larger with more equant form, and
were randomly distributed within olivine crystals.
Other experimental studies show the importance of isothermal growth and annealing
following rapid crystal growth. Kohut and Nielsen (2004) were able to form melt inclusions
in forsteritic olivine and anorthitic plagioclase in a primitive MORB melt through rapid
cooling followed by isothermal crystal growth. In this case isothermal conditions promoted
slower crystal growth and trapped inclusions during formation of tabular overgrowths over
dendritic or skeletal crystals. Goldstein and Luth (2006) formed melt inclusions in forsterite
crystallizing from a haplobasaltic composition over a range of cooling rates of 7-250 °C/hour
(but not at a cooling rate of 1 °C/hour). Inclusions were significantly more abundant, and
resulted in multiple inclusions forming within individual crystals, when cooling was followed
by a six-hour isothermal period prior to final quenching. Goldstein and Luth (2006) also
suggested that inclusions formed by trapping melt within interstices formed by agglomeration
or simultaneous growth of individual platy olivine crystals. Baker (2008) was also able to
form melt inclusions ranging from ~8-200 µm diameter in plagioclase and clinopyroxene at 1
GPa using a rapid cooling (50°/minute) to a temperature 75 °C below the liquidus (~1225 °C)
followed by isothermal crystallization for periods ranging 36 minute to 14 hours. Inclusions
were more common in the longer duration experiments, consistent with the importance of
isothermal crystal growth and textural equilibration following rapid cooling (e.g., Kohut and
Nielsen 2004), although embayments in grain boundaries (interpreted by the author as the
initial stages of inclusion formation) were common in all experiments.
Overall, the range of cooling histories used in experimental studies of melt inclusion for-
mation do not deviate significantly from those expected during the formation, transport and
eruption of basaltic magmas and thus there is good reason to believe that the mechanisms of
inclusion formation identified in these experimental studies are relevant to natural samples. Na-
kamura and Shimikata (1998) were also able to produce inclusions in plagioclase by promoting
rapid dissolution followed by subsequent slow crystal growth. In this case they reproduced tex-
tures similar to the “spongy” zones commonly observed in plagioclase from convergent margin
lavas and although these textures are observed in basaltic rocks this trapping mechanism may
be more relevant to intermediate and evolved compositions (Michael et al. 2002).
It is also important to note that the smooth walled equant or ovoid geometries commonly
observed in melt inclusions (Fig. 1) probably largely result from dissolution and reprecipitation
of olivine after trapping, as the inclusion seeks to minimize surface energy, and thus may
not provide petrographic information on the mechanism of inclusion trapping. Experimental
studies by Schiano et al. (2006) shows that such textural re-equilibration processes appear
isochemical, and may be relatively fast at the temperatures of basaltic magmas. Under thermal
gradients textural re-equilibration may also be accompanied by migration of inclusions within
the individual host mineral grains.
Changes in composition during inclusion formation. A key question for petrological
studies of melt inclusions is whether inclusion formation also promotes changes in composition
of the trapped melt. The primary concern is whether the preferential incorporation of boundary
layers that form at growing crystal-liquid interfaces result in trapped compositions that differ
from the composition of melt present at the time of trapping, and thus lie off the liquid line of
descent. Crystals growing rapidly with respect to the rates of component diffusion in silicate
melts depletes the adjacent liquid in compatible elements, and enriches it in incompatible
elements, forming a compositional boundary layer. Once crystallization ceases or slows
Melt Inclusions in Basaltic and Related Volcanic Rocks 295
then boundary layers will disperse at a rate controlled by diffusion within silicate liquids. If
boundary layers are relatively thick (>> ~10 µm) and inclusion trapping is rapid compared
to the time scales required for boundary layers to re-equilibrate with external melt following
crystal growth then trapped melt compositions may lie off the liquid line of descent. Examples
of this are evident from experimental studies. Faure and Schiano (2005) show that inclusions
hosted with rapidly growing dendritic and skeletal crystals have marked enrichments of Al2O3
over CaO, reflecting the slower diffusion rate of Al in silicate melts (Liang et al. 1996; Fig.
8A), whereas inclusions in slower growing polyhedral crystals lie close to the predicted
liquid line of descent for the experimental composition. Thus the rate of crystal growth and
mechanism of inclusion trapping may be important to understanding the composition of
natural melt inclusions, and interpretation of inclusions may benefit from careful petrography
and documentation of crystal growth mechanisms. This may be difficult to discern in many
olivine crystals due to textural equilibration and further growth. However mapping of P2O5
zoning in olivine may aid in recognition of rapidly grown portions of crystals (Milman-Barris
et al. 2008), and studies to date suggests that inclusions in olivine often occur in P2O5-rich
zones interpreted to form during rapid crystal growth.
Experimental studies also show that the extent to which boundary layers are incorporated
into inclusions is also dependent on the thermal history of samples. The experiments of Kohut
and Nielsen (2004) and Goldstein and Luth (2006), where isothermal crystallization times are
relatively long, show little deviation of trapped inclusion compositions away from the liquid line
of descent—even where inclusions occur in skeletal and dendritic crystals. Isothermal growth
(e.g., Fig. 8), was probably sufficient to allow boundary layers formed during rapid crystal
growth to dissipate prior to melt isolation. The importance of thermal history and isothermal
cooling is also highlighted in recent work by Baker (2008), who showed that boundary layers
enriched in S, Cl, P2O5, as well as in major elements such as Al2O3 and FeO formed at the
margins of rapidly growing plagioclase and pyroxene. Boundary layers (defined by zones of
enrichment in compatible elements and depletion in compatible elements) were narrower (but
showed greater changes in concentration) where isothermal crystallization times were shortest.
Longer isothermal periods revealed broader boundary layers with less enrichment or depletion,
consistent with collapse of steep gradients after the cessation of rapid crystal growth. Baker
(2008) modeled these variations in terms of two-stage cooling and diffusive equilibration. For
basaltic melts at 1200 °C, the elements that diffuse relatively slowly (P, D ≈ 2×10−13 m2/sec)
require a period of ~100 hours to drop to a value within ~10 % of that of the of the external melt,
whereas more rapidly diffusing elements (Baker 2008, uses the example of S, D ≈ 2×10−12 m2/
sec; Freda et al. 2005) require much shorter time periods—generally << 10 hours (Fig. 9).
Despite experimental demonstration of the potential for boundary layer contributions to
trapped melt inclusions (e.g., Faure and Schiano 2005; Baker et al. 2008), there is less evidence
from natural basaltic inclusion suites to suggest that this is a concern—particularly where
studies concentrate on inclusions that are sufficiently large for routine geochemical analysis
(typically ≥ ~30 µm in diameter). Efforts to recognize the effects of trapping boundary layers in
natural samples, by comparing the size of inclusions with elements with different diffusivities
(e.g., Nielsen et al. 1995, 1998; Kuzmin and Sobolev 2004) show little to no evidence for
preferential incorporation of boundary layers in smaller inclusions. A study of Icelandic picrites
showed that fractionation of CaO/Al2O3 attributable to boundary layer incorporation into melt
inclusions only occurred in inclusion 15 µm in diameter or less (Kuzmin and Sobolev 2004).
A summary of data from olivine-hosted melt inclusions from a number of primitive magmas
shown in Figure 8B-E also suggests fractionation of CaO/Al2O3 in response to boundary layer
trapping may be less common in natural basaltic melt inclusions. In particular there is little
evidence for large relative enrichments in Al evident in fast-trapped inclusions seen by Faure
and Schiano (2005). Where large deviations in CaO/Al2O3 are observed in inclusion suites these
typically extend to higher CaO/Al2O3 in inclusions than in the host glass or lavas (Kamenetsky
Figure 8. Comparison of CaO and
296
Al2O3 measured in experimental
and natural melt inclusions. (A)
The results of the experimental
study of Faure and Schiano (2005)
where inclusions formed via rapid
crystal growth (dendritic, skeletal
and dendritic polyhedral) show
large enrichments in Al2O3 over
CaO. Inclusions formed in more
slowly grown polyhedral olivine
have compositions close to the
liquid line of descent. Figure is
modified from Faure and Schiano
(2005). (B) Measured composi-
tions of melt inclusions formed
in olivine from the experimen-
tal study of Goldstein and Luth
(2006). Inclusion compositions
have higher CaO and Al2O3 than
Kent
matrix glass, interpreted to be due
to post-entrapment crystallization,
but still lie on the expect liquid
lines of descent. The differences
between the results of Faure and
Schiano (2005) and Goldstein and
Luth (2006) probably reflect dif-
ferent cooling paths chosen for
each experiment. (C), (D) and (E)
show data from natural inclusion
suites from Baffin Island, Canada
(Kent et al. 2004a; Kent, A.J.R.,
unpublished data), MORB (Sobo-
lev 1996; Shimizu 1998; Saal et
al. 2002) and Hawaii (Norman et al. 2002). Lines of constant CaO/Al2O3 and compositions of matrix
glasses are also shown. Although there are distinct differences in CaO/ Al2O3 between suites there are no
large departures from the expected liquid line of descent for inclusions within individual suites, similar
to those demonstrated by Faure and Schiano (2005), implying that little fractionation of this ratio occurs
during natural inclusion formation (see text).
Melt Inclusions in Basaltic and Related Volcanic Rocks 297
Figure 9. (A) Modeled changes in concentration of P, S and H2O (given as Cx/Co, the ratio of concentration
at distance x to that at infinite distance from the crystal) adjacent to a growing crystal immediately following
the end of rapid crystal growth (e.r.g.), and after 10 or 100 hours of isothermal crystal growth. Calculations
are for a basaltic melt at 1200 °C. Modified from Baker (2008). (B) Relation between inclusion size for
rehomogenized inclusions from Endeavour segment of the Juan de Fuca Ridge (Sours Page et al. 1999;
Nielsen et al. 1998). Nielsen et al. (1998) interpreted the lack of covariation as evidence against equilibration
between host and inclusion, but also argues against trapping of K2O-enriched boundary layers.
et al. 1998; Shimizu 1998; Schiano et al. 2000; Norman et al. 2002; Danyushevsky et al. 2004;
Ehlburg et al. 2007; Laubier et al. 2007), and are proscribed to petrological variations such as
differences in the degree of melting and amount of clinopyroxene present within the mantle
source (e.g., Schiano et al. 2000; Norman et al. 2002), interaction between mantle melts and
clinopyroxene during melt transport (Shimizu 1998) or melts produced by localized dissolution
and reaction (Danyushevsky et al. 2004). This potential for petrological variation limits the use
of the CaO/Al2O3 ratio as an unambiguous means of determining the presence of boundary
layer trapping, unless petrological variations in CaO/Al2O3 in host and associated lavas are well
defined (e.g., Vigouroux et al. 2008). Baker (2008) also suggest that Cl/P2O5 or S/P2O5 ratios be
used for detecting boundary layer effects, although the application of these ratios is also limited
by the possibility of Cl and S loss via degassing prior to trapping, the complex solubility of S
during changes in melt composition and redox state (e.g., Mathez 1976; Wallace and Carmichael
1992; Rowe et al. 2007) and difficulty obtaining sufficiently precise P2O5 measurements to
recognize enrichment in MORB and other basalts with naturally low P2O5 contents.
298 Kent
The approach of Michael et al.
(2002), where compositional parameters
are compared to diffusivity (or proxies
of diffusivity) for individual elements,
can be modified and extended to other
suites, and provides an alternate means
to investigate the role of boundary-layer
driven diffusion in natural melt inclusion
suites. An example is shown in Figure 10
as a comparison between composition and
the known diffusivity of these elements
in basaltic melts for olivine-hosted melt
inclusions from Baffin Island, Canada
and Theistareykir, Iceland. In Figure 10A,
where the standard deviation calculated
for individual elements is compared
to diffusivity, correlations are poor to
non-existent, suggesting a minimal
role for diffusion-driven elemental
fractionation in controlling the overall
trace element variability evident in these
melt inclusion suites. There are also poor
correlations evident between the degree
of enrichment or depletion relative
to primitive mantle and diffusivity in
representative individual melt inclusions
(Fig. 10B), suggesting that the lithophile Figure 10. (A) Comparison between diffusivity in
basaltic melt at 1200 °C with the standard deviation
trace element composition of individual calculated from individual trace element abundances
inclusions are not controlled by boundary in 25 olivine-hosted melt inclusions from Baffin Is-
layer trapping. land (Kent, A.J.R. unpublished data) and 47 olivine-
hosted inclusions from Theistareykir, Iceland (Slater
Although further investigation et al. 2001). Diffusion data are from LaTourette et al.
of the role of boundary layer trapping (1996, 1997) and Richter et al. (2003). Note that LREE
is required, the limited evidence for have all been assigned the same diffusivity as Nd and
variations in natural inclusion suites to heavy HREE as Yb. Data from Richter et al. (2003)
date may reflect the fact that isothermal were calculated at 1200 °C using the same activation
energy as Ba (181 kJ/mol; LaTourette et al. 1996). (B)
crystallization following rapid crystal Comparison of diffusivity with primitive-mantle-nor-
growth plays an important role in forming malized concentration for representative enriched and
melt inclusions within natural magmatic depleted melt inclusions from Baffin Island, Canada
systems. Further, even where trapping of (Kent, A.J.R. unpublished data). The poor correlations
boundary layers occurs, inclusions that in (B) and (C) suggest that diffusivity in basaltic melts
does not control the variations in melt inclusion com-
are small enough to record the effects positions evident in these suites. As shown in Figure
are also probably too small for routine 17 the standard deviation and mineral-melt partition
analysis (< ~30 µm). Selection of larger coefficients are more strongly correlated.
inclusions for analysis may preferentially
select those that formed from defects
at the crystal-liquid interface during slower growth, rather than through rapid crystal growth
(Faure and Schiano 2005). It is also possible that future studies could reveal fractionations
related to boundary layer trapping in elements that diffuse more slowly in melts, such as zr and
P2O5 (Baker 2008). Overall boundary layer trapping is probably of more concern for study of
lower temperature and more silicic magmas, where lower temperatures and higher degree of
polymerization promote slower melt diffusion rates. In silicic melts the lower limit of inclusion
Melt Inclusions in Basaltic and Related Volcanic Rocks 299
size, below which inclusions are effected by boundary layer effects is estimated at ~50 µm (e.g.,
Lu et al. 1995).
evolution of melt inclusions after trapping
Considerable effort has gone into understanding the processes that occur in melt inclusions
after trapping, and how these might effect the eventual measured inclusion compositions (e.g.,
qin et al. 1992; Tait 1992; Sobolev and Danyushevsky 1994; Sobolev 1996; Nielsen et al.
1998; Schiano and Bourdon 1999; Danyushevsky et al. 2000, 2002a,b; Kress and Ghiorso
2004; Gaetani and Watson 2000, 2002; Cottrell et al. 2002; Spandler et al. 2007; Portnyagin
et al. 2008).
The compositional evolution of a melt inclusion following trapping is related to the pressure
and temperature path followed by the individual inclusion, and thus may diverge significantly
from that experienced by the surrounding melt. At the time of formation the temperature and
pressure of the inclusion will exactly match that of the surrounding melt, and as the volumes
of melt inclusions are small, temperature within an inclusion will remain linked to that of
the surrounding melt. However, as the bulk modulus of silicate liquids (1.2×1010 Pa; Murase
and McBirney 1973) are about an order of magnitude less than the elastic moduli of silicate
minerals (1.97 x 1011 Pa; Gebrande 1982), and because dissolved volatile components exsolve
in response to decompression, pressure changes within trapped inclusions are relatively small
during magmatic decompression (Tait 1992; Schiano and Bourdon 1999). quantitative models
show that the pressure change experienced by a melt inclusion during decompression will
only be a small fraction (at least < 30%, and probably below 10%) of the total pressure change
experienced by the host crystal and magma (Fig. 11; e.g., Tait 1992; Schiano and Bourdon
1999). Moreover, if the degree of expansion of an inclusion exceeds the tensile strength of the
host mineral then cracking and decrepitation (loss of pressure and volatile components) of the
inclusion can also occur. The threshold pressure required to rupture the mineral host decreases
with inclusion radius (Fig. 11B) consistent with the common observation that large melt
inclusions often have cracks within the adjacent mineral host. Fracturing can lead to changes
in pressure within the inclusion and exsolution and loss of a vapor phase (either during initial
decompression or laboratory rehomogenization), as well as providing a pathway for subaerial
or submarine weathering to alter inclusion compositions (Nielsen et al. 1998).
Post entrapment crystallization (PEC). Once an inclusion has been trapped and the
temperature decreases, crystallization of the host phase on the inclusion wall will commence
(Sobolev and Kostyuk 1975; Roedder 1984; Sobolev 1996; Kress and Ghiorso 2004). This is an
inevitable consequence of cooling the melt-host system and will occur in all trapped inclusions.
As nucleation sites are readily available at the margin of the inclusion, newly crystallized
host will almost always occur as concentric shells at the inclusion wall. This may not be
detectable using optical microscopy, but is often apparent in BSE or compositional images
from the electron microprobe (Fig. 1). In the absence of significant re-equilibration between
the newly grown crystal and the mineral host, new mineral growth will be near fractional. For
olivine this results in crystallization of an increment of olivine that is more magnesian than
the trapped melt, following established Fe2+/Mg partitioning relationships, where KD (FeO/
MgO) olivine-melt = 0.30 ± 0.04 (Roedder and Emslie 1970; Ford et al. 1983; Falloon et al.
2007; Putirka et al. 2007). This will deplete the trapped melt in Mg relative to Fe2+ and the next
increment of olivine crystallized will be less Mg-rich (but still more Mg rich than coexisting
melt). This produces a zone of Fe-rich olivine surrounding the melt inclusion, depletes the
melt inclusion in Mg (Fig. 12) and other compatible elements, and increases concentrations
of olivine-incompatible elements. Crystallization in plagioclase-hosted inclusions likewise
results in progressive depletion of Ca over Na and formation of more albitic rims around
inclusions (Nielsen et al. 1995).
300 Kent
Figure 11. (A) Pressure changes in melt inclusions trapped at 1.0 GPa during decompression to
atmospheric pressure. Decompression paths are shown for basaltic and felsic composition melt inclusions.
The decompression of interstitial glass is shown for comparison. Decompression between points 1. and 2.
represent adiabatic decompression from 1 GPa to 1 atm and then cooling from 1200 °C to 25 °C (point
3). Felsic and basaltic trajectories differ slightly due to differences in the bulk moduli of these liquids.
Modified from Schiano and Bourdon (1999). (B) Calculated decrease in pressure within an olivine-
hosted melt inclusion, as a function of external pressure and different volumes of CO2 gas present at
the starting pressure. The presence of an initial gas phase leads to higher pressures within the inclusion
following decompression (Tait 1992). Inset shows increase in the cracking threshold pressure of olivine-
hosted inclusion as a function of inclusion radius. The curve for quartz-hosted inclusions is also shown
for reference. Calculated differences in cracking thresholds arise due to differences in surface energy (Tait
1992). Larger melt inclusions are more likely to crack their mineral host. Modified from Tait (1992).
In cases where there has been minimal diffusive re-equilibration between inclusion and
host after trapping (see 3.4.1 below), corrections for post entrapment crystallization in olivine
can be made via experimental reheating to the temperature of trapping (Sobolev and Kostyuk
1975; Nielsen et al. 1995, 1998; Danyushevsky et al. 2002a; Schiano 2003) or by numerical
Melt Inclusions in Basaltic and Related Volcanic Rocks 301
Figure 12. Measured and corrected MgO
contents for melt inclusions and matrix
glasses from two samples from Loihi
seamount. Also shown are MgO contents
for matrix glasses from Loihi reported in
the literature. Corrected MgO contents
in melt inclusions, calculated using the
numerical approach outlined in the text,
are greater than those in associated matrix
glasses, and are also greater than the
majority of glasses reported from Loihi
from the literature (includes both tholeiitic
and alkaline compositions). Corrected
MgO contents approach those estimated
for primary Hawaiian melts. Modified
from Kent et al. (1999a).
correction schemes. For the latter, olivine of a composition calculated to be in equilibrium with
the trapped melt is added iteratively to the measured trapped composition until equilibrium
between inclusion and host is obtained (Sobolev and Chaussidon 1996; Sobolev 1996). In most
instances cases this requires addition of less than 10-15% (by weight) olivine (e.g., Sobolev
and Chaussidon 1996; Kent et al. 1999a; Lassiter et al. 2002) and thus has little effect on
olivine incompatible elements beyond the uncertainties associated with measurements. This
procedure works well for olivine as FeO/MgO partitioning is relatively insensitive to pressure
and temperature (e.g., Roedder and Emslie 1979; Putirka et al. 2007) and so is robust over the
crystallization conditions of basaltic magmas, although if pressure of crystallization is known
independently then this can also be incorporated into the calculations (Falloon et al. 2007).
This approach does require additional constraints on the Fe2+/Fe3+ ratio. For MORB most
workers follow the formulations of Kilinc et al. (1983) and Kress and Carmichael (1991) for
a given oxygen fugacity.
Similar approaches have not been developed for plagioclase, as cation partitioning
relationships are more dependent on pressure and temperature. Nielsen et al. (1995)
emphasized the use of phase equilibria modeling to determine if experimentally reheated
inclusions were in equilibrium with their host. Kress and Ghiorso (2004) also present a more
thermodynamically rigorous procedure to correct for post entrapment crystallization using
the MELTS algorithm that takes into account changes in oxygen fugacity, temperature and
pressure during crystallization.
It should also be noted that the elements that are most sensitive to correction for post
entrapment crystallization are those that are compatible within the host mineral, and these
may thus be less robust in corrected melt inclusion compositions. A common example is the
MgO content of melt inclusions in basaltic rocks, which, because of the high MgO content of
forsterite-rich olivine, is often strongly depleted during post entrapment crystallization (e.g.,
Fig. 12). For example a melt inclusion with an uncorrected MgO content of 5.1 wt% (and FeO/
MgO ~2) has a corrected MgO content of 9.1 wt% (an increase of 180%) when calculated
302 Kent
to be in equilibrium with Fo89 olivine. Variation of just ± 1 Fo unit in the composition of the
host olivine is equivalent to a range of corrected MgO contents between 8.1-10.3 wt%. In
comparison, for the same correction to Fo89, incompatible element contents decrease by only
~15% due to the addition of olivine, and ratios of incompatible elements are unaffected by the
correction. For this reason incompatible elements and their ratios are often considered more
robust sources of information in melt inclusions than compatible elements.
re-equilibration with host and or external melt
The potential for melt inclusions to preserve magmatic compositions and remain
isolated during magmatic evolution has also lead to scrutiny of the potential for diffusive
exchange between the inclusion and the host mineral and/or the external melt (qin et al. 1992;
Danyushevsky et al. 2000, 2002a,b; Gaetani and Watson 2000, 2002; Cottrell et al. 2002;
Spandler et al. 2007). The central question is whether the composition of melt inclusions may
be altered significantly by post-entrapment re-equilibration, and if so whether this effect can
be reversed by numerical schemes or neutralized by restricting interpretations to elements or
indices that are less sensitive to re-equilibration. If an inclusion exchanges completely with
external melt then the trapped melt composition will only reflect the composition of the most
recent external melt the host crystal has encountered, if partial equilibration occurs then there
is the opportunity for elements or species with different equilibration rates to be fractionated
from one another (qin et al. 1992; Cottrell et al. 2002) and thus to induce chemical signals that
do not reflect real magmatic variations.
qin et al. (1992) laid the theoretical basis for understanding diffusive equilibration of
melt inclusions by developing analytic solutions to the diffusion equation for isothermal
equilibration of a spherical inclusion located concentrically within a spherical mineral host.
This approach has subsequently been further addressed using theoretical, experimental and/
or observational approaches (Danyushevsky et al. 2000, 2002a,b; Gaetani and Watson 2000,
2002; Cottrell et al. 2002; and Spandler et al. 2007). Inclusion equilibration rate is a function
of diffusivity of the species of interest within the host crystal, relative size of inclusion and
host (or proximity of the nearest crystal face), partition coefficient for species where KD < 0.1,
cooling rate, and degree of initial disequilibrium between the host and inclusion (e.g., Fig. 13).
These studies show that more rapid re-equilibration is expected in inclusions that are smaller,
have larger ratios of inclusion radius to host mineral radius, and that cool more slowly. Species
that are more compatible and/or have higher diffusivities, and that are further from initial
equilibrium will also equilibrate more rapidly. The rate of equilibration scales proportionately
with the diffusivity and with the inverse square of inclusion size—thus an inclusion that is
twice as large will equilibrate four times more slowly (Cottrell et al. 2002).
Iron loss in olivine-hosted inclusions. One specific concern is the effect of re-equilibration
between inclusion and host and/or external melt on the Fe2+ content and Fe/Mg ratio of trapped
melts within olivine-hosted melt inclusions (Danyushevsky et al. 2000; Gaetani and Watson
2000, 2002). Iron and Mg are compatible within olivine and Fe-Mg interdiffusion is relatively
rapid at temperatures and oxygen fugacities relevant to basaltic melts (at 1200 °C D = 10-15.0±0.6
m2/sec; Buening and Buseck 1973; Misener 1974; Jurewicz and Watson 1988). During eruption
and in other cases where crystallization and cooling rates are relatively rapid compared to Fe-
Mg interdiffusion, post entrapment crystallization occurs in a near fractional manner and can
be readily reversed by experimental heating or numerical methods described above (e.g., Fig.
12; Sobolev 1996; Danyushevsky et al. 2000).
Where cooling rates are slower relative to crystallization, then diffusive equilibration be-
tween the inclusion and host olivine and/or external melt can become significant (Danyush-
evsky et al. 2000, 2002b; Gaetani and Watson 2000, 2002). The net effect of this is Fe loss.
This is because, as Fe-rich olivine grown at the inclusion host interface equilibrates with more
magnesian olivine in the remainder of the host mineral, the inclusion also seeks to re-equilibrate
Melt Inclusions in Basaltic and Related Volcanic Rocks 303
Figure 13. Re-equilibration between melt in-
clusions and ambient melt. Modified from qin
et al. (1992). Axes show equilibration percent
versus τ, a dimensionless time parameter de-
fined as τ = Dt/b2. Where D is diffusivity, t is
time and b is host mineral radius (see inset).
(A) The effect of variation in mineral melt
partition coefficient (KD) showing that equili-
bration is faster where an element is more
compatible within the host. (B) The effect of
variation in the ratio of inclusion radius to host
radius (see inset in A), equilibration is faster
when this ratio is larger. C0 refers to the con-
centration of a given element within the melt
inclusions, KD is the mineral melt partition
coefficient.
with the rim by adding Fe to maintain Fe/Mg equilibrium. Such Fe-loss is generally recognized
by Fe contents of inclusions that are below those of the liquid line of descent followed by whole
rock and matrix glass compositions (Fig. 14B-D; Danyushevsky et al. 2000, 2002b). Further, as
inclusions hosted in more magnesian olivines are subject to a greater degree of Fe-loss, effected
inclusions will also often define a correlation between the forsterite content of the host olivine
and measured Fe contents within melt inclusions (Fig 14A; e.g., Danyushevsky et al. 2000,
2002b; Yaxley et al. 2004).
The effects of Fe loss are schematically illustrated in Figure 15. Due to volume consider-
ations Fe loss has little net effect on the bulk composition of the larger host crystal or external
melt, but will deplete the inclusion in Fe as it seeks to maintain Fe-Mg exchange equilibrium
with the host mineral (Danyushevsky et al. 2000, 2002b; Gaetani and Watson 2000, 2002). Iron-
loss has been documented in olivine-hosted inclusions in a number of basaltic and associated
suites, including those from subduction zones (where the effect is more common; Sobolev and
Danyushevsky 1994; Danyushevsky et al. 2000), and in komatiite, ocean island basalt and con-
tinental flood basalt suites (Danyushevsky and Sobolev 1994; Danyushevsky et al. 2000, 2002b;
Norman et al. 2002; Yaxley et al. 2004). An analogous process has also been suggested to occur
in slowly cooled olivine-hosted spinel mineral inclusions (Scowen et al. 1991).
Gaetani and Watson (2000) emphasize that Fe loss is irreversible through experimental
reheating of inclusions or other technique involving simple addition of olivine. Iron loss can result
in Fe2+/Mg ratios that are low, and if these are corrected by following normal olivine addition
procedures or experimental reheating then this will lead to overestimation of Mg# contents of
the trapped inclusion compositions. Inclusions may also be more oxidized through loss of Fe2+
(Rowe et al. 2007). In addition, experimental reheating may overestimate trapping temperatures
by as much as 50 °C in cases where complete re-equilibration occurs over a large interval
of temperature (Danyushevsky et al. 2000). Numerical techniques are available to estimate
304 Kent
the amount of Fe lost. Danyushevsky et
al. (2000) presented an iterative correction
scheme to reverse the effects of Fe loss in
homogenized inclusions which enables
calculation of homogenization temperatures,
although this requires an independent
estimate of the FeO content of the trapped
liquid and thus may be subject to additional
uncertainty. Danyushevsky et al. (2002b)
also present an alternate correction scheme
based on measurement and modeling of
Fe diffusion profiles in the host mineral
surrounding the inclusion. Vigouroux et al.
(2008) corrected for Fe-loss in naturally
quenched and un-rehomogenized inclusions
by adding Fe2+ back into each melt inclusion
until the FeO* of the inclusion was restored
to the MgO vs. FeO* trend defined by whole
rock and groundmass glass compositions.
Another promising approach may be to use
parameterizations for MgO partitioning
between olivine and basaltic liquid (Putirka
2005; Falloon et al. 2007; Putirka et al.
2007) to determine the original MgO
content of the trapped liquid and using Fe2+/
Mg partitioning relations to establish the
FeO content.
Danyushevsky et al. (2000) considered
the case where equilibration occurred
Figure 14. Indications of Fe loss in melt inclusions. between a melt inclusion and “infinite”
(A) Negative correlation between forsterite content host olivine, disregarding the possibility of
of the host olivine and Fe content of an inclusion
as the result of Fe-loss during re-equilibration
continued equilibration between the host
between olivine and melt (see text). Modified from and the external melt during continued
Yaxley et al. (2005). The extent of Fe loss is greater magmatic evolution. Gaetani and Watson
in the more magnesian olivine host phenocrysts (2000, 2002) used numerical modeling to
(Danyushevsky et al. 2000). (B) Anomalously low study equilibration with external melt and
FeO contents in melt inclusions in olivine-hosted
melt inclusions from Tonga boninites compared to
demonstrated that where cooling rates were
the liquid line of descent defined by host glasses and sufficiently slow equilibration with external
whole rock compositions. (C) Al2O3 contents lie on melt may also result in loss of Fe relative
the estimated liquid line of descent. Modified from to Mg. Cooling rates as rapid as 1-2 °C/
Danyushevsky et al. (2000). year are sufficient to maintain exchange
equilibrium between an inclusion and its
host melt during magmatic evolution, and these are considerably faster than estimates of
cooling rates within crustal basaltic magma chambers (e.g., Cooper et al. 2001).
The effect of re-equilibration on major element contents in other mineral hosts remains to
be explored. Diffusion rates of Ca and Na in plagioclase are relatively slow (Grove et al. 1984),
as they require coupled diffusion with Al, and thus might not be expected to re-equilibrate
with basaltic melt inclusions over timescales relevant to igneous systems. Danyushevsky et
al. (2000) also suggested that equilibration within plagioclase hosted inclusions in MORB
may produce lower FeO and TiO2 contents, based on melt inclusions from ODP Hole 896A,
although they note that there is no clear physical model apparent for this.
Melt Inclusions in Basaltic and Related Volcanic Rocks 305
Figure 15. The schematic effects of equilibration between melt inclusions and host olivine (modified from
Danyushevsky et al. 2000). Shaded field represents an inferred liquid line of descent. Path A-B represents
the effect of crystallization of olivine on the walls of the host mineral from a starting composition of A
during cooling, but without equilibration between the new olivine and the remainder of the host crystal.
Path A-B is reversible via rehomogenization or numerical treatments for olivine crystallization after
entrapment (see text). Path A-B2 represents full equilibration between new olivine grown on the interior
of the inclusion and the host olivine via Fe-Mg interdiffusion, path A-B1 represent an arbitrary partial
re-equilibration path. Paths B-C, B1-C1 and B2-C2 represent the paths followed by melt inclusions during
rapid cooling after eruption, where crystallization rate is rapid relative to the rate of Fe-Mg diffusion in
olivine and near-fractional growth of olivine occurs. From C, C1 or C2 post-eruptive olivine growth can be
corrected for to return inclusions to B, B1,or B2, but continued correction of partially or fully equilibrated
inclusions moves inclusions further along B1-A1 or B2-A2. The homogenized compositions A1, A2 have
lower Fe and higher Mg# than the original trapped inclusion and if fully equilibrated with host olivine (R1,
R2) also have higher MgO contents and Mg#‘s.
The inset (bottom right) shows the MgO and FeO variations expected for a 100 µm diameter melt
inclusion in a 2 mm diameter olivine trapped at 15 kb and transported to 5 kb with ~200 °C temperature
decrease (0.5 °C/year; Gaetani and Watson 2002). The entrapment pressure of the inclusion is preserved
during decompression. The dotted line shows the trend followed with no Fe-Mg equilibration (dotted line)
and the dashed line reflects full equilibration between inclusion host and external melt. Modified from
Gaetani and Watson (2002), axes have been adjusted to be the same scale as the larger plot.
Modification of incompatible trace elements by diffusive equilibration. As noted above
the rate of diffusive equilibration of a given species between a melt inclusion and the host crystal
or external melt depends on the diffusivity, inclusion and host size and mineral melt partition
coefficients (Fig. 13; qin et al. 1992). Although low mineral melt partition coefficients should
restrict the rate of re-equilibration for many incompatible lithophile trace elements, if significant
equilibration does occur then variations in diffusivities and partition coefficients could result in
changes in relative abundances—often a valuable source of petrological information. Although
many studies of melt inclusions have explicitly or implicitly assumed that negligible diffusive
modification of inclusions may have occurred after trapping, recent work have demonstrated
that changes in some compatible (e.g., Fe in olivine) and/or rapidly diffusing components
(e.g., H+, H2O) occur in some natural inclusion suites (e.g., Danyushevsky et al. 2000; Cottrell
306 Kent
et al. 2000; Hauri 2002; Rowe et al. 2007). Every inclusion suite studied should be assessed on
its own merits for possible diffusive modification using the approaches detailed below.
Cottrell et al. (2002) developed a numerical model to study the effect of re-equilibration
between host and inclusion on incompatible element abundances, and applied insight from
their results to interpret data from olivine and plagioclase hosted trace element abundances in
MORB suites. Although knowledge of diffusivity of some trace elements limited interpretations,
Cottrell et al. (2002) suggested that significant re-equilibration could result in fractionation of
REE contents within some melt inclusions, particularly in plagioclase where elements such
as Sr, Eu and Ba can be compatible or slightly incompatible (Fig. 16). Comparison of model
Figure 16. Results of modeling of diffusive equilibration between melt inclusions and olivine and
plagioclase hosts, modified from Cottrell et al. (2002). Panels A-D show the calculated effects of diffusive
re-equilibration for olivine and plagioclase hosted compositions with parallel primitive mantle normalized
patterns for REE (A and C). Panels B and D show the calculated effect after 2500 years. Panels E-J
show compiled data for olivine and plagioclase hosted melt inclusions from the literature. Data sources:
Kamenetsky et al. (1997); Slater et al. (2001); Kent, A.J.R. unpublished data.
Melt Inclusions in Basaltic and Related Volcanic Rocks 307
results for REE with compiled data sets appears to indicate that Eu, and possible the HREE
contents in plagioclase-hosted melt inclusions are effected by re-equilibration with the host for
number for plagioclase-hosted melt inclusion suites (Fig. 16). In contrast, predicted patterns
for REE re-equilibration in olivine-hosted melt inclusion suites do not resemble equilibrated
compositions and probably reflect magmatic diversity.
Cottrell et al. (2002) also emphasized comparisons of the standard deviation, calculated
for individual elements within suites of melt inclusions from the same samples, and mineral-
melt or peridotite-melt partition coefficients. The rationale is that bulk partition coefficients
(for mantle melting) and mineral melt partition coefficients (for crystal fractionation) drive
variations among incompatible trace elements in primitive melts. It has been recognized
for some time that standard deviations for trace elements in suites of MORB matrix glasses
correlate well with bulk partition coefficient for peridotite melting (Hofmann 1988). Thus if
melt inclusions sample melt diversity and preserve it unaltered by re-equilibration or other
inclusion-specific processes then the variance exhibited by melt inclusions should correlate
with bulk or mineral-melt partition coefficients. If partitioning also drives compositional
variations for the lavas that host melt inclusions via the same set of magmatic processes then
variances shown by trace elements in melt inclusion populations should correlate with those of
the host and associated lava or glasses, although this might not be a 1:1 correlation.
It is clear from the work of Cottrell et al. (2002) and from the additional comparisons shown
in Fig. 17 that plagioclase-hosted melt inclusion suites can show anomalous standard deviations
compared to partition coefficients or host lava compositions. Standard deviations calculated
for host glass or lavas suites are well correlated with bulk partition coefficients for peridotite
melting (the correlations with log D shown in Fig. 17 have r2 > 0.85) emphasizing the role of
partitioning in compositional variation. Standard deviations calculated for olivine hosted melt
inclusion suites also correlate well with both Dperidotite-melt (r2 > 0.8) and with standard deviations
for the same elements in host lava suites (not shown but r2 are > 0.8). However correlations
between Dperidotite-melt and plagioclase hosted inclusion suites are considerably worse (r2 < 0.5).
Moreover Sr and Eu, and Ba appear to have standard deviations in most plagioclase hosted
inclusion suites that are lower than elements of similar Dperidotite-melt, whereas zr has elevated
standard deviation (Fig. 17). Overall this suggests that careful examination of trace element
abundances in plagioclase for the effects of re-equilibration is required before compositional
variations are interpreted as the result of magmatic processes.
The above analysis and the work of Cottrell et al. (2002) suggests that trace element
contents of olivine-hosted inclusions remain relatively unaffected by re-equilibration after
trapping, at least for the timescale relevant for transport and storage of basaltic magmas, and
this appears consistent with observations from olivine-hosted melt inclusion suites. Cottrell et
al. (2002) used estimates for diffusivities of REE and other trace elements in olivine, as these
had not been measured. Recent experimental determinations of diffusivities of trace elements in
olivine show that they may be considerably faster than previously thought, with DREE at 1300 °C
~10-15 m2sec-1 (Spandler et al. 2007). On this basis Spandler et al. (2007) argued that diffusive
re-equilibration of REE and other trace elements was sufficiently fast, occurring on timescales
of years, relative to the longer timescales of ascent of basaltic magma and storage within the
crust (e.g., Faul 2001; Cooper et al. 2001; Condomines et al. 2003), to preclude preservation of
mantle-derived compositions in olivine and spinel hosted basaltic melt inclusions.
It is currently unclear whether the rapid equilibration envisaged by Spandler et al. (2007)
is relevant to natural melt inclusion suites. More recent measurements of the diffusion rate of
REE in olivine record values that are 3-4 orders of magnitude slower (Cherniak et al. 2007)
than those of Spandler et al. (2007), and if the diffusivities of Spandler et al. (2007) are used
for the model of Cottrell et al. (2002) then the pattern predicted by re-equilibration shown in
Fig 16 would occur within just 2-3 years (Cottrell, E. personal communication). This finding
308 Kent
Figure 17. Comparison of standard
deviations calculated for trace element
abundances measured in plagioclase and
olivine hosted melt inclusion suites and
matrix glass compositions from basaltic
samples from the Southwest Indian Ridge
(SWIR), Baffin Island and Theistareykir,
Iceland. Panels (A), (C), and (E) show
comparisons between percent standard
deviations for melt inclusion suites
compared to those for host glass or lava
other associated glasses or lavas from
each location. Panels (B), (D) and (F)
show percent standard deviations for melt
inclusions and host and associated glasses
or lavas compared to bulk D for peridotite
melting (from Cottrell et al. 2002, and
sources therein). Data sources are Slater
et al. (2001), Lund (2003), Meyzen et
al. (2003), Kent et al. (2004b), and Kent,
A.J.R. unpublished data.
is at odds with the compositions of natural melt inclusion suites, where, as discussed above,
REE profiles and standard deviations do not match those expected from equilibration, but
are often closely matched to those of the host and associated lava compositions. Moreover
compositional variations amongst REE and other lithophile trace elements in olivine-hosted
melt inclusions that are directly attributable to mantle variations are relatively common. One
clear example are the melt inclusion suites from subduction zone lavas that show trace element
signatures that relate to subduction fluxes introduced to the mantle wedge (e.g., Sobolev and
Melt Inclusions in Basaltic and Related Volcanic Rocks 309
Danyushevsky 1994; Kamenetsky et al. 1997; Sisson and Bronto 1998; Kent and Elliott 2002;
Cervantes and Wallace 2003a; Walker et al. 2003) and relate water contents (itself a rapidly
diffusing component within olivine; Mackwell and Kohlstead 1990; Portnyagin et al. 2008) to
mantle melting (e.g., Cervantes and Wallace 2003a). In other cases chemical variations relating
to mantle melting or mantle source variations can be preserved alongside chemical variations
relating to crustal residence of magmas (Kent et al. 2002a; Yaxley et al. 2004; Kamenetsky and
Gurenko 2007).
Improved analytical capabilities, specifically increased analytical sensitivity and access to
greater spatial resolution for in situ analyses, offer additional means to identify and constrain
the role of diffusive equilibration in altering the composition of trace element signatures within
melt inclusions. Although these approaches have not yet been widely applied, spatially-resolved
measurements of the compositions of the minerals that host inclusions can be used to search
for trace elements diffusion gradients adjacent to melt inclusions, as has been done for Fe/Mg
diffusion (e.g., Danyushevsky et al. 2000). Such gradients could be useful in establishing the
presence and extent of diffusive exchange. In addition, analyses of trace element abundances
in host minerals adjacent to melt inclusions, can provide a direct indication of whether trace
element distribution between host and inclusion conform to expected equilibrium partitioning
relationships. Mineral-melt partition coefficients are generally well documented for most of
the common phenocryst phases in basaltic melts (e.g., Nielsen et al. 1992; Rollinson 1993;
Bindemann et al. 1998), although abundances of highly incompatible elements in some host
minerals may be more difficult to measure. A recent study by Weinsteiger et al. (2007) shows
that Ti partitioning between plagioclase phenocrysts and contained melt inclusions conformed
to expected KD (plagioclase-melt) values (0.03 ± 0.01) over a large range of individual
inclusion TiO2 contents. This suggests that Ti variations in inclusions accurately reflect the
compositions of trapped liquids, and not chemical fractionation related to inclusion trapping or
re-equilibration.
A final important point is that regardless of the rate of diffusive equilibration of trace
elements with the external melt, significant equilibration will only occur in the presence of a
gradient in chemical potential. For melt inclusions trapped within crystals that remain within an
external melt of similar composition there is little to drive changes in trapped melt compositions,
unless magmatic evolution significantly modifies the composition of the host melt, or
partitioning changes dramatically with changes in pressure or temperature. Gaetani and Watson
(2002) argue that polybaric magmatic evolution will provide a gradient for re-equilibration of
Fe/Mg within olivine-hosted inclusions, but that alone would provide little chemical gradient
in incompatible element abundances. Thus it would be possible for an inclusion to preserve
its trapped composition almost indefinitely—even if the rate of equilibration were rapid. It is
argued below that in the majority of cases the diversity of melt inclusions evident in individual
lavas is a reflection of mixing of magmas and melts of different compositions to produce the
eventual erupted lava compositions. If this is the case, and if final mixing and blending occurs
relatively late in the processes of melting, melt transport and crustal storage prior to eruption,
then melt inclusions could easily preserve a range of near-pristine mantle-derived compositions.
Differences in gradients in chemical potential may also prove an important difference between
experimental results, where gradients are generally steep by design and natural samples, where
the concentration differences between inclusion and external melt are typically much lower.
Diffusive loss or gain of H2O. A special case when considering diffusive re-equilibration
is the loss or gain of water by melt inclusions after trapping. For melt inclusions trapped
at relatively high temperatures, and or with slow cooling rates the diffusion rate of H2O as
H+ in olivine is relatively rapid (e.g., Mackwell and Kohlstead 1990). There are also recent
experimental results that suggest that diffusion of molecular water may also be rapid enough to
promote changes in melt inclusion water contents (Portnyagin et al. 2008).
310 Kent
If external hydrogen fugacity differs from that within the inclusion, then H2O loss or gain
via diffusion can occur during slow cooling or laboratory reheating (Roedder 1979; Sobolev
and Danyushevsky 1994; Danyushevsky et al. 2000; Massare et al. 2002; Hauri et al. 2002;
Rowe et al. 2007). This may be especially germane to studies where samples come from the
massive interiors of thick lava flows, where cooling rates may be relatively slow relative to flow
tops or tephra. For this reason many studies, particularly for subduction-related volcanism, have
successfully focused on studies of scoria and other rapidly-cooled tephra (e.g., Anderson 1975;
Roggensack et al. 1997; Cervantes and Wallace 2003a; Benjamin et al. 2007; Wade et al. 2007)
and have measured high water contents. Unfortunately for studies based on older portions of
the geologic record, tephra and other rapidly cooled materials such as flow tops are often poorly
preserved and the massive portions of flows may be the only sample materials available.
Danyushevsky et al. (2002a) argued that loss or gain of H2O via H+ diffusion requires
dissociation of H2O and is limited by the amount of Fe2+ available, because few other species
are available to accommodate the dissociation. Thus the potential H2O loss is limited by
(Danyushevsky et al. 2002a):
H2Oloss (wt%) = 0.125 × FeO (wt%) (2)
and is typically < 1 wt% in most basaltic and related rocks (Danyushevsky et al. 2002a). Further
H loss is possible if Fe2+ is available in the host crystal, as in olivine, but this is marked by the
presence of fine magnetite “dust” adjacent to the inclusion wall which is readily identifiable in
reflected light or using BSE imagery (Nielsen et al. 1998; Danyushevsky et al. 2002a). Gain
of water via hydrogen diffusion is likewise limited by the amount of Fe3+ available to reduce
via (Danyushevsky et al. 2002a):
H2Ogain (wt%) = 0.113 × Fe2O3 (wt%) (3)
Given the lower Fe2O3 contents in most basaltic magmas potential H2O gain in this manner
is probably more limited (Danyushevsky et al. 2002a).
From these constraints H2O loss in this manner is probably proportionally less significant
in volatile rich magmas such as those found in subduction zones (e.g., Sisson and Layne 1993;
Cervantes and Wallace 2003a; Walker et al. 2003; Gurenko et al. 2005), but could be more
significant for inherently drier MORB or OIB magmas (e.g., Dixon et al. 1991, 2002; Kent et
al. 1999a; Saal et al. 2002).
Recent experimental work by Portnyagin et al. (2008) suggest that diffusion of molecular
water through olivine may also be sufficiently rapid (similar to H+ diffusion) to modify the
water contents of trapped melt inclusions, although the actual diffusion mechanism remains
unclear. Portnyagin et al. (2008) showed that “dry” melt inclusions (H2O < 0.5 wt%) in olivine
from the Galapagos Plateau were able to acquire as much as 2.5 wt% H2O after two days in
water-bearing melt at 200 MPa and 1140 °C. This water gain is very rapid, and suggests that
it could easily occur within natural inclusion suites if water contents of the external melt were
modified by crystal fractionation, magma mixing, degassing or other petrological processes.
This observation may also explain decoupling between water and components of similar bulk
partition coefficient evident in some MORB and subduction related melts (e.g., Sobolev and
Chaussidon 1996; Portnyagin et al. 2008), although in some cases this could also be caused by
diffusive decoupling of water and lithophile trace elements (e.g., REE) during long term mantle
storage prior to melting (Workman et al. 2006).
Many workers have used the covariation between water and other elements of either similar
volatility or mantle-melting behavior to investigate the veracity of water contents measured in
basaltic melt inclusions. Many studies show that systematic variations, related either to coherent
trends among volatile elements (CO2, Cl, S) related to degassing (e.g., Roggensack 1997; Cer-
vantes and Wallace 2003a; Kamenetsky et al. 2007) or consistent covariation between H2O and
Melt Inclusions in Basaltic and Related Volcanic Rocks 311
elements with similar bulk partition coefficients during mantle melting or crystal fractionation
(e.g., K2O, La, Ce, Nd, Ba; e.g., Cervantes and Wallace 2003b; Saal et al. 2005; Wade et al. 2006;
Sadofsky et al. 2008).
Other chemical parameters can also be used to recognize H2O mobility. Diffusion of H will
result in changes in hydrogen isotope composition due the slower diffusivities of D relative
to H. Loss of H2O equates with higher δD as H diffuses more rapidly than D. Hauri (2002)
reported results of a study of olivine hosted melt inclusions from Hawaiian subaerial lavas and
demonstrated that inclusions from the massive portions of lavas could record unmodified H2O
contents, but differentiation of these from those that had lost H2O via diffusion required careful
evaluation of H2O and δD, and moreover samples that preserved unmodified H2O contents were
not consistently related to flow morphology of other geologic relationships that could be used
in a predictive sense.
The oxidation state of sulfur, as determined by SKα peak shift methods is also a sensitive
indicator of redox changes in trapped melts related to water loss or gain via H+ diffusion—
particularly when H2O is considered in units of weight percent. This can potentially be used to
determine whether inclusions have lost water, particularly in companion with other indices of
redox state, such as V/Sc ratios (e.g., Canil 1997; Li and Lee 2004). Rowe et al. (2007) show
that loss or gain of just 0.1 wt% water would be associated with fO2 change of ~ 1 log unit,
which is well within the uncertainty of the SKα technique for most basaltic melt compositions
(Wallace and Carmichael 1994; Rowe et al. 2007).
OrIGIn OF COMPOSItIOnaL VarIatIOnS In MeLt InCLuSIOnS
A common and key observation concerning the chemical compositions of melt inclusions,
and one that has already been emphasized, is that they exhibit a high degree of compositional
diversity, even after correction for post-entrapment crystallization and diffusive Fe-loss. This
diversity is expressed in major and trace element abundances, in isotopic compositions, or a
combination of all of these, but in many cases variations in incompatible lithophile elements
or isotopic composition provide the most robust indications as they are less sensitive to the
effects of postentrapment crystallization and diffusive re-equilibration (see above). In many
cases the range of chemical compositions evident in melt inclusions from a single sample may
rival or exceed that present in the entire suite of host and associated lava or tephra samples (in
this instance the latter refers to lavas or tephra which are demonstrably products of the same
magmatic system as the host sample, although not necessarily part of a consanguinous series).
Examples of this phenomenon are shown in Figures 16-21 and are apparent from studies of
all environments where basaltic magma commonly erupt, including MORB (e.g., Sobolev and
Shimizu 1993; Shimizu 1998; Nielsen et al. 1995; Sours-Page et al. 1999, 2002; Slater et al.
2001; Font et al. 2007; Laubier et al. 2007); subduction zones (e.g., Kamenetsky et al. 1997;
Kent and Elliott 2002; Walker et al. 2003; Wallace and Cervantes 2003a; Rowe et al. 2008),
Ocean Islands (e.g., Saal et al. 1998; Kent et al. 1999a; Sobolev et al. 2000; Norman et al. 2002)
and Continental Flood Basalts (e.g., Kent et al. 2002a; Yaxley et al. 2005).
Despite the potential for modification of trapped melt compositions during entrapment or
post trapping equilibration, it is probable that the majority of compositional variations observed
in melt inclusion suites relate to directly variations in the compositions of the trapped melts
themselves. Evidence for this assertion come from a number of sources, with one of the most
compelling being that melt inclusion compositions often define compositional arrays that
overlie and/or extend those evident in host and associated lavas (e.g., Anderson and Brown
1972; Sobolev et al. 1983a,b, 1991; Sinton 1993; Sobolev and Danyushevsky 1994; Gurenko
and Chaussidon 1995, 1997; Kamenetsky et al. 1995, 1997; Kamenetsky 1995; Nielsen et al.
1995; Sobolev 1996; Portnyagin et al. 1997; Saal et al. 1998, 2002, 2005; Shimizu 1998; Kent
312 Kent
et al. 1999a, 2002a; Sours-Page 1999; Danyushevsky et al. 2000; Shimizu et al. 2001; Hansen
and Gronvold 2000; Slater et al. 2001; Kent and Elliott 2002; Norman et al. 2002; Cervantes
and Wallace 2003a,b; Gurenko et al. 2005; Jackson and Hart 2006; Benjamin et al. 2007; Kohut
et al. 2006; Ehlburg et al. 2007; Font et al. 2007; Kamenetsky and Gurenko 2007; Laubier et
al. 2007; Portnyagin et al. 2007; Rowe et al. 2007; Sadofsky et al. 2008; Johnson et al. 2008;
Vigouroux et al. 2008; Rowe et al. 2008 and many others). In addition melt inclusions may
also correlate with host mineral compositions (e.g., Kamenetsky et al. 2002; Kamenetsky and
Gurenko 2007; Elburg et al. 2007; Weinsteiger et al. 2007) and/or have phase equilibria that
infer that melt inclusion compositions are in equilibrium with the same source or phenocryst
mineralogy as the host lavas (Nielsen et al. 1995, 1998; Kamenetsky 1995; Sobolev 1996;
Shimizu 1998; Slater et al. 2001; Yaxley et al. 2005; Vigouroux et al. 2008).
If melt inclusions do inherit the bulk of their compositional diversity from the melts they
trap then two important questions are: (1) what is the ultimate source of this diversity, and
(2) why are inclusions more variable than the bulk compositions of the rocks that host them?
A view that is widely expressed in the literature is that the range of melt inclusions derives
directly from the diversity of melt compositions present within the host magmatic system, and
reflect the same range of well-documented petrological processes known to drive variations
within lavas, glasses and other bulk samples, including variations in mantle source composition
and mineralogy, differences in the degree and style of melting, melt mixing and variable melt
aggregation, crystallization and crystal-liquid separation, and assimilation and degassing. The
volumes of individual eruptive units in basaltic volcanoes vary, but even the smallest eruptions
are many, many orders of magnitude greater in volume than individual melt inclusions, which
rarely exceed a few thousand cubic microns. Erupted lava or tephra compositions are thus
considered to represent blended mixtures of a number of different melt compositions, whereas
melt inclusions sample this diversity at a smaller scale, consistent with their smaller volumes
and protect trapped melts from subsequent mixing and blending (e.g., Dungan and Rhodes
1978; Sinton et al. 1993; Sobolev and Shimizu 1993; Nielsen et al. 1995, 2000; Sobolev 1996;
Kent et al. 1999a, 2002a; Slater et al. 2001; Norman et al. 2002; Maclennan et al. 2003). In this
sense melt inclusions can be broadly considered the unmixed (or less-mixed) equivalents of
erupted lavas or tephra (e.g., Sobolev and Shimizu 1993; Sobolev 1996; Shimizu 1998; Kent et
al. 1999a; Slater et al. 2001; Maclennan et al. 2003). It is also important to note that although
the above model considers composition variations as spatial features of magma systems,
melt inclusions are also trapped at discrete times and thus may likewise preserve temporal
variations. Thus melt inclusions may preserve magmatic compositions from earlier stages in
the liquid line of descent relative to the bulk or glass compositions of erupted materials. It
is also probable that the analysis techniques used for bulk samples, which generally involve
fine crushing or dissolution of rock promote artificial “averaging” of material that may be
more heterogeneous at finer scales relevant to trapped melt inclusions. However, analyses
of matrix glasses in submarine basalts made using the same microbeam techniques used for
melt inclusions rarely shows compositional diversity at the 10-100 µm scale relevant to melt
inclusions (e.g., Kent et al. 1999a,b, 2002b, 2004b; Norman et al. 2002; Workman et al. 2006),
and glasses typically appear homogeneous within analytical uncertainties. A comparison of
the chemical variations evident in lava and melt inclusion samples from the Borgarhraun
flow of Theistareykir volcano, Iceland by Maclennan et al. (2003) showed that both showed
compositional variations related to incomplete mixing of the products of a single melting
column together with crystal liquid separation. The degree of variability in inclusions was
greater than that in whole rock lavas samples, consistent with inclusions sampling a volume of
material ~30 times less than lava samples. Variation in compatible element appeared to record
short length scale variation in crystal-liquid separation, presumably occurring within shallow
magma chambers, whereas incompatible element variations related to longer length scales
controlled by the outputs of the melting column.
Melt Inclusions in Basaltic and Related Volcanic Rocks 313
A general model where melt inclusions provide a greater sample of magmatic diversity is
supported by a number of observations. Firstly, as noted above, trends defined by melt inclusion
suites often extend and or overlie trends shown by bulk lava, tephra or matrix glass samples,
and compositional variations in individual melt inclusion suites have been related to almost
all of the various igneous processes that drive compositional variations in basaltic systems.
This includes variations in mantle source composition (e.g., Sinton et al. 1993; Nielsen et al.
1995; Sobolev and Chaussidon 1996; Kamenetsky et al. 1997; Saal et al. 1998; Sours-Page
et al. 1999, 2002; Sobolev et al. 2000; Kent and Elliott 2002; Wallace and Cervantes 2003a;
Elburg et al. 2007), variations in the degree and style of melting and in the mineralogy of the
mantle source (e.g., Sobolev and Shimizu 1993; Nielsen et al. 1995; Gurenko and Chaussidon
1995; Sobolev 1996; Shimizu 1998; Slater et al. 2001; Maclennan et al. 2003; Ehlburg et al.
2007; Lubier et al. 2007); mantle-melt interaction during melt transport (e.g., Shimizu et al.
1998), and crystal fractionation and assimilation (Kent et al. 1999a, 2002a; Lassiter et al.
2002; Danyushevsky et al. 2003, 2004; Yaxley et al. 2004; Kamenetsky and Gurenko 2007).
Secondly studies have also documented close correspondences between the calculated average
composition of melt inclusions from a given sample and that of their host lavas (e.g., Nielsen et
al. 1995; Kent et al. 1999a, 2002a; Kent and Elliot 2002; Slater et al. 2001; Norman et al. 2002;
Maclennan et al. 2003), which directly implies that the range of compositions sampled by melt
inclusions are also those that blend to produce individual lavas in which each inclusion suite
occurs (and where the correspondence is 1:1 that the proportion of each composition sampled
by inclusions is also approximately the same as its contribution to the host lava). An example is
shown in Figure 18 where K2O contents for lava compositions in MORB from the Endeavour
segment of the Juan de Fuca ridge correlate directly with the average K2O contents of melt
inclusions from each sample, despite the large range in K2O in inclusions from each individual
samples (Sours-Page 1999). Figure 19 shows trace element abundances in olivine hosted melt
inclusions from two crustally contaminated basaltic lavas from Yemen (Kent et al. 2002a).
Although trace elements, particularly those like Ba, K, U that are enriched in continental crust,
vary widely within individual inclusions, the average inclusion composition (after correction
for fractionation of olivine) is a close match to the host lava in each instance. In each of these
Figure 18. Comparison between K2O contents in host lavas versus those measured in plagioclase-hosted
melt inclusions for MORB from the Endeavour segment of the Juan de Fuca Ridge (modified from
Sours-Page et al. 1999). The average melt inclusion composition and overall range are shown as well as
representative histograms (dashed lines and grey fields) showing the relative proportions of K2O found in
melt inclusions from three different samples. Despite the large range of inclusion compositions there is a
close correspondence between the average composition and that of the host.
314 Kent
Figure 19. Trace element abundances measured in olivine hosted melt inclusions from two crustally
contaminated lavas (samples 281, 282) from Yemen. As shown in (A) and (B) inclusions show large
variations in primitive mantle normalized trace element abundances, including a small number of high-Sr
inclusions. (C) and (D) show average melt inclusion abundances normalized to primitive mantle and also
to Nb contents (the latter to correct for differences in crystal fractionation). Error bars reflect one standard
deviation in inclusion compositions. Despite the wide variation in trace element abundances the average
composition is a close match to the host lava. Modified from Kent et al. (2002a).
two examples the similarity of average inclusions to the host lava, even for elements that vary
widely in concentration, strongly suggest that inclusions and host sample the same range of
compositions, albeit at different scales. Although it has not yet been widely used within the
literature, comparisons between the variance or standard deviations of melt inclusion suites
and those determined for host and associated lavas or tephras can form an additional basis for
investigating the relationship of inclusions and bulk rock or matrix glass compositions (e.g.,
Fig. 17; Cottrell et al. 2002).
Some melt inclusion suites also appear to show changes in melt diversity that is consistent
with the above model and in so doing can also provide insight into the architecture and behavior
of basaltic magma systems. An example is given in Figure 20(A) where K2O/TiO2 and zr/Y
Melt Inclusions in Basaltic and Related Volcanic Rocks 315
Figure 20. (A) Comparison between K2O/TiO2 and Mg# (100xMg/[XMg + XFe2+] where X denotes mol
fraction) measured in plagioclase-hosted melt inclusions from the Southwest Indian Ridge (SWIR; Lund
2003). The inset shows zr/Y in MORB melt inclusions versus the forsterite content (Fo) of the host olivine
(modified from Sobolev 1996). (B) Average melt inclusion K2O/TiO2 ± 1s versus representative half
spreading rates for compiled MORB melt inclusion data sets for the Southwest Indian Ridge, Mid Atlantic
Ridge (MAR), Juan de Fuca Ridge (JdF), Gorda Ridge and East Pacific Rise (EPR). Data from Nielsen
et al. (1995); Kamenetsky 1996; Shimizu 1998; Sours-Page (1999, 2002); Lund (2003) and Nielsen, R.L.
(unpublished data).
(these elemental ratios are used here as indices of melt variation) are compared to measured
Mg#, or host olivine composition. Both examples show greater degrees of magmatic diversity in
the more primitive melts and reduced diversity in more evolved compositions. This is consistent
with a magmatic system where a range of diverse primitive melts are introduced and then mix
and homogenize as they differentiate to less evolved compositions. Not all melt inclusion suites
show this behavior, and this contrast may reflect differences in the timing of magmatic inputs,
mixing and eruption between different basaltic magma storage and transport systems. In other
cases melt inclusion diversity may also reflect the effect of external forcing factors. As an
example of the latter Figure 20(B) shows the apparent decrease in melt inclusion diversity (as
reflected by the standard deviation calculated for individual melt inclusion suites) with increasing
spreading rate for compiled MORB melt inclusion suites. The most diverse inclusion suites,
at least in terms of K2O/TiO2 come from the slower spreading centers. Although this might
also reflect variations in mantle source character between ridges, the difference is consistent
with the idea that established magma chambers in higher melt flux medium and fast spreading
centers promote greater degrees of magmatic processing and mixing. In slower spreading ridges
more transient magma supply limits the presence of permanent magma chambers, and thus
inclusions preserve more variable compositions (Michael and Cornell 1998; Sours-Page et al.
2002). Similar observations and arguments have been made for the comparison between melt
inclusion suites from ridge and off axis seamounts (Sours-Page et al. 2002).
From the above discussion it is now relatively straightforward to address the question
of what the diversity evident melt inclusions actually represents: in the context of the model
described above inclusion simply sample the compositional variations produced by the same
sets of petrological processes that drive compositional variations within erupted lavas, tephras
316 Kent
or matrix glasses. Melt inclusions sample this diversity at a finer scale, and may also sample
some compositions preferentially. Survival of melt inclusions with variable compositions
within crystals from the same hand sample, even with inclusions from the same phenocryst
(e.g., Sobolev et al. 2000; Kent et al. 2002a) shows that the rates of crystal transport within
a basaltic magmatic system (or the rates at which crystals from one melt are introduced into
another) are faster than the rates are which melts can mix and homogenize, or that inclusions
can re-equilibrate with external melt once trapped. It is interesting to note that an analogous
conclusion has also recently been made concerning the rates of crystal transport being
considerably more rapid than rates of crystallization and crystal re-equilibration in silicic
melts (Ruprecht et al. 2008). In addition, although we have argued that magmatic systems are
diverse and that individual erupted units may represent aggregation of a number of melts or
magmas of different composition, the relative homogeneity of matrix glasses in many samples
with variable melt inclusions shows that the mixing and blending of multiple melts does occur
efficiently prior to final quenching. This may occur relatively late in magmatic evolution (e.g.,
Kent et al. 1999a), and could well be related to final eruption and emplacement.
Finally, in the context of the above discussion it is worth restating some of the advantages
of using melt inclusions for studies of basaltic and related volcanic rocks. The greater degree of
variation evident in inclusion suites offers greater scope for using measured compositional to
characterize petrological processes and to test hypotheses. Inclusions may also provide access
to melts produced during stages of magmatic evolution that are not well represented in bulk
rocks samples—generally the initial stages of the liquid lines of descent, and to volumetrically
small or depleted melts that are likewise poorly represented in bulk compositions. This is not
to say that there will always be clear agreement as to the causes of melt inclusion variations—
even in apparently simple magmatic systems petrologists often argue over the causes of
compositional variations. There is no reason to expect this would not continue!
how representative are trapped melts?
The volumes of melt inclusions are exceedingly small compared to their host lavas
or tephras, and even smaller in comparison to overall size of the magmatic systems which
produce them. Thus extrapolation across many orders of magnitude is implicit when applying
petrologic information obtained from inclusions to entire magmatic systems.
One of the most difficult relationships to establish is that between host lavas and so
called “anomalous melt inclusions”. In contrast to previous definitions (e.g., Spandler et al.
2007) which consider any inclusions of different composition to the host rock as anomalous,
herein anomalous inclusion are defined as those which do not lie on compositional trends
defined by the majority of contained melt inclusions together with host and associated lavas
and tephra samples (e.g., Danyushevsky et al. 2003, 2004) and that can be reasonably be
ascribed to a common set of petrologic processes that drive both inclusions and whole rock or
glass compositions. Anomalous melt inclusions occur in many different melt inclusion suites
(e.g., Sobolev and Shimizu 1998; Sobolev et al. 2000; Kent et al. 2002a; Danyushevsky et
al. 2003, 2004) although they are subordinate within the majority of melt inclusion suites.
There are a number of possible origins for anomalous inclusions. They have been interpreted
as compositions that are present at volumetrically low fractions within the host magmatic
systems, and thus are effectively swamped by other compositions within bulk lava or
tephras, especially if the anomalous compositions have low trace element abundances (e.g.,
Sobolev and Shimizu 1993; Sobolev et al. 2000; Danyushevsky et al. 2003, 2004). In this
case inclusions may represent discrete melt compositions that are petrologically significant,
but volumetrically minor. Alternatively anomalous inclusions may have compositions that
have been strongly altered by inclusion-specific processes during entrapment and subsequent
diffusive equilibration and thus may not reflect any real melt composition present within the
host magmatic system (e.g., Michael et al. 2002). Finally anomalous melt inclusions may
Melt Inclusions in Basaltic and Related Volcanic Rocks 317
represent real melt compositions (i.e., melts that exist discretely within the magmatic system)
that are preferentially sampled by melt inclusions, and thus over-represented within inclusion
populations compared to their contributions to bulk magma compositions. One example of
the uncertainty concerning interpretation of the so-called “ultra-depleted melt” (UDM) and
“Sr-rich” melt inclusions found in some basaltic rocks. Sobolev and Shimizu (1993) reported
the first example of an UDM inclusion, characterized by extreme depletion in LREE and
other incompatible elements, from 9°N on the Mid Atlantic Ridge and interpreted this as
evidence for near-fractional melting of melt-depleted material from the top of a mantle melting
column, consistent with the compositions of clinopyroxene in abyssal peridotites (Johnson
et al. 1990). Subsequent studies have suggested that these compositions could also relate to
diffusion-driven fractionation during melt transport (Van Orman et al. 2002) or channelized
melt transport (Spiegelman and Kelemen 2003). Strontium-rich inclusions were reported from
Hawaiian lavas (Sobolev et al. 2000) but also occur in lavas from MORB, Icelandic basalts and
continental flood basalts (e.g., Figure 19, 21; Gurenko and Chaussidon 1995; Kamenetsky et
al. 1998; Kent et al. 2002a; Danyushevsky et al. 2004). Sobolev et al. (2000) interpreted Sr-rich
inclusions from Mauna Loa, Hawaii, to represent melts from gabbroic components recycled
within mantle plumes. The Sr-rich trace element signature is only apparent in small numbers
Figure 21. Examples of anomalous “ultra depleted” and “Sr-rich” melt inclusions from the (A) Mauna Loa,
Hawaii, and (B) the Mid Atlantic Ridge, and Iceland (Gurenko and Chaussidon 1995; Sobolev et al. 2000;
Danyushevsky et al. 2003). Normal composition melt inclusions are shown for comparison in both panels
and the field of Mauna Loa lava compositions is also shown in (A).
318 Kent
of inclusions, and for Hawaii is absent in whole rock compositions, although some MORB and
Icelandic basalts have Sr-rich characteristics (Fig. 21; Danyushevsky et al. 2003, 2004). For
example, only six out of 160 analyzed inclusions from Mauna Loa analyzed by Sobolev et al.
(2000) contain the Sr-rich signature, and a subsequent study by Norman et al. (2002) found no
Sr-rich compositions among an additional 80 analyzed inclusions from Mauna Loa. The low
abundance of Sr-rich and ultradepleted compositions complicates the interpretation of these as
mantle-derived compositions, and subsequent workers have suggested that they instead result
from interaction of basaltic melts and plagioclase-bearing crustal lithologies (Michael et al.
2002; Danyushevsky et al. 2003, 2004).
A final question to consider is whether the process of melt inclusion formation itself
can create a bias in the compositions of melts that are preserved. Crystallization is tied to the
thermal evolution of magmas, such that cooling promotes mineral growth, and melt inclusion
formation, and thus could also lead to preferential trapping and preservation of melts formed
in the cooler portions of magmatic systems within melt inclusion populations. Danyushevsky
et al. (2003, 2004) and Yaxley et al. (2004) have argued that melt compositions reflecting
mineral dissolution and assimilation of enriched melts may be preferentially trapped within
melt inclusions, as mineral dissolution, reaction and assimilation occurs together with cooling
and phenocryst growth in boundary layers where hot basaltic melts come into contact with
cooler wallrocks or crystal mushes and undergo rapid compositional change and crystallization
(e.g., Kent et al. 2002a). Importantly this could promote a bias where compositions reflect
interaction with wallrocks or crystal much zones being trapped in the most mafic phenocrysts.
This bias may also be exacerbated by selection of larger inclusions as more amenable to
microanalytical techniques (Danyushevsky et al. 2004).
MeLt InCLuSIOnS In BaSaLtIC and reLated rOCKS
The following is a brief summary of some the applications of melt inclusions to studies
of basaltic and related volcanic rocks, focused on the main geodynamic environments in
which basalts are found. The goal is not to provide a systematic or exhaustive survey, but to
give the reader a sense of the important contributions that melt inclusions have made towards
understanding of basaltic magmatism.
Mid-ocean ridges
Much of the early work on melt inclusions in basaltic rocks was focused on studies of MORB
and mid ocean ridge magma systems (e.g., Dungan and Rhodes 1978; Sinton et al. 1993), and
melt inclusions continue to play an important role in our understanding of the most voluminous
magmatic systems on earth. Sobolev and Shimizu (1993) in an influential paper identified the
presence of an “ultra-depleted” composition in an olivine-hosted melt inclusion (Fig. 21) from
the Mid Atlantic Ridge, and subsequently from other ridge localities (Sobolev et al. 1994).
These compositions were consistent with the highly depleted nature of clinopyroxenes from
abyssal peridotites (Johnson et al. 1990), and were interpreted to be depleted melts produced
during the final stages of near-fractional melting of adiabatically upwelling mantle beneath
ridges. Similar arguments suggesting that melt inclusions sample incompletely aggregated
melts have been made in other studies of MORB (Shimizu 1998; Laubier et al. 2007) and for
Iceland (Gurenko and Chaussidon 1995; Slater et al. 2001; Maclennan et al. 2003), and thus
melt inclusions provide important constraints on mantle melting and transport beneath ridges.
Subsequent studies have suggested alternate mechanisms for formation of ultra depleted melt
compositions, including fluxed melting of mantle harzburgite (Nielsen et al. 2000), diffusive
fractionation during melt transport (Van Orman et al. 2002), dissolution and reaction within
magma mush zones (Danyushevsky et al. 2003, 2004) or diffusional equilibration between
inclusions and trapped melts (Michael et al. 2002). However the results of Sobolev and Shimizu
Melt Inclusions in Basaltic and Related Volcanic Rocks 319
(1993) were also an early indication of the ability of basaltic melt inclusions to contain highly
variable trace element compositions, and provided a spur for subsequent work. Since then,
studies of MORB melt inclusions have focused on the nature of primitive MORB melts, and
their interaction with residual mantle and crystal-rich cumulates (e.g., Kamenetsky et al. 1998;
Kamenetsky and Crawford 1998; Shimizu 1998; Danyushevsky et al. 2003, 2004; Kamenetsky
and Gurenko 2006; Laubier et al. 2007; Font et al. 2007). In addition to MORB, melt inclusions
have also proven valuable to the study of ophiolite sequences (e.g., Sobolev et al. 1983b, 1996;
Portnyagin et al. 1997).
A number of studies of MORB focus on inclusions trapped within anorthite-rich plagio-
clase phenocrysts and megacrysts (e.g., Sinton et al. 1993; Nielsen et al. 1995; Sours-Page
et al. 1999, 2002; Hansen and Gronvold 2000; Lund 2003; Font et al. 2007) (e.g., Fig. 1G).
Although, as noted, questions remain about the ability of plagioclase to preserve some trace ele-
ment signatures (Cottrell et al. 2002; Danyushevsky et al. 2002a), studies of plagioclase provide
a tremendous amount of information concerning the diversity of compositions present within
MORB magma systems. Nielsen et al. (1995) exploited this diversity, together with the fact that
large numbers of inclusions could be analyzed, to devise statistical tests of models for melt pro-
duction and sampling beneath mid ocean ridges. This approach has not been widely employed;
however it should prove useful where data from large numbers of inclusions are available.
Melt inclusions have also contributed to an understanding of volatile contents in MORB
magmas and their depleted mantle source (Sobolev 1996; Sobolev and Chaussidon 1996; Saal et
al. 2002), demonstrating that MORB magmas are relatively dry. Saal et al. (2002) analyzed melt
inclusions from depleted magmas from the Siqueiros fracture zone and argued for constancy of
volatile/non volatile element ratios within magmas derived from the depleted mantle (e.g., CO2/
Nb, H2O/Ce, F/P, Cl/K and S/Dy). The constancy of these ratios is of assistance when studying
degassed samples as it allows estimation of primary volatile abundances using the compositions
of non-volatile trace elements, although it has subsequently been argued that CO2/Nb ratios are
more variable (Cartigny et al. 2008) than considered by Saal et al. (2002).
Subduction zones
Melt inclusion studies have had a large impact on understanding magma genesis in
subduction zones, as they provide a link between melt compositions and the abundances of
H2O and other volatile elements that are thought to drive partial melting (e.g., Stolper and
Newman 1994; Kelley et al. 2006). Anderson (1973, 1974) was the first to demonstrate high
water contents in melt inclusions from subduction-related basalts, and a large number of studies
have now used melt inclusions to constrain the pre-degassing contents of volatile elements
in basaltic and related melts within subduction zones (e.g., Sisson and Layne 1993; Sobolev
and Chaussidon 1996; Sisson and Bronto 1998; Newman et al. 2000; Cervantes and Wallace
2003a; Walker et al. 2003; Wallace 2005; Wade et al. 2006; Benjamin et al. 2007; Portnyagin
et al. 2007; Kamenetsky et al. 2007; Johnson et al. 2008; Sadofsky et al. 2008; Vigouroux et al.
2008). These studies provide convincing evidence that basaltic magmas from subduction zones
have significantly higher water contents (at least up to 5 wt% water) compared to basalts from
other geodynamic environments (e.g., Kovalenko et al. 2000; Wallace 2005)—a key observation
supporting models that consider that hydrous slab-derived fluxes to arc and back-arc magmas
responsible for initiating mantle partial melting. In addition, melt inclusion studies have also
demonstrated the presence of relatively dry magmas in some subduction zones (e.g., Sisson
and Layne 1993; Sisson and Bronto 1998), which are interpreted to result from decompression
melting related to flow within the mantle wedge.
Melt inclusions have been used to study changes in subduction related fluxes across or
along arc segments (e.g., Walker et al. 2003; Rowe et al. 2008) and in some cases show that
water contents may be decoupled from supposedly fluid-mobile elements (Walker et al. 2003),
challenging established views of element transport. Inclusions also provide evidence for the
320 Kent
complexity of subduction-related magma systems, highlighting variations produced during dif-
ferentiation (e.g., Sobolev and Danyushevsky 1994; Lee and Stern 1998; Elburg et al. 2007) and
demonstrating that significant variations in slab-derived trace element fluxes may occur within,
and between, individual arc melting systems (Kamenetsky et al. 1997; Kent and Elliott 2002;
Elburg et al. 2007). Finally melt inclusions also provide a means to relate volatile abundances to
eruption style (e.g., Roggensack et al. 1997; Cervantes and Wallace 2003a; Gurenko et al. 2005;
Kamenetsky et al. 2007), and so can be used to evaluate volcanic hazards.
Oceanic islands
Studies of melt inclusions in ocean island basalts (OIB) and related rocks are fewer
than those from MORB or subduction environments, but nevertheless are important for
understanding the nature and origin of mantle heterogeneity sampled by OIB magmas. Saal
et al. (1998), in one of the first applications of in situ Pb isotope studies in melt inclusions,
demonstrated that OIB source regions of individual basalts may sample material with highly
diverse isotopic compositions. Their results suggest tremendous heterogeneity in Pb isotopic
composition (comparable to the range shown by all terrestrial basalts in terms of 208Pb-207Pb-
206
Pb), at the scales sampled by individual lavas. This is an important insight given that the
range of isotopic compositions of OIB lavas largely defines the range of terrestrial basaltic
rocks, and is used to constrain the composition and evolution of the mantle (e.g., Hofmann
1997). Together with subsequent work (Saal et al. 2005) these measurements show that large
variations exist in melt inclusions from lavas representing the HIMU, EM-I and EM-II mantle
isotopic endmembers, suggesting that all lavas also probably contain a common Pb isotope
contribution from overlying depleted oceanic lithosphere. Subsequent studies of Pb isotopes in
OIB melt inclusions further emphasize the variability of mantle source regions that contribute
to individual lavas (Kobayashi et al. 2004; Yurimoto et al. 2004), and it has also been shown
that melt inclusions from individual OIB have measurable variations in H, Sr, B and Li isotope
signatures (Hauri 2002; Kobayashi et al. 2004; Jackson and Hart 2006).
Melt inclusions also provide a means to assess the chemical heterogeneity of OIB magmas
in more detail than possible using lava compositions alone. Norman et al. (2002) used a large
data set for olivine-hosted melt inclusions from Hawaiian picrites to argue for the effects of both
variations in degree of melting and source composition in controlling the composition of melts
erupted from Hawaiian volcanoes. They suggested that these inclusions sampled the Mauna
Loa isotopic component at larger degrees of melting. Sobolev et al. (2000) identified rare Sr-
rich melt inclusions from Mauna Loa lavas (e.g., Fig. 21) and argued that these represented
recycled gabbroic components within the Hawaiian plume—although subsequent studies
have argued that these may instead result from reaction or mineral dissolution within magma
transport systems (Danyushevsky et al. 2004).
Studies of OIB and related rocks also indicate contamination by assimilation of brine or
altered oceanic crustal components could be relatively common (Kent et al. 1999a; Lassiter et
al. 2002; Hauri 2002), and may play an important role in determining the eventual composition
of erupted OIB magmas. In addition, melt inclusions provide important constraints on the
volatile contents of primary OIB magmas (Wallace 1998; Lassiter et al. 2002; Davis et al.
2003), and show that degassing, crystal fractionation and contamination, as well as mantle
source variations, may all affect the contents of volatile elements. Some studies also show a
relationship between eruptive behavior and magmatic volatile contents in Hawaiian volcanoes,
and melt inclusions have been used to estimate pre-eruptive volatile contents and fluxes (e.g.,
Wallace and Anderson 1998).
Continental flood basalts
Even fewer studies of continental flood basalts (CFB) have been made using melt inclusions.
This may in part reflect the relatively limited appearance of olivine and other phenocryst phases
Melt Inclusions in Basaltic and Related Volcanic Rocks 321
in some CFB lavas sequences, although there is clear promise for future studies. Workers
from the former Soviet Union conducted studies on the Siberian Traps (e.g., Sobolev et al.
1981; Ryabchikov et al. 2001a,b, 2002) in an effort to constrain primary magma compositions
and conditions of magma formation. Elsewhere, studies have emphasized the role of crustal
interactions in controlling melt inclusion compositions (e.g., Kent et al. 2002a; Yaxley et al.
2004; Nielsen et al. 2006), although it is also clear that melt inclusions in CFB may preserve
primitive melts and mantle-derived compositional variations (e.g., Larsen and Pedersen 2000;
Kent et al. 2002a; Nielsen et al. 2006).
Melt inclusions have also been used to study the volatile contents of CFB melts. Thordarson
and Self (1996a) used measurements of S, Cl and F contents of melt inclusions in olivine
and plagioclase to estimate pre-degassing magmatic volatile contents, and together with a
volatile budget model (the so called “petrological method”) estimated the atmospheric release
of volatile elements during eruption of the voluminous Roza flow of the Columbia River
Flood Basalts. These authors applied this same melt-inclusion-based approach to estimate
atmospheric loading of volatile elements during other large basaltic fissure eruptions (e.g.,
Thordarson and Self 1996b).
COnCLudInG reMarKS
Melt inclusions constitute a powerful tool for studying basaltic magmas, and represent
a rich source of information about basaltic magma systems. Analysis of multiple inclusion
from just a single sample (or even in some cases a single crystal) can produce an array of melt
compositions that allow petrologists to study magmatic volatile contents, to better gauge the
range of compositions present within magmatic systems, to gain access to relatively primitive
melt compositions trapped in early-formed phenocrysts, and to study melt compositions in
altered rocks. In concert with careful petrological and geochemical study of the rocks and
minerals that host melt inclusions, as well as others from the same magmatic systems, melt
inclusions can provide unique information about the nature of magma generation and transport
and the compositions of produced melts and magmas.
Formation of melt inclusions appears to be a normal part of the evolution of most basaltic
magmas, with trapping probably occurring during slow crystal growth immediately after
episodes of rapid cooling and crystallization. Concerns exist that processes that lead to trapping
of melt inclusions may also alter the compositions of trapped melts and that inclusions may
subsequently equilibrate with their host mineral or external melts. A number of methods are
available for assessing the possibility and extent of modification of trapped melt inclusion
compositions, and the basis for most of these is a careful comparison between inclusions and
the compositions of rocks that host them. Such comparisons show that much of the variability
evident in most melt inclusion suites represents variations in the composition of trapped melts
themselves. The range of melt inclusion compositions observed in basaltic volcanic rocks tell
us that considerable compositional diversity exists within many basaltic magma systems, and
that this diversity is relatively poorly sampled by the bulk compositions of erupted lavas or
tephras, or by analyses of matrix glasses. Although it is often possible to reconstruct the major
and trace element compositions of inclusions after post entrapment crystallization and diffusive
modification, the most robust sources of information from melt inclusions are incompatible
elements that diffuse relatively slowly, and for most inclusions and mineral hosts this includes
most of the commonly used lithophile trace elements.
To date interest in the study of melt inclusions in basaltic rocks has been linked to
technical advances in the ability to make accurate chemical measurements in small sample
volumes. This should continue for the foreseeable future, in part driven by increasing access
to in situ trace element measurement techniques as LA-ICP-MS and (to a lesser extent) SIMS
322 Kent
instrumentation become more common. In addition, improvements and increased application
of techniques for more precise stable and radiogenic isotope measurements will provide new
dimensions for the study of melt inclusions.
Improvements in instrumentation and access are also leading to the generation of increasing
numbers of large multi-element (and in some cases isotopic) datasets from melt inclusions. In
addition online database efforts within the petrological community are making compilation of
large published datasets easier. This is creating a tremendous amount of data and a companion
effort to develop alternate techniques for analysis and synthesis of such large data sets is also
warranted. One option is expansion of the multivariate analysis techniques used by Slater et al.
(2001), although few other workers have explored this approach to date.
Finally, there is also an important role for continued studies of the processes of inclusion
formation and evolution using natural and experimental samples. Recent experiment studies
have raised important questions about the effect of trapping and diffusional re-equilibration
on melt inclusion compositions. However, to date there have not been equivalent attempts to
relate these observations to the compositional variations evident in natural inclusion suites.
Reproduction of complex natural kinetic phenomenon, such as those involved in formation
and evolution of melt inclusions, is difficult within the laboratory environment and on non-
geologic timescales. Thus, although continued experimental studies of melt inclusion formation
and evolution processes are vital, they are more valuable when accompanied by attempts to
reconcile their results with observations of the natural melt inclusion record.
aCKnOWLedGMentS
The author would like to thank colleagues and students past and present for discussions
concerning melt inclusions and related subjects, particularly Paul Wallace, Michael Wiedenbeck,
Doug Phinney, Michael Rowe, Marc Norman, Ian Hutcheon, Joel Baker, David Peate, Ingrid
Peate, Roger Nielsen, Ed Kohut, John Eiler, Jon Woodhead, Janet Hergt, Steve Eggins, Alex
Sobolev, Glenn Gaetani, Nobu Shimizu, Greg Yaxley, Erik Hauri and David Graham. The
author would also particularly like to thank Ed Stolper for providing early encouragement to
an economic geologist interested in petrology. The author also gratefully acknowledges the
ongoing support of the National Science Foundation.
Several reviewers, including Katie, Kelley, Liz Cottrell, Leonid Danyushevsky, Keith
Putirka and Benedetto de Vivo also provided valuable feedback on early versions of this
chapter.
reFerenCeS
Anderson AT (1973) The before-eruption H2O content of some high alumina magmas. B Volcanol 37:530-552
Anderson AT (1974) Evidence for a picritic, volatile-rich magma beneath Mt Shasta, California. J Petrol 15:243-
267
Anderson AT, Wright TL (1972) Phenocrysts and glass inclusions and their bearing on oxidation and mixing of
basaltic magmas, Kilauea Volcano, Hawaii. Am Mineral 57:188-216
Anderson AT, Brown GG (1975) CO2 and formation pressures of some Kilauean melt inclusions. Am Mineral
78:794-803
Baker DR (2008) The fidelity of melt inclusions as records of melt composition. Contrib Mineral Petrol doi:
10.1007/s00410-008-0291-3
Benjamin ER, Plank T, Wade JA, Kelley KA, Haun EH, Alvarado GE (2007) High water contents in basaltic
magmas from Irazu Volcano, Costa Rica. J Volcanol Geotherm Res 168:68-92
Bindeman IN, Davis AM, Drake MJ (1998) Ion microprobe study of plagioclase-basalt partition experiments at
natural concentration levels of trace elements. Geochim Cosmochim Acta 62:1175-1193
Blundy J, Cashman K (2008) Petrologic reconstruction of magmatic system variables and processes. Rev
Mineral Geochem 69:179-239
Melt Inclusions in Basaltic and Related Volcanic Rocks 323
Bombardieri DJ, Norman MD, Kamenetsky VS, Danyushevsky LV (2005) Major element and primary sulfur
concentrations in Apollo 12 mare basalts: The view from melt inclusions. Meteorit Planet Sci 40:679-693
Bonnin-Mosbah M, Metrich N, Susini J, Salome M, Massare D, Menez B (2002) Micro x-ray absorption near
edge structure at the sulfur and iron K-edges in natural silicate glasses. Spectrochim Acta B 57:711-725
Buening DK, Buseck PR (1973) Fe-Mg lattice diffusion in olivine. J Geophys Res 78:6852-6862
Canil D (1997) Vanadium partitioning and the oxidation state of Archaean komatiite magmas. Nature 389:842-
845
Cartigny P, Pineau F, Aubaud C, Javoy M (2008) Towards a consistent mantle carbon flux estimate: Insights
from volatile systematics (H2O/Ce, delta D, CO2/Nb) in the North Atlantic mantle (14 degrees N and 34
degrees N). Earth Planet Sci Lett 265:672-685
Cervantes P, Wallace P (2003a) Magma degassing and basaltic eruption styles: a case study of similar to 2000
year BP xitle volcano in central Mexico. J Volcanol Geotherm Res 120:249-270
Cervantes P, Wallace PJ (2003b) Role of H2O in subduction-zone magmatism: New insights from melt inclusions
in high-Mg basalts from central Mexico. Geology 31:235-238
Cherniak DJ (2007) REE diffusion in olivine. Eos Trans, Am Geophys Union 88:MR13C-1397
Clocchiatti R, Massare D (1985) Experimental crystal-growth in glass inclusions - the possibilities and limits of
the method. Contrib Mineral Petrol 89:193-204
Cooper KM, Reid MR, Murrell MT, Clague DA (2001) Crystal and magma residence at Kilauea Volcano,
Hawaii: Th-230-Ra-226 dating of the 1955 east rift eruption. Earth Planet Sci Lett 184:703-718
Cottrell E, Spiegelman M, Langmuir CH (2002) Consequences of diffusive re-equilibration for the interpretation
of melt inclusions. Geochem Geophys Geosys 3: doi: 000175371900001
Danyushevsky LV, Della-Pasqua FN, Sokolov S (2000) Re-equilibration of melt inclusions trapped by magnesian
olivine phenocrysts from subduction-related magmas: petrological implications. Contrib Mineral Petrol
138:68-83
Danyushevsky LV, Leslie RAJ, Crawford AJ, Durance P (2004) Melt inclusions in primitive olivine phenocrysts:
The role of localized reaction processes in the origin of anomalous compositions. J Petrol 45:2531-2553
Danyushevsky LV, McNeill AW, Sobolev AV (2002a) Experimental and petrological studies of melt inclusions
in phenocrysts from mantle-derived magmas: an overview of techniques, advantages and complications.
Chem Geol 183:5-24
Danyushevsky LV, Perfit MR, Eggins SM, Falloon TJ (2003) Crustal origin for coupled ‘ultra-depleted’ and
‘plagioclase’ signatures in MORB olivine-hosted melt inclusions: evidence from the Siqueiros Transform
Fault, East Pacific Rise. Contrib Mineral Petrol 144:619-637
Danyushevsky LV, Sokolov S, Falloon TJ (2002b) Melt inclusions in olivine phenocrysts: Using diffusive re-
equilibration to determine the cooling history of a crystal, with implications for the origin of olivine-phyric
volcanic rocks. J Petrol 43:1651-1671
Davis MG, Garcia MO, Wallace P (2003) Volatiles in glasses from Mauna Loa Volcano, Hawai’i: implications
for magma degassing and contamination, and growth of Hawaiian volcanoes. Contrib Mineral Petrol
144:570-591
Dixon JE, Clague DA, Stolper EM (1991) Degassing history of water, Sulfur, and carbon in submarine lavas
from Kilauea Volcano, Hawaii. J Geol 99:371-394
Dixon JE, Stolper EM, Holloway JR (1995) An experimental study of water and carbon dioxide solubilities in
mid ocean ridge basaltic liquids .1. Calibration and solubility models. J Petrol 36:1607-1631
Dixon JE, Leist L, Langmuir C, Schilling JG (2002) Recycled dehydrated lithosphere observed in plume-
influenced mid-ocean-ridge basalt. Nature 420:385-389
Dunbar NW, Hervig RL (1992) Volatile and trace element composition of melt inclusions from the lower
Bandelier Tuff: Implications for magma chamber proceses and eruptive style. J Geophys Res 97:15151-
15170
Dungan MA, Rhodes JM (1978) Residual glasses and melt inclusions in basalts from Dsdp Legs 45 and 46 -
evidence for magma mixing. Contrib Mineral Petrol 67:417-431
Durrant SF (1999) Laser ablation inductively coupled plasma mass spectrometry: achievements, problems,
prospects. J Anal At Spectrom 14:1385-1404
Eggins SM, Kinsley LK, Shelley JMG (1998) Deposition and elemental fractionation processes during
atmospheric pressure laser sampling for analysis by ICP-MS. Appl Surf Sci 127-129:278-286
Elburg MA, Kamenetsky VS, Foden JD, Sobolev A (2007) The origin of medium-K ankaramitic arc magmas
from Lombok (Sunda arc, Indonesia): Mineral and melt inclusion evidence. Chem Geol 240:260-279
Fahey AJ, Goswami JN, McKeegan KD, zinner E (1987) 26Al, 244Pu, 50Ti and trace element abundances in
hibonite grains from CM and CV meteorites. Geochim Cosmochim Acta 51:329-350
Falloon TJ, Danyushevsky LV, Ariskin A, Green DH, Ford CE (2007) The application of olivine geothermometry
to infer crystallization temperatures of parental liquids: implications for the temperature of MORB
magmas. Chem Geol 241:207-233
Faure F, Schiano P (2005) Experimental investigation of equilibration conditions during forsterite growth and
melt inclusion formation. Earth Planet Sci Lett 236:882-898
324 Kent
Font L, Murton BJ, Roberts S, Tindle AG (2007) Variations in melt productivity and melting conditions along
SWIR (70 degrees E-49 degrees E): evidence from olivine-hosted and plagioclase-hosted melt inclusions.
J Petrol 48:1471-1494
Ford CE, Russell DG, Craven JA, Fisk MR (1983) Olivine liquid equilibria - temperature, pressure and
composition dependence of the crystal liquid cation partition-coefficients for Mg, Fe2+, Ca and Mn. J
Petrol 24:256-265
Freda C, Baker DR, Scarlato P (2005) Sulfur diffusion in basaltic melts. Geochim Cosmochim Acta 69:5061-
5069
Frezzotti ML (2001) Silicate-melt inclusions in magmatic rocks: applications to petrology. Lithos 55:273-299
Gaetani GA, Watson EB (2000) Open-system behaviour of olivine-hosted melt inclusions. Earth Planet Sci Lett
183:27-41
Gaetani GA, Watson EB (2002) Modeling the major-element evolution of olivine-hosted melt inclusions. Chem
Geol 183:25-41
Gebrande H (1982) Elastic wave velocities and constant of elasticity of rock and rock forming minerals. In:
Physical properties of rocks. G Angenheister (ed) Springer Verlag, Berlin, Lanalt-Bornsetin series, group
V, vol. 1b, 3.1.2, p 8-34
Goldstein SB, Luth RW (2006) The importance of cooling regime in the formation of melt inclusions in olivine
crystals in haplobasaltic melts. Can Mineral 44:1543-1555
Grove TL, Baker MB, Kinzler RJ (1984) Coupled CaAl-NaSi diffusion in plagioclase feldspar - experiments
and applications to cooling rate speedometry. Geochim Cosmochim Acta 48:2113-2121
Gunther D, Heinrich CA (1999) Comparison of the ablation behaviour of 266 nm Nd : YAG and 193 nm ArF
excimer lasers for LA-ICP-MS analysis. J Anal At Spectrom 14:1369-1374
Gurenko AA, Belousov AB, Trumbull RB, Sobolev AV (2005) Explosive basaltic volcanism of the Chikurachki
Volcano (Kurile arc, Russia): Insights on pre-eruptive magmatic conditions and volatile budget revealed
from phenocryst-hosted melt inclusions and groundmass glasses. J Volcanol Geotherm Res 147:203-232
Gurenko AA, Chaussidon M (1995) Enriched and depleted primitive melts included in olivine from Icelandic
tholeiites: Origin by continuous melting of a single mantle column. Geochim Cosmochim Acta 59:2905-
2917
Gurenko AA, Chaussidon M (1997) Boron concentrations and isotopic composition of the Icelandic mantle:
Evidence from glass inclusions in olivine. Chem Geol 135:21-34
Gurenko AA, Chaussidon M (2002) Oxygen isotope variations in primitive tholeiites of Iceland: evidence from
a SIMS study of glass inclusions, olivine phenocrysts and pillow rim glasses. Earth Planet Sci Lett 205:63-
79
Gurenko AA, Chaussidon M, Schmincke HU (2001) Magma ascent and contamination beneath one intraplate
volcano: Evidence from S and O isotopes in glass inclusions and their host clinopyroxenes from Miocene
basaltic hyaloclastites southwest of Gran Canaria (Canary Islands). Geochim Cosmochim Acta 65:4359-
4374
Halter WE, Pettke T, Heinrich CA, Rothen-Rutishauser B (2002) Major to trace element analysis of melt
inclusions by laser-ablation ICP-MS: methods of quantification. Chem Geol 183:63-86
Hansen H, Gronvold K (2000) Plagioclase ultraphyric basalts in Iceland: the mush of the rift. J Volcanol
Geotherm Res 98:1-32
Hattendorf B, Latkoczy C, Günther D (2003) Laser ablation ICP-MS. Anal Chem A 75:341A-347A
Harlou R, Pearson DG, Nowell GM, Davidson JP, Kent AJR (2005) Sr isotope studies of melt inclusions by
TIMS. Geochim Cosmochim Acta 69:A380
Hauri E (2002) SIMS analysis of volatiles in silicate glasses, 2: isotopes and abundances in Hawaiian melt
inclusions. Chem Geol 183:115-141
Hauri E, Wang JH, Dixon JE, King PL, Mandeville C, Newman S (2002) SIMS analysis of volatiles in silicate
glasses 1. Calibration, matrix effects and comparisons with FTIR. Chem Geol 183:99-114
Heinrich CA, Pettke T, Halter WE, Aigner-Torres M, Audetat A, Gunther D, Hattendorf B, Bleiner D, Guillong
M, Horn I (2003) quantitative multi-element analysis of minerals, fluid and melt inclusions by laser-
ablation inductively-coupled-plasma mass-spectrometry. Geochim Cosmochim Acta 67:3473-3497
Hervig RL, Mazdab FK, Williams P, Guan Y, Huss GR (2006) Limits of quantitative analysis by secondary ion
mass spectrometry. Chem Geol 227:83-99
Hinton RW (1995) Ion microprobe analysis in geology. In: Microprobe techniques in the earth sciences. Potts
PJ, Bowles JFW, Reed SJB, Cave MR (eds) Chapman and Hall, London, p 235-290
Hofmann AW (1988) Chemical differentiation of the Earth: The relationship between mantle, continental-crust,
and oceanic crust. Earth Planet Sci Lett 90:297-314
Hofmann AW (1997) Mantle geochemistry: The message from oceanic volcanism. Nature 385:219-229
Humphreys MCS, Kearns SL, Blundy JD (2006) SIMS investigation of electron-beam damage to hydrous,
rhyolitic glasses: Implications for melt inclusion analysis. Am Mineral 91:667-679
Jackson MG, Hart SR (2006) Strontium isotopes in melt inclusions from Samoan basalts: Implications for
heterogeneity in the Samoan plume. Earth Planet Sci Lett 245:260-277
Melt Inclusions in Basaltic and Related Volcanic Rocks 325
Jackson SE, Gunther D (2003) The nature and sources of laser induced isotopic fractionation in laser ablation-
multicollector-inductively coupled plasma-mass spectrometry. J Anal At Spectrom 18:205-212
Jakobsen JK, Veksler IV, Tegner C, Brooks CK (2005) Immiscible iron- and silica-rich melts in basalt
petrogenesis documented in the Skaergaard intrusion. Geology 33:885-888
Jambon A, Lussiez P, Clocchiatti R, Weisz J, Hernandez J (1992) Olivine growth-rates in a tholeiitic basalt - an
experimental-study of melt inclusions in plagioclase. Chem Geol 96:277-287
Jochum KP, Dingwell DB, Rocholl A, Stoll B, Hofmann AW, Becker S, Besmehn A, Bessette D, Dietze H-J,
Dulski P, Erzinger J, Hellebrand E, Hoppe P, Horn I, Janssens K, Jenner GA, Klein M, McDonough WF,
Maetz M, Mezger K, Müker C, Nikogosian IK, Pickhardt C, Raczek I, Rhede D, Seufert HM, Simakin
HG, Sobolev AV, Spettel B, Straub S, Vincze L, Wallianos A, Weckwerth G, Weyer S, Wolf D, zimmer M
(2000) The preparation and preliminary characterisation of eight geological MPI-DING reference glasses
for in-situ microanalysis. Geostandard Geoanal Res 24:87-133
Jochum KP, Stoll B, Herwig K, Willbold M, Amini M, Hofmann AW, Aarburg S, Abouchami W, Alard O,
Bouman C, Becker ST, Brätz H, de Bruin D, Canil D, Cornell D, Dalpé C, Danyushevsky LV, Dücking M,
Groschopf N, Günther D, Guillong M, de Hoog C-J, Höfer H, Horz K, Jacob D, Kasemann SA, Kent AJR,
Klemd R, Lahaye Y, Latkoczy C, Ludwig T, Mason P, Meixner A, Misawa K, Nash BP, Pfänder J, Premo
WR, Raczek I, Rosner M, Stracke A, Sun W, Tiepolo M, Vannuci R, Vennemann T, Wayne D, Woodhead
JD, zack T (2006) MPI-DING reference glasses for in-situ microanalysis: New reference values for element
concentrations and isotope ratios. Geochem Geophys Geosys 7, q02008, doi:10.1029/2005GC001060
Jochum KP, Willbold M, Raczek I, Stoll B, Herwig K (2005b) Chemical characterisation of the USGS reference
glasses GSA-1G, GSC-1G, GSD-1G, GSE-1G, BCR-2G, BHVO-2G and BIR-1G using EPMA, ID-TIMS,
ID-ICP-MS and LA-ICP-MS. Geostandard Geoanal Res 29:285-302
Jochum KP, Stoll B, Herwig K, Amini A, Abouchami W, Hofmann AW (2005b) Lead isotope ratio measurements
in geological glasses by laser ablation-sector field-ICP mass spectrometry (LA-SF-ICPMS). Int J Mass
Spectrom 242:281-289
Johnson J, Nielsen RL, Fisk M (1996) Plagioclase-hosted melt inclusions in the Steens basalt, southeastern
Oregon. Petrology 4:247-254
Johnson ER, Wallace PJ, Cashman KV, Granados HD, Kent AJR (2008) Magmatic volatile contents and
degassing-induced crystallization at Volcán Jorullo, Mexico: Implications for melt evolution and the
plumbing systems of monogenetic volcanoes. Earth Planet Sci Lett 269:477-486
Johnson KTM, Dick HJB, Shimizu N (1990) Melting in the oceanic upper mantle - an ion microprobe study of
diopsides in abyssal peridotites. J Geophys Res-Solid Earth Planets 95:2661-2678
Jurewicz AJG, Watson EB (1988) Cations in olivine. 2. Diffusion in olivine xenocrysts, with applications to
petrology and mineral physics. Contrib Mineral Petrol 99:186-201
Kamenetsky MB, Sobolev AV, Kamenetsky VS, Maas R, Danyushevsky LV, Thomas R, Pokhilenko NP, Sobolev
NV (2004) Kimberlite melts rich in alkali chlorides and carbonates: A potent metasomatic agent in the
mantle. Geology 32:845-848
Kamenetsky V (1996) Methodology for the study of melt inclusions in Cr-Spinel, and implications for parental
melts of MORB from Famous areas. Earth Planet Sci Lett 142:479-486
Kamenetsky VS (1995) New method for determining primitive melts - example of melt inclusions in spinel from
tholeiites of the famous area, Mid-Atlantic Ridge. Petrology 3:473-477
Kamenetsky VS, Crawford AJ, Eggins S, Muhe R (1997) Phenocryst and melt inclusion chemistry of near-axis
seamounts, Valu Fa Ridge, Lau Basin: insight into mantle wedge melting and the addition of subduction
components. Earth Planet Sci Lett 151:205-223
Kamenetsky VS, Eggins SM, Crawford AJ, Green DH, Gasparon M, Falloon TJ (1998) Calcic melt inclusions
in primitive olivine at 43°N MAR: evidence for melt-rock reaction/melting involving clinopyroxene-rich
lithologies during MORB generation. Earth Planet Sci Lett 160:115-132
Kamenetsky VS, Gurenko AA (2007) Cryptic crustal contamination of MORB primitive melts recorded in
olivine-hosted glass and mineral inclusions. Contrib Mineral Petrol 153:465-481
Kamenetsky VS, Pompilio M, Metrich N, Sobolev AV, Kuzmin DV, Thomas R (2007) Arrival of extremely
volatile-rich high-Mg magmas changes explosivity of Mount Etna. Geology 35:255-258
Kamenetsky VS, Sobolev AV, Eggins SM, Crawford AJ, Arculus RJ (2002) Olivine-enriched melt inclusions
in chromites from low-Ca boninites, Cape Vogel, Papua New Guinea: evidence for ultramafic primary
magma, refractory mantle source and enriched components. Chem Geol 183:287-303
Kamenetsky VS, Sobolev AV, Joron JL, Semet MP (1995) Petrology and geochemistry of Cretaceous ultramafic
volcanics from eastern Kamchatka. J Petrol 36:637-662
Kelley KA, Plank T, Grove TL, Stolper EM, Newman S, Hauri E (2006) Mantle melting as a function of water
content beneath back-arc basins. J Geophys Res-Solid Earth 111: doi: 000240948900002
Kent AJR, Clague DA, Honda M, Stolper EM, Hutcheon ID, Norman MD (1999b) Widespread assimilation of a
seawater-derived component at Loihi Seamount, Hawaii. Geochim Cosmochim Acta 63:2749-2761
326 Kent
Kent AJR, Peate DW, Newman S, Stolper EM, Pearce JA (2002b) Chlorine in submarine glasses from the Lau
Basin: seawater contamination and constraints on the composition of slab-derived fluids. Earth Planet Sci
Lett 202:361-377
Kent AJR, Baker JA, Wiedenbeck M (2002a) Contamination and melt aggregation processes in continental
flood basalts: constraints from melt inclusions in Oligocene basalts from Yemen. Earth Planet Sci Lett
202:577-594
Kent AJR, Clague DA, Honda M, Stolper EM, Hutcheon ID, Norman MD (1999b) Widespread assimilation of a
seawater-derived component at Loihi Seamount, Hawaii. Geochim Cosmochim Acta 63:2749-2761
Kent AJR, Elliott TR (2002) Melt inclusions in Mariana arc lavas: Implications for the formation and evolution
of arc magmas. Chem Geol 183:265-288
Kent AJR, Jacobsen B, Peate DW, Waight TE, Baker JA (2004a) Isotope dilution MC-ICP-MS rare earth element
analysis of geochemical reference materials NIST SRM 610, NIST SRM 612, NIST SRM 614, BHVO-2G,
BHVO-2, BCR-2G, JB-2, WS-E, W-2, AGV-1 and AGV-2. Geostandard Geoanal Res 28:417-429
Kent AJR, Norman MD, Hutcheon ID, Stolper EM (1999a) Assimilation of seawater-derived components in an
oceanic volcano: evidence from matrix glasses and glass inclusions from Loihi seamount, Hawaii. Chem
Geol 156:299-319
Kent AJR, Stolper EM, Francis D, Woodhead J, Frei R, Eiler J (2004b) Mantle heterogeneity during the
formation of the North Atlantic Tertiary Province: Constraints from trace element and Sr-Nd-Os-O isotope
systematics of Baffin Island picrites. Geochem Geophys Geosys 5: doi 000225030200004
Kent AJR, Ungerer CA (2005) Production of barium and light rare earth element oxides during LA-ICP-MS
microanalysis. J Anal At Spectrom 20:1256-1262
Kent AJR, Ungerer CA (2006) Analysis of light lithophile elements (Li, Be, B) by laser ablation ICP-MS:
Comparison between magnetic sector and quadrupole ICP-MS. Am Mineral 91:1401-1411
Kilinc A, Carmichael ISE, Rivers ML, Sack RO (1983) The ferric-ferrous ratio of natural silicate liquids
equilibrated in air. Contrib Mineral Petrol 83:136-140
Kobayashi K, Tanaka R, Moriguti T, Shimizu K, Nakamura E (2004) Lithium, boron, and lead isotope
systematics of glass inclusions in olivines from Hawaiian lavas: evidence for recycled components in the
Hawaiian plume. Chem Geol 212:143-161
Kohut E, Nielsen RL (2004) Melt inclusion formation mechanisms and compositional effects in high-An feldspar
and high-Fo olivine in anhydrous mafic silicate liquids. Contrib Mineral Petrol 147:684-704
Kohut EJ, Stern RJ, Kent AJR, Nielsen RL, Bloomer SH, Leybourne M (2006) Evidence for adiabatic
decompression melting in the Southern Mariana arc from high-Mg lavas and melt inclusions. Contrib
Mineral Petrol 152:201-221
Kovalenko VI, Naumov VB, Yarmolyuk VV, Dorofeeva VA (2000) Volatile components (H2O, CO2, Cl, F, and S)
in basic magmas of various geodynamic settings: Data on melt inclusions and quenched glasses. Petrology
8:113-144
Kress VC, Carmichael ISE (1991) The compressibility of silicate liquids containing Fe2O3 and the effect of
composition, temperature, oxygen fugacity and pressure on their redox states. Contrib Mineral Petrol
108:82-92
Kress VC, Ghiorso MS (2004) Thermodynamic modeling of post-entrapment crystallization in igneous phases.
J Volcanol Geotherm Res 137:247-260
Kuzmin DV, Sobolev AV (2004) Boundary layer contribution to the composition of melt inclusions in olivine.
Geochim Cosmochim Acta 68:A544-A544
Larsen LM, Pedersen AK (2000) Processes in high-Mg, high-T magmas: Evidence from olivine, chromite and
glass in palaeogene picrites from West Greenland. J Petrol 41:1071-1098
Lassiter JC, Hauri EH, Nikogosian IK, Barsczus HG (2002) Chlorine-potassium variations in melt inclusions
from Raivavae and Rapa, Austral Islands: constraints on chlorine recycling in the mantle and evidence for
brine-induced melting of oceanic crust. Earth Planet Sci Lett 202:525-540
LaTourrette T, Wasserburg GJ (1997) Self diffusion of europium, neodymium, thorium, and uranium in
haplobasaltic melt: The effect of oxygen fugacity and the relationship to melt structure. Geochim
Cosmochim Acta 61:755-764
LaTourrette T, Wasserburg GJ, Fahey AJ (1996) Self diffusion of Mg, Ca, Ba, Nd, Yb, Ti, zr, and U in
haplobasaltic melt. Geochim Cosmochim Acta 60:1329-1340
Laubier M, Schiano P, Doucelance R, Ottolini L, Laporte D (2007) Olivine-hosted melt inclusions and melting
processes beneath the FAMOUS zone (Mid-Atlantic Ridge). Chem Geol 240:129-150
Layne GD, Godon A, Webster JD, Bach W (2004) Secondary ion mass spectrometry for the determination of
delta Cl-37 Part I. Ion microprobe analysis of glasses and fluids. Chem Geol 207:277-289
Lee J, Stern RJ (1998) Glass inclusions in Mariana arc phenocrysts: A new perspective on magmatic evolution
in a typical intra-oceanic arc. J Geol 106:19-33
Li zxA, Lee CTA (2004) The constancy of upper mantle fO2 through time inferred from V/Sc ratios in basalts.
Earth Planet Sci Lett 228:483-493
Melt Inclusions in Basaltic and Related Volcanic Rocks 327
Liang Y, Richter FM, Davis AM, Watson EB (1996) Diffusion in silicate melts .1. Self diffusion in CaO-Al2O3-
SiO2 at 1500 degrees C and 1 GPa. Geochim Cosmochim Acta 60:4353-4367
Longerich HP, Jackson SE, Gunther D (1996) Laser ablation inductively coupled plasma mass spectrometric
transient signal data acquisition and analyte concentration calculation. J Anal At Spectrom 11:899-904
Lowenstern JB (1994) Chlorine, Fluid immiscibility, and degassing in peralkaline magmas from Pantelleria,
Italy. Am Mineral 79:353-369
Lowenstern JB (1995) Application of silicate melt-inclusions to the study of magmatic volatiles. In: Magmas,
Fluids and Ore Deposits. Thompson J (ed) Mineral Assoc Canada Short Course, 23, p 71-100
Lowenstern JB (2003) Melt inclusions come of age: volatiles, volcanoes and Sorby’s legacy. In: Melt inclusions
in volcanic systems. de Vivo B, Bodnar RJ (eds) Elsevier, Amsterdam, p 258
Lu Fq, Anderson AT, Davis AM (1995) Diffusional gradients at the crystal/melt interface and their effect on the
composition of melt inclusions. J Geol 103:591-597
Lund MJ (2003) Plagioclase-hosted melt inclusion along the Southwest Indian Ridge. MS Thesis, University of
Copenhagen, Copenhagen
Mackwell SJ, Kohlstedt DL (1990) Diffusion of hydrogen in olivine: Implications for water in the mantle. J
Geophys Res 95:5079-5088
Maclennan J, McKenzie D, Hilton F, Gronvold K, Shimizu N (2003) Geochemical variability in a single flow
from northern Iceland. J Geophys Res- Solid Earth 108: doi: 000181501800001
Mangan M, Sisson T (2000) Delayed, disequilibrium degassing in rhyolite magma: decompression experiments
and implications for explosive volcanism. Earth Planet Sci Lett 183:441-455
Massare D, Metrich N, Clocchiatti R (2002) High-temperature experiments on silicate melt inclusions in olivine
at 1 atm: inference on temperatures of homogenization and H2O concentrations. Chem Geol 183:87-98
Mathez EA (1976) Sulfur solubility and magmatic sulfides in submarine basalt glass. J Geophys Res 81:4269-
4276
McDonough WF, Ireland TR (1993) Intraplate origin of komatiites inferred from trace elements in glass
inclusions. Nature 365:432-434
Métrich N, Wallace PJ (2008) Volatile abundances in basaltic magmas and their degassing paths tracked by melt
inclusions. Rev Mineral Geochem 69:363-402
Metrich N, Susini J, Foy E, Farges F, Massare D, Sylla L, Lequien S, Bonnin-Mosbah M (2006) Redox state
of iron in peralkaline rhyolitic glass/melt: x-ray absorption micro-spectroscopy experiments at high
temperature. Chem Geol 231:350-363
Meyzen CM, Toplis MJ, Humler E, Ludden JN, Mevel C (2003) A discontinuity in mantle composition beneath
the southwest Indian ridge. Nature 421:731-733
Michael PJ, Cornell WC (1998) Influence of spreading rate and magma supply on crystallization and assimilation
beneath mid-ocean ridges: Evidence from chlorine and major element chemistry of mid-ocean ridge
basalts. J Geophys Res-Solid Earth and Planets 103:18325-18356
Michael PJ, McDonough WF, Nielsen RL, Cornell WC (2002) Depleted melt inclusions in MORB plagioclase:
messages from the mantle or mirages from the magma chamber? Chem Geol 183:43-61
Milman-Barris MS, Beckett JR, Baker MB, Hofmann AE, Morgan z, Crowley MR, Vielzeuf D, Stolper E
(2008) zoning of phosphorus in igneous olivine. Contrib Mineral Petrol 155:739-765
Misener DJ (1974) Cation Diffusion in Olivine. In: Geochemical Transport and Kinetics. Hofman AW, Giletti
BJ, Yoder HS, Yund RA (eds) Carnegie Institution of Washington, Washington, 53, p 117-129
Moore G (2008) Interpreting H2O and CO2 contents in melt inclusions: constraints from solubility experiments
and modeling. Rev Mineral Geochem 69:333-361
Murase T, Mcbirney AR (1973) Properties of Some Common Igneous Rocks and Their Melts at High-
Temperatures. Geol Soc Am Bull 84:3563-3592
Nakamura M, Shimakita S (1998) Dissolution origin and syn-entrapment compositional change of melt inclusion
in plagioclase. Earth Planet Sci Lett 161:119-133
Nekvasil H, Filiberto J, McCubbin FM, Lindsley DH (2007) Alkalic parental magmas for chassignites? Meteorit
Planet Sci 42:979-992
Newman S, Stolper EM, Stern RJ (2000) H2O and CO2 from the Mariana arc and backarc systems. Geochem
Geophys Geosys 1: doi:1999GC000027
Nielsen RL, Crum J, Bourgeois R, Hascall K, Forsythe LM, Fisk MR, Christie DM (1995) Melt inclusions in
high-An plagioclase from the Gorda Ridge: an example of the local diversity of MORB parent magmas.
Contrib Mineral Petrol 122:34-50
Nielsen RL, Gallahan WE, Newberger F (1992) Experimentally determined mineral-melt partition-coefficients
for Sc, Y and Ree for olivine, ortho-pyroxene, pigeonite, magnetite and ilmenite. Contrib Mineral Petrol
110:488-499
Nielsen RL, Michael PJ, Sours-Page R (1998) Chemical and physical indicators of compromised melt inclusions.
Geochim Cosmochim Acta 62:831-839
328 Kent
Nielsen RL, Sours-Page R, Harpp K (2000) Role of a Cl-bearing flux in the origin of depleted ocean floor
magmas. Geochem Geophys Geosys 1, doi:1999GC000017
Nielsen TFD, Veksler IV (2002) Is natrocarbonatite a cognate fluid condensate? Contrib Mineral Petrol 142:425-
435
Nielsen TFD, Turkov VA, Solovova IP, Kogarko LN, Ryabchikov ID (2006) A Hawaiian beginning for the
Iceland plume: Modelling of reconnaissance data for olivine-hosted melt inclusions in Palaeogene picrite
lavas from East Greenland. Lithos 92:83-104
Niu HS, Houk RS (1996) Fundmental aspects of ion extraction in ICP-MS. Spectrochim Acta B 51:779-815
Norman MN, Garcia MO, Kamenetsky VS, Nielsen RL (2002) Olivine-hosted melt inclusions in Hawaiian
picrites: Equilibration, melting and plume source characteristics. Chem Geol 183:143-168
Paul B (2006) A new perspective on melt inclusions: Development of novel in-situ analytical protocols. Ph.D.
dissertation, University of Melbourne, Melbourne
Paul B, Woodhead JD, Hergt J (2005) Improved in situ isotope analysis of low-Pb materials using LA-MC-ICP-
MS with parallel ion counter and Faraday detection. J Anal At Spectrom 20:1350-1357
Pearce NJG, Perkins WT, Westgate JA, Gorton MP, Jackson SE, Neal CR, Chenery SP (1997) A compilation of
new and published major and trace element data for NIST SRM 610 and NIST SRM 612 glass reference
materials. Geostandard Newslett 21:115-144
Perkins WT, Pearce NJG (1995) Mineral microanalysis by laserprobe inductively coupled plasma mass
spectrometry. In: Microprobe Techniques in the Earth Sciences. Potts PJ, Bowles JFW, Reed SJB, Cave
MR (eds) Chapman and Hall, London, p 291-325
Pettke T, Halter WE, Webster JD, Aigner-Torres M, Heinrich CA (2004) Accurate quantification of melt inclusion
chemistry by LA-ICPMS: a comparison with EMP and SIMS and advantages and possible limitations of
these methods. Lithos 78:333-361
Portnyagin M, Almeev R, Matveev S, Holtz F (2008) Experimental evidence for rapid water exchange between
melt inclusions in olivine and host magmas. Earth Planet Sci Lett 272:541-552
Portnyagin M, Hoernle K, Plechov P, Mironov N, Khubunaya S (2007) Constraints on mantle melting and
composition and nature of slab components in volcanic arcs from volatiles (H2O, S, Cl, F) and trace
elements in melt inclusions from the Kamchatka Arc. Earth Planet Sci Lett 255:53-69.
Portnyagin MV, Danyushevsky LV, Kamenetsky VS (1997) Coexistence of two distinct mantle sources during
formation of ophiolites: a case study of primitive pillow-lavas from the lowest part of the volcanic section
of the Troodos Ophiolite, Cyprus. Contrib Mineral Petrol 128:287-301
Putirka KD (2005) Mantle potential temperatures at Hawaii, Iceland, and the mid-ocean ridge system, as inferred
from olivine phenocrysts: Evidence for thermally driven mantle plumes. Geochem Geophy Geosy 6, doi:
000229144800002
Putirka KD, Perfit M, Ryerson FJ, Jackson MG (2007) Ambient and excess mantle temperatures, olivine
thermometry, and active vs. passive upwelling. Chem Geol 241:177-206
qin z, Lu F, Anderson ATJ (1992) Diffusive re-equilibration of melt and fluid inclusions. Am Mineral 77:565-
576
Raczek I, Stoll B, Hofmann AW, Jochum KP (2001) High-precision trace element data for the USGS reference
materials BCR-1, BCR-2, BHVO-1, BHVO-2, AGV-1, AGV-2, DTS-1, DTS-2, GSP-1 and GSP-2 by ID-
TIMS and MIC-SSMS. Geostandard Newslett 25:77-86
Ramos FC, Tepley FJ III (2008) Inter- and intracrystalline isotopic disequilibria: techniques and applications.
Rev Mineral Geochem 69:403-443
Richter FM, Davis AM, DePaolo DJ, Watson EB (2003) Isotope fractionation by chemical diffusion between
molten basalt and rhyolite. Geochim Cosmochim Acta 67:3905-3923
Rocholl A (1998) Major and trace element composition and homogeneity of microbeam reference material:
Basalt glass USGS BCR-2G. Geostand Newslett - J Geostand Geoanal 22:33-45
Rocholl A, Dulski P, Raczek I (2000) New ID-TIMS, ICP-MS and SIMS data on the trace element composition
and homogeneity of NIST certified reference material SRM 610-611. Geostand Newslett - J Geostand
Geoanal 24:261-274
Roedder E (1979) Origin and significance of magmatic inclusions. Bull Mineral 102:487-510
Roedder E (1984) Fluid Inclusions. Rev Mineral 12:1-644
Roedder E, Emslie RF (1970) Olivine-liquid equilibrium. Contrib Mineral Petrol 29:275-289
Roggensack K, Hervig RL, McKnight SB, Williams SN (1997) Explosive basaltic volcanism from Cerro Negro
volcano: Influence of volatiles on eruptive style. Science 277:1639-1642
Rollinson HR (1993) Using Geochemical Data: Evaluation, Presentation, Interpretation. Longman, Singapore
Rose EF, Shimizu N, Layne GD, Grove TL (2001) Melt production beneath Mt. Shasta from boron data in
primitive melt inclusions. Science 293:281-283
Rowe MC, Kent AJR, Nielsen RL (2007) Determination of sulfur speciation and oxidation state of olivine
hosted melt inclusions. Chem Geol 236:303-322
Rowe MC, Kent AJR, Nielsen RL (2008) Subduction influence on basaltic oxygen fugacity and trace- and
volatile-elements across the Cascade Volcanic Arc. J Petrol (in press)
Melt Inclusions in Basaltic and Related Volcanic Rocks 329
Rowe MC, Nielsen RL, Kent AJR (2006) Anomalously high Fe contents in rehomogenized olivine-hosted melt
inclusions from oxidized magmas. Am Mineral 91:82-91
Ruprecht P, Bergantz GW, Dufek J (2008) Modeling of gas-driven magmatic overturn: Tracking of phenocryst
dispersal and gathering during magma mixing. Geochem Geophys Geosys doi:10.1029/2008GC002022
Russo RE, Mao xL, Liu HC, Gonzalez J, Mao SS (2002) Laser ablation in analytical chemistry - a review.
Talanta 57:425-451
Ryabchikov ID, Solovova IP, Ntaflos T, Buchl A, Tikhonenkov PI (2001) Subalkaline picrobasalts and plateau
basalts from the Putorana plateau (Siberian continental flood basalt province): II. Melt inclusion chemistry,
composition of “primary” magmas and P-T regime at the base of the superplume. Geochem Int 39:432-
446
Ryabchikov IA, Solovova IP, Kogarko LN, Bray GP, Ntaflos T, Simakin SG (2002) Thermodynamic parameters
of generation of meymechites and alkaline picrites in the Maimecha-Kotui province: Evidence from melt
inclusions. Geochem Int 40:1031-1041
Ryabchikov ID, Ntaflos T, Buchl A, Solovova IP (2001) Subalkaline picrobasalts and plateau basalts from the
Putorana plateau (Siberian continental flood basalt province): I. Mineral compositions and geochemistry
of major and trace elements. Geochem Int 39:415-431
Saal AE, Hart SR, Shimizu N, Hauri EH, Layne GD (1998) Pb isotopic variability in melt inclusions from
oceanic island basalts, Polynesia. Science 282:481-484
Saal AE, Hart SR, Shimizu N, Hauri EH, Layne GD, Eiler JM (2005) Pb isotopic variability in melt inclusions
from the EMI-EMII-HIMU mantle end-members and the role of the oceanic lithosphere. Earth Planet Sci
Lett 240:605-620
Saal AE, Hauri EH, Langmuir CH, Perfit MR (2002) Vapour undersaturation in primitive mid-ocean-ridge basalt
and the volatile content of Earth’s upper mantle. Nature 419:451-455
Sadofsky SJ, Portnyagin M, Hoernle K, van den Bogaard P (2008) Subduction cycling of volatiles and trace
elements through the Central American volcanic arc: evidence from melt inclusions. Contrib Mineral
Petrol 155:433-456
Schiano P (2003) Primitive mantle magmas recorded as silicate melt inclusions in igneous minerals. Earth Sci
Rev 63:121-144
Schiano P, Bourdon B (1999) On the preservation of mantle information in ultramafic nodules: glass inclusions
within minerals versus interstitial glasses. Earth Planet Sci Lett 169:173-188
Schiano P, Clocchiatti R, Boivin P, Medard E (2004) The nature of melt inclusions inside minerals in an
ultramafic cumulate from Adak volcanic center, Aleutian arc: implications for the origin of high-Al basalts.
Chem Geol 203:169-179
Schiano P, Clocchiatti R, Lorand JP, Massare D, Deloule E, Chaussidon M (1997) Primitive basaltic melts
included in podiform chromites from the Oman Ophiolite. Earth Planet Sci Lett 146:489-497
Schiano P, Eiler JM, Hutcheon ID, Stolper EM (2000) Primitive CaO-rich, silica-undersaturated melts in island
arcs: Evidence for the involvement of clinopyroxene-rich lithologies in the petrogenesis of arc magmas.
Geochem Geophys Geosys 2, doi: 1999GC00032
Schiano P, Provost A, Clocchiatti R, Faure F (2006) Transcrystalline melt migration and Earth’s mantle. Science
314:970-974
Scowen PAH, Roeder PL, Helz RT (1991) Reequilibration of chromite within Kilauea Iki Lava Lake, Hawaii.
Contrib Mineral Petrol 107:8-20
Seaman SJ, Dyar MD, Marinkovic N, Dunbar NW (2006) An FTIR study of hydrogen in anorthoclase and
associated melt inclusions. Am Mineral 91:12-20
Shimizu K, Komiya T, Hirose K, Shimizu N, Maruyama S (2001) Cr-spinel, an excellent micro-container for
retaining primitive melts - implications for a hydrous plume origin for komatiites. Earth Planet Sci Lett
189:177-188
Shimizu N (1998) The geochemistry of olivine-hosted melt inclusions in a FAMOUS basalt ALV519-4-1. Phys
Earth Planet Int 107:183-201
Shimizu N, Hart SR (1982) Applications of the ion micro-probe to geochemistry and cosmochemistry. Annu
Rev Earth Planet Sci 10:483-526
Sinton CW, Christie DM, Coombs VL, Nielsen RL, Fisk MR (1993) Near-primary melt inclusions in anorthite
phenocrysts from the Galapagos Platform. Earth Planet Sci Lett 119:527-537
Sisson TW, Bronto S (1998) Evidence for pressure-release melting beneath magmatic arcs from basalt at
Galunggung, Indonesia. Nature 391:883-886
Sisson TW, Layne GD (1993) H2O in basaltic and basaltic andesite glass inclusions from four subduction-
related volcanoes. Earth Planet Sci Lett 117:619-637
Slater L, McKenzie D, Gronvold K, Shimizu N (2001) Melt generation and movement beneath Theistareykir,
NE Iceland. J Petrol 42:321-354
Sobolev AV, Clocchiatti R, Dhamelincourt P (1983a) Variations of the temperature, Melt composition and water-
pressure during olivine crystallization in oceanitic rocks from the Piton Fournaise volcano (Reunion island,
1966 eruption). C R Acad Sci Ii 296:275-281
330 Kent
Sobolev AV (1996) Melt inclusions as a source of primary petrographic information. Petrology 4:209-220
Sobolev AV, Hofmann AW, Nikogosian IK (2000) Recycled oceanic crust observed in ‘ghost plagioclase’ within
the source of Mauna Loa lavas. Nature 404:986-990
Sobolev AV, Chaussidon M (1996) H2O concentrations in primary melts from suprasubduction zones and mid-
ocean ridges: Implications for H2O storage and recycling in the mantle. Earth Planet Sci Lett 137:45-55
Sobolev AV, Danyushevsky LV (1994) Petrology and geochemistry of boninites from the north termination of
the Tonga trench - constraints on the generation conditions of primary high-Ca boninite magmas. J Petrol
35:1183-1211
Sobolev AV, Gurenko AA, Shimizu N (1994) Ultra-depleted melts from Iceland: data from melt inclusion
studies. Mineral Mag 58A:860-861
Sobolev AV, Kamenetsky VS, Kononkova NN (1991) New data on petrology of Siberia Meymechites.
Geokhimiya 1084-1095
Sobolev AV, Migdisov AA, Portnyagin MV (1996) Incompatible element partitioning between clinopyroxene
and basalt liquid revealed by the study of melt inclusions in minerals from Troodos lavas, Cyprus. Petrology
4:307-317
Sobolev AV, Shimizu N (1993) Ultra-depleted primary melt included in an olivine from the Mid-Atlantic ridge.
Nature 363:151-154
Sobolev AV, Tsamerian OP, zakariadze GS, Shcherbovskii AJ (1983b) Compositions and conditions of the
crystallization of the lesser Caucasus Ophiolite Volcanogenic Complex melts according to the data of melt
inclusion study. Doklady Akademii Nauk Sssr 272:464-468
Sobolev VS, Kostyuk VP (1975) Magmatic crystallization based on a study of melt inclusions. Fluid Incl Res
9:182– 235
Sorby HC (1858) On the microscopic structures of crystals, indicating the origin of minerals and rocks. Geol
Soc London q J 14:453-500
Sours-Page R, Nielsen RL, Batiza R (2002) Melt inclusions as indicators of parental magma diversity on the
northern East Pacific Rise. Chem Geol 183:237-261
Sours-Page RL, Johnson KTM, Nielsen RL, Karsten JL (1999) Local and regional variation of MORB parent
magmas: Evidence from melt inclusions from the Endeavour segment of the Juan de Fuca ridge. Contrib
Mineral Petrol 134:342-363
Spandler C, O’Neill HSC, Kamenetsky VS (2007) Survival times of anomalous melt inclusions from element
diffusion in olivine and chromite. Nature 447:303-306
Spiegelman M, Kelemen PB (2003) Extreme chemical variability as a consequence of channelized melt
transport. Geochem Geophys Geosys 4: doi 000184337400001
Stolper E, Newman S (1994) The role of water in the petrogenesis of Mariana Trough magmas. Earth Planet
Sci Lett 121:293-325
Straub SM, Layne GD (2003) The systematics of chlorine, fluorine, and water in Izu arc front volcanic rocks:
Implications for volatile recycling in subduction zones. Geochim Cosmochim Acta 67:4179-4203
Sun W, Bennet VC, Eggins SM, Arculus RJ, Perfit MR (2003) Rhenium systematics in submarine MORB and
back-arc basin glasses: laser ablation ICP-MS results. Chem Geol 196:259-281
Sutton SR, Karner J, Papike J, Delaney JS, Shearer C, Newville M, Eng P, Rivers M, Dyar MD (2005) Vanadium
K edge xANES of synthetic and natural basaltic glasses and application to microscale oxygen barometry.
Geochim Cosmochim Acta 69:2333-2348
Tait S (1992) Selective preservation of melt inclusions in igneous phenocrysts. Am Mineral 77:146-155
Taylor RP, Jackson SE, Longerich HP, Webster JD (1997) In situ trace-element analysis of individual silicate
melt inclusions by laser ablation microprobe inductively coupled plasma-mass spectrometry (LAM-ICP-
MS). Geochim Cosmochim Acta 61:2559-2567
Thomas JB, Bodnar RJ (2002) A technique for mounting and polishing melt inclusions in small (< 1 mm)
crystals. Am Mineral 87:1505-1508
Thomas R, Kamenetsky VS, Davidson P (2006) Laser Raman spectroscopic measurements of water in unexposed
glass inclusions. Am Mineral 91:467-470
Thordarson T, Self S (1996) Sulfur, chlorine and fluorine degassing and atmospheric loading by the Roza
eruption, Columbia River Basalt Group, Washington, USA. J Volcanol Geotherm Res 74:49-73
Thordarson T, Self S, Oskarsson N, Hulsebosch T (1996) Sulfur, chlorine, and fluorine degassing and atmospheric
loading by the 1783-1784 AD Laki (Skaftar fires) eruption in Iceland. Bull Volcanol 58:205-225
Van Orman JA, Grove TL, Shimizu N (2002) Diffusive fractionation of trace elements during production and
transport of melt in Earth’s upper mantle. Earth Planet Sci Lett 198:93-112
Vigouroux N, Wallace PJ, Kent AJR (2008) Volatiles in high-K magmas from the western Trans-Mexican
Volcanic Belt: evidence for fluid-flux melting and extreme enrichment of the mantle wedge by subduction
processes. J Petrol 49:1589-1618
Wade JA, Plank T, Melson WG, Soto GJ, Hauri EH (2006) The volatile content of magmas from Arenal volcano,
Costa Rica. J Volcanol Geotherm Res 157:94-120
Melt Inclusions in Basaltic and Related Volcanic Rocks 331
Walker JA, Roggensack K, Patino LC, Cameron BI, Matias O (2003) The water and trace element contents of
melt inclusions across an active subduction zone. Contrib Mineral Petrol 146:62-77
Wallace P, Carmichael ISE (1992) Sulfur in basaltic magmas. Geochim Cosmochim Acta 56:1863-1874
Wallace PJ (1998) Water and partial melting in mantle plumes: Inferences from the dissolved H2O concentrations
of Hawaiian basaltic magmas. Geophys Res Lett 25:3639-3642
Wallace PJ, Anderson AT (1998) Effects of eruption and lava drainback on the H2O content of basaltic magmas
at Kilauea volcano. Bull Volcanol 59:327-344
Wallace PJ, Carmichael ISE (1994) S-speciation in submarine basaltic glasses as determined by measurements
of S Kα x-ray wavelength shifts. Am Mineral 79:161-167
Weinsteiger AB, Kent AJR, Nielsen RL, Tepley FJ (2007) Are variations in TiO2 contents in anorthite-hosted
MORB melt inclusions controlled by magma compositions or boundary layer phenomena? Eos Trans
88(52): V11B-058.
Workman RK, Hauri E, Hart SR, Wang J, Blusztajn J (2006) Volatile and trace elements in basaltic glasses from
Samoa: Implications for water distribution in the mantle. Earth Planet Sci Lett 241:932-951
Yaxley GM, Kamenetsky VS, Kamenetsky M, Norman MD, Francis D (2004) Origins of compositional
heterogeneity in olivine-hosted melt inclusions from the Baffin Island picrites. Contrib Mineral Petrol
148:426-442
Yurimoto H, Kogiso T, Abe K, Barsczus HG, Utsunomiya A, Maruyama S (2004) Lead isotopic compositions
in olivine-hosted melt inclusions from HIMU basalts and possible link to sulfide components. Phys Earth
Planet Interior 146:231-242
zinner E, Crozaz G (1986) A method for the quantitative measurement of rare-earth elements in the ion
microprobe. Int J Mass Spec 69:17-38
Reviews in Mineralogy & Geochemistry
Vol. 69 pp. 333-361, 2008 9
Copyright © Mineralogical Society of America
Interpreting h2O and CO2 Contents
in Melt Inclusions: Constraints from Solubility
experiments and Modeling
Gordon Moore
Department of Chemistry & Biochemistry
Arizona State University
Tempe, Arizona, 85287-1604, U.S.A.
[email protected]
IntrOduCtIOn
Due to their volatile nature and low solubility in silicate melts at the surface of the Earth,
the direct measurement of the volatile components H2O and CO2 in magmatic systems is
dependent on the presence of glass inclusions trapped in a crystal host prior to eruption. These
inclusions, along with the glassy rinds of submarine pillow lavas, represent one of the few
windows researchers have into the pre-eruptive chemical characteristics of magmatic systems
(e.g., Anderson et al. 2000; Roggensack 2001a; Metrich and Wallace 2008). This information
is critically important, as it is volatile exsolution and expansion that provides much of the
energy for explosive eruptions, plus it informs our broader understanding of geochemical
fluxes in igneous environments (Wallace 2005), and helps us understand specific magmatic
behavior such as pre-eruptive phase equilibria (Moore and Carmichael 1998). The dissolved
volatile concentration in magmas also strongly influences their physical properties such as
density and viscosity (Lange 1994; Ochs and Lange 1999), which in turn affect volcanological
behavior such as eruption style (Sparks et al. 1994; zhang et al. 2007).
In order to use melt inclusion volatile content measurements for petrologic interpretation,
experimental volatile solubility data are critically necessary. Experimental solubility
constraints allow the measured volatile contents in melt inclusions to be used to estimate
intensive properties such as a minimum depth of entrapment of the inclusion (i.e., a calculated
fluid saturation pressure), as well as to indicate the type of degassing process (i.e., open or
closed system) that occurred during the emplacement and eruption of the magma (Fig. 1).
These estimates assume, of course, that the magma is saturated in a fluid containing H2O and/
or CO2, and are not appropriate for a magma that is fluid-undersaturated. With the assumption
of fluid saturation however (see in this volume Blundy and Cashman 2008 and Metrich and
Wallace 2008 for discussion of evidence for mid- to lower-crustal fluid saturation in magmas),
solubility models place mass balance constraints on magmas, lead to estimates of fluid/magma
mass ratios, and to models for their evolution during crystallization and degassing (Wallace
et al. 1995; Papale 2005). Determining all of these characteristics is crucial for assessing
trends in pre-eruptive and eruptive behavior, and is essential information for volcano hazard
monitoring and prediction at any specific volcano.
The importance of volatiles such as H2O in influencing igneous processes was recognized
as early as Bowen (1928). quantification of their properties however did not significantly
begin until the late 1950’s with the efforts of researchers like C.W. Burnham and his pioneering
measurements of the high P-T intensive parameters of H2O (Burnham et al. 1969), including
its solubility in silicate melts (Burnham and Davis 1971, 1974). This laid the groundwork for
1529-6466/08/0069-0009$05.00 DOI: 10.2138/rmg.2008.69.9
334 Moore
3000
500 MPa
C. Negro melt inclusions
2500
400 MPa
2000
CO - ppm
300 MPa
Fig. 1A
1500
2
1000 200 MPa
Figure 1. a. An example of volatile con-
centrations of CO2 and H2O in calc-alkaline
500 100 MPa melt inclusions from a single eruption of
Cerro Negro volcano (data from Roggensack
2001a). Isobaric concentration lines calcu-
0 lated using the VolatileCalc solubility model
0 2 4 6 8
H O - wt%
for basalts (Newman and Lowenstern 2002).
2 B. Magnified view of same Cerro Negro melt
Closed (5% vapor)
inclusion data set showing calculated degas-
Open system sing paths calculated using VolatileCalc. The
400 MPa
bold solid line represents closed system de-
300 MPa
1200 gassing assuming the presence initially of
5% fluid. The dashed line is the calculated
1000 open-system degassing path. Note that while
the general trend of the data is explained by
closed system degassing, the variation in
CO - ppm
800 Fig. 1B
200 MPa the data remains difficult to interpret by any
single degassing process.
600
2
400
100 MPa
200
0
2 3 4 5 6 7
H O - wt%
2
further quantification and the development of new measurement techniques in the 1980’s.
In particular, Fourier transform infrared spectroscopy (FTIR) (Stolper 1982; Stolper et al.
1987) and secondary ion mass spectrometry (SIMS) (Hervig and Williams 1988; Hervig
1992) revolutionized both field and experimental igneous petrology by allowing precise
determination of H2O and CO2 contents in glassy samples at a spatial resolution of tens of
microns. These analytical improvements have also facilitated the development of models for
solubility behavior in melts as a function of pressure, temperature, and melt composition, and
contributed significantly to our understanding of volatile contents in magmas.
This review chapter will focus on the significant amount of new experimental and modeling
work over a broad range of melt and fluid compositions that has been performed since Carroll
and Holloway (1994) published their important RiMG volume entitled Volatiles in Magmas.
Although a significant amount of work on synthetic melts has also been conducted as part of
a successful effort to elucidate the various solubility mechanisms and dependencies at work
in silicate melts (e.g., Holloway et al. 1992a; Kohn and Brooker 1994; Behrens et al. 1996;
Miyagi et al. 1997; Brooker et al. 1999), this data has proved to be difficult to incorporate
into general solubility models for natural melts (Moore et al. 1998; Papale 1999; Papale et al.
2006). Therefore, this chapter will only include the recent work (i.e., since 1994) on pure and
H2O and CO2 Content in Melt Inclusions: Experiments & Modeling 335
mixed (H2O + CO2) solubility in natural melt compositions, and its application to interpreting
pre-eruption volatile contents measured in melt inclusions and submarine glasses.
exPerIMentaL VOLatILe SOLuBILIty
StudIeS In naturaL MeLtS
the “good” solubility experiment
Our ability to extract useful information from observed melt inclusion volatile contents
is certainly dependent on our understanding of the solubilities of those volatiles as a function
of pressure (P), temperature (T), and melt/fluid composition. This understanding comes only
from experimental measurement of the solubility behavior in various melt compositions. Once
obtained, the resulting experimental dataset can be used to develop models that allow interpola-
tion and extrapolation of the data to conditions and compositions that may or may not have been
investigated or be experimentally possible, but that are important nonetheless to understanding
and predicting the character of volcanic systems (e.g., VolatileCalc; Fig. 1). In order to build
the most precise and stable solubility models however, the best possible attainable experimental
data are required. The data must therefore meet several requirements in order to be suitable for
use in creating any empirical model, and these are discussed in the following sections.
Experimental apparatus for volatile solubility studies. The solubility of H2O and/or CO2 in
silicate liquids are experimentally determined by equilibrating a known melt composition under
fluid-saturated conditions at known P and T, and then quenching the liquid to a glass. These
studies can be conducted in a variety of high P-T apparatus that meet the requirements specific
to this type of experiment. These requirements include: 1) a relatively large sample volume
(0.3 to 1 cc) to accommodate the inherently large molar volume of fluids, and the subsequent
volume change associated with dissolution of fluid into the silicate liquid, especially at lower (<
300 MPa) P conditions, 2) a rapid quench from super-liquidus T to below the glass transition T,
particularly for mafic samples and any experimental condition resulting in high dissolved H2O
content (> 4 wt% H2O) in the liquid, and 3) near-hydrostatic P conditions to minimize capsule
failure due to shear, as well as to minimize error in estimating the experimental P.
The high P-T apparatus that meet these requirements include the rapid-quench cold seal
vessel (Ihinger 1991), the rapid-quench pressure vessel (Holloway et al. 1992b), and the
large volume piston cylinder (Baker 2004; Moore et al. 2008a). Each of these techniques has
advantages and disadvantages for the solubility experiment. For example, the cold seal (CS)
and pressure vessel (PV) offer hydrostatic and precise P control with large sample volumes,
unlike the relatively small volumes of the piston cylinder (PC). The problem of their relatively
slow quench rates has been overcome by the development of rapid-quench modifications (e.g.,
Ihinger 1991; Holloway et al. 1992b). The upper P-T ranges for the CS and PV methods are
constrained however (~1150 °C, ~300 MPa; Holloway and Wood 1988) by the material strength
of the vessel and other considerations such as short furnace life at high T, and the occurrence
of large T gradients at higher P. Thus, they are typically used for more silicic (i.e., at lower P
and T) compositions. While being able to achieve the higher T and extremely fast quench rates
(> 150 °C/s) needed for mafic compositions, the PC suffers from a disadvantage in that its
solid pressure media (as opposed to fluid or gas media in CS or PV) requires a minimum force
to overcome internal friction that leads to non-hydrostatic conditions, and thus is normally run
at pressures greater than 500 MPa. The PC also has relatively smaller sample volumes, making
fluid-saturated experiments at lower P difficult. These factors have resulted in an experimental
“gap” in P space for solubility studies, particularly for the more mafic compositions, between
300 MPa (upper region for the CS and PV) and 500 MPa (minimum P for PC). Recent work
with the large volume piston cylinder (Baker 2004; Moore et al. 2008a) has shown that with
a modified sample assembly (Fig. 2), this apparatus can be used for solubility determinations
336 Moore
19 mm
Capsule
NaCl
Solid Pyrex
Powdered Pyrex
Crushable MgO
45 mm
Pyrophyllite disk
Alumina TC sheath
Stainless Steel Base Plug
13mm
Graphite Furnace and Lid
Figure 2. Schematic of piston cylinder solid media assembly used for low pressure (< 500 MPa) experiments,
including H2O-CO2 fluid solubility and volatile-bearing phase equilibria experiments. From Moore et al.
(2008a). Figure 2
at pressures lower than previously thought (down to 300 MPa), and presents the capability to
bridge this experimental P-T region for volatile solubility studies.
Analytical techniques for determining melt and fluid compositions. Once a solubility
experiment has been successfully conducted, it is necessary to determine, at a minimum, the
dissolved volatile content of the glassy run product. If the experiment involves mixed fluid
components (e.g., H2O + CO2), it is critically important to also determine the composition of the
fluid (i.e., the H2O/CO2 ratio) that was in equilibrium with the melt. The analysis of the glassy
solid portion of the run product can be conducted using several methods, including the same
micro-beam techniques used for melt inclusion and submarine glass determinations (e.g., SIMS
and FTIR; see Ihinger et al. 1994; Metrich and Wallace 2008). For those experimentalists with
the luxury of large sample volumes, using more precise bulk techniques such as high T vacuum
manometry and Karl-Fischer titration is preferred due to their first-order nature (i.e., the calibra-
tions for these techniques do not depend on secondary standards; see Ihinger et al. 1994).
To determine the fluid composition in mixed volatile experiments, two techniques are
generally used; a mass balance/gravimetry technique and a low T vacuum manometry method.
The mass balance technique involves weighing the sample capsule, then lowering the T of the
capsule such that the H2O freezes but the CO2 remains a vapor (usually using liquid nitrogen).
The capsule is then punctured, releasing only the CO2 and then re-weighed. The capsule is then
heated above 100 °C to release all the H2O, and then weighed a final time. While quite a simple
technique, it requires a large enough amount of fluid to overcome any weighing errors, and
thus success is also dependent on the precision of the balance used. Unfortunately, the large
volume necessary to equilibrate a melt with such a large amount of fluid can be prohibitive,
and some studies using this technique report errors in the fluid composition between 20 and
100% (e.g., Kadik and Lukanin 1973; Tamic et al. 2001; Botcharnikov et al. 2006). The low
T vacuum manometry method consists of puncturing the sample capsule under vacuum, and
then collecting and separating the resulting fluid in the vacuum line. The amounts of the H2O
H2O and CO2 Content in Melt Inclusions: Experiments & Modeling 337
and CO2 are then measured using manometric techniques precise to ± 10 micromoles of fluid
(Moore et al. 2008a), enabling a far more precise determination of the fluid composition in
equilibrium with the melt, particularly for mixed fluid compositions near XfH2O ~ 0.5. The
error in the determination increases however as the pure end member fluid compositions are
approached (i.e., as the amount of the minor fluid component approaches 10 micromoles).
Thus, a “good” solubility experiment is conducted in an appropriate high P-T apparatus that
allows stable equilibration of the volatiles with the silicate melt. It also employs a fast enough
quench from super-liquidus T to form a glass and prevent any retrograde equilibration of the
volatiles with the silicate portion of the sample. Once a successful sample is retrieved, precise
characterization of both the fluid (if mixed volatiles are involved) and the melt composition is
critical. For researchers who have apparatus with large sample volumes at their disposal, bulk
analytical techniques for the dissolved volatile contents are desirable for their precision and
their first-order nature (Ihinger et al. 1994). Their disadvantage is that they usually destroy a
significant amount of the sample, particularly if duplicate measurements are desired. Non-
destructive micro-analytical techniques for dissolved H2O and CO2 in glasses such as FTIR
and SIMS can be successfully used, but must be used in conjunction with well-characterized
standards of similar composition to the sample being investigated (Hauri et al. 2002; King et
al. 2002; Ihinger et al. 1994; Moore et al. 2008b).
the solubility of pure h2O and CO2 in natural melts
Almost all of the early solubility experimental work was done on pure, single component
H2O or CO2 fluids in mafic to silicic melt compositions (e.g., Goranson 1931; Hamilton et al.
1964; Burnham and Davis 1971). Because H2O was recognized to have a significant impact on
the chemical and physical behavior of silicate melts, and because CO2 was discovered to have
a relatively small solubility (Wyllie and Tuttle 1959), much of the early work focused on H2O
rather than CO2. This focus on H2O was also due to the fact that dissolved volatile contents
were first determined using a mass balance technique that involved puncturing the capsule and
measuring the weight loss due to the evaporation of the undissolved fluid component. This re-
quired a significant amount of the volatile component to be dissolved in order to minimize the
associated weighing errors. Another early approach was to measure the size of the “dimples” in
the capsule walls, thereby estimating the amount of excess fluid present in vesicles during the
experiment. This technique, used mostly in PV work, required not only the presence of a large
amount of fluid, but also required a capsule material ductile enough to easily occupy the fluid
space during quench and the associated decrease in volume (e.g., gold or silver) of the fluid
(Burnham and Jahns 1962). The disadvantage of these capsule materials however, is their low
melting point that limits the upper T of the experiments, and thus the composition of the melt
that can be explored.
Spectroscopic techniques such as FTIR and SIMS have since supplanted these earlier
methods. These newer techniques have resulted in an increase in the precision of the volatile
content measurements, which is critical for volatile components with low solubility (e.g., CO2).
And in the case of FTIR, added information regarding the speciation of the dissolved volatile
in the melt is also obtained (e.g., zhang and xu 2007; Hui et al. 2008). The following section
will briefly discuss the results of the early single component fluid solubility measurements,
and then focus on the more recent determinations and their implications for understanding
natural magmatic systems.
Pure H2O. Despite the analytical constraints, early measurements on pure H2O
solubilities laid the groundwork for our current understanding of volatile behavior in melts.
For H2O, many solubility measurements were made in a broad variety of synthetic and natural
compositions (e.g., Hamilton et al. 1964; see also McMillan 1994 for references). From these
measurements, it became possible to define the P and T dependence of H2O solubility in mafic
to silicic melts, which led to predictions of the speciation of the dissolution reaction that were
338 Moore
later generally confirmed by spectroscopic measurements (Stolper 1982; McMillan 1994).
The development of the FTIR technique in the 1980’s and 1990’s for measuring dissolved
H2O content and speciation in silicate glasses was also accompanied by solubility studies on
natural compositions such as rhyolites (Silver et al. 1990; Ihinger 1991). These data were used
to develop new solubility models based on the behavior of volatile species (e.g., molecular
H2O) observed spectroscopically (Fig. 3). This work has ultimately led to a geothermometer/
speedometer based on the speciation of H2O in silicic glass (Newman et al. 1988; Stolper
1989; zhang et al. 2000; zhang and xu 2007), and has improved our understanding of the
diffusion properties of H components in melts and glasses (zhang 1999; zhang 2007).
4
Molecular H O
2
3.5 OH
3 Figure 3. Speciation of dissolved H2O
OH and H O - wt%
versus total dissolved H2O in rhyolite
2.5 as measured using FTIR spectroscopy.
Note the crossover point at ~2.5 wt%
total H2O where the concentration of
2
2
molecular H2O equals that of the OH
1.5 species. Data from Silver et al. (1990)
equilibrated at 850 °C and quenched
1
at 200 °C/min. See zhang et al. (1997)
and zhang (1999) for discussion of the
dependence of speciation of H2O on
0.5
cooling rate.
0
0 1 2 3 4 5 6
Total H O - wt%
2
Recent work measuring H2O solubility in natural compositions has consisted of
significantly expanding the compositional
Figure 3 range of the melts investigated, as well as detailed
studies at lower P (Table 1). For example, Carroll and Blank (1997) have measured the
solubility of H2O in a peralkaline phonolite, as well as calibrated its infrared absorptivity
coefficients. This study found that the phonolite composition behaves similarly to jadeitic
glass in terms of its H2O solubility. Di Matteo et al. (2004) explored the H2O solubility of a
similar, but more trachytic, composition and found similar behavior as the phonolite of Carroll
and Blank (1997). A very detailed, pure H2O solubility study on calc-alkaline rhyolite at low P
(0.098-25 MPa) was conducted by Liu et al. (2005), yielding very important data for late stage
degassing of hydrous rhyolites. Behrens and Jantos (2001) have also conducted a detailed
study on the effect of compositional variation in rhyolites on the H2O solubility across the
peralkaline-peraluminous join.
The determinations of H2O solubility in mafic compositions are unfortunately much more
limited than those for silicic melts. This is partially due to the fact that H2O-rich mafic melts
are famously difficult to quench to a glass, as well as that most mafic igneous systems generally
contain significant amounts of CO2, making the pure H2O solubility behavior somewhat
less applicable to natural systems. Nevertheless, several workers have tackled the necessary
experimental difficulties and made solubility measurements on mafic compositions. These
include the several pure H2O runs in the low P study of dissolved volatiles in a MORB tholeiite
by Dixon et al. (1995), as well as the study of Pineau et al. (1998) on a basaltic andesite.
