Principles of Igneous and Metamorphic Petrology

Winter
2e Principles of
ISBN 978-1-29202-153-9 Igneous and Metamorphic Petrology
John D. Winter
Second Edition
9 781292 021539
Pearson Education Limited
Edinburgh Gate
Harlow
Essex CM20 2JE
England and Associated Companies throughout the world

Visit us on the World Wide Web at: www.pearsoned.co.uk

© Pearson Education Limited 2014

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted
in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without either the
prior written permission of the publisher or a licence permitting restricted copying in the United Kingdom
issued by the Copyright Licensing Agency Ltd, Saffron House, 6–10 Kirby Street, London EC1N 8TS.

All trademarks used herein are the property of their respective owners. The use of any trademark
in this text does not vest in the author or publisher any trademark ownership rights in such
trademarks, nor does the use of such trademarks imply any affiliation with or endorsement of this
book by such owners.

ISBN 10: 1-292-02153-5

ISBN 13: 978-1-292-02153-9

British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library

Printed in the United States of America

 Magma Diversity

The previous analysis indicates that only mafic intru- occur without crystal settling. As we shall see in Section 4,
sions are susceptible to the effects of differentiation by grav- mixing of silicic (crustal melts) and mafic (mantle melts) is
ity settling of crystals. There is evidence, however, that a popular alternative interpretation to the evolutionary
many silicic bodies have evolved along a liquid line of de- trends in some of these systems.
scent. Several investigators have used the approach to the In addition to gravity settling, three other mechanisms
ternary eutectic in the systems albite–orthoclase–silica may facilitate the separation of crystals and liquid. Filter
(Figure 3) to indicate the evolution of late granitic liquids pressing (compaction), mentioned earlier in reference to
toward the thermal minimum, or eutectic, composition. Oth- partial melting, is also possible in crystal mushes that form
ers have used the same clustering of analyses around the as cumulates or crystal suspensions. The amount of trapped
minimum to indicate eutectic partial melts of sialic material intercumulus liquid between cumulate minerals may be as
in the continental crust. Certainly, a eutectic magma could high as 60 vol. % (Irvine, 1980b). With the added weight of
result from either process. Included in Figure 3 are the further accumulation, the crystal mush may be compacted
cation norm components of the successive intrusive phases (McKenzie, 1984), squeezing much of the liquid out into the
of the Tuolumne Intrusive Series. Note the progressive ap- main magma body. Another method of filter pressing in-
proach of these magmas to the eutectic composition along a volves the movement of a phenocryst-laden crystal mush.
trend that follows the decompression eutectic path. What- Any constriction in the conduit causes the crystals to inter-
ever the origin of the parental magma in this case, this series fere and slow with respect to the liquid.
appears to have evolved toward the low-pressure thermal Another similar mechanism by which crystals may
minimum. Harker-type bivariate variation diagrams of intru- be segregated from the liquid occurs when crystal-rich mag-
sive sequences also indicate evolutionary trends. Bateman mas flow in a laminar fashion near the walls of the magma
and Chappell (1979) interpreted the trends for the Tuolumne body. The process is known as flow segregation (or
Intrusive Series to be the result of fractional crystallization. flow[age] separation, or flow[age] differentiation). The mo-
Similar fractional crystallization-based interpretations for tion of the magma past the stationary walls of country rock
Sierran zoned granitoids have been proposed by Bateman (Figure 4) creates shear in the viscous liquid as a result of
and Nokelberg (1978) and Noyes et al. (1983). Although the the velocity gradient near the walls. The resulting differen-
compositional trends may be compatible with fractional tial motion forces the magma to flow around phenocrysts,
crystallization, crystal settling has been considered very
problematic in such viscous silicic magmas (Brandeis and
Jaupart, 1986; Sparks et al., 1984). Harper et al. (2005),
however, cited viscosity and field criteria supporting crystal Relative shear
settling in hydrous granites. Later in this chapter, we shall
explore methods by which fractional crystallization can

Qtz

Ptotal = PH O

Direction of flow
2

Country
Rock

1 atm

1 GPa Grain dispersive

3 GPa pressure

Magma
Ab Or

FIGURE 3 Position of the H2O-saturated ternary eutectic

(minimum melt composition) in the albite–orthoclase–silica FIGURE 4 Flow of magma adjacent to a wall of country
system at various pressures. The shaded portion represents the rock results in differential motion and shear in the magma.
composition of most granites. Included are the compositions Where such shear is constricted, as between adjacent
of the Tuolumne Intrusive Series, with the arrow showing the phenocrysts or between phenocrysts and the contact, a
direction of the trend from early to late magma batches. force (called grain-dispersive pressure) is generated and
Experimental data from Wyllie et al. (1976). pushes the phenocrysts apart and away from the contact.

