American Mineralogist, Volume 67, pages 877-8E5, I9E2

Sulfide saturation of basalt and andesite meltsat high pressures and temperatures

RrcHeno F. WBNoI-eNor

Lunar and Planetary Institute3303 NASA Road IHouston. TX 77058

Abstract

The solubility of sulfur in silicate melt coexisting with iron sulfide melt has beendetermined as a function of temperature and pressure for three silicate compositions: Mt.Hood andesite, Grande Ronde basalt and Goose Island basalt (FeO, : 5.4, I l.l and l7 .UVo,respectively). Experimental temperatures vary from 1300-1460' C and pressures vary from12.5-30 kbar. The oxygen fugacity in the experiments is near that of the C-CO-CO5O2butrer (10-e bar at P, : 20 kbar) and the sulfur fugacity is buffered internally at a valueabove the iron-troilite equilibrium 0G, : I bar at Pt : 20 kbar).

Sulfur solubility decreases with increasing pressure and increases with increasingtemperature and FeOl content of the silicate melt; observed solubilities range from a lowvalue of 0.05 wt.To at 30 kbar and 1420" C (andesite) to a high value of 0.26Vo at 12.5 kbarand 1420'C (Goose Island basalt). The FeO, content of the coexisting sulfide meltdecreases with increasing pressure (from a maximum of 5.5% at 12.5 kbar and 1420" C-Goose Island basalt), increases with FeO, content of the silicate melt and does not appearto be dependent on temperature. With increasing pressure, the miscibility gap betweennatural silicate melts and sulfide melts expands and the ratio of FeOiilicate meltTp"g,sulfide melt

increases.The sulfur solubility measurements establish a basis for evaluating the appearance of

immiscible sulfide in erupted magmas, core segregation processes, and subsequenttransition metal partitioning studies. Because of the strong negative pressure dependenceof sulfur solubility in silicate melts, a magma may be sulfide-saturated in the source regionbut not on eruption until the late stages of crystallization.

Introduction

When the sulfur content of an Fe-bearing magmaexceeds the saturation limit for the bulk composi-tion, an immiscible iron sulfide melt fraction sepa-rates. Chalcophile elements, including many transi-tion metals, partition preferentially into the sulfidemelt (Maclean and Shimazaki, 1976; Fleet et al.,1977; Medvedev and Al'mukhamedov, 1978; Raja-mani and Naldrett, 1978; Fleet et al., l98l) effectingmajor changes in the trace and minor elementdistributions of the silicate melt (Duke and Nal-drett, 1978) and possibly generating metal-sulfidedeposits (Naldrett, 1973:' Campbell, 1977). Chalco-phile element distributions will also be dependenton whether the silicate melts are sulfur-saturated attheir source: Elimination of the residual sulfide (bydissolution in the silicate melt) is required for theentry of many elements (e.g., the Pt-group ele-0003-0M>vE2l09 I 0-0877$02. 00

ments) into the silicate melt. To understand thegeochemistry of sulfur-bearing magmatic systems,better knowledge is required of the solubility ofmetal sulfide in silicate melt at its source, and howthat solubility changes as a function of changingintensive and extensive variables. In this investiga-tion, the sulfur saturation surface has been deter-mined between 12.5 and 30 kbar and 1300-1460" Cfor three silicate melt compositions representing arange of SiO2 and FeO, (total iron calculated asFeO) compositions: Mt. Hood andesite (MHA),Grande Ronde basalt (GRB) and Goose Islandbasalt (GIB).

Measurements of sulfur content in sulfur-under-saturated silicate melts, determined at low pressurein simple synthetic systems (Fincham and Richard-son, 1954; Richardson and Fincham, 1954; Mac-Lean, 1969; Shimazaki and Clark, 1973) and incomplex synthetic and natural systems (Haughton

877

WENDLANDT: SULFIDE SATURATION OF MELTS

et al., 1974; Katsura and Nagashima, 1974; Shimaand Naldrett, 1975; Buchanan and Nolan, 1979,Danckwerth et al., 1979), have indicated a strongdependence on temperature and composition: Apositive dependence of sulfur content on tempera-ture,1f5,, and TiO2 and FeO content of the silicatemelt and a negative correlation between sulfurcontent and /6,. Along the sulfide saturation sur-face, however, sulfur contents of silicate meltsdecrease with increasing /s, and decreasing /e,(Haughton et al., 1974; Dan6kwerth et al., 1979).Additionally, estimates of sulfur solubility in sili-cate melts have been made by analyzing sulfurcontents of rapidly quenched basalts and dacites(Moore and Fabbi, 1971; Mathez,1976 Czamanskeand Moore, 1977 Ueda and Itaya, 1981), of differ-entiated siliceous melt in a Hawaiian lava lake(Skinner and Peck, 1969), and of glass inclusionstrapped in crystals (Anderson, 1974). These datashow a large degree of variability as one wouldexpect of occurrences of variable compositionwhich quenched at different rates, temperatures,and pressures.