Other measurements have been made in the reconnaissance studies of Moore et al. (1995,
1998) in the effort to develop a broad compositional model for H2O solubility. Intermediate
table 1. Recent experimental studies containing data on the solubility of pure H2O and CO2 in natural melts.
reference Melt Composition P - MPa T - °C apparatus1 analytical2 Methods
Pure H2O
Behrens and Jantos 2001 rhyolite 50-500 800 EHPV KF
Botcharnikov et al. 2004* rhyodacite 200 850 CS ED
Botcharnikov et al. 2006 andesite 500 1200 Rq-IHPV FTIR
Botcharnikov et al. 2007* andesite 200 1050, 1200 Rq-IHPV KF, FTIR
Carroll and Blank 1997 phonolite 19-150 850-973 Rq-CS FTIR, Man
Di Matteo et al. 2004 trachyte 20-200 850 CS FTIR, KF
Dixon et al. 1995 MORB3 18-72 1200 Rq-IHPV FTIR, Man
Larsen and Gardner 2004 phonolite 50-200 825-850 CS FTIR
Liu et al. 2005 rhyolite 0.1-25 700-1200 IHPV FTIR
Moore et al. 1995, 1998 various 65-311 900-1130 Rq-IHPV Man
Pineau et al. 1998 basaltic andesite 50-300 1200-1250 IHPV Man
Tamic et al. 2001 rhyolite 200, 500 800, 1100 CS, IHPV FTIR, KF
Webster et al. 1999* basalt, andesite 200 1100-1170 IHPV SIMS
Webster and Rebbert 1998* rhyolite 50-200 700-975 IHPV SIMS
Yamashita 1999 rhyolite 22-100 850-1200 Rq-IHPV FTIR
Pure CO2
Botcharnikov et al. 2006 andesite 200, 500 1200, 1300 Rq-IHPV FTIR
Brooker et al. 2001 melilitite 1200-2700 1300-1600 PC FTIR, EA
Jendrzejewski et al. 1997 MORB3 1200, 1300 25-195 IHPV FTIR, Man
Thibault and Holloway 1994 leucitite 100-2000 1200-1600 IHPV, PC EA, FTIR, SIMS
1
Apparatus abbreviations are: CS – cold seal vessel; EHPV – externally heated pressure vessel; IHPV – internally heated pressure vessel; Rq – rapid quench.
2
H2O and CO2 Content in Melt Inclusions: Experiments & Modeling
Methods include: EA – COH elemental analyzer; ED – electron probe difference; FTIR – Fourier transmission infrared spectroscopy; KF – Karl Fischer titration;
Man – high temperature vacuum manometry; SIMS – secondary ion microprobe spectrometry.
3
Mid-ocean ridge tholeiitic basalt.
*- may contain significant amounts of halogens and/or sulfur.
339
340 Moore
compositions (i.e., andesitic melts) are also relatively lacking in terms of pure H2O solubility
data for similar reasons as the mafic melts (see Table 1 for recent references), but include
measurements by Botcharnikov et al. (2006, 2007), and Moore et al. (1998).
General H2O solubility behavior in natural melts. A plot of the H2O solubility for basalt
and rhyolite as a function of increasing P is shown in Figure 4. It can be seen that at low P
(>70 MPa) and their appropriate T, the pure H2O solubility of the two compositions are very
similar in terms of weight percent of the melt. To some this is often somewhat surprising, as
it is commonly understood that H2O solubility increases with silica content of the melt. It is,
however, purely a function of the inverse T dependence of H2O solubility, and the ~350 °C
difference between a basalt and rhyolite. At higher P however, the greater P dependence of
the H2O solubility in the rhyolite dominates, and the H2O contents of the two compositions
diverge quickly. The addition of certain components, such as alkalies, have also been shown to
greatly increase the H2O solubility (Fig. 5; Carroll and Blank 1997; Behrens and Jantos 2001;
Di Matteo et al. 2004; Larsen and Gardner 2004) in natural melts, and highlight the necessity of
measuring solubility behavior across a broad range of conditions and melt compositions. Another
important characteristic of the H2O component and its solubility that should not be overlooked
is its absolute concentration value. It has been measured in melt inclusions from mafic arc
magmas at amounts greater than 6 weight percent (e.g., Sisson and Layne 1993; Roggensack
2001a; Wallace 2005). Because H2O is such a light molecule, at these concentrations the H2O
component comprises 15-20 mol percent of these magmas (calculated on an oxide basis),
superseded only by SiO2 and Al2O3, and therefore becomes one of the dominant components of
the melt on a molecular basis.
Pure CO2. Investigations into the solubility of pure CO2 in natural silicate melts have
followed a similar track to that of H2O, with the added complication of its low (generally 100-
1000’s of ppm) dissolved concentration making it inherently difficult to precisely measure.
Wyllie and Tuttle (1959) made some of the first experiments on CO2 in geologically relevant
systems (granitic) and recognized that the amount of CO2 dissolved in the melts was small
relative to H2O. This fact, along with the imprecise measurement techniques used at the time
(e.g., gravimetric, beta-track radiography, and electron microprobe deficit; Blank and Brooker
1994) led to much debate in the 1970’s and early 1980’s as to what the solubility of CO2 and
its dependencies in natural melts were. Since then, significant and successful CO2 solubility
experimental studies, particularly of dissolution mechanisms and speciation in synthetic melts
300 Basalt H O (1200°C)
2
Rhyolite H O (850°C)
2
250
Figure 4. Dissolved H2O concentra-
Pressure - MPa
200
tions in pure H2O-saturated basalt
and rhyolite as a function of pressure.
150 The basalt data is at 1200 °C (Ham-
ilton et al. 1964; Dixon et al. 1995)
and the rhyolite is at 850 °C (Silver et
100 al. 1990). Note the similar concentra-
tions of H2O at low pressure (i.e., < 2
50 wt% H2O).
0
1 2 3 4 5 6
H O - wt%
2
H2O and CO2 Content in Melt Inclusions: Experiments & Modeling 341
(e.g., Kohn et al. 1991; Brooker et al. 2001), were done. More recently however, there have been
only a few studies conducted on the solubility of pure CO2 in natural melts. Similar to H2O, the
work that has been done in this area has focused on increasing the range of melt compositions,
as well as the P and T conditions studied (Table 2).
General solubility behavior of pure CO2 in natural melts. In general, but to varying
degrees, the concentration of total dissolved CO2 in natural melts increases with increasing P,
and decreases with increasing T. These dependencies have been systematically measured in
rhyolitic (Fogel and Rutherford 1990) and mafic (Brooker et al. 2001; Jendrzejewski et al. 1997;
Shilobreyeva and Kadik 1990; Pan et al. 1991; Thibault and Holloway 1994) compositions.
Other studies have determined pure or near pure CO2 solubilities in other compositions as well,
but they are often a small part of a larger study involving mixed volatiles (e.g., Dixon et al.
1995; Botcharnikov et al. 2006; King and Holloway 2002). The total solubility and speciation
Figure 5. Summary diagram taken from Behrens and Jantos (2001) showing the effect of compositional
changes in rhyolites on the H2O solubility at various pressures. XH2O is the mole fraction of H2O dissolved
in the melt. See text for definition of the compositional parameter (MCLNK-A)/O.
342
table 2. Recent experimental studies containing data on the solubility of mixed H2O-CO2 fluid in natural melts.
reference Melt Composition P - MPa T - °C apparatus1 analytical1 Methods
Behrens et al. 2004a dacite 100-500 1250 Rq-IHPV KF, SIMS, FTIR
Botcharnikov et al. 2006 andesite 500 1200 Rq-IHPV FTIR
Botcharnikov et al. 2007* andesite 200 1050, 1200 Rq-IHPV KF, FTIR
Dixon et al. 1995 MORB3 18-72 1200 Rq-IHPV FTIR, Man
Moore
Jakobsson 1997 icelandite 1000 1400 PC FTIR, EA
King and Holloway 2002 andesite 1000 1300 PC FTIR, SIMS
Moore et al. 2008b basalt 300-700 1200 PC FTIR, Man, SIMS
Tamic et al. 2001 rhyolite 200, 500 800, 1100 CS, IHPV FTIR, KF
1 - See Table 1 for abbreviations.
*- may contain significant amounts of halogens and/or sulfur.
H2O and CO2 Content in Melt Inclusions: Experiments & Modeling 343
of dissolved CO2 are both a strong function of melt composition, with silicic melts containing
only molecular CO2 (Behrens et al. 2004b), and mafic melts having only the carbonate complex
(Blank and Brooker 1994). The intermediate compositions are more complicated in that they
can contain both the molecular and carbonate species of dissolved CO2 (Behrens et al. 2004a;
King et al. 2002). Despite it being well established that CO2 solubility is strongly dependent
on composition, little experimental work has been done to constrain this dependence in natural
melts. Some experimental work has explored the compositional dependence of pure CO2 on
tholeiitic and alkalic basalts (Pan et al. 1991; Thibault and Holloway 1994), while Brooker et
al. (2001) investigated silica-undersaturated alkalic melts at high P. Unfortunately, no study
exists that explores the effect of composition for mafic or silicic compositions in a systematic
way similar to the detailed study for H2O solubility dependence in rhyolites of Behrens and
Jantos (2001).
the solubility of mixed fluids (h2O + CO2) in natural melts
Perhaps the most important experimental solubility studies for understanding and
interpreting melt inclusion data are those that involve mixed volatiles (H2O + CO2). This
is true because melt inclusions typically contain measurable amounts of both components.
And, because of the interplay between the mixing properties of a binary fluid and intensive
parameters such as P, T, and melt composition, experimental constraints on mixed solubility
behavior are critical to properly modeling the system. Unfortunately, a survey of the literature
shows that there is a relative dearth of these experiments, and not nearly enough to precisely
understand melt inclusion volatile contents given the number of variables, particularly for
mafic and intermediate compositions (Table 2). The situation is beginning to be recognized,
however, as more and more melt inclusion measurements are made, and the need grows to
precisely interpret them, particularly for subduction-related volcanoes (e.g., Roggensack
2001a,b; Cervantes and Wallace 2003a; Wade et al. 2006; Johnson et al. 2008; Metrich and
Wallace 2008). Several important experimental studies have also recently highlighted the
current inadequacies of the existing data and the resulting effect on the available mixed volatile
solubility models (Moore et al. 2005, 2008c).
The earliest mixed volatile experiments on natural melts were conducted by Wyllie and
Tuttle (1959), and pioneering work was also performed in the 1970’s and early 1980’s by several
groups (Kadik et al. 1972; Brey and Green 1973, 1975; Mysen et al. 1975; Jakobsson and
Holloway 1986). All of these researchers were stymied however by their inability to precisely
measure the dissolved components (particularly CO2) in the melts. As was mentioned before,
mixed volatile studies also require precise fluid composition determination, adding another
complication to these already difficult experiments. The modern analytical techniques of SIMS
and FTIR have been as important to advances in mixed volatile studies as they have been for the
pure end-member solubility studies. But, because of the complication of the presence of both
CO2 and H2O species dissolved in the melts, calibration of these techniques has proven more
difficult than for the pure component studies. This is especially true for the use of FTIR on
intermediate compositions, where CO2 is present in the melt as both a carbonate and molecular
species (Mandeville et al. 2002; King et al. 2002). Recent work using the SIMS technique has
also shown the possibility of a matrix effect of H2O that suppresses the ionization of carbon
in mafic glasses (Moore et al. 2008b), complicating the SIMS determination of CO2 in mafic
glasses containing greater than 3 wt% dissolved H2O. All of these complications arise from
the fact that there are not enough well-characterized glass standards available that contain
significant amounts of both H2O and CO2, and point out that much work remains to be done in
this area.
General behavior of mixed volatile solubility in natural melts. Despite the complicated
nature of mixed solubility experiments, there are several excellent studies that have elucidated
the solubility behavior of mixed volatiles in natural melts. In particular is the work of Tamic et
344 Moore
al. (2001) who conducted a detailed investigation of the H2O-CO2 solubility in a rhyolite up to
500 MPa (Table 2). Also included are the study of Behrens et al. (2004b) from 100-500 MPa
on a dacitic liquid, and the low-P study of a mid-ocean ridge basalt of Dixon et al. (1995). The
results of Dixon et al. (1995) at high T (1200 °C) and low P (<100 MPa) are suitably modeled
assuming that the solubility of both CO2 (as a carbonate ion) and molecular H2O behave in a
Henrian manner, and that the two volatile components mix linearly in the fluid. This linear
behavior breaks down however for both components at higher P (>150 MPa). This observed
behavior in basalt is similar to that seen in the high-P study on rhyolite of Tamic et al. (2001)
and the dacite of Behrens et al. (2004a), where linear solubility dependence for H2O is seen at
200 MPa. At higher P (500 MPa) however, both of these studies report a greater dependence on
the fluid composition (i.e., on H2O fugacity). CO2 solubility is also non-linear at all conditions
in these studies, in contrast to the linear behavior reported by Blank et al. (1993) at lower P.
Behrens et al. (2004a) also report a marked increase in CO2 solubility (as well as increasing
carbonate/molecular species ratio) with increasing H2O content, indicating that H2O stabilizes
CO2 in dacitic melts. All of these observations point to the conclusion that the influence of
H2O on CO2 solubility in natural melts is complicated, particularly at higher P (>100 MPa).
This conclusion is significant, particularly for mafic and intermediate magmas that are thought
to have relatively deep origins, and to contain significant amounts of H2O and CO2 prior to
eruption.
Other recent studies (Moore et al. 2005, 2008c; Roggensack and Moore 2008) have
investigated the influence of compositional variation on H2O-CO2 solubility in calcic and calc-
alkaline basalts, with the goal of directly determining solubility behavior in melt compositions
similar to those found in arc-related melt inclusions. One of the striking results from these
studies is the strong positive dependence of the CO2 solubility of these hydrous arc basalts on
their calcium content (Fig. 6). Unfortunately, many existing solubility models cannot account
for this dependence, leading to a significant overestimation of the minimum saturation P
and an incorrect estimate of the equilibrium fluid composition for calc-alkaline basaltic melt
inclusions. These discrepancies are due again to the lack of experimental data available for a
broad range of melt compositions on which to base solubility models, particularly for calcic
to calc-alkaline compositions. This topic will be discussed in more detail in the following
sections on solubility models and their application to melt inclusion data.
6000
Hydrous arc basalts
5000
Dissolved CO - ppm
4000 Figure 6. The influence of CaO
content on the CO2 solubility in cal-
2
cic to calc-alkaline basalt composi-
tions. Saturated H2O-CO2 experi-
3000
mental data at 400 MPa and 1200
°C from Roggensack and Moore
(2008) containing 3-5 wt% H2O.
2000
1000
6 8 10 12 14 16 18 20
CaO - wt%
H2O and CO2 Content in Melt Inclusions: Experiments & Modeling 345
VOLatILe SOLuBILIty MOdeLS FOr naturaL SILICate MeLtS
Many diverse mathematical approaches have been used to model the solubility of H2O and/
or CO2 in silicate melts (see Table 3). Most models are based on thermodynamic relationships,
and range from simple descriptions of P and T dependencies to those that make use of certain
solubility characteristics (e.g., linear Henrian behavior). Some are based on melt composition
dependencies that are inferred to be related to melt structure (e.g., the NBO/T polymerization
parameter of Mysen 1987). Others are simply ad hoc mathematical expressions (e.g., Taylor
series expansions) that were chosen only because they fit the data well. Each model type has
its particular strengths and weaknesses, and care must be taken with all of them so that the user
is not misled into believing a prediction that could be in serious error, or over-interpreting the
physical significance of a statistical fit parameter.
table 3. Volatile solubility models.
Melt
reference P - MPa T - °C Model type1
Composition
H2O solubility
Behrens and Jantos 2001 rhyolitic 0.1-200; 500 800 comp
Carroll and Blank 1997 phonolitic 0.1-191 850 reg sol
Di Matteo et al. 2004 trachytic 20-200 850 comp
Dixon et al. 1995 basaltic 0.1-98 1200 reg sol
Liu et al. 2005 rhyolitic 0.1-500 700-1200 emp
Moore et al. 1998b various 0.1-300 700-1200 comp
Papale 1997 various 0.1-1000 >730 reg sol
Yamashita 1999 rhyolitic 0.1-100 700-1200 reg sol
zhang 1999 rhyolitic 0.1-800 500-1350
CO2 solubility
Dixon et al. 1995 basaltic 0.1-98 1200 reg sol
Dixon et al. 1997 alkali basalt 0.1-500 1200 comp
Papale 1997 various 0.1-1000 >730 reg sol
Mixed volatile solubility
Behrens et al. 2004a2 dacitic-rhyolitic 100-500 850-1250 reg sol
Liu et al. 2005 rhyolitic 0.1-500 700-1200 emp
Newman and Lowenstern 2002 basalt, rhyolite various
Papale 1999 various <1000 >730 comp reg sol
Papale et al. 2006 various comp reg sol
3
Tamic et al. 2001 rhyolitic 75-500 800-1100 emp
1
- Model types include: comp – compositional, accounts for compositional variation, but may include an ad hoc fit equation
form; comp reg sol – compositional regular solution model; emp – empirical, generally an ad hoc form of fit equation; reg
sol – regular solution model, no compositional dependence; various – uses many different models for its calculations.
2
- CO2 solubility for H2O-CO2 bearing silicic compositions.
3
– H2O solubility for H2O–CO2 bearing rhyolite.
346 Moore
To simplify the discussion of the recent modeling work, each model is grouped in terms
of the volatile component that is modeled (Table 3). Although grouped in this way, any given
model may or may not be applicable to calculating solubility for a pure or mixed volatile fluid,
depending on how the model is constructed. For example, the general compositional model for
H2O by Moore et al. (1998) is intended for calculating the solubility of pure H2O. But because
it uses H2O fugacity as a parameter, it can also be used successfully to calculate H2O solubility
for mixed volatile cases (i.e., by assuming an linear mixing model for CO2 and H2O in the fluid,
where the fugacity of H2O in the fluid mixture is equal to the fugacity of pure H2O at the P and
T of interest times the mole fraction of H2O in the mixture), as long as the final H2O fugacity
value does not significantly exceed that used to calibrate the model. Other models, such as the
rhyolite/dacite-CO2 solubility model of Behrens et al. (2004a), are based on mixed hydrous-
CO2 bearing datasets and thus are only correctly used in the mixed volatile case.
h2O solubility models
Silicic models (>65 wt% SiO2). A majority of H2O solubility studies from the literature
prior to 1994 were conducted on silicic compositions (McMillan 1994). This is also true for
the recent work reviewed here (Table 1). There are also many other studies that fall under
the silicic classification that are not listed in Table 1 because they are synthetic or haploid
compositions (e.g., Holtz et al. 1995, 2000; Romano et al. 1996; Dingwell et al. 1997; Schmidt
et al. 1999; Behrens et al. 2001). All of this experimental attention has resulted in a good
understanding of the mechanisms of H2O dissolution in rhyolites, as well as many models for
H2O solubility in silicic melts. And, for rhyolitic melts specifically, it has resulted in two types
of models that are currently in general use. The first type is based on the regular solution model
developed first for the system albite-diopside-silica-H2O by Silver and Stolper (1985), and
then later applied to rhyolite melts (Silver et al. 1990). The second is a model that accounts for
changes in H2O solubility due to compositional variation in rhyolites by using a metal cation/
total oxygen ratio (Behrens and Jantos 2001).
The regular solution model for H2O solubility in rhyolites of Silver and Stolper (1985)
and Silver et al. (1990) is described and discussed by Holloway and Blank (1994). This type
of model has the characteristic that its fit parameters can be equated with thermodynamic
parameters (e.g., partial molar volume) for the particular component being dissolved in the
melt of interest. This type of interpretation has been shown by Ochs and Lange (1999) to
be incorrect however, as the regressed values are subject to significant error stemming from
the low P nature of the solubility data, and uncertainty in the activity model for H2O in the
melt. By using direct density determinations on hydrous melts, Ochs and Lange (1999) have
convincingly shown that there is only one value for the partial molar volume of H2O (22.89
±0.55 cc/mol) that is independent of melt composition, as well as H2O concentration.
Nevertheless, the formalism of Silver et al. (1990) has been applied to other solubility
data sets, and represent precise solubility models for various melt compositions. For example,
Carroll and Blank (1997) use the model successfully to develop an expression for the
solubility of H2O in a silicic phonolite. The H2O solubility model of Yamashita (1999) for
rhyolite incorporates the measured partial molar volume value for H2O of Ochs and Lange
(1999) and compositional independence into his regular solution model (i.e., it is not a variable
parameter) with great success at P up to 100 MPa (Fig. 7). zhang (1999) also addresses the
H2O partial molar volume inconsistency directly, and shows the difficulties in deriving partial
molar parameters using solubility data, as well as developing an excellent H2O solubility
model for rhyolites (Fig. 7) that can convincingly be extrapolated to high P and T. Despite all
of the complications over the interpretation of the fit parameters, the regular solution model
formalism of Silver and Stolper (1985) provides an H2O solubility model that has been well
used to recover the experimental solubility values and to precisely estimate H2O solubility
contents for particular rhyolitic compositions.
H2O and CO2 Content in Melt Inclusions: Experiments & Modeling 347
Figure 7. H2O concentration versus pressure for peralkaline (NSL), meta-aluminous (qV), and peraluminous
(EDF) rhyolitic compositions, including various predicted solubility curves. Note the stability of the zhang
(1999) model up to 500 MPa, demonstrating the utility of accounting for H2O speciation. Figure taken from
Behrens 7
Figure and Jantos (2001).
The second type of model for estimating the H2O solubility of rhyolites is that developed
by Behrens and Jantos (2001) that accounts for solubility changes due to variation in the bulk
melt composition of the rhyolite. This model is based on a series of solubility experiments
conducted over a range of P on a broad range of rhyolites (Table 1), including peralkaline,
peraluminous, and subaluminous silicic melts. This study convincingly shows that H2O
solubility has a minimum for subaluminous compositions, with an increase in solubility in
both the peralkaline and peraluminous direction (Fig. 5). This variation in H2O solubility
can be modeled using a compositional parameter (MCLNK-A/O = 100×(2Mg + 2Ca + Li
+ Na + K – Al)/O), although the degree of scatter in their data led them to conclude that
the model is only semi-quantitative. This is attributed by the authors to the fact that the
MCLNK-A/O parameter accounts for neither the concentration nor oxidation state of iron in
their experiments, implying a significant role of iron in determining H2O solubility in silicic
melts.
Mafic models (< 52 wt% SiO2). As there are significantly less data for H2O solubility in
mafic melts, there are subsequently far fewer models that have been developed to calculate
it. In fact, the only model specifically developed for estimating H2O solubility in basalts
is that of Dixon et al. (1995). This model uses the regular solution formalism discussed
above for rhyolitic compositions, but derives different values for the fit parameters. Other
models available for mafic melt calculations are those that fall into the general compositional
category discussed below.
General compositional models for H2O solubility. The effort of Burnham and workers
on developing a general H2O solubility model followed their measurements in silicic
compositions (Burnham and Davis 1971; Burnham and Davis 1974), as well as their unique
P-V-T measurements of supercritical H2O fluid at high P and T (Burnham et al. 1969). This
work culminated in a thermodynamic-based model that assumed a linear relationship between
the activity of the H2O component and the square of the mole fraction dissolved in the melt
348 Moore
(Burnham 1981), and equilibrium constants were developed to calculate H2O solubility for a
broad range of melt compositions (Burnham 1994) based on feldspar-like melt components. A
corollary of the main assumption of their model is that OH is the dominant dissolving species,
and that molecular H2O is not present in significant amounts. This assumption has been shown
to be true at lower P, but at higher P and H2O fugacities molecular H2O becomes the dominant
dissolving species present (Fig. 3; Stolper 1982; Silver et al. 1990). Nevertheless the Burnham
model, as it has come to be known, has been used productively to predict pure H2O solubility
in a broad range of natural melt compositions, and certainly provided a useful framework for
understanding H2O solubility.
The empirical model of Moore et al. (1998) represents a successful attempt to create a
simple model that defines the P, T, and compositional dependence of H2O solubility for a very
broad range of melts. Using a series of forty-one H2O solubility determinations on fourteen
melts chosen to independently maximize compositional differences, they fit the data to a ther-
modynamically-based equation of the form:
a P
2 ln X Hmelt
2O
= + ∑ bi Xi + c ln fHfluid
2O
+d (1)
T i T
fluid
where X Hmelt
2O
= the mole fraction of H2O dissolved in the melt; fH2 O = the fugacity of H2O
in the fluid; T = temperature in Kelvin; P = pressure in bars; and Xi = the anhydrous mole
fractions of the oxide components. The fit parameters for Equation (1) are: a, bi (i = Na2O,
FeOt, Al2O3), c, and d, and the fugacity of H2O is calculated using the modified Redlich
Kwong equation of state (Holloway and Blank 1994). This model has proven successful in
estimating H2O solubility across a spectrum of natural melt compositions, and is available
as a spreadsheet that facilitates calculation and use (Moore et al. 1998). Behrens and Jantos
(2001) correctly point out however that this model does not capture the compositional
dependence they observed in rhyolites, and suggest that another compositional parameter
similar to theirs could be incorporated as more data become available. Di Matteo et al.
(2004) go a step further and re-regress the Moore et al. model, finding a single compositional
parameter that reproduces their data in trachytic compositions. They find however that the
resulting expression is useful only for trachytic melts. Also, as is described by zhang (1999)
and the authors themselves, the Moore et al. (1998) model is strictly empirical in nature, and
thus should not be extrapolated beyond the maximum H2O fugacity for which it is calibrated
(~3000 bars).
In perhaps the most ambitious effort to date to develop a compositionally dependent
model for H2O, Papale (1997) combined the regular solution mixing model for the anhydrous
melt components of Ghiorso et al. (1983) with an exhaustive database of pure H2O and CO2
solubility measurements (640 for H2O, 263 for CO2) to derive the regular solution mixing
parameters for hydrous and CO2-bearing natural melts. One significant result from this
work is the observation that the CO2 solubility dataset suffers from significant inconsistency
in the measurements, and this leads to difficulty in creating a model that recovers all the
experimental observations. The H2O dataset however appears to be internally consistent,
although there are inconsistencies with the synthetic and mono-mineralic compositions. The
resulting expression for H2O solubility has ten interaction parameters between water and the
melt components, and reproduces the dataset to within ±10% of most of the data. The Papale
(1997) model also predicted the observation of a compositionally independent partial molar
volume for the H2O component that was subsequently confirmed by the work of Ochs and
Lange (1999). Unfortunately, because of the heavy computational nature of this model, it has
not been used to its fullest potential by other workers. It has also since been improved and
modified to include mixed fluid behavior by Papale (1999) and Papale et al. (2006), which
will be discussed in more detail below.
H2O and CO2 Content in Melt Inclusions: Experiments & Modeling 349
CO2 solubility models
As the amount of recent work determining the pure CO2 solubility in natural melts is
relatively small (Table 1), so are the number of studies that attempt to model its behavior
(Table 3). Much of the earlier work, particularly on silicic compositions, has been reviewed
by Blank and Brooker (1994) and Holloway and Blank (1994). Significant for the behavior of
dissolved CO2 in basalts though, are the more recent solubility studies of Dixon et al. (1995)
and Dixon (1997). Using their low-P determinations of CO2 solubility, Dixon et al. (1995)
developed a regular solution model for CO2 in a mid-ocean ridge tholeiite. In order to account
for the significant CO2 solubility changes due to compositional variation in basalts, Dixon
(1997) developed a compositional parameter, PI, (where PI= −6.50[Si+4 + Al+3] + 20.17[Ca+2
+ 0.8K+] + 0.7Na+ + 0.4Mg+2 + 0.4Fe+2) based on cation proportions and the solubility
determinations on an alkalic leucitite (Thibault and Holloway 1994) and basanite (Dixon and
Pan 1995). This allowed the estimation of CO2 solubility from tholeiitic to strongly alkalic
basalts as a function of composition. The compositional dependence was further simplified
with the observation that the PI compositional parameter varied linearly as a function of
the SiO2 content of the basalts, thus facilitating a simple estimate based on P, T, and SiO2
concentration (Dixon 1997). It is this parameterization that is incorporated into the mixed
volatile solubility basalt model of VolatileCalc (Newman and Lowenstern 2002; see further
discussion below). While this model works well for tholeiitic to alkalic basalts, it is not
able to precisely estimate the CO2 solubility of other compositions such as calcic and calc-
alkaline basalts (Moore et al. 2008b). One cause of this is that the PI compositional parameter
becomes negative for certain basalt compositions (typically, high Al2O3 and high CaO; see
Fig. 8), leading to nonsensical negative CO2 solubility concentrations. Unfortunately, these
types of basaltic compositions commonly occur in subduction-related volcanic arcs, where
there are many ongoing melt inclusion studies that require precise estimation of solubility
properties for CO2.
2.5
2 Leuc
Basalt compositions of High - Ca
Dixon (1997)
basalt
(tholeiite to leucitite)
1.5
Kilauea
PI factor
1 Basanite
Arc Basalt
0.5 MORB
(Nicaragua)
0
Arc Basalt
(SE Guatemala)
-0.5
High Alumina Basalt (V.Fuego)
-1
0.4 0.6 0.8 1 1.2 1.4
CaO/Al O - wt%
2 3
Figure 8. The compositional parameter (PI) of Dixon (1997) versus CaO/Al2O3 for arc basaltic compositions
(triangles). Arrow shows the range of PI for the tholeiitic and alkali-rich basalts used to calibrate the model.
Note the negative values of PI for Volcan Fuego and the Guatemalan basalt.
Figure 8
350 Moore
Papale (1997) also applied the approach of combining the anhydrous mixing properties
of melts with solubility data to modeling the behavior pure CO2 solubility in a wide range of
natural melts. As was noted in the previous H2O solubility modeling section, much of the pure
CO2 solubility data in the literature used by Papale (1997) at the time was inconsistent due to
the lack of precise measurement techniques. Therefore, the dataset was difficult to model with
this approach, and required the elimination of many outliers (Papale 1997). Perhaps because of
the scatter in the data used, the model required twice as many mixing parameters (twenty, as
opposed to ten for H2O) between CO2 and the melt components to fully describe the solubility
variation.
Mixed (h2O + CO2) solubility models
Modeling the dissolution of an H2O-CO2 fluid in natural silicate melts is a complicated
endeavor that requires a significant amount of precise experimental data. As was discussed
earlier, the dataset must not only include the precise measurement of the H2O and CO2 content
of the fluid-saturated glass, but also the composition of the H2O-CO2 fluid that was in equilib-
rium with the melt during the experiment.
There are several approaches that have been taken in developing the current mixed
volatile solubility models. One is to use fugacity-based, pure fluid solubility models and make
assumptions about the mixing behavior of the H2O and CO2 components in the fluid and melt.
Generally, these are simplifying assumptions such as linear mixing in the fluid, which leads
to dissolved solubility values that are calculated using a fugacity value equal to the pure fluid
fugacity multiplied by the mole fraction of the component of interest in the fluid (fH2O = XH2O
fH2O). This is the approach taken by Newman and Lowenstern (2002) in their mixed volatile
solubility algorithm for basalts and rhyolites known as VolatileCalc. This mode of calculation
assumes however that there is no influence of one component on the solubility of the other,
which we have seen earlier, is valid at low P (< 100 MPa) but is certainly incorrect for silicic
melts at higher P (Tamic et al. 2001; Behrens et al. 2004a). The lack of experimental data at
higher P for mafic compositions leaves these assumptions essentially untested for basaltic melts,
although they appear reasonable for low-P calculations (Dixon et al. 1995). Another approach is
that of Behrens et al. (2004a), Tamic et al. (2001), and Liu et al. (2005), all of whom based their
individual solubility expressions for H2O and CO2 in silicic melts on a dataset that included
mixed volatile measurements. A significantly more comprehensive approach has been taken
by Papale (1999) and Papale et al. (2006), who combined the available pure and mixed volatile
solubility data with their previous work using the regular solution model for the anhydrous melt
components of Ghiorso et al. (1983) to obtain the fluid/melt interaction parameters for both the
volatile components.
Models specific to silicic compositions. For the solubility of CO2 in hydrous silicic liquids,
the model of Behrens et al. (2004a) explores the behavior of dacitic and rhyolitic liquids.
Interestingly, they note the presence of detectable amounts of the dissolved carbonate species
present in dacite, the stabilization of which they attribute to the high concentration of H2O in
their experiments. This observation is supported by a similar one made in andesites and rhyolites
at higher P (King and Holloway 2002; Moore et al. 2006), and significantly complicates the
determination of total CO2 in these melts using FTIR by increasing the number of peaks that
must be identified and calibrated. As a result, Behrens et al. (2004a) resorted to the use of
SIMS for their CO2 measurements and report a significant increase of CO2 solubility with H2O
content. They also observe a significant change in CO2 solubility behavior of a high P non-
linear to low P linear dependence on fluid composition. The resulting regular solution model
from this work is therefore only useful in hydrous silicic magmas (although they present a
model for anhydrous melts as well), from 100-500 MPa, and over a range mole fraction of H2O
in the fluid of 0.1-0.9. The study of Liu et al. (2005) explores the apparently non-ideal behavior
that occurs in H2O-CO2 bearing rhyolites at high P in the extreme ends of the H2O-CO2 fluid
H2O and CO2 Content in Melt Inclusions: Experiments & Modeling 351
composition space, and develop an ad hoc empirical expression that can be used for pure H2O
and CO2 fluids, and reproduces the non-linear behavior in the mixed fluid compositional regions
not well constrained by the model of Behrens et al. (2004a). Although excellent at recovering
the experimental values from the dataset (± 20%), care must be taken when extrapolating the
CO2 solubility expression of Liu et al. (2005) to pressures or conditions that are outside of the
experimental conditions of the dataset because of its exponential nature.
The H2O solubility models of Tamic et al. (2001) and Liu et al. (2005) for mixed volatile-
bearing rhyolites take a very similar approach in that they both use series type equations to
model their H2O-CO2 solubility data. Tamic et al. (2001) chose a truncated Taylor series expan-
sion equation type that relates the H2O solubility to P, T, and fluid composition. This model
has eighteen fit parameters that reproduce a broad range of solubility data from the literature to
within ±5.5% relative. The expression for H2O solubility of Liu et al. (2005) is similarly depen-
dent on the fluid composition, but has fewer fit parameters (six), and reports a 2σ error of 15%.
As mentioned above, caution must be used with these types of empirical models, particularly
when extrapolating them beyond the bound of their data set (i.e., P > 500 MPa).
The rhyolite model in VolatileCalc (Newman and Lowenstern 2002) uses the regular
solution parameters for H2O from Silver et al. (1990), and those for CO2 are from Blank et
al. (1993). The fugacities of the two fluid components are assumed to mix linearly, and their
pure, end-member fugacities are calculated using the modified Redlich-Kwong equation of
state (Holloway 1977; Flowers 1979). The solubilities of each component in the melt are also
assumed to be independent of each other, and to behave linearly (Henrian) with respect to
changes in fluid composition. As the rhyolite model is based essentially on pure and mixed fluid
solubility data for meta-aluminous rhyolite at low P (<150 MPa), it is capable of generating
precise estimates for the solubility parameters, and essentially matches other models that only
calculate the pure component solubilities. There are, however, several caveats that must be
taken into account. First, there is no compositional dependence built in to the calculation for
rhyolite that can account for the H2O solubility behavior observed by Behrens and Jantos (2001)
across the peralkaline-peraluminous compositional spectrum of silicic melts. At the same time,
the compositional effect on CO2 in silicic magmas is also known to be important (Stolper et al.
1987), but remains unconstrained. Thus, the bulk composition of any silicic melt inclusion suite
must be shown to be close to meta-aluminous and relatively constant in order to properly apply
VolatileCalc to it. Secondly, we have also seen in the work of Tamic et al. (2001) and Behrens et
al. (2004a) that CO2 solubility appears to be dependent on H2O content at P equal to and greater
than 200 MPa, which should cause users of VolatileCalc to consider the output at these higher P
carefully, particularly for values at the ends of the fluid composition spectrum.
Mixed solubility models for basalts. The mixed volatile solubility calculation package
of Newman and Lowenstern (2002) can also be used to estimate minimum saturation P and
degassing trends for basaltic melt inclusion studies. VolatileCalc incorporates the regular
solution parameters for H2O solubility determined by Dixon et al. (1995), and uses the
CO2 solubility parameters from Dixon et al. (1995) and Pan et al. (1991). It also uses the
compositional parameterization of Dixon (1997) discussed in the earlier section on CO2
solubility. This results in a flexible model that allows users to calculate solubility parameters
from nephelinite to tholeiite by inputting only the SiO2 content of the melt. As this compositional
dependence is constructed using solubility data from tholeiitic and alkali-rich basalts, it is
capable of precisely estimating the solubility parameters within this compositional range.
Unfortunately, the model is not capable of correctly estimating the mixed solubility behavior
of calcic or calc-alkaline compositions, as these basalts are outside of the compositional
parameterization of Dixon (1997). In particular, the effect of increasing calcium content has
been shown to have an especially strong positive effect on the CO2 solubility in basalts (Fig.
6; Moore et al. 2008b). Because of the strong P dependence of CO2, applying VolatileCalc to
these compositions results in an overestimation of the minimum saturation P by as much as
352 Moore
35-50% relative, and typically outputs results that are beyond the range of P calibration of the
model itself. Similarly, the fluid compositions estimated by VolatileCalc are also significantly
incorrect compared to experimental values (see discussion below).
General compositional models for mixed volatile solubility. Currently, the only model
available that accounts for compositional variation in H2O-CO2 solubility from basalts to
rhyolites is that of Papale (1999), which was updated by Papale et al. (2006). This model uses
the entire available dataset for C-O-H-silicate liquid solubility to calibrate its multi-component
regular solution model. There are 865 measurements of H2O, 173 for CO2, and 84 measurements
for mixed volatiles, giving an indication of the breadth of the database used. One improvement
of this work over the earlier studies is that the number of recent CO2 measurements is now
large enough that the older (pre-1980’s), inconsistent determinations could be thrown out.
This has resulted in a more stable and systematic dataset to model, although they report that
the larger number of H2O determinations has resulted in higher confidence for the calculations
of that component. Nevertheless, the ambitious model of Papale et al. (2006) has reached the
point where its authors can begin contemplating effects such as how the oxidation state of iron
in Fe-rich melts might affect the solubility values.
the compositional variation in melt inclusions and its implications for interpreting
volatile contents
We have seen from the previous discussions that melt composition plays an important role
in determining the solubility of H2O and CO2 in magmas. Therefore, it must be accounted for
in any solubility model being used to interpret melt inclusion volatile contents. Unfortunately,
accounting for compositional effects on solubility is rarely performed in most melt inclusion
studies, due to the fact that there are no models readily available that have this capability (The
model of Papale et al. 2006 is an obvious exception, but has only recently become broadly
available due to the efforts of Dr. Mark Ghiorso of OFM Research who has created a website
calculator that allows several types of calculations using the model, as well as to Dr. Papale’s
generosity in making the code available.). While the basalt model of VolatileCalc can account
for compositional variation in alkali-rich basalts, it is not able to precisely calculate solubility
behavior for subduction-related basalts. The rhyolite model, as we have seen, also does not
have the ability to account for solubility changes due to compositional variation.
This section will examine the implications of compositional effects for interpreting melt
inclusion volatile contents using both VolatileCalc and the Papale et al. (2006) models. In
both cases, an attempt is made to test the model output for rhyolitic and basaltic compositions
with experimental data not used in their respective databases. This is a difficult task, as the
number of mixed volatile solubility measurements since the publication of Papale et al.
(2006) is limited. As a result, the data of Tamic et al. (2001) is used for rhyolites, even though
it is contained within the Papale et al. (2006) dataset. Plus, selected data on calcic and calc-
alkaline basalts from Moore et al. (2008) are used to benchmark the basalt models. We also
ask the question whether the scatter that is difficult to explain using reasonable degassing
histories in many melt inclusion suites (e.g., Fig. 1), could be due to compositional effects
on the melt solubility.
Compositional variation in silicic magmas. The effect of compositional variation on the
two solubility models (VolatileCalc; Newman and Lowenstern 2002 and Papale et al. 2006)
will be assessed here using the mixed solubility data of Tamic et al. (2001) and Behrens et
al. (2004a), representing a meta-aluminous rhyolite and dacite respectively. The test was
approached as if the experimental data were melt inclusion measurements, and the models
used accordingly to calculate minimum saturation P and fluid compositions. The resulting
values for the fluid composition and P are plotted against the measured experimental values in
Figure 9. It can be seen that both VolatileCalc and Papale do an excellent job of recovering the
H2O and CO2 Content in Melt Inclusions: Experiments & Modeling 353
0.8
H2O
0.6
Model X
0.4
0.2
0
0 0.2 0.4 0.6 0.8 1
Measured X
H2O
600
500
Model Pressure - MPa
400
300
200
Rhyolite - Papale
Rhyolite - VolCal
100 Dacite - Papale
Dacite - VolCal
0
0 100 200 300 400 500 600
Measured Pressure - MPa
Figure 9. Model versus measured values for pressure and fluid compositions (XH2O) in a rhyolite and dacite
using VolatileCalc and Papale et al. (2006). Rhyolite data taken from Tamic et al. (2001) and the dacite
values are from Behrens et al. (2004a). Neither dataset is used to calibrate VolatileCalc, while the Papale et
al. model does contain the Tamic et al. data.
200 and 500 MPa rhyolitic data of Tamic et al. (2001). VolatileCalc reproduces the 200 MPa
data better by a few percent, while Papale is slightly better at 500 MPa. The major distinction
between the two models occurs when the dacite data of Behrens et al. (2004a) is considered.
The Papale model performs as well with the dacite data as it does for the rhyolite, while
VolatileCalc overestimates the saturation P by a factor of two, and underestimates the H2O
content of the fluid by similar amounts. Of course, Newman and Lowenstern (2002) did not
design the rhyolite model in VolatileCalc to be applied to a dacite composition (66 wt% SiO2),
and the estimates for the appropriate rhyolite are quite good. Although somewhat extreme,
this exercise highlights the potential dangers inherent in using a model based on a single
composition to calculate solubilities for a range of melt compositions.
Compositional variation in arc basaltic melt inclusions. The compositional variation
found in melt inclusions from basalts and basaltic andesites in arc settings is represented in
Figure 10 (for further discussion of compositional variation in basaltic melt inclusions, see
354 Moore
Kent 2008 in this volume). Even within
High Mg basalts (Mexico) single eruptive suites (e.g., Cerro Negro,
Xitle volcano (Mexico)
Sulawesi Guatemala) significant total variation is
Volcan Fuego (Guatemala) found in both the calcium and total alkali
Mt. Etna (Italy)
Cerro Negro (Nicaragua) content, both of which influence the total
CO2 solubility in mafic melts. And, these
15
values can vary as much as 50% for any
given silica content, which is an impor-
tant observation given that the composi-
tional parameterization of Dixon (1997)
CaO - wt%
10 for basalts in VolatileCalc assumes a
linear dependence on silica content. The
significance of this is that any composi-
5
tional variation at a single value for silica
is not accounted for (except for the tholei-
ite to alkali-rich basalt trend) in the Dixon
(1997) model. Thus, before using these
0
40 45 50 55 60
models to calculate solubility parameters
SiO - wt% for melt inclusions found in arcs, they
2
should be benchmarked against experi-
6
mental solubility data conducted on com-
positions similar to the melt inclusions.
5 Unfortunately, there are very few studies
on arc compositions, and as a result, se-
4 lect data from Moore et al. (2008c) and
Na O - wt%
Botcharnikov et al. (2007) will be used
3 here. As the Botcharnikov et al. (2007)
data is andesitic, it is not appropriate to
2
2 use VolatileCalc for it, and only the Pa-
pale et al. (2006) model will be used.
1 Again approaching this test as if the
experimental data were melt inclusion
0
40 45 50 55 60
measurements, we calculated minimum
SiO - wt% saturation P and fluid compositions for
2
the mafic compositions of Moore et al.
7 (2008) using the basalt module of Volati-
leCalc. The results are plotted versus the
6 experimental values in Figure 11. Be-
cause the experiments were conducted
5
at 400, 500 and 600 MPa, some of the
dataset is beyond the stated capability of
K O - wt%
4
VolatileCalc of 500 MPa. Considering
3
2
2
Figure 10. Harker diagrams for melt inclusion
compositions taken from arc-related basalts. Note
the broad range of compositions within single suites
1
of inclusions. Data from: high-Mg basalts and xitle
(Cervantes and Wallace 2003a,b); Sulawesi (Elburg
0 et al. 2007); Volcan Fuego (Roggensack 2001b);
40 45 50 55 60
SiO - wt%
Mt. Etna (Spilliaert et al. 2006); Cerro Negro
2 (Roggensack 2001a).
H2O and CO2 Content in Melt Inclusions: Experiments & Modeling 355
only the 400 and 500 MPa data however, we see that except for one point the predicted satura-
tion P are overestimated by a minimum of 25%, extending up to 50%. Similar deviations from
the experimental values are seen in the fluid compositions as well.
The results of the same test for the Papale model, adding the andesite composition of
Botcharnikov et al. (2007), are shown in Figure 12. The Papale model appears to recover the
low-P andesite data and the high P calc-alkaline basalt values quite well. The other high-P
estimates suffer from significant scatter (average 22% residual) however. For the most part the
scatter in the P values are not systematic, except for that of the high calcium basalt which are
systematically overestimated (and may be outside the compositional range of the model). The
fluid composition estimates for the high calcium basalt experiments are also anomalously and
significantly low, while those for the andesite appear systematically high. Taking into account
the error in the measured experimental parameters for mixed solubility data (particularly
that for fluid composition measurements), and given the broad compositional range that this
model covers, these results are certainly acceptable and represent a significant and important
improvement in our ability to model the behavior of C-O-H fluids in volcanic systems where
magma compositions may vary significantly.
Implications of compositional effects on observed melt inclusion variation. Melt
inclusion volatile contents are often presented on a CO2 versus H2O plot similar to Figure 1.
While this is a simple way to represent the volatile content variation found in a given suite
of melt inclusions, what cannot be shown on this plot is the bulk compositional variation
of the melt inclusions themselves. Because of the observed variation in melt inclusion bulk
compositions (Fig. 10; Kent 2008), and the strong effect of melt composition on solubility,
which has been discussed above, modeled solubility curves such as CO2-H2O isobars, isopleths
(variation at constant fluid composition), and degassing trends for a single composition on a
CO2-H2O diagram are meaningless unless it can be shown that the compositional variation of
the melt inclusion suite is small, and similar to that used in the model.
Melt inclusion volatile content data may be better examined if the compositional effects
on the volatile solubility are accounted for. By plotting the calculated minimum saturation P
using a compositionally dependent model such as Papale et al. (2006) against the measured
values for CO2 and H2O in the melt inclusions (Fig. 13) it becomes possible to see the “P
trajectory” of any given suite of inclusions, and to locate depth regions important to the fluid
and melt evolution of the magma. This is difficult to accurately envision using the CO2-H2O
plot, as the saturation P of each inclusion is dependent on its individual composition. With
the melt inclusion data of Roggensack (2001a) for Cerro Negro volcano recast in this way,
it becomes possible to identify important pressures where a majority of the sampled melt
inclusions were trapped (assuming the system was saturated), and a more isobaric character
(between 150-250 MPa) to the fluid-melt evolutionary history becomes apparent. Future work
in this area will be to define the complicated behavior of degassing trends (open vs. closed
system, etc) using compositionally dependent models such as Papale et al. (2006).
aCKnOWLedGMentS
This work was supported by N.S.F. grants EAR-0409863 and EAR-0610005. The
thoughtful reviews of Rebecca Lange, Keith Putirka, Paul Wallace, and an anonymous reviewer
greatly improved the manuscript. I would also like to thank Kurt Roggensack for many
animated discussions about volatiles, melt inclusions, experiments, and what it all means. And
much credit goes to John R. Holloway for introducing me to volatile solubility experiments so
many years ago, and to Ian S.E. Carmichael who taught me that models can only be as good
as the data they are based on.
1 1
356
0.8 0.8
0.6 Figure 11 (to the left). Model 0.6
(fluid)
H2O
versus measured values for fluid
H2O
composition and pressure in H2O-
0.4 CO2 saturated arc basalts using 0.4
Model X
VolatileCalc. Dashed lines show
Model X
500 MPa limit of VolatileCalc. Note
0.2 the systematic overestimation of
0.2
pressure at 400 and 500 MPa by as
much as 50%. Data from Moore et
al. (2006) and Moore et al. (2008c).
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Measured X (fluid) Measured X
H2O H2O
800
Moore
800
Calcic basalt Figure 12 (to the right). Model Calcic basalt
Calc-alkaline basalt
700 versus measured values for fluid Calc-alkaline basalt
Arc basaltic andesite 700
composition and pressure in H2O- Arc basaltic andesite
CO2 saturated arc basalts using Andesite
600 600
Papale et al. (2006). Scatter in
data falls randomly around 1:1
500 line for both fluid composition and 500
pressure, excepting the systematic
400 underestimation of the fluid 400
composition for the calcic basalt
composition. Basalt data the same
300 300
Model Pressure - MPa
Model Pressure - MPa
as Figure 11, with the andesite from
Botcharnikov et al. (2007).
200 200
100 100
100 200 300 400 500 600 700 800 100 200 300 400 500 600 700 800
Measured Pressure - MPa Measured Pressure - MPa
Figure 12
H2O and CO2 Content in Melt Inclusions: Experiments & Modeling 357
300
Min. Saturation Pressure (Papale) - MPa 250
200
150
100
50
0
0.65 0.7 0.75 0.8 0.85 0.9 0.95 1
fluid
Calculated X (Papale)
H2O
300
Min. Saturation Pressure (Papale) - MPa
Figure 13. Calculated minimum saturation
pressure vs. calculated fluid composition,
250
measured CO2, and H2O content for melt
inclusions from Cerro Negro (Roggensack
200 2001a). Saturation pressures calculated
using Papale et al. (2006). By recasting
the melt inclusion bulk compositions and
150
volatile contents into saturation pressures,
it becomes possible to more precisely
100 identify pressure regions critically
important to the fluid/melt evolution of the
magma. For this case, note the essentially
50
isobaric evolution of both the fluid and
melt volatile composition between 250 and
0 150 MPa. Compare to the oft-used CO2 vs.
0 200 400 600 800 1000 1200 H2O plot (e.g., Fig. 1A and 1B) where such
Measured CO - wt% behavior is difficult to discern.
2
300
Min. Saturation Pressure (Papale) - MPa
Cerro Negro MI
250
200
150
100
50
0
2.5 3 3.5 4 4.5 5 5.5 6 6.5
Measured H O - wt%
2
Figure 13
358 Moore
reFerenCeS
Anderson AT, Davis AM, Lu F (2000) Evolution of Bishop Tuff rhyolitic magma based on melt and magnetite
inclusions, and zoned phenocrysts. J Petrol 41:449-473
Baker DR (2004) Piston-cylinder calibration at 400 to 500 MPa: A comparison of using water solubility in
albite melt and NaCl melting. Am Mineral 89:1553-1556
Behrens H, Jantos N (2001) The effect of anhydrous composition on water solubility in granitic melts. Am
Mineral 86:14-20
Behrens H, Meyer M, Holtz F, Benne D, Nowak M (2001) The effect of alkali ionic radius, temperature, and
pressure on the solubility of water in MAlSi3O8 melts (M = Li, Na, K, Rb). Chem Geol 174:275-289
Behrens H, Ohlhorst S, Holtz F, Champenois M (2004a) CO2 solubility in dacitic melts equilibrated with H2O-
CO2 fluids: Implications for modeling the solubility of CO2 in silicic melts. Geochim Cosmochim Acta
68:4687-4703
Behrens H, Romano C, Nowak M, Holtz F (1996) Near-infrared spectroscopic determination of water species in
glasses of the system MalSi3O8 (M=Li, Na, K): an interlaboratory study. Chem Geol 128:41-63
Behrens H, Tamic N, Holtz F (2004b) Determination of the molar absorption coefficient for the infrared
absorption band of CO2 in rhyolitic glasses. Am Mineral 89:301-306
Blank JG, Brooker RA (1994) Experimental studies of carbon dioxide in silicate melts: solubility, speciation,
and stable carbon isotope behavior. Rev Mineral 30:157-186
Blank JG, Stolper EM, Carroll MR (1993) Solubilities of carbon dioxide and water in rhyolitic melt at 850 °C
and 750 bars. Earth Planet Sci Lett 119:27-36
Blundy J, Cashman K (2008) Petrologic reconstruction of magmatic system variables and processes. Rev
Mineral Geochem 69:179-239
Botcharnikov RE, Behrens H, Holtz F (2006) Solubility and speciation of C-O-H fluids in andesitic melt at
T=1100-1300 °C and P=200 and 500 MPa. Chem Geol 229:125-143
Botcharnikov RE, Holtz F, Behrens H (2007) the effect of CO2 on the solubility of H2O-Cl fluids in andesitic
melt. Eur J Mineral 19:671-680
Bowen NL (1928) The evolution of igneous rocks. Princeton Univ Press, New Jersey
Brey G, Green D (1975) The role of CO2 in the genesis of olivine melilitite. Contrib Mineral Petrol 49:93-103
Brey GP, Green DH (1973) CO2 solubility and solubility mechanisms in silicate melts at high pressures. Contrib
Mineral Petrol 49:93-103
Brooker RA, Kohn SC, Holloway JR, McMillan PF (2001) Structural controls on the solubility of CO2 in silicate
melts Part I: bulk solubility data. Chem Geol 174:225-239
Brooker RA, Kohn SC, Holloway JR, McMillan PF, Carroll MR (1999) Solubility, speciation, and dissolution
mechanisms of CO2 in melts along the NaAlO2-SiO2 join. Geochim Cosmochim Acta 63:3549-3565
Burnham (1994) Development of the Burnham model for prediction of H2O solubility in magmas. Rev Mineral
30:122-129
Burnham CW (1981) The nature of multicomponent aluminosilicate melts. In: Chemistry and Geochemistry of
solutions at high temperatures and pressures. Phys Chem Earth 13&14:191-227
Burnham CW, Davis NF (1971) The role of H2O in silicate melts: I. P-V-T relations in the system NaAlSi3O8-
H2O to 10 kilobars, 700-1100 °C. Am J Sci 274:902-940
Burnham CW, Davis NF (1974) The role of H2O in silicate melts: II. Thermodynamic and phase relations in the
system NaAlSi3O8-H2O to 10 kilobars and 1000 °C. Am J Sci 274:902-940
Burnham CW, Holloway JR, Davis NF (1969) Thermodynamic properties of water to 1000 °C and 10000 bars.
Geol Soc Am Spec Paper 132:1-96
Burnham CW, Jahns RH (1962) A method for determining the solubility of water in silicate melts. Am J Sci
260:721-745
Carroll MR, Blank JG (1997) The solubility of H2O in phonolitic melts. Am Mineral 82:549-556
Carroll MR, Holloway JR (eds) (1994) Volatiles in Magmas. Reviews in Mineralogy. Vol 30, Min Soc America,
Fredericksburg, Virginia
Cervantes P, Wallace P (2003b) Magma degassing and basaltic eruption styles: a case study of ~2000 year BP
xitle volcano in central Mexico. J Volcanol Geotherm Res 120:249-270
Cervantes P, Wallace PJ (2003a) Role of H2O in subduction-zone magmatism: New insights from melt inclusions
in high-Mg basalts from central Mexico. Geology 31:235-238
Di Matteo V, Carroll MR, Behrens H, Vetere F, Brooker RA (2004) Water solubility in trachytic melts. Chem
Geol 213:187-196
Dingwell DB, Holtz F, Behrens H (1997) The solubility of H2O in peralkaline and peraluminous granitic melts.
Am Mineral 82:434-437
Dixon JE (1997) Degassing of alkali basalts. Am Mineral 82:368-378
Dixon JE, Pan V (1995) Determination of the molar absorptivity of dissolved carbonate in basanitic glass. Am
Mineral 80:1339-1342
H2O and CO2 Content in Melt Inclusions: Experiments & Modeling 359
Dixon JE, Stolper EM, Holloway JR (1995) An experimental study of water and carbon dioxide solubilities in
mid-ocean ridge basaltic liquids. Part I: Calibration and solubility models. J Petrol 36:1607-1631
Elburg M, Kamenetsky V, Nikogosian I, Foden J, Sobolev A (2006) Coexisting high- and low-calcium melts
identified by mineral and melt inclusion studies of a subduction-influenced syn-collisional magma from
South Sulawesi, Indonesia. J Petrol 47:2433-2462
Flowers GC (1979) Correction of Holloway’s (1977) adaptation of the modified Redlich-Kwong equation
of state for calculation of the fugacities of molecular species in supercritical fluids of geologic interest.
Contrib Mineral Petrol 69:315-318
Fogel RA, Rutherford MJ (1990) The solubility of carbon dioxide in rhyolitic melts: a quantitative FTIR study.
Am Mineral 75:1311-1326
Ghiorso MS, Carmichael ISE, Rivers ML, Sack RO (1983) The Gibbs free energy of mixing of natural silicate
liquids; an expanded regular solution approximation for the calculation of magmatic intensive variables.
Contrib Mineral Petrol 84:107-145
Goranson RW (1931) The solubility of water in granitic magmas. Am J Sci 22:481-502
Hamilton DL, Burnham CW, Osborn EF (1964) Solubility of water and effects of oxygen fugacity and water
content on crystallization in mafic magmas. J Petrol 5:21-39
Hauri E, Wang J, Dixon JE, King PL, Mandeville C, Newman S (2002) SIMS analysis of volatiles in silicate
glasses 1. Calibration, matrix effects and comparisons with FTIR. Chem Geol 183:99-114
Hervig RL (1992) Ion probe microanalyses for volatile elements in melt inclusions. EOS Trans Am Geophys
Union 73:367
Hervig RL, Williams P (1988) SIMS microanalyses of minerals and glasses for H and D. In: Secondary Ion
Mass Spectrometry, SIMS VI. Benninghoven A, Huber AM, Werner HW (eds) J. Wiley & Sons, New
York, p 961-964
Holloway JR (1977) Fugacity and activity of molecular species in supercritical fluids. In: Thermodynamics in
Geology. Fraser D (ed) Reidel, Boston Mass, p 161-181
Holloway JR, Behrens H, Dingwell DB, Taylor RP (1992a) Water solubility in aluminosilicate melts of
haplogranitic composition at 2 kbar. Chem Geol 96:289-302.
Holloway JR, Blank J (1994) Application of experimental results to C-O-H species in natural melts. Rev Mineral
30:187-230
Holloway JR, Dixon JE, Pawley AR (1992b) An internally heated, rapid-quench, high-pressure vessel. Am
Mineral 77:643-646
Holloway JR, Wood BJ (1988) Simulating the Earth: Experimental Geochemistry. Unwin Hyman, Winchester
Mass
Holtz F, Behrens H, Dingwell DB, Johannes W (1995) H2O solubility in haplogranitic melts: Compositional,
pressure, and temperature dependence. Am Mineral 80:94-108
Holtz F, Roux J, Behrens H, Pichavant M (2000) Water solubility in silica and quartzofeldspathis melts. Am
Mineral 85:682-686
Hui H, zhang Y, xu z, Behrens H (2008) Pressure dependence of the speciation of dissolved water in rhyolitic
melts. Geochim Cosmochim Acta 72:3229-3240
Ihinger PD (1991) An experimental study of the intersection of water with granitic melt. PhD thesis, California
Institute of Technology
Ihinger PD, Hervig RL, McMillan PF (1994) Analytical methods for volatiles in glasses. Rev Mineral 30:67-
121
Jakobsson S (1997) Solubility of H2O and CO2 in an icelandite at 1400 °C and 10 kilobars. Contrib Mineral
Petrol 127:129-135
Jakobsson S, Holloway JR (1986) Crystal-liquid experiments in the presence of a C-O-H fluid buffered by
graphite + iron + wustite: Experimental method and near-liquidus relations in basanite. J Volcanol
Geotherm Res 59:265-291
Jendrzejewski N, Trull TW, Pineau F, Javoy M (1997) Carbon solubility in Mid-Ocean Ridge basaltic melt at
low pressures (250-1950 bar). Chem Geol 138:81-92
Johnson ER, Wallace P, Cashman KV, Delgado Granados H, Kent, AJR (2008) Magmatic volatile contents
and degassing-induced crystallization at Volcan Jorullo, Mexico: Implications for melt evolution and the
plumbing systems of monogenetic volcanoes. Earth Planet Sci Lett 269:477-486
Kadik AA, Lukanin OA (1973) The solubility-dependent behavior of water and carbon dioxide in magmatic
processes. Geochem Int 10:115-129
Kadik AA, Lukanin OA, Lebedev YB, Kolovushkina EY (1972) Solubility of H2O and CO2 in granite and basalt
at high pressures. Geochem Int 9:1041-1050
Kent AJR (2008) Melt inclusions in basaltic and related volcanic rocks. Rev Mineral Geochem 69:273-331
King PL, Holloway JR (2002) CO2 solubility and speciation in intermediate (andesitic) melts: The role of H2O
and composition. Geochim Cosmochim Acta 66:1627-1640
360 Moore
King PL, Vennemann TW, Holloway JR, Hervig RL, Lowenstern JB, Forneris JF (2002) Analytical techniques
for volatiles: A case study using intermediate (andesitic) glasses. Am Mineral 87:1077-1089
Kohn SC, Brooker RA (1994) The effect of water on the solubility and speciation of CO2 in aluminosilicate
glasses along the join SiO2-NaAlO2. Mineral Mag 58:489-490.
Kohn SC, Brooker RA, Dupree R (1991) 13C MAS NMR: A method for studying CO2 speciation in glasses.
Geochim Cosmochim Acta 55:3879-3884
Lange RA (1994) The effect of H2O, CO2, and F on the density and viscosity of silicate melts. Rev Mineral
30:331-370
Larsen J, Gardner J (2004) Experimental study of water degassing from phonolite melts: implications for volatile
oversaturation during magmatic ascent. J Volcanol Geotherm Res 134:109-124
Liu Y, zhang Y, Behrens H (2005) Solubility of H2O in rhyolitic melts at low pressures and a new empirical
model for mixed H2O-CO2 solubility in rhyolitic melts. J Volcanol Geotherm Res 143:219-225
Mandeville CW, Webster JD, Rutherford MJ, Taylor BE, Timbal A, Faure, K (2002) Determination of molar
absorptivities for infrared absorption bands of H2O in andesitic glasses. Am Mineral 87:813-821
McMillan PF (1994) Water solubility and speciation models. Rev Mineral 30:131-156
Métrich N, Wallace PJ (2008) Volatile abundances in basaltic magmas and their degassing paths tracked by melt
inclusions. Rev Mineral Geochem 69:363-402
Miyagi I, Yurimoto H, Takahashi E (1997) Water solubility in albite-orthoclase join and JR-1 rhyolite melts at
1000 °C and 500 to 2000 bars, determined by micro-analysis with SIMS. Geochem J 31:57-61.
Moore G, Carmichael ISE (1998) The hydrous phase equilibria (to 3 kbar) of an andesite and basaltic andesite
from western Mexico: constraints on water content and conditions of phenocryst growth. Contrib Mineral
Petrol 130:304-319
Moore G, Roggensack K, Hervig RL, Vennemann T (2008b) Effect of variable H2O contents on calibrations of
carbon measurements in basaltic glass using SIMS and FTIR. Geochem Geophys Geosyst (in prep)
Moore G, Roggensack K, Holloway J (2006) Dissolved carbonate species in mixed-volatile rhyolitic melts:
carbon speciation correlates with dissolved H2O content. EOS Trans AGU Fall Meet 87:52
Moore G, Roggensack K, Klonowski S (2008a) A low-pressure – high-temperature technique for the piston-
cylinder. Am Mineral 93:48-52
Moore G, Roggensack K, Vennemann T (2005) Compositional variation in arc lavas and their associated melt
inclusions: Implications for magmatic volatile contents and degassing behavior. EOS Trans AGU Fall
Meet 86:52
Moore G, Roggensack K, Vennemann T, Hervig RL (2008c), Mixed (H2O + CO2) volatile solubility in calcic to
calc-alkaline basalts from 400 to 700 MPa. Contrib Mineral Petrol (in prep)
Moore G, Vennemann T, Carmichael ISE (1995) Solubility of water in magmas to 2 kbar. Geology 23:1099-
1102
Moore G, Vennemann T, Carmichael ISE (1998) An empirical model for the solubility of H2O in magmas to 3
kilobars. Am Mineral 83:36-42
Mysen B, Arculus RJ, Eggler DH (1975) Solubility of carbon dioxide in natural nephelenite, tholeiite and
andesite melts to 30 kbar pressure. Contrib Mineral Petrol 53:227-239
Mysen BO (1987) Magmatic silicate melts: Relations between bulk composition, structure, and propertites. In:
Magmatic Processes: Physicochemical Principles. Mysen BO (ed) Geochemical Society, University Park,
Pennsylvania, 375-399
Newman S, Epstein S, Stolper E (1988) Water, carbon dioxide, and hydrogen isotopes in glasses from the
ca. 1340 A.D. eruption of the Mono craters, California: constraints on degassing phenomena and initial
volatile contents. J Volcanol Geotherm Res 35:75-96
Newman S, Lowenstern JB (2002) VolatileCalc: a silicate melt-H2O-CO2 solution model written in Visual Basic
for excel. Comp Geosci 28:597-604
Ochs FA, Lange RA (1999) The density of hydrous magmatic liquids. Science 283:1314-1317
Pan V, Holloway JR, Hervig RL (1991) The pressure and temperature dependence of carbon dioxide solubility
in tholeiitic basalt melts. Geochim Cosmochim Acta 55:1587-1595
Papale P (1997) Modeling of the solubility of a one-component H2O or CO2 fluid in silicate liquids. Contrib
Mineral Petrol 126:237-251
Papale P (1999) Modeling of the solubility of a two-component H2O + CO2 fluid in silicate liquids. Am Mineral
84:477-492
Papale P (2005) Determination of total H2O and CO2 budgets in evolving magmas from melt inclusion data. J
Geophys Res 110:B03208
Papale P, Moretti R, Barbato D (2006) The compositional dependence of the saturation surface of H2O + CO2
fluids in silicate melts. Chem Geol 229:78-95
Pineau F, Shilobreeva S, Kadik A, Javoy M (1998) Water solubility and D/H fractionation in the system basaltic
andesite-H2O at 1250 °C and between 0.5 and 3 kbars. Chem Geol 147:173-184
H2O and CO2 Content in Melt Inclusions: Experiments & Modeling 361
Roggensack K (2001a) Sizing up crystals and their melt inclusions: a new approach to crystallization studies.
Earth Planet Sci Lett 187:221-237
Roggensack K (2001b) Unraveling the 1974 eruption of Fuego volcano (Guatemala) with small crystals and
their young melt inclusions. Geology 29:911-914
Roggensack K, Moore G (2008) Explosive arc basalts: H2O-CO2 solubility experiments and melt inclusion
evidence for a possible magmatic gas origin of the Tiscapa maar, Nicaragua. Geology (in prep)
Romano C, Dingwell DB, Behrens H, Dolfi D (1996) Compositional dependence of H2O solubility along the
joins NaAlSi3O8-KAlSi3O8, NaAlSi3O8-LiAlSi3O8, and KAlSi3O8-LiAlSi3O8. Am Mineral 81:452-461
Schmidt BC, Holtz F, Pichavant M (1999) Water solubility in haplogranitic melts coexisting with H2O-H2 fluids
Contrib Mineral Petrol 136:213-224
Shilobreyeva SN, Kadik AA (1990) Solubility of CO2 in magmatic melts at high temperatures and pressures.
Geochem Int 27:31-41
Silver L, Stolper E (1985) A thermodynamic model for hydrous silicate melts. J Geol 93:161-178
Silver LA, Ihinger PD, Stolper EM (1990) The influence of bulk composition on the speciation of water in
silicate glasses. Contrib Mineral Petrol 104:142-162
Sisson TW, Layne GD (1993) H2O in basalt and basaltic andesite glass inclusions from four subduction-related
volcanoes. Earth Planet Sci Lett 117:619-635
Sparks RSJ, Barclay J, Jaupart C, Mader HM, Phillips JC (1994) Physical aspects of magmatic degassing I.
Experimental and theoretical constraints on vesiculation. Rev Mineral 30:413-446
Spilliaert N, Allard P, Métrich N, Sobolev A (2006) Melt inclusion record of the conditions of ascent, degassing,
and extrusion of volatile-rich alkali basalt during the powerful 2002 flank eruption of Mount Etna (Italy).
J Geophys Res 111:B04203
Stolper E (1982) The speciation of water in silicate melts. Geochim Cosmochim Acta 46:2609-2620
Stolper E (1989) Temperature dependence of the speciation of water in rhyolitic melts and glasses. Am Mineral
74:1247-1257
Stolper E, Fine G, Johnson T, Newman S (1987) Solubility of carbon dioxide in albitic melt. Am Mineral
72:1071-1085
Tamic N, Behrens H, Holtz F (2001) The solubility of H2O and CO2 in rhyolitic melts in equilibrium with a
mixed CO2-H2O fluid. Chem Geol 174:333-347
Thibault Y, Holloway JR (1994) solubility of CO2 in a Ca-rich leucitite: effects of pressure, temperature and
oxygen fugacity. Contrib Mineral Petrol 116:216-224
Wade JA, Plank T, Melson WG, Soto GJ, Hauri E (2006) The volatile content of magmas from Arenal volcano,
Costa Rica. J Volcanol Geotherm Res 157:94-120
Wallace P (2005) Volatiles in subduction zon magmas: concentrations and fluxes based on melt inclusion and
volcanic gas data. J Volcanol Geotherm Res 140:217-240
Wallace PJ, Anderson AT, Davis AM (1995) quantification of pre-eruptive exsolved gas contents in silicic
magmas. Nature 377:612-616.
Wyllie PJ, Tuttle OF (1959) Effect of carbon dioxide on the melting of granite and feldspars. Am J Sci 257:548-
655
Yamashita S (1999) Experimental study of the effect of temperature on water solubility in natural rhyolite melt
to 100 MPa. J Petrol 40:1497-1507
zhang Y (1999) H2O in rhyolitic glasses and melts: measurement, speciation, solubility, and diffusion. Rev
Geophys 37:493-516
zhang Y (2007) Silicate melt properties and volcanic eruptions. Rev Geophys 45:RG4004
zhang Y, Jenkins J, xu z (1997) Kinetics of the reaction H2O + O = 2OH in rhyolitic glasses upon cooling:
geospeedometry and comparison with glass transition. Geochim Cosmochim Acta 61:2167-2173
zhang Y, xu z (2007) A long-duration experiment on hydrous species geospeedometer and hydrous melt
viscosity. Geochim Cosmochim Acta 71:5226-5232
zhang Y, xu z, Behrens H (2000) Hydrous species geospeedometer in rhyolite: Improved calibration and
application. Geochim Cosmochim Acta 64:3347-3355
Reviews in Mineralogy & Geochemistry
Vol. 69 pp. 363-402, 2008 10
Copyright © Mineralogical Society of America
Volatile abundances in Basaltic Magmas and their
degassing Paths tracked by Melt Inclusions
nicole Métrich
Laboratoire Pierre Sue
CNRS-CEA, CE-Saclay
Gif sur Yvette, 91191, France
[email protected]
Paul J. Wallace
Department of Geological Sciences
University of Oregon
Eugene, Oregon, 97403-1272, U.S.A.
[email protected]
IntrOduCtIOn
The abundances of CO2, H2O, S and halogens dissolved in basaltic magmas are strongly
variable because their solubilities and ability to be fractionated in the vapor phase depend
on several parameters such as pressure, temperature, melt composition and redox state.
Experimental and analytical studies show that CO2 is much less soluble in silicate melts
compared to H2O (e.g., Javoy and Pineau 1991; Dixon et al. 1995). As much as 90% of the
initial CO2 dissolved in basaltic melts may be already degassed at crustal depths, whereas H2O
remains dissolved because of its higher solubility such that H2O contents of basaltic magmas
at crustal depths may reach a few percents. Most subduction-related basaltic magmas are rich
in H2O (up to 6-8 wt%; Sisson and Grove 1993; Roggensack et al. 1997; Newman et al.
2000; Pichavant et al. 2002; Grove et al. 2005) compared to mid-ocean ridge basalts (<1 wt%;
Sobolev and Chaussidon 1996; Fischer and Marty 2005; Wallace 2005).
During magma movement towards the surface, exsolution of major volatile constituents
(CO2, H2O) causes gas bubble nucleation, growth, and possible coalescence that exert a strong
control on the dynamics of magma ascent and eruption (Anderson 1975; Sparks 1978; Tait et
al. 1989). Gas bubbles have the ability to move faster than magma (Sparks 1978), particularly
in low viscosity basaltic magmas. Bubble accumulation, coalescence and foam collapse give
rise to differential transfer of gas slugs and periodic gas bursting (Strombolian activity; Jaupart
and Vergniolle 1988, 1989) or periodic lava fountains (Vergniolle and Jaupart 1990; Philips
and Wood 2001) depending on magma physical properties and ascent rate. It is also thought
that strombolian and lava fountain activities critically depend on the magma rise speed, the
transition between the two styles being controlled by the bubble diameter for the same magma
ascent rate (Wilson and Head 1981; Parfitt and Wilson 1995; Parfitt 2004). A model of gas
segregation derived from laboratory experiments (<10 vol% gas fraction) was recently applied
to Stromboli suggesting that the transition between episodic explosive eruption of gas-poor,
relatively dense magma and more explosive, gas-rich, low viscosity magma would take place
for critical values of the magma discharge rate of 0.1-1 m3 s−1 (Menand and Philips 2007).
Whether the magma rises with its gas bubbles entrained (closed-system degassing) or the
gas bubbles are able to segregate from the melt (open-system degassing) further affects gas
1529-6466/08/0069-0010$05.00 DOI: 10.2138/rmg.2008.69.10
364 Métrich & Wallace
compositions and dissolved volatile contents of ascending melts. Accordingly, the abundances
of H2O, S, and halogens in the melt and the composition of the exsolving gas phase will change
in response to the pressure-dependent solubility and diffusivity of each volatile component.
Water is commonly the most abundant volatile constituent and has a direct influence on
magma properties and rheology. It drastically affects the silicate melt density because of its
relatively large partial molar volume (17-25 cm3/mol) when dissolved (between 750 and 1250
°C, 0-1 GPa) compared to other silicate melt components (Ochs and Lange 1999). The effect
of 1 wt% dissolved H2O on the density of basaltic melt would be equivalent to increasing the
temperature by ~400 °C or decreasing the pressure by ~500 MPa (Ochs and Lange 1999).
Dissolved H2O also has a strong effect on melt viscosity (e.g., Shaw 1972; Richet et al.
1996; Giordano and Dingwell 2003), such that loss of H2O during ascent causes melts to
become increasingly viscous. Water exsolution and loss also exert a main control on the melt
crystallization (temperature and sequence of mineral phase appearance). The effect of H2O
on phase equilibria is such that H2O loss enhances crystallization, particularly of plagioclase,
whereas adding H2O to a melt depresses plagioclase crystallization in basaltic systems (e.g.,
Sisson and Grove 1993; Métrich and Rutherford 1998; Grove et al. 2003; Di Carlo et al. 2006;
Pichavant and Macdonald 2007). Finally, increasing the H2O content of a melt also increases
the diffusivities of other volatiles such as S (Freda et al. 2005) and halogens (Aletti et al.
2007). Thus the amount of H2O dissolved in magmas is one of the key parameters that must be
known to model physical properties and crystallization paths during ascent.
Assessing the volatile abundances in basaltic magmas, their degassing paths during ascent,
and the magmatic volatile budgets of active volcanoes has been a major objective of many
studies in the last 15 years. Because of strong pressure-dependent solubilities, the volatile
components are almost totally lost from erupted volcanic products, and therefore melt droplets
(melt or glass inclusions) entrapped during crystal growth have been widely used for addressing
these questions. Recent developments in microanalyses of volatile elements in melt inclusions
and glasses have led to large data sets on CO2 and H2O, which complement data for other
volatile species (Cl, S, F), trace elements and isotopes.
In this chapter we first review information on how melt inclusions form, the analytical
techniques for quantitative determination of volatile concentrations, and the results of
experimental solubility studies of volatiles. Second, we provide a general summary of volatile
abundances in basaltic magmas, volatile behavior during magma ascent, and vapor exsolution
pressures on the basis of melt inclusions hosted in olivine crystals from natural basaltic samples.
Third, we examine the relationships between melt inclusion data, eruption styles, and volcanic
gas fluxes. Finally, we discuss systematics of the primary or initial volatile contents in basaltic
magmas from different tectonic environments. Based on our review of the extensive literature
on these topics, we conclude by summarizing a few key questions for future research.
What are MeLt InCLuSIOnS and hOW dO they FOrM?
Primary melt inclusions form in crystals when some process interferes with the growth of a
perfect crystal, causing a small volume of melt to become encased in the growing crystal. This
can occur from a variety of mechanisms, including (1) skeletal or other irregular growth forms
due to strong undercooling or non-uniform supply of nutrients, (2) formation of reentrants
by resorption followed by additional crystallization, and (3) wetting of the crystal by an im-
miscible phase (e.g., sulfide melt or vapor bubble) or attachment of another small crystal (e.g.,
spinel on olivine) resulting in irregular crystal growth and entrapment of that phase along with
silicate melt (Sobolev and Kostyuk 1975; Roedder 1984; Lowenstern 1995). Most melt inclu-
sions contain only silicate melt at the time of trapping, but this single phase may exsolve one or
more vapor bubbles and/or crystallize daughter crystals post-entrapment as a result of cooling
Volatile Abundances & Degassing Paths Tracked by Melt Inclusions 365
(Lowenstern 1995). Melt inclusions formed by the third mechanism described above contain
multiple phases at the time of trapping, though additional vapor exsolution or crystallization
may occur post-entrapment (Roedder 1984). Magma mixing may cause multiple populations
of primary melt inclusions to be present in phenocrysts from a single volcanic rock or tephra
layer (Anderson 1976).
Because mafic melt inclusions hosted in olivine phenocrysts are the main focus of this
review, we summarize recent experimental data that bear on the formation of inclusions in this
mineral. Having constraints on the olivine texture and melt inclusion formation are suitable
before an accurate interpretation of volatile evolution. Building on an extensive experimental
literature on the morphology of olivine (e.g., Donaldson 1976), there have been several recent
studies (Faure et al. 2003; Faure and Schiano 2004, 2005) in which dynamic crystallization
experiments involving forsterite in the CMAS system have been compared with natural oliv-
ine and melt inclusion morphologies. These studies found 5 different crystal morphologies at
increasing degrees of undercooling: polyhedral, tabular, skeletal (hopper), dendritic (swallow-
tail), and feather shape. More than one morphological type was commonly produced in a single
experiment. The summary below is based on these studies.
Polyhedral crystals form at relatively low cooling rates (~1 to 2 °C/h) and have well de-
fined crystal faces. Experimentally grown polyhedral crystals commonly contained spherical
or elongated melt inclusions (Fig. 1A). Natural melt inclusion-bearing olivines commonly
show similar morphology, providing evidence that at least some inclusions form by slow, near-
equilibrium growth (Fig. 1B). Slowly grown, polyhedral olivines also show embayed textures
(Faure and Schiano 2005); such textures in natural olivines are commonly interpreted to be the
result of disequilibrium caused by resorption, but the experimental results show that these form
during near-equilibrium growth.
Tabular shape, in which crystals are rectangular in sections parallel to [010], are a less
common form in dynamic crystallization experiments, and the crystals do not contain trapped
melt inclusions. Olivine evolves from tabular to skeletal (hopper) to dendritic (swallowtail) with
increasing degree of undercooling (Faure and Schiano 2005). Evolution from tabular to skeletal
(hopper) is particularly important in forming the U-shaped cavities in some hopper crystals, as
described below (Faure et al. 2003).
Skeletal crystals (hopper morphology) form at higher cooling rates and degrees of under-
cooling and appear either as hexagonal or hourglass shapes depending on the plane of orienta-
tion. A characteristic feature of hopper crystals is that the ends of the prismatic crystal created
in the experiments have large glass-filled cavities. In larger crystals, these cavities have a U-
shaped profile, whereas in smaller crystals they are V-shaped (Faure et al. 2003). These cavities
can become sealed by additional olivine growth if such crystals formed by rapid cooling are
subjected to short (~5 min.), isothermal annealing. This results in a closed, skeletal olivine in
which two melt inclusions are symmetrically distributed with regard to the center of the crys-
tal (Fig. 2A). Longer isothermal annealing times (several hours) cause the cavities to become
completely filled by the growth of olivine (Faure et al. 2003; Faure and Schiano 2005). Oliv-
ine crystals with pairs of melt inclusions similar in morphology to the experimentally grown,
closed skeletal crystals with U-shaped cavities occur in Kilauea olivines (Fig. 2B). Melt inclu-
sions with an hourglass morphology were described from Paricutin volcano (Luhr 2001). This
morphology is similar to closed skeletal crystals with V-shaped cavities (Fig. 2B), but in some
cases growth was sufficiently rapid to leave a central channel in the crystal unfilled such that the
two cavities at the ends of the prism are connected (Luhr 2001).
More complex olivine and inclusion textures are created in closed dendritic olivine, which
are formed when initial rapid cooling forms a swallowtail olivine followed by isothermal an-
nealing to form a final morphology that appears polyhedral (Fig. 3A; Faure and Schiano 2005).
This produces a well faceted crystal with abundant melt inclusions showing preferred orienta-
366 Métrich
Experimental and natural & Wallace
polyhedral olivine with melt inclusions
A Melt
B 100 m
inclusion
Olivine
100 µm
Keanakakoi Ash
Kilauea, Hawaii
100 m
Jorullo volcano, Mexico
Figure 1. Comparison of (A) experimentally produced and (B) natural polyhedral olivines and their
melt inclusions. (A) Upper left: sketch showing a polydedral forsterite crystal with a large rounded glass
inclusion near the center (based on SEM image of an experimental olivine produced at a cooling rate 2 °C/
hr; Faure et al. 2003). Lower left: backscattered electron image of forsterite crystal showing polyhedral
morphology with large embayment and skeletal (stair-like) overgrowths (marked by arrows) produced
in a second stage of more rapid cooling. [Photo used by permission of Elsevier Limited from Faure and
Schiano (2005) Earth Planet Sci Lett, Vol. 236, Fig. 8A, p. 890.] (B) Upper right: photomicrograph of small
polyhedral olivine with large, central, partly faceted melt inclusion from the Keanakakoi Ash Member,
Kilauea Volcano, Hawaii. Lower right: similar to above, but with large, central irregular inclusion. Olivine
is from Jorullo Volcano, Mexico.
tion. Natural examples of such textures are an indication of the complex cooling histories by
which some olivines form (Figs. 3B, C).
Compositional boundary layers can form adjacent to growing crystals due to competition
between diffusion in the silicate melt and crystal growth. A primary concern in melt inclusion
studies is whether the trapped melt accurately represents the bulk melt surrounding the crystal
at the time of entrapment or has been affected by boundary layer enrichment or depletion
(Roedder 1984; Baker 2008). In the dynamic crystallization experiments described above
(Faure and Schiano 2005), polyhedral olivines contained melt inclusions that were similar
to the parental melt regardless of inclusion size (Fig. 4). For such crystals, there were no
compositional gradients in the glass surrounding the olivine, suggesting that growth was
controlled by interface attachment rather than diffusion processes. In contrast, more rapid
growth morphologies (skeletal, dendritic) contain inclusions that are moderately to strongly
enriched in Al2O3 compared to the parental melt (Fig. 4). Because both CaO and Al2O3 are
incompatible in olivine, in situ crystallization in the experiments should cause correlated
increases in both elements. The anomalously high Al2O3 contents in melt inclusions in rapidly
grown olivines and the presence of diffusional gradients in glass adjacent to the experimental
crystals suggests that Al2O3 builds up in the boundary layer because it has a much lower
Volatile Abundances & Degassing Paths Tracked by Melt Inclusions
Experimental and natural skeletal (hopper morphology) olivine with melt inclusions
367
Paricutin, Mexico
Keanakakoi Ash 500 m
Kilauea, Hawaii
Figure 2. Comparison of (A) experimentally produced and (B) natural hopper (skeletal morphology)
olivines and their melt inclusions. (A) Formation of closed skeletal olivine with symmetrically distributed
melt inclusions. Sections are parallel to the (010) plane. The two photomicrographs on the upper right show
evolution from hopper (cooling rate 422 °C/hr) to closed hopper crystal (cooling rate 437 °C/hr; then charge
was reheated for 5 min. below the liquidus temperature. [Photos used by permission of Elsevier Limited
from Faure and Schiano (2005) Earth Planet Sci Lett, Vol. 236, Fig. 11A, p. 895.] (B) Photomicrographs
of closed-hopper morphology in olivine phenocrysts from the Keanakakoi Ash Member, Kilauea Volcano
(3 photos on left), and Paricutin Volcano, Mexico (photo on right).
diffusivity than CaO in silicate melts. Baker (2008) demonstrates that P2O5 has very low
diffusivity in silicate melts as well, and shows that the ratio Cl/P2O5 is useful as an indicator of
whether melt inclusions have been affected by boundary layer enrichment processes.
The dynamic crystallization experiments provide a framework for testing whether natural
melt inclusions have been affected by boundary layer enrichment processes by comparing
CaO and Al2O3 in inclusions to bulk rock or tephra and matrix glass. Shown in Figure 5A are
CaO and Al2O3 data for olivine-hosted melt inclusions from high-K minettes and basanites
from the Colima Graben in western Mexico (Vigouroux et al. 2008). The olivine hosts are
mostly polyhedral, and the inclusions vary from ellipsoidal to irregular in shape. Most of the
melt inclusions have CaO/Al2O3 ratios that are similar to values for the bulk tephra (Fig. 5A).
Although a few inclusions have anomalous CaO/Al2O3 ratios, these are more likely the result
of augite fractionation. The data show no evidence for the strongly enriched Al2O3 values seen
in melt inclusions in rapidly grown experimental forsterite crystals in the CMAS system (cf.
Fig. 4). Similarly, melt inclusions from Etna scoriae show no evidence of anomalously high
Al2O3 (Spilliaert et al. 2006a). Interestingly, olivine-hosted melt inclusions from Paricutin
Volcano, many of which have closed skeletal forms indicating rapid crystal growth (Luhr
2001), also do not show any anomalous behavior of Al2O3 (Fig. 5B).
The results summarized above show that compositions of natural olivine-hosted melt
inclusions are partly at odds with the results of dynamic crystallization experiments in that rapid
368 Métrich & Wallace
Experimental and natural closed dendritic olivine with melt inclusions
B 100 m
Blue Lake Maar, Oregon Cascades
Stromboli Volcano
Figure 3. Comparison of (A) experimentally produced and (B) natural dendritic polyhedral olivines and
their melt inclusions. (A) Swallowtail olivine that is reheated below the liquidus temperature develops a mor-
phology similar to polyhedral olivine, but with numerous melt inclusions showing preferential orientations.
[Photo used by permission of Elsevier Limited from Faure and Schiano (2005) Earth Planet Sci Lett, Vol.
236, Fig. 11B, p. 895.] (B) Photomicrograph of olivine crystal from Oregon Cascades showing numerous
melt inclusions with preferred orientation, similar to experimental textures in (A). (C) Olivine crystal from
Stromboli showing concentric zones rich in small melt inclusions showing preferred orientations. This likely
represents rapid cooling or H2O-loss induced pulses of swallowtail growth followed by slow crystallization.
growth morphologies do not show any Al2O3 enrichment compared to bulk sample values. One
possibility is that the presence of several wt% H2O and/or additional melt components such
as Fe increase the diffusivity of components such as Al2O3 in the melt. While we suggest that
all melt inclusion data sets should be scrutinized in terms of elements that are incompatible
in olivine, the lack of Al2O3 enrichment in natural olivine-hosted melt inclusions suggests
that H2O, which diffuses faster than Al2O3, should not have been affected by boundary layer
enrichment processes during melt inclusion formation. This interpretation is consistent with
recent modeling results (Baker 2008). Diffusivities of H2O and CO2 in mafic melts have only
been measured at a limited set of compositions and temperatures (Baker et al. 2005; zhang
et al. 2007). Recent experimental studies of S and Cl diffusion in silicate melts show that in
nominally anhydrous basaltic melts, Cl diffusivity is about 10 times lower than that of H2O,
similar to that of CO2, and a factor of 5 higher than that of S (Freda et al. 2005; Alletti et al.
2007). The presence of H2O increases the diffusivity of Cl by no more than a factor of 3 (Alletti
et al. 2007). Based on available data, models of incompatible element build-up in a boundary
layer adjacent to a growing crystal suggest that trapped basaltic melt inclusions may, if formed
by sufficiently rapid growth, be enriched in S, Cl, and CO2 as a result of boundary-layer
effects, but the enrichments are likely to be on the order of 10-30% or less (Baker 2008).
Different diffusivities for S and Cl should result in some scattering of S/Cl ratios in melt
inclusions. However, many basaltic melt inclusion data sets show no significant variations in
Volatile Abundances & Degassing Paths Tracked by Melt Inclusions 369
dendritic polyhedral
21 forsterite
20 dendritic
forsterite
19 of
d line a tion
i i
liqu eren t
di
skeletal
18
Al 2O3 (wt%)
forsterite
17
polyhedral
forsterite
16
Equilibrium host glass
Closed glass inclusions in polyhedral forsterite
Open glass inclusions in polyhedral forsterite
15 Open glass inclusions in skeletal forsterite
Open glass inclusions in dendritic forsterite
"Closed" glass inclusions in dendritic polyhedral forsterite
Open glass inclusions in dendritic polyhedral forsterite
14
17 18 19 20 21 22 23
CaO (wt%)
Figure 4. CaO vs. Al2O3 showing compositions of melt inclusions in polyhedral, dendritic polyhedral,
skeletal and dendritic forsterite produced experimentally in the CMAS system. Both CaO vs. Al2O3 are
incompatible during forsterite growth. The initial melt composition is shown by the solid diamonds in the
lower left, and the final quenched residual melts are shown by the gray diamonds. [Figure used by permission
of Elsevier Limited from Faure and Schiano (2005), Earth Planet Sci Lett, Vol. 236, Fig. 4, p. 887.]
the S/Cl ratio (before S degassing or sulfide immiscibility) for inclusions having contrasting
morphologies indicative of different crystallization rates for their host olivines. This is
illustrated by average values of the S/Cl ratios (±1σ) in H2O -rich, oxidized basaltic inclusions
of Vesuvius (S/Cl = 0.34±0.05; Marianelli et al. 2005); Stromboli (S/Cl =1.1±0.1; Métrich
et al. 2005), and Etna (2.0±0.2; Spilliaert et al. 2006a). Instead, diffusion might significantly
affect the S/Cl ratio in melt and gas during dynamic degassing controlled by bubble growth.
POSt-entraPMent MOdIFICatIOn OF MeLt InCLuSIOnS
A number of processes can potentially modify the compositions of melt inclusions after
they are trapped inside their crystal host (e.g., Danyushevsky et al. 2002). Most published
studies on post-entrapment modification are based on melt inclusions in olivine, and we
therefore review mechanisms of post-entrapment change and the ensuing limitations to data
interpretation (see also Kent 2008). During post-entrapment cooling, crystallization of the
included melt occurs along the melt–crystal interface, depleting the melt in constituents that
enter the crystalline phase and enriching it in elements incompatible in the crystal. For olivine-
hosted melt inclusions, post-entrapment crystallization of olivine causes a strong depletion in
MgO of the included melt, but other major elements are much less affected (Danyushevsky et
al. 2000). Diffusive exchange can also occur between the melt and crystal. This particularly
affects the FeOT contents of many melt inclusions in olivine (Danyushevsky 2001), but both
CaO (Gaetani and Watson 2000, 2002) and trace element abundances (Cottrell et al. 2002) can
also be modified.
370 Métrich & Wallace
Augite
fractionation
B Paricutin
Constant CaO/Al2O3
Figure 5. CaO vs. Al2O3 showing compositions of natural olivine-hosted melt inclusions from (A) high-K
minette and basanite tephra from cinder cones near Volcan Colima in the western Trans-Mexican Belt
(data from Vigouroux et al. 2008) and (B) basaltic andesite tephra from Paricutin volcano, Mexico (data
from Luhr 2001). Solid and dashed lines show constant CaO/ Al2O3 ratios for the bulk tephra samples
corresponding to each of the melt inclusion data sets. Some of the inclusions in (A) show evidence for
augite fractionation, but none of the inclusions show strong enrichments like those found in many of the
CMAS melt inclusions in Figure 4.
Of primary concern for studies of volatiles in basaltic systems is the potential for post-
entrapment diffusive loss of volatiles from melt inclusions through the olivine host. During 1-atm
heating experiments on olivine-hosted melt inclusions, systematic increases in homogenization
temperature occur over time (Sobolev et al. 1983) in response to the deformation of the host
crystal (Massare et al. 2002). FTIR measurements have verified that this is coupled with a
decrease in hydroxyl and molecular H2O concentrations in the inclusions, and almost total
dehydration of inclusions can occur within a few hours at high temperature (Massare et al. 2002).
It is likely that the dehydration is caused by rapid proton diffusion through the olivine due to
the very low H2 fugacity of the environment surrounding the olivine in the heating experiments
(Danyushevsky et al. 2002), or preferential diffusion of water-derived species (OH, molecular
H2O) through olivine dislocations and lattice defects (Massare et al. 2002). Hydrogen loss causes
oxidation of Fe in the melt inclusion that results in precipitation of magnetite in a characteristic
magnetite dust texture (Danyushevsky et al. 2002). Loss of water from olivine-hosted melt
Volatile Abundances & Degassing Paths Tracked by Melt Inclusions 371
inclusions by proton diffusion either during experimental heating studies or in slowly cooled
lava flows causes anomalously high D/H ratios because of the more rapid diffusion of H relative
to D (Hauri 2002).
The amount of H2O that can be lost from an inclusion by hydrogen diffusion was suggested
to be limited to ≤ 1 wt% H2O by the amount of FeOT in the inclusion and the stoichiometry of
the oxidation reaction (Danyushevsky et al. 2002), but larger H2O losses have been documented
experimentally, suggesting that other mechanisms may also operate (Massare et al. 2002).
Recent experiments have shown that melt inclusions with initially low H2O concentrations (<1
wt%) can gain substantial H2O in as little as 2 days at magmatic temperature when surrounded
by more H2O-rich melt (Portnyagin et al. 2008). In these experiments, the addition of H2O to
the system changes the mineral-melt equilibria leading to dissolution of pre-existing olivine
and MgO enrichment in the melt. These results have been suggested to indicate that molecular
H2O is a rapidly diffusing species through the olivine and that melt inclusions can either gain
or lose H2O depending on changes in the magmatic environment surrounding the crystal
(e.g., loss of H2O during ascent and degassing or increases in H2O during isobaric fractional
crystallization).
There is no evidence, as far as we are aware, for diffusive loss of CO2, Cl, or S from melt
inclusions (e.g., Portnyagin et al. 2008). However, diffusive loss of Fe from melt inclusions can
lead to exsolution of a sulfide globule in the inclusion and consequent decrease in dissolved S
content of the silicate melt (Danyushevsky et al. 2002).
Another important post-entrapment modification affecting melt inclusions is the formation
of shrinkage bubbles (Roedder 1984). During post-entrapment (but pre-eruptive) cooling, the
much greater thermal contraction of the melt relative to the crystal host results in formation
of a vapor bubble (shrinkage bubble) that can significantly deplete the melt in CO2 because of
its low solubility (Anderson and Brown 1993; Cervantes et al. 2002). For olivine-hosted melt
inclusions, a decrease in temperature of 100 °C is estimated to create a shrinkage bubble that
is 0.9 to 1.6 vol% of the inclusion (Anderson and Brown 1993; Riker 2005). Further expansion
of the bubble occurs during eruptive cooling, but this happens over such a short timescale that
little to no additional volatiles are likely to diffuse from the melt to the bubble (Anderson and
Brown 1993). In experiments on olivines (Fo88) from a Mauna Loa picrite, reheating to 1400
°C for <10 minutes rehomogenized the shrinkage bubble (≤3 vol%) into the melt, but even
with rapid quenching a vapor bubble commonly formed during quench (Cervantes et al. 2002).
Analyses of the rehomogenized melt inclusions by FTIR showed that ~80% of the total CO2
that was dissolved in the melt at the time of trapping had been lost to the shrinkage bubbles
during an ~100 °C cooling interval between entrapment and eruption. Smaller cooling intervals
would result in less loss of CO2 to shrinkage bubbles. Basaltic melt inclusions in Hawaiian
olivines that experience ≤40 °C cooling between entrapment and eruption generally do not
contain shrinkage bubbles (Riker 2005).
During ascent and eruption, volatiles can be lost if the host crystal ruptures around an
inclusion or if an inclusion is not fully sealed in the host. The resulting depressurization in the
inclusion forms one or more vapor bubbles, and such inclusions can typically be recognized
by thin films of glass that form where melt is injected into the cracks. Slow cooling of melt
inclusions in lava flows or domes causes crystallization of the trapped melt, and such inclusions
must be experimentally reheated for analysis (Danyushevsky et al. 2002). However, additional
volatiles can be lost if the inclusion host is cracked or if there is diffusion of H or molecular
H2O through the host, and these lost volatiles cannot be restored to the melt through reheating
and rehomogenization.
Given the potential for volatile loss from partially to fully crystallized melt inclusions
from slowly cooled lava flows and domes, it is preferable to analyze naturally glassy melt
inclusions from rapidly cooled tephra deposits (e.g., Anderson and Brown 1993; Roggensack
372 Métrich & Wallace
et al. 1997; Métrich et al. 2001) or submarine samples. When analysis of rehomogenized melt
inclusions is necessary, short heating times (preferably ≤ 15 min) are required to prevent H2O
loss by diffusion (Danyushevsky et al. 2002; Massare et al. 2002) or decreases in melt S and
changes in S6+/Stotal due to oxidation of the inclusion (Danyushevsky et al. 2002; Rowe et al.
2007). Finally, it should be remembered that even naturally glass melt inclusions from tephra
or submarine samples have the potential to have lost or gained some H2O by diffusion through
the olivine host during pre-eruptive storage or ascent. Cases where this has been unequivocally
documented are rare (e.g., Portnyagin et al. 2008).
anaLytICaL teChnIqueS FOr VOLatILeS
The main microanalytical techniques that are now routinely used to analyze the major
volatiles in silicate melt inclusions are secondary ion mass spectrometry (SIMS or ion
microprobe; H2O, CO2, S, Cl, F), Fourier Transform Infrared (FTIR) Spectroscopy (H2O, CO2),
electron microprobe (S, Cl, F, S6+/S2− ratio), and confocal micro-Raman spectroscopy (H2O). In
addition, a nuclear microprobe technique has been used to analyze C in glasses.
A thorough review of the theory and application of SIMS and FTIR for analysis of volatiles
can be found in Ihinger et al. (1994). More recent discussions of calibration issues, matrix
effects, and comparison of SIMS and FTIR for analysis of glasses is given by Hauri et al. (2002)
and Blundy and Cashman (2008). An advantage of SIMS relative to FTIR is that melt inclusions
to be analyzed only need to be exposed on one side of the host crystal, but the SIMS technique
is destructive whereas FTIR is not. Another advantage of SIMS for volatile analysis is that all
major volatiles (H2O, CO2, S, Cl, F) together with trace elements can be analyzed on the same
instrument. A comparison of SIMS, FTIR, and bulk extraction by hydrogen manometry for
andesitic glasses can be found in King et al. (2002). A summary of sample preparation issues
and the applications of SIMS and FTIR to volatile analysis in melt inclusions can be found in
Nichols and Wysoczanski (2007).
An essential part of analysis of H2O and CO2 by FTIR is determination of absorption (or
extinction) coefficients for the species of interest. The fundamental OH stretching vibration
that forms an assymetric band with maximum around 3570 cm−1 is commonly used to measure
total H2O in a sample. This band has an absorption coefficient that is only weakly dependent
on major element composition (Dixon et al. 1995), in contrast to the bands for molecular H2O
(5200 and 1630 cm−1) and hydroxyl (4500 cm−1; Dixon et al. 1995; King et al. 2002; Mandeville
et al. 2002), which have absorption coefficients that more strongly depend on bulk composition.
For basaltic melt inclusions with less than ~4 wt% dissolved H2O, using the band at 3570 cm−1
insures better precision on H2O measurements. The absorption coefficient for carbonate ion in
mafic glasses is also strongly dependent on major element glass composition (Dixon and Pan
1997).
For quantitative determination of H2O and CO2 by transmission FTIR, the thickness of the
melt inclusion-bearing wafer must be measured precisely, and this is normally done either with
a high-precision digital micrometer or by optically measuring the thickness under a microscope
(e.g., Wallace et al. 1999). A microprofilometer can also be used for this (Roggensack 2001).
Spectroscopic methods for determining the wafer thickness include a method based on ratioing
the absorbances at different regions of the background (Agrinier and Jendrzejewski 2000)
and using the spacing of interference fringes in reflectance spectra taken on the host crystal
(Wysoczanski and Tani 2006). Recently, a technique has been developed for analyzing small
melt inclusions that are fully enclosed inside a wafer but not actually intersected at the wafer
surface (Nichols and Wysoczanski 2007).
A technique for using confocal micro-Raman spectroscopy to measure H2O in melt
inclusions has recently been developed and calibrated (Thomas 2000; Behrens et al. 2006;
Volatile Abundances & Degassing Paths Tracked by Melt Inclusions 373
Di Muro et al. 2006). Besides being non-destructive, like FTIR, a major advantage of this
technique is that inclusions do not have to be double-face polished and in transparent crystals
they can be analyzed in situ without exposure to surface (Thomas et al. 2006). This feature
and the very small spatial resolution (1-2 µm) make it possible to analyze a large number of
inclusions within a single host crystal. With proper calibration, this technique is very promising
for analyzing H2O in glasses ranging from basaltic to rhyolitic in composition (Mercier et al.
2008). For most compositions the minimum detection limit is assessed to be ~0.1 wt% of H2OT,
or even lower in Fe-poor glasses.
Carbon concentrations can be determined using the 12C(d, p)13C nuclear reaction (e.g.,
Varela et al. 2000; Spilliaert et al. 2006a). The samples are irradiated with an incident deuteron
beam of 1.45 MeV and the protons detected between 2.520 and 3.036 MeV behind a 9 µm thick
Al screen with a surface barrier detector (1500 µm depleted depth). Any surface contamination
is identified by a high energy peak. At 1.45 MeV incident energy the main interferences are due
to (d, p) nuclear reactions with the major constituents of the matrix such as 24Mg, 28Si, 27Al.
In order to take into account the matrix effect, the background is measured on C-free silicate
glasses. In silicate glasses (density = 2.6 to 2.7 g/cm3), the analyzed depth is 8-9 µm. The
minimum detection limit for carbon varies from 40 to 15 ppm as the integrated charge increases
from 1 to 3 °C. The uncertainty (1σ) using reference basaltic glasses containing 300 and 100
ppm of carbon is 10%.
The electron microprobe is commonly used to analyze Cl, S and F in melt inclusions, as
well as the major element composition of inclusions and their host crystals. Technical details
regarding analysis of F and Cl in glasses by electron microprobe are reviewed by Carroll and
Webster (1994). Sulfur occurs in multiple valence states (S2−, S6+) in glasses, and the wavelength
of S Kα radiation changes as a function of the mean oxidation state (Carroll and Rutherford
1988; Wallace and Carmichael 1994; Jugo et al. 2005a). Therefore it is important to measure
the appropriate peak position to use during the electron microprobe analysis of S in silicate
glasses, and the peak position used for glass samples should be different from that used to
analyze sulfide or sulfate standards used for calibration. More details regarding the S Kα peak
shift technique and interpretation of the results in terms of oxygen fugacity are discussed by
Kent (2008).
Major element data for melt inclusions are often used to estimate the amount of H2O in
the inclusion by the difference of the major element total (including S, Cl and F) from 100%.
However, electron beam radiation causes permanent damage to hydrous glasses during electron
microprobe analysis because alkali ions migrate away from the surface of the irradiated sample
(Humphreys et al. 2006). If not properly corrected for, electron microprobe analyses of hydrous
glasses will result in values of Na2O that are far too low, and this effect seriously degrade
estimates of H2O by difference. Methods for accurate analysis of Na2O and H2O by difference
in hydrous glasses are discussed in Blundy and Cashman (2008).
VOLatILe SOLuBILItIeS and VaPOr MeLt PartItIOnInG
Volatile components occur as dissolved species in silicate melts, but they can also be
present in an exsolved vapor phase if a melt is vapor saturated. In laboratory experiments
it is possible to saturate melts with a nearly pure vapor phase (e.g., H2O saturated), though
the vapor always contains at least a small amount of dissolved solute. In natural systems,
however, multiple volatile components are always present, and data from volcanic gases, melt
inclusions, and submarine pillow rim glasses show that the major volatiles present in magma
are H2O, CO2, S, Cl and F. When the sum of the partial pressures of all dissolved volatiles in a
silicate melt equals the confining pressure, the melt becomes saturated with a multicomponent
(C-O-H-S-Cl) vapor phase (Verhoogen 1949). Referring to natural magmas as being H2O
374 Métrich & Wallace
saturated or CO2 saturated is thus, strictly speaking, incorrect because the vapor phase is never
pure and always contains more than one volatile component.
Earlier experimental studies on volatile solubility in silicate melts generally focused on
melts saturated with a single volatile component, but more recent studies on basaltic to andesitic
melts have examined solubility relations when multiple volatile components are present (Luhr
1992; Dixon et al. 1995; Webster et al. 1999; King and Holloway 2002; Botcharnikov et
al. 2007; Moore 2008). Though experimentally difficult, it is particularly useful to measure
the composition of the vapor phase coexisting with vapor-saturated melt so that vapor-melt
partitioning of volatiles can be determined as a function of temperature, pressure, melt
composition, and oxygen fugacity. When compared with experimental solubility studies, the
data on dissolved H2O and CO2 concentrations in olivine-hosted melt inclusions that we review
in this chapter provide strong evidence that basaltic magmas are vapor-saturated in the middle
to upper crust. In this section, we review recent experimental solubility data and thermodynamic
models useful for understanding vapor saturation and degassing in basaltic systems.
For H2O and CO2, the maximum solubility in a silicate melt is controlled by saturation
with the pure vapor phase. In contrast, for S, Cl, and F, the maximum solubility is limited
by saturation with a solid or immiscible liquid phase such as anhydrite, Fe-S-O liquid, alkali
chloride melt, or alkali fluoride melt (e.g., Luhr 1990; Wallace and Carmichael 1992; Carroll
and Webster 1994; Webster et al. 1999; Jugo et al. 2005a).
Because H2O is typically the most abundant volatile component in natural melts, and
H2O and CO2 together exert the largest contribution towards the total vapor pressure, many
experimental studies have focused on the solubility behavior of these two volatiles. H2O dissolves
in silicate melts to form both OH– and molecular H2O species (Silver et al. 1990), and CO2
dissolves in basaltic melts as carbonate ions (Fine and Stolper 1986). At pressures up to 1 kbar
(100 MPa), both dissolved carbonate and molecular H2O obey Henry’s Law solubility behavior
such that vapor saturation isobars on a diagram of H2O vs. CO2 show a negative diagonal
trend (Fig. 6; Dixon and Stolper 1995). This makes it possible to estimate the vapor saturation
pressures for melt inclusions using H2O-CO2 data (assuming the partial pressures of the other
volatile components are small in comparison), and these values can be equated to the trapping
pressure of the inclusion if it formed from vapor-saturated melt. However, the solubilities of
both H2O (Moore et al. 1998) and CO2 (Dixon 1997) in silicate melts are also dependent on
melt composition as well as temperature, so these effects must also be taken into account in
making vapor saturation pressure calculations. Solubility, degassing and vapor-melt partitioning
of H2O and CO2 can be calculated for nephelinitic to basaltic and rhyolitic melts with the
VolatileCalc program (Newman and Lowenstern 2002), which uses thermodynamic solubility
models calibrated with experimental data. Software for performing similar calculations using
a different solubility model (Papale et al. 2006), is now available (http://ctserver.ofm-research.
org/Papale/Papale.php). The accuracy of these models is evaluated in light of new H2O-CO2
solubility data by Moore (2008).
For basaltic melts with oxygen fugacities near or below the NNO buffer, S saturation is
controlled by an immiscible Fe-S-O liquid phase (Wallace and Carmichael 1992). Dissolved
S concentrations of sulfide-liquid-saturated melts show a strong positive dependence on melt
FeO and temperature and a negative dependence on pressure (Wallace and Carmichael 1992;
Mavrogenes and O’Neill 1999; O’Neill and Mavrogenes 2002; Liu et al. 2007). In these systems,
the saturation concentration of S in the melt is also a strong function of the ratio of network
modifying to network forming cations (Liu et al. 2007). There has been little experimental
investigation of the effect of dissolved H2O on S concentration of sulfide-liquid-saturated melts,
but Liu et al. (2007) report experimental data on an Etna basalt at 1 GPa in which increasing
H2O from ~0 to 4 wt% caused a decrease in melt S content by about 50%.
Volatile Abundances & Degassing Paths Tracked by Melt Inclusions 375
Figure 6. Inverse correlation of CO2 and H2O in vapor-saturated basaltic melt at various constant pressures
and 1200 °C. Solid curves are each for a different constant pressure (vapor saturation isobars). Dashed
curves show isopleths of constant vapor composition. Both the isopleths and isobars are calculated using a
thermodynamic solubility model calibrated with experimental data (Dixon and Stolper 1995). This figure
can be used to determine the pressure and composition of the vapor phase in equilibrium with a basaltic
melt that is vapor saturated. [Figure used by permission of Oxford University Press, from Dixon and
Stolper (1995), J Petrol, Vol. 36, Fig. 1, p. 1635.]
In more oxidized melts, maximum dissolved S is controlled by saturation with either
anhydrite (Carroll and Rutherford 1987; Luhr 1990) or immiscible sulfate melt (Jugo et al.
2005a). Melt S concentrations are much higher for sulfate-saturated melts than for sulfide-
liquid-saturated melts (Fig. 7), and dissolved S shows a strong positive dependence on both
temperature and pressure but much less dependence on melt composition. Measurements of S
Kα wavelength in glasses show that the dramatic increase of S solubility with oxygen fugacity
reflects a change from dominantly S2− in the melt to dominantly S6+ (Carroll and Rutherford
1988; Métrich and Clocchiatti 1996; Jugo et al. 2005b). It has been shown that the presence of
S4+ species in natural oxidized glass inclusions and low pressure silicate glasses is an analytical
artifact caused by electron microprobe beam damage (Métrich et al. 2003) or by irradiation
with the synchrotron x-ray beam during analysis (Wilke et al. 2008). Recent x-ray absorption
spectroscopy determination of the sulfur oxidation state in experimental glasses and inclusions
points to the fact that sulfur in reduced basaltic magmas is possibly oxidized upon cooling
(Métrich et al. 2008).
Empirical and thermodynamic models for prediction of S concentrations in sulfide-
saturated melts include models calibrated for anhydrous basaltic melts (Holzeid and Grove
2002; O’Neill and Mavrogenes 2002; Li and Ripley 2005), and anhydrous to hydrous basaltic
to rhyolitic melts (Moretti and Ottonello 2005; Scaillet and Pichavant 2005; Liu et al. 2007).
These models were compared by Liu et al. (2007), who concluded that their model predicted
S concentrations at sulfide saturation in hydrous melts better than the models calibrated for
anhydrous systems. They also found that the Scaillet and Pichavant (2005) model was more
successful at predicting S concentrations at sulfide saturation in Fe-poor rhyolitic melts, but
376 Métrich & Wallace
10.00
Jugo et al. (2005)
Carroll and Rutherford
(1985, 1987)
Sulfur content, (wt. %) 1.00
0.10
0.01
-4 -2 0 2 4 6
Oxidation state, ∆FMQ
Figure 7. Experimentally determined S concentration as a function of oxygen fugacity in basaltic melts at
1300 °C and 1 GPa and trachyandesite melts (with <17 wt% FeO) at 200 MPa and 927-1027 °C. The arrow
indicates the expected shift in the stability of sulfate with decreasing silica activity (Sisson 2003). [Figure
used by permission of Oxford University Press, from Jugo et al. (2005b), J Petrol, Vol. 46, Fig. 3, p. 794.]
less successful for other melt compositions. One advantage of the Scaillet and Pichavant (2005)
model for use in melt inclusion studies is that it explicitly includes the relationship between
melt S content and fS2. This makes it possible to calculate the mole fractions of H2S and SO2
in the vapor phase in equilibrium with melt if the oxygen fugacity, temperature, pressure and
H2O content are known.
Experimental studies show that Cl is highly soluble in basaltic melts, with maximum
dissolved values as high as 2.9 wt% for H2O-poor melts saturated with molten NaCl at 2 kbar
(200 MPa) and 1080-1210 °C (Webster et al. 1999). With increasing H2O in the system, the
molten NaCl contains some dissolved H2O and is therefore best described as a hydrosaline
melt. The dissolved Cl content does not vary much for melts containing as much as ~4 wt% H2O
(Fig. 8). At dissolved H2O contents of 5 to 6 wt%, the melt is in equilibrium with an H2O-Cl
vapor phase rather than hydrosaline melt. Within the resolution of the data, it appears that there
is no invariant point where hydrosaline melt and H2O-Cl vapor coexist, but rather that there is
complete miscibility between the two phases. Olivine-hosted basaltic melt inclusions generally
have Cl concentrations << 1 wt%, indicating that the melts would have been saturated with Cl-
bearing vapor rather than dense hydrosaline melt during storage, ascent and degassing.
The maximum Cl contents of andesitic melts determined experimentally under similar
conditions as those described above are slightly lower (2.5 wt% Cl) than for basaltic melt
(Botcharnikov et al. 2007). In contrast to the basalt experiments, however, the H2O-Cl data for
andesitic melts suggests coexistence of hydrosaline melt and H2O-Cl vapor for melts with high
dissolved H2O and Cl, and the addition of CO2 to the vapor appears to increase the extent of the
immiscibility gap between the two phases (Botcharnikov et al. 2007). This behavior, though, is
unlikely to affect natural andesitic melts because, like their basaltic counterparts, they generally
do not contain high enough Cl to reach saturation with hydrosaline melt.
Volatile Abundances & Degassing Paths Tracked by Melt Inclusions 377
Figure 8. Solubilities of H2O and Cl in basaltic melt saturated with Cl-bearing aqueous vapor or hydrosaline
melt at 200 MPa and 1040-1170 °C. Numbers next to data points give the measured wt% Cl in the vapor
or hydrosaline melt; numbers preceded by ≈ represent calculated values that may have very large errors.
Filled squares are melts with 9-10.9 wt% FeO and open squares are melts with 6-8.1 wt% FeO. Silicate
melt + vapor are stable along the horizontal part of the curve, and silicate melt + hydrosaline melt are stable
along the vertical part. Vapor and hydrosaline melt are inferred to be completely miscible along the curved
portion. [Figure used by permission of Elsevier Limited, from Webster et al. (1999), Geochim Cosmochim
Acta, Vol. 63, Fig. 2, p. 733.]
A detailed review of experimental solubility studies of F in silicate melts can be found in
Carroll and Webster (1994).
h2O and CO2 aBundanCeS In BaSaLtIC MaGMaS: CryStaLLIZatIOn
dePthS, MaGMa aSCent and deGaSSInG PrOCeSSeS
h2O-CO2 abundances in basaltic magmas
Direct assessment of the dissolved amounts of H2O and CO2 in magmas comes from
melt inclusions, although experimental phase equilibria also provide strong constraints on
magmatic H2O contents. A large literature is dedicated to basaltic melt inclusions, specifically
hosted in olivine, because this phase is usually present along the overall crystallization paths of
mafic magmas. The possible post-trapping evolution of olivine-hosted melt inclusions is also
relatively well studied (see section on post-entrapment modifications). In contrast, systematic
studies on basaltic melt inclusions in clinopyroxene are scarce.
Available data on CO2 contents of basaltic melt inclusions from arc settings indicate a
limited range of concentrations (0-≤0.25 wt%; Fig. 9A). Dissolved CO2 contents >0.3 wt%
have only rarely been reported in basaltic melt inclusions, whereas primary basaltic melts
are inferred to have high CO2 contents up to 0.6 wt% in MORB (Fisher and Marty 2005), 0.7
wt% at Kilauea (Gerlach et al. 2002), and possibly as high as 0.6-1.3 wt% in arcs (Wallace
2005). Similarly, olivine-hosted melt inclusions from Etna contain ≤0.4 wt% CO2 (Métrich
et al. 2004), whereas an original CO2 content of ~1.5 wt% for Etna primary magma (Allard
378 Métrich & Wallace
Figure 9. Compilation of CO2 (A)
and H2O (B) concentrations in melt
inclusions from subduction-related
basalts and basaltic andesites. Note that
CO2 concentrations <2000 ppm and
most of the low H2O concentrations
are likely the result of partial degassing
before inclusion entrapment (see text).
Data from Central America [Fuego
(Roggensack 2001; Sisson and Layne
1993); Popocatepelt and Colima (Atlas
et al. 2006); Cerro Negro (Roggensack
2001); Irazù (Benjamin et al. 2007);
Arenal (Wade et al. 2006); BVF and
Pacaya (Walker et al. 2003); Paricutin
(Lurh 2001); Mexico (Cervantes and
Wallace 2003a,b; Johnson et al. 2008)];
Lesser Antilles (Bouvier et al. 2008);
Kamchatlka (Portnyagin et al. 2007);
Aeolian arc [Stromboli (Métrich et al.
2001; Bertagnini et al. 2003); Vulcano
(Gioncada et al. 1998)]; Mariana
(Newman et al. 2000); and Galunggung
(Sisson and Bronto 1998).
et al. 1997) was inferred from the time-averaged CO2/S ratio in summit plume emissions
and the S content of the most primitive melt inclusions. Olivine-hosted melt inclusions from
high-K minettes and basanites from the Colima Graben in western Mexico contain ≤0.53 wt%
CO2 (Vigouroux et al. 2008). Primary magmas may therefore be vapor saturated at crustal
depths (Anderson 1995; Wallace 2005), and exsolved CO2-rich gas bubbles would enhance the
magma buoyancy and thus its eruptive style as discussed below.
Water concentrations recorded by melt inclusions in subduction settings show a broad
range from 0.2 to 6 wt% (Fig. 9B) that overlaps with values for back arc basin basalts (BABB).
In comparison, basaltic melt inclusions from non-arc settings average 0.12 and 0.5 wt% H2O,
respectively, for N-MORB and E-MORB (Sobolev and Chaussidon 1996), and vary mostly
between 0.2 and 0.8 wt% in Hawaiian basalts (Anderson and Brown 1993; Wallace and
Anderson 1998). Some lower values for Hawaiian basalts have probably been affected by H
diffusive loss from melt inclusions in slowly cooled lava flows (Hauri 2002).
The large range of H2O contents in arc basaltic melt inclusions results from several
factors: variable degassing at middle to upper crustal depths before inclusion entrapment, H2O
increase during magma differentiation, and ultimately, variable H2O content in the magma
source region. As an example of the effects of differentiation, basaltic andesite melt inclusions
from Fuego volcano (Sisson and Layne 1993) contain some of the highest H2O concentrations
(~6 wt%) in arc magmas but have less than 5 wt% of MgO and are trapped in Fe-rich olivine
(Fo~74). An example of the effects of low pressure degassing before inclusion entrapment
comes from Stromboli scoriae, in which olivine-hosted melt inclusions contain low H2O (<0.5
wt%; Métrich et al. 2001).
Volatile Abundances & Degassing Paths Tracked by Melt Inclusions 379
Focusing on melt inclusions that best represent magmatic values prior to degassing, the
H2O contents of subduction-related basaltic magmas varies from 0.2-0.3 wt% (Galunggung,
Indonesia arc; Sisson and Bronto 1998) to 5.2 wt% (Central Mexico; Cervantes and Wallace
2003a). These inclusions were trapped in Mg-rich olivines (Fo≥88) and contain high
concentrations of S (~0.2 and 0.6 wt%, respectively) and CO2 (0.075 and 0.2 wt%, respectively).
They likely record the primitive signature of magma prior to significant degassing of H2O
and S. Additional arguments support this conclusion. In Galunggung samples, melt inclusions
have an H2O content that remains identical and low (0.2-0.3 wt%) while their CO2 content
decreases from 0.075 wt% down to a few ppm upon decompression. Moreover, their basaltic
composition is very similar to that of the host magmas, even though melt inclusions suffered
some post-entrapment crystallization and Fe-loss (Sisson and Bronto 1998). Low H2O
concentrations (0.2 wt%) have also been reported in high-alumina basaltic melt inclusions
hosted in Fo~88 olivine from Black Crater basalt in the Cascade arc (Medicine Lake volcano;
Sisson and Layne 1993). In the latter case there is also agreement between low H2O content
of melt inclusion and their MORB-like composition. Similarly, low H2O contents (0.1-0.6
wt%) associated with low Cl (0.02-0.03 wt%) have been reported in picritic to basaltic melt
inclusions in Southern Mariana arc olivines (Fo89−91). These low values have been interpreted
as evidence for adiabatic decompression melting of a relatively anhydrous but metasomatized
mantle wedge (Kohut et al. 2006).
In contrast, inclusions recording high H2O concentrations from Central Mexico basalts
show the strongest enrichments of large ion lithophile elements (LILE) relative to high field
strength elements (HFSE), which is a characteristic of subduction-related magmas worldwide.
This pattern is consistent an origin related to mantle fluxing by an H2O-rich component
derived from the subducted slab (Cervantes and Wallace 2003a). Hence there is consistency
between the H2O content and the magma geochemical signature as discussed in the final
section below.
A large range in H2O concentrations is exemplified at the scale of the Mariana arc where
primitive magmas are believed to contain 1-3 wt% H2O prior to degassing (Newman et al.
2000). A similar range is also illustrated at the scale of a single volcano at Mt. Shasta, where
the pre-eruptive water contents of magmas range from <1 to possibly as high as 8-10 wt%
based on melt inclusion studies and experimental petrology (Grove et al. 2002, 2005).
As deduced from melt inclusion data reported in the literature, the H2O contents of
subduction-related basaltic magmas, prior to extensive degassing, is between <1 and 5.2 wt%,
after data filtering (excluding MgO < 5 wt%). It includes boninites (1.7 wt% on average) and
island arc tholeiites (1.6 wt% on average; Sobolev and Chaussidon 1996).
Pressures of crystallization and magma degassing
Estimation of vapor saturation pressures. Because of the low solubility of CO2 in
basaltic magmas (e.g., Dixon et al. 1995), most of them are saturated with CO2-rich vapor at
high pressure. This is in agreement with the larger CO2 concentrations in the magma sources
as deduced from volcanic and hydrothermal CO2 emissions and fluxes compared to estimates
based on CO2 dissolved in melt inclusions (e.g., Fisher and Marty 2005; Wallace 2005).
Accordingly, most olivine-hosted melt inclusions are probably trapped under vapor-saturated
conditions. Evidence of vapor saturation can come from the occurrence of heterogeneous
melt/gas trapping in olivines, the presence of CO2-vapor bubbles and the inability of melt
inclusions to be experimentally homogenized. Assuming vapor saturation and limited loss
of CO2 to shrinkage bubbles, the total vapor (or fluid) pressure (PCO2+PH2O) at the time of
melt entrapment and, therefore, the corresponding depth of magma ascent and degassing, can
be assessed from the CO2 and H2O contents dissolved in melt inclusions. These pressures
are commonly calculated using the program VolatileCalc (Newman and Lowenstern 2002).
However, the vapor saturation pressures calculated for H2O-rich basalts can be overestimated
380 Métrich & Wallace
by using this program at higher pressures because high H2O content strongly enhances CO2
solubility in basaltic melt by stabilizing carbonate species at pressure ≥400 MPa (Botcharnikov
et al. 2005; Moore 2008). In fact, only a few pressure estimates for melt inclusions apparently
reach 500 MPa (as reported by Roggensack et al. 1997; Walker et al. 2003 for Central America;
Cervantes and Wallace 2003a and Vigouroux et al. 2008, for Mexico). Instead, most of the
vapor saturation pressures for basaltic melt entrapment in olivine range from 300-400 to a few
MPa, at which pressures the effect of water on CO2 solubility would be less significant. An
evaluation of the accuracy of VolatileCalc and the method of Papale et al. (2006) is given by
Moore (2008).
Open-system vs. closed-system degassing at basaltic volcanoes. Models for degassing
of magmas saturated with H2O+CO2 vapor predict that the CO2/H2O ratio in the vapor phase
is higher than that of the coexisting melt, and this ratio in the melt decreases rapidly during
degassing (Newman et al. 2000).
Under open-system conditions, when the vapor is extracted from the melt (Rayleigh
distillation), the CO2 concentration in the melt decreases with negligible loss of H2O until
the melt reaches the vapor saturation pressure for pure H2O. An example of open-system
degassing is illustrated by Mariana Trough samples (Newman et al. 2000). Basaltic melt
inclusions hosted in olivine contain up to 0.09 wt% CO2 and their host glasses have as little as
0.02 wt% CO2 whereas the water contents of inclusions and host glasses are similar (2.23±0.7
vs. 2.14±0.13 wt%) as are their chemical compositions.
However, closed-system degassing—where the melt remains in equilibrium with its coex-
isting gas phase (Dixon et al. 1995)—is suggested by melt inclusion data from several volca-
noes. In such a situation, exsolution of H2O into the gas phase is more significant and ultimately
induces magma crystallization, as illustrated by the following examples. An example of H2O
loss and crystallization during magma ascent under closed-system conditions is given by the
basaltic-andesite scoria from the 1723 eruption of Irazù in Costa Rica (Benjamin et al. 2007). In
that case, the H2O-CO2 degassing path is calculated under closed-system conditions assuming
that the magma was coexisting with 2 wt% (~10 vol%) exsolved gas at ~150 MPa (Fig. 10A).
As the pressure decreased down to <20 MPa, and the temperature decreased by 30 °C, the oliv-
ines became richer in FeO and their melt inclusions in K2O. The overall story of crystallization
and magma degassing seems rather simple, although the olivine compositional range (Fo87−79)
is quite large and brackets the equilibrium olivine composition calculated for the host lava
(Fo83). All olivines have normal zoning and rims of Fo78. The most primitive melt inclusions
trapped in olivine Fo87 record the highest CO2 (445 ppm) and H2O (3.1 wt%) concentrations
and thus the highest vapor saturation pressure of melt entrapment (125-150 MPa).
Samples from prehistoric activity of Arenal in Costa Rica also illustrate coupled magma
degassing and polybaric crystal fractionation (Wade et al. 2006). The most mafic inclusions
have a high alumina basaltic composition considered to be the closest to parental Arenal
magmas. They show the highest H2O (3.9 wt%), CO2 (<0.03 wt%) and sulfur (0.3 wt%)
contents and the highest entrapment pressures (~200 MPa). Decreasing H2O contents (down to
1.1 wt%) correlate broadly with host olivine composition (Fo79−76) and extent of crystallization,
as indicated by increasing SiO2 and K2O in the melt inclusions. Overall, they record mafic
magma ascent and degassing from 200 to 20 MPa, under apparent closed-system conditions
that were simulated with 1 wt% exsolved gas and crystal fractionation.
Despite the good fit of the Irazu and Arenal melt inclusion data to closed-system degassing
models, we note that the required initial exsolved gas contents of 2 and 1 wt%, respectively,
require either a much higher initial CO2 content for the magma (on the order of 1 to 2 wt%
CO2) or some open-system process of gas enrichment. In the following section, we discuss the
growing body of evidence for open-system addition of gas to ascending magmas.
500 3500
Irazu-MI-Ol [1] Oct. 2002 - 1st phase 0.70
M.I., Emb. B 0.65
300
Irazu-MI-cpx [1] Fo87 A 400 MPa
Nov./Dec. 2002 - 2nd phase
Arenal-MI-Ol [2]
M.I., Emb.
200
400 Ar model parent [2] 3000
15
300 MPa
0
Closed-system
degassing
10
0
300 CSD 2% CSD 1%
2500
0.55
50
Fo85-87 Fo76-79 2000
200
CO2 ppm
200 MPa
20 Fo80-81 Fo76-77 1500
CO2 ppm
100
Fo79 CO2-flushed magma
ponding zone
Fo73
1000
0
0 1 2 3 4 5 100 MPa
500
H 2O wt% 1.5% Open
Matrix
0%
system
Figure 10. Diagram showing the variation of the CO2 and H2O 0
concentrations in melt inclusions in (A) olivine and clinopyroxene 0 1 2 3 4
from, respectively, the 1763 and 1963-5 eruptions of Irazù volcano
([1] Benjamin et al. 2007) and olivine from prehistoric tephra layers
H2O wt %
of Arenal ([2] Wade et al. 2006). The variations observed in the
diagrams are by far larger than the analytical error (see [1] and [2] for details on error and accuracy of SIMS analyses). For each sample set the composition of the host
olivine is indicated as Fo mol%. The closed-system degassing curves were calculated using VolatileCalc (Newman and Lowenstern 2002), assuming starting conditions
of 49 wt% SiO2, 1100 °C and 2 wt% exsolved vapor for Irazù samples (Benjamin et al. 2007) and 1 wt% exsolved vapor phase for Arenal samples (Wade et al. 2006); (B)
from the 2002 flank eruption of Mt Etna (Spilliaert et al. 2006a). Degassing path of the basaltic melt containing 3.4 ± 0.2 wt % dissolved H2O at 400 MPa is calculated in
either open system (OSD) or closed system (CSD) with 0 and 1.5 wt % initial gas phase. Isobars, degassing trends and curves of equilibrium between the melt and a vapor
phase of fixed composition (isopleths) are calculated using VolatileCalc (Newman and Lowenstern 2002). Data points for the first stage, in particular samples emitted
Figduring
10B
lava fountains at the onset of the eruption (black squares), fit with closed system magma degassing with 1.5 wt % initial gas phase. Melt inclusions having a large bubble
Volatile Abundances & Degassing Paths Tracked by Melt Inclusions
(dashed contour domain) suffered CO2-diffusion from the melt towards the bubble during cooling. Most of the erupted magma is represented by the cluster of inclusions
(second phase) trapped at between 220 and 170 MPa that depart from the CSD trend. Variable magma dehydration was regarded as the possible result of deep CO2 flushing
of magma during its ponding at an average pressure of 200 MPa (~ 8 km depth). This latter process can also account for a water depleted inclusions trapped at pressures of
300 MPa. Abbreviations are: M.I. for melt inclusions and Emb. for embayments.
381
382 Métrich & Wallace
Deviation from CO2-H2O equilibrium degassing trends. Melt inclusion H2O and CO2
data from some volcanoes show variations that cannot be explained by either open- or closed-
system degassing. Such features are exemplified by melt inclusions hosted in olivine (Fo80−82)
from scoriae emitted during explosive activity of the 2002 flank eruption at Mt Etna (Spilliaert
et al. 2006a). Most of the melt inclusions were trapped at pressures ranging from ~400 to 200
MPa (Fig. 10B). In a first population of melt inclusions, the CO2 and H2O variations follow a
closed-system degassing path that was described above as being the result of decompression
of the primitive basalt already coexisting with 1.5 wt% vapor phase at ~400 MPa. Closed-
system ascent of this basaltic magma gave rise to lava fountains at the onset of the eruption. In
contrast, melt inclusions from samples representative of the largest volume of trachybasaltic
magma delivered during the second phase of the 2002 eruption were trapped in olivine within
a narrow pressure range from 220 to 170 MPa. The inclusions delineate a “water-depletion”
trend where H2O decreases (from 2.9 to 2.2 wt%) while CO2 weakly varies (1140-1050 ppm).
The range of H2O concentrations is larger than the analytical error (6-8% relative). This trend
deviates from that of closed-system degassing unless the vapor amount reaches 15 to 20
wt%, which is physically implausible. A process of mixing between the H2O-rich primitive
ascending melt that was extruded previously, at the onset of the eruption, and already degassed
magma was rejected because of their high sulfur (0.29-0.32 wt%) and Cl (0.15-0.16 wt%).
The selective water-depletion in Etna basalt at high pressure was interpreted as resulting from
the re-equilibration of the magma with a deeper-derived CO2-rich gas phase, during ponding
and crystallization of magma at ~200±20 MPa (equivalent to 8 km depth). In order to buffer
a gas phase with 65 mol% of CO2 in equilibrium with the melt inclusions, either 5 wt% of
pure CO2 or ≥8 wt% of an H2O-CO2 mixture with ≤0.85 mol% CO2 would have to have
been added to the ponding magma (Spilliaert et al. 2006a). Continuous CO2 flushing of Etna’s
plumbing system is consistent with the high CO2 content (~1.5 wt%) inferred in primary Etna
magma (e.g., Allard et al. 1997). It was proposed that early CO2-melt separation may thus be
a common process at Mt Etna, except in the case of very fast magma ascent, which allows
closed-system rise of bubble-melt mixtures.
Another example where fluxing of CO2-rich gas appears to have played an important role
is the historic eruption (1759-1774) of Jorullo volcano in central Mexico (Johnson et al. 2008).
Jorullo melt inclusions trapped some of the most volatile-rich (≤5.3 wt% H2O, ≤0.1 wt% CO2),
primitive (≤10.5 wt% MgO; olivine up to Fo90.6) melts yet measured in an arc setting, as well
as more degassed, evolved compositions. About half of the analyzed melt inclusions deviate
from either open- or closed-system degassing trends, but these inclusions span a wider range
of pressures (150-350 MPa) than in the case of Etna. These differences are consistent with
differences in the plumbing systems. Jorullo was a short-lived, monogenetic volcano, in which
olivine crystallized over a wide range of depths from the middle to upper crust, probably in a
network of dikes and sills, rather than mostly crystallizing in a single, larger magma ponding
zone as occurs at Etna at ~8 km depth.
Deviation from calculated closed-system degassing trends was also reported at Mt
Vesuvius, where melt inclusions show H2O depletion coupled with limited variations in melt
CO2 contents (Marianelli et al. 2005). Note that the compositional variation of melt inclusions
also affects the degassing trend because the bulk composition can have a strong effect on CO2
solubility (Moore 2008). The evolution of CO2 vs. H2O in Arenal and Fuego samples reported
in Wade et al. (2006), and shown in Figure 10A, could also indicate some deviation - larger than
the analytical error - from closed-system conditions.
Gas fluxing related to deeply-derived magmatic CO2 could be a common process at basaltic
volcanoes. However, further evaluation of this hypothesis requires large systematic data sets on
melt inclusions from individual volcanic systems and assessment of the effects of disequilibrium
degassing (e.g., Gonnermann and Manga 2005) in such highly dynamic basaltic systems.
Volatile Abundances & Degassing Paths Tracked by Melt Inclusions 383
Gas fluxing, H2O loss, and crystallization. We described previously how closed-system
degassing can cause H2O loss from melt and thereby result in crystallization during ascent.
The process of gas fluxing involving CO2-rich gas also causes H2O loss from melt and can
therefore be effective in driving crystallization.
At Jorullo volcano, Mexico, the high MgO content of the early erupted magmas combined
with initially high H2O and the effects of gas fluxing caused olivine-only crystallization over
a wide range of depths (10-400 MPa), as indicated by melt inclusion data (Johnson et al.
2008). This effect can be seen by examining phase relations for both H2O-saturated and H2O-
undersaturated conditions (Fig. 11). Melt inclusion data demonstrate that olivine crystals
formed at a maximum pressure of 400 MPa under H2O-undersaturated conditions, although
the magma was vapor saturated because of the presence of CO2. Fluxing of CO2-rich gas
derived from deeper within the plumbing system would remove H2O from the melt, even
though the melts were H2O undersaturated. Thus the melts would have been below their
relevant H2O undersaturated liquidii (Fig. 11), forcing small quantities of olivine to crystallize
during ascent. Larger amounts of olivine would have crystallized once the melts crossed the
H2O-saturated liquidus (at ~150-200 MPa). Increases in melt inclusion K2O with decreasing
pressure and host olivine Fo content for early erupted Jorullo melt inclusions are consistent
with ≤14% crystallization, which matches the prediction based on the phase diagram. Later
in the eruption, low pressure degassing caused extensive shallow crystallization of olivine,
plagioclase, and minor clinopyroxene. The K2O increases in late erupted melt inclusions
require up to 29-36% crystallization before melt inclusion entrapment, again consistent with
phase equilibria.
450
olivine
H2O saturated H2O undersaturated
400
cpx
amph
350
Effect of
CO2 fluxing
5 wt% H O
300 plag
Pressure (MPa)
2
4 wt% H O
3 wt% H O
2
250
200
adiabatic
150 ascent
path
100
50
0
900 950 1000 1050 1100 1150 1200 1250 1300
Temperature (°C)
Figure 11. Calculated phase diagram for early Jorullo melt composition (11 wt% MgO). Circle with cross
indicates hypothetical starting melt based on the highest melt inclusion entrapment pressure of 400 MPa
with 5.3 wt% H2O. The adiabatic ascent path includes the effects of crystallization, gas exsolution, and gas
expansion. See Johnson et al. (2008) for a complete description of how the phase diagram was calculated.
[Figure used by permission of Elsevier Limited from Johnson et al. (2008), Earth Planet Sci Lett, Vol. 236,
Fig. 8, p. 483.]
384 Métrich & Wallace
Whether H2O loss occurs by open-system gas fluxing or closed-system degassing, in
either case it causes crystallization of magma during ascent or storage. As shown by some of
the examples given above, degassing-induced crystallization during basaltic magma ascent is
a common process. However, recognizing the effects of degassing-induced crystallization can
be complicated by several additional processes. These include: (i) magma mixing involving
degassed and undegassed magmas, as exemplified by Popocatépetl and Colima (Atlas et al.
2006); (ii) hybridization (mingling) in which inclusions and their host crystals are derived
from different magmas distributed vertically in the crust and thus record different entrapment
pressures (e.g., from 100 to 440 MPa in the case of the 1974 vulcanian eruption at Fuego,
Guatemala; Roggensack 2001); and (iii) assimilation at upper crustal levels as reported at
Paricutin (Mexico; Luhr 2001) and Jorullo (Johnson et al. 2008).
Sulfur and halogen degassing
Bulk degassing rates and exsolution pressures. The degassed fraction of a volatile species
can be derived from the difference between the volatile content of undegassed melt inclusions
and that of the degassed residual melt corrected for crystallization (Sisson and Layne 1993).
As with H2O and CO2, the initial dissolved S in a magma is almost totally lost prior to or
upon eruption. Residual matrix glasses commonly contain <50 to ~150 ppm S (e.g., Luhr 2001;
Métrich et al. 2001; Cervantes and Wallace 2003b; Witter et al. 2004, 2005; Spilliaert et al.
2006b), except in the case of products emitted during highly explosive subaerial and submarine
eruptions. Progressive exsolution of S from melt into the gas phase during decompression has
been widely described, and S generally decreases with decreasing H2O, as shown for Central
American samples (Fig. 12A). Sulfur starts degassing at rather high pressure in oxidized,
H2O-rich magmas in which S is dominantly dissolved as sulfate, as calculated from electron
microprobe S Kα wavelength shifts (Wade et al. 2006; Benjamin et al. 2007). Pressure of 140
MPa is reported for efficient degassing of S that is prevalently dissolved as sulfate in the H2O-
rich Etna basalt (Spilliaert et al. 2006b). In that case, ~80% of the initial S is lost in a pressure
5000 5000
Irazu-MI-Ol [1] Etna [3]
4000
Irazu-MI-Cpx [1]
Arenal-MI-Ol [2]
~ 150 MPa
A 4000 140 MPa
B
AR model parent [2]
S ppm
3000 3000
2000 2000
ed
at
tur
Sulfide saturation
sa
1000 1000
e
lfi d
Su
0 0
0 1 2 3 4 5 0 1 2 3 4
H2O wt% H2O wt%
Figure 12. Variation of H2O and S concentrations in melts inclusions from (a) Irazù samples ([1] Benjamin
et al. 2007) and Arenal ([2] Wade et al. (2006). The authors calculated the S degassing paths based on
the approach in Sisson and Layne (1993), using the K2O-H2O relationship to calculate the proportion of
H2O in the separating assemblage of crystals and bubbles and assuming no fractionation of K2O. Sulfur
degassing path was calculated from water mass balance and Dvap/melt = 110 and 70 for S in, respectively,
Irazù (Benjamin et al. 2007) and Arenal (Wade et al. 2006) samples; (B) from Etna olivine hosted melt
inclusions from samples of the 2002 flank eruption (filled circle) and of three lava fountains in 2000 (open
square). The pressure of S degassing is assessed from the CO2 and H2O content of melt inclusions. In a
CSD process, the S decrease is fitted with DS=αD*S values from 0 to 60, where D*S is the instantaneous
gas–melt partition coefficient and α the mass fraction of total vapor that is accumulated at each step of
degassing upon ascent ([3] Spilliaert et al. 2006b).
Volatile Abundances & Degassing Paths Tracked by Melt Inclusions 385
range between 140 and 10 MPa. This range of pressures is far higher than the low pressure of
S exsolution (≤10) MPa reported for submarine basalts (e.g., Dixon et al. 1991).
The evolution of S and H2O during decompression depends on magma ascent rate and
degassing conditions (open vs. closed system). Changes in redox state towards more reduced
conditions can drive the melt to sulfide saturation during late stage crystallization at shallow
depth, and this affects the S degassing path. Faster magma ascent inhibits these processes until
low pressure, causing S degassing to be controlled by vapor-melt partitioning. In cogenetic
basaltic and trachybasaltic magmas at Etna, the S content of sulfide-saturated melt inclusions
trapped in Fo<75 olivine representative of the shallow, partly degassed magma are only half
that of sulfide-free inclusions in Fo~80 olivine that crystallized during fast ascent of basaltic
magma under dominantly closed-system conditions (Fig. 12B). In contrast to the differences
in S content, the H2O contents of the different inclusions are comparable (~1.5 wt%).
The amount of initial Cl lost by degassing is in the range of 10-50% (e.g., Hawaii – Gerlach
and Graeber 1985; Etna – Spillaert et al. 2006b; Laki – Thordarson et al. 1996; Stromboli –
Métrich et al. 2001). Fluorine is much less affected by degassing, with typically <15% of the
initial amount lost (references as above). The lower extent of Cl and F degassing from magma
causes their concentrations to increase in melt inclusions with increasing differentiation, in
contrast to S, in agreement with the higher solubilities of Cl and F in mafic melts (Webster
2004). As an example, Cl positively correlates with U and F in Arenal melt inclusions (Wade et
al. 2006). Chlorine has been found to significantly degas at rather low pressure (<20-10 MPa)
at Etna even though it starts to be slightly exsolved into the gas phase at pressure ≤100 MPa
(Spilliaert et al. 2006b). Moreover, the rather large range of Cl concentrations demonstrates
that Cl degassing strongly depends on the dynamics of magma ascent and extrusion and
therefore on kinetic effects, in the absence of Cl-bearing mineral phases. Kinetic effects have
been demonstrated during decompression experiments on H2O-rich rhyolitic melts where the
amount of Cl retained in matrix glass depends on the size of erupted pumice clasts (Gardner
et al. 2006). Similar effects may occur in basaltic melts in which Cl diffusion would be faster
because the diffusion in H2O-rich basaltic melts is significantly higher than that of hydrous
rhyolite (Aletti et al. 2007). Actually, Cl loss is enhanced in highly fragmented clasts from lava
fountaining activity (Spilliaert et al. 2006b).
eruPtIOn StyLeS and PrOCeSSeS –
hOW dO they reLate tO MeLt InCLuSIOn data?
Rapid ascent of volatile-rich magma under closed-system conditions results in highly
explosive eruptions (possibly plinian) because there is no time for gas to escape (Cashman
2004). Large plinian eruptions of silicic magmas are common, whereas basaltic magmas
generally give rise to less explosive eruptions because of their lower viscosity, although plinian
basaltic eruptions, typical of a dispersed flow, can occur (Vergniolle and Mangan 2000). The
explosivity of basaltic eruptions, the transition between explosive and effusive style or the
transition from lava fountaining to strombolian activity depends on the regime of gas bubble
transfer and the magma supply and ascent rates (e.g., Pioli et al. 2008). Combining the
compositions of melt inclusions and their host crystals, dissolved volatile contents, and host
magma compositions sheds light on the processes that control magma eruptive style.
A rare example of a sub-plinian eruption involving deeply derived picritic magma has
been reported at Mt Etna (Coltelli et al. 2005). The olivine (Fo90-91) of the explosive products
contain melt inclusions rich in MgO (12.6 wt% on average) that are very similar in major
(including FeO) and trace elements to the bulk rocks after correction for post-entrapment
crystallization of the host olivine on the inclusion walls (Kamenetsky et al. 2007). The H2O
386 Métrich & Wallace
and CO2 contents of the melt inclusions average 3.2±0.3 wt% and 0.29±0.2 wt%, respectively,
excluding a few values affected by CO2 loss. The inclusion entrapment pressures—estimated
to be >400 MPa—are most likely underestimated because the entrapped glass coexists with
gaseous CO2 and carbonates present in the vapor bubble. The high explosivity of the eruption
has been attributed to the rapid ascent of H2O and CO2-rich picritic magma under dominantly
closed-system degassing conditions (Coltelli et al. 2005).
More commonly, basaltic volcanoes are characterized by vigorous lava fountains and/or
strombolian and violent strombolian activity. The Kilauea Iki (Hawaii) eruption of 1959 is a
well studied example of lava fountaining, but it is somewhat unusual in that it involved picritic
magma (Anderson and Brown 1993; Anderson 1995). In this example, the olivine (Fo87-88)
in equilibrium with the picritic magma mostly crystallized at relatively low pressure (<~100
MPa), as inferred from their melt inclusions. These data have been interpreted as resulting
from crystallization of a newly intruded, hot, MgO-rich magma that rose through and mixed
with cooler degassed magma stored at shallow depths (Anderson and Brown 1993). It has
been proposed that gassy picritic magma (15 wt% MgO in melt + 0.3 wt% total of exsolved
and dissolved CO2) becomes buoyant relative to the stored, degassed tholeiitic magma only at
pressures less than 200 MPa (Anderson 1995).
More typical, differentiated Kilauea magma erupted at Pu’u O’o along the east rift zone
giving rise to high lava fountains (up to 400 m high, Parfitt 2004) and less powerful spattering
activity. Melt inclusions in the former case have preserved relatively high H2O contents (0.4-
0.5 wt%), whereas inclusions from the spattering activity contain lower values of 0.1-0.3 wt%
H2O. These lower H2O values require degassing at pressures significantly lower than those
in Kilauea’s summit magma reservoir (Wallace and Anderson 1998). The low H2O values
have been interpreted as indicating that the magma was already degassed prior to olivine
crystallization and melt inclusion entrapment.
Closed-system ascent and degassing of Etna basaltic magma was inferred for the 100-300
m high lava fountaining that occurred at the onset of the 2002 flank eruption of Mt Etna, on
the basis of the CO2 and H2O degassing path recorded in melt inclusions (Fig. 10B). In this
case undegassed, primitive melt inclusions are representative of the erupted basaltic magma
for major elements and trace element patterns (Spilliaert et al. 2006a; Métrich et al. 2007).
This interpretation is consistent with the hypothesis that lava fountaining events are controlled
by magma speed rise, such that rapid ascent does not allow time for gas transfer or bubble loss
from rising magmas (e.g., Wilson and Head 1981; Parfitt and Wilson 1995).
Similarly, fast ascent of deep-seated, volatile-rich magma under closed-system conditions
gives rise to the most explosive activity and the emission of basaltic pumice at Stromboli
(Aeolian islands, Italy). These products contain MgO-rich olivines with undegassed melt
inclusions (on average: H2O = 2.8±0.2 wt%; S = 0.22±0.02 wt%; CO2 = 0.1-0.2 wt%) that were
trapped at pressures of 200-300 MPa (Métrich et al. 2001; Bertagnini et al. 2003). In contrast,
gas jets and mild explosions lasting a few seconds propel lava lumps and dense scoriae during
the rythmic, strombolian activity at the craters. These scoriae carry low pressure crystals with
melt inclusions (<1.2 wt% H2O, <0.01 wt% S, low CO2) that mirror the extensively degassed
and viscous magma stored in the uppermost parts of the conduits, capping the volcanic system
(Landi et al. 2004).
Mafic cinder cones commonly exhibit strongly pulsatory explosive activity described
as violent strombolian that produces both moderately high eruption columns (2-6 km) with
abundant fine ash and simultaneous lava effusion from lateral flank vents (Pioli et al. 2008).
Examples of such activity are the historic eruptions of Paricutin and Jorullo in Mexico, where
melt inclusion data show that basaltic to basaltic andesite magmas had initial H2O contents of
3.5-5.3 wt% and CO2 ≤ 1000 ppm (Luhr 2001; Johnson et al. 2008; Pioli et al. 2008).
Volatile Abundances & Degassing Paths Tracked by Melt Inclusions 387
VOLCanIC deGaSSInG: InteGratIOn OF VOLCanIC GaS
COMPOSItIOnS and GaS FLuxeS WIth MeLt InCLuSIOnS
Syn-eruptive magma degassing and volatile degassing budgets
The mass of volatiles released by syn-eruptive magma degassing can be estimated from
the initial dissolved volatile contents recorded by undegassed melt inclusions, the residual
volatile contents in bulk rock or matrix glass samples, and the volume of magma erupted.
This approach, known as the “petrologic method”, has been widely applied and debated
for assessing the S emissions from volcanic eruptions because S (primarily as SO2 and H2S
species) injected into the stratosphere is converted into H2SO4 aerosols that have an effect on
climate and the Earth’s atmosphere (see Robock and Oppenheimer 2003 for a review). The
petrological estimates of SO2 degassing have been compared with independent ground-based,
airborne and satellite measurements of SO2 flux at active volcanoes (see Wallace 2001 and Self
et al. 2006 for reviews).
The mass of SO2 (MSO2) released from a volcanic eruption is estimated using the petrologic
method according to the relation: MSO2 = 2 Mmagma (CMI−CWR), where Mmagma is the mass of
bulk magma derived from the volume of erupted products corrected for their vesicularity,
CMI is the S concentration in undegassed melt inclusions representative of the bulk erupted
magma, CWR is the S concentration in bulk rock, or in matrix glass that is corrected for the
extent of crystallization. The factor of 2 is included because the mass of SO2 is two times that
of S (SO2/S mass ratio = 2).
Large basaltic eruptions can deliver huge masses of volcanic gas into the atmosphere
during the emission of voluminous lava flows, along 10 to 100 km long fissures, at high effusion
rates, with occasionally violent lava fountain activity (Self et al. 2006). The 8 month long
1783-1784 Laki flood basalt eruption in Iceland released 14.7 km3 of quartz tholeiitic magma
and is a well studied example of syn-eruptive degassing. The large volume and homogeneous
composition of Laki products imply that they were erupted from a compositionally uniform
magma body in a deep-seated (>10 km) reservoir (Thordarson et al. 1996). The major and
trace element compositions of melt inclusions from the eruptive products are very similar to
those of the bulk erupted magma, but the recent discovery of heterogeneous oxygen isotope
values for individual olivine crystals from this eruption has called into question the origin of
the crystals (Bindeman 2008). The eruption released 122 Mt of SO2, 15 Mt of Cl and 7 Mt
of F as assessed from petrologic estimates (Thordarson et al. 1996). These authors concluded
that the estimates for the Laki eruption were reliable because similar values for H2SO4 (~200
Mt) were provided by independent estimates. By integrating data on the melt inclusions, lavas,
tephra and deposits, it is shown that 80% of the S mass was released during the explosive
eruptive phases at the vents and carried by eruption columns, which reached heights of 9-13
km, to the lower stratosphere (Thordarson and Self 2003). Less than 20% of the sulfur mass
was released from the lava to form localized haze within the lower troposphere.
Similarly, good agreement has been obtained between the petrologic estimates of SO2
released during the 94-day 2002 flank eruption at Etna and the total SO2 emissions derived from
optical correlation spectrometer (COSPEC) measurements during the eruption (Spilliaert et al.
2006a), considering the uncertainties in the erupted product volumes (±50%) and COSPEC
data (±20-25%). This agreement implies that there was no excess release of SO2 during the
eruption; the SO2 emissions were essentially supplied by syn-eruptive bulk degassing of the
extruded magma. Good agreement between petrologic and spectroscopic estimates of SO2
output has also been demonstrated at Arenal by considering melt inclusions that are most
representative of undegassed basaltic andesite magma (Wade et al. 2007).
The consistency between petrologic and other independent estimates of SO2 output at some
basaltic volcanoes demonstrates that the petrologic method is useful when the only source of
388 Métrich & Wallace
S comes from syn-eruptive magma degassing. As shown by the example of Laki above and
also by cases discussed in Wallace (2001), this is most commonly the case in non-subduction
related volcanoes. Petrologic estimates of SO2 emissions from ancient flood basalts are likely
to provide realistic values (Sharma et al. 2004).
excessive degassing and the role of degassed but unerupted magma
The S contents of undegassed melt inclusions have been widely used to assess the mass and
volume of degassed magma that is necessary to sustain measured SO2 fluxes during quiescent
degassing (only gas emissions) at active basaltic volcanoes. This assessment is based on the
relationship: Mmagma = ΦSO2/2xS, where M is the mass of degassed magma, ΦSO2 is the time-
averaged flux of SO2, and xS is the mass fraction of S degassed from magma based on melt in-
clusions and bulk rock or matrix glass samples, as described above for the petrologic method.
As an example, only a small fraction (<10%) of the degassed magma is actually erupted
at Stromboli (Allard et al. 2008). This interpretation is derived from the S concentrations (0.22
wt%) in high pressure melt inclusions trapped in Fo88−89 olivine (Bertagnini et al. 2003), the
time averaged flux of SO2 (250 tons d−1), and the bulk magma extrusion rate (0.018 m3 s−1)
over the last 30 years.
quiescent and intrusive degassing—where the amount of SO2 released is in excess com-
pared to estimates based on syn-eruptive degassing of erupted magma—is commonly reported
at open conduit basaltic volcanoes such as Etna (Allard 1997), Villarica (Witter et al. 2004),
and Popocatepetl (Witter et al. 2005), amongst others.
differential gas bubble transfer
Because gas bubbles can segregate from the melt at different depths, the volatile compo-
nents fractionate according to their solubilities. The composition of the exsolved gas phase will
thus change as a function of the pressure-related solubility of each volatile species. Knowing
the abundance of S, Cl, F, H2O and CO2 dissolved in melt inclusions and the pressure interval
over which degassing has occurred, as derived from the H2O and CO2 content, it is possible to
calculate the pressure-related evolution of S and halogens and the S/Cl and S/F ratios in the
gas phase as the pressure decreases.
