Available online at www.sciencedirect.

com

Geochimica et Cosmochimica Acta 73 (2009) 5778–5792

www.elsevier.com/locate/gca

Experimental constraints on Pt, Pd and Au partitioning

and fractionation in silicate melt–sulfide–oxide–aqueous fluid
systems at 800 °C, 150 MPa and variable sulfur fugacity
Aaron S. Bell a,*, Adam Simon a, Marcel Guillong b
a
Department of Geoscience and High Pressure Science and Engineering Center, University of Nevada Las Vegas, 4505 Maryland Pkwy,
Las Vegas, NV 89154-4010, USA
b
ETH Institute for Isotopengeologie/Mineral.Rohstoffe, Clausiusstrasse 25, Zürich CH 9082, Switzerland

Received 20 January 2009; accepted in revised form 25 June 2009; available online 19 July 2009

Abstract

We have performed experiments to constrain the effect of sulfur fugacity (fS2) and sulfide saturation on the fractionation
and partitioning behavior of Pt, Pd and Au in a silicate melt–sulfide crystal/melt–oxide–supercritical aqueous fluid
phase–Pt–Pd–Au system. Experiments were performed at 800 °C, 150 MPa, with oxygen fugacity (fO2) fixed at approximately
the nickel–nickel oxide buffer (NNO). Sulfur fugacity in the experiments was varied five orders of magnitude from approx-
imately log fS2 = 0 to log fS2 = 5 by using two different sulfide phase assemblages. Assemblage one consisted initially of
chalcopyrite plus pyrrhotite and assemblage two was loaded with chalcopyrite plus bornite. At run conditions pyrrhotite
transformed compositionally to monosulfide solid solution (mss), chalcopyrite to intermediate solid solution (iss), and in
assemblage two chalcopyrite and bornite formed a sulfide melt. Run-product silicate glass (i.e., quenched silicate melt) and
crystalline materials were analyzed by using both electron probe microanalysis and laser ablation inductively coupled plasma
mass spectrometry. The measured concentrations of Pt, Pd and Au in quenched silicate melt in runs with log fS2 values rang-
ing from approximately 0.0 to 5.0 do not exhibit any apparent dependence on fS2. The measured Pt, Pd and Au concentra-
tions in mss do vary as a function of fS2. The measured Pt, Pd and Au concentrations in iss do not appear dependent on fS2.
The data suggest that fS2, working in concert with fO2, via the determinant role that these variables play in controlling the
magmatic sulfide phase assemblage and the solubility of Pt, Pd and Au as lattice bound components in magmatic sulfide
phases, is a controlling factor on the budgets of Pt, Pd and Au during the evolution of magmatic systems.
Ó 2009 Elsevier Ltd. All rights reserved.

1. INTRODUCTION deposits. Magmatic differentiation processes play the criti-

cally instrumental role in amplifying, by orders of magni-
Quantitative constraints on the geochemical behavior of tude, the low abundance of most ore metals in crustal
precious and base metals in silicate melt–aqueous volatile reservoirs (cf., Rudnick and Gao, 2003) to ore-grade con-
phase(s) systems are intrinsically valuable in elucidating centrations. Understanding the character of these processes
the redistribution of metallic, siderophile and chalcophile is critical to the development of predictive models for ore
elements during the evolution of ore-forming magmatic sys- deposit formation. The partitioning of the platinum group
tems. Metallically fertile magmas are a necessary, albeit not elements (PGE) and base metals has been investigated in
sufficient, condition for the formation of magmatic ore natural and experimental assemblages in broadly mafic
systems, defined as silicate melt coexisting with immiscible
Fe–Ni–sulfide melt and/or crystals. Such studies have
* yielded both empirically estimated and experimentally
Corresponding author. Tel.: +1 702 895 3573.
E-mail address: [email protected] (A.S. Bell). determined partition coefficient values for the PGE and

0016-7037/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.gca.2009.06.037
 Experimental constraints on Pt, Pd and Au fractionation 5779

other metals between co-existing sulfide crystals and/or for the metals available for transport during sulfide frac-
sul=sil
liquid and silicate liquid, defined as DPGE . tionation and/or magmatic degassing at the level of ore de-
sul=sil
Published DPGE values for the entire group of platinoids posit formation.
range from 103 to 105 (Fleet et al., 1991; Peach et al., 1990;
Stone et al., 1990; Peach et al., 1994; Fleet et al., 1999), con- 2. PROCEDURES
sistent with the long recognized enrichment and association
of the PGE with sulfide phases in layered-mafic-intrusion- 2.1. Starting materials
hosted PGE–Ni–Cu deposits.
Significant effort, and concomitant controversy, has Starting materials utilized in this study were a synthetic
been aimed at quantifying the solubility and oxidation state Corning haplogranite glass (provided by Dr. David Lon-
of many noble and base metals in silicate melts. Experimen- don, University of Oklahoma), natural hexagonal pyrrho-
tal studies aimed at elucidating noble metal solubilities in tite (po), bornite (bn), and chalcopyrite (ccp; provided by
Fe-free, mafic-analog silicate melts have demonstrated con- Dr. George Harlow, American Museum of Natural His-
sistently that PGE concentrations at metal saturation vary tory), Calumet magnetite (mt; Calumet Skarn, Colorado),
proportionally with the oxygen fugacity (fO2) imposed and 10 wt.% NaCl eq. NaCl–KCl ± HCl aqueous solutions.
upon the system (Ertel et al., 1999). Laboratory studies The composition of the starting rhyolite glass is provided in
have demonstrated that Pt solubility in melts ranging from Table 1. The starting sulfide phases and mt were character-
basalt to rhyolite, excluding high Ca melts, is less than 1 ized by using electron probe microanalysis (EPMA) for ma-
lg/g at redox conditions (log fO2 6 NNO) approximating jor and most of the minor elements and laser ablation
those present in most natural magmatic systems (Borisov inductively coupled plasma mass spectrometry (LA-ICP-
and Palme, 1997; Ertel et al., 1999; Farges et al., 1999; MS) for Pt, Pd and Au. The concentration of these elements
Blaine et al., 2005). In addition to the dependence of is less than 1 lg/g in all phases. Small aliquots of Pt and Pd
PGE solubility on fO2, it is likely that PGE solubilities metal foil were added to serve as the Pt and Pd metal
are dependent on sulfur fugacity (fS2). Jugo et al. (1999) sources. Phase relations for the NaCl–KCl–H2O system
hypothesized that fS2 is important in controlling the solu- indicate that the aqueous fluid at run conditions exists as
bility of Cu and Au in sulfide phases during the evolution a one-phase supercritical aqueous fluid (Bodnar et al.,
of magmatic systems. While their experimental study was 1985; Chou et al., 1992).
performed at a single fO2 (NNO + 0.5) and fS2 (log
fS2 = 1) condition, they hypothesized that fS2 may con- 2.2. Experimental design
trol the partitioning of Au between silicate melt and sulfide,
in this case pyrrhotite (po), by one possible equilibrium Experiments were performed at 800 °C and 150 MPa in
such as René-41 cold seal pressure vessels. Rhyolite glass, sulfide
po sys po sys phases, mt and aqueous solution were loaded (Table 2) into
lmelt
AuO0:5 þ lFeS þ 0:5lS2 ¼ lAuFeS2 þ 0:25lO2 ð1Þ
Au capsules (5 mm OD  2.5 cm length with a 0.2 mm wall
where lji is the chemical potential of component i in thickness) and sealed with a carbon electrode arc welder.
phase j. Eq. (1) suggests that variation in fS2 and fO2, Three different assemblages were loaded as follows: Assem-
owing to processes such assimilation of country rock blage 1 initially contained cp + po + mt + glass + aqueous
(i.e., introduction of reduced and/or oxidized sulfur) fluid, the latter had an initial ratio of NaCl–KCl–HCl of
and degassing, and associated auto-oxidation of the mag- unity. Assemblage 2 contained cp + po + mt +
ma (cf., Candela, 1991), may play a determining role in glass + aqueous fluid, the latter had an initial NaCl–KCl
the metal enrichment process of the evolving magmatic ratio of unity and no HCl. Assemblage 3 contained
system. For example, assimilation of reduced country cp + bn + mt + glass + aqueous fluid, the latter had an ini-
rock will drive the magma toward lower fS2, hence favor- tial ratio of NaCl–KCl–HCl of unity. At run conditions cp
ing the mass transfer of ore metals from the silicate melt and po in assemblages 1 and 2 transformed to iss and mss;
into sulfide (e.g., Au into sulfide as written in Eq. (1)). assemblage 3 is interpreted to have contained a sulfide
Fractionation and subsequent removal of sulfide phases liquid formed from the reaction of bn and cp. Table 2
at depth during ascent of the magma will reduce the po-
tential for metals to be later scavenged and redistributed
by magma-evolved aqueous fluids. Table 1
In this study we investigated the role of fS2 on the solu- Haplogranite glass composition.
bilities of Pt, Pd, Au and Cu in an H2O-saturated rhyolite
Oxide constituent wt.%a
silicate melt, sulfide and oxide phase system at 800 °C and
150 MPa. Sulfur fugacity was controlled by using different SiO2 77.68
sulfide phase assemblages to impose fS2 that span the range TiO2 0.01
of fS2 values found in many natural systems. The results Al2O3 12.7
Fe2O3 0.28
demonstrate that changes in fS2 of an evolving sulfide-sat-
Na2O 4.49
urated magma shift the stable phase assemblage to an K2O 4.5
assemblage with a different capacity to sequester metals.
Such chemical changes affect the inherently finite metal Total 99.66
a
budget of the system and have important consequences Analyzed by ICP, EPMA and LA-ICP-MS.
5780 A.S. Bell et al. / Geochimica et Cosmochimica Acta 73 (2009) 5778–5792