217
 Magma Diversity

thereby exerting pressure on them at constrictions where for olivine with decreasing pressure requires that a lot of
phenocrysts are near one another or near the contact itself. olivine must form as the melt composition follows the liq-
The pressure, called grain dispersive pressure (Komar, uidus away from the olivine side of the diagram in a rising
1972), forces the grains apart and away from the contact. basaltic melt (see Problem 1). Thus, the relative amount of
This effect is greatest near the walls, and it drops off quickly olivine that crystallizes with a rising basaltic magma will
toward the magma interior, where the flow becomes uni- be far greater than the amount that forms during isobaric
form. Phenocrysts thus concentrate away from the walls to crystallization.
mitigate the pressure buildup. This concentration is most The broad acceptance of Bowen’s (1928) contention
apparent in dikes and sills, where the volume affected by that fractional crystallization is the predominant mechanism
the contact comprises a substantial proportion of the body, of magmatic differentiation is now being questioned. This
resulting in a distinct concentration of coarse phenocrysts one process cannot account for all of the diversity in the
toward the center (Figure 5). Flow segregation is an inter- broad spectrum of natural igneous rocks, even if we allow
esting, though localized, phenomenon and cannot be respon- for variations due to the influence of changes in pressure or
sible for the evolution of more than a small proportion of ig- associated fluids. Some observed chemical trends simply
neous rocks. cannot be accomplished by fractional crystallization. Other
A third mechanism involves the separation and rise of classical examples of fractional crystallization have not
buoyant liquids from boundary layers in which crystals form withstood more critical analysis and the test of time. For ex-
without themselves moving. This relatively new model has be- ample, the 300-m thick Triassic diabase Palisades Sill, on
come popular recently and will be introduced in Section 5. the eastern banks of the Hudson River, is commonly cited as
The majority of fractional crystallization models as- an example of a vertically differentiated sill with layers
sume that fractionation has taken place in a stationary formed by gravity settling. The overall composition of the
magma chamber at constant pressure. The rise of basaltic sill is tholeiitic basalt, as demonstrated by the upper and
magmas, as pointed out by O’Hara (1968b), may involve lower chill zones. The 10- to 20-m thick olivine-rich layer at
fairly continuous fractional crystallization as it rises, which the base is commonly attributed to differentiation by settling
must obviously be a polybaric fractionation process. One and accumulation of early-forming dense olivine crystals.
result is that the fractionating minerals vary as their stability Although vertical chemical trends in the sill are compatible
fields are crossed (e.g., garnet to spinel to plagioclase). with fractional crystallization of pyroxene and pyroxene ac-
Another is that the shift in the eutectic point with pressure cumulation zones occur near the bottom of the sill, the strik-
also causes the quantity of the liquidus phases that crystal- ing olivine layer is not compatible with the trends, and
lize to vary. In particular, the increase in the size of the field olivine is far too rare elsewhere in the sill to be consistent
with the concentration in the layer. The layer has recently
been reinterpreted as one of several late intrusions of magma
into the crystallizing tholeiitic liquid of the sill. This injected
pulse was olivine rich and dense, so it accumulated near the
base (Husch, 1990). Some magma series, such as the calc-
a alkaline series associated with subduction zones, may involve
mixing of components to a greater extent than fractional
crystallization trends.
Other cases against fractional crystallization were
b based on proportionality arguments. The great granite
batholith belts, for example, are thought to be too extensive
to have been created by fractional crystallization from a
basaltic parent. It would require approximately 20 parts of
c
original basalt to create 1 part late granitic liquid by frac-
tional crystallization. We need not walk long in places like
the Sierra Nevada, with so many square kilometers of
granitic rocks, to wonder where all of the basalt went! Mod-
ern theories that consider granite batholiths to be much thin-
d ner than originally thought (Hamilton and Myers, 1967, see
Chapter 4) reduce the magnitude of the problem, so we
could still follow Bowen (1948) and postulate that the lower
levels of the crust are composed of the denser fractionated
e gabbros. Seismic and gravity surveys, however, argue against
this possibility, making the origin of granite batholiths via
FIGURE 5 Increase in size and concentration of olivine fractionation from a basaltic parent untenable (Presnall,
phenocrysts toward the center of small dikes by flow
differentiation. Isle of Skye, Scotland. After Drever and
1979). These arguments serve only to demonstrate that frac-
Johnston (1958). Reproduced by permission of the Royal tional crystallization cannot lead to all of the magmatic
Society of Edinburgh. rocks now exposed at the surface of the Earth. It is still a