At Mt Etna, the S/Cl mole ratio (1.09 × weight ratio) in the gas phase has been estimated
from melt inclusion data as function of pressure, assuming closed- and open-system degassing,
with and without sulfide immiscibility (Spilliaert et al. 2006b; Fig. 13). The S/Cl ratio of the
gas phase is controlled by S outgassing until about 10-20 MPa and becomes influenced by Cl
exsolution only at lower pressures, when the melt is already depleted in S. Pure open-system
degassing fails to explain the S/Cl ratios (≤0.2) commonly measured in gas emissions during
Etna eruptions. Under closed-system conditions, the S/Cl ratio of the gas phase evolves more
progressively down to 3.7 (molar) in gas emitted at the surface. Any gas bubbles segregated
from the melt at depth will have a high S/Cl ratio. As an example, gas bubbles segregated from
~1.8 km have a high S/Cl value of 10 (Fig. 13), as measured in the gas emission during one of
the cyclic lava fountains on June 2000 at Etna crater (Allard et al. 2005). The lava fountaining
has been interpreted—on the basis of this high S/Cl ratio in gas emissions—as deriving from
bubble accumulation prior to eruption (Allard et al. 2005) according to the bubble foam model
(Vergniolle and Jaupart 1990).
GeOCheMICaL SySteMatICS OF the
PrIMary VOLatILe COntentS OF MaFIC MaGMaS
One of the most important conclusions that can be drawn from the data reviewed in this
chapter is that most of the observed phenocrysts in basaltic magmas crystallize during ascent,
Volatile Abundances & Degassing Paths Tracked by Melt Inclusions 389
5 15 30 50 80 105 140 MPa
30 10
~1.8 km 9
25
8
S/Cl - gas (molar)
20 7
15 6
e
lfid
5
Su
10 4
S/Cl melt (wt)
5 3.7 3
0.2 2
0
1
-5 0
0.45 0.55 0.65 0.75 0.85
Melt fraction (f)
Figure 13. Pressure-related evolution of the S/Cl ratio in the melt and the gas phase of Mt Etna, under
closed-system conditions (curves 1 and 2) and open-system conditions (curve 3), modified from Spilliaert et
al. (2006b). The S/Cl ratio in the gas phase is calculated following the relationship (xS/xCl)gas = (DS/DCl) ×
(xS/xCl)melt; where x and D are, respectively, the concentrations and the bulk partition coefficients of S and
Cl as determined from the melt inclusion data. The S/Cl ratios in melt are given in weight units, whereas
the calculated gas ratios are given in molar units in order to have direct comparison with values measured
in gas emissions. CSD curves and the final S/Cl ratio in surface gas are calculated with an average bulk Cl
loss of 55%. Pure OSD conditions produce much lower S/Cl ratio (<0.2) in surface gases than measured in
emissions of both eruption types (Spilliaert et al. 2006b). The CSD curves illustrate the significant influence
of sulfide immiscibility on the S/Cl degassing path. The lithostatic depth of ~1.8 km below vents is computed
from rock densities in the sedimentary basement of Etna. The main interpretation that gas bubbles segregated
from rather low pressure fits with a process of differential gas transfer (see text).
degassing, and shallow storage in the crust. As such, melt inclusions in crystals preserve a
record of these processes, but the inclusions represent melts that had variably degassed prior
to entrapment. This variable degassing hinders our ability to determine the primary volatile
contents of basaltic melts formed by partial melting of the Earth’s upper mantle. The extent
to which this variable degassing hides the primary volatile contents is inversely proportional
to the solubility of the component in question. Thus for low solubility components like CO2
(and He), melt inclusions formed in the middle to upper crust may never record values that are
close to primary values (Wallace 2005). The major volatile components that degas the least are
F and Cl, and they are least affected by this problem. H2O and S lie between the two extremes
and show variable degrees of degassing in melt inclusion data sets.
A standard approach has been to analyze as large a number of inclusions as is practical and
to then assume that the highest pressure inclusions are best representative of the undegassed
magma. In the case of H2O, an independent check on this can be made using experimental
phase equilibria (e.g., Di Carlo et al. 2006; Johnson et al. 2008), mineral-melt hygrometers
(Housh and Luhr 1991; Hamada and Fuji 2007), and partitioning of OH between melt and
nominally anhydrous minerals (e.g., Hauri et al. 2006). However, agreement between these
various approaches only confirms the melt H2O content during crystallization, but it does not
preclude higher primary magmatic H2O contents when melts are in the upper mantle or lower
crust. In this section, we review the systematics of the primary volatile contents of basaltic
magmas from arcs, as inferred from melt inclusion data, and compare them to what is known
from submarine glasses from mid-ocean ridges and oceanic islands, with the goal of better
understanding volatile recycling and magma formation in subduction zones.
390 Métrich & Wallace
Several recent studies have used melt inclusions to examine volatile variations either
with distance from the trench or with position along the arc where there are important along-
arc variations in subduction parameters. Variations in volatile content with distance from
the trench were studied in Guatamala by Walker et al. (2003). For arc front volcanoes, they
found maximum H2O contents of 6.2 wt% (Fuego) and 2 wt% (Pacaya). However, the Pacaya
inclusions have CO2 below detection limit, so the low H2O values are likely the result of
partial degassing before inclusion entrapment. Monogenetic basaltic cinder cones located up
to 80 km behind the volcanic front have maximum H2O contents similar to Pacaya (~2 wt%).
Overall, the results for Guatamala show no consistent across arc-variation, but the highest H2O
contents (Fuego) are found only at the arc front. Distinctly higher CO2 contents were found in
some of the monogenetic cones behind the front. Walker et al. (2003) speculated that this could
be the result of devolatilization of carbonate sediments with increasing slab depth, but given
how strongly CO2 degasses at crustal pressures, this could also be caused by greater average
crystallization depths for the back-arc cones.
Complementing this is a recent study of along-arc variations in volatiles in Central America
based on data from 11 volcanic centers in Guatamala, Nicaragua and Costa Rica (Sadofsky et
al. 2008). The thickness of the downgoing oceanic crust, composition of subducted sediments,
and slab dip angle all vary along this arc. For the northern part of the arc, the downgoing
plate is normal-thickness, 25 Ma oceanic crust, overlain by ~450 m of carbonate ooze and
hemipelagic clay, and probably underlain by strongly serpentinized mantle. In contrast, beneath
central Costa Rica the downgoing plate is relatively thick, 15-20 Ma oceanic crust formed at
the Galapagos spreading ridge and overprinted by the Galapagos hotspot. Estimated parental
magma H2O contents (for melts in equilibrium with Fo91 olivine) vary from 2.7 to 5.5 wt%,
but show no consistent variation along the arc despite the large differences in the nature of the
subducting slab (Sadofsky et al. 2008). The highest values were for Cerro Negro in Costa Rica
and Irazu in central Costa Rica. Parental magma S, Cl and F also show no systematic variations
along the arc, though Irazù volcano at the southern end of the arc contains much higher values
than the other volcanoes (Benjamin et al. 2007). Sadofsky et al. (2008) concluded that melting
beneath all parts of the Central American arc is triggered by addition of H2O-rich components
from the downgoing slab.
In an across arc study of olivine-hosted (Fo73−91) melt inclusions from the Kamchatka and
northern Kurile arcs, Portnyagin et al. (2007) calculated mean parental H2O concentrations
(melts in equilibrium with Fo90 olivine) for the different volcanoes studied ranging from
1.8 to 2.6 wt%, with values generally decreasing with increasing distance from the trench.
Interestingly, the parental H2O estimates based on the highest measured melt inclusion H2O
contents (rather than averages) at each volcano show no such pattern but are instead constant
at 2.5-2.8 wt% H2O. However, about two-thirds of the data set are for olivines from lava flows
that were experimentally rehomogenized, and the H2O values are lower than those found for
olivine-hosted melt inclusions from rapidly cooled tephra at Kluchevskoy (≤3.9 wt% H2O in
Fo88 olivine and ≤7.1 wt% H2O in Fo80 olivine; Auer et al. 2008). Mean parental S, Cl, and F
based on the melt inclusion data all show increases with increasing distance from the trench.
In a similar across arc study of olivine-hosted (Fo82−91) melt inclusions from nine cinder
cones in the central Trans-Mexican Volcanic Belt (TMVB), Johnson et al. (2008.) found parental
magma H2O contents (for melts in equilibrium with Fo90 olivine) that are high at the volcanic
front (3.5-5.2 wt% H2O), but remain high (3.0-4.5 wt%) in the cones over 130 km behind the
front (3.0-4.5 wt% H2O). The center farthest from the front, a tuff ring, has much lower water
(~1 wt%). Parental magma S contents are variable across the arc, ranging from 850-2150 ppm,
and parental Cl concentrations remain high over most of the arc (~860-1360 ppm), decreasing
to 390 ppm in the tuff ring far behind the front. The high volatile concentrations across the
central TMVB are surprising given that the subducting Cocos plate is young (~11-15 Ma at the
Volatile Abundances & Degassing Paths Tracked by Melt Inclusions 391
trench), and therefore relatively hot, and thus significant devolatilization should occur beneath
the forearc region. Low B concentrations in whole rock samples are consistent with shallow
slab dehydration (Hochstaedter et al. 1996), so the data from the TMVB show that H2O can be
strongly decoupled from highly fluid-mobile elements such as B.
To better understand the recycling of volatiles in subduction zones and their role in mantle
melting, it is necessary to compare volatile concentrations to non-volatile trace elements of
both similar and contrasting behavior. For example, H2O and Ce have similar bulk partition
coefficients during partial melting of mantle peridotite, making the H2O/Ce ratio a valuable
indicator of sources that have become enriched in H2O (e.g., Michael 1995). Some incompatible
trace elements in arc magmas, such as Y and the high field strength elements (HFSE), Ti,
zr, Nb, Ta, have similar abundances to values for depleted and enriched MORB, leading to
the interpretation that these elements are supplied primarily by the mantle wedge (Pearce
and Peate 1995). This suggests that these elements are generally not mobile in the H2O-rich
component(s) transferred from the slab to the wedge, though they can be mobile under certain
circumstances (Kessel et al. 2005; Barry et al. 2006). Comparison of HFSE and Y systematics
for arc and non-arc magmas shows that the mantle wedges beneath arcs are variable in
composition, with compositions ranging from more depleted than the depleted MORB mantle
(DMM) to values requiring a source similar to that for EMORB magmas (Pearce and Peate
1995; Wallace and Cervantes 2003; Portnyagin et al. 2007). The more depleted sources form
as a result of previous partial melting and melt extraction, and heterogeneities in the wedge can
exist even on relatively small spatial scales (e.g., Wallace and Carmichael 1999). The enriched
vs. depleted character of the mantle source is well illustrated by the Nb/Y ratio because both
these elements are relatively fluid immobile, but Nb is more highly incompatible and thus is
depleted during partial melting (Pearce and Peate 1995).
Volatiles and fluid-mobile elements (e.g., K, Ba, Sr, U, Pb) transferred from the slab to the
wedge overprint and selectively enrich the variably depleted mantle wedge (Fig. 14A). Partial
melting of DMM to form depleted MORB results in magmas with low H2O/Y (≤100; le Roux
et al. 2006) whereas partial melts from an OIB-type mantle source yield higher values (≤1000)
as well as much higher Nb/Y. Data from olivine-hosted melt inclusions from Kamchatka,
Central America, and the TMVB have H2O/Y varying from ~1000 to 5400. The mantle wedge
for Kamchatka appears to be as or more depleted than DMM (Portnyagin et al. 2007) whereas
the wedges beneath Central America and the TMVB are less depleted and more heterogeneous
(Fig. 14A). Interestingly, the data for Kamchatka suggest that the mantle wedge is less strongly
enriched in H2O than beneath Central America and the TMVB, despite the fact that the plate
subducting beneath Kamchatka is much older.
The high values of H2O/Y for basaltic arc magmas require mantle source H2O
concentrations in the range from 0.3 to nearly 1 wt%, compared to 70-160 ppm that is estimated
for DMM (Workman and Hart 2005). Grove et al. (2006) presented a model for hydrous flux
melting in arcs. H2O-rich fluids rising from the subducted slab or from breakdown of chlorite
in the mantle wedge just above the slab rise through an inverted thermal gradient in the wedge
caused by corner flow of hot asthenospheric mantle. When the fluid encounters mantle at
the temperature of the wet peridotite solidus, a small mass fraction of H2O-saturated melt
forms, with ~20-30 wt% dissolved H2O. As these melts continue to rise through the inverted
thermal gradient, they react with the peridotite, extracting basaltic components and diluting
the initially very high H2O contents. By the time the melts reach the hottest part of the wedge,
at temperatures ~1400 °C, the partial melts contain ~5 wt% H2O, similar to the highest values
for primitive arc magmas.
Some arc basaltic magmas are relatively H2O-poor. In the TMVB, the H2O-poor types
have high Nb/Y (Fig. 14A; Cervantes and Wallace 2003(a or b)), and their isotopic and trace
element characteristics are similar to basalts erupted in the Mexican Basin and Range province
392 Métrich & Wallace
5000 Kamchatka
Cent. Am.
4000 Arenal
5% Mexico
Mex High K
H2O/Y
3000 DMM + 0.5% H2O
MORB
DMM
2000 20% DMM 0.5%
OIB
1000 Stromboli
OIB
DMM
5%
Etna
20% 5% 20%
0
0.01 0.1 1 10
Nb/Y
150 Kamchatka
Cent. Am.
Arenal
Mexico
Cl/Y
100
Mex High K
5%
MORB
DMM + 100 ppm Cl
DMM
50 DMM 150
20% OIB
OIB 5% Stromboli
DMM
20% 5% 20%
0
0.01 0.1 1 10
Nb/Y
250
200 Kamchatka
Cent. Am.
S/Y
150 Arenal
Mexico
Mex High K
100
Stromboli
MORB
50
0
0.01 0.1 1 10
Nb/Y
Figure 14. Relationship between major volatiles and “conservative” trace elements, which are not likely
to be mobile in fluids released from subducted slabs (Pearce and Peate 1995). (A) Comparison of Nb/Y
vs. H2O/Y in arc basaltic magmas and MORB. Red line shows 5 to 20% batch partial melts of DMM
(Workman and Hart 2005) containing 110 ppm H2O calculated using partition coefficients from Kelley et
al. (2006). Black line near middle of figure shows 5 to 20% batch partial melts of DMM with 0.5 wt% H2O.
Black line in lower right shows 5 to 20% batch partial melts of an OIB source with 0.2 wt% H2O. Data are
shown for Kamchatka (Portnyagin et al. 2007), Central America (Sadofsky et al. 2008), Arenal volcano,
Costa Rica (Wade et al. 2006), central Mexico (Johnson et al. 2008), high-K minettes and basanites in
western Mexico (Vigouroux et al. 2008), Stromboli (Bertagnini et al. 2003), Etna (Kamenetsky et al. 2007),
and East Pacific Rise MORB (le Roux et al. 2006). (B) Nb/Y vs. Cl/Y. Partial melting curves as in (A),
with 0.38 ppm Cl in DMM, a fluid enriched DMM source with 100 ppm Cl, and an OIB source with 50
ppm Cl. (C) Nb/Y vs. S/Y.
Volatile Abundances & Degassing Paths Tracked by Melt Inclusions 393
well behind the arc (Luhr 1997). The H2O/Y values of these intraplate-type alkali basalts are
similar to other OIB (Fig. 14A), indicating that the mantle source region for these magmas
has not been significantly enriched in H2O and trace elements by subduction processes. In the
Cascades of Northern California, the H2O-poor types are high-alumina olivine tholeiites with
trace element characteristics more similar to MORB (Sisson and Layne 1993; Grove et al.
2002). Another well documented example of H2O-poor basaltic magma in arcs comes from
the 1982-1983 eruptions of Galunggung volcano, Indonesia, where basaltic melt inclusions
contain 0.2 to 0.3 wt% H2O (Sisson and Bronto 1998). In general, H2O-poor arc magmas are
inferred to be the result of decompression melting caused by upwelling in the mantle wedge
(Sisson and Layne 1993), probably caused by advection of hot mantle from behind the arc due
to corner flow (Grove et al. 2006).
The Cl concentrations of basaltic arc magmas are also much higher than those of depleted
MORB, EMORB, and most OIB-type magmas (Fig. 14B). Cl/Y ratios for arc magmas show
similar patterns as H2O/Y, and the Cl concentrations require ~100 to 400 ppm Cl in the mantle
source compared to 0.38 ± 0.25 ppm that is present in DMM (Workman and Hart 2005).
The wide range of Cl/H2O ratios for arc and back-arc magmas requires that Cl and H2O are
strongly fractionated from one another either during devolatilization of the slab or during
migration of slab-derived fluids through the mantle wedge (Fig. 15; Kent et al. 2002; Wallace
2005). Fluid inclusions in eclogites from exhumed subduction zone complexes can be used to
understand the compositions of fluids released during dehydration of subducted oceanic crust.
Philippot et al. (1998) used such data to suggest that fluids released from the upper oceanic
crust, which is much more strongly altered during hydrothermal alteration near spreading
ridges, have salinities of 3.1-4.0 wt% NaCl. In contrast, fluids derived from deeper in the
oceanic crust, which has been hydrothermally altered at higher temperatures than the upper
crust, are more saline, with 17-45 wt% NaCl. Thus the salinities of fluids released from the
dehydrating oceanic crust appear to bracket most values for arc and back-arc magmas, though
the data from Cerro Negro in Nicaragua appear somewhat anomalous in requiring low salinity
fluids. Fluids released during breakdown of serpentinite in the downgoing slab are estimated to
have 4-8 wt% NaCl (Scambelluri et al. 2004), also similar to many values for arc and back-arc
magmas. However, fluids derived from serpentinite dehydration in the downgoing slab have
light δ18O values, and evidence for strong involvement of such fluids has only been found in
the northern Central American arc (Eiler et al. 2005).
Olivine-hosted melt inclusions from basaltic arc magmas typically have 900-2500 ppm
S, but values extend as high as 4000-6000 ppm (Wallace 2005; Benjamin et al. 2007). These
values are higher than for MORB magmas of the same FeOT but are comparable to some
OIB (Wallace and Carmichael 1992). As with H2O and Cl, arc magmas also have higher S/Y
than MORB and OIB (Fig. 14C). The higher S/Y ratios may reflect addition of subduction-
recycled S from the slab to the wedge, oxidation of the mantle wedge by fluids such that
residual sulfide is no longer stable during melting, or large degrees of melting that exhaust
residual sulfide (Wallace 2005; Righter et al. 2008). However, the higher S contents relative
to MORB of comparable FeOT cannot be derived from low–fO2 DMM because of solubility
constraints imposed by sulfide saturation (Mavrogenes and O’Neil 1999), so breakdown of
sulfide by oxidation or large degrees of mantle melting is required. Estimated mantle source
concentrations for basaltic arc magmas are 250-500 ppm S (Métrich et al. 1999; de Hoog et al.
2001), higher than DMM estimates of 80-300 ppm S (Chaussidon et al. 1989).
COnCLudInG reMarKS
We have provided here a review of published literature showing the relevance of studies
on olivine-hosted melt inclusions for determining the volatiles dissolved in magma and their
394 Métrich & Wallace
2500
50% 15% 10%
2000 Stromb
Mexico 5%
1500
Cl (ppm)
Etna
Fuego
Kamchatka
G
1000 Guat Cerro
(BVF) Negro
500 Lau
Kilauea Mar
Central America 1% NaCl
N-MORB
0
0 1 2 3 4 5 6 7
H2O (wt.%)
Figure 15. H2O vs. Cl for olivine-hosted melt inclusions from arc basalts. Data are shown for Cerro Negro
(Roggensack et al. 1997), Fuego (Roggensack 2001), Guatamalan volcanoes behind the volcanic front
(Guat BVF; Walker et al. 2003), the Trans-Mexican Volcanic Belt (Luhr 2001; Cervantes and Wallace
2003a), Stromboli (Metrich et al. 2001; Bertagnini et al. 2003), Galunggung volcano, Indonesia (G; Sisson
and Bronto 1998), parental melt compositions from Kamchatka (Portnyagin et al. 2007), parental melt
compositions for Central America (Sadofsky et al. 2008), and undegassed Etna inclusions (Spilliaert
et al. 2006a). For Cerro Negro, only inclusions from the 1992 explosive eruption are plotted; partially
degassed inclusions from the effusive 1995 eruption are not shown. Note that a Mexican sample with
OIB-like composition has low H2O (open circle with 1.3 wt% H2O). Field for Lau Basin back-arc basalts
is from Kent et al. (2002). The average Cl/H2O ratio for Mariana Trough basalts (Stolper and Newman
1994) is shown as a dashed line and coincides with the low end of values for the Lau Basin. Lines show
basaltic melt compositions produced by addition of fluid with 1, 5, 10, 15, and 50 equivalent wt% NaCl
to an N-MORB mantle source. Most of the data have not been corrected for varying degrees of olivine
fractionation, and thus do not represent primary magma values, but such correction should only affect the
absolute concentrations and not the Cl/H2O ratios.
evolution during magma ascent and decompression. Based on this, we summarize the following
main points and raise some questions for future work:
1. Olivine-hosted melt inclusions lead to pressure estimates that only rarely exceed
300-400 MPa, and provide a powerful tool for tracking degassing between these
pressure and surface. Such a range of pressures is generally lower than those deduced
from CO2-rich fluid inclusions (Hansteen and Klügel 2008). This discrepancy raises
an important question for future research. Presently, the volatile data for melt
inclusions hosted in high pressure clinopyroxenes are very scarce, and these should
be investigated further. Regarding olivine, its crystallization in basaltic magmas
depends on temperature and the melt water content. Its growth rate is strongly
influenced by the rate of decompression and undercooling and thus the conditions
of magma ascent. Melt inclusion formation is favored by temperature and pressure
gradients and undercooling. Olivine-hosted melt inclusions from rapidly cooled
explosive products are the most suitable for avoiding extensive post-entrapment
interaction between the inclusions and their host crystals. Data from such inclusions
provide a rich record of dynamic processes. They allow us to infer the conditions of
Volatile Abundances & Degassing Paths Tracked by Melt Inclusions 395
magma ascent, crystallization, and degassing. For future work, combining textural
studies of natural and experimental olivines, their melt inclusion morphologies, and
the dissolved volatile contents of the inclusions would allow a better estimate of the
timescales of magma crystallization and ascent.
2. quantifying the H2O and CO2 contents of melt inclusions is a difficult task. The
inclusions must have suffered limited post-entrapment evolution. Their interpretation
must be tightly associated with a precise study of both the inclusions and their host
mineral composition and also that of the bulk eruptive products. The efforts dedicated
in the last 15 years to developing analytical techniques to get quantitative analyses
of both the major (CO2, H2O) and minor (S, halogens) volatile components, with
uncertainties that on average are ≤10%, have been very successful. These efforts have
led to a rapid increase in the quality and quantity of melt inclusion volatile data that
in turn have been integrated with trace element and isotope geochemistry, surface gas
emission measurements, and geophysical data relevant to understanding depths of
magma storage.
3. Because CO2 is extensively lost to the gas phase at high pressure, melt inclusions
generally are not representative of the CO2 concentrations of magmas at high pres-
sure. However, a new idea that has emerged is that magma flushing by CO2-rich gas
has the potential to cause magma dehydration and affect the degassing path of H2O.
Efforts must be made to improve the modeling of CO2-H2O evolution during decom-
pression. Improvements to models of magma degassing require experimental and
thermodynamic data on the solubility of CO2 in H2O-bearing basaltic melts.
4. Melt inclusion data can be used to model degassing of S and halogens from melt as a
function of pressure. Segregation of gas bubbles from the melt at depth will result in
high ratios of S/Cl in the gas phase released by erupting magmas.
5. The petrologic method based on melt inclusions has proven to be reliable for estimat-
ing the SO2 output from syn-eruptive degassing of large flood basalts and basaltic
magmas ascending in closed system conditions, provided there is no differential gas
transfer (gas loss) prior to eruption.
6. For volcanic systems underdoing quiescent degassing, the S contents of undegassed
melt inclusions can be combined with SO2 fluxes measured in surface gas emissions
to estimate the volume of underlying, non-erupted magma that has degassed.
7. The effects of shallow degassing processes must be considered in studies that attempt
to elucidate the primary volatile contents of magmas based on olivine-hosted melt
inclusions. Systematic volatile and trace element analyses in subduction-related
magmas make it possible to (i) distinguish magmas formed by decompression melting
of MORB- or OIB-type mantle sources from those formed by partial melting of a
mantle source fluxed by slab-derived H2O-rich components, (ii) address specific
questions on the relative fractionation of H2O and Cl during slab dehydration, and (iii)
investigate the origins of the high S concentrations in some arc basalts. Future progress
in understanding volatile recycling in arcs requires integration of melt inclusion data,
geodynamic models for the thermal structure in the downgoing plate and mantle
wedge, thermodynamic and experimental data on H2O storage and its effect on partial
melting in the wedge, and models of fluid and melt transport through the mantle.
Finally, one of the main points that arises here is that combining melt inclusion data
with (i) accurate studies of their host olivines and the mineralogy of the host magmas, (ii)
experimental data, and (iii) field work and gas measurements is a necessity and represents a
main challenge for the next few years.
396 Métrich & Wallace
aCKnOWLeGMentS
We are grateful to K. Putirka, J. Lowenstern, P. Papale and G. Moore for their comments
and thorough reviews that strongly contributed to improving this manuscript. We would also like
to thank Emily Johnson for providing several of the photomicrographs of melt inclusions.
reFerenCeS
Agrinier P, Jendrzejewski N (2000) Overcoming problems of density and thickness measurements in FTIR
volatile determinations: a spectroscopic approach. Contrib Mineral Petrol 139:265-272
Alletti M, Baker DR, Freda C (2007) Halogen diffusion in a basaltic melt. Geochim Cosmochim Acta 71:3570-
3580
Allard P, Jean-Baptiste P, D’Alessandro W, Parello F, Parisi B, Flehoc C (1997) Mantle-derived helium and
carbon in groundwaters and gases of Mount Etna, Italy. Earth Planet Sci Lett 148:501-516
Allard P (1997) Endogenous magma degassing and storage at Mount Etna. Geophys Res Lett 24:2219-2222
Allard P, Burton M, Muré F (2005) Spectroscopic evidence for a lava fountain driven by previously accumulated
magmatic gas. Nature 43:407-410
Allard P, Aiuppa A, Burton M, Caltabiano T, Federico C, Salerno G, La Spina A. (2008) Crater gas emissions
and the magma feeding system of Stromboli volcano. Learning from Stromboli. AGU Geophys
Monograph (in press)
Anderson AT (1975) Some basaltic and andesitic gases. Rev Geophys Space Phys 13:37-55
Anderson AT (1976) Magma mixing: Petrological process and volcanological tool. J. Volcanol Geotherm Res
1:3-33
Anderson AT (1995) CO2 and the eruptibility of picrite and komatiite. Lithos 34:19-25
Anderson AT, Brown GG (1993) CO2 contents and formation pressures of some Kilauean melt inclusions. Am
Mineral 78:794-803
Atlas zD, Dixon JE, Sen G, Finny M, Martin-Del Pozzo AL (2006) Melt inclusions from volcan Popocatépelt
and volcan Colima, Mexico: melt evolution due to vapor-saturated crystallization during ascent. J
Volcanol Geotherm Res 153:221-240
Auer S, Bindeman I, Wallace P, Ponomareva V, Portnyagin M (2008) The origin of hydrous, high δ18O
voluminous volcanism: Diverse oxygen isotope values and high magmatic water contents within the
volcanic record of Klyuchevskoy Volcano, Kamchatka, Russia. Contrib Mineral Petrol, doi:10.1007/
s00410-008-0330-0
Baker DR, Freda C, Brooker RA, Scarlato P (2005) Volatile diffusion in silicate melts and its effects on melt
inclusions. Ann Geophys 48:699-717
Baker DR (2008) The fidelity of melt inclusions as records of melt composition. Contrib Mineral Petrol,
doi:10.1007/s00410-008-0291-3
Barry TL, Pearce JA, Leat PT, Millar IL, le Roex AP (2006) Hf isotope evidence for selective mobility of high-
field-strength elements in a subduction setting: South Sandwich Islands. Earth Planet Sci Lett 252:223-
244
Benjamin ER, Plank T, Wade JA, Kelley KA, Hauri EH, Alvarado GE (2007) High water contents in basaltic
magmas from Irazù volcano, Costa Rica. J Volcanol Geotherm Res 168:68-92
Behrens H, Roux J, Neuville DR, Siemann M (2006) quantification of dissolved H2O in silicate glasses using
confocal microRaman spectroscopy. Chem Geol 229:96-112
Bertagnini A, Métrich N, Landi P, Rosi M (2003) Stromboli volcano (Aeolian Archipelago, Italy): An open
window on the deep-feeding system of a steady state basaltic volcano. J Geoph Res 108(B7):2336,
doi:10.1029/2002JB002146
Bindeman I (2008) Oxygen isotopes in mantle and crustal magmas as revealed by single crystal analysis. Rev
Mineral Geochem 69:445-478
Blundy J, Cashman K (2008) Petrologic reconstruction of magmatic system variables and processes. Rev
Mineral Geochem 69:179-239
Botcharnikov RE, Holtz F, Behrens H (2007) The effect of CO2 on the solubility of H2O-Cl fluids in andesitic
melt. Eur J Mineral 19:671-680
Botcharnikov R, Freise M, Holtz F, Berhens H (2005) Solubility of C-O-H mixtures in natural melts: new
experimental data and application range of recent models. Ann Geophys 48:633-646
Bouvier AS, Métrich N, Deloule E (2008) Slab-derived fluids in magma sources of St Vincent (Lesser Antilles
Arc): volatiles and light elements imprints. J Petrol 49:1427-1448, doi:10.1093/petrology/egn031
Carroll MR, Rutherford MJ (1987) The stability of igneous anhydrite: experimental results and applications for
sulfur behaviour in the 1982 El Chichon trachyandesite and other evolved magmas. J Petrol 28:781-801
Volatile Abundances & Degassing Paths Tracked by Melt Inclusions 397
Carroll MR, Rutherford MJ (1988) Sulphur speciation in hydrous experimental glasses of varying oxidation
states: results from measured wavelength shifts of sulphur x-ray. Am Mineral 73:845-849
Carroll MR, Webster JD (1994) Solubilities of sulfur, noble gases, nitrogen, chlorine and fluorine in magmas.
Rev Mineral Geochem 30:231-279
Cashman KV (2004) Volatile controls on magma ascent and eruption. In: The State of the Planet: Frontiers
and Challenges in Geophysics. AGU Geophys Monogr Series 150. Sparks RSJ, Hawkesworth CJ (eds),
Washington DC, p 109-124
Cervantes P, Kamenetsky V, Wallace P (2002) Melt Inclusion Volatile Contents, Pressures of Crystallization
for Hawaiian Picrites, and the Problem of Shrinkage Bubbles. EOS Trans, AGU, 83(47), Fall Meet Suppl
Abstract V22A-1217
Cervantes P, Wallace P (2003a) Role of H2O in subduction-zone magmatism: New insights from melt inclusions
in high-Mg basalts from central Mexico. Geology 31:235-238
Cervantes P, Wallace P (2003b) Magma degassing and basaltic eruption styles: a case study of 2000 year BP
xitle volcano in Central Mexico. J Volcanol Geotherm Res 120:249-270
Chaussidon M, Albarède F, Sheppard SM (1989) Sulphur isotope variations in the mantle from ion microprobe
analyses of micro-sulphide inclusions. Earth Planet Sci Lett 92:144-156
Coltelli M, Del Carlo P, Pompilio M, Vezzoli L (2005) Explosive eruption of a picrite: The 3930 BP subplinian
eruption of Etna volcano (Italy). Geophys Res Lett 32:L23307, doi:10.1029/2005GL024271
Cottrell E, Spiegelman M, Langmuir CH (2002) Consequences of diffusive re-equilibration for the interpretation
of melt inclusions. Geochem Geophys Geosys 3, doi:10.1029/2001GC000205
Danyushevsky LV, Della-Pasqua FN, Sokolov S (2000) Re-equilibration of melt inclusions trapped by
magnesian olivine phenocrysts from subduction-related magmas: petrological implications. Contrib
Mineral Petrol 138:68-83
Danyushevsky LV (2001) The effect of small amounts of H2O on crystallization of mid-ocean ridge and bark-
arc basin magmas. J Volcanol Geotherm Res 110:265-280
Danyushevsky LV, McNeill AW, Sobolev AV (2002) Experimental and petrological studies of melt inclusions
in phenocrysts from mantle-derived magmas: an overview of techniques, advantages and complications.
Chem Geol 183:5-24
de Hoog JCM, Taylor BE, van Bergen MJ (2001) Sulfur isotope systematics of basaltic lavas from Indonesia:
implications for the sulfur cycle in subduction zones. Earth Planet Sci Lett 189:237-252
Di Carlo I, Pichavant M, Rotolo S, Scaillet B (2006) Experimental Crystallization of a High-K Arc Basalt: the
Golden Pumice, Stromboli Volcano (Italy), J Petrol 1-27 doi:10.1093/petrology/egl011
Di Muro A, Villemant B, Montagnac G, Scaillet B, Reynard B (2006) quantification of water content and
speciation in natural silicic glasses (phonolite, dacite, rhyolite) by confocal MicroRaman spectroscopy.
Geochim Cosmochim Acta 70:2868-2884
Dixon JE (1997) Degassing of alkalic basalts. Am Mineral 82:368-378
Dixon JE, Clague DA, Stolper EM (1991) Degassing history of water, sulfur, and carbon in submarine lavas
from Kilauea volcano, Hawaii. J Geol 99:371-394
Dixon JE, Stolper EM (1995) An experimental study of water and carbon dioxide solubilities in mid-ocean
ridge basaltic liquids. Part II: Applications to degassing. J Petrol 36:1633-1646
Dixon JE, Stolper EM, Holloway JR (1995) An experimental study of water and carbon dioxide solubilities in
mid-ocean ridge basaltic liquids. Part I: calibration and solubility models. J Petrol 36:1607-1631
Dixon JE, Pan V (1997) determination of the molar absorptivity of dissolved carbonate in basaltic glass. Am
Mineral 80:1339-1342
Donaldson CH (1976) An experimental investigation of olivine morphology. Contrib Mineral Petrol 57:187-
213.
Eiler JM, Carr MJ, Reagan M, Stolper E (2005) Oxygen isotope constraints on the sources of Central American
arc lavas. Geochem Geophys Geosys 6: q07007, doi:10.1029/2004GC000804
Faure F, Trolliard G, Nicollet C, Montel J-M (2003) A developmental model of olivine morphology as a
function of the cooling rate and the degree of undercooling. Contrib Mineral Petrol 145:251-263
Faure F, Schiano P (2004) Crystal morphologies in pillow basalts: implications for mid-ocean ridge processes.
Earth Planet Sci Lett 220:331-344
Faure F, Schiano P (2005) Experimental investigation of equilibration conditions during forsterite growth and
melt inclusion formation. Earth Planet Sci Lett 236:882-898
Fine GJ, Stolper EM (1986) Carbon dioxide in basaltic glasses: concentrations and speciation. Earth Planet
Sci Lett 76:263-278
Fisher T, Marty B (2005) Volatile abundances in the sub-arc mantle: insights from volcanic and hydrothermal
gas discharges. J Volcanol Geotherm Res 140:205-216
Freda C, Baker DR, Scarlato P (2005) Sulfur diffusion in basaltic melts. Geochim Cosmochim Acta 69:5061-
5069
Gaetani GA, Watson EB (2000) Open system behavior of olivine hosted melt inclusions. Earth Planet Sci Lett
83:27-41
398 Métrich & Wallace
Gaetani GA, Watson EB (2002) Modeling the major-element evolution of olivine-hosted melt inclusions.
Chem Geol 183:25-41
Gardner JE, Burgisser A, Hort M, Rutherford M, (2006) Experimental and model constraints on degassing of
magma during ascent and eruption. In: Neogene-quaternary continental margin volcanism: A perspective
from Mexico. Siebe C, Macias JL, Aguirre-Diaz GJ (eds) Geol Soc Am Spec Paper 402:99-113
Gerlach TM, Graeber EJ (1985) Volatile budget of Kilauea volcano. Nature 313:273-277
Gerlach TM, McGee KA, Elias T, Sutton AJ, Doukas MP (2002) The carbon dioxide emission rate of Kïlauea
Volcano; implications for primary magma CO2 content and summit reservoir dynamics. J Geophys Res
107:2189
Gioncada A, Clocchiatti R, Sbrana A, Bottazzi P, Massare D, Ottolini L (1998) A study of melt inclusions at
Vulcano (Aeolian islands, Italy): insights on the primitive magmas and on the volcanic feeding system.
Bull Volcanol 60:286-306
Giordano D, Dingwell DB (2003) Viscosity of hydrous Etna basalt: implications for Plinian-style basaltic
eruptions. Bull Volcanol 65:8-14
Gonnermann HM, Manga M (2005) Nonequilibrium magma degassing: Results from modeling of the ca. 1340
A.D. eruption of Mono Craters, California. Earth Planet Sci Lett 238:1-16
Grove TL, Chatterjee N, Parman SW, Medard E (2006) The influence of H2O on mantle wedge melting. Earth
Planet Sci Lett 249:74-89
Grove TL, Baker M, Price RC, Parman SW, Elkins-Tanton LT, Chatterjee N, Muentener O (2005) Magnesian
andesite and dacite lavas from Mt. Shasta, northern California: products of fractional crystallization of
H2O-rich mantle melts. Contrib Mineral Petrol 148:542-565
Grove TL, Elkins LT, Parman SW, Chatterjee N, Müntener O, Gaetani GA (2003) Fractional crystallization and
mantle-melting controls on calc-alkaline differentiation trends. Contrib Mineral Petrol 145:515-533
Grove TL, Parman SW, Bowring SA, Price RC, Baker MB (2002) The role of an H2O-rich fluid component in
the generation of primitive basaltic andesites and andesites from Mt. Shasta region, N. California. Contrib
Mineral Petrol 142:375-396
Hansteen TH, Klügel A (2008) Fluid inclusion thermobarometry as a tracer for magmatic processes. Rev
Mineral Geochem 69:143-177
Hamada M, Fujii T (2007) H2O-rich island arc low-K tholeiite magma inferred from Ca-rich plagioclase-melt
inclusion equilibria. Geochem J 41:437-461
Hauri E (2002) SIMS analysis of volatiles in silicate glasses, 2: isotopes and abundances in Hawaiian melt
inclusions. Chem Geol 183:15-141
Hauri E, Wang J, Dixon JE, King PL, Mandeville C, Newman S (2002) SIMS analysis of volatiles in silicate
glasses, 1: Calibration, matrix effects and comparisons with FTIR. Chem Geol 183:99-114
Hauri EH, Gaetani GA, Green TH (2006) Partitioning of water during melting of the Earth’s upper mantle at
H2O-undersaturated conditions. Earth Planet Sci Lett 24:715-734
Hochstaedter AG, Ryan JG, Luhr JF, Hasenaka T (1996) On B/Be ratios in the Mexican Volcanic belt. Geochim
Cosmochim Acta 60:613-628
Holzeid A, Grove TL (2002) Sulfur saturation limits in silicate melts and their implications for core formation
scenarios for terrestrial planets. Am Mineral 87:227-237
Housh TB, Luhr JF (1991) Plagioclase-melt equilibria in hydrous systems. Am Mineral 76:477-492
Humphreys MCS, Kearns SL, Blundy JD (2006) SIMS investigation of electron-beam damage to hydrous,
rhyolitic glasses: Implications for melt inclusion analysis. Am Mineral 91:667-679
Ihinger PD, Hervig RL, McMillan PF (1994) Analytical methods for volatiles in glasses. Rev Mineral 30:67-
121
Jaupart C, Vergniolle S (1988) Laboratory models of Hawaiian and Strombolian eruptions. Nature 331:58-60
Jaupart C, Vergniolle S (1989) The generation and collapse of a foam layer at the roof of a basaltic magma
chamber. J Fluid Mechanic 203:347-380
Javoy M, Pineau F (1991) The volatiles record of a “popping’ rock from the Mid-Atlantic Ridge at 14°N:
chemical and isotopic composition of gas trapped in the vesicles. Earth Planet Sci Lett 107:598-611
Johnson EJ, Wallace PJ, Cashman KV, Delgado Granados H, Kent A (2008) Magmatic volatile contents and
degassing-induced crystallization at Volcan Jorullo, Mexico: Implications for melt evolution and the
plumbing systems of monogenetic volcanoes. Earth Planet Sci Lett 269:477-486
Johnson EJ, Wallace PJ, Delgado Granados H, Manea VC, Kent A, Bindeman I, Donegan C (2008) The origin
of H2O-rich subduction components beneath the Michoacán-Guanajuato Volcanic Field, Mexico: insights
from magmatic volatile contents, oxygen isotopes, and 2-D thermal models for the subducted slab and
mantle wedge. J Petrol (in review)
Jugo PJ, Luth RW, Richards JP (2005a) Experimental data on the speciation of sulfur as a function of oxygen
fugacity in basaltic melts. Geochim Cosmochim Acta 69:477-503
Jugo PJ, Luth RW, Richards JP (2005b) An experimental study of the sulfur content in basaltic melts saturated
with immiscible sulphide or sulfate liquids at 1300 °C and 1.0 GPa. J Petrol 46:783-798
Volatile Abundances & Degassing Paths Tracked by Melt Inclusions 399
Kamenetsky V, Pompilio M, Métrich N, Sobolev AV, Kuzmin DV, Thomas R (2007) Arrival of extremely
volatile-rich high-Mg magmas changes explosivity of Mount Etna. Geology 35:255-258, doi:10.1130/
G23163A.1
Kelley KA, Plank T, Grove TL, Stolper EM (2006) Mantle melting as a function of water content beneath back-
arc basins. J Geophys Res 111, B09208, doi:10.1029/ 2005JB003732
Kent AJR (2008) Melt inclusions in basaltic and related volcanic rocks. Rev Mineral Geochem 69:273-331
Kent AJR, Peate DW, Newman S, Stolper EM, Pearce JA (2002) Chlorine in submarine glasses from the Lau
Basin: seawater contamination and constraints on the composition of slab-derived fluids. Earth Planet Sci
Lett 202:361-377
Kessel R, Schmidt MW, Ulmer P, Pettke T (2005) Trace element signature of subduction-zone fluids, melts and
supercritical liquids at 120-180 km depth. Nature 437: doi:10.1038/nature03971
King PL, Vennemann TW, Holloway JR, Hervig RL, Lowenstern JB, Forneris JF (2002) Analytical techniques
for volatiles: A case study using intermediate (andesitic) glasses. Am Mineral 87:1077-1089
King PL, Holloway JR (2002) CO2 solubility and speciation in intermediaite (andesitic) melts: the role of H2O
and composition. Geochim Cosmochim Acta 66:1627-1640
Kohut EJ, Stern R, Kent AJR., Nielsen RL, Bloomer SH, Leybourne M (2006) Evidence of adiabatic
decompression melting in the Southern Marianna Arc from high-Mg lavas and melt inclusions. Contrib
Mineral Petrol, doi:10.1007/s00410-006-0102-7
Landi P, Métrich N, Bertagnini A, Rosi M (2004) Dynamics of magma mixing and degassing recorded in
plagioclase at Stromboli (Aeolian Archipelago, Italy). Contrib Mineral Petrol 147:213-237
Le Roux PJ, Shirey SB, Hauri EH, Perfit MR, Bender JF (2006) The effects of variable sources, processes and
contaminants on the composition of northern EPR MORB (8-10°N and 12-14°N): Evidence for volatiles
(H2O, CO2, S) and halogens (F, Cl). Earth Planet Sci Lett 251:209-231
Li C, Ripley EM (2005) Empirical equations to predict the sulfur content of mafic magmas at sulfide saturation
and applications to magmatic sulfide deposits. Mineral Deposit 40:18-230
Liu Y, Samaha N-T, Baker DR (2007) Sulfur concentration at sulfide saturation (SCSS) in magmatic silicate
melts. Geochim Cosmochim Acta 71:1783-1799
Lowenstern JB (1995) Application of silicate-melt inclusions to the study of magmatic volatiles. In: Magmas,
Fluids and Ore Deposits. Thompson JFH (ed) Mineral Assoc Canada Short Course Series 23:71-100
Luhr JF (1990) Experimental phase relations of water and sulfur saturated arc magmas and the 1982 eruptions
of El Chichon volcano. J Petrol 31:1071-1114
Luhr JF (1992) Slab-derived fluids and partial melting in subduction zones - insights from 2 contrasting
Mexican volcanoes (Colima and Ceboruco). J Volcanol Geothem Res 54:1-18
Luhr JF (1997) Extensional tectonics and the diverse primitive volcanic rocks in the western Mexican Volcanic
Belt. Can Mineral 35:473-500
Luhr JF (2001) Glass inclusions and melt volatile contents at Paricutin volcano, Mexico. Contrib Petrol Mineral
142:261-283
Mandeville CW, Webster JD, Rutherford MJ, Taylor BE, Timbal A, Faure K (2002) Determination of molar
absorptivities in infrared absorption bands of H2O in andesitic glasses. Am Mineral 87:813-821
Marianelli P, Sbrana A, Métrich N, Cecchetti A (2005) The deep feeding system of Vesuvius involved in recent
violent strombolian eruptions. Geophys Res Lett 32:L02306, doi:10.1029/2004GL021667
Massare D, Métrich N, Clocchiatti R (2002) High-temperature experiments on silicate melt inclusions in
olivine at one atmosphere. Inference on temperatures of homogenization and H2O concentrations. Chem
Geol 183:87-98
Mavrogenes JA, O’Neill HC (1999) The relative effects of pressure, temperature and oxygen fugacity on the
solubility of sulfide in mafic magmas. Geochim Cosmochim Acta 63:1173-1180
Menand T, Philips JC (2007) Gas segregation in dykes and sills. J Volcanol Geotherm Res 159:393-408
Mercier M et al. (2008) Influence of glass polymerization and oxidation on microRaman water analysis in
alumino-silicate glasses. Geochim Cosmochim Acta (accepted)
Métrich N, Clocchiatti R (1996) Sulfur abundance and its speciation in oxidized alkaline melts. Geochim
Cosmochim Acta 60:4151-4160
Métrich N, Ruhterford MJ (1998) Low pressure crystallization paths of H2O-saturated basaltic-hawaiitic melts
from Mt Etna: Implications for open-system degassing of basaltic volcanoes. Geochim Cosmochim Acta
62:1195-1205
Métrich N, Schiano P, Clocchiatti R, Maury RC (1999) Transfer of sulfur in subduction settings. An example
from Batan island (Luzon volcanic arc, Philippines). Earth Planet Sci Lett 67:1-14
Métrich N, Bertagnini A, Landi P, Rosi M (2001) Crystallization driven by decompression and water loss at
Stromboli volcano (Aeolian island) Italy. J Petrol 42:1471-1490
Métrich N, Susini, Galoisy L, Calas G, M. Bonnin-Mosbah.M., Menez B (2003) x-ray microspectroscopy of
sulphur in basaltic glass inclusions. Inference on the volcanic sulphur emissions. J Phys 4 104:393-398
400 Métrich & Wallace
Métrich N, Allard P, Spilliaert N, Andronico D, Burton M (2004) 2001 flank eruption of the alkali- and volatile-
rich primitive basalt responsible for Mount Etna’s evolution in last three decades. Earth Planet Sci Lett
228:1-17
Métrich N, Berry AJ, O’Neill H St C, Susini J (2008) The oxidation state of sulfur in synthetic and natural
glasses determined by x-ray absorption spectroscopy. Geochim Cosmochim Acta (submitted)
Métrich N, Bertagnini A, Landi P, Rosi M, Belhadj O (2005) Triggering mechanism at the origin of paroxysms
at Stromboli (Aeolian archipelago, Italy): the 5 April 2003 eruption. Geophys Res Lett 32:L103056,
doi:10.1029/2004GL022257
Métrich N, Kamenetsky V, Allard P, Pompilio M (2007) Trace element and volatile signature of Etna magma
source(s): a melt inclusion approach. 2007 Goldschmidt Conference Cologne. Geochim Cosmochim Acta
Suppl 71(15):A658
Michael PJ (1995) Regionally distinctive sources of depleted MORB: evidence from trace elements and H2O.
Earth Planet Sci Lett 131:301-320
Moore G, Vennemann T, Carmichael ISE (1998) An empirical model for the solubility of H2O in magmas to 3
kilobars. Am Mineral 83:36-42
Moore G (2008) Interpreting H2O and CO2 contents in melt inclusions: constraints from solubility experiments
and modeling. Rev Mineral Geochem 69:333-361
Moretti R, Ottonello G (2005) Solubility and speciation of sulfur in silicate melts: The conjugated Toop-Samis-
Flood-grjotheim (CTSFG) model. Geochim Cosmochim Acta 69:801-823
Newman S, Lowenstern JB (2002) VolatileCalc: a silicate melt-H2O-CO2 solution model written in Visual
Basic Excel. Comp Geosci 2:597-604
Newman S, Stolper EM, Stern RJ (2000) H2O and CO2 in magmas from Mariania arc and back arc systems.
Geochem Geophys Geosyst 1:1999GC000027
Nichols ARL, Wysoczanski RJ (2007) Using micro-FTIR spectroscopy to measure volatile contents in small
and unexposed inclusions hosted in olivine crystals. Chem Geol 242:371-384
Ochs FA, Lange RA (1999) The density of hydrous magmatic liquids. Science 283:1314-1317
O’Neill HSC, Mavrogenes JA (2002) The sulfide capacity and sulfur content at sulfite saturation of silicate
melts at 1400 °C and 1b. J Petrol 43:1049-1087
Papale P, Moretti R, Barbato D (2006) The compositional dependence of the saturation surface of H2O + CO2
fluids in silicate melts. Chem Geol 229:78-95
Parfitt EA (2004) A discussion of the mechanisms of explosive basaltic eruptions. J Volcanol Geotherm Res
134:77-107
Parfitt EA, Wilson L (1995) Explosive volcanic eruptions - Ix. The transition between Hawaiian-style lava
fountaining and strombolian explosive activity. Geophys J Int 121:226-232
Pearce JA, Peate DW (1995) Tectonic implications of the composition of volcanic arc magmas. Annu Rev
Earth Planet Sci 23:251-285
Pichavant M, Mysen BO, Macdonald R (2002) Source and H2O content of high-MgO magmas in island arc
settings: an experimental study of a primitive calc-alkaline basalt from St. Vincent, Lesser Antilles arc.
Geochim Cosmochim Acta 66:2193-2209
Pichavant M, Macdonald R (2007) Crystallization of primitive basaltic magmas at crustal pressures and genesis
of the calc-alkaline igneous suite: experimental evidence from St Vincent, Lesser Antilles arc. Contrib
Mineral Petrol 2007, doi10.1007/s00410-007-0208-6
Philippot P, Agrinier P, Scambelluri M (1998) Chlorine cycling during subduction of altered oceanic crust.
Earth Planet Sci Lett 161:33-44
Philips JC, Woods AW (2001) Bubble plume generated during recharge of basaltic magma reservoirs. Earth
Planet Sci Lett 186:297-309
Pioli L, Erlund E, Johnson E, Cashman K, Wallace P, Rosi M, Delgado Granados H (2008) Explosive dynamics
of violent Strombolian eruptions: The eruption of Parícutin Volcano 1943-1952 (Mexico). Earth Planet
Sci Lett 271:359-368
Portnyagin M, Hoernle K, Plechov P, Mironov N, Khununaya S (2007) Constraints on mantle melting and
composition and nature of slab components in volcanic arcs from volatiles (H2O, S, Cl, F) and trace
elements in melt inclusions from the Kamchatka Arc. Earth Planet Sci Lett 255:53-69
Portnyagin M, Almeev R, Matveev S, Holtz F (2008) Experimental evidence for rapid water exchange between
melt inclusions in olivine and host magma. Earth Planet Sci Lett 272:541-552
Richet P, Lejeune AM, Holtz F, Roux J (1996) Water and viscosity of andesite melts. Chem Geol 128:185-
197
Righter K, Chesley JT, Calazza CM, Gibson EK, Ruiz J (2008) Re and Os concentrations in arc basalts: The
roles of volatility and source region fO2 variations. Geochim Cosmochim Acta 72:926-947
Riker J (2005) The 1859 eruption of Mauna Loa Volcano, Hawaii: Controls on the development of long lava
channels. Unpublished M.S. Thesis, University of Oregon
Volatile Abundances & Degassing Paths Tracked by Melt Inclusions 401
Robock A, Oppenheimer C (2003) Volcanism and the Earth’s Atmosphere. Geophys Monograph 139,
Washington DC
Roedder E (1984) Fluid Inclusions, Am Mineral Soc, Chelsea, Michigan
Roggensack K, Hervig RL, McKnight SB, Williams SN (1997) Explosive basaltic volcanism from Cerro
Negro volcano: influence of volatiles on eruptive style. Science 277:1639-1642
Roggensack K (2001) Unraveling the 1974 eruption of Fuego volcano (Guatemala) with small crystals and
their young melt inclusions. Geology 29:911-914
Rowe MC, Kent AJR, Nielsen RL (2007) Determination of sulfur speciation and oxidation of olivine-hosted
melt inclusions. Chem Geol 236:303-322
Sadofsky SJ, Portnyagin M, Hoernle K, van den Bogaard P (2008) Subduction cycling of volatiles and trace
elements through the Central American volcanic arc: evidence from melt inclusions. Contrib Mineral
Petrol 155, doi:10.1007/s00410-007-0251-3
Scaillet B, Pichavant M (2005) A model of sulphur solubility for hydrous mafic melts: application to the
determination of magmatic fluid compositions of Italian volcanoes. Ann Geophys 48:671–698
Scambelluri M, Fiebig J, Malaspina N, Muntener O, Pettke T (2004) Serpentinite subduction: Implications for
fluid processes and trace-element recycling. Int Geol Rev 46:595-613
Self S, Widdowson M, Thordarson T, Jay AE (2006) Volatile fluxes during flood basalt eruptions and potential
effects on global environment: A Deccan perspective. Earth Planet Sci Lett 248:517-531
Sharma K, Blake S, Self S, Krueger AJ (2004) SO2 emissions from basaltic eruptions, and excess sulfur issue.
Geophys Res Lett 31:L13612, doi:10.1029/2004GL0.19688
Shaw HR (1972) Viscosities of magmatic liquids: an empirical method of prediction. Am J Sci 272:870-893
Silver LA, Ihinger PD, Stolper E (1990) The influence of bulk composition on the speciation of water in silicate
glasses. Contrib Mineral Petrol 104:142-162
Sisson TW (2003) Native gold in a Hawaiian alkalic magma. Econ Geol 95:643-648
Sisson TW, Bronto S (1998) Evidence for pressure-release melting beneath magmatic arcs from basalt at
Galunggung, Indonesia. Nature 391:883-885
Sisson TW, Layne GD (1993) H2O in basalt and basaltic andesite glass inclusions from 4 subduction-related
volcanoes. Earth Planet Sci Lett 117:619-635
Sisson TW, Grove TL (1993) Temperatures and H2O contents of low-MgO high-Alumina basalts. Contrib
Mineral Petrol 113:167-184
Sobolev AV, Clocchiatti R, Dhamelincourt P (1983) Les variations de la température, de la composition
du magma et l’estimation de la pression partielle d’eau pendant la cristallisation de l’olivine dans les
océanites du Piton de la Fournaise (Réunion, éruption de 1966). C R Acad Sci Paris 296:275-280
Sobolev AV, Chaussidon M (1996) H2O concentrations in primary melts from supra-subduction zones and
mid-oceanic ridges: Implications for H2O storage and recycling in the mantle. Earth Planet Sci Lett
137:45-55
Sobolev VS, Kostyuk VP (1975) Magmatic Crystallisation as Based on the Study of Melt Inclusions. “Nauka”,
Novosibirsk, pp 232
Sparks RSJ (1978) The dynamics of bubble formation and growth in magmas. J Volcanol Geotherm Res 3:1-
37
Spilliaert N, Allard P, Métrich N, Sobolev A (2006a) Melt inclusion record of the conditions of ascent, degassing
and extrusion of volatile-rich alkali basalt during the powerful 2002 flank eruption of Mount Etna (Italy).
J Geophys Res 111:B04203, doi:10.1029/2005/JB003934
Spilliaert N, Métrich N, Allard P (2006b) S-Cl-F degassing pattern of water-rich alkali basalt: modelling and
relationship with eruption styles on Mount Etna volcano. Earth Planet Sci Lett 248:772-786
Stolper EM, Newman S (1994) The role of water in the petrogenesis of Mariana trough magmas. Earth Planet
Sci Lett 121:293-325
Tait S, Jaupart C, Vergniolle S (1989) Pressure, gas content and eruption, periodicity in a shallow crystallizing
magma chamber. Earth Planet Sci Lett 92:107-123
Thomas R (2000) Determination of water contents of granite melt inclusions by confocal laser Raman
microprobe spectroscopy. Am Mineral 85:868-872
Thomas R, Kamenetsky VS, Davidson P (2006) Laser Raman spectroscopic measurements of water in
unexposed glass inclusions. Am Mineral 91:467-470
Thordarson T, Self S (2003) Atmospheric and environmental effects of the 1783-1784 Laki eruption: A review
and reassessment. J Geophys Res 108(D1):4011, doi:10.1029/2001JD002042
Thordarson T, Self S, Oskarsson N, Hulsechosch T (1996) Sulfur, chlorine and fluorine degassing and
atmospheric loading by the 1783-1784 AD (Skaftar Fires) eruption in Iceland. Bull Volcanol 58:205-
225
Vergniolle S, Jaupart C (1990) Dynamics of degassing at Kilauea Volcano, Hawaii. J Geophys Res 95:2793-
2809
402 Métrich & Wallace
Vergniolle S, Mangan S (2000) Hawaian and strombolian eruption. In: Encyclopedia of Volcanoes. Sigurdsson
H (ed) Academic Press, p 447-461
Verhoogen J (1949) Thermodynamics of a magmatic gas phase. Univ Calif Bull Dept Geol Sci 28:91-136
Varela ME, Métrich N, Bonnin-Mosbah M, Kurat G (2000) Carbon in glass inclusions of Allende, Vigarano,
Bali and Kaba (CV3) olivines. Geochim Cosmochim Acta 64:3923-3930
Vigouroux N, Wallace PJ, Kent AJR (2008) Volatiles in high-K magmas from the western Trans-Mexican
Volcanic Belt: Evidence for fluid-flux melting and extreme enrichment of the mantle wedge by subduction
processes. J Petrol, doi:10.1093/petrology/egn039
Wade JA, Plank T, Melson WG, Soto GJ, Hauri AH (2006) Volatile content of magmas from Arenal volcano,
Costa Rica. J Volcanol Geotherm Res 157:94-120
Walker J, Roggensack K, Patino LC, Cameron BI, Matias O (2003) The water and trace element contents of
melt inclusions accross an active subduction zone. Contrib Mineral Petrol 146:62-77
Wallace PJ (2001) Volcanic SO2 emissions and the abundance and distribution of exsolved gas in magma
bodies. J Volcanol Geotherm Res 108:85-106
Wallace P (2005) Volatiles in subduction zone magmas: concentrations and fluxes based on melt inclusion and
volcanic gas data. J Volcano Geotherm Res 140:217-240
Wallace P, Carmichael ISE (1992) Sulfur in basaltic magmas. Geochim Cosmochim Acta 56:1863-1874
Wallace PJ, Carmichael ISE (1994) S speciation in submarine basaltic glasses as determined by measurements
of SKα x-ray wavelength shifts. Am Mineral 79:161-167
Wallace P, Anderson AT (1998) Effects of eruption and lava drainback on the H2O contents of basaltic magmas
at Kilauea volcano. Bull Volcanol 59:327-344
Wallace PJ, Carmichael ISE (1999) quaternary volcanism near the Valley of Mexico: Implications for subduction
zone magmatism and the effects of crustal thickness variations on primitive magma compositions. Contrib
Mineral Petrol 135:291-314
Wallace P, Anderson AT, Davis AM (1999) Gradients in H2O, CO2 and exsolved gas in a large-volume silicic
magma system: interpreting the record preserved in melt inclusions from the Bishop Tuff. J Geophys Res
104:20,097-20,122
Webster JD, Kinzler RJ, Mathez EA (1999) Chloride and water solubility in basalt and andesite melts and
implication for magmatic degassing. Geochim Cosmochim Acta 63:729-738
Webster JD (2004) The exsolution of magmatic hydrosaline chloride liquids. Chem Geol 210:33-48
Wilke M, Jugo PJ, Klimm K, Susini J, Botcharnikov R, Kohn SC, Janousch M (2008) The origin of S4+
detected in silicate glasses by xANES. Am Mineral 93:235-240
Wilson L, Head JW (1981) Ascent and eruption of basaltic magma on the Earth and Moon. J Geophys Res
86:2971-3001
Witter JB, Kress VC, Delmelle P, Stix J (2004) Volatile degassing, petrology, and magma dynamics of the
Villarica lava lake, Southern Chile. J Volcanol Geotherm Res 134:303-337
Witter JB, Kress VC, Newhall CG (2005) Volcan Popocatépelt, Mexico. Petrology, Magma mixing and
immediate sources of volatiles for the 1994-present eruption. J Petrol 46:2337-2366
Workman RK, Hart SR (2005) Major and trace element composition of the depleted MORB mantle (DMM).
Earth Planet Sci Lett 231:53-72
Wysoczanski RJ, Tani K (2006) Spectroscopic FTIR imaging of water species in silicic volcanic glasses and
melt inclusions: an example from the Izu-Bonin arc. J Volcanol Geotherm Res 156:302-314
zhang Yx, xu xJ, zhu MF (2007) Silicate melt properties and volcanic eruptions. Rev Geophys 45:RG4004
Reviews in Mineralogy & Geochemistry
Vol. 69 pp. 403-443, 2008 11
Copyright © Mineralogical Society of America
Inter- and Intracrystalline Isotopic disequilibria:
techniques and applications
Frank C. ramos
Department of Geological Sciences
New Mexico State University
Las Cruces, New Mexico, 88003-8001, U.S.A.
[email protected]
Frank J. tepley III
College of Oceanic and Atmospheric Sciences
Oregon State University
Corvallis, Oregon, 97331-5503, U.S.A.
[email protected]
IntrOduCtIOn
In the last ~20 years, we have seen a significant expansion of techniques related to geo-
chemical and isotopic microsampling of materials in the earth sciences. From constraining
pre-eruptive histories of flood basalt magmas to identifying the natal rivers of origin of
anadromous fishes, these techniques have had significant impacts in a wide variety of scientific
fields. Nowhere has the impact been greater than in identifying the sources, processes, and
timing of processes involved in igneous magmatic systems. Both technique refinements and
the development of new technologies have aided in advancing microsampling applications,
thus allowing for a better understanding of the sources and mechanisms responsible for
changing geochemical and isotopic signatures in natural systems. In this chapter, we focus on
the techniques and technologies associated with radiogenic isotope microsampling and review
applications of these techniques as utilized in scientific investigations.
Isotope microsamping is a logical extension of earlier studies that evaluated individual
components of magmas and magma systems, including melts and minerals. From the use of pe-
trographic microscopes and the later introduction of the electron microprobe, the focus on inter-
nal chemical variations in melts and minerals is critical to assessing the petrogenetic histories of
igneous rocks. Even today, these technologies are used to ensure that further trace element and
isotopic analyses are undertaken in a textural and major element context. For trace elements and
isotopes, early studies (e.g., Cortini and van Calsteren 1985) confirmed variations in the melt
and mineral components of many igneous rocks but focused on mineral or glass separates. Po-
tential information associated with isotopic variations retained by individual crystals or internal
variations within individual crystals was lost. Later studies such as Geist et al. (1988) focused
on isotope variations within single megacrysts to constrain mixing scenarios between basaltic,
andesitic, and rhyolitic magmas, and Davidson et al. (1990) used microdrilling to document
isotopic variations across an enclave-host magma interface and to identify xenocrystic crystals
from within enclaves. These represent early studies that truly focused on “internal” isotopic
variations of magmatic components from igneous systems and foreshadowed the potential of
future studies utilizing isotopes at the microscale to identify magmatic processes and constrain
their impacts on magmatic systems.
1529-6466/08/0069-0011$05.00 DOI: 10.2138/rmg.2008.69.11
404 Ramos & Tepley
Isotope microsampling can be separated into two broad categories: 1) microspatial analyses
and 2) microanalyte analyses. For microspatial analyses, samples commonly originate from
micromilling, laser ablation sampling, or secondary ionization mass spectrometry (SIMS). For
microanalyte analyses, samples are commonly obtained from single crystals or micromilling.
As implied by the name, microspatial analyses target a limited area/volume of material such as
a limited number of rings of a fish otolith or a single growth layer of a magmatic phenocryst,
with the ultimate intent of constraining isotopic variability in the growth history of the target
material. Constraining magmatic growth histories of phenocrysts exemplifies the utility of
microspatial analyses. Phenocrysts grow from core to rim by the addition of sequential layers
of material. Plagioclase, for example, typically contains high concentrations of Sr that allow
for small portions of mineral to be successfully analyzed for 87Sr/86Sr ratios. By analyzing
sequential layers, changes in the isotopic character of the magma in which the crystal grew,
imposed by open-system processes, can be constrained (Tepley et al. 2000; Ramos et al. 2005).
Thus, microspatial 87Sr/86Sr isotope analyses are well-suited to identify and track the effects of
open-system processes occurring during mineral growth.
With microanalyte techniques, spatial resolution may be sacrificed to obtain the minimum
required amount of analyte needed for a successful analysis. An example is that of Sr or Pb
in quartz. The mineral structure and composition of quartz (SiO2) does not readily allow for
Sr and Pb elemental substitution, two elements commonly used as radiogenic tracers. As a
result, Sr and Pb concentrations in quartz are low and typically result from melt or mineral
inclusions captured during growth of individual crystals. Low Sr and Pb concentrations and the
small size of quartz-hosted inclusions usually preclude them from being individually analyzed
by microspatial techniques. However, these crystals can be approached from a microanalyte
perspective using single grain analyses where all analyte is assumed to result from inclusions
rather than the quartz host. Such single crystals must however be large enough to contain
enough inclusions to yield the minimum required amount of analyte for successful analyses.
Thus, microspatial and microanalyte techniques overlap to some degree but they also diverge so
as to become very different in nature and, thus, scientific interpretation.
To comprehensively address isotope microanalysis techniques, we introduce the types
of isotope analyses undertaken, the technologies involved in these isotope analyses, the
critical aspects of these isotope analyses, and review a range of applications which have been
undertaken. Our intent is to offer a broad review of successful applications, not an exhaustive
review of all studies.
MICrOSPatIaL ISOtOPe anaLySIS
Early petrographic studies focusing on mineral phenocrysts from volcanic rocks
documented textural evidence for highly variable magmatic histories recorded during mineral
growth. Assessments of these varied histories were initially evaluated by measuring in situ
major element variations (e.g., core to rim transects) using the electron microprobe targeted
toward individual mineral phenocrysts. Such measurements were critical in assessing major
element chemical changes of magmas occurring during mineral crystallization. These changes
could result from a myriad of magmatic processes including, but not limited to, crystal
fractionation, magma mixing, magma recharge, and/or processes associated with assimilation/
fractional crystallization (AFC). Major element variations, however, could only account
for broad assessments supporting major element modification of magmas in which crystals
grew. Major elements alone could not, for example, confirm the presence of external inputs
into magmatic systems such as required by AFC processes. Thus, identifying endmembers
responsible for such external changes was more problematic.
Inter- and Intracrystalline Isotopic Disequilibria 405
In contrast to major elements, radiogenic isotopes offer an intriguing alternative in assessing
the influence of either internal (closed-system) or external (open-system) inputs into magmatic
systems and constraining the isotopic composition of the endmembers involved. Combining
major element and isotopes has vast potential in understanding igneous petrogenesis in addition
to other scientific fields. Initially, microdrilling (referred to as micromilling in this chapter)
using diamond-embedded bits designed for the dental and engineering industries was used
to obtain small samples in a microspatial context. Early attempts obtained samples on scales
of 300 µm diameter by 300 µm depth and typically targeted Sr isotopes (e.g., Tepley et al.
1999). Fortunately, for most systems, even this level of microspatial sampling documented
large isotopic variations in magmatic phenocrysts resulting from open-system processes
affecting phenocryst-related host magmas. Early targets included minerals such as plagioclase
and calcite, minerals generally characterized by high Sr contents (≥500 ppm). After milling,
samples were dissolved and purified using chromatographic techniques and analyzed using
thermal ionization mass spectrometry (TIMS). Analyte amounts were large (≥100 ng Sr) and
required purification to removed interfering elements (e.g., Rb) prior to analysis. Partly as a
result, micromilling became accepted as a tool for better understanding processes associated
with magmagenesis and the popularity of the technique increased to the point that commercial
micromills were built.
Since the mid- to late-1990s, micromilling has been a relatively common technique for
acquiring samples, especially for materials requiring purification using column chromatography
(e.g., Rb-rich potassium feldspar, Knesel et al. 1999). In addition, bits are now either diamond-
embedded or made from highly pure tungsten carbide. Both are commonly used for silicate
and carbonate applications that target a range of problems in igneous petrology. In addition,
applications associated with biology and ecology have used isotopes to trace mammal
movements (e.g., Hoppe et al. 1999) and identify the natal rivers of origin of marine caught
salmon (e.g., Barnett-Johnson et al. 2005). Even mineral formation rates in meteorites have
been examined using micromilling (Bizzarro et al. 2004; Thrane et al. 2006).
In addition to micromilling, two in situ techniques are commonly utilized for microspatial
isotope analyses. The first uses secondary ionization mass spectrometry (SIMS) to analyze
material that is removed from the surface of samples bombarded by oxygen ions, cesium ions,
or electrons. Such analyses are typically used to measure radiogenic Pb isotope ratios that
do not require high measurement precisions in minerals such as zircon (Davis et al. 2003;
Ireland and Williams 2003). More recently, common Pb isotope ratios have been measured to
variable precisions in U- and Th-poor materials such as melt inclusions. These measurements
are generally limited to ratios involving the more abundant 206Pb, 207Pb, and 208Pb isotopes and
have been useful in identifying mantle endmember components involved in ocean island basalt
(OIB) volcanism (e.g., Saal et al. 1998). In addition to Pb isotopes, recent work by Weber et
al. (2005) measured Sr isotopes in aragonite otoliths to evaluate natal river sources of salmon
(Bacon et al. 2004). As such, expansion of SIMS-related techniques have also allowed for a
broader spectrum of microspatial studies.
Microspatial analyses have also benefited from the attainment of increased accuracies
and greater measurement precisions using laser ablation sampling. Advances in laser sampling
systems and multi-collector inductively coupled mass spectrometry have combined to greatly
expand potential targets for Sr and Pb isotope analyses to include highly elementally complex
materials such as clinopyroxene and volcanic rock groundmass. In the last 5-7 years, great
strides have been made in obtaining highly accurate Sr and Pb isotope measurements with
variable but consistently greater measurement precisions which on occasion rival those typical
of TIMS (e.g., Ramos et al. 2004; Kent 2008). Below, we review microspatial techniques and
applications.
406 Ramos & Tepley
Micromilling
Technique. Historically, the concept of removing growth zones or intracrystalline sub-
samples from larger samples for chemical analyses was governed by the scientific benefit of
such analyses. All sub-disciplines in geology have gained some advantage from applying such
techniques from evaluating sulfur isotopes in sulfur-bearing phases in breccias (Lambert et al.
1982), to obtaining x-ray diffraction data from micromilled minerals in calc-silicate granulites
(Maaskant et al. 1980), to identifying fish provenance by micromilling otoliths for Sr isotopic
analyses (Kennedy et al. 2002). Applying micromilling in igneous petrology for the purposes
of evaluating isotopes in magma-related materials was first accomplished by Davidson et al.
(1990) in which the cores of plagioclase feldspar crystals, and a small droplet of chilled mafic
melt in the same hand sample, were milled using small bits and a basic drill press. Since then,
micromilling has been and continues to be used successfully in a range of igneous-related
applications.
Applications to igneous petrogenesis, however, are limited by the capability to physically
sample intended target materials and to efficiently capture micromilled materials for chemical
processing. As early as 1945, an experimental motorized microdrill was developed for extract-
ing hard minerals from polished surfaces of large samples (Wagner 1945). In 1977, a device
using a series of levers and joints that allowed for fine movements of a probe or hypodermic
needle called a micromanipulator was used for extracting small crystals, 10-20 µm in diameter,
from thin sections for further chemical analysis (Rickwood 1977). It was also common to use
either hand-held drills (e.g., Dremel™) or mounted drill presses (e.g., Sherline™ vertical mill)
assisted with either binocular or petrographic microscopes to mill material from a variety of
target materials as long as spatial resolution was not critical. Advances in micrometer stages,
optical systems, and finally, computer-assisted or -driven milling devices allowed for sampling
on fine spatial scales that are used today (e.g., NewWave™ MicroMill™ or New Hermes™
Vanguard VMC).
Davidson et al. (1990), and many subsequent microdrilling studies, successfully analyzed
Sr isotopes in plagioclase. Plagioclase, by far the most common mineral in arc-related volcanic
rocks, has great potential to record temporal chemical and isotopic changes associated with
magma chamber dynamics (Anderson 1983; Stamatelopoulou-Seymour et al. 1990; Blundy
and Shimizu 1991; Singer et al. 1995). It is usually one of the first minerals on the liquidus and
commonly crystallizes during magmatic differentiation of basalts and andesites. Compositional
zoning in plagioclase typically reflects primary growth owing to slow CaAl-NaSi diffusive
exchange within the crystal structure (Grove et al. 1984), and textural features preserved in
phenocrysts help identify changing physical and chemical parameters in the host magma from
which it crystallizes (e.g., Ginibre et al. 2002, 2004). Combined chemical and textural zoning
may be used to reconstruct the conditions in which plagioclase grew, and by proxy, the chemical
and physical evolution of the magma itself. Lastly, plagioclase usually contains high concentra-
tions of Sr and relatively low concentrations of Rb which make it ideally suited for Sr isotope
evaluations targeting initial Sr isotope ratios (i.e., 87Sr/86Sr ratios that have not been measur-
ably influenced by radiogenic ingrowth). In addition to plagioclase, phases such as potassium
feldspar, pyroxene, hornblende, apatite, glass, etc., may also be analyzed for Sr isotopes or
integrated with additional constraints such as Pb or Nd isotopes from these same phases.
Ultimately an optimal mineral phase and isotope system must be chosen that will yield
the most productive results for addressing the specific scientific problem of interest. To
accomplish this goal, a balance between elemental concentrations, the precision of the milling
technique, the ability to remove milled material from the crystal surface, and the ability of
the analytical technique (TIMS or solution MC-ICPMS) to analyze the chemically processed
sample accurately and precisely (as the precision required may be a function of the specific
scientific application) must be pursued. As noted by Davidson et al. (2007), there is a trade-off
Inter- and Intracrystalline Isotopic Disequilibria 407
between the concentration of the target element and the required minimum amount of material
milled and processed such that ample analyte must be acquired to provide for precise and
accurate results. As Figure 1 illustrates, higher concentrations of the target element require
less overall amounts of material to be sampled, which offers the potential for finer spatial
resolution. Alternatively, lower concentrations require greater amounts of milled sample, and
thus, coarser spatial resolution.
For example, Sr readily substitutes for Ca in plagioclase feldspars, thus concentrations
typically range from a few hundred to a few thousand ppm. Approximately 1-3 ng of Sr are
needed for a highly precise TIMS analysis (Charlier et al. 2006), although processing such
small samples requires exceptional care. Therefore, a plagioclase with a relatively low Sr
concentration of 100 ppm would require ~0.01 mg of sample to yield 1 ng Sr. Alternatively, for
a plagioclase with 1000 ppm Sr, only 0.001 mg of sample is required. Clearly the limiting step
may be the ability to recover, handle, and process such small amounts of material for analysis.
To do so requires having a consistent blank and good knowledge of the blank concentration
and isotopic composition (at such low analyte levels, blank corrections become increasingly
important). The ideal circumstance is to have a well-characterized sample-blank relationship
where blank corrections are minimal.
Micromilling techniques vary between 1) milling individual holes in crystal growth zones
to 2) milling shallow troughs along growth zones, in each case generating enough material to
undertake appropriate isotopic analyses. Early milling efforts used individual holes in which
several small (<300 µm diameter x <300 µm depth) holes were drilled into a phenocryst with
overlapping holes used to obtain finer spatial resolution (e.g., Siebel et al. 2005; Tepley et al.
1999, 2000). One disadvantage associated with this method is the lack of lateral or depth-
related spatial resolution. Knesel et al. (1999) used a variant of this technique by utilizing a
single, large-diameter hole (~600 µm) drilled into sanidine phenocrysts with material removed
10000
apatite
Concentration of Sr (ppm)
1000
plagioclase
hornblende
10
0n
K-feldspar
clinopyroxene
go
10
100
fS
ng
r
of
1n
Sr
go
fS
r
10
Typical Sr concentrations
Concentration
1 of element X (ppm)
0.001 0.01 0.1 1 10 100
mass (milligrams)
Figure 1. Diagram of concentration versus mass associated with different mineral phases and the masses
of material needed to produce 1 ng, 10 ng and 100 ng of target analyte for analysis. Also shown are
ranges of Sr concentrations in selected minerals found in typical volcanic rocks. [Used by permission of
Annual Reviews, from Davidson et al. (2007), Annual Review of Earth and Planetary Sciences, Vol. 35,
Figure
Supplemental Material, Fig. 1, p. 1]. 1.
408 Ramos & Tepley
at various controlled depths. So, although surficial lateral resolution was lost, depth resolution
was maintained by carefully controlling milling depths. The shallow-trough method of milling,
described in Charlier et al. (2006), uses a computer-controlled milling machine equipped with
a binocular microscope, an adjustable speed motor, and a computer controlled x-Y stage
reproducible to ±1 micron. The computer-controlled milling machine allows for sampling
multiple holes within growth zones and for designating the milling depth of each hole, such
that sufficient material is removed from the sample using multiple holes rather than a single
deep hole. The advantage is that lateral- and depth-related spatial resolution is increased in
comparison to milling single holes.
Depending on the milling device, the user will have to determine the desired milling
location. Hand-operated milling machines are best used with a micrometer stage so as to
maximize lateral spatial resolution. Previous workers used a detachable micrometer stage
(Fig. 2) that was secured to both the vertical mill and a petrographic microscope (e.g., Tepley
et al. 1999, 2000). The advantage with this technique is the ability to precisely determine
locations of drill sites under the microscope, record locations, re-fit the stage to a vertical mill,
and then precisely return to pre-determined drilling sites. Vertical spatial resolution, however,
suffers using this method because of the imprecision of determining when the bit is at the
sample surface. Computer-controlled milling machines are more precise for both lateral and
vertical spatial resolutions, and software accompanying milling machines helps to determine
drill tip location, sample surface location, and milling depth (Charlier et al. 2006). Options
such as milling depth, location, holes versus troughs, and rate of milling are controlled using
accompanying software.
Following milling, recovering material from the sample surface can be accomplished either
by the dry method, collecting material with a small plastic scoop, or the wet method, collecting
material using water. In the wet method, a drop of ultra-pure (18.2 MΩ) water is used as both a
cooling agent for the bit and sample collection medium. As material is milled from the sample
surface, milled particulates collect in the water creating a slurry. Charlier et al. (2006) describe
the added technique of placing a warmed square of Parafilm™ with a small ~4 mm hole cut
into it on the sample surface. The hole in the Parafilm™ is placed over the desired milling site
and aids in slurry collection by confining the water to the drilling area and preventing it from
draining into possible surface cracks of the polished thick section. Both methods use a micro-
pipette with 0.5-10 µL tip to suction/collect the slurry. Suctioning additional water drops from
the sample surface may be needed to ensure greater recovery of milled material.
If determining the isotopic compositions of samples is the primary goal, samples can be
placed directly into digestion beakers and be dissolved. However, if isotope dilution is required,
samples must be dried, weighed, and spiked. In this case, a sample weighing method must be
devised. Early studies (Tepley et al. 1999, 2000) use a small square piece of clean aluminum
foil, shaped around a circular mold, to make a flat-bottomed, aluminum collection vessel into
which the sample is pipetted and dried. Once dry, the aluminum vessel containing the sample
is weighed and the weight recorded. Sample is then carefully scraped into a digestion beaker.
The aluminum vessel, less sample, is then re-weighed; the difference between the two is the
sample weight. In general, this method results in relatively poor sample recoveries as a result
of remaining sample adhering to the aluminum vessel or loss resulting from static electricity.
Errors in sample weight affect element concentrations determined by isotope dilution but will
not compromise the isotopic composition of the sample.
Charlier et al. (2006) developed a more precise method of weighing micromilled samples.
These authors used a small receptacle made of gold foil that was weighed, loaded with sample,
dried, re-weighed, and placed into the Teflon digestion beaker along with digestion acids. The
gold foil was impervious to digestion acids and was removed and re-weighed after digestion
where the difference in weight was the sample weight. Similarly, Ramos et al. (2005) used pre-
Inter- and Intracrystalline Isotopic Disequilibria 409
(a) Sherline microdrill.
Petrographic microscope with
micrometer stage attached.
Microdrill with micrometer stage Microdrill with manual Z-direction control and
attached. plate for micrometer attachment.
(b) New Wave Instruments Micromill.
(a) Video camera
(b)
Moveable
binocular zoom
microscope
head
(c)
High speed
mill
Microscope
optical axis
X-Y offset
X-Y stage
Figure 2. Illustrations of different milling devices. a) Sherline™ vertical mill shown with removable
Figure 2.
micrometer stage attached to petrographic microscope and to the mill. The sample is secured in a removable
upper portion of the micrometer stage. This allows the operator to maintain a centered stage while being
able to remove the sample for washing between milling events. The upper portion can then be re-attached to
the stage in the same orientation and settings. b) Images illustrating various components of the NewWave™
MicroMill™ including the moveable stage, binocular microscope and video camera, tungsten carbide mill
bit in sample collection water slurry, and illustrations of how the sample is pipetted into a beaker for
chemical analysis. [Used by permission of Elsevier from Charlier et al. (2006), Chemical Geology, Vol.
232, Fig. 1, p. 117].
410 Ramos & Tepley
weighed, pre-cleaned Teflon disks on which samples were dried and re-weighed. The samples
plus Teflon disks were placed into digestion beakers. The advantage of these methods was
that the user could theoretically recover the entire dried sample for a fairly accurate sample
weight.
Samples sizes vary according to bit diameters and milling depths and, most importantly,
are controlled by how much material is needed for high precision analysis. Because samples
are generally small, additional precautions must be taken to ensure cleanliness of beakers and
reagents used in the dissolution process. Using multiply-distilled reagents and freshly distilled
acids during dissolution is recommended. The amount of reagents used in the digestion
process is limited by 1) the minimal amount of reagent used to maintain low blanks, and 2) the
minimum amount of reagents required to prevent samples from drying during the digestion
process. If too small a volume of reagent is used, heating will vaporize the reagent and require
an extra cooling stage for re-condensing the sample or require re-digestion. Approximately
0.3-0.5 ml of reagent is minimally required to meet these conditions. Typical beakers used
in this process are Savillex™ 3 ml Teflon screw-top beakers which are cleaned in a thorough
fashion. Further sample processing and column chromatography is described in the Single
Crystal Analyses section below.
Standard dissolution techniques for most rock digestions can also be used for micromilled
samples: HF and concentrated HNO3 for the first stage, 7N HNO3 for the second stage, and 6N
HCl for the third stage (Ramos 1992). Samples can then be re-dissolved in the acid required
for chromatography. Samples are generally spiked prior to digestion unless sample splitting
is required. In general, splitting while samples are dissolved in 6N HCl is best. Mixed spikes
(e.g., Rb/Sr, Sm/Nd) are typically used to ensure that critical element ratios are preserved
even if weighing errors have occurred. Required spike amounts can be calculated using a
spreadsheet and estimated concentrations of elements of interest in the sample.
Applications to volcanic rocks. Whole-rock isotopic measurements are important in
understanding broad petrogenetic histories of volcanic systems. However, whole-rock sampling
produces average chemical and isotopic characteristics that either obscure or completely
eliminate potential variations in mineral and melt components of rocks both isotopically and
geochemically. Micromilling techniques were developed to extract isotopic and geochemical
traits that whole-rock processing and whole-crystal analyses cannot obtain, thus allowing
researchers to more thoroughly constrain the petrogenetic histories of volcanic and plutonic
systems. Here, we review micromilling studies that have aided and, in some cases, complicated
our understanding of crustal-level volcanic processes.
Micromilling, sample processing, and undertaking TIMS analyses are time consuming,
therefore a thorough sampling methodology is required prior to initiating studies. Micromilling
can be combined with a range of additional techniques including Nomarski Differential
Interference Contrast (NDIC), a reflected light technique (Anderson 1983; Pearce and Clark
1989) used to enhance compositionally dependent textural features, and electron microprobe
analysis to elucidate potential compositional variations in target phenocrysts. The combined
use of Nomarski interferometry, electron microprobe analysis, and Sr-isotopic microanalysis
to provide textural, chemical and isotopic information is termed Crystal Isotope Stratigraphy
(CIS; Davidson and Tepley 1997; Davidson et al. 1998). These techniques help define a more
thorough sampling strategy and, once accomplished, provide information that would have
otherwise been lost as a result of the averaging effects of whole-rock techniques.
One of the first studies to employ these techniques was that of Tepley et al. (2000) who
investigated the 1982 eruption of El Chichón Volcano in Mexico. Tilling and Arth (1994) had
previously found that mineral separates (plagioclase, clinopyroxene, hornblende, anhydrite,
apatite and groundmass) from the same hand sample of the 1982 eruption were isotopically
heterogeneous. Tepley et al. (2000) found texturally complex plagioclase phenocrysts
Inter- and Intracrystalline Isotopic Disequilibria 411
characterized by multiple dissolution zones, large variations in An contents associated with
dissolution surfaces, and monotonic decreases in 87Sr/86Sr from core to rim (Fig. 3). They
interpreted these data as reflecting crystal growth in an increasingly contaminated magma
reservoir that was recharged with similar trachyandesitic magma with lower 87Sr/86Sr. The
lower 87Sr/86Sr recharge magma presumably fluxed the reservoir with hotter and more
volatile-saturated melt causing textural discontinuities in the plagioclase, with consequent
large magnitude changes in An contents, and eventual eruption. Decreasing 87Sr/86Sr ratios in
plagioclase resulted from crystal growth following re-equilibration in newly mixed magma.
Thus, utilizing crystal isotope stratigraphy, Tepley et al. (2000) were able to elucidate a more
detailed picture of the petrogenetic evolution of El Chichón Volcano that otherwise would have
been overlooked as a result of whole-rock chemical and isotope averaging.
Numerous other studies have used crystal isotope stratigraphy to evaluate the petrogenetic
histories of a range of different volcanic rocks. Tepley et al. (1999) investigated mixing
0.7054
0.7052
Sr/86Sr 0.7050
Sr
86
0.7048
Sr/
87
87
0.7046
0.7044
0.7042
core rim
core to rim
0.7054
0.7052
Sr/86Sr
0.7050
0.7048
87
0.7046
0.7044
core rim
0.7042
core to rim
Figure 3. Cartoon illustrating two plagioclase feldspar phenocrysts from the 1982 eruption of El Chichón
volcano (Mexico). The drawings show Figure 3. zones (stippled areas), dissolution boundary (dashed
dissolution
line), micromilling holes, and electron microprobe traverse locations. Graphs to the right show changing
87
Sr/86Sr ratios from cores to rims for each crystal with filled rectangles representing results from individual
micromill holes. In both cases, 87Sr/86Sr ratios decrease from cores to rims resulting from crystal growth in
magma experiencing recharge with a lower 87Sr/86Sr melt. [Used by permission of Oxford University Press,
from Tepley et al. (2000), Journal of Petrology, Vol. 41, Fig. 3, p. 1402].
412 Ramos & Tepley
relationships and mafic inclusion disaggregation at Chaos Crags, a series of rhyodacite and
dacite domes containing variably quenched basaltic andesite inclusions on the flank of Lassen
Peak in Lassen Volcanic National Park, California (USA). The principal findings from this
study were that three styles of mixing occurred in the system: 1) limited mixing at the interface
between rhyodacite magma and a ponded basalt layer to form a basaltic andesite, 2) mechanical
dispersal of blobs of this material into the rhyodacite magma, and 3) disaggregation of the
inclusions through shear forces in the resident magma. In the initial mixing episode, rhyodacitic
magma containing large plagioclase phenocysts mixed into the basalt. These plagioclase
phenocrysts reacted in the hotter and more mafic liquid and formed reaction textures and upon
cooling, grew rims in equilibrium with the predominant magma composition. These rims had
higher An contents and lower 87Sr/86Sr ratios than cores. Disaggregation of the basaltic andesite
inclusions dispersed the contents of the inclusions into the more felsic magma in a later mixing
event. In this study, the complex pathways of mixing were easily discerned through combined
textural analyses, crystal chemistry variations, and isotopic analyses using micromilling. It was
suggested that these types of mixing processes were more common than previously thought and
that they may be responsible for the creation of intermediate lavas in arc environments.
Hora (2003) and Davidson et al. (2007) also documented the complex mixing of magmas
and movement of crystals within a volcanic system through use of single crystal isotope analyses
and intra-crystalline micromilling and isotopic analyses at Ngauruhoe Volcano (New zealand).
Micromilled plagioclase crystals in basaltic andesite lavas showed increases in 87Sr/86Sr from
cores to rims which were consistent with crystals growing in magmas that became progressively
more contaminated with higher 87Sr/86Sr melts. Accompanying clinopyroxene phenocrysts with
low 87Sr/86Sr ratios suggested that they crystallized in a low 87Sr/86Sr melt prior to plagioclase
crystallization. Subsequent remobilization, prior to eruption, combined the two minerals in
later-erupted volcanic rocks. Just as in Tepley et al. (1999), micromilling allowed researchers
to document changes in isotopic values within crystals and, when integrated with textural
and compositional information and single crystal isotopic analyses, to trace the pathways of
interaction of magmas in evolving volcanic systems.
Similar magma/crystal dynamics were illustrated by Perini et al. (2003). These authors
showed that potassium feldspar megacrysts hosted in mafic alkaline potassic, ultrapotassic,
and differentiated rocks retained both initial Sr isotopic homogeneity and heterogeneity and
could be seen in eruptive products at nearby volcanoes in central Italy. At Monte Cimino,
isotopic results from micromilling potassium feldspar megacryst cores and rims suggested that
they nucleated and grew in a trachytic magma but were subsequently mixed into latite and
olivine latite melts. At Vico volcano, micromilling and isotopic analyses of potassium feldspar
megacrysts and isotopic analyses of whole-rock hosts showed a complex relationship. Three
scenarios were observed; 1) a megacryst was isotopically homogenous from core to rim and in
isotopic equilibrium with its host trachyte; 2) a megacryst was isotopically heterogeneous from
core to rim and not in isotopic equilibrium with its host tephri-phonolite; and 3) a megacryst
was isotopically homogenous from core to rim but not in isotopic equilibrium with its host
olivine latite. As a result, Perini et al. (2003) created a complex model in which some crystals
were derived from an older trachytic melt similar in composition to the host trachytic melt,
some crystals were co-genetic with their current host, and some crystals were xenocrysts.
As these studies have illustrated, crystals commonly record complex textural, compositional,
and isotopic characteristics that result from volcanic systems that recycled previously crystallized
phenocrysts, commonly referred to as antecrysts, from crystal-rich mush piles genetically
related to the system but not actively crystallizing from the prevalent magma (e.g., Jerram and
Davidson 2007). An excellent example that illustrated these complex interactions was Armienti
et al. (2007). This study compared whole-rock 87Sr/86Sr ratios with those of rims and cores of
clinopyroxene “phenocrysts” and a hornblende megacryst from an historic eruption of Mount
Etna (Italy). Results showed that pyroxenes and hornblende were in isotopic disequilibrium
Inter- and Intracrystalline Isotopic Disequilibria 413
with their hosts, except for a very thin rim on a single clinopyroxene crystal that was in 87Sr/86Sr
equilibrium with the host rock. To further justify an antecrystic origin, these authors evaluated
crystal size distributions (CSDs) (see Armienti 2008 in this volume) of the mineral phases
hosted in lavas erupted in 2001. Clinopyroxenes with disequilibrium isotopic ratios were part
of a group of larger crystals that defined a separate trend on the CSD diagram which supported
a different growth history in comparison to smaller phenocrystic minerals in the lavas. Together,
evidence suggested that crystals in isotopic disequilibrium with their host rocks grew in a
different, but related, magma compared to the prevalent magma feeding the current eruption.
The authors suggested that these “disequilibrium” crystals originated from a cumulate mush
that existed at depth, and that some were entrained in recently erupted magma.
Chadwick et al. (2007) also applied crystal isotope stratigraphy to plagioclase-bearing
basaltic andesite lavas from Merapi volcano. They documented significant Sr isotope variations
in cognate plagioclase phenocrysts. Chadwick et al. (2007) attributed the isotopic variability
in phenocrysts to extensive magma-crust interaction in two forms: 1) direct incorporation of
high-An content, high-87Sr/86Sr plagioclase xenocrysts originally from a sedimentary protolith
which were present as crystal cores that were subsequently overgrown by plagioclase in
equilibrium with resident magma, and 2) plagioclase crystal growth in resident magma that
was progressively contaminated by calcareous and volcaniclastic xenoliths and carbonate
crust with concurrent crystal fractionation. Constraining the exchange of mass between
various reservoirs in this magmatic system through micromilling allowed Chadwick et al.
(2007) to speculate that incorporation of the proposed crustal material (limestone) may have
major effects on the volatile budget of the volcano. They suggested that this interaction, which
is more significant than previously thought, may have significant consequences on the eruptive
behavior at Merapi Volcano and other volcanoes built on carbonate platforms.
Lastly, Morgan et al. (2007) combined textural analysis in the form of crystal size
distributions (CSD) and microsampling techniques to understand magma supply and mixing
processes at Stromboli Volcano (Italy). Plagioclase phenocrysts from a 26,000 year-old sample
were targeted in which time and growth rate data were determined through CSD, and changing
isotopic information of the supply magma was determined through micromilling and TIMS.
The authors constructed a model in which isotopic micromilling results and respective sampling
positions were overlaid and then stretched and shifted to generate an isotopic timeline. The
end result was an isotopic evolution line that represented, by proxy, magma evolution of the
evolving Stromboli magmatic system. Combining the information, Morgan et al. (2007) were
able to determine the timescales of mixing, crystallization and contamination prior to eruption
of Stromboli. In this study, as with Waight et al. (2000a), it was shown that micromilling was
a powerful tool for determining magma chamber processes. However, when combined with
other process-related techniques or contrasting in situ isotopic analyses, the technique became
even more useful.
Applications to plutonic rocks. Unlike volcanic rocks, plutonic rocks cool slowly and
are thus affected by diffusive re-equilibration associated with long-term high-temperature
environments, or by variable isotopic exchange associated with hydrothermal fluids.
Additionally, plutonic age determinations and high parent/daughter ratios in potential target
isotopic systems can yield sources of error that must be evaluated. Discussions of these
problems, some of which are outlined here, can be found in Davidson et al. (2008) and
Davidson et al. (2005). Given these caveats, we discuss notable studies that apply micromilling
to plutonic rocks and host minerals that show a range of alteration from little or no diffusive
re-equilibration to full isotopic resetting.
One of the first to utilize intracrystalline isotopic micromilling on plutonic rock
components was Waight et al. (2000a) who determined initial 87Sr/86Sr and εNd(i) on cm-
scale feldspars in a 395 Ma S-type granite from the Wilson’s Promontory Batholith of the
414 Ramos & Tepley
Lachlan Fold Belt (SE Australia). This research tested whether isotope variations could be
preserved in plutonic rocks with obvious magma mixing/mingling textures that remained at
moderate temperatures for millions of years. Additionally, this study was an excellent plutonic
analogue to volcanic equivalents in which the interactions between different magmas and
crystals were explored (e.g., Tepley et al. 1999). Evidence for magma mingling included
numerous mafic microgranular enclaves hosted in a monzogranite, and feldspar megacrysts
in both the host and enclaves. Micromilling of megacrysts revealed significant decreases in
initial 87Sr/86Sr and complementary increases in εNd(i) from cores to rims in both alkali and
plagioclase feldspars (Fig. 4). High-temperature diffusive or hydrothermal re-equilibration
would effectively produce homogeneous isotope signatures rather than a complementary
divergent trend, therefore such processes could be ruled out. Thus, this research verified not
only that micromilling techniques could be applied to slow-cooling, old plutonic systems
when evaluated for the possibility of hydrothermal alteration, but also that application of these
techniques to petrogenetic problems could trace the movement of crystals between plutonic
systems with different isotope characteristics.
Not all plutonic systems are immune to alteration processes however. Waight et al. (2000b)
documented both initial Sr and Nd isotopic heterogeneity and homogeneity in plagioclase
megacrysts in comparison with their respective host magmas residing in enclaves in two I-type
plutons from the Lachlan Fold Belt (SE Australia). In the Swifts Creek pluton, 87Sr/86Sr and
143
Nd/144Nd isotope ratios of plagioclase megacrysts were distinct from their host enclaves
which confirmed results from earlier studies that concluded an admixed nature to the enclaves
(Eberz and Nicholls 1988; Eberz et al. 1990). In this case, Waight et al. (2000b) modeled
feldspars as having crystallized in a primitive granitoid which were then transferred to their
current host. In the Bridle Track pluton, 87Sr/86Sr ratios of feldspars, the enclaves in which
they resided, and the granitic host of the enclaves were indistinguishable. 143Nd/144Nd ratios of
the same crystals and hosts were slightly different which suggested a mingled magma origin,
as did field evidence, petrology and whole-rock data. Waight et al. (2000b) concluded that a
0.7110 4
3
0.7105
2
0.7100 1
initial 87Sr/86Sr
0.7095 0
-1
εNd
0.7090
-2
0.7085 -3
0.7080 -4
-5
0.7075 sample size -6
0.7070 -7
0 1 2 3 4
Distance (mm)
Figure 4. Sr and Nd isotope traverses from a potassium feldspar megacryst hosted in a mafic enclave from
Wilson’s Promontory Batholith (Australia). Unfilled squares represent initial 87Sr/86Sr ratios and filled
squares represent initial εNd valuesFigure 4. micromilled holes. Filled boxes represent initial isotope
for the same
range of host granite and unfilled boxes represent initial isotope ratio of enclave for 87Sr/86Sr (left) and εNd
(right), respectively. Modified after Waight et al. (2000a).
Inter- and Intracrystalline Isotopic Disequilibria 415
significant Sr isotope equilibration process occurred to homogenize Sr isotopes of megacrysts,
enclaves, and granitoid host, but was unable to homogenize Nd isotopes completely. In these
cases, the use of multiple isotope systems constrained the altered nature of the samples and
allowed for appropriate interpretation.
Siebel et al. (2005) determined that sub-solidus thermo-metamorphic or hydrothermal
alteration affected 87Sr/86Sr ratios in potassium feldspars residing in various Late Paleozoic
plutonic rocks of the Bavarian Forest. In some cases, potassium feldspars were homogenized
along with bulk groundmass via a post-crystallization fluid phase. In other cases, interaction
of post-magmatic hydrothermal fluids produced isotopic heterogeneities. In all cases, the
propagation of errors associated with assumed crystallization ages, parent/daughter ratio
measurements, and isotopic ratio measurements were evaluated to ensure accurate initial age
calculations.
In many of these studies, the use of micromilling allowed workers to follow the paths of
crystal migration and growth in various magmas. Halama et al. (2002) focused on plagioclase
megacrysts from gabbroic dikes in the Gardar Province (South Greenland) to trace the path
of plagioclase-rich mushes, postulated to migrate from the crust-mantle interface, through
the crust eventually ponding within the crust and forming massive anorthosites. Halama et al.
(2002) used major element zoning patterns, in situ trace element variations, and 87Sr/86Sr ratio
variations in the plagioclase megacrysts to formulate a model for anorthosite formation.
Another plutonic micromilling study that elucidated processes similar to magma
interaction processes in volcanic environments was Perlroth (2000), documented in Davidson
et al. (2005, 2008). This study explored Sr isotope contrasts in potassium feldspar megacrysts in
the ~400 Ma Shap Granite (NW England). The Shap Granite has abundant potassium feldspar
megacrysts and mafic inclusions, many of which contain potassium feldspar megacrysts rimmed
by plagioclase. A leading hypothesis was that potassium feldspar megacrysts grew in the Shap
Granite and were then transferred into a boundary layer mixture along an interface between
the two magmas which eventually produced mafic enclaves of a mixed nature. Thermal and
chemical equilibration of the potassium feldspar with the inclusion mafic magma allowed the
plagioclase rims to grow. There is some debate (Pidgeon and Aftalion 1978; Rundle 1992)
about the age of the Shap Granite, and this led to equivocation regarding age-corrected initial
Sr isotope ratios of samples micromilled from a potassium feldspar megacryst. However, if a
younger age were accepted for the Shap Granite, then the isotopic data became consistent with
megacryst growth in the granite followed by crystal transfer and rim re-equilibration within
mafic enclaves, thus mimicking processes involved in volcanic systems.
The significance of these plutonic studies is threefold: 1) original isotopic disparities in
minerals of old plutonic rocks can persist for extended time periods, and can be measured using
micromilling, however this is not always the case, 2) multiple isotopic systems (Sr, Nd, Pb) can
be used in a complementary fashion in some cases to rule out the effects of both slow cooling,
high-temperature alteration and/or hydrothermal isotopic resetting, and 3) interpretations
based on micromilling techniques require precise and accurate measurements of the age of
the plutonic body, the parent/daughter ratio and isotopic ratio of the sample in question, and
appropriate error propagation of all measurements.
In long-lived volcanic and plutonic systems, sub-solidus isotopic diffusive re-equilibration
is a process that warrants careful evaluation. For example, Gagnevin et al. (2005) not only
documented Sr isotopic zoning in potassium feldspar megacrysts in the 7 Ma Monte Capanne
monzogranite (Elba, Italy), but also evaluated whether diffusive re-equilibration may have
affected the system. Micromilling transects from core to rim of several potassium feldspar
megacrysts revealed high to low initial 87Sr/86Sr values. These observations suggested crystal
growth in a crustally contaminated magma with high initial 87Sr/86Sr ratios followed by growth
in magma recharged with mafic melt with lower initial 87Sr/86Sr. However, to test whether
416 Ramos & Tepley
changes in 87Sr/86Sr ratios were due to isotope re-equilibration, Gagnevin et al. (2005a)
modeled the isotopic data to determine timescales required for diffusive exchange. Results
showed that at least 20 m.y. were needed to produce the isotopic profiles of the megacrysts.
Thus, they considered this amount of time unrealistic given other published results for similar
granite crystallization histories.
Isotopic variations in plutonic minerals also allowed researchers to constrain cooling times
based on the lack of diffusive equilibration in minerals with intracrystalline isotope contrasts.
An example was the Rum Intrusion (NW Scotland) where Tepley and Davidson (2003)
demonstrated that select plagioclases crystallized in liquids contaminated by isotopically diverse
wallrocks whereas other plagioclases crystallized in homogeneous uncontaminated magmas.
Both types of crystals were then transported to the floor of the magma reservoir where they
cooled together. Preservation of isotopic heterogeneity within some of these crystals allowed
Tepley and Davidson (2003) to calculate a maximum amount of time that the Rum magma
chamber was held at high temperature based on the diffusivity of Sr in plagioclase feldspar.
Figure 5 illustrates how cooling rate (linear or exponential) affects the isotopic equilibration
of crystals of various lengths. Given that isotopic disequilibrium (regardless of the magnitude)
in the Rum plagioclases existed on length scales of 1-2 mm, only cooling rates below the 2
3
linear cooling: T=T 0-kt
-α t
exponential cooling: T=T e
0 -1
0.01°C yr
equilibration length (mm)
2
α = 0.00001
(τ ~ 69,000 yrs)
1/2
-1
0.05°C yr
1 -1
0.1°C yr
α = 0.0001
(τ ~ 6,900 yrs)
1/2
-1 α = 0.0005
1°C yr (τ1/2 ~ 1,400 yrs)
0
0 2000 4000 6000 8000 10000
time (years)
Figure 5. Diffusive re-equilibration models used to determine the effective scale of Sr exchange owing to
cooling from an initial temperature T0. Both linear and exponential models are considered. In the linear
cooling models, k represents linear cooling rates of 0.01 to 1°C yr−1 and is represented in the diagram
by solid lines. In the exponential cooling models, represented in the diagram by dashed lines, α is the
effective temperature decay constant and τ1/2 is the time taken for the temperature to drop from T0 to T0/2
(see Reddy et al. (1996) for mathematic formulations of these equations). Diffusive exchange is dependent
on plagioclase composition. In this case, the diagram illustrates labradorite (An50-70) at 1200°C. Since
all plagioclase crystals evaluated are ≤2 mm, only the modeled cooling rates within the shaded box are
acceptable solutions. Note that diffusive exchange effectively ceases when temperature cools to below
1000 °C. This is indicated by the solid dot on each cooling curve. [Used by permission of Springer-Verlag,
from Tepley and Davidson (2003), Contributions to Mineralogy and Petrology, Vol. 145, Fig. 11, p. 639].
Inter- and Intracrystalline Isotopic Disequilibria 417
mm equilibration length line (shaded area) were acceptable solutions to the cooling rate
determination. Ultimately, the modeling indicated that the cooling rate for the Rum Intrusion
was ~0.1 °C yr−1 or faster and Sr isotope equilibration essentially ceased at T = 1000 °C (circle
on cooling path line) after which there could be no further appreciable diffusion or addition to
the equilibration length.
The issue of in situ crystallization versus crystal transport was revisited by Davidson et al.
(2008) who documented isotopically distinct initial 87Sr/86Sr ratios in plagioclase and pyroxene
phenocrysts in the Dias Intrusion (Dry Valleys, Antarctica). This work evaluated whether
minerals crystallized in situ (in which case the assumption is that different crystals share the
same isotopic composition owing to growth in the same magma) or in magmas with different
isotopic compositions and were transported to their current hosts. Scatter in age-corrected
isotopic results suggested that plagioclase and pyroxene crystallized in different environments
and were transported to their current position in the Dias Intrusion.
As these studies demonstrate, the possibility of extracting petrogenetic information in
plutonic rock mineral phases through micromilling is possible, however, several issues must
be evaluated. As Davidson et al. (2008) point out, one issue is whether the system under study
can support mineral-scale isotopic heterogeneity. Small, upper crustal intrusions tend to cool
more quickly than large, high-temperature, slow-cooling magmas. The added time associated
with slow-cooling magmas tends to force isotopic equilibration. Another issue is confidence
in age corrections of the magmatic body. Measurement errors associated with the age of the
body and parent/daughter ratios (e.g., Rb/Sr, Sm/Nd) magnify resulting errors associated with
determining whether initial, mineral-scale isotopic heterogeneity existed. Lastly, effects owing
to post-emplacement alteration need to be evaluated. As Waight et al. (2000a,b) have shown,
using multiple isotopic systems may alleviate the uncertainty associated with data obtained
through singular isotopic system analyses.
Secondary ionization mass spectrometry (SIMS)
Technique. Microspatial measurements of Pb and Sr isotopes have been undertaken
using secondary ionization mass spectrometry (SIMS). SIMS based analyses can be targeted
to micron-length sample sites although analyses can vary widely in measurement precisions
as a result of relatively low ion beam intensities (e.g., compared to TIMS). For Pb isotope
measurements using SIMS, analyses typically focus on the more abundant 206Pb, 207Pb, and
208
Pb masses and are only minimally affected by potential interfering elements/molecules (e.g.,
HfSi+) as a result of the use of variable mass resolutions (e.g., 1800-2000 for feldspars and 3000
for glasses; Layne and Shimizu 1997) which still allow for high ion beam transmissions. For
melt inclusions, a 16O− ion beam and high resolution (~3500) is used to ensure minimal influence
of potential interfering elements at the Pb isotope masses (Saal et al. 1998, 2005). In addition,
typical spot sizes are ~20-30 microns. And although precisions are generally poor (0.2 to 1.2%),
Pb isotope ratios have proven to vary widely in target materials, such that SIMS analyses have
successfully identified large Pb isotope variations in, for example, melt inclusions.
Applications. Saal et al. (1998) first used SIMS to measure Pb isotope ratios in olivine-
hosted melt inclusions to identify an isotopically diverse suite of magmas involved in ocean
island volcanic plumbing systems that could not be resolved by conventional whole-rock
analyses. These authors found that melt inclusions retained a range of EMII and HIMU mantle
endmember components that spanned 50% of the worldwide Pb isotope variation observed
in ocean island basalts in only a few sampled basalts. This study also demonstrated that large
ranges of Pb isotope signatures were retained by olivine-hosted melt inclusions in individual
basalt flows. These signatures were only retained in melt inclusions because melt aggregation
had homogenized variable melt fractions prior to eruption. Saal et al. (2005) expanded this
investigation to include three endmember mantle compositions (HIMU, EMI, and EMII) and
presented a model in which melts of an ocean lithospheric component (i.e., Pacific MORB)
418 Ramos & Tepley
were thought to mix with wallrock components originating from either HIMU, EMI, and EMII
magmas as reflected by linear Pb isotope trends.
Kobayashi et al. (2004) also analyzed melt inclusions for Pb isotopes in conjunction with
lithium and boron isotopes at Hawaii and concluded that recycled materials from ancient sub-
ducted crust were incorporated into the mantle sources generating Hawaiian basaltic volcanism.
Results again confirmed that melt inclusions retained greater isotopic diversity, as compared to
whole rocks, and successfully demonstrated the utility of combining SIMS-measured Pb iso-
tope ratios with additional isotopic constraints to identify and constrain the chemical nature of
the sources contributing to ocean island volcanism. And although these studies focused on melt
inclusions, Layne and Shimizu (1997) demonstrated that Pb isotopes could also be successfully
measured in plagioclase, potassium feldspar, pyrite, and chalcopyrite using SIMS.
In addition to Pb isotopes, Weber et al. (2005) developed a procedure to measure Sr isotopes
in fish otoliths using the Sensitive High Resolution Ion Microprobe (SHRIMP). Compared to
laser ablation and TIMS, SHRIMP generated 87Sr/86Sr ratios had relatively poor precisions
(87Sr/86Sr ±0.001 to 0.0001) but even at these levels Bacon et al. (2004) effectively identified
the natal rivers of salmon originating from North American cratonic regions and western North
American Paleozoic accreted terrains. SIMS Sr isotope measurements usually had greater
amounts of interfering elements and molecules, however, and beam intensities were generally
lower than TIMS due the limited volume sampled.
Laser ablation sampling
Sr isotope technique. Christensen et al. (1995) first demonstrated the utility of measuring
Sr isotopes in carbonate and plagioclase using laser ablation (LA) sampling in conjunction with
multi-collector inductively coupled plasma mass spectrometry (MC-ICPMS). The advantages of
laser sampling were clear: samples were easily prepared and analyses could be obtained rapidly,
bypassing time-intensive digestion and purification procedures. Although true for some minerals
such as carbonate, accurate analyses of other materials still offer significant challenges. Since
the Christensen et al. (1995) evaluation, refinements in the utility and longevity of commercial
laser ablation systems and advances in multi-collector inductively coupled mass spectrometers
have combined to greatly expand the use of laser sampling associated with microspatial isotope
analyses. Further studies by Davidson et al. (2001) evaluated plagioclase glasses with variable
Rb/Sr ratios, while Waight et al. (2002), Bizzarro et al. (2003), and Schmidberger et al. (2003)
expanded the range of analyzed materials to include apatite, clinopyroxene, and sphene. Ramos
et al. (2004) also demonstrated that highly accurate Sr isotope measurements could be obtained
on less Sr enriched carbonates (≤1000 ppm Sr), plagioclases (≤750 ppm Sr), and clinopyroxenes
(≤50 ppm Sr), in addition to elementally complex materials such as basaltic and basaltic andesite
fine-grain groundmass (Figs. 6, 7 and 8). Additionally, accurate assessments of Sr isotopes using
laser ablation sampling of aragonite otoliths have greatly benefited ecological studies of fishes
(e.g., Thorrold and Shuttleworth 2000; Woodhead et al. 2005; Barnett-Johnson et al. 2005).
Two requirements are generally necessary for successful laser ablation isotope measure-
ments: 1) analyte abundances must be relatively high, although precisions associated with ma-
terials characterized by low analyte abundances may still be sufficient to address investigations
involving large dynamic isotope ranges, and 2) potential interfering elements and molecules
must be minimal or their respective impact on measured ratios must be accurately removed. If
high analytical precisions are not required (e.g., 87Sr/86Sr of ≥0.00004), laser sampling can be a
very efficient means of acquiring data from a time perspective, and thus, is highly advantageous
as compared to undertaking micromilling and purification using column chromatography.
Since laser ablation analyses involve direct sampling of unpurified materials, measurement
accuracies may be highly dependent on the presence of interfering elements, thus target
materials with low interfering element abundances are generally preferred. Such effects are
Inter- and Intracrystalline Isotopic Disequilibria 419
0 .7 0 4 7 0
0 .7 0 4 6 0
80µm pits
87Sr/86Sr
0 .7 0 4 5 0
0 .7 0 4 4 0
0 .7 0 4 3 0
80µm x 500µm troughs
0 .7 0 4 2 0
Rim Core
0 .7 0 4 1 0
0 5 10 15 20 25 30 35
0 .0 5 6 6 0
0 .0 5 6 4 0
84Sr/86Sr
0 .0 5 6 2 0
0 .0 5 6 0 0
Mean
0 .0 5 5 8 0 0.05627±0.00018 (2σ)
0 .0 5 5 6 0
0 5 10 15 20 25 30 35
Number of Analyses
Figure 6. Core to rim 87Sr/86Sr and 84Sr/86Sr variations of a single plagioclase crystal from Pisgah Crater,
California (USA) determined using laser ablation sampling. Gray filled squares and stippled field indicate
ratios measured using micromilling and TIMS. Thirty-four 87Sr/86Sr ratios measured using laser ablation
sampling (filled squares with 2σ error bars in upper diagram) using troughs and pits follow the trend defined
by micromilled samples and confirm variations resulting from assimilation of materials in crustal-level
magma chambers (Ramos and Reid 2005). In contrast to 87Sr/86Sr ratios, 84Sr/86Sr ratios are consistently
lower than TIMS determined values with a mean of 0.05627±0.00018. This may result from potential
interfering elements or molecules, or inaccurate Kr corrections. [Used by permission of Elsevier, from
Ramos et al. (2004), Chemical Geology, Vol. 211, Fig. 8, p. 149].
best exemplified in Sr isotope measurements. Sr isotope ratios can be influenced by a variety
of interfering elements including 87Rb, 84Kr, 86Kr, 168Er2+(mass/charge = 84), 168Yb2+, 172Yb2+,
174
Yb2+, and 176Yb2+ (Davidson et al. 2001; Ramos et al. 2004). Many potential molecular
species such as Ca dimers (e.g., 44Ca43Ca, Waight et al. 2002), Ca argides (e.g., 44Ca40Ar,
Woodhead et al. 2005), and Fe oxides (e.g., 56Fe16O2+, Schmidberger et al. 2003) may also
influence Sr isotope ratios (Fig. 9). In addition, isotopes such as 85Rb, which is monitored
for 87Rb corrections, may also be influenced by interfering elements such as 170Yb2+ and
170 2+
Er . Thus, even isotopes that are measured for interference corrections may be affected,
potentially compromising accurate interference corrections. For the VG Element AxiomTM
and Nu PlasmaTM mass spectrometers, Ca-dimers and Ca-argides pose significant challenges
to obtaining accurate Sr isotope measurements in Ca-rich minerals such as carbonate and
plagioclase. Careful interference correction procedures, however, have been successfully
used to obtain accurate Sr isotope measurements (e.g., Woodhead et al. 2005). In contrast,
Sr isotope measurements on Ca-doped NBS 987 standard solutions and Ca-rich minerals,
including carbonate and plagioclase, are only minimally affected (Hart et al. 2005) or not
affected (Ramos et al. 2004) by Ca dimers and Ca argides when using the ThermoFinnigan
NeptuneTM. Although it is unclear why Ca-dimers and Ca-argides are either present or absent,
differences in the design of plasma interfaces or the standard running parameters used (per
420 Ramos & Tepley
0 .7 0 5 6
0 .7 0 5 4 Grain 7
87Sr/86Sr
0 .7 0 5 2
0 .7 0 5 0
0 .7 0 4 8
0 .7 0 4 6
Rim Core Rim
0 .7 0 4 4
1 2 3 4 5
0 .7 0 5 6
0 .7 0 5 4
87Sr/86Sr
0 .7 0 5 2
0 .7 0 5 0
0 .7 0 4 8
Grain 6
0 .7 0 4 6
Rim Core Rim
0 .7 0 4 4
1 2 3 4 5
Number of Analyses
Figure 7. 87Sr/86Sr ratios of two single clinopyroxene crystals from a primitive Grand Canyon basalt (MgO
wt% of ~10%) from the western United States determined using laser ablation sampling. Filled squares
with 2σ error bars represent multiple laser ablation trough analyses and gray squares with 2σ error bars
represent micromilled samples from the same crystals. Stippled fields reflect 87Sr/86Sr ratios measured from
four separate single crystals using column chromatography and TIMS. 87Sr/86Sr ratios measured using
160x500 µm laser ablation troughs in grain 6 overlap those of single crystals (lower diagram) confirming
accuracy of measured 87Sr/86Sr ratios. Results for grain 7, however, reflect variable 87Sr/86Sr ratios that lie
outside the field of single crystals illustrating the capability of laser ablation sampling to identify within
crystal variations that are not detectable using single crystal analyses. [Used by permission of Elsevier,
from Ramos et al. (2004), Chemical Geology, Vol. 211, Fig. 9, p. 151].
machine/manufacturer) may result in variations in ionization potentials and in the interfering
molecules and ions present. And although not directly evaluated, similarities between Sr
isotope ratios obtained on plagioclase using laser sampling in conjunction with the GV
IsoProbeTM compared to those obtained using micromilling and TIMS (Davidson et al. 2001),
also allow for only a minimal influence of Ca dimers and Ca argides during analyses. Tables 1a
and 1b show two different collector configurations for Sr isotope measurements and relevant
first order interferences for the ThermoFinnigan NeptuneTM and Nu PlasmaTM, respectively.
Select interfering elements may be present during all analyses, Rb and Kr for instance, but
the presence of others may depend on the specific mass spectrometer being used. Great care
must be taken to identify and constrain all potential interferences. Ultimately, correcting for
interfering elements affecting Sr isotope masses, and relevant corrections, is both the most
challenging and most critical aspect of obtaining accurate Sr isotope ratios using laser ablation.
Thus, it is beneficial to confirm the accuracy of results using alternative, proven methods such
as micromilling/TIMS.
Inter- and Intracrystalline Isotopic Disequilibria 421
0 .7 0 5 6
0 .7 0 5 4
160µmx500µm
troughs
87Sr/86Sr
0 .7 0 5 2
0 .7 0 5 0
0 .7 0 4 8
0 .7 0 4 6
80µm pits
0 .7 0 4 4
1 2 3 4 5 6
0 .0 5 6 6 0
0 .0 5 6 4 0
84Sr/86Sr
0 .0 5 6 2 0
80µm pits
160µmx500µm
0 .0 5 6 0 0 troughs
0 .0 5 5 8 0
0 .0 5 5 6 0
1 2 3 4 5 6
Number of Analyses
Figure 8.
Figure 8. 87Sr/86Sr and 84Sr/86Sr variations of fine-grained Pisgah Crater (USA) basaltic groundmass. Filled
squares with 2σ error bars result from laser ablation sampling using 160×500 µm troughs and 80 µm
diameter pits. Gray squares and stippled field represent results from four separate micromilled holes of
fine-grained groundmass. 87Sr/86Sr results reflect variable errors but results overlap those from micromilled
samples and attest to the accuracy of 87Sr/86Sr obtained using laser ablation sampling. 84Sr/86Sr variations
deviate from micromilled samples analyzed using TIMS and partly reflect variations presumably resulting
from pit-induced fractionation effects. [Used by permission of Elsevier, from Ramos et al. (2004), Chemical
Geology, Vol. 211, Fig. 8, p. 149].
To offer a general starting point in which to undertake Sr isotope analyses using laser
ablation, Tables 2a and 2b are presented to illustrate machine and laser settings. These
parameters are used in conjunction with the ThermoFinnigan NeptuneTM/UP213 New WaveTM
laser and the Nu PlasmaTM/Lambda Physik Compex 110ARFTM excimer laser, respectively,
and will change depending on the specific mass spectrometer and laser being used. Many
of the previously mentioned studies utilize different mass spectrometers and lasers and will
include basic information for machine and laser settings. Additionally, the Faraday collector
configurations and major interfering elements described above can be used initially and refined
as necessary. Further evaluations should be undertaken to ensure that additional interferences
are not present. This can be accomplished in many ways, but Ramos et al. (2004) had success
in using NBS987 standard solutions doped with potential interferences such that specific
elements or molecules could be individually evaluated as potential interferences.
Once interferences are identified, a correction procedure must be designed and applied
to remove their impact from the desired ratios of interest. This may be complex. For the
ThermoFinnigan NeptuneTM, all Faraday collectors are used to measure the isotopes of Sr and
interfering elements and molecules. If more isotopes were needed to be measured, a peak-
hopping routine would be required. It is critical that ion intensities from at least one isotope per
interfering species is measured, independent of the effects of additional interfering elements
422 Ramos & Tepley
Figure 9. A) Mass scans at resolving power of 3000 illustrating the presence and magnitude of a range of
interfering elements on the VG Axiom MC-ICPMS at or near Sr and Rb masses of interest. B) Mass scan
at normal resolving power illustrating the presence and magnitude of a range of half-mass and whole-mass
interfering elements. [Used by permission of Elsevier, from Waight et al. (2002), International Journal of
Mass Spectrometry, Vol. 221, Fig. 4, p. 234].
or molecules. Ramos et al. (2004) offered a detailed description of the off-line correction
sequence utilized for Sr isotope analyses on the ThermoFinnigan NeptuneTM.
Firstly, Kr contributions are removed by measuring on-peak baselines during normal gas
flow through the ablation cell without firing the laser. Background Kr signals, present as a
result of Kr impurities in the He and Ar gas supplies, are directly measured as elevated ion
table 1a. Collector block configuration of the ThermoFinnigan Neptune™ MC-ICPMS
(modified after Ramos et al. 2004) with potential interferences illustrated.
Collector L4 L3 L2 L1 C h1 h2 h3 h4
Mass 83 83.5 84 85 85.5 86 86.5 87 88
83 167 84 85 171 86 173 87 88
Isotope of interest Kr11.5% Er2+ Sr0.56% Rb72.2% Yb2+ Sr9.86% Yb2+ Sr7.00% Sr82.6%
84 86 87
Isobaric interferences Kr57.0% Kr17.3% + Rb27.8%
166 168 170
Er2+ Er2+ Er2+ Er2+
168 170 172 174 176
Yb2+ Yb2+ Yb2+ Yb2+ Yb2+ Yb2+
table 1b. Collector block configuration of the Nu Plasma™ MC-ICPMS (modified
after Woodhead et al. 2005) with potential interferences illustrated.
Collector L4 L3 L2 L1 C h1 h2 h3 h4 h5 h6
Mass 82 83 84 85 86 87 88 89
82 83 84 85 86 87 88 89
Isotope of interest Kr11.5% Kr Sr0.56% Rb72.2% Sr9.86% Sr7.00% Sr82.6% Y100%
Inter- and Intracrystalline Isotopic Disequilibria
84 86 87
Isobaric interferences Kr57.0% Kr17.3% Rb27.8%
42 40 43 40 44 40 46 40 48
Argides Ca Ar Ca Ar Ca Ar Ca Ar Ca40Ar
42 43 44 46 48
Ca Dimers Ca40Ca Ca40Ca Ca40Ca Ca40Ca Ca40Ca
423
424 Ramos & Tepley
table 2a. Operating parameters of the ThermoFinnigan NeptuneTM MC-ICPMS and Nu
PlasmaTM MC-ICPMS (modified from Ramos et al. 2004 and Woodhead et al. 2005).
neptune nu Plasma
RF power 1200 W 1350 W
Argon cooling gas flow rate 15 L/min 13 L/min
Auxiliary gas flow rate 0.8 L/min 0.9 L/min
Interface cones Nickel Nickel
Acceleration voltage 10 kV 4 kV
Mass resolution 400 400
−9
Mass analyzer pressure 5-8×10 mbar 4×10−9 mbar
Detection system Nine Faraday collectors Eight Faraday collectors
Sampling mode Three or ten blocks of 10 × 8 s
integrations for laser and solution
analysis, respectively
Background/baseline 3 min on peak in 2.0% HNO3 on peak
Nebulizer Glass cyclonic spray chamber fitted ?
with a Micromist PFA nebulizer
Uptake mode Free aspiration Free aspiration
Sample uptake rate 50 µL/min ?
Typical sensitivity on 88Sr 55-60 V/ppm (10-11 ohm resistors)
Ar sample gas flow rate 0.60-1.05 L/min, optimized to 0.85 L/min, optimized to
maximize 88Sr signal maximize 88Sr signal
Beam dispersion (Dispersion quad) −25
table 2b. Operating parameters of UP213 New WaveTM and Helex/ Lambda Physik
CompexTM 110ArF excimer laser (modified from Ramos et al. 2004 and Woodhead et al. 2005).
Laser uP213 new WavetM Laser helex Lambda Physik
Compex 110arF excimer
Wavelength 213 nm 193 nm
Spot dimensions 160 µm × 500 µm trough or 80 µm 105 µm diameter spot
× 500 µm trough or 80 µm diameter
spot
Energy density ~7-10 J
Pulse rate 10 Hz 5 Hz
Carrier gas Helium Helium
He auxiliary gas flow rate 0.85 to 0.95 L/min, optimized every 0.2 L/min, optimized every run
(for analysis of Sr isotopes in run session session
plagioclase)
Inter- and Intracrystalline Isotopic Disequilibria 425
signals (compared to electronic baselines) at the 83, 84, and 86 mass positions and removed
during baseline correction. Typical signals were always less than 3 mV of 83Kr. Although this
methodology worked for Ramos et al. (2004), Hart et al. (2005) questioned whether on-peak
baselines accurately account for Kr contributions and instead used the 84Kr intensity corrected
for 84Sr contributions to calculate 86Kr contributions. Thus, as is typical for laser ablation
analyses, the user will have to evaluate which procedure works more effectively in this regard.
After Kr, REE2+ corrections are undertaken. The 167Er2+ and 171Yb2+ ion intensities can be
measured in the low 3 and center Faraday collectors, respectively, which are set at half-mass
positions (Tables 1a and 1b). From the 167Er2+ ion intensity, interfering 166Er2+, 168Er2+ and 170Er2+
intensities can be calculated using the 166Er2+/167Er2+, 168Er2+/167Er2+, and 170Er2+/167Er2+ ratios
measured from reference solutions. These contributions can then be mathematically removed
from the mass 83, 84, and 85 ion intensities prior to correcting for the next interfering element/
molecule. Yb interference corrections can then be pursued in the same fashion.
After REE2+, 87Rb corrections can then be undertaken. Ramos et al. (2004) applied a
procedure that used the interference corrected 85Rb intensity and the mass bias factor determined
from the interference corrected 86Sr/88Sr ratio to determine the 87Rb contribution to the ion
intensity at mass 87. In contrast, Hart et al. (2005), which intended to measure glasses with
significantly higher Rb/Sr ratios, determined that 87Rb contributions could not be accurately
assessed to the degree needed for samples with high Rb/Sr by using 86Sr/88Sr determined mass
bias corrections. As a result, these authors applied Rb corrections that incorporated 85Rb/87Rb
ratios determined on an external set of glasses in which respective 87Sr/86Sr ratios had previously
been measured.
Although a complete set of correction procedures are reviewed here, selected materials may
not require correction for all potential interferences, thus appropriate correction procedures will
need to be determined for each type of material analyzed. In addition, if analyses are undertaken
at poor vacuum conditions, peak-tailing corrections may also need to be applied.
After correcting for Rb, standard normalization of Sr isotopes to 86Sr/88Sr = 0.1194 can
be undertaken to correct for mass spectrometer mass bias/mass fractionation. Note that this
normalization is ~10 times larger than that for TIMS, thus inaccurate normalization can have
a much greater impact on resulting normalized Sr isotope ratios. Ultimately, this correction
procedure proved effective in generating accurate 87Sr/86Sr isotope ratios but only variably
accurate 84Sr/86Sr ratios in a broad array of target materials (Ramos et al. 2004). For analyses
using the Nu PlasmaTM, it is unclear how separation between Ca dimers, Ca argides, and REE2+
would be accomplished, and as of yet, successful Sr isotope analyses of REE-rich minerals such
as clinopyroxene are absent. For the GV IsoprobeTM, no detailed evaluation of interferences
has been undertaken probably because most are removed by use of a collision cell. It has been
demonstrated, however, that relatively accurate Sr isotope ratios can be obtained on REE-rich
minerals such as clinopyroxene (Schmidberger et al. 2003), so it is assumed that REE interfering
effects are generally minor. Vroon et al. (2008) offers a review of correction and reduction
procedures from a range of studies using laser ablation for further reference.
To analyze Sr isotopes using laser ablation, we suggest starting with previously determined
mass spectrometer settings, laser settings, and collector configurations. A program should
then be initiated to evaluate all potential elemental and molecular interferences and create
appropriate correction procedures. Once done, users should analyze “known” samples with
previously determined Sr isotope ratios. Modern marine carbonates are commonly used to
evaluate measurement accuracies as modern seawater 87Sr/86Sr is well-known. Overall however,
there is a dearth of reference materials for laser ablation applications, thus it is commonly
easiest to analyze potential target materials using micromilling, chromatography, and TIMS
or purified solution MC-ICPMS to obtain accurate Sr isotope ratios of reference materials.
Reproducing “known” results will give users a means in which to evaluate measurement
426 Ramos & Tepley
accuracy and reproducibility and an opportunity to test correction procedures for a range of
materials. Such tests should be done for each type of material to be analyzed (i.e., carbonate,
plagioclase, clinopyroxene, apatite, etc) and will be useful for evaluating at which concentration
levels potential interferences can be accurately corrected (e.g., Rb/Sr ≤ 0.05).
Sr isotope applications. Studies measuring 87Sr/86Sr ratios of minerals using laser ablation
sampling encompass a wide range of applications. Carbonates, for example, offer a generally
Sr-rich mineral with few interferences. For evaluating analytical accuracies and precisions
associated with carbonates, shells characterized by modern marine 87Sr/86Sr ratios are commonly
measured (e.g., Christensen et al. 1995; Outridge et al. 2002; Ramos et al. 2004; Woodhead
et al. 2005). Alternative methods include directly comparing 87Sr/86Sr ratios measured using
laser ablation sampling to results determined on purified samples, including freshwater otoliths,
acquired using micromilling and TIMS (Barnett-Johnson et al. 2005). For otolith studies, spot
and raster sampling have been used to document sequential changes in 87Sr/86Sr ratios that
correlate to fish movements from freshwater to marine environments (Fig. 10; e.g., Woodhead
et al. 2005), between freshwater environments (e.g., Outridge et al. 2002), and within natal
rivers of origin (e.g., Barnett-Johnson et al. 2008).
Studies identifying the natal rivers of fishes include Atlantic croaker (Thorrold and
Shuttleworth 2000), tropical shad (Milton and Chenery 2003), barramundi (McCulloch et
al. 2005), and California salmon (Barnett-Johnson et al. 2005). Barnett-Johnson et al. (2008)
also correlated the spatial distribution of 87Sr/86Sr ratios with distinctive microtextural features
indicative of wild or hatchery origins (Barnett-Johnson et al. 2007) to evaluate wild and hatchery
salmon population dynamics in California offshore marine fisheries. This study demonstrated
that natal rivers associated with both wild and hatchery fish could be accurately determined
using 87Sr/86Sr ratios measured by laser ablation sampling. The combination of otolith studies to
date offers a foundation in which individual otolith growth rings can be used to estimate times
associated with fish movements through different freshwater environments prior to outmigration
to marine environments. These time constraints may aid in evaluating ecological and biological
factors to ensure the greatest chance for successful marine migration and future reproduction of
endangered anadromous fish species.
In contrast to otoliths, Bizzarro et al. (2003) focused on analyses of magmatic apatite and
carbonate phenocrysts to identify open-system processes affecting alkaline igneous rocks of
the Sarfartoq province (Greenland). This study successfully combined variable textural and
mineral 87Sr/86Sr isotopic compositions, and concluded that similar 87Sr/86Sr variations in both
apatite and carbonate crystals result from coeval crystallization occurring during continuous
modification of the host magma as a result of magmatic recharge. Magmas acquired variable
87
Sr/86Sr ratios that originated from melts added to the magma reservoir from a range of
subcontinental lithospheric- and plume-related carbonate-rich mantle components. Similar
processes occurred throughout the history of the Sarfartoq province.
For plagioclase, early studies focused on identifying and tracking open-system processes
reflected in 87Sr/86Sr variations retained in phenocrysts from basalts at Long Valley, California
(USA) (Christensen et al. 1995) and El Chichón (Mexico) (Davidson et al. 2001). Christensen
et al. (1995) used 87Sr/86Sr ratios acquired by laser ablation sampling to demonstrate that
plagioclase phenocrysts, which otherwise were characterized by uniform major element
compositions, retained variable 87Sr/86Sr ratios resulting from magma mixing associated with
post-caldera basaltic magmatism at Long Valley. Additionally, Davidson et al. (2001) used laser
rastering to measure 87Sr/86Sr profiles along traverses across plagioclase phenocrysts from El
Chichón, an arc-related volcano. These traverses were shown to be generally symmetric, were
correlated to textural features revealed by Nomarsky imaging, and demonstrated differential
rim and core 87Sr/86Sr signatures confirmed by micromilling and TIMS. El Chichón variations
resulted from sequential growth of plagioclase cores in magma with higher 87Sr/86Sr and growth
Inter- and Intracrystalline Isotopic Disequilibria 427
Figure 10. Sr ionFigure 1 0 . 84Sr/86Sr, and 87Sr/86Sr transect profiles of an otolith. Profiles reflect
beam intensity,
marine (0.70916) and freshwater (0.70653) 87Sr/86Sr signatures. Such profiles can be used to identify the
87
Sr/86Sr signatures of natal rivers and track fish outmigration to marine environments, inmigration to
freshwater environments, or migration between freshwater environments. [Reproduced by permission of
The Royal Society of Chemistry, from Woodhead et al. (2005), Journal of Analytical Atomic Spectrometry,
Vol. 20, Fig. 2, p. 26].
428 Ramos & Tepley
of rims in magma with lower 87Sr/86Sr. For both studies, not only were the effects of open-system
processes such as magma mixing identified, but the isotope compositions of the different mixed
magmas were constrained.
In contrast to smaller volcanic systems, 87Sr/86Sr ratios of plagioclase obtained using
laser ablation sampling were also used to 1) confirm the presence of a wide range of 87Sr/86Sr
variations in all major formational units of flood basalts erupted on the Columbia Plateau
(USA) and 2) constrain the processes associated with continental, plume-related volcanism.
Early-erupted Imnaha basalts, previously thought to result from melting of Yellowstone
plume mantle without significant external inputs, exhibited a wide range of 87Sr/86Sr ratios
in plagioclase phenocrysts while host basalts exhibited only subtle variations (Eckberg et al.
2006). The overall range of 87Sr/86Sr variations, best observed in plagioclase, originated from
plume-related magma interacting with either accreted fluid-fluxed subcontinental mantle/crust
or cratonic crust encountered during ascent (Wolff et al. 2008). Where magmas interacted with
fluid-fluxed materials, plagioclase 87Sr/86Sr ratios were lower than plume-related signatures.
Where magmas interacted with cratonic crust, presumably while residing in crustal chambers,
plagioclase crystals retained highly elevated 87Sr/86Sr ratios.
Further plume-related magmatism on the Columbia River plateau generated Picture Gorge
basalts and the volumetrically immense Grande Ronde basalts. Both whole-rocks and plagioclase
crystals from these formations retained wider ranging 87Sr/86Sr signatures in comparison to
earlier-erupted Imnaha basalts. These variations resulted from more extensive interactions with
either fluid-fluxed accreted materials (Picture Gorge basalts) or greater amounts of assimilation
of cratonic crust (Grande Ronde basalts) and reflected the same processes affecting Imnaha
basalts, but on a massive scale. Interestingly, laser ablation analyses of plagioclase crystals were
key in identifying the processes that influenced isotopic and trace element signatures of early
erupted magmas, processes that would greatly influence the geochemical and isotopic nature
of future eruptions of large-volume flood basalts such as those comprising the Grande Ronde
formation.
In addition to tracing the sources and processes generating Columbia River flood basalts,
87
Sr/86Sr ratios measured using laser ablation sampling were also used to evaluate crystal
residence times for all major Columbia River basalt formations (Ramos et al. 2005). Results
indicated that most plagioclase phenocrysts, and where available clinopyroxene, reflected 1)
disequilibrium with host melts and 2) phenocryst incorporation into large-volume flood basalt
magmas at varied times prior to eruption, some less than five years prior (Fig. 11).
Open-system processes influencing the products of the 1915 Lassen volcano (USA)
eruption were also evaluated by Salisbury et al. (2008). This study pursued an integrated
approach using crystal size distributions (CSDs), textural, and in situ analyses including 87Sr/86Sr
ratios measured by laser ablation sampling. Andesite- and dacite-hosted plagioclase crystals
with 87Rb/86Sr ratios as high as 0.20 were successfully analyzed. Sequential trough analyses
documented subtle core to rim 87Sr/86Sr variations that resulted from the complex interplay
of a range of dacite, andesite, and basaltic andesite magmas. Phenocrysts of plagioclase, with
uniform 87Sr/86Sr ratios, were thought to have experienced simple crystallization histories over
100s to 1000s of years in a dacitic magma chamber. Microphenocrysts, however, grew over
months in an environment in which andesite and basaltic andesite were mixed, while microlites
reflected equilibrium growth during ascent and eruption from the resulting hybrid magma. This
study exemplified the utility of integrating a range of chemical and physical parameters to
understand the histories of individual eruptive episodes of arc volcanoes.
For clinoproxene, Schmidberger et al. (2003) targeted individual crystals from a single
mantle xenolith and documented a large range in 87Sr/86Sr ratios. This range resulted, at least
partially, from the infiltration and reaction with kimberlitic host melt during magmatic ascent
but may have also been largely present before magmatism. If so, highly variable 87Sr/86Sr ratios
Inter- and Intracrystalline Isotopic Disequilibria 429
Imnaha Formation; An70 Grande Ronde Formation; An70
0.7045 0.7065
A B
TResidence=50 years
0.7040 0.7060
1150°C An75 TResidence=5 years (1100°C)
1100°C
An65
1050°C
0.7035 0.7055
-1.0 -0.5 0 0.5 1.0 -1.0 0 1.0
Core (mm) Core (mm)
Wanapum Formation Saddle Mountains Formation
Roza Member (South); An65 Elephant Mountain Member; An60
0.7135
0.7055 D
C
0.7130
0.7050 0.7125
0.7120
TResidence=250 years (1100°C)
TResidence=5 years (1100°C)
0.7045 0.7115
-2.0 -1.0 0 1.0 2.0 -1.0 -0.5 0 0.5 1.0
Core (mm) Core (mm)
Figure
Figure 11.
11. Selected plagioclase phenocryst 87Sr/86Sr profiles from Columbia River flood basalts (modi-
fied from Ramos et al. 2005) compared to host rock 87Sr/86Sr (gray fields). Phenocrysts in Columbia River
flood basalts generally reflect isotope disequilibrium with host magmas. Profiles are from A) early Imnaha
basalt, B) Grande Ronde basalt, C) Roza basalt, and D) late-erupted Saddle Mountain basalt. These profiles
are modeled using simple one-dimensional diffusion to determine maximum phenocryst residence ages.
Curves in A) reflect temperatures of 1150°C, 1100°C, and 1050°C (left side) and curves for An75 and An65
(right side). Dashed curves in B), C) and D) are for 1100°C and An75. Modeling suggests that plagioclase
phenocrysts reside in Columbia River flood basalts for only short times (<250 years), indicating that mineral
and melt components of flood basalt magmas were undergoing assembly immediately before or during
eruption.
present at the single xenolith scale would indicate direct evidence for mantle heterogeneity at
small scales.
Although not all-inclusive, these studies reflect the broad utility of 87Sr/86Sr isotope analyses
obtained using laser ablation samples in a microspatial isotope context. The potential for these
analyses to address new questions associated with biological and petrological fields is extensive
and future studies will undoubtedly use such analyses in even more creative ways.
Pb isotope technique. In contrast to Sr, Pb isotopes measured using laser ablation have
far fewer apparent interfering elements or molecules. By introducing thallium into the sample
stream via a nebulizing spray chamber prior to the sample entering the torch of the mass spec-
trometer, analytical precisions associated with Pb isotope ratios are increased substantially as a
result of mass fractionation correction using 203Tl/205Tl normalization (Horn et al. 2000). As a
result, Tl normalized Pb isotope ratios sampled using laser ablation attain precisions and accu-
racies that approach those of double-spike TIMS measurements although some variances have
been documented (Thirwall 2002). Alternatively, sample-standard bracketing can be effectively
used, commonly using NBS610 glass as a standard. Thus, laser ablation sampling offers an op-
portunity to acquire accurate and precise Pb isotope ratios in a highly time-efficient fashion.
Figure 12 illustrates results from six separate 50×250 µm troughs analyzed from a
single Valles caldera potassium feldspar crystal using the ThermoFinnigan NeptuneTM (Wolff
430 Ramos & Tepley
Valles caldera
pot assium f eldspar
Single Cryst al
LT: 1 Analysis
LT: 5
2 0 8 Pb/ 2 0 4 Pb: 3 7 .5 5 5 (5)
LT: 2
2 0 7 Pb/ 2 0 4 Pb: 1 5 .4 8 9 (2)
2 0 6 Pb/ 2 0 4 Pb: 1 7 .8 0 2 (2)
2 0 8 Pb/ 2 0 6 Pb: 2 .1 0 9 6 (2)
LT: 3
2 0 7 Pb/ 2 0 6 Pb: 0 .8 7 0 1 (1)
LT: 6
LT: 4
5 0 0 µm
Laser Trough: 1 ( LT: 1 ) 2 ( LT: 2 ) 3 ( LT: 3 ) 4 ( LT:4 ) 5 ( LT: 5 ) 6 ( LT: 6 )
2 0 8 Pb/ 2 0 4 Pb: 3 7 .5 9 ( 8 ) 3 7 .4 6 ( 5 ) 3 7 .5 6 ( 7 ) 3 7 .6 7 ( 8 ) 3 7 .5 3 ( 6 ) 3 7 .4 7 ( 7 )
2 0 7 Pb/ 2 0 4 Pb: 1 5 .5 0 ( 3 ) 1 5 .4 5 ( 2 ) 1 5 .4 9 ( 3 ) 1 5 .5 4 ( 3 ) 1 5 .4 8 ( 3 ) 1 5 .4 6 ( 3 )
2 0 6 Pb/ 2 0 4 Pb: 1 7 .8 1 ( 4 ) 1 7 .7 5 ( 2 ) 1 7 .8 0 ( 3 ) 1 7 .8 6 ( 4 ) 1 7 .7 9 ( 3 ) 1 7 .7 6 ( 3 )
2 0 8 Pb/ 2 0 6 Pb: 2 .1 0 9 9 ( 5 ) 2 .1 0 9 8 ( 4 ) 2 .1 0 9 4 ( 5 ) 2 .1 0 8 9 ( 7 ) 2 .1 0 9 8 ( 4 ) 2 .1 0 9 4 ( 4 )
2 0 7 Pb/ 2 0 6 Pb: 0 .8 7 0 3 ( 2 ) 0 .8 7 0 3 ( 2 ) 0 .8 7 0 0 ( 2 ) 0 .8 6 9 7 ( 5 ) 0 .8 7 0 1 ( 1 ) 0 .8 7 0 1 ( 1 )
2 0 8 Pb: 1 4 1 mV 1 5 7 mV 1 6 2 mV 1 4 2 mV 1 5 0 mV 1 7 2 mV
Figure
Figure 12. 1 2 . of a Valles caldera potassium feldspar crystal sampled for Pb isotopes using laser ablation.
Image
Six 250 µm troughs were sampled with results of each analysis illustrated (LT 1-6). After sampling, the
single crystal was dissolved, purified, and analyzed using the procedures of Wolff and Ramos (2003).
Results attest to the accuracy of Pb isotope ratios measured using laser ablation sampling in respect to
ratios of both the more abundant Pb masses and ratios normalized to the least abundant 204Pb mass.
and Ramos 2004). Analyses of isotope ratios involving the more abundant 206Pb, 207Pb, and
208
Pb isotopes average 2.1095 and 0.8701 for 208Pb/206Pb and 207Pb/206Pb respectively, with
all individual analyses lying within in-run error of these averages. To evaluate accuracy, the
crystal was then dissolved, Pb was purified using anion exchange chromatography, and the
sample was analyzed using Tl-doping following the procedures of Wolff and Ramos (2003).
Results for this crystal are 2.1096±0.00002 and 0.8701±0.0001 for 208Pb/206Pb and 207Pb/206Pb,
respectively. For potassium feldspar, laser and single crystal results yield identical 208Pb/206Pb
and 207Pb/206Pb ratios and attest to the accuracy of laser ablation, Tl-normalized Pb isotope
measurements.
To evaluate ratios normalized to less abundant 204Pb, 206Pb/204Pb, 207Pb/204Pb, and 208Pb/204Pb
average 17.80, 15.49, and 37.55 respectively for the same six laser ablation analyses. These
ratios are characterized by poorer measurement precisions, as reflected by two decimal place
precisions, as a result of much lower 204Pb ion intensities. And unlike larger peaks, 204Pb is
corrected for interfering 204Hg which is present in minor amounts presumably as impurities in
He or Ar carrier gases. In addition, very minor signals originate from the potassium feldspar
crystal itself, potentially near the crystal surface. To minimize such surface contributions for
both Hg and Pb, ~3µm of the surface was ablated along the targeted sampling area prior to
analysis. To correct for Hg, 202Hg is monitored and used to calculate 204Hg contributions using
the 202Hg/204Hg ratio measured in separate Hg solutions. Results for the single crystal analysis
are 17.802±0.002, 17.489±0.002, and 37.555±0.005 for 206Pb/204Pb, 207Pb/204Pb, and 208Pb/204Pb
respectively. Thus, laser ablation measurements relative to 204Pb have poorer overall analytical
precisions but still lie within in-run error of the single crystal results, again demonstrating
measurement accuracy for potassium feldspar.
Inter- and Intracrystalline Isotopic Disequilibria 431
In contrast to Sr, Pb contents of most magmatic materials are substantially lower. In this
case, the Pb content of the potassium feldspar is generally between 10-15 ppm, a concentration
level that still yields highly accurate and precise 208Pb/206Pb and 207Pb/206Pb ratios and highly
accurate and relatively precise 206Pb/204Pb, 207Pb/204Pb, and 208Pb/204Pb ratios. In contrast to the Tl
normalized analytical procedure described above, Simon et al. (2007) pursue a sample/standard
bracketing methodology that normalizes laser ablation results to nebulized NBS981 solutions
to correct for mass bias. This procedure is reserved, however, for measuring ratios involving the
larger 206Pb, 207Pb, and 208Pb isotopes from potassium feldspars and related glasses from Long
Valley rhyolites where interferences are presumed to be minor. Further studies by Kent (2008)
evaluate Pb isotope accuracy, reproducibility, and interference corrections using both Faraday
and ion multipliers, and sample/standard bracketing using standard/reference glasses.
Pb isotope applications. There are a range of studies that focus on Pb isotopes analyzed
using laser ablation. Below, selected studies are reviewed that either focus on uncommon
radiogenic materials or materials characterized by common Pb isotope signatures. Studies
targeting highly radiogenic materials such as zircon are not included. For example, Willigers et
al. (2002) obtained Pb ages using laser ablation sampling, including 204Pb results using an ion
multiplier, for apatite, monazite, and sphene. While sphene ages diverged from those determined
by TIMS, apatite and monazite ages were within error of TIMS results for equivalent mineral
separates. These authors attributed this sphene age discrepancy to the presence of interferences
across the Pb mass spectrum, although individual interferences were not identified.
For materials characterized by common Pb isotope ratios, Mathez and Waight (2003) used
laser ablation sampling to determine 207Pb/206Pb and 208Pb/206Pb ratios of sulfide and plagioclase
crystals to evaluate the evolution of the Bushveld Complex (South Africa), a large layered mafic
intrusion. These authors found Pb isotope disequilibrium between sulfides and plagioclase,
and between plagioclases from different layers of the intrusion. For sulfide-plagioclase pairs,
sulfides typically had lower 208Pb/206Pb and higher 207Pb/206Pb ratios than plagioclases that were
attributed to the addition of Pb from surrounding country rocks. Pb isotope variations between
plagioclase crystals from different layers of the mafic intrusion resulted from the inputs of
multiple sources of Pb during crystallization. Different Pb isotope ratios between sulfides and
plagioclase however, resulted from Pb addition to sulfides after temperatures dropped below
those needed for Pb exchange in plagioclase. And, these different Pb sources were thought
to have obliterated any potential isotopic signatures of the mantle sources associated with the
original magma.
Pb isotope ratios of Bishop Tuff-hosted potassium feldspars obtained by Simon et al. (2007)
were used to confirm that feldspar crystals grew from magmas associated with the Bishop Tuff,
not magmas related to precaldera glass mountain lavas. Similar to Wolff and Ramos (2003),
Simon et al. (2007) used Pb isotopes of potassium feldspar to discount the possibility that
the Bishop Tuff resulted from long-lived rhyolitic magmatism and instead concluded that the
Bishop Tuff more-likely resulted from the accumulation of many small, different magmas. In
addition, Pb isotope ratios defined linear trends between young Long Valley basalts and local
evolved crust undermining any potential for accurate age determinations using Rb/Sr isotopes.
In addition to igneous applications, Tyrrell et al. (2007) utilized the efficiency of laser
ablation sampling to analyze a large suite of potassium feldspar crystals to trace sediment
provenance associated with the breakup of Pangea. Potassium feldspar crystals in selected
Triassic and Jurassic sandstones were targeted to identify crystal populations based on Pb
isotopes. Both sandstones had two main populations of crystals with the Jurassic sandstone
also having one additional outlier population. Crystals from each population were found
together independent of facies and stratigraphic position. Isotope ratios from these distinct
populations were compared with potential source rocks found throughout the North Atlantic
region. Sources were then either discounted or included and paleodepositional directions were
432 Ramos & Tepley
determined for sediments contributing to these sandstones. In addition to constraining sediment
transport directions, dispersal distances were also determined, as well as identification of
different sediment sources associated with different periods of North Atlantic rifting.
Pb isotopes analyzed using laser ablation sampling also aided geologic structure-based
studies. Connelly and Thrane (2005) utilized laser ablation sampling to rapidly measure a large
suite of potassium feldspar crystals from currently adjoining early Proterozoic orthogneiss
terranes in west Greenland to evaluate a potential suture zone between them. These authors
focused on initial Pb isotope ratios of potassium feldspars to distinguish whether these terranes
shared similar magmatic sources and were thus related. Two populations of feldspars were
present, each with distinctive initial Pb isotope ratios, which suggested that these orthrogneiss
bodies did not originate from similar mantle/crustal sources. As a result, these authors
concluded that a cryptic suture must be present between the two west Greenland orthogneiss
terranes.
Overall, Pb isotopes sampled using laser ablation offer an accurate means in which to
acquire data in a time-efficient manner. Current applications attest to the potential utility of
Pb isotopes results measured using laser ablation sampling for a wide variety of scientific
applications in the earth sciences.
MICrOanaLyte anaLySIS
Single crystal analyses
Technique. Together with micromilling, single crystal isotope analyses target a limited
sample microspatially (i.e., a single crystal) but unlike micromilling, single crystals usually
contain low overall amounts of analyte because either the crystal is small or the crystal is
characterized by low analyte concentrations. In addition to single crystals, multi-crystal
samples may also be targeted using the same technique because many multi-crystal samples
(e.g., quartz) also have low analyte abundances. For Sr and Pb isotopes, olivine and quartz best
demonstrate the utility of single crystal analyses.
Olivine, (Mg,Fe)2SiO4, and quartz, SiO2, are found in a wide variety of igneous rocks
and have little tolerance for incorporating Sr or Pb into their respective mineral structures as a
result of elemental size and charge considerations. As a result, both typically have very low Sr
and Pb concentrations. In contrast, mineral or melt inclusions captured during olivine or quartz
growth may contain significant amounts of Sr and/or Pb. The generally small size of these
inclusions, however, may require large amounts of host mineral to obtain minimal analyte
abundances. For olivine, inclusions of basaltic melt can contain Sr concentrations ≥1000 ppm.
