Matjuschkin, V., Blundy, J. D., & Brooker, R. A. (2016). The effect ofpressure on sulphur speciation in mid to deepcrustal arc magmas andimplications for the formation of porphyry copper deposits. Contributions toMinerology and Petrology, 171, [66]. DOI: 10.1007/s00410-016-1274-4

Peer reviewed version

Link to published version (if available):10.1007/s00410-016-1274-4

Link to publication record in Explore Bristol ResearchPDF-document

This is the author accepted manuscript (AAM). The final published version (version of record) is available onlinevia Springer at http://link.springer.com/article/10.1007/s00410-016-1274-4. Please refer to any applicable termsof use of the publisher.

University of Bristol - Explore Bristol ResearchGeneral rights

This document is made available in accordance with publisher policies. Please cite only the publishedversion using the reference above. Full terms of use are available:http://www.bristol.ac.uk/pure/about/ebr-terms.html

1

The effect of pressure on sulphur speciation in mid- to deep-crustal arcmagmasandimplicationsfortheformationofporphyrycopperdepositsVladimirMatjuschkin1*,JonD.Blundy1andRichardA.Brooker1SchoolofEarthSciences,UniversityofBristol,WillsMemorialBuilding,QueensRoad,BristolBS81RJ,UnitedKingdom*vladimir.matjuschkin@bristol.ac.ukvladimirgeo@googlemail.comAbstractPistoncylinderexperimentsareusedtoinvestigatetheeffectofoxygenfugacity(ƒO2)onsulphurspeciationandphaserelationsinarcmagmasat0.5to1.5GPaand840-950°C. The experimental starting composition is a synthetic trachyandesitecontaining 6.0 wt% H2O, 2880 ppm S, 1500 ppm Cl and 3880 ppm C. Redoxconditionsrangingfrom1.7logunitsbelowtheNi-NiObuffer(NNO-1.7)toNNO+4.7were imposed by solid state buffers: Co-CoO, Ni-NiO, Re-ReO2 and Hematite-Magnetite. All experiments are saturatedwith a COH-fluid. Experiments producedcrystal-bearing trachydaciticmelts (SiO2 from60 to69wt%) forwhichmajorandvolatileelementconcentrationsweremeasured.Experimentalresultsdemonstrateapowerful effect of oxidation stateonphase relations. For example, plagioclasewasstable above NNO, but absent at more reduced conditions. Suppression ofplagioclasestabilityproduceshigherAl2O3andCaOmelts.Thesolidsulphur-bearingphases and sulphur speciation in themelt are strong functionsof ƒO2, as expected,butalsoofpressure.At0.5GPatheanhydritestabilityfieldisintersectedatNNO≥+2,but at 1.0 and 1.5 GPa experiments at the same ƒO2 produce sulphides and thestability field of sulphatemoves towardshigher ƒO2 by~1 log unit at 1.0GPa and~1.5logunitsat1.5GPa.Asaresult,modelsthatappealtohighoxidationstateasanimportant control on the mobility of Cu (and other chalcophiles) during crustaldifferentiationmustalsoconsidertheenhancedstabilityofsulphideindeep-tomid-crustal cumulates even for relatively oxidised (NNO+2) magmas. Experimentalglasses reproduce the commonlyobservedminimum in sulphur solubilitybetweentheS2-andS6+stabilityfields.ThesolubilityminimumisnotrelatedtotheFecontent

2

(Fe2+/Fe3+ or total) of themelt. Insteadwepropose thisminimumresults fromanunidentified,butrelativelyinsoluble,Sspeciesofintermediateoxidationstate.Keywords: sulphur solubility and speciation, oxygen fugacity, phase relations,hydrousmagma,porphyrycopperdeposit1.Introduction