Table 2
Summary of all experimental assemblages at run conditions. All starting solutions had a NaCl/KCl ratio of unity.
Assemblage Run nos. Phases at run conditions NaCl/ P T log10 log10
HCla (MPa) (°C) fO2 fS2b
One 12, 13, 15 iss + mss + mt + rhyolite melt + aqueous supercritical 1 150 800 NNO 0.19
fluid + Pt + Pd + Au (0.29)
Two 46, 47, 48, iss + mss + mt + rhyolite melt + aqueous supercritical 0 150 800 NNO 0.73
49, 50 fluid + Pt + Pd + Au (0.11)
Three 32, 33, 34 Intergrown bn and iss + mt + rhyolite melt + aqueous 1 150 800 NNO 5.16
supercritical fluid + Pt + Pd + Au (NA)
a
This is the starting NaCl/HCl ratio of the aqueous fluid.
b
Nominal sulfur fugacity values are reported as averages calculated with the Toulmin and Barton (1964) algorithm with 1r errors and
should be considered estimates for a given set of experimental runs. See text in Section 2.3 sulfur and oxygen fugacities for a discussion of
calculation and interpretation of the experimental meaning of these values.

provides the details of each assemblage. Water was used as Eq. (2). As a consequence of fixing fH2 and fS2, the fH2S
the pressure medium. The fO2 of each experiment was buf- of the experimental system must also be fixed by the
fered at approximately Ni–NiO by relying on the intrinsic equilibrium
buffering capacity of the René-41 pressure vessels (cf.,
H2 þ 1=2S2 ¼ H2 S: ð3Þ
Chou, 1987). The fO2 buffering capacity of the vessel has
been evaluated by using external Au capsules loaded with We hypothesize that the progressive sulfidation of the Pt
Ni and NiO to verify the stability of both phases over run metal chip, added as the Pt source, dynamically removed
times exceeding 1000 h. Temperature was monitored with sulfur from the magmatic volatile phase (mvp) over the
factory-calibrated Omega type K (Chromel–Alumel) ther- course of the run via the equilibrium
mocouples. Run pressure was monitored with a bourdon 
Pt þ 1=2Smvp
2 ¼ PtS: ð4Þ
tube pressure gauge which was calibrated against a fac-
tory-calibrated Heise gauge. Thermal gradients in the pres- EPMA-analysis of the Pt chips recovered from runs of
sure vessels are 5–7 °C over the length of the experimental different durations indicates that a sulfidation front propa-
capsule. Experimental run duration was varied from 67 to gated toward the center of the Pt chip. The sulfidation
672 h; run times are provided in Table 3. Experiments were front was observed as a rind of PtS which thickens pro-
terminated by a nearly isobaric quench in a stream of com- gressively with increasing run time. The subtly changing
pressed air from 800 to 200 °C and then immersed in an experimental fS2 conditions are reflected compositionally
ambient temperature water bath. Capsules were removed in the dissolved sulfur content of the quenched silicate melt
from the vessels, cleaned with lab-grade H2O, examined and also the metal content of the recovered mss. In light of
optically and weighed to determine if the capsules remained the relatively small range in fS2 values for a given assem-
sealed during the experiment. Only capsules that exhibited blage, with respect to the long experimental run durations
mechanical integrity, evinced a mass change of 60.3 mg, (a maximum difference of 1.0 log unit for run durations
and yielded a strong hiss when pierced with a hypodermic in excess of 600 h is observed), we postulate that the com-
syringe, evincing high internal capsule pressure, were pro- position of the experimental run products represents a
cessed for analysis. steady state (on the time scale of 2–3 days) or perhaps
an equilibrium state, reflective of the ‘‘instantaneous” fS2
2.3. Sulfur and oxygen fugacity of the system. Thus, the quenched-in compositions of the
experimental run products represent a snapshot of a sys-
Two different starting sulfide phase assemblages were tem continuously evolving toward its final equilibrium
used in an attempt to control fS2 at a fixed fO2. The starting fS2 condition defined by the Pt–PtS buffer. The sluggish
sulfide assemblages were po plus cp in assemblages 1 and 2, decline in fS2 values (approximately 0.04 log units day1)
and cp + bn in assemblage 3. Despite the interdependent coupled with the relatively high reactivity of the sulfide
nature of the equilibria controlling fO2 and fS2, it is rou- and silicate melt phases as observed in other studies at
tinely assumed that the vastly greater buffering capacity similar P–T conditions (e.g., Ballhaus and Ulmer, 1995;
of the Ni-rich pressure vessel fixes the fO2 (cf. Chou, Clemente et al., 2004) suggests that the fS2 of the system
1987) of the experimental assemblage via the equilibrium, at the time of the termination of the run, is accurately re-
in the case of po and mt, flected in the run products (i.e., the run products closely
maintained an equilibrium relationship with system fS2
6FeSpo þ 4Osys mt mvp
2 ¼ 2Fe3 O4 þ 3S2 ð2Þ
as this parameter slowly changed). While the experiments
In the absence of large quantities of sulfide, the fO2 im- did not reach a final ‘‘equilibrium” sulfidation state, the
posed by the vessel in conjunction with the water pressure relative differences in the experimental partitioning behav-
medium will fix the fH2 imposed on the experimental ior for Pt, Pd, Cu and Au place important constraints on
system, thus the vessel-controlled fH2 imposed on the the geochemical behavior of these metals in magmatic-
experimental system controls fS2 via the equilibrium in hydrothermal systems.
 Experimental constraints on Pt, Pd and Au fractionation 5781

2.4. Analytical procedures 2.4.2. LA-ICP-MS analyses

LA-ICP-MS analyses were performed at the Institute for
2.4.1. EPMA analyses Isotopengeologie and Mineral Rohstoffe at ETH in Zürich,
Recovered glass (i.e., quenched melt) and sulfides were Switzerland by using a homogenized prototype Excimer
analyzed by using wavelength dispersive spectrometry ArF laser (k = 193 nm) system, similar to the Geolas Sys-
(WDS) with a JEOLÒ JXA-8900 SuperProbe at the Univer- tem, coupled with a Perkin-Elmer Elan 6100 DRC quadru-
sity of Nevada, Las Vegas. Glasses were analyzed for major pole ICP mass spectrometer. To improve the limits of
elements using a 15 kV accelerating voltage and a 5 nA Far- detection for Pt, Pd and Au, a carrier gas mixture com-
aday cage current. A 10 lm de-focused beam was used in prised of a helium (1.15 L/min) and hydrogen (6 ml/min)
conjunction with the low beam current in order to minimize mixture was utilized (Guillong and Heinrich, 2007). The
current–flux–density-induced alkali diffusion (cf., Morgan diameter of the ablation crater was varied from 90 lm for
and London, 1996). quenched glasses and mt to 30 lm for sulfide run products;
The concentration of S in the quenched glasses was quan- ablations were carried out at an energy density of 25 J cm1
tified by using a 10 lm diameter beam and a 50 nA Faraday and a repetition rate of 10 Hz for silicate glasses and mt. To
cage current with counting times of 90 s on peak and 90 s on improve the transient signal from the sulfides, the energy
the background. The position of the S Ka peak was deter- density was decreased to 10 J cm2 and 3 Hz repetition rate.
mined with a PET spectrometer for individual aliquots of NBS-610 was utilized as a standard reference material for
experimental glass by using a statistical, automated peak all investigated phases. Additionally, a synthetic PGE-
search function in the JEOLÒ software. Natural ZnS was doped Cu–Fe sulfide standard (Wohlgemuth-Ueberwasser
used as a standard for the S analysis. The applied ZAF ma- et al., 2007) was utilized for Pt and Pd in some sulfide
trix correction procedure for the S analyses employed the phases. The polyatomic interference from 65Cu40Ar+ pro-
averaged major element composition of the glass as deter- duced in the plasma on the isotope 105Pd is problematic
mined by the previously described analytical conditions. for the analysis of Pd in Cu rich phases (i.e., iss). The
65
Calculated analysis-specific 3r limits of detection (LOD) Cu40Ar+ species production rate was quantified by ablat-
for S in the glasses range from 60 to 70 lg/g for the entire ing 99.999% Cu and measuring the resultant count rates on
suite of analyses. The low concentration of S in the experi- mass 65 and 105; CuAr+ interference was measured to be
mental glasses prevented us from using the peak shift in 0.05% of the counts measured on mass 65. Based on this,
the S Ka radiation, determined by EPMA, to estimate the the raw counts from 105Pd can be corrected for interfering
65
proportion of sulfide to sulfate species dissolved in the glass. Cu40Ar+ species if necessary. The data indicate that the
However, considering the relatively low oxidation state of use of NBS-610 and the synthetic PGE sulfide standards
the runs (NNO), we estimate that greater than 95% of to reduce and calculate Pd and Pt concentrations in sulfide
the sulfur present in the melt is dissolved as the reduced spe- phases yield statistically similar, albeit slightly different re-
cies S2 (cf., Carroll and Rutherford, 1988). Secondary stan- sults. As discussed below, both the EPMA- and LA-ICP-
dards VG-A99 and VG-2 were analyzed to evaluate the MS-determined concentrations of Pd and Pt are presented
performance of the S analytical routine. Using our analytical to allow a comparison of the use of the silicate glass stan-
routine, the concentration of S in standard glasses VG-2 and dard (i.e., NBS-610) vs. the synthetic PGE-doped standard
VG-A99 were quantified at 1300 ± 44 and 155 ± 9 lg/g, to quantify Pd, Pt and Au concentrations. Data processing
respectively. Previous studies that used EPMA to quantify and reduction of the raw LA-ICP-MS data were performed
the S content of VG-2 report 1403 ± 44 lg/g (O’Neill and by using SILLS (Guillong et al., 2008) using EPMA-deter-
Mavrogenes, 2002) and 1340 ± 80 lg/g (Dixon et al., mined element concentrations as the internal standard. Alu-
1991). The concentration of S in VG-2 has been reported minum was used as the internal standard for the silicate
as 1320 ± 50 lg/g when quantified by wet chemical methods glass analyses. Iron was utilized as the internal standard
(Wallace and Carmichael, 1992). The S content of VG-A99 for mt, mss and iss.
as determined by EPMA has been reported as 138 ± 13 lg/g
(Witter et al., 2005) and 170 ± 80 lg/g (Dixon et al., 1991). 3. RESULTS
The agreement between our EPMA-determined S values
and extant literature values suggests that the S concentra- 3.1. Silicate glass from assemblages 1, 2 and 3
tions reported herein are accurate.
Recovered Cu–Fe sulfide crystals were analyzed by 3.1.1. Major element and chlorine concentrations in the
using a 30 kV accelerating voltage and a 30 nA beam silicate glass
current. Iron, S, Ni and Cu were quantified by using Ka The silicate glass recovered from all runs is crystal free
X-ray emission lines, whereas Pt, Pd and Au were quanti- and homogenous within a given experiment with respect
fied by using La X-ray emission lines in order to mitigate to the concentrations of Si, K, Na, Al and Cl (Table 3).
spectral interferences observed in the M series lines for The concentration of Fe exhibits slight variability; however,
these latter elements. A beam diameter of 15 lm was used the range of Fe concentrations in silicate glass is consistent
in order to reintegrate the composition (presumably at with that reported in previous experimental studies at sim-
run conditions) of the fine scale (<1 lm) exsolution features ilar conditions (e.g., Simon et al., 2008). The concentration
of the recovered sulfide material. Quenched sulfide melts of S in some recovered glasses exhibits slight heterogeneity;
were analyzed with a broad 40 lm beam to obtain the aver- these values are attributed to uncertainty associated with
age composition of the sulfide melt at run conditions. analyzing low sulfur concentrations that are near the
5782 A.S. Bell et al. / Geochimica et Cosmochimica Acta 73 (2009) 5778–5792