If, for example, melt inclusions account for 1% of the total weight of an olivine crystal and
contain 1000 ppm Sr, a 1 mg (0.001 g) single crystal would yield ~10 ng of Sr, an amount
easily analyzed using TIMS. If, however, the crystal contained only 0.1% melt inclusions, a
1 mg crystal would yield ~1 ng of Sr. A ~1 ng Sr sample can still be analyzed to a reasonable
precision (87Sr/86Sr ≤0.0001) using TIMS but uncertainty of whether the measured isotope
ratio reflects that of melt inclusions magnifies as the amount of Sr contamination resulting
from the digestion and purification process becomes a greater percentage of the total amount
of analyte analyzed. This “processing” contamination is measured as a total process blank. If
such blanks can be limited to ≤10 pg Sr, small (~1 ng) samples can be analyzed successfully
while maintaining blank levels of ≤1%. Maintaining low blank levels is the most critical
aspect of accurately analyzing single crystals, or small micromilled samples, characterized
by low analyte abundances. While Sr samples as small as 0.5 ng can be successfully analyzed
(e.g., Wolff et al. 1999), uncertainties in blank corrected ratios increase dramatically when
processing blanks exceed 1% of total analyte.
Inter- and Intracrystalline Isotopic Disequilibria 433
For quartz crystals from high-silica rhyolites, captured melt inclusions may have low
Sr but high Pb concentrations. Current studies targeting comendites and pantellerites from
Baitoushan volcano (Ramos et al. 2007, 2008) determine quartz-hosted melt inclusion Sr
concentrations of ≤1 ppm while Pb concentrations are ≥40 ppm. These quartz crystals are
characterized by high melt inclusion contents, typically >20%. For a 3 mg quartz crystal, the
Sr yield would be <0.6 ng while the Pb yield would be >24 ng. While analyzing 0.6 ng of Sr
is challenging, especially given that associated blank would have to be ≤6 pg to maintain a 1%
blank contribution, analyzing 24 ng of Pb is easily accomplished using solution MC-ICPMS in
conjunction with Tl-normalization (Wolff and Ramos 2003). Thus, single quartz crystals could
be analyzed for Pb isotopes but multiple crystals would be required for Sr isotope analyses.
To undertake low blank single crystal or micromilled Sr isotope measurements, a method
of column chromatography must be pursued which offers an opportunity to attain total process
blanks of generally less than ~20 pg. Two methods are prevalent: 1) a stripping method using Sr
specTM resin and 2) a chromatographic method using standard cation exchange resin. The first
uses a small volume of Sr specTM resin loaded onto ~1 ml shrink-fit Teflon columns with a frit
inserted at the column tip (e.g., Davidson et al. 1998; Charlier et al. 2006). This method removes
unwanted ions by flushing the resin with ~3N HNO3 and removing Sr with pure ultrafiltered
water. Total process blanks using this method can be <10 pg (e.g., Harlou et al. 2005).
During the Wolff et al. (1999) study, the presence of organic ion signals that appeared at
both Sr and Rb masses was identified in samples processed using batches of Sr specTM resin
made prior to 1998. As a result, pyrex columns with high aspect ratios (height/width >25)
were designed that use ~0.5 ml of 200-400 mesh cation exchange resin to undertake low
blank Sr chromatography. These small-volume, high-aspect ratio columns were fitted with
microporous frits and minimized blank while maximizing Rb/Sr chemical separations. The
procedure for Rb and Sr sequestration was essentially the same as larger cation exchange
columns commonly used for whole-rock purifications, but the resin and eluant volumes
were dramatically reduced. These small-volume columns were used to undertake column
purifications of quartz and potassium feldspar crystals (Wolff et al. 1999; Knesel et al. 1999)
and plagioclase, clinopyroxene, amphibole and olivine crystals (Ramos and Reid 2005) and
maintained blanks that were typically <20 pg and commonly <10 pg.
To obtain low blanks using either Sr specTM or cation exchange resins, the resins must
be extensively pre-cleaned prior to use and all acid components used in the digestion and
chromatography procedures must be fresh and doubly distilled. Normally these acids are
distilled the night before use and extracted directly from the distillery. In addition, 3 ml Teflon
digestion vessels are used and must be cleaned extensively. All should undergo cleaning in
Teflon bottles containing distilled aqua regia that sit on a warm hot plate for more than a week.
All pipette tips and bottles containing distilled acids should be cleaned prior to undertaking
chemistry. For pre-cleaning cation exchange resin, a ~20 ml disposable Bio Rad Econo-
columnTM is filled with resin and flushed with freshly distilled 2.5N HCl, 6.0N HCl, and
ultrafiltered H2O for at least two days prior to use. For Sr specTM resin, hot water is commonly
flushed through the resin in a similar fashion.
To determine the total amount of Sr in the samples to be analyzed, isotope dilution using
a high-purity Rb/Sr spike is required. Spikes should be added to samples and blanks prior to
undergoing digestion. A high Rb/Sr spike and a low Rb/Sr spike are commonly used depending
on the Rb/Sr ratio of the target material (e.g., plagioclase would require low Rb/Sr spike and
potassium feldspar would require high Rb/Sr spike). For crystals from high-silica rhyolites
(e.g., potassium feldspar and quartz), a spike with very high Rb/Sr may be required. Isotope
dilution results will 1) yield Rb and Sr concentrations and 87Rb/86Sr ratios for age dating or
age correction if required and 2) allow for the determination of total analyte abundances so
that samples can be stripped of blank contributions. Such contributions will be minor with
434 Ramos & Tepley
blanks that are <1% of the analyte but may become more significant when blanks exceed 1%.
For actual blank measurements, minimal amounts of spike should be used (commonly ~1 µL)
so that pre-spiked 87Sr/86Sr ratios of blank can be determined and samples can be corrected
for the effects of blank. It may also be useful to determine blank 87Sr/86Sr ratios from samples
collected from flushing greater amounts of acid (i.e., greater than used in normal column
chromatography) through columns without adding spike. Accurately constraining the 87Sr/86Sr
ratio of the blank will be critical for correcting samples in which blank exceeds 1% of the total
analyte.
For analyses of small samples, thermal ionization mass spectrometry (TIMS) is usually
preferred to MC-ICPMS. Loading of samples requires TaO2 and dilute phosphoric acid that
must also be devoid of Sr (see Charlier et al. 2006). Standard loading procedures include
adding 1 µL of TaO2 suspended in a solution of 5% nitric acid and 1 µL of 5% phosphoric acid.
Filaments for analyses should be made of rhenium and brought to temperatures that exceed
normal Sr analytical conditions and be manually checked for Sr and Rb signals in the mass
spectrometer prior to loading. Commonly, filaments have to set at high temperatures to allow
for all Sr and especially Rb to be depleted. Care should be used for analyzing small Sr samples
and NBS987 Sr standards should be analyzed at similar concentration levels to evaluate mass
spectrometer analytical accuracy and reproducibility at low analyte levels.
For Pb purifications, anion exchange resin and 1N HBr can be used in similarly small
columns as that used for Sr microchemistry. Resin again needs to be pre-cleaned prior to
use and HBr needs to be doubly distilled to maintain low blanks. Teflon cleaning procedures
should be followed as described for Sr and columns should be cleaned accordingly. Additional
chromatography procedures may also be used. Studies associated with single zircons have also
been successful in purifying Pb while minimizing blank, thus select procedures from these
studies could also be pursued. Ultimately, Pb isotope analyses should be undertaken using
either double-spiking and TIMS or MC-ICPMS. Alternatively, Tl-doping or standard sampling
bracketing can be used with MC-ICPMS.
Sr and Pb applications. Single crystal isotope analyses were used in a range of applications.
The earliest focused on a wide variety of minerals to identify and constrain the effects of open-
system processes and discriminate between competing models describing the petrogenesis
of S-type rhyolites from Italy (Feldstein et al. 1994). Two separate rhyolites were involved:
a more radiogenic 87Sr/86Sr endmember and a less radiogenic 87Sr/86Sr endmember. Single
crystal analyses included plagioclase, potassium feldspar, biotite, and cordierite, in addition to
glass. For plagioclase, textural evidence was also integrated with isotope results. Coarse and
medium, highly sieved plagioclase crystals with low initial 87Sr/86Sr ratios were determined
to be associated with clinopyroxene megacrysts and mafic enclaves, materials not directly
related to host rhyolites. All other plagioclase crystals had higher initial 87Sr/86Sr ratios and
were considered phenocrystic. These phenocrystic plagioclase 87Sr/86Sr ratios were generally
lower than those of phenocrystic potassium feldspar crystals but higher than accompanying
glass (assumed to reflect the melt component). In addition, individual biotite crystals retained
initial 87Sr/86Sr ratios that spanned most of the total observed range in both rhyolitic glasses
and crystals. Thus, a complex petrogenetic history was involved. Feldstein et al. (1994)
concluded that this history included crystallization of phenocrysts occurring at a mixing
interface between two layers of rhyolitic magma in which minerals grew in an environment
that offered a range of 87Sr/86Sr ratios. Additionally, diffusion modeling was used to constrain
possible crystal residence times and suggested that this process occurred relatively quickly
(<10,000 years) as much of the observed isotopic heterogeneity would have been eliminated
if time periods were longer. This study exemplified the use of single crystal isotope analyses
to evaluate the petrogenetic history of highly complicated rocks involving a range of melt and
mineral components derived from multiple sources.
Inter- and Intracrystalline Isotopic Disequilibria 435
More recent studies targeting large rhyolitic magma systems included Wolff et al. (1999)
and Knesel et al. (1999). These studies used Sr isotopes of single potassium feldspar crystals
to evaluate open-system processes affecting Bandelier Tuff and Taylor Creek (New Mexico,
USA) rhyolites. Wolff et al. (1999) focused on linear trends defined by potassium feldspar
crystals in the Otowi member of the Bandelier Tuff (Valles caldera). These trends could either
have been Rb/Sr isochrons or mixing lines (line “A” in Fig. 13 defined by potassium feldspars).
If isochronous, results suggested extensive residence of Bandelier high-silica rhyolitic magma
for ~270,000 years prior to eruption. Alternatively, linear trends could have also resulted from
two-component mixing without age significance.
To further discriminate between these possibilities, Wolff and Ramos (2003) evaluated
similar potassium feldspar crystals using single crystal Pb isotope analyses. Single crystal Pb
isotope results also defined a linear array that, if isochronous, yielded an age in excess of 10 Ma,
an age that far exceeded any reasonable magma residence time. Thus, Pb isotope variations must
have resulted from mixing. Sr was highly depleted (~3 ppm) in Valles magmas compared to Pb
(~40 ppm), therefore Sr isotopes must have been modified long before Pb isotopes given any
reasonable contaminant (e.g., local crustal lithologies), and both must then have resulted from
mixing not aging. Thus, results from combined single crystal Sr and Pb isotopes undermined
the possibility of long-lived magma residence at Valles caldera and called into question similar
assessments for other large rhyolitic magma systems. Additional work focused on 87Sr/86Sr
variations in glasses from glomerocrysts and melt inclusions in quartz. Overall, results suggested
mixing (Fig. 13) between three magmatic components: 1) magmas with crustal Sr and Pb isotope
signatures, 2) magmas with more-mantle like Sr and Pb isotope signatures, and 3) minor amounts
of local Proterozoic crust. Mixing between these components resulted in a complex array of Sr
and Pb isotope variations that may be typical for large rhyolitic magma systems.
Knesel et al. (1999) also focused on 87Sr/86Sr ratios of potassium feldspar crystals derived
by both single crystal and micromilling analyses. This study integrated crystal size parameters
from Taylor Creek rhyolites to constrain the evolutionary history of a single dome from another
0 .7 1 0 Local Crust Single Feldspar Cryst als
Quart z w/ Melt Inclusions
Single Feldspar f rom Glomerocryst s
Glasses f rom Glomerocryst s
0 .7 0 9
8 7 Sr/ 8 6 Sr ( at 1 .5 8 Ma)
0 .7 0 8
B
0 .7 0 7
0 .7 0 6
A
0 .7 0 5
0 200 400 600 800 1000
8 7Rb/ 8 6 Sr
Figure 13. Single crystal feldspar, quartz separates, and glass 87Rb/86Sr and 87Sr/86Sr results from pumice
and pumice hosted glomerocrysts of the Bandelier Tuff. Variations result from mixing 1) highly-evolved,
rhyolitic magmas with crustal isotope characteristics with 2) highly-evolved, magmas with more mantle-
like 87Rb/86Sr and 87Sr/86Sr characteristics (double-headed, bold arrow indicated by “A”). In addition,
crystals and glasses that diverge to the upper left of this diagram (dashed arrows indicated by “B”) originate
from melt, present as glass in glomerocrysts or melt inclusions in quartz, that has also incorporated local
Proterozoic crust after mixing (Wolff and Ramos 2004).
436 Ramos & Tepley
large rhyolitic magma system. Figure 14a shows that as individual crystal size increased, initial
87
Sr/86Sr values increased to a maximum and then decreased to a minimum. Interestingly, initial
87
Sr/86Sr profiles of variably sized sanidine crystals showed the same pattern, in this case core-to-
rim, of increasing to a maximum then decreasing to a minimum with the added observation that
the smallest crystals lacked the initial increasing trend (Fig. 13b). Knesel et al. (1999) postulated
that these trends resulting from crystal growth initially in a magma chamber contaminated
by incongruent melts of crustal components with higher 87Sr/86Sr (initial increasing 87Sr/86Sr
trend) followed by progressive recharge of a similar composition magma with lower 87Sr/86Sr
(decreasing 87Sr/86Sr trend). Larger, presumably older crystals found throughout the chamber
grew during contamination and magma recharge events and recorded sequentially increasing
and then decreasing initial 87Sr/86Sr ratios from core to rim, whereas smaller, presumably
younger crystals grew during the magma recharge phase only and recorded only decreasing
initial 87Sr/86Sr ratios. In this study, the combined use of micromilling, single crystal, and crystal
size techniques proved to be critical to identifying a complex petrogenetic history.
In contrast to successful studies focused primarily on single potassium feldspar crystals,
additional mineral and melt components were used by Ramos et al. (2007) to evaluate magma
generation processes at Baitoushan volcano (China/North Korea). At Baitoushan, minerals and
melt obtained from at least three recent (<10 ka) comendite and pantellerite eruptions reflected
disequilibrium at many levels. Figure 15 illustrates potassium feldspar and pumice fragment
87
Sr/86Sr variations found in highly-evolved, alkali-rich comendite and trachyte components
comprising the 980 AD eruption which emitted ~100 km3 of volcanic material. Initially the
(a) Whole crystal Sr-isotopic variations
0.717
0.716
inital 87Sr/86Sr
Figure 14. a) Initial 87Sr/86Sr versus
0.715 crystal size in sanidines from the Taylor
Creek Rhyolite complex. This diagram il-
0.714
lustrates increasing then decreasing initial
87
0.713
Sr/86Sr ratios in crystals with decreas-
ing size. Smaller crystals do not record
0.712 increasing initial 87Sr/86Sr ratios because
they reflect later growth in a magma
0.711 chamber that was initially contaminated
by wallrock melt with high 87Sr/86Sr and
1 2 3 4 5 6 7
then progressively fluxed with rhyolitic
Crystal size (mm) magma with low 87Sr/86Sr. b) Diagram of
changing 87Sr/86Sr values in three micro-
(b) Intracrystalline Sr-isotopic variations milled sanidine phenocrysts of different
0.718 sizes. The largest crystal showed increas-
ing 87Sr/86Sr near the core to a maximum
0.716 core 87
Sr/86Sr, and then decreasing 87Sr/86Sr
inital 87Sr/86Sr
from the middle of the crystal to the rim.
0.714
Smaller crystals recorded only portions
0.712 core of this trend owing to later growth in the
rims
magma chamber. [Used by permission of
0.710 Oxford University Press, from Knesel et
al. (1999), Journal of Petrology, Vol. 40,
7 mm crystal
0.708 5.5 mm crystal Fig. 4 and 7, p. 779 and 781].
core 3.8 mm crystal
0.706
-3.0 -2.5 -2.0 -1.5 -1.0 -0.5 0.0
Distance from crystal rim (mm)
Figure 14.
Inter- and Intracrystalline Isotopic Disequilibria 437
0 .7 0 5 3
Single Comendit e Pumice Fragment
Single Comendit e Feldspar Cryst al
Single Trachyt e Pumice Fragment
Single t rachyt e Feldspar Cryst al
0 .7 0 5 2
8 7 Sr/ 8 6 Sr 0 .7 0 5 1
0 .7 0 5 0
0 100 200 300 400
8 7Rb/ 8 6 Sr
Figure 15. Feldspar and pumice 87Sr/86Sr and 87Rb/86Sr results from the ~980 AD Baitoushan comendite
eruption (Ramos et al. 2007, 2008). The eruption involved two distinct magma components: 1) a early-
erupted, highly-evolved comendite and 2) a later-erupted, less-evolved trachyte. Unfilled squares and
diamonds represent results from individual comendite potassium feldspar crystals and pumice fragments,
respectively. Filled squares and diamonds represent results from individual trachyte potassium feldspar
crystals and pumice fragments, respectively. Comendite and trachyte crystals and pumices retain distinct
isotope characteristics. Trachyte potassium feldspar crystals define a linear trend (also in 87Sr/86Sr versus
1/Sr) that results from 2-component mixing that does not involve comendite as a mixing endmember.
Accompanying trachyte pumices are characterized by greater 87Sr/86Sr variations imposed after feldspar
fractionation. In contrast, comendites have potassium feldspar crystals that retain even greater 87Sr/86Sr
variations than trachytes that are both higher and lower that accompanying comendite pumice fragments.
Results reflect open-system magmatic processes most-likely involving a range of highly-evolved, variably
radiogenic magmas rather than variably enriched crustal materials.
eruption produced comendite, while the final stages were comprised of trachyte. For comendites,
potassium feldspar crystals retained 87Sr/86Sr ratios which were both higher and lower than
accompanying pumices while the reverse was seen in trachyte components where potassium
feldspar crystals retained a limited range in 87Sr/86Sr while trachyte pumices spanned most of
the range of comendite potassium feldspars. These relationships reflected the complex interplay
of highly-evolved magmas and most-likely resulted from the mixing of a range of melt and
mineral components which were not directly identified in the eruptive products.
In addition to the 980 AD eruption, a pantellerite was erupted from Baitoushan at ~0
AD. This pantellerite retained a range of mineral components including potassium feldspar,
clinopyroxene, and quartz. These different components retained a range of 87Sr/86Sr ratios that
were different than accompanying ratios of pumices (Fig. 16A). Melt, as reflected by pumice,
generally defined a ~400 ka isochron that suggested an extended magma residence age.
Mineral components, however, did not lie along this trend and reflected more consistent, less
radiogenic 87Sr/86Sr ratios. In addition, Pb isotope ratios of the same crystals defined negative
trends which undermined any age relevance for Rb/Sr isotopes and must have resulted from
the mixing of magmas with highly variable Pb isotope signatures that were potentially related
to the incorporation of local Archean lithologies assimilated at deep crustal levels. These
variations are different than those reflected in the 980 AD eruption and must have resulted
from either 1) a range of magmas whose presence was limited in time such that magmas with
similar signatures were not seen in later eruptions or 2) that a range of magmas existed but
did not communicate potentially as a result of the involvement of a complex magma plumbing
system that prevented potential interactions.
438 Ramos & Tepley
0 .7 1 1
A)
0 .7 1 0
0 .7 0 9
8 7 Sr/ 8 6 Sr
0 .7 0 8
0 .7 0 7
Pumice Fragment
0 .7 0 6
Feldspar
Clinopyroxene
0 .7 0 5 Quart z
0 .7 0 4
0 100 200 300 400 500 600
8 7 Rb/ 8 6 Sr
1 7 .9
B)
1 7 .8
2 0 6 Pb/ 2 0 4 Pb
1 7 .7
1 7 .6
Pumice Fragment
1 7 .5 Feldspar
Clinopyroxene
Quart z
1 7 .4
1 7 .3
0 0 .5 1 .0 1 .5 2 .0
2 3 8 U/ 2 0 4 Pb
Figure 16. Rb-Sr and U-Pb systematics of minerals and pumices from the ~0 AD Baitoushan pantellerite
eruption (Ramos et al. 2007). A) Illustrates differences in 87Sr/86Sr ratios of individual potassium
feldspar crystals, individual clinopyroxene crystals, and a quartz separate (~20 crystals) as compared to
accompanying pumice fragments. quartz has high 87Sr/86Sr and 87Rb/86Sr (>1000) and lies to the left of this
graph (as indicated by the arrow). Pumice fragments generally define a linear trend that, if isochronous,
suggests ~400,000 years of magma residence prior to eruption. Potassium feldspar and clinopyroxene
crystals, however, have relatively consistent 87Sr/86Sr ratios that are significantly lower than pumice
fragments. B) Pb isotope results from the same crystals (except quartz symbols represent two individual
crystals) that define a negative trend with all minerals and pumice fragments having variable signatures
that cannot result from extended magma residence. Results are consistent with Pb addition after the bulk
of crystallization of clinopyroxene and potassium feldspar, or the scavenging of crystals from unrelated
magmas by the magma represented by pantellerite pumice fragments.
In contrast to rhyolites, Ramos and Reid (2005) identified assimilation processes occurring
during basaltic magmagenesis at Pisgah Crater, California (USA) using single olivine crystals,
in addition to plagioclase, clinopyroxene, and amphibole. This study clearly demonstrated
that melt inclusions in olivine crystals from alkalic, continental basalts reflected both the 1)
greatest overall Sr isotope variations of all magmatic components present and 2) sequentially
increasing Sr isotope ratios in progressively erupted lavas. Results confirmed that basalts
acquired a range of isotope characteristics that resulted from the incorporation of materials
during residence at crustal depths (Glazner and Farmer 1992) and not as a result of the mixing
of mantle-derived magmas (Reiners 2002). Further work by Mickiewicz (2007) combined
crystal size distributions (CSDs) and 87Sr/86Sr ratios obtained by micromilling to evaluate the
crustal residence history and magma chamber dynamics associated with these basalts. This
work also clearly identified individual sieve-textured plagioclase xenocrysts which retained
highly radiogenic 87Sr/86Sr ratios that confirmed an origin involving crustal assimilation.
Inter- and Intracrystalline Isotopic Disequilibria 439
In contrast in igneous applications, Muller et al. (2000) used micromilling to mill out small
pieces of synkinematc minerals (not milling them directly) from mylonites and dykes from the
Eastern Alps. Using this technique, samples could be inspected prior to milling thick sections
to ensure exclusion of mineral inclusions that could potentially impact Rb and Sr budgets.
Although this study mainly targeted white micas that grew in pressure shadows of feldspar
and andalusite during deformation, potassium feldspar, plagioclase, biotite, and andalusite
were also included. Multiple white mica samples, and some additional minerals, were used
to generate isochrons dating 1) mica formation resulting from mylonitization affecting their
respective host rocks and 2) mica formation ages associated with small local dikes in the
area. Most mylonitization ages were similar to nearby intrusion ages, suggesting that most
white micas, and thus mylonitization, resulted from deformation due to pluton emplacement.
In addition, dikes yielded similar ages as both intrusion and mylonitization ages. Thus,
micromilling was successfully used to obtain portions of single crystals that required low
blank chromatography to obtain dates of deformation related to local intrusive rocks.
Overall, single crystal isotope studies have proven valuable in addressing questions
related to igneous and metamorphic petrogenesis and discriminating between potential
models describing how volcanic and metamorphic rocks and minerals obtain their chemical
and isotopic characteristics. Further studies will undoubtedly integrate even more chemical,
textural, and size parameters to better identify and constrain the effects of processes involved
in magmagenesis and metamorphism.
SuMMary
As reviewed above, the utility of microscale isotope sampling is only beginning to be
appreciated. Whether micromilling, single crystal analyses, or in situ analyses using SIMS
or laser ablation sampling, these techniques offer vast potential for understanding a range of
natural systems. Although this manuscript mainly focuses on applications targeting magmatic
minerals and melt components, investigations of biological materials have also benefited
greatly. As such, additional fields will also benefit from such isotope sampling techniques in
the future. As has been the case for the last ~20 years, advances in both sampling systems and
measuring technologies will continue, increasing the possibilities and range of opportunities
to apply multiple isotope systems in scientific applications. Ultimately, these techniques will
allow evaluations of natural systems in new and innovative ways, ways that are bound to
expand our understanding of natural processes in revolutionary ways.
aCKnOWLedGMentS
The ideas and techniques illustrated in this chapter have been developed over time by a
number of researchers and collaborators. We would like to express our gratitude to these people,
too numerous to list, who aided our understanding of systems reviewed in this manuscript.
For discussions about microanalytical topics, we would like to acknowledge J. Wolff, J. Gill,
J. Davidson, B. Charlier, P. Holden, D. Tollstrup, W. Bohrson, S. Hart, K. Knesel, and R.
Barnett-Johnson. In addition, J. Wolff, J. Gill, and A. Kent, allowed access to unpublished
data and S. Rodgers, A. Kinch, B. Shurtleff and R. Wilson assisted with analyses. We would
also like to thank K. Putirka, J. Wolff, and T. Waight for thorough reviews of the manuscript.
Investigations of and unpublished data for Baitoushan and Valles caldera high-silica rhyolites
were supported by National Science Foundation grants to Ramos (EAR-0538214) and Wolff
(EAR-001013).
440 Ramos & Tepley
reFerenCeS
Anderson AT (1983) Oscillatory zoning of plagioclase: Nomarski interference contrast microscopy of etched
polished thin sections. Am Mineral 68:125-129
Armienti P (2008) Decryption of igneous rock textures: crystal size distribution tools. Rev Mineral Geochem
69:623-649
Armienti P, Tonarini S, Innocenti F, D’Orazio M (2007) Mount Etna pyroxene as a tracer of petrogentic
processes and dynamics of the feeding system. In: Cenozoic Volcanism in the Mediterranean Area.
Beccaluva L, Bianchini G, Wilson M (eds) Geol Soc Am Spec Paper 418:265-276
Bacon CR, Weber PK, Larsen KA, Reisenbichler R, Fitzpatrick JA, Wooden JL (2004) Migration and rearing
histories of chinook salmon (Oncorhynchus tshawytscha) determined by ion microprobe Sr isotope and
Sr/Ca transects of otoliths. Can J Fish Aquat Sci 61:2425-2439
Barnett-Johnson R, Grimes CB, Royer CF, Donohoe C (2007) Identifying the contribution of wild or hatchery
Chinook salmon to the ocean fishery using otolith microstructure as natural tags. Can J Fish Aquat Sci
64:1-10
Barnett-Johnson R, Pearson, TE, Ramos FC, Grimes CB, MacFarlane RB (2008) Tracking natal origins of
salmon using isotopes, otoliths, and landscape geology. Limno Oceanograph 53:1633-1642
Barnett-Johnson R, Ramos FC, Grimes CB, MacFarlane RB (2005) Validation of Sr isotopes in otoliths by
laser ablation multicollector inductively coupled plasma mass spectrometry (LA-MC-ICPMS): opening
avenues in fisheries science applications. Can J Fish Aquat Sci 62:2425-2430
Bizzarro M, Baker JA, Haack, H (2004) Mg isotope evidence for contemporaneous formation of chondrules
and refractory inclusions. Nature 431: 275-278
Bizzarro M, Simonetti A, Stevenson RK, Kurszlaukis S (2003) In situ 87Sr/86Sr investigation of igneous apatites
and carbonates using laser-ablation MC-ICP-MS. Geochim Cosmochim Acta 67:289-302
Blundy JD, Shimizu N (1991) Trace element evidence for plagioclase recycling in calc-alkaline magmas. Earth
Planet Sci Lett 102:178-19
Chadwick JP, Troll VR, Ginibre C, Morgan D, Gertisser R, Waight TE, Davison JP (2007) Carbonate assimilation
at Merapi Volcano, Java, Indonesia: insights from crystal isotope stratigraphy. J Petrol 48:1793-1812
Charlier BLA, Ginibre C, Morgan D, Nowell GM, Pearson DG, Davidson JP, Ottley CJ (2006) Methods for the
microsampling and high-precision analysis of strontium and rubidium isotopes at single crystal scale of
petrological and geochronological applications. Chem Geol 232:114-133
Christensen JN, Halliday AN, Lee D, Hall CM (1995) In situ Sr isotopic analysis by laser ablation. Earth Planet
Sci Let 136:79-85
Connelly JN, Thrane K (2005) Rapid determination of Pb isotopes to define Precambrian allochthonous
domains: An example from west Greenland. Geology 33:953-956
Cortini M, van Calsteren PWC (1985) Lead isotope differences between whole-rock and phenocrysts in recent
lavas from southern Italy. Nature 314: 343-345
Davidson JP, Charlier BLA, Hora JM, Perlroth R (2005) Mineral isochrons and isotopic fingerprinting: pitfalls
and promises. Geology 33:29-32
Davidson JP, De Silva SL, Holden P, Halliday A (1990) Small-scale disequilibrium in a magmatic inclusion
and its more silicic host. J Geophys Res 95:17,661-17,675
Davidson JP, Font L, Charlier BLA, Tepley FJ III (2008) Mineral-scale Sr isotope variation in plutonic rocks-a
tool for unraveling the evolution of magma systems. Proc R Soc Edinb 97:357-367
Davidson JP, Morgan D, Charlier BLA, Harlou R, Hora JM (2007) Microsampling and isotopic analysis of
igneous rocks: Implications for the study of magmatic systems. Annu Rev Earth Sci 35:273-311
Davidson JP, Tepley FJ III (1997) Recharge in volcanic systems: Evidence from isotopic profiles of phenocysts.
Science 275:826-829
Davidson JP, Tepley FJ III, Knesel KM (1998) Isotopic fingerprinting may provide insights into evolution of
magmatic systems. EOS Trans Am Geophys Union 79:185, 189, 193
Davidson JP, Tepley FJ III, Palacz z, Meffan-Main S (2001) Magma recharge, contamination and residence
times revealed by in situ laser ablation isotopic analysis of feldspar in volcanic rocks. Earth Planet Sci
Let 184:427-442
Davis DW, Williams IS, Krogh TE (2003) Historical development of zircon geochronology. Rev Mineral
Geochem 53:145-181
Eberz GW, Nicholls IA (1988) Microgranitoid enclaves from Swifts Creek Pluton SE-Australia: textural
and physical constraints on the nature of magma mingling processes in the plutonic environment. Geol
Rundsch 77:713-736
Eberz GW, Nicholls IA, Maas R, McCulloch MT, Whitford DJ (1990) The Nd- and Sr-isotopic composition
of I-type microgranitoid enclaves and their host rocks from the Swifts Creek Pluton, southeast Australia.
Chem Geol 85:119-134
Inter- and Intracrystalline Isotopic Disequilibria 441
Eckberg A, Wolff JA, Ramos FC, Hart GL, Tollstrup DL (2006) The purity of the COSYMA in the Imnaha
basalt: Strontium isotope ratio variations in plagioclase phenocrysts. EOS Transactions 52: Abstract
V51D-1703
Feldstein SN, Halliday AN, Davies GR, Hall CM (1994) Isotope and chemical microsampling: Constraints on
the history of an S-type rhyolite, San Vincenzo Tuscany, Italy. Geochim Cosmochim Acta 58:943-958
Gagnevin D, Daly JS, Poli G, Morgan D (2005) Microchemical and Sr isotopic investigation of zoned
K-feldspar megacrysts: insights into the petrogenesis of a plutonic system and disequilibrium processes
during crystal growth. J Petrol 46:1689-1724
Geist DJ, Myers JD, Frost CD (1988) Megacryst-bulk rock isotopic disequilibrium as an indicator of the
contamination processes: The Edgecumbe Volcanic Field, SE Alaska. Contrib Mineral Petrol 99:105-
112
Ginibre C, Kronz A, Worner G (2002) High-resolution quantitative imaging of plagioclase compositions using
accumulated Back-Scattered Electron images: new constraints on oscillatory zoning. Contrib Mineral
Petrol 142:436-448
Ginibre C, Wörner G, Kronz A (2004) Structure and dynamics of the Laacher See magma chamber (Eifel,
Germany) from major and trace element zoning in sanidine: a catholuminescence and electron microprobe
study. J Petrol 45:2197-2123
Glazner AF, Farmer GL (1992) Production of isotopic variability in continental basalts by cryptic crustal
contamination. Science 55:72-74
Grove TH, Baker MB, Kinzler RJ (1984) Coupled CaAl-NaSi diffusion in plagioclase feldspar: experiments
and applications to cooling rate speedometry. Geochim Cosmochim Acta 48:2113-2121.
Halama R, Waight T, Markl G (2002) Geochemical and isotopic zoning patterns of plagioclase megacrysts in
gabbroic dykes from the Gardar Province, South Greenland: implications for crystallization processes in
anorthositic magmas. Contrib Mineral Petrol 144:109-127
Harlou R, Pearson DG, Nowel GM, Davidson JP, Kent AJR (2005) Sr studies of melt inclusions by TIMS.
Geochim Cosmochim Acta 69/10S:A380
Hart SR, Ball L, Jackson M (2005) Sr isotopes by laser ablation PIMMS: Applications to cpx from samoan
peridotite xenoliths. WHOI Plasma Facility Open File Technical Report 11
Hoppe KA, Koch PL, Carlson RW, Webb SD (1999) Tracking mammoth and mastodons: Reconstruction of
migratory behavior using strontium isotope ratios. Geology 27:439-442
Hora JM (2003) Magmatic differentiation processes at Ngauruhoe Volcano, New zealand: constraints from
chemical, isotopic and textural analysis of plagioclase crystal zoning. MSc Dissertation, University of
California, Los Angeles, Los Angeles, California
Horn I, Rudnick RL, McDonough WF (2000) Precise elemental and isotope ratio determination by simultaneous
solution nebulization and laser ablation-ICP-MS: application to U-Pb geochronology. Chem Geol
164:281-301
Ireland TR, Williams IS (2003) Considerations in zircon geochronology. Rev Mineral Geochem 53:215-241
Jerram DA, Davidson JP (2007) Frontiers in textural and microgeochemical analysis. Elements 3:235-238
Kennedy BP, Klaue A, Blum JD, Folt CL, Nislow KH (2002) Reconstructing the lives of fish using Sr isotopes
in otoliths. Can J Fish Aquat Sci 57:2280-2292
Kent AJR (2008) In situ analysis of Pb isotope ratios using laser ablation MC-ICP-MS: Controls on precision
and accuracy and comparison between Faraday cup and ion counting systems. JAAS, doi: 10.1039/
b801046c
Knesel KM, Davidson JP, Duffield WA (1999) Evolution of silicic magma through assimilation and subsequent
recharge: Evidence from Sr-isotopes in sanidine phenocrysts, Taylor Creek Rhyolite, NM. J Petrol
40:773-786
Kobayashi K, Tanaka R, Moriguti T, Shimizu K, Nakamura E (2004) Lithium boron and lead isotope
systematics of glass inclusions in olivines from Hawaiian lavas: evidence for recycled components in
Hawaiian plume. Chem Geol 212:143-161
Lambert IB, Drexel JF, Donnelly TH, Knutson J (1982) Origin of breccias in the Mount Painter area, South
Australia. J Geol Soc Austral 29:115-125
Layne GD, Shimizu N (1997) Measurement of lead isotope ratios in common silicate and sulfide phases using
the CAMECA IMS 1270 Ion Microprobe. In: Secondary Ion Mass Spectrometry, SIMS xI. Gillen G et
al. (eds) John Wiley, p 63-65
Maaskant P, Coolen JJMMM, Burke EAJ (1980) Hibonite and coexisting zoisite and clinozoisite in a calc-
silicate granulite form southern Tanzania. Mineral Mag 43:995-1003
Mathez EA, Waight TE (2003) Lead isotope disequilibrium between sulfide and plagioclase in the Bushveld
Complex and the chemical evolution of large layered intrusions. Geochim Cosmochim Acta 67:1875-
1888
McCulloch M, Cappo M, Aumend J, Muller W (2005) Tracing the life history of individual barramundi using
laser ablation MC-ICP-MS Sr-isotopic and Sr-Ba ratios in otoliths. Mar Freshwater Res 56:637-644
442 Ramos & Tepley
Mickiewicz S (2007) Constraining continental basaltic magma chambers processes: Textural and in situ
geochemical investigation of plagioclase from Pisgah Crater, California. MSc Dissertation, Central
Washington University, Washington
Milton DA, Chenery SR (2003) Movement patterns of the tropical shad (Tenualosa ilisha) inferred from
transects of 87Sr/86Sr isotope ratios in their otoliths. Can J Fish Aquat Sci 60:1376-1385
Morgan DJ, Jerram DA, Chetkoff DG, Davidson JP, Pearson DG, Kronz A, Nowell GM (2007) Combining
CSD and isotopic microanalysis: magma supply and mixing processes at Stromboli volcano, Aeolian
Islands, Italy. Earth Planet Sci Lett 260:419-431
Muller W, Mancktelow NS, Meier, M (2000) Rb-Sr microchrons of synkinematic mica in mylonites: an
example from the DAV fault of the Eastern Alp. Earth Planet Sci Lett 180:385-397
Outridge PM, Chenery SR, Babaluk JA, Reist JD (2002) Analysis of geological Sr isotope markers in fish
otoliths with subannual resolution using laser ablation-multicollector-ICP-mass spectrometry. Environ.
Geology 42:891-899
Pearce TH, Clark AH (1989) Nomarski interference contrast observations of textural details in volcanic rocks.
Geology 17:757-759
Perini G, Tepley FJ III, Davidson JP, Conticelli S (2003) The origin of K-feldspar megacrysts hosted in alkaline
potassic rocks from central Italy: a track for low-pressure processes in mafic magmas. Lithos 66:223-
240
Perlroth R (2000) An investigation of crystal transfer between melts. MSc Dissertation, University of California,
Los Angeles, Los Angeles, California
Pidgeon RT, Aftalion M (1978) Co-genetic and inherited zircon U-Pb systems in granites. Palaeozoic granites
of Scotland and England. Geol J Special Issue 10:183-220
Ramos FC, Reid MR (2005) Distinguishing melting of heterogeneous mantle sources from crustal contamination:
Insights from Sr isotopes at the phenocryst scale, Pisgah Crater California. J Petrol 46:999-1012
Ramos FC, Rodgers SL, Gill JB (2007) Evaluating the timing of volcanism at Baitoushan volcano (North
Korea/China) in the context of open-system effects: Insights from Sr, Nd, and Pb isotopes at the single
grain scale. AGU Fall Meeting abs. #V51C-0712
Ramos FC, Rodgers, SL, Wolff JW, Gill JB (2008) Connecting high silica rhyolite-hosted phenocrysts to
potential crystal mush intrusive complements: Insights from Baitoushan volcano and Valles caldera. GSA
Cordilleran/Rocky Mountains Section, Abs. w/ Programs, p18-7 p 84
Ramos FC, Wolff JA, Tollstrup DL (2004) Measuring 87Sr/86Sr variations in minerals and groundmass from
basalts using LA-MC-ICPMS. Chem Geol 211:135-158
Ramos FC, Wolff JA, Tollstrup DL (2005) Sr isotope disequilibrium in Columbia River flood basalts: Evidence
for rapid shallow-level open-system processes. Geology 33:457-460
Ramos FC (1992) Isotope geology of the Central grouse Creek Mountains, Box Elder County, Utah. MS Thesis
University of California, Los Angeles, Los Angeles, California
Reddy SM, Kelley SP, Wheeler J (1996) A 40Ar/39Ar laser probe study of micas from the Seisa zone, Italian
Alps: Implications for metamorphic and deformation histories. J Metamorph Geol 14:493-508
Reiners P (2002) Temporal-compositional trends in intraplate eruptions: Implications for mantle heterogeneity
and melting processes. Geochem Geophys Geosys 10.1029/2001GC000250
Rickwood PC (1977) A technique for extracting small crystals from thin sections. Am Mineral 62:382-384
Rundle CC (1992) Review and assessment of isotopic ages from the English Lake District. British Geological
Survey, Onshore Geology Series, Technical Report WA/92/38
Saal AE, Hart SR, Shimizu N, Hauri EH, Layne GD (1998) Pb isotopic variability in melt inclusions from
oceanic island basalts, Polynesia. Science 282(5393):1481-1484
Saal AE, Hart SR, Shimizu N, Hauri EH, Layne GD, Eiler JM (2005) Pb isotopic variability in melt inclusions
from the EMI-EMII-HIMU mantle end-members and the role of oceanic lithosphere. Earth Planet Sci
Lett 240:605-620
Salisbury MJ, Bohrson WA, Clynne M, Ramos FC, Hoskin P (2008) Plagioclase crystal populations identified
by crystal size distribution and in situ chemical data: implications for timescales of magma chamber
processes associated with the 1915 Eruption of Lassen Peak, CA. J Petrol 2008; doi: 10.1093/petrology/
egn045
Schmidberger SS, Simoneti A, Francis D (2003) Small-scale Sr isotope investigations of clinopyroxenes from
peridotite xenoliths by laser ablation MC-ICP-MS-implications for mantle metasomatism. Chem Geol
199:317-329
Siebel W, Reitter E, Wenzel T, Blaha U (2005) Sr isotope systematics of K-feldspars in plutonic rocks revealed
by the Rb-Sr microdrilling technique. Chem Geol 222:183-199
Simon JI, Reid MR, Young ED (2007) Lead isotopes from LA-MC-ICPMS: tracking the emergence of mantle
signatures in an evolving silicic magma system. Geochim Cosmochim Acta 71:2014-2035
Singer BS, Dungan MA, Layne GD (1995) Textures and Sr, Ba, Mg, Fe, K, and Ti compositional profiles in
volcanic plagioclase: clues to the dynamics of calc-alkaline magmas chambers. Am Mineral 80:776-798
Inter- and Intracrystalline Isotopic Disequilibria 443
Stamatelopoulou-Seymour K, Vlassopoulos D, Pearce TH, Rice C (1990) The record of magma chamber
processes in plagioclase phenocrysts at Thera volcano, Aegean volcanic arc, Greece. Contrib Mineral
Petrol 104:73-84
Tepley FJ III, Davidson JP (2003) Mineral-scale Sr-isotope constraints on magma evolution and chamber
dynamics in the Rum layered intrusion, Scotland. Contrib Mineral Petrol 145:628-641
Tepley FJ III, Davidson JP, Clynne MA (1999) Magmatic interactions as recorded in plagioclase phenocrysts
of Chaos Crags, Lassen Volcanic Center, California. J Petrol 40:787-806
Tepley FJ III, Davidson JP, Tilling RI, Arth JG (2000) Magma mixing, recharge and eruption histories recorded
in plagioclase phenocrysts from El Chichón volcano, Mexico. J Petrol 41:1397-1411
Thirwall M (2002) Multicollector-ICPMS analysis of Pb isotopes using a 207Pb-204Pb double spike demonstrates
up to 400ppm/amu systematic errors in Tl normalization. Chem Geol 184:255-279
Thorrold SR, Shuttleworth S (2000) In situ analysis of trace elements and isotope ratios in fish otoliths using
laser ablation sector field inductively coupled plasma mass spectrometry. Can J Fish Aquat Sci 57:1232-
1242
Thrane K, Bizzarro M, Baker JA (2006) Extremely brief formation interval for refractory inclusions and
uniform distributions of 26 Al in the early solar system. Astrophys J 646:L159-L162
Tilling RI, Arth JG (1994) Sr and Nd isotopic compositions of sulfur-rich magmas of El Chichón volcano,
Mexico. IAVCEI abstract
Tyrrell S, Haughton PD, Daly JS (2007) Drainage reorganization during the breakup of Pangea revealed by
in-situ Pb isotopic analysis of detrital k-feldspar. Geology 35:971-974
Vroon Pz, Wagt BV, Koornneef JM, Davies GR (2008) Problems in obtaining precise and accurate Sr isotope
analysis from geological material using laser ablation MC-ICPMS. Anal Bioanal Chem 390:465-476
Wagner WR (1945) An experimental micro-drill. Science 101:127-128
Waight TE, Baker J, Peate D (2002) Sr isotope ratio measurements by double-focusing MC-ICPMS: techniques,
observations and pitfalls. Int J Mass Spectrom 221:229-244
Waight TE, Dean AA, Maas R, Nicholls IA (2000b) Sr and Nd isotopic investigations towards the origin of
feldspar megacrysts in microgranular enclaves in two I-type plutons of the Lachlan Fold Belt, southeast
Australia. Austral J Earth Sci 47:1105-1112
Waight TE, Maas R, Nicholls IA (2000a) Fingerprinting feldspar phenocrysts using crystal isotopic composition
stratigraphy: implications for crystal transfer and magma mingling in S-type granites. Contrib Mineral
Petrol 139:227-239
Weber PK, Bacon CR, Hutcheon ID, Ingram BL, Wooden JL (2005) Ion microprobe measurement of strontium
isotopes in calcium carbonate with application to salmon otoliths. Geochim Cosmochim Acta 69:1225-
1239
Willigers BJA, Baker JA, Krogstad EJ, Peate DW (2002) Precise and accurate in situ Pb-Pb dating of apatite,
monazite, and sphene by laser ablation multi-collector ICP-MS. Geochim Cosmochim Acta 66 6:1051-
1066
Wolff JA, Ramos FC (2003) Pb isotope variations among Bandelier Tuff feldspars: No evidence for a long-
lived silicic magma chamber. Geology 31:533-5363
Wolff JA, Ramos FC (2004) Sr and Pb in high-silica rhyolites: Lessons from the Bandelier Tuff. EOS Trans.
AGU 85(17), Jt. Assem. Suppl. Abstract V12A-02
Wolff JA, Ramos FC, Davidson JP (1999) Sr isotope disequilibrium during differentiation of the Bandelier
Tuff: Constraints on the crystallization of a large rhyolitic magma chamber. Geolology 27:495-498
Wolff JA, Ramos FC, Hart GL, Patterson JD, Brandon AD (2008) Columbia River flood basalts from a
centralized crustal magmatic system. Nature Geosci 1:163-167
Woodhead J, Swearer S, Hergt J, Maas Roland (2005) In situ Sr-isotope analysis of carbonates by LA-MC-
ICP-MS: interference corrections, high spatial resolution and an examples from otolith studies. J Anal
At Spectrom 20:22-27
Reviews in Mineralogy & Geochemistry
Vol. 69 pp. 445-478, 2008 12
Copyright © Mineralogical Society of America
Oxygen Isotopes in Mantle and Crustal Magmas as
revealed by Single Crystal analysis
Ilya Bindeman
Department of Geological Sciences
University of Oregon
Eugene, Oregon, 97403-1272, U.S.A.
[email protected]
Part I: IntrOduCtIOn
Oxygen is the most abundant element in the Earth’s crust, mantle, and fluids and therefore
its isotopic composition provides robust constraints on magma genesis. Application of oxygen
isotope geochemistry to volcanology and igneous petrology provides a much needed foundation
for radiogenic isotope and trace element approaches. Since isotope fractionations at high
temperature are small, there is a demand for high analytical precision in order to recognize
and interpret small (tenths of permil) variations in isotopic composition. Recently improved
analytical techniques involving lasers and ion microprobes, and reduction in sample and spot
size, has painted a picture of isotope complexity on a single crystal scale that is helpful in
interpreting magma genesis and evolution. In this chapter a review is provided for several
classic examples of silicic and basic magmatism, including Yellowstone and Iceland, that shows
isotope zoning and heterogeneity reaching several permil. Isotope heterogeneity fingerprints
crystal sources and provides constraints on diffusive and recrystallizational timescales. These
new lines of evidence reveal that magma genesis happens rapidly, at shallow depths, and through
batch assembly processes.
Oxygen isotope geochemistry spans more than 50 years of investigation and is the most
developed among other traditional (e.g., C, N, H, S, Li, B) and less-traditional (e.g., Fe, Mo,
Cu) stable isotope systems. While we provide basic concepts of isotope fractionation below, the
reader is referred to three prior RiMG volumes on stable isotopes (Valley, Taylor, and O’Neil
1986 – volume 16; Valley and Cole 2001 – volume 43; Johnson et al. 2004 – volume 55), and
the Hoefs (2005) and Sharp (2006) textbooks for greater treatment and historic perspective.
Finally this Chapter does not deal with oxygen isotopic variations in meteorites and planetary
igneous materials, and interested readers are referred to RiMG volume 68, “Oxygen in the Solar
System” (MacPherson et al. 2008).
The remaining novel aspects of oxygen isotope geochemistry of igneous and metamorphic
rocks are related to continuing application of laser fluorination analysis to phenocrysts that
provide superior analytical precision to other methods. In particular, 1) Many topical studies
should include major element oxygen measured at new analytical levels as a component in
multi-isotope and trace elemental investigation, 2) Older studies that relied on whole-rock
methods should be reassessed, 3) Many igneous systems such as continental and oceanic
flood basalts, large silicic igneous provinces, and island arc magmas should be reevaluated,
4) Oxygen isotopes should be studied on a single crystal level. This Chapter outlines the
perspective that magmatic crystals should be studied individually when possible, or by size
fractions, to demonstrate whether crystals are isotopically homogeneous and is in equilibrium
with other minerals within common host magma. As is demonstrated below, in many classic
1529-6466/08/0069-0012$05.00 DOI: 10.2138/rmg.2008.69.12
446 Bindeman
examples of igneous systems around the world, magmatic crystals are not in equilibrium and
thus bulk phenocryst oxygen isotope analysis does not reveal the full details of petrogenesis,
nor do they provide a proxy for magmatic values.
Basics of oxygen isotope variations in nature and their causes
Oxygen consists of three isotopes: 16O (99.76%), 17O (0.04%), and 18O (0.2%). Mass-
dependent, oxygen isotopic variations are described through a ratio of 18O to 16O and the delta
notation that is traditionally used in stable isotope geochemistry:
δ18O = (RSA/RST − 1) × 1000 (1)
where RSA and RST are absolute ratios of 18O/16O in sample and standard and the standard
is Vienna Mean Standard Ocean Water (VSMOW) with 18O/16O absolute ratios of 0.020052
(Baertschi 1976). A ∆18O measure of variation between the sample and the standard is in
permil (‰), or part per thousand. It is not recommended to use any other standard (e.g., PDB)
when describing oxygen isotopic variations in nature since it only causes confusion. The
multiplication factor of 1000 is used to display natural isotope variations in whole numbers
and not fractions; it is used because isotope variations due to chemical or physical processes in
nature are small, in the third or second decimal place of the 18O/16O ratio.
Oxygen isotopic variations on Earth span about 100‰ (Fig. 1). Half of these variations,
and nearly all negative δ18O values are occupied by meteoric waters that are isotopically-light
as a result of Rayleigh distillation upon vapor transport and precipitation. Silicate rocks and
magmas occupy the positive part of this diagram with the absolute majority of mantle rocks,
basaltic magmas and chondritic meteorites plotting in a narrow range of 5.5 to 5.9‰. Silicate,
oxide, carbonate, and phosphate minerals and igneous, sedimentary, and metamorphic rocks
such as carbonates and especially diatoms that precipitate from water at low temperatures
are the highest δ18O materials because of large positive isotope fractionation factors between
silica, carbonates and water at low temperatures (e.g., Friedman and O’Neil 1977; Chacko
et al. 2001; Hoefs 2005). Metasedimentary rocks and igneous rocks such as S-type granites
inherit high-δ18O supracrustal signature from the source, while rocks that represent remelting of
hydrothermally altered rocks that interacted with low-δ18O meteoric water at high-temperature
are low-δ18O (see below).
Equilibrium isotope fractionation factors between minerals and melts. The equilibrium
isotope fractionation factor is defined as:
α = RA/RB (2)
where RA and RB are absolute isotope ratios of individual, coexisting minerals at equilibrium.
Since α variations are typically in the second or third decimal point, a more convenient
parameter, 1000lnα is used.
Because of the difference in chemical bonds affecting vibrational frequencies of oxygen
in minerals, a heavier isotope of oxygen partitions itself into a mineral with stronger (more
covalent) Si-O-M bonds. As a general rule, isothermal high-T distribution of 18O among
coexisting igneous phases is explained by the proportion of Si-O and M-O bonds (Taylor 1968;
zheng et al. 1993a,b; Chiba et al. 1989; Hoefs 2005), while the cation identity in homovalent
substitution (e.g., Fe vs. Mg) plays an insignificant third order role. Thus the common igneous
minerals will become progressively lighter from quartz (pure silicate) to magnetite (pure M-O
oxide). For example, a polymineralic granite at 850 °C with the whole-rock value of 7.8‰ will
have decreasing δ18O values of its constituent minerals in this sequence: quartz (8.2‰) > albite
≈ K-Fsp (7.5‰) > anorthite (6.6‰) > zircon (6.4‰) ≥ pyroxene (6.3‰) ≈ amphibole ≥ biotite
≥ garnet ≈ olivine (6.1‰) > sphene (5.4‰) ≥ ilmenite (4.9‰) > apatite ≥ magnetite (3.5‰).
The quoted δ18O values were calculated using experimental and empirical fractionation factors
(e.g., Taylor and Sheppard 1986; Chiba et al. 1989; zheng 1993a,b; Chacko et al. 2001; Valley
Single Crystal Analysis of Oxygen Isotopes in Magmas 447
Figure 1. Oxygen isotopic variations in magmas and rocks on Earth. Gray band between 5.5 and 5.9 Figur
represents “normal-δ18O magmas” for ultramafic and basaltic rocks that constitute predominant mantle and
products of is melting; closed system differentiation of such magma into andesite and rhyolite that show
normal δ18O values of 5.8 to 6.5 permil (Fig. 4). High-δ18O rocks and magmas result from low-temperature
precipitation or interaction with sea water involving large positive 1000lnαmineral-water isotope fractionation
(Fig. 2a). Low-δ18O rocks and magmas result from high-temperature interaction with, or subsequent re-
melting of, materials that interacted with meteoric waters at high-temperatures involving small isotope
fractionations. Data are from both original sources and compilations by Muehlenbachs (1998); Eiler
(2001), Valley et al. (2005); Hoefs (2005); Sharp (2006). Selected areas emphasized in this review are
shown as open boxes.
et al. 2003). Establishing the equilibrium fractionation factors for igneous minerals is still an
active area of experimental research, especially for refractory accessory minerals, for which
direct experiments are difficult (e.g., Krylov et al. 2002; Valley et al. 2003; Trail et al. 2008).
For isotopic fractionation between minerals and melt, most variation is caused by the
changing composition of melt. In an attempt to quantify the mineral-melt fractionation factors,
the easiest approach is to treat the melt as a mixture of normative minerals and then calculate the
weighted average of individual mineral-melt (normative sum) fractionation factors (Matthews
et al. 1994; Palin et al. 1996; Eiler 2001; Appora et al. 2002). Although the natural silicate
melts are more complicated mixtures of various components those chemical variations are
often treated by a regular solution model (Ghiorso and Sack 1995), for the purpose of isotope
448 Bindeman
fractionation, non idealities of melt mixing functions play a minor role. The normative mineral
approach assumes that a normative mineral in melt has the same partition function ratios for
isotope distribution; it has been demonstrated to be largely correct assumption for a variety
of quartzofeldspathic mixtures, despite a few tenths of one permil disagreement between
crystalline quartz and quartz-melt as measured in experiments (Matthews et al. 1998).
When δ18O values are relatively similar and not extremely low or high in δ18O, the
fractionation factor 1000lnα is numerically very close to the simple difference between
measured δ18O values, or ∆18O (e.g., Hoefs 2005; Sharp 2006):
1000lnαquartz-zircon ≈ ∆18O quartz-zircon = δ18O quartz − δ18Ozircon (3)
Isotope fractionation of oxygen between coexisting minerals, like any other stable isotope, is a
strong function of temperature; for solid substances at temperatures higher than ~50-100 °C, the
isotope fractionation factor 1000lnα is a linear function of 1/T2 (Biegielsen and Meyer 1947;
Urey 1947). The reverse quadratic rather than 1/T dependence that is common in thermodynamics
of element partitioning is explained by the fact that the factor that causes isotope fractionations-
isotope vibrational frequencies, are themselves increasing with temperature. Figure 2 presents
isotope fractionations between major igneous minerals and water as a function of T. Note
that isotope fractionations are largely pressure-independent for common upper mantle-crustal
pressures because isotope substitutions do not have significant volume effects. In 1000lnα vs.
1/T2, isotope exchange between minerals, and most likely minerals and silicate melt, can be
described by a single A factor:
1000lnα = 106A/T 2 (4)
defining linear dependence passing through the origin at infinite temperature at which isotope
fractionations are zero.
Because of the reverse quadratic relationship between 1000lnα and temperature (Fig. 2)
isotope fractionations between coexisting minerals, melts and fluids are small at high magmatic
temperatures, typically less than 2-3‰. It underlies the importance of high precision, better
than 0.1‰, to resolve subtle isotopic differences between different sources for magmas. Small
(1-2‰) isotope fractionation between minerals were traditionally used to predict that isotope
fractionation are consistent with magmatic temperatures and not with subsolidus exchange (e.g.,
Taylor 1986), meaning that they reflect primary sources rather than secondary isotope effects.
Isotope thermometry. An important aspect of oxygen isotope analysis of coexisting mineral
assemblages is the ability to use the experimentally-determined A factors and measured ∆18O
as input parameters to calculate temperature in Equation (4). Temperature estimates are more
easily determined for mineral pairs with large A factors such as quartz and magnetite (Fig. 2B).
Several assumptions should be evaluated in order for isotope thermometry to reflect primary
magmatic values. First, minerals should be fresh and unaltered, as can be demonstrated under
an electron microscope. Second, the rock or magma should have cooled rapidly so that the
measured isotope temperature reflects quenched pre-eruptive conditions and does not reflect the
effects of slow plutonic cooling with sequential mineral closure to isotope exchange (e.g., Eiler
et al. 1993; Farquhar et al. 1993). Third, minerals should be phenocrysts and not isotopically-
distinct xenocrysts.
Because the first two conditions are rarely met for metamorphic and likewise plutonic
rocks, the initial optimism to be able to use pressure-independent oxygen isotope thermometry
(where P and T can be correlative based on chemical geothermometers and geobarometers)
nearly disappeared, except for simplest bimineralic assemblages involving refractory
accessory phases where possibility for exchange is limited (e.g., Valley 2001). However, for
fresh volcanic rocks that quenched rapidly, measured ∆18O values can be used to determine
preeruptive temperature. Young pyroclastic volcanic rocks such as single pumice or tephra
Single Crystal Analysis of Oxygen Isotopes in Magmas 449
15 (a)
δ18O (Quartz)
10
δ18O ‰
Normal- δ18O rocks and magmas
δ18O (Fsp)
0 Δ18O(Fsp-H2O)
δ18O (H2O)
-10
25° 900°
Temperature, °C
T, °C
1000 700 500 400 300
13 Figure 2. Oxygen isotope equilibrium between
minerals, melts and water as a function of
12 (b) temperature. a) Schematic diagram illustrating
δ18O as a function of temperature for a closed-
11 system exchange between minerals and water.
Isotope variations at high temperature are
10 Quartz-Magnetite smaller, and water becomes isotopically heavier
Figure 2a
than rocks (∆18OFsp-H2O < 0) after a crossover
9 temperature of 400-550 °C, and that high-T
water-rock interaction will lead to depletion of
8 rocks with respect to 18O (Friedman and O’Neil
1000 lnα
Albite-Magnetite
7 1977; Chacko et al. 2001). When sea water
with 0‰ is considered, low-T hydrothermal
6
Anorthite-Magnetite alteration may lead to higher-δ18O values
while interaction with meteoric water will
5 lead to rock δ18O depletion at nearly every
Quartz-Zircon temperature. Hydrothermal alteration will lead
4 to highly heterogeneous distribution of oxygen
Anorthite-Olivine isotopes in rocks depending on local (i.e., mm-
3 Rhyolite-zircon scale) water-rock ratios. b) Mineral-mineral
and mineral-melt isotope fractionation as a
2 Basalt-Olivine function of 1/T2 . Data sources: Thin lines are
Quartz- experimental data (Chiba et al. 1989). Thicker
1 rhyolite Albite-Muscovite
lines represent empirical measurements (Eiler
2001; Bindeman and Valley 2002; Valley et al.
0 1 2 3 4 2003; Bindeman et al. 2004).
106/T2, K-2
Figure
clasts perhaps present the best examples of quenched eruptive products2b
that preserve magmatic
∆ O values, and thus temperatures (e.g., Bindeman and Valley 2002, Fig. 3).
18
Mineral vs. whole rock analysis. Oxygen isotope ratios are used in much the same way
as radiogenic isotopes or trace elemental ratios to define or identify different geochemical
reservoirs or sources of magmas. In this sense, whole rock oxygen isotope analysis should
theoretically represent magma and its source, and such analysis was a common practice in
early studies. The early investigation of oxygen isotopes in igneous rocks relied on whole rock
analysis by conventional methods; however, large oxygen isotopic variations put forward by
450 Bindeman
(a) Figure 3. a) Recognition
δ18O, ‰ of ∆18Omelt-mineral isotope
equilibria and disequilibria.
mineral 2
D With decreasing tempera-
ture, isotope fractionations
(divergent lines) between
C minerals (letter boxes) and
melt increase up until the
m e l t temperature of eruptive
quench, Tq, oxygen diffu-
E
Δ18Omelt-mineral1
sion is expected to operate
without closure. Crystal A
(mineral 1) and crystal D
A (mineral 2) are in isotope
mineral 1 equilibrium among them-
selves and the melt, while
B individual crystals E, B
and C of mineral 1 record
T, °C disequilibrium ∆18Omelt-min-
Tq eral values. Crystal B comes
from a lower-δ18O source,
while crystal C comes from a higher-δ18O source,
(b) and they are therefore xenocrysts. Crystal E is only
9 Quartz subtly, by a few tenths of one permil, heavier than
the equilibrium value, and can be interpreted to rep-
8 resent crystallization at higher-temperatures and then
Calculated Melt capture into a lower-temperature magma differentiate
7 similar to cumulate entrapment. Crystal E could be
δ18 O ‰ VSMOW
6 a “proto”- or “ante” cryst. b) An example of isotope
equilibria among minerals and melt in Bishop tuff,
5 Zircon CPx Figure 3a
a large volume rhyolite deposit in which individual
pumice clasts record equilibrium ∆18O melt-mineral
4 isotope fractionations, divergent with decreasing
temperature (from Bindeman and Valley 2002). Each
3 vertical column of symbols represent analyses from a
single pumice clast erupted from different depth of the
2 Magnetite thermally-zoned magma chamber ranging from 6 to
11 km (Wallace et al. 1999) thus quenching specific
1 magmatic ∆18Omelt-mineral values. The obtained isotopic
700 725 750 775 800 825
temperatures agree very well with Fe-Ti oxide tem-
T, °C, Δ 18O(Qz-Mt) peratures in the Bishop tuff clasts.
these early researchers were later reinterpreted as being mostly due to secondary alteration
effects with surface waters (Taylor 1986). Studies of minerals instead of rocks, and in particular
refractory phenocrysts such as quartz, zircon and olivine, yielded a picture of subtle subpermil
variations in δ18O within the mantle (Valley et al. 1998; Eiler 2001) and narrower ranges in
many crustal rocks (e.g., Valley et al. 2005). Fresh glass of modern or recent volcanic rocks that
Figure
is unaltered can be used to directly infer primary 3b δ18O values. However, volcanic
magmatic
glass in old rocks is metastable and exchanges oxygen with waters even before there is clear
petrographic evidence of devitrification or development of microscopic clays (e.g., Taylor
1986). Iridescence in sanidine and glass, which may signal the appearance of microscopic
clays can be used as an indicator of δ18O modification.
Unaltered phenocrysts in igneous rocks provide a better view of true magmatic δ18O value,
but they are different from the δ18O value of the magma by a fractionation factor ∆18Omineral-melt
that is a function of temperature, melt composition, and, for the case of plagioclase and similar
heterovalent solid solution, their Ca/Na and Al/Si ratios. In studies involving comparison of
δ18O between different magmas or sections in a drillcore, it is better to present data for a single
Single Crystal Analysis of Oxygen Isotopes in Magmas 451
mineral if possible (e.g., Wang et al. 2003). In some other applications, a calculated magma
value can be generated and compared, especially if δ18O values of multiple phenocrysts are
obtained. Additionally, when relying on δ18O values of crystals in old volcanic or plutonic
rocks, it is better to avoid feldspars and micas, and rely on “refractory” phases that are resistant
to secondary isotope exchange, and do not have solid solution that fractionates oxygen isotopes.
For that matter, quartz and zircon in silicic rocks, and olivine in basic rocks, provide reliable
proxies for their parental melts, while plagioclase and K-feldspar are less reliable.
The use of phenocrysts requires extra time for mineral separation—a task that was tedious
for conventional resistance-furnace fluorination that necessitates 5-25 mg of monomineralic
separate. With the reduction of sample size by ten-fold, mineral separation time is now trivial or
minimal for laser fluorination and ion microprobe methods. This translates into better quality,
inclusion-free, alteration-free separate. Most importantly, the 0.5-2 mg typical sample size for
the laser fluorination analysis overlaps with sizes of typical phenocrysts of most rock-forming
minerals, thus allowing them to be studied individually, an approach utilized in this Chapter.
Oxygen isotopes in mantle-derived rocks, normal-δ18O, high-δ18O, and low-δ18O
magmas
Basaltic magmas of mid-ocean ridges and most common island arc basalt are characterized
by a relatively narrow 5.7±0.2‰ range of δ18O values calculated from phenocrysts or measured
directly in fresh glasses (Fig. 1). The OIB basalts document greater range of +4 to +6‰ and
reflect recycling of materials that interacted with surface waters at low or high temperature
into the mantle (Fig. 1). Eclogite nodules and serpentinized ultramafic rocks exposed on the
surface document greater δ18O ranges (Kyser et al. 1982; Harmon and Hoefs 1995). High-δ18O
magmatic values are seen in intermediate and silicic rocks in continental arcs and collision
zones, in magmas that were derived from or interacted with metasedimentary protoliths; low-
δ18O values characterize caldera settings and rift zones that involve shallow crustal recycling of
rocks interacted with meteoric waters at high-T. Reviews by Taylor (1986) and recent reviews
by Eiler (2001) and Valley et al. (2005) provide a comprehensive account of whole-rock and
phenocryst-based oxygen isotope geochemistry of the mantle and the crust. This chapter
challenges these studies somewhat by documenting new complexity revealed by individual
phenocryst studies, which shows far greater ranges in individual phenocryst δ18O values that
reflect crystallization from or exchange with the diverse melts (open boxes in Fig. 1).
It is convenient to define most common basic magma as “normal-δ18O” and consider
products of its differentiation-crystallization as a “normal-δ18O differentiation array” (Fig. 4).
It has been noted in many natural examples of closed-system differentiation series that
differentiation of basalt leads to a small subpermil increase in δ18O melt value, with a particular
magnitude and trajectory of increase weakly dependent on the sequence of phase appearances
that affect ∆18Omineral-melt (Anderson et al. 1971; Taylor and Sheppard 1986; Eiler 2001).
In order to better understand mineral-melt oxygen isotope partitioning, and calculate
basaltic magma differentiation trends based on experimental mineral-mineral and mineral-
melt partitioning, a computational approach can be taken. The rich existing database on
mineral-mineral isotope fractionation (Friedman and O’Neil 1977; Chiba et al. 1989; Chacko
et al. 2001, and others), is now complemented by isotope exchange experiments involving CO2
gas as an exchange medium and silicate melts of variable composition (see summary in Eiler
2001). We recommend continuing with the practice of treating melt as a mixture of normative
mineral components and calculate ∆18Ophenocryst-melt isotope fractionation as a weighed sum of
∆18Ophenocryst-normative mineral in melt between the phenocryst and each normative component of the
melt for a given temperature. This approach has been used by Eiler (2001) to demonstrate small
(<0.2-0.3‰) isotopic differences between variably differentiated basic rocks as a function of
basalt MgO content, by zhao and zheng (2003) for rocks of variable composition, and by
Bindeman et al. (2004) for extended island arc series as a function of SiO2. The latter study
452 Bindeman
Differentiation of High-Mg basalt, Differentiation of variable starting
6.1 1 wt% water, different pressures 6.1 compositions
high-Al,
δ 18O(melt), ‰
1kb 4 kb 3 wt% H2O
6.0 Shoshonite
6.0
1 bar
5.9 5.9
8 kb high-Al,
1 wt% H2O
5.8 5.8 tholeiite
20 kbar
(a) (b)
5.7 5.7
45 50 55 60 65 70 75 45 50 55 60 65 70 75
SiO2, wt%, melt SiO2, wt%, melt
6.8
Closed system Assimilation of high-δ18O
differentiation:: supracrustal materials
high-Mg series
high-Al series and
alkali-rich series
δ 18O(melt), ‰
6.4
(c) High-δ18O magmas
array
18 O
6.0 l-δ
ma
Nor
Low-δ18O magmas
MORB Assimilation of low-δ18O
5.6 hydrothermally-altered rocks
45 50 55 60 65 70 75
SiO2, wt%, melt Figure 4
Figure 4. Oxygen isotope effects of closed system igneous differentiation that produce insignificant 0.2-
0.5‰ increase and concave-down trend in δ18O-SiO2 coordinates. a) Differentiation of high-Mg basalts
in MELTS program at different pressures; b) differentiation of various starting compositions; c) results
of multiple numerical crystallization experiments for the two most common island arc series, labeled
as “normal-δ18O” magma differentiation array, that separates fields of high-δ18O magmas and low-δ18O
magmas, whose genesis requires open system behavior. After Bindeman et al. (2004).
demonstrated that for a typical igneous rock, a smooth, concave downward trajectory, and an
increase of 0.3-0.4‰ is expected in δ18Omelt-SiO2 coordinates from basalt to rhyolite (Fig. 4).
Only some rare magmatic series involving unusual Fe-enrichment in the residual melt such as
the Skaergaard intrusion yield a nearly flat trajectory (e.g., Anderson et al. 1971; Kalamarides
1984; Bindeman et al. 2008a). A trajectory of δ18O increase with SiO2 is steeper from basalt
to andesite because low-δ18O minerals such as olivine and pyroxene fractionate early, driving
melt toward higher δ18O values (Fig. 4). With the appearance of feldspar, the trend flattens
while crystallization of near-eutectic quartz+feldspar leads to a nearly constant δ18O value of
melt and separating cumulate assemblage. The latter is important for high-silica rhyolites that
are near-eutectic systems; since the chemical composition of bulk cumulates is the same as the
residual melt, the ∆18Ocumulate-melt fractionation is zero.
Single Crystal Analysis of Oxygen Isotopes in Magmas 453
High-δ18O magmas. The δ18O value of crustal rocks is diverse (Fig. 1, 4), and the majority
of crustal igneous rocks have δ18O values higher than the mantle magma differentiation array.
The majority of cases of isotopically-heavy magma values are related to the exchange, or
derivation from, a high-δ18O metasedimentary silicate rock (such as metapelites, carbonates,
or metagreywackes), that originally crystallized from, or exchanged with, sea water at surface
temperatures. Oxygen isotopes serve as an important parameter of crustal assimilation and
derivation, especially when coupled with other monitors of these processes such as 87Sr/86Sr
isotopes (Taylor 1980, 1986), U-series, and trace elements (Finney et al. 2008).
Low-δ18O magmas. A subset of igneous rocks, silicic and basic, that represent remelting,
assimilation or exchange with hydrothermally-altered rocks (altered by heated meteoric waters)
represent δ18O magmas. The low-δ18O rocks were earlier known to occur in only a few loca-
tions such as Yellowstone and Iceland (e.g., Taylor 1986). The author of this Chapter is a strong
believer of far greater abundance of low-δ18O magmas than is currently thought both in terms
of volume and the number of volcanic units; however, low-δ18O magmas are likely underrepre-
sented in pre-Tertiary geologic record because being of shallow genesis they are eroded away.
Part II: SInGLe PhenOCrySt ISOtOPe StudIeS
A recent upsurge in interest in single crystal studies reflects the advent of trace elemental
and isotopic microbeam and microdrilling techniques that are suitable for dating and
fingerprinting single crystals. These approaches have led to the realization that many, if not
most, igneous systems contain isotopically diverse and chemically distinct crystal populations
that are “heterogeneous on all scales” (e.g., Dungan and Davidson 2004). Minerals that are
found in magmas as “pheno”crysts may not necessarily have crystallized from their host melt
and may be “proto”crysts or cumulates that are entrained into the more differentiated host
product, “ante”crysts capured from chamber walls that represent prior episode(s) of magmatism
in the same place, or “xeno”crysts captured from rocks that are much older and are not related
to the current cycle of magmatism. For review of these topics see Ramos and Tepley (2008)
of this volume. Collectively, there is a growing consensus that histories of crystals and crystal
populations (“crystal cargo”) and their host melt may be decoupled and thus isotope disequilibria
may provide insights into the origin of both crystals and magma.
Many examples of isotopic zoning and disequilibria in volcanic phenocrysts, with
emphasis on volcanic arcs, have been described, and a small level of residual zoning may in fact
characterize the majority of igneous rocks, even in slowly cooled plutonic examples (Tepley and
Davidson 2003). Therefore, the presence of isotopically-zoned phenocrysts provides “blessing
rather than a curse” into a potentially important record of magma sources and timescales of
magmatic processes.
Modern methods of oxygen isotope analysis of crystals
Sizes of typical “pheno”crysts in volcanic rocks are in the 0.5-2 mm scale, with their
relative abundance, expressed through crystal size distribution (Fig. 5), an important textural
characterization of a rock that is pertinent to the kinetics of its crystallization (Streck 2008;
Armenti 2008 and references therein) and crystal inheritance. A certain mass of material is
required for a precise single crystal oxygen isotope analysis, and as seen on Figure 6b, the
precision of an isotopic measurement decreases with the method and the amount of material
available. For example, the quoted 0.5-2mm range corresponds to 0.35 to 22 mg of crystal mass
with density 2.8 g/cm3.
The most precise method for oxygen isotope analysis—laser fluorination—was developed
in the 1990s (Sharp 1990; Valley et al. 1995). This method relies on fluorination of ~0.3-2 mg
material followed by the multiple-cycle analysis of generated micromol quantities of O2 or
454 Bindeman
11 10
LBT HRT-A
ln(n/V L) (cm-4 ) (a) UBT
LCT-A (b) HRT-C
MFT
LCT-B LCT-A
9 post-LCT 8 post-LCT
7 6
5 4
3 2
0 0.8 1.6 2.4 3.2 0 100 200 300
Quartz size (mm) Zircon l ength (µm)
Figure 5. Crystal Size distributions of quartz and zircons in rapidly quenched pumice clasts in large volume
tuffs and smaller volume lavas. There are significant differences in crystal sizes and CSD even within
the same tuff units. The abundant concave down, lognormal CSDs suggests solution and re-precipitation
recycled smaller crystals and the outermost rims of large crystals (e.g., Simakin and Bindeman 2009).
CSDs can potentially be used to correlate different magma batches with different crystallization conditions.
Abbreviations: HRT- Huckleberry Ridge tuff, LCT- Lava Creet tuff, MFT- Mesa Falls tuff, LBT and UBT
are Lower and Upper Bandelier tuffs.
Spot diameter (x depth), μm Single crystal radius or spot size, mm
1x1 10 30x6 0.1 0.4 2 Figure 5
NanoSIMS
Precision, ‰, 1 st dev
1.0
Small radius
0.5 Large
ion Continuous flow
radius
di GSMS analysis
microprobes
ion
microprobes Dual inlet GSMS analysis
0.3 UV laser
fluorination
Conventional
Ni rod bombs
0.1 CO2 laser
fluorination
0.05
4 100 0.01 0.1 1 10
picograms
Sample size, mg
Figure 6. Precision of various analytical methods for oxygen isotopes as a function of sample weight or
spot size; precision ranges are current and are based on informed prejudices of the author. Dashed lines
represent “best day” precisions. Figure 6
CO2 gas on a mass spectrometer against a standard gas of known isotopic composition (dual
inlet on Fig. 6). The dual inlet mode of sufficiently large micromol quantities of gas remains
the base method of stable isotope mass spectrometry (as TIMS for radiogenic isotopes) analysis
with precision improving as a square root of the number of unknown gas-reference gas compar-
ison cycles, commonly better than 0.01-0.02‰ for a typical 6-10 cycle analysis. Most analytical
error in oxygen isotope analysis is related to the chemistry of oxygen extraction by fluorination.
Single Crystal Analysis of Oxygen Isotopes in Magmas 455
Overall external precision in CO2 laser fluorination labs in Wisconsin, Caltech, and Oregon is
“better than 0.1‰” 1 standard deviation, with the uncertainty on standards of ±0.05-0.07‰
common for many analytical sessions (Valley et al. 1995; Eiler 2001; Bindeman et al. 2008c).
The advantage of CO2 laser fluorination over other methods include: 1) quantitative, 100%
reaction of silicate material at temperatures approaching 2000 °C; at such temperatures isotope
fractionations between generated gas and potential O-bearing residue are much smaller (ca
Fig. 3) than that at 500 °C conventional resistance furnace analysis. 2) High temperature pro-
vides the ability to heat and react even the most refractory minerals such as olivine and zircon;
3) reduction of sample size to a single phenocryst that generate enough gas for a multiple cycle,
dual inlet analysis. 4) Mineral separation is much less tedious and the quality control is better
since it is possible to consistently rely on the highest purity concentrate.
The most important current in situ analytical technique for light stable isotopes is ion mi-
croprobe analysis. This method uses secondary ions generated by the collision of a primary
beam with a target; secondary ions are then analyzed on a mass spectrometer using a variety
of secondary ion collection and filtering techniques (Hinton 1995; Ireland 1995; Riciputi et al.
1998; Kita et al. 2004). Spot size can vary from ~2 to 20 µm on large multicollector (e.g., Cam-
eca 1270/80, Page et al. 2007) and small (Cameca 7f) radius ion microprobes, and the analysis
time can be remarkably short (6-10 min per spot). However, achieving an appropriate flat pol-
ish of the surface to be analyzed and other aspects of sample preparation requirements can be
time consuming and challenging. The author’s own use of several large-radius, dual collector
instruments yielded external precision of 0.15-0.24‰ on a single spot analysis. NanoSims, an
instrument that is used chiefly for nanoscale imaging (e.g., Badro et al. 2007), provides permil
precision of submicron particles and areas.
The UV laser fluorination methods (Rumble et al. 1997; Young et al. 1998, 2008) have
similar to the ion microprobe precision but greater spot sizes; the advantage of UV laser fluori-
nation is the ability to precisely measure 17O within single minerals that is important for extra-
terrestrial applications (e.g., Young et al. 2008 and references therein).
While CO2 laser fluorination remains the most precise technique for single crystal isotope
studies, the improved techniques of large radius ion microprobe analysis for single spots makes
the latter appropriate for studying 1-2‰ level isotope variations required for high-temperature
igneous environments. These two methods are likely to complement each other for quite some
time in the future.
δ18O heterogeneity and ∆18Ocrystal-melt disequilibria
At equilibrium (Figs. 2-3), a volcanic rock should have consistent δ18O values of minerals
and ∆18Omineral-mineral and ∆18Omineral-melt isotope fractionations. However, this chapter documents
that examples of isotope disequilibria recognized by single phenocryst isotope studies are more
common than previously thought.
Oxygen isotope disequilibria may characterize selected xenocrystic crystals in a popula-
tion or be the property of the majority of crystals in a rock. The second case clearly requires
interpretation of melt and crystal origin in the course of a single petrogenetic process. It is
also possible that one mineral may have disequilibrium ∆18O values while others are in high-
temperature equilibrium.
The sign and magnitude of ∆18Ocrystal-melt disequilibria may be variable, i.e., magma may be
lighter or heavier than the measured or computed values of magma required for equilibrium with
each mineral that occurs in it (Fig. 2-3). If a low-δ18O magma mixes with a high-δ18O magma,
or by assimilating high-δ18O country rock, the ∆18Ocrystal-melt may be small or negative. As an ex-
ample, magmatic values of refractory minerals that survive slow plutonic cooling, such as zircon
and garnet (Lackey et al. 2006), may have a variable sign for the ∆18Omineral-mineral fractionation.
456 Bindeman
Petrogenetic processes leading to diverse δ18O values and ∆18Ocrystal-melt disequilibria.
Unlike many radiogenic isotope ratios that can suffer significant changes at <1% addition of
isotopically-contrasting component, oxygen is a major element. Modification of δ18O value by
more than several tenths of 1 permil requires volumetrically-significant mass transformations,
by many percents to tens of percent. Such proportions bring mass and heat balance problems
to the forefront of the discussion of crystal and magma origins. The following processes are
proposed to affect both δ18O values, and cause ∆18Ocrystal-melt disequilibria.
(i) Interaction with waters or brines derived from country rocks or stoped blocks (e.g.,
Friedman et al. 1974; Muehlenbachs et al. 1974; Hildreth et al. 1984; Taylor 1986)
(ii) Rapid assimilation of rocks significantly different in δ18O that affects magma and not
protocrysts (Taylor 1986; Balsley and Gregory 1998; Spera and Bohrson 2004).
(iii) Partial melting of rocks with groundmass that has suffered hydrothermal alteration
followed by mixing with more normal magma (e.g., Bacon et al. 1989).
(iv) Complete or bulk melting of shallow hydrothermally-altered rocks followed by an
eruption of magma at the surface (e.g., Bindeman and Valley 2001).
It should be noted that sometimes oxygen isotopic values are the only evidence of petrogenetic
processes; this is particularly true when magma exchanges with predecessor rocks of the same
chemical or isotopic values (e.g., Bindeman et al. 2008b). Lack of radiogenic ingrowth often
renders radiogenic systems to be unable to resolve sources that were in supracrustal environments
before being melted (e.g., Lackey et al. 2005). In order to be preserved, transformation of melt
δ18O that leads to ∆18Ocrystal-melt disequilibria should proceed faster relative to oxygen diffusive
timescales for the largest crystal size, so that reequilibration processes by intracrystalline oxygen
diffusion and/or solution-reprecipitation have not erased the memory of the processes described
above. For that matter, volcanic rocks and especially pyroclastic igneous rocks preserve more
evidence of disequilibria since volcanic eruption quenches their isotope values (Auer et al.
2008). Fresh tephra, scoria, and pumice clasts may provide the best material to study since they
undergo the most rapid cooling, while other volcanic products can suffer posteruptive growth
and reequilibration.
Toward isotopic equilibrium: diffusion vs. solution reprecipitation. It is likely that both
intracrystalline diffusion and solution reprecipitation play a role in isotope equilibria. Figures 7
and 11 present two contrasting examples on how to recognize these two processes given intrac-
rystalline zoning patterns. Element mapping, backscatter-electron, and cathodoluminescence
imaging are appropriate tools to recognize one from another in the course of isotope investiga-
tion. Furthermore, crystal size distribution of minerals aids in deciding the presence or absence
of solution-reprecipitation episodes and on quantifying the amount of material redeposited
(Simakin and Bindeman 2008). In particular, abundant lognormal crystal size distributions of
zircon and quartz (Fig. 5) suggests that tens of percent of zircon mass was dissolved and repre-
cipitated, and that the smallest crystals, as well as the rims of the larger crystals, were recycled
more than once. The reprecipitation of the rim will not affect the cores of the largest crystals
that should retain distinct isotopic values, which are only possible to anneal by intracrystalline
diffusion.
Rates of intracrystalline oxygen diffusion vary by 4 orders of magnitude for different
minerals (Fig. 8). Increases in oxygen and water fugacity typically decreases oxygen diffusion
rates for zircon, olivine, and quartz (Ryerson et al. 1989; Farver and Yund 1991; Watson and
Cherniak 1997). Rates of solution reprecipitation are highly variable and depend strongly on
the driving forces—undersaturation caused by changes in temperature or composition (e.g.,
Watson 1996), and are kinetically-limited by the diffusion of the components such as silica
through the melt (zhang et al. 1989). When driving forces for dissolution are great (e.g., at large
undersaturations) dissolution may erase evidence of early intracrystalline isotope diffusion and
Single Crystal Analysis of Oxygen Isotopes in Magmas 457
(a)
6
0.69±0.03Ma
4 1
7 4
δ18O, ‰
2
3
0 2 8
5 6
-2
(b) 0 50 100
Distance, µm
(b)
5
9 Core
4
1 5
3
δ18O, ‰
2
10
1
0Rim 8
-1 4 3
2 11
-2
0 50 100 150 200 250 Figure 7a
Distance, µm
Figure 7. Cathodoluminescence image of isotopically zoned, older zircons from low-δ18O rhyolites of
Yellowstone and ion microprobe profiles. Sharp boundaries in isotopic composition indicate that solution
and re-precipitation in (b) played more important role than intracrystalline diffusion (e.g., bell-shaped
profile) in (a). The solution and re-precipitation generated step-function isotope profiles, with higher δ18O
cores surrounded by the low-δ18O rims; faster rates of solution and re-precipitation are capable of erasing
the evidence of intracrystalline diffusion. Modified after Bindeman et al. (2008b).
leave sharp boundaries of oxygen isotope distribution within zircon (Fig. 7). Figure 7b
The amount of δ O heterogeneity and the ∆ Omineral-melt disequilibria can be used to
18 18
predict mineral diffusive timescales if it is assumed that exchange between mineral and melt is
rate limited by slow intracrystalline oxygen diffusion. If there is textural evidence that solution-
reprecipitation played a more important role in isotope reequilibration, then timescales obtained
using diffusion coefficients and intracrystalline diffusion provides the maximum time the
crystals could have resided in isotopically distinct melt.
Part III: CaSe StudIeS OF OxyGen ISOtOPe dISequILIBrIa In
IGneOuS rOCKS BaSed On SInGLe CryStaL ISOtOPe anaLySIS
Below a review is provided on the examples of oxygen isotope disequilibria found in
archetypal examples of basic and silicic magmatism around the world, with emphasize on
olivine-basalt, plagioclase-basalt, zircon-rhyolite, and quartz-rhyolite disequilibria that
458 Bindeman
T, °C pertain to the origin of mafic and
1200 1000 750 silicic magmas and their crystals.
The examples given here clearly
demonstrate the need to explore
many more igneous systems for
-15 Cations in olivine disequilibria relationships, and that
Fe-Mg, Mn, Ca, Ni Quartz single phenocryst isotopic studies of
refractory minerals such as olivine
and zircon, that can only crystallize
-16 Olivine from, or exchange with, the high-
Diffusion Coefficient, m2/s
Albite
temperature melt, document the
tremendous variety of δ18O melts
Anorthite that are present in the crust and the
-18 mantle.
Diopside
Sphene Basaltic igneous systems: olivine
in basalts
-20 Iceland. The most extreme
Zircon example of heterogeneous olivine
δ18O values (3‰ range), and
∆18Omelt-olivine and ∆18Omelt-plagioclase
-22 disequilibria is within the large-
6 7 8 9 10 volume basalts of Iceland, as
described by Bindeman et al.
10000/T (2006b; 2008c). The greatest 2.2-
Figure 8. Experimental studies of diffusion with common 5.2‰ range of olivine δ18O values
igneous minerals show four orders of magnitude difference is measured in the products of
in oxygen diffusion coefficients, and that in olivine cation
1783-1784AD, 15 km3 Laki fissure
diffusion is faster than oxygen diffusion. As a result,
isotopically-zoned xenocrysts with annealed compositional eruption (Fig. 9) and individual
Figure 8
zoning should be common. Data sources: zircon, Watson and plagioclase phenocrysts in the same
Cherniak (1997), wet; Sphene, Morishita et al. (1996); quartz, samples also exhibit disequilibrium
Farver and Yund (1991), wet; Olivine, Ryerson et al. (1989), relations with olivine and melt.
qFM; Diopside, Farver (1989); Anorthite and albite, Elphick
et al. (1988, 1986) wet. Furthermore, the basaltic tephra
erupted over the last 8 centuries
and as late as in November 2004
from the Grímsvötn central
volcano, which together with Laki are a part of a single volcanic system, is indistinguishable
in δ18Omelt from Laki glass. All basalts have exceptionally low-δ18O melt value of 3.1 ± 0.1‰,
homogeneous among 15 km3 of Laki basalt. This suggests that they tapped a homogeneous
and relatively long-lived, well mixed low-δ18O magma reservoir that retained its distinct δ18O
as well as (226Ra/230Th), (230Th /232Th), 87Sr/86Sr and trace elemental values (Sigmarsson et al.
1991). In order to generate such low-δ18O value, digestion of tens of percent of low-δ18O crust
is required.
The preservation of oxygen isotope disequilibria between olivine, melt and plagioclase
(Figs. 9, 10) suggests that no more than a hundred years has elapsed since incorporation of
plagioclase into basaltic melt because plagioclase has relatively fast oxygen isotope diffusion
(Elphick et al. 1986, 1988; Fig. 8). Likewise, time estimates based on mineral diffusive
timescales of oxygen and cations in olivine crystals suggests crystal residence for hundreds of
years in a well-mixed reservoir under Grimsvotn (e.g., Bindeman et al. 2006b). Similarly short
(<102 yr) time estimates were obtained by Gurenko and Chaussidon (2002) and Gurenko and
Sobolev (2006) in their oxygen isotope study of isotopically-zoned olivines from Midfell area.
Single Crystal Analysis of Oxygen Isotopes in Magmas 459
δ18 OOlivine :
96-Laki MORB
L36 Iceland
5 Plume
L14 Glass
δ18O, ‰ 4 Plag
Olivine
Cpx
3 Eq. Plag
Equilibrium Olivine
2
1450 1783-4 1996 2004
1200 1500 1998
Calendar Year, AD
Figure 9. Extreme isotope heterogeneity and disequilibria of individual olivine and plagioclase crystals
in the Laki–Grímsvötn magma system based on historical tephra and products of the Laki eruption. There
have been relatively constant δ18Omelt values for eight centuries and extreme diversity in δ18O values of
“pheno”crysts. The δ18O values of An75 plagioclase and olivine in equilibrium with 3.1±0.1‰ melt areFigure 9
shown. In the 1996 eruption, plagioclase spans a far wider range than in prior events, which is interpreted here
to indicate shorter residence time and later entrainment of feldspars in magma (Bindeman et al. 2006b).
Olivines in other large-volume Holocene basalts of Iceland also display remarkable
variety in their δ18O and Fe-Mg zoning patterns (Fig. 11). The preservation of Fe-Mg and
Ni zoning in some of those olivines suggests that these grains spend less than 100 years in
the magma prior to eruption (e.g., Costa and Dungan 2005). Ion microprobe oxygen isotope
profiling of single olivine crystals has revealed the presence of crystals that are not zoned but
are surrounded by a thin rim (preeruptive xenocrysts); crystals that are zoned with respect
to Fe-Mg and δ18O (zoned crystals); and crystals that are only zoned with respect to δ18O
but not composition (compositionally equilibrated, annealed crystals). Moreover, while the
majority of crystals have high-δ18O cores and low-δ18O melt-equilibrated rims, there were a
few crystals with the reverse relationship- crystals with high-δ18O cores that are diffusively
grading into a lower δ18O rim, and crystals with relatively abrupt compositional and δ18O
overgrowth boundaries (Fig. 11). The proportion of “short-residence” zoned crystals with thin
overgrowth rims is variable from unit to unit. The majority of crystals in Laki with relatively
constant Fo75 composition and subtle residual Ni, Ca, Mn zoning requires longer residence
(100s of years).
It appears that the best model to explain isotope disequilibria and heterogeneous δ18O
values in minerals in large volume Holocene basalts from Iceland include magmatic digestion
and erosion of Pleistocene hyaloclastites (Bindeman et al. 2008c). Mechanical and thermal
magmatic erosion characterizes many lava tubes (Williams et al. 2001, 2004) and it is likely
to occur upon basaltic magma flow in the complex network of dikes and sills of the Icelandic
rift-related magma plumbing system.
Hawaii. Interesting evidence on ∆18Omelt-olivine disequilibria is coming from another classic
example of basaltic volcanism—Hawaii. Garcia et al. (1998) analyzed coexisting bulk olivine
and matrix glass from the 1983-1997 Puu Oo eruption episode of Kilauea volcano in Hawaii
460 Bindeman
6 6 6 6
96-Laki
IC-10
IC-28
Olivine, δ18O, ‰ 96-Laki
IC-18
IC-10
IC-28
IC-18
IC-29
IC-29
IC-11
IC-13
IC-11
IC-13
IC-17
IC-27
IC-17
IC-27
Olivine, δ18O, ‰ IC-14
IC-14
Thordarhyrna,
Thordarhyrna, IC-29 IC-29
Olivine, δ18O, ‰
5
Olivine, δ18O, ‰
5 5
5 Eldgja IC-10,11
Eldgja IC-10,11
Veidivotn,
Veidivotn, IC-13, 15IC-13, 15
4 4
Veidivotn,
Veidivotn, IC-14 IC-14
4 4
6 6 Veidivotn,
Veidivotn, IC-18 IC-18
96-Laki
IC-10
IC-28
IC-18
Veidivotn,
IC-17 IC-17
IC-29
Veidivotn,
IC-11
IC-13
3
IC-17
IC-27
3
IC-14
equilibrium
equilibrium Pre-Laki, IC-27-28
3 3 Thordarhyrna, IC-29 Pre-Laki, IC-27-28
Olivine, δ18O, ‰
5 Olivine, δ18O, ‰
5 a a b b Laki, 96-Laki
Laki, 96-Laki
2 ±1 SD ±1 SD Eldgja IC-10,11
2
36 3 3.5 3.5 4 4 4.5 4.5 65 65
70 70
75 75 Veidivotn,
80 80
85 85
90
IC-13, 1590
IC-10
Groundmass, δ18O, ‰
δ18O, ‰ Forsterite, %
IC-29
4 Groundmass, Forsterite, %
IC-11
IC-17
Veidivotn, IC-14
IC-14
4
SIMS olivine
SIMS olivine Thordarhyrna, IC-29 Veidivotn, IC-18
Olivine, δ18O, ‰
5 Laser Olivine
Laser Olivine Veidivotn, IC-17
3 Eldgja IC-10,11
equilibrium Laser Plagioclase
Laser Plagioclase Pre-Laki, IC-27-28
3 Veidivotn, IC-13, 15
±1 SD
a b Laki, 96-Laki
Veidivotn, IC-14
2 4
3 3.5 4 4.5 65 70 75 Veidivotn,
80 IC-18 90
85
Figure 10. Widespread
18 ‰ isotopic heterogeneity of olivine and plagioclase crystals in basalts from Iceland
and ion microprobe Forsterite, % IC-17
Groundmass, Veidivotn,
analyzed by δlaserO,fluorination (SIMS). Sample number and corresponding source
brium center are shown; each symbol corresponds to separate lava flow.
Pre-Laki, a) δ18O values of glass and olivines
IC-27-28
SIMS olivine18 3
vs. δ
a Olivine O value of their host groundmass (measured
b by laser) arranged
Laki, 96-Laki by sample (unit) number. Olivines
±1 SD Laser in each sample are lower than the mantle olivine by more than 1‰, and are higher than the expected
4 equilibrium
Laser Plagioclase
4.5 65 value70of olivines
75 with80their host
85 groundmass.
90 The field labeled equilibrium denotes δ18Oolivine
values that would be in equilibrium with their host melt; calculated equilibrium fractionation between
dmass, δ18O, ‰ Forsterite, %
olivine and these Icelandic melt compositions of 0.7±0.3‰, defining the range of 0.4-1‰ were obtained
using the parameterization of Bindeman et al. (2004). b) δ18O values vs. forsterite content of olivines Figure
Figure 10 10
analyzed by ion microprobe and electron microprobe in the same spot. Average standard deviation (±1 SD)
on single spot analysis by the ion microprobe are ±0.24‰, and ±0.1‰ by the laser fluorination. There is a
lase positive correlation of δ18O vs. %Fo for some, while in others there is a distinct bimodal distribution. From
Bindeman et al. (2008c).
and observed moderate ∆18Omelt-olivine disequilibria deviating by up to 0.8‰ from the estimated
equilibria ∆18Omelt-olivine = 0.7±0.1‰ (Fig. 12a). The majority of variation characterizes
Figure 10 glass
that is too light for rather constant δ18Oolivine values. Both olivine and matrix glass in this
prolonged eruption are variably depleted by 0.3-0.8‰ relative to the normal mantle and thus
require assimilation or exchange with low-δ18O hydrothermally-altered rocks inside the Kilauea
edifice. It appears that little or no correlation exists between radiogenic isotopes, trace element
ratios and oxygen isotope parameters with the exception that late
Figure 10erupted, more magnesian
basalts of the Puu Oo eruption were closer to mantle δ18O values and had smaller ∆18Omelt-
olivine disequilibria. Garcia et al. (1998) noticed that the lavas that traveled in the shallow, near
horizontal, magma pathways for longer periods of time (Fig. 12b) exhibited greater levels of
depletion and ∆18Omelt-olivine disequilibria. Although these authors had difficulties finding the
right mass balance or a particular mechanism of 18O/16O preeruptive exchange, the fact that
variable ∆18Omelt-olivine and δ18Omelt values occurred in the same eruption limits the timescale of
magma-rock interaction to a duration of preeruptive residence, of months to years.
In a subsequent study of the Kilauea volcanic record, Garcia et al. (2008) described even
larger variability in ∆18Omelt-olivine and δ18Omelt values in products of 1650-2000AD historic
eruptions from Kilaeua summit. In contrast to the Grimsvotn-Laki system described by
Bindeman et al. (2006b), the Hawaiian record indicated variable δ18Omelt values and relatively
constant δ18Oolivine values that led Garcia et al. (2008) to suggest that melt depletion happened
Single Crystal Analysis of Oxygen Isotopes in Magmas 461
Figure 11. Mg- elemental
maps showing various types of
elemental and isotopic zoning
in olivines from large-volume
basalts from Iceland (Binde-
man et al. 2008c). Ovals are
pits left after ion microprobe
analyses of O isotopes. (a-b)
grains with normal compo-
sitional and isotopic zoning
with decreasing Fo and δ18O
toward the rim; (c) grain
with compositional zoning
but without isotope zoning;
(d) compositionally unzoned
grain with subtle increase in
δ18O toward the rim. [Used
with permission of Oxford
University Press from Garcia
et al. (1998) Journal of Pe-
trology, Vol. 39, p. 803-817.]
after olivine grew, perhaps shortly pre- or syneeruptively (Fig. 12b), and that the difference
between Kilauea and Grimsvotn is related to the much smaller magma chamber under the Figure
former. There appears to be a correlation of δ18Omelt with Pb and Sr isotopes of hydrothermally-
altered more radiogenic Mauna Loa lavas, and thus they may serve as a viable assimilant for
historic Kilauea lavas.
Isotopically homogenous olivines in Kilauea provide a nice natural analogue to constrain
rates of oxygen exchange in olivine. This example suggests that days to months of residence of
originally normal-δ18O olivines in low-δ18O melt was insufficient to cause heterogeneous δ18O
values in olivine. The examples of large volume basalts from Iceland excluding Laki (Figs. 10-
11) continue the trend of exchange and demonstrate that longer olivine residence time in these
basalt has led to olivines acquiring low-δ18O values and preserving δ18O heterogeneity. In
Laki, the maximum δ18Oolivine heterogeneity and ∆18Oolivine-melt disequilibria indicate the longest
residence of hundreds of years (e.g., Bindeman et al. 2006b).
Kamchatka. In a study of olivine-matrix relationship of Klyuchevskoy volcano
in Kamchatka, the largest in Eurasia, Auer et al. (2008) found a large ∆18Omelt-olivine range
(Fig. 13) with both positive and negative deviations from equilibrium. Unlike Hawaii and
Iceland with low-δ18O values, the prolific and tall Klyuchevskoy volcano displays the highest
known δ18Oolivine and δ18Omelt values (Dorendorf et al. 2000; Auer et al. 2008). While most
magmas in Klyuchevskoy and the surrounding Central Kamchatka Depression area are high-
(b)
462 Bindeman
(b)
(a)
(a) Figure 12. Olivine-Matrix dis-
equilibria in products of Kilaeua
eruptions. a) Disequilibria in 1983-
1997 Puu Oo eruption. b) Magma
pathways within Kilauea volcano.
[Used by permission of Oxford
University Press from Garcia et al.
(1998), J Petrol, Vol. 38, Fig. 1, p.
805.]
(b)
Figure 12a
Figure 12a
δ18O (Portnyagin et al. 2007), the Mg rich component is more normal-δ18O, while high-Al
Figure 12b
basalts exhibit the highest δ18Oolivine values of 7.6‰. Both high-Mg and high-Al basalt are
hydrous with up to 7 wt% water in Fo80 melt inclusions (Auer et al. 2008).
Given that variable in δ18O olivines also exhibit Fe-Mg disequilibria, and given the
correlation of δ18Oolivine with Al2O3 and MgO content of the parental melt, the best explanation
for Klyuchevskoy is pre-eruptive mixing between variable δ18O cumulates and high-Al and
high-Mg basalts, as well as direct magma mixing between high-δ18O, high-Al, and lower-δ18O,
high-Mg basalts that are observed in several historic eruptions. Isotopic and compositional
heterogeneity of Klyuchevskoy basalts (Fig. 13) is explained by the short time of weeks these
olivines spend in magma.
Summary. A histogram of δ18Oolivine values measured in the samples discussed above (single
crystals and bulk of several or size fraction crystals) are plotted in Fig. 14 and compares them
with the two previously published compilations of bulk analysis: δ18Owhole rock (Harmon and
Hoefs 1995, conventional methods), and δ18Obulk olivine values (Eiler 2001, laser fluorination). It
can be seen that the single crystal based studies document far greater δ18O range than could have
possibly been anticipated by bulk analysis of either whole rocks or multi-crystal concentrates.
Olivine is a refractory, early-crystallizing, high-temperature phenocryst, and thus the existence
Single Crystal Analysis of Oxygen Isotopes in Magmas 463
Figure 13. Oxygen isotopic heterogeneity of olivine phenocrysts and their host groundmass for historic
or recent (<7 ka) tephra and lava samples from Klyuchevskoy volcano, Kamchatka, Russia. Notice that
olivines in a single sample (defined by horizontal arrays) spans up to a few permil to the left and right
of equilibrium ∆18Oolivine-groundmass values shown at respective temperatures by black lines. This extreme
δ18Oolivine diversity is accompanied by significant Fe-Mg disequilibria between olivine and groundmass.
The disequilibria and heterogeneity in this case is explained by short-term pre-eruptive mixing between
variable-δ18O magma with variable δ18O olivines, and by preeruptive mixing between cumulate and
magma. From Auer et al. (2008).
Figure 13
40
Figure 14. Histogram of δ18O values of HH95
forsteritic olivine measured largely as
single phenocrysts in selected examples E01
30
of basaltic igneous rocks discussed in this
work. Hatched pattern marked E01 indicates
Frequency
the majority of olivine values in mantle
derived magmas based largely on bulk mono- Hawaii
20
mineralic separates summarized by Eiler
(2001), while thin horizontal line indicates
δ18O values of basalt (-0.5 permil for basalt- Kamchatka
olivine fractionation) from Harmon and 10
Hoefs (1995). Notice that refractory single
olivine records great variety of δ18O values of
melt from which it crystallized or exchanged Iceland
with exceeding either laser-based (E01) or 0
conventional, whole-rock values (HH95). 1 2 3 4 5 6 7 8 9
δ18 O(Olivine), ‰
of such heterogeneity demonstrates the existence of basaltic melts either in the mantle or the
crust, that span 6‰.
Silicic igneous systems
Figure
Isotopic analysis of individual phenocrysts or intra-crystalline domains is a novel tool to
fingerprint crystal sources and to recognize separate magma batches of silicic magmas or solid
sources from which the crystals once crystallized or were trapped (Tepley et al. 1999; Binde-
man and Valley 2001; Wolff and Ramos 2003). Below, we discuss oxygen isotope disequilibria
464 Bindeman
with the examples of quartz and zircon, two common phases of constant composition and thus
simple ∆18Omineral-melt fractionation (e.g., Fig. 3). zircon has a particular advantage for oxygen
isotope studies: it is one of the most refractory, alteration-resistant minerals that retains oxygen
isotopic values (Watson 1996; Valley 2003), it is a prime mineral for U-Th-Pb geochronology,
and it is a prime mineral for other isotopic (e.g., Hf) and trace elemental studies (Hanchar and
Hoskin 2003).
Isotope zoning and Residual memory in Zircon. U-Pb ion microprobe dating of zircon
crystallization in volcanic samples demonstrates that zircon crystallization (and closure to
subsequent U-Pb exchange) commonly predates the eruption of host lavas as determined by Ar-
Ar dating. This has been documented for a range of different sized magma systems. Miller and
Wooden (2004) found that zircons in Devils Kitchen rhyolite (Coso volcanic center, California)
are up to 200 k.y. older than corresponding K-Ar ages, and several zircon populations with
different pre-eruptive histories could be identified, signifying an origin from different magma
batches. Simon and Reid (2005) suggested that zircons in the Glass Mountain rhyolites (Long
Valley, California) record episodes of punctuated and independent evolution rather than the
periodic tapping of a long-lived magma chamber. U-Pb zircon ages in combination with Ar-
Ar potassium feldspar ages for the Geysers pluton (California Coast Ranges) indicate that the
shallow portions of the pluton cooled to <350 °C within ~200 ka, whereas at the same time
zircons from just solidified granitoids at deeper levels became remobilized by the heat of newly
intruded magma (Schmitt et al. 2003a,b). Charlier et al. (2005) found zircon ages spanning
100 k.y. in the Taupo volcanic zone in New zealand, and interpreted zircons to be derived by
bulk remobilization of crystal mush and assimilation of metasediment and/or silicic plutonic
basement rocks. In their study of Crater Lake volcanic rocks, Bacon and Lowenstern (2005)
identified parental rocks as a source for antecrystic zircons and plagioclase in the form of co-
erupted granodioritic blocks and magmas. These studies did not reach a universal conclusion
on the state of silicic magma bodies in the crust, but favored variably “long” zircon residence in
“mushy” upper crustal magma chambers (e.g., Vazquez and Reid 2002; Reid 2003).
Oxygen isotope analysis of U-Pb dated zircon populations provides an important additional
insight into the origin of zircons and their host magmas. Many large silicic magma systems
display remarkably different degrees of oxygen isotope disequilibria and diversity in their
phenocryst populations.
Case studies of large silicic magma systems
Bishop tuff. On one extreme, there is a nearly homogeneous δ18O magma body parental to
Bishop tuff (Fig. 3b, Bindeman and Valley 2002) in which zircon appears to be in perfect ∆18O
equilibrium with other minerals; on the other extreme there are rhyolites from Yellowstone,
with tremendous diversity in age, δ18O and ∆18Ozircon-melt values (Fig. 15). The erupted Bishop
tuff magma body was equilibrated with respect to the major element oxygen (Fig. 3b) and
did not retain zircon or other phenocrystal evidence of inheritance from the Glass Mountain
magmas (Simon and Reid 2005). The pre-Bishop Tuff, Glass Mountain rhyolites, erupted over
a time-span exceeding 1 Ma, exhibit heterogeneity with respect to Sr, Nd, Pb, and O isotopes
between different domes (Davies and Halliday 1998; Bindeman and Valley 2002; Simon
and Reid 2005). Subsequent accretion of these magma batches led to the formation of the
Bishop Tuff magma body, which averaged isotopic differences in the melt, equilibrated δ18O
in minerals, and rejuvenated U-Pb ages of zircons (Reid and Coath 2000). Several other large
volume rhyolites: Cerro Galan, Toba, Fish Canyon, both monotonous-intermediates, and high-
silica rhyolites, exhibit relative isotope homogeneity between early and late eruption products
(Bindeman and Valley 2002).
Yellowstone. On the other end of the spectrum, Yellowstone intracaldera volcanic rocks
belonging to the Upper Basin lavas are an example where almost the entire population of zircons
is inherited from precaldera source rocks spanning 2 m.y., based on both U-Pb zircon ages and
Single Crystal Analysis of Oxygen Isotopes in Magmas 465
LCT Zircon cores Glass
8
Quartz Sanidine
HRT
C
BC
6 B MBB DR
SBB
CF
δ18O ‰ VSMOW
MFT
4 SP
SCL
-2 Central Plateau
±0.2‰ Upper Basin lavas
lavas
-4
2.0 1.8 1.3 0.60 0.55 0.50 0.45 0.25 0.20 0.15 0.10
Eruption Age, Ma
Figure 15. Isotope disequilibria and diversity of δ18O values of zircons and quartz from selected rhyolitic
lavas and tuffs from Yellowstone Plateau volcanic field, Wyoming, USA measured by the large radius ion
microprobe. Analyses of δ18O are plotted vs. Ar-Ar eruptive age (Gansecki et al. 1996; Lanphere et al.
2002); the age of the SBB is by U/Th dating of zircon rims, (Watts et al., unpublished data); BC age is by
K-Ar (Obradovich (1992). HRT, MFT, and LCT are Huckleberry Ridge (Members B and C), Mesa Falls,
and Lava Creek tuff caldera forming eruptions; Unit abbreviations: BC - Blue Creek, MBB and SBB are
Middle and South Biscuit Basin, DR- Dunraven Road, SCL- Scaup Lake, SP- Solfatara Plateau. Modified
after Bindeman et al. (2008b).
δ18O values (Bindeman et al. 2001; 2008b). Tremendous ranges of δ18O values of phenocrysts
and variable ∆18Omineral-melt and ∆18Oquartz-zircon disequilibria in the majority of 6000 km3
Figure 15
volcanic rocks in Yellowstone (Fig. 15) are revealed by individual quartz phenocryst analysis
by laser fluorination, and individual zircon analysis by ion microprobe. These rhyolites are the
products of nearly wholesale remelting and recycling of hydrothermally altered materials from
earlier eruptive cycles, and they preserve extreme oxygen isotopic variability and zoning in
phenocrysts, including quartz (Bindeman and Valley 2001).
Timber Mt. Caldera Complex, Nevada. Intermediate between Yellowstone and Long
Valley, are the large volume low-δ18O Ammonia tanks rhyolites of Timber Mountain caldera
in Nevada (Fig. 16) that exhibit ∆18Ozircon-melt disequilibria but demonstrate ∆18Oquartz-melt and
∆18OSphene-melt equilibria. Neither different parts of the large Ammonia Tanks tuff magma
body, exemplified by the study of individual pumice clasts dispersed by the caldera-forming
eruptions (Mills at al. 1997; Tefend et al. 2007), nor crystal populations, exemplified by the
study of phenocrysts (Bindeman et al. 2006a), were unable to achieve whole-rock isotopic
and chemical equilibration. Diversity of δ18O zircon values in Ammonia Tanks rhyolites is
supported by the diversity of their ages that spans the entire history of Timber Mountain
caldera complex (Bindeman et al. 2006a).
Recognizing magma batches using individual phenocrysts. At Yellowstone and Timber
Mt. caldera complexes, lavas and tuffs with poly-age and diverse δ18O zircon populations
represent accretion of independent homogenized magma batches that were generated rapidly by
remelting of source rocks of various ages and δ18O values. Sufficient magma residence time may
allow annealing δ18O and ∆18O values of quartz, then sphene, and then zircon. However, isotope
zoning in zircon in these two caldera complexes within even large-volume rhyolites suggest
that “residual memory” of accretion and magma derivation from low-δ18O hydrothermally-
altered rocks has not yet been erased (Fig. 17). Large volume tuffs of Yellowstone and Timber
466 Bindeman
3 TS TC RM AT
Δ 18O(Qz-Zrc):
bulk Zrc
Δ18O(Qz-Zrc), (Melt-Zrc) Equilibrium Zrc cores,
Δ18O(Qz-Zrc) Small zircons
2 <53um
Equilibrium Δ18O(melt-Zrc):
Δ18O(melt-Zrc) bulk Zrc
Zrc cores
1 Small zircons
<53um
Less than equilibrium
Δ 18 O(Qz-Zrc), and Δ 18 O (melt-Zrc)
of the Ammonia Tanks cycle
0
12.8 12.7 11.6 11.45
Eruption Age, Ma
Figure 16. zircon-quartz and zircon-melt oxygen isotope disequilibria in large volume rhyolites from
Timber Mt/Oasis Valley caldera complex, Nevada, using size fraction analysis of zircons by laser
fluorination. Large volume units Topopah Spring (TS), Tiva Canyon (TC), and Rainier Mesa (RM) show
equilibrium relations, while zircons in all four of Ammonia Tanks (AT) cycle samples have smaller than
equilibrium fractionations vs. quartz, due to the inherited high-δ18O cores which are present even in smaller
zircons; Smaller zircons show closer equilibrium with quartz and melt than larger zircons, because the latter
have higher- δ18O cores. Sphene and quartz in these samples are in isotopic equilibrium. Four samples in
the Ammonia Tanks cycle are plotted in accordance with their relative eruption sequence at around 11.45
Ma. Modified after Bindeman et al. (2006a).
8 5
Mesa Falls tuff Huckleberry Ridge tuff
Member-C
6
4
±1st dev
Figure 16
Number
Number
3
rims
4
±1st dev
2
cores
2
1
0 0
2.5 3 3.5 4 4.5 3.5 4 4.5 5 5.5 6
δ18 O ‰ δ18 O ‰
8 6
Lava Creek tuff Ammonia Tanks tuff
cores 5
6
Number
4
Number
±1st dev
4 3 ±1st dev
rims
rims
2
2 cores
0 0
2.5 3 3.5 4 4.5 3 4 5 6 7
δ18 O ‰ δ18 O ‰
Figure 17
Figure 17. Oxygen isotope diversity and zoning of zircons from major tuff units from Yellowstone and
Ammonia Tanks tuff of Timber Mt. caldera complex. Dashed vertical line represents equilibrium value
with the host glass. Data from Bindeman et al.. (2006a, 2008b).
Single Crystal Analysis of Oxygen Isotopes in Magmas 467
Mt. exhibit the ∆18Oquartz-melt equilibria, small ∆18Ozircon-melt disequilibria, and diverse δ18Ozircon
values (Figs. 15-16, Bindeman et al. 2006a, 2008b).
When δ18Ozircon values are plotted against the eruptive volume in three large silicic systems
(Fig. 18) there appears to be an empirical correlation that suggests decreasing ∆18Ozircon-melt
disequilibria with increasing volume to ~100-300 km3. This correlation may indicate that
magma bodies of ~100-300 km3 in size represent well-mixed reservoirs which exist for a
long enough time to anneal isotope disequilibria in zircons and other minerals. However,
greater heterogeneity that appears in larger, supervolcanic volumes of magma, may reflect
late stage or preeruptive magma batch addition or stratification. The only exception from this
rule is provided by the Kilgore tuff, 1800 km3 eruptive unit in the Heise volcanic field in
Idaho (Morgan and McIntosh 2005; Bindeman et al. 2007), which appears to have greater
homogeneity in zircon age and δ18O values.
Phenocryst heterogeneity in plutonic rocks
Plutonic rocks have undergone prolonged cooling that should in theory anneal isotope
disequilibria, and evidence of batch segregation given diffusion coefficients and times
involved (Fig. 8). However, zircon, a mineral with the slowest oxygen diffusion, demonstrates
∆18Ozircon-other mineral disequilibria in studied examples of plutonic rocks. King and Valley (2001)
and Lackey et al. (2006) (Fig. 19) studied the oxygen isotopic composition of minerals in
the Idaho Batholith and Sierra Nevada Batholith respectively, using two refractory minerals
with slow oxygen diffusion—garnet and zircon. At equilibrium, both orthosilicates should
have nearly identical δ18O values and ∆18Ozircon-garnet = 0±0.1‰ (Valley et al. 2003). King and
Valley (2001) and Lackey et al. (2006) observed that near the contact with high-δ18O country
rocks, zircon retained its average intraplutonic δ18O values, while other minerals, including
garnet, increased their 18O from the assimilation of high-δ18O country rocks. In both examples,
lower than equilibrium δ18O zircon values, and correspondingly higher than zero ∆18Ogarnet-
18
zircon values indicated late stage contamination by a high-δ O wallrock that did not affect
the previously crystallized zircon. However, it is also possible that their data indicate batch
6
Batch
residence
diversity in ‰
time
δ18O zircon n
Number of Batches
UB, Canyon flow
3
AT
CT
LC
CP, earlyy
HRT
CP, late
MFT
0
40 100 300 1000 2500
Erupted volume km3
Figure 18. Oxygen isotope diversity of zircons from major tuff units from Yellowstone and Timber Mt.
Calderas, and smaller volume rhyolitic lavas as determined by ion microprobe analysis of individual
zircons. Achieving isotope equilibrium is interpreted to represent batch residence time that increases with
unit size and seems to reach least heterogeneity for units several hundred cubic kilometers erupted as a Figure
single cooling unit such as Mesa Falls tuff. More voluminous units develop greater heterogeneity due to
the number of erupted Members. Unit abbreviations are the same as in Figures 15 and 16; CP and UB are 18
Central Plateau and Upper Basin post Lava Creek tuff intra-caldera lavas of Yellowstone.
468 Bindeman
(b)
(a)
Figure 19. Two examples of plutonic oxygen isotope disequilibria between zircon and other minerals.
a) Idaho Batholith (King and Valley 2001) showing progressively divergent ∆18Ogarnet-zircon fractionation
closer to the contact with high δ18O country rock; b) Variations in δ18O and ∆18Ogarnet-zircon across Dinkey
Dome pluton, Sierra Nevada Batholith. In both examples lower than equilibrium δ18O zircon values, and
correspondingly greater than zero ∆18Ogarnet-zircon values indicate late stage contamination by a high-δ18O
wall-rock that did not affect already crystallized zircon. Equilibrium ∆18Ogarnet-zircon = 0 value is expected
to be ~0‰ (Valley et al. 2003). [Used with kind permission of Springer Science+Business FigureMedia19 from
Lackey et al. (2005) Contrib Mineral Petrol, Vol. 151, p. 20-44, and King and Valley (2001) Contrib
Mineral Petrol, Vol. 142, p. 72-88.]
accretion similar to the volcanic examples outlined above: it is noteworthy that ∆18Ozircon-
garnet disequilibria was only preserved near intrusive contacts because these likely involved
intraplutonic quench. Both King and Valley (2001) and Lackey et al. (2006) emphasized slow
oxygen diffusion and high closure temperature for zircon, and thus the ability of zircon to
retain its original crystallized magmatic value.
In their ion microprobe study of zircons in plutonic rocks, Bolz (2001), Appleby (2008),
and Appleby et al. (2008) documented further examples of δ18O heterogeneity. The latter paper
presents detailed ion microprobe data for zircons in two diorites associated with the Scottish
Caledonian Lochnagar pluton. The diorites are very similar with respect to whole-rock
chemical composition and age but have very different zircon oxygen isotope compositions.
One diorite sample forms a homogeneous zircon population but the other covers a large range
of δ18O values of several permil. Variations mostly occur between individual zircon crystals,
but δ18O zoning is also present within a few crystals. While in the majority of cases δ18O
increases with zircon growth, as is found in the studies of Lackey et al. (2006) and King
and Valley (2001), there are individual zircon crystals with the reverse relationship in the
same sample, suggesting that several mixing events of variable δ18O batches occurred. The
same diversity and zoning pattern is also found in a larger sample suite from Lochnagar and
Etive plutons (Appleby 2008), and δ18O zoning and disequilibria, supported by U-Pb age,
differences in Hf isotopic values of zircons, and their REE values, appears to be a common
phenomenon in these two plutons.
Summary. Highly diverse δ18O zircon values of rhyolites in the examples discussed in
this Chapter (Fig. 15 and 17) can be complemented by analyses of δ18O values in single zircon
and other refractory igneous minerals (see Bolz 2001; Bindeman et al. 2008a,c; Appleby et
al. 2008) that demonstrate a picture of diverse silicic magma sources. Since zircons can only
Single Crystal Analysis of Oxygen Isotopes in Magmas 469
exchange oxygen upon crystallization from melt or through a prolonged exchange with melt,
such diverse δ18Ozircon values record the diversity of crustal melts that are coming from high-
δ18O and low-δ18O sources and then mix effectively. It is likely that plutons and large volume
batholiths are assemble by pulses (Grunder 1995; Davidson et al. 2001; Coleman et al. 2004;
Glazner et al. 2004; de Silva and Gosnold 2007) and we now find single crystal isotopic
evidence of these assembly processes. The lack of isotope zoning and disequilibria in some
large-volume rhyolites such as Bishop tuff (Fig. 3b) suggest convection and annealing due to
magma residence and these topics are discussed below.
Part IV: ISOtOPe dISequILIBrIa:
What haVe We Learned aBOut MaGMa GeneSIS?
Mineral-diffusive timescales and magma residence times
At high magmatic temperatures, timescales of diffusive equilibration of isotopes and
trace elements between crystals and magma are highly variable but are usually short (from
<1 to 100s years) relative to magma generation and segregation timescales (e.g., Costa et al.
2008). In this sense, they provide a chronological resolution that is unattainable by most other
methods, especially for older rocks in which U-series methods cannot be used (e.g., Bindeman
and Valley 2001; Costa et al. 2008). For the youngest volcanic rocks, the age of the melt could
be independently constrained by the short-lived U series isotopes, such as (226Ra/230Th) and
(210Pb/226Ra) activity ratios those half-lives of 1600 yr and 22 yr respectively are comparable
with the cation and oxygen diffusion times (Bindeman et al. 2006b; Cooper and Reid 2008).
Accessory minerals with slow diffusion coefficients, and particularly zircon (Fig. 8, Watson
and Cherniak 1997; Valley 2003) are valuable probes of transient changes by diffusion and
solution reprecipitation at high magmatic temperatures, although there appears to be growing
evidence that solution-reprecipitation plays a more important role (e.g., Page et al. 2007) and
thus diffusion timescales provide maximum time for isotope transformation.
rapid magma genesis and high magma production rates
The existence of isotopically-zoned and disequilibrium crystals that constitute the majority
of crystal populations (Figs. 9, 10, 15, 17) suggests rapid magma genesis. Hundreds to thousands
of years time seems reasonable to generate small volume igneous bodies of a few cubic kilometers
or less, given average magma production rates in arcs and plumes of 0.001-0.01 km3/yr (Dufek
and Bergantz 2005; McBirney 2006). The remarkable observation which is coming from single
crystal isotope studies is that large volume (40-1000 km3) rhyolites (Figs. 15, 17) and large
volume (3-20 km3) basalts (Figs. 9-13), all have isotope disequilbria for the majority of their
crystals. For example, the voluminous 1000 km3 Ammonia Tanks rhyolite from the Timber Mt.
caldera complex demonstrate that individual, isotopically-distinct melt batches generated by
reheating of hydrothermally-altered rocks were able to coalesce over time-scales of <150 ky, or
likely <10 ka (Fig. 20) into a ~1,000 km3 size magma body (Mills et al. 1997; Tefend et al. 2007;
Bindeman et al. 2006a). In this period of time, hundreds of cubic kilometers of hydrothermally-
altered low-δ18O protolith were melted and digested, and inherited zircons with older ages and
higher-δ18O cores survived the hydrothermal alteration and melting. High magma production
rates of ~0.01 km3/yr are thus estimated, which require intrusion of substantial volumes of
basaltic magmas on the order of many hundreds of km3, that preceded the melting process and
may require ignimbrite “flare-up” (e.g., de Silva and Gosnold 2007).
Likewise, in order to generate a Laki-size basaltic flow of 15 km3 in 1000 to 8000 years
based on mineral-diffusive timescales of isotopically-zoned crystal populations, and excess of
(226Ra/230Th), high magma production and accumulation rates of 0.002-0.02 km3/yr are required.
It should also be noted that these are minimum estimates based on the erupted volumes and
470 Bindeman
10000 yr
100 Kilgore tuff
53µm zircon bulk
Ammonia
105µm zircon cores
% exchanged, Zircon
Tanks Tuff
75
and lavas
1000 yr
50 10000 yr
300 yr
25 Yellowstone
lavas
100 yr
0
100 yr 300 yr 1000 yr
0 25 50 75 100
% exchanged, 4 mm Quartz cores
Figure 20. Timescale of magma genesis based on isotope re-equilibration between quartz in zircon within
rhyolitic melt at 850 °C using diffusion coefficients from Watson and Cherniak (1997) and Farver et al.
(1991), and diffusion in a sphere equation from Crank (1975). A mineral core constitutes the inner 50%
of diameter. The field for isotopically-diverse Yellowstone zircons and quartz is shown and requires a
500-5000 yr residence time of xenocrysts in magma. In Ammonia Tanks tuff, quartz δ18O zoning is totally
annealed, while the remaining subtle 1.5‰ zoning in zircons is consistent with 15,000-20,000 yr of
residence at the same conditions. This model predicts that disequilibria due to unequal exchange between
quartz and zircon would persist for up to 2000 yr; after that only zircon would show internal zoning lasting
for up to ~30,000 yr. Diffusion is assumed to become important when temperatures increase during melting
and to stop during eruption. The range of zircon zoning in Ammonia Tanks tuff is consistent with ~10,000
Figure 20
yr from melting to eruption, but for the Kilgore ignimbrite (Heise volcanic field) is longer (Bindeman et al.
2007). The range of zircon and quartz zoning in Yellowstone’s low-δ18O rhyolites (Bindeman and Valley
2001) and from Timber Mt. (Bindeman et al. 2006b) are consistent with shorter residence times in smaller-
volume low-δ18O lavas. While this model is based on diffusion, this solution provide maximum timescales
if solution and re-precipitation played more important role.
assuming complete evacuation of magma reservoirs and thus could be several times greater
if the erupted proportion is only a fraction of the total magma generated. The quoted magma
production rates are high but lower than Hawaii’s hot spot eruption rates of ~ 0.01-0.1 km3/yr.
Very shallow petrogenesis and the role of hydrothermal carapaces
Abundant low-δ18O signatures of crystals that we found in great abundance uniquely
fingerprint shallow magma petrogenesis since meteoric water does not penetrate deeper than
~10 km, where the closed porosity in rocks no longer allows low-δ18O waters to circulate (e.g.,
Taylor and Shappard 1986). Caldera settings and rift zone environments facilitate hydrothermal
fluid flow and shallow petrogenesis through assimilation of hydrothermally-altered carapace
around the magma chambers, which we call “crustal cannibalization.” Documented examples
of shallow level magma genesis presents significant challenges for heat and mass balance.
Mechanically, shallow crustal cannibalization processes may involve reactive bulk assimilation
(Dungan and Davidson 2004; Beard et al. 2005) which starts when interstitial melt forms
interconnected networks, causing the roofrock to collapse and to disintegrate into individual
phenocrysts or crystal clusters. The resulting crystal mush then becomes homogenized by
convection. Isotopic diffusion and re-equilibration of crystals will last until the eruptive
quench. In this model, magma cannibalizes earlier erupted rocks and the old hydrothermal
system that penetrated through them. Self-cannibalization generates a low-δ18O component
Single Crystal Analysis of Oxygen Isotopes in Magmas 471
that lowers the δ18O value of the magma. At Yellowstone and Crater Lake, large scale, nearly
wholesale melting of hydrothermally altered crust occurred rapidly and effectively enough that
a significant portion of the old crust with abundant inherited “antecrysts” of zircons, quartz,
and plagioclase (Bacon and Lowenstern 2005; Bindeman et al. 2008b) are preserved in the
erupted magmas. Likewise, oxygen isotopic heterogeneity of olivine and plagioclase crystals,
and disequilibrium relations between crystals and melt in large volume basaltic eruptions
in Iceland and basalts of Kilauea (Figs. 11-12) suggest shallow, sometimes preeruptive
modification of basalts in the upper few kilometers of crust.
Abundant evidence of shallow magma petrogenesis contradicts common knowledge
that it is easy to generate magma in the middle or lower crust where ambient heat balance
around magma bodies prevents rapid heat dissipation. Magma genesis by remelting in shallow
environments is aided by: 1) rocks in caldera and rift environments that are pre-heated by
prior long-term volcanic activity and subsidence that brings surface-altered rocks down to the
underlying heat sources (Fig. 21); 2) erupted rocks are porous, fractured and thus relatively
easy to disintegrate by a reactive assimilation processes compared to their plutonic equivalent;
3) shallowly emplaced eruptive products contain abundant volcanic glass which means that
no latent heat of fusion is required for their remelting (i.e., melting efficiency is greater for
volcanic rocks than for plutonic rocks); 4) shallow volcanic rocks in hydrothermal settings are
pre-conditioned and saturated with water that promotes flux melting.
Accretion of large silicic magma bodies
The generation and eruption of silicic magmas continues to present challenges in answering
questions about the size and longevity of crustal magma bodies, their physical state as either
stagnant, near-solidus cumulate mushes, near-liquidus convecting liquids, or initially solid
rocks, and the time and depth of their segregation (e.g., Annen and Sparks 2002; Bachmann
and Bergantz 2003, 2004; Dufek and Bergantz 2005; Bindeman et al. 2008b).
The Recycling Machine
Timescale 2: STACKING
Burial
B Figure 21. Magmatic cannibaliza-
U tion in shallow crustal environments
R and timescales of crystal recycling.
I Progressive burial by caldera col-
A lapses, rifting and/or overloading
L
brings erupted volcanic rocks and
their sub-volcanic equivalent down
deeper to the melting zone. Hydro-
thermal alteration serves as a flux
to cause melting, and glassy, po-
rous nature of these rocks makes
re-melting and disaggregation me-
chanically easier. Timescale 2 is
the age between the core and the
rim of zircon, while timescale 1 is
REMELTING the crystal diffusive/recrystalliza-
REACTIVE ASSIMILATION tion age.
Timescale 1:
LOOSE COHESIVENESS
RECYCLING Remelting-
Eruption
Figure 21
472 Bindeman
The common theme of documented examples of isotope disequilibria presented in this
chapter is batch assembly of magma bodies with initially diverse phenocryst populations,
followed by increased mixing and annealing with increased residence time. There is plenty
of radiogenic isotope evidence that the accretion of large silicic magma bodies includes
batches formed by differentiation of basaltic magmas in the crust, and by partial melting of
the radiogenic crust due to heat transport by basaltic intrusions (e.g., see Grunder 1995; Annen
and Sparks 2002, and Dufek and Bergantz 2005 for a review of the current literature). Partial
melting of the variable-δ18O crust will very often generate isotopically-distinct individual
magma batches, and isotopes of major element oxygen nicely reflect these processes, especially
when constrained by zircons. High-δ18O batches reflect supracrustal materials, while low-
δ18O values document batches that are generated by melting of hydrothermally-altered crustal
carapace. Oxygen isotope evidence from the Yellowstone and Timber Mt calderas (Figs. 15-
17) indicates that source rocks were cooled below the solidus, altered by heated meteoric
waters that imprinted the rocks with low δ18O signatures, and then remelted in distinct pockets
by intrusion of mafic magmas. Each pocket of new melt was variable in δ18O, but inherited
zircons retained earlier age and δ18O values.
The model of remelting and batch accretion presented in this chapter contradicts
the commonly held model of a single, large-volume, mushy-state magma chamber that is
periodically reactivated and produces rhyolitic off-springs.
the origin of “phenocrysts” and interpretation of their melt inclusions
The presence of δ18O zoning and variability among phenocryst populations, as discussed
in numerous examples above, suggest that crystals are captured from different δ18O sources and
thus record different pre-eruptive residence times. For example, phenocryst studies of Icelandic
basalts (Figs. 11-12) reveal crystal-scale complexity that requires olivine recycling at different
times prior to eruption. The evidence presented above suggests that prolonged annealing and
isotopic homogenization of xenocrystic or antecrystic crystal populations is capable of erasing
magma memory of crystal origin. Two views on crystal origin are possible: (1) crystals grew
from the melt in which they occur, (2) crystals are “annealed beyond recognition” and mixed
in with phenocrystic populations that grew from the melt. From the point of view of an isotope
geochemist, it is still possible to recognize diverse crystal origins, their fragmentation and
recrystallization histories (Fig. 22), and therefore fingerprint magma batches. These annealed
crystals may or may not be in chemical equilibria with the host melt and exhibit subtle isotope
evidence of diverse sources.
Heterogeneous magmatic origins of olivines in basalts clearly requires the need for
reassessment of their melt inclusions, including volatile content (e.g., Thordarson and Self
2003). The same applies to melt inclusion in isotopically annealed quartz or zircon xenocrysts
in silicic rocks, which may record melt compositions from previous eruption episodes that
predate the host magma by many hundreds thousands years. Melt inclusion compositional
studies, combined with identification of melt δ18O sources and ages (e.g., melt inclusions in
zircon), would provide powerful insights into the chemistry of assembled magma batches.
In this chapter, we presented evidence that annealed crystals are more common than
phenocrysts in many basaltic and silicic magma systems around the world. This means that
magmas and their crystals cargo may be difficult to produce, but are easy to recycle. Taken as
a whole, these results beg the question: are there any “pheno”crysts left?
aCKnOWLedGMentS
Author’s prior work with Alfred Anderson, John Valley, John Eiler, and prior and ongoing
collaborations with Leonid Perchuk, Olgeir Sigmarsson, Paul Wallace, Vera Ponomareva,
Single Crystal Analysis of Oxygen Isotopes in Magmas 473
(a) (b)
1 mm
Figure 22. Melt Inclusions in pheno(xeno)crysts: reheated, re-homogenized, decrepitated, and exchanged
with the external melt. a) Fragmented host quartz crystal from the Bishop tuff after experimental reheating
that lead to melt inclusion decrepitation and fragmentation of crystal. b) Cathodoluminescence image
Late Bishop tuff quartz showing sharp boundaries between the dark-CL core and bright-CL rim with melt
inclusions in both places (see Wark et al. 2007). Bright CL rims and dark CL cores are nearly identical
isotopically (Bindeman and Valley 2002). Image courtesy: Julie Roberge
Figure 22
Axel Schmitt, Andrey Gurenko, and Maxim Portnyagin has led to formulation and maturation
of ideas presented here. Informal reviews by Jim Palandri, Erwan Martin, Kathryn Watts,
comments by John Valley, and formal reviews by Mike Garcia, Jade Star Lackey, Keith
Putirka and an anonymous reviewer have dramatically improved presentation of this chapter.
Supported by NSF (EAR0537872) and the University of Oregon.
reFerenCeS CIted
Anderson AT Jr., Clayton RN, Mayeda TK (1971) Oxygen isotope geothermometry of mafic igneous rocks. J
Geol 79:715-729
Annen C, Sparks RSJ (2002) Effects of repetitive emplacement of basaltic intrusions on thermal evolution and
melt generation in the crust. Earth Planet Sci Lett 203:937-955
Appleby SK (2008) The origin and evolution of granites: an in-situ study of zircons from Scottish Caledonian
granites, PhD Dissertation, University of Edinburgh, Scotland
Appleby SK, Graham CM, Gillespie MR, Hinton RW, Oliver GJH, EIMF (2008) A cryptic record of magma
mixing in diorites revealed by high-precision SIMS oxygen isotope analysis of zircons. Earth Planet Sci
Lett 269:105-117
Armienti P (2008) Decryption of igneous rock textures: crystal size distribution tools. Rev Mineral Geochem
69:623-649
Auer S, Bindeman I, Wallace P, Ponomareva V, Portnyagin M (2008) The origin of hydrous, high-δ18O
voluminous volcanism: diverse oxygen isotope values and high magmatic water contents within the
volcanic record of Klyuchevskoy volcano, Kamchatka, Russia. Contrib Mineral Petrol DOI 10.1007/
s00410-008-0330-0
Bachmann O, Bergantz GW (2003) Rejuvenation of the Fish Canyon magma body: a window into the evolution
of a large-volume silicic magma systems. Geology 71:789-792
Bachmann O, Bergantz GW (2004) On the origin of crystal-poor rhyolites: extracted from batholithic mushes
J Petrology 45:1565-1582
Bacon CR, Adami LH, Lanphere MA (1989) Direct evidence for the origin of low-δ18O silicic magmas:
quenched samples of a magma chamber’s partially-fused granitoid walls, Crater Lake, Oregon. Earth
Planet Sci Lett 96:199-208
Bacon CR, Lowenstern JB (2005) Late Pleistocene granodiorite source for recycled zircon and phenocrysts in
rhyodacite lava at Crater Lake Oregon. Earth Planet Sci Lett 233:277-293
474 Bindeman
Badro J, Ryerson FJ, Weber PK, Ricolleau A, Fallon SJ, Hutcheon ID (2007) Chemical imaging with
NanoSIMS: A window into deep-Earth geochemistry, Earth Planet Sci Lett: 262:543-551
Baertschi P (1976) Absolute 18O Content of standard mean ocean water. Earth Planet Sci Lett 31:341-344
Balsley SD, Gregory RT (1998) Low-δ18O magmas: why they are so rare? Earth Planet Sci Lett 162:123-136
Beard JS, Ragland PC, Crawford ML (2005) Reactive bulk assimilation: A model for crust-mantle mixing in
silicic magmas. Geology 33:681-684
Biegelsen J, Mayer MG (1947) Calculation of equilibrium constants for isotope exchange reactions. J Chem
Phys 15:261-267
Bindeman IN, Brooks CK, McBirney AR, Taylor HP (2008a) The low-δ18O, late-stage ferrodiorite magmas
in the Skaergaard Intrusion: Result of liquid immiscibility, thermal metamorphism, or meteoric water
incorporation into magma? J Geology, (in press)
Bindeman IN, Fu B, Kita N, Valley JW (2008b) Origin and evolution of Yellowstone silicic magmatism based
on ion microprobe analysis of isotopically-zoned zircons. J Petrol 49:163-193
Bindeman IN, Gurenko AA, Sigmarsson O, Chaussidon M (2008c) Oxygen isotope heterogeneity and
disequilibria of olivine phenocrysts in large volume basalts from Iceland: Evidence for magmatic digestion
and erosion of Pleistocene hyaloclastites. Geochim Cosmochim Acta doi: 10.1016/j.gca.2008.06.010
Bindeman IN, Ponomareva VV, Bailey JC, Valley JW (2004) Volcanic arc of Kamchatka: a province with
high-δ18O magma sources and large-scale 18O/16O depletion of the upper crust. Geochim Cosmochim
Acta 68: 841-865
Bindeman IN, Schmitt AK, Valley JW (2006a) U-Pb zircon geochronology of silicic tuffs from Timber Mt
Caldera complex, Nevada: rapid generation of large volume magmas by shallow-level remelting. Contrib
Mineral Petrol 152:649-665
Bindeman IN, Sigmarsson O, Eiler JM (2006b) Time constraints on the origin of large volume basalts derived
from O-isotope and trace element mineral zoning and U-series disequilibria in the Laki and Grímsvötn
volcanic system. Earth Planet Sci Lett 245:245-259
Bindeman IN, Valley JW (2001) Low-δ18O rhyolites from Yellowstone: Magmatic evolution based on analyses
of zircon and individual phenocrysts. J Petrol 42:1491-1517
Bindeman IN, Valley JW (2002) Oxygen isotope study of Long-Valley magma system: Isotope thermometry
and role of convection. Contrib Mineral Petrol 144:185-205
Bindeman IN, Valley JW, Wooden JL, Persing HM (2001) Post-caldera volcanism: In situ measurement of
U-Pb age and oxygen isotope ratio in Pleistocene zircons from Yellowstone caldera. Earth Planet Sci Lett
189:197-206
Bindeman IN, Watts KE, Schmitt AK, Morgan LA, Shanks PWC (2007) Voluminous low-δ18O magmas in
the late Miocene Heise volcanic field, Idaho: Implications for the fate of Yellowstone hotspot calderas.
Geology 35:1019-1022
Bolz V (2001) The oxygen isotope geochemistry of zircon as a petrogenetic tracer in high temperature contact
metamorphic and granitic rocks. MS Thesis, University of Edinburgh, Scotland.
Chacko T, Cole DR, Horita J (2001) Equilibrium oxygen, hydrogen and carbon isotope fractionation factors
applicable to geologic systems. Rev Mineral Geochem 43:1-81
Charlier BLA, Wilson CJN, Lowenstern JB, Blake S, Van Calsteren PW, Davidson JP (2005) Magma generation
at a large hyperactive silicic volcano (Taupo New zealand) revealed by U-Th and U-Pb systematics in
zircons. J Petrol 46:3-32
Chiba H, Chacko T, Clayton RN, Goldsmith JR (1989) Oxygen isotope fractionations involving diopside,
forsterite, magnetite, and calcite: Application to geothermometry, Geochim Cosmochim Acta 53:2985-
2995
Coleman DS, Gray W, Glazner AF (2004) Rethinking the emplacement and evolution of zoned plutons:
Geochronologic evidence for incremental assembly of the Tuolumne Intrusive Suite, California. Geology
32:433-436
Cooper KM, Reid MR (2008) Uranium-series crystal ages. Rev Mineral Geochem 69:479-554
Costa F, Dohmen R, Chakraborty S (2008) Time scales of magmatic processes from modeling the zoning
patterns of crystals. Rev Mineral Geochem 69:545-594
Costa F, Dungan M (2005) Short time scales of magmatic assimilation from diffusion modeling of multiple
elements in olivine. Geology 33:837-840
Crank J (1975) The Mathematics of Diffusion, 2nd Edition. Oxford: Oxford University Press
Davidson JP, Tepley F III, Palacz z, Meffan-Main S (2001) Magma recharge, contamination and residence
times revealed by in situ laser ablation isotopic analysis of feldspar in volcanic rocks. Earth Planet Sci
Lett 184:427-442
Davies GR, Halliday AN (1998) Development of the Long Valley rhyolitic magma system: Sr and Nd isotope
evidence from glasses and individual phenocrysts. Geochim Cosmochim Acta 62:3561-3574
de Silva SL, Gosnold WD (2007) Episodic construction of batholiths: insights from the spatiotemporal
development of an ignimbrite flare-up. J Volcanol Geotherm Res 167:320-335
Single Crystal Analysis of Oxygen Isotopes in Magmas 475
Dorendorf F, Wiechert U, Worner G (2000) Hydrated sub-arc mantle: A source for the Klyuchevskoy volcano,
Kamchatka/Russia. Earth Planet Sci Lett 175:69-86
Dufek J, Bergantz GW (2005) Lower crustal magma genesis and preservation: A stochastic framework for the
evaluation of basalt-crust interaction. J Petrology 46: 2167-2195
Dungan MA, Davidson J (2004) Partial assimilative recycling of the mafic plutonic roots of arc volcanoes: An
example from the Chilean Andes. Geology 32:773-776
Eiler JM (2001) Oxygen isotope variations in basaltic lavas and upper mantle rocks. Rev Mineral Geochem
43:319-364
Eiler JM, Valley JW, Baumgartner LP (1993) A new look at stable isotope thermometry. Geochim Cosmochim
Acta 57:2571-2583
Elphick SC, Dennis PF, Graham CM (1986) An experimental-study of the diffusion of oxygen in quartz and
albite using an overgrowth technique. Contrib Mineral Petrol 92:322-330
Elphick SC, Graham CM, Dennis PF (1988) An ion microprobe study of anhydrous oxygen diffusion in
anorthite - a comparison with hydrothermal data and some geological implications. Contrib Mineral
Petrol 100:490-495
Farquhar J, Chacko T, Frost BR (1993) Strategies for high-temperature oxygen-isotope thermometry - a worked
example from the Laramie anorthosite complex, Wyoming, USA. Earth Planet Sci Lett 117:407-422
Farver JR (1989) Oxygen self-diffusion in diopside with application to cooling rate determinations. Earth
Planet Sci Lett 92:386-396
Farver JR, Yund RA (1991) Oxygen diffusion in quartz: dependence on temperature and water fugacity. Chem
Geol 90:55-70
Finney B, Turner S, Hawkesworth C, Larsen J, Nye C, George R, Bindeman I, Eichelberger J (2008) Magmatic
differentiation at an island-arc caldera: Okmok volcano, Aleutian Islands, Alaska. J Petrol doi:10.1093/
petrology/egn008
Friedman I, Lipman PW, Obradovich JD, Gleason JD, Christiansen RL (1974) Meteoric water in magmas.
Science 184:1069-1072
Friedman I, O’Neil JR (1977) Compilation of stable isotope fractionation factors of geochemical interest
(Geological Survey Professional Paper 440-KK. In: Data of Geochemistry, 6th Ed. Fleischer M (ed)
Washington, D.C., U.S. Gov Print Off p. 1-12
Gansecki CA, Mahood GA, McWilliams MO (1996) 40Ar/39Ar geochronology of rhyolites erupted following
collapse of the Yellowstone caldera, Yellowstone Plateau volcanic field: implications for crustal
contamination. Earth Planet Sci Lett 142:91-107
Garcia MO, Ito E, Eiler JM (2008) Oxygen isotope evidence for chemical interaction of Kilauea historical
magmas with basement rocks. J Petrol 49:757-769
Garcia MO, Ito E, Eiler JM, Pietruszka AJ (1998) Crustal contamination of Kilauea Volcano magmas revealed
by oxygen isotope analyses of glass and olivine from Puu Oo eruption lavas. J Petrol 39:803-817
Ghiorso MS, Sack RO (1995) Chemical mass transfer in magmatic processes. 4. A revised and internally
consistent thermodynamic model for the interpolation and extrapolation of liquid-solid equilibria in
magmatic systems at elevated temperatures and pressures. Contrib Mineral Petrol 119:197-212
Glazner AF, Bartley JM, Coleman DS, Gray W, Taylor Rz (2004) Are plutons assembled over millions of years
by amalgamation from small magma chambers? GSA Today 14:4-11
Grunder AL (1995) Material and thermal roles of basalt in crustal magmatism: A case study from eastern
Nevada. Geology 23:952-956
Gurenko AA, Chaussidon M (2002) Oxygen isotope variations in primitive tholefites of Iceland: evidence
from a SIMS study of glass inclusions, olivine phenocrysts and pillow rim glasses. Earth Planet Sci Lett
205:63-79
Gurenko AA, Sobolev AV (2006) Crust-primitive magma interaction beneath neovolcanic rift zone of Iceland
recorded in gabbro xenoliths from Midfell, SW Iceland. Contrib Mineral Petrol 151:495-520
Hanchar JM, Hoskin PWO (eds) (2003) zircon. Reviews in Mineralogy and Geochemistry Vol. 53, Mineralogical
Society of America, Washington, D.C.
Harmon RS, Hoefs J (1995) Oxygen-isotope heterogeneity of the mantle deduced from global O-18 systematics
of basalts from different geotectonic settings. Contrib Mineral Petrol 120:95-114
Hildreth W, Christiansen RL, O’Neil JR (1984) Catastrophic isotopic modification of rhyolitic magma at times
of caldera subsidence, Yellowstone Plateau Volcanic Field. J Geophys Res 89:8339-8369
Hildreth W, Halliday AN, Christiansen RL (1991) Isotopic and chemical evidence concerning the genesis and
contamination of basaltic and rhyolitic magmas beneath the Yellowstone Plateau Volcanic Field. J Petrol
32:63-138
Hinton RW (1995) Ion microprobe analysis in geology. In: Microprobe Techniques in the Earth Sciences. PJ
Potts, JFW. Bowles, SJB. Reed and MR Cave (eds). London, Chapman and Hall: 235-289
Hoefs J (2005) Stable Isotope Geochemistry, Springer, 5th edition
476 Bindeman
Ireland TR (1995) Ion microprobe mass spectrometry: techniques and applications in cosmochemistry,
geochemistry, and geochronology. In: Advances in Analytical Geochemistry. Hyman M, Rowe MW
(eds). London, JAI Press Inc. 2:1-118
Johnson CM, Beard BL, Albarede F (eds) (2004) Geochemistry of Non-Traditional Stable Isotopes. Rev
Mineral Geochem, volume 55. Mineralogical Society of America
Kalamarides RI (1984) Kiglapait geochemistry VI: Oxygen isotopes. Geochim Cosmochim Acta 48:1827-
1836
King EM, Valley JW (2001) The source, magmatic contamination, and alteration of the Idaho batholith. Contrib
Mineral Petrol 142:72-88
Kita NT, Ikeda Y, Togashi S, Liu Yz, Morishita Y, Weisberg MK (2004) Origin of ureilites inferred from a
SIMS oxygen isotopic and trace element study of clasts in the Dar al Gani 319 polymict ureilite. Geochim
Cosmochim Acta 68: 213-4235
Krylov DP, zagnitko VN, Hoernes S, Lugovaja IP, Hoffbauer R (2002) Oxygen isotope fractionations between
zircon and water: Experimental determination and comparison with quartz-zircon calibrations. Eur J
Mineral 14:849-853
Kyser TK, O’Neil JR, Carmichael ISE (1982) Genetic relations among basic lavas and ultramafic nodules:
Evidence from oxygen isotope compositions. Contrib Mineral Petrol 81:88-102
Lackey JS, Valley JW, Hinke HJ (2006) Deciphering the source and contamination history of peraluminous
magmas using delta 18-O of accessory minerals: examples from garnet-bearing plutons of the Sierra
Nevada batholith. Contrib Mineral Petrol 151:20-44
Lackey JS, Valley JW, Saleeby JB (2005) Supracrustal input to magmas in the deep crust of Sierra Nevada
batholith: Evidence from high-delta O-18 zircon. Earth Planet Sci Lett 235:315-330
Lanphere MA, Champion DE, Christiansen RL, Izett GA, Obradovich JD (2002) Revised ages for tuffs of the
Yellowstone Plateau volcanic field: Assignment of the Huckleberry Ridge Tuff to a new geomagnetic
polarity event. Geol Soc Am Bull 114:559-568
MacPherson GJ, Mittlefehldt DW, Jones JH, Simon SB, Papike JJ, Mackwell S (eds) (2008) Oxygen in the
Solar System. Rev Mineral Geochem, volume 68. Mineralogical Society of America
Marsh BD (1998) On the interpretation of crystal size distributions in magmatic systems. J Petrol 39:553-599
Matthews A, Palin JM, Epstein S, Stolper EM (1994) Experimental study of 18O/16O partitioning between
crystalline albite, albitic glass and CO2 gas. Geochim Cosmochim Acta 58:5255-5266
Matthews A, Stolper EM, Eiler JM, Epstein S (1998) Oxygen isotope fractionation among melts, minerals and
rocks. In: Proceedings of the 1998 Goldsmith Conference, Toulouse, pp. 971–972. London Minerological
Society.
McBirney AR (2006) Igneous Petrology, 3rd Edition, Johns and Bartlett
Miller JS, Wooden JL (2004) Residence resorption and recycling of zircons in Devils Kitchen rhyolite Coso
Volcanic Field, California. J Petrol 45:2155-2170
Mills JG Jr., Saltoun BW, Vogel TA (1997) Magma batches in the Timber Mountain magmatic system SW
Nevada volcanic field Nevada, USA. J Volcanol Geotherm Res 78:185-208
Morgan LA, McIntosh WC (2005) Timing and development of the Heise volcanic field, Snake River Plain,
Idaho, western USA. Geol Soc Am Bull 117:288-306
Morishita Y, Giletti BJ, Farver JR (1996) Volume self-diffusion of oxygen in titanite: Geochem J 30:71-79
Muehlenbachs K, Anderson AT, Sigvaldasson GE (1974) Low-δ18O basalts from Iceland. Geochim Cosmochim
Acta 38:577-588
Muehlenbachs K (1998) The oxygen isotopic composition of the oceans, sediments and the seafloor. Chem
Geol 145:263-273
Obradovich JD (1992) K-Ar ages of Yellowstone Volcanic Field volcanic rocks. US Geol Surv Open File Rep
92408
Page Fz, Ushikubo T, Kita NY, Riciputi LR, Valley JW (2007) High precision oxygen isotope analysis of
picogram samples reveals 2-µm gradients and slow diffusion in zircon. Am Mineral 92:1772-1775
Palin JM, Stolper EM, Epstein S (1996) Oxygen isotope partitioning between rhyolitic glass/melt and CO2: an
experimental study at 550–950 degrees C and 1 bar. Geochim Cosmochim Acta 60:1963-1973
Portnyagin M, Bindeman IN, Hoernle K, Hauff F (2007) Geochemistry of primitive lavas of the Central
Kamchatka Depression. In: Magma Generation at the Edge of the Pacific Plate. AGU Monograph series
172 “Volcanism and Subduction: the Kamchatka Region” p. 199-239
Ramos FC, Tepley FJ III (2008) Inter- and intracrystalline isotopic disequilibria: techniques and applications.