The vast majority of porphyry copper deposits (PCDs) form in metallogenic beltssub-paralleltosubduction-relatedvolcanicarcs(e.g.CamusandDilles2001;Sillitoe2012).Asaresult,therehasbeenconsiderablespeculationthattheiroriginmightbea consequence of the oxidized nature of arc magmas and/or the sub-arc mantleenvironment(seeFig.1a),possiblyinfluencedbytheincreasedstabilityofsulphaterelativetosulphides(e.g.Mungall2002;Jugo2009;Sato2012).Copperitselfismuchless sensitive to variable oxidation state and is monovalent for most magmaticconditions(CandelaandHolland1984).However,copperisconsideredachalcophileelement1 and can form solid sulphide phases or partition strongly into immisciblesulphideliquids.Assulphurismultivalent,itsspeciationisstronglycontrolledbytheoxidationstateofmagma,oftenexpressed in termsof theoxygen fugacity (ƒO2).Atreduced conditions Fe-sulphide saturation during magmatic differentiationsequesters copper into a sulphide-bearing cumulate assemblages, depleting theevolved silicate melts in copper. At more oxidized conditions, calcium sulphate(anhydrite) forms instead of sulphides. Copper does not partition into anhydrite,consequentlyinoxidisedsystemscopperbehavesasanincompatibleelementduringdifferentiation,becomingenrichedintheevolvingsilicatemelt.Previousexperimentalstudiesonbasaltsat200MPa(Jugoetal.2010;Botcharnikovetal.2010)suggestthatthetransitionfromsulphidetosulphatestability(andS2-toSO42- species in the melt) occurs as ƒO2 is increased from the nickel-nickel oxidebuffer(NNO)to1.5logunitsabove,suchthatsulphateistheonlystableS-speciesatƒO2>NNO+1.5 (Fig. 1b).This is at thehighendof the ƒO2 rangeobserved inmostbasalts eruptedat theEarth's surface, butwithin the rangeof arcmagmas (Figure1a).Atagivenpressure,theƒO2ofthesulphide-sulphatetransitionhasbeenshown

1Technicallychalcophilemeans“copper-loving”,althoughitiswidelyusedtoindicatea“sulphur-loving”(thiophile)elementandthatisthemeaningintendedhere

3

toincreaseslightlywithincreasingsilica(by~0.1logunits)foratrachyandesiteandbyafurther0.5logunitsforrhyolite(MassotaandKeppler2015),althoughCarrollandRutherford(1987)reportanhydritestabilitybyNNO+0.8foratrachyandestite.It has also been established that oxidized conditions produce a sulphate solubility(S6+speciesinthemelt)thatis2-5timeshighercomparedtothesulphidesolubility(S2-speciesinthemelt)prevalentatreducedcondition(e.g.CarrollandRutherford1987;Luhr1990;MassotaandKeppler2015).Thehighersulphursolubilityandtheinhibition of dense sulphide precipitation implies that high oxidation state canpromotebothSandCuenrichmentinevolvingsilicatemeltsfavouringtheeventualformationofporphyrycopperdeposits(e.g.Mungall2002;Sunetal.2013).Onemainlineofevidencethatarcmagmasareonaveragemoreoxidisedthanthoseproducedatmid-oceanridgesorinothersettings(Fig.1a)isbasedonacomparisonofFe3+andFe2+ratiosinlavaseruptedattheEarth’ssurface(e.g.Carmichael1991;Ballhaus1993;KelleyandCottrell2009;Leeetal.2010,Foley2011).Thisisfurthersupportedbytheoccurrenceofanhydriteinmanyoxidized,eruptedarcmagmasuchas Pinatubo, Philippines (Bernard et al. 1991) or El Chichón,Mexico (Luhr, 1990).However,thereisconsiderabledebateregardingthemechanismthatproducestheseoxidizedconditions(e.g.KelleyandCottrell2009,2012;Humphreysetal.2015). Ithas long been believed that the transfer of hydrous fluids (± sulphur) from the‘oxidized’subductingslabwastheprimarycauseandtheadditionofthisfluidtothemantlewedgemustleadtoahigherfO2forthesub-arcmantlecomparedwithmore‘normal’mantle that is the sourceofMORbasalts (Evansetal.2012;CrabtreeandLange2012).Thisviewissupportedbycomparisonbetweenmantlexenolithsfromthevarioustectonicenvironments,alsoshowninFig.1a(Woodetal.1990;BrandonandDraper 1992; Johnson et al. 1996; Parkinson andArculus 1999).However, analternate view has been put forward recently based on redox-sensitive elementpartitioningorisotopefractionationsthatareinconsistentwithmeltgenerationfromanoxidized,sub-arcmantlesource(seeFig.1a;Dauphasetal.2009;Leeetal.2005;2010; Mallmann and O’Neill 2009). The proponents of this view reason that arcmagmasmustacquiretheiroxidizedcharacterenroutefromtheirmantlesourcetothe Earth's surface. In particular they appeal to major differences in magmaticprocessesandtimescalesbetweendryMORBandvolatile-richarcsystems.

