0361-0128/01/3491/157-8 $6.00 157

IntroductionThe compositions of magmatic and anatectic sulfide melts,

particularly volatile contents, are difficult to determine fromthe resultant crystalline products. The original volatile con-tent of a sulfide melt is obscured by the inability of major sul-fide phases to incorporate anions other than sulfur and the ex-tensive low-temperature textural reequilibration of sulfideminerals. The dissolution of volatiles in sulfide melts is a po-tentially important phenomenon, in terms of lowering thesolidus of sulfide melts, changing their physical properties(density, viscosity), and the formation of metasomatic alter-ation halos following the exsolution of volatiles. The recentexperimental study of Mungall and Brenan (2003) hasdemonstrated the solubility of halogens in sulfide melts andthus provides a plausible explanation for the coincidence ofelevated halogen contents in the rocks surrounding manymagmatic massive sulfide deposits. Kress (1997) investigatedthe solubility of O in Fe-S-O liquids and suggested that thepresence of C and H may alter the surface tension of silicateliquid-sulfide liquid interfaces and the addition of H2O to asilicate melt containing immiscible sulfide liquid may en-hance separation and coalescence of the sulfide liquid.

Anatectic sulfide melting, or partial melting of existing sul-fide occurrences during high-temperature metamorphism,has been suggested on the basis of phase relationships (Brettand Kullerud, 1967; Mavrogenes et al., 2001; Frost et al.,2002) and field evidence (Lawrence, 1967; Hofmann, 1994;Hofmann and Knill, 1996; Sparks and Mavrogenes, 2003;Tomkins and Mavrogenes, 2002, 2003; Tomkins et al., 2004).As a process, anatectic sulfide melting of a hydrothermallyderived sulfide protolith containing low melting point chal-

cophile elements such as Sb and As is capable of producingsulfide liquids at significantly lower temperatures than thoseat which conventional magmatic Fe-Ni-Cu-S-(O) melts form.Typically, magmatic sulfide melts form through saturation ofa parental silicate melt in an immiscible sulfide liquid. Assuch, the composition of the sulfide liquid is dictated by par-titioning between the sulfide liquid and the much larger vol-ume of silicate melt and the fO2 it imposes. As demonstratedby Mungall and Brenan (2003), the halogen contents of a sul-fide melt are lower that that of the silicate melt with which itis in equilibrium. Anatectic sulfide melts may form throughdirect melting of sulfide minerals in metamorphic environ-ments where fluids are common. Therefore, anatectic sulfidemelts have the potential to contain significant volatiles andwill be subject to the associated effects, such as a loweredsolidus temperature and decreased density and/or viscosity.This paper reports experiments conducted to determine theeffect of H2O on sulfide melting.

The limitations of experimental techniques have hamperedexperimental investigations into hydrous sulfide melts, ren-dering results contradictory and inconclusive. The investiga-tion of Naldrett and Richardson (1967) into the effect of H2Oon FeS-Fe3O4 at 2 kbars encountered many experimental dif-ficulties, outlined in some detail by the authors. Loss of ironfrom the sample to the gold capsule and the unknown effectof gold and pressure on the melting point of FeS-Fe3O4 weresome of the problems encountered. To overcome the effect ofgold on melting, hydrous and anhydrous FeS-Fe3O4 assem-blages were run concurrently. However, the H2O-free samplemelted, whereas the H2O-bearing sample did not, which Nal-drett and Richardson (1967) ascribed to imprecise tempera-ture control in their internally heated pressure vessel. Theyconcluded that H2O has little influence on the melting pointof FeS-Fe3O4 and oxide-free Fe-bearing sulfide melts. Kon-nikov (1997) investigated the effect of H2O on the melting of

HYDROUS SULFIDE MELTING: EXPERIMENTAL EVIDENCE FOR THE SOLUBILITY OF H2O IN SULFIDE MELTS

JEREMY L. WYKES†,*Department of Earth and Marine Sciences, Australian National University, Canberra, ACT 0200, Australia

AND JOHN A. MAVROGENES

Department of Earth and Marine Sciences, and Research School of Earth Sciences, Australian National University, Canberra, ACT 0200, Australia

AbstractThe effect of H2O on sulfide melting temperatures has been investigated in the FeS-PbS-ZnS system at 1.5

GPa, revealing that the addition of H2O results in a 35°C drop in melting temperature from 900° to 865°C.In addition to the melting point depression, the solubility of H2O is confirmed by the presence of vesicles inthe quenched melt. No oxide phases were present in any of the run products, ruling out oxidation as a causeof the melting point depression. Confirmation of the solubility of H2O in sulfide melts is consistent with therecent suggestion by Mungall and Brenan (2003) of a magmatic origin for halogen-rich alteration associatedwith magmatic sulfide ore deposits, as the hydrous component of the alteration may similarly originate in thefractionating sulfide melt. Anatectic sulfide melts could be expected to contain more H2O than magmatic sul-fide melts, owing to the lack of a parental silicate melt that buffers the H2O content of magmatic sulfide melts.Fluids expelled from the cooling anatectic melts, such as that present during granulite facies metamorphismof sulfide deposits at Broken Hill, Australia, may have been responsible for associated retrograde hydrother-mal alteration.