Rev Mineral Geochem 69:403-443
Reid MR (2003) Timescales of magma transfer and storage in the crust. In: The Crust, Treatise on Geochemistry,
Vol. 3. Rudnick RL (ed) Elsevier, Oxford, UK. p 167-193
Reid MR Coath CD (2000) In situ U-Pb ages of zircons from the Bishop Tuff: No evidence for long crystal
residence times. Geology 28:443-446
Single Crystal Analysis of Oxygen Isotopes in Magmas 477
Riciputi LR, Paterson BA, Ripperdan RL (1998) Measurement of light stable isotope ratios by SIMS: Matrix
effects for oxygen, carbon, and sulfur isotopes in minerals. Int J Mass Spectrom 178:81-112
Rumble D, Farquhar J, Young ED, Christensen CP (1997) In situ oxygen isotope analysis with an excimer laser
using F2 and BrF5 reagents and O-2 gas as analyte. Geochim Cosmochim Acta: 61:4229-4234
Ryerson FJ, Durham WD, Cherniak DJ (1989) Oxygen diffusion in olivine- effect of oxygen fugacity and
implications for creep. J Geophys Res Solid Earth 94:4105-4118
Schmitt AK, Grove M, Harrison TM, Lovera O, Hulen JB, Walters M (2003a) The Geysers - Cobb Mountain
Magma System California (Part 1): U-Pb zircon ages of volcanic rocks conditions of zircon crystallization
and magma residence times. Geochim Cosmochim Acta 67:3423-3442
Schmitt AK, Grove M, Harrison TM, Lovera O, Hulen JB, Walters M (2003b) The Geysers-Cobb Mountain
Magma System California (Part 2): Timescales of pluton emplacement and implications for its thermal
history. Geochim Cosmochim Acta 67:3443-3458
Sharp z (2006) Principles of Stable isotope Geochemistry. Prentice Hall
Sharp zD (1990) A laser-based microanalytical method for the in situ determination of oxygen isotope ratios
of silicates and oxides. Geochim Cosmochim Acta 54:1353-1357
Sigmarsson O, Condomines M, Gronvold K, Thordarson T (1991) Extreme magma homogeneity in the 1783-
84 Lakagigar eruption: origin of a large volume of evolved basalt in Iceland. Geophys Res Lett 18:2229-
2232
Simakin AG, Bindeman IN (2008) Evolution of crystal sizes in the series of dissolution and precipitation events
in open magma systems. J Volcanol Geotherm Res doi: 10.1016/j.jvolgeores.2008.07.012
Simon JI, Reid MR (2005) The pace of rhyolite differentiation and storage in an ‘archetypical’ silicic magma
system Long Valley. Earth Planet Sci Lett 235:123-140
Spera FJ, Bohrson WA (2004) Open-system magma chamber evolution: an energy-constrained geochemical
model incorporating the effects of concurrent eruption, recharge, variable assimilation and fractional
crystallization. J Petrol 45:2459-2480
Streck MJ (2008) Mineral textures and zoning as evidence for open system processes. Rev Mineral Geochem
69:595-622
Taylor HP (1968) The oxygen isotope geochemistry of igneous rocks. Contrib Mineral Petrol 19:1-71
Taylor HP Jr. (1980) The effects of assimilation of country rocks by magmas on O18-O16 and Sr87/Sr86
systematics in igneous rocks. Earth Planet Sci Lett 47:243-254
Taylor HP Jr. (1986) Igneous rocks: II. Isotopic case studies of circumpacific magmatism. Rev Mineral 16:273-
316
Taylor HP Jr., Sheppard SMF (1986) Igneous rocks: I. Processes of isotopic fractionation and isotopic
systematics. Rev Mineral 16:227-272
Tefend KS, Vogel TA, Flood TP, Ehrlich R (2007) identifying relationships among silicic magma batches
by polytopic vector analysis: A study of the Topopah Spring and Pah Canyon Ash-flow Sheets of the
Southwest Nevada Volcanic Field. J Volcanol Geotherm Res 167:198-211
Tepley FJ, Davidson JP (2003) Mineral-scale Sr-isotope constraints on magma evolution and chamber dynamics
in the Rum layered intrusion, Scotland. Contrib Mineral Petrol 145:628-641
Tepley FJ, Davidson JP, Clynne MA (1999) Magmatic interactions as recorded in plagioclase phenocrysts of
Chaos Crags, Lassen Volcanic Center, California. J Petrol 40:787-806
Thordarson T, Self S (2003) Atmospheric and environmental effects of the 1783-1784 Laki eruption: A review
and reassessment. J Geophys Res.-Atm D1 Article Number: 4011
Trail D, Bindeman IN, Watson EB (2008) Experimental determination of quartz-zircon oxygen isotope
fractionation. (in prep)
Urey HC (1947) The thermodynamic properties of isotopic substances. J Chem Soc (Lond) 1947:562-581
Valley JW (2001) Stable isotope thermometry at high temperatures. Rev Mineral Geochem 43:365-413
Valley JW (2003) Oxygen isotopes in zircon. Rev Mineral Geochem 53:343-385
Valley JW, Bindeman IN, Peck WH (2003) Calibration of zircon-quartz oxygen isotope fractionation. Geochim
Cosmochim Acta 67:3257-3266
Valley JW, Cole D (eds) (2001) Stable Isotope Chemistry. Rev Geochem Mineral, volume 43. Mineralogical
Society of America
Valley JW, Kinny PD, Schulze DJ, Spicuzza MJ (1998) zircon megacrysts from kimberlite: oxygen isotope
variability among mantle melts. Contrib Mineral Petrol 133:1-11
Valley JW, Kitchen N, Kohn MJ, Niendorf CR, Spicuzza MJ (1995) UWG-2, a garnet standard for oxygen
isotope ratio: strategies for high precision and accuracy with laser heating. Geochim Cosmochim Acta.
59:5223-5231
Valley JW, Lackey JS, Cavosie AJ, Clechenko C, Spicuzza MJ, Basei MAS, Bindeman IN, Ferreira VP, Sial
AN, King EM, Peck WH, Sinha AK, Wei CS (2005) 4.4 billion years of crustal maturation: oxygen
isotope ratios of magmatic zircon. Contrib Mineral Petrol 150:561-580
478 Bindeman
Valley JW, Taylor HP Jr., O’Neil JR (eds) (1986) Stable Isotopes in High Temperature Geological Processes.
Rev Geochem Mineral, volume 16. Mineralogical Society of America
Vazquez JA, Reid MR (2002). Time scales of magma storage and differentiation of voluminous high-silica
rhyolites at Yellowstone caldera, Wyoming. Contrib Mineral Petrol 144:274-285
Wallace PJ, Anderson AT, Davis AM (1999) Gradients in H2O, CO2, and exsolved gas in a large-volume silicic
magma system: Interpreting the record preserved in melt inclusions from the Bishop Tuff. J Geophys Res
104:20097-20122
Wang zG, Kitchen NE, Eiler JM (2003) Oxygen isotope geochemistry of the second HSDP core. Geochim
Geophys Geosystems 4 Article Number: 8712
Wark DA, Spear FS, Cherniak DJ, Watson EB (2007) Pre-eruption recharge of the Bishop magma system.
Geology 35:235-238
Wark DA, Watson EB (2006) Titaniq: A titanium-in-quartz geothermometer. Contrib Mineral Petrol 152:743-
754
Watson EB (1996) Dissolution, growth and survival of zircons during crustal fusion: Kinetic principles,
geological models and implications for isotopic inheritance. Trans R Soc Edinburgh: Earth Sci 87:43-56
Watson EB, Cherniak DJ (1997) Oxygen diffusion in zircon. Earth Planet Sci Lett 148:527-544
Williams DA, Kadel SD, Greeley R, Lesher CM, Clynne MA (2004) Erosion by flowing lava: geochemical
evidence in the Cave Basalt, Mount St. Helens, Washington. Bull Volcanol 66:168-181
Williams DA, Kerr RC, Lesher CM, Barnes SJ (2001) Analytical/numerical modeling of komatiite lava
emplacement and thermal erosion at Perseverance, Western Australia. J Volcanol Geotherm Res 110:27-
55.
Wolff JA, Ramos FC (2003) Pb isotope variations among Bandelier Tuff feldspars: no evidence for a long-lived
silicic magma chamber. Geology 31:533-536
Young ED, Fogel ML, Rumble D, Hoering TC (1998) Isotope-ratio-monitoring of O2 for microanalysis of O18/
O16 and O17/O16 in geological materials. Geochim Cosmochim Acta 62:3087-3094
Young ED, Kuramoto K, Marcus RA, Yurimoto H, Jacobsen SB (2008) Mass-independent oxygen isotope
variation in the solar nebula. Oxygen in the solar system: Rev Mineral Geochem 68:187-218
zhang Yx, Walker D, Lesher CE (1989) Diffusive crystal dissolution. Contrib Mineral Petrol 102:492-513
zhao z, zheng Y-F (2003) Calculation of oxygen isotope fractionation in magmatic rocks. Chem Geol 193:59-
80
zheng YF (1993a) Calculation of oxygen isotope fractionation in anhydrous silicate minerals. Geochim
Cosmochim Acta 57:1079-1091
zheng YF (1993b) Calculation of oxygen-isotope fractionation in hydroxyl-bearing silicates. Earth Planet Sci
Lett 120:247-263
Reviews in Mineralogy & Geochemistry
Vol. 69 pp. 479-544, 2008 13
Copyright © Mineralogical Society of America
uranium-series Crystal ages
Kari M. Cooper
Department of Geology
University of California
Davis, California, 95616, U.S.A.
[email protected]
Mary r. reid
Department of Geology
Northern Arizona University
Flagstaff, Arizona, 86011, U.S.A.
[email protected]
IntrOduCtIOn
As the collected papers in this volume aptly demonstrate, chemical and textural records
contained within crystals have become increasingly useful tools for understanding the evolution
of magmatic systems. An essential component of studies of the dynamics of magmatic systems is
the ability to place the thermal, physical, and compositional evolution of magmas in a temporal
context. Studies of the kinetics of crystal growth and reaction, and of diffusive re-equilibration
of crystals and melts, can provide information about the duration of magmatic processes
such as magma ascent (Rutherford 2008), crystal growth (Hammer 2008), and residence of
crystals in melts other than the ones in which they crystallized (Costa 2008). However, the
only method of extracting information from young crystals about the absolute age of magmatic
processes is isotopic dating. The timescales of many sub-volcanic magmatic processes such
as magma transport, differentiation, crystallization, and storage within the crust, appear to
be commensurate with the half-lives of uranium-series (U-series) nuclides, and therefore
this isotopic system offers the greatest potential to quantify timescales and rates of magmatic
processes. We focus here on crystal dating using the decay products of 238U and 235U, including
238
U-230Th-226Ra-210Pb disequilibria, 235U-231Pa disequilibria, and U-Pb dating of young crystals.
In this chapter, we review the principles of U-series dating of crystals, the types of information
that have been gained so far from U-series dating, and the potential for combining U-series
crystal ages with other crystal-scale and magma-scale information to unravel the dynamics of
magmatic systems.
Principles of u-series dating
U-Pb dating is based on the accumulation of 206Pb and 207Pb over time due to decay of 238U
235
and U, respectively, and is conceptually similar to most isotopic dating techniques. U-Pb
dating of zircon, both by TIMS and in situ techniques, has been reviewed recently (Parrish and
Noble 2003; Ireland and Williams 2003; Kosler and Sylvester 2003), and we refer readers to
these papers for fundamentals of the techniques. Here, we are concerned primarily with U-Pb
dating of young zircon (<1 Ma), and the complications inherent in dating young crystals and
techniques to address these factors are detailed in the second section. U-series disequilibrium
dating is less widely-applied and often is less intuitive; therefore, we briefly review here the
principles relevant to U-series dating of crystals in magmatic systems. For a comprehensive
1529-6466/08/0069-0013$10.00 DOI: 10.2138/rmg.2008.69.13
480 Cooper & Reid
view of U-series geochemistry, we refer the reader to two earlier reviews of the principles of
U-series geochemistry and its application to geological problems (Ivanovich and Harmon 1992;
Bourdon et al. 2003a).
238
U decays to 206Pb through a series of intermediate daughter nuclides, including 234U,
230
Th, 226Ra, and 210Pb; similarly, 235U decays to 207Pb through a series of intermediate nuclides,
including 231Pa (Fig. 1). Each of the intermediate daughter nuclides is both radiogenic and
radioactive, and half-lives of the intermediate daughters range from seconds to 245 k.y. (Fig.
1). The currently-accepted values for half-lives of the daughters discussed in this chapter are
compiled in Table 1. Left undisturbed for a sufficient period of time, any U-bearing system will
attain a state known as secular or radioactive equilibrium, where the number of decays per unit
time (i.e., the activity) of each of the intermediate daughters is equal to that of the parent isotope
of uranium. Secular equilibrium is in fact a dynamic steady state, where the declining activity
of each daughter nuclide tracks the slowly decreasing activity of 238U or 235U. Once attained,
a system will remain in secular equilibrium until some process fractionates one or more of the
daughter nuclides from their parents. Because the intermediate daughters have a wide range of
geochemical behaviors in natural systems, geological events will, more often than not, produce
disequilibrium in at least some parent-daughter pairs in the decay chain. In the case of magmatic
systems, processes such as partial melting, assimilation of wall rocks, crystallization of minerals,
and degassing have all been shown to produce disequilibria in the various phases present. After
disequilibria have been produced, the system (and each sub-system; e.g., individual phases) will
238 234 235
U U U
4.47x109y 2.45x105y 7.04x108y
234 231
Pa Pa
6.69 h 3.28x104y
234 230 231 227
Th Th Th Th
24.1 d 7.6x104y 1.06 d 18.7 d
227
Ac
21.8 y
226 223
Ra β Ra
1599 y 11.4 d
222
α 219
Rn Rn
3.823 d 3.96 s
218 214 210 215
Po Po
-4
Po Po
3.04 min 1.6x10 s 138.4 d 1.8x10-3 s
214 210 211
Bi Bi Bi
19.7 min 5.01 d 2.14 min
214 210 206 211 207
Pb Pb Pb Pb Pb
26.9 min 22.6 y stable 36.1 min stable
207
Tl
4.77 min
Figure 1. Schematic decay chains for 238U and 235U. The isotopes used in the U-series dating discussed in
this chapter are shown as black boxes. The half-lives shown are those compiled in Bourdon et al. (2003).
Uranium-series Crystal Ages 481
table 1. Half-lives of U-series nuclides discussed in this chapter.
Parent-daughter half-life of approximate
pair daughter useful timescale
(yr) (yr)
238
U-230Th 75,690 ± 230 1000-400,000
230 226
Th- Ra 1599 ± 4 100-10,000
226
Ra-210Pb 22.6 ± 0.1 1-100
235 231
U- Pa 32,760 ± 220 500-175,000
Half-lives are those compiled in Bourdon et al. (2003b). Note that useful timescale
for each parent-daughter pair is only a broad estimate and the accessible timeframe
can vary depending on the initial disequilibria and parent/daughter ratio.
return to secular equilibrium through decay of excess daughter or ingrowth of daughter that is
in deficit relative to the activity of the parent. In all of the parent-daughter pairs discussed here,
the half-life of the daughter is significantly shorter than that of the parent; in this case, secular
equilibrium will be attained at a rate governed by the half-life of the daughter product. If the
initial abundance of the daughter after a disturbance is known or can be estimated, the degree
of disequilibrium measured at present can be used to calculate the time since disequilibrium
was generated. The timescale over which disequilibrium can be detected—and therefore the
useful time range for each parent-daughter pair for dating—depends on the initial degree of
disequilibrium and the accuracy of the measurement of the abundances of parent and daughter.
A useful rule of thumb is that disequilibria can be detected for ~5 times the half-life of the
daughter (Table 1); however, with the increased precision of modern mass spectrometric
analyses, and when analyzing phases such as zircon which are expected to have very large
parent/daughter ratios and therefore extreme disequilibria at the time of crystallization, dating
in favorable cases can be extended to significantly longer timescales. A lower limit to the range
of time accessible with each parent-daughter pair is given by the time required for daughter
abundances to deviate measurably from the initial conditions, and therefore is also a function of
the analytical uncertainty and the confidence with which the initial daughter abundance can be
estimated. Approximate time ranges accessible with each parent-daughter pair discussed here
are compiled in Table 1.
The mathematics governing parent-daughter abundances in a decay chain are presented in
some detail in Ivanovich and Harmon (1992), and we will not reproduce the full derivation of
the equations here. For decay of a parent to daughter where the daughter is also radioactive, the
general equation governing the abundance of the daughter product is given by
λ1 0 − λ1t
N2 = (
−λ t 0 −λ t
N1 e − e 2 + N 2 e 2 ) (1)
λ 2 − λ1
where N2 is the number of atoms of the daughter present at time t, N20 is the number of atoms of
the daughter present initially (t0), N10 is the number of atoms of the parent present initially, and
λ1 and λ2 are the decay constants of the parent and daughter, respectively. Assuming that the time
of interest is short with respect to the half-life of the parent and that the half-life of the parent
is significantly longer than the half-life of the daughter, and making use of the relation that the
decay rate (activity) for a given nuclide is equal to Nλ, Equation (1) can be simplified to give
( )
( N 2 ) = ( N1 )0 1 − e − λ2 t + ( N 2 ) e − λ2 t
0
(2)
482 Cooper & Reid
where parentheses around the chemical symbols, by convention, denote activities (the number
of decay events per unit time).
u-series dating of crystals
U-series measurements can thus be used to date crystals in igneous systems, and can provide
information about absolute ages of events or processes that occurred within the past few years
to decades (e.g., 226Ra-210Pb disequilibria) to thousands of years to hundreds of thousands of
years (238U-230Th disequilibria). U-series dating of crystals can be applied through bulk mineral
separates for major phases, which include material from hundreds to thousands of grains, or
through in situ analyses of sub-grain domains in accessory minerals with high concentrations
of U or Th. We present information about the details of requirements for sample size, other
analytical considerations, and techniques for calculating ages in the second section (“Analytical
and practical considerations”). Once an analysis has been obtained, the interpretation of a
U-series date for a crystal or bulk mineral aggregate depends on the geologic context, including
the extent to which the analyzed grains represent a simple crystallization history or more
complex histories involving protracted crystal growth, recycling of material from earlier
phases of magmatic activity and/or incorporation of crustal material during magma storage.
U-series crystal ages have been used thus far to examine the timescales and dynamics of a wide
variety of magmatic processes, including magma storage within the crust, crystallization and
differentiation, recycling of cogenetic crystal material from young plutonic bodies or crystal
mush zones, thermal and chemical evolution of magmas during storage, and dating eruptions,
among other applications. We discuss these various interpretations of U-series crystal ages,
and the information that can be gained in different scenarios of varying complexity, in the
following sections. As is increasingly obvious from studies such as those reviewed in the third
section (“What do crystal ages tell us?”), coupling U-series crystal ages with other crystal-scale
information has the potential to provide unique insights into the evolution of magmatic systems,
and in the fourth section (“Resolving ages from complex crystal populations: Incorporating
other observations”) we discuss various combinations of types of information and how they
may provide insights well beyond what is available from crystal ages or crystal-scale chemical
information in isolation. In the final section (“Key questions and future directions”), we discuss
the potential to capitalize on and expand on these types of combined studies, and propose some
areas that are likely to be fruitful avenues for further research.
anaLytICaL and PraCtICaL COnSIderatIOnS
Introduction
The ability to measure U-series ages of minerals depends ultimately on the number of
atoms of each nuclide present in the analyzed volume relative to the sensitivity of the analytical
technique, which in turn reflects a combination of the concentration of U-series elements within
the mineral of interest, the abundance of that mineral in a rock (and therefore the ability to
purify a mineral separate of sufficient size for analysis), and the analytical technique used.
Measurements of many of the U-series nuclides in whole-rock samples are now routine,
but our ability to measure crystal ages is still largely limited by analytical considerations.
The analytical challenges of U-series mineral analyses largely result from two factors. First,
the U-series elements are incompatible in most major minerals in igneous rocks, requiring
significantly larger sample sizes for analysis of minerals relative to whole-rocks or glass.
Second, the abundance of each nuclide at radioactive equilibrium is inversely proportional to its
half-life, which means that the nuclides with shorter half-lives (e.g., 226Ra or 210Pb) are orders
of magnitude less abundant in magmas than are 238U or 230Th. The result in practice is that
while in situ measurements of U-series nuclides are now possible by secondary ionization mass
spectrometry (SIMS) and potentially by laser ablation- inductively coupled mass spectrometry
Uranium-series Crystal Ages 483
(LA-ICP-MS; Stirling et al. 2000; Bernal et al. 2005), they are limited to trace phases with high
concentrations of U and/or Th. Measurement of 238U-230Th-226Ra-210Pb disequilibria in major
phases requires preparation of a bulk mineral separate and measurement by thermal ionization
mass spectrometry (TIMS), or multicollector inductively coupled mass spectrometry (MC-ICP-
MS; also known as plasma ionization multicollector mass spectrometry or PIMMS), or, in the
case of 210Pb, alpha spectrometry.
This section covers the practical side of U-series crystal dating; here, we review the
requirements for successful analysis in terms of sample size and preparation for both bulk
separate and in situ analyses. We then review the methods (and time investment required) for
analysis by TIMS, MC-ICP-MS, and SIMS, followed by a discussion of methods for calculating
ages from the different types of data, along with some of the complications that can arise and
methods for addressing them.
Practical limitations: sample preparation and analysis
Sample size requirements. Figure 2 illustrates concentrations of U, Th, and Ra, and
210
Pb activities expected for different minerals precipitated from magmas with a range of U
concentrations. Many variables influence the exact abundances of parent and daughter nuclides
present in a mineral and therefore the amount required for analysis (e.g., age of the mineral
separate, the degree of U-Th or Th-Ra disequilibria in the magma, the mineral major-element
composition and temperature of crystallization, which in turn affect the partitioning behavior,
etc.), but this figure can be used to estimate in a broad sense how large a mineral sample
would be required for analysis of a given element. Partition coefficients used to construct
Figure 2 are given in Table 2. In situ measurements require the highest concentrations, and
are only possible for trace phases where U and/or Th are compatible (e.g., zircon, allanite).
Although concentrations of U and Th in apatite may be high enough to analyze 238U-230Th
disequilibria by SIMS, uranium and thorium will not be greatly fractionated during apatite
crystallization, which would lead to difficulties in correcting for initial 230Th present in the
crystals and therefore imprecise ages. 238U-230Th dating of bulk mineral separates requires ~10
ng of Th and U, and is feasible for most of the minerals shown.
Current in situ techniques are not sensitive enough to allow analysis of radium; even
analysis of 230Th by in situ techniques is only feasible for high-U phases such as zircon, where
230
Th abundances will be high. Radium abundances in igneous rocks are generally 6-7 orders
of magnitude lower than elemental abundances of Th and U. However, given the relatively high
ionization efficiency of Ra by TIMS (and similar although somewhat lower total efficiency by
MC-ICP-MS), it is possible to measure radium concentrations in samples containing as little
as only a few femtograms of Ra, and measurements of ~10 fg generally produce good results
(albeit with somewhat larger uncertainties than is typical for whole-rock samples, where a
total of tens to hundreds of femtograms of Ra are analyzed). The calculations of concentrations
necessary for 0.1 g-5 g mineral separates shown in Figure 2c assume that a total of 10 fg
of radium is analyzed. Minerals where radium is relatively less incompatible (plagioclase,
amphibole, and phlogopite) or even compatible (K-feldspar) are attractive targets for 230Th-
226
Ra dating and feldspar was one of the first minerals dated using Ra-Th (e.g., Reagan et al.
1992). Relatively few analyses of amphibole have been reported in the literature (e.g., Turner
et al. 2003a,b), and to our knowledge no analyses of mica, but these minerals have significant
potential in future dating applications. As is apparent from Figure 2c, concentrations of Ra
in olivine, clinopyroxene, orthopyroxene, and zircon are expected to be extremely low at the
time of crystallization, and Ra in pure separates of very young minerals could not reliably be
measured. However, because of the high Th/Ra ratios in these minerals, 226Ra ingrowth will
be rapid, and concentrations of Ra expected in equilibrium with Th in the same minerals are
several orders of magnitude higher (Fig. 2). Thus the age of the mineral being analyzed will
have a significant effect on the expected concentration and therefore on the amount required
Figure 2. Plots of U concentration vs.
484
daughter concentration in various minerals,
intended to be used as a way to broadly
estimate the volume of mineral separate 106
104 zrn
necessary for analysis. Panels show A) U A C
concentration, B) Th concentration (both in phl/kfs
ap 104 pl
terms of 230Th and total Th; 230Th calculated 102 zrn
amp
assuming 232Th/230Th = 210,000, derived from 102
assuming secular equilibrium in the melt and amp
a Th/U weight ratio in the melt of 3.5), C) 100 cpxx
226
op 100
Ra concentration, and D) 210Pb activity c px
in minerals at the time of crystallization. o px
10-2 10-2 c px o px
All calculations assume secular equilibrium ol
between 238U, 230Th, 226Ra, and 210Pb in the phl zrn
10-4 ol
[U] in mineral (ppm)
[226Ra] in mineral (fg/g)
melt from which minerals are crystallizing. 10-4
pl
Also shown are concentrations required for
ol 10-6
analysis of 0.1g (short-dash grey lines), 1g 10-6
(long-dash grey lines), and 5 g (solid grey 0.1 1 10 100 0.1 1 10 100
lines) mineral separates (calculated assuming [U] in melt (ppm) [U] in melt (ppm)
that 10 ng of U and Th, 10 fg of 226Ra, and 1 zr n 104 102
dpm total activity for 210Pb are required for B D pl
106 ap 101
analysis; in favorable circumstances analysis 102
Cooper & Reid
ampphl
of smaller quantities may be possible). No
cpx
attempt has been made to show the full range 104
ampx
cp 100
of concentrations encompassed by variations pl opx 100 opx
in partition coefficients as a function ol
10-1
of temperature, crystal composition, or 102
10-2
uncertainties in partition coefficients; instead,
10-2
an intermediate D value has been chosen
100 phl
[230Th] in mineral (fg/g)
[Th] in mineral (ppm)
for each mineral. Blundy and Wood (2003) 10-4
10-3
(210Pb) in mineral (dpm/g)
present a comprehensive review of partition
coefficients for U-series nuclides; we use 10-2 ol
their compilation as a basis and the specific 10-6 10-4
0.1 1 10 100 0.1 1 10 100
partition coefficients used in calculations [U] in melt (ppm)
are given in Table 2. Also shown in panel C
[U] in melt (ppm)
(dashed lines with labels) are concentrations
of Ra in olivine, opx, cpx, and zircon at
equilibrium with 230Th in the minerals.
Cooper&Reid Figure 2
Uranium-series Crystal Ages 485
table 2. Partition coefficients used to construct Figure 2.
Mineral du dth dBa dra dra/dBa dPb
plagioclase 6.0×10−4 0.003 0.42 0.079 0.19 0.61
K-feldspar, Or50 — — 13 5 0.38 —
amphibole 0.019 0.018 0.41 0.033 0.08 0.10
phlogopite 1×10−3 1×10−3 5 5 1 0.1
−5 −7
clinopyroxene 0.017 0.016 1.5×10 2.0×10 0.01 0.05
orthopyroxene 0.005 0.002 1.2×10−5 1.8×10−7 0.015 0.009
olivine 6.0×10−5 9.5×10−6 1.0×10−5 1.2×10−7 0.02 0.004
apatite 2.21 1.68 0.27 — — —
zircon 100 16.7 0.004 4.0×10−9 1.0×10−6 —
All partition coefficients derived from values or equations in Blundy and Wood (2003) unless otherwise noted; Ra
partition coefficients calculated using Ba as a proxy as described for the different minerals in Blundy and Wood
(2003). Dashes where no data are available. Calculations for Ba, Ra, and Pb in plagioclase assume An50 at 950
°C. U, Th, and Pb partitioning data for amphibole from Tiepolo et al. (2000). U, Th, and Pb in phlogopite from
LaTourrette et al. (1995). DBa in phlogopite from Blundy and Wood (2003) Figure 22, corresponding to ~900 °C.
DBa in clinopyroxene from calculations in Cooper et al. (2001). DBa in orthopyroxene estimated using experimental
data for Sr reported in Blundy and Wood (2003) and their estimate that DBa is 50-100 times lower. DBa for olivine
chosen to be within the range reported in the GERM database, yet not higher than DBa for pyroxene. DU, DTh, and
DBa for apatite are averages of the data from empirical studies of Mahood and Stimac (1990) and Luhr et al. (1984);
we did not use the Ds of Bea et al. (1994) in the calculations for U and Th because they are an order of magnitude
higher than the other two estimates.
for analysis. Several studies have successfully analyzed 230Th-226Ra disequilibria in pyroxene
(e.g., Volpe and Hammond 1991; Cooper et al. 2001; Turner et al. 2003a,b). When analyzing
a given mineral separate for 238U-230Th-226Ra disequilibria, the concentration of radium is
generally the limiting factor dictating sample size.
The very short half-life of 210Pb (22.6 yr) and incompatible behavior of Pb in most rock-
forming minerals combine to produce extremely low abundances of 210Pb in most rock-forming
minerals. Therefore, 210Pb in minerals remains one of the most challenging of the measurements
of U-series nuclides, and generally requires preparation of larger samples than is required for
other analyses. Analysis of 210Pb generally requires ~1-5 g of mineral separate, in addition to
that required for U-Th-Ra analyses. Nevertheless, with alpha counting by 210Pb proxy, 210Pb
activity can be analyzed to better than 1% accuracy (1 sigma; e.g., Reagan et al. 2006).
Sample preparation for SIMS. It is possible to analyze accessory phases directly in
rocks and thin sections (“in situ” sensu stricto) but it is more efficient to cluster a number of
grains in a single mount. Accessory phases like zircon can be concentrated from crushed rock
material using conventional heavy liquid and magnetic separation techniques. Alternatively,
an HF-resistant mineral like zircon can be isolated by selective dissolution of the host rocks.
If applied exclusively to glass shards, the latter method is a potential means of isolating only
zircons that are not included within phenocrysts (e.g., Bindeman et al. 2006). Grains can be
handpicked from the concentrate, mounted onto a surface (typically a piece of sticky tape)
along with standards, and fixed in epoxy. Most workers then polish the mount using Al2O3 or
diamond paste (0.25 to 1 micrometer diameter) so that grain interiors are exposed. Analyses
of unpolished zircon rims can be performed by placing the faces of zircon grains flush to the
mount surface (Reid and Coath 2000; Bindeman et al. 2006). This permits a spatial resolution
of near-rim heterogeneity that is an order of magnitude lower than conventional spot analyses.
486 Cooper & Reid
Depth profiling either by continuous analysis (Vazquez 2004; Reid 2008) or by incrementally
polishing samples to discrete depth intervals (Bindeman et al. 2006) can also be performed.
Once the grain mount is prepared, the grains are mapped in reflected light and/or by back-
scattered electron (BSE) imaging. It is advantageous to also characterize internal compositional
heterogeneities and/or included mineral/melts before analysis, using either cathodoluminesence
(CL; zircon) or BSE (allanite, chevkinite). Either before or after grain mapping, the mount is
ultrasonically cleaned—sometimes in HCl—to remove surface Pb contamination and a thin
coat of gold (10-100 nm) is then applied to minimize charging of the sample surface during
ion beam sputtering. Some studies (e.g., Bacon et al. 2000; Bacon and Lowenstern 2005;
Bindeman et al. 2006) focused their studies on U-rich zircons that had been pre-selected using
BSE, CL, or scanning ion imaging. Others focused on grains that were relatively small and
homogeneous (Sano et al. 2002). These approaches maximize the age resolution by minimizing
the uncertainties for 206Pb and 230Th and the chances that the ion beam averages multiple age
domains, but could favor a specific zircon population to the exclusion of others.
Preparation of mineral separates for bulk analyses and chemical separation. Preparation
of mineral separates of 0.1-5 g is generally sufficient for U-series dating of minerals, with
larger mineral separates required for 226Ra and 210Pb analysis and/or for mafic samples and
smaller mineral separates required for 238U-230Th analysis and/or silicic samples. In practice,
the amount of time required to produce the mineral separate will also depend on the purity
required for the mineral separate. Most isochron dating of minerals (including 238U-230Th
dating) can tolerate the inclusion of a significant fraction (~10%) of impurities in the mineral
separates, because the effect will be to move data points closer together along the same
isochron line while keeping the slope and therefore the calculated age the same. However,
as detailed below, accurate 230Th-226Ra and 226Ra-210Pb dating requires correction of the data
for the effects of impurities in the mineral separate, and therefore it is generally necessary to
have mineral separates that contain at most a few percent impurities. Therefore, in addition
to the requirement of larger sample sizes, 230Th-226Ra and 226Ra-210Pb dating require mineral
separates of higher purity, and are thus quite time-consuming.
Once mineral separates have been prepared, chemical separation of the elements of
interest requires a significant effort, again scaled by the technique and specific nuclides
analyzed. Chemical separation and mass spectrometric techniques for U-series nuclides have
recently been reviewed by Goldstein and Stirling (2003), and we refer readers to that paper
for analytical details. Analytical procedures vary from one lab to another; we summarize here
information from several labs which have produced data on bulk mineral separates (Turner
et al. 2003b; Rogers et al. 2004; Tepley et al. 2006; Cooper and Donnelly 2008). In brief,
238
U-230Th analyses alone of mineral separates generally require dissolution, sample spiking
and equilibration, followed by 1-4 ion-exchange column steps; all told, dissolution, chemistry
and preparation for mass spectrometry requires on the order of two to three weeks per batch
of 8-10 samples. Radium and 210Pb analyses require additional effort in chemical separation.
However, most labs have adopted techniques that allow separation of U, Th, and Ra from the
same sample, which increases efficiency in chemical processing (e.g., Goldstein and Stirling
2003; Turner et al. 2003a,b; Rogers et al. 2004; Tepley et al. 2006; Cooper and Donnelly
2008). Radium separation for mineral separates requires an additional 4-5 ion-exchange
column steps beyond what is required for U-Th preparation, and given the increased sample
size requirements (which lead to large columns for the initial steps), the total process for
sample dissolution, spiking and equilibration, and column chemistry, for U-Th-Ra analyses
requires approximately 4-6 weeks for a batch of samples.
To date, no techniques have been developed to allow separation of 210Pb from the same
aliquot used for U-Th-Ra measurements, so analysis of 226Ra-210Pb disequilibria requires a
mineral separate that is approximately twice as large as analysis of 238U-230Th-226Ra alone.
Uranium-series Crystal Ages 487
Column chemistry for Pb requires one ion-exchange column and preparation for analysis by
alpha spectrometry (autoplating) requires approximately 6-12 hours per sample, which totals
approximately 1-2 weeks for sample dissolution, chemistry, autoplating, and counting per
batch of samples.
Mass spectrometry: in situ analyses. Many accessory phases have sufficiently high U
and/or Th contents so that domains within individual minerals can be dated by in situ methods,
especially those of secondary ion mass spectrometry. Ages may be based on the magnitude of
238
U-230Th disequilibrium or on the amount of in-grown 206Pb, the ultimate daughter of 238U.
Recently, protocols for 235U-231Pa dating of zircon have been developed (Schmitt 2007). In
situ dating of young accessory phases has mainly focused on zircon and has been performed
at UCLA (e.g., Reid et al. 1997; Dalrymple et al. 1999), University of Western Australia (e.g.,
Brown and Fletcher 1999), Stanford University (e.g., Bacon et al. 2000; Lowenstern et al.
2000), and Hiroshima University (Sano et al. 2002). Relevant analytical conditions for U-Pb
and Th-U SIMS dating of zircon in the various laboratories are described by Reid et al. (1997),
Brown and Fletcher (1999), Sano et al. (2002), Schmitt et al. (2003a), Bacon and Lowenstern
(2005), and Schmitt (2006). Dating of magmatic allanite and chevkinite is described by
Vazquez and Reid (2004) and Vazquez (2008).
In the ion microprobe, an accelerated (primary) beam of ions is focused onto a sample
surface in order to eject atomic and molecular species from the target minerals (Fig. 3). The
resulting secondary ions are then accelerated into the mass spectrometer. For U-Pb analyses,
flooding of the sample surface with O2 can enhance Pb ion yields under favorable circumstances
Figure 3. Schematic illustration of the approach of secondary ionization mass spectrometry. A primary
ion beam (black circles) bombards the sample, generating secondary species from a sample’s surface (gray
circles). The secondary ions are extracted into the mass spectrometer where the relative mass abundances
are measured.
488 Cooper & Reid
by factors of ~2 (zircon; Ireland and Williams 2003) to 10 (baddelyite; Chamberlain et al. 2007)
in certain instruments (Cameca ims 1270/1280). Areas sputtered from the sample surfaces
typically range from 25-45 micrometers in diameter (Fig. 4), although results for beam spots
as small as 10 micrometers have been reported (e.g., Brown and Fletcher 1999). The depth
resolution of SIMS spots is, as noted above, more than an order of magnitude smaller than
the lateral dimensions of the spot and this effect can be exploited to advantage (e.g., Reid and
Coath 2000).
Before data acquisition can begin, sample surfaces are typically sputtered or rastered for
2-4 minutes to remove the gold coat and possible surface contamination in the analysis area.
This means that the outermost few tenths of a micrometer of a mineral rim cannot be analyzed.
Even with this precautionary measure, the more slowly sputtered margins of the analysis pit
may contribute common Pb as the analysis proceeds. One solution to this is to reduce an
aperture diameter in the secondary ion path so that ions only from the central portion of the
analysis pit are emitted into the mass spectrometer. However, the reduction in the common Pb
correction may not compensate for the resulting loss in precision arising from the attendant
reduction in overall secondary beam intensity.
For U-Pb dating, secondary ion beam intensities are measured for zr2O4 (90zr2O4,
92
zr zrO4, or 94zr2O4), 206Pb, 207Pb, 238U, 232ThO, and 238UO (with “O” referring to 16O),
94
along with a background analysis near mass 204Pb. Collection of 204Pb, 208Pb, and 232Th beam
intensities may also be included. For U-Th dating, a zr2O4 peak, one or more backgrounds
(e.g., at mass/charge = 246.3), and 230ThO, 232ThO, and 238UO are measured. Tuning the
spectrometer to a mass resolution of >4000 (U-Th) to >4800 (U-Pb) ensures that molecular
interferences (HfO2 and REE oxides) on the masses of interest are well-resolved when dating
of zircons is being performed. Under these conditions, 206Pb count rates in young zircons range
from 0.5 to 16 counts per second (cps); 230ThO count rates range from 0.4-5 cps. For allanite,
a mass resolution of ~9000 is required for U-Th analyses but even at this resolution 230ThO
intensities in allanite can still be as high as several cps (Vazquez and Reid 2004). Counts
Figure 4. Electron back-scattered image of a zircon from the Torihama dacite, Japan, after ion microprobe
analysis. Note the well-defined and flat-bottomed crater created by ion sputtering and the presence of glass
and apatite inclusions and glass selvages on the grain. [Used with permission of Elsevier from Sano et al.
(2002) Journal of Volcanology and Geothermal Research, Vol. 117, p. 288.]
Uranium-series Crystal Ages 489
for key peaks are typically integrated over several data cycles totaling hundreds of seconds.
Background noise count rates are less than 0.05 cps (except where unusually high primary
beam currents have been applied), negligible for most species of interest except for 230ThO.
U, Th, Pb and their oxides have different ionization yields during sputtering and so a
correction factor must be applied to measured ratios to obtain true ratios of 20xPb/23xU and
230
Th/238U. The relative sensitivities of Pb, Th, and U in zircon have been determined using
various standards, as extensively discussed by Ireland and Williams (2003). For U-Th
dating, ThO/UO relative sensitivity factors can be obtained from the difference between the
230
ThO/238U ratio measured in “ancient” (t > 400 k.y.) zircons and that expected at secular
equilibrium (e.g., Reid 1997; Charlier et al. 2005). Recent studies (Bacon and Lowenstern
2005; Charlier et al. 2005) have propagated the uncertainty in this fractionation factor into the
reported age uncertainty. A somewhat more convoluted but more precise approach is to obtain
relative sensitivity factors by comparing the 230ThO/238U ratio measured to time-integrated
Th/U of zircon standards based on 208Pb/206Pb (Reid et al. 1997).
Uranium contents are typically estimated by dividing the measured ratios of a uranium
species to a zirconium oxide species (e.g., 238UO/94zr2O4) on the unknown by the same
ratio obtained on the reference standard and multiplying that quotient by the standard’s U
concentration. Standards sufficiently homogeneous in U abundance include SL13 (238 ppm
U), Cz3 (550 ppm U), and 91500 (81.2 ppm U). An analogous approach may be applied to
determine Th concentrations or the Th concentrations can be determined directly from the U
concentration by multiplying the latter by the measured Th/U.
Mass spectrometry: bulk mineral separates. Techniques for analysis of U-series nuclides
by alpha spectrometry, TIMS, MC-ICP-MS, or single-collector, high-resolution ICP-MS, have
been reviewed elsewhere (Goldstein and Stirling 2003), and we focus this discussion on the
practical considerations of time and effort required to perform these analyses using different
techniques. The time required for analysis of bulk mineral separates after chemical separation
depends greatly on the technique used, and also can vary significantly from one sample to the
next. TIMS analyses require approximately two hours each for U and Th concentration mea-
surements by isotope dilution with an additional 4-8 hours for measurement of Th isotopic com-
positions (if run separately from the Th ID measurement) and an additional 2-8 hours for radium
concentration measurement. MC-ICP-MS offers a significant advantage in sample throughput,
with for example a single measurement of thorium isotopic composition in a sample requiring
as little as 15-20 minutes. However, this is a somewhat misleading comparison, as instrument
calibration, tuning, and analysis of standards (especially in the case of sample-standard bracket-
ing) all require significant amounts of time. Perhaps a more appropriate comparison is that, for
example, one can expect to complete analysis of Th isotopic compositions of 1-2 samples per
day of instrument time by TIMS, and 8-10 samples per day by MC-ICP-MS.
Analysis of radium in mineral separates can be done by TIMS (see refs in Goldstein and
Stirling 2003), by MC-ICP-MS (Pietruszka et al. 2002; Ball et al. 2008; Cooper and Donnelly
2008; Sims et al. 2008) or by single-collector, high-resolution ICP-MS (Sims et al. 2003;
Cooper et al. 2005). Radium measurements by TIMS require ~4-8 hours of instrument time per
sample, whereas radium measurements by MC-ICP-MS require ~1 hour per sample (when time
for washout, calibration, and standards is included). The lower ionization efficiency of radium
by TIMS compared to ICP-MS is counterbalanced by the lower transmission efficiency of ICP-
MS, leading to a somewhat smaller effective ion yield by ICP-MS (0.5%, compared to 1-10% by
TIMS; Goldstein and Stirling 2003). Because of the tradeoff between time required for analysis
and effective ion yield, it is our experience that with small samples (10’s of femtograms of Ra)
more precise results are obtained by TIMS than by MC-ICP-MS, whereas with larger samples
(~100 fg), results with good precision (~1% relative errors, 2-sigma) can be obtained by using
MC-ICP-MS, with the advantage that the analyses are much more rapid than by TIMS.
490 Cooper & Reid
Alpha spectrometry. Analysis of 210Pb by alpha spectrometry is the most time-consuming
of the 238U-230Th-226Ra-210Pb measurements. 210Pb activity is measured by alpha spectrometry
using 210Po as a proxy. Samples are spiked with 209Pb and counted for approximately one
week. 210Pb activities are determined either by assuming secular equilibrium between 210Pb
and 210Po for samples older than ~2 years (with a half-life of 138 days, 210Po will be in
equilibrium with its parent 210Pb within two years)(Turner et al. 2004; Reagan et al. 2006) or,
for younger samples, by repeated measurements of 210Po over a period of months at various
stages of 210Po ingrowth toward secular equilibrium, which are extrapolated to equilibrium
with 210Pb (Reagan et al. 2005, 2006). Despite the analytical challenges, a few 226Ra-210Pb data
for plagioclase have been published (Reagan et al. 2006, 2008) and further measurements hold
promise to unravel the most recent crystallization history of volcanic minerals.
Summary. The net result is that the amount of time and effort required for preparation of
a mineral separate, preparation of the separate for analysis (mounting and polishing for SIMS
analysis or chemical separation of the elements of interest for TIMS/MC-ICP-MS), and analysis
time, vary widely between the different parent-daughter pairs, different compositions of rocks,
and different analytical techniques. However, even the most straightforward of these analyses
requires a significant effort in terms of time, effort, and funding, which provides an incentive
for careful consideration during sample selection. Nevertheless, with well-chosen samples,
the investment is worthwhile considering that the U-series system provides information about
timing and duration of magmatic events that is simply not possible to gain by other techniques
and that contributes a unique perspective on magmatic processes.
age calculations
U-Th-Pb ages by SIMS. For young accessory phases, 207Pb/206Pb deviations from the
modern-day production ratio of 0.04605 reflect contributions from sources other than in situ
U decay. An important source of this deviation can be contributions from “common Pb”. For
U-bearing minerals (e.g., zircons) of identical age, variable mixtures of radiogenic and common
Pb will define an array on a plot of 207Pb/206Pb versus 238U/206Pb (Fig. 5). The intercept with
concordia (the locus of points defining the covariation in 207Pb/206Pb versus 238U/206Pb over
time; Tera and Wasserburg 1974) defines the age of the zircons and the ordinate intercept
defines the common 207Pb/206Pb (e.g., Baldwin and Ireland 1995). If zircons are of various
ages, it may be more meaningful to calculate model ages by assuming a common Pb value,
e.g., the (initial) isotopic composition of the host rock. In this case, the common Pb correction
equates to extrapolating from the common Pb value on the ordinate through the data point to
the concordia curve (Fig. 5). In some cases, rather than a magmatic value for common Pb,
corrections or allowances for a contaminant Pb may be justified (Schmitt 2003b). Analyses
with a large common Pb component could be the result of the primary beam overlapping or
intercepting inclusions (glass or mineral) and rejection of these results may be justified.
A second source of deviation of 207Pb/206Pb from the modern-day production ratio arises
from initial disequilibria in the U-series decay chains. Of these disequilibria, only 238U-230Th
disequilibrium is likely to have a detectable effect on the U-Pb age but the superimposed effect
of radioactive decay means that the initial 230Th/238U of accessory phases cannot be measured
directly. One approach for estimating the initial 230Th/238U fractionation is to compare the
232
Th/238U for each ionprobe spot to that of the host or, alternatively, to a putative parental melt
composition. In instances where a grain has returned to secular equilibrium, the true age (T) can
be calculated from the measured 206Pb*/238U from the following relationship (Schärer 1984):
λ
206
Pb* = 238
( )
U eλ238T − 1 + 238 ( f − 1)
λ 230
(3)
where f = [Th/U]crystal/[Th/U]melt and 230Th is assumed to fractionate relative to 238U to the same
Uranium-series Crystal Ages 491
Figure 5. Tera-Wasserburg concordia diagram. Shown are 13 analyses of zircon from the Geysers granite,
California. Line fit to the data gives a y-intercept 0.80 ± 0.06, the common 207Pb/206Pb ratio for the zircons
if they are isochronous in age. The isochron intersects concordia (horizontal line with tic marks at discrete
age increments) at 1.11 +0.05/−0.06 (2σ) Ma. This age agrees with an age of 1.13 ± 0.04 Ma obtained by
Dalrymple et al. (1999) for the same zircons. Note that neither age is corrected for an initial, crystallization-
induced 230Th deficit in the zircons. [Used with permission of the Geological Society of America from
Bacon et al. (2000) Geology, Vol. 28, p. 468.]
degree as 232Th. For zircon, a Th/U fractionation factor predicted from Th and U zircon:melt
partition coefficients, f = 0.17 (Blundy and Wood 2003), equates to a maximum (at secular
equilibrium) deficit between measured and true ages of 90 k.y. while average f values of
0.19-0.26 estimated from a large number of igneous rocks (Bindeman et al. 2006) equate
to maximum age deficits of 80-88 k.y. Younger crystals and more complex disequilibrium
scenarios are addressed by Ludwig (1977).
U-Th isochron and model ages for ion microprobe spot analyses can be calculated from
the activities of 230Th, 232Th, and 238U. Activities can be calculated from ion ratios using the
decay constants in Table 1. The calculation of U-Th isochron ages is described in the following
section and is warranted when minerals form at the same time with the same initial Th isotope
ratio. Individual spot analyses can alternatively be expressed as model ages by correcting the
accessory phase activity ratios for an initial Th isotope composition. One approach for this is
to calculate two-point mineral-melt isochrons using the present-day (230Th/232Th)-(238U/232Th)
characteristics of the melt. This assumes, in effect, that the minerals either crystallized from
their host melt or from melts very similar to it. Ratios for the melts may be obtained by TIMS or
MC-ICP-MS following the procedures already outlined. Ten-fold and greater internal variations
in U/Th are often observed in and between zircons (e.g., Brown and Smith 2004; Miller and
Wooden 2004; Charlier et al. 2005; Simon and Reid 2005) which suggests that zircons grow un-
der variable conditions involving widely varying melts. Under favorable circumstances it may
be possible to determine U/Th ratios in melt (now glass) entrapped in the zircons. An alternative
approach is to fix the initial 230Th/232Th and assume that there is no 238U-230Th disequilibrium
in the melt (i.e., assuming (230Th/232Th) = (238U/232Th)). The typical zircon has a large enough
U/Th ratio so that uncertainty in the initial ratio is a modest contribution to the overall model
age uncertainty except for zircon analyses with low (238U/232Th). At the opposite extreme, the
U/Th ratio of allanite is so low that the variation in (230Th/232Th) depends only on its initial Th
isotope ratio and the amount of time elapsed (i.e. the second term on the right in Equation (4)
below can be neglected).
492 Cooper & Reid
We are aware of two interlaboratory comparisons of SIMS ages on young accessory
phases. Bacon and Lowenstern (2000) used a SHRIMP-RG ion microprobe to analyze zircons
from the Geysers granite that had previously been analyzed using a CAMECA ims 1270 by
Dalrymple et al. (1999). They obtained a 238U/206Pb age of 1.11 ± 0.06 Ma (95% confidence
level; MSWD = 0.60) and 207Pb/206Pb ratio of 0.80 ± 0.06 for common Pb (Fig. 5), compared
to a 238U/206Pb age of 1.13 ± 0.04 Ma (MSWD = 1.1) and a 207Pb/206Pb ratio of 0.91 ± 0.09
for common Pb obtained previously. Schmitt and Vazquez (2006) used a CAMECA ims 1270
ion microprobe to analyze zircons from Salton Buttes that had previously been analyzed by
Brown et al. (2004) using a SHRIMP-RG and found closely overlapping U-Th ages between
~10 and ~18 ka.
238
U-230Th-226Ra-210Pb ages of bulk separates. The classical approach to U-series dating
using mineral separates is an isochron approach (e.g., Kigoshi 1967; Allegre 1968). The most
common form of the isochron diagram used for the 238U-230Th parent-daughter pair is shown
in Figure 6a. As with all isochron diagrams, this is a plot of parent vs. daughter (in this case
expressed as activities), both normalized to an isotope of the daughter whose abundance does
not change over the time of interest (in this case, 232Th, which, while not stable, has a half-life
that is long enough that decay can be neglected on the timescales of interest). The form of
Equation (2) relevant to calculating 238U-230Th isochron ages is:
238
(230 Th ) (230 Th )0 − λ230 t ( U) 0
= e + (
1 − e − λ230 t ) ( 4)
( 232
Th) ( 232
Th ) 232
Th ( )
where, because uranium activity will not change measurably over the timescales of interest, Uo
can be assumed to be equal to Umeasured.
A t2
t1
(230Th)/(232Th)
Figure 6. Schematic isochron diagrams for
pl melt A) 238U-230Th and B) 230Th-226Ra. Shown are
t0 plagioclase (triangles), melt (circles) and
cpx clino-pyroxene (squares); the sense of U/Th,
Ra/Ba and Th/Ba fractionation in minerals
is consistent with known partitioning
behavior, but the extent of fractionation is
schematic. Melt was assumed to start with
modest 230Th-excess and 226Ra-excess, but
(238U)/(232Th) the overall behavior for minerals will be the
same in the case of the opposite sense of
B disequilibria in the melt. Tie lines between
symbols are isochrons representing the
time of initial crystallization (t0) and two
t2
melt subsequent times (t1 and t2) corresponding
to approximately one and two half-lives of
(226Ra)/[Ba]
the daughter isotope, respectively. Note that
t1
isochrons on Ba-normalized 230Th-226Ra
pl isochrons (B) are in general kinked lines,
cpx although they will approach a straight line
as the phases approach secular equilibrium.
t0
(230Th)/[Ba]
Uranium-series Crystal Ages 493
The typical assumptions for age calculations using an isochron approach are required: 1)
the duration of crystallization must be short relative to the half-life of the daughter isotope; 2)
all phases used to calculate ages must have precipitated at the same time from the same magma;
and 3) each phase must have remained a closed system with respect to parent and daughter since
the time of crystallization. Minerals with the same crystallization age will form an inclined
array with slope proportional to the crystallization age. The 238U-230Th isochron has a number
of distinctive features in comparison to ‘conventional’ isochron diagrams (such as Rb-Sr). For
example, radioactive equilibrium is represented on the 238U-230Th isochron diagram by a line
of unity slope (known as the equiline), and any phase with initial disequilibrium will move
toward the equiline via ingrowth or decay of the daughter 230Th. This means that phases can
move vertically upward or downward, rather than the continual increase in daughter isotope
abundance (and corresponding decrease in parent isotope abundance) shown by ‘conventional’
systems. These trajectories will be near-vertical on a U-Th isochron diagram, because decay
of 238U and 232Th are slow enough that their activities will not change measurably over the
timescales of interest. Any phases that have aged enough to reach radioactive equilibrium will
remain on the equiline until some additional geologic event fractionates parent from daughter;
this means that a mineral array that coincides with the equiline has an indeterminate age. In
addition, as each phase exponentially approaches equilibrium the rate of vertical movement
on an isochron diagram will slow; this leads to asymmetric age uncertainties, as a symmetric
analytical uncertainty in 230Th/232Th ratio will represent a longer period of time on the side of
the data point closest to the equiline than on the side of the data point farther from the equiline.
Finally, the initial (230Th)/(232Th) ratio of the array is given by the intersection of the mineral
array with the equiline, rather than the intersection with the y-axis. The 238U-230Th isochron
approach has been used successfully to date bulk mineral separates in a large number of cases
(e.g., Allegre 1968; Volpe and Hammond 1991; Reagan et al. 1992; Volpe 1992; Schaefer et
al. 1993; Bourdon et al. 1994; Black et al. 1997, 1998a,b; Heath et al. 1998; Bourdon et al.
2000; zellmer et al. 2000; Cooper et al. 2001; Heumann and Davies 2002; Charlier et al. 2003;
Turner et al. 2003a,b; Tepley et al. 2006; Snyder et al. 2007), suggesting that the assumptions
required for its use are often justified. We discuss in later sections cases where one or more of
these assumptions may have been violated, and the information that may still be derived from
an isochron approach even in these more complex scenarios.
An analogous diagram can be constructed for 230Th-226Ra parent-daughter pair, with the
additional complication that Ba is used as a normalizing element because no longer-lived
isotope of radium exists (Fig. 6b). In the case of the Ba-normalized 230Th-226Ra isochron
diagram, an additional assumption that Ba is an exact chemical analog for Ra is required (e.g.,
Volpe and Hammond 1991; Reagan et al. 1992). As has been predicted for most minerals
(Blundy and Wood 1994, 2003) and recently shown experimentally in the case of plagioclase
(Miller et al. 2007a), there are in fact significant differences between the partitioning behavior
of radium and barium in most rock-forming minerals. In most cases, radium is predicted
to behave more incompatibly than barium (or, in the rare cases like K-feldspar where Ra
and Ba are both compatible, less compatibly), leading to the preferential exclusion of Ra
from the crystal lattice and to a correspondingly low Ra/Ba ratio in the mineral at the time of
crystallization relative to Ra/Ba in the liquid (Blundy and Wood 2003). This effect will lead to
deviations of co-precipitating phases from a straight-line array on a Ba-normalized 230Th-226Ra
isochron diagram (Fig. 6b), and age calculations assuming two-point straight-line isochrons
can be either older or younger than the true crystallization age. In order to circumvent this
problem, Cooper and Reid (2003) developed the use of a radium evolution diagram to provide
a method of graphically estimating ages (Fig. 7). This diagram plots (226Ra)/[Ba] vs. time in
mineral phases and groundmass/whole rock/glass. (226Ra)/[Ba] at the time of measurement
plots along the y-axis, and curves representing the evolution of (226Ra)/[Ba] over time are
calculated based on measured (230Th)/[Ba] ratios. To this point, the treatment of the data is
494 Cooper & Reid
A
meltpl
meltpx
226
Ra)/[Ba]
gm
(
pl
px
years before present
5000
B
4000
(226Ra)/[Ba]
3000
meltau meltpl
2000
gm
1000
0
0 500 1000 1500 2000 2500 3000
years before present
Figure 7. Ra evolution diagrams. A) Schematic Ra evolution diagram showing (226Ra)/[Ba] ratios in
minerals and melt projected back in time as predicted by their measured (230Th)/[Ba] ratios. The time
of measurement is at the y-axis. Shown are curves for plagioclase (dashed line), pyroxene (dotted line)
and groundmass (assumed to represent the melt composition; thin solid line); compositions of pyroxene
and plagioclase must be corrected for the effects of impurities before plotting. The intersection of the
curves for pyroxene and plagioclase with the curve for groundmass represent the conventional two-point
isochron ages uncorrected for Ra/Ba fractionation. Also shown are curves for melt in equilibrium with
plagioclase and pyroxene (lableled meltpl and meltpx, respectively), calculated at each point in time by
dividing (226Ra)/[Ba] in each mineral at that point in time by the ratio DRa/DBa for that mineral. The age
of crystallization is given by the point in time where the composition of the melt was equal to that in melt
in equilibrium with the minerals, shown graphically by the intersection of curves for meltpl and meltpx
with the curve for groundmass. Note that the two-point mineral-melt isochron ages uncorrected for Ra/Ba
fractionation can be either too old (plagioclase) or too young (pyroxene) depending on the initial (226Ra)/
(230Th) ratio and Ra/Ba ratio in the minerals. B) Example evolution diagram constructed using data for a
sample of the 1955 eruption of Kilauea volcano, Hawaii (modified from Cooper et al. 2001). Solid curves
for melt in equilibrium with plagioclase correspond to An70 at 1160 °C (upper curve) and An60 at 1100 °C
(lower curve); the dotted curves allow for 10% uncertainty in calculated DRa/DBa for plagioclase. Dashed
lines for groundmass evolution through time indicate propagation of measurement uncertainties. Box with
diagonal ruling represents age range of intersection of curves for groundmass andCooper&Reid
melt in equilibrium
Figure 7
with plagioclase. Curves for melt in equilibrium with augite (meltau) are also shown; solid and dot-dash
lines delimit range of curves permitted by ± 35% uncertainty in partition coefficients and variations due
to temperature. Data are consistent with co-precipitation of pyroxene and plagioclase from a melt with
composition of the groundmass at an age of 1000 +300/−400 y (Cooper et al. 2001).
Uranium-series Crystal Ages 495
conceptually the same (although graphically different), from an isochron approach, and the
intersection of curves for mineral and liquid (e.g., plagioclase or pyroxene and groundmass
in Fig. 7) represents the conventional two-point isochron age. However, the true age must
take into account differential partitioning of Ra and Ba, and therefore at each point in time,
(226Ra)/[Ba] in the mineral is corrected for the effects of fractionation by dividing by the ratio
DRa/DBa. This new curve represents the composition of a melt in chemical equilibrium with
the mineral (as defined by the partition coefficients) at any given point in time, and the model
age of crystallization is given by the intersection of this curve with the curve for evolution of
(226Ra)/[Ba] in the melt (represented by whole-rock, glass, or groundmass data, whichever is
most appropriate in a given case). We consider this to be a model age because it is subject to
uncertainties in the partition coefficients for Ra and Ba. At this point, only one experimental
determination of radium partition coefficient (in anorthite) has been published (Miller et al.
2007a). Therefore, we rely on calculations of radium partition coefficients based on an elastic-
strain model using barium as the proxy element. Some confidence in this approach is given by
the fact that the elastic-strain model agrees with the experimental data for anorthite (Miller et
al. 2007a). A full discussion of calculation of these partition coefficients is beyond the scope
of this paper, and we refer the reader to Blundy and Wood (2003) for a complete review of
partitioning behavior of U-series elements and discussion of how best to estimate radium
partition coefficients for minerals.
A further complication of calculating 230Th-226Ra ages—relative to calculating 238U-230Th
or other ages where an isochron approach is feasible—is that impurities in the mineral separates
will result in differences in the calculated ages. As is apparent from Figure 6b, mixtures of
different phases will generally move data points away from a tie-line connecting pure minerals
and melt, and will therefore affect the calculated ages. In effect, the model age calculations
assume that the mineral separates are pure because they rely on partition coefficients to correct
for Ra/Ba fractionation during crystallization; therefore, the composition of the pure mineral
phases must be estimated in order to calculate ages. Two approaches can be used to improve
estimates of Ra, Th and Ba concentrations in the pure minerals: one method is to prepare
mineral separates that are as pure as possible, which will minimize the necessity for correction.
However, due to the common presence of inclusions of other minerals and melt within phases
such as plagioclase, K-feldspar, pyroxene, or amphibole, it is generally not practical to prepare
separates that are 100% pure. Instead, it is necessary to correct the measured U-series element
concentrations for the effects of impurities. This can be accomplished by taking a split of the
dissolved mineral separate prior to U-Th-Ra analysis and analyzing its trace-element composition
by ICP-MS, and comparing this to in situ measurements of trace elements in the pure mineral
phase by SIMS or LA-ICP-MS. The trace-element concentrations of the bulk mineral solution
can then be compared to trace-element data for other phases which may be present (e.g., melt,
other mineral phases) in order to correct the measured U, Th, and Ra concentrations for the
effects of these impurities. To date, this approach has generally relied on only a few elements
(e.g., Cooper et al. 2001; Cooper and Reid 2003; Turner et al. 2003a,b), but by incorporating a
suite of elements (e.g., Tepley et al. 2006) and by taking a least-squares approach to fitting the
data, these uncertainties could be further minimized. Even with the relatively crude approach of
using only a few elements for correction of impurities, the magnitude of the uncertainty on the
ages contributed by the correction for impurities is of the same order as the uncertainty due to
the choice of pre-eruptive temperature and average crystal composition used when calculating
partition coefficients to use for correction for initial Ra/Ba fractionation. The age calculations
are sensitive only to the ratio DRa/DBa, not to the absolute magnitude of DRa and DBa, both of
which change in concert with changes in the partitioning parameters (site radius and Young’s
modulus), and therefore uncertainties in these variables affect the absolute values of the partition
coefficients much more than the ratio (as discussed in more detail in Cooper et al. 2001). As
a result, the uncertainties in 230Th-226Ra model ages are dominated by uncertainties in the
496 Cooper & Reid
pure mineral compositions and in the pre-eruptive temperature, with measurement uncertainties
for the U-series nuclides and uncertainties in the partitioning parameters playing a relatively
minor role. This offers the potential that by improving our ability to quantitatively account for
impurities in the future, we can significantly improve the uncertainty in model ages.
There are few analyses of 226Ra-210Pb disequilibria in mineral separates reported in the
literature; two examples are analyses of plagioclase separates in lavas from Arenal (Reagan et
al. 2006) and Mount St. Helens (Reagan et al. 2008). An isochron approach has not been applied
to 226Ra-210Pb crystal ages, in part because the ages of crystallization (for at least some part of
crystals) in systems analyzed to date are not short with respect to the half-life of 210Pb (e.g.,
Reagan et al. 2006, 2008). Lead is incompatible in calcic plagioclase, becoming compatible
in sodic plagioclase (Bindeman et al. 1998), and ratios of DPb/DRa will be greater than unity
in all plagioclase compositions (Blundy and Wood 2003). Using the method of Blundy and
Wood (2003 and references therein) to calculate partition coefficients for Ra, and comparing to
those for Pb calculated using the data of Bindeman et al. (1998), initial (210Pb)/(226Ra) ratios in
plagioclase can be estimated and 226Ra-210Pb model ages calculated from the time required for
decay from estimated initial ratios to the measured (210Pb)/(226Ra) ratios (Reagan et al. 2006,
2008). As discussed in more detail below, this approach captures an average crystallization
age of the crystals (albeit weighted toward young crystal growth) and therefore can be used to
estimate the proportion of crystal growth that occurred within decades of eruption.
Effect of mixing on mineral ages. Mixing arrays will produce straight lines on isochron
diagrams which can be mistaken for arrays with age significance, therefore it is useful here
to assess the effects of mixing on mineral ages. Mixing in this context can refer to several
different scenarios: 1) mixing between different phases of the same age in the same rock; 2)
mixing between old and young components of a given mineral in the same rock, where all
mineral fractions precipitated from the same liquid; 3) mixing between different components
of a given mineral which may have the same or different ages but which precipitated from
different liquids. Scenarios 2 and 3, and the implications for age interpretations, are discussed
below; here we briefly discuss the effects on calculated ages of mixing between different
phases within the same system. In the case of the 238U-230Th system, minerals of the same
age define straight lines on an isochron diagram. Mixing between any two phases also results
in a straight-line array where the mixed data lie between the two end members. Therefore,
mixing between phases in the same rock will not affect the slope of the isochron array (or the
calculated age). However, mixing will generally decrease the spread of the array in terms of
U/Th ratios, which will lead to more uncertainty on the slope of the line for a given analytical
uncertainty, and therefore will increase the uncertainty of the age.
230
Th-226Ra and 226Ra-210Pb ages are both model age calculations, dependent on the
assumption that calculated partition coefficients applied to the measured ratios will accurately
reflect interelement fractionation during mineral growth. This will only be true if the measured
ratios reflect the composition of the mineral in question, rather than a mixture of the mineral
being analyzed and any impurities in the bulk separate. As shown on Figure 6, mixtures between
different phases of the same age can significantly change both (226Ra)/[Ba] and (230Th)/[Ba] in
the mineral separate. Depending on the mineral in question and the nature of the impurities,
impurities can have the effect of either increasing or decreasing the model ages compared to the
true crystallization age. This effect is exacerbated in the case of minerals where radium, thorium
and barium are highly incompatible (e.g., pyroxene), in which case even small mass fractions
of impurities can dominate the U-series budgets of the mineral separate. For example, Cooper
et al., (2001) calculated that of the 1.37 fg/g Ra in the pyroxene separate, the pure pyroxene
contains only ~0.21 fg/g, indicating that ~85% of the radium present in the pyroxene separate
originated from impurities, even though the same calculations estimated <1% impurities
(melt+plagioclase) in the separate. As discussed above, this highlights the need for taking care
Uranium-series Crystal Ages 497
to minimize impurities during preparation of mineral separates, and the need to develop more
robust methods of estimating the composition and percentage of impurities in order to correct
the bulk separate data for these effects. Exactly how best to go about this will depend on the
details of the sample and must ultimately be treated on a case-by-case basis. Even in the absence
of mass-balance calculations using many elements in multiple phases, in most cases maximum
and minimum concentrations can be estimated by identifying the phases that will have the most
leverage on the system and by quantifying their contributions using mass-balance calculations
(e.g., Cooper et al. 2001). Similar arguments apply to 226Ra-210Pb ages.
effect of diffusion on crystal ages
A chemical potential for U-series elements could potentially be imposed between crystals
and liquid if crystal cores are isolated by continued crystal growth as the melt continues to
differentiate, or if crystals are entrained by a new host melt with different composition from the
original melt. In addition, chemical potentials could develop in minerals simply due to changes
in concentration due to decay. Whatever the mechanism producing chemical gradients, they
could drive diffusion of parent and daughter nuclides into or out of crystals, which in turn could
affect U-series age calculations and could either increase or decrease apparent ages. The effect
of diffusive re-equilibration on crystal ages depends on the time that a crystal spends in contact
with the melt and on the diffusivity of the parent and daughter elements, which generally are
slow at magmatic temperatures. For example, diffusion of U, Th, and Pb in zircon at magmatic
temperatures is slow enough that diffusive mobility is negligible over the timescales discussed
here (Cherniak and Watson 2003). Temperatures high enough to allow significant mobility of
U and Th in zircon would be above the zircon saturation temperature for most silicic magmas,
and therefore zircon would likely dissolve before diffusively re-equilibrating with a new host
liquid. Vazquez and Reid (2004) estimated that Th diffusion in allanite would only affect a few
micrometers in allanite over a period of 100,000 years at temperatures that characterize allanite-
bearing rhyolites (<800 °C).
Diffusion of U and Th in plagioclase is very slow at magmatic temperatures, based on diffu-
sion parameters calculated using the model of Van Orman et al. (2001) which predicts diffusion
coefficients of order 10−21 m2/sec at 1200 °C (Saal and Van Orman 2004; Van Orman et al. 2006).
Diffusivity of radium in plagioclase has not been measured, but can also be estimated using
experimentally-derived Ba diffusion parameters (Cherniak 2002) and the elastic strain model of
Van Orman et al. (2001, 2006); this estimate is of the same order of magnitude as Ba (Cherniak
2002), which is approximately three orders of magnitude faster than U and Th (Van Orman et
al. 2006). Given these estimates of diffusion parameters for U, Th, and Ra in plagioclase, U-Th
crystal ages are not likely to be affected by diffusion during tens to hundreds of thousands of
years of residence at magmatic temperatures. Ra-Th ages will be robust for tens to hundreds of
thousands of years at temperatures less than ~1000 °C, but ages could potentially have been af-
fected by thousands of years of residence at temperatures >1100 °C (Cooper and Reid 2003).
Diffusion of U and Th in diopside is also very slow (diffusion coefficients on the order
of 10−20 m2/s at 1200 °C; Van Orman et al. 1998, 2006), and U-Th ages of clinopyroxene are
therefore likely to be robust at magmatic temperatures for the timescales of interest for U-series
dating. Radium diffusion parameters for clinopyroxene have been estimated by Van Orman et
al. (2006), and are expected to be 1-3 orders of magnitude faster than U and Th, suggesting
that (as with Ra in plagioclase) protracted residence at relatively high magmatic temperatures
(>1000 °C) could result in some diffusive effects on Ra-Th ages for pyroxene.
Although the diffusivity of Ra and Ba is predicted to be faster than that of U and Th
in plagioclase, it is still relatively slow (diffusion coefficients on the order of 10−20 m2/s at
1000 °C; Cherniak 2002). Watson (1996a) and Watson and Liang (1995) proposed a model
for trace-element enrichment in a crystal lattice due to surface effects when crystal growth
498 Cooper & Reid
rates are rapid relative to diffusion of the element of interest. Growth entrapment refers to
the incorporation of concentrations of incompatible elements in a crystal that are higher than
concentrations predicted by bulk crystal-liquid equilibrium, due to the differences in crystal
lattice sites at the surface monolayer compared to sites in the interior of the crystal. These
higher surface concentrations can be trapped within a growing crystal if the element of interest
cannot diffuse outward faster than the crystal-liquid interface grows outward (Watson and Liang
1995; Watson 1996a). Given estimates for Ra and Ba diffusion rates compared to estimates of
plagioclase crystal growth rates, Ra/Ba ratios (and therefore calculated ages) could be affected
by growth entrapment during crystallization at temperatures less than 1000 °C (Fig. 8; Cooper
and Reid 2003). The relative effects of such growth entrapment on different elements are not
well known, but, given that both D’s will be closer to unity, it is likely that Ra/Ba ratios would
be increased relative to bulk chemical equilibrium due to a decreased ability of the crystal to
discriminate between the two elements, which would lead to a decrease in apparent crystal
age (or even an undefined crystal age; see Cooper and Reid 2003). The concordance between
U-Th and Ra-Th ages of plagioclase in most cases, together with the concordance of Ra-Th
ages calculated from different phases, suggests that this is not a general phenomenon, but
it may be important to consider in the case of rapid decompression-induced crystallization
of plagioclase, which may be orders of magnitude faster than cooling-induced crystallization.
In summary, diffusive re-equilibration between crystals and liquids is unlikely to affect
U-series crystal ages, except in the case of 230Th-226Ra ages of plagioclase and pyroxene
10-8 Pe
=1
0
=2
Pe
.1
10-10 Pe
=0
V (m/sec)
10-12
Ra 1000
10-14
1200
1000
900
10-22 10-21 10-20 10-19 10-18 10-17
Diffusion coefficient (D) m2/sec
Figure 8. Plot of diffusivity vs. growth rate illustrating conditions of potential growth entrapment (modified
after Cooper and Reid 2003). Solid lines represent contours of constant growth Peclet number Pe = Vl/Di,
where V is the crystal growth rate, l is the half-thickness of the enriched surface layer (assumed to be 50
angstroms), and Di is the lattice diffusivity of element i (Watson and Liang 1995). Growth entrapment (i.e.,
concentrations higher than bulk crystal-liquid equilibrium) of element i will be measurable at Pe = 0.1,
significant for Pe > 2, and ~100% effective (i.e., concentrations on the interior will be equal to those in the
surface layer) at Pe = 10 (Watson and Liang 1995). Therefore, these values delimit regions of inefficient,
partial, and efficient growth entrapment, respectively; the region where growth entrapment is possible is
shaded. Estimates of plagioclase growth rates for the 1980-1986 dacite dome are shown by ruled fields.
Experimentally-determined diffusion coefficients for Ba in labradorite (An67) from Cherniak (2002) are
shown by vertical dashed lines (labeled with temperatures in °C). Also shown is diffusion coefficient for Ra
at 1000 °C (dotted line) calculated by Cherniak using the method of Van Orman and others (2001). Based
on these diffusivities and growth rates, growth entrapment is expected for crystallization at temperatures
less than 1000 °C, and could occur to some degree at temperatures up to 1200 °C if growth rates were
similar to maximum estimates. See text for discussion.
Uranium-series Crystal Ages 499
held at high temperatures (>1000 °C) for thousands to tens of thousands of years. Growth
entrapment of high concentrations of incompatible trace elements, which will be controlled by
the rate of diffusion of cations, is unlikely to affect 230Th-226Ra ages in the case of relatively
slow, thermal-driven crystallization. In contrast, this may be an important factor to consider in
the case of rapid, decompression-induced crystallization of plagioclase, and would have the
effect of making calculated plagioclase model ages too young.
What dO CryStaL aGeS teLL uS?
Introduction
Early analyses of U-series disequilibria in minerals were performed in an attempt to
date the eruption age of a lava (Kigoshi 1967; Taddeucci et al. 1967); some examples of
successful dating of eruptions using this method are presented below. However, it was also
recognized early on that crystallization of minerals in a volcanic rock could pre-date eruption
by a significant time period (e.g., Allegre 1968). The number of U-series analyses of minerals
in volcanic rocks has increased dramatically over the past two decades, primarily driven by
analytical advances which have decreased the sample size requirements and made analysis
of minerals more practical. This applies both to bulk mineral analyses, where multicollector
mass spectrometry (TIMS or MC-ICP-MS) have allowed analysis of smaller samples at higher
precision than was possible by alpha spectrometry (e.g., Goldstein and Stirling 2003), and to
the development of in situ dating of accessory minerals with relatively high U and Th contents
(Reid et al. 1997; Vazquez and Reid 2004). The greater analytical and interpretive challenges
of 230Th-226Ra and 226Ra-210Pb mineral data have limited the number of 230Th-226Ra-210Pb
mineral data compared to 238U-230Th analyses, but the database of 230Th-226Ra and even 226Ra-
210
Pb crystal ages has also grown significantly over the past decade.
The recognition that many, if not most, crystals in volcanic rocks pre-date eruption by
significant time periods (thousands to hundreds of thousands of years; Reid 2003), coupled
with increasing evidence of mineral-melt disequilibrium from techniques such as single-crystal
or sub-crystal trace-element and isotopic analyses, textural information, and diffusion studies
(e.g., Ramos and Tepley 2008; Streck 2008; Costa 2008), indicates that many crystals contain
records of magmatic processes that pre-date their residence in the host magma. However,
given the complexity of magmatic processes recorded in crystals, the traditional distinctions
between “phenocrysts” precipitated from the host magma and “xenocrysts” incorporated from
wall rocks unrelated to the magmatic system are insufficient to describe the diversity of origins
of crystal material within a magma. To distinguish phenocrysts from cogenetic crystals that are
recycled from earlier phases of magmatism within the same system, Wes Hildreth (presentation
at the Penrose Conference on “Longevity and Dynamics of Rhyolitic Magma Systems”, 2001)
proposed adoption of the term “antecrysts,” which we use here. Miller et al. (2007b) further
proposed that the term “phenocryst” is a textural term that does not apply to, for example,
zircon crystals that may have grown from the host liquid but which are nevertheless quite small
relative to other crystals in the rock. They propose the term “autocryst” to identify crystals
grown from the host liquid, and we adopt that terminology here.
Especially in the case of bulk mineral separates, where ages are averaged over many
hundreds or thousands of crystals, U-series ages may reflect contributions from crystals
with different origins ranging from phenocrysts sensu stricto (i.e., autocrysts) to antecrysts
scavenged from mush zones or young solidified intrusions produced by earlier magma batches
that are part of the magmatic activity at the same center, to true xenocrysts incorporated from
wall rocks that are unrelated to the magmatic activity at a given center. Interpretation of crystal
ages in terms of the thermal and chemical history of a given magmatic system thus depends
to a large degree on whether the crystals analyzed show evidence of antecrystic or xenocrystic
500 Cooper & Reid
components, or whether they are dominated by autocrysts. In the latter case, crystal ages can
provide information on magma residence times, average crystal growth rates, and differentiation
timescales of magmas (Reid 2003). In more complex scenarios, where there is independent
evidence of antecrystic or xenocrystic components in the minerals analyzed, crystal ages
can still provide information about the age and proportions of various crystal components in
a magma, and, when linked with crystal thermometry, barometry, textural information, and
trace-element or isotopic information, can provide unique insights toward unraveling magmatic
histories. In particular, U-series ages provide time information over timescales that are not
accessible through any other technique, and therefore can be used to place other chemical and
textural information in a temporal context. For example, as discussed in more detail below,
U-series data have been used to identify different crystal components that are present in related
magmas; to identify magmas within the same system that were present at the same time but did
not apparently interact with each other chemically; to quantify the age and proportion of old and
young crystal populations present in a magma; to identify young plutonic material that is related
to erupted magmas and to identify crystals originating from these young plutons which were
incorporated into young magmas; to track the thermal history of a magma body through time;
and to identify ages of prominent differentiation events, among other applications. Crystal ages
are thus crucial in understanding the rates and dynamics of magmatic processes such as crystal
growth, storage, recycling, magma differentiation and mixing.
using u-Pb and decay series ages to date eruptions
U-Pb analyses, especially of zircon, are a longstanding and accurate means of dating
igneous rocks (see Davis et al. 2005, for a historical overview); regular improvements in
analytical sensitivity continue to reduce age uncertainties and refine interpretations (e.g.,
Crowley et al. 2007). Early on, U-series analyses of igneous rocks likewise focused on its
potential for dating volcanic rocks, albeit necessarily restricted to eruptions that occurred in the
past <350 k.y. Even so, this method provides a means of directly dating volcanic rocks that is
complementary or an alternative to other methods. The latter includes 40Ar/39Ar dating (which
may be subject to inheritance, especially in mafic rocks) and the more indirect approaches
of 14C-dating and thermoluminescence-dating, and cosmogenic exposure age dating (which
depend on recovery of organic materials or soils, or minimization of the effects of erosion,
respectively). As described above, the age sensitivity of dating volcanic rocks using U-series
disequilibrium depends on magnitude of parent-daughter fractionation and on the duration of
return to secular equilibrium.
In a series of recent studies, Jicha and others (2005, 2007, 2008) obtained internal
U-Th mineral isochrons for major and minor minerals (e.g., plagioclase, clinopyroxene,
orthopyroxene, magnetite) and for groundmass separated from lava flows and pyroclastic rocks
in arc settings (Seguam Island, Aleutians; Puyehue–Cordón Caulle volcanic complex, Chile;
and Mount Adams volcanic field, Washington). In virtually every case, the U-Th isochron
ages are indistinguishable from high precision 40Ar/39Ar ages and, by extension, from the
time of eruption. An important conclusion to be drawn from these studies is that the major
mineral phases must have grown shortly (<1000 yr, based on analytical uncertainties) before
eruption. Dating of accessory phases from young volcanic systems has also yielded ages in
some cases that are concordant with those obtained by other methods. In their study of the
Bishop Tuff, Crowley et al. (2007) chemically processed and analyzed individual zircons and
obtained 238U-206Pb ages for most that, at 767.1 ± 0.9 ka, are largely concordant with the 770.4
± 3.6 ka eruption age obtained by 40Ar/39Ar dating (Sarna-Wojcicki et al. 2000, corrected to
Taylor Creek sanidine neutron fluence monitor). 235U-207Pb ages are, however, ~52 k.y. older
and further work is needed to resolve this difference. A U-Th isochron defined by zircon and
apatite aliquots from Puy du Dome, France (Condomines et al. 1997) also passes through data
points for potassium feldspar and biotite and the derivative age is in excellent agreement with
thermoluminescence and 14C age dating. Most crystallization must have occurred within a
Uranium-series Crystal Ages 501
thousand years before eruption.
As this review explores in greater detail below, the timing of phenocryst growth can
be resolvably older than that of eruption. Accordingly, a better approach to dating volcanic
eruptions would seem to be to date groundmass phases. This presents the difficult challenge of
separating sufficient mineral quantities for precise analysis while avoiding the contaminating
effects of phenocryst fragments and/or glass (which can easily swamp the U-series nuclide
signatures of the mineral phases). Nonetheless, Sims et al. (2007) used the internal isochron
defined by groundmass and groundmass plagioclase to estimate an eruption age of 68 +24/−20
ka (2σ) for the Bluewater Flow, New Mexico. Discrepancies between this age and ages from
40
Ar/39Ar (2.3-5.6 Ma) and cosmogenic nuclide abundances (39-56 ka) can be explained by the
effects of inherited Ar and of erosion, respectively.
Pre-eruptive crystal growth—‘simple’ scenarios
In situ ages. Over the past decade, a number of studies have dated individual accessory
phases, and subdomains within those phases, from young (most <2 Ma), mainly silicic,
systems. Instead of being eruption-aged, most zircon and allanites give crystallization ages—
and attendant closure to Th and Pb isotope exchange (Cherniak and Watson 2003; Vazquez
and Reid 2004)—in excess of those of eruption, as gauged mainly by the difference between
the crystallization ages and Ar-Ar and/or K-Ar eruption ages. Thus rather than focusing on
dating eruption, the attention has shifted to using accessory phase ages to gain insights into the
magma evolution. Simon et al. (2008) recently summarized existing results for U/Th and U/
Pb ion microprobe analyses of young zircons in terms of 1) the oldest zircon age and 2) mean
pre-eruption ages, calculated as the difference between mean zircon age (tzir ) and the eruption
age (terup) (reproduced here as Table 3). More detailed information about many of these results
is provided later in this chapter.
Volumes of erupted magma that have been studied range from <1 km3 (e.g., Reid et al.
1997; Simon and Reid 2005) to ≥1000 km3 (e.g., Bindeman et al. 2001; Bindeman et al. 2006).
Many studies find zircon ages that are several hundreds of k.y. older than the eruption age and
mean pre-eruption intervals, ∆ tzir-terup , for virtually all eruptions lie in the range of several tens
to a few hundreds of k.y. (Table 3; Reed 2003; Simon et al. 2008). Smaller volume effusive
eruptions tend to have longer pre-eruption intervals than larger explosive eruptions but there
is considerable overlap in this aspect of the systems. A straightforward interpretation of these
results is that magmas are stored for tens to hundreds of thousands of years before erupting. On
the other hand, age distributions commonly show peaks in the timing of crystallization (e.g.,
Reid et al. 1997; Brown and Fletcher 1999; Reid and Coath 2000; Vazquez and Reid 2002;
Miller and Wooden 2004; Charlier et al. 2005). Sequential eruptions from a given center may
exhibit common age populations (e.g., Reid et al. 1997; Vazquez and Reid 2002; Charlier et
al. 2005), suggestive of repeated remobilization of the same magma reservoir or subvolcanic
intrusion. In other cases, temporally and geographically proximal eruption centers may emit
eruptions with distinct zircon age populations (e.g., Charlier et al. 2003). U-Pb dating of igneous
xenoliths (e.g., Bacon and Lowenstern 2005; Schmitt 2006) and of shallow plutons (Schmitt
et al. 2003b) reveal zones in which solidification was complete or nearly complete at the same
time as inferred magma storage. Such results beg questions about the geologic significance
of the long pre-eruption intervals. Do large pre-eruption ages reflect long storage intervals of
crystals in liquid-rich systems and therefore the residence time of what might conventionally
be considered “magma”? Or do crystals become remobilized and/or recycled from crystalline
mush or subvolcanic plutons by new introductions of melt, in which the ages reflect a mean
crystal age for the magma and the evolution of the subvolcanic system? Considered in this
light, the apparently “long” mean pre-eruption ages of most silicic systems (Table 3; Simon et
al. 2008) could integrate an evolutionary history during which the accessory phases may have
been immersed in liquid for only a fraction of this time.
502 Cooper & Reid
table 3. Summary of age characteristics of silicic extrusions (from Simon et al. 2008). [Used with permission of
Elsevier from Simon et al. (2008) Earth and Planetary Science Letters, Vol. 266, p. 190.]
eruption agea Oldest zircon ageb Mean pre-eruption interval (ka)
dating
Sample Locality
(ka) (ka) (ka) (se) Calculatedc 2se Modeled sd method
Explosive large volume
Oruanui Taupo, Nz 27 27 178 4 18 5 52 3 U/Th
(218) (125) U/Th
Las Tres Baja,Mx 36 36 168 48 85 12 74 9 U/Th
(289) (100) U/Th
Tihoi Taupo, Nz 45 45 166 58 27 8 62 5 U/Th
(227) (280) U/Th
Earthquake Taupo, Nz 62 63 151 71 47 19 62 9 U/Th
(258) (98) U/Th
Rotoiti Taupo, Nz 62 64 229 89 15 15 60 8 U/Th
Okaia Taupo, Nz 29 29 158 48 27 15 71 6 U/Th
Kos Plateau Aegean Arc 161 164 287 40 64 14 80 3 U/Th
479 74 U/Pb
Whakamaru Taupo, Nz 340 338 608 20 93 22 92 7 U/Pb
Rockland Tuff Lassen Peak, CA 576 584 814 69 70 20 134 13 U/Pb
Lava Creek Yellowstone, WY 639 655 710 30 8 23 62 8 U/Pb
Early Bishop Long Valley, CA 760 777 926 36 64 8 104 6 U/Pb
Late Bishop Long Valley, CA 760 777 926 25 34 7 84 9 U/Pb
Effusive small volume
South Deadman Long Valley, CA 0.6 0.6 271 79 43 29 184 14 U/Th
Crater Lake Crater Lake, OR 27 27 175 79 7 5 52 7 U/Th
Pitchstone Plateaud Yellowstone, WY 60 61 233 47 40 11 33 5 U/Th
West Yellowstoned Yellowstone, WY 109 110 306 21 53 20 66 8 U/Th and
U/Pb
Deer Mountain Long Valley, CA 115 118 285 89 38 25 90 9 U/Th
Solfatara Plateaud Yellowstone, WY 116 117 199 30 50 12 55 7 U/Th
Dry Creek” Yellowstone, WY 159 161 200 70 6 26 40 8 U/Th
Devils Kitchen Coso, CA 587 602 800 20 84 22 123 8 U/Pb
Tyler Valley Gey Geysers, CA 670 689 1148 48 261 30 181 18 U/Pb
YA Glass Mountain Long Valley, CA 885 894 1119 64 87 18 94 17 U/Pb
YG Glass Mountain Long Valley, CA 900 916 1215 91 17 17 130 13 U/Pb
CM0002 ACR Gey Geysers, CA 1150 1169 1488 36 151 20 93 10 U/Pb
ACR Gey Geysers, CA 1150 1169 1566 96 191 20 124 13 U/Pb
CM0004 rhyodacite Gey Geysers, CA 1100 1128 1499 39 182 20 143 16 U/Pb
K3154 ACR Gey Geysers, CA 1180 1205 1628 32 155 20 153 11 U/Pb
CM0003 dacite Gey Geysers, CA 1000 1017 1408 39 243 20 216 18 U/Pb
OD Glass Mountain Long Valley, CA 1686 1712 2031 64 141 13 223 13 U/Pb
OL Glass Mountain Long Valley, CA 1926 1945 2284 55 154 33 119 16 U/Pb
OC Glass Mountain Long Valley, CA 2045 2066 2463 61 166 29 247 20 U/Pb
Pine Mt Gey Geysers, CA 2170 2208 2672 105 252 30 145 23 U/Pb
Data sources included in Simon et al. 2008
a
Eruption ages listed first as originally reported and then after correction for K-Ar decay constant error and choice and ascribed age of neutron
influence monitor (cf. Renne et al. 1998 ); b Parenthetical dates reflect zircon growth that is probably older than the oldest reported zircon dates
because they are within secular equilibrium and imprecise.; c Weighted mean pre-emption ages calculated from U/Th data will be bias towards a
younger age because U/Th ages have larger age uncertainty as dates approach secular equilibrium.; d Not small volume
Bulk separate analyses. Mineral ages determined from bulk analyses represent an average
of the ages of all crystals/zones within the population analyzed, weighted by volume of the
components of different ages and by the concentration of parent in each of the components.
Interpretations of the geologic significance of such an age depend on exactly what scenario one
envisions for crystal growth and the relation between crystals and host liquid. In the simplest
case, the assumption is that phenocryst-sized crystals precipitated from the host liquid during
a short time interval relative to the half-lives of the isotopes of interest. The extent to which a
given system conforms to this simple scenario depends in part on the half-life of the daughter
isotope in a given parent-daughter pair: a ‘short’ crystallization interval may be a more accurate
Uranium-series Crystal Ages 503
approximation for U-Th ages than for Ra-Th or 226Ra-210Pb ages. Documenting bulk chemical
equilibrium between crystals and liquid is a necessary (though not sufficient) condition for
demonstrating crystallization from the host liquid, and therefore for interpreting crystal ages
in terms of magma residence. In addition, a simple and rapid crystallization history will lead
to concordant ages between different parent-daughter pairs and between phases expected to
be crystallizing at the same time; thus, as discussed in more detail below, discordant ages are
themselves an indication of a more complex history recorded in the crystals. One somewhat
more complex scenario is that of progressive crystal growth from the same magma, in which
case the crystal age will represent an average of the duration of crystallization (albeit weighted
by growth rate and by the parent concentration in different zones). Bulk sample crystallization
ages in these kinds of relatively simple scenarios can be interpreted in terms of different geologic
processes such as crystal and magma residence times or differentiation ages.
For example, whether interpreted as a ‘burst’ of crystallization of short duration or as an
average age of progressive crystal growth, if the crystals precipitated in equilibrium with the
host liquid (i.e., autocrysts), the U-series crystal age can provide a minimum residence time for
the host magma (e.g., Reagan et al. 1992; Schaefer et al. 1993, Heath et al. 1998; Bourdon et al.
2000; Cooper et al. 2001; Heumann et al. 2002; Tepley et al. 2006). For example, in the case of
the 1952 eruption of Kilauea, the case for interpreting plagioclase and pyroxene ages in terms
of minimum magma residence time is bolstered by (1) concordant 230Th-226Ra model ages
for pyroxene and plagioclase (after correction for the effects of Ra-Ba fractionation) because
the two phases are expected to co-precipitate in evolved Hawaiian magmas, (2) concordant
238
U-230Th and 230Th-226Ra ages for each of the bulk separates, and (3) evidence for major-
element chemical equilibrium between crystals and liquid (Cooper et al. 2001). Mixed crystal
ages have been observed in bulk separates of a trace phase as well: Charlier and zellmer (2000)
measured 238U-230Th ages of bulk separates of zircon of different sizes, finding systematically
older ages for larger size fractions. They interpreted this observation to reflect episodic or
continuous crystallization, emphasizing that the age represents an average crystallization age
rather than the onset of crystallization.
Bulk crystal ages for crystals with relatively simple histories can also be interpreted in
terms of differentiation ages—i.e. the average crystallization age records the age of a dominant
interval of crystallization/differentiation. For example, Turner et al. (2003b) measured
238
U-230Th-226Ra disequilibria in whole-rock samples and mineral separates from lavas and
cogenetic cumulate xenoliths collected from Sangeang Api, Indonesia. Ra-Th ages (corrected
for the effects of Ra-Ba fractionation) are equivalent for cumulates and minerals separated
from the lavas, which Turner et al. (2003b) interpret to reflect the timescale (~2000 yr) of the
dominant crystallization/differentiation interval.
Crystal recycling—autocrysts vs. xenocrysts. vs. antecrysts
With increasing evidence for complex interactions between different magmas, or between
magmas and crystal mush or solidified melts from a prior injection of magma in a given system
(for example, see other chapters in this volume: Bindeman 2008; Costa 2008; Kent 2008; Streck
2008; Ramos and Tepley 2008), it is possible—or even likely—that simple crystallization from
a single magma, as described in the previous section, is the exception rather than the rule.
Especially in the case of bulk mineral separates where ages are averaged over many hundreds or
thousands of crystals, crystal ages may reflect contributions from crystals with different origins
ranging from phenocrysts sensu stricto (autocrysts) to antecrysts scavenged from remnants of
earlier phases of the magmatic activity at the same center, to true xenocrysts. Although the
diversity of origins of minerals within bulk separates may complicate interpretation of U-series
data in terms of ages, the data can provide unique constraints on the origin and age of different
components contributing to the crystal population in a given magma.
504 Cooper & Reid
Xenocryst identification by in situ dating. An outstanding feature of in situ mineral
dating is its ability to isolate contributions from xenocrysts—or xenocrystic cores—from those
of autocrysts. Restitic or recycled crystals could skew the results of age dating of mineral
separates. Potential contaminants acquired during sample or in laboratory preparations,
while not strictly xenocrystic in the normal sense, might also be identified. The presence
of xenocrysts could also signal a role for crustal melting and/or assimilation and can be a
means of fingerprinting the sources of those inputs. For example, a potential contribution
from greywacke sources to the generation of the Taupo volcanics, New zealand, was provided
by discovery of Cretaceous-aged zircons (Charlier et al. 2005). Alternatively, if independent
evidence for assimilation exists, the absence of xenocrysts could provide insights into the
conditions of magma evolution by requiring that xenocrysts are either completely resorbed
during melting and/or are very effectively isolated from the melt.
When considered in total, a notable conclusion from the work to date is that zircons that
predate the onset of a given center’s magmatic activity are remarkably rare in volcanic rocks.
Less than 1% (10 of >200 zircons) analyzed in nine units erupted over the past 45 k.y. from
Taupo volcano, New zealand, were found to be older than quaternary in age (Charlier et al.
2005). Brown and Smith (2004), in particular, targeted their efforts at recognizing xenocrystic
material in the 1.21 Ma Ongatiti ignimbrite of the Taupo Volcanic zone. Cathodoluminescence
imaging revealed only four resorbed cores—out of 300 zircons studied—and these yielded
119 to 340 Ma U-Pb ages that link the xenocrysts to Mesozoic and Paleozoic metasedimentary
rocks (Fig. 9). Young, 15-30-µm-thick rims on those zircons are inferred to represent growth
in the thousands of years before eruption, indicating that the cores are not simply accidental
material eroded from the conduit during or immediately prior to eruption (Brown and Smith
2004). Bindeman et al. (2006) noted that, of more than 200 individual zircons analyzed in
studies of Yellowstone volcanism (Bindeman et al. 2001; Vazquez and Reid 2002), only three
Cretaceous and Triassic—and no Archean basement—xenocrysts were found. Similarly, fewer
than 2% of the >120 Long Valley zircons studied so far (Reid and Coath 2000; Simon and Reid
2005; Simon et al. 2007) are xenocrysts and these are from underlying Mesozoic crust. Many
other studies report no xenocrysts (e.g., Lowenstern et al. 2000; Charlier et al. 2003; Bacon
and Lowenstern 2005; Bachmann et al. 2007).
Perhaps less surprisingly, settings in which silicic magmas are normally in the minority may
have relatively few xenocrysts. Lowenstern et al. (2006) found a single Pan-African basement
zircon amongst the otherwise young zircons that characterize granophyre xenoliths entrained
in rift-related basalts from the Alid Volcanic Center, Eritrea (Fig. 10). Schmitt and Vazquez
(2006) dated zircons in rhyolites and magmatic xenoliths associated with continental rupture
and development of nascent oceanic crust in the Salton Trough area of southern California.
Two of 44 zircons studied yielded Proterozoic ages; a single xenocryst of Jurassic age has also
been found (Fig. 9; Brown et al. 2004). The Proterozoic zircons are interpreted to be detrital
grains incorporated during the eruption since visible overgrowths on these zircons are absent
and sedimentary fragments in the lavas are commonly smeared out and intermingled with the
obsidian. Pilot et al. (1998) reported 330 and 1,600 Ma old zircons from gabbros exposed near
the Kane fracture zone of the Mid-Atlantic Ridge. In contrast, Moeller et al. (2006) obtained
U-Pb dates on zircons found within polished petrographic thin sections of gabbroic rocks from
similar samples from the same locations. They found only ~1 Ma and younger zircons and
attributed the previously reported Paleozoic-Proterozoic ages to contamination during sample
preparation.
Only a few studies report substantial evidence for xenocrysts. More than two thirds
(fourteen of 20) of analyzed zircons from the ~13 k.y. lower Laacher See Tuff yielded pre-
quaternary ages ranging in age from ~390 Ma to 2.7 Ga (Schmitt 2006). While some are
magma-hosted xenocrysts, most lack adhering glass and are inferred to be crystals derived
Uranium-series Crystal Ages 505
Figure 9. Cathodoluminescence images of selected Ongatiti zircon grains showing spot ages for xenocrystic
cores surrounded by magmatic-aged rims. Black ellipses show approximate analysis areas. [Used with
permission of Elsevier from Brown and Smith (2004) Journal of Volcanology and Geothermal Research,
Vol. 135, p. 252.]
from disseminated country rock fragments. Three of 7 zircon grains dated from the 7 k.y. old
rhyolite eruption of Cotopaxi volcano, Ecuador, gave Proterozoic ages (0.8-1.2 Ga; Garrison
et al. 2006). In that case, younger ages on the remaining zircons (18-30 Ma) are inferred to
represent hybrid grains wherein young rims mantle old cores. Five of 14 zircons in the ~36
ka La Virgen tephra from Las Tres Vírgenes composite cone in Baja, California, yielded ages
>25 Ma (Schmitt et al. 2006). In contrast, none of the 30 ignimbrite zircons from two adjacent
calderas yields an age that deviates from mean zircon ages of <1.4 Ma.
In summary, while it is difficult to generalize, it nonetheless appears that xenocrystic
zircon, and especially xenocryst-cored zircons, are an uncommon feature of many volcanic
rocks.
Autocrysts versus antecrysts: in situ dating. Accessory phase dating of young plutonic
inlusions shows that individual intrusions can have a range of crystallization ages. xenocrysts
notwithstanding, the age distributions may be continuous (e.g., Charlier et al. 2003; Bachmann
et al. 2007) or discontinuous (e.g., Lowenstern et al. 2000; Vazquez et al. 2007) but generally
overlap the duration of magmatic activity at a given volcanic center. As with xenoliths, volcanic
506 Cooper & Reid
Figure 10. Cathodoluminescence images for zircons from a granitic xenolith from a 15.2 rhyolite pumice
deposit at Alid volcanic center, Eritrea. Note the generally euhedral grain shapes and the growth-related
oscillatory zoning represented by the dark and light bands throughout some grains (e.g., A) or mantling
rounded cores (e.g., C). White and black ellipses show approximate analysis areas. [Used with permission
of Oxford University Press from Lowenstern et al. (2006) Journal of Petrology, Vol. 47, p. 2114.]
rocks exhibit a variety of age distributions, from unimodal to multimodal. In some instances,
the ages correspond to known periods of earlier activity (Bacon and Lowenstern 2005).
Disequilibrium mineral assemblages (e.g., Nakada et al. 1994; Gardner et al. 2002; Bachmann
et al. 2002; Schwartz et al. 2005) have been used to suggest that so-called “phenocrysts” are
actually antecrysts, crystals disaggregated or otherwise remobilized from young crystal-rich
domains, whether from earlier crystallized magmas or cumulate mush at the margins of magma
bodies (e.g., Mahood 1990; Nakada et al. 1994; Harford and Sparks 2001). Recycling may occur
by assimilation or rejuvenation of these subvolcanic intrusions. While age distributions alone
are usually insufficient to distinguish antecrysts from phenocrysts (other relevant observations
are described in the next section), they are often key arbiters of such a distinction.
One line of evidence for the presence of antecrysts has been identification of U-Th
isotopic affinities between melt- and xenolith-hosted zircons. Ca. 27 ka rhyodacite tephra
and lavas, interpreted to be early leaks from the magma chamber responsible for the climatic
eruption of Mount Mazama, yield an approximately unimodal zircon age distribution that
peaks at 45 k.y. (~20 k.y. before eruption; Fig. 11) and matches the zircon age distribution of
a granodiorite block (Fig. 12; Bacon and Lowenstern 2005). Tailing of the age distribution to
>200 ka and a possible clustering of ages around ~110 k.y. show that older zircons are also
present. Exclusive of xenocrysts, U-Th isotope characteristics of zircons in the Lower Laacher
See phonolite, Germany, are isochronous with those of syenite xenoliths and define a single
age population that crystallized within ~5 k.y. of eruption (Fig. 12; Schmitt 2006). The older
peak in a bimodal zircon age distribution for a Salton Sea obsidian matches a discrete ∼18
ka zircon population in a felsite xenolith, suggesting episodic growth of zircon and possible
recycling of pre-existing zircon-bearing rocks (Schmitt and Vazquez 2006).
Uranium-series Crystal Ages 507
Figure 11. Cumulative age probability curves and, for most, histograms for U-Pb or U-Th age distributions
obtained by SIMS dating of domains within zircons. Distributions in Figures 11A and 11E are plotted with
respect to U-Th isochron slopes; corresponding ages are shown by dashed vertical lines. Age distributions
are unimodal, bimodal, and multimodal. A) U-Th age distribution for zircons from an ~27 ka Crater Lake
rhyodacite (Bacon and Lowenstern 2005); B) U-Pb age distribution for zircon rims from the ~600 ka
Devils Kitchen rhyolite (Miller and Wooden 2004); C) U-Pb age distribution for zircons from the ~340 ka
Whakamaru Ignimbrite (from Charlier et al. 2005 based on data for Brown and Fletcher 1999); D) U-Th age
distribution for zircons from the ~115 ka Deer Mountain rhyolite (from Charlier et al. 2005 based on data
for Reid et al. 1997); and E) U-Th age distributions in various Taupo rhyolites (eruption ages shown by dark
vertical bar; from Charlier et al. 2005). Also shown for reference by vertical white bars are TIMS ages.
508 Cooper & Reid
Figure 12. A) Histograms and cumulative age probability curves for slopes of 238U–230Th isochrons for
SIMS zircon data from 4 granodiorite samples (heavy lines) and a rhyodacite (thin line) from Mt. Mazama.
Vertical dashed lines are reference ages. ENE, etc., in parentheses give approximate azimuths of collection
sites relative to center of Crater Lake caldera. B) (230Th)/(232Th) versus (238U)/(232Th) diagram for Laacher
See zircons including best-fit ages and initial (230Th)/( 232Th)0 values. A xenocryst (excluded from the
regression) and another spot were omitted from this plot for clarity. Inset shows close-up with bulk glass
separate data for Laacher See phonolitic pumice (Bourdon et al. 1994), recalculated ages, and recalculated
(230Th)/(232Th)0. LLST—Lower Laacher See tephra. MSWD—mean square of weighted deviates. Modified
after Bacon and Lowenstern (2005) and Schmitt (2006).
Uranium-series Crystal Ages 509
Three of the younger and more voluminous Central Plateau Member lavas of Yellowstone
caldera, Wyoming, have nearly identical unimodal zircon age populations even though they
erupted at times separated by as much as ~50 k.y. (from 160 to 110 ka). Vazquez and Reid (2002)
conclude that these lavas tapped a new magma reservoir that differentiated approximately 40
k.y. before the first eruption. In contrast, over half of the zircons in the somewhat older (0.4-0.5
Ma) post-caldera rhyolites are similar to those in older Yellowstone volcanic rocks and could
represent crystals recycled from earlier volcanic and/or plutonic rocks (Bindeman et al. 2001).
zircons in the pre-, syn-, and postcaldera eruptions at Ammonia Tanks, Nevada, are similarly
interpreted to be recycled from older eruptive cycles (Bindeman et al. 2006).
Antecrysts have also been identified on the basis of considering zircon age distributions in
the context of textural observations. Based on the crystal-rich nature of and presence of crystal
clots within the ~10 km3 Earthquake Flat rhyolite, New zealand, a unimodal zircon population
that crystallized ~60 k.y. before eruption is thought to represent antecrysts remobilized not
long before eruption from a largely frozen crystalline pluton (Charlier et al. 2003). Similarly,
textural characteristics of the 160 ka Kos Plateau Tuff, Greece, were taken to indicate that a
broadly unimodal zircon age population that peaks at ~230 ka (70 k.y. before eruption) but tails
to significantly older ages (~500 ka; Bachmann et al. 2007) reflects rejuvenation of crystal mush
zircons (Fig. 13). The general concordance in age between these zircons and pumice-hosted
granitic clasts led Bachmann et al. (2007) to further suggest that the latter were excavated either
from the marginal “rind” of a magma chamber or from co-genetic plutonic rocks.
Multimodal age distributions have also been taken as evidence that crystal populations
represent mixtures of antecrysts and autocrysts. Even though U-Pb ages for nearly half of
dated zircon cores and rims from the 587 ka (K-Ar age) Devils Kitchen rhyolite of the Coso
Volcanic Field, California, cluster around 600 ka, the rim ages span a minimum of 200 k.y.
(Fig. 11; Miller and Wooden 2004). Peaks in the zircon age distributions are present at 620
and 670 ka, as are subordinate peaks at 730 and 800 ka. Miller and Wooden infer that zircons
with relatively old rims (>700 ka) were probably recycled from earlier intrusions whereas the
~670 ka cores that are rimmed by younger growth probably represent zircons incorporated into
younger, zircon-saturated melts. zircon age distributions for the various dacites, rhyodacites
and rhyolites associated with the Taupo Volcanic Field, New zealand, range from unimodal to
multimodal (Fig. 11; Brown and Fletcher 1999; Charlier et al. 2003; Charlier et al. 2005). Core
to rim age variations in excess of 250 k.y.in some Whakamaru zircons were attributed by Brown
and Fletcher (1999) to long magma residence times, with a role for crystal recycling. Charlier
et al. (2005) noted the clustering of ages in the Whakamaru results and in the Taupo volcanics
they studied and attributed the episodic nature of zircon crystallization to remobilization of
zircons from mush instead of prolonged residence in a liquid-dominated magma body. In
particular, a prominent cluster of ~100 ka model ages in the pre-26.5 ka Oruanui eruptives
corresponds to an earlier phase of modest volume dome-building and is interpreted to reflect
crystals derived from partly to wholly solidified plutonic bodies.
Complex crystal populations and bulk mineral ages. The presence of crystals with
multiple origins within a bulk mineral separate precludes a simple interpretation of the bulk
crystal age in terms of magma residence or average crystal growth rate. However, U-series ages
can provide unique insights about the origin and especially the age of different components.
U-series age data are particularly useful when combined with other types of chemical or
textural information about the crystals, which can in turn provide independent information
about the origin and abundance of different components. Measuring ages of mineral separates
may be one of the only ways to recognize antecrysts in the case of a magmatic system with
limited compositional range. The isotopic compositions of U-series nuclides and, in some
cases, the crystal ages themselves can be used as tracers of different populations within a
given magmatic system. The effect on U-series ages of mixtures of different minerals of the
510 Cooper & Reid
A. Granitic
Inclusions
Figure 13. Isochron slopes and
age distributions of zircons from
Pumices Kos Plateau Tuff, Greece, show-
ing presence of continuous age
distributions. A) Model isochron
slopes and corresponding 238U-
230
Th age distributions for grani-
toid xenoliths (top) and pumices
(bottom). B) 238U-206Pb age dis-
tributions for granitoid xenoliths.
Weighted U-Th mean ages are
228±14 ka (1σ; n = 70; MSWD
= 1.08) and 194±10 ka (1σ; n =
74; MSWD = 1.6) for pumice
and granitic clasts, respectively.
Weighted U-Pb age is 213±5 (2σ;
Isochron Slope n = 33; MSWD = 17.8) for pum-
ice and clasts combined. Eruption
age of 161 ± 1 ka shown for refer-
B. 35
Eruption age ence. Based on Bachmann et al.
161 ± 1 ka (2007).
30
20 213 ± 5 (2σ) ka
15 MSWD = 17.8
10
5 Pumices ( )
Inclusions ( )
0 Errors are 2σ
0 100 200 300 400 500 600
Age (k.y.)
same age, and of progressive crystal growth for a given mineral from the same magma, was
discussed earlier; here, we address somewhat more complex scenarios of mixing of minerals
and the effects on ages calculated for bulk separates.
Where multiple mineral phases define an isochron on a 238U-230Th isochron diagram, it is
generally difficult to maintain a straight line array by mixing of different mineral populations.
For example, Heath et al. (1998) present several scenarios of mixing between phyric magmas,
between phyric and aphyric magmas, and between a cumulate and a magma (Fig. 14). In
general, producing a linear array with appropriate slope requires a fortuitous balance between
the age, proportion, and initial 230Th/232Th of each component, and where multiple phases are
Uranium-series Crystal Ages 511
present mixing of mineral populations
is unlikely to explain a linear array.
However, in cases where a mineral array
can be dominated by one component,
such a model could be appropriate, and
in some cases the combination of 230Th-
226
Ra and 238U-230Th data are easiest to
explain by mixing scenarios (see below).
In the case of accessory phases that
are inclusions in major phases, the U-Th
budget of the mineral separate may be
so dominated by even small amounts of
trace phases that the bulk mineral separate
age essentially reflects accessory phase
crystallization (e.g., Heumann et al. 2002;
Charlier et al. 2003). However, it is still
possible to derive some age information
from the bulk separate in some of these
cases. For example, plagioclase separates
from rhyolite pumice and dome samples
of the 1305 AD Kaharoa eruption of
Tarawera Volcano, New zealand, plot on
a zero-age isochron but to the right of the
equiline on a 238U-230Th isochron diagram,
which is interpreted to reflect the presence
of young zircon inclusions within the
bulk separate (observed in thin section);
nevertheless, the plagioclase must still be
contemporaneous with or younger than
the zircon inclusions, requiring a young
age for plagioclase as well (Klemetti and
Cooper 2007).
In the case of 230Th-226Ra ages,
mixtures of different minerals may have
more profound effects on calculated ages
than in the case of 238U-230Th ages. As
discussed earlier, the 230Th-226Ra data for Figure 14. Schematic 238U-230Th isochron diagrams
bulk mineral separates must be corrected illustrating consequences of mixing different crystal
for the presence of other phases in order populations in several scenarios. Model 1: mixing
of two populations of zero-age minerals crystallized
to allow calculation of model ages. It from different magmas. In this case, it would require
is still possible, however, that even a a fortuitous combination of proportions of mixing for
mineral separate containing only a single each mineral phase to produce an inclined linear array.
phase may represent a mixture of crystals Model 2: Mixing of an aphyric lava with old crystals;
of different ages and origins. The extent no isochron will be produced. Model 3: Mixing of
zero-age crystals with older crystals formed from melt
to which model ages calculated from of similar composition. Like Model 1, this requires
such mixed populations are meaningful a fortuitous proportion of old and young crystals to
depends on how much factors such as produce a linear array. Model 4: mixing of a zero-
composition (in particular, (226Ra)/[Ba] age lava with old cumulate minerals via overgrowths
and (230Th)/[Ba]) of the parent liquid containing melt inclusions. In this case the apparent
age is meaningless. [Used with permission of Elsevier
and the crystallization temperature may from Heath et al. (1998) Earth and Planetary Science
have varied between the different crystal Letters, Vol. 160, p. 58.]
512 Cooper & Reid
populations. In the case where crystals of different ages precipitate from the same magma,
the model age will represent the average age of crystallization. In somewhat more complex
cases where there is evidence for contributions of crystals from multiple magmas, it can still be
possible to calculate average ages by assuming a limited range for (226Ra)/[Ba] in the liquids
from which the minerals crystallized; this assumption can be tested in cases where 238U-230Th-
226
Ra disequilibria in multiple lavas of diverse ages have been analyzed (although there is always
some degree of uncertainty about whether erupted lavas are representative of the diversity of
compositions sub-surface). However, model age calculations in many cases may be relatively
robust to assumptions about liquid (226Ra)/[Ba], within the range measured for a given volcanic
system (e.g., Cooper and Reid 2003; Tepley et al. 2006). Overall, model ages that represent
averages of mixed crystal populations can still be estimated, albeit with somewhat larger
uncertainties due to the uncertainty in (226Ra)/[Ba] of the magma(s) from which the minerals
crystallized.
Xenocrysts in bulk mineral separates. A unique feature of U-series ages compared to
other geochronologic techniques is that once a given parent-daughter pair has attained secular
equilibrium, the ratio of parent to daughter remains constant. Therefore, very old crystals
have no more leverage on the average age of a bulk population than do crystals that have just
recently attained secular equilibrium. This is in contrast to systems where the parent/daughter
ratio continually decreases over time (e.g., U-Pb), and it means that xenocrysts with ages of
millions or hundreds of millions of years have relatively little influence on the average age of
a mineral separate. However, it also means that other techniques (e.g., U-Pb dating or crystal-
scale measurements of long-lived isotopic tracers such as Sr or Nd) must be used in order to
definitively distinguish xenocrysts from antecrysts.
Discordant ages and antecrysts in bulk mineral separates. In the simple scenario of
crystallization over a short duration followed by closed system aging, dates of the same samples
using different parent-daughter pairs should yield the same age. Relatively few studies exist
where multiple parent-daughter pairs have been measured for the same samples, although the
data set is growing. In many cases where 238U-230Th and 230Th-226Ra ages have been measured
on the same mineral separates, the two systems yield different apparent ages (e.g., Volpe and
Hammond 1991; Volpe 1992; Cooper and Reid 2003; Turner et al. 2003a). Furthermore, in all
cases where 226Ra-210Pb disequilibria have been measured on plagioclase separates, measurable
disequilibrium yields model crystallization ages of decades (Reagan et al. 2006, 2008), even
though 238U-230Th and 230Th-226Ra model ages for plagioclase in the same samples are thousands
to tens of thousands of years old (Tepley et al. 2006; Cooper and Donnelly 2008). Similarly,
anorthoclase separates from Mt. Erebus show 228Th-232Th disequilibria, indicating that some
crystal growth occurred within 30 yr of eruption (Reagan et al. 1992), despite 230Th-226Ra ages
of ~2500 years (note that although these isochron ages were not corrected for the effects of Ra-
Ba fractionation, and therefore the true ages are likely somewhat younger than the quoted ages,
the crystal ages are not likely to be within error of eruption age).
Although some cases of discordant ages may be explained as mixing arrays with old ap-
parent ages, a more general explanation is that the discordant ages reflect protracted or episodic
crystal growth spanning thousands to tens of thousands of years, where the different parent-
daughter pairs will record different average ages for the same proportion of old and young
material (Cooper and Reid 2003). An effect of the unchanging parent/daughter ratio at secular
equilibrium is that all crystals in secular equilibrium, whatever their true age, appear to be the
same ‘age’, as determined by the time required for a given parent-daughter pair to reach secular
equilibrium—approximately 350 k.y. for 238U-230Th, 10 k.y. for 230Th-226Ra , and 100 yr for
226
Ra-210Pb. One implication of this effect is that the average age of a crystal population is al-
ways biased toward young crystal growth, but that the apparent age of a mixture will vary with
the age of the old component and between different parent-daughter pairs. Mixtures of crystals
Uranium-series Crystal Ages 513
where the old component is not in secular equilibrium for either parent-daughter pair would be
expected to produce concordant apparent ages (unless the concentration of parent varies in a
different way for each parent-daughter pair, or the liquid from which the minerals crystallized
had a different composition, between the two components). If the older component is in secular
equilibrium for the shorter-lived daughter, discordant apparent ages between different parent-
daughter pairs are an expected result of mixing crystal populations of different ages (Fig. 15).
106
apparent age (y)
A
70 75
5
10 50
30
20
10
104
Ra-Th secular equilibrium
5
2.5
1
103 0k
a
10 ka
230Th-238U
50 ka
30
102 ka
10 a
5 k
10
10 102 103 104 105
226Ra-230Th apparent age (y)
104
apparent age (y)
B
90 1000 y
70
103 50 750 y
30 500 y
20 250 y
10 100 y
102 5 50 y
226Ra-230Th
10
1
1 10 102 103
210Pb-226Ra apparent age (y)
Figure 15. Apparent ages produced by mixing old and young plagioclase components in a bulk mineral
separate. A) 230Th-226Ra apparent age vs. 230Th-238U apparent age (modified from Cooper and Reid 2003). B)
226
Ra-210Pb apparent age vs. 230Th-226Ra apparent age. In each panel, fine straight line highlights concordant
ages, and thick grey line represents secular equilibrium for the shorter-lived parent-daughter pair. Solid
curves are constructed by calculating apparent ages (with reference to the young liquid) of a mixed crystal
population, where the young component is assumed to be zero age and where the age of the old component
is varied to produce the different solid curves on the diagrams. Dashed lines represent contours of a constant
percentage of old plagioclase, where the percentage is labeled. The scenario modeled is that some plagioclase
crystals (‘old’ crystals) precipitated from a melt, aged, and then were incorporated into a new liquid with
similar composition to the original magma, followed by a second phase of crystallization that occurred in
equilibrium with the new magma immediately prior to eruption and analysis. Both magmas are assumed to
have (230Th)/(238U) = 1.10, (226Ra)/(230Th) = 1.30, and (210Pb)/(226Ra) = 1.0 at the time of crystal growth and
the same concentrations of U, Th, Ra, Ba, and 210Pb, leading to the equal concentrations of U, Th, and Ba in
old and young plagioclase. Abundances of Ra and 210Pb, and 230Th/232Th of old plagioclase are calculated by
Cooper&Reid Figure 15
accounting for U, Th, Ra, and 210Pb decay since the time of crystallization.
514 Cooper & Reid
There is a tradeoff between the age of the old component and the percentage of old mate-
rial that can account for any given combination of two apparent ages (Fig. 15). However, if
independent evidence of the proportion of old and young crystal growth exists (e.g., textural
information or other chemical evidence fingerprinting different components), the age of the
older component can be estimated. In the case where 238U-230Th, 230Th-226Ra , and 226Ra-210Pb
data all exist for the same samples, the combination of apparent ages for the three systems may
be enough to uniquely estimate the age and proportion of old and young components (although
it should be noted that this would not be possible in the case of more than two growth episodes).
Thus there is additional information to be gained from analysis of multiple parent-daughter
pairs in the same samples.
For example, discordant 238U-230Th and 230Th-226Ra ages in plagioclase have been measured
in two samples of the ~2000-ybp Castle Creek eruptive period at Mount St. Helens (Fig. 16;
Volpe and Hammond 1991; Cooper and Reid 2003). Textural evidence for disequilibrium
between plagioclase and liquids, combined with major-element and trace-element zoning in
plagioclase beyond that expected for crystallization from a single magma, suggest that these
discordant ages reflect mixing of multiple crystal populations where the older component is
tens of thousands of years old and makes up on the order of 40-50% of the bulk mineral separate
(Cooper and Reid 2003). It is interesting in this context that the four other Mount St. Helens
samples analyzed in the same study (with eruption ages between 2000 ybp and 1982) also show
textural and trace-element evidence for mixed crystal populations, but do not have discordant
238
U-230Th and 230Th-226Ra ages (Cooper and Reid 2003). One possible interpretation is that the
older component in the case of the four younger samples is not old enough to create discordant
ages. Furthermore, all of the average plagioclase 230Th-226Ra model ages overlap, regardless
of whether the Ra-Th and U-Th ages are discordant, indicating that some crystal component in
each magma must have been present beneath the volcano at the same time (Fig. 16).
Another example of discordant ages is seen in samples from Tonga and from Soufriere,
Lesser Antilles, where 238U-230Th plagioclase ages are 30-50 ka whereas the same mineral
separates show 230Th-226Ra disequilibria (Fig. 17; Turner et al. 2003a). An initial 238U-230Th
study of minerals from Soufriere lavas and cumulates concluded that, whereas they could not
rule out mixing producing the linear array of 238U-230Th data that they measured in samples
from Soufriere, they preferred a model where the crystals represented phenocrysts precipitated
from the magma (i.e., autocrysts) and therefore magma residence times on the order of tens
of thousands of years (Heath et al. 1998). Turner et al. (2003a) present additional analyses of
samples from Soufriere, including 230Th-226Ra data for minerals separated from a cumulate
xenolith. They found that the cumulate minerals defined a linear array on a 238U-230Th isochron
with the same apparent age as the earlier samples (40-50 ka), yet all of the mineral separates
preserved 230Th-226Ra disequilibria. Based on these discordant ages and on crystal size
distribution (CSD) evidence for complex crystal populations, Turner et al. (2003a) concluded
that the data best fit a model where the cumulate minerals have old cores (tens of thousands
of years) with young overgrowths produced during entrainment in a new magma. Turner et
al. (2003a) also present 238U-230Th-226Ra data for Tonga lavas and one cumulate xenolith, in
addition to crystal size distribution data which shows evidence for multiple crystal populations,
and they interpret the combined data set to reflect storage of plagioclase as cumulates, which
were remobilized and incorporated into subsequent magmas passing through the same system.
In addition, Turner et al. (2003a) also present calculations of apparent ages of mixtures, finding
that discordant apparent ages are expected in the case of old cores mixed with young rims.
A similar pattern of discordant data is also shown in data for the 2004-2008 eruption of
Mount St. Helens (Cooper and Donnelly 2008). Whole-rock samples of dacite erupted from
October 2004-April 2005 overlap with samples erupted in the 1980’s on a 238U-230Th isochron
diagram. However, plagioclase separates in two samples erupted only a month apart (October
2004 and November 2004) have 230Th/232Th ratios that are significantly different from each other
Uranium-series Crystal Ages 515
1.40
A 6±4 ka
1.35
90-9 90-7
(230Th)/(232Th)
1.30 2±15 ka 90-3 90-5
90-8 90-2
ka
1.25
2
±1
4±13 ka gm wr px pl ol mt
27
1.20
a
6k
4±3 ka
±1
34
1.15
1.10
0.5 1.0 1.5 2.0 2.5 3.0 3.5
(238U)/(232Th)
B
90-9
90-3
90-8
90-7
90-5
90-2
0 1000 2000 3000 4000 5000 6000 7000
years before present
Figure 16. Comparison of 238U-230Th isochron ages and fractionation-corrected 230Th-226Ra ages for
minerals separated from Mount St. Helens samples (modified after Cooper and Reid 2003). A) 238U-230Th
isochron; data from Volpe and Hammond (1991). Apparent ages are indicated by numbers along dashed
isochron arrays. Abbreviations as follows: groundmass (gm), whole rock (wr), pyroxene (px), plagioclase
(pl), olivine (ol), magnetite (mt). Solid line represents the equiline. B) Summary of fractionation-corrected
226
Ra -230Th ages for plagioclase (solid lines) and pyroxene (dashed lines). Filled diamonds represent “best-
estimate” plagioclase model ages using mean plagioclase compositions and pre-eruptive temperatures
based on bulk composition. Lines indicate range of plagioclase model ages allowing for variations in An
(± An10) and temperature (900-1000 °C for dacites, 1000-1100 °C for andesites, and 1000-1200 °C for
the basalt). Arrow pointing to right for MSH 90-3 plagioclase age indicates that the upper limit is >8 ka.
Range of pyroxene model ages are shown by fine dashed lines; this range primarily
Cooper&Reid, reflects
Figure 16 propagation
of uncertainties through mass-balance calculations of Ra in pure pyroxene; uncertainties in model DRa/
DBa have a very minor effect. Grey circles represent eruption ages for each sample (sample numbers along
vertical axis). Note that apparent U-Th ages are within error of Ra-Th ages for all samples except 90-5 and
90-2, likely indicating an old crystal component in these two samples. See text for discussion.
and from plagioclase in a sample of 1982 dacite (Volpe and Hammond 1991). The apparent
whole-rock/plagioclase age for one sample is ~40 ka, yet plagioclase in this sample also has
significant 230Th-226Ra disequilibria (Cooper and Donnelly 2008) and 226Ra-210Pb disequilibria
(Reagan et al. 2008). These data are most consistent with mixing of plagioclase with diverse
230
Th/232Th ratios and different ages, ranging from tens of thousands of years to a few decades
prior to eruption.
The presence of 226Ra-210Pb disequilibria or 228Th-232Th disequilibria in other feldspar
separates with 230Th-226Ra apparent ages of thousands of years (Reagan et al. 1992, 2006,
516 Cooper & Reid
284 S. Turner et al. / Earth and Planetary Science Letters 21
2008) suggests that a very young crystal Fig. 1 and d
component, perhaps reflecting crystalli- magnitude o
zation during ascent and eruption, may be
growth rates
ubiquitous in volcanic rocks but is rarely
which are n
recognized when using other mineral dat-
ing techniques. estimates [1
end. This do
Overall, there are many cases where taken for su
discordant ages and/or other textural and the 10s of ky
chemical information about the crystal over, if cryst
populations indicate recycling of crystals
nously with
within the same magmatic system. Espe-
tal settling is
cially in cases where there is limited com-
few hundred
positional variability between the erupted
products (e.g., Mount St. Helens 2004- pally on the
2008 eruption compared to the 1980-86 temperature
eruption), age information may be the Late and M
most effective way not only to recognize sities and vis
crystal recycling, but also to quantify 842 and 2599
the proportion and ages of various com- sized crystal
ponents. Collectively, these cases where km3 magma
discordant ages are observed suggest that tively [26]. W
by combining age information with tex- ring around
tural and crystal-chemical information it accumulation
may be possible to reconstruct a signifi- es such as
cant part of a volcano’s history from the could lead to
crystal record. Thus, since c
Insights from dating of subvolcanic to be much q
plutons by the appar
Many lavas and pyroclastic deposits cores resided
contain plutonic blocks that could rep- from Raoul,
resent magmas or crystal mush that so- magma, you
lidified in subvolcanic domains or, in the entrained the
case of anatexis, could be the remnants of duced overgr
magma sources. Dating of these blocks is cess from th
a means of distinguishing between these In this mo
possibilities and in the case of the former, way re£ect
providing insights into the emplacement growth, crys
and growth of subvolcanic intrusions. The storage as a
more complete volcano-plutonic record entrainment
provided by considering the crystalliza- to eruption a
tion history of plutonic as well as volcanic Fig. 2. (a) U^Th equiline diagram for Late and238Metis Shoal. the cores gre
Figure 17. Diagrams showing discordant U-230Th
rocks can better illuminate the dynamical (b) 226 Ra/Ba versus time evolution diagram
230 226 for Late ground-
(A) and fractionation-corrected Th- Ra ages (B, erupted grou
evolution of a volcanic system. massC) and
forcalculated liquid
samples from in equilibrium
Tonga. [Used withwith the plagioclase
permission of tained from
226
separate. (c) from
Elsevier Ra/Ba versus
Turner et al.time evolution
(2003) diagram
Earth and for Me-
Planetary
Plutonic zircons studied to date have tis Shoal
Sciencegroundmass
Letters, Vol.and
214, calculated
p. 284.] liquid in equilibrium
such as thos
come mainly from dioritic to granodiorit- with the plagioclase separate. Errors in panels b and c allow logical mean
ic blocks but zircons have been obtained for 15% uncertainty in the value of DRa /DBa for the minerals ence of core
from syenitic and gabbroic inclusions as welland (e.g., Schmitt and
the propagation Vazquez
of the 2006;
analytical Bacon
errors et al.
in the case of lines that m
the groundmass (dashed curves).
2007; Schmitt 2007). The presence of interstitial melt in plutonic blocks has been attributed to drawn betwe
partial melting (Schmitt and Vazquez 2006; Bachmann et al. 2007) and to partial solidification the equiline
(Charlier et al. 2003). Granophyric intergrowths of quartz and feldspar are common in grani-
Uranium-series Crystal Ages 517
toid xenoliths, as are intercrystalline voids and miarolitic cavities (Charlier et al. 2003; Low-
enstern et al. 2006; Schmitt 2006; Bacon et al. 2007). Small to millimeter-sized zircon crystals
may occur in and adjacent to the inclusion vesicles (Fig. 18). Micrometer-sized cavities and
tunnels in some zircons suggest phase changes (solidification or resorption) in the presence of
aqueous fluid (Bacon et al. 2007) while other zircons exhibit evidence of sub-solidus exsolu-
tion (Fig. 18; Schmitt 2006). Ages for zircon populations dominated by these crystals may be
biased towards the latest stages of crystallization.
Figure 18. Backscatter electron images of Lower Laacher See Tuff zircon (zrn) with A) ThSiO4 (Tho) rods
present versus B) homogeneous zircon. Dotted ellipses indicate U-Th ion microprobe spot locations. C)
Secondary electron image of vesicle-grown zircon in a syenitic subvolcanic nodule. [Used with permission
of the Geological Society of America from Schmitt (2006) Geology, Vol. 34, p. 598.]
zircons and major phases in nodules associated with dominantly mafic volcanic centers
provide little evidence for protracted crystallization histories. For example, well-resolved U–Th
zircon ages for Salton Buttes granophyre xenoliths indicate crystallization within a few 1000
yr (Schmitt and Vazquez 2006). Batches of dioritic magma may have evolved for ≤104 years
beneath Hawaiian volcanoes based on the similarity between core and rim ages on individual
crystals (Vazquez et al. 2007). Low-U zircons in the Alid, Eritrea, granophyre may have grown
from less differentiated magma <15,000 years prior to final solidification (Lowenstern et al.
2006). Differences between ages of growth zones or between individual grains in individual
nodules are not statistically significant in granitoid xenoliths from Medicine Lake, California
(Lowenstern et al. 2000), except for a single Medicine Lake crystal with a core that crystallized
before all others and ~50 k.y. before its rim. The same generalities apply to Crater Lake
xenoliths (Bacon and Lowenstern 2005) where multiple analyses of zircon grains mostly yield
similar ages on the same grains. Some of the granodiorite zircons yield younger ages at their
ends or margins than nearer to or in the center and one grain with a core in U-series secular
equilibrium has multiple younger ages towards rims. U-Th-Ra dating of minerals in cumulate
mafic and ultramafic xenoliths entrained in potassic lavas from Sangeang Api, Indonesia,
yield ages that are both young (<2 ka) and concordant with mineral ages for one of the lavas,
suggesting that a dominant interval of differentiation produced subvolcanic cumulates which
were entrained in the magmas during eruption (Turner et al. 2003b). In addition, as discussed
518 Cooper & Reid
in a previous section, crystal size distributions together with discordant U-Th and Th-Ra ages
of minerals in cumulate xenoliths and lavas from Tonga and Soufriere suggest multiple stages
of crystal growth and storage as cumulate or plutonic bodies, followed by entrainment of
crystals and additional crystal growth within the host magma (Turner et al. 2003a).
Some comagmatic (sensu lato) blocks yield relatively simple zircon age distributions
(e.g., Alid granophyres, Eritrea; Lowenstern et al. 2006). Others, such as dioritic to leucocratic
xenoliths from Hawaii (Vazquez et al. 2007), yield bimodal age populations with peaks that
are separated by 60 to 200 k.y. zircons from four granodiorite blocks from Crater Lake yield a
broad range in crystallization ages (between ~20 ka and ~300 ka) that cluster at 50–70, ~110,
and ~200 ka (Bacon and Lowenstern 2005). The older core in the Medicine Lake granitoid
grain noted above may also be derived from an earlier intrusion. Thus, magmas that solidify
beneath young volcanic centers apparently recycle zircons from earlier intrusions.
Comparisons between accessory phase ages for volcanic rocks and the plutonic blocks
they host provide a means of linking volcano-plutonic records, as in the case of zircons in a
~27 ka Crater Lake rhyodacite lava that apparently crystallized at the same time as zircons in
an associated granodiorite block (Bacon and Lowenstern 2005). U-Th zircon ages corroborate
evidence from crystal sizes and cathodoluminescence patterns for a common zircon population
between an erupted Salton Sea rhyolite and a possible shallow intrusive (Schmitt and Vazquez
2006). The age concordance between zircons extracted from Kos, Greece, granitic clasts and
pumice zircons shows that the granitic inclusions were derived from co-genetic intrusive rocks
(Fig. 13; Bachmann et al. 2007). Relatively uniform U-Th isotope characteristics for zircon
crystals present in the vesicles of juvenile syenitic nodules from Laacher See cluster close to
the isochron defined by zircons from the Lower Laacher See tephra and similarly suggest a
cogenetic origin for plutonic and volcanic zircon populations (Fig. 12; Schmitt 2007).
In some cases, the timing of zircon crystallization in xenoliths correlates with episodes
of volcanic activity rather than with volcanic zircon populations. Petrographic, chemical, and
isotopic similarities between the Alid, Eritrea, granophyres and host tephras suggest that the
granophyres may be wall fragments from the shallow magma chamber responsible for the Alid
tephra. zircons apparently grew late in the crystallization history of granophyre blocks (and are
absent in the tephra) and virtually all crystallized within the same time period as doming and
volcanism at the Alid volcanic center (36–15 ka; Fig. 10). The timing of zircon crystallization
in Crater Lake granodiorite blocks also clusters at intervals that correspond to known periods of
dacitic volcanism (Bacon and Lowenstern 2005).
The more complete volcano-plutonic record provided by dating plutonic nodules can better
reveal the complexity of subvolcanic plumbing systems. For example, some of the glass-bearing
granitoid fragments from Rotoiti, New zealand, had been interpreted on mineralogical and
chemical grounds to be the incompletely crystallized portion of a magma system that erupted
>200 k.y. earlier (Charlier et al. 2003). Model 238U-230Th zircon age distributions and means
that are very similar to those found in a host ignimbrite pumice and by TIMS analyses of two
pumice zircon mineral separates, respectively, are inferred to contradict this. The ~55 ka age
yielded by another mineralogically distinct granitoid is essentially eruption-aged and dissimilar
to the main Rotoiti zircon population. Thus the Rotoiti granitoid clasts apparently crystallized
synchronously with the main Rotoiti magma body but in separate pockets of magma.
Before they were dated, melt-bearing granitic inclusions had been taken as evidence for
an anatectic origin for the Kos Plateau Tuff (Keller 1969). However, even the anhedral cores
observed in both pumice and clast zircons proved to be no older than 500 k.y., negating the pos-
sibility that the clasts are restitic from melting of pre-Kos granites (Fig. 13). In contrast, ∼30
and ∼9 ka zircons in inferred partial melt pockets in mafic xenoliths attest to melting and recy-
cling of preexisting basaltic crust associated with extension of the Salton Trough (Schmitt and
Vazquez 2006). Similarly, based on zircon dating of xenoliths, two brief but voluminous epi-
Uranium-series Crystal Ages 519
sodes of mafic volcanism at Medicine Lake volcano, California, at approximately 90 and 30 ka
were accompanied by intrusion of intermediate to silicic granitoids thought to represent melts
of young crust formed in association with the high-heat flux (Lowenstern et al. 2000). Lastly,
zircon ages for highly fractionated nodules provide evidence for a more complex interplay
between alkaline and subalkaline magmas beneath Hawaii than previously inferred (Vazquez
et al. 2007).
reSOLVInG aGeS FrOM COMPLex CryStaL POPuLatIOnS:
InCOrPOratInG Other OBSerVatIOnS
Introduction
Taken in isolation, it can be nearly impossible to insightfully interpret U-Pb or U-series
ages in terms of magma evolution because of the possible effects of crystal inheritance or
recycling and mixing of different crystal populations. In some instances, it may be possible to
resolve questions about the extent to which a particular date represents a single geologic event
simply by applying more than one age dating technique. Discordant ages from different parent-
daughter pairs, for example, can themselves be an indication that crystals of more than one
age are present in a bulk mineral separate (see previous section). More often, petrologic and
geochemical observations at both the rock and mineral scale will be informative. It is becoming
increasingly commonplace to couple trace element, oxygen isotope, and thermometry analysis
with in situ crystal ages. These help to subdivide crystal populations and/or fractions of crystals
according to the conditions under which they may have grown. It is less common to couple
other observations of in situ crystal chemical or isotopic data with bulk mineral crystal ages,
but other textural (e.g., CSD or petrographic studies), chemical, or isotopic information about
the magma as a whole has been used to provide interpretive context for bulk mineral ages.
In this section, we review studies that have combined multiple approaches to understanding
the implications of crystal ages in terms of magma evolution, including combining multiple
dating techniques in the same sample, dating multiple size fractions of the same mineral,
combining U-series ages with trace-element or isotopic data from the same crystals, combining
age information and crystal thermometry, and incorporating textural information. Each of these
approaches offers different perspectives on the processes affecting the chemical and thermal
evolution of magmas and, using crystal ages to provide a temporal context, these types of
combined studies can be very powerful tools for studying the dynamics of magmatic systems.
Combining u-th and u-Pb SIMS dating
Accessory phases that give 238U-230Th results within error of secular equilibrium could
be no more than a few hundreds of thousands of years old or could be Archean in age. If the
former, minerals may have too little 206Pb in-growth for acquisition of a precise 238U-206Pb
age. As an illustration of this, results for a zircon from La Vírgen volcano, Mexico, are within
error of possible secular equilibrium (289+/−100 ka; 1σ) and give a consistent but poorly
constrained U-Pb age of 127±81 ka (Schmitt et al. 2006). Nonetheless, reconnaissance U-Pb
surveys of zircons on the ion microprobe can be used to show whether there is little 206Pb,
supporting inferences that the zircons are all young and that xenocrysts are absent (Lowenstern
et al. 2000; Charlier et al. 2003; Bacon and Lowenstern 2005; Bacon et al. 2007). A particularly
instructive age comparison that emphasizes the importance of identifying the age of old zircon
is provided by the discordant U-Pb vs. U-Th results of Garrison et al. (2006) for a rhyolite
from Cotopaxi volcano, Ecuador. Four zircons that are not in 238U-230Th equilibrium at the
2σ level give 238U -206Pb model ages of 18 to 30 Ma, considerably younger than the 0.8-1.2
Ga zircons also detected in the same rhyolite. This example demonstrates that young rims
on old zircons can produce 238U-230Th disequilibrium while preserving old, if geologically
meaningless, U-Pb ages.
520 Cooper & Reid
In those cases where xenocrysts are identified, U-Pb dating can identify possible sources
and provide insights into the role of crustal recycling and the duration of zircon dissolution
(Vazquez and Reid 2002; Lowenstern et al. 2006; Schmitt 2006; Schmitt and Vazquez 2006;
Schmitt et al. 2006). Cores of West Yellowstone zircons that give 238U-230Th close to or within
error of secular equilibrium yield three 207Pb-corrected 238U -206Pb isotope ages between ~200-
350 ka and two older ages of 770±109 ka and 1192±88 ka (Table 3). In especially favorable
circumstances, it can be possible to obtain either model or isochron ages using both U-Th
and U-Pb systems. For example, Vazquez et al. (2007) obtained a 238U -206Pb isochron age of
261±28 ka (2σ; n = 24; MSWD = 1.8) that is concordant with a 238U-230Th isochron age of 257
+60/-46 ka (2σ; n = 24; MSWD = 0.69), both on zircons from Hualalai leucocratic xenolith
(Fig. 19). Bachmann et al. (2007) obtained continuous U-Th and U-Pb age distributions for the
Kos Plateau rhyolite zircons, from eruption to 340 ka, with concordant weighted mean ages:
228±14 ka (1σ; n = 70; MSWD = 1.08) and 194±10 ka (1σ; n = 74; MSWD = 1.6) for U-Th
on pumice and granitic clasts, respectively, and 213±5 (2σ; n = 33; MSWD = 17.8) for U-Pb
on pumice and clasts combined (Fig. 13).
80
150
U/206Pb (x 10-3)
1.0 60
200
40
0.8 300 350
20 400
207Pb/206Pb
238
600
0 ∞
0.6 0.7 0.8 0.9 1.0
m(238U)-(230Th)
0.4
0.2
300 ka 200 ka 150 ka
600 ka
0.0
0 10 20 30 40 50 60 70 80 90
238U/206Pb x 10-3
Figure 19. Tera-Wasserburg concordia diagram for zircons from Hualalai, Hawaii, leucocratic xenoliths.
Data have not been corrected for common Pb. Concordia curve has been modified to account for the initial
230
Th deficit of zircon assuming f = 0.28 (see Section on ‘Analytical and Practical Considerations’ for
further details). Inset shows that U-Pb ages are concordant with the U-Th age (m(238U)-(230Th) signifies
model age isochron slope). Modified after Vazquez et al. (2007).
u-th/u-Pb spots versus mineral separates
Zircon spot analyses versus TIMS analyses of zircon aliquots. In cases of unimodal age
populations, fitting of in situ data to a single isochron may be warranted. Thermal ionization
mass spectrometric analyses of zircon populations may be justified as well, and can provide
additional information on whether the distribution of spot ages is representative of the overall
zircon population (e.g., Charlier and zellmer 2000; Charlier et al. 2005). The TIMS data can,
in turn, help to assess the match between the initial Th or Pb isotope ratio for the zircons to
Uranium-series Crystal Ages 521
that for the host melt (e.g., Schmitt 2006). However, care must be taken to ensure that there
is not a subtle age variation with parent-daughter ratio which could produce an array with a
slope that is different from the crystallization age and thus yield an erroneous estimate of the
initial ratio.
Zircon spot analyses combined with bulk analyses of major phases. In a few cases,
zircon spot ages and 238U-230Th dating of bulk mineral separates (major phases +/− zircon)
have been done on the same samples, providing additional information for interpretation. For
example, U-series disequilibria have been used to examine processes beneath Laacher See
volcano, Germany, both through analysis of bulk mineral separates (Bourdon et al. 1994)
and in situ analyses of zircon (Schmitt 2006). Bulk mineral-glass U-series ages for minerals
separated from the Lower Laacher See tephra and from syenite cumulate nodules are either
significantly younger (by ca. 5 k.y.) or older (by ca. 12, ca. 22, and ca. 400 k.y.) than the
eruption age (Bourdon et al. 1994). The older ages are interpreted as contamination of the
mineral separates by secular equilibrium xenocrysts that shifted bulk (230Th)/(238U) closer
to the equiline (Bourdon et al. 1994), whereas heterogeneous initial (230Th)/(232Th) for melt
and crystals could account for apparent crystallization ages younger than eruption (Schmitt
2006). Schmitt (2006) recalculated ages for the samples analyzed by Bourdon et al. (1994)
using an estimated (230Th)/(232Th) in melt derived from the initial (230Th)/(232Th) inferred from
zircon analyses. These recalculated ages are in agreement with the eruption age and are also
close to zircon crystallization ages. The new zircon ages, along with recalculated mineral-
glass isochron ages, lead to a significant revision of the interpretation of magma storage and
differentiation times beneath Laacher See, from tens of thousands of years (Bourdon et al.
1994) to only a few thousand years prior to eruption (Schmitt 2006).
A second example comes from Long Valley Caldera, California. Long Valley Caldera
formed during the 760-770 ka eruption responsible for the Bishop Tuff; this eruption was both
preceded by and followed by eruption of rhyolitic domes and lava flows (the precaldera Gass
Mountain rhyolites and the postcaldera Inyo and Mono rhyolites). zircon U-Th ages by SIMS
(Reid 1997) and bulk mineral ages (Heumann et al. 2002) have been measured for postcaldera
rhyolites. zircon in the Bishop Tuff and the precaldera Glass Mountain rhyolites has been
dated by using SIMS U-Pb analyses (Reid and Coath 2000: Simon and Reid 2005; Simon et al.
2007) and TIMS U-Pb analyses (Crowley et al. 2007). Individual feldspar and melt inclusion-
bearing quartz from the Bishop Tuff and Glass Mountain rhyolites have been analyzed for
their Sr (by TIMS) and Pb (by laser-ablation MC-ICPMS) isotope signatures (Christensen and
DePaolo 1993; Christensen and Halliday 1996; Davies and Halliday 1998; Simon et al. 2007),
and Rb-Sr model ages have been reported.
Heumann et al. (2002) present bulk 238U-230Th mineral separate ages for combinations of
glass, sanidine, plagioclase, and amphibole separated from post-caldera rhyolite domes and
lava flows at Long Valley. One sample produces a mineral array with slope corresponding
to an age that is 135 k.y. before eruption, similar to a clustering in zircon spot ages (Fig.
11; Reid et al. 1997) for the same eruption. However, the glass and whole-rock analyses are
not collinear with the other mineral data, suggesting that the minerals did not have a simple
crystallization history. The glass and mineral data for other samples in their study do not in
general form linear arrays, and high concentrations of U and Th in some mineral separates
suggest that the U-Th systematics of mineral separates for major phases are dominated by the
presence of zircon and/or allanite inclusions. They conclude that their data are best modeled
by multiple crystallization events for trace phases (zircon and allanite), some of which were
captured in zircon single-crystal analyses and spot ages, and some of which are reflected in
mineral-glass bulk analyses which are dominated by zircon or allanite inclusions. Rb-Sr dating
of feldspars in the same samples suggests feldspar fractionation at ~250 ka, followed by zircon
crystallization from ~250 ka to ~140 ka.
522 Cooper & Reid
Reid and Coath (2000) and Simon and Reid (2005) presented in situ U-Pb dating of
zircon from the Bishop Tuff and pre-caldera Glass Mountain rhyolites. zircon from the Glass
Mountain rhyolites yielded ages that cluster between 2.0 and 1.7 Ma and between 1.1 and 0.85
Ma and support differentiation and crystallization in these magmas hundreds of k.y. before
eruption. They found that Bishop Tuff zircons had mean crystallization ages that preceded
eruption (at 760-770 ka) by 90 k.y. and concluded that the Glass Mountain and Bishop Tuff
eruptions tapped different magma bodies. This conclusion is supported by additional Pb isotope
data for feldspars separated from Glass Mountain and Bishop Tuff rhyolites (Simon et al.
2007) which show distinct Pb isotope compositions for feldspars from precaldera and caldera-
related rhyolites. In contrast, Crowley et al. (2007) found 238U-206Pb crystallization ages for
the Bishop Tuff that are essentially concordant with the 40Ar/39Ar eruption age, as described
earlier in this chapter. Additional work is needed to resolve the discrepancies between the
SIMS and TIMS age results.
A similar conclusion, that eruptions from the same system tapped distinct magmas, was
shown for New zealand rhyolites by Charlier et al. (2003). Charlier et al. (2003) present
238
U-230Th data for major phase separates (magnetic separate, plagioclase, and biotite) along
with 238U-230Th age of zircon separates by TIMS and in situ zircon 238U-230Th ages by SIMS
from two rhyolitic eruptions from the Okataina volcanic center that likely occurred within
months of each other (the Rotoiti and Earthquake Flat eruptions). Differences in zircon age
spectra and mean zircon ages for the two eruptions indicate that they tapped distinct magma
bodies (Charlier et al. 2003). In this case, the U-Th systematics for major phases appear to be
dominated by inclusions of accessory minerals (monazite or zircon), such that isochron ages
essentially give average accessory phase crystallization; these bulk mineral ages agree with
TIMS ages for zircon separates and with the mean of zircon spot ages (Charlier et al. 2003),
and thus appear not to provide information about the evolution of the magmatic system beyond
what was available from the zircon analyses.
However, in other cases, comparisons of bulk mineral separate ages with in situ ages
can provide a more complete view of the geochemical evolution of a magma. For example,
Charlier and coworkers (Charlier and zellmer 2000; Charlier et al. 2005) analyzed 238U-230Th
disequilibria in bulk separates (plagioclase, orthopyroxene, hornblende, magnetite, and
zircon separates of different size fractions), as well as in situ 238U-230Th data for individual
zircon grains separated from samples of the Oruanui eruption of Taupo, New zealand. The
zircon separates analyzed by TIMS give average ages that vary systematically with size
fraction, which is interpreted to reflect progressive crystallization where the young growth is
preferentially sampled in the smaller size fractions (Charlier and zellmer 2000). A 238U-230Th
isochron derived from glass and major phases is permissive of co-crystallization of the major
phases with the youngest zircons at ~33 ka (Charlier et al. 2005), yet the zircon spot ages
extend to much older ages, suggesting that many of the zircons are antecrystic and record a
longer period of magmatic evolution than the major phases.
Recently, Klemetti and Cooper (2007) present in situ 238U-230Th data for zircons and
238
U-230Th-226Ra data for bulk plagioclase separated from the ca. 1300AD Kaharoa eruption
of Tarawera volcano in the Okataina Volcanic Center, New zealand. zircon spot ages range
from within error of the eruption age to >300 ka (within error of secular equilibrium), though
the majority of ages fall in the range of 20-50 ka. Plagioclase 238U-230Th data yield a two-point
plagioclase-glass isochron within error of the eruption age. High U/Th and high zr measured
in splits of the dissolved bulk plagioclase separate, together with petrographic observation
of zircon inclusions in plagioclase, suggest that the 238U-230Th data for the bulk plagioclase
primarily reflect the presence of zircon inclusions. However, the plagioclase itself must still
be dominantly young in order to host young zircon inclusions. Further evidence for youthful
plagioclase comes from 230Th-226Ra disequilibrium measured in the plagioclase separates,
Uranium-series Crystal Ages 523
which requires that the plagioclase and any included zircon must be less than a few thousand
years old. Thus, the combined data require that the plagioclase (plus zircon inclusions)
crystallized shortly before eruption, whereas the bulk of the zircon grains measured by SIMS
have cores that are tens of thousands of years old, suggesting that most of the zircons were
antecrystic, whereas the plagioclase appears likely to have grown from the host liquid.
In other cases, the bulk mineral ages may be dominated by mixing signatures, where the
zircon ages provide age information. Garrison et al. (2006) analyzed 238U-230Th disequilibria in
bulk mineral separates from two andesites and two rhyolites from Cotopaxi Volcano, Ecuador,
in addition to the U-Th spot ages for seven at least partially xenocrystic zircons described
previously in this section. The bulk mineral age for one of the andesites is undefined, with
the other yielding an apparent age of 28 ka; however, based on textural evidence, Garrison et
al. (2006) interpret this array to reflect mixing rather than having age significance. 238U-230Th
isochron ages for the rhyolites are 76 and 112 ka, compared to eruption ages of 7.2 and 6.3
ka. Collectively, these data suggest melt generation in multiple stages (including melting of
crust adding xenocrysts to the rhyolites), mixing, and extended residence (75-100 ka) of silicic
magmas within the crust prior to eruption.
In summary, zircon spot ages combined with bulk mineral ages for major phases and
zircon can provide significantly more insights about the evolution of magmas than can either
technique alone. Although the details of the information each provides may vary from one
magmatic system to another, the combination of ages for major phases and on trace phases
allows a more complete history of the magma to be determined, including the temporal
relationship between crystallization of major and trace phases.
Multiple size fractions
As discussed previously, Charlier and zellmer (200) and Charlier et al. (2005) analyzed
bulk zircon separates of multiple size fractions, and found in some cases that there were
systematic differences in age between different sizes of crystals. This type of study suggests
that, similar to the additional information about magma evolution provided by dating multiple
phases within a given system, multiple size fractions of a given phase can provide additional
insights about the duration of crystallization for a given phase, or perhaps the proportion of
xenocrysts and antecrysts that may be present in a magma. Few studies to date have attempted
U-series dating of multiple size fractions of major phases. Donnelly and Cooper (2006) present
U-Th data for three size fractions of plagioclase in a sample erupted from Mount St. Helens
in 2004 (Fig. 20). All size fractions have 230Th/232Th much lower than that in the host whole-
rock, suggesting either that the plagioclase is tens of thousands of years old or, more likely,
that the bulk separate represents a mix of older plagioclase (which might also have crystallized
from a lower-230Th/232Th magma) with young plagioclase in equilibrium with the host magma
(Donnelly and Cooper 2006; Cooper and Donnelly 2008). The three size fractions define a
linear array, with the largest (250-500 µm) size fraction having the lowest 230Th/232Th ratio,
but the intersection of the array with the 230Th/232Th ratio in the melt also has higher U/Th than
any of the measured plagioclase or than partitioning would predict for zero-age plagioclase
(Fig. 20). One possibility is that the different size fractions represent mixing between an old,
large, low-230Th/232Th antecrystic plagioclase population and a second end-member that is
itself a mix between zero-age plagioclase and whole-rock material adhering to the outside of
grains or melt inclusions. The U/Th ratio of the largest size fraction is similar to that predicted
from partitioning (if the crystals precipitated from a magma with similar U/Th to the host
whole rock), suggesting that this fraction may contain a large percentage of old plagioclase.
The addition of 230Th-226Ra data for the separates (in progress) will test whether the larger size
fraction does indeed contain a smaller proportion of zero-age plagioclase, which will then
allow an assessment of the characteristics and likely origin of the old plagioclase population.
524 Cooper & Reid
1.4
1.3 zero-age
(230Th)/(232Th)
1.2
1.1
125-250
1.0
75-125
250-500
0.9
0.9 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
(238U)/(232Th)
Figure 20. 238U-230Th isochron showing data for whole rocks and multiple size fractions of plagioclase
in Mount St. Helens dacite. Shown are analyses of two splits of SH304-2A1 whole-rock (black squares),
three size fractions of plagioclase separated from SH304-2A1 (black triangles; labeled with size fraction
in micrometers), additional whole-rock analyses of other dacites erupted from Mt St. Helens during 2004-
2006 (open squares), whole-rock analyses of samples erupted from 1980-1986 (open diamonds), and
calculated composition of zero-age plagioclase in equilibrium with SH304-2A1 whole rock (grey triangle).
All whole-rock data from Cooper and Donnelly (2008); analyses of plagioclase size fractions reported in
Cooper and Donnelly (2008; 125-250 µm size fraction) and Donnelly and Cooper (2006). Solid line is the
equiline, dashed lines represent mixing lines between zero-age plagioclase and whole-rock, and a mixing
line defined by the three size fractions of plagioclase. See text for discussion.
textural perspectives
Textural features of crystals can provide insights into the pre-eruption paragenesis of the
minerals. These include internal chemical heterogeneities within individual crystals and the
distribution of grains, grain sizes, and grain shapes in a crystal population.
Cooper&Reid Figure 20 One important
distinction is whether grains have selvages of glass and therefore were probably immersed in
liquid rather than being isolated from liquid as inclusions in other phenocrysts or in xenoliths
before eruption and mineral separation. For zircon, instances have been reported where
the absence of adhering glass correlates to pre-quaternary grains and where these grains
are interpreted to be accidental crystals derived from disseminated country rock fragments
(Schmitt et al. 2006). The presence of glass selvages on other zircon xenocrysts shows that
those grains were magma-hosted.
Grain shapes and textural features of zircons and/or associated “phenocrysts” can help to
elucidate the paragenesis and evolution of a crystal population. Variable crystal shapes, from
euhedral to anhedral crystal shapes, have been suggested to indicate growth in conditions
in that ranged from melt-rich to melt-poor (Vazquez et al. 2007). Some studies have noted
evidence of rounding of zircons, indicative of partial resorption (e.g., Schmitt et al. 2003b;
Bacon and Lowenstern 2005; Schmitt 2006). These are generally taken as evidence of near-
eruption injection of heat into a system, ostensibly by mafic recharge, and possible derivation
of zircon from more crystalline portions of a system. For example, Bacon and Lownstern
(2005) found that zircons from granodiorite blocks are similar in size, shape, and CL zoning to
zircons from the rhyodacite that entrained the blocks, suggesting a strong affinity between the
two. Plagioclase crystals in the granodiorites exhibit evidence of subsolidus exsolution in the
form of crystallographically oriented Fe-oxide needles. Identification of these same features
in at least 80% of plagioclase crystals in the rhyodacite suggests that they are dominantly
antecrysts derived from plutonic rocks. Schmitt (2006) correlated the presence of ThSiO4
Uranium-series Crystal Ages 525
rods in Laacher See tephra zircon (Fig. 18), tentatively interpreted as exsolution features, to
the same features in associated syenite subvolcanic nodules. Finally, studies have noted that
granophyre zircons have unusual “spongy” textures (Schmitt and Vazquez 2006; Bacon et al.