Experiments involving sulfur at high pressuresare few in number because of experimental difficul-ties, primarily those of encapsulation and of con-trolling.fo, and fs,. Helz and Wyllie (1979) deter-mined sulfur solubility in carbonatitic melts (iron-absent) at 1 kbar. Helz (1977) determined the firstappearance of sulfide melt in natural basalts to 20kbar. Experiments were done in Pt capsules with-out controlled/6. and/s,; although iron loss to thecapsule was minimal (within approximately 3% ofthe original amount), sulfur mass balance was notreported. Huang and Williams (1980) investigatedportions of the system Fe-Si-S-O to 32 kbar anddemonstrated that the sulfide melt-silicate meltmiscibility gap expands with pressure. Mysen(1977) presented preliminary results for SO2 solubil-ity in albite melt, but with no details on experimen-tal technique. Mysen and Popp (1980) reportedsulfur solubilities increasing markedly with pres-sure and only slightly with temperature in diopsideand albite melt; however, applications of their re-sults to natural systems are limited by the Fe-absentcompositions they investigated. The current inves-tigation was undertaken to reconcile the variable,and, in some instances, conflicting results of previ-ous investigators and to establish a basis for futurepartitioning studies of platinum metals betweensulfide and silicate melts.

Experimental methods

All experiments were done with a piston-cylinderapparatus using 1.27 cm talc-Pyrex furnace assem-blies and a piston-out technique with a -5 kbarcorrection for friction as calibrated against thequartz-coesite transition. Nominal pressures weremaintained at -r0.1 kbar. Temperatures were mea-sured with Pt-Ptl0Rh thermocouples and automati-cally controlled to -r2" C. No corrections for theeffect of pressure on the emf of the thermocouplehave been made.

Starting materials were 60Vo silicate and 40Vopyrrhotite by weight. The compositions of the sili-cate fractions are shown in Table 1 The Mt. Hoodandesite, Grande Ronde basalt, and Goose Islandbasalt were selected for the ranges of SiO2 and FeO,they encompassed. Melting phase relations of theandesite at high pressures have been determined byEggler and Burnham (1973); those of the GooseIsland basalt have been determined by Helz (1978).High pressure phase relations have not been estab-lished for the Grande Ronde basaltl however.

Table l. Compositions of silicate starting materials

GRB** GIB*

si02

Ti 02

Al 203Fer0,

Fe0

Mn0

Mgo

Ca0

Nar0

\0Pzoss

Hzo

Total

60 . 78

0 . 8 3

1 7 . 3 8

a . z 1

3 . 3 8

0 . l 0

3 . 1 9

6 . 0 9

4 . 1 6

L . 2 0

0 . 1 8

n . d .

0 . 4 4

99.97

54 .38

I . 9 1

L4.71

n . d .

l l . l l * * *

0 . 1 9

4 . 5 8

8 . 0 1

2 . 6 7

1 . 1 4

0 . 2 9

n . o .

45 .93

3 . 7 3

lL.67

4. 06

13 .36

0 . 30

4 . 2 3

8 .81

2 .40

t . 2 8

r . 7 7

0 . 13

1 .03

98 .709 8 . 9 9

*Co l lec ted by G, E . Lo fgren .* *Co i lec ted by P. Sch i f fman.***Total iron determined as FeO.Analyses determined by XRF at Johnson Space Center;

NarO determined by INAA; FeO determined by t i trat ion.

WENDLANDT: ST]LFIDE SATURATION OF MELTS 879

Schiffman and Lofgren (1982) have determined thel-atm relations at the QFM buffer. The silicatefractions were equilibrated at the QFM buffer at900" C and l-atm for two days prior to mixing withthe sulfide fraction.

Pyrrhotite was synthesized in evacuated silicaglass tubes at 700'C for 18 hours using Fe wire(9995% purity) and S (99.999+Vo purity). The ironsulfide composition, determined by the X-ray dif-fraction technique of Toulmin and Barton (1964), isFese35S. The desired mixtures of silicate:sulfide(60:40, by weight) were prepared by grinding inalcohol to less than 3 pm grain size and drying at120" C in an evacuated oven.

The capsule configuration in all the experimentsinvolved loading 5-10 mg of the starting mixtureinto a graphite capsule which was then loaded into aPt capsule (3 mm diameter) and sealed. This designprovided a closed system, buffered the oxygenfugacity at a known value, and prevented iron andsulfur loss to the precious metal capsule. All experi-ments were at pressures and temperatures abovethe liquidus of the bulk composition except for onerun at 20 kbar and 1300'C involving the GRB inwhich approximately 2% crystals were detected.Two immiscible liquid phases, one predominantlysilicate and the other sulfide, were observed in allruns. The sulfide melt typically segregated to form asingle spherical bleb at the bottom of the charge andquenched to a fine grained pyrrhotite aggregate.The silicate melts quenched to glass.

Compositions of the coexisting melts were deter-mined by electron microprobe analysis using troiliteand basalt glass standards. Within the limits ofresolution, no solution of silicate melt componentsin the sulfide melt, with the exception of iron, wasdetected. Multiple analyses of iron and sulfur in thesilicate melt and sulfide melt were obtained and thedata averaged; values are presented with t 1o er-rors.