4

AsnotedbyHumphreysetal. (2015)an intriguingpoint in supportofa change inoxidation state during crustal processes is the occurrence of Cu-Fe-sulphideinclusions in early-stage phenocrysts (or ‘antecrysts’) that are found in magmaseruptedwithanoxidationstateapparentlytoohightostabilizesulphidephases(e.g.ShiveluchVolcano,Kamchatka,Humphreysetal.2006;Santiaguito,Guatemala,Scottetal.2013;Satsuma-Iwojima,Japan,UedaandItaya1981;andPopocatépetl,Mexico,Schaafetal.2005),andsometimesalsointheassociatedcumulates(Leeetal.2012).ThisevenseemstobetrueforMountPinatubo,wheresulphateisclearlythestablesulphur-bearing phase in the final erupted magma (Pallister et al. 1996; Hattori1993).Humphreysetal.(2015)suggestthisbehaviormightberelatedtoachangeinwateractivity(aH2O),butnotethatmagnetitecrystallisationhasbeenproposedasan alternative option (Carmichael andGhiorso 1986; Sun et al. 2004; Jenner et al.2010), leading to a complex interplay between Fe3+/Fe2+ and S6+/S2- ratios in themelt(Sunetal.2004;MorettiandPapale2004).Unfortunately,ourunderstandingofthe relationship between ƒO2 and sulphur oxidation state is limited to relativelyshallow conditions (<500 MPa) making interpretation of any early, deeperdifferentiationprocessesverydifficult.The aim of this study is to explore the behaviour of sulphur at high-pressureconditions representing lower to middle crustal magmatic processes. We useexperimentsperformedinpistoncylinderapparatusandƒO2conditionsimposedbysolid-state buffers (Fig. 1b); Cobalt-Cobalt Oxide (Co-CoO), Nickel-Nickel Oxide(NNO), Rhenium-Rhenium Oxide (Re-ReO2), and, Haematite-Magnetite (H-M orFe2O3-Fe3O4). In particular we determine the ƒO2 required to transition from thesulphide-dominated to thesulphate-onlystability fieldsatpressuresup to1.5GPa,byvaryingtheƒO2fromNNO-1.5toNNO+2andinsomecasesuptoNNO+4.7.Thisencompassesalmost theentire ƒO2rangeofsubductionzonemagmas(Fig1a).Theinfluence of deep crustal sulphur speciation on Cu mobility and subsequent PCDformation is assessed. In addition we present some basic textural observations,phase relations and melt composition for each ƒO2 investigated, to identifydifferences between oxidized and reduced fractionation trends in hydrous arcmagmas.2.Experimentalmethodology

5

2.1StartingcompositionThe starting composition (Table1)wasprepared fromvarious syntheticmaterials(albite,anorthite,orthoclase,fayalite,quartz,rutileandhematite).Itschemistry(onanavolatile-freebasis)approximatesthatofaphyricbasaltictrachyandesite‘FM37’(Freyetal.1984)fromLagunadelMaule,anactivevolcaniccalderacomplexontheChile-ArgentinaboarderintheAndeanSouthernVolcanicZone.Theratiooffayalite(Fe2+) to hematite (Fe3+),was used to fix the Fe3+/∑Fe of the startingmix at~0.3corresponding to ∆NNO ≈ +1.2 at 200 MPa and 1000 °C (Carmichael 1991). Thestartingmixwasdopedwith1500ppmCladdedasNaCland2880ppmS,onehalfofwhichwasaddedasFeS(S2-)andtheotherasCaSO4•2H2O(S6+). Inaddition,3800ppmcarbonwasaddedasCaCO3bycontrolledreactionofCa(OH)2withCO2 inair.Thisamountofcarbonisatthehighendofconcentrationsexpectedforarcmagmas(seeLowenstern2001;Blundyetal.2010),butwasdesignedtoensureexcessfluidwaspresentinallruns(seeResults).Finalconcentrationsofcarbonandsulphurinthe starting mix were determined using an ELTRA CS800 analyser at LeibnizUniversität Hannover. In order to examine the homogeneity of S and C, threedifferentbatchesofstartingmaterialwereanalysed;theconcentrationsareidenticalwithin the measurement uncertainty (±30 ppm). Water contents of 6 wt.% wereadded as hydroxides Mg(OH)2, Al(OH)3, Ca(OH)2. Batches of 10-20 capsules werepreparedatonetime,whilsttheremainingmaterialwasstoredinasealedglassvialtopreventfurtherreactionwithatmosphericCO2.2.2Experimentalset-upHigh-pressure experiments were carried out in an end-loaded piston cylinderapparatusatUniversityofBristol. Runsat1.0and1.5GPawereperformedusinga12.5mm salt-Pyrex assembly, whilst the run at 0.5 GPa (nominal) used a 19mm"Kushiro" talc-Pyrex assembly (see McDade et al. 2002). Capsule configurationfollowed the design of Matjuschkin et al. (2015) to maintain the required ƒO2.Informationaboutquench-rate,attainmentandmaintenanceofƒO2equilibriumandits high accuracy and precision, as well as detailed sketches of the experimentalsetupareprovidedinMatjuschkinetal.(2015).Thetriplecapsuleassemblyconsistsof (i) an outer 10 mm-long gold outer capsule (Ø=3 mm) containing crushablealuminapowderand6µlH2O,(ii)aninner5mm-longplatinumcapsule(Ø=2mm)