detection limit. EPMA data indicate that all glasses are

Elemental composition of run-product experimental glasses. Major element concentrations were quantified by using EPMA. The concentrations of Pt, Pd, Au and Cu were quantified by using LA-

1.09
1.01
1.09
1.14
1.07

0.99

1.02
1.05
1.05

1.03

0.97
ASI
slightly peralkaline to peraluminous; the aluminum satura-
tion index (ASI), the ratio of Al/Na + K, ranges from 0.97
Total

91.52
91.88
91.64
91.53
92.72

91.97

91.98
92.69
92.83

91.77

91.62
to 1.14. The solubility of water in the melt, estimated by
using the EPMA difference method, ranges from 7 to
8.5 wt.%. This estimate is consistent with the expected
0.038 (0.01)
0.086 (0.04)

0.043 (0.02)

0.100 (0.02)

0.019 (0.01)
0.35 (0.30)
0.65 (0.49)

0.06 (0.03)

0.14 (0.11)

0.18 (0.13)
Au (lg/g)

water solubility of 6–8 wt.% in rhyolite melt at 800 °C

and 150 MPa (McMillan and Holloway, 1987).

NA
3.1.2. Cu, Au, Pt and Pd concentrations in the silicate glass

(0.046)
(0.027)
(0.05)
The major and trace element concentrations including

(0.01)

0.027 (0.01)
0.02 (0.01)
Pd (lg/g)

0.34 (.31)
0.16 (.09)
0.21 (.22)

Cu, Au, Pt and Pd in the recovered silicate glasses are pro-

vided in Table 3. The concentration of Cu in silicate glass
0.074

0.046
0.033
0.010

NA
NA

ranges from 5 to 22 lg/g and is homogeneously distributed

within glass beads recovered from all runs. These values are
0.893 (0.03)
0.266 (0.37)

0.046 (0.03)

0.027 (0.02)

consistent with the published Cu concentrations of 4 lg/g

1.03 (0.15)

0.17 (0.11)
0.01 (0.01)
1.2 (0.17)

1.1 (0.11)

0.6 (0.04)
Pt (lg/g)

(Lynton et al., 1993), 26 lg/g (Jugo et al., 1999) and 1

lg/g (Simon et al., 2006) for similar bulk assemblages at
NA

similar experimental conditions. In this study, melts reacted

with low sulfidation Cu-rich phase assemblages have the
Cu (lg/g)

(0.89)

(1.9)
(1.4)

highest Cu concentrations, whereas high sulfidation Cu-

13 (1.4)
13.6 (4)

22 (21)
6 (1.4)

(2)
(1)

poor phase assemblages generally yield lower Cu concentra-

8 (2)

NA
4.8
9.6
6.7
4.8
5.2

tions in rhyolite melt.

ICP-MS. ASI totals represent the molar ratio of Al2O3divided by the product of Na2O and K2O. Uncertainties are 1r.

equals one

The quantification of the concentrations of Au, Pt and

152 (39)
70.87 (0.49) 0.55 (0.05) 4.96 (0.17) 2.99 (0.10) 11.69 (0.17) 0.15 (0.15) 233 (48)
245 (44)

152 (49)
227 (49)
184 (34)
153 (14)

151 (19)
S (lg/g)

83 (27)
74 (17)
143 (9)

Pd in recovered silicate glass is complicated owing to the

Iss + po + magnetite + rhyolite melt + aqueous supercritical fluid + Pt + Pd + Au; no HCl added to starting solution

observed heterogeneous distribution of these metals in glass

(Fig. 1a and b). Several previous studies have discussed the
Iss + po + magnetite + rhyolite melt + aqueous supercritical fluid + Pt + Pd + Au; Na/K/H of starting solution

inherent difficulty in quantifying the solubility of noble met-

70.45 (0.52) 0.90 (0.43) 5.90 (0.10) 3.17 (0.09) 11.54 (0.22) 0.18 (0.04)
71.55 (0.60) 0.94 (0.05) 5.83 (0.11) 2.99 (0.09) 11.29 (0.18) 0.18 (0.02)
3.06 (0.08) 11.67 (0.21) 0.21 (0.03)
72.68 (0.30) 0.59 (0.07) 5.42 (0.17) 2.85 (0.09) 11.10 (0.21) 0.13 (0.01)
72.34 (0.26) 0.71 (0.04) 5.63 (0.11) 2.92 (0.08) 11.28 (0.27) 0.15 (0.01)

0.14 (0.03)
0.09 (0.02)
0.18 (0.10)
71.11 (0.75) 0.67 (0.10) 5.26 (0.21) 3.04 (0.19) 11.71 (0.21) 0.2 (0.02)

71.89 (0.75) 0.73 (0.03) 5.32 (0.21) 3.13 (0.06) 11.65 (0.28) 0.2 (0.01)

als in quenched silicate melts with oxidation states lower

than the mt–hematite buffer. The LA-ICP-MS spectra from
Cl

ablation of silicate glasses are routinely complicated by the

presence of spatially heterogeneous distributions of some
2.93 (0.34) 11.51 (0.18)
0.34 (0.04) 5.68 (0.14) 2.57 (0.32) 11.26 (0.20)
2.99 (0.15) 10.55 (0.6)

metals (e.g., Ag; Simon et al., 2008). This so called

‘‘nano-nugget effect” (cf., Borisov and Palme, 1997; Ertel
Al2O3

et al., 1999; Simon et al., 2007), as observed in LA-ICP-

MS spectra has been interpreted previously as either a
Iss + bn + magnetite + rhyolite melt + aqueous supercritical fluid + Pt + Pd + Au

quench-exsolution phenomenon (e.g., Cottrell and Walker,

2006) or as a product of redox reactions within the charge
Na2O

(e.g., Ertel et al., 1999). Stable PtO species within the melt
structure are reduced to elemental Pt in colloidal suspen-
sion within the melt (Ertel et al., 1999) as a function of
70.78 (0.41) 0.96 (0.05) 5.9 (0.10)

5.89 (0.2)

5.49 (0.3)

the reduction of the experiment oxidation state during

run up. Simon et al. (2008) suggest that initial noble metal
K2O

solubilities in the melt are relatively high, approaching the

10 lg/g range, over the first 12–24 h of a run before the
experimental charge comes to equilibrium with the intrinsic
0.52 (0.2)

1.91 (0.2)

redox state of the pressure vessel; hence, an excess of Pt is

FeO

initially dissolved in the melt. As the charge approaches os-

motic redox equilibrium, the fO2 of the metal-oversaturated
melt decreases. During the period of reduction the melt
71.08 (0.4)
72.46 (0.7)
71.09 (0.5)

gradually exsolves excess Pt, Pd and Au, forming nano-

nuggets, until the melt fO2 reaches the buffered value of
SiO2

NNO. Ertel et al. (1999) demonstrated that during a per-

iod of fO2 reduction, Pt and Rh solubilities decrease in the
melt, causing the melt to exsolve the excess metals and
NA, not analyzed.
Duration

forming nano-nuggets until the melt fO2 reaches its final

state. We specifically note the congruency of the Pd, Au
167
360

160
336
483
607
672

140
231
281
69

and to a much lesser extent Cu in LA-ICP-MS signals from

Table 3

the nugget bearing analyses (Fig. 1a). Such signals have the
Run

appearance that the nugget-forming metals are either

12
13
15

46
47
48
49
50

32
33
34
 Experimental constraints on Pt, Pd and Au fractionation 5783

alloyed or enjoy an intimate, mechanical association. Only xPtmetallic cPtmetallic 1=2