The oxygen fugacity in the experiments musthave been near the C-CO-CO2-O2 buffer (Frenchand Eugster , 1965; French, 1966). The initial redoxstate of the silicate component was at QFM; thus, aslight excess of oxygen was available at run condi-tions. The C-CO-CO2-O2 buffer is about 1 log unitmore oxidizing than the IW buffer at all run condi-tions.

The sulfur fugacity in the experiments was buff-ered internally by the bulk composition and varieswith pressure and temperature. The lower limit of

possible values of /5, is constrained by the iron-troilite buffer: Using the free energy data for liquidFeS of Sharma and Chang (1979) and extrapolatedmolar volume data for sulfur liquid from West(1950) and Toulmin and Barton (1964), the minimumlog.fs. at 20 kbar and 1420" C is about -0.5, andprobably changes less than I log unit in eitherdirection between 12.5 and 30 kbar. The upper limiton /s, values in the experiments can be no greaterthan ihe total pressure (log/5, : 4.3 at 20 kbar) andis probably much less.

Experimental results

In Table 2, the experimental conditions and sulfurcontents of the silicate melts are summarized. Ex-periments were at 12.5, 20, and 30 kbar and atemperature range of 1300-1460'C' The tempera-ture and pressure dependences of sulfur contents insilicate melts are plotted in Figures I and2, respec-tively. Sulfur saturation shows a negative pressuredependence and a positive temperature dependencefor all three silicate compositions. Lowest solubili-ties are observed in the MHA which has high SiO2and low FeO, contents; highest solubilities occur inthe GIB which has low SiOz and high FeOt. There isno basis for claiming linear temperature and pres-

sure dependences other than noting that a straight

Table 2. Results of exPeriments

P( kbar)

T du ra t i on S so lub i l i t y(oc ) ( h r ) (w t%) *

(M t . Hood Andes i t e )uO(Fe .geSS)+o

Runnumber

740745734735744736

747709747737t q o

/ 5 5

757759751749

742013401380142014601420

(Grand Ronde Basal t )uo(Fe. t tUS)OO

142013001380t42014607420

(Goose I s l and Basa I t ) 60 (Fe .985s )40

t2.52020zo2030

66oR

+I . Z J

1 2 q

2020202030

72.52020202030

142013401380r42014607420

676 . 554 . 1 56

Y . J J

15 . 758 . 54 . 55

0 . 1 4 + 0 . 0 1 40 . 0 5 + 0 . 0 1 00.07 + 0 .0200.09 + 0 .0130 . 1 2 + 0 . 0 1 10 . 0 5 + 0 . 0 1 2

0 . 1 9 + 0 . 0 0 90 . 0 9 + 0 . 0 0 90 . 1 4 T o . o 1 20 . 1 5 + 0 . 0 0 70 . i 7 + 0 . 0 1 0o . t l F o . 0 1 6

0 . 2 6 + 0 . 0 1 70.17 + 0 .0090 . 1 9 + 0 . 0 1 20 . 2 2 + 0 . 0 ! 40 . 2 5 + 0 . 0 1 90 . 1 8 + 0 . 0 1 0

*Errors are + Io.

880 WENDLANDT: SULFIDE SATURATION OF MELTS

Fig. l. Temperature dependence of sulfur saturation in Mt.Hood andesite (MHA), Grande Ronde basalt (GRB) and GooseIsland basalt (GIB) at 20 kbar. The brackets are !1o.

line fits the limited number of points; in fact,Richardson and Fincham (1954) and Haughtonet al.(1974) have shown that a log sulfur concentrationyr. reciprocal temperature plot defines a straightline.

Figure 3 shows sulfur ys. FeO, contents of thesilicate melts at 1420' C for three pressures, toemphasize the opposing effects of pressure andFeO, content on sulfur saturation. At constantpressure and temperature, the sulfur fugacity andthe oxygen fugacity are fixed and, thus, the ferric/ferrous ratio is approximately constant for the threemelt compositions. Between 12.5 and 30 kbar,sulfur contents in the silicate melts decrease by lessthan 0.1 wt.Vo for all melts. The ferric/ferrous ratioalso decreases with increasing pressure but by onlya slight amount as the oxygen fugacity associatedwith the C-CO-CO2-O2 buffer increases at a lesserrate than that associated with the I-W buffer. In-creasing FeO, content of the silicate melt is accom-panied by an increase in the sulfur content ofabout0.12 wt.%.

A trend of increasing FeOl contents of the silicatemelts with increasing pressure for a given bulkcomposition is also evident in Figure 3 (emphasizedby the dashed control lines). Such trends cannot beexplained by the assumption that all iron is divalent.Even if all the iron were calculated as ferric, thedifference would not account for the "lost" iron inthe silicate melt at 12.5 kbar. Nor can it be surmisedthat iron is being incorporated into the sulfide meltto compensate for the deviation from stoichiometrictroilite: for the amount of sulfide present, the addi-tion of only 0.4 wt.Vo Fe is needed to satisfy thestoichiometry. The likely explanation is that FeO isdissolving in the sulfide melt preferentially at lowerpressures. A similar decrease in FeO solubility inFeS-rich melts with increasing pressure in the sys-tem Fe-S-O was reported by Wendlandt andHuebner (197q.