† Corresponding author: e-mail, jeremy.wykes@ems.anu.edu.au*Current address: Department of Earth and Space Sciences, University of

California at Los Angeles, Los Angeles, California 90095-1567.

©2005 Society of Economic Geologists, Inc.Economic Geology, v. 100, pp. 157–164

pyrrhotite at 1 kbar and reported a maximum melting pointdepression of 80° to 90°C at 10 wt percent H2O. The resultsof Konnikov (1997) were reinterpreted by Mungall and Bre-nan (2003) as reflecting oxidation through H2 loss and subse-quent melting on the FeS-Fe3O4 cotectic, an interpretationfavored here.

Thus, poor characterization of the pressure dependence ofsulfide (and sulfide-oxide) melting temperatures, interactionbetween sulfide melt and noble metal capsules, and H2 diffu-sion are all factors that have limited the ability of previousstudies to investigate the fluxing effect of H2O. In addition, itis not possible to directly measure the H2O content of aquenched sulfide melt due to the inability of sulfides toquench to glass.

We have reinvestigated the solubility of H2O in sulfidemelts and here report experiments designed to measure theeffect of H2O on the temperature of melting of a FeS-PbS-ZnS assemblage at 1.5 GPa. A depression of the melting pointand textural features are presented as evidence for the solu-bility of H2O. The temperature and composition of theternary eutectic in the H2O-free FeS-PbS-ZnS system at 1.5GPa was also determined. The pressure of 1.5 GPa was cho-sen for this study to maximize the melting point depressiondue to the addition of water, based on the assumption that themagnitude of such a melting point depression increases withpressure.

Methods

Capsule method

Noble metals (Ag, Au, Pd, Pt) commonly employed as cap-sule materials are highly chalcophile and dissolve readily into asulfide melt, making such capsules unsuitable for sulfide melt-ing experiments. Large quantities of sulfide melt are also diffi-cult to contain in other materials, such as graphite or boron ni-tride (BN). Thus, a novel capsule technique was devised forthis investigation. Capsules were manufactured by cold press-ing ~0.25 g of coarse galena (PbS) powder (natural galena fromthe West Fork mine, Viburnum Trend, Missouri) in a 4-mm-diameter pellet press, to produce a bucket-shaped capsule, and~0.10 g to form a lid. The advantage of such a capsule is that itdoes not add an additional component to the system but acts asa ready supply of reagent. A backscattered electron image of atypical galena capsule is shown in Figure 1.

Polycrystalline, cold-pressed capsules are an unproventechnique, and their ability to contain water was a prerequi-site for their use in this investigation. To test this, a series ofexperiments involving a silicate starting composition ofknown H2O content were conducted. The H2O content of theresulting glass was then quantified using Fourier transforminfrared (FTIR) spectroscopy, to evaluate the performance ofgalena and troilite capsules. The silicate composition was alsorun in a welded 2.3-mm (outside diam) Pt capsule for com-parison.

Piston-cylinder technique

All experiments were conducted at 1.5 GPa, utilizing a12.7-mm end-loaded piston cylinder apparatus located at theResearch School of Earth Sciences, Australian National Uni-versity. Galena capsules were placed in BN sleeves with BN

disks at the top and bottom and were positioned inside thegraphite heater with MgO pieces. An MgO sleeve and topdisk were employed for the Pt capsule. A low-friction pyrex-NaCl-teflon foil pressure assembly was utilized. Temperaturewas monitored using a type B, Pt94Rh6-Pt70Rh30, thermocou-ple, housed in 2-bore mullite tubing. Temperature measure-ments were accurate to within ±10°C; precision was esti-mated to be less than ±5°C. No pressure correction wasapplied to the electromotive force output of the thermocou-ple. The piston-out technique was employed in all experi-ments, as cold pressurization was required to press the lidonto the capsule. Microscopic inspection at the completion ofsuccessful experiments confirmed annealing of the lid to thecapsule. In a typical 1.5 GPa experiment, 1.0 GPa was appliedto the sample before the furnace was connected, and the sam-ple was heated at 150°C/min. The final 0.5 GPa was appliedonce the temperature reached 400°C. The sample was alwaysoverpressured by ~0.1 GPa (determined by experience) to ac-count for settling of the assembly. Following the experiment,capsules were extracted from the assembly and mounted inepoxy before polishing and examination under reflected light.Epoxy mounts of silicate melting experiments were ground to300- to 1,200-µm thickness and doubly polished. Thicknesseswere measured by a micrometer and are accurate to ±10 µm.

Starting materials

Starting compositions for sulfide melting experiments weremixtures of FeS, ZnS (Aldrich 99.9%), and natural PbS. Thegalena was the same as that used to construct the capsulesand contains very low amounts of Ag and Sb (confirmed byEMP (electron microprobe analysis) and LA-ICP-MS (laserablation-inductively coupled plasma-mass spectrometry;Wykes, 2003). FeS was synthesized by heating stoichiometricproportions of high-purity S and Fe metal in an evacuated sil-ica tube at 800°C overnight. Stoichiometry was confirmed via

158 SCIENTIFIC COMMUNICATIONS

0361-0128/98/000/000-00 $6.00 158

1 mm

tr + sptr + sp

galena

FIG. 1. Backscattered electron image of a polished section from a typicalgalena capsule experiment (JW02-34). The galena capsule has been loadedwith a pellet of sphalerite (sp) and troilite (tr).