6

containingsolidredoxbufferand3µlH2Oand(iii)aninner3mm-longgoldcapsule(Ø=2mm)containing5mgofstartingmaterial.ThereasonforsuchasmallamountofstartingmaterialwastoensureearlyachievementofƒO2equilibriumbetweenthebufferandthesample(seeMatjuschkinetal.2015,experimentMTB/RRO-24L).Allcapsules were pre-annealed at 1000°C for about 1 hour, then loaded with therelevant material, wrapped in a wet paper tissue, cooled with freezer spray andweldedshutusingamicro-spotwelder.Finally, thiscapsuleassemblywascheckedforwaterleakagebyweighingbeforeandafterstoringina200°Cdryingoven.Thecapsule assembly was placed into a solid Pyrex sleeve, with any space filled withPyrexpowderandø=3mm,0.1mm-thicksolidPyrexdiscswereplacedatbothends.Pyrex isused inorder tominimizebothhydrogenexchangewith thepressure celland carbon infiltration from the furnace (Brooker et al. 1998; Matjuschkin et al.2015). Experimentswerecarriedoutat0.5,1and1.5GPaandtemperatures from840-950 °C. Further experiments at temperatures >975° all failed, probably as aresult of the formation of an immiscible sulphide melt that attacks the capsulematerial, unlike at lower temperature where a solid sulphide phase forms. Theresulting loss of sulphur in failed experiments produced a strong H2S smell whenopening the outer Au capsule. Further attempts to run experiments replacing AuwithAuPdcapsulesfailedduetoformationofPdS,whichremovesalmosttheentiresulphur content of the experiment within <1 hour (even shorter than the <6 hrreported by Jugo et al. 2005). It appears unlikely that shorter runs would reachequilibrium,especiallyusingourparticularsolid-statebuffertechnique.Ourapproachwas to control the redox stateof eachexperimental chargebyusingthe different solid redox buffers described above (Fig. 1b). Buffer materials wereplaced into Pt capsules in proportion 1molmetal to 1molmetal oxide, with theexceptionofthemostreducedCo-CoObufferedexperiments,wherethedecisiontouseonlyCometalwas justifiedbyrelativehighƒO2differencebetweenthestartingmaterial and buffer itself. Pt is used for its high melting point and its high H2permeabilityallowingthebuffertorapidlyimposetherequiredƒO2(morecorrectly,theƒH2).Experimental temperatures were monitored with type D W3Re97-W25Re75thermocouples. These thermocouples showed evidence of oxidation in the HMexperimentsleadingtoerroneoustemperatures.Forthoserunsthetemperaturewas

7

controlledbythepoweroutputalone.Theactualtemperatureoftheseexperimentswas cross-checked using the hornblende-plagioclase thermometer of Holland andBlundy(1994)appliedtotherunproducts.Pressurewas controlledautomaticallywithinof±3MPaof the setpoint anda3%friction correction was applied following McDade et al. (2002) for the 12.5 mmassembly (1.0 and1.5GPa). The19mmassembly is not calibrated at 0.5GPa andalthoughthereisnocorrectionrequiredat1.0GPa(McDadeetal.2002)it is likelythe true pressure was as low as 0.43 GPa (see below), so the quoted 0.5 GPa is'nominal'.Rundurationswere~4days.Experimentswerequenchedbycutting-offpowertothefurnace.Recoveredcapsuleassemblieswereembeddedinepoxyresin,piercedwitha sharpblade toensure thepresenceofwater in theouterand innercapsules, then groundwith silicon carbide and finally polished, all using oil ratherthan water-based lubricants to prevent dissolution of any run-product sulphateminerals. Redox bufferswere examined and only runswith bothmetal andmetaloxidecomponentspresentattheconclusionoftherunarereportedhere.2.3FluidcompositionandexperimentalredoxconditionsInourexperimentstheƒO2ofthesampleisonlyequaltotheƒO2ofthebufferiftheactivityofwater(aH2O )inthesampleis1.Otherwise,thedeviationisgivenby

∆logƒO2 = 2 log(aH2Osample ) (e.g.,Sissonetal.2005;Matjuschkinetal.2015).