To test the suggestion that Fe-loss in the silicatemelts is related to FeO gain in the sulfide melt, massbalance calculations for iron in the sulfide melt weremade (summarized in Table 3). Averaged iron andsulfur analyses of the sulfide melts at 12.5 and 30kbar and 1420" C indicate iron in excess of thatneeded for a stoichiometric FeS melt. When oxygenis added to the excess iron in the amount appropri-ate for FeO, the summed weights are comparable tothe FeOl deficits observed in the silicate melts,suggesting that there can be appreciable FeO sol-ubility in the sulfide melts at the lower pressures.High FeO contents of the silicate melt correlate

=

lo

r l o=o

05 'o , , rJ i

" ,

20 25 30

Fig. 2. Pressure dependence of sulfur saturation in the threesilicate melts at 1420" C. Brackets are ilc.

os 'o ,,11 ,,

'o 2s 30

Fig. 3. Effect of FeO, content of the silicate melt on thesaturation sulfur content ofthe silicate melt at constant pressureand temperature for the three melts (solid lines). At fixedpressure and temperature the/6, and/5" are fixed; thus, the Fe3+/Fe2* ratio is approximately conitant. tiashed lines show the lossof FeO, component from the silicate melt at low pressure(discussed in the text). Changing temperatures at fixed pressurefor a given composition (not shown) alters the sulfur content butnot the FeO, content of the silicate melt.

WENDLANDT: SULFIDE SATURATION OF MELTS

Table 3. Calculated FeO solubilitv in sulfide melt

E8l

Run P T Xro* xs* Excess Fenumber (kbar) (oc) (m;T) (mdl ) (wt%)

Ca l cu la ted 0 w i t hexcess Fe (wt%)

Calculated FeO FeO def ic iencYso lub i l i t y (w t%) i n s i l i ca te me l t

7 ( 1

749I Z . J

30

t 2 . 530

t 2 . 530

t420t420

t4201420

7420t420

Goose I s l and Basa l t - F . . 9g5S

1 .1276 1 .0495 4 .36 1 .25t . 7277 1 .0770 2 .83 0 .81

Grand Ronde Basal t - Fe.9g5S

7 . t 3 2 t 1 . 0 9 1 2 2 . 3 2 0 . 6 6t . i 33s 1 . 1084 1 .40 0 .40

Mt. Hood Andesi te - F".985S

t . 1442 t . 7277 0 .92 0 .26t . L25e 1 .1306 ( - 0 .26 ) ( - - )

5 . 6 13 . 6 4

2 . 9 81 .80

6 . 14J . O a

3 . 5 02 . 4 9

r 0 6

0 . 4 1

74t7t7

740736

1 . 1 8( - - )

r Molar proponti-ons of Fe and S in eulfide melt.

with high FeO contents in the sulfide melt. Becausethe calculation assumes stoichiometric troiliticmelt, which would only occur in equilibrium withiron metal, minimum estimates of FeO solubility insulfide melt are obtained.

Using the experimental data for sulfur solubilityin silicate melt and the calculated data for FeOsolubility in sulfide melt, a FeS-SiO2-"FeO" dia-gram, projected from the remaining components,can be constructed. Figure 4 depicts the tielinesbetween the coexisting melts at 1420" C and 12.5and 30 kbar for the three bulk compositions. Theexpansion of the miscibility gap with increasingpressure is evident in the SiO2-rich region; theexpansion is a consequence of decreased sulfursolubility in silicate melt at high pressure. Huangand Williams (1980) previously noted an expansionof the miscibility gap in the system Fe-Si-S-O withincreasing pressure. Also with increasing pressurethe ratio of FeO, in sulfide melt to FeOl in silicatemelt decreases, causing a counter-clockwise rota-tion of the tielines. Because a similar sense ofrotation is observed for all three compositions it islikely that the neglected components have littleimpact on the phase relations, although they repre-sent a compositional variable which is not evaluatedin this study. For example, Thornber et al. (1980)have demonstrated that the ferric/ferrous ratio ofmelts is affected by bulk composition, particularlyalkali contents, at constant P, T and fs,.

Discussion

Sulfur saturation of silicate melts

A comparison of existing data with the results ofthis investigation is difficult because of different

experimental conditions and melt compositions. Inthis study it was observed that sulfur solubility atsaturation increased with temperature and FeOcontents of the silicate melt and decreased withincreasing pressure. Because fo, and, A, increasewith increasing pressure in the experimental design,it is impossible to separate their relative effects; thatis. to claim that the increase in sulfur solubility isonly a pressure effect. Although calculations of /5"and/e, at run conditions (discussed earlier) suggestonly slight changing of these variables between 12.5and 30 kbar, Haughton et al. (1974) showed that atl-atm an increase of/s" by 1 log unit decreases thesulfur solubility at saturation by approximately ll2

14200c

PROJECIED FROM,,oTHERS" (WT %)

FeO 1

Fig. 4. Tie lines between coexisting immiscible sulfide andsilicate melts at 1420" C and 12.5 and 30 kbar. Melt compositionsare projected onto the FeS-FeO-SiO2 surface from the othercomDonents.