EMP analysis. This troilite also was used to construct capsulesfor the silicate melting experiments. Experimental mixtureswere homogenized by grinding in an agate mortar and pestle.The compositions of the different starting mixtures used inthis study are listed in Table 1. The terms binary and ternaryrefer to the number phases present in the particular mixture;e.g., a binary mixture contains only FeS and ZnS, whereas aternary mixture contains FeS, ZnS, and PbS. However, ex-periments utilizing binary mixtures involve FeS, ZnS, andPbS, as galena (PbS) is provided by the capsule. Capsuleswere loaded with a 2.5-mm pressed pellet of starting mater-ial. Initial experiments were conducted using the FeS-ZnSmixtures, because at a temperature below the FeS-ZnS binaryeutectic, but above the FeS-PbS-ZnS ternary eutectic, meltcan only form where all three phases (galena, troilite andsphalerite) are in contact (e.g., at the contact between theFeS-ZnS pellet and the PbS capsule). This method was em-ployed to aid in the identification of small degrees of melting,as melt is restricted to the edge of the FeS-PbS sample. Oncethe approximate temperature of the ternary eutectic wasidentified, an FeS-PbS-ZnS composition was used, which wasbased on the measured composition of quenched melt fromearlier experiments. By proceeding in this manner it washoped that the eutectic could be dramatically demonstratedby the entire sample melting over a finite temperatureinterval.

A haplogranite mixture (listed in Table 2) was preparedfrom fired high-purity carbonates and oxides, to whichAl(OH)3 was added, before grinding under acetone. The H2Ocontent of the Al(OH)3 was determined by mass differenceafter heating an aliquot of Al(OH)3 at 900°C for 1 h.

Electron microscopy

All samples were examined using backscattered electron(BSE) imaging with a JEOL JSM6400 scanning electron mi-croscope (SEM) at the Electron Microscopy Unit, ResearchSchool of Biological Sciences, Australian National University.Quantitative analyses of mineral phases and quenched melts

were obtained using the JSM6400 Oxford Link ISIS EDS(energy dispersive spectrometry) accessory. An acceleratingvoltage of 15 kV and a beam current of 1 nA were used andthe counting time was 120 s. Mineral phases (pyrrhotite,galena, and sphalerite) were analyzed, using a focused 1-µmspot, whereas melt compositions were obtained from areascans, ranging from 10 to 500 µm across, to accommodate theheterogeneity of the sulfide melt quench intergrowth.

Infrared spectroscopy

The H2O contents of hydrous glasses were calculated fromIR spectra collected using a Bruker IFS28 infrared spectrom-eter equipped with a Bruker A590 infrared microscope. Theinstrument comprises a Globar light source, KBr beamsplitter,and an MCT (HgCdTe) detector. Absorption spectra were col-lected between 600 and 5,500 cm–1 at a spectral resolution of2 cm–1. Spot sizes ranged from 50 to 200 µm and 32 scans wereaccumulated for each spectrum. Background spectra weremeasured frequently. Between four and six spectra wererecorded for each sample, and attempts were made to analyzeonly homogeneous and transparent portions of the glass. How-ever, this was not always possible due to the presence ofheterogeneities. Spectra showing obvious attenuation werediscarded. Baseline correction of the spectra was performedby fitting a flexicurve baseline (Sowerby and Keppler, 1999).Calculations of peak area by fitting different baselines andpeak boundaries to a single spectrum showed the repeatabilityof the area determinations to be ±1.3 percent. The density(2,330 g ⋅ L–1) and integrated molar extinction coefficients(278 ± 28 L ⋅ mol–1 ⋅ cm–2 for the 4,500-cm–1 peak and 2,480± 10 L ⋅ mol–1 ⋅ cm–2 for the 5,200-cm–1 peak) determined byWithers and Behrens (1999) for a synthetic Ab37Or29Qz34 rhy-olite were used for quantification of the H2O content.

Results

Silicate melt experiments

Table 3 contains run conditions, calculated H2O contents,sample thickness, and the standard deviation of the watercontents. Uncertainty is associated with the choice of densityand molar extinction coefficients, although these parametersare not listed in Table 3 as this study is concerned with therelative differences in H2O content between experimentsrather than the absolute H2O content. Vesicle-free glass waspresent in all experiments, although transmitted and/or re-flected light microscopy and SEM imaging revealed that allexperiments contained sulfide inclusions and crystalline ma-terial, presumably unmelted starting material, although thedegree of melting was estimated to be >80 percent in all ex-periments. All glasses contained quenched kyanite crystals.