nugget free analyses of silicate glasses (Fig. 1b) have been K3 ¼  fO2 ð6Þ
xPtOmelt cPtOmelt
considered in this study, as these LA-ICP-MS signals are
assumed to approximate best the equilibrium solubility of where x and c are the mol fraction and activity coefficient,
the noble metals in the melt. respectively. The numerator of the expression may be writ-
The measured solubility of Au in silicate glass from indi- ten as:
vidual runs and among runs in the present study is variable xPt  cPt ¼ aPt ð7Þ
(see Table 3). The data in the present study indicate that the
best estimate for Au solubility in S-bearing rhyolite melt is where a is the activity of Pt in Eq. (5). The solubility of Pt,
on the order of a few tens to a few hundreds of ng/g (Table as expressed in Eq. (4), is a function of the activity of Pt in
3). Frank et al. (2002) reported that Au solubility at 800 °C the source phase, asource
Pt , (i.e., native metal, metallic alloy, or
and 100 MPa in a S-free Au-saturated rhyolite melt is 1 lg/g. sulfide) and the fO2 imposed on the system. The choice of
As discussed by Frank et al. (2002), it is possible that the PtO as the stable Pt species in the melt is based on the
quenched silicate glasses produced in their experiments con- strong positive correlation observed between Pt solubility
tained submicroscopic gold particles or regions of increased and fO2 reported by Ertel et al. (1999). Eq. (4) governs
Au concentration within the glass that formed during the solubility of Pt as a divalent oxide species within the sil-
quenching. Frank et al. (2002) discussed the nugget effect, icate melt structure.
with specific reference to Instrumental Neutron Activation The observed decrease in the measured abundances of Pt
Analysis (INAA) and Secondary Ion Mass Spectrometry and Pd in the quenched glasses (Fig. 2) with increasing run
(SIMS), and concluded that the best estimate of Au solubil- duration may reflect the decreasing activity of Pt and Pd in
ity in S-free rhyolite melt was on the order of 1 lg/g. Simon the experimental system as governed by the interaction of
et al. (2005) used nugget-free LA-ICP-MS analyses to quan- sulfur in the aqueous phase with alloys of the Pt and Pd
tify that the solubility of Au in S-free rhyolite is on the or- foils via the following sulfidation reaction:
der of 500 ng/g. Jugo et al. (1999) reported Au solubility in xPtsource þ ð1  xÞPdsource þ 1=2Svapor
2 ¼ ðPtx Pd1x ÞS ð8Þ
S-bearing Au-saturated rhyolite melt, saturated with iss and
The activity of Pt in the system may be further decreased
po at 850 °C and 100 MPa, to be on the order of 2–4 lg/g.
via the addition of Pd and Fe in solid solution on the PtS
Simon et al. (2007) report that the solubility of Au in
lattice. A similar relationship was also noted by Li et al.
S-bearing rhyolite, equilibrated with arsenopyrite (FeAsS),
(2003) for Ni in S bearing magmas. The asource is defined
is 0.6 lg/g. All of these experimental studies were con- Pt
by the equilibrium expression in Eq. (6) where:
ducted at broadly similar fS2, fO2, pressure and tempera-
!1=x
ture conditions. We suggest that the higher reported Au
aðPt;PdÞS
solubility values may reflect the nature of the analytical asource
Pt ¼ 1x
ð9Þ
technique. Jugo et al. (1999) and Frank et al. (2002) used ðametal
Pd Þ  ðf Svapor
2 Þ1=2  K5
the bulk analytical technique INAA which would incorpo- Substitution of the expression for Pt activity derived in Eq.
rate Au nuggets and include them in the analytical totals. (7) into the numerator of the equation governing Pt solubil-
The solubilities of Au in assemblages 1, 2 and 3 are ity as an oxide species (Eq. (4)) clearly shows that Pt solu-
0.26 ± 0.34, 0.11 ± 0.13 and 0.16 ± 0.03 lg/g, respectively; bility in the silicate melt phase in the investigated system is
the average solubility is 0.18 ± 0.07 lg/g. As stated above, not only proportional to the oxygen fugacity of the system,
the uncertainties do not reflect the presence of nuggets in but a complex function of the fS2 imposed on the experi-
the integrated portion of the signal. In light of the afore- mental system and the composition of the phase acting as
mentioned issues, we suggest that the data in the current the Pt source to the melt. The asource , as defined in Eq.
Pt
study constrain the solubility of Au in S-bearing rhyolite (7), via progressive reaction of elemental Pt with Pd, S
melt, at the PTX conditions in this study, to no more than and Fe, will decrease until the composition of the Pt–Pd–
a few 100 ng/g, consistent with previous studies. Fe–S overgrowths on the Pt metal come to equilibrium with
The concentrations of Pt and Pd in silicate glass are re- the system. Only then will the asysPt be fixed. Thus, it is pre-
ported in Table 3. A systematic decrease in the abundances dicted that Pt solubility in the melt in sulfide–oxide–silicate
of Pt and Pd in the quenched melt occurs as a function of melt systems should vary with changes in the fO2, fS2 and
increasing run duration for each experimental assemblage the asource of the experimental system. Platinum solubility
Pt
(Fig. 2). Decreasing metal concentrations may be a function in the experimental system is thusly controlled by the vari-
of the temporal evolution of the redox state of the system ables fS2 and fO2 where ametal of the source is a function of
Pt
(initially) and/or by the additional possibility of decreasing the fS2 imposed on the experimental system.
the activities of the soluble metals or metal alloys in the
experimental system via sulfidation and progressive alloy- 3.1.3. Mss composition
ing of elemental metallic species (i.e., the Pt and Pd metal Recovered mss varies remarkably little in the S to R
chips used as the source of Pt and Pd). Platinum dissolution Fe–Pd–Pt–Au–Cu ratio within a set of analyses for multiple
into the silicate melt structure may occur via the reaction grains from individual experiments. Mss from nearly all runs
proposed by Borisov and Palme (1997) contains measurable quantities of Ni, an element which is not
Ptmetallic þ 1=2Osystem ¼ PtOmelt ð5Þ present at measurable concentrations in any of the starting
2
materials. Nickel may be introduced via the diffusion of Ni
with the corresponding equilibrium constant written as: through the Au capsule from the pressure vessel via the water
5784 A.S. Bell et al. / Geochimica et Cosmochimica Acta 73 (2009) 5778–5792

Fig. 1. (a) Nano-nugget bearing LA-ICP-MS signal from quenched silicate melt. Note the congruency of Cu, Au and Pd signals; additionally
note the homogenous Fe signal. The homogenous nature of the Fe signal suggests that the nano-nuggets observed in the signal are not an
artifact induced by small scale crystalline sulfide inclusions, but are in fact nano-nuggets analogous to those observed in other experimental
studies. (b) Nano-nugget ‘‘free” LA-ICP-MS signal from quenched silicate melt. All metal signals display some small degree of heterogeneity,
however, these signals are comparatively homogenous with respect to the signal in Fig. 1a.

pressure medium and direct mechanical contact of po with function of temperature (Ballhaus and Ulmer, 1995). The
the Au metal capsule as observed in previous experimental presence of quench-exsolution texture in our experimental
studies (Scaillet and MacDonald, 2001). The composition mss is consistent with the down-temperature decrease in
of mss is provided in Table 4. The solubility of Au in mss in Pt and Pd solubility observed by Ballhaus and Ulmer
the current study is of 170 ± 71 lg/g, similar to the value of (1995). Mss at run conditions is expected to take on the
470 lg/g for Au solubility in mss reported by Jugo et al. hexagonal structure with a fully disordered Fe sublattice;
(1999) at similar pressure–temperature-composition however, upon termination of the run, ordering of the
(PTX) conditions; log fS2 values overlap statistically in both vacancies on the Fe sublattice and eventual conversion or
studies. partial conversion to a monoclinic form, is speculatively
Ubiquitous exsolution features (Fig. 3) less than 1 lm in responsible for the pervasive exsolution textures.
width were observed in most of the recovered mss crystals. Integration of the composition of the exsolution features
Although the exsolution features were too small to be with the composition of the matrix mss is required to accu-
individually resolved quantitatively with EMPA-WDS or rately quantify the concentration of these elements at run
LA-ICP-MS, the exsolution features have been qualita- conditions. We used a 30 lm beam diameter for both
tively characterized chemically by using energy dispersive EPMA and LA-ICP-MS analyses of mss. The data reported
X-ray spectrometry (EDS) and identified as predominately in Table 4 indicate that both methods effectively reintegrate
Pd–Cu–Ni-rich-sulfide relative to the host mss. Exsolution the mss and exsolution lamellae and yield similar metal
features observed in mss are likely the product of a rela- concentrations. The reported EPMA-determined concen-
tively slow quench rate; i.e., 2 min. The solubilities of Pt trations of Fe and Cu suggest that both metals are homog-
and Pd in mss have been demonstrated to be a strong enously distributed in the mss.
 Experimental constraints on Pt, Pd and Au fractionation 5785

10

Composition of recovered monosulfide solid solution. Fe, S, Cu and Ni quantified by EPMA. Pd, Pt and Au were quantified by using both EPMA and LA-ICP-MS. The LA-ICP-MS spectra

100.07 (0.7)
100.28 (0.5)
100.35 (0.3)
100.31 (0.5)
100.53 (0.5)
100.37 (0.2)
100.69 (0.3)
99.54 (0.7)
Pt 195
Pd 105

Pd (EPMA) Pd (sulfide) Pd (NBS-610) Pt (EPMA) Pt (sulfide) (lg/g) Pt (NBS-610) Au (NBS-610) Au (EPMA) Total
1 Au 197
log10 concentration (µg/g)

for Pd and Pt were reduced using both a synthetic Pt-, Pd-bearing sulfide standard and also NBS-610; denoted as EPMA and NBS-610, respectively. Uncertainties are 1r.
0.1

(lg/g)

<DL
<DL
<DL
<DL
<DL
<DL
<DL
<DL
0.01

9.1 (5.3)
286 (68)
134 (24)
131 (79)
(lg/g)
0.001

NA
0 100 200 300 400 500 600 700 800

108
136
230
Run Duration (Hours)

Fig. 2. The concentration of Pt, Pd and Au in the silicate melt as a

1789 (170)

(195)
function of run duration. The uncertainties on each datum are

(52)
(40)
(24)
13 (.79)
smaller than the symbol size.