882 WENDLANDT: SULFIDE SATURATION OF MELTS

log unit, which is well within the range of variationdeserved in this study (e.g., for GRB, log S concen-tration is -0.72 at 1420" C and 12.5 kbar, and -0.96at 1420'C and 30 kbar). Opposing this effect ofincreasing /5, is the experimental observation thatincreasing/s. enhances the sulfur solubility ar satu-ration (e.g., Flaughton et al., 1974). The magnitudeof sulfur solubilities determined in this study (0.05-0.25 wt.%) overlap analyses of solubilities in naturaloccurrences (approximately 0.1 wt.%) as well asranges of values determined in 1-atm experiments(0.04-0.32 wt.%).

This study documents an apparent negative pres-sure dependence for sulfur solubility as previouslysuggested by Helz (1977) and Huang and Williams(1980). Mysen and Popp (1980) reported much high-er sulfur solubilities (up to 1.4 wt.% S) in iron-absent melts of diopside and albite compositions inaddition to solubilities that increased with increas-ing pressure. Their observations are difficult toreconcile with those of this study; however, it mustbe noted that the oxygen fugacity of their experi-ments was buffered by the boron nitride innercapsule of their assembly at a value several log unitsbelow the IW buffer (Wendlandt et al., 1982), andnot near the C-CO-COz-Ozbuffer as intended. Thelow/6,, in conjunction with high/s, of their experi-ments, probably stabilizes the association of sulfidewith calcium in the silicate melt, accounting for theenhanced solubilities. Furthermore, because Mysenand Popp's (1980) experiments involve iron-absentsystems, increases inl5, relative tole. with increas-ing pressure probably enhance the sulfur solubilityin the silicate liquid. The effect would be analogousto that observed in iron-bearing systems beforeseparation of an immiscible FeS-rich liquid. There-fore, whereas the experimental results of Mysenand Popp (1980) and this investigation are apparent-ly discrepant, in fact, the differences can be ex-plained adequately in terms of the different experi-mental systems, and specifically, by the absence orpresence of an immiscible iron sulfide fraction.

Whether silicate melts are sulfur-saturated intheir source regions and/or when sulfide saturationoccurs during magma ascent remain significant un-knowns in our understanding of sulfur-bearing mag-matic systems. A thorough understanding requires,among other things, knowledge of sulfur solubilitiesand the sulfur saturation surface for silicate melts asa function of P, T, and melt composition, as well asknowledge of sulfur abundances in melt sourceregions. The latter information is model dependent,

sulfur abundances and distributions being a func-tion of earth accretion and core segregation pro-cesses in addition to such variables as the /e.conditions of the protoearth. Sulfur contents ofmantle xenoliths are probably not reliable indica-tors of sulfur abundances in the upper mantlebecause much of the sulfide in xenoliths can beinterpreted as having been introduced or remobi-lized by late stage processes (Pasteris, 1981).

Helz (1977) suggested, on the basis of pervasivetextural evidence of immiscible sulfide melts ineruptive basalts coupled with the experimental ob-servation of sulfide saturation defining a P-Z sur-face parallel to and very close to volatile-absentbasalt liquidi, that sulfide liquid is probably presentduring melt formation and that sulfide remains as aresidual phase in the source region. The results ofthis study confirm sulfur saturation isopleths rough-ly paralleling anhydrous basalt liquidi (e.g., Thomp-son, 1972). Depending on the P-T trajectory of anascending magma, however, it is conceivable that amagma may be saturated with respect to sulfur inthe source region but not on eruption until the latestages of crystallization. For example, a liquidusinfluenced by the presence of volatiles, H2O and/orCO2 will have a steeper slope, dPldT, than thesulfur solubility isopleths determined in this studyfor anhydrous basalt melts. Accordingly, a magmainitially sulfur-saturated at its source would beundersaturated throughout an ascent that paralleledthe slope of its liquidus or was in any way steeperthan an anhydrous basalt liquidus. Perhaps the onlydirect evidence of sulfide-saturated silicate melt atelevated pressures is that reported by Roedder(1981): He observed sulfide-saturated melt inclu-sions in olivine of dunite inclusions in basalts fromLoihi seamount and estimated (using P-V-T datafor CO2 extrapolated to 1200' C) that the inclusionswere trapped at 10-18 km depth (3-5.5 kbar).Whereas these depths are above the magma sourceregion, the evidence is supportive of sulfide-satura-tion at pressures at least intermediate between thesource region and the surface.

Residual sulfide in the source region will influ-ence chalcophile element abundances in a magmaby acting as a sink for the chalcophile elements andpreventing their enrichment in the silicate partialmelt. Inasmuch as higher Ni contents of Archaeanmafic rocks relative to modern basalts for a givenMg-number have been ascribed to the influence oftransition metal monosulfides in the source regionsof the older rocks (Arculus and Delano, l98l), it

WENDLANDT: SULFIDE SATURATION OF MELTS 883

may also be necessary to stipulate that partialmelting eliminated sulfide as a residual phase in thesource reglon.