Experiment JW02-36 (PbS capsule) was the first silicatemelt experiment performed, and the product had a signifi-cantly lower H2O content than that of the other PbS-encap-sulated experiment (JW02-41). Experiment JW02-36 wasloaded with an insufficient amount of oxide mix, which re-sulted in cracking of the capsule during pressurization, andthe capsule presumably lost most of its H2O. Comparisonbetween the H2O contents of the successful experiments inpyrrhotite and galena capsules with that of the Pt capsulereveals surprisingly good agreement. There is a maximum

SCIENTIFIC COMMUNICATIONS 159

0361-0128/98/000/000-00 $6.00 159

TABLE 1. Starting Compositions for Sulfide Experiments

Mixture FeS ZnS PbS FeS/ZnS

Binary 5 70.0 30.0 2.3:1Binary 8 83.3 16.6 5:1Ternary 3 55.2 4.6 40.2 12:1Ternary 4 45.0 5 50.0 9:1

Notes: All values in mol percent

TABLE 2. Haplogranite Composition

Mol % Wt %

SiO2 67.37 72.57Na2O 6.28 6.98K2O 1.86 3.14Al2O3 6.24 11.41H2O 18.25 5.90

Total 100.00 100.00

relative difference of only 12.5 percent between the H2Ocontents of the experiments in the sulfide and Pt capsules.These results demonstrate that sulfide capsules are capable ofcontaining H2O to produce hydrous melts, approximating theefficiency of noble metal capsules. Additional experiments inthe PbS capsule were performed in which free H2O wasadded to an Al(OH)3-bearing oxide mix. These runs producedvesicle-rich H2O-oversaturated glasses, demonstrating thatthe PbS capsule also can seal free H2O. Thus cold-pressedpolycrystalline capsules are suitable for the investigation ofhydrous sulfide melting.

Anhydrous experiments

The run conditions and results of H2O-free experimentsare given in Table 4. The onset of melting in the FeS-PbS-ZnS system was observed between 895° and 900°C at 1.5GPa, using an FeS-ZnS sample, and was conclusively demon-strated at 900°C, using an FeS-PbS-ZnS mix. Complete melt-ing of the sample was observed between 900° and 905°C.BSE images of sample textures from runs over the tempera-ture range 895° to 905°C are presented in Figure 2. Almostcomplete melting of the sample over a 10°C temperature in-terval is proof of a eutectic melting relationship. The 900° ±10°C temperature of the FeS-PbS-ZnS eutectic at 1.5 GPa isin excellent agreement with the 60°C per GPa pressure de-pendence of the eutectic determined by Mavrogenes et al.(2001).

Unusual textures were observed in experiment JW02-47(Fig. 2C), in which the melt patch was separated from theunmelted portion of the sample by a band of galena contain-ing small amounts of interstitial melt. This texture is inter-

preted to be the result of a small vertical temperature gradi-ent within the capsule. The upper portion of the sample wasbelow the eutectic temperature, whereas the lower portionwas above the eutectic temperature. In response to this ther-mal gradient the melt migrated up temperature, dissolvinggalena on the hotter side of the melt patch (in this instance,the lower boundary) and precipitating galena on the coolerside. Small amounts of melt were trapped between galenagrains during this process. Thermal migration of this kind hasbeen observed in salt-hosted fluid inclusions (Roedder, 1984)and in experiments simulating sulfide migration in iron mete-orites (Buchwald et al., 1985). Melts are typically very sensi-tive to small temperature gradients (Wark and Watson, 2002),and the magnitude of the temperature gradient is also likelyto be small as its effects are only visible in experiments within5°C of the eutectic. As all experiments utilized a sample as-sembly of the same dimensions, the temperature gradient isassumed to have been similar in all cases.

The composition of the eutectic melt obtained from EDSarea scans is listed in Table 5, with the 1-bar eutectic fromMavrogenes et al. (2001) for comparison. Melt compositionsare reported in terms of monosulfides as the sulfur content ofthe melt from experiment JW02-47 was 50.03 mol percent(1σ: 0.37) and similar for all other experiments. Increasingpressure from 1 bar to 15 kbars reduced the ZnS content ofthe melt by ~50 percent and increased the PbS content.

Hydrous experiments

The results and experimental conditions of the H2O-bear-ing experiments are listed in Table 6. Large amounts of H2Owere added to allow for the expected migration of an un-known amount of H2O into the capsule walls over the courseof the experiment. Figure 3A-C presents BSE images of tex-tures from experiment JW02-56. The addition of H2O pro-duced a 35°C depression in the eutectic temperature and

160 SCIENTIFIC COMMUNICATIONS

0361-0128/98/000/000-00 $6.00 160

TABLE 3. Fourier Transform Infrared Analyses of Hydrous Glasses from Silicate Melting Experiments

Run Capsule Thickness (µm) H2O4500 H2O5200 H2Ototal (wt %) σ n

JW02-36 PbS 320 0.51 0.58 1.09 0.13 6JW02-37 FeS 840 1.58 4.98 6.56 0.23 5JW02-39 FeS 570 1.43 4.37 5.80 0.28 5JW02-41 PbS 1140 1.43 5.69 7.13 0.36 4JW02-42 Pt 840 1.55 4.78 6.34 0.10 5

Notes: H2O content of starting mixture was 5.90 wt %; abbreviations: H2O4500 = wt % conc. of H2O from 4,500-cm–1 peak, H2O5200 = wt % conc. of H2Ofrom 5,200-cm–1 peak, H2Ototal = sum of H2O4500 and H2O5200; σ = standard deviation based on n spectra, n = number of spectra collected