(lg/g)

NA

NA
108
236
561
910
3.1.4. Iss composition
Intermediate solid solution was observed in all run prod-
ucts; however, the iss observed in runs 32, 33 and 34 is in-

470 (40)
24 (1.3)
ter-grown with bn and is thought to have exsolved from the

NA

NA
NA
NA
NA
NA
sulfide melt, discussed below, during quench. Iss from these
three runs is thus not reported as it was reintegrated into

0.18 (0.04)
0.22 (0.02)
0.33(0.18)
the sulfide melt composition. The measured compositions

(lg/g)
of iss reported in Table 5, is similar in all experiments.

<DL

<DL
<DL

<DL
0.06
The composition of run product iss deviates significantly
from the stoichiometry of the starting chalcopyrite. Total
S content of recovered iss appears to vary little from the

1.41 (0.02)
3.3 (0.18)

0.5 (0.13)

1.9 (0.08)
2.54 (0.1)
average value of 35.0 wt.%. Platinum and Pd concentra-

1.2 (0.1)
(lg/g)

tions in iss were generally near or below the EPMA limit NA

NA
of detection, whereas Au concentrations within individual
iss crystals in a given experiment exhibited variability. We
2.87 (0.15)

2.31 (0.06)

hypothesize that the observed heterogeneity of Au within

(lg/g)

recovered iss is a quench phenomenon; i.e., the presence

NA

NA
NA
NA
NA
NA
of Au lamellae in iss, observed by using back scattered elec-
tron (BSE) imaging, indicate exsolution of Au during
55.08 (0.87) 40.18 (0.38) 2.16 (0.71) 0.05 (0.007) 2.38 (0.50)
53.17 (0.23) 38.9 (0.51) 4.83 (0.77) 0.05 (0.004) 2.44 (1.73)
2.33 (0.69)
(0.42)
(0.26)
(0.56)
(0.25)

quench as reported by Jugo et al. (1999). The concentration

of Au in run-product iss in the current study ranges from
<DL
wt.%

0.90
1.31
1.18
1.79

0.37 to 0.72 wt.%, with an average on the order of

0.5 wt.%. This is similar, albeit lower than the experimen-
(0.05)
(0.17)
(0.07)
(0.02)

tally determined value (±1r) of 1.9 ± 0.4 wt.% for Au sol-

49.71 (1.6) 39.69 (0.36) 2.77 (NA) 5.3 (0.36)
Ni wt.%

ubility in mss reported by Jugo et al. (1999).

58.78 (0.46) 37.32 (0.23) 4.10 (0.61) <DL
0.14
0.43
0.51
0.32

3.1.5. Sulfide melt

4.14 (1.09)
5.01 (2.62)
5.41 (2.67)
5.10 (2.06)

The run products from charges loaded initially with cp

and bn contain a single sulfide entity that we interpret to
wt.%

be a quenched sulfide liquid (Fig. 4). The sulfide entity

Cu

encapsulates entire mt crystals and the boundary between

38.91 (0.55)

38.98 (0.48)
38.63 (0.7)
38.9 (0.22)

the two appears to be a wetting surface. The EPMA-deter-

S wt.%

mined composition of the sulfide melt phase is provided in

Table 6. This phase exhibits a complex intergrowth texture
between Cu-rich and Fe-rich sulfide phases. The average
55.08 (1.2)

53.42 (1.9)
53.31 (1.4)

composition provided in Table 6 reflects the inherent dif-

53.5 (1.1)
Run Fe wt.%

ficulty in reintegrating such a complex quench phase. Sul-

fide quench textures in our runs are consistent with the
Table 4

observed sulfide quench textures from other studies of sul-

fide melts (cf., Ripley et al., 2002). Recently revised phase
12
13
15
46
47
48
50
34
#

relations for the Cu–Fe–S–O system from Tsujimura and

5786 A.S. Bell et al. / Geochimica et Cosmochimica Acta 73 (2009) 5778–5792

Composition of recovered intermediate solid solution Fe, S, Cu and Ni quantified by EPMA. Pd, Pt and Au were quantified by using both EPMA and LA-ICP-MS. The LA-ICP-MS spectra for

100.24
100.51
100.29
100.17

100.02
Total

99.73
99.76
99.14

(0.2)
(0.2)
(0.1)
(0.7)
(0.1)
(0.4)
(0.4)

(0.4)
Au (EPMA)

1400 (1400)
2100 (2500)

1100 (2400)
3080 (490)
(lg/g)
Pd and Pt were reduced using both a synthetic Pt-, Pd-bearing sulfide standard and also NBS-610; denoted as EPMA and LA-ICP-MS, respectively. Uncertainties are 1r.

<DL
<DL

<DL

NA
Au (NBS-610)

6771 (1895)
4404 (1153)
3706 (1483)
7220 (1256)

5693 (1739)
5585 (14)

107 (55)
(lg/g)

NA
Pt (NBS-610)

1.99 (0.4)
0.62 (0.5)

4.3 (0.8)
7.1 (1.9)

862 (16)

62 (117)
19 (29)
(lg/g)

NA
Fig. 3. BSE image of fine scale exsolution textures in mss;
exsolution features are a Pd–Ni–Cu rich sulfide. The high Z#

Pt (sulfide)
phase in the lower left hand corner is metallic Au. The scale bar is
20 lm.

(lg/g)

NA
NA
NA
NA
NA
NA
NA

NA
Kitakaze (2004) place the melting temperature for the dry
iss-bornite system at less than 800 °C, the temperature of Pt (EPMA)
experiments in the current study. Experiments in this (lg/g)

<DL
<DL
<DL
<DL
<DL
<DL
<DL

<DL
study were performed at water-saturated conditions and
the effect of fH2O on sulfide melt–solid phase relations is
unknown. We did not attempt to quantify the oxygen con-
Pd (NBS-610)

tent of the sulfide melt phases owing to the relatively slow

172 (196)
1541 (94)
311 (387)

276 (183)
quench rate which likely promoted the extensive crystalli-

114 (49)
93 (46)
64 (76)
(lg/g)

zation of the sulfide melt phase which renders it unlikely

NA
that the original oxygen content of the sulfide liquid was
preserved.
Pd (sulfide)

3.1.6. Magnetite
(lg/g)

Magnetite was recovered in all successful experimental

NA
NA
NA
NA
NA
NA
NA

NA
runs. The concentrations of all analyzed metals (Au, Cu,
Pd and Pt) in mt are on the order of a few ng/g. EPMA
Pd (EPMA)

analyses of run-product mt indicate that the final composi-

tion is near stiochiometric end-member mt.
<DL
<DL
<DL
<DL
<DL
<DL
<DL

<DL
wt.%

3.2. Evaluation of experimental sulfur fugacity

Sulfur fugacity was calculated by using the measured S

(0.01)
(0.01)
<DL
<DL
<DL
<DL
<DL

<DL
wt.%

0.08
0.12

solubilities of the recovered experimental glasses following

Ni

Toulmin and Barton (1964) and the empirical model of Cle-

Cu wt.%

mente et al. (2004). In the Clemente et al. (2004) model, fS2

(0.38)
(0.17)
(0.40)
(0.71)
(0.07)
(0.04)
(1.01)

(0.33)
26.78
25.82
26.96
28.19
27.64
26.43
27.24

26.97

values are calculated relative to the metallic Fe-troilite sulf-

idation (FFS) buffer, utilizing an internally consistent fO2
corresponding to NNO + 0 for all of the experimental runs.
S wt.%

Eq. (6) was used to calculate fS2 values

(0.15)
(0.13)
(0.35)
(0.38)
(0.15)
(0.25)
(0.28)

(0.22)
35.45
34.83
34.82
35.13
34.74
34.98
35.28

35.2

log SðppmÞ- 0:001T þ 0:2567ðDNNOÞ

DFFS ¼ : ð10Þ
0:1713 þ 0:0034ðDNNOÞ
Fe wt.%

(0.35)
(0.15)
(0.15)
(0.11)
(0.19)
(0.91)

(0.44)
37.39
38.75
37.85
35.48
38.03
38.54
37.23

37.62
(1.3)

where D FFS is the log fS2 value defined relative to the FFS
buffer. Sulfur fugacity was also calculated by using the
Table 5