Oxygen content of suffide melt

Naldrett (1969) has demonstrated that apprecia-ble oxygen (7-8 wt.%) can dissolve in sulfide liquidsin the system Fe-S-O at l-atm. Wendlandt andHuebner (19'79) subsequently showed that the sol-ubility of oxygen in the same system at 30 kbar isless than 1wt.Vo. Coupled with observations in thisstudy, of decreasing FeO contents in sulfide meltwith increasing pressure, it is suggested that theoxygen content of naturally occurring sulfide meltsmay be indicative of the depth of sulfur saturationor, at least, depth of equilibration.

Skinner and Peck (1969) noted that the sulfide-rich immiscible liquids coexisting with basalt in theAlae pit crater, Kilauea, contain about 4Vo oxygenby weight suggesting a near surface equilibration ofsilicate and sulfide melts. Mathez /1976) noted thatthe oxide content of immiscible sulfide blebs in avariety of submarine basalts was variable; occa-sionally coarse magnetite-sulfide intergrowths wereobserved with as much as 2.2Vo oxygen, but usuallymagnetite was very fine grained and often absentfrom the sulfide. The former textural occurrencescould be interpreted as indicating shallow depths oforigin whereas the latter imply attainment of sulfursaturation at high pressures, perhaps in the magmasource region. If the oxygen content (as magnetite)of immiscible sulfide blebs is a reliable indicator ofdepth of saturation, then the identification of oxy-gen-poor blebs lends support to suggestions ofresidual sulfide in the mantle after anatexis andsome melts may remain sulfur-saturated fromsource to surface.

A further application of the sulfide melt composi-tions observed in this study is to core segregationhypotheses. Because the density of the earth's coreis approximately llVo less than that of Fe-Ni at theappropriate pressures and temperatures (Birch,1964), the presence of as much as 20Vo of anadditional element is required. Brett and Bell (1969)proposed that sulfur was a possibility by virtue ofthe low temperature of the Fe-FeS eutectic over alarge pressure interval. Ringwood (1977 a,b) arguedagainst sulfur as the light element in the core on thegrounds that the earth would have had to accrete asmuch as one-half the primordial abundance. Heargued instead that oxygen was a better candidateby virtue of presumed large oxygen solubilities in

iron melt at core pressures and temperatures.Wendlandt and Huebner (1979), comparing the sys-tem Fe-S-O at upper mantle pressures with that atl-atm (Hilty and Crafts, 1952; Naldrett, 1969),observed a decrease in oxygen solubility in theeutectic melt from about 8 wt.Vo at low pressure toless than I wt.Vo at 30 kbar. They surmised that ifthe system Fe-S-O was applicable then sulfur couldnot be discounted as the light element in the core.Huang and Williams (1980) investigated portions ofthe system Fe-Si-S-O at 30 kbar, showing that themiscibility gap between silicate and sulfide meltsexpands with pressure; they predicted that thesinking of a FeS-rich melt through the mantle wouldcarry very little oxygen and silicon with it. Theresults of the current investigation in systems ofgreater geologic complexity support both previousobservations. Although oxygen or silicon in thecore cannot be discounted, ifearly terrestrial differ-entiation involved a magma ocean (e.g., Anderson,1981) then it is likely that a large reservoir of sulfurexists in the lower mantle if not in the core.

Conclusions

Analyses of coexisting immiscible sulfide andsilicate melts quenched from experiments between12.5-30 kbar and 1300-1460'C indicate that: (1)temperature and pressure have opposing effects onthe sulfur contents of silicate melts at saturation;with increasing temperature sulfur contents in-crease whereas with increasing pressure (and in-creasing /e. and /5) sulfur contents decrease; (2)high FeOt contents of silicate melts enhance sulfursolubilities; and (3) the FeO solubility in sulfidemelt decreases with increasing pressure.

These observations suggest that immiscible sul-fide blebs occurring in erupted magmas may havesegregated at either high or low pressures, or both:the P-T paths of ascending magmas and the slopesof the magmas' liquidi are important variables indetermining when sulfur saturation occurs. A fur-ther implication of the data is that if planetarydifferentiation processes involve magma oceans,then it is likely that a reservoir of sulfide segregatesinto the lower mantle if not into planetary cores.This sulfide melt will be oxygen and silicon depletedat high pressures.

AcknowledgmentsThe author is grateful for critical reviews by Wendy Harrison,

Marc Norman, Jill Dill Pasteris, and Robert Popp. The assist-ance of O. Mullins and the use of the experimental petrology labat the Johnson Space Center is appreciated. The research

884 WENDLANDT: SULFIDE SATURATION OF MELTS

reported in this paper was performed at the Lunar and planetaryInstitute (operated by the Universities Space Research Associa-tion under Contract No. NASW 3389 with the National Aeronau-tics and Space Administration). This paper is Lunar and plane-tary Institute Contribution No. 467

ReferencesAnderson, A. T. (1974) Chlorine, sulfur, and water in magmas

and oceans. Geological Society of America Bulletin, 85, 1485-1492.