TABLE 4. Run Conditions and Results of Anhydrous FeS-PbS-ZnS Melting Experiments

Run T (°C) Mixture Melt

JW02-29 840 Binary 5 NoJW02-30 870 Binary 5 NoJW02-31 900 Binary 5 YesJW02-32 885 Binary 5 NoJW02-33 890 Binary 5 NoJW02-34 895 Binary 5 NoJW02-45 905 Binary 8 YesJW02-46 895 Ternary 3 SlightJW02-47 900 Ternary 3 YesJW02-66 905 Ternary 4 Yes

Notes: All experiments conducted at 1.5 GPa for 2 h

TABLE 5. Melt Compositions from Hydrous and Anhydrous Melting Experiments

FeS ZnS PbS

JW02-47 (900°C, 1.5 GPa) 47.8 4.6 47.62σ (15 analyses) 3.2 2.6 3.2JW02-56 (865°C, 1.5 GPa, +H2O) 52.4 5.3 44.62σ (15 analyses) 4.2 1.4 3.3Mavrogenes et al. (2001) (800°C, 1 bar) 51.3 10.6 38.0

Note: Values in mol percent

resulted in the development of vesicles, typically rimmed bygalena, in samples containing quenched melt. Fluid pores werepresent in the sample and capsule walls of all experiments,melt-bearing or otherwise. Despite extensive searching, using

BSE imaging and X-ray mapping, no oxide phases were iden-tified in any experiment. Isolated occurrences of small, (<5-µm) round blebs of a Pb-rich phase (presumably a quenchedPb melt) were identified within the quenched intergrowth ofthe sulfide melt in experiment JW02-56 (Fig. 3D).

The presence of fluid pores is taken as evidence that H2Owas present within the capsule for the duration of the exper-iment. In experiments in which melting occurred, the de-pression in the melting point relative to the anhydrous eutec-tic and the occurrence of vesicles in melt patches isinterpreted as evidence for the solubility of H2O in the sulfidemelt. As sulfide melts quench to a fine intergrowth of sulfidephases (troilite, sphalerite, and galena) that do not accommo-date H2O, H2O is thus exsolved on quenching. The absenceof oxide phases and the presence of very minor amounts(<0.01 vol %) of Pb melt suggest that insignificant S was lostfrom the melt to form H-O-S volatiles. In experiments inwhich melting was not observed, neither PbS nor ZnS was ob-served outside the sample area, even though fluid pores were

SCIENTIFIC COMMUNICATIONS 161

0361-0128/98/000/000-00 $6.00 161

100 µm 100 µm

100 µm 100 µm

A B

C D

tr + sptr + sp

galenacapsulewall

quenched melttr + sp + melttr + sp + melt

quenched melt

recrystallizedtr + sp + gn

tr + sp + gntr + sp + gn

FIG. 2. Backscattered electron images from H2O-free melting experiments in the FeS-PbS-ZnS system. Abbreviations: gn= galena, sp = sphalerite, tr = troilite. A. Textures from experiment JW02-34 performed at 895°C, below the FeS-PbS-ZnSeutectic. B. Textures from experiment JW02-31 performed at 900°C. The onset of melting is indicated by the pockets ofquenched melt formed between the troilite + sphalerite sample and the wall of the galena capsule. Small amounts of meltare present along tr-sp grain boundaries. C. Textures from experiment JW02-47 loaded with a tr-gn-sp sample pellet, as op-posed to the tr-gn pellet of experiment JW02-31. This experiment dramatically demonstrated the presence of a small-tem-perature gradient within the capsule. Originally loaded with a sample geometry identical to that in Figure 1, the lower halfof the sample, at temperatures high enough to melt, has migrated uptemperature toward the bottom of the capsule. D. Tex-tures from experiment JW02-66, performed at 905°C. The 5°C increase in temperature results in almost complete meltingof the sample.

TABLE 6. Run Conditions and Results of Hydrous FeS-PbS-ZnS MeltingExperiments

Run T (°C) Duration (h) Mixture H2O (wt %) Melt

JW02-48 895 2 Ternary 3 ~7 YesJW02-49 890 2 Ternary 3 ~22 YesJW02-50 875 1 Ternary 3 ~21 YesJW02-51 850 1 Ternary 3 ~15 NoJW02-52 865 1 Ternary 3 ~10 SlightJW02-53 870 1 Ternary 3 ~12 NoJW02-56 865 1 Ternary 4 ~33 YesJW02-59 855 1 Ternary 4 ~25 No

Notes: All experiments conducted at 1.5 GPa, H2O is listed as wt percentrelative to the amount of FeS-PbS-ZnS mixture added

well developed in the capsule walls. This suggests that fluidtransport of PbS and ZnS was insignificant.

EDS analyses of quenched melt from experiment JW02-56(S content 49.44 mol %, 1σ: 0.44; Table 5) demonstrate thatthe addition of H2O did not significantly alter the relative pro-portions of FeS, PbS, and ZnS in the eutectic melt (identicalwithin 2σ).