Fe(1x)S indicator method, formulated by Toulmin and

Run

50
49
48
47
46
15
13
12

Barton (1964) as
#
 Experimental constraints on Pt, Pd and Au fractionation 5787

log10 f S2 ¼ ð70:03  85:83xÞð1000=T  1Þ þ 39:30

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 1  :9989x  11:91 ð11Þ
where x represents the mole fraction of the FeS compo-
nent of mss. Palladium, Pt, Au, Cu and Ni have been as-
sumed to substitute ideally for Fe in the mss lattice on a
one to one basis (cf., Jugo et al., 1999). The effect of
these metals on the mss solution behavior is assumed
to be ideal given the low measured concentrations of
these metals in mss. Values for fS2 calculated by both
methods are reported in Table 7. The fS2 values from
experimental runs have been plotted relative to the liquid
sulfur condensation curve, Pt–PtS buffer, and the metallic
FFS buffer curve in Fig. 5. We suggest that the observed
discrepancies between fS2 values calculated from the two
methods may reflect the kinetically sluggish nature of S
dissolution into the melt (i.e., minor disequilibrium with
respect to S solubility in the melt phase). Additionally,
in runs 33 and 34, some individual EPMA analyses of
S in the glass were at or below the detection limit. The Fig. 4. BSE image of a sulfide entity interpreted to be a re-
average sulfur concentration for these runs included only crystallized sulfide liquid. Dark and light grey intergrowths are iss
values that were above the EPMA limit of detection; and bn. The scale bar is 100 lm. The dark grey crystal at lower left
is magnetite which coexisted with the sulfide liquid at run
thus, it is likely that fS2 values calculated from the
conditions.
EPMA averages for these runs are systematically biased
toward higher values. We suggest that using the mss
composition to calculate the fS2 of a system is a more ro- 4. DISCUSSION
bust method than using the sulfur content of the silicate
melt owing to the much higher degree of complexity in 4.1. Fractionation of Pt, Pd and Au in nature: the effect of
the solution behavior of the latter. Additionally the mss fO2 and fS2
composition appears to be temporally more responsive
to changing fS2 conditions. The new data place constraints on the relative fraction-
Sulfur fugacity values calculated from both methods ation effects induced by the precipitation of mss, iss and
appear to converge in experimental runs of intermediate sulfide–oxide melts on the abundances of Pt, Pd and Au
duration. This reflects the attainment of a steady state in silicate magma. Silicate melts can reach sulfide saturation
with respect to both S solubility in the silicate melt and owing to magmatic fractionation or the addition of S
the metal composition of the recovered mss. We interpret sourced from outside of the magma (e.g., assimilation of
this decrease in calculated fS2 values to be a function of biogenic pyrite from a sedimentary sequence). Both
the progressive sulfidation of the Pt and Pd metal loaded processes result in an increase in the aS2 in the melt, hence
initially as a metal source for the experiments. The reac- promoting the saturation of the silicate melt with a sulfide
tion of the finite reservoir of sulfur within the experimen- phase. It is the fO2–fS2 condition at which sulfide saturation
tal charge with the metal sources (i.e., the Pt and Pd foil occurs that ultimately controls which sulfide phase will be
chips) should result in depletion of the sulfur within the stable (cf., Jugo et al., 1999), and how it will affect the
supercritical aqueous fluid phase with increasing progress Pd–Pt–Au budget of the evolving magmatic system prior
of favorable Pt and Pd sulfidation reactions. Evidence for to the exsolution of an aqueous magmatic volatile phase.
the progressive removal of S from the supercritical aque- Here we discuss the effects of mss, iss and sulfide melt on
ous fluid phase was observed in the form of the develop- modifying the metal budget of magmatic systems.
ment of a PtS ‘‘rind” on the recovered pieces of Pt metal
foil. Evidence for the removal of S from the supercritical 4.1.1. The effect of mss saturation
aqueous fluid phase is also observed in the systematic de- The data in this study suggest that saturation of a mag-
crease of the calculated fS2 values for the runs. Analo- matic system with mss should strongly fractionate Pd from
gous Cu–sulfide overgrowths have been observed Pt, and both of these PGE from Au. In addition to fraction-
encapsulating molten Cu metal blebs in a study of Cu ation effects, the ability for mss to scavenge Pt and Pd from
solubility in mafic silicate melts (Holzheid and Lodders, a silicate melt is dependent upon the fS2 of the system; i.e.,
2001). The advancing sulfidation front (Figs. 6 and 7) higher fS2 yields higher Pt and Pd solubilities in mss. This
is truncated at the interface with elemental Pt. This ele- relationship is consistent with the results of Ballhaus and
mental Pt forms the core of the recovered foil chip. No Ulmer (1995); however, the absolute concentrations of both
elemental Pd was recovered in any of the runs indicating Pt and Pd in mss in this study are lower than the values re-
that all Pd present was either lattice bound in mss and ported by Ballhaus and Ulmer (1994) for experiments per-
Cu–Fe sulfide phases, alloyed with Au and Pt, or bound formed at identical temperatures similar fS2 values.
in separate, ubiquitous PdS phases. Discrepancies between the solubility data in the two studies
5788 A.S. Bell et al. / Geochimica et Cosmochimica Acta 73 (2009) 5778–5792

Table 6
Individual EPMA point analyses of quenched sulfide liquid for Run 32, 33 and 34. Note the heterogeneous distribution of Au. Each point
analysis was performed using a broad, 40 lm beam in an attempt to reintegrate the composition of the sulfide liquid to that of run conditions.
Fe S Ni Pd Pt Au Cu Total
Run 32 quenched sulfide
20.6 30.21 0.09 <DL <DL 0.04 49.3 100.31
18.4 29.05 0.08 0.04 <DL 1.5 51.38 100.5
22.27 30.95 0.13 0.12 <DL 6.41 41.63 101.54
18.62 29.28 0.05 <DL <DL 0.07 53.02 101.06
21.76 30.63 0.1 <DL <DL 1.33 45.81 99.64
20.97 30.24 0.06 <DL <DL 0.15 48.63 100.09
22.44 31.1 0.12 0.13 <DL 4.71 42.77 101.28
19.87 29.7 0.15 0.02 <DL 0.92 48.52 99.2
23.53 30.81 0.11 0.02 <DL 1.49 42.98 98.97
18.92 28.64 0.07 0.05 <DL 2.44 50.53 100.67
Mean 20.74 30.06 0.10 0.06 NA 1.91 47.46 100.33
1r 1.77 0.86 0.03 0.05 NA 2.11 3.95 0.86
Run 33 quenched sulfide liquid
26.96 30.8 <DL <DL <DL <DL 41.81 99.6
26.44 30.29 <DL 0.05 <DL <DL 43.2 100.01
26.61 30.25 <DL <DL <DL 0.02 43.13 100.05
27.59 31.86 <DL 0.02 <DL 0.16 41.12 100.78
26.44 30.85 <DL <DL <DL 0.03 42.83 100.19
26.91 30.93 <DL 0.04 <DL 0.57 42.24 100.72
26.43 29.83 <DL 0.05 <DL 0.09 44.41 100.85
26.58 30.09 <DL 0.04 <DL 0.1 44.34 101.18
Mean 26.75 30.61 NA 0.04 NA 0.16 42.89 100.42
1r 0.40 0.64 NA 0.01 NA 0.21 1.15 0.54
Run 34 quenched sulfide liquid
23.56 28.65 <DL 0.02 <DL 0.17 47.84 100.26
33.03 30.73 <DL 0.3 <DL 0.41 35.13 99.62
35.87 31.46 <DL 0.24 <DL 0.19 30.72 98.51
19.44 27.41 <DL <DL <DL 0.07 52.02 98.96
18.33 26.57 <DL 0.1 <DL <DL 53.48 98.5
17.95 26.66 <DL <DL <DL 0.14 54.3 99.05
28.75 29.63 <DL <DL <DL 0.06 42.27 100.72
32.33 30.41 <DL 0.38 <DL 0.05 37.95 101.13
19.81 27.34 <DL 0.02 <DL <DL 53.08 100.26
22.64 28.22 <DL <DL <DL 0.17 49.67 100.73
18.17 26.99 <DL <DL <DL 0.11 54.31 99.59
Mean 24.53 28.55 NA 0.18 NA 0.15 46.43 99.76
1r 6.74 1.75 NA 0.15 NA 0.11 8.52 0.93

Table 7
may be attributed to the more complex nature of the solid
Sulfur fugacity of experimental run. Uncertainties are 1r.
solutions utilized in the current study. Ballhaus and Ulmer
(1994) equilibrated mss with either Pt or Pd, but never both Run Duration S in glass log fS2b log fS2c
simultaneously; additionally, their study did not include # (h) (lg/g) (bar) (bar)
additional competing metals. The current study has utilized 12 69 152 (39) 1.05 0.07
assemblages containing a more complex metal assemblage, 13 167 233 (48) 0.02 0.52
thus, we hypothesize that more complex assemblages may 15 360 245 (24) 0.15 0.14
be the cause of the variability in measured solubility values. 46 160 152 (49) 1.05 0.92
47 336 227 (49) 0.03 0.87
The presence of Cu at several wt.% and Ni at slightly less
48 483 185 (34) 0.56 0.86
than 1 wt.% may inhibit the solubility of Pt and Pd on 49 607 153 (14) 1.02 0.90
the Fe sublattice of the mss lattice. Future work is planned 50 672 143 (10) 1.19 0.76
to quantify the crystallographic residency of metals in mss 32 140 151 (19) 1.06 NRa
solid solutions. 33 231 87 (20) 3.21 NRa
To calculate Nernst-type partition coefficients for Pt, 34 281.5 73 (17) 3.64 5.16
Pd and Au between rhyolite melt and mss, taking into a
NR indicates that mss was not recovered from the charge.
consideration the variation in values determined in this b
Sulfur fugacity calculated by using the Clemente et al. (2004)
study, we used lower and upper values for each metal empirical algorithm.
as a limiting case. Using a lower and upper solubility va- c
Sulfur fugacity calculated by using the Toulmin and Barton
lue of 0.1 and 0.5 lg/g for Pt, Pd and Au in the rhyolite (1964) algorithm.
 Experimental constraints on Pt, Pd and Au fractionation 5789