2007), inferred to reflect zircon growth in the presence of a volatile-rich melt or fluid.
Crystal Size Distributions (CSD). A sensitive test of whether the distributions of crystal
sizes reflect those of growth and not selective accumulation and/or fractionation is to compare
the size distribution to the crystal population densities (see review by Armienti 2008). If
estimates for crystal growth rates and its temporal dependence (e.g., linear) are also available,
such crystal size distributions (CSD) could be used to fingerprint crystal populations and
changes in the populations (if sampled by multiple eruptions), to identify the presence of
mixed or inherited crystal populations (where breaks in the CSD slope are found), and to
estimate the residence time of crystal populations at nominal growth rates.
As discussed above, Turner et al. (2003a) presented crystal size distributions in lavas
and cumulate xenoliths from Tonga and from Soufriere volcano, Lesser Antilles, combined
with 238U-230Th-226Ra data for bulk mineral separates from the same samples. The CSDs for
these lavas are kinked, indicating more than one crystal population within the lavas, consistent
with discordant 238U-230Th-226Ra data suggesting that the crystals in Tonga and Soufriere lavas
contain more than one component (Turner et al. 2003a).
zircon dissolution experiments suggest growth rates on the order of 10−15-10−17 cm/s
for silicic rocks (Watson 1996b). Residence times for zircon populations calculated using
these growth rates are of the order of 100–200 k.y. (Bindeman et al. 2001; Bindeman and
Valley 2002). Both conventional separation procedures and extraction of zircons by selective
HF-dissolution have been applied, with no apparent differences. However, a key element of
uncertainty in these interpretations is the extent to which progressive recrystallization has
modified the CSD (Fig. 21; Bindeman et al. 2001, 2006; Bindeman and Valley 2002, 2003;
Bindeman 2003).
Cathodoluminescence. Back-scattered, secondary electron, and cathodoluminescence
images of polished grain mounts can reveal chemical variations within minerals. Of these,
cathodoluminescence (CL), the detection of light emitted in the visible portion of the spectrum
when a sample is bombarded with electrons, provides the finest-scale detail for zircon (Fig. 10).
Differences in shading reflect relative differences in REE and U concentrations; accordingly,
CL can be used to detect chemically distinct zones within individual grains and so is often
used to guide the choice of analytical spots. CL can also supplement the optical identification
of glass selvages on crystals (Fig. 4), thus providing evidence that the grains were surrounded
by melt rather than having been occluded in other minerals.
Within any given zircon populations, the CL record of crystal growth may range from
simple to complex. In most cases, CL imaging of young zircon provides evidence for fine-
scale oscillatory (magmatic) zoning (e.g., Bacon et al. 2000, Vazquez and Reid 2002, Schmitt
et al. 2003b, Miller and Wooden 2004; Schmitt 2006, Schmitt and Vazquez 2006). Sector
zoning in zircon is also relatively common, resulting from significant differences in the trace
element contents of the sectors and identified by patches of relative CL brightness. Fine-
scale oscillatory zoning may be superimposed on an overall pattern of U and Th growth
zoning (e.g., Miller and Wooden 2004). More rarely zircons may contain internal boundaries
where zoning is truncated, indicative of periods of where crystals experienced episode(s) of
resorption followed by subsequent growth. In more pronounced cases, distinct cores, often
anhedral and embayed and therefore indicative of more pronounced dissolution, are present.
These cores may be associated with ages that are demonstrably older than the rims (Vazquez
and Reid 2002; Schmitt et al. 2003b,c; Brown and Smith 2004; Miller and Wooden 2004).
Bindeman et al. (2001) reported that ~5–20% of zircons from the 0.4-0.5 Ma post-caldera
rhyolites at Yellowstone caldera had resorbed cores whereas Brown and Smith (2004) found
526
Figure 21. quartz (A) and
zircon (B) size distributions
in various caldera-related
lavas. Abbreviations are
LBT: Lower Bandelier
Tuff, BT: Bishop Tuff;
LCT: Lava Creek Tuff;
HRT: Huckleberry Ridge
Tuff; MFT: Mesa Falls Tuff;
AT: Ammonia Tanks Tuff;
YTT: Youngest Toba Tuff.
150-400 quartz and 250-
700 zircon crystals were
measured for each CSD.
Y-axis shows population
density, expressed in terms
Cooper & Reid
of number of nuclei (#) of a
characteristic length (L) in
a sample volume (V). CSD
slopes are also given and
these can be used to derive
residence times if a constant
growth rate is assumed.
See Armienti (2008) for
further discussion of CSD.
Modified after Bindeman
(2003).
Uranium-series Crystal Ages 527
only four crystals with resorbed cores among more than 300 zircons from the 1.21 Ma Ongatiti
ignimbrite, Taupo Volcanic zone (Fig. 9), and Bacon and Lowenstern (2005) found no CL
evidence for inherited zircon cores in a Crater Lake granodiorite block despite searching
hundreds of grains. A relative lack of clearly distinct cores would seem to be consistent with
the general lack of xenocrysts in most volcanic systems studied to date. On the other hand,
more subtle internal resorption boundaries may separate episodes of growth where multimodal
age distributions are found (e.g., Miller et al. 2004; Charlier et al. 2005; Schmitt et al. 2006).
Using a combination of CL imaging and age results, Bindeman et al. (2001) identified
four populations (excluding xenocrysts) in the aforementioned post-caldera rhyolites at
Yellowstone. Schmitt and Vazquez (2006) noted that the patchy CL images of Salton Buttes
rhyolite zircons were similar to those observed in zircons from associated and similarly-aged
granophyre zircons. Brown and Fletcher (2004) found that many of the larger zircons in the
Ongatiti ignimbrite have a common internal structure consisting of featureless rims of 10-50
µm surrounding finely zoned internals, often separated by a dark CL zone with sparse melt
inclusions. Thus CL patterns can enable episodes of growth to be recognized and may help to
identify different zircon populations.
Crystal-scale major element/trace element variations
Major elements. U-Th dating of the allanite can be especially informative since this
mineral, an accessory epidote found in some silicic magmas, exhibits diagnostic interelement
substitutions. Vazquez and Reid (2004) found that variations in MnO/MgO and La/Nd in
allanite are sensitive indicators of magmatic differentiation and that the complexly zoned
allanites from the Youngest Toba Tuff, Indonesia, revealed crystallization environments that
alternated between less and more differentiated magmas. Euhedral to subeuhedral intragrain
zone boundaries suggest unrestricted growth at conditions that were more liquid than solid.
Minerals may have been exchanged and mixed between magmas of different compositions.
By pairing these chemical variations with U-Th ages, Vazquez and Reid found a pronounced
increase in the chemical heterogeneity of the grains that suggested strong chemical
diversification of the magma within a few tens of k.y. of eruption.
U, Th, and Th/U. zircons from volcanic and nonmiarolitic plutonic rocks typically have
a few tens to several thousand of parts per million (ppm) each of U and Th (Belousova et al.
2002). In their recent compilation of magmatic zircon Th/U (Fig. 22), Bindeman et al. (2006)
corroborated the earlier characterization of Hoskin and Schaltegger (2003): most igneous
zircons have Th/U ~ 0.5, with a range from ~0.2 to ~0.9. Exceptions include carbonatitic
zircons (Th/U up to 9,000; Hoskin and Schaltegger 2003) and, as described further here, some
granophyre zircons (Th/U up to 150; Schmitt 2006).
Many workers have compared Th/U and/or U and Th concentrations within and between
zircons, and between zircons and their host melts to assess the affinities amongst them
(Schmitt 2006; Bacon et al. 2007). For example, Timber Mountain zircons have comparatively
anomalous Th/U ratios (in excess of unity and up to 4.7) but most can broadly be explained by
normal Th/U partitioning as they crystallized from their high Th/U host magmas (Bindeman et
al. 2006). Commonly, several-fold to more than an order of magnitude variations in U contents
within and between zircons from a given sample are observed (Brown and Smith 2004; Miller
and Wooden 2004; Charlier et al. 2005; Simon and Reid 2005). Interface kinetics and the
compositional- and temperature-dependences of partition coefficients will tend to amplify
the variations in zircon relative to the chemical variations in melts from which it grows.
Additionally, it can be difficult to assign a normal trend to the variation in U and Th/U because
of the heterogeneity—sometimes local—in accessory phase assemblages. U-rich cores may
be statistically older than their U-poor rims (Bindeman et al. 2001; Miller and Wooden 2004;
Schwartz et al. 2005) or the opposite trend in U concentration may occur. Extreme variations
in U concentration may characterize zircons that show evidence of multiple periods of growth
528 Cooper & Reid
Typical igneous zircon
(A)
100 High-Si rhyolites
(~75 wt.% SiO2)
50 Average DTh/U =
0.19±0.09
0
(B)
Intermediates
(65-75 wt.% SiO2)
50 Average DTh/U =
0.24±0.10
0
20 (C)
Intermediates
(55-65 wt.% SiO2)
10 Average DTh/U =
0.26±0.16
0
(D)
200 Anorthosites/gabbros
Average DTh/U =
100 0.19±0.07
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Th/U
Figure 22. Histograms showing Th/U ratios in zircons from a variety of host rocks. Data for high-silica
rhyolites and intermediate rocks are dominated by results for Cenozoic igneous rocks from the western
U.S. Low silica rocks include granodiorites, gabbroic anorthosites, and gabbros from MORB and
intracontinental rift settings. Modified after Bindeman et al. (2006).
and resorption, where magmatic zoning is clearly truncated by multiple internal dissolution
surfaces (e.g., Brown and Fletcher 1999; Miller and Wooden 2004), and these variations
enhance the chances of recognizing resorption surfaces. Most workers conclude that a given
zircon population can grow in a variety of chemical environments, including ones representing a
spectrum of melt fractions (Brown and Fletcher 1999; Schmitt et al. 2003c; Miller and Wooden
2004; Charlier et al. 2005).
High Th (up to ~5 wt%) and Th/U (up to 150) characterizes zircons from miarolitic plutonic
rocks from some volcanic centers (Schmitt 2006; Bacon et al. 2007). Other granophyre zircons
that apparently grew in the presence of hydrothermal fluids (Schmitt and Vazquez 2006) have
moderately high Th but the average to low Th/U observed in hydrothermal veins (Hoskin 2005).
The former may be related, in part, to microinclusions of monazite (Schmitt and Vazquez 2006)
or to advanced degrees of melt differentiation in the presence of a highly oxidizing vapor phase
that reduces the tendency of U to partition into zircon (Bacon et al. 2007).
Other trace elements. As with Th and U, there can be significant variations in the other
trace element contents in zircons and particularly in the rare earth elements (REE). Yttrium
concentration variations of up to an order of magnitude are correlated with Th and U variations
in Yellowstone zircons (Bindeman et al. 2001) and with zones that are dark and light under
Uranium-series Crystal Ages 529
cathodoluminescence. Such large variations cannot be explained by equilibrium crystal growth
from a single, mostly liquid-rich domain and support inferences from U and Th concentrations
and from Th/U variations for growth of zircons in environments that range from melt-rich to melt-
poor magma, including in plutonic (subvolcanic) rocks (e.g., Bacon et al. 2000, 2007; Bacon
and Lowenstern 2005; Schmitt 2006). A factor of two variation in trace elements, especially Y,
concentrations was also found between different crystal sectors of single Yellowstone crystals,
showing that, as noted above, crystal growth kinetics are important in determining absolute
abundances (Watson and Liang 1995; Watson 1996a). These observations suggest that trace
element concentrations in zircon are dependent on local variations in concentration in the
source and on the kinetics of zircon crystallization (e.g., Hoskin and Ireland 2000).
The effects of kinetics and local trace element concentrations notwithstanding, the relative
trace element characteristics of zircon may help to assess the affinity of a zircon to its host
and, as warranted, to fingerprint possible sources of zircon inheritance (e.g., Belousova et al.
2002; Hoskin 2005). Diagnostic features may include HfO2 contents, the ratio of LREE:HREE
(zircons are generally strongly HREE-enriched), and the presence and magnitude of anomalies
in Ce (Ce/Ce*, related to the preferential incorporation in zircon of Ce4+ compared to Ce3+) and
Eu (Eu/Eu*, mirroring the influence of feldspar fractionation on the host melts; e.g., Bacon
et al. 2007; Vazquez et al. 2007). Vazquez et al. (2007) found that REE concentrations in
Hualalai zircons are typical of zircon from evolved alkalic magmas. Schmitt (2006) matched
REE patterns of inherited Laacher See zircons (Fig. 23) to syenite (negative Eu anomalies)
rather than carbonatite (little or no Eu anomalies). Elevated HfO2 concentrations, relatively
high Th/U and LREE/HREE (and weak or absent positive Ce anomalies) may be diagnostic
of crystallization in the presence of an oxidizing fluid (Hoskin 2005). These same features are
observed in the spongy textured granophyric zircons and suggests late-stage crystallization of
these zircons in the presence of an aqueous fluid (Fig. 23; Schmitt and Vazquez 2006; Bacon
et al. 2007). As noted above, Schmitt and Vazquez (2006) attribute these characteristics at
least in part to the presence of small (~5 µm) monazite inclusions whereas Bacon et al. (2007)
suggest that oxidation of late-stage melts causes zircons to exclude U5+ and U6+, leading to high
Th/U. Ongoing investigations aimed at linking volcanoes and plutons will undoubtedly lead to
additional means of fingerprinting zircon origins.
Trace elements in major minerals. Compared to zircon, relatively little effort has focused
on coupling trace-element variations with ages of major phases. This likely reflects in part
the fact that it is easier to make connections between information from trace elements at the
sub-crystal scale with ages at a sub-crystal scale than it is to link sub-crystal trace-element
information with bulk separate ages. Despite this difference in spatial scale sampled, it is still
possible to make some broad inferences about the timescales of the magmatic processes that
are reflected in trace-element zoning. For example, as discussed earlier, sub-crystal zoning
in barium in plagioclase from Mount St. Helens—beyond that predicted from major-element
variations and partitioning behavior—provides evidence that the cores of many crystals are
antecrystic and, by extension, that crystals have not been held at high temperatures long
enough for diffusion to equilibrate crystals with the host liquid. One further implication is that
230
Th -226Ra crystal ages have not been affected by diffusive exchange with the host liquid.
Similarly, 238U-230Th data for minerals separated from dacite at Kameni, Santorini yield a U-Th
isochron age of 18 +19/−16 ka (zellmer et al. 2000), whereas Sr diffusion profiles in Kameni
lavas indicate crystal residence times of less than 100-450 years (zellmer 1999). These data
collectively suggest that a relatively small proportion of the overall history of antecrysts was
spent suspended in the host magma at high temperatures.
Crystal ages and thermometry
Experimental determinations of zircon solubility make it possible to bracket the
conditions under which zircon will be stable in silicic magmas (Watson and Harrison 1983).
530 Cooper & Reid
Figure 23. Chondrite-normalized rare earth element patterns for zircons from: A) Lower Laacher See
tephra (LLST) and subvolcanic nodules and B) Veniaminoff volcano xenoliths. LLST tephra and syenite
zircons both share REE patterns characterized by heavy rare earth element (HREE) enrichment, prominent
positive Ce anomalies, and significant negative Eu anomalies. These patterns are similar to those for
evolved syenitic zircon but differ from those of carbonatitic zircons where negative Eu anomalies caused
by feldspar fractionation are typically absent (e.g., Belousova et al. 2002). Most Veniaminof zircons also
have chondrite-normalized rare earth element (REE) patterns characteristic of igneous zircon (Belousova
et al. 2002), with increasingly negative Eu anomalies being consistent with crystallization differentiation
involving plagioclase. Bacon et al. (2007) attribute 1) high REE concentrations and large Eu anomalies
to crystallization from highly differentiated interstitial melts; 2) unusually large Ce/Ce* in diorite and
granodiorite zircon to crystallization in the presence of an oxidizing fluid; and 3) elevated light REE
concentrations and REE patterns in two spongy-textured granodiorite zircons to hydrothermal zircons
(Hoskin 2005). After Schmitt et al. (2006) and Bacon et al. (2007).
zircon saturation mainly depends on the zr content and temperature of a magma but the bulk
chemical composition of a magma (especially the relative proportion of network modifiers to
network formers) also affects zircon solubility. Pressure, including water pressure, does not
seem to directly affect the stability of zircon to an appreciable degree (Watson and Harrison
1983) but may have an indirect effect by influencing the stability of other phases. Saturation
temperatures obtained from the glass or groundmass compositions of a volcanic rock may be
Uranium-series Crystal Ages 531
those closest to the pre-eruptive conditions
but estimates based on whole-rock Western Moat, Long Valley, CA
compositions are usually not significantly (A)
different at low degrees of crystallinity.
Some studies have compared the
zircon saturation temperature (Tzirc) to Crater Lake, OR
estimated magma temperatures (Tmag), (B)
particularly those from Fe-Ti oxides, in
order to gain insights in thermal history
in the lead up to eruption (Fig. 24). For Taupo Volcanics, NZ
those cases where Tmag > Tzirc, zircons will
Subgroup 3
be dissolving and the model of Watson (C)
(1996b) can be used to calculate the time
Subgroup 2
necessary for zircon dissolution to occur.
The model assumes that zr diffusion is the
Subgroup 1
rate-limiting process for zircon dissolution
and thus that there is a static chemical
boundary layer in melt around a zircon. Dacites
Based on a preponderance of ~200 ka
zircons in the coarsely porphyritic rhyolites Oruanui rhyolites
of Deer Mountain and Inyo Domes, Long
Valley, California, Reid et al. (1997) argue Okaia
that the magma responsible for them could
not have been at T > Tzirc for more than 1 Tihoi
k.y. since the zircon crystallized. For those
domes, iron–titanium oxide temperatures 750 790 830 870 910 950
(809 ±4 °C) are similar to those for zircon Temperature °C
saturation (795-805 °C; Fig. 24). In
contrast, iron–titanium oxide temperatures Fe-Ti oxide T Zircon saturation T
(~850–880 °C) for the 27 ka rhyodacite
eruption at Mt. Mazama exceed the zircon Figure 24. Comparison between Fe-Ti oxide
saturation temperature by ~50 °C (Bacon temperatures (closed bars and circles) and zircon
saturation temperatures (open bars) in Long Valley
and Lowenstern 2005). Rhyodacite western moat rhyolites, Mt. Mazama rhyodacites
zircons, inferred to be antecrysts based on and granodiorites, and Taupo volcanics. Instances
age and phase similarities to granodiorite where paired temperatures overlap and where they
blocks, must have been dissolving. A are discrepant are found. See text for discussion.
modest amount of dissolution (5%) could Modified after Charlier et al. (2005) with data from
Reid et al. (1997), Bacon and Lowenstern (2005),
explain the subtle rounding of zircon that and references therein.
is observed and requires dispersal of the
granodiorite into the magma within a few
tens of years of eruption. Even after that eruption, dissolution would have continued in the
reservoir so that by the time the climactic eruption of Mount Mazama responsible for the
Crater Lake caldera occurred, no zircons remained. Complete dissolution would have only
required ~700 yr.
Among the Taupo Volcanic Field dacites to rhyolites, the unit Ω dacite has Tmag > Tzirc
(Fig. 24; Charlier et al. 2005), indicating that zircons should have been dissolving in the lead
up to eruption. Unit Ω is one of the only Taupo magmas in which pre-350 ka quaternary and
older inherited zircons are found. Calculations based on Watson (1996b) show that the largest
zircons (250 µm) will dissolve in <100 yr in the Unit Ω dacite. This and the subhedral to
euhedral morphologies of the zircons imply that the Unit Ω dacite must have been erupted
within tens of years of the zircons being introduced into the dacite magma.
532 Cooper & Reid
Somewhat different variants on the case of Tmag > Tzirc are those where, rather than
immediately before eruption, some earlier interval in the magmatic history is inferred to have
been at temperatures greater than Tzirc. In one instance, quartz-hosted glass (melt) inclusions are
significantly undersaturated in zircon at the equilibration temperatures obtained for the major
mineral phases (Schmitt et al. 2003a). Thus, zircon growth must postdate melt-entrapment in
quartz. zircon saturation temperatures similar or less than those inferred for magma liquidus
temperatures have also been used to explain the absence of zircon xenocrysts (e.g., Schmitt et
al. 2003c).
Other studied volcanic eruptions have Tmag ≤ Tzir. In these cases, zircons can reasonably
be assumed to preserve the last time that temperatures passed through the zircon saturation
temperature, except where thermal excursions over the zircon liquidus have been modest and/or
short-lived. Cases where zircon saturation temperatures are greater than or are similar to Fe-Ti
oxide temperatures have been used to suggest that 1) zircons date the attainment of conditions
that correspond to zircon saturation (e.g., Reid et al. 1997; Reid and Coath 2000; Vazquez and
Reid 2002); 2) magmas fluctuated between zircon saturated and undersaturated conditions as
they cycled between cooling, crystallization, and magma mixing or recharge events (e.g., Brown
and Smith 2004; Miller and Wooden 2004); and/or 3) transient heat pulses could incorporate
zircons that then persist, resulting in zircon inheritance and zr excesses (Bindeman et al. 2006).
Magmatic fluctuations could account for growth that was episodic rather than continuous and
thus explain multimodal age populations.
The recent development of a Ti-in-zircon geothermometer (Watson et al. 2006) means that
temperatures can be obtained for individual zircons and, significantly, for zones within zircons.
Activities for TiO2 and SiO2 in the melt must be obtained or assumed and, if temperatures have
varied, may include allowances for activity variations in response to changing accessory phases
and mineral proportion. Uncertainties in the Ti activity notwithstanding, Schmitt and Vazquez
(2006) could use the Ti contents of the Salton Trough zircons to estimate that partial melting of
hydrothermally altered basalts produced rhyolites at temperatures between ∼770 and 835 °C.
Vazquez and Reid (2004) also obtained crystal-scale temperature information, in this case
for allanite, by correlating the chemical variations within individual allanites to those expected
for equilibrium with experimental melts produced at temperatures that ranged from 700 to
770 °C (Fig. 25). In this way, they were able to show that the thermochemical evolution of
individual allanites from Toba caldera, Indonesia, can be complex and reflect crystallization
under conditions that could have included mingling between recharge and resident magmas
and/or during intrareservoir self-mixing.
Single-crystal or sub-crystal oxygen isotopic data
zircons are resistant to weathering and to isotopic exchange at surface and near-surface
conditions, and therefore can preserve oxygen-isotopic compositions imparted at the time
of crystallization. Coupling of in situ oxygen isotope analyses with U-Pb radiometric age
determinations has proven to be a potent means for investigating the crustal evolution of Earth
(e.g., Valley 2003; Kemp et al. 2006). Specific to the dynamics of magmatic processes, the
oxygen isotope compositions of zircon mirror those of the melts from which they crystallize.
It is therefore possible to use zircon-melt isotopic fractionation factors to determine whether
the isotopic composition of a zircon’s host is different from the one that was responsible for
zircon growth. In particularly amenable cases, this can be used to distinguish antecrysts from
true xenocrysts. Where isotopic disequilibrium between a zircon and its hosts pertains, it is
additionally possible to use the intragrain O isotope gradient in zircon to estimate the residence
time of zircons in the new host. Core-rim variations in the oxygen isotope values of zircons
can be obtained by successively abrading and analyzing zircons grains or by ion microprobe
analyses (e.g., Bindeman and Valley 2000).
Uranium-series Crystal Ages 533
Figure 25. Temporal changes in the crystal growth environment of the Younger Toba Tuff, Indonesia, as
inferred from internal covariations between MnO/MgO and U-Th ages in allanites. Predicted variations
in temperature (right-hand axis) associated with changes in MnO/MgO are based on experimental
crystallization of Younger Toba Tuff magma at 100-200 MPa (Gardner et al. 2002). Also shown are trends
of fractional crystallization for different fractionation rates, using two starting compositions. Increase
in the diversity of allanite composition between 110 and 75 ka is inferred to reflect interaction between
and mingling of variably fractionated magma bodies with temperatures between ~760 and 715 °C. MnO/
MgO obtained by electron microprobe analyses; U-Th ages obtained by ion microprobe analyses. [Used
with permission from The American Association for the Advancement of Science from Vazquez and Reid
(2004) Science, Vol. 305, p. 993.]
Some eruptions that occurred in the aftermath of caldera collapses are characterized by
low-δ18O lavas and tephras (e.g., the 0.4 -0.5 Ma postcaldera eruptions at Yellowstone caldera
complex, Wyoming, and the Ammonia Tanks pre- to post-caldera cycle that followed in the
wake of the Rainier Mesa caldera eruption, Timber Mountain/Oasis Valley, Nevada), with δ18O
ranging to more than 5‰ lower than mantle values. These 18O-depleted melts are generally
inferred to be derived by remelting of older, hydrothermally altered, but otherwise chemically
similar, rhyolites (e.g., Bacon et al. 1989; Balsley and Gregory 1998; Bindeman et al. 2001b).
Cores of zircons associated with these eruptions have normal rather than low δ18O values and
their ages overlap the timing of earlier eruptions, suggesting that the zircons are inherited in
some fashion (Bindeman and Valley 2000, 2001; Bindeman et al. 2001, 2006, 2008). Core-
to-rim decreases in δ18O (by as much as 5‰; Fig. 26; Bindeman et al. 2001) indicate that the
534 Cooper & Reid
Figure 26. Model isotope profiles (thin curves)
for 52.5 µm radius zircons compared to δ18O
values obtained by successive air abrasion of
zircons from Middle Biscuit Basin, Yellowstone
(bold vertical bars). Isotope values relative to
VSMOW (Vienna standard mean ocean water).
Durations of diffusional re-equilibration at
850 °C between cores with δ18O = +4‰ and
rims with δ18O ~ −1‰ indicated; associated
whole zircon δ18O values shown by italicized
numbers. Bold vertical bars connect measured
core (left-oriented horizontal marks) and rim
(right-oriented horizontal marks) δ18O values;
rim compositions obtained by difference
between whole zircon and core values. The
sharp contrast between the core and rim
compositions of the zircons suggests residence
times of <102-103 yr at 850 °C or 103-104 yr
at 750 °C. An inferred zoning profile, shown
by the thick grey line, yields a weighted δ18O
that corresponds to the measured δ18O of
whole grains. Also shown for reference are the
isotopic composition of host obsidian (striped
box) and the rim isotopic composition of four
zircons obtained by ion microprobe analyses
(large filled circle). See Bindeman and Valley
(2000) for additional details.
zircons grew from and/or partially exchanged oxygen with 18O-depleted melts. The failure of
zircon to equilibrate with its host limits the residence time of zircons in the melt: the duration
of melting must be on the order of 10 k.y. or less at Yellowstone, based on solid-state diffusion
modeling, or less than a few thousand years at Ammonia Tanks Tuff, Nevada, based on the
dissolution calculations by Watson (1996). Lower Laacher See Tuff zircons have δ18O values
up to 1.3‰ lower than expected for equilibrium with glass in their host phonolite (Schmitt
2006); as noted elsewhere, extrapolated initial Th isotope ratios for the zircons are also out of
equilibrium with their host. Oxygen (and extrapolated initial Th) isotope similarity between
zircons in the tephra and in syenite nodules lead to the conclusion that the zircons date partial
to complete crystallization of subvolcanic apophyses. The absolute values of the thorium and
oxygen isotope ratios in zircon also support evidence from the presence of abundant crustal
xenocrysts for extensive crustal contamination.
Melts with δ18O values of +5.4 to +6.6‰ are estimated to be responsible for the Salton
Trough zircons based on their oxygen isotope compositions and these match values obtained
for Salton Buttes obsidians. Schmitt and Vazquez (2006) show that these values are too low
for silicic magma genesis by melting of continental crust or by fractional crystallization of
tholeiitic basalt and conclude that melting of basalts in the presence of geothermal fluids could
be responsible for generation of silicic magmas in the Salton Trough (Schmitt and Vazquez
2006). In support of this interpretation, zircons located within plagioclase-rich domains in the
basaltic xenoliths have δ18O values that overlap the range of rhyolite zircons.
Summary
Various combinations of crystal age information and crystal-scale chemical or isotopic
data can thus serve several purposes: first, the addition of other information can help interpret
age data, and in particular can help distinguish antecrysts from xenocrysts and identify mixed
crystal populations that may have distinct ages. Second, the addition of age information to
Uranium-series Crystal Ages 535
studies involving thermometry (whether by using zircon saturation temperatures or Ti-in-
zircon thermometry) can replace qualitative thermal histories of magmas with quantitative
thermal histories. In cases where sub-crystal chemical information is also available, the
thermal and chemical evolution of magmas can be quantified. Finally, in cases where isotopic
and/or chemical disequilibrium between crystals and melts has been preserved, crystal ages
and diffusion modeling can be combined to delimit the duration of crystal residence at high
temperatures and/or within a given melt. Thus, the combination of crystal age information
with other textural, chemical, and isotopic data represents one of the most promising pathways
toward furthering our understanding of magma dynamics.
Key queStIOnS and Future dIreCtIOnS
What can crystal ages tell us about magmatic processes?
The unique contribution of U-series crystal ages to studies of magmatic processes is the
ability to quantify absolute ages of crystal populations or even individual zones within crystals,
on timescales commensurate with many magmatic processes. What is simultaneously the great
strength and the interpretive challenge of using crystal age information to understand magmatic
processes is that what is dated is the record of magma evolution contained within the crystals,
not the magma as a whole. This has the advantage that the crystals record processes and events
that are averaged and therefore obscured in the liquid fraction of the magma, but the weakness
is that the crystal only contains a record of what was happening while it was growing. For
example, there could be substantial periods of time that are represented only by discontinuities
in the crystal zoning record. In cases where crystal size and age have been combined to calculate
apparent average crystal growth rates, these rates are much slower than experimentally-
determined crystal growth rates, suggesting significant periods of time where the crystals were
not growing or were even resorbed. In the case of bulk crystal aggregates, any interpretation
must take into account the fact that analyses are averaging over many grains and may also
reflect the influence of melt inclusions or accessory phase inclusions, so the crystal age may not
correspond to the age of the major phase constituting the bulk of the separate. We can evaluate
the extent to which this is the case by in situ (e.g., LA-ICP-MS or SIMS) measurements of trace
elements in the crystal compared to trace-element contents of the dissolved bulk separates.
These interpretive challenges notwithstanding, the studies described in earlier sections
demonstrate the key role that crystal ages can play in understanding processes operating
beneath the surface in magmatic systems. For example, crystal ages provide some of the
strongest evidence for the longevity of magmatic systems, which have forced a re-evaluation of
the physical nature of magmas when stored for long periods of time, the interactions between
different magmas passing through the same system, and the implications of common (if not
ubiquitous) crystal recycling (see section “Crystal recycling” above). Crystal ages have also
provided evidence that magma bodies erupted close together in space and time have distinct
crystal ages, which requires that the magmas were not interconnected—or at the least, not well-
mixed (e.g., Charlier et al. 2003). Conversely, in different settings, crystal ages have shown
that magmas which erupted from the same volcanic system, thousands of years apart, contain
crystals with the same age spectra and therefore likely were tapping the same magma bodies
and/or cumulate bodies (e.g., Reid et al. 1997). Interpretations such as these would simply not
have been possible without age information.
Although the foregoing examples show that crystal ages even in (relative) isolation
can provide insights into magmatic processes, crystal ages are most powerful as tools for
unraveling magma dynamics when combined with other information about the dated crystals
and their relation to host melts. For example, even in simple cases, crystal ages may be
related to magma residence times or to differentiation timescales, but these interpretations
536 Cooper & Reid
rely on having corroborating evidence for chemical equilibrium between crystals and host
melts. Independent sources of information at the crystal scale are even more important to
understanding crystal age information in cases where the crystal history may be more complex;
additional information is necessary in order to test multiple interpretations that are permitted
by the age data alone. As demonstrated by the examples presented above, the addition of
age information from multiple parent-daughter pairs and/or multiple size fractions, and the
addition of textural and crystal-scale geochemical data (trace-element or isotopic data) to age
information, can provide the means to take the crystal ages beyond simply understanding
when crystals grew to understanding the rates and timescales of processes such as the thermal
and chemical evolution of magmas, and both temporal and trace-element or isotopic evidence
for recycling of crystalline material from earlier phases of magmatism.
Most U-series ages, however, are not suited to capturing some very rapid or short-lived
magmatic processes. The most commonly-used parent-daughter pairs (238U-230Th and 230Th-
226
Ra) yield ages that capture magmatic events on timescales of centuries to hundreds of
thousands of years, and thus do not record processes such as re-equilibration between crystals
and new host liquids (diffusion ages of residence) or degassing-induced crystallization (one
caveat being that anomalously high Ra/Ba ratios may provide a clue that crystallization was too
rapid to allow trace-element equilibration). An exception, however, is 226Ra-210Pb crystal ages,
which yield information about crystal growth on years to decades timeframe. This timescale
is similar to the timescales of magma storage given by diffusion studies (e.g., Costa 2008),
and there is a great deal of promise in combining 226Ra-210Pb data with diffusion modeling to
access the youngest part of a magma’s crystallization and crystal recycling history.
What do we get from the different scales of analyses?
The strength of in situ analysis is the ability to focus in on a relatively restricted period
of crystal growth, and potentially to identify ages of more than one growth episode in a single
crystal. In situ ages also give information about the distribution of ages within a sample,
rather than a single age for a large number of crystals. On the other hand, a large number of
analyses is required in order to characterize the age of a population within a sample, and this
is especially true if there is more than one population of crystals. With our current analytical
capabilities, in situ analyses are also restricted to accessory phases. This can be an advantage
in that the conditions of zircon growth are more restricted than many of the major phases,
so that relating ages to parameters such as temperature may be more straightforward than
with major phases. However, it can also be a disadvantage in that crystal growth may be
more subject to kinetic limitations on diffusion of trace elements through silicic melts, and the
trace-element compositions of zircon or allanite may therefore be more likely to reflect kinetic
effects than are major phases.
In contrast, bulk mineral separates—especially of major rock-forming minerals—require
analysis of hundreds to thousands of crystals, and crystal ages thus derived are necessarily
averaging over the entire population of crystals sampled. The disadvantage is that it is more
difficult to identify multiple phases of crystal growth or multiple crystal populations, and an
average age derived from a mixed-age mineral separate may not correspond to any geological
event. On the other hand, one advantage is that the influence of a few crystals of distinct age
(for example, xenocrysts in secular equilibrium) on the average age is relatively limited. In
addition, bulk crystal data are likely to be representative of the overall crystal population(s)
within a given magma, as long as an appropriate size range is sampled. Furthermore, fractional
crystallization of major minerals is one of the main controls on the chemical evolution of
magmas; therefore, bulk crystal ages may be more easily related to processes that operate at
the magma reservoir scale. In situ ages and bulk mineral separates thus offer complementary
information, and it can be particularly powerful to combine the two techniques.
Uranium-series Crystal Ages 537
Future directions
Crystal ages are clearly an important part of the picture in understanding the dynamics
of magmatic systems. Equally clearly, these data are most powerful when combined with
other types of information about the crystals that are dated, and the links between the crystals
and both the host magmas and any other magmas from which at least some of the crystals
precipitated. Progress has been made in combining age information with other textural and
chemical information, but the potential for combined studies is only starting to be tapped. In
particular, what is necessary in order to interpret crystal ages in the context of the dynamics and
evolution of a given magmatic system is additional information with which to link the crystal
ages to the chemical and textural evidence for magma evolution. Some examples (though not
a complete list, by any means) of potentially fruitful directions for future work follow.
Crystal-scale trace-element and isotopic data. Additional measurements combining
trace-element information at the sub-crystal scale with age information could be particularly
useful in tracking mixed magmas and crystal recycling. For example, trace element data could
potentially be used to identify the overall degree of evolution of a magma during crystallization,
to identify any discontinuities in trace-element zoning that could reflect magma recharge or
incorporation of crystals into a new magma, and in general to assess the extent to which crystals
have attained chemical equilibrium with their host melts. In combination with absolute age
data, studies of trace-element diffusion can provide information as to the duration of storage of
crystals at high temperatures (i.e., in magmas as opposed to storage in mush zones or as young
plutonic bodies). Sub-crystal isotopic data in combination with age data could be especially
useful for identifying antecrystic or xenocrystic components to crystal populations, and to
quantifying the proportion of crystals that represent different populations; in the case where
discordant ages are obtained from different parent-daughter pairs, this kind of information
could be used to quantify the age of old and young components of crystallization, and therefore
the timescales over which crystals are stored and recycled.
Thermometry. A number of new types of thermometers have recently been developed,
and their potential in combination with crystal age data to develop time-temperature (+/−
composition) paths recorded in crystals has only started to be explored. In particular, Ti-in-
zircon thermometry together with spot ages in the same crystals can allow the use of zircon
crystals as tracers of the thermal state of a magma over time. There is also great potential
in combining these temperature-time studies with sub-crystal trace-element data to obtain
information about the chemical state of the magma over time.
Melt inclusions. Little has been done to date in combining information from melt
inclusions with age data for the same crystals. Study of melt inclusions in dated crystals could
be particularly useful for constraining the volatile history of magmas (especially for putting
the volatile contents of melt inclusions within a temporal framework) and for examining the
isotopic and trace-element evolution of magmas.
Numerical modeling. Numerical modeling could provide a conceptual framework for
interpreting crystal-scale age data as well as trace-element data. For example, numerical models
have recently been developed that allow tracking of crystals that are treated individually rather
than as part of a continuum fluid (Ruprecht et al. 2008). These models can allow crystal growth
and interactions with the host magma through kinetic and trace-element partitioning laws, and
have the potential to provide synthetic crystal zoning profiles that are produced over a known
timeframe, which can be compared to data from real samples in order to understand what
scale of trace-element variation reflects magma-wide properties and what scale may primarily
reflect the local chemical environment of each crystal. By combining this information with
age data, we could potentially examine the timescales over which different kinds of processes
may be recorded in crystals.
538 Cooper & Reid
Kinetics of crystal growth. The effects of rapid crystallization may affect the way that
parent and daughter are incorporated into crystals, and thus the calculated ages. It is important
to understand these effects, and especially the circumstances under which these effects are
likely to be important, in order to have confidence in interpretations based on age data and
trace-element data.
Final thoughts
Overall, the geochemical/petrological community is moving toward a more integrated
approach to the study of crystals in magmas, and the records of magmatic processes contained
within them. This volume summarizes a wide variety of approaches to understanding crystal
records, and each chapter contains examples of ways in which multiple data types may be
combined, yet these integrated studies are just beginning to show their full potential. Magmatic
systems are themselves complex, and this complexity is reflected in the crystals contained within
them. Such complexity demands a comprehensive approach in order to make progress, and the
next decade promises to yield many more insights into the dynamics of magmatic systems.
aCKnOWLedGMentS
This work was supported in part by NSF grants EAR 0714455 and EAR 0649295 to KMC
and EAR 0538309 to MRR. We thank Mark Reagan, Axel Schmitt, and Simon Turner for
constructive reviews which improved the manuscript. We also thank editors Keith Putirka and
Frank Tepley for further editorial comments and suggestions, and for their efforts in putting
together this volume.
reFerenCeS
Allègre CL (1968) 230Th dating of volcanic rocks: a comment. Earth Planet Sci Lett 5:209-210
Armienti P (2008) Decryption of igneous rock textures: crystal size distribution tools. Rev Mineral Geochem
69:623-649
Bachmann O, Charlier BLA, Lowenstern JB (2007) zircon crystallization and recycling in the magma chamber
of the rhyolitic Kos Plateau Tuff (Aegean arc). Geology 35:73-76
Bachmann O, Dungan MA, Lipman PW (2002) The Fish Canyon magma body, San Juan volcanic field,
Colorado: rejuvenation and eruption of an upper-crustal batholith. J Petrol 43:1469-1503
Bacon CR, Adami LH, Lanphere MA (1989) Direct evidence for the origin of low-18O silicic magmas: quenched
samples of a magma chamber’s partially-fused granitoid walls, Crater Lake, Oregon. Earth Planet Sci Lett
96:199-208
Bacon CR, Lowenstern JB (2005) Late Pleistocene granodiorite source for recycled zircon and phenocrysts in
rhyodacite lava at Crater Lake, Oregon. Earth Planet Sci Lett 233:277-293
Bacon CR, Persing HM, Wooden JL, Ireland TR (2000) Late Pleistocene granodiorite beneath Crater Lake
caldera, Oregon, dated by ion microprobe. Geology 28:467-470
Bacon CR, Sisson TW, Mazdab FK (2007) Young cumulate complex beneath Veniaminof caldera, Aleutian arc,
dated by zircon in erupted plutonic blocks. Geology 35:491–494
Baldwin SL, Ireland TR (1995) A tale of two eras; Pliocene-Pleistocene unroofing of Cenozoic and late Archean
zircons from active metamorphic core complexes, Solomon Sea, Papua New Guinea. Geology 23:1023-
1026
Ball L, Sims KWW, Schwieters J (2008) Measurement of 234U/238U and 230Th/232Th in volcanic rocks using the
Neptune MC-ICP-MS. J Anal At Spectrom 23:173-180
Balsley SD, Gregory RT (1998) Low-18O silicic magmas: why are they so rare? 162:123-136
Bea F, Pereira MD, Stroh A (1994) Mineral/leucosome trace-element partitioning in a peraluminous migmatite
(a laser-ablation-ICP-MS study). Chemical Geology 117:291-312
Belousova EA, Griffin WL, O’Reilly SY, Fisher NI (2002) Igneous zircon: trace element composition as an
indicator of source rock type. Contrib Mineral Petrol 143:602-622
Bernal J-P, Eggins SM, McCulloch MT (2005) Accurate in situ 238U–234U–234Th–230Th analysis of silicate
glasses and iron oxides by laser-ablation MC-ICP-MS. J Anal At Spectrom 20:1240-1249
Bindeman I (2008) Oxygen isotopes in mantle and crustal magmas as revealed by single crystal analysis. Rev
Mineral Geochem 69:445-478
Uranium-series Crystal Ages 539
Bindeman IN (2003) Crystal sizes in evolving silicic magma chambers. Geology 31:367-370
Bindeman IN, Davis AM, Drake MJ (1998) Ion microprobe study of plagioclase-basalt partition experiments at
natural concentration levels of trace elements. Geochim Cosmochim Acta 62:1175-1193
Bindeman IN, Schmitt AK, Valley JW (2006) U–Pb zircon geochronology of silicic tuffs from the Timber
Mountain/Oasis Valley caldera complex, Nevada: rapid generation of large volume magmas by shallow-
level remelting. Contrib Mineral Petrol 152:649-665
Bindeman IN, Valley JW (2000) Formation of low-18O rhyolites after caldera collapse at Yellowstone, Wyoming,
USA. Geology 28:719-722
Bindeman IN, Valley JW (2001) Low-δ18O rhyolites from Yellowstone: magmatic evolution based on analyses
of zircons and individual phenocrysts. J Petrol 42:1491-1517
Bindeman IN, Valley JW, Wooden JL, Persing HM (2001) Postcaldera volcanism: in situ measurement of
U–Pb age and oxygen isotope ratio in Pleistocene zircons from Yellowstone caldera. Earth Planet Sci Lett
189:197-206
Bindeman IN, Valley JW (2002) Oxygen isotope study of the Long Valley magma system, California: isotope
thermometry and convection in large silicic magma bodies. Contrib Mineral Petrol 144:185-205
Bindeman IN, Valley JW (2003) Rapid generation of both high and low-delta O-18 large-volume silicic magmas
at the Timber Mountain/Oasis Valley caldera complex Nevada. Geol Soc Am Bull 115:581-595
Bindeman IN, Fu B, Kita N, Valley JW (2008) Origin and evolution of Yellowstone silicic magmatism based on
ion microprobe analysis of isotopically-zoned zircons. J Petrol 49:163-193
Black S, Macdonald R, Barreiro B, Dunkley PN, Smith M (1998a) Open system alkaline magmatism in northern
Kenya: evidence from U-series disequilibria and radiogenic isotopes. Contrib Mineral Petrol 131:364-
378
Black S, Macdonald R, DeVivo B, Kilburn CRJ, Rolandi G (1998b) U-series disequilibria in young (A.D. 1944)
Vesuvius rocks: Preliminary implications for magma residence times and volatile addition. J Volcanol
Geotherm Res 82:97-111
Black S, Macdonald R, Kelly MR (1997) Crustal origin for peralkaline rhyolites from Kenya: Evidence from
U-series disequilibria and Th-isotopes. J Petrol 38:277-297
Blundy J, Wood B (1994) Prediction of crystal-melt partition coefficients from elastic moduli. Nature 372:452-
454
Blundy J, Wood B (2003) Mineral-melt partitioning of uranium, thorium and their daughters. Rev Mineral
Geochem 52:59-123
Bourdon B, Henderson GM, Lundstrom CC, Turner SP (eds) (2003a) Uranium-Series Geochemistry. Reviews in
Mineralogy and Geochemistry Vol. 52. Mineralogical Society of America. Washington, DC
Bourdon B, Turner S, Henderson GM, Lundstrom CC (2003b) Introduction to U-series geochemistry. Rev
Mineral Geochem 52:59-123
Bourdon B, Worner G, zindler A (2000) U-series evidence for crustal involvement and magma residence times
in the petrogenesis of Parinacota volcano, Chile. Contrib Mineral Petrol 139:458-469
Bourdon B, zindler A, Worner G (1994) Evolution of the Laacher See magma chamber: Evidence from SIMS
and TIMS measurements of U-Th disequilibria in minerals and glasses. Earth Planet Science Lett 126:75-
90
Brown SJA, Fletcher IR (1999) SHRIMP U–Pb dating of the preeruption growth history of zircons from the
340 ka Whakamaru ignimbrite, New zealand: evidence for >250 ky magma residence times. Geology
27:1035-1038
Brown K, Carter C, Fohey N, Wooden J, Yi K, Barth A (2004) A study of the origin of rhyolite at mid-ocean
ridges; geochronology and petrology of trachydacite and rhyolite from Salton Sea, California, and
Torfajokull, Iceland. Geol Soc Am Abstr Programs 36:79
Brown SJA, Smith RT (2004) Crystallisation history and crustal inheritance in a large silicic magma system:
206
Pb/238U ion probe dating of zircons from the 1.2 Ma Ongatiti ignimbrite, Taupo Volcanic zone. J
Volcanol Geotherm Res 135:247-257
Chamberlain KR, Harrison TM, Schmitt AK, Heaman LM, Swapp SM, Khudoley AK, Sears JW, Prokopiev AV
(2007) In situ micro-baddeleyite U-Pb dating method: an age from any dike. Geol Soc Am Abstr Programs
39:97
Charlier B, zellmer G (2000) Some remarks on U-Th mineral ages from igneous rocks with prolonged
crystallization histories. Earth Planet Sci Lett 183:457-469
Charlier BLA, Peate DW, Wilson CJN, Lowenstern JB, Storey M, Brown SJA (2003) Crystallization ages in
coeval silicic magma bodies: 238U-230Th disequilibrium evidence from the Rotoiti and Earthquake Flat
eruption deposits, Taupo Volcanic zone, New zealand. Earth Planet Sci Lett 206:441-457
Charlier BLA, Wilson CJN, Lowenstern JB, Blake S, Van Calsteren PW, Davidson JP (2005) Magma generation
at a large, hyperactive silicic volcano (Taupo, New zealand) revealed by U-Th and U-Pb systematics in
zircons. J Petrol 46:3-32
Cherniak DJ, Watson EB (2003) Diffusion in zircon. Rev Mineral Geochem 53:113-143
Cherniak DJ (2002) Ba diffusion in feldspar. Geochim Cosmochim Acta 66:1641-1650
540 Cooper & Reid
Christensen JN, DePaolo DJ (1993) Time scales of large volume silicic magma systems: Sr isotopic systematics
of phenocrysts and glass from the Bishop Tuff, Long Valley, California. Contrib Mineral Petrol 113:100-
114
Christensen JN, Halliday AN (1996) Rb-Sr ages and Nd isotopic compositions of melt inclusions from the
Bishop Tuff and the generation of silicic magma. Earth Planet Sci Lett 144:547-561
Condomines M (1997) Dating recent volcanic rocks through 230Th-238U disequilibrium in accessory minerals:
Example of the Puy de Dôme (French Massif Central). Geology 25:375-378
Cooper KM, Donnelly CT (2008) 238U-230Th-226Ra disequilibria in dacite and plagioclase from the 2004-2005
eruption of Mount St. Helens In: A Volcano Rekindled: The First Year of Renewed Eruption at Mount St.
Helens, 2004-2006. Sherrod DR, Scott WE, Stauffer PH (eds) US Geological Survey Professional Paper
1750 (in press)
Cooper KM, Reid MR (2003) Re-examination of crystal ages in recent Mount St. Helens lavas: Implications for
magma reservoir processes. Earth Planet Sci Lett 213:149-167
Cooper KM, Reid MR, Murrell MT, Clague DA (2001) Crystal and magma residence at Kilauea Volcano,
Hawaii: 230Th-226Ra dating of the 1955 east rift eruption. Earth Planet Sci Lett 184:703-718
Cooper KM, Sims KWW, Eiler JM (2005) Time scales of assimilation and mixing at Krafla volcano, Iceland.
EOS: Trans AGU v. 86: Fall Meet Suppl, Abstract V13B-0539
Costa F, Dohmen R, Chakraborty S (2008) Time scales of magmatic processes from modeling the zoning
patterns of crystals. Rev Mineral Geochem 69:545-594
Crowley JL, Schoene B, Bowring SA (2007) U-Pb dating of zircon in the Bishop Tuff at the millennial scale.
Geology 35:1123-1126
Dalrymple GB, Grove M, Lovera OM, Harrison TM, Hulen JB, Lanphere MA (1999) Age and thermal history
of The Geysers plutonic complex (felsite unit), Geysers geothermal field, California; a 40Ar/39Ar and U-Pb
study. Earth Planet Sci Lett 173:285-298
Davis DW, Williams IS, Krogh TE (2003) Historical Development of zircon Geochronology. Rev Mineral
Geochem 53:145-181
Davies GR, Halliday AN (1998) Development of the Long Valley rhyolitic magma system: strontium and
neodymium isotope evidence from glasses and individual phenocrysts. Geochimica Cosmochimica Acta
62:3561-3574
Donnelly CT, Cooper KM (2006) Comparison of U-Th-Ra disequilibria in multiple crystal populations in lava
from the current eruption of Mt. St. Helens. EOS: Trans AGU v. 87: Fall Meet Suppl, Abstract V54B-03
Gardner JE, Layer PW, Rutherford MJ (2002) Phenocrysts versus xenocrysts in the youngest Toba Tuff:
Implications for the petrogenesis of 2800 km3 of magma. Geology 30:347-350
Garrison J, Davidson J, Reid M, Turner S (2006) Source versus differentiation controls on U-series disequilibria:
Insights from Cotopaxi Volcano, Ecuador. Earth Planet Sci Lett 244:548-565
Goldstein SJ, Stirling CH (2003) Techniques for measuring uranium-series nuclides: 1992-2002. Rev Mineral
Geochem 52:23-57
Hammer JE (2008) Experimental studies of the kinetics and energetic of magma crystallization. Rev Mineral
Geochem 69:9-59
Harford CL, Sparks RSJ (2001) Recent remobilisation of shallow-level intrusions on Montserrat revealed by
hydrogen isotope composition of amphiboles. Earth Planet Sci Lett 185:285-297
Heath E, Turner SP, Macdonald R, Hawkesworth CJ, van Calsteren P (1998) Long magma residence times at an
island arc volcano (Soufriere, St. Vincent) in the Lesser Antilles: evidence from 238U-230Th isochron dating.
Earth Planet Sci Lett 160:49-63
Heumann A, Davies GR (2002) U-Th disequilibrium and Rb-Sr age constraints on the magmatic evolution of
peralkaline rhyolites from Kenya. J Petrol 43:557-577
Hoskin PWO (2005) Trace element composition of hydrothermal zircon and the alteration of Hadean zircon
from the Jack Hills, Australia. Geochim Cosmochim Acta 69:637-648
Hoskin PWO, Ireland TR (2000) Rare earth element chemistry of zircon and its use as a provenance indicator.
Geology 28:627-630
Hoskin PWO, Schaltegger U (2003) The composition of zircon and igneous and metamorphic petrogenesis. Rev
Mineral Geochem 53:27-62
Ireland TR, Williams IS (2003) Considerations in zircon geochronology by SIMS. Rev Mineral Geochem
53:215-241
Ivanovich M, Harmon RS (eds) (1992) Uranium-series Disequilibria: Applications to Earth, Marine, and
Environmental Sciences. Oxford University Press, Oxford
Jicha BR, Singer BS, Beard BL, Johnson CM (2005) Contrasting timescales of crystallization and magma
storage beneath the Aleutian Island arc. Earth Planet Sci Lett 236:195-210
Jicha BR, Singer BS, Beard BL, Johnson CM, Moreno-Roa H, Naranjo JA (2007) Rapid magma ascent and
generation of 230Th excesses in the lower crust at Puyehue–Cordón Caulle, Southern Volcanic zone, Chile.
Earth Planet Sci Lett 255:229-242
Uranium-series Crystal Ages 541
Jicha BR, Johnson CM, Hildreth W, Beard BL, Hart GL, Shirey SB, Singer BS (2008) Deciphering crust vs.
mantle inputs and the timescales of magma genesis at Mount Adams using 238U-230Th disequilibria and Os
isotopes. Earth Planet Sci Lett (in press)
Keller J (1969) Origin of rhyolites by anatectic melting of granitic crustal rocks; the example of rhyolitic pumice
from the island of Kos (Aegean Sea). Bull Volcanol 33:924-959
Kemp AIS, Hawkesworth CJ, Paterson BA, Kinny PD (2006) Episodic growth of the Gondwana supercontinent
from hafnium and oxygen isotopes in zircon. Nature 439:580-583
Kent AJR (2008) Melt inclusions in basaltic and related volcanic rocks. Rev Mineral Geochem 69:273-331
Kigoshi K (1967) Ionium dating of igneous rocks. Science 156:932-934
Klemetti EW, Cooper KM (2007) Cryptic young zircon and young plagioclase in the Kaharoa Rhyolite,
Tarawera, New zealand: Implications for crystal recycling in magmatic systems. EOS Transactions AGU,
Fall Meet Suppl, Abstract V41F-03
Kosler J, Sylvester PJ (2003) Present trends and the future of zircon in geochronology: laser ablation ICPMS.
Rev Mineral Geochem 53:243-275
LaTourrette T, Hervig RL, Holloway JR (1995) Trace element partitioning between amphibole, phlogopite, and
basanite melt. Earth Planet Sci Lett 135:13-30
Lowenstern JB, Charlier BLA, Clynne MA, Wooden JL (2006) Extreme U–Th disequilibrium in rift-related
basalts, rhyolites and granophyric granite and the timescale of rhyolite generation, intrusion and
crystallization at Alid Volcanic Center, Eritrea. J Petrol 47:2105–2122
Lowenstern JB, Persing HM, Wooden JL, Lanphere M, Donnelly-Nolan J, Grove TL (2000) U-Th dating of
singe zircons from young granitoid xenoliths: new tools for understanding volcanic processes. Earth Planet
Sci Lett 183:291-302
Ludwig KR (1977) Effect of initial radioactive-daughter disequilibrium on U-Pb isotope apparent ages of young
minerals. J Res U. S. Geol Survey 5:663–667
Luhr JF, Carmichael ISE, Varekamp JC (1984) The 1982 eruptions of El Chichon Volcano, Chiapas, Mexico:
mineralogy and petrology of the anhydrite-bearing pumices. J Volcanol Geotherm Res 23:69-108
Mahood GA (1990) Reply to comment on “evidence for long residence times of rhyolitic magma in the Long
Valley magmatic system: the isotopic record in precaldera lavas of Glass Mountain”. Earth Planet Sci Lett
99:395-399
Mahood GA, Stimac JA (1990) Trace-element partitioning in panterllerites and trachytes. Geochim Cosmochim
Acta 54:2257-2276
Miller SA, Burnett DS, Asimow PD, Phinney DL, Hutcheon ID (2007a) Experimental study of radium
partitioning between anorthite and melt at 1 atm. Am Mineral 92:1535-1538
Miller JS, Matzel JEP, Miller CF, Burgess SD, Miller RB (2007b) zircon growth and recycling during the
assembly of large, composite arc plutons. J Volcanol Geotherm Res 167:282-299
Miller JS, Wooden JL (2004) Residence resorption and recycling of zircons in Devils Kitchen Rhyolite Coso
Volcanic Field, California. J Petrol 45:2155-2170
Moeller A, Hellebrand E, Whitehouse M, Cannat M (2006) Trace elements in young oceanic zircons. EOS
Transactions, American Geophysical Union 87: Suppl. 26, Dec. 2006
Nakada S, Bacon CR, Gartner AE (1994) Origin of phenocrysts and compositional diversity in pre-Mazama
rhyodacite lavas, Crater Lake, Oregon. J Petrol 35:127-162
Parrish RR, Noble SR (2003) zircon U-Th-Pb geochronology by isotope dilution-thermal ionization mass
spectrometry (ID-TIMS). Rev Mineral Geochem 53:183-213
Pietruszka AJ, Carlson RW, Hauri EH (2002) Precise and accurate measurement of 226Ra–230Th–238U disequilibria
in volcanic rocks using plasma ionization multicollector mass spectrometry. Chem Geol 188:171-191
Pilot J, Werner C-D, Haubrich F, Baumann N (1998) Palaeozoic and Proterozoic zircons from the Mid-Atlantic
Ridge. Nature 393:676-679
Ramos FC, Tepley FJ III (2008) Inter- and intracrystalline isotopic disequilibria: techniques and applications.
Rev Mineral Geochem 69:403-443
Reagan M, Tepley III FJ, Gill J, Wortel M, Hartman B (2005) Rapid time scales of basalt to andesite differentiation
at Anatahan volcano, Mariana Islands. J Volcanol Geotherm Res 146:171-183
Reagan MK, Cooper KM, Pallister JS, Thornber CR, Wortel M (2008) Timing of degassing and plagioclase
growth in lavas erupted from Mount St. Helens, 2004-2005, from 210Po-210Pb-226Ra disequilibria. In: A
volcano rekindled: the first year of renewed eruption at Mount St. Helens, 2004–2006. Sherrod DR, Scott
WE, Stauffer PH (eds) USGS Professional Paper 1750 (in press)
Reagan MK, Tepley III FJ, Gill JB, Wortel M, Garrison J (2006) Timescales of degassing and crystallization
implied by 210Po-210Pb-226Ra disequilibria for andesitic lavas erupted from Arenal volcano. J Volcanol
Geotherm Res 157:135-146
Reagan MK, Volpe AM, Cashman KV (1992) 238U- and 232Th-series chronology of phonolite fractionation at
Mount Erebus, Antarctica. Geochim Cosmochim Acta 56:1401-1407
542 Cooper & Reid
Reid MR (2003) Timescales of magma transfer and storage in the crust In: The Crust. Rudnick RL (ed) Elsevier,
Oxford, pp 167-193
Reid MR (2008) How long does it take to supersize an eruption? Elements 4:23-28
Reid MR, Coath CD (2000) In situ U-Pb ages from zircons from the Bishop Tuff: No evidence for long crystal
residence times. Geology 28:443-446
Reid MR, Coath CD, Harrison TM, McKeegan KD (1997) Prolonged residence times for the youngest rhyolites
associated with Long Valley Caldera: 230Th-238U ion microprobe dating of young zircons. Earth Planet Sci
Lett 150:27-39
Renne PR, Swisher CC, Deino AL, Karner DB, Owens T, DePaolo DJ (1998) Intercalibration of standards,
absolute ages and uncertainties in 40Ar/39Ar dating. Chem Geol 145:117–152
Rogers NW, Evans PJ, Blake S, Scott SC, Hawkesworth CJ (2004) Rates and timescales of fractional
crystallization from U-238-Th-230-Ra-226 disequilibria in trachyte lavas from Longonot volcano, Kenya.
J Petrol 45:1747-1776
Ruprecht P, Bergantz GW, Dufek J (2008) Modeling of gas-driven magmatic overturn: Tracking of
phenocryst dispersal and gathering during magma mixing. Geochemistry Geophysics Geosystems 9: doi
10.1029/2008GC002022
Rutherford MJ (2008) Magma ascent rates. Rev Mineral Geochem 69:241-271
Saal AE, Van Orman JA (2004) The Ra-226 enrichment in oceanic basalts: Evidence for melt-cumulate diffusive
interaction processes within the oceanic lithosphere. Geochem Geophys Geosys 5:2003GC000620
Sano Y, Tsutsumi Y, Terada K, Kaneoka I (2002) Ion microprobe U-Pb dating of quaternary zircon: implication
for magma cooling and residence time. J Volcanol Geotherm Res 117:285-296
Sarna-Wojcicki AM, Pringle MS, Wijbrans J (2000) New 40Ar/39Ar age of the Bishop Tuff from multiple sites
and sediment rate calibration for the Matuyama-Brunhes boundary. J Geophys Res 105:21,431–21,443
Schaefer SJ, Sturchio NC, Murrell MT, Williams SN (1993) Internal 238U-series systematics of pumice from the
November 13, 1985, eruption of Nevado del Ruiz, Colombia. Geochim Cosmochim Acta 577:1215-1219
Scharer U (1984) The effect of initial 230Th disequilibrium on young U-Pb ages: the Makalu case, Himalaya.
Earth Planet Sci Lett 67:191-204
Schmitt A (2006) Laacher See revisited: High-spatial-resolution zircon dating indicates rapid formation of a
zoned magma chamber. Geology 34:597-600
Schmitt AK (2007) Ion microprobe analysis of (231Pa)/(235U) and an appraisal of protactinium partitioning in
igneous zircon. Am Mineral 92:691-694
Schmitt AK, Grove M, Harrison TM, Lovera O, Hulen JB, Walters M (2003a) The Geysers-Cobb mountain
magma system California (Part 2): timescales of pluton emplacement and implications for its thermal
history. Geochim Cosmochim Acta 67:3443-3458
Schmitt AK, Grove M, Harrison TM, Lovera OM, Hulen J, Walters M (2003b) The Geysers–Cobb Mountain
Magma System, California (Part 1): U–Pb zircon ages of volcanic rocks, conditions of zircon crystallization
and magma residence times. Geochim Cosmochim Acta 67:3423-3442
Schmitt AK, Lindsay JM, de Silva S, Trumbull RB (2003c) U-Pb zircon chronostratigraphy of early-Pliocene
ignimbrites from La Pacana, north Chile: implications for the formation of stratified magma chambers. J
Volcanol Geotherm Res 120:43-53
Schmitt AK, Stockli DF, Hausback BP (2006) Eruption and magma crystallization ages of Las Tres Virgenes
(Baja California) constrained by combined 230Th/238U and (U–Th)/He dating of zircon. J Volcanol
Geotherm Res 158:281–295
Schmitt AK, Vazquez JA (2006) Alteration and remelting of nascent oceanic crust during continental rupture:
Evidence from zircon geochemistry of rhyolites and xenoliths from the Salton Trough, California. Earth
Planet Sci Lett 252:260-274
Schwartz JJ, John BE, Cheadle MJ, Miranda EA, Grimes GB, Wooden JL, Dick HJB (2005) Dating the growth
of oceanic crust at a slow-spreading ridge. Science 310:654-657
Simon JI, Reid MR (2005) The pace of rhyolite differentiation and storage in an ‘archetypical’ silicic magma
system, Long Valley, California. Earth Planet Sci Lett 235:123-140
Simon JI, Reid MR, Young ED (2007) Lead isotopes by LA-MCICPMS: tracking the emergence of mantle
signatures in an evolving silicic magma system. Geochim Cosmochim Acta 71:2014-2035
Simon JI, Renne PR, Mundil R (2008) Implications of pre-eruptive magmatic histories of zircons for U–Pb
geochronology of silicic extrusions. Earth Planet Sci Lett 266:182-194
Sims KWW, Blichert-Toft J, Fornari D, Perfit MR, Goldstein S, Johnson P, DePaolo DJ, Hart SR, Murrell MT,
Michael P, Layne G, Ball L (2003) Aberrant youth: Chemical and isotopic constraints on the young off-
axis lavas of the East Pacific Rise. Geochem Geophys Geosys 4:8621, doi: 8610.1029/2002GC000443
Sims KWW, Ackert Jr. RP, Ramos FC, Sohn RA, Murrell MT, DePaolo DJ (2007) Determining eruption ages
and erosion rates of quaternary basaltic volcanism from combined U-series disequilibria and cosmogenic
exposure ages. Geology 35:471-474
Uranium-series Crystal Ages 543
Sims KWW, Gill JB, Dosseto A, Hoffman DL, Lundstrom CC, Williams RW, Ball L, Tollstrup D, Turner S,
Prytulak J, Glessner JJG, Standish JJ, Elliott T (2008) An inter-laboratory assessment of the thorium
isotopic composition of synthetic and rock reference materials. Geostand Geoanal Res 32:65-91
Snyder DC, Widom E, Pietruszka AJ, Carlson RW, Schminke H-U (2007) Time scales of formation of zoned
magma chambers: U-series disequilibria in the Fogo A and 1563 A.D. trachyte deposits, Sao Miguel,
Azores. Chem Geol 239:138-155
Streck MJ (2008) Mineral textures and zoning as evidence for open system processes. Rev Mineral Geochem
69:595-622
Stirling CH, Lee D-C, Christensen JN, Halliday AN (2000) High-precision in situ 238U–234U–230Th isotopic
analysis using laser ablation multiple-collector ICPMS. Geochim Cosmochim Acta 64:3737–3750
Taddeucci A, Broecker WS, Thurber DL (1967) 230Th dating of volcanic rocks. Earth Planet Sci Lett 3:338-
342
Tepley III FJ, Lundstrom CC, Gill JB, Williams RW (2006) U-Th-Ra disequilibria and the time scale of fluid
transfer and andesite differentiation at Arenal volcano, Costa Rica (1968-2003). J Volcanol Geotherm Res
157:147-165
Tera F, Wasserburg GJ (1974) U–Th–Pb systematics on lunar rocks and inferences about lunar evolution and
the age of the moon, Proceedings from the 5th Lunar Science Conference (Supplement 5). Geochim
Cosmochim Acta 2:1571-1599
Tiepolo M, Vannucci R, Bottazzi P, Oberti R, zanetti A (2000) Partitioning of rare earth elements, Y, Th, U, and
Pb between pargasite, kaersutite, and basanite to trachyte melts: Implications for percolated and veined
mantle. Geochemistry Geophysics Geosystems 1: 2000GC000064
Turner S, Black S, Berlo K (2004) Pb210-Ra226 and Ra228-Th232 systematics in young arc lavas: implications for
magma degassing and ascent rates. Earth Planet Sci Lett 227:1-16
Turner S, George R, Jerram DA, Carpenter N, Hawkesworth C (2003a) Case studies of plagioclase growth and
residence times in island arc lavas from Tonga and the Lesser Antilles, and a model to reconcile discordant
age information. Earth Planet Sci Lett 214:279-294
Turner S, Foden J, George R, Evans P, Varne R, Elburg M, Jenner G (2003b) Rates and processes of potassic
magma evolution beneath Sangeang Api Volcano, East Sunda Arc, Indonesia. J Petrol 44:491-515
Valley JW (2003) Oxygen isotopes in zircon. Rev Mineral Geochem 53:343-385
Van Orman JA, Grove TL, Shimizu N (1998) Uranium and thorium diffusion in diopside. Earth Planet Sci Lett
160:505-519
Van Orman JA, Grove TL, Shimizu N (2001) Rare earth element diffusion in diopside: Influence of temperature,
pressure and ionic radius, and an elastic model for diffusion in silicates. Contrib Mineral Petrol 141:687-
703
Van Orman JA, Saal AE, Bourdon B, Hauri EH (2006) Diffusive fractionation of U-series radionuclides during
mantle melting and shallow-level melt-cumulate interaction. Geochim Cosmochim Acta 70:4797-4812
Vazquez JA (2004) Time scales of silicic magma storage and differentiation beneath caldera volcanoes from
uranium-238-thorium-230 disequilibrium dating of zircon and allanite. PhD. thesis (unpubl), Earth and
Space Sciences, UCLA, pp 284
Vazquez, JA (2008) Allanite and chevkinite as absolute chronometers of rhyolite differentiation. Geol Soc Am
Abstr Programs 40:74
Vazquez JA, Reid MR (2002) Time scales of magma storage and differentiation of voluminous high-silica
rhyolites at Yellowstone caldera, Wyoming. Contrib Mineral Petrol 144:274-285
Vazquez JA, Reid MR (2004) Probing the accumulation history of the voluminous Toba magma. Science
305:991-994
Vazquez JA, Shamberger PJ, Hammer JE (2007) Plutonic xenoliths reveal the timing of magma evolution at
Hualalai and Mauna Kea, Hawaii. Geology 35:695–698
Volpe AM (1992) 238U-230Th-226Ra disequilibrium in young Mt. Shasta andesites and dacites. J Volcanol
Geotherm Res 53:227-238
Volpe AM, Hammond PE (1991) 238U-230Th-226Ra disequilibria in young Mount St. Helens rocks: time constraint
for magma formation and crystallization. Earth Planet Sci Lett 107:475-486
Watson EB (1996a) Surface enrichment and trace-element uptake during crystal growth. Geochim Cosmochim
Acta 60:5013-5020
Watson EB (1996b) Dissolution, growth and survival of zircons during crustal fusion: kinetic principles,
geological models and implications for isotopic inheritance. Trans Royal Soc Edinburgh Earth Sci 87:43-
56
Watson EB, Harrison TM (1983) zircon saturation revisited: temperature and compositional effects in a variety
of crustal magma types. Earth Planet Sci Lett 64:295-304
Watson EB, Liang Y (1995) A simple model for sector zoning in slowly grown crystals; implications for growth
rate and lattice diffusion, with emphasis on accessory minerals in crustal rocks. Am Mineral 80:1179-
1187
544 Cooper & Reid
Watson EB, Wark DA, Thomas JB (2006) Crystallization thermometers for zircon and rutile. Contrib Mineral
Petrol 151:413-433
zellmer G, Turner S, Hawkesworth C (2000) Timescales of destructive plate margin magmatism: new insights
from Santorini, Aegean volcanic arc. Earth Planet Sci Lett 174:265-281
zellmer GF, Blake S, Vance D, Hawkesworth C, Turner S (1999) Plagioclase residence times at two island arc
volcanoes (Kameni Islands, Santorini, and Soufriere, St. Vincent) determined by Sr diffusion systematics.
Contrib Mineral Petrol 136:345-357
Reviews in Mineralogy & Geochemistry
Vol. 69 pp. 545-594, 2008 14
Copyright © Mineralogical Society of America
time Scales of Magmatic Processes from
Modeling the Zoning Patterns of Crystals
Fidel Costa
Institut de Ciencies de la Terra Jaume Almera, CSIC
c/ LLuis Sole i Sabaris s/n
Barcelona 08028, Spain
[email protected]
ralf dohmen, Sumit Chakraborty
Institut für Geologie, Mineralogie und Geophysik
Ruhr Universität Bochum
D-44780 Bochum, Germany
[email protected]; [email protected]
IntrOduCtIOn
The advent of polarized light microscopy in the middle of the 19th century allowed
mineralogists and petrologists interested in igneous rocks to recognize the widespread
occurrence of fine-scale heterogeneities in the optical properties of minerals (e.g., see Young
2003 for details). The interpretation of mineral zoning patterns as archives of magmatic
processes has been with us for some time (e.g., Larsen et al. 1938; Tomkeieff 1939). The
development of the electron microprobe in the 1960’s allowed mineral zoning profiles to be
quantitatively analyzed and modeled (e.g., Bottinga et al. 1966; Moore and Evans 1967).
Enhanced textural observations permitted recognition of many kinds of detailed structures
during the growth and dissolution of magmatic minerals (Fig. 1; e.g., Anderson 1983; Pearce
and Kolisnik 1990), and these processes were also explored using experimental and numerical
models (e.g., Albarède and Bottinga 1972; Lofgren 1972; Kirkpatrick et al. 1976; Loomis
1982). Major element zoning patterns are routinely measured, and with the arrival of the ion
microprobe, the identification of heterogeneous distribution of trace elements opened the
window for more realistic and sophisticated scenarios and models (Fig. 1; e.g., Kohn et al.
1989; Blundy and Shimizu 1991; Singer et al. 1995). We now measure abundances of naturally
occurring isotopes at the scale of tens of micrometers (e.g., Davidson et al. 2007a; Ramos and
Tepley 2008) and this allows in situ dating of crystals (e.g., Cooper and Reid 2008).
Major and trace element or isotopic zoning in igneous minerals has been used to unravel
the details of a large range of processes: from magma mixing, to magma degassing, magma
transport, and fractionation (e.g., Blundy and Shimizu 1991; Singer et al. 1995; Umino and
Horio 1998; Ginibre et al. 2007; Davidson et al. 2007b; Streck 2008). Crystal stratigraphy has
also been used to construct lineages and genealogies between different crystals from the same
or from different geological units (Wallace and Bergantz 2002). zoning in crystals is also im-
portant because it may record the paths to irreversible magma mixing which cannot be inferred
from studies of glasses alone (but see for example Perugini et al. 2004; Petrelli et al. 2006).
The advances in the spatial resolution and precision of in situ measurements of mineral
compositions have also been beneficial to the experimental determination of the rates at which
elements diffuse through crystals. The availability of large data sets of zoning patterns of
1529-6466/08/0069-0014$05.00 DOI: 10.2138/rmg.2008.69.14
546 Costa, Dohmen & Chakraborty
(a)
100 mm
(b)
100
80 Pl2n 2000
An mol%
Sr ppm
60 1500
40
1000
20 rim core
0 500
600 200
Mg ppm
150
Ba ppm
400
100
200
50
0 0
4000
3000
Fe* ppm
K ppm
3000
2000
2000
1000
1000
0
200 12
Ti ppm
La ppm
150 8
100
50 4
0 0
0 200 400 600 800 0 200 400 600 800
distance from the rim mm distance from the rim mm
Figure 1. (a) Nomarski DIC image of a plagioclase crystal from a gabbroic xenolith of Volcán San Pedro
(Costa et al. 2002). The crystal shows a well-defined core and a rim separated by an oscillatory-zoned
interval. Ion microprobe analysis pits and the trace of the electron microprobe traverse between the pits are
evident. (b) Major (An = anorthite mol%) and trace element profiles across this crystal. Note the abrupt
change in the An mol% which coincides with the position of the core-rim boundary (location marked by
a thin vertical line). Similar profiles are recorded by all trace elements except Mg. These relations were
interpreted by Costa et al. (2003) in other crystals of the same rock suite to indicate that Mg was partially-
equilibrated by diffusion between the core and the rim. Measurement and modeling of these trace element
profiles provides information about magmatic processes and their time scales. The data are from Costa
(2000), where more details can be found.
Time Scales of Magmatic Processes 547
various elements in natural crystals combined with reliable diffusion coefficients for these
same minerals have shown that diffusion within initially zoned minerals may partially or
completely erase such zoning. This has led to a new perspective wherein the zoning of crystals
is a record of magmatic variables and processes (e.g., Ginibre et al. 2007; Streck 2008), and
the extent of erasure is a measure of the time during which changes in these variables occur.
The first applications were on meteorites (e.g., Wood 1964; Goldstein and Short 1967; Short
and Goldstein 1967) and extraterrestrial igneous rocks from the Moon or Mars (e.g., Taylor
et al. 1973; Takeda et al. 1975; Walker et al. 1977), primarily to determine cooling rates of
these magmas. However, the fundamental temporal information that can be extracted from
compositional zoning is the duration of a particular process (e.g., see Turner and Costa 2007;
Chakraborty 2008 for recent reviews). Compositional zoning in a crystal may be used to obtain
time if: (1) one or more of the magmatic environmental variables such as temperature (T),
pressure (P), volatile fugacities, or the composition of the liquid change (creating zoning),
and (2) the diffusion of the elements of interest in the mineral is fast enough to partially erase
the zoning, but slow enough for the crystal to not fully equilibrate to the new set of conditions
(Fig. 2). Once a zoning pattern is established in a crystal it is possible under certain conditions
to obtain a time constraint and correlate this directly to the process that is recorded in the
crystal. Such constraints are most reliable when they are based on multiple determinations of
durations from a single traverse by modeling the diffusion of multiple elements (e.g., Costa
and Dungan 2005; Morgan and Blake 2006). As different elements diffuse at different rates,
it is possible to obtain time scales for more than one process from a single thin section. Using
multiple elements in a given crystal, multiple crystals of the same mineral from a single section,
and different minerals from the same thin section makes obtaining large numbers of time
determinations in a single thin section an opportunity that still needs to be fully exploited.
This chapter is designed to provide the reader with the knowledge necessary to go from
measuring a concentration profile in a mineral to obtaining the duration of a process by
applying the diffusion equations. This includes (i) presenting the basics of the equations that
govern diffusion, (ii) clarifying the steps involved in extracting temporal information using
(a) (b) (c)
P1,T1,X1 P2,T2,X2 P2,T2,X2
melt
+ time
concentration
concentration
concentration
Distance Distance Distance
Figure 2. (a) Schematic representation of a crystal growing from a melt with a given set of variables
(pressure: P, temperature: T, composition: X). A possible compositional traverse direction through the
crystal is marked with an arrow and the corresponding concentration profile is shown below. (b) A change
in one of the environmental variables is recorded in this case as a zone of different composition. (c) If
diffusion in the crystal occurs fast enough, the original zoning is modified with time. One can use such
information to constraint the processes and their durations.
548 Costa, Dohmen & Chakraborty
these equations (including providing a brief introduction to numerical methods and some
computer programs that use the method of finite differences to do such calculations), (iii)
describing the parameters that control diffusion coefficients (with an Appendix that explains
how the dependencies arise), (iv) considering how crystal growth and diffusion are combined
to create observed profiles, and (v) discussing to what extent the necessary information can be
obtained from nature. We have also compiled information on magmatic rates using the diffusion
modeling approach and we compare these to the information from radiogenic isotopes.
equatIOnS tO deSCrIBe dIFFuSIOn PrOCeSSeS
In this section we introduce some basic concepts related to the physical and mathematical
description of diffusion. Our presentation is oriented towards the application of the equations
to obtaining time scales of magmatic processes and in some cases we have preferred to discuss
the material from an intuitive perspective rather than in a rigorous manner. The reader is
referred to Watson (1994), Chakraborty (1995, 2008), Lasaga (1998), Ganguly (2002) and
Watson and Baxter (2007) for general aspects of diffusion in the Earth sciences, and to the
textbooks of Manning (1968), Flynn (1972), Philibert (1991), Allnatt and Lidiard (1993), and
Glicksman (2000) for a rigorous treatment of diffusion in solids.
diffusion, flux and constitutive laws
Diffusion can be defined, following Onsager (1945), as the relative motion of one or more
particles of a system relative to other particles of the same system. Random motions result
from the thermal energy contained in a system, and they occur in all materials at all times
at temperatures above the absolute zero. The diffusive motion of particles in air around us is
perceptible and large (we can smell things pretty quickly), whereas the diffusive motion of Mg
atoms in an olivine crystal sitting on our desk is imperceptibly slow. The important point is
that the existence of a concentration gradient or driving force of any kind is not necessary for
diffusion. What we are often interested in, however, is the quantity called flux, J. It is defined
as the amount of material (expressed as the number of moles, mass of a substance, volume of
a substance, and so forth) that crosses a plane of unit surface area in a given direction per unit
time. This is a vector quantity and has units such as mol·m−2·s−1. To obtain a net flux or directed
motion due to diffusion, a driving force that coaxes particles (e.g., atoms, molecules, ions) to
preferentially move in one direction is required. In transport of matter, the commonly applicable
driving force is a chemical potential gradient—matter flows in the direction of decreasing
chemical potential. Onsager (1945) assumed that the flux, Ji, of a diffusing component i, is
proportional to the gradient of the corresponding chemical potential, ∂µi/∂x (diffusion in one
dimension), and the two are related to each other via a proportionality constant, Li:
∂mi
J i = − Li (1)
∂x
This relationship is intuitive, but it is not very practical because chemical potential is not an
easily observable or measurable quantity. Chemical potentials can be related to measurable
concentrations by the definition of a reference state and its associated reference chemical
potential, µ°: µi = µ° + RT ln(γiCi), where γi is the activity coefficient. In the case of an ideal
solution or a diluted component (for exceptions see below), where the derivative ∂ln(γi)/∂x
is zero, this relationship can be shown to be equivalent (e.g., Schmalzried 1981; Ganguly
2002) to the empirical relationship observed by A. Fick about 100 years earlier (Fick 1855). It
connects the flux, Ji, to the concentration gradient of the species, (∂Ci/∂x):
∂Ci
J i = − Di (2)
∂x
Time Scales of Magmatic Processes 549
The proportionality constant, Di, is the diffusion coefficient of component i. This relationship,
often referred to as Fick’s first law, is analogous to other laws relating fluxes of different
kinds to relevant driving forces (e.g., electrical current to electrical potential gradient: Ohm’s
Law; heat flux to the gradient of temperature: Fourier’s Law) and is of practical use because
concentration is an easily measurable quantity. The connection to the chemical potential
gradient is the reason that Fick’s law is applicable to diffusion in such disparate media as
gases, liquids, and solids, even though the mechanisms of diffusion in these media are entirely
different. The homogenization of a concentration gradient, a process that is a major concern of
this chapter, is usually the result of a directed flux and not just random diffusion. Diffusion is
ubiquitous, directed flux occurs only in the presence of a suitable driving force.
It is possible to obtain other kinds of mathematical relationships between directed flux
and the measurable quantity, concentration. Some examples that will concern us in this chapter
are the following:
(i) Diffusion in an anisotropic medium (e.g., a non-cubic crystal):
∂Ci ∂C ∂C
J ix = − Dixx − Dixy i − Dixz i
∂x ∂y ∂z
∂C ∂C ∂C
J iy = − Diyx i − Diyy i − Diyz i (3)
∂x ∂y ∂z
∂Ci ∂C ∂C
J iz = − Dizx − Dizy i − Dizz i
∂x ∂y ∂z
Total flux, J ix, along a direction x is not just proportional to the concentration gradient along
this direction. It is the sum of three terms, each of which has the form of Fick’s law. This set
of equations can be simplified if the axes are transformed into the principal axis system of
the diffusion tensor (e.g., see Ganguly 2002). These principal axes may coincide with the
crystallographic symmetry axes in some crystal classes, as is the case for orthorhombic olivine.
Equation (3) can be expressed in matrix notation by the compact form Ji = −di·∇Ci.
(ii) Diffusion accompanied by growth or dissolution—the moving boundary problem:
∂Ci
J i = − Di + v ⋅ Ci ( 4)
∂x
Here the total flux, Ji, is the result of flux due to diffusion plus another flux due to the net growth
or dissolution of a crystal in the surrounding medium (e.g., melt) at a rate, v . This equation
describes flux of matter in a growing / dissolving crystal if the x-axis is defined to be such that
the origin (x = 0) is always at the surface of the growing / dissolving crystal.
(iii) Flux due to a combination of forces; e.g., chemical and electrical forces acting on
charged particles (e.g., ions) placed in a chemical as well as an electrical potential gradient:
∂m ∂ψ
J i = − Li i − zi ⋅ e ⋅ (5)
∂x ∂x
where zi is the charge of the diffusing component i, ∂ψ/∂x is the gradient of the mean electrical
potential and e is the elemental charge. This is particularly important with respect to diffusion in
ionic crystals, such as silicates, because it ensures that a charge imbalance does not develop due
to diffusion of ions. Such constraints do not apply to metallic substances. Care should be taken
when extrapolating the analysis of diffusion phenomena from such systems to silicates.
(iv) Diffusion in a one-dimensional, multicomponent system with n components, wherein
there is coupling between the fluxes of different species (e.g., when charge balance constraints
550 Costa, Dohmen & Chakraborty
need to be fulfilled in a system containing ionic particles). The coupling of charged particles
can be formally considered with the electric field imposed by the flux of the charged particles
as in Equation (5) (for more details see Lasaga 1979; Schmalzried 1981):
∂C1 ∂C ∂C
J1 = − D11 − D12 2 − D1n −1 n −1
∂x ∂x ∂x
∂C ∂C ∂C
J 2 = − D21 1 − D22 2 − D2 n −1 n −1 (6)
∂x ∂x ∂x
∂C ∂C ∂C
J n −1 = − Dn −11 1 − Dn −12 2 − Dn −1n −1 n −1
∂x ∂x ∂x
This kind of expression was developed intuitively by Onsager as an analogy to the expression
derived by Fick for simple binary systems. This set of equations can be written in matrix nota-
tion as J = −d·∂C/∂x, where J = {J1, J2, …, Jn−1} and C = {C1, C2, …, Cn−1}. The flux of the n-th
component directly follows from the flux of the other components by an additional constraint
such as mass balance in a closed system. Onsager demonstrated that this form of the equation
fulfils some basic requirements of thermodynamics, particularly when stated in terms of the
more physically reasonable (but practically difficult to measure) chemical potential gradients.
Experimental studies in different kinds of systems (ranging from aqueous solutions through sil-
icate crystals and melts) have shown that such functions are an adequate description of diffusion
in multi-component systems (e.g, Cooper 1965; Chakraborty et al. 1995; Liang et al. 1997).
Expressions that relate fluxes to various measurable parameters (concentration in the most
common cases of interest to us) are known as constitutive relations. Generally speaking, these
are relationships between fluxes and the forces that cause them, the connection between the
two being made through parameters that are characteristic of specific media, and which are
generally termed material constants. The diffusion coefficient, D, (Eqn. 2) and the constant
L (Eqn. 1) are examples of material constants. The mathematical form of these equations
describes the nature of the process; the values of the material constants are the causes of
different behavior in different systems. This chapter is concerned with the material constant
termed diffusion coefficient. It is necessary to ensure that D has the dimensions [L]2·[T]−1
(e.g., m2·s−1) irrespective of the choice of units for measuring concentration, flux etc.
time dependence: the continuity equation
The reader would be left wondering why the variable time, t, does not appear explicitly
in any of the equations shown above. Time is introduced by employing one of two universal
constraints that any process must follow—the requirement of mass balance (the 2nd being the
requirement that total entropy must increase in any natural process). In terms of fluxes, this
is stated most simply as: when a specified system is observed over any given length of time,
the difference between what comes in and what goes out is the net change in the amount
contained in that system. This is a loose statement of a rigorous mathematical law called the
Gauss Divergence Theorem (Fig. 3). The principle is the same that we use to balance our bank
accounts (Input – Output = Change of balance).
Consider the volume of a rectangular parallelepiped where each side has incrementally
small lengths e.g., dx, dy, dz. We observe the number of particles entering and leaving it through
its faces. To simplify matters, we can specify that the flux occurs only in the x-direction. If the
change of flux, J, in the x-direction is (∂J/∂x), then we can describe the flux at position x + dx
to the first order as:
∂J
J ( x + dx ) = J ( x ) + ⋅ dx ( 7)
∂x
Time Scales of Magmatic Processes 551
z
J(x) J(x + dx)
dz
y
dy
dx
x x + dx
Figure 3. Nomenclature for the mass balance of fluxes in a rectangular solid. Volume of a rectangular
parallelepiped with each side has incrementally small lengths e.g., dx, dy, dz. J(x) and J(x+dx) are the flux
parallel to the x-direction at positions x, and x+dx, respectively.
The total number of particles (atoms, ions) entering and leaving through the sides of the
parallelepiped is:
∂J
( J ( x + dx ) − J ( x ) ) ⋅ dy ⋅ dz = ⋅ dx ⋅ dy ⋅ dz (8)
∂x
The number of particles in the incremental volume element dx⋅dy⋅dz, defined as C(x,t), changes
to C(x,t+dt) over an infinitesimal time interval, dt (short time where the flux is effectively
constant) due to the flux imbalance:
∂J
C ( x, t + dt ) − C ( x, t ) = − ⋅ dt ( 9)
∂x
The minus sign follows from the fact that the concentration has to increase when J(x + dx) < J(x).
The time derivative of C is defined mathematically as the limit of (C ( x, t + dt ) − C ( x, t )) / dt when
dt goes to zero and Equation (9) can be also expressed in the following differential form:
∂C ( x, t ) ∂J
=− (10)
∂t ∂x
This derivation, based only on mass balance, is totally independent of the form of the consti-
tutive equation we have chosen. Equation (10) is the continuity equation and replacing J in
this equation using one of the expressions from the previous section allows us to obtain the
corresponding time dependent equation. The simplest form of this relationship, for a binary
exchange with a constant diffusion coefficient, is obtained by substituting Equation (2) on
Equation (10) yields:
∂C ( x, t ) ∂ 2C ( x , t )
=D (11a )
∂t ∂x 2
This is often described as the diffusion equation, or Fick’s second Law. Equation (11a) is valid
only when D is independent of C or x; otherwise, the diffusion equation takes the following
(expanded) form:
∂C ( x, t ) ∂D ∂C ( x, t ) ∂ 2C ( x, t )
= ⋅ + D⋅ (11b)
∂t ∂x ∂x ∂x 2
The combination of a constitutive equation related to fluxes with the continuity equation
leads to an equation such as (11a), which relates the change of concentrations in time to the
concentration gradients in the form of a partial differential equation. It describes the evolution
552 Costa, Dohmen & Chakraborty
of concentration, C, as a function of time, t, at different spatial coordinates, C(x,t). To obtain the
specific concentration distribution (= concentration profile), C(x), at any point of time t after the
initiation of diffusion, we need three additional equations for each medium (e.g., crystal, melt) in
which diffusion occurs. The first of these describes the initial condition (= concentration profile
in our case) from which diffusion commences. The remaining two equations provide additional
constraints that allow the concentrations at the boundaries of the diffusion system (= boundary
conditions) to be specified. The boundary conditions may be formulated by specifying either
the concentration or the flux at the boundaries of the system. If the system of interest is a crystal,
a boundary condition is needed to specify the concentration or flux evolution at its surface. If
this crystal is in equilibrium with a large body of melt at constant temperature, the concentration
at the surface of the crystal would remain constant and obey the conditions of equilibrium, at the
same time that diffusion strives to homogenize the concentration gradients in its interior. This
condition may be expressed as Cs = K where Cs is the concentration at the surface and K is a
constant. Various forms of initial and boundary conditions are discussed below.
Once the initial and boundary conditions have been specified, an analytical solution may
be derived in the form of a formula: C(x,t) = .......with the diffusion coefficient, features of the
initial concentration distribution, and the variables x and t appearing in some form on the right
hand side. If the diffusion coefficient and initial conditions are known, the concentration at any
position (x) and time (t) can be calculated. The analytical solutions for many different initial and
boundary conditions for different geometries can be found in Crank (1975) and Carslaw and
Jaeger (1986). These solutions are exact and calculations using these are fast, but they are limited
to simple shapes for the diffusion medium (e.g., sphere, slab) and relatively simple boundary
conditions (e.g., constant composition, variation according to some prescribed function). These
simplifications introduce approximations when modeling shapes and concentration variations
in real systems.
An alternative to analytical solutions (Appendix I) is to discretize the continuity equations
using numerical methods. This chapter is accompanied by two computer programs that illustrate
the implementation of these methods for a natural case: an Excel© spreadsheet (olivine_diffu_
excel.xls), and a Mathematica© notebook (olivine_diffu_mathema.nb). These can be found
on the internet at http://www.minsocam.org/MSA/RIM. For most cases of interest, speed is not
a problem for typical desktop computers, and the loss of accuracy (compared to analytical
solutions) is inconsequential. Major benefits of this approach are that realistic shapes of crystals,
variations in initial and boundary conditions, and the form of the continuity equations may
be implemented with a minimum demand on mathematical abilities. Numerical methods are
becoming the method of choice for modeling natural systems due to these advantages.
MOdeLInG naturaL CryStaLS
We follow a forward modeling approach to obtain information on time (= duration) by
modeling concentration gradients observed in natural crystals. Petrological constraints and
intuition are applied to infer an initial concentration distribution, and the suitable form of the
continuity equation. Equations (11a or 11b) are two possibilities, but others may be obtained
by combining Equation (10) with any one of Equations (3, 4, 5, or 6). Petrological analysis
should be employed to formulate suitable boundary conditions. The relevant questions are of
the type: Was the crystal closed to element exchange with its surroundings during diffusion?
And was the composition at its surface held constant during diffusion? A discretized form of
the suitable continuity equation (Appendix I), in combination with these initial and boundary
conditions, allows concentration profiles to be calculated for different durations of diffusion if
diffusion coefficients are known. The progressive evolution of the concentration profile from
the chosen initial condition is tracked, until a profile is obtained that matches the observed
Time Scales of Magmatic Processes 553
concentration profile or two-dimensional element map. The duration for which the calculated
profile matches the observed profile is the length of time during which the natural crystal
experienced diffusion, and this is the time scale of interest. Uncertainties arise from the choice
of various models and parameters as they do in any modeling approach. These include the
choice of the form of the continuity equation, initial and boundary conditions, and diffusion
coefficients (see section about uncertainties below for more details). The diffusion equations
formulated above typically hold for cases where temperature remains constant during diffusion,
but non-isothermal diffusion is common in nature. Some of the aspects discussed here have
been addressed by Ganguly (2002) and Chakraborty (2006) in the context of similar modeling
in metamorphic rocks.
The key parameter in such models is the diffusion coefficient. A detailed discussion of this
quantity follows below but for purposes of illustration in the examples that follow immediately
we use the following expression (Dohmen and Chakraborty 2007a,b) for the calculation of the
diffusion coefficient (in m2·s−1) for Fe-Mg diffusion in olivine parallel to [001]:
fO 2 3( X Fe − 0.1) 201000 + ( P − 10 5 ) ⋅ 7 ⋅ 10 −6
DFe-Mg = 10 −9.21 ⋅ ( )1/ 6
⋅ 10 exp −
10 −7 RT
where T is in Kelvin, P and fO2 (pressure and oxygen fugacity, respectively) are in Pascals, XFe
is the mole fraction of the fayalite component, and R is the gas constant in J·mol−1·K−1. Note
that the diffusion coefficient depends (a) exponentially on temperature and (b) on intensive
thermodynamic variables such as fO2. After discussing the modeling approaches, we discuss
how such dependencies arise in the section on Diffusion Coefficients. It is shown that such
expressions are now based on a microscopic understanding of diffusion mechanisms so that
uncertainties in our knowledge of diffusion coefficients have been considerably reduced and
time scales retrieved from diffusion modeling are more robust.
Initial and boundary conditions
There is no unique way of estimating the initial conditions, as this depends on the mineral
and the nature of the problem. Figure 4 shows some possible initial distributions that may be
associated with commonly observed chemical zoning in crystals: gradual (Fig. 4a), abrupt
(Fig. 4b), homogeneous (Fig. 4c), and oscillatory (Fig. 4d). In the figure are also shown how
such initial distributions evolve by diffusion (one profile at an arbitrary point of time in each
(a) (b)
Concentration
Concentration
Figure 4. End-member types of
initial zoning patterns used for
obtaining time scales (dotted lines)
and distribution after a given diffusion
Distance Distance time (solid lines). (a) Smooth zoning
that might be created during magma
(c) differentiation. (b) Abrupt zoning
(d) could be created during magma
mixing. (c) Homogenous initial
zoning where diffusion is driven by
Concentration
Concentration
later change of concentrations at the
boundary. (d) Oscillatory zoning.
Distance Distance
554 Costa, Dohmen & Chakraborty
case). The shapes of these profiles are good indicators of different initial distributions, although
inferences concerning initial conditions are not always straightforward. Some strategies that
have been employed are:
(1) Comparisons of zoning profiles for elements with very different diffusivities.
Unmodified or less modified profiles of slowly diffusing elements can give clues to
the initial distributions of rapidly diffusing elements. The profile shapes of the very
slowly diffusing major element compositions CaAl-NaSi (Grove et al. 1984) in zoned
plagioclase may be used to constrain the initial concentration profile (Costa et al.
2003) of a rapidly diffusing element such as Mg (LaTourrette and Wasserburg 1998).
This approach also works in sanidine (Morgan and Blake 2006), where the slowly
diffusing Ba may allow the initial distribution of faster Sr to be constrained and used
to obtain time scales. A similar framework is applicable to olivine if zoning of slowly
diffusing elements like P (Millman-Barris et al. 2008) can be used as a proxy for the
initial profile shapes of the much faster diffusing divalent cations (e.g., Fe-Mg, Ni,
Mn). This approach is powerful for distinguishing among the types of initial profiles
(Fig. 4), provided it can be established that (i) the slower and faster diffusing elements
had similar pre-diffusion distributions (not always the case, e.g., see the distribution
of REE vs. Fe, Mg etc. in garnets, Denison and Carlson 1997), (ii) diffusive coupling
(see below) has not modified the profile of the slower diffusing species.
(2) Using an arbitrary maximum concentration range for natural samples as an extreme
initial concentration to obtain a maximum time for the diffusion process (e.g.,
zellmer et al. 1999). A variant of this approach is to start with the arbitrary initial
case of oscillatory zoning and track the progressive elimination of short-wavelength
features by diffusion. The smallest surviving wavelength in the measured profile
is an indication of the upper limit of diffusion-duration that the crystal may have
experienced (Trepmann et al. 2004). A related method has been to use an extremely
sharp step-like change and assume that the measured gradient is due entirely to
diffusion (e.g., Fig. 4b; Morgan et al. 2004, 2006; Wark et al. 2007). Although all
of these choices of initial conditions would yield maximum times, it is particularly
useful for short diffusion profiles. These approaches have been used to model zoning
gradients in garnet, clinopyroxene, feldspars, and quartz.
(3) An initially homogeneous concentration profile is often a good assumption for
modeling olivine (e.g., Fig. 4c; Costa and Chakraborty 2004; Costa and Dungan
2005). The fact that robust time estimates have been obtained may reflect the growth
of unzoned crystals from a large reservoir of liquid, or that they equilibrated at a high
temperature before the final magmatic event that drove diffusion. The assumption
of simple initial profiles, with a posteriori verification through the robustness of
retrieved time scales, is therefore a useful variant.
(4) Smooth zoning profiles may be produced by magma fractionation during crystal
growth (Fig. 4a), with or without accompanying diffusion. Algorithms such as those
based on minimization of free energy allow such growth profiles to be simulated and
used as initial conditions for diffusion. This has been done for metamorphic minerals
(e.g., Loomis 1986; Spear et al. 1990) and we illustrate it below in an igneous system
using the MELTS algorithm (Ghiorso and Sack 1995).
All other factors remaining the same, the use of different initial profiles gives different
equilibration times. This effect is not always large. We have illustrated this with a series of
numerical models that use two extreme possibilities for the shape of the initial profile: one
is a homogenous crystal (e.g., Fig. 5a) and the other is a core with a rim overgrowth (e.g.,
Fig. 5b). The differences in retrieved time scales in these cases are less than a factor of 1.5 as
can be seen in Figure 5. Initial conditions should be constrained as well as possible, but the
uncertainties related to a lack of knowledge of this condition are not critical.
Time Scales of Magmatic Processes 555
100
Flux at boundary, (a)
Init
initially homogeneous 80
concentration
60
40
Equi
20
1.6 100
R (b) 80 E %
Flux at boundary, 1.4
Init 60
initially zoned, 80 1.2 40
20
complex history
concentration
1.0 0
0 50 100150200
60 time 0 50 100 150200
time
40
20
te
0
20 40 60 80
distance
Figure 5. Illustration of results from numerical calculations that show the effect of two very different
initial conditions on calculated times. Left panels show a hypothetical crystal surrounded by a matrix.
The arrow marks the position of the compositional traverse. The right panel shows the time evolution
of diffusion calculations for two crystals with different initial profiles at the same diffusion times. The
initial profile (Init) is shown as a dotted line, and the equilibrium concentration (Equi) in shown as dashed
line. (a) The initial concentration is homogeneous and diffusion is driven by the change in composition
at the boundary. (b) The initial concentration profile is defined by step-function zoning and diffusion is
driven a change in the boundary and by the original zoning profile. The inset at the right of the figure
compares the time evolution for the two initial profiles as measured by the percentage of equilibration E%
(= 100 − [100×{C-Equi}/{Init-Equi}]) at the center of the crystal. The crystal with the high homogenous
concentration (black line) takes longer to equilibrate than the one with an initial step (grey line). The inset
in the left side of the panel shows that the E% difference between the two initial profiles is at most a factor
of 1.5 as measured by the value of R (= [E% of crystal b]/ [E% of crystal a]). This means that if we are
able to constrain the maximum initial concentration, the precise shape of the initial profile introduces a
limited uncertainty. The composition at the boundary is considered to be constant with time (for cases with
changing conditions at the boundary see figures in next section).
The boundary conditions arise from the nature of exchange of the elements of interest
at the boundary of the crystal with its surrounding matrix. Concentration at the boundary is
defined if local and instantaneous equilibrium between the mineral surface and matrix occurs,
which commonly appears to be the case (but not always; e.g., Dohmen and Chakraborty 2003).
The boundary may be closed to exchange of matter (no flux) if the mineral is surrounded by
a phase where the element of interest does not partition significantly or if the diffusion rate of
the element in the surrounding matrix is much slower than in the mineral. Open boundaries
are expected to be far more common in magmatic systems where minerals are in diffusive
communication via melts. Open boundaries may result in the concentration at the rim of a
crystal remaining fixed (e.g., when the crystal sits in a large volume of melt in which the
concentration and diffusion rate of the element of interest is significant) or variable (as when
the melt reservoir has limited volume, or when the ambient conditions such as temperature,
556 Costa, Dohmen & Chakraborty
oxygen fugacity, etc. change on the relevant time scale). It is important to differentiate between
these scenarios because the times obtained by modeling a system with closed boundaries are
much shorter than those obtained with open boundaries (Fig. 6; see also Chakraborty and
Ganguly 1991 for some analytical solutions).
These considerations allow isothermal diffusion models to be evaluated, and these
are often adequate for volcanic systems. Such models yield the duration of time that a
heterogeneous crystal experienced high temperatures; i.e., prior to quenching by eruption.
Most studies on volcanic systems have used isothermal models. However, crystals may have
experienced thermal oscillations, such as those that may result from convection in a magma
reservoir or due to multiple thermal pulses (e.g., Singer et al. 1995). As long as there is no
overall heating or cooling trend, the results of modeling the duration of diffusion at a constant
intermediate temperature are the same as that when oscillations are taken into account (e.g.,
Lasaga and Jiang 1995). The next level of sophistication is to incorporate the effect of changing
temperature during diffusion. This is crucial for modeling plutonic systems with protracted
thermal histories.
100
Isolating boundary, (a)
Init
concentration
initially zoned 80
60 Equi
40
20
0
Flux at boundary, (b) 100
80
Init 60
initially zoned 80
concentration
40
20 E%
0
60 0 50 100150 200
time
40
Equi
20
0
0 20 40 60 80
distance
Figure 6. Examples of calculations to show the effects of using a closed vs. an open boundary, on time
scales that are obtained. The main abbreviations and symbols are those used in Figure 5. The evolution
of the concentration profiles for the two crystals with different boundary conditions but identical initial
profiles are shown for the same diffusion times. (a) The crystal is closed to flux with the surrounding
environment. (b) The crystal is open to flux and has a constant composition at the boundary. The inset at
the right of the figure compares the time evolution for the two initial profiles as measured by the percentage
of equilibration E% at the center of the crystal. A closed-system model reaches equilibrium much faster.
Incorrectly applying this case to nature could underestimate the time scales by more than one order of
magnitude, as shown in the inset where the percentage equilibration (%E) is plotted against time.
Time Scales of Magmatic Processes 557
non-isothermal diffusion
The diffusion models outlined above change in two important ways when non-isothermal
diffusion is considered. “Non-isothermal” refers to systems undergoing a progressive change in
temperature (e.g., heating, cooling) rather than modest oscillations about a mean temperature
(thermal cycling). To model diffusion in a non-isothermal system we need to consider that:
(i) diffusion coefficients change because they are strong functions of temperature, and (ii) the
boundary conditions (e.g., the concentrations at the rims of crystals) change as a consequence of
changing temperature. The magnitude of effect (i) is determined by the temperature dependence
of diffusion coefficients (activation energy, see below) whereas the magnitude of effect (ii) is
determined primarily by the enthalpy change of the exchange reaction between the mineral
and its matrix for the element of interest. The modal proportions of mineral and matrix and the
diffusion rates of the element in the two media influence the effect of (ii) as well.
The non-isothermal diffusion problem commonly may be reduced to an isothermal diffu-
sion problem. This is accomplished by identifying a “constant,” characteristic temperature of
diffusion, TCh, such that the duration obtained by modeling isothermal diffusion at this tempera-
ture is the same as the one that would be obtained from a non-isothermal model. Chakraborty
and Ganguly (1991) show that for diffusion processes with activation energies on the order of
200 kJ·mol−1, there is a simple relationship between the peak temperature attained in a variety
of non-isothermal histories, TPk, and the characteristic temperature: TPk ~ 0.95 TCh in Kelvin. If
it is known that a pluton cooled slowly from a peak temperature of 1100 °C, a diffusion calcula-
tion carried out at the constant characteristic temperature of 1030 °C would yield a very good
approximation of the duration of cooling to the temperature where diffusion effectively froze.
This simplification is a consequence of the fact that diffusion rates depend exponentially on
temperature. Much of the total diffusion during any thermal history occurs at temperatures very
close to the thermal peak.
A second consequence of this exponential dependence allows us to use diffusion as a tool
to measure time scales; diffusion effectively stops and the system (= concentration profiles
in our case) ‘freezes’ with falling temperature. Analytical solutions were derived for systems
undergoing cooling histories in a number of early works that made use of this behavior (Dodson
1973, 1976, 1986; Lasaga et al. 1977; Lasaga 1983). Dodson (1973) provided a formulation for
the closure temperature (Tc); i.e., the temperature at which a system ‘freezes’ with respect to
diffusion:
Q A Do RTc2
= ln − (12)
RTc Q s a2
Q is the activation energy for diffusion, Do is the pre-exponential factor, A is a geometric factor,
s is the cooling rate in the vicinity of the closure temperature, and a is the diffusing distance.
This equation applies to the mean concentration in a zoned crystal (e.g., Tc is really a mean
closure temperature), it only uses a single cooling rate, the reservoir of the crystal is infinite
in size, and diffusion within it is taken to be infinitely fast. It is also assumed that sufficient
diffusion has occurred to completely erase the initial composition. Ganguly and Tirone (1999,
2001) present formulations in which the mean closure temperature can be calculated for
minerals that still record their initial concentrations and these should be of wide applicability
to igneous systems. Comparable formulations are used to determine the closure temperatures
of radiogenic isotope systems, and thereby to determine to which thermal events the dates
obtained from these systems pertain. In the context of this manuscript it is relevant to point out
that one can also calculate closure temperature profiles (Dodson 1986; Ganguly and Tirone
1999, 2001), where the system closes earlier and at higher temperatures in the crystal core than
in the rim. Thus, there is not a unique closure temperature of a mineral and radiogenic system.
558 Costa, Dohmen & Chakraborty
It depends on the position and also on the precision of the measurements of the elements
or isotopes of interest (e.g., Charlier et al. 2007). These statements can be understood by
considering the diffusion equation that describes the in-growth of the radiogenic daughter
isotope (in one dimension; e.g., Albarède 1996):
∂C DA ∂ 2C DA
= D(t ) DA + λN o exp(−λt )
∂t ∂x 2
where CDA is the concentration of the daughter, No is the number of parent atoms present
initially, D(t)DA is the diffusion coefficient which depends on time through the thermal history,
but not on composition or position, λ is the decay constant, and x is distance.
Numerical methods allow us to explore variations that are free from the restrictions noted
above. To illustrate the effects of a cooling history on the calculated diffusion durations and
the types of profiles that should result, we have performed a series of numerical calculations.
The thermal history and cooling rates at four different positions within a spherical 1 km3
of magma intrusion (for details about the parameters used see caption of Fig. 7) are used
to calculated how the diffusion coefficients change with time at each point. We have used
the relation x = 4 Dt for characterizing and comparing the diffusion distance x at different
locations and for different elements (see Fig. 7). The diffusion distance varies depending
on the thermal history and on position of the crystal within the pluton. Longer histories are
recorded in the interior of the pluton. These numerical simulations illustrate that: (1) the use of
a single temperature and crystal zoning data from a single location within the pluton is likely
to give ambiguous results. The best location for using a single temperature might be close
to the pluton rims because these undergo faster cooling, analogous to the volcanic systems
considered above (this is only valid for a single intrusion event), and, (2) the zoning patterns
of multiple types of crystals at multiple locations within the pluton should provide a picture of
the temperature time evolution of the plutons, and thus allow discrimination between different
scenarios of pluton growth and formation (e.g., Glazner et al. 2004). Mineral zoning has rarely
been applied to obtain time scales for magmatic or hydrothermal processes in plutonic rocks
(e.g., Coogan et al. 2002, 2005a, 2007; Tepley and Davidson 2003; Gagnevin et al. 2005;
Davidson et al. in press), but it seems that such modeling could provide interesting insights
into the evolution of plutons.
The results shown in Figure 7 also illustrate the effect of closure temperature on diffusion
length scales. After a given time and temperature, the subsequent cooling history does not have
any effect on the concentrations (e.g., diffusion distance vs. time relations are horizontal, as
in Fig. 7e to f). In Figure 8 we show calculated closure temperatures for a number of elements
in different minerals. These may be used as guidelines for first order estimates of how much
diffusion to expect (length of diffusion zones). The plot can help to evaluate which elements
in which minerals are expected to show significant diffusion zoning or erasure of magmatic
zoning in different volcanic and plutonic settings.
Variable boundary conditions – a numerical simulation using MeLtS
The preceding section focuses on the first of two complications that are introduced by
non-isothermal diffusion—the variation of diffusion coefficient with changing temperature.
These considerations allow first order estimates of length and time scales during non-isothermal
Figure 7. (caption continued from facing page)
using diffusion data of Dohmen and Chakraborty (2007a) and Dohmen et al. (2007). The relation x = 4 Dt
describes the penetration distance of diffusion for a semi-infinite medium from an infinite source, e.g., in the
analytical solution: (Cs − C ) (Cs − C∞ ) = erf[( x ) (2 Dt )], with Cs= concentration at the boundary, C∞= con-
centration at the infinite medium. This would correspond to the profile schematically shown in Figure 4c.
V = 1 km , r = 620 m,
3
(a)
Ti= 950 C, Tr= 200 C (b)
o o
1000
Time
total time = 2500 yr Scales of Magmatic Processes 559
Temperature (oC)
4
800 1
3 2
3
V = 1 km , r = 620 m, 600 3
(a) 2 4
Ti= 950 C, Tr= 200 C (b)
o o
400
1000
total time 1= 2500 yr
(oC)
200
800
4 1
Temperature
3
0 2
600
10 3
2 4
(d)
yr-1)
400
) (degree
1 1
10 -12
(c) 200 4
10-17 1
10-10 3
DPb (mm s-1)
yr-1rate
10-22 10 2
4 3
(d)
Cooling
10-27
101
-2
1
Cooling rate (degree
10
10
-12
-32
(c) 4
10-37
10
-17
1 10
-3
10 0 -1
3
DPb (mm s-1)
0 500 1000 2
1500 2000 2500 500 1000 1500 2000 2500
10-22 2
4 3 Time in years Time in years
10 -27
10-2 1
10-32
10
-3
10-37
0 500 1000 1500 2000 2500 0 500 1000 1500 2000 2500
Time in years Time in years
1
(e) (f)
1
2 1
Distance in mmDistance in mm
2
10
-1
3
4 3
1-2 4
10 (e) (f)
1
2 1 in Plag
Pb in zircon CaAl-NaSi
2
10 -3
-1
10 3
10
3
4 3 1
24
10-2 (g) 1
2
Distance in mmDistance in mm
3
10
2
Pb in zircon CaAl-NaSi in Plag
4
3
10-3 4 (h)
10
3
1
10 (g) 2
1
2
Sr in Plag (An65) 3
10
2
Fe-Mg
4
in Oli (Fo80)
1 3
0 500 10004 1500 2000 2500 0 500 1000 1500 2000 (h)
2500
10 Time in years Time in years
Figure 7. Numerical simulations
Sr in Plagillustrating
(An65) the effect of thermal history on the diffusion distances and
Fe-Mg in Oli (Fo80)
on the time1 scales that may be retrieved from crystal zoning patterns of plutonic rocks. (a) Parameters
used in the0calculations
500 (T i = initial
1000 temperature
1500 2000 of the 0magma,
2500 500Tr = initial
1000temperature
1500 2000of the country
2500 rock,
r = radius of the sphere, VTime = volume of
in years the sphere) and illustration of the sphere and
Time in years locations at which
the temperatures were retrieved. We used a solution for a conductive cooling of a sphere with a thermal
diffusivity of 10−6 m2·s−1. Position 1 is at the centre of the sphere, position 2 is at half the radius of the
sphere (310 m from boundary), position 3 is 62 m from wall-rock boundary, and position 4 is 25 m from the
boundary. The calculation was run for a total of 2500 years. (b) Temperature-time plot for the 4 positions.
Note how the temperature close to the wall-rock boundary (positions 3 and 4) changes much faster than
those at the interior. (c) Diffusion rate–time evolution for Pb in zircon (data of Cherniak and Watson 2001)
to illustrate the effect of temperature-time path on diffusion rates. (d) Cooling rate evolution with time at
the four different positions. The amount of diffusion that will occur also depends on the cooling rate, which
is very variable depending on the position. (e) to (f) amount of diffusion distance for selected elements or
components in different minerals. CaAl-NaSi in plagioclase (= Plag) using diffusion data of Grove et al.
(1984), Sr in plagioclase using diffusion data of Cherniak and Watson (1994), Fe-Mg in olivine (= Oli)
(continued on facing page)
560 Costa, Dohmen & Chakraborty
1200
O
1000 Fe-Mg Cr Ca
Fe-Mg-Mn
Temperature (oC) 800
Ni Fe-Ti oxides
600
H
400
200
(a)
0
An
1200 Pb
Ba
La
1000
Sr
Sr
Temperature ( C)
Mg
800
o
600
400 Li
200
(b)
0 -4
10 -3
10 -2
10 10
-1
1
Cooling rate ( C yr )
o -1
Figure 8. Mean closure temperatures calculated following Dodson (1973) for a number of elements in
olivine, plagioclase, clinopyroxene, Fe-Ti oxides, and zircon. (a) Labels are elemental diffusion in olivine
of composition Fo90 at 1 bar, NNO oxygen buffer, and parallel to [001] when applicable. Grey line for
clinopyroxene (diffusion data from Dimanov and Sautter 2000), and dotted line for Fe-Ti oxides (diffusion
data from Freer and Hauptman 1978). Data sources for olivine: Fe-Mg from Dohmen and Chakraborty
(2007a, b), Ca from Coogan et al. (2005b), Mn, Ni from Petry et al. (2004), H from Demouchy and
Mackwell (2006), Cr from Ito and Ganguly (2006), O from Gérard and Jaoul (1989), Ryerson et al. (1989),
and Dohmen et al. (2002b). (b) Labels are diffusion of elements or anorthite in plagioclase (~ An70) and
also of Pb in zircon (dashed line; diffusion data from Cherniak and Watson 2001), and Sr in sanidine (grey
line; diffusion data from Cherniak 1996). Diffusion data sources for plagioclase: Sr from Cherniak and
Watson (1994), Ba from Cherniak (2002), La from Cherniak (2003), Mg from LaTourrette and Wasserburg
(1998) extrapolated to An67 using Costa et al. (2003) equation, Li from Giletti and Shanahan (1997), and
anorthite from Grove et al. (1984). All calculations for a sphere of 500 µm radius.
diffusion. Detailed descriptions of concentration profile shapes require adequate treatment of
varying boundary conditions due to temperature changes. The early analytical models cited in the
last section, as well as the work in the 1970’s and 80’s on extraterrestrial igneous processes cited
in the introduction, considered variation of boundary conditions with temperature. However,
the simplified thermal histories (e.g., linear or exponential cooling rates) and variations in
boundary conditions considered in these analytical models are not always capable of capturing
the complexities of real systems. Complications in natural systems may include non-linear
changes of temperature with time, and changes in the coexisting phase assemblage during the
course of magmatic evolution. The availability of software that allows the total free energy of
Time Scales of Magmatic Processes 561
a system to be minimized (e.g., MELTS: Ghiorso and Sack 1995; PERPLEx: Connolly 1990;
DOMINO-THERIAK: de Capitani and Brown 1987) has opened a window of opportunity
for diffusion modeling. These thermodynamic calculations yield the modal abundances and
compositions of all phases in assemblages as they evolve along selected P-T-fO2-fH2O paths.
Such information may be stored in discrete, digitized forms and implemented as boundary
conditions in numerical models of diffusion. Diffusion modeling carried out in combination
with such thermodynamic modules has potential for accurately describing profile shapes that
develop during non-isothermal diffusion.
We have performed a series of numerical simulations wherein we have used MELTS
(Ghiorso and Sack 1995) to calculate the equilibrium mineral-melt assemblage in a basaltic
composition. We have focused on how the composition of olivine changes with decreasing T
(from near liquidus at 1220 °C to 1170 °C) at constant P (100 MPa). We have chosen a crystal
size of ~500 µm and a total diffusion time of 6 months (Fig. 9). These conditions have been used
to generate three types of diffusion simulations: (i) constant temperature and composition at the
boundary, (ii) cooling history but constant composition at the boundary, and (iii) cooling history
and variable composition at the boundary (Fig. 9). We highlight the following observations:
(1) The profile shapes for cases (i) and (ii) are very similar. The only difference is that in
case (ii) the extent of diffusion is smaller (shorter diffusion profiles) because the effect
of decreasing diffusion coefficient due to decreasing temperature has been ignored in
case (i). If we model the profile obtained in case (ii) using a single temperature as in
case (i), we obtain very similar durations of diffusion; e.g., 5 months instead of 6. This
shows that even if a volcanic rock has experienced a cooling history, the errors are
typically not large if we use only a single temperature to model the process. Moreover,
in a natural case we would obtain the temperature from a geothermometer that is also
subject to the closure condition (Eqn. 12) and may not necessarily record the highest
T, and therefore in practice the time difference would probably be even smaller. Such
an effect could also be accounted for by using the characteristic temperature concept
discussed above. The extent of this mismatch increases with increasing duration and/
or extent of diffusion.
(2) The profile shapes produced in case (iii) are different from those produced in the other
two cases. The shape of the profile for Fo mol%, for example, cannot be fitted in de-
tail by any model in which the composition at the boundary is held constant. The Ca
profile shape is impossible to obtain using any model similar to those considered in
cases (i) and (ii). This means that detailed modeling of natural crystals provide hints on
the importance or necessity of addressing variations in boundary conditions. The time
scales obtained from case (iii) and case (i) are different, the latter being a factor of 4
shorter. This is mainly because the concentration gradient that is observed is a product
of diffusion as well as variations in boundary conditions, and the latter effect is not
accounted for in case (i). However, the shapes of the calculated profiles would provide
a poor description of the observed profile shapes and this mismatch should be taken as
an indication of the inadequacy of a constant-composition boundary condition model.
(3) All the three diffusion models lead to similar and linear relations when the
concentrations of the different endmembers of the olivine are plotted against each
other (Fo vs. Ni, Fo vs. Ca, and Fo vs. Mn; Fig. 9i-j). These linear trends are very
different from the kinks shown by some of the equilibrium compositions. These plots
show that combination of thermodynamic packages and diffusion models permit
us to distinguish between olivine compositions that have been largely affected by
diffusion and those that are unmodified after the creation of growth zoning. The
specific relations may vary from one natural case to another and to be meaningful the
calculations should be tailored (e.g., particular P, T, composition) to the magmatic
system under study.
562 Costa, Dohmen & Chakraborty
84 (a) (b)
82 84 (a) (b)
Fo mol%
80 82
Fo mol%
78 80
76 78
0.8 76 (c) (d)
Tephro mol%
0.6 0.8
(c) (d)
Tephro mol%
0.4 0.6
0.2 0.4
0.0 0.2
1.3 (e)
0.0 (f)
Mont mol%
1.3 (e)
(f)
1.2
Mont mol%
1.2
1.1
0.3
(g)1.1 (h)
Ni-Ol mol%
0.3
0.2 (g) (h)
Ni-Ol mol%
0.2
0.1
0.1
0.0
1160 1180 1200 1220 0 100 200 300 400 500
0.0 Temperature (oC) Distance (mm)
1160 1180 1200 1220 0 100 200 300 400 500
o
Temperature ( C) Distance (mm)
0.8 1.4
(i) (j)
Monti mol%
Tephro mol%
0.8 1.3 1.4
0.6 (i) (j)
Monti mol%
Tephro mol%
1.2 1.3
0.6
0.4 1.2
1.1
0.4
0.2 1.0 1.1
76 78 80 82 84
0.30 0.2 1.0
(k) Fo mol%
76 78 80 82 84
0.25 0.30
Ni-Ol mol%
(k) Fo mol%
MELTS equilibrium
0.20 0.25
Ni-Ol mol%
Constant T and boundary (i)
MELTS equilibrium
Cooling and constant boundary (ii)
0.15 0.20 Constant T and boundary (i)
Cooling and variable boundary (iii)
Cooling and constant boundary (ii)
Total time 6 months
0.10 0.15 Cooling and variable boundary (iii)
Cooling rate = 1-0.5 ·10-2 oC yr-1
76 78 80 82 84 Total time 6 months
0.10
Fo mol% Cooling rate = 1-0.5 ·10-2 oC yr-1
76 78 80 82 84
Fo mol%
Figure 9. Numerical simulations to show the effects of cooling history and change in boundary compositions
on the shapes of diffusion profiles. The temperature vs. composition of the olivine were calculated using
MELTS (Ghiorso and Sack 1995) and correspond to the fractionation mode (equilibrium mode gave very
similar results) of a basaltic composition at the quartz-fayalite-magnetite oxygen buffer (qFM) and 100 MPa
(panels a, c, e, g of the figure). The olivine end members are Fo mol% = 100· Mg/(Mg+Fe+Ca+Mn+Ni),
Tephro mol% = 100·Mn/(Mg+Fe+Ca+Mn+Ni), Mont = 100·Ca/(Mg+Fe+Ca+Mn+Ni), Ni-Ol mol% =
100·Ni/(Mg+Fe+Ca+Mn+Ni). Fractionating minerals were olivine, clinopyroxene and plagioclase. The
changes in slope in the T vs. olivine composition plots occur where the two Ca-rich minerals start to co-
(continued on facing page)
Time Scales of Magmatic Processes 563
diffusion and crystal growth – the moving boundary problem
A scenario that probably occurs commonly in nature is diffusion concurrent with growth
or dissolution of a crystal. This case belongs to the category of moving boundary problems,
and in pure mathematical terms, it simply involves using constitutive Equation (4) and the
continuity equation for describing how the concentration changes with time. If we focus on
using numerical methods, then the difficulty of obtaining analytical solutions for different
cases is of little consequence. However, major problems arise from (i) ascertaining how much
growth or dissolution occurs at what stage of the evolution of a crystal, and (ii) the rate law that
applies to growth or dissolution. The numerical implementation of a dissolution + diffusion
model is more problematic than the implementation of a growth + diffusion model despite
their mathematical similarities. The diffusion process itself may be modified in different ways
during dissolution, but it is still a poorly understood topic. Consequently, we will focus here on
the combination of growth and diffusion. Until now, applications of diffusion with growth or
dissolution models have been rather limited (Nakamura 1995b: olivine growth; Watson 1996:
zircon growth; Ozawa and Nagahara 2000: olivine dissolution).
The thermodynamic free energy minimization algorithms help to alleviate some of
the problems associated with moving boundaries. With a given P-T-fO2-fH2O path for a bulk
composition, the conditions at which a particular phase is produced can be obtained. If growth
is instantaneous (i.e., there is no kinetic delay) and is uniformly distributed over all available
crystals in a system (constrained, for example, by a crystal size distribution), the change in size of
a crystal can be tracked numerically along with changes in modal abundances and composition.
The impact of open-system behavior and resulting changes in bulk composition (e.g., due to
magma mixing) are easily incorporated in stepwise-discretized numerical systems.
To illustrate the process, we have made a series of numerical simulations using the MELTS
algorithm. The change in modal abundance of olivine indicated by the MELTS simulation is
translated into growth of olivine and this adds layers of different compositions over the calculated
range of P-T conditions. The size of each layer is determined by the amount of olivine that
crystallized in the thermodynamic model; this olivine was plated onto a pre-existing olivine
of spherical shape and size as in the simulations considered in the last section. We varied the
growth rates from ~ 10−8 to 10−10 cm·s−1. These are within the limits that have been estimated
from natural examples where the crystallization times are known or can be estimated (Cashman
1990). We highlight the following observations from the results shown in Figure 10:
(1) The profiles produced by only growth or by diffusion plus growth, with growth rates
between 10−8 to 10−9 cm·s−1, have shapes that are quite different from those produced
Figure 9. (caption continued from facing page)
crystallize. The length of the crystal and total time (6 months) were chosen arbitrarily but are similar to
those relevant for magmatic systems. Cooling rates are 1-0.5×10−2 °C·yr−1. The three diffusion models give
different profiles (panels b, d, f, h). Single temperature and boundary composition produces the longest
diffusion distance because we have used the initial temperature (grey line). The model with a cooling
history (black dotted line) and constant boundary gives a similar shape, only shorter diffusion distance
due to the lower temperatures (solid black line). The model that incorporates cooling and change at the
boundary (about 12 mol% per year) shows a smaller diffusion distance yet, and a different profile shape
which should make such profiles recognizable. Inset in panel (b) shows the detailed evolution of the non-
isothermal and variable boundary model [referred to case (iii) in the main text]. Also shown are binary plots
of olivine composition using the end-members (i-k) where it can be seen that the kinked trends generated
from magma fractionation are different from the linear ones produced by diffusion alone. The symbols (i),
(ii) and (iii) refer to the different cases explained in the text. Diffusion data for olivine come from: Fe-Mg
from Dohmen and Chakraborty (2007a, b), Ca from Coogan et al. (2005b), Ni-Mn from Petry et al. (2004)
all parallel to [001], and at qFM buffer and 1 atmosphere.
564 Costa, Dohmen & Chakraborty
1 0.8
84 Fo mol% Tephro mol%
2
82 3
0.6
80
78 0.4
(a) (b)
76
0.2
0.8 1.4
Tephro mol% Monti mol%
0.6 1 1.3
3 2
0.4 1.2
3
2 1
0.2 1.1
(d)
(c) 1.0
0.0
0.3
Ni-Ol mol%
1.4 Monti mol%
3
1.2 3 2
2 0.2
1
1.0 1
0.8 (e) (f)
0.1
76 78 80 82 84
0.3 1
Ni-Ol mol% Fo mol%
2
0.2 3 Growth only
Growth plus diffusion
(1) 3·10 cm·s
-8 -1
0.1
(2) 3·10 cm·s
-9 -1
growth rates
(g) (3) 3·10 cm·s-1
-10
0.0 Total time 6 months
0 100 200 300 400 500
Distance (mm)
Figure 10. Numerical simulations of growth plus diffusion models. The composition of crystallizing
olivine is that calculated by the MELTS thermodynamic algorithm (Ghiorso and Sack 1995) under the
conditions explained in the caption of Figure 9 and in the text. The concentration profile without diffusion
is shown as a thick grey line. The other lines are for growth plus diffusion at various growth rates. The same
diffusion data and conditions were used as in the previous section.
by a diffusion only model in our sample calculation (Fig. 9). This means that they
should be easily recognizable in that such shapes, if observed in nature, cannot be
fitted by a diffusion-only model. It should be possible to carry out a generalized
analysis based on non-dimensional parameters such as Peclet numbers to evaluate the
nature of interaction between growth rates and length of diffusion profiles. Such an
analysis geared explicitly to the determination of time scales from diffusion profiles
in magmatic systems does not exist yet, but see Ozawa and Nagahara (2001) for a
related example.
(2) As the growth rate decreases diffusion has a larger impact on the shape of the compo-
sitional profile. This departure occurs at a growth rate of ca. 10−10 cm·s−1 or less. The
growth plus diffusion profiles of Fo, Ni and Mn, can be fitted with a diffusion-only
Time Scales of Magmatic Processes 565
equation if we use a homogeneous starting profile (with the homogeneous concentra-
tion equivalent to the maximum measured concentration) and a constant-composition
boundary condition at the rim (e.g., like model (i) shown in Fig. 9). The times we ob-
tain are about twice as long as those used to produce the profiles in the diffusion plus
growth models. Although this is a significant difference, it is reasonable considering
the fact that we have used a simple model for a complex situation involving growth
plus diffusion with changes in temperature and boundary conditions with time.
The growth plus diffusion profile for Ca cannot be fitted with a diffusion-only model
without invoking a more complex initial or boundary condition than that assumed for
the other elements. Moreover, we would find a very large difference in retrieved times
compared to that obtained by fitting profiles of the other elements. These observations
show that careful study of multi-element profiles permits an assessment of the assumed
boundary conditions.
(3) The effects of concurrent diffusion and growth can also be recognized in binary plots
of olivine composition as noted in the previous section. The Fo vs. Mn plot does
not discriminate between growth and diffusion because Fe-Mg and Mn have similar
geochemical behaviors during magma fractionation and their diffusion rates are also
practically identical. The Fo vs. Ca plot is much more complex, and although there is
a change in its pattern that depends on the growth and diffusion rates it is not clear that
this should be easily recognizable in natural crystals. The most useful plot is that of
Fo vs. Ni. The relation between these two appears to be non-linear in a fractionating
magma, but with increasing diffusive modification the trend becomes more linear.
The proposed relations above that should allow one to distinguish between zoning
produced by growth and diffusion are illustrated with examples from some natural olivine
crystals (Fig. 11). One is from the Upper Placeta San Pedro sequence of the Tatara San Pedro
volcanic complex. The compositional zoning in this crystal was shown by Costa and Dungan
(2005) to have been the result of diffusion (Fig. 11a). The other crystal is from a lava of the
Middle Estero Molino sequence, also from the Tatara San Pedro volcanic complex (qCNE.1;
M. Dungan unpublished data). This crystal shows a non-linear relationship between Fo and Ni
(Fig. 11b), indicating that there is a significant component of growth zoning retained in this
crystal. Other criteria to distinguish between growth and diffusion zoning are based on the
anisotropy of diffusion, which is discussed in more detail below. Linear relations between Fo
and Ni are a necessary but not unique condition for identifying zoning produced by diffusion.
Linear relations can also be obtained from growth alone, in particular for the low-Fo part of the
trend (Fig. 10f). We note that these models are only illustrative and for application to a specific
case the particular P-T-composition relations of the system should be used. If a diffusion-only
calculation is used to model a concentration profile produced by growth + diffusion, time scales
are overestimated. The effect of dissolution, to the extent that it can be modeled at all (see
above), will be to underestimate time scales.
dIFFuSIOn COeFFICIentS
The use of the diffusion equation to obtain time scales of magmatic processes is critically
dependent on the accuracy of the relevant diffusion coefficients and whether experimentally
determined diffusion coefficients apply to the natural processes that we want to model.
Although here we do not review the experimental approach to obtaining diffusion data (see
Baker 1993; Watson 1994; Chakraborty 1995, 2008), we want to make two observations.
The first is that the advancements in experimental and analytical equipment based on thin
film technology and nanometer-sized diffusion zones allow the determination of diffusion
coefficients in many cases directly at the relevant P, T, and other environmental variables
566 Costa, Dohmen & Chakraborty
0.4
(a) (b)
0.3
NiO wt%
0.2
0.1
0.0
(c) (d)
0.5
MnO wt%
0.4
0.3
0.2
0.1
0.0
0.25
(e) (f)
0.20
CaO wt%
0.15
0.10
0.05
0.00
72 74 76 78 80 82 84 72 74 76 78 80 82 84 86 88
Fo mol% Fo mol%
Figure 11. Plots of data from natural crystals to show effect of diffusion vs. growth. (a-c) Plot of crystal
Ol1_T6 from Upper Placeta San Pedro lava from the Tatara-San Pedro volcanic complex. The crystal is
normally zoned (i.e., decreasing core to rim concentrations) of Fo, Ni, and Mn (traverses found in Costa
and Dungan 2005). They show linear relations between Fo and other elements, which is indicative (but not
proof) of profiles controlled mainly by diffusion (e.g., compare with Fig. 10f). This is in accord with the fact
that Costa and Dungan (2005) found almost identical times when they modeled the profiles of Fo, Ni, Ca
and Mn as being produced only by diffusion. (d-f) Crystal qCNE.1 Oliv-15 from the package G of Middle
Estero Molino Sequence of the Tatara San Pedro volcanic complex (unpublished data of M. Dungan). This
crystal is also normally zoned in Fo, Ni, and Mn. This crystal shows kinked relations between Fo and Ni
and Fo and Ca. These indicate that diffusion has not been able to erase the growth zoning, particularly in
the Fo-rich core (e.g., compare with plot Fig. 10f). If one would fit the zoning profiles of this crystal using
diffusion models, different times would be obtained for different elements which would also indicate the
presence of a significant component of growth zoning.
of interest for magmatic processes (e.g., Dohmen et al. 2002a). This removes the need for
extrapolation from experimental data to the natural conditions in many cases. The second
is that the number and the quality (e.g., exploring the full range of environmental variables)
of diffusion coefficients has increased dramatically, and we understand the mechanisms of
diffusion better. It is currently possible to use almost any igneous rock to obtain time scales of
magmatic processes with a level of certainty that was not possible a couple of decades ago.
We first present the different types of diffusion coefficients, beginning with a discussion
of atomistic mechanisms, and present examples to illustrate which diffusion coefficients to
use in different situations. We discuss the intensive variables and other parameters that control
the diffusion coefficients with a detailed analysis of Fe-Mg diffusion in olivine for the purpose
of clarifying why (and how) the diffusion coefficients depend on specific thermodynamic
variables (Appendix II). This also allows us to explore what is buried within the activation
energies and pre-exponential factors in Arrhenius-type descriptions of the diffusion coefficient.
We hope that these sections will allow the user of diffusion coefficients to use them with more
confidence, with an understanding of their limitations in the applications to natural situations.
Other publications that have described some of these aspects in the geological literature include
Watson (1994), Chakraborty (1995, 2008), Ganguly (2002), Watson and Baxter (2007).
Time Scales of Magmatic Processes 567
types of diffusion coefficients
Diffusion in single crystals requires point defects. The most common diffusion mechanism
involves vacancies but other possibilities are the interstitial mechanism, direct exchange, ring
mechanism, the interstitialcy mechanism (see e.g., Philibert 1991). The vacancy mechanism
implies that each atom surrounding a vacant crystallographic site in an otherwise homogeneous
medium has a probability to jump into it, thereby leaving a new vacant site behind. This
process is self diffusion and does not require the presence of a concentration gradient on the
macroscopic scale. The atomic jumps can be treated statistically. For a particular atom, i,
sitting on a particular crystallographic site a theoretical diffusion coefficient can be defined
(Flynn 1972; Philibert 1991) as:
Di = w ⋅ l 2 (13)
where w is the probability per unit time (or frequency) of a jump to any of the adjacent crystal-
lographic sites at a distance l. This diffusion coefficient is called the self diffusion coefficient
(or random walk diffusion coefficient). Equation (13) relates a macroscopically measurable
property (Di) and parameters that characterize the diffusion mechanism on an atomistic scale.
The most common experimental methods used to study self diffusion employ isotopic
tracers to make the diffusion process measurable. An isotopic concentration gradient is
artificially created in an experimental charge and the concentration distribution evolves with
time as predicted by Fick’s 2nd law (Eqn. 11a,b). The diffusion coefficient of an isotope j of
an element i that describe its flux solely in response to the isotopic concentration gradient in a
chemically homogeneous medium will be called the tracer diffusion coefficient of the element,
D*(j i )= f·Di. The factor f considers effects of the geometry (coordination number, symmetry) and
the diffusion mechanism (e.g., probability that the atom jumps back without contributing to
a net flux; for details see Flynn 1972; Schmalzried 1981; Philibert 1991); it is effectively a
constant with a value close to unity in most cases (see examples in Schmalzried 1981; Philibert
1991). The terms self- and tracer diffusion coefficients have been used interchangeably in
the literature. The tracer diffusion coefficient is not necessarily the diffusion coefficient of a
trace element, although it might be justified to use a tracer diffusion coefficient to describe the
re-equilibration of a trace element concentration gradient (see below). The important point
from the perspective of a modeler is that the tracer diffusion coefficients can be, and often
are, functions of composition; i.e., the tracer diffusion coefficient of Mg in a Fo90 olivine is
different from that in a Fo75 olivine.
Tracer diffusivities may apply strictly to only few natural examples of magmatic systems,
such as those involving zoning in oxygen (e.g., Bindeman 2008) and Sr isotopes in crystals
(although in this case the absence of an overall Sr concentration gradient needs to be demon-
strated; Ramos and Tepley 2008). A concentration gradient of a trace or minor element, or a
major element, in a solid solution will be involved in most other applications, and these gradi-
ents have two consequences: First, this implies the existence of a chemical potential gradient
which must be used to describe the flux of the element (Eqn. 1). This process is called chemi-
cal diffusion and chemical diffusion coefficients (interdiffusion in binary, or multi-component
systems) are used to describe it. The numerical magnitudes of chemical and tracer diffusion
coefficients can be very similar, but they are different physical quantities. Secondly, as various
crystallographic sites are occupied by more than one element, the flux of the element of inter-
est cannot be treated independently from those of the other elements occupying the same site
and formalisms to handle multi-component diffusion are required.
Multi-component diffusion in silicates
In an ionic crystal or solution the flux of an ion implies a flux of charge that creates
an electrochemical potential gradient (Eqn. 5). This potential must be compensated by a
568 Costa, Dohmen & Chakraborty
corresponding counter-flux in the absence of an external electrical field in order to maintain
charge neutrality for the bulk crystal. The diffusion of an element in a mineral is an interdiffusion
process in which at least two elements (or other structural units of a single crystal, e.g.,
electrons, vacancies, see below) are involved. The formulation of Lasaga (1979) accounts for
the coupling of diffusion in an ionic crystal. This framework specifies the coupled fluxes of
cations in a silicate wherein the Si sub-lattice remains fixed, providing a fixed reference frame
for the description of the fluxes of other ions. The constitutive equation describing the fluxes of
n diffusing components or elements in one dimension is given by Equation (6). The individual
elements of the diffusion matrix (Dij) are related to the tracer diffusion coefficients of the
components and the thermodynamic properties of the silicate (Lasaga 1979) by:
z j ∂ ln γ i ∂ ln γ i
Dij = Di∗δij + Di∗Ci −
zn ∂C j ∂Cn
n
n ∂ ln γ k zi ∂ ln γ k
(
− Di∗ ziCi ) ∑ ( )
zk 2Ck Dk∗ × z j D∗j − Dn∗ + ∑ zkCk Dk∗
∂C j
−
zn ∂Cn
(14)
k =1 k =1
Here i and j are indices for the component, zi – charge of component i, Di∗ – tracer diffusion
coefficient of component i, Ci – concentration of component i, δij – the Kronecker delta symbol
(1 for i = j, 0 otherwise). The concentration gradient of the nth component is determined from
that of the other components according to the equation:
∂Cn n −1 ∂C
= −1 / zn ⋅ ∑ z j ⋅ j (15)
∂x j =1 ∂x
Garnet was one of the first silicates or minerals where multi-component diffusion was experi-
mentally studied (Elphick et al. 1985; Loomis et al. 1985; Chakraborty and Ganguly 1992) and
satisfactorily described by the application of the Lasaga (1979) model.
Example I: Chemical interdiffusion of divalent cations. Let us consider diffusion of
divalent Fe and Mg in a binary solid solution such as olivine. In the absence of any other
mechanism of charge compensation, the net flux of the two ions at any given point must be
zero (i.e., there must be equal number of Fe2+ and Mg2+ crossing a point in opposing directions
per unit area per unit time). A single equation can describe the flux (consider Eqn. 6) where
Mg has been chosen as the dependent component:
∂CFe
J Fe = − DFeMg ⋅ = − J Mg (16)
∂x
The interdiffusion coefficient DFeMg can be derived from Equation (14) to be:
*
DFe ⋅ DMg
*
∂ ln γ Fe
DFeMg = ⋅ 1 + (17)
X Fe ⋅ DFe
*
+ X Mg ⋅ DMg
* X Mg
Where XFe was introduced as the molar fraction of the Fe-endmember defined as CFe/(CFe
+ CMg) with an analogous definition for the Mg-endmember, XMg [= (1 − XFe) in this binary
case]. The term in parenthesis in Equation (17) is called the thermodynamic factor and is equal
to one if the solid solution behaves ideally, which is a good first approximation for olivine.
Tracer diffusion coefficients ( Di*) may depend on the major element composition, XFe. This
must be considered when Equation (14) and its form for the binary system (Eqn. 17) are used.
The interdiffusion coefficient in a binary system approaches the tracer diffusion coefficient of
the dilute component if one component is very diluted (e.g., XFe approaches 0); it follows that
*
DFeMg ~ DFe (Fig. 12). If a major component has a much smaller tracer diffusion coefficient
than the dilute component, then the interdiffusion coefficient deviates significantly from the
Time Scales of Magmatic Processes 569
10
2.5 ppm 250 ppm
10
1 1
DFeMg/D*Fe
0.1
10-1
0.01
10
-2
0.001
10
-3
10
-4
10-2 1 100
Fa mol%
Figure 12. Plot of the chemical diffusion coefficient for FeMg interdiffusion (DFeMg) normalized to the
* * *
tracer diffusion coefficient of Fe (DFe ) against the molar fraction of Fe for various ratios of DMg /DFe , which
are indicated by the numbers on the curves. The chemical diffusion coefficient becomes effectively equal
* *
to DFe if XFe is below a critical value for all ratios. The critical value becomes smaller for decreasing DMg
* * * *
/DFe . If DMg would be even larger than DFe the chemical diffusion process could be modeled by DFe for
binary solid solutions with XFe < 0.1. The solid square is a value calculated with experimental Mg tracer
diffusion and Fe-Mg diffusion data for olivine with XFe = 0.1 (Chakraborty et al. 1994 and Chakraborty
* *
1997, respectively), which indicates that DFe is ten times larger than DMg .
tracer diffusion coefficient even for relatively small concentrations of the dilute component
(Fig. 12). Therefore, it is not always justified to use the tracer diffusion coefficient for a trace
element, and some estimation of expected diffusion rates can help to decide whether tracer
diffusion coefficients are applicable. Experimental diffusion data for ~Fo90-olivine show that
the tracer diffusion coefficients of Fe and Mg differ roughly by a factor of 10 (Fig. 12), which
*
implies that for XFe < 0.01 the approximation DFeMg = DFe is justified.
These considerations can be extended to multi-component systems by imposing the
condition that Ci tends to zero for any given i in Equation (14). The algebra is more involved,
but the basic principle is the same; i.e., the process of diffusion of a trace element in a multi-
component system is decoupled from the diffusion and concentration gradients of the other
components as long as the diffusion rate of one other (preferably major) component is fast
enough to ensure charge balance, and the transport rate is well described by the tracer diffusion
coefficient. The tracer diffusion coefficient of some trace elements may also depend on the
composition of the multi-component solution and therefore depend on the instantaneous
concentrations and gradients of the other components in the system.
Natural olivine containing Mn, Ni and Ca as trace elements is a multi-component phase
which can be described as (Fe,Mg,Mn,Ni,Ca)2SiO4. The transport of trace elements that occur
as divalent cations in the same octahedral sites as the major elements Fe and Mg can be modeled
(Costa and Dungan 2005) using their respective tracer diffusion coefficients (e.g., Ni, Mn: Petry
et al. 2004; Ca: Coogan et al. 2005b), without consideration of diffusive coupling (note the
contrast with Fe-Mg diffusion). Nonetheless, the values of these tracer diffusion coefficients
depend on the composition (and hence on instantaneous concentration gradients) in the olivine;
*
e.g., DNi in an olivine of composition Fo90 is ~ 4.0 × 10−17 m2·s−1 at 1000 °C, whereas the same
tracer diffusion coefficient is ~ 4.0 × 10−16 m2·s−1 in an olivine of composition Fo40.
Example 2: Chemical diffusion of Sr2+ in plagioclase. Sr occurs at trace element
concentration levels in most natural plagioclases and its diffusivity has been experimentally
investigated in the presence of concentration (Cherniak and Watson 1994; Giletti and Casserly
570 Costa, Dohmen & Chakraborty
1994) and isotopic gradients (LaTourette and Wasserburg 1998). Cherniak and Watson (1994)
demonstrated that chemical diffusion of Sr occurs mainly by an exchange with Ca in anorthite-
rich plagioclase. Equation (14) for the binary form with the two components Ca2+ and Sr2+ can
be used (analogous to Eqn. 17) to describe the chemical diffusion coefficient of Sr2+:
DSr* ⋅ DCa
*
∂ ln γ Sr
DSrch = ⋅ 1 + (18)
XSr ⋅ DSr* + X Ca ⋅ DCa
*
XSr
It is valid to assume that the thermodynamic factor is 1 for low concentrations of Sr in an unzoned
plagioclase (i.e., constant anorthite content), because the activity coefficient is constant (Henry’s
*
Law limit). LaTourette and Wasserburg (1998) determined that DCa ~ 4.6 × 10−17 m2·s−1 and DSr*
~ 7.1 × 10−17 m2·s−1 at 1400 °C and XAn = 0.95. As the ratio of these tracer diffusion coefficients
is close to 1, we can directly conclude from Figure 12 that DSrch = DSr* for molar fractions of Sr
smaller than 0.01. The chemical diffusion coefficient of Sr varies from 7.1 × 10−17 m2·s−1 to
6.7 × 10−17 m2·s−1 for a variation of 3 orders of magnitude of Sr/(Sr+Ca) from 0.001 to 0.1 at
1400 °C. Given that typical Sr contents in plagioclase are 500-1000 wt ppm, chemical diffusion
of Sr in anorthite-rich plagioclase can be modeled using its tracer diffusion coefficient.
The situation described above will change when the anorthite content of the plagioclase
changes: not only does the chemical diffusion coefficient of Sr in plagioclase depend strongly
on the anorthite content, but the mechanism of exchange which accounts for the charge balance
in the crystal (as shown by Cherniak and Watson 1994) changes as well. In addition to the
dependence of DSr* on anorthite content, a more complicated situation arises if the plagioclase
is zoned in the major components (e.g., anorthite content). The activity coefficient γi of a trace
element i is in most cases constant for a fixed major element composition, but this may vary
within a single crystal if it is zoned with respect to major elements. This means that the term
∂ln(γi)/∂x is non-zero, and Equation (1) cannot be transformed to Equation (2). An extension
of Equation (2) which considers this additional derivative has been proposed by Costa et al.
(2003) and they have applied this approach to the modeling of Mg gradients in plagioclase.
∂Ci A ∂X
J i = − Di* ⋅ + Di* ⋅ Ci ⋅ i ⋅ An (19)
∂x RT ∂x
A formulation such as this is applicable to any other trace element diffusing in a crystal that
is zoned with respect to major elements. The additional driving force for the diffusion flux is
related to the concentration gradient of the anorthite content, which has been treated as frozen
for the duration being modeled because diffusivity of the anorthite component is much slower
compared to that of trace elements (e.g., Grove et al. 1984). The factor Ai is a thermodynamic
parameter related to activity coefficients. Its value can be empirically approximated to a
first degree from the compositional dependence of the trace element partitioning between
plagioclase and melt (i.e., dependence on anorthite content; Costa et al. 2003).
In summary, the tracer diffusion coefficients of all diffusing components plus the
thermodynamic mixing properties of the solid solution must be known to model concentration
profiles in crystals via the diffusion matrix (Eqn. 14). Some simplifications are possible in the
absence of the full body of required information. The assumptions that mixing is ideal or that
some components are effectively immobile considerably simplifies the calculations without
introducing significant errors in time scales retrieved from modeling.
Parameters that determine tracer diffusion coefficients
Any factor that affects tracer or self diffusion coefficients affects all of the other diffusion
coefficients as well. The values of diffusion coefficients in common silicate systems are
mainly controlled by temperature, pressure, water fugacity, oxygen fugacity, composition,
and crystallographic direction. The types and numbers of variables that need to be used
Time Scales of Magmatic Processes 571
to describe diffusion coefficients are determined by the Gibbs phase rule; i.e., there is a
connection between transport properties and an equation from equilibrium thermodynamics.
Temperature always has a strong effect because diffusion is a thermally activated process.
Pressure and composition normally affect diffusion coefficients, although the consequences
may be undetectably small in many systems. The activities of components and volatile
fugacities may, but do not always, influence diffusion. An Arrhenius equation that describes the
temperature and pressure dependencies is commonly used to calculate diffusion coefficients
from experimental data:
−Q − ∆V ( P − 105 )
D = Do ⋅ exp (20)
R ⋅T
Here Q is the activation energy at 105 Pa, ∆V is the activation volume, P is pressure in Pascals,
R is the gas constant, and Do the pre-exponential factor.
The magnitude of the dependencies of the diffusion coefficient on T and P depend on the
size of Q and ∆V. The lowest activation energies for elements diffusing in silicates are those for
H and Li diffusion in olivine, pyroxenes, and plagioclase, which are around 100-200 kJ·mol−1
(e.g., Giletti and Shanahan 1997; Coogan et al. 2005a; Ingrin and Blanchard 2006). Divalent
cations for the same silicates tend to have activation energies between 200-300 kJ·mol−1 (e.g.,
Giletti and Casserly 1994; Cherniak and Watson 1994; LaTourrette and Wasserburg 1998;
Dimanov and Sautter 1999; Petry et al. 2004; Coogan et al. 2005b; Dohmen and Chakraborty
2007a; Dohmen et al. 2007). Increasingly larger activation energies in the ranges of 300 to
500 kJ·mol−1 have been found for trivalent and tetravalent cations (e.g, van Orman et al. 2001;
Dohmen et al. 2002b; Cherniak 2003; Ito and Ganguly 2006; Costa and Chakraborty 2008).
Knowing the activation energy, the effect of changing temperature at a given P can be estimated
using lnD(T2) − lnD(T1) = −Q [R(T2 − T1)]−1. For an activation energy of 200 kJ·mol−1, a decrease
of temperature from 1200 °C to 1000 °C lowers the diffusion coefficient by a factor of ~50. The
effect of a change of 200 °C is much larger at lower T due to the exponential function, and a
drop from 800 to 600 °C varies D by about 4 orders of magnitude.
The effect of pressure on diffusion rates has been investigated experimentally to a lesser
degree, and so the values of the activation volumes are less constrained. The size of the
activation volume is commonly relatively small, so experimentally challenging measurements
of diffusion coefficients over a large pressures ranges are required. Most activation volumes
for silicates are positive and within the ranges of 0 to 16 × 10−6 m3·mol−1 (e.g., review in
Béjina et al. 2003; Holzapfel et al. 2007). Positive activation volumes imply that diffusion
rates become slower at higher pressures at a given temperature. The effect of pressure alone at
a given temperature can be obtained using lnD(P2) – lnD(P1) = −∆V (P2 − P1) [RT]−1. Using
a ∆V of 7 × 10−6 m3·mol−1 (Fe-Mg in olivine; Holzapfel et al. 2007) at 1200 °C, an increase
of pressure from 1 atmosphere to 1 GPa decreases the Fe-Mg diffusion by less than a factor
of 2. The implication is that pressure dependence may be ignored for modeling many crustal
processes, but it is crucial to consider these effects when dealing with mantle processes. It is
apparent that the effect of pressure on diffusion rates is much stronger at lower temperatures
because of the smaller contribution from the T term in the denominator.
The influence of other variables such as fO2, fH2O, and composition are commonly included
in the pre-exponential factor (e.g., Hier-Majumder et al. 2005; Dohmen and Chakraborty
2007a; Costa and Chakraborty 2008). The importance of volatile fugacities and compositions
vary from case to case and it is not possible to generalize, but in some situations their effects
can be more significant than the pressure effect noted above for crustal processes. Appendix II
includes a derivation to demonstrate how these variables influence diffusion coefficients. This
analysis, based on point defect thermodynamics, shows that a generalized form of a diffusion
coefficient is
572 Costa, Dohmen & Chakraborty
∆H f + ∆H m + P ⋅ (∆Vf + ∆Vm ) m1
Di* = D° ⋅ exp − ⋅ a1 ⋅ a2 m2 ⋅ ... ⋅ an-1mn−1 (21)
RT
where the various terms are defined in detail in Appendix II, but by comparing this equation
to Equation (20) we observe that multiple energies are involved in the exponential, and that
the pre-exponential includes the activities or fugacities of components (the a terms in equation
above). We discuss below the effects of impurities (e.g., trace elements), oxygen fugacity, and
water fugacity on diffusion rates, using Fe-Mg diffusion in olivine as an example.
Effect of water and impurities (trace elements). A fluid, or any impurity, can influence
diffusion in a medium only if it affects the properties of that medium in some manner.
Consideration of point defects and how they influence diffusion rates tells us that the two
possibilities are if fluids or impurities alter (1) the concentration and/or (2) the mobility of point
defects responsible for diffusion. Recalling that point defects in silicates are charged entities
(a missing Mg2+ ion in the structure is effectively a doubly negatively charged vacancy), we
recognize that incorporation of fluid species or other impurities has the potential to balance
or create charge imbalances. This will affect diffusion rates by altering the concentrations of
point defects (e.g., vacancies).