218
 Magma Diversity

common and important process, particularly in the early phase is released. The vapor phase may contain unusually
crystallization of mafic liquids, but there are other important high concentrations of volatile constituents such as H2O,
differentiation processes and other primary magmas. CO2, S, Cl, F, B, and P, as well as a wide range of incompatible
and chalcophile elements.
The volatile release and concentration associated with
2.2 Volatile Transport
pluton rise or resurgent boiling may momentarily increase
Chemical differentiation can also be accomplished when a the pressure at the top of the intrusion and fracture the roof
separate vapor phase coexists with a magma and liquid–vapor rocks in some shallow intrusions (it may also initiate vol-
fractionation takes place. A vapor phase may be introduced canic eruptions). Both the vapor phase and some of the late
in any of three principal ways. First, a fluid may be released silicate melt are likely to escape along a network of these
by heating of hydrated or carbonated wall rocks. We shall fractures as dikes of various sizes. The silicate melt com-
discuss some ramifications of this process in later sections monly crystallizes to a mixture of quartz and feldspar. It is
of this chapter. typically found in small dikes with a sugar-like texture,
Second, as a volatile-bearing but undersaturated magma which is informally called aplite. The vapor phase is typi-
rises and pressure is reduced, the magma may eventually be- cally concentrated as dikes or pods in, or adjacent to, the
come saturated in the vapor, and a free vapor phase is released. parental granitic pluton, where it crystallizes to form a char-
Because the vapor phase has a lower density than the melt, it acteristically magmagenic form of pegmatite.
rises, diffusing through the magma, and concentrates near the Although pegmatite is used as a textural classification
top of the magma chamber. Such concentrated fluid may even term for very coarse grain size, and there are other methods of
permeate into the roof rocks. This process usually involves an creating large crystals, the type of pegmatite described above
H2O-rich fluid, and it produces a variety of hydrothermal alter- is the most common. The large grain size in magmagenic peg-
ation effects. For example, the alkali metasomatism known as matites is not due to a slow cooling rate but is a result of poor
fenitization above nephelinite– carbonatite bodies has been at- nucleation and very high diffusivity in the H2O-rich phase,
tributed to alkali-rich fluids derived from the highly alkaline which permits chemical species to migrate readily and add to
intrusives. rapidly growing minerals. The size of crystals in pegmatites
A third mechanism for generating a separate fluid can occasionally be impressive, such as spodumene, micro-
phase is a result of late-stage fractional crystallization. Most cline, or mica crystals 6 to 10 m across. Most pegmatites are
early-formed igneous minerals are anhydrous (even hydrous “simple,” essentially very coarse granites. Others are more
minerals are less so than associated melts), so their segrega- complex, with a tremendous concentration of incompatible
tion from a hydrous melt enriches the melt in H2O and other elements and a highly varied mineralogy, commonly display-
volatile phases. Eventually the magma reaches the satura- ing a concentric zonation (Jahns and Burnham, 1969; C̆erný,
tion point, and a hydrous vapor phase is produced. This 1991; Simmons et al., 2003), as shown in Figure 6. Because
somewhat paradoxical “boiling off” of water as a magma the late fluid segregation concentrates several unusual ele-
cools has been called retrograde (or resurgent) boiling. ments, pegmatites are important economic resources and are
Of course, the three processes by which a vapor can be mined for Li, Be, the rare earths, W, Zr, and a host of others
produced need not be entirely separate, and all three may elements that are rarely concentrated in other environments.
contribute to saturation and volatile release from a magma, They are also a major source of gems.
depending upon the composition of the original magma, the Vapors that completely escape the magma and move to
rates of cooling and rise, the initial volatile content, the ex- higher levels may cool further and precipitate low-temperature
tent of fractional crystallization, the temperature, the nature minerals, such as sulfides in a hydrothermal system (commonly
of the wall rocks, etc. As a separate vapor is produced, the mixed in part with meteoric water).
chemical constituents in the system partition themselves be- Miarolitic pods, or cavities, are smaller fluid segre-
tween the liquid and vapor phases in appropriate equilib- gations trapped in the plutonic host. When finally exposed at
rium proportions, some remaining preferentially in the melt the surface, they are coarse mineral clusters (usually a few
and others becoming enriched in the vapor phase. The result centimeters across), the centers of which are typically hol-
is a silicate-saturated vapor phase in association with a low voids from which the fluid subsequently escaped. The
vapor-saturated silicate liquid phase. hollow cavities have euhedral crystals (of the same minerals
The cation sites in minerals are much more con- comprising the pluton) that extend inward, where they grew
strained and selective than in melts, so the chemical con- into the fluid, unimpeded by other minerals. Like complex
stituents in minerals are generally much simpler. As a result, pegmatites, some miarolitic cavities or pods have a concen-
the process of fractional crystallization tends to remove only tric structure consisting of layers of different mineralogy
a few elements from the liquid in significant quantities, and (Jahns and Burnham, 1969; McMillan, 1986.
a number of incompatible, LIL, and non-lithophile elements Because the addition of H2O lowers the melting point
become concentrated in the latest liquid fraction. Many of of magmas, the release of hydrous fluid into the country
these will further concentrate in the vapor, once formed. rocks causes the liquidus temperature in the main magma
This is particularly true in the case of resurgent boiling body to rise suddenly, resulting in rapid crystallization of
because the melt already is evolved by the time the vapor much of the liquid remaining with the previously formed