Fig. 5. Plot of fS2 buffer curves relevant to calculated experimental

sulfur fugacity values. Experimental fS2 values are schematically
represented by the stippled ellipse. Buffer curves modified after
Fleet et al. (1999).
Fig. 7. Thin sulfidation rim (dark gray rimming the light grey) on a
Pt foil chip in run of 67 h.
melt yields apparent Nernst-type partition coefficients
that range from a low of Dmss=melt
Pd = 4.4  104 to a high
mss=melt
of DPd = 2.2  10 . Similarly for Pt, Dmss=melt
5
Pt ranges
from a low of 6  103 to a high of 3  104. These calcu- with those of Ballhaus and Ulmer (1994) suggests that
lated Dmss=melt values indicate that crystallization of mss the calculated partition coefficient values may increase
leads to significant fractionation of Pt from Pd during sil- with decreasing Cu content of the mss and additionally,
icate melt solidification. Thus, during mss fractionation, increase with increasing fS2 of the system.
the absolute concentration of Pd in the silicate melt The absolute magnitude of the depletion of Pt and Pd
may decrease by a factor 10 relative to the concentration from the silicate melt phase ultimately depends on the fS2
of Pt in the silicate melt. Gold is the most refractory of of the magmatic system, as the stability field for mss at
the three metals with measured concentrations in the sil- magmatic conditions spans nearly 10 orders of magni-
icate melt ranging from 0.35 to 0.01 lg/g with an average tude. Thus, it is conceivable that a silicate melt could
on the order of 0.1 lg/g. These Au solubility values yield precipitate mss, yet retain significantly less Pt and Pd
a minimum Dmss=melt = 2.3  103 and a maximum than are soluble in mss at much higher fS2, significantly
Au
mss=melt
DAu 4
= 1  10 . A comparison of these new results lowering the apparent partition coefficient values. Simi-
larly, mss saturation may be inhibited entirely given the
appropriate fO2, fS2 and fH2O conditions. Hydrous sili-
cate melts containing dissolved S at a fixed fO2 and
fH2O will posses a defined and fixed fH2S, given the
appropriate low magmatic fO2. Sulfidation reactions
involving ferrous iron dissolved in the silicate melt or
the transformation of mt to mss may be described by
the following equilibria:
FeOmelt þ H2 S ¼ FeSmss þ H2 O ð12Þ
mss
1=3Fe3 O4 þ H2 S ¼ FeS þ H2 O þ 1=6O2 ð13Þ
Assuming the mss in Eqs. (8) and (9) and mt in Eq. (9)
are pure phases, these reactions work in concert with the
iron activity in the melt, fS2 and fO2 to buffer the fugacity
of the gaseous species H2S and H2O in the melt.
fH2 O 1
K8 ¼  ðcmelt
FeO Þ ð14Þ
Xmelt
FeO  fH 2 S

where X melt
FeO is the mole fraction of dissolved FeO in the sil-
icate melt, fH2 O is the fugacity of water, fH2 S is the fugacity
of H2S, and cmelt
FeO is the activity of FeO in the melt. The fH2 O
Fig. 6. An advancing sulfidation front in Pt foil chip in run of is related to fO2 and fH2 S via the equilibrium:
360 h. Outer rim becomes (Pt, Pd, Fe) S in composition, where as
the inner portion of the sulfidation front is nearly pure PtS grading
f H2 O  f O1=6
2
K9 ¼ ð15Þ
into metallic Pt. f H2 S
5790 A.S. Bell et al. / Geochimica et Cosmochimica Acta 73 (2009) 5778–5792

If the intrinsic values of fH2O and fH2S fixed by the 5. APPLICATION TO NATURAL SYSTEMS
water content, sulfur content and oxidation state of the
melt are out of equilibrium with the buffered values re- The observed fractionation trends induced by the
quired by Eqs. (8) and (9), mss will be either resorbed crystallization of mss with respect to the metal content
and consumed until it is exhausted, or crystallize until the of the residual melt are not in agreement with the high
ratio of fH2S and fH2O of the silicate melt reach the equi- Pt/Pd ratios observed in some porphyry ore systems
librium buffered values. Alternatively in sulfide unsaturated (Auge et al., 2005). The discord between Pt/Pd ratios ob-
systems, high fH2O values may impede entirely the crystal- served in PGE-rich porphyry ore deposits (Werle et al.,
lization of a sulfide phase given that the melt is sufficiently 1984; Mutschler et al., 1985; Hulbert et al., 1988), Mulja
oxidized (i.e., at oxidation states where H2S formation is and Mitchell, 1991; Cassidy et al., 1996; Ohnenstetter
limited, as reduced melts with high fH2O will have a high and Watkinson, 1998) and the Pt/Pd ratios caused by
propensity to crystallize an oxide phase). The ‘‘instanta- sulfide saturation, as constrained in this study, suggest
neous” water fugacity of the melt at any point during mag- that magma oxidation and subsequent sulfide resorption
matic evolution (prior to vapor saturation by second and/or chemical exchange of metallic species between
boiling) exerts a first order control on the stability and crystalline magmatic sulfide material and an evolved
modal abundance of crystalline mss present in the system. aqueous magmatic fluid play dominant roles in the deter-
The reduction in amssFeS by the incorporation of additional mination of the final Pt/Pd ratio of the ore system. Fur-
metal components (i.e., Cu, Ni and Co) into the mss lattice thermore, sulfide saturation early in the evolution of a
will shift the pure end member equilibrium dictated by Eqs. rhyolite silicate magma and the subsequent removal of
(8) and (9) in a direction that represses mt crystallization. the sulfide phase via Rayleigh fractionation may reduce
The model equilibria discussed above indicate that early the concentrations of Pt and Pd in the silicate melt to
saturation of a silicate melt with mss (given an appropri- levels too low for the PGE to be effectively scavenged
ately high fS2), provided that the mss is removed from the and concentrated by a magmatic volatile phase in por-
system via a Rayleigh type fraction process, may effectively phyry ore-forming systems.
reduce the PGE budget of the system levels that preclude
the development of a PGE rich porphyry system. However, 6. CONCLUDING REMARKS
mss saturation at low fS2 values may be possible given a
high amelt
FeO and appropriate fO2, fH2O and fH2S. Under such Our experimental results constrain fractionation trends
low fS2 conditions, the ability of mss to deplete the melt of amongst Pd, Pt and Au in rhyolite magmatic systems.
Pt and Pd should be drastically reduced relative to high fS2 The data suggest that the timing of sulfide saturation and
conditions. Additionally, at low fS2 conditions conducive to the nature of magmatic sulfide phase stability may control
the saturation of a magmatic sulfide phase the modal abun- the ultimate abundances and ratios of Pd, Pt and Au trans-
dance of crystalline mss should be extremely low due to the ported by magmatic hydrothermal fluids to porphyry-type
low concentrations of S acting as a limiting reagent. ore forming systems. Early crystallization of a magmatic
sulfide phase in an evolving rhyolite melt system (and the
4.1.2. The effect of iss saturation removal of this phase from further chemical interaction
Crystallization of iss also appears to fractionate Pt, Pd with the rest of the magmatic system) may effectively stifle
and Au in the silicate melt. The effect of iss crystallization volatile-phase transport of the PGE and their enrichment
is to fractionate Au and Pd from Pt in the silicate melt with in the magmatic-hydrothermal environment with metal ra-
Pt being retained most strongly in the melt phase. Gold and tios consistent with those observed in nature. Magmatic sul-
Pd concentrations in iss range from 1000 to 5000 lg/g, fide phases may, however, play an important temporal role
whereas the concentration of Pt in iss is consistently below in liberating sequestered Pt and Pd via oxidation and res-
the EPMA limit of detection. It is interesting to note the orbtion into the melt or via direct sulfide–aqueous fluid
similarity of the geochemical behavior of Pd and Au in chemical interaction late in the magmatic-hydrothermal
iss when contrasted with the distinctly, dissimilar behavior transition phase.
of these metals in recovered mss. There appears to be no
systematic variation of the Pt, Pd or Au solubility in iss ACKNOWLEDGMENTS
with variation in either fS2 or the Fe/Cu ratio of the recov-
ered iss. Gold concentrations reported in this study are sim- This work was supported by the National Science Founda-
ilar to, albeit larger, than those reported in Simon et al. tion NSF EAR 0609550. We thank Ed Mathez, Chusi Li
and an anonymous reviewer for comments that significantly
(2000). The fractionation trend of Pt and Pd, induced by
improved the manuscript, both substantively and stylistically;
the stability of magmatic iss, mirrors the mss fractionation also James Brenan and Thomas Pettke are thanked for informal
trend where the concentration of Pt in the silicate melt ex- reviews. We thank Wohlgemuth-Ueberwasser and Chris Ballhuas
ceeds that of Pd. The relatively compatible behavior of Au for providing sulfide standards for LA-ICP-MS analyses, Dave
in iss differs distinctly from that observed in mss at the same London for providing the starting glass, and George Harlow
experimental conditions. The compatible behavior of Au in for the starting bn, cp and po. AB thanks the Society of
the iss lattice should result in an inverse fractionation effect Economic Geology, Geological Association of Canada/Mineral-
with respect to the ratio of Au to Pt and Pd in the residual ogical Association of Canada, the UNLV Department of
silicate melt. Such a situation should strip Au from the melt Geoscience and the UNLV Graduate and Professional Student
relative to Pd and Pt. Association.
 Experimental constraints on Pt, Pd and Au fractionation 5791