We have developed a very simplified point defect model for olivine considering only
two point defects for the charge neutrality condition; vacancies, and Fe3+ on the metal site
(Appendix II). This constraint can be easily extended to the case of natural olivine where
several impurities are present:
[Al′Si ] + [Li′M ] + [Na′M ] + 2 ⋅ [VM′′ ] = [ Fe•M ] + [CrM• ] + [Al•M ] + [Sc•M ] + [OH •O ] (22)
The parentheses [...] denote the number of defects per formula unit of olivine: [VM′′ ] is a vacancy
on a M site (e.g., missing Fe2+ or Mg2+); [Fe•M ] is Fe3+ on a M site; [Al′Si ] is Al3+ on the Si site;
[Li′M ] is Li+ on the metal site; [Na′M ] is Na+ on the metal site; [CrM• ] is Cr3+ on the metal site; [Al•M ]
is Al3+ on the metal site. Negatively charged defects must exactly balance positively charged de-
fects (Eqn. 22). The concentrations of such defects are thermodynamically defined by chemical
potentials or fugacities of the chemical components in the system or environment of the min-
eral. Only the vacancies on the metal site are relevant for Fe-Mg diffusion. Following Dohmen
and Chakraborty (2007a), the effect of impurities on the concentration of these metal vacancies
can be evaluated by defining the equivalent charge of the impurities, [IP3•M ], as follows:
[IP3•M ] = [CrM• ] + [Al•M ] + [Sc•M ] + [OH •O ] − [Al′Si ] − [Li′M ] − [Na′M ] (2
23)
where some impurities are mutually charge balanced. The effective concentrations of the
impurities, as far as charge balance of vacant metal sites is concerned, are much lower than
the absolute concentrations of each of the impurities. The charge balance condition can be
expressed here in the reduced form:
2 ⋅ [VM′′ ] = [ IP3•M ] + [Fe•M ] (24)
Considering the formation reaction of metal vacancies and Fe3+ (Eqn. A5), we can calculate
(equilibrium constant and activity coefficients from Tsai and Dieckmann 2002) the concentration
of these two defects for a given value of [ IP3•M ]. Figure 13a shows that for “pure” olivine the
logarithm of VM′′ increases with the logarithm of fO2 with a slope of 1/6 (compare with Eqn.
A11). This dependence becomes weaker with increasing [ IP3•M ] until it vanishes and the metal
vacancies become directly proportional to [ IP3•M ] with a constant of proportionality of 1/2
(compare with Eqn. 24). In the case of San Carlos olivine, such a transition from fO2-controlled
defect chemistry to impurity-controlled defect chemistry occurs at ~800 °C (Dohmen and
Chakraborty 2007a). The degree to which the trace element composition may influence the
diffusion rate of the major elements can be calculated and predicted by the point defect model.
Time Scales of Magmatic Processes 573
This is important when considering whether experimentally determined diffusion coefficients
apply to the natural case to be modeled. The effect of water on diffusion rates can be evaluated
using the expressions shown above, provided that the amount of water incorporated in olivine
(as some form of defect species) is known as a function of water fugacity (Fig. 13b,c).
If sufficiently high concentrations of H are incorporated in a crystal with a given type of
impurity, the charge neutrality condition may change (vacancies may be charge balanced by
H species rather than by Fe3+). This would lead to a change in diffusion behavior that would
manifest itself macroscopically as a “dry” to “wet” transition (e.g., Costa and Chakraborty
2008). Other factors that are not treated here can influence the value of D. These include the
effects of non-hydrostatic pressure and transport along fast diffusion paths (e.g., dislocations).
Some discussion of these aspects can be found in Chakraborty (2008).
-2.0
(a)
-3.0 1000 ppm
Figure 13. (a) Concentrations of the metal
log [VMe]
vacancies in units of atoms per formula unit
-4.0 100 ppm of olivine (M2SiO4) as calculated from the
1100 oC point defect thermodynamic data of Tsai
10 ppm 1 MPa and Dieckmann (2002). Concentration
-5.0 ‘pure’ in ppm for three different values of the
IW QFM NNO equivalent charge of impurities are shown
as a function of fO2. Arrows mark the
-6.0 position of different solid oxygen fugacity
-15 -13 -11 -9 -7 -5 buffers: NNO = fO2 at the Ni-NiO buffer,
log fO2 (MPa) qFM = fO2 at the quartz-fayalite-magnetite
buffer, IW = fO2 at the iron-wüstite buffer.
-2.0 Equations of buffers were taken from
1100 oC (b) Huebner and Sato (1970), and Eugster
1 MPa 0.05 0.2 0.5 1.0 and Wones (1962), (b) Concentration as
a function of fH2O (equivalent to Psat.) for
-3.0 various fO2. The numbers at the individual
curves denote the log of the fO2 (in bar).
log [VMe]
-4 Solubility of H2O in olivine is calculated
-6
-8 after zhao et al. (2004). The numbers
-4.0 -10 on the dotted lines are pressures in GPa
and indicate the water fugacity at water
saturation (water fugacity calculated after
-5.0 Pitzer and Sterner 1994). The calculation
5 6 7 8 9 was done for 1 MPa and 1100 °C. For
the calculation procedure see Dohmen
Log fH20 (Pa) and Chakraborty (2007a). (c) Ratio
of the diffusion coefficient obtained
30 under water-bearing conditions (data of
Fe-Mg in Fo86 (c) 2.0
Hier-Majumder et al. 2005) and water-
25 fO2 = NNO o C absent conditions (data of Dohmen and
1 00 o Chakraborty 2007 a,b). The numbers
20 C
1.0 1 00 on the dotted lines are pressures in GPa
Dwat.sat/Ddry
1 2 and indicate the water fugacity at water
15
saturation. This corresponds to the
10 0.5 maximum effect of water for the given
0.05 0.1 0.2
total pressure. The effect of T on fH2O is
5 not shown but it is small for the T range
used here.
0
7.0 7.5 8.0 8.5 9.0 9.5 10.0
Log fH20 (Pa)
574 Costa, Dohmen & Chakraborty
dIFFuSIOn MOdeLInG – POtentIaL PItFaLLS and errOrS
We now have a complete toolbox of modeling methods and an understanding of the
diffusion coefficients that is necessary to make use of these methods. The application of these
methods is, however, subject to potential but avoidable pitfalls. It is also necessary to determine
the magnitude of uncertainties from various sources to which the modeling results are subject.
Multiple dimensions, sectioning effects and anisotropy
Compositional data available for diffusion modeling are typically obtained from thin
sections of rocks, thereby requiring the use of one-dimensional concentration profiles, or two-
dimensional images of concentration distribution. One could access the three dimensional
zoning distribution by repeated polishing and analysis but this is extremely time consuming.
Extracting time scales from one-dimensional or two-dimensional models of a process that
occurred in 3-D can introduce errors because the diffusive flux from dimensions that are
not being modeled are neglected (e.g., Eqn. 3). This leads to an overestimate of the time
required to obtain a given extent of diffusive modification of a concentration distribution. It
is difficult to generalize because the effects depend on the shape and size of the crystal, on
the diffusion coefficient, and on the duration of the diffusion process (Fig. 14). Smaller errors
arise from the neglect of additional dimensions if the amount of diffusion that has occurred
is small (e.g., shorter the diffusion profiles) relative to the size of the crystals. It is possible
to overcome this effect by using an analytical solution for a three-dimensional geometry to
model the concentration profiles observed in one or two dimensions. The standard geometries
(e.g., sphere, cylinder) that are typically considered in these solutions constitute an over
simplification of real crystal morphologies. The numerical methods we are discussing in this
work make it possible to extract time scales from fitting two-dimensional composition maps
rather than one-dimensional profiles. Such applications would improve the robustness of the
modeling results (Fig. 15).
Another related effect is that of the orientation of the crystal and the zoning with respect
to the plane of the thin section. The ideal sections in which to study zoning are those that go
through the centre of the crystal and are also perpendicular to one or more crystal faces (Pearce
1984). Mineral grains exposed in thin sections rarely conform to both of these requirements.
Compositional traverses should be taken perpendicular to crystal margins because inclined
traverses or sections that are not perpendicular to the diffusion front yield artificially longer
diffusion profiles (= longer time scales from modeling these profiles; e.g, Fig. 14; Ganguly et
al. 2000; Costa et al. 2003; Costa and Chakraborty 2004). Criteria that may be used to identify
suitable sections or traverse directions include: (i) Choosing the strongest gradient available
in a thin section for modeling diffusion of a given element in a given mineral (e.g., Fig. 1,
An profile), (ii) If the zoning is normal or reverse, then the maximum (minimum for reverse)
concentration found at the core of a crystal is likely to be from the most central section of the
mineral available in the thin section, (iii) Using the normalization criteria of Pearce (1984)
where the concentration is plotted against the diameter raised to the power of three. This
method applies only to cubic or spherical crystals.
Diffusion in non-cubic crystal classes is anisotropic in the same way that optical properties
are. Therefore, the diffusion coefficient along an arbitrary direction can be calculated if the
diffusion coefficients along the three principal axes are known (Eqn. 3; which may or may not
coincide with crystallographic axes, just as in the case of optic axes):
Ditraverse = Dia (cos α)2 + Dib (cos β)2 + Dic (cos γ )2
Here Ditraverse is the diffusion coefficient parallel to the traverse direction in which composition
is measured, and Dia, Dib, and Dic are the diffusion coefficients parallel to the a, b, and c principal
axes; and α, β, and γ are the angles between the traverse and each axis, respectively. Although
Time Scales of Magmatic Processes 575
(a) nce (b) nce
dista 20 30 dista 20 30
10 10
300 300
Mg ppm
Mg ppm
200 200
100 100
0 0
20 20
distan 40 distan 40
ce ce
60 60
(c) 100 (d)
melt c axis Trav2, 2-dim.
80 (F1+F2)
Trav1, 2-dim.
F1 a 60 (F1+F2)
E%
125 mm
250 mm
Trav2 Trav1 axis
40
Trav1, 1-dim.
F2
250 mm 20 (only F1)
500 mm
0
0 100 200 300 400 500 600 700 800 900
years
Figure 14. Illustration of numerical calculations that show that two-dimensional effects on diffusion lengths
are important for elongate crystals (such as plagioclase) or for elements that show a marked anisotropy of
diffusion (such as Fe-Mg in olivine). (a) initial and (b) diffused two-dimensional zoning distribution of the
plagioclase (Costa et al. 2003) Ignoring the elongated shape of plagioclase tends to overestimate the time
scales by more than an order of magnitude. (c) Olivine crystal with the horizontal dimension parallel to the
c axis and the vertical dimension parallel to the a axis. Also shown are the positions of two traverses (Trav1,
solid and dotted lines; Trav2, gray line). Note that in a real situation the crystal would exchange matter with
the surrounding melt (shown as gray zone outside the crystal) in all three dimensions. The contributions
of the fluxes (J) from two of these dimensions are shown as large black arrows (F1, F2). F1 represents the
flux from the direction parallel to the traverses and F2 is the flux from a direction perpendicular to them.
Diffusion parallel to the c axis is 6 times faster than that parallel to a or b, and this leads to a contribution
from F2 that is 6 times greater than that of F1. (d) Plot of the time evolution of the concentration at
the center of the crystal E% showing the effects two dimensions and that of the position of the traverse
(compare Trav1 with Trav2). Neglecting the flux from the horizontal direction can lead to an overestimate
of the time by an order of magnitude or more.
this anisotropy has not always been characterized, it is known for many important mineral groups
such as olivine, clinopyroxene, and apatite (see Brady 1995). Modeling concentration profiles
with diffusion coefficients that are not appropriate for the direction in which the compositions
were measured will yield incorrect time scales. However, it is relatively straightforward to
characterize the orientation of a crystal in a thin section with electron backscatter diffraction
(EBSD). If anisotropic diffusion data are available, compositional data may be combined with
EBSD observations to calculate the appropriate diffusion coefficient for modeling a profile
measured in an arbitrary direction (but perpendicular to a crystal margin – see above).
Aside from reducing artifacts that result when anisotropy of diffusion is not considered
in the course of modeling, diffusion anisotropy is also an excellent tool for identifying
compositional gradients that have been produced by diffusion. Two-dimensional element
maps reveal the differences in width of the compositional zoning in different directions. If
anisotropic diffusion data are available, the expected difference in diffusion widths along
different directions may be calculated and compared with the observed differences in the
element map. Using the relationship x2 ~ Dt, one obtains that x a x c ≈ D a D c . If zoning has
been produced by diffusion, the observed ratio of widths of zoning along different directions
576 Costa, Dohmen & Chakraborty
(a) (b) Mg X-ray map
70
82
[001]
[010] 150 mm
Figure 15. In addition to acquiring multiple one-dimensional traverses in a single crystal it is also useful
to obtain and model two-dimensional x-ray distribution maps. This can be done using the method of finite
differences, although the complex geometry of many crystals can be better modeled with finite elements
rather than finite differences. (a) Simplified two-dimensional model of the olivine crystal shown in (b).
Note the asymmetry of the zoning in the modeled and natural crystal. This is due to the fact that anisotropy
of diffusion of Fe-Mg is such that diffusion parallel to [001] is about 6 times faster than that parallel to
[010] or [100]. The numbers in (a) indicate the Fo content and increase by 2%, with the lowest value at 70
and up to 82.
should be close to that calculated from the diffusion data, once sectioning effects have been
accounted for (see above). Alternately, if the zoning is produced by growth or dissolution, the
widths are more likely to be independent of crystallographic direction, or at least the ratio
would not correspond to that predicted from diffusion data. This method was applied by Costa
and Dungan (2005) to establish that the compositional profiles they modeled were produced
by diffusion with little or no contribution from growth or dissolution.
uncertainties in the calculated times from the diffusion models
Uncertainties associated with calculated time scales are mainly of two types: (a) those linked
to the parameters that enter into the calculations (e.g., diffusion coefficients, temperature), and
(b) those related to the inferred conditions or model at which diffusion took place (e.g., initial
or boundary conditions).
Most geothermometers have an uncertainty of ± 30 °C; if results of direct phase equilibria
experiments are available for the studied system then the uncertainty can be reduced to ± 10
°C. How uncertainty in the determination of temperature translates into uncertainties in the
calculated time scales depends on the activation energy and the specific temperature. Errors
increase with increasing Q and decreasing temperature. At magmatic temperatures (e.g., T >
900 °C): for Q between 200 and 300 kJ·mol−1, a 50 °C uncertainty changes the calculated
times by a factor of ~2; for 500 kJ·mol−1 by as much as a factor of ~4. For 100 kJ·mol−1, this
uncertainty reduces to a factor of 1.3. These factors are all <1.5 for uncertainties of 20 °C in
cases where the temperature can be experimentally determined (e.g., Costa and Chakraborty
2004). If D also depends strongly on other parameters (e.g., P or fO2), the related uncertainties
should also be considered. These are typically much smaller than the contribution related
to temperature. Uncertainties in the diffusion coefficients themselves are variable. Recently
published data have uncertainties for activation energies of about 10%, and experimentally
Time Scales of Magmatic Processes 577
determined diffusion coefficients are within a factor of two at any given temperature. Another
potential source of uncertainty is related to the precision of the compositional data in the
traverse that is modeled. This could take the form of spatial resolution or profiles affected by
convolution effects (e.g., Ganguly et al. 1988), or simply compositional data with low precision
(e.g., for some trace elements from electron microprobe analyses). Profiles fit to even scattered
data yield uncertainties in retrieved time scales of less than a factor of 2. A factor of 2 to 4 could
be considered to be the overall uncertainty for the determination of time scales from diffusion
modeling in magmatic systems.
Other uncertainties may arise from the choice of the form of the constitutive equation,
estimation of initial and boundary conditions, from sectioning, and from ignoring additional
dimensions. These uncertainties are variable and need to be evaluated in each case, but if
maximum care is taken during each step described above, they can be considerably reduced.
We favor modeling multiple elements in single crystals, acquiring data from multiple crystals
in a single sample, and examining multiple samples from a given system. This approach should
limit the uncertainty in retrieved time scales by minimizing artifacts. Such datasets also provide
a test for reproducibility and allow for a statistical treatment of the results.
a CheCKLISt FOr OBtaInInG tIMe SCaLeS OF
MaGMatIC PrOCeSSeS By dIFFuSIOn MOdeLInG
(1) It is important to develop independent constraints on the nature of the magmatic process
whose duration or time scale one hopes to obtain from diffusion modeling. A working
hypothesis of what caused the zoning (magma mixing?, crystal fractionation?) and
the physical situation under which the zoning developed (in a large melt reservoir?,
surrounded by which specific minerals?) is useful for identifying suitable constituative
equations and initial and boundary conditions to be used in modeling. For example, if
field or microscopic evidence indicates that magma mixing took place, the nature of
the mixing end-members may help to define initial conditions.
(2) Expected time scales and knowledge of diffusion coefficients (Fig. 8, Fig. 13) may be
used to obtain a preliminary estimate of which elements in which minerals to focus
on. Once a set of elements and minerals are identified, it is necessary to evaluate the
availability of diffusion coefficients. Diffusion coefficients that have been measured
using well-defined experimental protocols at conditions close to those to be modeled
(T, fO2, P, composition of the crystal) are ideal. Once the diffusion coefficients have
been selected, it is necessary to identify: (a) on which parameters they depend, on
the basis of experimental data, and (b) on which parameters they should depend,
according to theoretical considerations related to point defects. A series of tables with
diffusion coefficients that have been used to obtain time scales can be found at www.
ruhr-uni-bochum.de/planet-diffusion.
(3) A petrological evaluation should be carried out to determine the relevant parameters
that influence the diffusion coefficients e.g., T, fO2.
(4) Careful selection of the locations and orientations of concentration profiles or element
maps is essential to avoid sectioning effects and to ensure that concentration profiles
are measured perpendicular to crystal margins in two dimensions. If diffusion in the
crystal is anisotropic, then it is necessary to obtain orientation data (e.g., EBSD) and
to calculate diffusion coefficients in the directions of the compositional traverses.
If possible, (a) two-dimensional element maps should be acquired to make use of
diffusion anisotropy and to identify true diffusion-related zoning, (b) concentration
profiles of multiple elements should be measured to verify the validity of initial and
boundary conditions and to evaluate reproducibility, and (c) compositional traverses
578 Costa, Dohmen & Chakraborty
in at least two directions at high angles to each other should be measured to make use
of diffusion anisotropy as a crosscheck on calculations.
(5) Identify the form of the constitutive equation appropriate for modeling and formulate
the relevant continuity equation. The appropriate type of diffusion coefficient (tracer-,
chemical-, multi-component- etc.) also needs to be chosen. Even if trace element
concentration profiles are to be modeled, it is essential to determine if diffusive
coupling may be important (e.g., plagioclase) or if tracer diffusion coefficients can
be used (see discussion of Sr diffusion in plagioclase earlier).
(6) Determine the best estimate of the nature of the initial concentration distribution and
flux at the boundary on the basis of the physical scenario being visualized. Insert these
as initial and boundary conditions in the continuity equation. Various approaches for
identifying initial and boundary conditions have been discussed. At this stage, it is
useful to discretize the equations (continuity equation, initial condition, boundary
condition) and write out the appropriate form for the numerical method of choice
(e.g., explicit finite difference, implicit methods).
(7) Decide on isothermal or non-isothermal conditions for constructing the model. A
preliminary estimate for non-isothermal conditions may be obtained by using an
isothermal model calculated at the characteristic temperature. Preliminary estimates
of time scales may be obtained at this stage, and checks for consistency may be
made (e.g., between traverses in multiple directions, traverses of multiple elements,
and results from one-dimensional vs. two-dimensional models etc.). The shapes of
calculated profiles should closely match those of the measured profiles.
(8) Diffusion models may be combined with thermodynamic free energy minimization
programs, if more details are desired (e.g., concentration profile shapes not
well described by model calculations although the zoning is shown to be due to
diffusion).
(9) These steps need to be applied to multiple crystals from single and multiple thin
sections in order to obtain a suitable statistics and confidence in the retrieved time
scales. Crosschecks are provided by the consistency of retrieved time scales from (a)
one-dimensional vs. two-dimensional models, (b) data from multiple traverses and
multiple crystals, taking anisotropy into account, (c) data from multiple elements
in the same crystal, and (d) data from multiple minerals, if possible. In view of the
various possible sources of artifacts, time scales calculated from only one or two
concentration profiles may be spurious.
tIMe SCaLeS OF MaGMatIC PrOCeSSeS that haVe Been OBtaIned
FrOM MOdeLInG the ZOnInG OF CryStaLS and theIr reLatIOn tO
Other InFOrMatIOn On tIMe SCaLeS OF PrOCeSSeS
A number of recent publications have used diffusion modeling to understand magmatic
processes, but only a few studies have specifically addressed the duration over which a process
has operated. Most studies have focused on modeling the zoning patterns in crystals from
volcanic rocks, but some studies have examined plutonic processes (Coogan et al. 2002, 2005a,
2007; Tepley and Davidson 2003; Gagnevin et al. 2005; Davidson et al. in press). We have
summarized below some studies that have used diffusion modeling of different minerals and
elements to determine the durations of magmatic processes. We have concentrated on relatively
recent works wherein determination of time scales was a significant focus. The times obtained
for a variety of processes range from a few days to a few years, and rarely extend to more than
about a hundred years. Following this compilation, we compare these time scales with those
obtained using radioactive isotopes. We note a few examples where both methods have been
Time Scales of Magmatic Processes 579
applied and discuss whether we should expect the same answer from these different methods.
Recent reviews of the time scales of magmatic processes can be found in Condomines et al.
(2003), Reid (2003), Hawkesworth et al. (2004), Peate and Hawkesworth (2005), Turner and
Costa (2007) and Costa (2008). The two last references contain the most information about
time scales obtained from diffusion modeling.
time scales from diffusion modeling
Open-system magmatic processes include magma assimilation, metasomatism, magma
mixing or mingling, and the time between mafic melt intrusion into a reservoir and the
subsequent eruption. The relatively large number of modeling studies devoted to this subject
probably reflects the importance of such processes for understanding magma differentiation
(e.g., Eichelberger 1978) and their potential for triggering eruption and volcanic hazards (e.g.,
Sparks et al. 1977). This focus also reflects the fact that such magmatic events typically leave
a clearly identifiable record in the crystals (e.g., abrupt zoning patterns; e.g., Fig. 4b). This
means that in these cases diffusion modeling with reliable initial and boundary conditions is
possible through identification of initial conditions via characterization of the end members
involved in mixing or assimilation. Many studies have used Fe-Ti zoning in magnetite (e.g.,
Nakamura 1995a; Venezky and Rutherford 1999; Coombs et al. 2000; Chertkoff and Gardner
2004). Although compositional traverses in magnetite might be used for interpreting some
processes (e.g., Devine et al. 2003), there is only very limited diffusion data (e.g., Freer and
Hauptman 1978); the retrieved time scales should be used with some caution until all the
parameters that control diffusion are fully investigated. Other studies have used major and
minor element zoning in olivine (e.g., Nakamura 1995b; Coombs et al. 2000; Pan and Batiza
2002; Costa and Chakraborty 2004; Costa and Dungan 2005). Diffusion in olivine is well
constrained (e.g., see caption of Fig. 10) and diffusion data for many elements, including
the effects of different environmental variables are known. Fe-Mg zoning in clinopyroxene
has been used for determining the time between magma intrusion and eruption (e.g., Costa
and Streck 2004; Morgan et al. 2004, 2006), although the diffusion data for these elements
is also limited (e.g., Brady and McCallister 1983; Dimanov and Sautter 2000). zoning of
trace elements (Sr, Mg, Ba) in plagioclase has also been applied (e.g., Costa et al. 2003,
2008) and there exists a number of recent determinations of diffusion coefficients which have
been reported for a large range of thermodynamic variables and numbers of elements (e.g.,
caption of Fig 10). Wolff et al. (2002) and Bindeman et al. (2006) used the disequilibrium
between oxygen isotope composition of feldspars and quartz or olivines to obtain time scales
for assimilation. There is a significant amount of experimental data available for O diffusion
(see review by Cole and Chakraborty 2001), but the fact that it is still not possible to measure
detailed profiles of the relevant oxygen isotopes in natural crystals (e.g., only core and rim
separates) make this type of time determination more prone to uncertainties (e.g., there is no
fitting of complete profiles). A recent development is the use of Ti zoning in quartz (Wark et al.
2007), which is also based on new determinations of Ti diffusion coefficients (Cherniak et al.
2007), and the development of a new geothermometer (Wark and Watson 2006). This should
open new avenues for understanding the time-temperature histories of silica-rich magmas.
The studies that we have highlighted include a variety of compositions and geological settings
from mafic-silicic mixing in arc volcanoes to mixing between different types of basalts in
mid-ocean ridges. Despite this variety of process and magma types, the majority of time scales
obtained lie between a few days to a few years and they rarely exceed 100 years.
Another rate that has been investigated is the time of magma transport from the mantle
to the surface, via studies of mantle xenoliths and xenocrysts. Klügel (2001), Shaw (2004),
Shaw et al. (2006), and Costa et al. (2005), used Fe-Mg zoning in olivine to obtain transport
times that are variable, from some months to a few hundred years. Such large ranges of times
appear to reflect rapid and direct transport of some mantle xenoliths to the surface, and the
580 Costa, Dohmen & Chakraborty
residence of others in intermediate crustal reservoirs before reaching the surface (e.g, Klügel
2001). A recent new finding is that H is zoned in mantle olivines (Demouchy et al. 2006;
Peslier and Luhr 2006). Such crystals also have been used to infer mantle-to-surface transport
rates of several hours to days. Although there are some determinations of H diffusion in natural
olivine (e.g., Kohlstedt and Mackwell 1998; Demouchy and Mackwell 2006), the experimental
studies have identified multiple diffusion mechanisms that operate at different rates and it is
still not clear which of them applies to natural crystals. More investigations of H diffusion are
necessary before it can be used confidently for obtaining time scales.
Time scale for the generation of silica-rich magmas by partial melting or remobilization
of crustal rocks have also been estimated with oxygen isotopes in quartz and zircon (Bindeman
and Valley 2001), and Sr and Ba zoning in plagioclase (zellmer et al. 1999, 2003). These
studies found time scales somewhat longer than those noted up to now, with a range between
few decades to a few thousand years. This is much longer than the days to several weeks
recorded by the Fe-Ti oxides for the same processes (Devine et al. 2003). This could reflect
that the Fe-Ti oxides record only the last thermal pulse involved in the process. The evidence
for previous temperature perturbations in the Fe-Ti zoning may have been erased due to the
very fast diffusion rates (e.g., Freer and Hauptman 1978). Morgan and Blake (2006) and
zellmer and Clavero (2006) have used Sr and Ba zoning in sanidine. Morgan and Blake (2006)
modeled magma residence and differentiation times for a large silicic magma to be ca. 100
k.y. There is significant diffusion data on sanidine (e.g., Cherniak 1996 and 2002), but the
difficulty is to model a suitable differentiation history. This requires a model that considers
changing composition at the crystal boundary with time, as well as growth during diffusion
(e.g., see previous sections). Morgan and Blake (2006) exploited the different diffusion rates
of Sr and Ba to partly overcome such complexities.
Relation of time scales obtained from modeling crystal zoning to radioactive isotopes.
The time scales calculated with diffusion modeling are short in geological terms and are rarely
more than a few hundred years. This is in contrast with the time determinations from radioactive
isotopes obtained from bulk-rocks, mineral separates, or multiple age determination in a single
crystal, which are in the range of a few 10’s to 100’s of kyr (e.g., Turner and Costa 2007). This
difference is probably due in part to the fact that even using U-series disequilibria it is difficult
to access the days to a hundred years range of magmatic processes, except for the rare cases
when 210Pb-226Ra disequilibria can be used (e.g., Sigmarsson 1996; Condomines et al. 2003).
We are not aware of any study that has applied diffusion modeling and radiogenic isotope
dating to exactly the same crystals, so we cannot quantitatively evaluate differences between
the two clocks. Such exercises have been undertaken on metamorphic rocks and the calculated
durations of processes of interest are the same using both methods (e.g., Ducea et al. 2003;
Spear 2004; Hauzenberger et al. 2005). However, these long-lived radiogenic isotopes and
long-time scales processes are quite different from the ones we have been discussing.
A magmatic system to which many radiogenic isotopic dating methods and diffusion
studies have been applied is the Bishop Tuff of Long Valley, California (e.g., Hildreth and
Wilson 2007; for a review see Costa 2008). If we leave aside age determinations with the
Rb-Sr system, dating of zircons indicates residence times on the order of 50 to 100 kyr (Reid
and Coath 2000; Simon and Reid 2005). This would correspond to the time between zircon
crystallization and a much later eruption. Morgan and Blake (2006) obtained similar times
by modeling the Sr and Ba zoning patterns in sanidine (~100 kyr). This is in agreement with
the higher end of the spectra of residence times for zircon and would suggest that zircon
and sanidine are coeval, that they crystallized during the development of the zonation of the
Bishop Tuff, and that the times obtained from diffusion zoning and radioactive isotopes are
compatible. Notwithstanding this apparent coincidence, recent investigations and diffusion
modeling of the Ti zoning patterns in quartz by Wark et al. (2007) indicate maximum time
Time Scales of Magmatic Processes 581
gap of about 100 years between the creation of the Ti rich rims and eruption. This high-Ti
zone can be interpreted as a high temperature event (Wark et al. 2007) and the short calculated
times may correspond to the timing of the last intrusion that triggered the eruption. This
example illustrates that even where different times are obtained by different methods it is not
necessary that one of them is wrong. For the case of the Bishop Tuff it seems plausible that
the accumulation of several hundreds of km3 of erupted magma, took the tens of thousands of
years suggested by the zircon data and the sanidine diffusion model, whereas the Ti-zoning
in quartz rims records only the last, short-lived event immediately prior to eruption. Detailed
petrographic analysis is crucial to interpret the different durations because different minerals
or even different parts of some crystals may record different aspects of a magma’s history.
SuMMary and PrOSPeCtS
(1) The zoning patterns of crystals are a valuable source of information that can be used
to reconstruct the processes which have occurred in magmatic reservoirs and volcanic
conduits. We stress that detailed analyses of the mineral zoning and modeling based
on chemical diffusion laws also allows the determination of the durations and rates of
some magmatic processes. This potential has been exploited by relatively few studies,
but it is capable of providing data that contribute to progress in the understanding and
quantification of magmatic systems.
(2) Modeling and interpretation of zoning patterns can be anchored within the existing
theoretical framework and experimental data for diffusion in solid silicates. There are
still many unknowns with respect to the atomistic mechanisms of diffusion and the
influences of some experimental variables. Nonetheless, the available information
allows robust time estimates to be obtained for some systems (olivine and plagioclase
mainly) and certain magmatic processes that characterize open systems.
(3) The ranges of time scales that can be accessed from modeling mineral zoning patterns
depend on the diffusion rates and on our capacity to resolve concentration gradients.
As the ranges of diffusion coefficients may vary by more than 13 orders of magnitude
in a single mineral (e.g., from CaAl – NaSi exchange in anorthite to Li diffusion in
plagioclase) and we can measure concentration gradients in crystal zones ranging
over about 6 orders of magnitude in length (some tens of nm to cm), the ranges of
time scales that can be accessed is almost unlimited. The combination of measuring
multiple elements in a single crystal, and in multiple crystals and minerals, offers
the opportunity to retrieve a large number of time determinations from a single thin
section. The availability and precision of diffusion coefficients will improve in the
future, as will our capacity to analyze increasingly low concentrations and resolve
increasingly smaller gradients. An increasing array of processes and their time scales
will become accessible in the future.
(4) The time scale determinations that have been performed so far on igneous systems
focus on volcanic rocks: i.e., much interesting research remains to be done in
modeling zoning patterns in plutonic crystals. Processes such as magma mixing
or assimilation occur rapidly and may predate eruption by a few months to about
100 years. These times are ~ two to three orders of magnitude shorter than those
obtained from radioactive isotopes. However, detailed analysis of the two data sets
shows that the two methods are retrieving times associated with different magmatic
processes or events. These results are more complementary than contradictory. More
studies focussing on the same types of materials are needed to test the degree of
difference and complementariness between diverse approaches. This will require
precise determination of diffusion coefficients plus isotopic and element abundances
on short length scales combined with insightful petrographic scrutiny.
582 Costa, Dohmen & Chakraborty
aCKnOWLedGMentS
FC thanks Michael Dungan and Brad Singer for introducing him to the fun and interesting
science of looking at the zoning patterns of crystals. M. Dungan is also thanked for many
discussions, the data shown for the qCNE1 crystal, and a careful review of this manuscript.
Reviews by Dan Morgan and Keith Putirka were useful to improve the clarity of the manuscript.
The sections about crystal growth and diffusion were motivated by discussions with the
participants of the Tatara-San Pedro volcanic complex GSA Field Forum, especially by the
criticisms of Jon Blundy. Many thanks to Keith Putirka and Frank Tepley for making this
short course happen and their careful editing. FC is supported by a Ramon y Cajal Fellowship
from the Spanish MEC. RD and SC acknowledge generous support of their research from the
German Science Foundation (DFG) over the years. Most recently they have been funded by
the SFB 526 program of the DFG.
reFerenCeS
Albarède F (1996) Introduction to geochemical modeling. Cambridge University Press
Albarède F, Bottinga Y (1972) Kinetic disequilibrium in trace element partitioning between phenocryst and
host lava. Geochim Cosmochim Acta 35:141-256
Allnatt AR, Lidiard AB (1993) Atomic Transport in Solids. Cambridge University Press, Cambridge
Anderson AT (1983) Oscillatory zoning of plagioclase: Nomarski interference contrast microscopy of etched
polished section. Am Mineral 68:125-129
Baker DR (1993) Measurement of diffusion at high temperatures and pressures in silicate systems. In:
Experiments at high pressure and applications to the earth’s mantle. Mineral Assoc Canada Short Course
21:305-355
Béjina F, Jaoul O, Liebermann RC (2003) Diffusion in minerals at high pressure: a review. Phys Earth Planet
Inter 139:3-20
Bindeman I (2008) Oxygen isotopes in mantle and crustal magmas as revealed by single crystal analysis. Rev
Mineral Geochem 69:445-478
Bindeman IN, Sigmarsson O, Eiler J (2006) Time constraints on the origin of large volume basalts derived
from O-isotope and trace element mineral zoning and U-series disequilibria in the Laki and Grímsvötn
volcanic system. Earth Planet Sci Lett 245:245-259
Bindeman IN, Valley JW (2001) Low δ18O rhyolites from Yellowstone: magmatic evolution based on analyses
of zircons and individual phenocrysts. J Petrol 42:1491-1517
Blundy JD, Shimizu N (1991) Trace element evidence for plagioclase recycling in calc-alkaline magmas. Earth
Planet Sci Lett 102:178-197
Bottinga Y, Kudo A, Weill D (1966) Some observations on oscillatory zoning and crystallization of magmatic
plagioclase. Am Mineral 51:792-806
Brady JB (1995) Diffusion data for silicate minerals, glasses, and liquids. In: Mineral Physics and
Crystallography: A Handbook of Physical Constants. AGU Reference Shelf 2. Ahrens TH (ed) Am
Geophys Union, Washington, D.C., p 269-290
Brady JB, McCallister RH (1983) Diffusion data for clinopyroxenes from homogenization and self-diffusion
experiments. Am Mineral 68:95-105
Carslaw HS, Jaeger JC (1986) Conduction of Heat in Solids. Oxford University Press, New York
Cashman K (1990) Textural constraints on the kinetics of crystallization of igneous rocks. Rev Mineral 24:259-
314
Chakraborty S (1995) Diffusion in silicate melts. Rev Mineral Geochem 32:411-504
Chakraborty S (1997) Rates and mechanisms of Fe-Mg interdiffusion in olivine at 980 °C – 1300 °C. J Geophy
Res 102:12317-12331
Chakraborty S (2006) Diffusion modelling as a tool for constructing timescales of evolution of metamorphic
rocks. Mineral Petrol 88:7-27
Chakraborty S (2008) Diffusion in solid silicates: a tool to track timescales of processes comes of age. Ann
Rev Earth Planet Sci 36:153-190
Chakraborty S, Dingwell DB, Rubie DC (1995) Multicomponent diffusion in ternary silicate melts in the system
K20-A1203-Si02: II. Mechanisms, systematics, and geological applications. Geochim Cosmochim Acta
59:265-277
Chakraborty S, Farver JR, Yund RA, Rubie DC (1994) Mg tracer diffusion in synthetic forsterite and San
Carlos olivine as a function of P, T, and fO2. Phys Chem Mineral 21:489-500
Time Scales of Magmatic Processes 583
Chakraborty S, Ganguly J (1991) Compositional zoning and cation diffusion in garnets. In: Diffusion, Atomic
Ordering and Mass Transport. Advances in Physical Geochemistry, Vol. 8. Ganguly J (ed) Springer-
Verlag, New York, p 120-175
Chakraborty S, Ganguly J (1992) Cation diffusion in aluminosilicate garnets: experimental determination
in spessartine-almandine diffusion couples, evaluation of effective binary diffusion coefficients, and
applications. Contrib Mineral Petrol 111:74-86
Charlier BLA, Bachmann O, Davidson JP, Dungan MA, Morgan DJ (2007) The upper crustal evolution of a
large silicic magma body: Evidence from crystal-scale Rb–Sr isotopic heterogeneities in the Fish Canyon
magmatic system, Colorado. J Petrol 48:1875-1894
Cherniak DJ (1996) Strontium diffusion in sanidine and albite, and general comments on strontium diffusion
in alkali feldspars. Geochim Cosmochim Acta 60:5037-5043
Cherniak DJ (2002) Ba diffusion in feldspar. Geochim Cosmochim Acta 66:1641-1650
Cherniak DJ (2003) REE diffusion in feldspar. Chem Geol 193:25-41
Cherniak DJ, Watson EB (1994) A study of strontium diffusion in plagioclase using Rutherford backscattering
spectroscopy. Geochim Cosmochim Acta 58:5179-5190
Cherniak DJ, Watson EB (2001) Pb diffusion in zircon. Chem Geol 172:5-24
Cherniak DJ, Watson EB, Wark DA (2007) Ti diffusion in quartz. Chem Geol 236:65-74
Chertkoff DG, Gardner JE (2004) Nature and timing of magma interactions before, during, and after the
caldera-forming eruption of Volcán Ceboruco, Mexico. Contrib Mineral Petrol 146:715-735
Cole DR, Chakraborty S (2001) Rates and mechanisms of isotope exchange. Rev Mineral Geochem 43:83-
223
Condomines M, Gauthier P-J, Sigmarsson O (2003) Timescales of magma chamber processes and dating of
young volcanic rocks. Rev Mineral Geochem 52:125-174
Connolly JAD (1990) Multivariable phase-diagrams - an algorithm based on generalized thermodynamics. Am
J Sci 290:666-718
Coogan LA, Hain A, Stahl S, Chakraborty S (2005b) Experimental determination of the diffusion coefficient
for calcium in olivine between 900 °C and 1500 °C. Geochim Cosmochim Acta 69:3683-3694
Coogan LA, Jenkin GRT, Wilson RN (2002) Constraining the cooling rate of the lower oceanic crust: a new
approach applied to the Oman ophiolite. Earth Planet Sci Lett 199:127-146
Coogan LA, Jenkin GRT, Wilson RN (2007) Contrasting cooling rates in the lower oceanic crust at fast- and
slow-spreading ridges revealed by geospeedometry. J Petrol 48:2211-2231
Coogan LA, Kasemann SA, Chakraborty S (2005a) Rates of hydrothermal cooling of new oceanic upper crust
derived from lithium-geospeedometry. Earth Planet Sci Lett 240:415-424
Coombs ML, Eichelberger JC, Rutherford MJ (2000) Magma storage and mixing conditions for the 1953-1974
eruptions of Southwest Trident volcano, Katmai National Park, Alaska. Contrib Mineral Petrol 140:99-
118
Cooper AR (1965) Model for multi-component diffusion. Phys Chem Glasses 6:55-61
Cooper KM, Reid MR (2008) Uranium-series crystal ages. Rev Mineral Geochem 69:479-554
Costa F (2000) The petrology and geochemistry of diverse crustal xenoliths, Tatara-San Pedro Volcanic
Complex, Chilean Andes. Terre & Environnement 19, 120 pp. Ph.D. thesis University of Geneva
Costa F (2008) Residence times of silicic magmas associated with calderas. In: Caldera Volcanism: Analysis,
Modelling and Response. Gottsmann J, Marti J (eds) Developments in Volcanology 10:1-55
Costa F, Chakraborty S (2004) Decadal time gaps between mafic intrusion and silicic eruption obtained from
chemical zoning patterns in olivine. Earth Planet Sci Lett 227:517-530
Costa F, Chakraborty S (2008) The effect of water on Si and O diffusion rates in olivine and their relation to
transport properties and processes in the upper mantle. Phys Earth Planet Int 166:11-29
Costa F, Chakraborty S, Dohmen R (2003) Diffusion coupling between trace and major elements and a model
for calculation of magma residence times using plagioclase. Geochim Cosmochim Acta 67:2189-2200
Costa F, Chakraborty S, Fedortchouk Y, Canil D (2005) Crystals in Kimberlites: Where are the Phenocrysts?
Eos Trans AGU, 86(52), Fall Meet. Suppl., Abstract V13B-0545
Costa F, Coogan L, Chakraborty S (2008) The time scales of magma mixing and mingling at mid-ocean ridges.
Geophys Res Abst 10:6229
Costa F, Dungan M (2005) Short time scales of magmatic assimilation from diffusion modelling of multiple
elements in olivine. Geology 33:837-840
Costa F, Dungan M, Singer B (2002) Hornblende and phlogopite-bearing gabbroic crustal xenoliths from Volcán
San Pedro (36° S), Chilean Andes: evidence for melt and fluid migration and reactions in subduction-
related plutons. J Petrol 43: 219-241
Costa F, Streck M (2004) Periodicity and timescales for basaltic magma replenishment at Volcán Arenal (Costa
Rica). IAVCEI General Assembly 2004 abstracts
Crank J (1975) The Mathematics of Diffusion. 2nd edition. Oxford Science Publication, Oxford
584 Costa, Dohmen & Chakraborty
Davidson JP, Font L, Charlier BLA, Tepley FJ III (2008) Mineral-scale Sr isotope variation in plutonic rocks-a
tool for unravelling the evolution of magma systems. Transactions of the Royal Society of Edinburgh:
Earth Sciences 97:357-367
Davidson JP, Morgan DJ, Charlier LA (2007a) Isotopic microsampling of magmatic rocks. Elements 3:261-
266
Davidson JP, Morgan DJ, Charlier LA, Harlou R, Hora JM (2007b) Microsampling and isotopic analysis of
igneous rocks: implications for the study of magmatic systems. Annu Rev Earth Planet Sci 35:273-311
de Capitani C, Brown TH (1987) The computation of chemical equilibrium in complex systems containing
non-ideal solutions. Geochim Cosmochim Acta 51:2639-2652
Demouchy S, Jacobsen SD, Gaillard F, Stern CR (2006) Rapid magma ascent recorded by water diffusion
profiles in mantle olivine. Geology 34:429-432
Demouchy S, Mackwell S (2006) Mechanisms of hydrogen incorporation and diffusion in iron-bearing olivine.
Phys Chem Minerals 33:347-355
Denison C, Carlson WD (1997) Three-dimensional quantitative textural analysis of metamorphic rocks using
high-resolution computed x-ray tomography: Part II. Application to natural samples. J Metamorph Geol
15:45-57
Devine JD, Rutherford MJ, Norton GE, Young SR (2003) Magma storage region processes inferred from
geochemistry of Fe-Ti oxides in andesite magma, Soufrière Hills volcano, Montserrat, W.I. J Petrol
44:1375-1400
Dimanov A, Sautter V (2000) ‘Average’ interdiffusion of (Fe, Mn)-Mg in natural diopside. Eur J Mineral
12:749-760
Dodson MH (1973) Closure temperature in cooling geochronological and petrological systems. Contrib
Mineral Petrol 40:259-274
Dodson MH (1976) Kinetic processes and thermal cooling of slowly cooling solids. Nature 259:551-553
Dodson MH (1986) Closure profiles in cooling systems. Mater Sci Forum 7:145-154
Dohmen R, Becker HW, Chakraborty S (2007) Fe-Mg diffusion coefficients in olivine, Part I: Experimental
determination between 700 and 1200 °C as a function of composition, crystal orientation and oxygen
fugacity. Phys Chem Mineral 34:389-407
Dohmen R, Becker HW, Meissner E, Etzel T, Chakraborty S (2002a) Production of silicate thin films using
pulsed laser deposition (PLD) and applications to studies in mineral kinetics. Eur J Mineral 14:1155-
1168
Dohmen R, Chakraborty S (2003) Mechanism and kinetics of element and isotopic exchange mediated by a
fluid phase. Am Mineral 88:1251-1270
Dohmen R, Chakraborty S (2007a) Fe-Mg diffusion in olivine II: point defect chemistry, change of diffusion
mechanisms and a model for calculation of diffusion coefficients in natural olivine. Phys Chem
Mineral34:409-430
Dohmen R, Chakraborty S (2007b) Fe-Mg diffusion in olivine II: point defect chemistry, change of diffusion
mechanisms and a model for calculation of diffusion coefficients in natural olivine (vol 34, pg 409, 2007).
Phys Chem Mineral34:597-598
Dohmen R, Chakraborty S, Becker HW (2002b) Si and O diffusion in olivine and implications for characterizing
plastic flow in the mantle. Geophys Res Lett 29: doi:10.1029/2002GL015480
Ducea MN, Ganguly J, Rosenberg EJ, Patchett PJ, Cheng W, Isachsen C (2003) Sm-Nd dating of spatially
controlled domains of garnet single crystals: a new method of high temperature thermochronology. Earth
Planet Sci Lett 213:31-42
Eichelberger JC (1978) Andesitic volcanism and crustal evolution. Nature 275:21-27
Elphick SC, Ganguly J, Loomis TP (1985) Experimental determination of cation diffusivities in aluminosilicate
garnets: I. Experimental methods and interdiffusion data. Contrib Mineral Petrol 90:36-44
Eugster HP, Wones DR (1962) Stability relations of the ferruginous biotite, annite. J Petrol 3: 82-125
Fick A (1855) On liquid diffusion. Phil Mag J Sci 10:31-39
Flynn CP (1972) Point Defects and Diffusion. Clarendon Press, Oxford
Freer R, Hauptman z (1978) An experimental study of magnetite-titanomagnetite interdiffusion. Phys Earth
Planet Int 16:223-231
Gagnevin D, Daly JS, Poli G, Morgan D (2005) Microchemical and Sr isotopic investigation of zoned
K-feldspar megacrysts: Insights into the petrogenesis of a granitic system and disequilibrium crystal
growth. J Petrol 46:1689-1724
Ganguly J (2002) Diffusion kinetics in minerals: principles and applications to tectono-metamorphic processes.
Eur Mineral Union 4:271-309
Ganguly J, Bhattacharya RN, Chakraborty S (1988) Convolution effect in the determination of compositional
zoning by microprobe step scans. Am Mineral 73:901-909
Time Scales of Magmatic Processes 585
Ganguly J, Dasgupta S, Cheng W, Neogi S (2000) Exhumation history of a section of the sikkim Himalayas,
India: records in the metamorphic mineral equilibria and compositional zoning of garnet. Earth Planet
Sci Lett 183:471-486
Ganguly J, Tirone M (1999) Diffusion closure temperature and age of a mineral with arbitrary extent of
diffusion: theoretical formulation and applications. Earth Planet Sci Lett 170:131-140
Ganguly J, Tirone M (2001) Relationship between cooling rate and cooling age of a mineral: theory and
applications to meteorites. Meteorit Planet Sci 36:167-175
Gérard O, Jaoul O (1989) Oxygen diffusion in San-Carlos olivine. J Geophys Res 94:4119-4128
Ghiorso MS, Sack RO (1995) Chemical mass transfer in magmatic processes IV. A revised and internally
consistent thermodynamic model for the interpolation and extrapolation of liquid-solid equilibria in
magmatic systems at elevated temperatures and pressures. Contrib Mineral Petrol 119:197-212
Giletti BJ, Casserly JED (1994) Strontium diffusion kinetics in plagioclase feldspars. Geochim Cosmochim
Acta 58:3785-3797
Giletti BJ, Shanahan TM (1997) Alkali diffusion in plagioclase feldspar. Chem Geol 139:3-20
Ginibre C, Wörner G, Kronz A (2007) Crystal zoning as an archive for magmatic evolution. Elements 3:261-
266
Glazner AF, Bartley JM, Coleman DS, Gray W, Taylor Rz (2004) Are plutons assembled over millions of years
by amalgamation from small magma chambers. Geology Today 14:4-11
Glicksman ME (2000) Diffusion in Solids. John Wiley & Sons, Inc. New York
Goldstein JI, Short JM (1967) Cooling rates of 27 iron and stony-iron meteorites. Geochim Cosmochim Acta
31:1001-1023
Grove TL, Baker MB, Kinzler RJ (1984) Coupled CaAl-NaSi diffusion in plagioclase feldspar: experiments
and applications to cooling rate speedometry. Geochim Cosmochim Acta 48:2113-2121
Hauzenberger CA, Robl J, Stüwe K (2005) Garnet zoning in high pressure granulite facies metapelites,
Mozambique belt, SE-Kenya: constraints on the cooling history. Eur J Mineral 17:43-55
Hawkesworth C, George R, Turner S, zellmer G (2004) Time scales of magmatic processes. Earth Planet Sci
Lett 218:1-16
Hier-Majumder S, Anderson IM, Kohlstedt DL (2005) Influence of protons on Fe-Mg interdiffusion in olivine.
J Geophys Res 110, doi: 10.1029/2004JB003292
Hildreth W, Wilson CJN (2007) Compositional zoning of the Bishop Tuff. J Petrol 48:951-999
Holzapfel C, Chakraborty S, Rubie DC, Frost DJ (2007) Effect of pressure on Fe-Mg, Ni and Mn diffusion in
(FexMg1−x)2SiO4 olivine. Phys Earth Planet Int 162:186-98
Huebner JS, Sato M (1970) The oxygen fugacity temperature relationships of manganese and nickel oxides
buffers. Am Mineral 55:934-952
Ingrin Y, Blanchard M (2006) Diffusion of hydrogen in minerals. Rev Mineral Geochem 62:291-320
Ito M, Ganguly J (2006) Diffusion kinetics of Cr in olivine and 53Mn-53Cr thermochronology of early solar
system objects. Geochim Cosmochim Acta 70:799-809
Kirkpatrick RJ, Robinson GR, Hays FJ (1976) Kinetics of crystal growth from silicate melts: anorthite and
diopside. J Geophys Res 81:5715-5720
Klügel A (2001) Prolonged reactions between harzburgite xenoliths and silica-undersaturated melt: Implications
for dissolution and Fe-Mg interdiffusion rates of orthopyroxene. Contrib Mineral Petrol 141:1-14
Kohlstedt DL, Mackwell SJ (1998) Diffusion of hydrogen and intrinsic point defects in olivine. z Phys Chem
207:147-162
Kohn SC, Henderson, CMB, Mason (1989) Element zoning trends in olivine phenocrysts from a supposed
primary high magnesian andesite: an electron and ion-microprobe study. Contrib Mineral Petrol 103:242-
252
Kröger FA, Vink HJ (1965) Relation between the concentrations of imperfections in crystalline solids. Solid
State Phys 3:307-435
Larsen ES, Irving J, Gonyer FA, Larsen ES III (1938) Petrologic results of a study of the minerals from the
Tertiary volcanic rocks of the San Juan region, Colorado. Am Mineral 23:227-257
Lasaga AC (1979) Multicomponent exchange and diffusion in silicates. Geochim Cosmochim Acta 43:455-
469
Lasaga AC (1983) Geospeedometry: an extension of geothermometry. In: Kinetics and Equilibrium in mineral
reactions. Advances in Physical Geochemistry, Vol 3. Saxena SK (ed) New-York, Springer-Verlag, p
81-114
Lasaga AC (1998) Kinetic theory in the earth sciences, Princeton University Press, Princeton
Lasaga AC, Jiang J (1995) Thermal history of rocks; P-T-t paths for geospeedometry, petrologic data, and
inverse theory techniques. Am J Sci 295:697-741
Lasaga AC, Richardson SM, Holland HD (1977) The mathematics of cation diffusion and exchange between
silicate minerals during retrograde metamorphism. In: Energetics of Geological Processes, Saxena SK,
Bhattachanji S (eds) p 353-388. Springer-Verlag, New York
586 Costa, Dohmen & Chakraborty
LaTourrette T, Wasserburg GJ (1998) Mg diffusion in anorthite: implications for the formation of early solar
system planetesimals. Earth Planet Sci Lett 158:91-108
Liang Y, Richter FM, Chaberlin L (1997) Diffusion in silicate melts: III. Empirical models for multicomponent
diffusion. Geochim Cosmochim Acta 61:5295-5312
Lofgren G (1972) Temperature induced zoning in synthetic plagioclase feldspar. In: The Feldspars. MacKenzie
WS (ed) Univ. of Manchester Press, Manchester, p 367-377
Loomis TP (1982) Numerical simulations of crystallization processes of plagioclase in complex melts: the
origin of major and oscillatory zoning in plagioclase. Contrib Mineral Petrol 81:219-229
Loomis TP (1986) Metamorphism of metapelites: calculations of equilibrium assemblages and numerical
simulations of the crystallization of garnet. J Metam Geol 4:201-229
Loomis TP, Ganguly J, Elphick SC (1985) Experimental determination of cation diffusivities in aluminosilicate
garnets: II. Multicomponent simulation and tracer diffusion coefficients. Contrib Mineral Petrol 90: 45-
51
Manning JR (1968) Diffusion Kinetics for Atoms in Crystals. Princeton University Press, Princeton
Milman-Barris MS, Beckett JR, Michael MB, Hofmann AE, Morgan z, Crowley MR, Vielzeuf D, Stolper E
(2008) zoning of phosphorus in igneous olivine. Contrib Mineral Petrol 155:739-765
Moore JG, Evans BW (1967) The role of olivine in the crystallization of the prehistoric Makaopuhi tholeiitic
lava lake, Hawaii. Contrib Mineral Petrol 15:202-223
Morgan DJ, Blake S (2006) Magmatic residence times of zoned phenocrysts: introduction and application of
the binary element diffusion modelling (BEDM) technique. Contrib Mineral Petrol 151:58-70
Morgan DJ, Blake S, Rogers NW, DeVivo B, Rolandi G, Davidson J (2006) Magma recharge at Vesuvius in the
century prior to the eruption AD 79. Geology 34:845-848
Morgan DJ, Blake S, Rogers NW, DeVivo B, Rolandi G, Macdonald R, Hawkesworth J (2004) Timescales
of crystal residence and magma chamber volume from modeling of diffusion profiles in phenocrysts:
Vesuvius 1944. Earth Planet Sci Lett 222:933-946
Nakamura A, Schmalzried H (1983) On the nonstoichiometry and point defects of olivine. Phys Chem Minerals
10:27-37
Nakamura M (1995a) Continuous mixing of crystal mush and replenished magma in the on going Unzen
eruption. Geology 23:807-810
Nakamura M (1995b) Residence time and crystallization history of nickeliferous olivine phenocrysts from the
northern Yatsugatake volcanoes, Central Japan: application of a growth and diffusion model in the system
Mg-Fe-Ni. J Volcanol Geotherm Res 66:81-100
Onsager L (1945) Theories and problems of liquid diffusion. Ann N Y Acad Sci 46: 241-265
Ozawa K, Nagahara H (2000) Kinetics of diffusion-controlled evaporation of Fe-Mg olivine: experimental
study and implication for stability of Fe-rich olivine in the solar nebula. Geochim Cosmochim Acta
64:939-955
Ozawa K, Nagahara H (2001) Chemical and isotopic fractionations by evaporation and their cosmochemical
applications. Geochim Cosmochim Acta 65:2171-2199
Pan Y, Batiza R (2002) Mid-ocean ridge magma chamber processes: Constraints from olivine zonation in lavas
from the East Pacific Rise at 9°30′N and 10°30′N. J Geophys Res 107, DOI 10.1029/2001JB000435
Pearce TH (1984) The analysis of zoning in magmatic crystals with emphasis on olivine. Contrib Mineral
Petrol 86:149-154
Pearce TH, Kolisnik AM (1990) Observations of plagioclase zoning using interference imaging. Earth Sci Rev
29:9-26
Peate DW, Hawkesworth CJ (2005) U series disequilibria: insights into mantle melting and the timescales of
magma differentiation. Rev Geophys 43, 2004RG000154
Perugini D, Ventura G, Petrelli M, Poli G (2004) Kinematic significance of morphological structures generated
by mixing of magmas: a case study from Salina Island (southern Italy). Earth Planet Sci Lett 222:1051-
1066
Peslier AH, Luhr JF (2006) Hydrogen loss from olivines in mantle xenoliths from Simcoe (USA) and Mexico:
mafic alkalic magma ascent rates and water budget of the sub-continental lithosphere. Earth Planet Sci
Lett 242:302-319
Petrelli M, Perugini D, Poli G (2006) Time-scales of hybridization of magmatic enclaves in regular and chaotic
flow fields: petrologic and volcanological implications. Bull Volcanol 68:285-293
Petry C, Chakraborty S, Palme H (2004) Experimental determination of Ni diffusion coefficients in olivine
and their dependence on temperature, composition, oxygen fugacity, and crystallographic orientation.
Geochim Cosmochim Acta 68:4179-4188
Philibert J (1991) Atom Movements: Diffusion and Mass Transport in Solids. Les Éditions de Physique. Les
Ulis
Pitzer KS, Sterner SM (1994) Equations of state valid continuously from zero to extreme pressures for H2O
and CO2. J Chem Phys 101:3111-3116
Time Scales of Magmatic Processes 587
Press WH, Teukolsky SA, Vetterling WT, Flannery BP (2007) Numerical Recipes 3rd Edition: The Art of
Scientific Computing, Cambridge University Press, Cambrigde
Ramos FC, Tepley FJ III (2008) Inter- and intracrystalline isotopic disequilibria: techniques and applications.
Rev Mineral Geochem 69:403-443
Reid M (2003) Timescales of magma transfer and storage in the crust, In: Treatise on Geochemistry, Volume
3: The Crust. Holland HD, Turekian KK (eds) Elsevier, p 167-193
Reid MR, Coath CD (2000) In situ U-Pb ages of zircons from the Bishop Tuff: no evidence for long crystal
residence times. Geology 28:443-446
Ryerson FJ, Durham WB, Cherniak DJ, Lanford WA (1989) Oxygen diffusion in olivine - effect of oxygen
fugacity and implications for creep. J Geophys Res 94:4105-4118
Sato (1986) High temperature a.c. electrical properties of olivine single crystal with varying oxygen partial
pressure: implications for the point defect chemistry. Phys Earth Planet Inter 41:269-282
Schmalzried H (1981) Solid State Reactions. Verlag Chemie, Weinheim
Shaw CSJ (2004) The temporal evolution of three magmatic systems in the West Eifel volcanic field, Germany.
J Volcanol Geotherm Res 131:213-240
Shaw CSJ, Heidelbach F, Dingwell DB (2006) The origin of reaction textures in mantle peridotite xenoliths
from Sal Island, Cape Verde: the case for metasomatism by the host lava. Contrib Mineral Petrol 151:681-
697
Short M, Goldstein JI (1967) Rapid methods of determining cooling rates of iron and stony iron meteorites.
Science 165: 59-61
Sigmarsson O (1996) Short magma chamber residence time at an Icelandic volcano inferred from U-series
disequilibria. Nature 391:440-442
Simon JI, Reid MR (2005) The pace of rhyolite differentiation and storage in an ‘archetypical’ silicic magma
system, Long Valley, California. Earth Planet Sci Lett 235:123-140
Singer BS, Dungan MA, Layne GD (1995) Textures and Sr, Ba, Mg, Fe, K, and Ti compositional profiles in
volcanic plagioclase: clues to the dynamics of calc-alkaline magma chambers. Am Mineral 80:776-798
Sparks RSJ, Sigurdsson H, Wilson L (1977) Magma mixing: a mechanism for triggering acid explosive
eruptions. Nature 267:315-318
Spear FS (2004) Fast cooling and exhumation of the Valhalla metamorphic core complex, southeastern British
Columbia. Int Geol Rev 46:193-209
Spear FS, Kohn MJ, Florence F, Menard T (1990) A model for garnet and plagioclase growth in pelitic schist:
implications for thermobarometry and P-T path determinations. J Metamorph Geol 8:683-696
Streck MJ (2008) Mineral textures and zoning as evidence for open system processes. Rev Mineral Geochem
69:595-622
Takeda H, Miyamoto M, Ishii T, Lofgren GE (1975) Relative cooling rates of mare basalts at the Apollo 12 and
15 sites as estimated from pyroxene exsolution data. Proc Lunar Sci Conf 6th, 987-996
Taylor LA, McCallister RH, Sardi O (1973) Cooling histories of lunar rocks based on opaque mineral
geothermometers. Proc Lunar Sci Conf 4th, 819-828
Tepley FJ III, Davidson JP (2003) Mineral scales Sr-istope constraints on magma evolution and chamber
dynamics in the Rum layered intrusion, Scotland. Contrib Mineral Petrol 145:628-641
Tomkeieff SI (1939) zoned olivines and their petrogenetic significance. Mineral Mag 25:229-251
Trepmann CA, Stöckhert B, Chakraborty S (2004) Oligocene trondhjemitic dikes in the Austroalpine basement
of the Pfunderer Berge, Südtirol- level of emplacement and metamorphic overprint. Eur J Mineral 16:641-
659
Tsai TL, Dieckmann R (2002) Variation of the oxygen content and point defects in olivines, (FexMg1−x)2SiO4,
0.2 = x = 1.0. Phys Chem Mineral 29:680-694
Turner S, Costa F (2007) Measuring timescales of magmatic evolution. Elements 3:267-272
Umino S, Horio A (1998) Multistage magma mixing revealed in phenocryst zoning of the Yunokuchi pumice,
Akagi volcano, Japan. J Petrol 39:101-124
Van Orman JA, Grove TL, Shimizu N, Graham DL (2001) Rare earth element diffusion in diopside: influence
of temperature, pressure and ionic radius, and an elastic model for diffusion in silicates. Contrib Mineral
Petrol 141:687-703
Venezky DY, Rutherford MJ (1999) Petrology and Fe-Ti oxide reequilibration of the 1991 Mount Unzen mixed
magma. J Volcanol Geotherm Res 89:213-230
Walker D, Longhi J, Lasaga AC, Stolper EM, Grove TL, Hays TL (1977): Slowly cooled microgabbros15555
and 15065. Proc Lunar Sci Conf 8th, 1521-1547
Wallace GS, Bergantz GW (2002) Wavelet-based correlation (WBC) of zoned crystal populations and magma
mixing. Earth Planet Sci Lett 202:133-145
Wark DA, Hildreth W, Spear FS, Cherniak DJ, Watson EB (2007) Pre-eruption recharge of the Bishop magma
system. Geology 35:235-238
588 Costa, Dohmen & Chakraborty
Wark DA, Watson EB (2006) Titaniq: a titanium-in-quartz geothermometer. Contrib Mineral Petrol 152:743-
754
Watson EB (1994) Diffusion in volatile-bearing magmas. Rev Mineral 30:371-411
Watson EB (1996) Dissolution, growth and survival of zircons during crustal fusion: kinetic principles,
geological models and implications for isotopic inheritance. Trans R Soc Edinburgh Earth Sci 87:43-56
Watson EB, Baxter EF (2007), Diffusion in solid-Earth systems. Earth Planet Sci Lett 253:307-327
Wolff JA, Balsley SD, Gregory RT (2002) Oxygen isotope disequilibrium between quartz and sanidine from the
Bandelier Tuff, New Mexico, consistent with a short residence time of phenocrysts in rhyolitic magma. J
Volcanol Geotherm Res 116:119-135
Wood JA (1964) The cooling rates and parent bodies of several iron meteorites. Icarus 3:429-459
Young DA (2003) Mind Over Magma. Princeton University Press, Princeton
zellmer GF, Blake S, Vance D, Hawkesworth C, Turner S (1999) Plagioclase residence times at two island arc
volcanoes (Kameni Islands, Santorini, and Soufriere, St. Vincent) determined by Sr diffusion systematics.
Contrib Mineral Petrol 136:345-357
zellmer GF, Clavero JE (2006) Using trace element correlation patterns to decipher a sanidine crystal growth
chronology: an example from Taapaca volcano, Central Andes. J Volcanol Geotherm Res 156:291-301
zellmer GF, Sparks RSJ, Hawkesworth C, Wiedenbeck M (2003) Magma emplacement and remobilization
timescales beneath Montserrat: insights from Sr and Ba zonation in plagioclase phenocrysts. J Petrol
44:1413-1431
zhao Y-H, Ginsberg SB, Kohlstedt DL (2004) Solubility of hydrogen in olivine: dependence on temperature
and iron content. Contrib Mineral Petrol 147:155-161
aPPendIx I
ShOrt GuIde tO the uSe OF the FInIte dIFFerenCe MethOd FOr
nuMerICaL CaLCuLatIOnS WIth the dIFFuSIOn equatIOn
The first step is to discretize the space and time coordinates into finite but small increments
to make a grid. This is shown in Figure A1. We have a step size ∆x on the space coordinate,
and ∆t on the time coordinate. In the simplest implementations these are taken to be constants,
but are not required to be so. These define the smallest variations that can be “recognized”
by the computer at that particular point of space and time. For example, if we let ∆x = 5 µm
and ∆t = 3600 s, then distances smaller than 5 µm and time scales smaller than 3600 s cannot
be resolved by the numerical calculation. Any numerical calculation is a trade off between
speed (larger the individual steps, fewer of these are required to get to any particular value)
and accuracy (smaller the step, closer it is to being truly infinitesimal, as required by rigorous
mathematics). We are concerned with describing concentrations as a function of space and
time—the variable C(x,t) used in the text. In the numerical implementation, we can associate
each value of concentration with two indices written as subscripts—one for space (i) and the
other for time (j), e.g., Ci,j. If we write, for example, C2,2 we are referring to the concentration
at grid point number 2 on the spatial scale at time step number 2. For our chosen value of ∆x
= 5 µm and ∆t = 3600 s, C2,2 is the concentration at a distance of 10 µm from our arbitrarily
chosen zero position after 2 hours (Fig. A1).
The next step is to convert the differential expressions into discrete forms using the Taylor
Series approximation (see Crank 1975, for a detailed exposition geared to the solution of
diffusion problems). There are several levels of approximation that may be used (first order,
second order etc.) and different options about the choice of neighborhood in which to carry out
the Taylor expansion (forward difference, backward difference, central difference etc.). The
simplest of these, in a forward difference scheme, can be written as
∂ C Ci +1, j − Ci , j
=
∂x ∆x
Time Scales of Magmatic Processes 589
(a)
3
C2,2
time
2
Dt
1
Dx
1 2 3
distance
(b) tim
e
concentration
distance
Figure a1. Visualization of the discretization of time and space used in finite difference algorithms.
(a) There is a value of concentration associated with each point in space and time. For example, C2,2
stands for the value of concentration at a distance of 2•∆x at the point of time 2•∆t. (b) Three-dimensional
visualization of how the concentration changes due to diffusion using the discretization of space and time.
See also text for more explanation.
The most important point to note here is that this expression converts a differential expression
into a simple algebraic expression that can be manipulated by straightforward rules of addition
and multiplication. Higher derivatives may be written by repeating the same approach. For
example,
∂ 2C ∂ ∂C 1 Ci +1, j − Ci , j Ci , j − Ci −1, j Ci +1, j − 2Ci , j + Ci −1, j
= = − =
∂x ∂x ∂x ∆x ∆x ∆x ( ∆x )
2
Similarly, for the time derivative we have
∂ C Ci , j +1 − Ci , j
=
∂t ∆t
If we put it all together and move terms around using algebra, the continuity Equation (11a)
in the text translates to
590 Costa, Dohmen & Chakraborty
Ci +1, j − 2Ci , j + Ci −1, j
Ci , j +1 = Ci , j + D∆t ( ) ( A1)
∆x 2
for calculating the concentration at a grid point i at the next time step, j+1. This formula
can be iterated over all grid points, shifting the index i by 1 each time, to obtain a complete
concentration profile at the time step j+1. The entire procedure may be then repeated to calculate
the concentration profile at the next time step, j+2, using the values determined at step j+1 as the
“old” values. If this is done until the index j reaches the value 2 (with the time step, ∆t = 3600
s, as above), we would have calculated a concentration profile that would be measured after 2
hours of diffusion. This numerical method is called the explicit finite difference method.
These simple formulae and iterations can be implemented on a standard spreadsheet
program such as Excel®. On the MSA WWW site (http://www.minsocam.org/MSA/RIM), we
provide a spreadsheet (olivine_diffu_excel.xls) and a Mathematica® notebook (olivine_diffu_
mathema.nb) with an example for Fe-Mg diffusion in olivine, but it is an instructive exercise
to consider the various constitutive equations described in the text, derive the corresponding
continuity equations, and then digitize them to the discrete numerical form as above. For
example, if the diffusion coefficient depends on concentration we need to solve another form
of the diffusion equation (Eqn. 11b), which in finite difference form is:
Di +1, j − Di , j Ci +1, j − Ci , j C − 2Ci , j + Ci −1, j
Ci , j +1 = Ci , j + ∆t ( )( ) + Di , j ∆t ( i +1, j ) ( A2)
∆x ∆x ∆x 2
For actual numerical implementation it is necessary to pay attention to a few other details.
For any explicit numerical scheme to be stable, it is necessary that the so-called Courant
condition be fulfilled. For Equation (A1), D, ∆x and ∆t can be combined together into one
parameter, r = D∆t (∆x )2. The Courant condition is that r < 0.5 (e.g, see Crank 1975). If the
scheme has been implemented on a spreadsheet, it is easy to see what happens when r exceeds
a value of about 0.5. If D, ∆x and ∆t are not constant during the course of a calculation (e.g.,
Eqn. A2 where D change with position or concentration, or for a non-isothermal case, where
D changes with temperature, and hence, time), then it is necessary to ensure that the Courant
condition is fulfilled at each grid point. With these equations and the boundary and initial
conditions for the case at hand we can solve the diffusion equation and fit compositional
profiles. For example, for the case of no flux at the boundary, we write C1,j(all) = C2,j(all), and
likewise Cn,j(all) = Cn−1,j(all) (where n is the total number of space grid points and j(all) means
for all times; this is the condition used in Fig. 6a). For the case that there is exchange, but a
constant composition at the boundary (Cb) we have C1,j(all) = Cn,j(all) = Cb (this is the condition
used in Fig. 6b, with Cb = 40).
Inspection of the finite difference formula provides several insights into the advantages
and limitations of the numerical method. Some of these are:
(1) If an explicit method is to be used, we cannot choose D, ∆x and ∆t arbitrarily. As D
and ∆x are usually specified by the problem at hand, the choice of ∆t is limited by
the Courant condition. If it turns out that ∆t needs to be 3600 s or less, for example,
it means that to calculate a concentration profile after 100 years of diffusion one
would need to carry out the steps noted above 876000 times at each spatial grid
point. For calculations to millions of years (e.g., cooling of a pluton), the difficulty
escalates. This is why it is necessary to use a programming tool or language to carry
out realistic calculations and “iteration by hand” on a spreadsheet can become a little
too cumbersome. Packages such as Mathematica®, Mathcad®, Maple®, or Matlab®
allow such implementation of automated iteration without full-fledged programming
using a high level language (e.g., C, Fortran). The latter are in general faster.
(2) Inspection of the formula above shows why it is necessary to have the two boundary
Time Scales of Magmatic Processes 591
conditions and one initial condition. To calculate the concentration at a grid point i,
it is necessary to know the concentrations at grid points i−1 and i+1. These formulae
cannot be used to calculate concentrations at the first (no i−1 grid point available)
and the last (no i+1 grid point available) positions. The two boundary conditions
provide the additional information to calculate concentrations at these points. To
obtain the concentrations at time step j+1, we need information about concentrations
at time step j. But these formulae cannot be used for the first time step and therefore
we need the initial conditions to start the iterations. It is an instructive exercise to try
to implement various forms of boundary conditions (concentration held constant, no
flux etc.) in the discrete form.
(3) As these calculations proceed by explicitly determining the concentrations at each
position at each time step, it is easy to add variations in the model. For example, to
start off a calculation we need to define the concentrations at each grid point (the
initial condition). Because each concentration is specified individually, it is irrelevant
whether the concentration profile is homogeneous, has a regular geometric form,
or is of an arbitrary shape. Similarly, if we wish to change the boundary conditions
subject to which diffusion occurs after, say, the 20th time step, we can directly do so.
This is where the flexibility of the numerical method noted in the text comes from.
This is also how we can take outputs from thermodynamic programs such as MELTS
and track diffusion along complex P-T-fO2-fH2O paths, with boundary conditions and
diffusion coefficients that vary as the coexisting phase assemblage and equilibrium
conditions change. More about discretization and numerical solution can be found in
for example Press et al. (2007).
aPPendIx II
POInt deFeCt therMOdynaMICS and hOW IntenSIVe VarIaBLeS
aFFeCt dIFFuSIOn COeFFICIentS
For diffusion to occur within a solid it is necessary to have point defects (e.g., see Flynn 1972
for a discussion), the simplest of which are vacancies (e.g., an atom missing in a crystallographic
site) and interstitials (an additional atom in a non-regular crystallographic site). The diffusivities
and concentration of such point defects is what mainly controls and explains the diffusivities of
the cations. Point defects need to exist in solids because thermodynamic constraints show that a
lower free energy (through an increase in entropy) is obtained by their presence at temperatures
above 0 K (e.g., Flynn 1972). This means that the presence of point defects in single crystals is
an equilibrium feature. Point defects can be treated as quasi-chemical species having a chemical
potential like any other solid solution component and hence the concentrations of the individual
point defects depend, in addition to P and T, on the activities of the thermodynamic components
(e.g., Kröger and Vink 1965; Schmalzried 1981).
We consider that diffusion of an atom i occurs by a vacancy mechanism; a treatment for
diffusion via interstitials would be analogous. A specific atom i can make a jump only if one
of the adjacent sites is vacant. In contrast, from the perspective of a vacancy the probability of
a jump to an adjacent occupied site is always high. The vacancy does not have to “wait” for
another vacancy to arrive at the adjacent site. We define the vacancy diffusion coefficient as: DV
= wV·l2. We define the tracer diffusion coefficient of the element by recognizing that the prob-
ability of a jump of i to an adjacent site is equal to the jump frequency (wv) times the probability
of i to be next to a vacancy. This latter probability is simply XV – the fraction of vacancies on the
crystallographic sub-lattice of the element i. The tracer diffusion coefficient of i is:
592 Costa, Dohmen & Chakraborty
Di* = fi ⋅ XV ⋅ DV = fi ⋅ XV ⋅ wV ⋅ l 2 ( A3)
where fi is the correlation factor for the diffusion of element i (see main text for the significance
of fi). The correlation factor fi and the jump distance can be treated as constants for a given
diffusion mechanism, but the concentration of the vacancies, XV, and the jump frequency of
the vacancy, wV, are strongly sensitive to a number of parameters. Temperature is the dominant
one, since both variables are thermally activated, and the dependence of the jump frequency
can be summarized by an Arrhenian equation (for exceptions see Flynn 1972):
∆H m + P ⋅ ∆Vm
wv = wv ⋅ exp − ( A4)
R ⋅T
where ∆Hm and ∆Vm are called migration enthalpy and migration volume, respectively. The
next task is to describe the concentration of vacancies. If n is the number of the thermodynamic
components of a crystal, the molar fraction of the vacancy may depend on P and T plus the
activities of n−1 chemical components:
X V = f ( P, T , a1 , a2 ,..., an −1 ) ( A5)
In most cases the influence of several component activities can be effectively ignored and only
one component controls the point defect chemistry.
Example of point defect chemistry of olivine and the relation to Fe-Mg diffusion. To
define the stoichiometry of (Fe,Mg)2SiO4-olivine (structural formula M2[SiO4], with M = Fe2+,
Mg2+; it is ignored that the M1 and M2 sites can be distinguished), allowing for the possibility
of existence of defects, four chemical components are required and thus three components
have to be fixed. These may be, for example, aSiO2, fO2 and aFe2SiO4. The common convention to
describe the structural units of a crystal is the Kröger-Vink notation (Kröger and Vink 1965),
which for the present case leads to:
VM′′ – vacancy on a M site; Fe•M–Fe3+ on a M site; Fe×M– Fe2+ on a M site
Here the main symbol stands for the element (exception: V for vacancy), the subscripted symbol
stands for the crystallographic site (e.g., M for an octahedral metal site or i for interstitial) and
the superscripted symbol stands for the effective charge relative to the ideal crystal, where •
stands for one positive effective charge, | stands for one negative effective charge and × is used
for a neutral structural unit, e.g., a regular structural unit like Fe×M.
To understand how point defects affect the diffusion of ions we write quasi-chemical
reactions for the formation of such defects. The terms and dependencies in these reactions
explicitly incorporate the relations with the diffusion coefficients. The equations need to
conserve mass, charge, and site balance. One kind of positive point defect and one kind of
negative point defect are commonly much more abundant than any other point defect and
effectively control the point defect chemistry of the crystal. In olivine VM′′ and Fe•M are probably
the majority point defects (e.g., Nakamura and Schmalzried 1983; Sato 1986). The effective
charge balance condition is:
2 ⋅ X V = X Fe3+ ( A6)
with XV and XFe3+ as the molar fractions of vacancies and Fe3+ on the metal site, respectively.
For the formation of the metal vacancies and Fe3+ in olivine the following reaction has been
considered (e.g., Nakamura and Schmalzried 1983):
6 ⋅ Fe xM + SiO2 + O2 (g) = 2 ⋅ VM’’ + 4 ⋅ Fe•M + Fe 2SiO 4 (A7)
This is a net transfer reaction, where metal vacancies are formed by oxidation of Fe2+ on
Time Scales of Magmatic Processes 593
octahedral sites to Fe3+ and a new formula unit of fayalite is formed on the surface by transfer
of solid SiO2 (e.g., from orthopyroxene or a melt) and gaseous O2 from an external source. We
introduce activity coefficients of the structural units using
mi = mi + RT ln( γ i Xi ) (A8)
where mi is the chemical potential of structural unit i in the reference state—the perfect single
crystal. In the reference state the crystallographic sites are fully occupied with the regular
structural units without any point defect. For a solid solution like olivine the selection of a
reference state for the point defects is not as straightforward within this formalism, an olivine
with a fixed forsterite content has to be chosen (see discussion in Dohmen and Chakraborty
2007a for more details). In thermodynamic equilibrium at constant P and T the mass action
law of the above point defect reaction (Eqn. A7) can be derived:
∆G°( P, T ) aVM’’ ⋅ aFe•M ⋅ aFe2 SiO4
2 4
K = exp − = a 6 ⋅a ⋅a (A9)
RT x
Fe M SiO2 O2
Because the point defects are highly diluted we can assume that these activity coefficients are
constant for a fixed forsterite content of the olivine. If we assume in addition that Fe and Mg
in olivine mixes ideally on the metal site it follows from Equation (A9):
∆G° X V 2 ⋅ X Fe3+ 4
K = exp − = K γ ( X Fe ) X 4 ⋅ a ⋅ a (A10)
RT Fe SiO2 O2
Here a Kγ has been defined, which incorporates the effects of the activity constants and their
dependence on the forsterite content (equivalent to the molar fraction of Fe, see Dohmen
and Chakraborty 2007a, for an explicit expression for Kγ). In combination with the charge
neutrality condition (Eqn. A6) an expression can be derived from Equation (A10), where the
molar fraction of vacancies is directly given as a function of P, T, fO2, aSiO2, and XFe:
∆G° X Fe 2 / 3
X V = exp − 161/ 6 K ( X )1/ 6 ⋅ aSiO2 ⋅ aO2
⋅ 1/ 6 1/ 6
( A11)
6 RT γ Fe
∆H ° − T ⋅ ∆S ° + P ⋅ ∆V ° X Fe 2 / 3
⇒ X V = exp − 161/ 6 K ( X )1/ 6 ⋅ aSiO2 ⋅ aO2
⋅ 1/ 6 1/ 6
( A12)
6 RT γ Fe
∆Sf ∆H f + P ⋅ ∆Vf X Fe 2 / 3
⇒ X V = exp ⋅ exp − ⋅ 1/ 6 1/ 6
⋅ aSiO2 1/ 6 ⋅ aO2 1/ 6 ( A13)
R RT 16 K γ ( X Fe )
with ∆Hf = ∆H°/6, ∆Vf = ∆V°/6 and ∆Sf = ∆S°/6. These quantities are called the formation
enthalpy, formation volume and formation entropy of vacancies on the metal site,
respectively (note that the specific values of these formation energies depends on the choice of
majority point defects). Using Equations (A3, A4, and A13), a final equation for the tracer
diffusion coefficient of Mg (and analogously for Fe) in olivine for a given forsterite content is
obtained:
∗ ( ∆H f + ∆H m ) + P ⋅ ( ∆Vf + ∆Vm )
⇒ DMg = D°( X Fe ) ⋅ exp − ⋅ aSiO2 1/ 6 ⋅ fO2 1/ 6 ( A14)
RT
The above equation is strictly applicable if VM′′ and Fe•M are the majority point defects, aSiO2, fO2
are held constant at a given value (i.e., buffered), and XFe is kept constant. D° depends on the
major element composition (XFe), the migration terms may depend on this as well. Comparing
this equation with Equation (20) we can know clearly identify the parameters that are buried
within the pre-exponential and exponential factors of the Arrhenius equation.
594 Costa, Dohmen & Chakraborty
The above example can be generalized for any crystal with a charge neutrality condition
that considers only two point defects, and a fundamental equation for the tracer diffusion
coefficient can be derived which summarizes the dependencies on the thermodynamic
parameters:
∆H f + ∆H m + P ⋅ (∆Vf + ∆Vm ) m1
Di* = D° ⋅ exp − ⋅ a1 ⋅ a2 ⋅ ... ⋅ an-1
m2 mn −1
(A15)
RT
Here the exponents mi are rational numbers determined by the charge neutrality condition
and the stoichiometry of the vacancy forming reaction. In deriving such expressions it is
implicitly assumed that the vacancies equilibrate instantaneously compared to the element
diffusion time scales. The justification for this can be found in Equation (A3) which shows
that diffusion of vacancies is orders of magnitude faster than diffusion of the elements (the
factor is simply 1/XV the fraction of vacancies on the respective site). However, the kinetics of
point defect equilibration is not always rate limited by their diffusion. The rate limiting factor
could, in principle, be the kinetics of the vacancy forming reactions such as Equation (A7).
These require a net transfer of elements from or to the crystal. While macroscopic diffusion
coefficient measurements as a function of temperature yield combined values of energies and
volumes of migration and formation (e.g., Q = ∆Hf + ∆Hm), computer simulations may provide
information on these parameters separately and provide insights on diffusion mechanisms.
Reviews in Mineralogy & Geochemistry
Vol. 69 pp. 595-622, 2008 15
Copyright © Mineralogical Society of America
Mineral textures and Zoning as evidence
for Open System Processes
Martin J. Streck
Department of Geology
Portland State University
Portland, Oregon, 97207, U.S.A.
[email protected]
IntrOduCtIOn
Investigations of mineral textures and zoning as evidence for open system processes
during magmatic evolution have always been a centerpiece of petrological studies and have
provided some of the best evidence for magma mixing and crustal contamination for many
decades (e.g., Milch 1905; Kuno 1936; Eichelberger 1975; Sato 1975; Anderson 1976). In
fact, evidence from mineral studies was instrumental in the acceptance of magma mixing as
important petrological process ever since it was initially proposed by Bunsen (1851). In recent
years, mineral studies are invigorated by the development of high-precision, high-resolution
analytical instruments and techniques through which textural information and compositional
data are combined (e.g., Jerram and Davidson 2007). As minerals respond texturally and
compositionally to changing magmatic environments, they preserve in their crystal growth
stratigraphy a wealth of information regarding their past history of magmatic processes
and compositions (cf. Ginibre et al. 2007). On the other hand, magmatic liquids (melts)
are snapshots of current magmatic states, and provide less direct evidence of the processes
responsible for their evolution. In addition, liquids may crystallize thereby destroying the
direct evidence they provide. With an appropriate set of observations and measurements, we
can correlate textures with mineral compositions and thus produce a richer composite picture
of magmatic evolution than compositional data of minerals alone. This contribution reviews
the zoning and textures of minerals frequently encountered in volcanic rocks (many of which
apply equally to plutonic rocks) and the interpretations ascribed to these features in terms of
open system magmatic processes.
First, I will address the analytical tools that are used to acquire textural and compositional
data, followed by a brief review of open system processes and a summary of the relevant
textures. From that point on, I will address some mineral specific aspects followed by a
discussion of practices that yielded evidence for open system behavior in magmatic systems
based on naturally occurring mineral assemblages and populations.
MethOdS tO IMaGe textureS and quantIFy COMPOSItIOn
Standard transmitted light microscopy remains at the heart of all mineralogical
investigations and provides us with important information on mineral abundances, distributions,
sizes, and textures. Although the optical microscope reveals crystal zoning, it seldom easily
differentiates between different chemical compositions. Standard light microscopes can be
equipped with reflected light Nomarski Differential Interference Contrast (NDIC) capabilities.
Nomarski interference contrast is a reflected light, beam-splitting technique useful for
imaging surface relief. Anderson (1983) was the first to systematically apply this method to
1529-6466/08/0069-0015$05.00 DOI: 10.2138/rmg.2008.69.15
596 Streck
the study of textures of plagioclase. In addition to a standard analyzer, the technique requires
reflected polarized light and a Nomarksi prism, a double-crystal prism, which can fit either
in the “accessory plate” slot or can be placed elsewhere above the objectives. To apply this
method, samples need to be etched to develop a micro-relief (Anderson 1983 and references
therein). The micro-relief develops because different compositional zones within a crystal etch
differentially. The developed micro relief is typically less than 1 µm, and a relief of ~500 Å is
sufficient for NDIC microscopy. Etching is performed with concentrated fluoboric (HBF4) acid
in the case of plagioclase (Anderson 1983) and with HF or HCl for pyroxene and olivine (Clark
et al. 1986). Typical etch times are 20-40 seconds for plagioclase and several minutes for olivine
and pyroxene (Clark et al. 1986). Etching procedures for pyroxene and olivine will strongly
damage the rest of the thin section, while etching for plagioclase is short enough in duration
to have little impact on other minerals or glasses. This is probably the main reason why use
of NDIC microscopy is most commonly applied to plagioclase and very limited with respect
to other phases. Etching is performed on polished surfaces (thin section or grain mount), and
the better the polish the better the quality of NDIC images. Samples etched for plagioclase for
+/− 30 seconds can be used directly for microprobe analysis. I have not found any correlation
between a surface showing strong textural variation and low analysis totals. Advantages of NDIC
microscopy are: 1) low cost, 2) speed (allows for a survey of large number of crystals), and 3)
superb textural resolution, particularly for recognition of dissolution surfaces (see below).
Back-scattered electron (BSE) images obtained with a scanning electron microscope
(SEM) or electron microprobe (EMP) provide compositional data since observed intensity
of the grayscale images directly reflect differences in mean atomic weight of the material
observed (Blundy and Cashman 2008). BSE images are widely used to show zoning patterns of
a variety of minerals where compositional changes lead to changes in the mean atomic weight.
In addition, high-resolution BSE images of plagioclase can be calibrated to provide high-
resolution profiles of Anorthite (An) content profiles (see Ginibre et al. 2002 for description of
technique). Catholuminescence (CL) images can in some instances (e.g., alkali feldspar, Ginibre
et al. 2007) provide more details on zoning than BSE images or provide zoning information not
revealed by other methods (e.g., quartz, Müller et al. 2003; see review of Blundy and Cashman
2008). Elemental x-ray maps obtained most commonly with the EMP can provide superb
details concerning compositional zoning and information on the distribution of several elements
simultaneously, however, elemental x-ray maps require considerable more time to acquire.
Analytical instruments used to obtain quantitative analyses at spatial resolutions sufficiently
small to be compatible with the above techniques are: electron microprobe (for major and
minor elements and occasionally trace elements), several laser ablation mass spectrometery
techniques (mostly for trace elements but also for major elements and isotopic ratios),
secondary ion mass spectrometry (SIMS) (isotopic ratios, trace element and volatile species
abundances) and the proton microprobe (PIxE) (see reviews of Ginibre et al. 2007; Jerram and
Davidson 2007; Blundy and Cashman 2008; Kent 2008; Ramos and Tepley 2008).
BrIeF reVIeW OF OPen SySteM PrOCeSSeS
Open system magmatic behavior involves a range of processes that lead to mass and
energy exchanges between a magma body and its external environment. This includes
degassing and loss of volatiles from an ascending magma and migration of volatiles to one
magma from another degassing at depth. Here, I focus on evaluating the mineralogical record
for open system processes of magmatic recharge (mixing) and the addition of country rock
(contamination). Both processes often leave a marked record in the texture and zoning of
single minerals and in the observed mineral populations of the resident magmas as crystals and
melts are added from recharging magma or from country rock.
Mineral Textures & Zoning: Evidence for Open System Processes 597
Mixing
Mixing of magmas is now recognized to play a major role in the petrogenesis of magmas
and to operate at a multitude of spatial scales. In fact, most phyric magmas seem to indicate
some form of mixing recorded by their minerals (see Bindeman 2008; Ramos and Tepley
2008). An important question in evaluating the significance of magma mixing processes is to
determine, based upon the mineralogical record, whether a recharge event occurred and, if so,
to what extent. For example, if a magma batch “self-mixes” (e.g., Couch et al. 2001) due to
temperature differences then this may occur without new magma input (i.e., recharge), yet this
process is likely to leave traces in the mineral record as an increase in temperature may cause
mineral resorption followed by renewed crystallization (e.g., Singer et al. 1995).
Magmas involved in mixing processes often exhibit distinctive ranges in their crystal-
liquid proportions and physical properties that influence the mixing process resulting in
important consequences for the mineral record. The following are two end-member type
scenarios capturing variable conditions of mixing. During the events leading up to the 1991
Pinatubo eruption, crystal-poor basaltic magma intruded into cool, crystal-rich dacitic magma
yielding hybrid andesite and causing quenching of the mafic magma as evidenced by quenched
magmatic inclusions (Pallister et al. 1996). On the other hand, mixing of high-silica rhyolite
and basaltic andesitic magmas to yield hybrid dacite in the Rattlesnake Tuff involved magmas
that were both crystal poor and seemingly equally fluid, without immediate quenching of the
basaltic andesitic magma (Streck and Grunder 1999). Despite the differences in both of these
examples, the ranges of magma compositions involved were quite similar. In general, the
greater the dissimilarities in composition, temperature, volatile and crystal content of mixing
magmas, the greater the difference in viscosity, which strongly influences the rheological and
fluid dynamical aspects of the mixing process (see summary in Russell 1990 and references
therein). The physical properties of the mixing magmas and the mixing proportions will also
determine the extent of liquid-state blending versus solid-state disintegration of chilled blobs
of magma. One extreme case was shown by Feeley and Dungan (1996) in which magma
mixing occurred, in part, by disintegration and dispersion of crystals derived from quenched
basaltic andesitic inclusions of a mafic magma that intruded into a dacitic magma (also Clynne
1999; Tepley et al. 1999).
Mixing may also involve multiple end-members and multiple mixing events. In some
cases, the rock record may allow petrologists to disentangle a simple mixing history into a
sequence of events and discrete compositions of end-members, however in many cases this is
likely to be just an approximation of what really occurred. For example, Streck et al. (2005b)
proposed that the compositionally monotonous yet mineralogically complex basaltic andesites
erupting from Arenal volcano during the ongoing 1968-2008 eruption are a product of multi-
stage mixing during ascent.
Contamination (assimilation)
The process of contamination or assimilation can be described as the incorporation of
country rock into a magma. Most commonly, this is viewed as the incorporation of solids
that subsequently melt after incorporation (also known as bulk assimilation, cf. Edwards
and Russell 1996a). However, a partially molten country rock mush or extraction of partial
melts obtained from the country rock (known as selective assimilation, cf. Edwards and
Russell 1996a) may fall into the same category. During bulk assimilation, crystals from the
incorporated country rock may melt or dissolve and thus may blend with the resident melt.
The country rock may also simply disintegrate and disperse its crystals. Most often the actual
process involved will likely be a combination of both (addition of crystals and melts) and
will depend on composition and mineralogy of the country rock and the temperature and
composition of the formerly uncontaminated magma.
598 Streck
MIneraL reCOrd FOr MIxInG OF MaGMaS and COuntry rOCK –
PhenOCyrSt, anteCrySt, xenOCrySt
Mineralogical responses induced by open system processes can be recognized in
contrasting mineral assemblages, or in textural or composition variations including major and
trace elements or isotopic ratios within individual crystals. Larger crystals embedded in finer
crystalline or glassy matrix are traditionally called phenocrysts. Upon closer inspection, these
are defined as phenocrysts, antecrysts, and xenocrysts. Phenocrysts are crystals that grow in-
situ from a magma. Antecrysts are cognate crystals that originate from a magma genetically
related to the one in which they are found, but are older and did not grow from the liquid in
which they are found (Hildreth pers. comm.). Finally, xenocrysts are crystals derived from
country rock that have no genetic relationship to the magma in question.
Exotic mineral compositions, textures, or mineral types are usually used for determining
the incorporation of xenocrystic material, but in practice this can be difficult to apply. For
example, positive identification of xenoliths and xenocrysts, based on compositional criteria
alone, is complicated or impossible when the xenocrysts are similar in composition to
crystals growing from the uncontaminated magma. In this case, older crystal ages may reveal
xenocrystic material (Cooper and Reid 2008).
Individual crystals may have multiple contributions. xenocrystic minerals can be over-
grown by new crystal material disguising their origin further (this applies equally to antecrysts)
(Fig. 1). Further, to distinguish crystals as antecrysts requires knowing the exact conditions of
crystallization for both phenocrysts and antecrysts. Where do we draw the line between phe-
nocrysts, which record crystallization conditions of the resident magma batch, but under vari-
able conditions, and antecrysts? Evidence provided by zoning and textures within single crys-
tals and relationships among crystals of distinct crystal populations should guide our assess-
ment of whether crystals or portions of crystals are phenocyrsts, antecrysts, or xenocrysts.
500 µm
Figure 1. BSE image of a clinopyroxene crystal from a high-K basalt from the Boring volcanic field,
(B) a normally zoned, euhedral 30-80 µm wide overgrowth on resorbed, cellular-
Portland, Oregon showing
textured core. Gray shading inside from the outermost 10 µm wide, light (Fe-richer) rim, corresponds to
Mg# 88 for darkest shade and to Mg# 84 for lighter gray. Normally zoned overgrowth is phenocrystic
part while resorbed core is likely an older, 200 µm portion as suggested by compositional similarlity
antecrystic
between overgrowth and core and highlights the difficulty in classifying single crystals “wholesale” as
phenocryst, antecrsyst, or xenocryst.
Fig. 1
Mineral Textures & Zoning: Evidence for Open System Processes 599
SuMMary OF ZOnInG and textureS
In this section, I will summarize common types of zoning and textures relevant to the
discussion of open system processes. In general, zoning is the expression of compositional
variability within a single crystal. Intra-grain mineral textures refer to the internal arrangement
of morphological or other features that lead to spatial patterns within single crystals. In some
cases, both terms are nearly inseparable and thus are often used synonymously. The summary
below draws from a variety of sources with some additions by the author and attempts to contrast
terms that are opposites. Notable sources are the works of Pearce and coworkers (Pearce and
Kolisnik 1990; Pearce 1994) who have extensively studied plagioclase textures, and Hibbard
(1995) and Vernon (2004) who both provide a general treatment of zoning and textures. Other
notable contributions to zoning and textures are Vance (1963, 1965) and Anderson (1984) for
oscillatory and patchy zoning, Downes (1974) for distinguishing fine (oscillatory) from coarse
banding in pyroxene, and Shore and Fowler (1996) for discussing a large array of minerals with
oscillatory zoning.
Crystal zoning
Normal zoning vs. reverse zoning. Normal zoning includes compositional changes in
the mineral from core to rim (i.e., along the crystal growth stratigraphy) as would be induced
by compositional changes in the melt as it follows a liquid line of descent during cooling.
The changing mineral composition during crystallization reflects the progressively evolving
composition of the melt during the solidification process in a closed system. Reverse zoning is
thus any compositional inversion within the growth stratigraphy of single crystals from what
would be expected to result from crystallization in a closed system.
Step zoning vs. progressive zoning. Normal or reverse zoning can manifest itself in the
form of discrete compositional steps along analytical core-rim traverses. This is called step-
zoning. Progressive zoning is the other, more typical case that produces smooth profiles where
compositions progressively change with position along an analysis profile. Step zoning is
always associated with textural transitions (imaged by one of the techniques mentioned above),
but textural transitions can have internally concordant (euhedral) or discordant (anhedral)
relationships (Fig. 2). On the other hand, progressive zoning occurs without textural transitions
unless superimposed by oscillatory zoning. Another exception may exist if diffusional relaxation
modifies an original step zoning into progressive zoning (Costa et al. 2008).
Oscillatory zoning vs. monotonous zoning. Oscillatory zoning is defined as “composi-
tionally varying growth-shells [or layers] that are generally parallel to crystallographic planes
of low Miller indices and have thicknesses ranging from tens of nanometers to several tens
of micrometers” (Shore and Fowler 1996), and is probably the zoning style that received the
most attention. Oscillatory zoning is common to many minerals (Shore and Fowler 1996).
Furthermore, oscillatory zoning is divided into “fine banding” and “coarse banding” based
on the width of the zoning (Downes 1974). I believe separating “fine” from “coarse” banding
is a useful distinction because fine banding on the scale of single microns may in fact have a
generally different origin than “coarse banding,” as recently suggested by Ginibre et al. (2002)
for plagioclase. This distinction is also supported by findings of multiple authors on coarse
banding observed in pyroxene (see below). Fine banding may be largely kinetically controlled
while coarse banding reflects dynamic magmatic processes. Therefore, for this contribution,
oscillatory zoning is taken to be the “fine banding” (i.e., on the scale of 1-10 µm) that is as-
sociated with small-scale compositional oscillations of various forms (sinuous or irregular).
Coarse banding is also known as “growth zones,” “growth bands,” and “mantles” (e.g., Duda
and Schmincke 1985); the term “mantle,” however, should be reserved for a reaction zone or
overgrowth of a different mineral as indicated by Hibbard (1981; 1995 p. 125). In plagioclase,
fine-banding is found either as straight (euhedral) or convolute conformable layers (Pearce
600 Streck
R
(A)
90 •
• •• • •• •• •• • ••
• •
55
• • • •• •• ••
• •
•• • • • •• ••
80 • • 50 • •
•• •
45 ••• • •
An 70 •
• •••
• • •
•
• 40 •••
60 • •
•
•• 35
•
50 30
150 130 110 90 70 50 30 10 • 80 70 60 50 40 30 20 10 0
distance to rim, microns
(B)
Figure 2. Reverse and step zoning styles as exemplified by plagioclase on Nomarski interference contrast
images. (A) Mount St. Helens plagioclase of the 2004-05 dacite with dissolution surfaces (R) and saw-
toothed zoning pattern; note reverse zoning is associated with crossing the dissolution surface. (B)
Plagioclase from a basaltic andesite of Arenal volcano, Costa Rica showing progressive reverse zoning in
growth bands (segments) that are separated by sharp drops in An content without notable dissolution within
growth bands and between adjacent growth bands.
Mineral Textures & Zoning: Evidence for Open System Processes 601
1994). The opposite of oscillatory zoning is monotonous (or textureless) zoning. The develop-
ment of oscillatory vs. monotonous zoning may be directly related to the growth conditions. In
the case of plagioclase, it has been suggested, based on the observation that oscillatory zoning
is virtually absent in experimentally produced plagioclase, that formation of oscillatory zoning
(fine banding) requires growth rates smaller than ~10−10 m/s and is favored by slow growth
rates of 10−11 to 10−13 m/s (Shore and Fowler 1996). This in turn implies that monotonous (tex-
tureless) zoning may result from faster growth rates. This is in accordance with the example
of plagioclase from Arenal volcano where monotonically zoned growth bands are located at
the outer margin of many crystals that likely grew rapidly during final ascent and solidification
(Streck et al. 2005b). A similar correlation of compositions and textural relations has been
observed in a number of other locations (Shore and Fowler 1996).
Patchy zoning vs. concentric zoning. Irregularly distributed zoning patterns are commonly
referred to as “patchy” zoning (Vance 1962), in contrast to concentric zoning. Terms like
“normal” or “reverse” zoning are typically not applied to areas of patchy zoning. Although
normal and reverse zoning relate compositional information to clear relative positions within
the crystal growth stratigraphy of a crystal, similar relations can be difficult to establish for
patchy zoned areas. Patchy zoning can result when crystallization into open spaces of skeletal
crystals produced compositions that differ from the ones of the skeletal crystal host. Patchy
zoning can also be generated when a crystal is in the process of re-equilibration through
diffusion. Determination of which process is responsible for this zoning pattern is a matter of
careful analysis of the mineral in question. A possible discriminator between crystal growth
and diffusional re-equilibration as cause for patchy zoning is the sharpness of the compositional
transition (step vs. progressive) between “patches.” If the compositional change associated
with going from one patch to the neighboring patch is sharp (step-like), crystal growth is likely
(Stewart and Pearce 2004) while a progressive transition suggests diffusional re-equilibration
(Tomiya and Takahashi 2005; Streck et al. 2007) (Fig. 3).
Crystal textures
Dissolution surface vs. pervasive resorption. Resorption processes lead to removal of
previously crystallized material and they occur in different forms. In one case, resorption leads
to dissolution of crystal material progressively from the exterior of euhedral crystals leading
to various degrees of rounded, anhedral appearances. The dynamics of dissolution in a variety
of studies have been discussed by Edwards and Russell (1996b), and the interested reader is
referred to their review. Peripheral dissolution may range from subtle rounding of edges to
forming strongly rounded or embayed crystals. Pervasive resorption may affect the entire crystal
leading to an open, cellular mineral structure, which can have a wide range of appearances
and textures. Also known as “sieve” textures, cellular textures range in size from those that
are clearly discernible (on a petrographic microscope at low to medium magnification; e.g.,
Stewart and Pearce 2004) (Fig. 4) to a crystal containing a nearly sub-microscopic “haze” of
variable opacity that represent miniscule glass droplets (vitric or devitrified); for such cases,
the term “dusty” texture is also used (e.g., Tsuchiyama 1985) (Fig. 4b). Cellular textures can
have a boxy or a spongy appearance (also synonymous with vermicular and wormy). Boxy-
cellular textures are typically related to rapid growth and represent a form of quenched textures
of which hopper-shaped or dendritic crystals are well known examples (cf. Lofgren 1980).
In plagioclase, boxy-cellular texture can also be due to dissolution and recrystallization in a
reaction zone (Nakamura and Shimakita 1998). On the other hand, spongy cellular textures are
commonly attributed to pervasive dissolution.
Breakdown mantle (reaction rim) vs. growth mantle. Break-down textures result when
an existing mineral is out of equilibrium and begins to re-crystallize into a new set of minerals
instead of dissolving (e.g., Rutherford and Hill 1993). Breakdown textures may be restricted
to the rim areas of crystals producing reaction rims or may affect the entire crystal generating
602 Streck
P MI
MI
MI
P
MI
MI
100 µm (A)
(B)
200 µm
Figure 3. Two examples of patchy zoning. (A) Nomarksi image of the core of a plagioclase crystal from
Arenal volcano, Costa Rica shows skeletal higher An plagioclase (high relief; ~An89) infilled by lower
An plagioclase (denoted by “P” and lower relief; ~An82). Note sharp transition between infilled and host
plagioclase. MI denotes some of the melt inclusions present. (B) BSE image with two orthopyroxene
crystals from the Whaleback volcano near Mount Shasta, California show diffuse patchy zoning; lighter
Fig. 3 shades reflect more iron-rich portions (lightest shade ~Mg# 79 ) and darker higher Mg# (darkest shade
~Mg#86). Note younger thin, euhedral rim of renewed high Mg# pyroxene, best seen in left crystal.
Mineral Textures & Zoning: Evidence for Open System Processes 603
200 µm (A)
500 µm (B)
Figure 4. Examples of spongy celluar textures. (A) coarsely spongy cellular texture in orthopyroxene from
the Whaleback volcano near Mount Shasta, California showing resorption channels. (B) finely spongy
Fig. 4
cellular texture (also called “sieved” or “dusty”) in center and near the rim of plagioclase crystals from
Narcondam volcano, Andaman Sea.
604 Streck
pseudomorphs. On the other hand, growth mantles are overgrowths on existing minerals.
Growth occurs either because nucleation on preexisting grains is more energetically favorable
than nucleating new crystals or because the dissolution of preexisting grains produces
compositional boundary layers that facilitate growth (e.g., Coombs and Gardner 2004).
MIneraL SPeCIFIC textureS and ZOnInG
There is a vast amount of data on mineral textures and zoning that pertain to the topic
discussed here. Hence, it is only possible to showcase some aspects of mineral specific
textures and zoning features and how they are commonly interpreted with regard to open
system processes.
Plagioclase
Plagioclase displays a wide range of textures and zoning patterns, and these have been
explored as tools for investigating magma dynamics for more than a century (e.g., Milch 1905).
The diversity and details of preservation of intra- and inter-grain variability of observed zoning
and textures in plagioclase, particularly of calc-alkaline andesites to dacites, can be very large
that make the study of plagioclase challenging. When plagioclase is present, it is typically very
abundant with volumetric phenocryst percentages often between 10-30%, corresponding to
a few to many hundreds of crystals per thin section. A proper documentation of textures and
zoning of the plagioclase crystals present in a single sample can be a considerable endeavor
on its own, requiring considerable investment of time and money. Single plagioclase crystals
often exhibit a multi-stage growth and resorption history (e.g., leading to a complex array of
textures and compositions) (Fig. 5) that can be difficult to sort into phenocrystic, antecrystic
or xenocrystic portions. Kinetic controls on plagioclase growth can also lead to texturally
pronounced oscillatory zoning (e.g., Low amplitude oscillations (LAO) of Ginibre et al.
2002) superimposed on zoning and textures induced by other factors (e.g., compositional
changes during recharge or temperature increase due to self-mixing) further complicating the
assessment of features of open system processes. Here, I focus on textures and zoning that
have been ascribed to open system processes based on evidence within single crystals.
Resorption in plagioclase is evident as individual dissolution surfaces (Fig. 2A) or as a
pervasive feature resulting in boxy- or spongy-cellular textures with a diverse array of textural
forms (Fig. 4B). The reader is referred to Pearce and Kolisnik (1990) and Pearce (1994)
for detailed descriptions of these textures. Although open system processes are commonly
invoked to explain such features, verifying this through documentation and observation may
be nontrivial. When confronted with plagioclase showing cellular textures, the first question
should be to what extent the texture might be due to rapid growth. Figure 6 is one case where
a resorbed core is overgrown by skeletal plagioclase and thus provides an example where
skeletal growth can be distinguished from a resorption feature. When shapes of remaining
plagioclase in cellular zones are rounded rather than squared, skeletal growth is unlikely, as
rapid growth would favor crystallographic directions leading to straight edges (Lofgren 1980).
On the other hand, boxy shapes can still be due to resorption if dissolution was followed
by (re)crystallization, leading to square-shaped melt inclusions and a boxy-cellular texture.
Nakamura and Shimakita (1989) produced these textures by placing a seed crystal of An59
into a melt that was in equilibrium with higher An plagioclase, thus simulating magma mixing
or contamination scenarios. Based on their work, this process would result in Ca-richer
plagioclase lining the melt inclusion wall and overprinting the original Na-rich plagioclase.
An alternative mechanism to induce pervasive resorption is decompression, as postulated by
Vance (1965) and experimentally produced by Nelson and Montana (1992) in a dry system.
Decompression of water-saturated magmas would crystallize plagioclase as volatiles are
exsolved, while resorption is possible during ascent prior to water saturation (Blundy and
Mineral Textures & Zoning: Evidence for Open System Processes 605
Figure 5. A single plagioclase crystal of the current (1968-2008) eruption of Arenal volcano showing
a multi stage crystallization history as expressed by texturally different zones (after Lunney 2001).
Compositions associated with zones are as follows: core: An 95-92, hieroglyphic oscillatory z.: An 90-72,
parallel oscillatory z: An 72-60-70, boxy cellular zone: An 70-80, rim growth band: An ~60-55.
Fig. 5
(A) (B)
Figure 6. Plagioclase crystal from Kalama age andesite of Mount St. Helens with spongy cellular texture
resulting from resorption (center) and with boxy cellular textured overgrowth on left and right side
originating from skeletal growth. B highlights marked area by rectangle in (A). Light areas surrounding
crystals are vesicles and dark areas are glass. Image A is about 400 microns in horizontal dimension.
Fig. 6
Cashman 2001). Finely sieved or coarsely cellular textured resorption zones can be located
anywhere between core and rim of crystals or may affect the entire crystal (Figs. 4B, 5)
Dissolution surfaces are found in most plagioclase crystals and interpretation of their
relevance regarding open system processes strongly depends on the nature of compositional
gradients on either side of a dissolution surface (see below).
Patchy zoning of plagioclase appears to be closely related to cellular textures in as much as
spongy-cellular texture may become patchy zoning during continued crystallization. Stewart
and Pearce (2004) studied one case where plagioclase of higher An content “infilled” spaces in
pre-existing cellular plagioclase. Other studies have found infilling plagioclase to be of lower
An content (e.g., Streck et al. 2005b; Humphreys et al. 2006).
606 Streck
Reverse zoning, expressed in terms of increased An content towards the rim of a crystal,
can indicate that plagioclase grew from a more calcic melt. In turn, this can be a sign of open
system behavior in the form of influx of higher Ca/Na melt, often considered to be synonymous
with addition of a more mafic melt. However, reverse zoning can also be induced by increase
in temperature or by increase in pressure under water-saturated conditions. On the other hand,
increase in pressure under water-undersaturated conditions would decrease the An content
(Blundy and Cashman 2001). For example, water-saturated experiments using Mount St.
Helens dacites as starting material have shown that raising the temperature by ~50 °C at isobaric
conditions increases the An content by about 20 mol% (mainly because of higher Ca/Na of the
melt at lower crystallinity), while an increase in pressure of 100 MPa (corresponding to ~3.5 km)
would raise the An content by only ~10 mol% (Rutherford et al. 1985; Rutherford and Devine
2008). From this, we see that an increase in temperature is more effective in inducing changes in
An content. Recently, Blundy et al. (2006) have also shown that increase in An content related to
temperature increase can occur during ascent of water-saturated magmas, where decompression
drives crystallization. This in turn causes the temperature to increase due to release of latent
heat of crystallization. Other increases in temperature can be induced by self-mixing (Couch et
al. 2001) or due to mixing with hotter, recharging magma. To differentiate between options, it
is important to correlate changes in An content with minor or trace elements (e.g., Ruprecht and
Wörner 2007) or isotopic composition (Tepley et al. 1999; Hart et al. 2007). More specifically,
Ruprecht and Wörner (2007) suggest that when increased An content correlates with increased
Fe concentrations, it is a sign for magma mixing with more calcic magmas. If Fe concentrations
stay constant, then an increase in temperature is the likely cause. On the other hand, increases in
Fe concentrations may also be a sign of increased DFe due to higher oxygen fugacity (Wilke and
Behrens 1999) or could be achieved by dissolution of ferromagnesian phases. The evidence for
magma mixing is stronger where increases in An content correlate with changes in 87Sr/86Sr, as
this will not be affected by changes in temperature. This behavior has been observed in dacites
of Chaos Crags, Lassen volcano (Tepley et al. 1999), and in a number of other cases (see review
of Ramos and Tepley 2008). In general, the larger the increase in An content, the less likely that
temperature alone could have induced the changes.
Compositional reverse zoning observed in plagioclase comes in two different textural
associations. First, the increase in An content occurs as progressive zoning (see above) within
one growth band that may be oscillatory textured (expressed in Nomarksi images as parallel
lines) (Fig. 2B). There are no major dissolution surfaces found within the growth band,
indicating no growth hiatus during which resorption occurred. Several such progressively
zoned growth bands may be found adjacent to each other. Transition between growth bands
can be conformable or may be bound by a dissolution surface; in either case, An content
steps back down before it increases again in the next growth band. Alternatively, such growth
bands can be bound by texturally different zones, such as bands of spongy cellular (sieved)
plagioclase. In the second form of reverse zoning, the An content increase is abrupt and occurs
when stepping across a dissolution surface (Fig. 2A). Anorthite contents outbound from the
dissolution surface towards the rim are commonly highest immediately after the dissolution
surface and decrease thereafter (e.g., Tepley et al. 2000; Streck et al. 2008). Association of
several such zones leads to the well-known saw-tooth zoning pattern of plagioclase (Fig.
2A) (e.g., Pearce 1994; Ginibre et al. 2002; and references therein). Recently, Ginibre (2002)
investigated oscillatory zoning of plagioclase and found that saw-toothed shaped An content
fluctuations of larger amplitude (~ > 5 mol% An) (STR pattern of Ginibre et al. 2002) are not
true oscillations because An content increase happens as a step function associated with an
obvious growth hiatus during which an unknown amount of plagioclase was resorbed followed
by a progressive decrease before crossing the next dissolution surface. This was also noted by
Pearce (1994) for his Type 2 zones. On the other hand, low-amplitude oscillations (e.g., LAO
pattern of Ginibre et al. 2002) with An changes of ≤ 2 mol% may possibly be true oscillations,
Mineral Textures & Zoning: Evidence for Open System Processes 607
and reflect kinetic effects as opposed to saw tooth zoning that is induced by processes as
mentioned above. Furthermore, if indeed reverse zoning associated with crossing a dissolution
surface marks growth from a more calcic melt, then the more calcic melt may in fact not
entirely be due to influx of new melt. The dissolution surface is evidence that an unknown
amount of plagioclase was resorbed. This dissolution of plagioclase could have led to a rise in
the Ca/Na ratio of the surrounding melt thus causing higher An plagioclase as crystallization
resumes to overgrow dissolution surface (see discussion in Streck et al. 2008).
Pyroxene
Reverse zoning and resorption textures associated with compositional step zoning or
progressive zoning are quite common in Ca- and Na-clinopyroxenes and in low-Ca pyroxene.
In many instances, these can be interpreted as signs for open system processes, such as magma
recharge and contamination, when used in the context of the full range of pyroxene compositions
and in the context of compositional data of melts and bulk rock (i.e., data serving as test for
equilibrium). In addition, step-zoning in normally zoned segments may also be indicative of
more complex growth histories involving open systems processes (e.g., Tomiya and Takahashi
2005). Normal zoning of pyroxene from basaltic to basaltic andesitic magmas typically means
core-to-rim changes indicating less magnesian compositions that can be coupled with lower Cr
and higher Ti and Mn contents and reflect compositional changes observed in the liquid line of
descent of these magmas. The opposite is the case for reverse zoning. On the other hand, normal
vs. reverse zoning in pyroxene from intermediate to silicic (felsic) melts is not recorded by the
distribution of major and minor elements consistently in the same way, and pyroxenes may not
show significant zoning at all. For example, crystallization may lead to increases in Mg contents
of pyroxene with differentiation instead of decreases, as exemplified by clinopyroxenes from a
zoned high-silica rhyolitic tuff (Streck and Grunder 1997). Addition of the trace-element record
may also be needed to unambiguously establish the existence and style of zoning.
Reverse zoning is evident using the optical microscope when compositional contrast is
strong and leads to different color intensities with the greater color saturation in the iron-
richer pyroxene (Fig. 7). The best examples are found among clinopyroxenes from alkaline
magmas, known also as “green core” pyroxenes (e.g., Brooks and Printzlau 1978; Barton and
van Bergen 1981; Duda and Schmincke 1985; Dobosi 1989; Barton et al. 1992; see Dobosi for
earlier reports on “green core” pyroxenes). The same workers attributed the observed reverse
zoning in “green-core” pyroxene to open system processes (magma mixing or contamination).
Similar examples of reverse zoning in pyroxenes as expressed by colorless rims (rim growth
bands) on colored cores are also found among sub-alkaline varieties (Fig. 7). Anderson (1974)
and Pe-Piper (1984) interpreted reverse zoning of clinopyroxene from sub-alkaline magmas as
evidence for magma mixing. In less optically evident zoning examples, BSE images provide
the evidence for the existence of zoning in major components – as changes in Fe/Mg of the
pyroxene often lead to strong contrast differences (x-ray element maps may provide improved
resolution of zoning characteristics although these are more time consuming). Contrast
differences in BSE images may not be solely due to changes in the Fe/Mg ratio; variations in
other elements such as Ca in Ca-pyroxene may contribute to contrast differences, the extent of
which can easily be verified by elemental quantification with the electron microprobe.
In single crystals, reverse zoning can take the form of growth bands located anywhere
between the core and rim of crystals (e.g., Dobosi 1989; Nakagawa et al. 2002; Streck et al.
2002, 2005b), 2007; Tomiya and Takahashi 2005; Ohba et al. 2007) and have been interpreted
to indicate growth from more mafic magmas after initially growing from more differentiated,
more silicic (or generally felsic) magmas. Typically, these compositional changes associated
with such growth bands occur in a step-zoning fashion, unless subsequently modified by
diffusion (cf. Costa et al. 2008). Such growth bands can have euhedral (conformable) or anhedral
(unconformable) interfaces with neighboring pyroxene and can repeat themselves in a single
85
608
Mg# •••
•••••
•••••• •
(A) •••
(B)
80 ••• •
••
• •• •
• •
••• •
••• ••••••
•• ••
• ••••
• • • ••••
• •
75 • • •• • ••••••••••• •••••••• •••••••
••
••••••••••••••• •
••••• •• ••
••
•
•••••••••••••••••••••••••••••••••••••••••••••••••••••••••••• ••••••••••••
•
• ••••• •••••••••
•• ••••• •• •• • ••• ••••••••• • •••••• •• ••
•• ••••••••• • • • • •• • ••
• •• • •••••• • ••••••••••••••• ••••••• •• ••• •••••••••••••• •• •••• ••••••
••• ••••••••• •••• •• ••• •••••• ••• •• •
• ••••• • •••
•••
•
end 70 •
•
•
start •
end
•
65 start •
•
••• • • •
•
• •••• •• •
•• ••
0.2 Cr2O3 , wt% • •• • •
•
• • •• •
•
Figure 7. Reverse and step zoning •
•
0.15 •
styles in pyroxene observable by
various methods. (A) high Mg#, 0.1
200 µm wide growth band in •
•
~middle of crystal is seen under •• • •
•
0.05 • ••• • • • • • • • •
• • • •••
• •• • • • • • • • •• • • • •• ••• •
• • •• •• ••
• • • •••• • •••• •••••• •••••
• •• •
• • •• • • • • • ••
cross-polarized light; high Mg# • •
• •
• • ••• • • •• • •••••
• •• •
• • • • •••••••••• •••••••• •••••• •• •• •••• •• ••
•• •• • • • •• • • • • • • • • • •••• •• • • • •
•••••••••••••••••••••••••••••••••••••••••••• ••••••••••••••••••••••••••••••••••••••••••••••• •••••••••• •••••• •••••• • • • • •••••••• •••• •••••••••••••••••••• •••• • ••••• ••••• •••••••••
growth band (40 µm wide) at rim 0
of crystal viewed in BSE image 0 500 1000 1500
in (B) and under plain light in Mg# distance, microns
Streck
90 ••••••• ••••••• •• (E)
(D). Composition profiles along •
black line are shown in (B) of (C ••••••••
•
D)
crystal in (A) (clinopyroxene 80
(cpx) from Arenal volcano) and
in (E) for crystal seen in (C, D) 70 •
(cpx from Whaleback volcano, •••••••
California). end
60 start •
start 1.2 Cr2O3 , wt%•• •
• • •
1 • •
•
end 0.8 •• •
0.6 ••• •
0.4 •
•••••••
0.2 • •
0 •••••••
0 10 20 30 40 50
distance, microns
Mineral Textures & Zoning: Evidence for Open System Processes 609
crystal, indicating multiple growth events from more mafic magmas, or occur consistently only
as rim growth bands (e.g., Duda and Schmincke 1985; Streck et al. 2007). The interpretation
that a more magnesian composition corresponds to a more mafic melt composition, and is not
primarily due to other factors (e.g., increase in fO2 reducing the amount of Fe2+ in the melt), is
supported when changes in Mg/Fe correlate with changes of other elements. For example, Mg/
Fe can positively correlate with Cr content of the pyroxene providing additional evidence for
growth from more mafic (higher Mg, Cr) magmas.
Another zonation type that can occur within single crystals and that may be used as
evidence for open system processes is patchy zoning instead of clear growth bands. As with
the interpretation of growth bands, the extent of compositional contrast and the correlation of
changing Mg/Fe with other chemical parameters needs to be evaluated, whether the zoning
records open system processes or not. In the case of patchy zoning, however, it is often
more difficult to find evidence as to whether the open system process is magma mixing or
contamination with solid crystal debris. Patchy zoning can result when step-zoned crystals
with originally continuous growth bands or homogenous core areas are in the process of
being slowly eliminated by intra-crystal diffusional processes over time (Fig. 8) (Tomiya and
Takahashi 2005). Patchy zoning may also result from melt-crystal re-equilibration processes
of originally unzoned crystals. Such re-equilibration processes can take place when single
crystals are suspended in liquids with which they are not in equilibrium or where crystals
interact with liquids via cracks and small channels along the margins of magmatic reservoirs.
end
start
85
• •
•••• •
•
•• •••••••
• • ••
80 •
•
••• •
• • • ••
• • ••• ••
•• ••
•• • •
••• •
Mg#
• ••••••••••
75 •• •
•••••• •••••••••••• ••
•
••
• •
• •• •
••
• ••
70
end start
65
350 250 150 50
distance in microns
Figure 8. Patchy zoning as revealed on BSE image of a clinopyroxene crystal from Arenal volcano. Lower
Mg# cpx seemingly overprints higher Mg# core. Such irregular shaped patchy zoning of higher Mg# core
is interpreted to result from diffusional re-equilibration rather than by infilling and overgrowth on spongy
cellular crystal.
610 Streck
Tomiya and Takahashi (2005) and Streck et al. (2007) interpreted patchy zoning observed in
strongly resorbed orthopyroxene crystals to result from diffusional re-equilibration processes,
in the latter example caused by magmas infiltrating country rocks. I’m not aware of a case
where patchy zoning resulted from infilling of newly crystallized pyroxene into a framework
pyroxene. To decide which is the original pyroxene composition and which is the overprinting
one in patchy zoning is an important aspect for temporal relationships. This aspect should
be clarified by observations based on multiple crystals through which the overall range of
zoning patterns and textures can be established. Textural relationships of individual crystals
may be equivocal or apparent crystal stratigraphic aspects (i.e., core = older, rim = younger)
may even suggest opposite relationships. For example, crystal stratigraphic aspects of the
orthopyroxene crystal of Figure 9 suggest that patchy zoning is due to higher Mg# composition
(darker) orthopyroxene overprinting lower Mg# pyroxene because lower Mg# pyroxene also
constitutes the core of the crystal. Rather, closer inspection of the textural details of the patchy
zoned area suggests the opposite. This is also in accordance with compositional relationships
of other patchy zoned crystals of the same sample. The observed “inverted” stratigraphy of the
pyroxene of Figure 9 may result in xenocrysts with complex growth histories when they are
broken up from country rock and incorporated into a liquid dominated magmatic system.
Resorption can have different textural expressions in pyroxene. Resorption can lead to
zones of spongy texture (Fig. 7A), or pyroxene grains can also simply be rounded (e.g., Dobosi
300 µm
Figure 9. Inverted stratigraphy observed in an orthopyroxene (opx) crystal from the Whaleback volcano,
California. Lighter shaded core (~Mg# 80) of crystal grew after darker shaded opx (~Mg# 86) based on
superposition of lighter spotted area across high Mg# opx and based on zoning of opx crystals across thin
section. Very thin euhedral rim grew last.
Mineral Textures & Zoning: Evidence for Open System Processes 611
1989). Rounding of corners can be readily induced if pyroxene is exposed to a different melt
composition (P. Ulmer person. communication). Spongy textured pyroxene can be associated
with patchy zoning of the remaining pyroxene (Tomiya and Takahashi 2005; Streck et al. 2007)
and this association is probably more common than recognized to date. Resorption in pyroxene
in the form of a spongy cellular texture is a clear indication of open system behavior, while
dissolution surfaces may only be significant in the context of compositional zoning steps.
Olivine
Single olivine crystals can respond rapidly to non-equilibrium conditions by dissolution
(e.g., Donaldson 1990), diffusional re-equilibration (Costa et al. 2008) or by forming reaction
rims (commonly orthopyroxene) (Coombs and Gardner 2004). All of the above processes
can indicate open system processes. However, dissolution can also be induced by higher
temperature. Likewise, profiles of lower Fo content toward the rim do not necessarily need to be
diffusion profiles, but could also represent normal zoning induced by fractional crystallization.
Skeletal olivine can also form in response to rapid growth and needs to be distinguished from
embayed or cellular structures that result from dissolution. The presence of olivine crystals
in intermediate to silicic magmas is often taken by itself as evidence for mafic recharge and
can be supported by evaluating olivine-liquid equilibria conditions via Roeder and Emslie
(1970). Careful analysis of compositional and textural relationships among olivines of entirely
basaltic systems can yield evidence for magma mixing and inheritance. For example, Helz
(1987) shows two types of inheritance. She attributes strained olivines, similar to the one
shown in Figure 10B, to be derived from the mantle while fragmented olivine originated from
brittle failure of a prior olivine cumulate within the volcanic edifice. In either case, these are
distinct from phenocrystic olivines. Thornber (2001) also in a study on olivine from Kilauea
lavas could identify magma mixing between rift-stored and summit-derived basaltic magmas
based on the existence of reversely zoned olivines. Compositionally reverse zoning with
higher Fo content toward the rim of crystals also occurs in olivine from Stromboli and serves
as evidence for open system processes (e.g., Landi et al. 2007), but, as is the case with other
ferromagnesian minerals, the correlation with other elements (e.g., Ni, in case of olivine) is
important to exclude other factors such as changes in fO2.
Resorption of olivine most commonly leads to embayed olivine crystals. Although cases
of resorption where the olivine is pervasively affected by breakdown reactions occur (Fig.
10A), they are rare with the exception of iddingsite formation, which is a common, mostly low
temperature alteration of olivine. There are studies that have found evidence that iddingsite
forms also at high temperature (Haggerty and Baker 1967; Clément et al. 2007) but because
of the prevalence of iddingsite resulting from surficial processes, it can be challenging to
unequivocally prove a high-T origin for these samples.
amphibole
Amphiboles are compositionally complicated, and the composition reflects dependencies
on parameters such as temperature, pressure, composition, and volatile budget. This often
makes zoning patterns of amphibole difficult to interpret and difficult to relate to open system
processes. For example, higher Mg# in hornblende can readily be induced by higher fO2
in the melt (Scaillet and Evans 1999) instead of growth from higher Mg# melts. Reverse
zoning has also been related to temperature (e.g., Bachmann and Dungan 2002; Rutherford
and Devine 2003) and pressure fluctuations within the magmatic reservoir (e.g., Rutherford
and Devine 2008). Thus, reverse zoning in amphibole may best express magmatic recharge if
zoning reflects a temperature increase, as postulated for the Fish Canyon magma (Bachmann
and Dungan 2002) and if this increase exceeds values estimated by other processes such
as by magma chamber turnover. On the other hand, to infer mafic recharge from increased
Mg# requires to exclude the other factors mentioned above. To emphasize this complexity,
Rutherford and Devine (2003) found contradictory evidence in experimental amphibole of
612 Streck
200 µm
(A)
(B)
1 mm
Figure 10. Textures of olivine: (A) pervasive breakdown reaction of olivine in a fresh basaltic sample from
the Boring Volcanic Field, Portland without any sign of iddingsite alteration. (B) undulatory extension
of olivine xenocryst derived from disintegrated mantle xenoliths in an alkali basalt from Borée, Massif
Central, France.
Fig. 10
Soufriere Hills magma, where lower temperature amphiboles had higher Mg# and the higher
temperature amphiboles lower Mg#.
Resorption or recrystallization expressed as breakdown features are very commonly found
in amphibole. Rounding of amphibole or embayment features consistent with dissolution
appears less common although it can occur (e.g., Fig. 11b in Humphreys et al. 2006).
Breakdown features in amphibole leading to reaction rims of variable thickness or pervasive
alteration are very common but are difficult to resolve from those caused by destabilization
of amphibole during decompression and ascent from subvolcanic reservoirs (e.g., Rutherford
and Hill 1993; Rutherford 2008). Most reaction rims of amphibole found in volcanic rocks
are interpreted to result from such ascent-related breakdown reactions and not from open
Mineral Textures & Zoning: Evidence for Open System Processes 613
system processes. However, magma recharge can similarly cause disequilibrium conditions by
changing temperature or compositional requirements (major element, volatiles) for continued
amphibole crystallization. Distinguishing between these two scenarios can be a difficult
task. For example, a population of amphiboles in Soufriere Hills magma has unusually thick
reaction rims interpreted to result from a heating event of the magmatic reservoir prior to the
events leading to the ongoing eruption, since most other amphiboles have thin (mostly <50
µm) or no reaction rims and lower Al contents (Rutherford et al. 1998). Thus, it seems that
decompression induced breakdown reactions tend to be near rim features, while amphibole
breakdown, due to open system processes, often leads to large scale breakdown that may affect
the entire crystal or juxtaposition of nonaltered and altered amphiboles (cf. Rutherford 2008).
Another example is made by some andesitic lavas from Narcondam volcano (Streck et al.
2005a), which only contain pseudomorphs after amphibole, whereas other andesitic samples,
which are slightly more silicic, contain populations of both fresh and reacted amphiboles and
pseudomorphs after amphibole next to each other (Fig. 11). In addition, there is independent
mineralogical evidence that magma mixing was important. These samples also record an
instance where breakdown reactions are in large part due to magma mixing, and not related to
ascent from the eruption-feeding reservoir to the surface.
Overgrowth of clinopyroxene as an optically continuous single crystal on amphibole (Fig.
12), and not as a reaction zone (e.g., Rutherford and Devine 2003), seems to be a clear sign
of open system processes. It is unlikely that such a feature forms when amphibole becomes
unstable during final ascent, but overgrowth of clinopyroxene on a pseudomorph after
amphibole is consistent with crystallization during an up-temperature gradient (Rutherford
et al. 1998), possibly combined with changes in volatile saturation level as can be induced by
magmatic recharge. In many cases, however, such overgrown amphiboles are often completely
decomposed, in which case it is difficult to identify the interior of the clinopyroxene as
originally having been an amphibole (Fig. 12). In fact, samples lacking amphibole entirely but
containing clinopyroxenes with cores consisting of a microcrystalline patchwork of pyroxene,
oxide and feldspar likely indicate that this patchwork was amphibole before the amphibole
became unstable and was subsequently overgrown by clinopyroxene. The amphiboles either
originated from country rock material or from an earlier magma that crystallized amphibole.
Other major minerals
Other major minerals where zoning has been explored for inferring open system processes
are alkali-feldspar and quartz. Repetitious concentric zoning in Barium of alkali feldspar has
been imaged on BSE pictures and CL maps (Lipman et al. 1997; Ginibre et al. 2004, 2007).
This zoning may be indicative of interaction of variably evolved felsic magmas. Barium (Ba)
is rapidly depleted in the felsic melts once alkali feldspar (sanidine or anorthoclase) starts
crystallizing, thus a return to high Ba concentrations outward from a more depleted Ba core
indicates replenishment of less evolved, Ba-rich melt (cf. Ginibre et al. 2004).
zoning in quartz is revealed by CL images where brighter zones are correlated with Ti-
rich compositions that in turn are a function of higher temperature conditions (Müller et al.
2003; Wark et al. 2007 and references therein). For example, in CL-images of quartz, darker,
lower temperature cores overgrown by brighter, higher temperature quartz are interpreted to
indicate a magma heating event due to recharge of a hotter rhyolitic magma than the one that
crystallized the inner portion of the quartz (e.g., Wark et al. 2007). The recent development of
being able to tie zoning patterns in quartz to temperature estimates of crystallization sheds new
light on the interpretation of the evolution and interaction of rhyolitic magmas.
accessory minerals
Accessory phases are also subject to resorption and zoning. Their textural images, combined
with compositional information are explored for open system behavior as well. The most
614 Streck
400 µm
(A)
(B)
1 mm
Figure 11. Amphibole in two andesites from Narcondam volcano, Andaman Sea requiring disequlibrium
conditions prior final ascent; (A) juxtaposition of fresh amphibole crystal without any reaction rim (right
crystal) and amphibole pseudomorph (left crystal) implies magma mixing; (B) sample in which all
amphibole crystals are pervasively broken down into a very fine opaque mass while preserving characteristic
Fig. 11
shape; the extent of reaction is suggestive of prolonged disequilibrium conditions either as magma carrying
amphiboles was heated or ascended to a storage depth outside amphibole stability.
widespread accessory mineral used in this regard is clearly zircon, and much has been written
about its zonation patterns, which are correlated with crystal age information (e.g., Charlier et
al. 2005) or other compositional information (e.g., Claiborne et al. 2006; Bindeman et al. 2007).
The interested reader is referred to RiMG volume 53, particularly to the contribution by Cofu
et al. (2003) who compile many zoning features of zircons. Other accessory minerals showing
zonation that can be explored for open system processes are apatite (e.g., Streck and Dilles
1998; Tepper and Kuehner 1999; Broderick et al. 2007), allanite (Vazquez and Reid 2004;
Beard et al. 2006), sphene (titanite) (Robinson and Miller 1999) and others.
Mineral Textures & Zoning: Evidence for Open System Processes 615
200 µm
(A)
(B)
200 µm
Figure 12. Pseudomorphs after amphibole found in cores of clinopyroxene crystals from one Kalama age
andesitic sample of Mount St. Helens; (A) preservation of euhedral shape of former amphibole allows
identification of the core of the clinopyroxene as pseudomorph; core of crystal seen in (B), although identical
in origin as core in (A), is difficult to identify as former amphibole and instead the crystal patchwork could
have been interpreted as disequilibrium feature of a former pyroxene composition (see text).
Fig. 12
MIneraL POPuLatIOnS aS eVIdenCe FOr OPen SySteM BehaVIOr
While textures and compositional zoning of individual crystals hold a great deal of
information on whether, and to what extent, open system processes have occurred, important
insight into open system behavior can also be obtained from investigating mineral populations
(Fig. 13). Investigations of mineral populations are typically performed in concert with
investigations of textures and zoning of individual crystals that are the basis for recognizing
different crystal “types.” Investigating mineral populations can be pursued for single phases
or for phase assemblages where different mineral phases are compositionally (e.g., Mg# of
ferromagnesian silicates) or spatially (e.g., in form of mutual mineral inclusions, e.g., Freundt
616 Streck
500 µm
A
A
P1
A A
P2
A
Figure 13. Striking example for two plagioclase populations in an andesite from Narcondam volcano,
Andaman Sea. P1 type plagioclase crystals show finely sieved (spongy cellular) textured interiors rimmed
by clear plagioclase; P2 type plagioclase crystals lack any sieved portion (with one or two exceptions). “A”
Fig. 13
denotes some of the amphibole phenocrysts.
and Schmincke 1992) cross-correlated. Streck et al. (2005b) investigated mineral populations
in compositionally monotonous basaltic andesites from the current eruption at Arenal volcano
by correlating the composition of mineral inclusions with host mineral compositions in the light
of extensive data on compositional variability of plagioclase, clinopyroxene, orthopyroxene,
spinel, and olivine. Their data suggest the existence of four different mineral assemblages
that resulted from variations in liquid compositions and physical conditions of crystallization.
Several of these mineral assemblages can be found at different growth stratigraphic positions
in single crystals, and thus require open system processes as an explanation. Although Arenal
Mineral Textures & Zoning: Evidence for Open System Processes 617
is remarkable for its limited bulk compositional (basaltic andesitic) compositions, similar
examples are found at many other arc related intermediate volcanic centers where complex
textural and compositional mineral relationships indicate disequilibrium assemblages (e.g.,
Nixon 1988; Clynne 1999).
The study by Helz (1987) is an example where mostly textural (or morphological) data
from a single phase (in this case, olivine), and only to a lesser extent the distinction by chemistry,
allowed the author to establish five types of olivine that could be interpreted in terms of open
system processes. The different olivines have a relatively narrow compositional range (~Fo83-
88) and so textures were used as the basis for interpreting the origins of crystals that include
mantle inheritance (xenocrysts), inheritance from cumulitic olivine of older crystallization
episodes (~antecrysts), and phenocrystic olivine.
On the other hand, crystal populations can be purely established on compositional grounds.
Suppose clinopyroxene crystals in one dacite sample are nearly unzoned but show distinct
compositions, e.g., one at higher and one at lower Mg# (or any other compositional parameter,
e.g., trace element, isotopic ratio, etc.). In this fictive case, it is the juxtaposition of crystals
from both populations, which could not have co-precipitated, that provides the evidence for
mixing or for contamination (if one crystal type is derived from country rock). Mixing of two
or more unzoned crystal populations occurs in explosive eruptions where mixing of liquids
shortly predates or is concurrent with eruption (e.g., Boden 1989; Streck and Grunder 1999;
Cathey and Nash 2004). The short time frame during which crystals experience disequilibrium
can prevent any re-equilibration or overgrowth of minerals that otherwise occur in systems
with compositional hybridization before eruption.
Another discriminator that can be used to establish open system processes based on mineral
populations is size of crystals through the technique of crystal size distributions (CSD) (Marsh
1988; Armienti 2008). Mixed crystals populations can be interpreted from kinks, changes
in slope, or disturbances (i.e., scatter in data) in the crystal size distribution (e.g., Cashman
1990). A recent example where this was demonstrated is work by Armienti et al. (2007) on
clinopyroxenes form Mt. Etna volcano. They show distribution patterns that are disturbed
at sizes around 800 and 3000 µm. The authors interpreted this as input of larger crystals as
explanation for the disturbance at 3000 µm while the lower size disturbance is less constrained
and could signify a burst in the nucleation rate or input of crystals.
When crystals are complexly zoned, it can be difficult to find criteria to be used for
evaluation of crystal populations. For these reasons, attempts have been made to find
mathematical solutions to express the extent of similarities (e.g., in their zoning profile) that
crystals share and to establish a common magmatic heritage (Wallace and Bergantz 2005).
Crystal populations have also been investigated using cluster analysis applied to the observed
range of mineral compositions (Cortes et al. 2007). Such an approach may be most fruitful
where compositional mineral data are investigated without the context of textural information.
On the other hand, select textural or compositional criteria can be used to distinguish crystal
populations. For example, de Silva et al. (2008) contrasted reversely zoned plagioclase crystals
from complexly oscillatory zoned and unzoned crystals.
In general, most, if not all investigations dealing with mineral populations are faced with
the difficult task of deciding which individual crystals are being taken as “representative” of
a population, which then are analyzed in greater detail. Typically, visual inspection across the
spectrum of crystals observed is the basis to select a few “nice” examples for further analysis.
The more crystal rich the sample and the more diverse styles of zoning that exist within one
mineral type, the more daunting it is to pick representative examples.
To circumvent the limitation of being qualitative regarding the distribution of crystal
types in a sample, crystal populations can be quantitatively surveyed using consistent criteria
618 Streck
distinguishing crystal types – similar to measuring sizes of crystals for CSD studies but
using textural and/or compositional criteria. For example, Streck et al. (2008) distinguished
plagioclase types in Mount St. Helens samples of the 2004-05 eruption based on whether
crystals contained dissolution surfaces or not, and whether crystals contained acicular
orthopyroxene inclusions. Using these criteria, they surveyed pre-selected areas across thin
sections and “mapped” each individual plagioclase phenocryst to establish the distribution and
proportionalities among crystal types. Their study shows that ~10% of plagioclase phenocysts
of the 2004-05 eruption are void of near-rim dissolution surfaces and that those were likely
supplied to the erupting magma reservoir during the last magmatic recharge event. In a similar
study, Streck and Leeman (2008) mapped all pyroxene crystals (> ~50 µm) of one entire thin
section with regards to textural and compositional features. In this study, they were able to
show that 50% of pyroxenes (>200 µm) initially grew from silicic melt followed by growth
from high-Mg magma while the second most abundant (~37%) population had xenocrystic
characteristics; both features combined attest to the important role of open system processes
in a magma that was previously identified as being little affected by mixing and contamination
(cf. Streck et al. 2007). Depending on the scope of the study, quantitative evaluation of crystal
populations, such as the ones just mentioned above, can be used to demonstrate the volumetric
importance of a population (e.g., crystals inherited from country rock) or to establish differences
in the populations among samples (e.g., magmas). How significant the differences are, in turn,
depends on the criteria used to define populations.
OutLOOK and COnCLuSIOnS
When minerals are present, it is likely that textures and compositions reveal some features
that require mixing processes in the form of turnover, magma mixing or mixing with country
rock contaminants. Use of the mineralogical features discussed above, as mileposts to interpret
the evolutionary paths of magmas that erupt to form volcanic rocks has revealed, in many
cases, an incredible degree of complexity. This complexity seems bottomless, because as new
analytical tools are developed, additional scales of heterogeneity, which also require explanation,
are discovered. There are several areas of research that I believe need to be pursued in tandem to
continue progress towards our ultimate goal to understand how magmas form and evolve. One
area is to pursue more experimental studies geared toward understanding compositional zoning
patterns, e.g., reverse zoning, in the context of growth and resorption textures and toward how
resorption textures develop, similar to studies by Tsuchiyama (1985), Nakamura and Shimakita
(1998), and Nelson and Montana (1992) to name a few. With the lessons learned from such
experiments, we will be in a better position to determine conditions that yield observed features
of mixing and contamination. The other area that deserves attention is better quantification
of petrographic observations of crystal bearing natural samples, not only regarding size or
composition, but to combine textural records as expressed by zoning and resorption features
observed in single crystals and crystal populations. No doubt, the minimum requirement for all
this is time, if we want to reach beyond a reconnaissance study regarding information preserved
in single crystals, within single minerals, and in crystal assemblages. Surely, not every sample
can be scrutinized at the same level of detail but those that will be investigated in detail will
advance our understanding of how we can end up with what appears sometimes as a nearly
unexplainable array of crystals across a thin section of only ~4 cm in length.
aCKnOWLedGMentS
I would like to thank the editors Keith Putirka and Frank Tepley for the invitation to
contribute to this RiMG issue. Reviews by Shan de Silva, Catherine Ginibre Adam Kent and
Mineral Textures & Zoning: Evidence for Open System Processes 619
by the editors are greatly acknowledged and considerably helped to strengthen this review.
Editorial suggestions by the reviewers, the editors as well as by Karen Carroll and Sue Wacaster
are also much appreciated and improved the English.
reFerenCeS
Anderson AT (1974) Evidence for a picritic, volatile-rich magma beneath Mt. Shasta, California. J Petrol
15:243-267
Anderson AT (1976) Magma mixing: petrological process and volcanological tool. J Volcanol Geotherm Res
1:3-33
Anderson AT (1983) Oscillatory zoning of plagioclase: Nomarksi interference contrast microscopy of etched
sections. Am Mineral 68:125-129
Anderson AT (1984) Probable relations between plagioclase zoning and magma dynamics, Fuego volcano,
Guatemala. Am Mineral 69:660-676
Armienti P (2008) Decryption of igneous rock textures: crystal size distribution tools. Rev Mineral Geochem
69:623-649
Armienti P, Tonarini S, Innocenti F, D’Orzio M (2007) Mount Etna pyroxene as tracer of petrogenetic processes
and dynamics of the feeding system. Geol Soc Am Spec Paper 418:265-276
Bachmann O, Dungan MA (2002) Temperature-induced Al-zoning in hornblendes of the Fish Canyon magma,
Colorado. Am Mineral 87:1062-1076
Barton M, van Bergen MJ (1981) Green clinopyroxenes and associated phases in a potassium-rich lava from
the Leucite Hills, Wyoming. Contrib Mineral Petrol 77:101-114
Barton M, Varekamp JC, van Bergen MJ (1982) Complex zoning of clinopyroxene in the lavas of Vulsini,
Latium, Italy: evidence for magma mixing. J Volcanol Geotherm Res 14:361-388
Beard, JS, Sorensen, SS, Gieré, R (2006) REE zoning in allanite related to changing partition coefficients during
crystallization: implications for REE behaviour in an epidote-bearing tonalite. Mineral Mag 70:419-435
Bindeman IN, Watts KE, Schmitt AK, Morgan LA, Shanks PWC (2007) Voluminous low delta18O in the late
Miocene Heise volcanic field, Idaho: Implications for the fated of Yellowstone hotspot calderas. Geology
35:1019-1022
Bindeman I (2008) Oxygen isotopes in mantle and crustal magmas as revealed by single crystal analysis. Rev
Mineral Geochem 69:445-478
Blundy J, Cashman K (2001) Ascent driven crystallization of dacite magmas at Mount St. Helens, 1980-1986.
Contrib Mineral Petrol 140:631-650
Blundy J, Cashman K (2008) Petrologic reconstruction of magmatic system variables and processes. Rev
Mineral Geochem 69:179-239
Blundy J, Cashman K, Humhreys M (2006) Magma heating by decompression-driven crystallization beneath
andesite volcanoes. Nature 443:76-80
Boden DR (1989) Evidence for step-function zoning of magma and eruptive dynamics, Toquima caldera
complex, Nevada. J Volcanol Geotherm Res 37:39-57
Broderick CA, Streck MJ, Halter WE (2007) Sulfur-rich apatites in silicic, calc-alkaline magmas: inherited or
not? EOS Trans AGU 88(52), Fall Meet. Suppl., Abstract V11B-0592
Brooks CK, Printzlau I (1978) Magma mixing in mafic alkaline volcanic rocks: the evidence from relict
phenocryst phases and other inclusions. J Volcanol Geotherm Res 4:315-331
Bunsen R (1851) Ueber die Prozesse der vulkanischen Gesteinsbildungen von Islands. Annalen Phys Chem
83:197-272
Cashman KV (1990) Textural constraints on the kinetics of crystallization of igneous rocks. Rev Mineral
Geochem 24:260-314
Cathey HE, Nash BP (2004) The Cougar Point Tuff: Implications for thermochemical zonation and longevity
of high-temperature, large-volume silicic magmas of the Miocene Yellowstone Hotspot. J Petrol 45:27-
58
Charlier BLA, Wilson CJN, Lowenstern JB, Blake S, van Calsteren PW, Davidson JP (2005) Magma generation
at a large hyperactive silicic volcano (Taupo, New zealand) revealed by U-Th and U-Pb systematics in
zircons. J Petrol 46:3-32
Claiborne LL, Miller CF, Walker BA, Wooden JL, Mazdab FK, Bea F (2006) Tracking magmatic processes
through zr/Hf ratios in rocks and Hf and Ti zoning in zircons: An examples from the Spirit Mountain
batholith, Nevada. Mineral Mag 70:517-543
Clark AH, Pearce TH, Roeder PL, Wolfson RI (1986) Oscillatory zoning and other microstructures in magmatic
olivine and augite: Nomarski interference contrast observations on etched polished surfaces. Am Mineral
71:734-741
620 Streck
Clément JP, Caroff M, Dudoignon P, Launeau P, Bohn M, Cotten J, Blais S, Guille G (2007) A possible link
between gabbros bearing High Temperature Iddingsite alteration and huge pegmatoid intrusions: The
Society Islands, French Polynesia. Lithos 96:524-542
Clynne MA (1999) A complex magma mixing origin for rocks erupted in 1915, Lassen Peak, California. J
Petrol 40:105-132
Coombs ML, Gardner JE (2004) Reaction rim growth on olivine in silicic melts: Implications for magma
mixing. Am Mineral 89:748-758
Cooper KM, Reid MR (2008) Uranium-series crystal ages. Rev Mineral Geochem 69:479-554
Corfu F, Hanchar JM, Hoskin PWO, Kinney P (2003) Atlas of zircon textures. Rev Mineral Geochem 53:469-
495
Cortes JA, Palma JL, Wilson M 2007 Deciphering magma mixing: The application of cluster analysis to the
mineral chemistry of crystal populations. J Volcanol Geotherm Res 165:163-188
Costa F, Dohmen R, Chakraborty S (2008) Time scales of magmatic processes from modeling the zoning
patterns of crystals. Rev Mineral Geochem 69:545-594
Couch S, Sparks RSJ, Carroll MR (2001) Mineral disequilibrium in lavas explained by convective self-mixing
in open magma chambers. Nature 411:1037-1039
de Silva S, Salas G, Schubring S (2008) Triggering explosive eruptions – The case for silicic magma recharge
at Huaynaputina, southern Peru. Geology 36:387-39
Dobosi G (1989) Clinopyroxene zoning patterns in the young alkali basalts of Hungary and their petrogenetic
significance. Contrib Mineral Petrol 101:112-121
Donaldson CH (1990) Forsterite dissolution in superheated basaltic, andesitic and rhyolitic melts. Mineral
Mag 54:67-74
Downes MJ (1974) Sector and oscillatory zoning in calcic augites from M. Etna, Sicily. Contrib Mineral Petrol
47:187-196
Duda A, Schmincke H-U (1985) Polybaric differentiation of alkali basaltic magmas: evidence from green core
clinopyroxenes (Eifel, FRG). Contrib Mineral Petrol 91:340-353
Edwards BR, Russell JK (1996a) Influence of magmatic assimilation on mineral growth and zoning. Can
Mineral 34:1149-1162
Edwards BR, Russell JK (1996b) A review and analysis of silicate mineral dissolution experiments in natural
silicate melts. Chem Geol 130:233-245
Eichelberger JC (1975) Origin of andesite and dacite: evidence of mixing at Glass Mountain in California and
at other circum-Pacific volcanoes. Geol Soc Am Bull 86:1381-1391
Feeley TC, Dungan MA (1996) Compositional and dynamic controls on mafic-silicic magma interactions at
continental arc volcanoes: evidence from Gordón El Guadal, Tatara-San Pedro complex, Chile. J Petrol
37:1547-1577
Freundt A, Schmincke H-U (1992) Mixing of rhyolite, trachyte and basalt magma erupted from a vertically and
laterally zoned reservoir, composite flow P1, Gran Canaria. Contrib Mineral Petrol 112:1-19
Ginibre C, Kronz A, Wörner G (2002) High-resolution quantitative imaging of plagioclase composition using
accumulated backscattered electron images: new constraints on oscillatory zoning. Contrib Mineral
Petrol 142:436-448
Ginibre C, Wörner G, Kronz A (2004) Structure and dynamics of the Laacher See magma chamber (Eifel,
Germany) from major and trace element zoning in sanidine: a cathodoluminescene and electron
microprobe study. J Petrol 45:2197-2223
Ginibre C, Wörner G, Kronz A (2007) Crystal zoning as an archive for magma evolution. Elements 3:261-
266
Haggerty SE, Baker I (1967) The alteration of olivine in basaltic and associated lavas. Contrib Mineral Petrol
16:233-257
Hart GL, Streck MJ, Wolff JA (2007) Sr isotope and compositional zoning of plagioclase in the active Mount
St. Helens lava dome. EOS Trans AGU, 88(52), Fall Meet. Suppl., abstract V32C-03
Helz RT (1987) Diverse olivine types in lavas of the 1959 eruption of Kilauea volcano and their bearing on
eruption dynamics. In: Volcanism in Hawaii. Decker RW, Wright TL, Stauffer PH (eds) US Prof Paper
1350:691-722
Hersum TG, Marsh BD (2007) Igneous textures: on the kinetics behind the words. Elements 3:247-252
Hibbard MJ (1981) The magma mixing origin of mantled feldspars. Contrib Mineral Petrol 76:158-170
Hibbard MJ (1995) Petrography to Petrogenesis. Prentice Hall, New Jersey
Humphreys MCS, Blundy JD, Sparks RSJ (2006) Magma evolution and open-system processes at Shiveluch
volcano: insights from phenocryst zoning. J Petrol 47:2303-2334
Jerram DA, Davidson JP (2007) Frontiers in textural and microgeochemical analysis. Elements 3:235-238.
Kent AJR (2008) Melt inclusions in basaltic and related volcanic rocks. Rev Mineral Geochem 69:273-331
Kuno H (1936) Petrological notes on some pyroxene-andesites from Hakone volcano, with special reference to
some types of pigeonite phenocrysts. Jap J Geol Geograph 13:107-140
Mineral Textures & Zoning: Evidence for Open System Processes 621
Landi P, Francalanci L, Pompilio M, Rosi M, Corsaro RA, Petrone CM, Nardini I, Miraglia L (2007) The
December 2002–July 2003 effusive event at Stromboli volcano, Italy: Insights into the shallow plumbing
system by petrochemical studies. J Volcanol Geotherm Res 155:263-284
Lipman PW, Dungan MA, Bachman O (1997) Comagmatic granophyrice granite in the Fish Canyon Tuff,
Colorado: implications for magma-chamber processes: Geology 25:915-918
Lofgren G (1980) Experimental studies on the dynamic crystallization of silicate melts. In: Physics of Magmatic
Processes. Hargraves RB (ed) Princeton University Press, p 487-551
Lunney M (2001) Andesite magma evolution based on textural and compositional analysis of plagioclase
phenocrysts of Arenal volcano, Costa Rica. MS thesis, Portland State University, Portland, Oregon
Marsh BD (1988) Crystal size distribution (CSD) in rocks and the kinetics and dynamics of crystallization I.