Anderson, D. L. (1981) Hotspots, basalts, and the evolution ofthe mantle. Science, 213,82-89.

Arculus, R. J. and Delano, J. W. (1981) Siderophile elementabundances in the upper mantle: evidence for a sulfide signa-ture and equilibrium with the core. Geochimica et Cosmo-chimica Acta. 45. 1331-1343.

Birch, Francis (1964) Density and composition of mantle andcore. Journal of Geophysical Research, 69, 4i77-43,88.

Brett, Robin and Bell, P. M. (1969) Melting relations in the Fe-rich portion of the Fe-FeS system at 30 kb pressure. Earth andPlanetary Science Letters, 6, 479-482.

Buchanan, D. L. and Nolan, Jack (1979) Solubility of sulfur andsulfide immiscibility in syntheric tholeiitic melts and theirrelevance to Bushveld-complex rocks. Canadian Mineralogist,t'7, 483-494.

Campbell, l. H. (1977) A study of macro-rhythmic layering andcumulate processes in the Jimberlana intrusion, Western Aus-tralia. Part I: The upper layered series. Journal of petrology,18, r83-215.

Czamanske, G. K. and Moore, J. G. (1977) Composition andphase chemistry of sulfide globules in basalt from the Mid-Atlantic Ridge rift valley near 37'N lat. Geological Society ofAmerica Bulletin. 88. 587-599.

Danckwerth, P. A., Hess, P. C. and Rutherford, M. J. (1979)The solubility of sulfur in high-TiO, mare basalrs. proceedingsLunar and Planetary Science Conference, l0th, 517-530.

Duke, J. M. and Naldrett, A. J. (1978) A numerical model of thefractionation of olivine and molten sulfide from komatiitemagma. Earth and Planetary Science Letters, 39,255-266.

Eggler, D. H. and Burnham, C. W. (1973) Crystallization andfractionation trends in the system andesite-H2O-CO2-O2 atpressures to l0 kb. Geological Society of America Bulletin, 84,2517-2532.

Fincham, C. J. B. and Richardson, F. D. (1954) The behavior ofsulphur in silicate and aluminate melts. Proceedings of theRoyal Society of London, 223A, 40-61.

Ffeet, M. E., MacRae, N. D. and Herzberg, C.'1. (1977) parti-tion of nickel between olivine and sulfide: A test for immisci-ble sulfide liquids. Contributions to Mineralogy and petrology,65, t9l-197.

Fleet, M. E., MacRae, N. D. and Osborne, M. D. (1981) Thepartition of nickel between olivine, magma and immisciblesulfide liquid. Chemical Geology, 32, ll9-127.

French, B. M. (1966) Some geological implications of equilibri-um between graphite and a C-H-O gas phase at high tempera-tures and pressures. Reviews of Geophysics and Space phys-ics. 4.223-253.

French, B. M. and Eugster, H. P. (1965) Experimental control ofoxygen fugacities by graphite-gas equilibrium. Journal ofGeo-physical Research, 70, 1529-1539.

Haughton, D. R., Roeder, P. L. and Skinner, B. J. (1974) Sol-ubility of sulfur in mafic magmas. Economic Geology, 69,451-

467.Helz, G. R. and Wyllie, P. J. (1979) Liquidus relationships in the

system CaCO3-Ca(OH)z-CaS and the solubility of sulfur incarbonatite magmas. Geochimica et Cosmochimica Acta, 43,259-265.

Helz, R. T. (1977) Determination of the P-T dependence of thefirst appearance of FeS-rich liquid in natural basalts to 20 kb.(abstr.) Transactions American Geophysical Union, EOS, 58,523.

Helz, R. T. (1978) The petrogenesis of the Ice Harbor member,Columbia Plateau, Washington: A chemical and experimentalstudy. Ph.D. Thesis, The Pennsylvania State University.

Hilty, D. C. and Crafts, Walter (1952) Liquidus surface of theFe-S-O system. Journal of Metals, 4, 1107-1312.

Huang, W.-L. and Williams, R. J. (1980) Melting relations ofponions of the system Fe-S-Si-O to 32 kb with implication tothe nature of the mantle-core boundary. (abstr.) Lunar andPlanetary Science XI, Lunar and Planetary Institute, Houston,486-488.

Katsura, Takashi and Nagashima, Shigeru (1974) Solubility ofsulfur in some magmas at I atmosphere. Geochimica etCosmochimica Acta, 38, 517-531.

Maclean, W. H. (1969) Liquidus phase relations in the FeS-FeO-Fe3Oa-SiO2 system and their application in geology.Economic Geology, 64, 865-884.

Maclean, W. H. and Shimazaki, Hidehiko (1976) The partitionof Co, Ni, Cu, and Zn between sulfide and silicate liquids.Economic Geology, 7 l, 1049-1057.

Mathez, E. A. (1976) Sulfur solubility and magmatic sulfides insubmarine basalt glass. Journal of Geophysical Research, 81,4269-4276.