The results of experiment JW02-53 (870°C, no melting) arenot consistent with the results of experiments JW02-52 andJW02-56 (865°C, melting). Fluid-filled pores were present inthe sample and capsule walls of experiment JW02-53, con-firming the presence of a fluid phase. Given that melting wasobserved in two experiments at 865°C, the results of experi-ment JW02-53 are considered anomalous.

The addition of H2O appears to have changed the physicalproperties of the sulfide melts, such that the H2O-bearingmelts were more mobile within the capsule, in some cases mi-grating through the capsule wall and ponding against the BNsleeve (experiment JW02-50). In experiments JW02-49 andJW02-50, separate domains of very coarse and fine quenched

products, respectively, were present in the sample (Fig. 3D).Dissolved H2O may have altered the structure of the sulfidemelt, resulting in the different quenching behavior.

DiscussionThe results of this experimental study clearly demonstrate

that the addition of H2O to a sulfide assemblage results in amelting point depression, which is interpreted as evidence forthe solubility of H2O in the melt. These findings encouragefurther experimentation as well as a search for supporting ev-idence in natural systems.

There are reports in the literature supporting the solubilityof H2O in sulfide melts from the Copper zone of the Frasermine (formerly the Strathcona Deep Copper zone), SudburyIgneous Complex, Canada, where highly fractionated sulfidemagma was injected into the footwall country rocks, isolatingit from the parent silicate magma (Li et al., 1992). Li et al.(1992), Farrow and Watkinson (1992), and Li and Naldrett(1993) described veinlets that splay off the Cu sulfidestringers. These veinlets are zoned from proximal sulfide rich

162 SCIENTIFIC COMMUNICATIONS

0361-0128/98/000/000-00 $6.00 162

A

100 µm 100 µm

gn + fluid poresgn + fluid pores

vesicles

quenched melt

100 µm

coarse intergrowthcoarse intergrowth'normal'

quenchedmeltPb-rich materialPb-rich material

10 µm

C

B

DC

FIG. 3. Backscattered electron images from H2O-bearing melting experiments in the FeS-PbS-ZnS system. A. and B. Ex-periment JW02-56 at 865°C. The addition of H2O depresses the melting temperature in the FeS-PbS-ZnS system and pro-duces patches of quenched melt with large vesicles, partially rimmed with galena. Much of the troilite and sphalerite hasbeen recrystallized to textures similar to those in Figure 2C. Intergranular fluid pores are present throughout the capsule,confirming that the experiments were fluid saturated. C. Detail from (B), showing typical quench textures and small blobsof Pb-rich material within the quenched melt. D. Experiment JW02-50 at 875°C. This experiment contained two distincttypes of quench textures. In the lower right corner of the sample there is a pocket of regular quench, whereas the rest of thesample consists of a coarse skeletal intergrowth of galena, sphalerite, and troilite.

to distal quartz rich and are surrounded by concentric halosof epidote and chlorite alteration. Li et al. (1992) suggestedthat the veinlets and surrounding alteration were formed byfluids exsolved from the cooling sulfide magma. Severalworkers (Jago et al., 1994; Molnar et al., 2001; McCormick etal., 2002; Hanley and Mungall, 2003; Mungall and Brenan,2003) have identified elevated halogen concentrations andhigh Cl/Br and Cl/F ratios in alteration halos associated withmagmatic sulfide deposits, including in the Fraser mine.Through experiments, Mungall and Brenan (2003) con-firmed that the halogen contents of a sulfide melt would in-crease during fractionation, leading to the exsolution of asaline fluid. However, on the basis of the existing experi-mental studies of Naldrett and Richardson (1967) and Kon-nikov (1997), Mungall and Brenan (2003) concluded that theexsolved fluid would most likely be anhydrous. The results ofthe present study confirm the possibility that a hydrosalinefluid phase derived from a sulfide melt could be an end-member fluid responsible for high Cl/Br hydrous alterationassociated with magmatic sulfide deposits. The controls onhalogen partitioning between sulfide and silicate melts out-lined by Mungall and Brenan (2003) are likely also applica-ble to H2O, and we suspect that, like Cl, H2O will partitionpreferentially into a silicate melt rather than a sulfide melt.Thus, significant H2O contents will only be observed in mag-matic sulfide melts that have undergone fractionation iso-lated from the parent silicate melt.

The solubility of other nonchalcophile components (e.g.,SiO2) in natural sulfide melts may be more extensive thancurrently recognized. The overwhelming majority of experi-ments on sulfide phase relationships have been conductedusing evacuated silica tubes, and little evidence exists for thesolubility of SiO2 in (slowly quenched, low-pressure) sulfidemelts. However, Kalinowski (2002) reported the growth ofeuhedral Mn silicates in experiments on the melting Mn-bearing sulfides, and Mungall and Brenan (2003) observed anunidentified Fe-Si-Cl phase in the quenched products of Cl-bearing sulfide melts. In these cases we might speculate thatthe presence of small amounts of one nonchalcophile compo-nent (e.g., Mn or Cl) may result in increased solubility of an-other nonchalcophile component (e.g., SiO2), similar to theinfluence of Cl on sulfur solubility in silicate melts(Botcharnikov et al., 2004).