219
 Magma Diversity

FIGURE 6 Schematic sections of three zoned fluid-phase deposits (not at the same scale). (a) Miarolitic pod in granite (several
centimeters across). (b) Asymmetric zoned pegmatite dike with aplitic base (several tens of centimeters across). (c) Asymmetric zoned
pegmatite with granitoid outer portion (several meters across). From Jahns and Burnham (1969).

minerals. This is an alternative way of generating porphyritic There are two problems with applying the forsterite–
texture and is common in many silicic plutons. silica liquid immiscibility gap to natural magmas. First, the
temperature of liquid immiscibility is far too high (over
1700°C) to represent a reasonable crustal process. Of course,
2.3 Liquid Immiscibility
the Mg-Si-O system is rather restricted, leading one to ask
Two liquids that don’t mix seems an unlikely occurrence, whether the addition of other components, required to create
and it is. Yet most of us are familiar with salad oil and oil more natural magmas, would lower the temperature of the
slicks, so we have some concept of the phenomenon. Many solvus. The effect, however, of adding alkalis, alumina, etc. is
oils do not mix with water, and, because they are less dense, to eliminate the solvus completely. When this was experi-
the oil floats to the top of the water, forming a distinct layer. mentally demonstrated, liquid immiscibility was relegated to
Most immiscible phases, whether liquids or solids, homoge- the compost pile of magmatic processes.
nize at elevated temperatures due to the increased entropy Interest was renewed when Roedder (1951) discovered a
and molecular vibrational energy, although for oil–water at low-temperature immiscibility gap in the central portion of the
atmospheric pressure, the homogenizing temperature is fayalite–leucite–silica system (Figure 7) at temperatures and
above the boiling point of water. The solvus, representing compositions that are conceivable for some Fe-rich natural
liquid or solid immiscibility on a phase diagram, is therefore magmas. Roedder (1979) provided a review of liquid immis-
convex upward on a temperature–composition diagram. We cibility in silicate magmas, citing dozens of references in
have already encountered immiscible liquids in the which natural occurrences of immiscible liquids were de-
forsterite–silica system, where, on the high-silica side of the scribed, including a significant proportion of the lunar sam-
diagram, a highly silica-rich liquid separates from a less sil- ples returned by the Apollo program.
ica-rich one. Three natural magmatic systems are widely recognized
Throughout the 20th century, geologists appealed to liq- as having immiscible liquids in some portion of their compo-
uid immiscibility as a mechanism for magmatic differentia- sitional range. The first is the system mentioned above,
tion, thinking that it might be responsible for the separation of which most commonly translates to natural Fe-rich tholeiitic
a granitic liquid from an evolving system (presumably from an basalts, which experience an initial trend toward iron enrich-
initial basaltic parent). Such a separation into contrasting liq- ment. In the later stages of fractionation, a “granitic” melt
uid systems was also used to explain enigmatic cases of bi- ( 775% SiO2) separates from a basaltic melt ( ' 40% SiO2).
modal volcanism, such as the basalt–rhyolite occurrences of Once separated, the silicic liquid must have a much
the Snake River–Yellowstone area, or the Basin-and-Range of lower density than the Fe-rich mafic liquid, and we would
the southwestern United States. expect it to rise and collect near the top of the magma