Medvedev, A. Ya. and Al'mukhamedov, A. I. (1978) Partition ofchalcophile elements between sulfide and silicate melts, asshown by experimental data. Doklady Akademii Nauk SSSR,240,7 t8J2 t .

Moore, J. G. and Fabbi, B. P. (1971) An estimare of the juvenilesulfur content of basalt. Contributions to Mineralogy andPetrology, 33, 118-127.

Mysen, B. O. (1977) Solubility of volatiles in silicate melts underthe pressure and temperature conditions of partial melting inthe upper mantle. In H. J. B. Dick, Ed., Magma Genesis,Chapman Conference on Partial Melting, p. 1-14. AmericanGeophysical Union.

Mysen, B. O. and Popp, R. K. (1980) Solubility of sulfur inCaMgSirO6 and NaAlSi3OS melts at high pressure and tem-perature with controlled /e. and /s.. American Journal ofScience. 280.78-92.

Naldrett, A. J. (1969) A portion of the system Fe-S-O between900 and 1080'C and its application to ore magmas. Journal ofPetrology, 10, 17 l-201.

Naldrett, A. J. (1973) Nickel sulfide deposits-their classifica-tion and genesis, with special emphasis on deposits ofvolcanicassociation. Canadian Institute of Mining and Metallurgy,Ottawa, Transactions, 7 6, 183-201.

Pasteris, J. D. (1981) Significance of sulfide textures and miner-alogy in mantle xenoliths. (abstr.) Geological Society of Amer-ica Abstracts with Programs, 13, 526.

Rajamani, V. and Naldrett, A. J. (1978) Partitioning of Fe, Co,Ni, and Cu between sulfide liquid and basaltic melts and thecomposition of Ni-Cu sulfide deposits. Economic Geology,71, 82-93 .

Richardson, F. D. and Fincham, C. J. B. (1954) Sulphur in

WENDLANDT: SULFIDE SATURATION OF MELTS 885

silicate and aluminate slags. Journal of the lron and SteelInst i tute. 178.4-15.

Ringwood, A. E. (1977a) Composition and origin of the earth.Research School of Earth Sciences. A.N.U.. no. 1299.

Ringwood, A. E. (1977b) Composition of the core and implica-tions for origin ofthe earth. Geochemical Journal, I l, I I 1-135.

Roedder, Edwin (1981) CO2-, sulfide-melt, and silicate-meltinclusions in olivine nodules from Loihi Seamount, Hawaii.EOS,62, 1083.

Schiffman, Peter and Lofgren, G. E. (1982) Dynamic crystalliza-tion studies on the Grande Ronde pillow basalts, CentralWashington. Journal of Geology, 90, 49J8.

Sharma, R. C. and Chang, Y. A. (1979) Thermodynamics andphase relationships of transition metal-sulfur systems: Part III.Thermodynamic properties of the Fe-S phase diagram. Metal-lurgical Transactions, l0B, 103-108.

Shima, H. and Naldrett, A. J. (1975) Solubility of sulfur in anultramafic melt and the relevance of the system Fe-S-O.Economic Geology, 70, 960-967 .

Shimazaki, Hidehiko and Clark, L. A. (1973) Liquidus relationsin the FeS-FeO-SiOrNazO system and geological implica-tions. Econornic Geology, 68, 79-96.

Skinner. B. J. and Peck. D. L. (1969) An immiscible sulfide meltfrom Hawaii. In H. D. B. Wilson, Ed., Magmatic Ore Depos-its, Economic Geology Monograph 4,310-322.

Thompson, R. N. (1972) Melting behavior of two Snake Riverlavas at pressures up to 35 kb. Carnegie Institution of Wash-

ington Year Book, 71,406-410.Thornber, C. R., Roeder, P. L. and Foster, J. R. (1980) The

effect of composition on the ferric-ferrous ratio in basalticliquids at atmospheric pressure. Geochimica et CosmochimicaActa,44, 525-532.

Toulmin, Priestley, III and Barton, P. B., Jr. (1964) A thermody-namic study of pyrite and pyrrhotite. Geochimica et Cosmo-chimica Acta. 28, Ul-671.

Ueda, Akira and Itaya, Tetsumaru (1981) Microphenocrysticpyrrhotite from dacite rocks of Satsuma-Iwojima, SouthwestKyushu, Japan and the solubility of sulfur in dacite magma.Contributions to Mineralogy and Petrology, 78,2l-26.

Wendlandt, R. F. and Huebner, J. S. (1979) Melting relations ofportions of the system Fe-S-O at high pressures and applica-tions to the composition ofthe earth's core. (abstr.) Lunar andPlanetary Science X, Lunar and Planetary Institute, Houston,1f29-r331.

Wendlandt, R. F., Huebner, J. S. and Harrison, W. J. (1982) theredox potential ofboron nitride and implications for its use asa crucible material in experimental petrology. American Min-eralogist, 67, 170-174.

West, J. R. (1950) Thermodynamic properties of sulfur. Industri-al and Engineering Chemistry , 42,713-718.

Manuscript received, January 22, 1982;accepted for publication, June 7, 1982.