Theory. Contrib Mineral Petrol 99:277-291
Milch L (1905) Über magmatische Resorption und porphyrische Struktur. Neues Jb Mineral Geol Paläont
22:1-32
Müller A, Wiedenbeck M, Van Den Kerkhof AM, Kronz A, Simon K (2003) Trace element in quartz – a combined
electron microprobe, secondary ion mass spectrometry, laser-ablation ICP-MS, and cathodoluminescenc
study. Eur J Mineral 15:747-763
Nakagawa M, Wada K, Wood CP (2002) Mixed magmas, mush chambers and eruption triggers: evidence from
zoned clinopyroxene phenocrysts in andesitic scoria from the 1995 eruptions of Ruapehu volcano, New
zealand. J Petrol 43:2279-2303
Nakamura M, Shimakita S (1998) Dissoultion origin and syn-entrapment compositional change of melt
inclusion in plagioclase. Earth Planet Sci Lett 161:1998
Nelson ST, Montana A (1992) Sieve-textured plagioclase in volcanic rocks produced by rapid decompression.
Am Mineral 77:1242-1249
Nixon GT (1988) Petrology of the younger andesites and dacites of Iztaccíhuatl volcano Mexico: I.
Disequilibrium phenocyrst assemblages as indicators of magma chamber processes. J Petrol 29:213-264
Ohba T, Kimura Y, Fujimaki H (2007) High-magnesian andesite produced by two-stage magma mixing: a case
study from Hachimantai, Northern Honshu, Japan. J Petrol 48:627-645
Pallister JS, Hoblitt RP, Meeker GP, Knight RY, Siems DF (1996) Magma mixing at Pinatubo: Petrographic
and chemical evidence for the 1991 deposits. In: Fire and mud: eruptions and lahars of Mount Pinatubo,
Philippine. Newhall CG, Punongbayan RS (eds) PHIVOLCS, quezon City, Philippines and University
of Washington Press, p 687-731
Pearce TH (1994) Recent work on oscillatory zoning in plagioclase. In: Feldspars and their Reactions. Parsons
I (ed) Kluwer Academic Publisher, p 313-349
Pearce TH, Clark AH (1989) Nomarski interference contrast observations of textural details in volcanic rocks.
Geology 17:757-759
Pearce TH, Kolisnik AM (1990) Observations of plagioclase zoning using interference imaging. Earth Sci Rev
29:9-26
Pe-Piper G (1984) zoned pyroxenes from shoshonite lavas of Lesbos, Greece: Inferences concerning shoshonite
petrogenesis. J Petrol 25:453-472
Ramos FC, Tepley FJ III (2008) Inter- and intracrystalline isotopic disequilibria: techniques and applications.
Rev Mineral Geochem 69:403-443
Robinson DM, Miller CF (1999) Record of magma chamber processes preserved in accessory mineral
assemblages Aztec Wash pluton, Nevada. Am Mineral 84:1346-1353
Roeder PL, Emslie, RF (1970) Olivine-liquid equilibrium. Contrib Mineral Petrol 29:275-289
Ruprecht P, Wörner G (2007) Variable regimes in magma systems documented in plagioclase zoning patterns:
El Misti stratovolcano and Andahua monogenetic cones. J Volcanol Geotherm Res 165:142-162
Russell JK (1990) Magma mixing processes: insights and constraints from thermodynamic calculations. Rev
Mineral Petrol 24:153-190
Rutherford MJ, Devine D, Barclay J (1998) Changing magma conditions and ascent rates during the Soufriere
Hills eruption on Montserrat. GSA Today 8:1-7
Rutherford MJ, Devine JD (2003) Magmatic conditions and magma ascent as indicated by hornblende phase
equilibria and reactions in the 1995-2002 Soufriére Hills magma. J Petrol 44:1433-1454
Rutherford MJ, Devine JD III (2008) Magmatic conditions and processes in the storage zone of the 2004–2006
Mount St. Helens dacite. In: A Volcano Rekindled: The Renewed Eruption of Mount St. Helens, 2004–
2006. Sherrod DR, Scott WE, Stauffer PH (eds) U.S. Geol Survey Prof Paper 1750 (in press)
Rutherford MJ, Hill PM (1993) Magma ascent rates from amphibole breakdown: An experimental study
applied to the 1980-1986 Mount St Helens eruptions. J Geophys Res 98(B11):19667-19685
Rutherford MJ, Sigurdson H, Carey S, Davis A (1985) The May 18, 1980 eruption of Mount St. Helens. 1. Melt
composition and experimental phase equilibria. J Geophys Res 90:2929-2947
Rutherford MJ (2008) Magma ascent rates. Rev Mineral Geochem 69:241-271
622 Streck
Sato H (1975) Diffusion coronas around quartz xenocrysts in andesite and basalt from Tertiary volcanic region
in north eastern Shikoku, Japan. Contrib Mineral Petrol 50:49-64
Scaillet B, Evans BW (1999) The 15 June 1991 eruption of Mount Pinatubo. I. Phase equilibria and pre-
eruption P-T-fO2-fH2O conditions of the dacite magma. J Petrol 40:381-411
Shore M, Fowler AD (1996) Oscillatory zoning in minerals; a common phenomenon. Can Mineral 34:1111-
1126
Singer BS, Dungan MA, Layne GD (1995) Textures and Sr, Ba, Mg, Fe, K, and Ti compositional profiles in
volcanic plagioclase; clues to the dynamics of calc-alkaline magma chambers. Am Mineral 80:776–798
Smith JV, Brown WL (1988) Feldspar Minerals. Springer-Verlag, Berlin
Stewart ML, Pearce TH (2004) Sieve-textured plagioclase in dacitic magma: Interference imaging results. Am
Mineral 89:348-351
Streck MJ, Broderick CA, Thornber CR, Clynne MA, Pallister JS (2008) Plagioclase populations and zoning
in dacite of the 2004-2005 Mount St. Helens eruption: constraints for magma origin and dynamics. In:
A Volcano Rekindled: The Renewed Eruption of Mount St. Helens. Sherrod DR, Scott WE, Stauffler PH
(eds), USGS Prof Paper 1750 (in press)
Streck MJ, Browning-Craig H, Haldar D, Ramos FC, Duncan R (2005a) Hornblende andesites/dacites in an
oceanic arc setting at Narcondam volcano, Andaman Sea, S.E. Asia. Geochim Cosmochim Acta Suppl.
69, issue 10 Goldschmidt Conference A647
Streck MJ, Dilles JH (1998) Sulfur evolution of oxidized arc magmas as recorded in apatite from a porphyry
copper batholith. Geology 26:523-526
Streck MJ, Dungan MA, Bussy F, Malavassi E (2005b) Mineral inventory of continuously erupting basaltic
andesites at Arenal volcano, Costa Rica: Implications for interpreting monotonous, crystal-rich, mafic arc
stratigraphies. J Volcanol Geotherm Res 140:133-155
Streck MJ, Dungan MA, Malavassi E, Reagan MK, Bussy F (2002) The role of basalt replenishment in the
generation of basaltic andesites of the ongoing activity at Arenal volcano, Costa Rica: evidence from
clinopyroxene and spinel. Bull Volcanol 64:316-327
Streck MJ, Grunder AL (1997) Compositional gradients and gaps in high-silica rhyolites of the Rattlesnake
Tuff, Oregon. J Petrol 38:133-163
Streck MJ, Grunder AL (1999) Enrichment of basalt and mixing of dacite in the rootzone of a large rhyolite
chamber: inclusions and pumices from the Rattlesnake Tuff, Oregon. Contrib Mineral Petrol 136:193-
212
Streck MJ, Leeman WP (2008) Complex crustal assembly of “Mt. Shasta” high-Mg andesite (HMA): Evidence
from mineral componentry. Geochim Cosmochim Acta 72:A906-A906 Suppl 1
Streck MJ, Leeman WP, Chesley J (2007) High-magnesian andesite from Mount Shasta: A product of magma
mixing and contamination, not a primitive mantle melt. Geology 35:351-354
Tepley FJ III, Davidson JP, Tilling RI, Arth JG (2000) Magma mixing, recharge and eruption histories recorded
in plagioclase phenocrysts from El Chichón volcano, Mexico. J Petrol 41:1397-1411, doi:10.1093/
petrology/41.9.1397
Tepley FJ III, Davidson JP, Clynne MA (1999) Magmatic interactions as recorded in plagioclase phenocrysts
of Chaos Crags, Lassen volcanic center, California. J Petrol 40:787-806
Tepper JH, Kuehner SM (1999) Complex zoning in apatite from the Idaho Batholith; a record of magma
mixing and intracrystalline trace element diffusion. Am Mineral 84:581-595
Thornber CR (2001) Olivine-liquid relations of lava erupted by Kilauea volcano from 1994 to 1998:
Implications for shallow magmatic processes associated with the ongoing East-Rift-zone eruption. Can
Mineral 39:239-266
Tomiya A, Takahashi E (2005) Evolution of the magma chamber beneath Usu volcano since 1663: a natural
laboratory for observing changing phenocryst compositions and textures. J Petrol 46:2395-2426.
Tsuchiyama A (1985) Dissolution kinetics of plagioclase in the melt of the system diopside-albite-anorthite,
and origin of dusty plagioclase in andesite. Contrib Mineral Petrol 89:1-16
Vance JA (1962) zoning in igneous plagioclase: normal and osciallatory zoning. Am J Sci 260:746-760
Vance JA (1965) zoning in igneous plagioclase: patchy zoning. J Geol 73:636-651
Vazquez JA, Reid MR (2004) Probing the accumulation history of the voluminous Toba magma. Science
305:991-994
Vernon RH (2004) A practical guide to rock microstructure. Cambridge Univ. Press, Cambridge
Wallace GS, Bergantz GW (2005) Reconciling heterogeneity in crystal zoning data: An application of shared
characteristic diagrams at Chaos Crags, Lassen volcanic center, California. Contrib Mineral Petrol
149:98-112
Wark DA, Hildreth W, Spear FS, Cherniak DJ, Watson EB (2007) Pre-eruption recharge of the Bishop magma
system. Geology 35:235-238
Wilke M, Behrens H (1999) The dependence of the partitioning of iron and europium between plagioclase and
hydrous tonalitic melt on oxygen fugacity. Contrib Mineral Petrol 137:102-114
Reviews in Mineralogy & Geochemistry
Vol. 69 pp. 623-649, 2008 16
Copyright © Mineralogical Society of America
decryption of Igneous rock textures:
Crystal Size distribution tools
Pietro armienti
Dipartimento di Scienze della Terra
via S. Maria 53
Pisa, Italy
[email protected]
In this overview it is planned to show how quantitative determination of mineral modes
data, by means of Crystal Size Distribution (CSD), may help in deciphering the kinetics of
magmatic crystallization both in natural and experimental conditions. Some stereological
considerations will be discussed to provide criteria for the choice of numbers and lengths of
size intervals for CSD computations. Equations for the balance of number and size of crystals,
in steady state crystallization and in dynamic regime, will be used to define the time frame of
magmatic processes, in the context of the case studies of Stromboli and Etna volcanoes.
The functional dependence of crystal size (L) and crystal number from time is explicitly
derived by interpreting experiments of olivine crystallization that report CSD data, thus
obtaining evaluation of thermodynamic parameters from CSD measurements.
CryStaL SIZe dIStrIButIOnS
Modal data obtained from rock sections can be represented as a Particle Size Distribution:
N(L) that is the number of objects (crystals, bubbles) of given size L contained in the unit
volume. N(L) is usually displayed in semi-logarithmic diagrams vs. a characteristic length of
the grains (Fig. 1; Marsh 1988). In the following discussion the size L is the diameter D for
spherical particles or the diameter of the sphere with the equivalent volume of the particle. In
igneous petrology, Crystal Size Distributions (CSD) and Bubble Size Distributions (BSD) can
be related to the cooling and degassing history of a magmatic body since:
25 Sample 170392
Plagioclase CSD Figure 1. Particle Size
Distribution: N(L) represents
20
the number of objects (crystals,
Ln (N(L))
bubbles) of given size L
contained in the unit volume.
15 N(L) is usually displayed in
semi-logarithmic diagrams
vs. a characteristic length of
10 the grains. Sample 170392
is taken from the flowing
lava flow during the 1991-93
eruption of Mt. Etna (Armienti
5 et al. 1994).
0.00 0.05 0.10 0.15 0.20
D (cm)
1529-6466/08/0069-0016$05.00 DOI: 10.2138/rmg.2008.69.16
624 Armienti
• crystal and bubble populations are generated by nucleation and growth processes
that are strongly dependent on the degree of undercooling (or supersaturation) and,
ultimately, by the temporal evolution of pressure and temperature paths of magma
bodies.
• the number of crystals and bubbles which nucleated and grew in the magma can
be balanced to account for the loss or gain of grains due to some mechanism of
transport.
This allows the extraction of temporal information from lava modes if we know how crystals
and bubbles nucleate and grow during magma transport and cooling processes.
Since the number of crystals depends on the process of rock formation, CSDs may be
profitably used to describe and model also metamorphic and sedimentary rocks, accounting
for the laws that allow grain nucleation and growth or transport and deposition.
deterMInatIOn OF the CurVeS OF POPuLatIOn denSIty
Population density N(L) is defined as the number of particles in each size interval divided
by the size of the interval, like in Figure 2—shortening the interval, the histograms tend to
a continuous function. Note that N(L) is to be determined for each class of size for the unit
volume. So, N(L) will depend on ∆L and will tend to zero for infinitesimal intervals. Often
measurements of N(L) for very small classes are difficult, thus it is preferred to determine the
cumulative crystal number
L
n( L ) = ∫ N ( L )dL (1)
0
Obtaining N(L) through derivation of the cumulative curve
dn( L )
N (L) = (2)
dL
10000
8000
6000
N(L)
4000
2000
0
10 9 8 7 6 5 4 3 2 1
L (size)
Figure 2. Population density N(L) represents the number of particles
in each size interval divided by the size of the interval.
Crystal Size Distribution Tools 625
A graphic example of the
20000
derivation of the cumulative curve
N(L) is reported in Figure 3. The
most common methods to obtain
cumulative curves of particle sizes
are by sieving of loose materials
or by stereologic correction of a
n(L)
10000
distribution of areas measured in
rock sections to get a distribution
of sizes in the space.
N(L)
Planar distributions are mea-
sured by collecting data related to
features of the particles such as 0
diameters or radii (for spherical 12 10 8 6 4 2 0
particles), cord lengths (for generic L (size)
shapes), and areas (for generic
shapes). All these features may be Figure 3. Derivation of the cumulative curve N(L).
obtained through handling an im-
age of a rock, in which we distin-
guish the phase of interest from the
groundmass and the other phases. Image analysis software provides methods of segmentation of
the images which puts in evidence only the phase of interest. Commercial packages like Visilog
® or open source codes like Image-J are particularly suitable to handle images of rocks that may
be acquired by cameras mounted on a microscope, or from normal scanners. These may be eas-
ily modified to hold polished rock slabs, while film scanners may be equipped with polaroids to
acquire images from rock thin sections (Armienti and Tarquini 2002).
analysis of distributions of areas
Segmentation puts in evidence the pixels belonging to the phase of interest. Segmentation
is easier if there is a strong contrast between the phase and the background. In Figure 4
is a BW image of Hawaiitic lava from Mt. Etna where magnetite is black. Selection of a
suitable threshold promptly allows acquisition of a binary map of its distribution where pixels
belonging to magnetite are set to equal 1 (black) and the remaining are set to equal 0 (white).
In the example of Figure 5, plagioclase, olivine, clinopyroxene and vesicles appear to be light
colored and the lower and high grey level thresholds were chosen to separate them (white)
from the glass and magnetite. More complex situations (e.g., distinction among light colored
phases) require some manipulation of images and different techniques of acquisition.
The final goal of measuring size distribution requires a further step, involving assumptions
of the grain shapes and a suitable stereologic treatment of planar data must be made to access
information in the space.
Stereologic reconstruction of CSd: monodispersed distributions
The reconstruction of the spatial distribution of crystals in rocks is based on the collection
of planar data collected on the surface of rock sections. This requires that some assumptions on
the grain shapes must be made a priori since spatial distributions of particles with different sizes
and shapes can generate the same planar size distribution (Fig. 6). We will assume in the fol-
lowing treatment that minerals and bubbles can be approximated as spherical particles, and the
analysis which follows is applied to the radii of the circles of equivalent area. This approxima-
tion involves relevant deviation from reality only for strongly elongated minerals but is sufficient
for most of the minerals occurring in lavas (Higgins 2000, 2002; Morgan and Jerram 2006).
626 Armienti
FIG 4
Figure 4. An example of segmentation through thresholding. (left) A BW image of Hawaiitic lava from
Mt. Etna where magnetite is black. (right) Acquisition of a binary map of its distribution where pixels
belonging to magnetite are set to equal 1 (black) and the remaining are set to equal 0 (white).
FIG 5
Figure 5. An example of segmentation through thresholding. (left) A BW image of Hawaiitic lava from
Mt. Etna where magnetite is black. Plagioclase, olivine, clinopyroxene and vesicles appear to be light
colored. (right) Lower and high grey level thresholds were chosen to separate them (white) from the glass
and magnetite.
Crystal Size Distribution Tools 627
Figure 6. Case A is that of a mono-
dispersed distribution of spheres (all
have the same diameter); this situation
may occur in well sorted sedimentary
rocks, in thermometamorphic rocks in
correspondence of a single nucleation
event, as well as in artificial aggregates
of materials, like concrete. Case
B illustrates how the same planar
B distribution can be generated by a poly-
dispersed distribution of diameters. Case
C illustrates the distribution of areas is
the same as in A and B, but only one
section cannot give information on the
spatial distribution.
FIG 6
For the simple case of a mono-dispersed distribution of spheres of diameter Dj the
following relation holds:
NV = NA / Dj (3)
where NV is the number of particles per unit volume, NA is the number of particles per unit area,
and Dj is the constant diameter of spheres. In Figure 7 only the spheres whose center is at a
distance less than Dj/2 are cut by the section (black line) of unit area. The number of sections
per unit area is NA and consequently the number per unit volume results to be NA / Dj.
Stereologic reconstruction of CSD: poly-dispersed distributions
When dealing with crystals and vesicles nucleated at different times, we can assume that
planar distributions are generated by particles of different sizes cut at different distances from
their center (case B in Fig. 6). There are several methods to determine, in the space, a CSD of
a dispersed phase from 2D information (Russ 1986). A simple conversion from apparent 2D
to volumetric size population may be performed considering the planar distribution as if it was
collected in the space, by rising NA to the power 3/2 (Kirkpatrick 1977a)
NV = NA3/2
628 Armienti
In this way, the contribution of larger spheres to smaller areas is ignored. In fact, small circles
on a section can be originated by small spheres but also by larger ones, cut at distances from
their centers almost equal to their radii (Fig. 8). Trivially smaller spheres cannot contribute to
larger size classes.
A method of “unfolding” of the areal size distribution was independently by Schwartz
(1939) and Saltykov (1949) and is now identified as the Schwartz-Saltykov method (DeHoff
and Rhines 1972). It is particularly suitable for digitized images where particle areas can
assume only discrete values, as some multiple of the pixel resolution. The algorithm is based
on the consideration that the diameter of a particle in a plane depends on where the section
cuts it, thus a frequency histogram for circle sizes obtained by random sectioning of a sphere
can be computed.
If we measure in the plane a section of size di, let:
• p(i,j) be the probability that a sphere of diameter Dj > di is cut in such a way to
produce a section of diameter di.
• NA(di,Dj) be the number of sections of diameter di (measured) per unit area obtained
from spheres of diameter Dj.
The true number of spheres of diameter Dj per unit area is then given by
N A (d j , D j )
N A (Dj ) = ( 4)
p( j , j )
and the problem to solve is to determine the values of p.
Assuming a “suitable” class interval ∆ we can generate j classes by cutting a sphere
of diameter Dj at a distance ∆ from each other. The probability of cutting a section with a
D1
D2
D3
D4
D5
D6
Figure 8. Smaller spheres (marked with the cross) cannot contribute to the population of larger ones.
FIG 8
Crystal Size Distribution Tools 629
diameter between di = i × ∆ and d(i−1) = (i × 1) × ∆ (with 0 < i ≤ j), is directly proportional to
the thickness h of the sections and inversely proportional to the radius of the sphere:
h h(i −1) − hi
p(i, j ) = = (5)
Dj Dj
2 2
where h(i−1) and hi are the distances from the center of the sphere to which diameters d(i−1) and
di respectively correspond (Fig. 9). On the basis of this probability, the Schwartz-Saltykov
method to “unfolds” the contribution of larger spheres to the population of smaller ones by
using a technique of iterative subtraction.
In order to express Equation (5) as a function of i, j and Di from Figure (5) above we
observe that
2 2
D d
hi = j − i (6a )
2 2
from which
2 2
j i
hi = ∆ ⋅ − ∆ ⋅ (6 b)
2 2
and after substitution in Equation (5) and a little manipulation
2 2
i −1 i
p(i, j ) = 1 − − 1− (7a )
j j
2 2 2 2
D j hi −1 D
− j − i
h
2 − 2 2 2
p(i, j ) = (7b))
Dj
2
d(i −1 )
Dj 2
h( i −1 )
hi d i
h
Figure 9. Relations between surfaces cut from a sphere and probability of cutting definite values.
FIG 9
630 Armienti
In Figure 10, the values of p(i,j) are computed for twenty distinct classes (j = 20). It is
evident that the largest contribution of class j is to the class j itself.
By substitution of p, Equation (4) then becomes:
N A (d j , D j ) N A (d j , D j )
N A (Dj ) = = (8)
p( j, j ) j −1
2
j
2
1− − 1−
j j
which provides the number of spheres of diameter dj = Dj per unit volume, while the number
of spheres with apparent diameter di < Dj provided by spheres of diameter Dj is
N A ( di , D j ) N A ( di , D j )
N A ( di ) = = (9)
p(i, j ) i −1
2
i
2
1− − 1−
j j
That guarantees the possibility of calculating the values on NV(Dj) for each class given that at
least one NV(Di) is known. Consider Figure 7 with six classes of spheres D1, D2, … , D6: we do
know the value of NV(D6) of the largest class (the one that contributes only to itself)
N A (d6 , D6 ) 1
NV ( D6 ) = (10)
p(6, 6) D6
Equation (8) then allows one to evaluate the contribution of the class Dj to the class Dj−1 and
so on, until the smallest class D1. If this operation is accomplished for each size interval the
unfolding will yield the spatial distribution of each class of spheres. For class 5
N A ( d5 , D5 ) N A ( d5 , D6 )
N A ( d5 ) = − (11)
2 2 2 2
5 −1 5 5 −1 5
1− − 1− 1− − 1−
5 5 6 6
0.4
0.3
probability
0.2
0.1
0.0
1
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
0.05
0.15
0.25
0.35
0.45
0.55
0.65
0.75
0.85
0.95
i/j
Figure 10. Distribution of probability of a class to contribute to the class itself.
FIG 10
Crystal Size Distribution Tools 631
For class 1
N A ( d1 ) = N A ( d1 , D1 ) + N A ( d1 , D2 ) + N A ( d1 , D3 ) + N A ( d1 , D4 ) + N A ( d1 , D5 ) + N A ( d1 , D6 ) (12)
and in general
N A (di ) = ∑ N A (di , D j ) (13)
which simply states that the contribution to class i in the section is due the number of sections
of class 1 plus the sum of contributions due to larger spheres. Since NA(d6,D6) is measured we
can obtain NA(d6,D6) from Equation (10) and Equation (7) provides NV of the particles of group
6 (Eqn. 10), which in turn allows one to calculate NA(d5,D5). Then
N A ( d5 , D5 ) = N A ( d5 ) − N A ( d5 , D6 )
is computed until unfolding is completed for the contribution to the smallest class of all the
other particles.
the “best” CSd for a set of unfolded planar data
It is evident that the shape assumed by the CSDs of the same set of planar sections is
strongly controlled by the number of intervals and by the size of the first class. Figure 11
and Table 1 report the different CSDs obtained for the same set of 3497 planar sections of
crystals (total area of 3693 mm2) measured on a slab of 6587 mm2. The Delesse theorem
(1847) ensures that the volume fraction of the measured particles corresponds to the area
fraction (vol% = 100 × 3693/6587 = 56.10%). On the basis of which elements can one choose
the most suitable partitions of a data set?
We can start considering that two fundamental measured quantities must be conserved
in the CSD, namely the volume fraction and the total number of particles. However when
arranging the binning of our data set, these two values must be conserved. We can obtain
these two quantities from CSDs: in each class of size Li, the number of sections of that size is
given by the product of their volumetric number density [NV(Li)] and the sampled volume for
particles of that size [Area × Li ]. Thus the total number of particles of a given CSD is:
N tot = Area ∑ Li NV (Li ) (14)
0.250
CSDs of table 1
0.200
ln (Nv(L)mm-4)
0.150
0.100
0.050
0.000
0 0.5 1 1.5 2 2.5 3
L (mm)
# of classes 13 16 22 27
Figure 11. Choice of the best possible distribution.
632 Armienti
table 1. Different CSDs computed from the same areal distributions.
∆L = 0.210 ∆L = 0.168 ∆L = 0.120 ∆L = 0.101
L Na(L) Nv(L) L Na(L) Nv(L) L Na(L) Nv(L) L Na(L) Nv(L)
−3 −4 −3 -4 −3 −4
class (mm) (mm ) (mm ) (mm) (mm ) (mm ) (mm) (mm ) (mm ) (mm) (mm ) (mm−4)
−3
1 0.105 91 0.000 0.084 91 0.062 0.060 91 0.114 0.050 91 0.233
2 0.315 438 0.165 0.252 190 0.006 0.180 190 0.083 0.151 0 0.142
3 0.525 275 0.010 0.420 523 0.182 0.300 248 0.076 0.252 190 0.052
4 0.735 562 0.100 0.588 283 0.043 0.420 275 0.068 0.352 248 0.067
5 0.945 517 0.085 0.756 279 0.014 0.540 283 0.060 0.453 275 0.071
6 1.154 438 0.075 0.923 517 0.109 0.660 279 0.052 0.554 283 0.074
7 1.364 186 0.006 1.091 230 0.012 0.779 267 0.046 0.655 279 0.097
8 1.574 308 0.037 1.259 394 0.073 0.899 250 0.040 0.755 0 0.071
9 1.784 236 0.027 1.427 164 0.015 1.019 230 0.034 0.856 267 0.046
10 1.994 192 0.025 1.595 144 0.006 1.139 208 0.029 0.957 250 0.042
11 2.204 80 0.005 1.763 236 0.035 1.259 186 0.025 1.057 230 0.038
12 2.414 117 0.015 1.931 101 0.004 1.379 164 0.021 1.158 208 0.037
13 2.624 57 0.009 2.099 171 0.025 1.499 144 0.017 1.259 186 0.045
14 2.267 66 0.007 1.619 126 0.015 1.359 0 0.033
15 2.435 51 0.005 1.739 110 0.012 1.460 164 0.022
16 2.602 57 0.010 1.859 101 0.011 1.561 144 0.019
17 1.978 91 0.010 1.662 126 0.016
18 2.098 80 0.010 1.762 110 0.015
19 2.218 66 0.008 1.863 101 0.020
20 2.338 51 0.007 1.964 0 0.015
21 2.458 33 0.005 2.064 91 0.011
22 2.578 24 0.005 2.165 80 0.010
23 2.266 66 0.009
24 2.366 51 0.008
25 2.467 33 0.007
26 2.568 0 0.006
27 2.669 24 0.005
Computed values
Ntot 3497 3497 3497 2937
% vol 60.70 58.63 53.69 31.8
Measured values
Area = 6587 mm2 Ntot = 3497 Area Particles = 3693 mm2 % area = % vol = 56.10%
Notes: Ntot = Σ[NV(Li) × Li] × Area; % vol = 100 × Σ[NV (Li) × Li3] × π/6
Similarly, the volume fraction of particles of size Li is the product of their volume [4/3π(Li / 2)3]
and their number in the volume sampled for that class size [Area Li NV(Li)] divided the sampled
volume [Area Li]. Thus the volume fraction of all the particles of a given CSD is:
3
4 L π
V f = ∑ NV (Li ) π i = ∑L
3
NV (Li ) (15)
3 2 6 i
or
100 π
∑L
3
V% = i NV (L i ) (16)
6
Equations (14) and (16) allow an easy choice of the “best” CSD that results so that the
difference between the number of computed sections and the number particles derived from
Crystal Size Distribution Tools 633
CSD (Eqn. 14) and the difference between the volume fraction derived from the section areas
and that derived from CSD (Eqn. 16) are minima.
Truncation effects
CSD data may show truncation effects arising from resolution (left-hand truncation - LHT)
of the adopted measurement device and from the size of the specimen (right-hand truncation -
RHT). Evaluation of left-hand and right-hand truncation effects on the CSD is essential before
attempting any interpretation (Pickering et al. 1995). Left-hand truncation arises from the reso-
lution of the image analysis procedure as a whole and is revealed by a decrease in the number
density of the counted grains. This truncation is reduced when data are acquired with a greater
resolution (Armienti et al. 1994). It is important to observe that resolution is controlled both by
the resolution of the starting image and by errors introduced by image processing. Thus, due to
the heavy image manipulations needed for image segmentation, in spite of the high resolution
of starting images, often CSDs lack LHT effects only for crystal sizes much larger than resolu-
tion (Fig. 12). Any interpretation of a ‘‘true’’ reduction of the number density of the smallest
grains as due to some processes require the acquisition of images with a higher resolution.
Right-hand truncation produces the flattening of size distributions on nearly horizontal
trends at large size ranges in which not all the classes of size occur in the sections (Fig. 12). This
results in a constant value of the number density for classes that contain at most one particle.
A logarithmic binning of size classes (i.e., logarithmic variation of the width of the class
size) allows to partly avoid this effect by counting a larger number of grains in larger (and
wider) classes, but the effect still appears at the limit of the distribution with a decrease of
the slope at the right hand of the distribution. Right-hand truncation effects can be limited by
enlarging the size of the measured sample. This is obtained by measuring CSDs of several
images of the rock in different positions. In general the statistic reliability of data has always
to be checked before attempting any interpretation. In the case of particle size distributions we
must be sure to analyze an area that is sufficient to guarantee that the largest particle detected
has the “true” number density and is not affected by RHT effects, otherwise the CSD plots will
be “sparsely” populated by points at larger diameters. In this case the N(L) does not represent
measurement of the number density but is related to “isolated” casual findings of particles of
that size in the explored section.
17.5
15
12.5
Ln n(L)
10
7.5
5
0.0 0.2 0.5 0.8 1.0
D(cm)
Figure 12. Right-hand truncation effect.
634 Armienti
The minimum area to be analyzed depends on the number of particles and the width of the
classes L according to the formula (Armienti et al. 1994)
1
Area = (17)
2 N ( L ) ∆L L ∆L
for the data in Figure 13a, in which larger bubbles have L = 0.15 cm, ∆L = 0.0036 cm, and N(L)
≈ e8 cm−4. The minimum area must not be less than 2 cm2 to have a statistical significance. In that
case the analyzed area was 3 cm2. This implies that the continuous distribution up to 0.2 cm is
statistically reliable, that the isolated point at about 1 cm must be discarded and that a larger area
is needed to obtain the “true” N(L) of bubbles larger than 0.2 cm. The problem can be solved
by using different samples for different grain sizes—thin sections, polished bricks and photos
of the outcrops—then merging the size distributions obtained at different scales. This was done
FIG 13
Figure 13. Effects of left-hand and right-hand truncation shown for samples LANz 24 and LANz 15. (a)
The typical effect of left-hand truncation is to reduce the number of counted individual crystals near the
resolution limit of the acquisition procedure. This can be observed both in the CSDs computed from one
thin section (circles) and from 10 thin sections (dots). Right-hand truncation arises from the limited size
of the explored thin section that allows the detection of only a few (at the limit only one) particles of large
size in each size interval: this results in the constant value of N(D) revealed by the flattening of the CSD
trend (empty circles). For the same choice of size interval, the effects may be avoided by analyzing a larger
population of crystal (dots). A different choice of size intervals would also reduce this effect as is evident
in the CSDs obtained from computed logarithmic binning of the size interval (squares). (b) Right-hand
truncation may not disappear, even when handling a large population of crystals (small diamonds) if the
number of size intervals is too large: this is particularly evident in sample LANz 15, where CSD is plotted
against D. In this case, right-hand truncation is revealed by the constant values of log(N(D)) and by the shift
of linearly binned CSD with respect to the logarithmic one. From Armienti and Tarquini (2002).
Crystal Size Distribution Tools 635
for the data in Figure 14, which reports the Bubble Size Distributions in a lava flow from Mt
Etna. Note the different sizes at which RHT effects appears when larger areas are acquired. The
adoption of a logarithmic binning for the calculation of CSD on a single section may also ensure
good results since it increases the value of the denominator of Equation (17), but it is important
to check that CSDs satisfy the conditions imposed by Equations (14) and (16).
5
Thin section (7 cm2)
0 Polished rock slab (122 cm2)
Ln N(L) av 3
Photo of outcrop (808 cm2)
-5
-10
-15
0 5 10 15 20 25 30
D (mm)
Figure 14. Bubble Size Distributions in a Mt. Etna lava flow collected with different tools (thin section,
a polished slab, a photo of the outcrop) that explore different total area, providing different RHT. Note
overlap of dominions not affected by under-sampling.
CSds and the PetrOGenetIC COntext
Crystal Size Distributions (CSD) are suitable to be interpreted in a petrologic context
since i) crystal populations are generated by nucleation and growth processes that are strongly
dependent on the degree of undercooling (and thus on the thermal history); ii) the number of
crystals which nucleated and grew in the magma can be balanced to account for the loss or
gain of grains due to some mechanism of transport (fractionation, mixing).
The first step in the interpretation of CSD diagrams is to write down equations that balance
the numbers of crystals
Number of crystals Growth Input − Growth Output
in unit volume of size = +
between L and L + ∆L Flux In − Flux Out
or
(V2 N 2 − V1N1 ) ∆L = ( G1V1N1 − G2V2 N 2 ) ∆t + (Qi Ni − Qo N o ) ∆L∆t
that has the differential form
∂(VN ) ∂(GVN )
+ = Qi N i − Qo N o (18)
∂t ∂L
where N(L) = crystal density (N(L)dL = number of crystals per unit volume with size between
636 Armienti
L and L+ dL), G = crystal growth rate; V = volume; L = crystal size (e.g. diameter); t = time;
Qini − Qono = Flux in − Flux out.
Equation (18) can be solved under different assumptions which result in different analytical
solutions. I will illustrate two different approaches to the solution of Equation (18) which can
be related to different petrogenetic mechanisms.
Continuously taped and refilled magma chamber: the Stromboli example
This situation may occur in stationary systems like a fractionating magma chamber,
continuously refilled and taped by eruptions. The typical case is represented by the activity of
Stromboli (Armienti et al. 2007).
In this case, Equation (18) may be solved with the following assumptions:
a) Crystallization takes place in a magma chamber of volume V
b) which is refilled at the rate Q
c) by a crystal free magma
d) and is continuously taped by eruptions at rate Q.
In this case,
In flux of crystals = QN i = 0( crystal free magma ⇒ N = 0 ) i
Out flux of crystals = QN o = VN o / τ ( τ = time of recharge )
which allows us to simplify Equation (18) into
∂(GVN ) N
= (19)
∂L τ
For constant G (Marsh 1988)
N = No exp (−L / Gτ) or ln(N) = ln(No) − L / Gτ
where τ = V / Qout is defined as the time whereupon the system of volume V is emptied or
totally exchanged at a flux rate of Qout, growth rate G does not depend on L, and the integration
constant No represents the density of crystal nuclei (i.e. the value of N as L > 0). In this case, in
a plot of N(L) respect to L, the slope of a linear distribution of N(L) is linked to the product Gτ
by the relation slope= −1/(Gτ) and, if G or τ are known, the other variable may be estimated by
measuring the slope of the distribution. The linear distribution of CSDs provides information
on the characteristic time of recharge of the system. In the example of Figure 14 (Stromboli)
the slope is −6.6242 mm−1 that for a growth rate on the order of 10−9 cm sec−1 would give a
residence time of ~1.5 × 107 sec, which corresponds to about 174 days.
The linear relationship of Figure 15 is typical of Stromboli and and additional data reveal
almost constant slopes for at least 1000 years (Armienti et al. 2007). However, even in such
an apparently simple system, the stationary condition of equilibrium is achieved through
continuous shifts from equilibrium and determination of the magma residence time through
the evaluation of the CSD slope may not be so direct as it appears.
The residence time of the Stromboli shallow feeding system has been estimated to be
19 years on the basis of Sr isotopic data (Francalanci et al. 1999). This estimate of residence
time would imply a plagioclase growth rate on the order of 2 × 10−11cm s−1. This seems too
low considering the results obtained on basaltic systems which average 10−9-10−10 cm sec−1
(Kirkpatrick et al. 1976; Lofgren 1980; Armienti et al. 1994). However, the obtained low growth
rate of the plagioclase must be the result of an average behavior which accounts for the peculiar
textural and chemical zoning of the plagioclase. In fact, plagioclase crystals in the crystal-rich
scoriae show alternating layers of labradoritic and bytownitic composition, that allow one to
recognize distinct recurrent phases of growth and dissolution (Landi et al. 2004).
Crystal Size Distribution Tools 637
6 Stromboli
lava flow Jan 15 2003
4 Plagioclase CSD
18745 crystals
2
Ln (N(D)) ( mm-4 )
0 Ln (N(L)) = -6.6242 D + 4.6224
-2
-4
-6
-8
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
D (mm )
Figure 15. CSD of a Stromboli lava flow. Data were collected on several sections. Note the RHT effect on
larger crystals in spite of the arithmetic increase of the binning of size classes.
The process suggested by Figure 16 leads one to assume that the growth rate G derived
from CSD data is a mean value resulting from growth and dissolution episodes:
G = G+ f + + G− f − (20)
where G+ and G− are mean growth and dissolution rates, respectively, and f + and f − are the
volume fractions of the system in which crystals grow or dissolve respectively. For a system
that behaves like a crystallizing system, we can assume f + to be proportional to the volume
fraction of magma supplied in a characteristic time:
f + = (V − Q × ∆t ) / V (21)
where Q is the magma supply rate, which in a steady system is equal to the mass eruption
rate, and ∆t is a characteristic time during which resorption events may occur due to mixing of
undegassed magma with the resident magma.
From the obvious relations f + + f − = 1 and V / Q = τ, it follows:
f + = 1 − ∆t / τ and f − = ∆t / τ (22)
from which
∆t ∆t
G = G + 1 − + G − (23)
τ τ
which states that the net growth rate is a mean of positive and negative (i.e., resorption) growth
rates weighted for the times in which the two conditions hold. The internal consistency of
Crystal Size Distribution and residence times based on isotopic disequilibria may be assessed
by means of the above equation. Assigning different values for G+, the value of G− may
be computed from Equation (23) for different choices of ∆t. In correspondence with these
characteristic times of dissolution, we can compute the amount of dissolution allowed for
crystals (G− × ∆t). Thus a plot ∆t / τ vs. (−G− × ∆t) for different G+ (Fig. 17) will provide an
estimate of the characteristic size of the crystal that will be completely resorbed. It is worth
638 Armienti
Figure 16. Back-scattered electron
images of plagioclase phenocrysts.
a) Characteristic zoning of the
plagioclase, consisting of alternating
layers of labradorite (dark grey)
and sieve-textured, bytownite (clear
grey). b) Enlargement showing small-
scale (1-5 µm) oscillatory zoning in
labradoritic layers and sieved textures
with abundant glass inclusions in
bitownitic layers. A dissolution
surface which indents for several tens
of µm a labradoritic layer is evident in
the external part. From Armienti et al.
(2007).
FIG 16
Fig 16
1.20
1.00
0.80
∆t/ τ
0.60
G+ 10^-10
0.40 G+ 10-9
g+ 10^-8
0.20
0.00
1.E-04 1.E-03 1.E-02 1.E-01 1.E+00 1.E+01
maximum dissolution (cm)
Figure 17. Plot of ∆t/τ vs. –G− × ∆t from Equation (23). The diagram provides the maximum size of
crystals that can be completely resorbed for given values of ∆t/τ. τ = 19 years and different values for
plagioclase growth rate. See text for major details. From Armienti et al. (2007).
Crystal Size Distribution Tools 639
noting that crystals, with sizes smaller than the amount of dissolution allowed, will not show
resorption features in so far as they are completely consumed during this interval.
Armienti et al. (2007) report that for G = 10−8 cm/s crystals of 1 cm size would be
completely resorbed. Thus it would be impossible to observe any resorption features in
phenocrysts less than 1 cm in diameter. Or, if 1 cm crystls are not completely resorbed, then
G must be 10−9 cm/s for most ∆t / τ values. To resorb only millimeter-sized crystals the volume
fraction in which the process should take place would be too large. Since only crystals in the
size range of a few hundreds of µm do not show resorption, an order of magnitude of G+ =
10−10 cm/s must be assumed for plagioclase growth rates in the Stromboli feeder conduit. This
implies also that the plagioclase nucleation rates result from a mean behavior of the system
and that the observed intercepts of CSDs are the balance of resorption and nucleation bursts.
Cooling of a magma batch: the case of Mount etna
CSDs evolve in a completely different way if the crystallizing system is not in a stationary
state but evolves due changes of temperature and/or undercooling. Such a system can be
described with the following conditions:
a) Magma is a batch of volume V
b) With no migration of crystals (no fractionation).
c) The only input of crystals to a given class is due to nucleation and growth.
d) Growth rate does not depend upon size L of crystals.
Under these limiting assumptions Equation (18) reduces to:
∂(VN ) ∂(VN )
+G = JVd ( L ) (24)
∂t ∂L
where:
G = growth rate (here not dependent on crystal size L)
J = nucleation rate
d(L) = Dirac function (= 0 for L > 0 and =1 for L = 0)
(thus accounting for crystal input only due to nucleation effects).
By solving the system
d(L) d (VN )
dt = = (25)
G J (t ) V (t ) dL
we find
J (t L ) V (t L )
N (t , L ) = (26)
G (t L ) V (t )
where tL is a time such that in the interval t-tL crystals belonging to the class between L and
L+∆L at time t, grew from 0 to L. V(t) and V(tL) are the volumes of the liquid fractions in which
crystallization took place at t and tL respectively.
The ratio V(tL) / V(t) is a function of the total crystallinity and in many cases can be
neglected. For example, for Etnean lavas the Porphiritic Index is always between 20 and 30%
(Armienti et al. 1994), thus 1.25 < V(tL) / V(t) < 1.43 and in the worst case ln(1.43) = 0.36 that
implies a correction of 0.36 for J / G. Thus the effect of the factor V(tL) / V(t) in most lavas can
be safely neglected and Equation (26) is simplified to
J (t L )
N (t , L ) = (27)
G (t L )
640 Armienti
All the crystals simultaneously undergo the same variations of growth rate and this implies
that crystals born in different times cannot have the same size. This trivial observation has many
important consequences and allows one to use Equation (27) for petrologic interpretations. In
fact, the condition that crystal size depends on time spent to grow and that N(L) reflects the ratio
J / G at the time in which crystals of size L appeared, allow one to conclude that: different trends
of N(L) vs. L reflect changes in the ratio J / G during crystallization and, due to dependence of J
and G on undercooling, a plot of N(L) vs. L corrresponds to a plot of undercooling vs. time.
Dependence of growth and nucleation rates from the undercooling ∆T may be explicitly
defined (Cashman 1990; Toramaru 1991):
−E * −16 πγ Vc2 To2
J ∝ exp exp (28)
R gTm 3k T ∆h2 ( T − T )2
m m o
−E ∆h ( Tm − To )
G ∝ exp 1 − exp (29)
RgTm
k To Tm
where Vc is the volume of one molecule of the crystallizing phase, Rg is the gas constant, γ is
the interfacial tension between crystal and melt, Tm is magma temperature, To is the liquidus
temperature, ∆T = Tm − To, ∆h is the enthalpy of fusion per molecule (= k∆H / Rg, where ∆H is
the molar enthalpy of fusion), k is the Boltzman constant, E* and E are the activation energies
per mole. The shape of the two functions is reported in Figure 18. The two expressions rapidly
rise reaching maxima which do not necessarily occur at the same value of ∆T, then rapidly
drop when the undercooling is very high and the system acquires a large viscosity.
At small undercooling (the stippled area of Fig. 18) Maaløe et al. (1989) suggest, for
growth and nucleation rates respectively, empirical laws to substitute for the above equations
G = A [exp(B∆T − 1)]
and
J = C [exp(D∆T − 1)]
from which
J C [ exp(D∆T ) − 1]
N (L) = = (30)
G A [ exp(B∆T ) − 1]
where A, B, C and D are constants. If D > B, then as undercooling increases, so does N(L).
As undercooling further increases, the two equations are no longer valid and the shapes of J
and G are better described by Eqns. (28) and (29) (Fig. 18). If the maximum of nucleation
occurs at the undercooling ∆TJ and that of growth rate at ∆TG, with ∆TJ > ∆TG, then N(L) will
display a marked increased after ∆TJ. For undercooling greater than ∆TG the J / G ratio also
will decrease, with increasing undercooling. We can interpret the trends of measured CSDs in
Etnean lavas observing that at constant and small undercooling Equation (30) guarantees that
N(L) is constant; thus we expect, in principle, a horizontal trend for CSD in systems at constant
temperatures. In fact
∆T = constant, implies J = constant, G = constant and N(L) = constant
The interpretation of porphyritic structures of volcanic rocks identify a population of
intratelluric crystals reaching larger size in deep seated conditions, then joined by a population
of microphenocrysts and microlites whose occurrence is attributed to the rapid cooling of
lavas as a consequence of rise and eruption. In the hypothesis that no crystal is subtracted
from a magmatic body by crystal migration mechanisms or added by processes different from
Crystal Size Distribution Tools 641
Figure 18. Shapes of the functions representing the dependence of growth (G) and nucleation (J) rates
from undercooling.
FIG 18 The values of J and G suddenly rise reaching maxima which not necessarily occur at
the same value of ∆T, then rapidly drop when the undercooling is very high and the system acquires a large
viscosity. At small undercooling (stippled area), Maaløe et al. (1989) suggest that growth and nucleation
rates follow exponential laws.
nucleation (e.g., fractionation, aggregation, xenolith addition), Equation (25) directly provides
N(L) as a function of the ratio between nucleation and growth rates at the time when crystals of
size L were formed. In this way CSDs can be seen as a continuous record of the cooling history
of the magma.
Figure 19 displays the CSD of plagioclase resulting from the simultaneous treatment
of binary images of feldspar collected on a data set of about 25 thin sections (~150 cm2).
This computation allows one to explore the continuity of populations belonging to the largest
classes and puts in evidence the main features of CSDs of all the minerals of Mt. Etna lavas.
Trend A is referred to a period of deep seated growth at low undercooling. Trend B and C
are related to the ascent of magma at increasing undercooling (the kink between B and C is
probably due to an increasing rate of undercooling). Trend D isperhaps generated after the
emission of lavas.
The effects of different cooling regimes on trend D are more evident in Figure 20,
which reports plagioclase CSDs determined on images collected at a higher magnification
from samples cooled at various rates. Samples were directly taken from the lava flow while
still yellow hot. The rate of cooling cannot be quantified, but increases in cooling rates for
samples 140392 and 170392 can be observed. Sample 140392 is a stubby chunk about 20 cm
in diameter, cooled in air; sample 170392 is 5-cm in size and was drop quenched in water.
Glassy groundmass characterizes the sample quenched in water, while sample 140392 has a
microcrystalline groundmass with scarce interstertial brown glsss. As shown in Figure 20, the
number of quench crystals is heavily influenced by the mode of quenching (rate of cooling) and
by the degree of undercooling reached before the glass transition. In some samples quenched
in water, particularly when CSDs are measured at a high resolution (or the glassy groundmass
allows detection of even the smallest crystals), it is possible to observe the occurrence of a
maximum of CSDs in the size interval D. Pointing out the significance of N(L) as the ratio
J / G, this feature can be interpreted as being due to a condition of undercooling during which
both the nucleation and growth rates decrease after a stage when J increases and G decreases.
642 Armienti
25
Trend D
Plagioclase total population
20 Trend C January - May 1992
Ln (N(L))
Trend B
15
Trend A
10
5
0.00 0.05 0.10 0.15 0.20
D (cm)
Figure 19. Interpretation of porphyritic structures of volcanic rocks.
30
1991 - 92 eruption
quenching effects
on plagioclase
25
Ln (N)
20
15
170392 quenched in water
140392 quenched in air
10
0.000 0.002 0.004 0.006 0.008 0.010 0.012 0.014
D (cm)
Figure 20. Effect of the quenching procedures on microlite crystallization.
Comparison between the total populations of different crystals in Mt. Etna lavas from
the 1991-93 eruption (Fig. 21) reveals that the shapes of the distributions are the same, thus
indicating the same cooling history. CSDs displayed in Figure 21 are related to samples
collected at the vent and 6 km downhill at the lava front. There is a good overlap for regions
A and B and a slight increase for sizes less than 0.003 cm in the sample gathered at the front,
except for olivine.
The largest increment in the number of crystals is observed for Ti-magnetite, whose
crystallization may be influenced by surface oxidation of lava. Since only the quenching
crystals (trend D) show slight changes in their number density, CSDs confirm that the flux
in lava tubes takes place at constant T (there is no change in undercooling). This implies that
Crystal Size Distribution Tools 643
25 20
Sample 230592 Sample 230592
Plagioclase CSD 18
Clinopyroxene CSD
20
Sampled at the vent 16
Sampled at the front 14
Ln(N)
Ln (N)
15
12
10 10
8
5
6
0.00 0.05 0.10 0.15 0.20 0.00 0.10 0.2
D(cm)
D(cm)
25 22
Sample 230592 Sample 230592
20 Magnetite CSD
Olivine CSD
20 18
Ln (N)
Ln (N)
16
15
14
10 12
10
5 8
0.00 0.05 0.10 0.00 0.01 0.02 0.03 0.04
D (cm) D (cm)
Figure 21. Comparison between the total populations of different
crystals in Mt. Etna lavas of the 1991-93 eruption.
major variations of undercooling are met by magma before eruption and are such to pervade
the whole batch of magma contemporaneously. Otherwise, we should observe strong variations
in crystallinity in different portions of the magma batch that achieve different undercooling in
response to the thermal gradient. Pressure, through the control exerted on the water solubility
(Simakin et al. 1999) may provide such a control on undercooling.
estimate of mineral growth rates for Mount etna
An interesting application of Equation (30) is to the estimate of growth rates of minerals
and of the time elapsed since the nucleation of the largest crystals.
At small (<10 °C) and constant undercooling J / G is constant and the largest crystals
belonging to trend A can be used to assess growth times. This implies that the sizes of the
largest crystals may be interpreted to reflect the time they spent in a deeper seated reservoir.
The long-lasting 1991-1993 eruption of Mt. Etna allowed this kind of determination since
it was sampled, in the first year, every day at the eruptive vents and at the lava fronts 6 km
downstream (Armienti et al. 1994). Figure 22 reports a set of measurements made on the largest
plagioclase of each sample in a time span of five months. Two sections of each specimen were
examined, corresponding to about 15 cm2 and measurements of the largest linear dimension
crystals were performed using digitized images. More sections representing several days were
examined to find the largest crystals, since insufficient sections are available for a given day. In
fact for N(L) ≈ e5 cm−4, L ≈ 0.1 cm and ∆L = 0.002 cm, the area to investigate must be greater
than 5 cm2 and becomes of the order of 102 cm2 for a density N(L)≈e5 cm−4. The maximum
crystal sizes increased continuously during the five month interval examined in this work. A
maximum growth rate at depth can thus be assessed from the envelope of the maximum sizes
shown in the figure, under the simplified assumption of a constant growth rate.
644 Armienti
0.6
xenocryst
Maximum axes of elongation of plagioclase
0.5
0.4
0.3
0.2
0.1
y = 0.27564 + 6.0910e-4x R^2 = 0.687
0.0
0 100 200
Days from the beginning of the eruption
Figure 22. Estimate of plagioclase growth rate in intratelluric conditions. Black squares are the largest
crystals detected in the collection of thin sections from the first 6 months of the eruption.
Trends of the envelope give values of 7.1×10−9 cm/s for plagioclase. This estimate of
growth rate of plagioclase at low undercooling is one order of magnitude greater than those
reported by Kirkpatrick (1977b) for much more crystalline basalts, but compares well to those
found by Kirkpatrick (1978) for pure anorthite and those estimated by Maaløe et al. (1989) for
basalts at undercooling of less than one degree. The obtained value for deep plagioclase growth
rate also matches those interpreted by Cashman (1990) for the Mauna Loa 1984 eruption. The
obtained growth rate allows an appraisal of the minimum times of the onset of crystallization
for plagioclase. To derive this datum we have to purge measured diameters for the growth
during the rise, as estimated by the kink of range A. Net deep growth is to be corrected by 0.05
cm. The resulting intercept indicates that plagioclase growth started 470±30% days before the
onset of the eruption.
teMPOraL eVOLutIOn OF CSds: SIZe aS FunCtIOn OF tIMe
The observation that in the case of a crystallizing batch of magma the equation
J(t L )
N (t ,L ) =
G(t L )
guarantees that a plot of N(L) vs. L is equivalent to a plot of the temporal evolution of
undercooling, opens the possibility of extracting temporal information from the dependence
of crystal size from time. In fact, the length attained by crystals in the time interval t-t0 can be
provided by the equation
t
Lt = ∫ G(∆T (t ) dt (31)
to
if we know the functional dependence of G from undercooling (G(∆T)) and of undercooling
from time (∆T(t)).
Crystal Size Distribution Tools 645
The experimental determinations of CSDs provided by zieg and Lofgren (2006) report a
set of data that may illustrate a method of general applicability to obtain this kind of dependence
using CSD data. These authors performed experiments of nucleation and growth of olivine in
a basaltic melt, starting from liquidus temperature and cooling their experimental charges
at a rate of 92 °C/h to induce different final undercooling by simply adopting different time
intervals. Their results are summarized as CSDs in Figure 23 where different symbols refer to
experiments of different duration and different final size of crystals. The computed CSD and
numerical solutions found for time t from Equation (31) are also reported as continuous and
dashed lines respectively. The procedure followed to obtain the explicit dependence of L and
N(L) from time is illustrated below.
The zieg and Lofgren experimental setting seems to imply a linear dependence of ∆T
from time
∆T (t) = α + β t (32)
where α (°C) and β (°C s−1) are constants to be computed. The features of the CSDs may be
constrained with Equation (26)
J (t L ) V (t L )
N (t , L ) =
G (t L ) V (t )
and adopting relations derived from Equations (28) and (29) to describe variations of J and G
with undercooling (Kirkpatrick 1977a):
( (
J = A exp ( − B / Tm ) exp − B′T0 2 Tm ⋅ ∆T 2 ( ))) (33)
(
G = C exp ( − D / Tm ) 1 − exp ( D′∆T ( T0Tm ) ) ) (34)
the unknown constants A, B, B′, C, D, and D′ describe the kinetics of crystallization. We can
Time
64’ 108’ 143’ 241’
14
12 calc.
10
240'
Ln (N(L) ) mm-4
143'
8
64'
6
108'
4
-2
0 0.5 1 1.5
L (mm)
Figure 23. zieg and Lofgren (2006) CSD measurements for different experimental duration. The fitting of
data excludes the largest crystals obtained for experiments of 240 min that could be affected by right hand
truncation. The correspondence between size and time is provided in the upper horizontal axis.
646 Armienti
write a set of boundary conditions derived from the CSDs of zieg and Lofgren (2006), namely
t
∫ Gdt = L
to
t (maxiumum length attained by crystals from time t -to ) (35
5)
∫ Jdt = N
to
tot ( total number of crystals) (36)
t
J 4
∫ G 3 πL dt = volume fraction
3
(Marsh 1998) (37)
to
J0
N ( L0 ) = at t = to (38)
G0
J t max
N ( L max) = (39)
Gt max
Equations (32)-(39) were numerically solved for the four zieg and Lofgren (2006) CSD
data sets reported in Figure 23, providing solutions and fittings of parameters for CSDs
characterization (Table 2 and Figs. 24, 25). The close agreement between the model and the
experimental data sets suggests that it is possible to use the obtained constants to estimate
relevant thermodynamic and kinetic parameters. For the studied case we get Table 3. These
1.2E-04
(a) 1.4E+03
(b)
1.0E-04 1.2E+03
J vs DT
8.0E-05 1.0E+03
G (mm s )
-1
J(s )
8.0E+02
-1
6.0E-05 240'
6.0E+02 143'
4.0E-05 G vs DT
4.0E+02
2.0E-05 240'
2.0E+02
143'
0.0E+00 0.0E+00
0 5 10 15 20 0 5 10 15 20
÷ T (°C) ÷ T (°C)
20 0.7
(c) 0.6 (d)
Underccoling (°C)
15 0.5
OL FRACT
0.4
10
0.3
oss
0.2
5 calc
0.1
0 0
0 50 100 150 200 250 300 0 100 200 300
t (min)
t (min)
Figure 24. Reconstruction of the zieg and Lofgren (2006) experiments: a) and b) report the dependence of growth
and nucleation rates for the experiments at 240 and 143 minutes. As stated in the paper initial undercooling is
in the order of 5° and nuclei practically do not form during the experiments as the early drop of J to 0 clearly
shows. c) reports the dependence of undercooling from time. d) compares the observed and computed evolution
of crystallinity accounting for the initial crystal fraction of 0.13 reported by the authors.
Crystal Size Distribution Tools 647
table 2. Fitting parameters and comparison between observed and computed
experimental data mathematicl procedure of minimization.
Fitting parameters calculated observed (240′)
A= 0.361 ∆T final (°C) 16
B= 1.06×10−4 Gfin (mm/sec) 7.43×10−5
B′ = 0.221 Lmax (mm) 1.30 1.3
C= 1.61×104 ttot (min) 241 240
D= 1.15×104 ln (N(Lmin)) 16.38 13.5
D′ = −1.67 % volume 0.50 0.6
α= 5.35 L(J/G = 6) 0.41 0.48
β= 8.21×10 −4 ln(Jo/Go) 1.72 1.72
300
t vs L from CSD
250 L and t measured
L measured t calculated
t (min)
200
150
100 size - time relation
t
Lt ∫ G ∆T ( t ) dt
50 to
0
0.00 0.50 1.00 1.50
L mm
Figure 25. Relation between L and t. Diamonds are the maximum lengths measured at the end of
experiments. Circles are lengths obtained from CSD measurements. The curve is obtained by the above
discussed mathematical procedure of minimization.
table 3. Computed thermodynamic parameters
Parameter From eqns. (28) and (29) Value units
E* (nucl.) = B Rg = −8.813 × 104 J mol−1
Y ol = B′ 3 k ∆h2/(16 π Vc2) = 4.10 × 10-3 J m−2
E (growth) = B′ × Rg = −9.554 × 104 J mol−1
∆h = k D′ = -2.310 × 10-23 J/ molecule
Adopted constants:
Vc (cell volume for fo94) = 291 × 10−30 m3
k (Boltzman’s constant) = 1.382 × 10−23 J K−1
Rg (gas constant) = 8.310 J K mol−1
648 Armienti
values (Table 3) also closely compare with literature data for olivine in basaltic melts (Dowty
1980; Cooper and Kohlstedt 1982) confirming, independently, the validity of this new model.
COnCLudInG reMarKS
CSD is an important tool in deciphering the crystallization history of magma, but it derives
from heavy data manipulation. Such manipulations require great care when choosing boundary
conditions for the system. Boundary conditions can only be derived from a profound petrologic
comprehension of the system under examination. This means that we have somehow have a
priori knowledge of which kind of CSD we expect and this controls the kind of model to apply.
CSDs are not able to tell us which kind of petrologic processes are under examination, so much
as they offer a key to understanding how known processes develop.
reFerenCeS
Allard B, Sotin C (1988) Determination of mineral phase percentages in granular rocks by image analysis on a
microcomputer. Comput Geosci 14:261-269
Armienti A, Francalanci L, Landi P (2007) Textural effects of steady state behaviour of the Stromboli feeding
system. J Volcanol Geotherm Res 160:86-98
Armienti P, Innocenti F, Pareschi MT, Pompilio M (1994) Effects of magma storage and ascent on the kinetics
of crystal growth. The case of the 1991-92 Mt. Etna eruption. Contrib Mineral Petrol 115:402-414
Armienti P, Innocenti F, Pareschi MT, Pompilio M, Rocchi S (1991) Crystal population density in not stationary
volcanic systems: estimate of olivine growth rate in basalts of Lanzarote (Canary Islands). Mineral Petrol
44:181-196
Armienti P, Pareschi MT (1992) Cooling hystory of basalts and its effects on the kinetics of olivine
crystallization. Acta Volcanol 2:7-15
Armienti P, Tarquini S (2002) Power-law olivine crystal size distributions in lithospheric mantle xenoliths.
Lithos 63(3-4):273-285
Brandeis G, Jaupart C (1987) The kinetics of nucleation and crystal growth and scaling laws for magmatic
crystallization. Contrib Mineral Petrol 96:24-34
Cashman KV (1988) Crystallization of Mount St. Helens 1980-1986 dacite: A quantitative textural approach.
Bull Volcanol 50:194-209
Cashman KV (1990) Textural constraints on the kinetics of crystallization of igneous rocks. Rev Mineral
24:259-314
Cashman KV (1992) Groundmass crystallization of Mount St. Helens dacite, 1980-1986: a tool for interpreting
shallow magmatic processes. Contrib Mineral Petrol 109:431-449
Cashman KV, Marsh BM (1988) Crystal size distribution (CSD) in rocks and the kinetics and dynamics of
crystallization. II: Makaopuhi lava lake. Contib Mineral Petrol 99:292-305
Cashman KV, Marsh BM (1988) Crystal size distribution (CSD) in rocks and the kinetics and dynamics of
crystallization. II: Makaopuhi lava lake. Contrib Mineral Petrol 99:292-305
Chayes F (1950) On the bias of grain-size measurements made in thin section. J Geol 58:156-160
Chayes F (1956) Petrographic Modal Analysis - An Elementary Statistical Appraisal. John Wiley & Sons, Inc.,
New York
Cooper RF, Kohlstedt DL (1982) Interfacial energies in the olivine basalt system. In: High-pressure Researches
in Geophysics. Akimoto S, Manghnami MH (eds) Center Academic Publication. Tokyo. Adv Earth
Planetary Sci 12:217-228
DeHoff RT, Rhines FN (1972) quantitative Microscopy. Masson et Cie, Paris
Delesse MA (1847) Procédé mécanique pour déterminer la composition des roches. Comptes Rendus
Hebdomadaires des Sciences de l’Academie de Sciences 25:544-545
Donaldson CH (1976) An experimental investigation of olivine morphology. Contrib Mineral Petrol 57:187-
213
Dowty E (1980) Crystal growth and nucleation theory and numerical simulation of igneous crystallization. In:
Physics of Magmatic Processes. Hargraves RB (ed) Princeton University Press, p 385-417
Francalanci L, Tommasini S, Conticelli S, Davies GR (1999) Sr isotope evidence for short magma residence
time for the 20th century activity at Stromboli volcano, Italy. Earth Planet Sci Lett 167:61-69
Gray N (1970) Crystal growth and nucleation in two large diabase dikes. Can Jour Earth Sci 7:366-375
Gray N (1978) Nucleation and growth of plagioclase, Makaopui and Alae lava lakes, Kilauea Volcano, Hawaii:
Discussion. Geol Soc Am Bull 89:797-798
Crystal Size Distribution Tools 649
Higgins MD (2000) Measurement of crystal size distributions. Am Mineral 85:1-12
Higgins MD (2002) Closure in crystal size distribution (CSD), verification of CSD calculations and the
significance of CSD fans. Am Mineral 87:160-164
Jambon A, Lussiez P, Clocchiatti R, Weisz J, Hernandez J (1992) Olivine growth rate in a tholeiitic basalt: an
experimental study of melt inclusions in plgioclase. Chem Geol 96:277-287
Kilburn C (1989) Surface of Aa flow-fields on Mount Etna, Sicily: morphology, rheology, crystallization and
scaling phenomena. In: Lava Flows and Domes. Fink J (ed) Springer-Verlag, Berlin, p 129-156
Kirkpatrick RJ (1977a) Crystal growth from melt: a review. Am Mineral 60:798-814
Kirkpatrick RJ (1977b) Nucleation and growth of plagioclase, Makaopuhi and Alae lava lakes, Kilauea
Volcano, Hawaii. Geol Soc Am Bull 88:78-84
Kirkpatrick RJ (1978) Nucleation and growth of plagioclase, Makaopui and Alae lava lakes, Kilauea Volcano,
Hawaii. Reply. Geol Soc Am Bull 89:799-800
Kirkpatrick RJ, Gilpin RR, Hays JF (1976) Kinetics of crystal growth from silicate melts: anorthite and
diopside. J Geophys Res 81:5715-5720
Landi P, Métrich N, Bertagnini A, Rosi M (2004) Dynamics of magma mixing and degassing recorded in
plagioclase at Stromboli (Aeolian Archipelago, Italy). Contrib Mineral Petrol 147:213-227
Lippman PW, Banks NG, Rhodes JM (1985) Degassing induced effects on lava rheology. Nature 317:604-
607
Lofgren GE (1974) An experimental study of plagioclase crystal morphology: isothermal crystallization. Am
J Sci 274:243-273
Lofgren GE (1980) Experimental studies on the dynamic crystallization of silicate melts. In: The Physics of
Magmatic Processes. Hargraves RB (ed) Princeton Univ Press, Princeton, p 487-551
Maaløe S, Tumyr O, James D (1989) Population density and zoning of olivine phenocrysts in tholeiites from
Kauai, Hawaii. Contrib Mineral Petrol 101:176-186
Mangan MT (1990) Crystal size distribution systematics and the determination of magma storage times: the
1959 eruption of Kilauea volcano, Hawaii. J Volcanol Geotherm Res 44:295-302
Marsh DB (1988) Crystal size distribution (CSD) in rocks and the kinetics and dynamics of crystallisation. I.
Theory. Contrib Mineral Petrol 99:277-291
Morgan DJ, Jerram DA (2006) On estimating crystal shape for crystal size distribution analysis. J Volcanol
Geotherm Res.154:1-7
Muncill GE, Lasaga AC (1987) Crystal-growth kinetics of plagioclase in igneous systems: one atmosphere
experiments and approximation of a simplified growth model. Am Mineral 72:299-311
Pareschi MT, Pompilio M, Innocenti F (1990) Automated evaluation of volumetric grain-size distribution
density from thin-section images. Comput Geosci 16:1067-1084
Pickering G, Bull JM, Sanderson DJ (1995) Sampling power law distribution. Tectonophys 248:1-20
Russ JC (1986) Practical Stereology. Plenum Press, New York
Saltykov SA (1949) Calculation of the distribution curves for the size of dispersed grains. Plant Laboratory
15:1317-1319
Schwartz HA (1939) Metallographic determination of the size distribution of tempered carbon nodules. Metals
Alloys 5(6):139-140
Simakin AG, Armienti P, Epel’baum MB (1999) Coupled degassing and crystallization: experimental study in
continuos pressure drop, with application to volcanic bombs. Bull Volcanol 61:275-287
Toramaru A (1991) Model of nucleation and growth of crystals in cooling magmas. Contrib Mineral Petrol
108:106-117
zieg M, Lofgren G (2006) An experimental investigation of texture evolution during continuous cooling. J
Volcanol Geotherm Res 154:74-88
Reviews in Mineralogy & Geochemistry
Vol. 69 pp. 651-674, 2008 17
Copyright © Mineralogical Society of America
deciphering Magma Chamber dynamics
from Styles of Compositional Zoning
in Large Silicic ash Flow Sheets
Olivier Bachmann and George W. Bergantz
Department of Earth and Space Sciences
University of Washington
Mailstop 351310
Seattle, Washington, 98195-1310, U.S.A.
[email protected], [email protected]
IntrOduCtIOn
The understanding of the dynamic processes in magmatic systems has grown and changed
markedly in the last decade. Old models for magmatic systems as vats of near-liquidus material
have been revised by observations from seismology (Sinton and Detrick 1992), crystal chemistry
and zoning (Davidson et al. 2007) and the geochronology and geochemistry of both plutonic and
volcanic systems (Hildreth 2004; Charlier et al. 2007; Miller et al. 2007; Peressini et al. 2007;
Walker et al. 2007). New views emphasise magmatic systems as temporally dominated by crystal
mushes (magma bodies with a high fraction of solid particles; see definition in Miller and Wark
2008), that wax and wane in temperature and crystallinity, and are subject to significant open
system processes (Charlier et al. 2007; Hildreth and Wilson 2007; Walker et al. 2007; Bachmann
and Bergantz 2008). These processes, such as magma reintrusion, mixing, gas sparging,
and subsequent thermal rejuvenation, may be significantly more important in producing the
characteristics of a magmatic system than the previous closed-system, near-liquidus behaviour
would predict. There are a number of recent reviews that summarize these observations (for
example see Eichelberger et al. 2006; Bachmann et al. 2007b; Lipman 2007). Our aim here
is to illustrate how this new perspective to magma dynamics is motivated by observations of
heterogeneities (or lack thereof) in erupted rocks (and to a lesser amount in plutons).
Owing to rapid withdrawal and quenching of magma during explosive volcanic eruptions
(hours to a few days), large-volume (>1 km3) pyroclastic deposits (also referred to as ignimbrites
or ash-flow tuffs) provide an instant image of the state of the magma chamber evacuated during
eruption. A first-order observation that characterizes these pyroclastic deposits of intermediate
to silicic composition is that many do not tap into chemically homogeneous reservoirs but show
compositional and thermal zoning, with early-erupted material differing significantly from late-
erupted material (e.g., Lipman et al. 1966; Lipman 1967; Smith 1979; Hildreth 1981). In detail,
nearly every eruptive product is different (see Hildreth 1981 for a comprehensive classification),
but three general groups can be defined as follows (Table 1 and Fig. 1): 1) deposits showing
quasi monotonic (or linear) gradients in composition and/or temperature, 2) deposits showing
abrupt gradients in composition, with gaps that can reach several wt% in major elements over
a very narrow stratigraphic interval, and 3) deposits showing no significant compositional
or thermal gradients. Our objective is to provide a synopsis of zoning patterns preserved in
ignimbrites as a means for understanding magma dynamics in silicic systems. We will argue
that retaining and/or producing homogeneity by convective stirring is actually more challenging
than preserving gradients, particularly in viscous, crystal-rich, silicic magma chambers, and
1529-6466/08/0069-0017$05.00 DOI: 10.2138/rmg.2008.69.17
652 Bachmann & Bergantz
table 1. Characteristics of some well-studied ignimbrites.
abruptly zoned Linearly zoned homogeneous units
units units
Examples Normally zoned: Purico Bishop Tuff, Bandelier Fish Canyon Tuff,
ignimbrite, Crater Lake Tuff, Huckleberry Lund Tuff, Cerro Galan
ignimbrite, Katmai Ridge, Tuff La ignimbrite, Atana
1912, Los Humeros, Primavera Tuff, Taylor ignimbrite, Oranui
Timber Mountain/Oasis Creek Rhyolite, Loma ignimbrite
Valley caldera complex Seca Tuff, Laacher See
Reversely zoned: ignimbrite, Campanian
Whakamaru ignimbrite, Ignimbrite, Aso
Bonanza Tuff Ignimbrite
Gradients in:
Major element Abrupt gaps (up to 20 Generally limited (few Not measurable
wt% SiO2) wt% at most)
Trace element variations Significant Significant Not measurable
Isotopic ratios Limited or absent Significant, and Some heterogeneities at
in some cases, and often correlated with the whole-rock scale
important in others Differentiation Index
Crystallinity Abrupt change (few Significant increase Not measurable
percent to >30 %). from early-erupted to
More mafic endmember late-erupted (~1 to >20
generally more crystal- vol%)
rich
Temperature Up to 100 °C in cases Reported (up to 100 °C) Not measurable
but absent in others
Gas content Top richer in volatiles Top richer in volatiles Not measurable
Linearly zoned units: Hildreth 1981; Christiansen 1984; Grunder and Mahood 1988; Mahood and Halliday 1988; Johnson
1989; Duffield et al. 1995; Christiensen and Halliday 1996; Davies and Halliday 1998; Knesel et al. 1999, Lipman 1967;
Wolff and Storey 1984; Worner and Schmincke 1984b; Worner and Schmincke 1984a; Wolff 1985; Wolff et al. 1990;
Civetta et al. 1997
Abruptly zoned units: Whitney et al. 1988; Chesner 1998; Lindsay et al. 2001; Schmitt et al. 2001 Bacon 1983; Bacon
and Druitt 1988; Hildreth and Fierstein 2000; Lipman et al. 1966; Varga and Smith 1984; Brown et al. 1998; Bindeman
and Valley 2003
Homogeneous units: Francis et al. 1989; Lindsay et al. 2001; Bachmann et al. 2002; Maughan et al. 2002; Dunbar et al.
1989; Wilson et al. 2006.
require two conditions: (1) density filtering of material allowed to enter the growing silicic
reservoir, and (2) thermal buffering close to the haplogranite eutectic.
types of gradients in ignimbrites
Precisely reconstructing the zonation in a magma chamber from erupted deposits is
difficult. Low sampling density, complexities in the depositional patterns induced by shifting
eruptive vent sites and fluctuations in eruption rates (e.g., Wilson and Hildreth 1997), partial
mixing during magma ascent in conduits (Blake and Ivey 1986a,b; Trial and Spera 1992)
and the time-dependence of the dispersal of fragmented materials (e.g., Neri et al. 2003) will
scramble many of the finer details of any spatial patterns of composition within the chamber
prior to eruption. Hence only first-order observations, such as how early-erupted material
differs from late-erupted material, can be considered as robust expressions of pre-eruptive
Deciphering Magma Chamber Dynamics 653
1. Linear gradient
2. Abrupt gradient
Ignimbrite sheet
3. Lack of gradient
Figure 1 1. Schematic illustration of the three types of zoning patterns that commonly occur
Figure
in ignimbrite (or ash-flow sheets) (modified from Bachmann and Bergantz 2008).
conditions for inter-deposit comparison. However, by sampling unfragmented material
(pumice or lava blocks) from multiple localities in a given ignimbrite, an approximation of the
pattern of gradients that existed in the magma chamber (whether abrupt, nearly continuous or
absent, Fig. 1) can be obtained.
It is important to appreciate that by using the term “gradient,” we are referring to a horizon-
tally-averaged quantity. In detail the variations in composition, temperature, and crystallinity
associated with convective stirring can locally vary significantly over short distances (envisage
mixing two different flavors of honey with different colors), but these intensive variables will
on-average produce a vertical gradient in a gravity field if the different phases have variable
densities. Below we provide a short synopsis of the different types of gradients, then assess the
role of convection in magma chambers and combine the two to explain how distinct types of
zoning, or the lack thereof, emerge.
Abrupt gradients
Step-like changes in the whole-rock chemical composition of pumices up to several wt% for
certain major elements has been documented in many deposits of explosive volcanic eruptions
(e.g., Brophy 1991, Fig. 2). Striking color changes make these abrupt variations visibly manifest
(see Bacon and Druitt 1988; Druitt and Bacon 1989; Eichelberger et al. 2000 and Fig. 2 for
pictures), and they are commonly associated with large changes in crystallinity (the more mafic
deposits being more crystal-rich, Bacon and Druitt 1988; Hildreth and Fierstein 2000).
Linear gradients
When plotted on geochemical variation diagrams, many silicic ignimbrites present a nearly
continuous array in pumice composition (e.g., Smith and Bailey 1966; Hildreth 1979; Halliday
et al. 1984; Grunder and Mahood 1988; Streck and Grunder 1997; Wolff et al. 1999; Milner
et al. 2003, Fig. 3) from most evolved, early-erupted to least-evolved late-erupted deposits,
although, in some volcanic units, small compositional gaps have been reported (Fridrich and
Mahood 1987; Streck and Grunder 1997). The observed chemical variations are particularly
striking for trace elements, but also involve major elements, isotopic rations, crystallinity,
654 Bachmann & Bergantz
Crater Lake climactic eruption (7700 BP)
3.5
e
olit
3 Rhy
ite
Dac
2.5 te
desi
An
K2O (wt%)
2 c
alti
Bas site
hK e
1.5
H ig and
lt
1 Basa
0.5
K
Low
0
45 50 55 60 65 70 75 80
SiO2 (wt%)
Figure 2
Figure 2. Compositional gap in the Crater Lake ignimbrite, obvious both on
chemical plots and in the field. (modified from Bacon and Lanphere 2006).
Pyroclastic flow Figure 3. Compositional spread
in pumices of the Bishop Tuff,
showing progressive depletion in
Ba and FeO* as a function of SiO2
Air fall (modified from Hildreth and Wil-
son 2007).
Main Bishop Tuff pumice type (xr, xm, xp) in Bishop Tuff
700 2
FeO* 1.8
600
Ba 1.6
500 1.4
FeO* (wt%)
Ba (ppm)
1.2
400
1
300
0.8
200 0.6
0.4
100
0.2
0 0
70 72 74 76 78 80
SiO2 (wt%)
Figure 3
Deciphering Magma Chamber Dynamics 655
temperature, and gas content (Table 1). These variations in most magmatic tracers imply that
fairly continuous compositional and thermal gradients are present in the magma reservoir
tapped by the eruption (no information on their spatial distribution though).
As an example of a typical linearly-zoned system, the Bishop Tuff shows: (1) an increase
in crystal content and decrease in volatile content from early- to late-erupted material (Hildreth
1979; Wallace et al. 1999), (2) lack of significant major element zoning, but more than twofold
variation in incompatible trace elements and extreme depletion in compatible trace element
(e.g., Sr, Ba) in early-erupted material (Hildreth 1979; Michael 1983; Hervig and Dunbar
1992; Knesel and Davidson 1997; Anderson et al. 2000), (3) variations in εNd, 206Pb/204Pbi,
and 87Sr/86Sri, but homogeneity in oxygen isotopes (Bindeman and Valley 2002), and (4) a ∆T
of ~100 °C from early to late-erupted material, documented with Fe-Ti oxides (Hildreth and
Wilson 2007), oxygen isotope thermometry (Bindeman and Valley 2002), and Ti-in-quartz
thermometry (Wark et al. 2007).
Lack of gradients
Among a number of well-described zoned ignimbrites, several examples stand out as
having very homogeneous whole-rock characteristics (including major and traces elements,
isotopic ratios, crystallinity, and temperature; Fig. 4). Most of these homogeneous units belong
to the Monotonous Intermediates, a group of large (>1,000 km3), crystal-rich (up to 45 vol%
crystals), dacitic units erupted in mature continental arcs (Hildreth 1981, Francis et al. 1989; de
Silva 1991; Chesner 1998; Lindsay et al. 2001; Bachmann et al. 2002; Maughan et al. 2002).
However, some homogeneous examples also occur in rhyolitic, crystal-poor units (Dunbar et
al. 1989; Wilson et al. 2006).
Figure 4. A section of intracaldera Fish Canyon Tuff (in Canyon Diablo, CO), showing a lack of any kind of
Figure 4 top to bottom. Up to 1 km thick section is exposed on the northern flank of the resurgent dome
gradient from
(La Garita mountains), with no base and top exposed, implying a thickness significantly greater than 1 km
(estimated at an average of 2 km for the entire 80×30 km La Garita caldera collapse area; Lipman 2000).
656 Bachmann & Bergantz
MeChanISMS FOr GeneratInG GradIentS
In SILICIC MaGMa ChaMBerS
Following the recognition that many large-volume ignimbrites originate in compositionally
complex magma reservoirs (Lipman et al. 1966; Lipman 1967), numerous models have been
proposed to explain the origin of these gradients. Below we summarize some key elements
of these models, and discuss them in light of recent advances in our understanding of the
dynamics of viscous, multi-phase mixtures (Bergantz and Ni 1999; Jellinek and Kerr 1999;
Jellinek et al. 1999; Bergantz 2000; Bergantz and Breidenthal 2001; Phillips and Woods 2002;
Burgisser et al. 2005; Dufek and Bergantz 2005). We will first summarize some aspects of the
kinematics of convection in magmatic systems.
the role of convection
As noted by Grout almost a century ago (Grout 1918) and many workers since (Morse
1986; Marsh 1988; Rudman 1992; Simakin et al. 1997; Simakin and Botcharnikov 2001),
convection in magma reservoirs is unavoidable as long as a dominantly-liquid (but not
necessarily crystal-poor) batch of magma is present. The most significant source of density
changes that provides the potential energy for convection are associated with phase change;
examples include the formation of crystals and/or bubbles. Magmatic systems are also open to
new injections (with both similar and different chemical compositions) and can assimilate their
wall-rocks (DePaolo 1981; Davidson and Tepley 1997; Bohrson and Spera 2001; Thompson
et al. 2002; Beard et al. 2005), leading to additional temperature and crystallinity (and hence
buoyancy) fluctuations within the reservoir (Keller 1969; Bergantz 2000; Murphy et al. 2000;
Bachmann et al. 2002). Thus, convective dynamics within magma reservoir are the result of
both (1) in situ density changes arising from cooling and/or volatile exsolution and (2) magma
reintrusion and/or volatile flux.
Convective stirring in magmatic systems is likely to be highly intermittent and arise from
many or all of the above mentioned buoyancy sources during the lifetime of a reservoir (which
can last several hundreds of thousands of years; e.g., Reid et al. 1997; Brown and Fletcher
1999; Vazquez and Reid 2004; Bacon and Lowenstern 2005; Simon and Reid 2005; Bachmann
et al. 2007a). Geological data require that any particular model must accommodate the
following observations: a) magma bodies cool on conductive timescales and rarely assimilate
large amounts of their margins by rapid heat transfer (e.g., Carrigan 1988; Barboza and
Bergantz 2000; Dufek and Bergantz 2005; Petcovic and Dufek 2005), b) heterogeneous crystal
populations (particularly in age population of zircons) at the margins and core of plutonic
suites generally preclude simple, monotonic sidewall and roof crystallization (Eichelberger et
al. 2006a; Miller et al. 2007; Walker et al. 2007), c) open system processes and secular cooling
produces intermittent convection with Rayleigh numbers high enough (>105) where conditions
for chaotic convection are (at least temporarily) obtained.
In these complex, open-system magma reservoirs, convection has a dual nature: it reduces
the scales and intensity of heterogeneities (e.g., Oldenburg et al. 1989; Coltice and Schmalzl
2006), but can also produce gradients in intensive variables and composition if the source of
buoyancy in itself generates heterogeneities (Bergantz and Ni 1999; Jellinek and Kerr 1999;
Jellinek et al. 1999, Figs. 5 to 8). The following paragraphs attempt to clarify this dual nature.
It is widely understood that mixing requires advection of material, which acts as a source
of localized high strain, reducing length scales of heterogeneities and allowing a significant
volume of material to be serviced by the mixing process. Subsequently, diffusion can operate
effectively once the average distance between heterogeneities have been reduced below the
diffusive length scale (e.g., Coltice and Schmalzl 2006). The most efficient mixing systems will
be those that can produce the greatest amount of strain at the widest range of scales, typically by
the rotational motion of eddy-like structures that both entrain fluid and wind them.
Deciphering Magma Chamber Dynamics 657
Jellinek et al. 1999 experiments Final set-up
Stable
No mixing
Initial set-up Convective
E=0
Unstable Overturn
Fluid 2: Low density
Fluid 1: High density
Partial mixing
Fluid 1: High density 0<E<1
Permeable
plate
Fluid 2: Low density
Perfect mixing
E =1
Figure Figure 5
5. Schematic diagram of mixing experiments described by Jellinek et al. (1999), starting from an
initial, gravitationally unstable set-up, inducing an overturn, and leading to a final mixture, whose mixing
efficiency will depend on the density and viscosity ratios of the two fluids.
The Reynolds and Rayleigh numbers are the most important diagnostics of the dynamic
and kinematic conditions within in a magma body. A very general form of the Rayleigh number
is (Jellinek et al. 1999):
BH 4
Ra = (1)
νκ2
where B is the buoyancy flux, H the height of the system, ν the kinematic viscosity of the
ambient magma and κ the diffusivity. The Reynolds number is traditionally defined as:
uL
Re = (2)
ν
where u is a scaled velocity, L the matching length scale and ν the kinematic viscosity. It can
also be written as (Jellinek et al. 1999):
B1/ 3 H 4 / 3
Re = (3)
ν
(assuming that the velocity is scaled on a balance of inertia and buoyancy, a problematic
assumption for small Reynolds numbers but that does not impact our overall conclusions).
The buoyancy flux appearing in both the Rayleigh and Reynolds numbers will be a complex
function of the conduction-controlled rate of cooling, the geometry (as it impacts cooling rate),
the resulting density changes associated with phase change, and thermal buffering with latent
heat (see, for example, Jellinek and Kerr 2001).
658 Bachmann & Bergantz
Schematic tank diagram
Convective
overturn
Ambient Fluid
(Fluid 1)
Permeable plate
Introduction of low density
Fluid 2 from below
Picture of tank mixing experiment Picture of tank mixing experiment
µambient/µinput = 0.017 and Re = 1290 µambient/µinput = 724 and Re = 0.33
permeable plate permeable plate
Introduction of low density Introduction of low density
Fluid 2 from below Fluid 2 from below
Figure 6. Illustration of the overturn process in the mixing experiments of Jellinek et al. (1999).
Figure 6
There are generally three regimes where the Reynolds number produces a distinct flow
template: (1) laminar, (2) transitional and (3) fully turbulent. In the laminar regime (Re <1),
buoyancy forces are balanced by viscous forces (viscosity controls the dynamics of the flow),
while in the turbulent regime (Re ~>1000), the buoyancy force is balanced by inertial forces
(onset of true turbulence). Intermediate values between these two (1 < Re < 1000) indicate that
both viscosity and inertia play a role, and this is the most difficult range to generalize because
the flows are not self-similar. This means that the structure of the flow is not ordered at the
largest scales: a small change in Reynolds number can produce an intermittent flow regime
that is not similar to the one that preceded it.
At high values of the Reynolds number (~1000), the mixing transition is reached, and
there is a full spectrum of eddy sizes from largest scales set in part by the geometry of the
system (initial and boundary conditions), to the smallest scales of eddies set by the viscosity.
For systems with Reynolds numbers at or above the mixing transition, mixing is fast and
Deciphering Magma Chamber Dynamics 659
1.0
0.9
0.8
0.7
Normalized height
0.6
No mixing
0.5 Complete mixing
0.4
Mixing efficiency
0.3
0 0.74
0.2 0.1 0.91
0.3 1
0.1 0.5
0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
(ρfinal-ρinput)/(ρambient-ρinput)
FigureFigure
7. Mixing
7 efficiency for fluids at different densities and viscosities as defined by
experiments of Jellinek et al. (1999), illustrated in a 2D cross section of a tank.
0.4
Figure 8. Horizontally-averaged vertical
gradient in crystallinity within a reservoir
0.3 Convecting
Normalized height
(modified from Bergantz and Ni 1999).
magma These results were obtained by numerically
simulating cooling of an originally solid-
0.2
free liquid, triggering precipitation of dense,
Horizontally-averaged solid phases at the roof. Once the dense roof
vertical solid volume fraction layer reaches a critical crystallinity, solid-rich
0.1
plumes collapse, leading to the growth of a
stable gradient in solid fraction within the
convecting, liquid body. See Bergantz and Ni
0.0
0 1 2 3 (1999) for details.
Solid volume fraction (%)
Figure 8
thorough, as mixed material is quickly passed among eddies of many sizes (Broadwell and
Mungal 1991; Dimotakis 2005), thus efficiently breaking-down existing gradients. Jets and
plumes produced in explosive eruptions reach this Reynolds number.
For Re <1000 (low to transitional Reynolds numbers), there is not a statistically complete
set of eddy sizes in the flow. Only a few eddy sizes dominate and the high-strain stirring is
of limited efficiency. If the Rayleigh number is high enough, such situations are prone to
“chaotic” convection (e.g., Ottino 1989; Ferrachat and Ricard 1998; Petrelli et al. 2006). The
onset of chaotic conditions can be determined by the appropriate Rayleigh numbers, although
the values will depend on the specific system under consideration. Generally, however, as long
as the Rayleigh number is above about 105, chaotic conditions can be obtained, even with a
low Reynolds number (Coltice 2005).
Chaotic mixing leads, on the short term, to islands of well-stirred material in a domain
of otherwise poorly stirred material (see overturn illustration in Fig. 5). If kept active long
enough, this kind of flow can lead to efficient mixing (unless there are strong contrasts in
660 Bachmann & Bergantz
viscosity and density; Ottino 1989). However, if the source of buoyancy and convective stirring
itself induces heterogeneities, then complete homogeneity cannot be reached.
Consider discrete, mechanically passive, chemical heterogeneities such as a local patch
within the magma system that has a distinct character, such as a small difference in crystallinity,
temperature or composition (or all three). If there is chamber-wide overturn driven by cooling,
heating or reintrusion from the boundaries, this patch will be homogenized with the rest of
the magma reservoir in a few overturns in a chaotically convecting system (e.g., Coltice 2005)
even at low Reynolds number; it does not require sustained vigorous fluid circulation. If the
viscosity contrast between the patch and the magma reservoir is large, the homogenization
may take longer but even crystal multiphase aggregates can be eroded and dissipated by shear
(Powell and Mason 1982; Petrelli et al. 2006).
On the other hand, if new heterogeneities are the sources of the convective process itself
(such as dense crystal plumes originating from a cold ceiling, or reintrusion of low density
magma from below), the system cannot reach perfect homogeneity (except at extremely high
mixing efficiencies in high Re number situations). As heat exchange with the surroundings
never completely stops, new heterogeneities in temperature, inducing variations in crystal
content are constantly created, impeding complete homogeneity in crystallinity (and therefore
composition); new gradients are re-established as old ones are consumed. More generally, any
process taking place at the chamber boundaries that leads to variations in trace elements, crystal
content, or temperature can export that variation into the magma body by convection (e.g.,
Marsh 1989) and produce a gradient in these quantities, while also homogenizing pre-existing
gradients. Therefore, even in a chemically closed (but cooling) system, perfect homogeneity
cannot be achieved for waning systems.
As illustrated in numerical and laboratory experiments (Figs. 5 to 8), convective processes
can readily generate gradients as they transport material and heat across a magma system. Any
density instability that originates at a boundary (e.g., crystal-rich plume falling from the top, a
gas-rich region that rises from the bottom, or a reintrusion of less dense material from below)
will all produce a gradient as it passes through the system. As material at the boundary rises
or falls, this process will entrain surrounding material (see Bergantz and Breidenthal 2001),
become diluted, and deliver this now mixed material to the other boundary; it also leaves a
“wake” of material that is a mix of the initial falling or rising plume and the surrounding
magma. If the process is repeated, it will produce and sustain a (horizontally averaged)
gradient, as it has been exemplified for a number of applications to magmas (Bergantz and Ni
1999; Jellinek and Kerr 1999; Jellinek et al. 1999; Ruprecht et al. 2008).
The ability of rising or falling plumes to create gradients in the surrounding magma has
been called the “mixing efficiency” (Linden and Redondo 1991; Jellinek and Kerr 1999;
Jellinek et al. 1999). It is defined on the basis of an initially unstable stratification, such as a
less dense layer entering the bottom, allowed to mix with the over-lying fluid as it penetrates
upward. One endmember is that the plumes rise or fall through the ambient magma with no
mixing, simply changing position, and the least amount of initial gravitational potential energy
is retained in the final configuration. The second endmember is perfect mixing, leaving no
gradient but a uniform mixture over the entire vertical span of the system. Between these two
extremes, the overturn process produces a stably stratified system and the vertical variation in
the density, and hence concentration, from the mixing of the penetrating and ambient fluids
produces a gradient (Fig. 7). To quantify this, the mixing efficiency, E, is defined as:
Pf − Pmin
E= ( 4)
Pmax − Pmin
where Pmax is the final, maximum value of the potential energy associated with perfect mixing
(which yields no gradient), Pmin is the final value associated with no mixing as the two layers
Deciphering Magma Chamber Dynamics 661
simply exchange places and Pf is the actual final potential energy of the system following
overturn:
Pf = ∫ g ( ρa − ρ( z ) ) zdz
H
(5)
0
g the scalar acceleration of gravity, ρa the density of the ambient, or resident fluid and ρ(z)
the horizontally averaged vertical density profile after overturn. It is important to note that
this definition of the mixing efficiency is a measure of the (horizontally averaged) vertical
gradient. The mixing efficiency has a value for zero if the two fluids slide past each other with
no mixing, and unity if the two fluids mix perfectly creating a system with no horizontally-
averaged vertical gradient. So a value of the mixing efficiency of say 0.5, does not imply that
50% of fluid A and 50% of fluid B are everywhere well mixed. It means that there is a gradient
in density (and hence concentration) from top to bottom with a slope of 0.5 and with equal
amounts of the resident and intruding material at only one vertical position.
For silicic magmas, there will be a wide range of buoyancy flux B depending on the source
of buoyancy (thermal, compositional, phase change), but reasonable values will vary between
10−7 and 10−11. Thus, for a given height, kinematic viscosity and diffusivity, one can estimate
the approximate Rayleigh and Reynolds numbers across a range of compositions (using Eqns.
1 and 2). As shown in Figure 9, the Reynolds number generally remains low (<10) even for
the largest possible examples (up to 5 km thick) and most dynamic scenarios. Therefore, the
mixing efficiency will always be low in magmatic systems, and for heterogeneities related to
magma cooling, perfect homogeneity cannot be reached.
Re > 103
104 Above mixing transition
Never reached in
onset of true turbulence
Rapid mixing
magmatic systems
103
102 Height of magma reservoir
= 5 km Re: 10-1-102
10 Occasional, short-lived
high-energy excursions,
Re 1 Realm of chaotic conditions
10-1 Height of magma reservoir
= 1 km Re < 10-1
Typical kinematic
10-2 conditions
recorded in plutons
10-3
50 60 70
Wt. % SiO2 (proxy for viscosity)
Figure 9
Figure 9. Reynolds number as a function of composition for different magmas reservoirs.
Origin of heterogeneity in silicic magmas: convective stirring vs. phase separation
On the basis of the arguments expressed above, the presence of zoning or heterogeneities
in a volcanic unit can represent two endmembers: (1) incomplete, or partial convective mixing
of two magma batches or (2) chemical differentiation or distillation by crystal-liquid separation
of an initially more homogeneous system. These two scenarios are similar to the concepts of
662 Bachmann & Bergantz
“arrested homogenization” and “progressive unmixing” of Eichelberger et al. (2000), although
we stress that convective stirring of different magma batches does not necessarily lead to
homogenization (and therefore does not need to be “arrested”).
Convective interactions of compositionally variable magma batches present in the same
reservoir are clearly common in the geological record. Magmatic plumbing systems favour
it, and a number of ignimbrites record the mechanical encounters between magmas of from
different sources (Smith 1979; Hervig and Dunbar 1992; Mills et al. 1997; Bindeman and Valley
2003; Eichelberger et al. 2006b; Davidson et al. 2007; Knesel and Duffield 2007). However,
many volcanic deposits display chemical heterogeneities that have been interpreted as being
dominantly induced by crystal fractionation within a given reservoir (in situ differentiation,
e.g., Thompson 1972; Hildreth 1981; Michael 1983; Brophy 1991; Francalanci et al. 1995;
Thompson et al. 2001; Hildreth 2004). A magma residing in a cooler reservoir will undergo
crystallization and volatile exsolution. As the new phases produced have different densities, the
system will have the tendency to differentiate, or unmix, into two or more batches with distinct
characteristics. Therefore, zoning patterns present in deposits with an in situ differentiation
signature represent progressive unmixing of an initially more homogeneous magma body (e.g.,
McBirney 1980; Eichelberger et al. 2000).
Both scenarios of heterogeneization can occur at different time scales depending on
the physico-chemical conditions in the magma reservoirs. They are explored below (when
applicable) in the development of the different types of gradients in magma chambers.
Mechanisms to generate abrupt gradients
An abrupt juxtaposition of distinct magma compositions (and/or crystal content) can easily
be explained by the interaction of two distinct magma batches with different characteristics
(e.g., Eichelberger et al. 2006b). However, abrupt compositional gaps are also found in units
showing geochemical evidence for in situ differentiation (Bacon and Druitt 1988; Druitt and
Bacon 1989; Brophy 1991; Hildreth and Fierstein 2000). Due to the small size of crystals in
magmatic systems (typically around 0.1 to 5 mm) and the high viscosity of SiO2-rich melts (104
– 106 Pa·s), separating crystals from its melt is a slow process (Sparks et al. 1984; Reid et al.
1997; Anderson et al. 2000; Eichelberger et al. 2006a), particularly if the magma is undergoing
some convective stirring (Martin and Nokes 1989; Burgisser et al. 2005). Therefore, crystal-
melt separation is enhanced when convection stops, which usually occurs when a crystal-
bearing magma transforms into a locked crystalline mush (at ~50 vol% crystals for low strain
rates; Vigneresse et al. 1996; Petford 2003; Rosenberg and Handy 2005; Caricchi et al. 2007;
Champallier et al. 2008). Once the crystallinity is high enough (> ~70%), permeability becomes
so low that crystal-melt separation by compaction cannot occur on geologically reasonable
timescales (McKenzie 1985; Bachmann and Bergantz 2004). Hence, the most favourable
crystallinity window appears to be around the rheological threshold (~50-60 vol% crystals),
when convection has stopped but the system remains permeable enough for the interstitial
melt to be expelled from the compacting crystalline framework (Thompson 1972; Brophy
1991; Thompson et al. 2001). The low-density interstitial melt can then accumulate above the
mush, and generate a nearly crystal-free cap (Bacon and Druitt 1988; Bachmann and Bergantz
2004; Hildreth and Wilson 2007; Walker et al. 2007, Fig. 10). In the event of the formation
of a crystal-poor cap by interstitial liquid extraction from a crystalline mush, eruption of both
the crystal-poor cap and its underlying mush during the same eruptive episode will lead to the
observed abrupt gradient present in many systems (Fig. 11).
the CaSe OF the aBruPtLy ZOned Crater LaKe eruPtIOn
Despite the wide compositional gap between the nearly-aphyric Crater Lake rhyodacite
and the co-erupted crystal-rich andesite and clear evidence for a chemically open-system (e.g.,
Deciphering Magma Chamber Dynamics 663
106
Lifetime of large
Timescales to generate crystal-poor
magma reservoirs
105 rhyolite caps from silicic mushes
(with 50-60 vol% crystals)
Years
104
Low density silicic interstitial melt
migrating upward from a non-convecting crystal mush
103
Hindered settling Compaction
Figure
Figure 10.10Range of timescales of interstitial liquid extraction (by hindered settling and compaction) for
typical silicic mushes with crystallinities of 50-60 vol% (for more details, see Bachmann and Bergantz
2004).
Figure 11. Development of a crystal-
poor horizon above a crystalline mush
(and below a crystallizing roof layer).
When such a systems in tapped by
eruption, the least-viscous, crystal-
poor material is expected to erupt
first. When exhausted, the viscous-
crystal-rich portion can be drawn-out,
rapidly leading to chocking of the
eruption. In natural deposits, such as
the 7700 BP Crater Lake ignimbrite,
one expects the commonly observed Erupted volume
sequence of a crystal-poor base
covered by a relatively thin, crystal-
rich, co-magmatic layer.
low and high Sr magma types), theFigure 11
two compositional layers have strong affinities, which
suggest a genetic link by crystal fractionation (Bacon and Druitt 1988; Druitt and Bacon 1989):
(1) the rhyodacite has trace element contents that require extensive fractional crystallization.
(2) The interstitial glass trapped within the andesitic layer is very similar in composition to the
rhyodacite. (3) Temperature and oxygen fugacity are nearly identical in both compositional
layers. These chemical and thermal affinities, in addition to the physical proximity, indicate
that the eruption tapped a magma reservoir similar to the setting illustrated in Figure 11. The
zoning pattern was induced by crystal-melt separation.
Mechanism to generate linear gradients
As for the abrupt gradient case, the two models for generating linear gradients in magma
chambers are: in situ differentiation processes (Hildreth 1981; McBirney et al. 1985; Trial and
664 Bachmann & Bergantz
Spera 1990; Marsh 2002) and magma mixing by new addition from below (Smith 1979; Hervig
and Dunbar 1992; Eichelberger et al. 2000; Knesel and Duffield 2007). The dynamic template
for the latter hypothesis has been investigated by a number of authors (Sparks and Marshall
1986; Frost and Mahood 1987; Oldenburg et al. 1989); incomplete mixing leads to imperfect
chemical blending (often referred to as mingling), and stratification in the chamber (Jellinek et
al. 1999). However, the mechanisms leading to stratification by the in situ evolution process is
more controversial. The most commonly accepted mechanism is the “sidewall crystallization”
hypothesis, involving the extraction of interstitial liquid from a crystallizing boundary layer
at the cold margin of a magma chamber (Chen and Turner 1980; McBirney 1980; Rice 1981;
Huppert and Sparks 1984; Spera et al. 1984; McBirney et al. 1985; Wolff et al. 1990; de Silva
and Wolff 1995; Spera et al. 1995). Although this mechanism may play a role in the generation
of chemical and physical heterogeneities in some cases, several considerations preclude sidewall
crystallization being the dominant differentiation template in crustal magma chambers.
a. Bodies of eruptible magmas in large silicic chambers are sill-like, having a low wall to
roof (or floor) ratio, rendering sidewall crystallization inefficient (de Silva and Wolff
1995)
b. Little evidence for sidewall crystallization is preserved in extensively mapped and
studied plutonic bodies (e.g., McNulty et al. 2000; Barnes et al. 2001; zak and Paterson
2005; Eichelberger et al. 2006a; Miller et al. 2007; Walker et al. 2007).
c. The fact that most crystals found in silicic magmas (except for some rhyolites and
aplites) are complexly zoned requires that crystals undergo complex transport paths
and circulation. This is inconsistent with the monotonic sidewall crystallization
hypothesis, where crystals would not be available to circulate and respond to changes
in the magmatic environment.
d. The vertical stacking of several convecting magma “layers” is inherently unstable
as drag and entrainment should occur at the interfaces between the different magma
batches, leading to re-blending at a rate similar to estimated differentiation rate (see
Davaille 1999; Gonnermann et al. 2002; Bachmann and Bergantz 2004 for more
details in this process).
Thus, the generation of a linear gradient by a “box filling” mechanism of double-diffusive
convection in a largely fluid reservoir appears unlikely in magmatic situations. We argue that
mixing of heterogeneities by sluggish convection is the most common process of generating
linear gradients in magma chambers. Linear gradients will certainly develop when two
different magmas interact, but such gradients can also develop in situ as crystallization
and differentiation continue even during periods of closed-system evolution, generating
heterogeneities in crystallinity and composition (e.g., Bergantz and Ni 1999; Bergantz 2000;
Couch et al. 2001) that will be progressively smeared over the entire chamber.
the CaSe OF the LInearLy ZOned BIShOP tuFF
Numerous models have been proposed to explain the distinctive characteristics of the
Bishop Tuff. The most recent and comprehensive treatment of this unit is Hildreth and Wilson
(2007). They propose a model for the zonation of the Bishop Tuff in which different pockets of
rhyolitic melt are segregated from a large, subjacent crystal mush. As the conditions within the
mush vary over time, extracting interstitial liquid at different stages would have lead to rhyolitic
pockets with slightly different chemical characteristics. This model appears plausible, but does
not take into account the effect of convection, which must have occurred in a system showing
a thermal gradient of ~100 °C (Hildreth 1979; Hildreth and Wilson 2007; Ghiorso and Evans
2008).
Deciphering Magma Chamber Dynamics 665
On the basis of the observations mentioned above and new constraints provided by
chemical and thermal heterogeneities in quartz crystals, which require timescales of less than
102-104 years (Bindeman and Valley 2002; Wark et al. 2007), the characteristics of the Bishop
Tuff seem best explained by a model very similar to the one proposed by Hildreth and Wilson
(2007), but with the requirement of some late convection stirring in the rhyolite cap induced by
a hotter reintrusion (leading to the development of a thermal gradient and the growth of bright
rims on quartz from the late-erupted material; Wark et al. 2007). So both convective stirring
just prior to eruption, and progressive unmixing to produce the rhyolite, seems to have played
a role in the evolution of the Bishop Tuff magma (see Fig. 12).
Shallow silicic magma chamber
1. Stirring by dense crystal
Crystal-poor cap plumes from roof
~10 km
2. Stirring due to interaction of two magmas with different physical properties
(induced by different composition, temperature, gas content)
Figure 12.12
Figure Schematic illustration of the two main possible sources of heterogeneities in
shallow, crystal-poor, silicic magma chambers, such as the Bishop magma reservoir.
the generation of homogeneity
In light of the discussion above, homogenizing at the whole-rock scale a crystal-rich
silicic magma, the most viscous silicate liquid on Earth (Scaillet et al. 1998), appears unlikely.
We consider below possible mechanisms to reach homogeneity in viscous silicic magmas.
These ideas are largely based on studies of homogeneous volcanic systems, such as the Fish
Canyon magma body (see following section), but stress that more work is needed to better
constrain some of these issues.
The largest silicic magmas bodies (10,000+ km3) are certainly constructed incrementally
(e.g., Deniel et al. 1987; Petford et al. 2000; Lipman 2007). To obtain these large homogeneous
masses of magma in such a clearly open-system situation, new magmatic additions need to
be either (1) similar in composition to the growing reservoir, or (2) efficiently blended by
mechanical stirring. Furthermore, crystal-melt separation (which will recreate heterogeneities)
should be hampered (Fig. 13).
All three requirements can be met if the following conditions are obtained. As a new
magma batch intrudes in the upper crust, it will quickly start crystallizing in the cold, low-
pressure environment. Once crystallinity increases, further changes in temperature will be
slowed by the latent heat release by crystallization, and increasingly ineffective conductive
cooling (Koyaguchi and Kaneko 1999). Therefore, magma batches are expected to remain as
crystal mushes (with crystallinities > 50 vol%) for most of their time above the solidus.
666 Bachmann & Bergantz
Road to homogeneity in open-system magma reservoirs
(a) Mechanical stirring (b) Adding same material (c) Keep it from separating
Figure 13. Illustrations
Figure 13 of possible ways to render and/or keep open-system magma bodies homogeneous.
Magmatic additions to this growing silicic magma reservoir are restricted to either (1)
similar composition (Hervig and Dunbar 1992; Eichelberger et al. 2000), released by a filtering
lower crustal MASH zone (Hildreth and Moorbath 1988) or (2) more mafic composition (hotter,
less viscous, and generally denser), which will mostly pond beneath the low-density, silicic
mush (Miller and Miller 2002; Wiebe et al. 2002; Harper et al. 2004). This latter situation is
prone to rejuvenation and self-mixing as the mafic magma acts as a hot plate (Couch et al.
2001; Bachmann and Bergantz 2006). As the magma body never remains for long in a crystal-
poor situation, crystal-melt separation can only occur by interstitial melt extraction.
the CaSe OF the FISh CanyOn MaGMa BOdy
The Fish Canyon magma erupted during the largest known silicic volcanic eruption, the
crystal-rich Fish Canyon Tuff (45 vol% crystals) and its satellite units erupted from the same
magma chamber (the Pagosa Peak Dacite and Nutras Creek Dacite, Lipman et al. 1997).
This system is well known for its whole-rock homogeneity (at the scale of the hand sample
and larger), in major and trace element composition, mineralogy, modal abundance, isotopic
ratios, temperature, and water content (Whitney and Stormer 1985; Johnson and Rutherford
1989; Bachmann et al. 2002; Charlier et al. 2007). Similarly homogeneous units have been
described in detail in the Andes (Francis et al. 1989; Lindsay et al. 2001), and in the Great
Basin, USA (Maughan et al. 2002). However, at the mm scale, The Fish Canyon magma
shows an enormous range in composition (both major, trace and isotopic), implying complex
open system behaviour and excursions in intensive variables (Bachmann and Dungan 2002;
Bachmann et al. 2005; Charlier et al. 2007). Of particular note: (1) different patches of glass
analyzed by microdrilling with a mm spacing are heterogeneous in 87Sr/86Sri; also, (2) biotite
crystals have 87Sr/86Sri higher than any other component in the magma (including the glass
around them) and thus indicate a provenance from the Precambrian wall rocks surrounding
the magma chamber (Charlier et al. 2007). Therefore, as these biotites were derived from
the disaggregation of blocks of material from the walls and are now found in the Fish
Canyon magma, some crystal dispersal by convection is required. On the basis of Sr diffusive
reequilibration of these Precambrian biotites, this dispersal had to occur in less than 10,000
years prior to eruption.
One possibility to dissipate gradients in crystal-rich situations, as anticipated by the
“defrosting” hypothesis (Mahood 1990), is reheating of locked silicic mushes by interaction
with hot, intrusive magmas. As it has been described in multiple natural systems (Murphy et al.
Deciphering Magma Chamber Dynamics 667
2000; Couch et al. 2001; Bachmann et al. 2002; Bachmann and Bergantz 2006; Hildreth and
Wilson 2007), cooling silicic magmas reach a crystal mush state (crystal content >50-60%)
that becomes rigid at low strain rates (convecting currents stop). If sufficient heat is added by
a more mafic re-intrusion from below (a late reheating event of about 50 °C is recorded by the
mineral phases in the Fish Canyon magma (Bachmann and Dungan 2002; Bachmann et al.
2002; Bachmann et al. 2005), the crystalline framework will progressively re-melt from the
bottom until becoming liquid again (able to sustain convection). The stirring agent in such case
is a buoyancy force generated not only by heating, but more importantly by the melting of the
high crystallinity barrier at the bottom of the chamber (thermal expansion accounts for only
1-10% of density variation due to crystallinity changes) AND by the injection of volatiles from
the commonly gas-rich new magma (Bachmann and Bergantz 2003, 2006). This re-heating
can create an inverted density stratification that could evolve to full-scale self-mixing if the
reheating event can supply enough energy.
An important observation made in rejuvenated systems is that they appear not to have
significantly mixed with the more mafic magma that acted as a heat source; convective mixing
is limited to the reactivated mush (self-mixing of Couch et al. 2001). In the Fish Canyon
system, only a few mafic enclaves have been preserved in the late erupted intracaldera facies
of the Fish Canyon Tuff (Bachmann et al. 2002) although a large basaltic andesite unit (the
Huerto Andesite) erupted immediately after the Fish Canyon Tuff eruption (Parat et al. 2005,
2008). This absence of chemical mixing can be understood if buoyancy terms are compared.
As density variations due to compositional changes are of the order of a few percent (~10
times larger than buoyancy forces driven by thermal expansion during a reheating of around
100 °C), reheating alone will not be sufficient to invert the compositional density gradients and
enable large scale convection involving both mafic and silicic magmas.
the CaSe OF the hOMOGeneOuS GranItOIdS
Most large silicic plutons (such as those found in the Sierra Nevada Batholith) are very
similar to the Monotonous Intermediates. We consider the largest of them (e.g., the Half Dome
Granodiorite in the Tuolumne Intrusive Suite; Bateman and Chappell 1979) as unerupted
equivalents of units such as the Fish Canyon magma. We argue that they follow the same
incremental growth scenario, undergoing periodic intrusions of (a) evolved magma of similar
composition blending in the growing reservoirs and (b) more mafic intrusions ponding at the
base of the mush (Wiebe and Collins 1998; Robinson and Miller 1999; Waight et al. 2001;
Miller and Miller 2002; Wiebe et al. 2002), which induce convective stirring due to addition of
heat and gas from below (Wiebe et al. 2007). The only difference between plutonic and volcanic
rocks is that the former cooled slowly to full solidification, allowing time to undergo some local
crystal-liquid separation. Such late crystal-melt separation, leading to evolved granitic cupolas
on top of large granodioritic bodies, is observed in several cases of well-exposed plutonic
sections (Johnson et al. 1990; Barnes et al. 2001; Walker et al. 2007; Wiebe et al. 2007).
COnCLuSIOnS
Many caldera-forming (large-volume) ignimbrite sheets display chemical and thermal
heterogeneities, reflecting evacuation from a compositionally zoned shallow reservoir. These
chemically and thermally complex reservoirs are an expected consequence of open-system
processes that are common in magmatic systems. Mixing of magma batches with different
physical properties (density, viscosity), assimilation of wall rocks, and internal phase changes
(crystallization, gas exsolution) related to cooling and decompression all lead to chemical and
thermal gradients within the reservoirs.
668 Bachmann & Bergantz
In contrast to common belief, convection is not an efficient homogenizing agent at
all scales in viscous magmatic systems and can produce long-lasting gradients. Recent
experiments and numerical studies have shown that heterogeneities (in composition, crystal
content, temperature) will arise (or be preserved) as sluggish convection stirs the system. If
left active long enough in a chaotic mode, convection can produce homogeneity, except in the
case of thermally-induced heterogeneities, which are continuously re-established in cooling
magma bodies. Therefore, in most magmatic situations, gradients are inescapable. This
inference is on par with volcanic deposits from explosive eruptions (ignimbrites), which, in
many cases, show heterogeneities in their whole-rock composition, crystallinity, temperature
and volatile contents. Looking at the ignimbrite record, we conclude that most crystal-poor
magma chambers will preserve heterogeneities as the system undergoes sluggish convection.
Paradoxically, the most viscous of these magmas (the crystal-rich Monotonous
Intermediates) and many of the exposed fossil magma chambers (upper crustal silicic plutons)
are remarkably homogeneous at the hand sample scale. If complete homogenization is the
result of stirring by low Re convection, one would expect the most viscous magmas to be
the least homogeneous. This apparent contradiction can be resolved by recognizing that
homogeneous units are evolved, crystal-rich mush zones. Many silicic magmas persist as
near-solidus crystal mushes, due to a combination of slow conductive cooling and latent heat
buffering temperature close to the eutectic. We argue that large silicic crystal mushes mostly
grow by addition of compositionally similar magmas, released by lower to mid-crustal MASH
zones. As a result, the base-line homogeneity is dictated by processes occurring in source
regions, and is modified by open-system events in the upper crust. These mushes block denser
mafic reintrusions at their bases, and keep them from thorough mixing. This hot underplating
triggers periodic stirring through large-scale overturn of the reservoirs by self-mixing and
gas sparging. Such periodic chamber-wide stirring is well illustrated in erupted crystal-rich
units (the Monotonous Intermediates) that show thorough convective stirring in the last 1,000-
10,000 years prior to eruption.
Plutons cannot record a high-energy state, as they must cool slowly to their solidii.
They, nonetheless, commonly appear fairly homogeneous at the hand sample scale, and we
propose that (1) they are also stirred periodically by reintrusion events, and (2) are kept fairly
homogeneous by the sluggishness of crystal-liquid separation in silicic systems. However,
plutons do partially unmix by crystal-liquid separation late in their histories, as evidenced
by highly evolved granitic caps on top of several large plutonic bodies (Johnson et al. 1990;
Barnes et al. 2001; Miller and Miller 2002; Walker et al. 2007).
aCKnOWLedGMentS
Swiss NSF grant #200021-111709 provided support to O.B. and NSF grants EAR-
0440391 and EAR-0711551 to G.W.B. during the completion of this paper. We thank Christian
Huber, Josef Dufek and Philipp Ruprecht for numerous lively discussions on topics covered
by this paper. We are grateful to Calvin Miller, Peter Lipman, Guil Gualda and Fidel Costa for
comments on earlier drafts of this manuscript, and to Keith Putirka for the time and effort he
invested in this contribution and volume.
reFerenCeS
Anderson AT, Davis AM, Fangqiong L (2000) Evolution of the Bishop Tuff rhyolitic magma based on melt and
magnetite inclusions and zoned phenocrysts. J Petrol 41:449-473
Bachmann O, Bergantz GW (2003) Rejuvenation of the Fish Canyon magma body: a window into the evolution
of large-volume silicic magma systems. Geology 31:789-792
Deciphering Magma Chamber Dynamics 669
Bachmann O, Bergantz GW (2004) On the origin of crystal-poor rhyolites: extracted from batholithic crystal
mushes. J Petrol 45:1565-1582
Bachmann O, Bergantz GW (2006) Gas percolation in upper-crustal silicic crystal mushes as a mechanism for
upward heat advection and rejuvenation of near-solidus magma bodies. J Volcanol Geotherm Res 149:85-
102
Bachmann O, Bergantz GW (2008) The magma reservoirs that feed supereruptions. Elements 4:17-21
Bachmann O, Charlier BLA, Lowenstern JB (2007a) zircon crystallization and recycling in the magma chamber
of the rhyolitic Kos Plateau Tuff (Aegean Arc). Geology 35:73-76
Bachmann O, Dungan MA (2002) Temperature-induced Al-zoning in hornblendes of the Fish Canyon magma,
Colorado. Am Mineral 87:1062-1076
Bachmann O, Dungan MA, Bussy F (2005) Insights into shallow magmatic processes in large silicic magma
bodies: the trace element record in the Fish Canyon magma body, Colorado. Contrib Mineral Petrol
149:338-349
Bachmann O, Dungan MA, Lipman PW (2002) The Fish Canyon magma body, San Juan volcanic field,
Colorado: rejuvenation and eruption of an upper crustal batholith. J Petrol 43:1469-1503
Bachmann O, Miller CF, de Silva S (2007b) The volcanic-plutonic connection as a stage for understanding
crustal magmatism. J Volcanol Geotherm Res 167:1-23
Bacon CR (1983) Eruptive history of Mount Mazama and Crater Lake caldera, Cascade Range, U.S.A. J
Volcanol Geotherm Res 18:57-115
Bacon CR, Druitt TH (1988) Compositional evolution of the zoned calcalkaline magma chamber of Mount
Mazama, Crater Lake, Oregon. Contrib Mineral Petrol 98:224-256
Bacon CR, Lanphere MA (2006) Eruptive history and geochronology of Mount Mazama and the Crater Lake
region, Oregon. Geol Soc Am Bull 118:1331–1359
Bacon CR, Lowenstern JB (2005) Late Pleistocene granodiorite source for recycled zircon and phenocrysts in
rhyodacite lava at Crater Lake, Oregon. Earth Planet Sci Lett 233:277-293
Barboza SA, Bergantz GW (2000) Metamorphism and anatexis in the mafic complex contact aureole, Ivrea
zone, Northern Italy. J Petrol 41:1307-1327
Barnes CG, Burton BR, Burling TC, Wright JE, Karlsson HR (2001) Petrology and Geochemistry of the Late
Eocene Harrison Pass Pluton, Ruby Mountains Core Complex, Northeastern Nevada. J Petrol 42:901-929
Bateman PC, Chappell BW (1979) Crystallization, fractionation, and solidification of the Tuolumne Intrusive
Series, Yosemite National Park, California. Geol Soc Am Bull 90:465-482
Beard JS, Ragland PC, Crawford ML (2005) Reactive bulk assimilation: A model for crust-mantle mixing in
silicic magmas. Geology 33:681-684
Bergantz GW (2000) On the dynamics of magma mixing by reintrusion: implications for pluton assembly
processes. J Struct Geol 22:1297-1309
Bergantz GW, Breidenthal RE (2001) Non-stationary entrainment and tunneling eruptions: A dynamic link
between eruption processes and magma mixing. Geophys Res Lett 28:3075-3078
Bergantz GW, Ni J (1999) A numerical study of sedimentation by dripping instabilities in viscous fluids. Int J
Multiphase Flow 25:307-320
Bindeman IN, Valley JW (2002) Oxygen isotope study of the Long Valley magma system, California: isotope
thermometry and convection in large silicic magma bodies. Contrib Mineral Petrol 144:185-205
Bindeman IN, Valley JW (2003) Rapid generation of both high- and low- δ18O, large volume silicic magmas at
the Timber Mountain/Oasis Valley caldera complex, Nevada. Geol Soc Am Bull 115:581-595
Blake S, Ivey GN (1986a) Density and viscosity gradients in zoned magma chambers, and their influence
withdrawal dynamics. J Volcanol Geotherm Res 30:201-230
Blake S, Ivey GN (1986b) Magma-mixing and the dynamics of withdrawal from stratified reservoirs. J Volcanol
Geotherm Res 27:153-178
Bohrson WA, Spera FJ (2001) Energy-constrained open-system magmatic processes. II: Application of energy-
constrained assimilation-fractional crystallization (EC-AFC) model to magmatic systems. J Petrol
42:1019-1041
Broadwell JE, Mungal MG (1991) Large-scale structures and molecular mixing. Phys Fluid A 3:1193-1206
Brophy JG (1991) Composition gaps, critical crystallinity, and fractional crystallization in orogenic (calc-
alkaline) magmatic systems. Contrib Mineral Petrol 109:173-182
Brown SJA, Fletcher IR (1999) SHRIMP U-Pb dating of the preeruption growth history of zircons from the
340 ka Whakamaru Ignimbrite, New zealand: Evidence for >250 k.y. magma residence times. Geology
27:1035-1038
Brown SJA, Wilson CJN, Cole JW, Wooden J (1998) The Whakamaru group ignimbrites, Taupo Volcanic zone,
New zealand: evidence for reverse tapping of a zoned silicic magmatic system. J Volcanol Geotherm Res
84:1-37
Burgisser A, Bergantz GW, Breidenthal RE (2005) Addressing complexity in laboratory experiments: the scaling
of dilute multiphase flows in magmatic systems. J Volcanol Geotherm Res 141:245-265
670 Bachmann & Bergantz
Caricchi L, Burlini L, Ulmer P, Gerya T, Vassali M, Papale P (2007) Non-Newtonian rheology of crystal-bearing
magmas and implications for magma ascent dynamics. Earth Planet Sci Lett 264:402-419
Carrigan CR (1988) Biot number and thermos bottle effect: Implications for magma-chamber convection.
Geology 16:771–774
Champallier R, Bystricky M, Arbaret L (2008) Experimental investigation of magma rheology at 300 MPa:
From pure hydrous melt to 76 vol.% of crystals. Earth Planet Sci Lett 267:571-583
Charlier BLA, Bachmann O, Davidson JP, Dungan MA, Morgan D (2007) The upper crustal evolution of a
large silicic magma body: evidence from crystal-scale Rb/Sr isotopic heterogeneities in the Fish Canyon
magmatic system, Colorado. J Petrol 48:1875-1894
Chen CF, Turner JS (1980) Crystallization in double-diffusive system. J Geophys Res 85:2573-2593
Chesner CA (1998) Petrogenesis of the Toba Tuffs, Sumatra, Indonesia. J Petrol 39:397-438
Christiansen RL (1984) Yellowstone magmatic evolution: Its bearing on understanding large-volume explosive
volcanism. In: Explosive Volcanism: Its Inception, Evolution, and Hazards. Nat Res Council Studies in
Geophysics. Nat Acad Press, Washington D.C.:84-95
Christiensen JN, Halliday AN (1996) Rb-Sr and Nd isotopic compositions of melt inclusions from the Bishop
Tuff and the generation of silicic magma. Earth Planet Sci Lett 144:547-561
Civetta L, Orsi G, Pappalardo L, Fisher RV, Heiken G, Ort M (1997) Geochemical zoning, mingling, eruptive
dynamics and depositional processes-the Campanian Ignimbrite, Campi Flegrei, Italy. J Volcanol Geotherm
Res 75:183-219
Coltice N (2005) The role of convective mixing in degassing the Earth’s mantle. Earth Planet Sci Lett 234:15-
25
Coltice N, Schmalzl J (2006) Mixing times in the mantle of the early Earth derived from 2-D and 3-D numerical
simulations of convection. Geophys Res Lett 33, L23304, doi:10.1029/2006GL027707
Couch S, Sparks RSJ, Carroll MR (2001) Mineral disequilibrium in lavas explained by convective self-mixing
in open magma chambers. Nature 411:1037-1039
Davaille A (1999) Two-layer thermal convection in miscible fluids. J Fluid Mech 379:223-253
Davidson JP, Morgan DJ, Charlier BLA, Harlou R, Hora JM (2007) Microsampling and isotopic analysis of
igneous rocks: implications for the study of magmatic systems. Annu Rev Earth Planet Sci 35:273-311
Davidson JP, Tepley FJ III (1997) Recharge in volcanic systems: evidence from isotope profiles of phenocrysts.
Science 275:826-829
Davies GR, Halliday AN (1998) Development of the Long Valley rhyolitic magma system: Strontium and
neodymium isotope evidence from glass and individual phenocrysts. Geochim Cosmochim Acta 62:3561-
3574
de Silva SL (1991) Styles of zoning in the central Andean ignimbrites: Insights into magma chamber processes.
In: Andean Magmatism and its Tectonic Setting. Harmon RS, Rapela CW (eds) Geol Soc Am Spec Paper
265:233-243
de Silva SL, Wolff JA (1995) zoned magma chambers; the influence of magma chamber geometry on sidewall
convective fractionation. J Volcanol Geotherm Res 65:111-118
Deniel C, Vidal P, Fernandez A, Fort P, Peucat J-J (1987) Isotopic study of the Manaslu granite (Himalaya,
Nepal): inferences on the age and source of Himalayan leucogranites. Contrib Mineral Petrol 96:78-92
DePaolo DJ (1981) Trace element and isotopic effects of combined wallrock assimilation and fractional
crystallization. Earth Planet Sci Lett 53:189-202
Dimotakis PE (2005) Turbulent mixing. Annu Rev Fluid Mech 37:329-356
Druitt TH, Bacon CR (1989) Petrology of the zoned calcalkaline magma chamber of Mount Mazama, Crater
Lake, Oregon. Contrib Mineral Petrol 101:245-259
Dufek J, Bergantz GW (2005) Lower crustal magma genesis and preservation: a stochastic framework for the
evaluation of basalt–crust interaction. J Petrol 46:2167-2195
Duffield WA, Ruiz J, Webster JD (1995) Roof-rock contamination of magma along the top of the reservoir for
the Bishop Tuff. J Volcanol Geotherm Res 69:187-195
Dunbar NW, Kyle PR, Wilson CJN (1989) Evidence for limited zonation in silicic magma systems, Taupo
Volcanic zone, New zeland. Geology 17:234-236
Eichelberger JC, Chertkoff DG, Dreher ST, Nye CJ (2000) Magmas in collision; rethinking chemical zonation
in silicic magmas. Geology 28:603-606
Eichelberger JC, Izbekov PE, Browne BL (2006) Bulk chemical trends at arc volcanoes are not liquid lines of
descent. Lithos 87(1-2):135-154
Ferrachat S, Ricard Y (1998) Regular vs. chaotic mantle mixing. Earth Planet Sci Lett 155(1-2):75-86
Francalanci L, Varekamp JC, Vougioukalakis G, Defant MJ, Innocenti F, Manetti P (1995) Crystal retention,
fractionation and crustal assimilation in a convecting magma chamber, Nisyros Volcano, Greece. Bull
Volcanol 56:601-620
Francis PW, Sparks RSJ, Hawkesworth CJ, Thorpe RS, Pyle DM, Tait SR, Mantovani MS, McDermott F (1989)
Petrology and geochemistry of the Cerro Galan caldera, northwest Argentina. Geol Mag 126:515-547
Deciphering Magma Chamber Dynamics 671
Fridrich CJ, Mahood GA (1987) Compositional layers in the zoned magma chamber of the Grizzly Peak Tuff.
Geology 15:299-303
Frost TP, Mahood G (1987) Field, chemical, and physical constraints on mafic-felsic magma interaction in the
Lamarck Granodiorite, Sierra Nevada, California. Geol Soc Am Bull 99:272-291
Ghiorso MS, Evans BW (2008) Thermodynamics of rhombohedral oxide solid solutions and a revision of the
Fe-Ti oxide geothermometer and oxygen-barometer. Am J Sci (in press)
Gonnermann HM, Manga M, Jellinek AM (2002) Dynamics and longevity of an initially stratified mantle.
Geophys Res Lett 29; doi:10.1029/2002GL014851
Grout FF (1918) Two-phase convection in igneous magmas. J Geol 26:481-499
Grunder AL, Mahood GA (1988) Physical and chemical models of zoned silicic magmas: the Loma Seca Tuff
and Calabozos caldera, southern Andes. J Petrol 29:831-867
Halliday AN, Fallick AE, Hutchinson J, Hildreth W (1984) A Nd, Sr, O isotopic investigation into the causes of
chemical and isotopic zonation in the Bishop Tuff, California. Earth Planet Sci Lett 68:378-391
Harper B, Miller C, Koteas C, Cates N, Wiebe R, Lazzareschi D, Cribb W (2004) Granites, dynamic magma
chamber processes and pluton construction: the Aztec Wash pluton, Eldorado Mountains, Nevada, USA.
Trans R Soc Edinburgh: Earth Sci 95:277-296
Hervig RL, Dunbar NW (1992) Cause of chemical zoning in the Bishop (California) and Bandelier (New
Mexico) magma chambers. Earth Planet Sci Lett 111:97-108
Hildreth W (1979) The Bishop Tuff: evidence for the origin of the compositional zonation in silicic magma
chambers. Geol Soc Am Spec Pap 180:43-76
Hildreth W (1981) Gradients in silicic magma chambers: Implications for lithospheric magmatism. J Geophys
Res 86:10153-10192
Hildreth W (2004) Volcanological perspectives on Long Valley, Mammoth Mountain, and Mono Craters: several
contiguous but discrete systems. J Volcanol Geotherm Res 136:169-198
Hildreth W, Fierstein J (2000) Katmai volcanic cluster and the great eruption of 1912. Geol Soc Am Bull
112:1594-1620
Hildreth WS, Moorbath S (1988) Crustal contributions to arc magmatism in the Andes of Central Chile. Contrib
Mineral Petrol 98:455-499
Hildreth WS, Wilson CJN (2007) Compositional zoning in the Bishop Tuff. J Petrol 48:951-999
Huppert HE, Sparks RSJ (1984) Double-diffusive convection due to crystallization in magmas. Annu Rev Earth
Planet Sci 12:11-37
Jellinek AM, Kerr RC (1999) Mixing and compositional stratification produced by natural convection: 2.
Applications to the differentiation of basaltic and silicic magma chambers and komatiite lava flows. J
Geophys Res 104:7203-7218
Jellinek AM, Kerr RC, Griffiths RW (1999) Mixing and compositional stratification produced by natural
convection: 1. Experiments and their applications to Earth’s core and mantle. J Geophys Res 104:7183-
7201
Jellinek AM, Kerr RC (2001) Magma dynamics, crystallization, and chemical differentiation of the 1959 Kilauea
Iki lava lake, Hawaii, revisited. J Volcanol Geotherm Res 110:235-263
Johnson CM (1989) Isotopic zonations in silicic magma chambers. Geology 17:1136-1139
Johnson CM, Czamanske GK, Lipman PW (1990) H, O, Sr, Nd, and Pb isotope geochemistry of the Latir
volcanic field and cogenetic intrusions, New Mexico, and relations between evolution of a continental
magmatic center and modifications of the lithosphere. Contrib Mineral Petrol 104:99-124
Johnson M, Rutherford M (1989) Experimentally determined conditions in the Fish Canyon Tuff, Colorado,
magma chamber. J Petrol 30:711-737
Keller J (1969) Origin of rhyolites by anatectic melting of granitic crustal rocks; the example of rhyolitic pumice
from the island of Kos (Aegean sea). Bull Volcanol 33:942-959
Knesel KM, Davidson JP (1997) The origin and evolution of large-volume silicic magma systems: Long Valley
caldera. Int Geol Rev 39:1033-1052
Knesel KM, Davidson JP, Duffield WA (1999) Evolution of silicic magma through assimilation and subsequent
recharge: Evidence from Sr isotopes in sanidine phenocrysts, Taylor Creek Rhyolite, NM. J Petrol 40:773-
786
Knesel KM, Duffield WA (2007) Gradients in silicic eruptions caused by rapid inputs from above and below
rather than protracted chamber differentiation. J Volcanol Geotherm Res 167:181-197
Koyaguchi T, Kaneko K (1999) A two-stage thermal evolution model of magmas in continental crust. J Petrol
40:241-254
Linden PF, Redondo JM (1991) Molecular mixing in Rayleigh-Taylor instability. Part 1: global mixing. Phys
Fluid A 3:1269-1277
Lindsay JM, Schmitt AK, Trumbull RB, De Silva SL, Siebel W, Emmermann R (2001) Magmatic evolution
of the La Pacana caldera system, Central Andes, Chile: Compositional variation of two cogenetic, large-
volume felsic ignimbrites. J Petrol 42:459-486
672 Bachmann & Bergantz
Lipman PW (1967) Mineral and chemical variations within an ash-flow sheet from Aso caldera, South Western
Japan. Contrib Mineral Petrol 16:300-327
Lipman PW (2000) The central San Juan caldera cluster: Regional volcanic framework. In: Ancient Lake
Creede: Its Volcano-Tectonic Setting, History of Sedimentation, and Relation of Mineralization in the
Creede Mining District. Bethke PM, Hay RL (eds) Geol Soc Am Spec Paper 346:9-69
Lipman PW (2007) Incremental assembly and prolonged consolidation of Cordilleran magma chambers:
Evidence from the Southern Rocky Mountain volcanic field. Geosphere 3:1-29
Lipman PW, Christiansen RL, O’Connor JT (1966) A compositionally zoned ash-flow sheet in southern Nevada.
USGS Prof Paper 524-F:1-47
Lipman PW, Dungan MA, Bachmann O (1997) Comagmatic granophyric granite in the Fish Canyon Tuff,
Colorado: Implications for magma-chamber processes during a large ash-flow eruption. Geology 25:915-
918
Mahood GA (1990) Second reply to comment of R.S.J. Sparks, H.E. Huppert, and C.J.N. Wilson on “Evidence
for long residence times of rhyolitic magma in the Long Valley magmatic system: the isotopic record in
precaldera lavas of Glass Mountain”. Earth Planet Sci Lett 99:395-399
Mahood GA, Halliday AN (1988) Generation of high-silica rhyolite: a Nd, Sr, and O isotopic study of Sierra La
Primavera, Mexican neovolcanic belt. Contrib Mineral Petrol 100:183-191
Marsh BD (1988) Crystal capture, sorting, and retention in convecting magma. Geol Soc Am Bull 100:1720-
1737
Marsh BD (1989) Magma chambers. Annu Rev Earth Planet Sci 17:439-474
Marsh BD (2002) On bimodal differentiation by solidification front instability in basaltic magmas, part 1: basic
mechanics. Geochim Cosmochim Acta 66:2211-2229
Martin D, Nokes R (1989) A fluid-dynamical study of crystal settling in convecting magmas. J Petrol 30:1471-
1500
Maughan LL, Christiansen EH, Best MG, Gromme CS, Deino AL, Tingey DG (2002) The Oligocene Lund Tuff,
Great Basin, USA: a very large volume monotonous intermediate. J Volcanol Geotherm Res 113:129-157
McBirney AR (1980) Mixing and unmixing of magmas. J Volcanol Geotherm Res 7:357-371
McBirney AR, Baker BH, Nilson RH (1985) Liquid fractionation. Part 1: Basic principles and experimental
simulations. J Volcanol Geotherm Res 24:1-24
McKenzie DP (1985) The extraction of magma from the crust and mantle. Earth Planet Sci Lett 74:81-91
McNulty BA, Tobish OT, Cruden AR, Gilder S (2000) Multi-stage emplacement of the Mount Givens pluton,
central Sierra Nevada batholith, California. Geol Soc Am Bull 112:119-135
Michael PJ (1983) Chemical differentiation of the Bishop Tuff and other high-silica magmas through
crystallization processes. Geology 11:31-34
Miller CF, Miller JS (2002) Contrasting stratified plutons exposed in tilt blocks, Eldorado Mountains, Colorado
River Rift, NV, USA. Lithos 61:209-224
Miller CF, Wark DA (2008) Supervolcanoes and their explosive supereruptions. Elements 4:11-16
Miller JS, Matzel JEP, Miller CF, Burgess SD, Miller RB (2007) zircon growth and recycling during the
assembly of large, composite arc plutons. J Volcanol Geotherm Res 167:282-299
Mills J, James G., Saltoun BW, Vogel TA (1997) Magma batches in the Timber Mountain magmatic system,
Southwestern Nevada Volcanic Field, Nevada, USA. J Volcanol Geotherm Res 78:185-208
Milner DM, Cole JW, Wood CP (2003) Mamaku Ignimbrite: a caldera-forming ignimbrite erupted from a
compositionally zoned magma chamber in Taupo Volcanic zone, New zealand. J Volcanol Geotherm Res
122:243-264
Morse SA (1986) Thermal structure of crystallizing magma with two-phase convection. Geol Mag 123:205-
214
Murphy MD, Sparks RSJ, Barclay J, Carroll MR, Brewer TS (2000) Remobilization of andesitic magma by
intrusion of mafic magma at the Soufrière Hills Volcano, Montserrat, West Indies. J Petrol 41:21-42
Neri A, Esposti Ongaro T, Macedonio G, Gidaspow D (2003) Multiparticle simulation of collapsing volcanic
columns and pyroclastic flow. J Geophys Res 108(B4):2202, doi:2210.1029/2001JB000508
Oldenburg CM, Spera FJ, Yuen DA, Sewell G (1989) Dynamic mixing in magma bodies: Theory, simulations,
and implications. J Geophys Res 94:9215-9236
Ottino JM (1989) The Kinematics of Mixing: Stretching, Chaos, and Transport. Cambridge University Press,
Cambridge
Parat F, Dungan MA, Lipman PW (2005) Contemporaneous trachyandesitic and calc-alkaline volcanism of the
Huerto Andesite, San Juan Volcanic Field, Colorado, USA. J Petrol 46(5):859-891
Parat F, Holtz F, Feig S (2008) Pre-eruptive Conditions of the Huerto Andesite (Fish Canyon System, San Juan
Volcanic Field, Colorado): Influence of volatiles (C-O-H-S) on phase equilibria and mineral composition.
J Petrol 49:911-935
Peressini G, quick JE, Sinigoi S, Hofmann AW, Fanning M (2007) Duration of a large mafic intrusion and heat
transfer in the lower crust: a SHRIMP U–Pb zircon Study in the Ivrea–Verbano zone (Western Alps, Italy).
J Petrol 48:1185-1218
Deciphering Magma Chamber Dynamics 673
Petcovic HL, Dufek J (2005) Modeling of magma flow and cooling dikes: implications for emplacement of
Columbia River Flood Basalts. J Geophys Res 110:1-15
Petford N (2003) Rheology of granitic magmas during ascent and emplacement. Annu Rev Earth Planet Sci
31:399-427
Petford N, Cruden AR, McCaffrey KJW, Vigneresse J-L (2000) Granite magma formation, transport and
emplacement in the Earth’s crust. Nature 408:669-673
Petrelli M, Perugini D, Poli G (2006) Time-scales of hybridisation of magmatic enclaves in regular and chaotic
flow fields: petrologic and volcanologic implications. Bull Volcanol 68:285-293
Phillips JC, Woods AW (2002) Suppression of large-scale magma mixing by melt-volatile separation. Earth
Planet Sci Lett 204:47-60
Powell RL, Mason SG (1982) Dispersion by laminar flow. AiChE J 28:286-293
Reid MR, Coath CD, Harrison TM, McKeegan KD (1997) Prolonged residence times for the youngest rhyolites
associated with Long Valley Caldera: 230Th-238U microprobe dating of young zircons. Earth Planet Sci Lett
150:27-39
Rice A (1981) Convective fractionation: a mechanism to provide cryptic zoning (macrosegregation), layering
crescumulates, banded tuffs and explosive volcanism in igneous processes. J Geophys Res 86:405-417
Robinson DM, Miller CF (1999) Record of magma chamber processes preserved in accessory mineral
assemblages, Aztec Wash pluton, Nevada. Am Mineral 84:1346-1353
Rosenberg CL, Handy MR (2005) Experimental deformation of partially melted granite revisited: implications
for the continental crust. J Metamorph Geol 23:19-28
Rudman M (1992) Two-phase natural convection: implications for crystal settling in magma chambers. Phys
Earth Planet Int 72:153-172
Ruprecht P, Bergantz GW, Dufek J (2008) Modeling of Gas-Driven Magmatic Overturn: Tracking of
Phenocryst Dispersal and Gathering During Magma Mixing. Geochem. Geophys. Geosyst 9:q07017,
doi:10.1029/2008GC002022
Scaillet B, Holtz F, Pichavant M (1998) Phase equilibrium constraints on the viscosity of silicic magmas 1.
Volcanic-plutonic comparison. J Geophys Res 103:27257-27266
Schmitt AK, de Silva SL, Trumbull RB, Emmermann R (2001) Magma evolution in the Purico ignimbrite
complex, nothern Chile: evidence for zoning of a dacite magma by injection of rhyolite melts following
mafic recharge. Contrib Mineral Petrol 140:680-700
Simakin A, Botcharnikov R (2001) Degassing of stratified magma by compositional convection. J Volcanol
Geotherm Res 105:207-224
Simakin A, Schmeling H, Trubissyn V (1997) Convection in melts due to sedimentary crystal flux from above.
Phys Earth Planet Int 102:185-200
Simon JI, Reid MR (2005) The pace of rhyolite differentiation and storage in an ‘archetypical’ silicic magma
system, Long Valley, California. Earth Planet Sci Lett 235:123-140
Sinton JM, Detrick RS (1992) Mid-Ocean Ridge Magma Chambers. J Geophys Res 97:197-216
Smith RL (1979) Ash-flow magmatism. Geol Soc Am Spec Paper 180:5-25
Smith RL, Bailey RA (1966) The Bandelier Tuff—A study of ash-flow eruption cycles from zoned magma
chambers. Bull Volcanol 29:83-104
Sparks RSJ, Huppert HE, Turner JS (1984) The fluid dynamics of evolving magma chambers. Phil Trans R Soc
London 310:511-534
Sparks RSJ, Marshall LA (1986) Thermal and mechanical constraints on mixing between mafic and silicic
magmas. J Volcanol Geotherm Res 29:99-124
Spera FJ, Oldenburg CM, Christiensen C, Todesco M (1995) Simulations of convection with crystallization in
the system KAlSi2O6-CaMgSi2O6: Implications for compositionally zoned magma bodies. Am Mineral
40:1188-1207
Spera FJ, Yuen DA, Kemp DV (1984) Mass transfer rates along vertical walls in magma chambers and marginal
upwelling. Nature 310:764-767
Streck MJ, Grunder AL (1997) Compositional gradients and gaps in high-silica rhyolites of the Rattlesnake Tuff,
Oregon. J Petrol 38:133-163
Thompson, Smith, Malpas (2001) Origin of oceanic phonolites by crystal fractionation and the problem of the
Daly gap: an example from Rarotonga. Contrib Mineral Petrol 142:336-346
Thompson AB, Matile L, Ulmer P (2002) Some thermal constraints on crustal assimilation during fractionation
of hydrous, mantle-derived magmas with examples from Central Alpine Batholiths J Petrol 43:403-422
Thompson RN (1972) Evidence for a chemical discontinuity near the basalt-andesite transition in many
anorogenic volcanic suites. Nature 236:106-110
Trial AF, Spera FJ (1990) Mechanisms for the generation of compositional heterogeneities in magma chambers.
Geol Soc Am Bull 102:353-367
Trial AF, Spera FJ (1992) Simulations of magma withdrawal from compositionally zoned bodies. J Geophys
Res 97:6713-6733
674 Bachmann & Bergantz
Varga RJ, Smith BM (1984) Evolution of the early Oligocene bonanza caldera, Northeast San Juan volcanic
field, Colorado. J Geophys Res 89:8679-8694
Vazquez JA, Reid MR (2004) Probing the Accumulation History of the Voluminous Toba Magma. Science
305:991-994
Vigneresse J-L, Barbey P, Cuney M (1996) Rheological transitions during partial melting and crystallization
with application to felsic magma segregation and transfer. J Petrol 37:1579-1600
Waight TE, Wiebe RA, Krogstad EJ, Walker RJ (2001) Isotopic responses to basaltic injection into silicic
magma chambers: a whole-rock and microsampling study of macrorhythmic units in the Pleasant Bay
layered gabbro-diorite complex, Maine, USA. Contrib Mineral Petrol 142:323-335
Walker BJ, Miller CF, Lowery LE, Wooden JL, Miller JS (2007) Geology and geochronology of the Spirit
Mountain batholith, southern Nevada: implications for timescales and physical processes of batholith
construction. J Volcanol Geotherm Res 167:239-262
Wallace PJ, Anderson AT, Davis AM (1999) Gradients in H2O, CO2, and exsolved gas in a large-volume silicic
magma chamber: interpreting the record preserved in the melt inclusions from the Bishop Tuff. J Geophys
Res 104:20097-20122
Wark DA, Hildreth WS, Spear FS, Cherniak DJ, Watson EB (2007) Pre-eruption recharge in the Bishop Tuff
magma chamber. Geology 35:235-238
Whitney JA, Dorais MJ, Stormer JC, Kline SW, Matty DJ (1988) Magmatic conditions and development of
chemical zonation in the Carpenter Ridge Tuff, Central San Juan volcanic field, Colorado. Am J Sci
288:16-44
Whitney JA, Stormer JC Jr. (1985) Mineralogy, petrology, and magmatic conditions from the Fish Canyon Tuff,
central San Juan volcanic field, Colorado. J Petrol 26:726-762
Wiebe R, Wark D, Hawkins D (2007) Insights from quartz cathodoluminescence zoning into crystallization of
the Vinalhaven granite, coastal Maine. Contrib Mineral Petrol 154:439-453
Wiebe RA, Blair KD, Hawkins DP, Sabine CP (2002) Mafic injections, in situ hybridization, and crystal
accumulation in the Pyramid Peak granite, California. Geol Soc Am Bull 114:909-920
Wiebe RA, Collins WJ (1998) Depositional features and stratigraphic sections in granitic plutons: implications
for the emplacement and crystallization of granitic magma. J Struct Geol 20:1273-1289
Wilson CJN, Blake S, Charlier BLA, Sutton AN (2006) The 26.5 ka Oruanui Eruption, Taupo Volcano, New
zealand: Development, characteristics and evacuation of a large rhyolitic magma body. J Petrol 47:35-69
Wilson CJN, Hildreth W (1997) The Bishop Tuff: new insights from eruptive stratigraphy. J Geol 105:407-439
Wolff JA (1985) zonation, mixing and eruption of a silica-undersaturated alkaline magma: a case study from
Tenerife, Canary Islands. Geol Mag 122:623-640
Wolff JA, Ramos FC, Davidson JP (1999) Sr isotope disequilibrium during differentiation of the Bandelier Tuff:
Constraints on the crystallization of a large rhyolitic magma chamber. Geology 27:495-498
Wolff JA, Storey M (1984) zoning in highly alkaline magma bodies. Geol Mag 121:563-575
Wolff JA, Worner G, Blake S (1990) Gradients in physical parameters in zoned felsic magma bodies: implications
for evolution and eruptive withdrawal. J Volcanol Geotherm Res 43:37-55
Worner G, Schmincke H-U (1984a) Mineralogical and chemical zonation of the Laacher See tephra. J Petrol
25:805-835
Worner G, Schmincke H-U (1984b) Petrogenesis of the Laacher See tephra. J Petrol 25: 836-851
zak J, Paterson SR (2005) Characteristics of internal contacts in the Tuolumne Batholith, central Sierra Nevada,
California (USA): Implications for episodic emplacement and physical processes in a continental arc
magma chamber. Geol Soc Am Bull 117:1242-1255
You might also like
- Structural Controls On Ore Genesis - Review in Economic High100% (1)Structural Controls On Ore Genesis - Review in Economic High190 pages
- The Interpretation of Igneous Rocks by K. G. Cox, J. D. Bell, R. J. Pankhurst (Auth.)No ratings yetThe Interpretation of Igneous Rocks by K. G. Cox, J. D. Bell, R. J. Pankhurst (Auth.)459 pages
- Introduction To Igneous and Metamorphic Rocks JD Winter PDFNo ratings yetIntroduction To Igneous and Metamorphic Rocks JD Winter PDF668 pages
- Characteristics of Porphyry Copper Deposits100% (2)Characteristics of Porphyry Copper Deposits4 pages
- T.M. Porter (Editor) - Super Porphyry Copper and Gold Deposits - A Global Perspective. 2-PGC Publishing (2005) PDF100% (4)T.M. Porter (Editor) - Super Porphyry Copper and Gold Deposits - A Global Perspective. 2-PGC Publishing (2005) PDF276 pages
- Special Publication No.4 of The Society For Geology Applied To Mineral Deposits100% (1)Special Publication No.4 of The Society For Geology Applied To Mineral Deposits602 pages
- Depositional Environment and Gross Lithologies Interpretation100% (2)Depositional Environment and Gross Lithologies Interpretation120 pages
- Texture and Structure of Hydrothermal Ores100% (2)Texture and Structure of Hydrothermal Ores34 pages
- Sillitoe. 1995. Exploration and Discovery of Mineral Deposits Through The CircumpacificNo ratings yetSillitoe. 1995. Exploration and Discovery of Mineral Deposits Through The Circumpacific132 pages
- Reading 3 - Minerals and Rock Forming ProcessesNo ratings yetReading 3 - Minerals and Rock Forming Processes35 pages
- Geochemical Analyses For Mapping Geological ProcessesNo ratings yetGeochemical Analyses For Mapping Geological Processes92 pages
- Bastin. Interpretation of Ore Textures PDFNo ratings yetBastin. Interpretation of Ore Textures PDF142 pages
- Introduction To Earth Science: Fourth EditionNo ratings yetIntroduction To Earth Science: Fourth Edition112 pages
- Geol Resources Lecture 15 (Greisen SN-W and Pegmatites)No ratings yetGeol Resources Lecture 15 (Greisen SN-W and Pegmatites)29 pages
- Day 1 Brief ioGAS Intro Revised and Course Schedule PDFNo ratings yetDay 1 Brief ioGAS Intro Revised and Course Schedule PDF11 pages
- Evolution of Regoliths - Geochemical Exploration100% (1)Evolution of Regoliths - Geochemical Exploration12 pages
- Transport of Metals by Hydrothermal FluidsNo ratings yetTransport of Metals by Hydrothermal Fluids13 pages
- MagmaticArcs&Cu AuMineralizationsIndonesia PDFNo ratings yetMagmaticArcs&Cu AuMineralizationsIndonesia PDF52 pages
- Thogeochemistry & Gold Lithogeochemistry & Gold - Prendergast KylieNo ratings yetThogeochemistry & Gold Lithogeochemistry & Gold - Prendergast Kylie32 pages
- The Geological Characteristics Geochemical Signature and Geophysical Expression of Porphyry Copper Gold Deposits in The Circum Pacific RegionNo ratings yetThe Geological Characteristics Geochemical Signature and Geophysical Expression of Porphyry Copper Gold Deposits in The Circum Pacific Region5 pages
- A Geochemical Classification For Granitic RocksNo ratings yetA Geochemical Classification For Granitic Rocks16 pages
- Empresa Minera Especial Tintaya S. A.: Santiago-FuleNo ratings yetEmpresa Minera Especial Tintaya S. A.: Santiago-Fule54 pages
- 1london Granitic Pegmatites - Scientific Wonders and Economic BonanzasNo ratings yet1london Granitic Pegmatites - Scientific Wonders and Economic Bonanzas6 pages
- Characteristics of High Temperature Carbonat Hosted Massive Sulphide Ores United States, Exico and PeruNo ratings yetCharacteristics of High Temperature Carbonat Hosted Massive Sulphide Ores United States, Exico and Peru30 pages
- Linnen - 2014 - Geochemistry of The Rare-Earth Element, NB, Ta, HF, and ZR Deposits100% (1)Linnen - 2014 - Geochemistry of The Rare-Earth Element, NB, Ta, HF, and ZR Deposits26 pages
- 07 AGS22 Hoschke (Geophysics) Final - 2No ratings yet07 AGS22 Hoschke (Geophysics) Final - 216 pages
- Gaillard Et Al (2020) - Use of Lithogeochem in Delineating Hydrothermal Fluid Pathways & Vectoring Towards Au Mineralization in Malartic DistrictNo ratings yetGaillard Et Al (2020) - Use of Lithogeochem in Delineating Hydrothermal Fluid Pathways & Vectoring Towards Au Mineralization in Malartic District20 pages
- Geology of The Nairobi Region, Kenya PDFNo ratings yetGeology of The Nairobi Region, Kenya PDF1 page
- Porphyry To Epithermal Transition in The Altar Cu - (Au-Mo) Deposit, Argentina, Studied by Cathodoluminescence, LA-ICP-MS, and Fluid Inclusion AnalysisNo ratings yetPorphyry To Epithermal Transition in The Altar Cu - (Au-Mo) Deposit, Argentina, Studied by Cathodoluminescence, LA-ICP-MS, and Fluid Inclusion Analysis36 pages
- Resource Geology - 2008 - Rosana - Cikidang Hydrothermal Gold Deposit in Western Java IndonesiaNo ratings yetResource Geology - 2008 - Rosana - Cikidang Hydrothermal Gold Deposit in Western Java Indonesia12 pages
- Interpretation of Multi-Element Geochemistry For The Classification and Zoning of Intrusion-Related Mineral Systems100% (1)Interpretation of Multi-Element Geochemistry For The Classification and Zoning of Intrusion-Related Mineral Systems14 pages
- Using Mineral Chemistry To Aid Exploration: A Case Study From The Resolution Porphyry Cu-Mo Deposit, ArizonaNo ratings yetUsing Mineral Chemistry To Aid Exploration: A Case Study From The Resolution Porphyry Cu-Mo Deposit, Arizona28 pages
- Nickel Laterite Ore Deposits: Weathered Serpentinites: Charles R. M. Butt and Dominique CluzelNo ratings yetNickel Laterite Ore Deposits: Weathered Serpentinites: Charles R. M. Butt and Dominique Cluzel6 pages
- Intrusion Related Gold Deposits Morrison GNo ratings yetIntrusion Related Gold Deposits Morrison G20 pages
- An Evaluation of Plate Tectonics Models PDFNo ratings yetAn Evaluation of Plate Tectonics Models PDF28 pages
- Geological Models in Epithermal Porphyry Exploration Terry Leach Legacy100% (3)Geological Models in Epithermal Porphyry Exploration Terry Leach Legacy11 pages
- Sedimentary Model of K-Successions Sandstones in HNo ratings yetSedimentary Model of K-Successions Sandstones in H17 pages
- Lateralandverticalfacieschangesof Paleocene Kolosh FormationNo ratings yetLateralandverticalfacieschangesof Paleocene Kolosh Formation14 pages