Field relationships at the Noril’sk-Talnakh deposits suggestthat the sulfide melt was mobilized and emplaced in its cur-rent location after solidification of the parent silicate melt(Kunilov, 1994; Yakubchuk and Nikishin, 2004). The role ofvolatiles in this process is unknown, although the basal con-tact between the massive sulfides and the underlying Devon-ian evaporite-bearing sequences would have provided ampleopportunity for incorporation of volatiles (SO4, Cl) into thesulfide melt. A full understanding of the effect of dissolvedvolatiles on solidus temperatures, density, and viscosity, andthe effects of their subsequent exsolution on the observed re-lationships between sulfide and silicate rocks at Norils’k-Tal-nakh and other magmatic sulfide deposits awaits further ex-perimental and field investigation.

Recently identified partial melting of sulfides at BrokenHill (NSW, Australia; Mavrogenes et al., 2001; Sparks andMavrogenes, 2003) provides an explanation for features such

as Ag-rich “droppers” (Sparks and Mavrogenes 2004) and thelocalized presence of Ag-rich galena in fold hinges (Plimer,1986). We speculate that during prograde metamorphismthe incipient sulfide melt at Broken Hill may have incorpo-rated significant H2O due to the lack of a parental silicatemelt to buffer the H2O content at a low level. Crystallizationof anhydrous sulfide phases during retrograde cooling alsowould have increased the H2O content to saturation, possiblycausing hydrothermal alteration of the host silicate rocks.Thus, we suggest that some of the retrograde alteration asso-ciated with the Broken Hill ore may have been the result offluids expelled from a cooling sulfide melt. Unfortunately,the amount of H2O that can be dissolved in a sulfide melt re-mains unknown, so the importance of sulfide melts as asource for the hydrous component in alteration zones is stillsomewhat of an open question. Hopefully, the results of thisstudy will provoke further research into the solubility of H2Oin sulfide melts and the geologic significance of hydrous sul-fide melts.

AcknowledgmentsThis work comprises the bulk of JLW’s B.Sc.(Honours) the-

sis, which was partly supported by a Society of Economic Ge-ologists student research grant. Frank Brink and Nick Wareare thanked for their assistance with SEM and EMP analyses.Andrew Berry is thanked for guidance using FTIR spec-troscopy. The invaluable help and knowledge offered by BillHibberson and Dean Scott eased JLW’s introduction to ex-perimental petrology. This paper was much improved follow-ing informal reviews by Bear McPhail, Steve Beresford, andbenefited from comments by Economic Geology reviewersJim Mungall, James Brenan, and Mark Hannington.May 28, November 8, 2004

REFERENCESBotcharnikov, R.E., Behrens, H., Holtz, F., Koepke J., and Sato, H., 2004,

Sulfur and chlorine solubility in Mt. Unzen rhyodacitic melt at 850°C and200 MPa: Chemical Geology, v. 113, p. 207–225.

Brett, R., and Kullerud, G., 1967, The Fe-Pb-S system: ECONOMIC GEOLOGY,v. 62, p. 354–369.

Buchwald, V.F., Kjer, T., and Thorsen, K.A., 1985, Thermal migration: Orhow to transport iron sulfide in solid iron meteorites: Meteoritics, v. 20, p.617–618.

Farrow, C.E.G., and Watkinson, D.H., 1992, Alteration and the role of fluidsin Ni, Cu and platinum-group element deposition, Sudbury Igneous Com-plex contact, Onaping-Levack area, Ontario: Mineralogy and Petrology, v.46, p. 67–83.

Frost, B.R., Mavrogenes, J.A., and Tomkins, A.G., 2002, Partial melting ofsulfide ore deposits during medium- and high-grade metamorphism: Cana-dian Mineralogist, v. 40, p. 1–18.

Hanley J.J., and Mungall, J.E., 2003, Chlorine enrichment and hydrous al-teration of the Sudbury Breccia hosting footwall Cu-Ni-PGE mineraliza-tion at the Fraser mine, Sudbury, Ontario, Canada: Canadian Mineralogist,v. 41, p. 857–881.

Hofmann, B.A., 1994, Formation of a sulfide melt during Alpine metamor-phism of the Lengenbach polymetallic sulfide mineralization, Binntal,Switzerland: Mineralium Deposita, v. 29, p. 439–442.

Hofmann, B.A., and Knill, M.D., 1996, Geochemistry and genesis of theLengenbach Pb-Zn-As-Tl-Ba-mineralisation, Binn Valley, Switzerland:Mineralium Deposita, v. 31, p. 319–339.

Jago, B.C., Morrison, G.G., and Little, T.L., 1994, Mineral zonation patternsand microtextural and micromineralogical evidence for alkali- and halogen-rich fluids in the genesis of the Victor Deep and McCreedy East footwallcopper orebodies, Sudbury Igneous Complex: Ontario Geological SurveySpecial Volume 5, p. 65–75.

SCIENTIFIC COMMUNICATIONS 163

0361-0128/98/000/000-00 $6.00 163

Kalinowski, A., 2002, An experimental investigation into the causes and ef-fects of sulfide partial melting at Broken Hill, N.S.W., Australia: Unpub-lished B.Sc. Honours thesis, Canberra, ACT, Australian National Univer-sity, 82 p.