220
 Magma Diversity

1200 Fa an iron–sulfide melt that is also rich in Cu, Ni, and other
1100 chalcophile elements. Small, round, immiscible sulfide
1000
droplets in a silicate glass matrix, similar to Philpotts’
SiO2 Fa (1982) granitic–tholeiitic examples above, have been ob-
served in a number of quenched ocean basalt glasses. Eco-
nomically important massive sulfide segregations in large,
layered mafic complexes have formed by separation and ac-
Two liquids
cumulation of immiscible sulfide melts.
A third liquid immiscibility gap occurs in highly alka-
line magmas that are rich in CO2. These liquids separate into
two fractions, one enriched in silica and alkalis and the other
in carbonate. These give rise to the nephelinite–carbonatite
association.
Although these are the three generally recognized oc-
currences of immiscible liquids, other magmas might sepa-
rate into two liquid phases under certain circumstances.
Lc Kfs SiO2 SiO2 These possibilities include lamprophyres (Philpotts, 1976;
Eby, 1980), komatiites, lunar mare, and various other vol-
FIGURE 7 Two immiscibility gaps in the system
fayalite–leucite–silica. The central one is of a composition, and
canics (see Roedder, 1979, for a summary).
at a low enough temperature (see the section in the upper left) The close spatial and temporal association of contrasting
to be attainable in some Fe-rich natural magmas (after Roedder, liquids may result from a number of processes in addition to liquid
1979, copyright © the Mineralogical Society of America). immiscibility. We can apply three tests to juxtaposed rocks to eval-
Projected into the simplified system are the compositions uate them as products of immiscible liquids. First, the magmas
of natural immiscible silicate pair droplets from interstitial
Fe-rich tholeiitic glasses (Philpotts, 1982).
must be immiscible when heated experimentally, or they must
plot on the boundaries of a known immiscibility gap, as in Figure
7. Second, immiscible liquids are in equilibrium with each
chamber. Crystallization of the magma must be advanced by other, and thus they must also be in equilibrium with the same
the time liquid separation occurs, however, and both liquids minerals. If the two associated liquids crystallized different miner-
are likely to become trapped in the already-formed crystal als or the same mineral with different compositions, they cannot
network. Philpotts (1982) described the textures of some Fe- be an immiscible pair. Finally, we may be able to use the pattern of
rich Hawaiian basalts in which small droplets of the two im- trace element fractionation between the two liquids to evaluate
miscible liquids are mingled in the interstitial glass trapped them as immiscible. Partitioning of minor and trace elements be-
between plagioclase and augite crystals. The separate droplet tween Fe-rich mafic liquids and granitic liquids, for example, can
compositions may be determined by microprobe and are pro- be distinctive when compared to the more common mafic mag-
jected into the Fa-Lc-silica system in Figure 7, along with mas with less Fe. Some incompatible elements (P, for example)
the liquid immiscibility gap of Roedder (1951). The actual are preferentially incorporated into an Fe-rich mafic liquid over
liquid compositions plot slightly outside the experimental the complimentary silicic one. A granitic rock relatively depleted
gap, probably because of the effects of Fe2O3, TiO2, and in these incompatible trace elements may be a product of liquid
P2O5, which expand the immiscible field. The low oxygen immiscibility. Of course, a low concentration in a particular trace
fugacity of the lunar basalts is the probable reason that im- element can also result if the liquid was derived from a similarly
miscible liquids are so common in them. depleted source. It is far more reliable, then, if rocks representing
Observing immiscible droplets is clear evidence of the both of the immiscible liquids can be evaluated. This has been ac-
process, but evidence is far less obvious that immiscible complished for some mixed dike rocks (Vogel and Wilband,
granitic liquids have separated and formed substantial segrega- 1978), but no one has yet succeeded in identifying a medium-
tions from Fe-rich tholeiites that are over 70% crystallized. Per- sized or larger granite as derived from an immiscible liquid.
haps filter pressing may aid the process, and the granophyric Although liquid immiscibility is now widely accepted
layers and lenses at the top of many mafic intrusions, including as a phenomenon in natural magmas, the extent of the
the Palisades Sill and the Skaergård intrusion (McBirney, 1975) process is still in question, and its importance in generating
may be the products of immiscible liquids. In such cases, liquid large bodies or a significant proportion of evolved magmatic
immiscibility is a late-stage addition to a more extensive rocks is doubtful.
process of fractional crystallization in these mafic intrusions.
Granitic bodies and other large-scale evolved liquids, however,
are unlikely products of immiscible liquids.
3 MAGMA MIXING
A second system displaying immiscible liquid behav-
ior is the separation of a sulfide-rich liquid from a sulfide- Magma mixing is a bit like liquid immiscibility in reverse,
saturated silicate magma. Less than one-tenth of a percent of and so was some of the reasoning behind its historical origins.
sulfur is sufficient to saturate a silicate magma and release The reigning paradigm of fractional crystallization implies

221