REFERENCES gold with application to magmatic sulfide-PGE deposits. Lithos

47, 127–142.
Auge T. M., Petrunov R. and Bailly L. (2005) On the origin of the Frank M. R., Candela P. A., Piccoli P. M. and Glascock M. D.
PGE mineralization in the elatsite porphyry Cu–Au deposit, (2002) Gold solubility, speciation, and partitioning as a
Bulgaria: comparison with the Baula–Nuasahi complex, India, function of HCl in the brine-silicate melt-metallic gold system
and other alkaline PGE-rich porphyries. Can. Mineral. 43, at 800 °C and 100 MPa. Geochim. Cosmochim. Acta 66, 3719–
1352–1375. 3732.
Ballhaus C. and Ulmer P. (1995) Platinum-group elements in the Guillong M., Meier D. M., Allan M. M., Heinrich C. A. and
Merensky reef: II. Experimental solubilities of platinum and Yardley B. (2008) SILLS: A MATLAB-Based Program for the
palladium in Fe1-xS from 950 to 450 °C under controlled fS2 Reduction of Laser Ablation ICP–MS Data of Homogeneous
and fH2. Geochim. Cosmochim. Acta 59, 4881–4888. Materials and Inclusions: Mineralogical Association of Canada
Blaine F. A., Linnen R. L., Holtz F. and Brügmann G. E. (2005) Short Course 40. Vancouver, B.C., pp. 328–333.
Platinum solubility in a haplobasaltic melt at 1250 °C and Guillong M. and Heinrich C. A. (2007) Sensitivity enhancement in
0.2 GPa: the effect of water content and oxygen fugacity. laser ablation ICP-MS using small amounts of hydrogen in the
Geochim. Cosmochim. Acta 69, 1265–1273. carrier gas. J. Anal. Atomic Spectrom. 22, 1488–1494.
Bodnar R. J., Burnham C. W. and Sterner S. M. (1985) Synthetic Holzheid A. and Lodders K. (2001) Solubility of copper in silicate
fluid inclusions in natural quartz. III. Determination of phase melts as function of oxygen and sulfur fugacity, temperature,
equilibrium properties in the system H2O–NaCl to 1000 °C and and silicate composition. Geochim. Cosmochim. Acta 65, 1933–
1500 bars. Geochim. Cosmochim. Acta 49, 1861–1873. 1951.
Borisov A. and Palme H. (1997) Experimental determination of the Hulbert L. J., Duke J. M., Eckstrand O. R., Lydon J. W. and
solubility of platinum in silicate melts. Geochim. Cosmochim. Scoates R. F. J. (1988) Geological environments of the
Acta 61, 4349–4357. platinum group elements. Geol. Surv. Canada Open File 1440,
Candela P. A. (1991) Physics of aqueous phase exsolution in 148.
plutonic environments. American Mineralogist 76, 1081–1091. Jugo P. J., Candela P. A. and Piccoli P. M. (1999) Magmatic
Carroll M. R. and Rutherford M. J. (1988) Sulfur speciation in sulfides and Au:Cu ratios in porphyry deposits: an experimental
hydrous experimental glasses of varying oxidation states. study of copper and gold partitioning at 850 °C, 100 MPa in a
Results from measured wavelength shifts of sulfur X-rays. haplogranitic melt-po-intermediate solid solution-gold metal
Am. Mineral. 73, 845–849. assemblage, at gas saturation. Lithos 46, 573–589.
Cassidy K. F., Lang, Jr., Lueck, B. A., Mortensen J. K., Russel J. Li C., Ripley E. M. and Mathez E. A. (2003) The effect of S on the
K., Thompson J. F. H. (1996). Geochemical and isotopic partitioning of Ni between olivine and silicate melt in MORB.
characteristics of Mesozoic alkalic intrusions in the Cordillera Chemical Geol. 201, 295–306.
nf British Columbia: tectonic significance and implications for Lynton S. J., Candela P. A. and Piccoli P. M. (1993) An
Cu–Au metallogeny. Geol. Soc. Austral., 13th Australian experimental study of the partitioning of copper between
Geological Convection, Canberra, February, 1996. #41 (abstr.). pyrrhotite and a high silica rhyolitic melt. Econ. Geol. 88,
Chou I.-M. (1987). Oxygen buffer hydrogen sensor techniques at 901–915.
elevated pressures and temperatures. In Hydrothermal Exper- McMillan P. and Holloway J. (1987) Water solubility in alumino-
imental Techniques (eds. G. C. Ulmer and H. L. Barnes), pp. silicate melts. Contrib. Mineral. Petrol. 97, 320–332.
61–99. Morgan G. B. and London, VI, D. (1996) Optimizing the electron
Chou I.-M., Sterner S. M. and Pitzer K. S. (1992) Phase relations in microprobe analysis of hydrous alkali aluminosilicate glasses.
the system NaCl KCl–H2O: IV. Differential thermal analysis of Am. Mineral. 81, 1176–1185.
the sylvite liquidus in the KCl–H2O binary, the liquidus in the Mulja T. and Mitchell R. H. (1991) The Geordie Lake intrusion,
NaCl–KCl–H2O ternary, and the solidus in the NaCl–KCl Coldwell complex, Ontario: a palladium- and tellurium-rich
binary to 2 kb pressure, and a summary of experimental data disseminated sulphide occurrence derived from an evolved
for thermodynamic-PTX analysis of solid–liquid equilibria at tholeiitic magma. Econ. Geol. 86, 1050–1069.
elevated P–T conditions. Geochim. Cosmochim. Acta 56, 2281– Mutschler F. E., Griffin M. E., Scott S. D. and Shannon S. S.
2293. (1985) Precious metal deposits related to alkaline rocks in the
Clemente B., Scaillet B. and Pichavant M. (2004) The solubility of North American Dordillera – an interpretative review. Trans.
sulphur in rhyolitic melts. J. Petrol. 45, 2171–2196. Geol. Soc. S. Africa 88, 355–377.
Cottrell E. and Walker D. (2006) Constraints on core formation Ohnenstetter D. and Watkinson D. H. (1998) Low-temperature
from Pt partitioning in mafic silicate liquids at high tempera- evolution of the platinum group mineralogy, Two Duck Lake
tures. Geochim. Cosmochim. Acta 70, 1565–1580. intrusions, Coldwell Complex, Ontario, Canada. In Interna-
Dixon J. E., Clague D. A. and Stolper E. M. (1991) Degassing tional Platinum 1998 (eds. N. P. Laveroux and V. V. Dislter).
history of water, sulfur, and carbon in submarine lavas from Theophrastus Publishing, St. Petersburg, pp. 116–118.
Kilauea Volcano, Hawaii. J. Geol. 99, 371–394. O’Neill H. St. C. and Mavrogenes J. A. (2002) The sulfide capacity
Ertel W., O’Neill H., Sylvester P. J. and Dingwell D. B. (1999) and the sulfur content at sulfide saturation of silicate melts at
Solubilities of Pt and Rh in a haplobasaltic silicate melt at 1400 °C and 1 bar. J. Petrol. 43, 1049–1087.
1300 °C. Geochim. Cosmochim. Acta 63, 2439–2449. Peach C. L., Mathez E. A. and Keays R. R. (1990) Sulfide–silicate
Farges F., Neuville D. R. and Brown G. E. (1999) Structural melt distribution coefficients for noble metals and other
investigation of platinum solubility in silicate glasses. Am. chalcophile elements as deduced from MORB implications for
Miner. 84, 1562–1568. partial melting. Geochim. Cosmochim. Acta 54, 3379–
Fleet M. E., Crocket J. H. and Stone W. E. (1991) Partitioning of 3389.
palladium, iridium, and platinum between sulfide liquid and Scaillet B. and MacDonald R. (2001) Phase reations of peralkaline
basalt melt: effects of melt composition, concentration, and silicic magmas and petrogenetic implications. J. Petrol. 42, 825–
oxygen fugacity. Geochim. Cosmochim. Acta 55, 2545–2554. 845.
Fleet M. E., Crocket J. H., Liu M. and Stone W. E. (1999) Simon G., Kesler S. E. and Essene E. J. (2000) Gold in porphyry
Laboratory partitioning of platinum-group elements (PGE) and copper deposits: experimental determination of the distribution
5792 A.S. Bell et al. / Geochimica et Cosmochimica Acta 73 (2009) 5778–5792

of gold in the Cu–Fe–S system at 400 to 700 °C. Econ. Geol. 95, tion of a sulfide liquid. Neues Jahrbuch für Mineralogie
259–270. Monatshefte 10, 433–444.
Simon A. C., Pettke T., Candela P. A., Piccoli P. M. and Heinrich Ripley E. M., Brophy J. G. and Li C. (2002) Copper solubility in a
C. A. (2005) Gold partitioning in melt–vapor–brine systems. basaltic melt and sulfide liquid/silicate melt partition coeffi-
Geochim. Cosmochim. Acta 69(332), 1–3335. cients of Cu and Fe. Geochim. Cosmochim. Acta 66, 2791–2800.
Simon A. C., Pettke T., Candela P. A., Piccoli P. M. and Heinrich Rudnick R. L. and Gao S. (2003) The composition of the
C. (2006) Copper partitioning in sulfur bearing magmatic continental crust. In The Crust, Treatise on Geochemistry (ed.
systems. Geochim. Cosmochim. Acta 70, 5583–5600. R. L. Rudnick), pp. 1–64. vol. 3. Elsevier-Pergamon, Oxford.
Simon A. C., Pettke T., Candela P. A., Piccoli P. M. and Heinrich Wallace P. and Carmichael I. S. E. (1992) Sulfur in basaltic
C. (2007) The partitioning behavior of As and Au in a magmas. Geochim. Cosmochim. Acta 56, 1683–1874.
haplogranite – vapor at magmatic conditions in sulfur-free and Werle J. L., Ikramuddin M. and Mutschler M. E. (1984) Allard
sulfur-bearing systems. Geochim. Cosmochim. Acta 71, 1764– Stock, La Plata Mountains, Colorado—an alkaline rock-hosted
1782. porphyry copper—precious metal deposit. Can. J. Earth Sci. 21,
Simon A. C., Pettke T., Candela P. A., Piccoli P. M. and Heinrich 630–641.
C. (2008) The partitioning behavior of silver in a vapor–brine– Witter J. B., Kress V. C. and Newhall C. G. (2005) Volcán
rhyolite melt assemblage. Geochim. Cosmochim. Acta 72, 1638– Popocatépetl, Mexico: Petrology, magma mixing, and immedi-
1659. ate sources of volatiles for the 1994–present eruption. J. Petrol.
Stone W. E., Crocket J. H. and Fleet M. E. (1990) Partitioning of 46, 2337–2366.
palladium, iridium, platinum, and gold between sulfide liquid Wohlgemuth-Ueberwasser C. C., Ballhaus C., Berndt J., Stotter
and basalt melt at 1200 °C. Geochim. Cosmochim. Acta 54, nee Paliulionyte V. and Meisel T. (2007) Synthesis of PGE
2341–2344. sulfide standards for laser ablation inductively coupled plasma
Toulmin P. and Barton P. B. (1964) A thermodynamic study of mass spectrometry (LA-ICP-MS). Contrib. Mineral. Petrol.
pyrite and pyrrhotite. Geochim. Cosmochim. Acta 64, 641–671. 154, 607–617.
Tsujimura T. and Kitakaze A. (2004) New phase relations in the
Cu–Fe–S system at 800 °C; constraint of fractional crystalliza-
Associate editor: Edward M. Ripley