Konnikov, E.G., 1997, Experimental studies of the melting processes in apyrrhotine-water system: Geology of Ore Deposits, v. 39, p. 474–479.

Kress, V., 1997, Thermochemistry of sulfide liquids. I. The system O-S-Fe at1 bar: Contributions to Mineralogy and Petrology, v. 127, p. 176–186.

Kunilov, Y.V., 1994, Geology of the Noril’sk region: the history of the discov-ery, prospecting, exploration and mining of the deposits: Ontario Geologi-cal Survey Special Volume 5, p. 203–216.

Lawrence, L.J., 1967, Sulphide neomagmas and highly metamorphosed sul-phide deposits: Mineralium Deposita, v. 2, p. 5–10.

Li, C., and Naldrett, A.J., 1993, High chlorine alteration minerals and cal-cium-rich brines in fluid inclusions from the Strathcona Deep Copperzone, Sudbury, Ontario: ECONOMIC GEOLOGY, v. 88, p. 1780–1796.

Li, C., Naldrett, A.J., Coats, C.J.A., and Johannessen, P., 1992, Platinum, pal-ladium, gold, and copper-rich stringers at the Strathcona mine, Sudbury:Their enrichment by fractionation of a sulfide liquid: ECONOMIC GEOLOGY,v. 87, p. 1584–1598.

Mavrogenes, J.A., MacIntosh, I.W., and Ellis, D.J., 2001, Partial melting ofthe Broken Hill galena-sphalerite ore: Experimental studies in the systemPbS-FeS-ZnS-(Ag2S): ECONOMIC GEOLOGY, v. 96, p. 205–210.

McCormick, K.A., Lesher, C.M., McDonald, A.M., Fedorowich, J.S., andJames, R.S., 2002, Chlorine and alkali geochemical halos in the footwallbreccia and sublayer norite at the margin of the Strathcona embayment,Sudbury structure, Ontario: ECONOMIC GEOLOGY, v. 97, p. 1509–1519.

Molnar, F., Watkinson D.H., and Jones P.C., 2001, Multiple hydrothermalprocesses in footwall units of the North Range, Sudbury Igneous Complex,Canada, and implications for the genesis of vein-type Cu-Ni-PGE deposits:ECONOMIC GEOLOGY, v. 96, p. 1645–1670.

Mungall, J.E., and Brenan, J.M., 2003, Experimental evidence for the chal-cophile behaviour of the halogens: Canadian Mineralogist, v. 41, p.207–220.

Naldrett, A.J., and Richardson, S.W., 1967, Effect of water on the melting ofpyrrhotite-magnetite assemblages: Carnegie Institution of WashingtonYearbook, v. 66, p. 429–431.

Plimer, I.R., 1986, Remobilisation in high-grade metamorphic environments:Ore Geology Reviews, v. 2, p. 231–245.

Roedder, E., 1984, Fluid inclusions: Reviews in Mineralogy, v. 12, 646 p.Sowerby, J.R., and Keppler, H., 1999, Water speciation in rhyolitic melt de-

termined by in-situ infrared spectroscopy: American Mineralogist, v. 84, p.1843–1849.

Sparks, H.A., and Mavrogenes, J.A., 2003, Sulfide partial melting at BrokenHill, Australia, in Eliopoulos, D.G., et al., eds, Mineral exploration and sus-tainable development: Rotterdam, Millpress, p. 1027–1030.

Tomkins, A.G., and Mavrogenes, J.A., 2002, Mobilization of gold as a poly-metallic melt during pelite anatexis at the Challenger deposit, South Aus-tralia: A metamorphosed Archean gold deposit: ECONOMIC GEOLOGY, v. 97,p. 1249–1271.

——2003, Generation of metal-rich felsic magmas during crustal anatexis:Geology, v. 31, p. 765–768.

Tomkins, A.G., Pattison, D.R.M., and Zaleski E., 2004, The Hemlo gold de-posit, Ontario: An example of melting and mobilization of a precious metal-sulfosalt assemblage during amphibolite facies metamorphism and defor-mation: ECONOMIC GEOLOGY, v. 99, p. 1063–1084.

Wark, D.A., and Watson, E.B., 2002, Grain-scale channelization of pores dueto gradients in temperature or composition of intergranular fluid or melt:Journal of Geophysical Research, v. 107B, p. 1–5.

Withers, A.C., and Behrens, H., 1999, Temperature-induced changes in theNIR spectra of hydrous albitic and rhyolitic glasses between 300 and 100K: Physics and Chemistry of Minerals, v. 27, p. 119–132.

Wykes, J.L., 2003, Just add H2O… hydrous sulfide melting in the FeS-PbS-ZnS system at 1.5 GPa: Unpublished B.Sc. Honours thesis, Canberra, ACT,Australian National University, 87 p.

Yakubchuk, A., and Nikishin, A., 2004, Noril’sk-Talnakh Cu-Ni-PGE de-posits: A revised tectonic model: Mineralium Deposita, v. 39, p. 125–142.

164 SCIENTIFIC COMMUNICATIONS

0361-0128/98/000/000-00 $6.00 164