Inthisstudy,allrunproductsarefluid-saturated,butduetothepresenceofCO2in

thestartingmix,the aH2Osample <1 andmustbeestimatedfrom

aH2O =ƒH2Ofl ⋅γH2O

fl

ƒH2O0

where ƒH2Ofl ,γH2O

fl arefugacityandactivitycoefficientofH2Oinmixedfluidand ƒH2O0 is

the fugacity of pure H2O at pressure and temperature of interest. ƒH2Ofl can be

calculatedforeachrunvia

ƒH2Ofl = P ⋅X fl

H2O⋅ ΓH2O ,

where P is pressure; X flH2O is the mol fraction of water in fluid; and ΓH2O is the

fugacity coefficient of H2O at pressure, temperature conditions. The latter wascalculatedusingtheequationofstate(EOS)ofChurakovandGottschalk(2003) for

8

eachexperimentalpressureandtemperature.Dependingontheoxygenbufferused,XH2O

and aH2O influidwereeithercalculatedbymassbalance,estimatedfromH2O-

CO2 solubility relationships, or determined from phase diagrams provided byWoermann and Rosenhauer (1985). The following volatile components werecalculatedtobestableatthevariousexperimentalrunconditions:H2,H2O,CO2,CO,CH4,SO2andH2S.Concentrationsofotherelementoxides(e.g.SiO2,Al2O3) in fluidsarenotwellknownforoursystemattheseconditions(NewtonandManning2008;Sverjenskyetal.2014),butarethoughttoberelativelylow,giventhat aSiO2 and aAl2O3inrunsarelessthanone(quartzandcorundumarenotpresent).Nevertheless,errorbarsonƒO2valueswereextrapolatedinaconservativemannerandgivesomeideaoftheeffectofminorfluidcomponents(incl.S,Cl).2.3.1RunsbufferedatRe-ReO2andH-M.

The Churakov and Gottschalk (2003) model suggests that CO2 and H2O are thedominantconstituentsoffluidsatƒO2≥NNO+2.Hence,themasses(m)ofH2OandCO2in fluid (fl) canbeestimatedbymassbalancecalculationsknowing: theamountofH2O and CO2 added to the startingmix (st); the concentrations of H2O and CO2 inquenchedglasses(gl);theamountofH2Oinhydrousphases(hydr);andthefractions(X)ofmeltandhydrousphases:

mH2Ofl =mH2O

st −mH2Ogl ⋅Xgl −∑mH2O

hydr ⋅Xhydr (1)

mCO2fl =mCO2

st −mCO2gl ⋅Xgl (2)

XH2Ofl / XCO2

fl = (mH2Ofl /mCO2

fl ) ⋅2.44 (3)

Thefractionsofminerals,meltandfluid foreachrunwerecalculatedusingMINSQ(HermannandBerry2002).2.3.2RunsbufferedbyNi-NiOandCo-CoO.

Equilibrium reactions at ƒO2≤NNO suggest the composition of COH-fluid shiftstowardsincreasing ƒH2O , ƒCH4 , ƒCO , aC withdecreasing ƒCO2 owingtoincreasing ƒH2inthesample:C+2H2O=CO2+2H2 (4)

whereK4 = (ƒCO2 ⋅ ƒH22 ) / (aC ⋅ ƒH2O

2 )

9

C+2H2=CH4 (5)

whereK5 = (ƒCH4 ) / (aC ⋅ ƒH22 )

C+CO2=2CO (6)

whereK6 = (ƒCO2 ) / (aC ⋅ ƒCO2

2 )

K4,K5andK6areequilibriumconstantsofreactions4,5and6respectively.AccordingtoEOScalculationsforfluidcompositionsstableat ƒH2 correspondingto

NNO buffer conditions, the total mol fractions of CH4 and CO in the fluid do notexceed 0.05%. These results are corroborated by high-pressure studies of Luth(1989) and Armstrong et al. (2015), who demonstrated that CH4 only becomessignificant at ƒO2 below NNO-3. Although CO2 and H2O are still the dominantconstituents of these COH-fluids, theirmol fractions cannot be calculated bymassbalanceasaresultofCO2breakingdowntographiteandH2Oviareaction4.Hence,

molar fractions of CO2 and H2O in the fluid were obtained from XH2Ofl isobars

estimatedfromCO2-H2OsolubilitycurvescalculatedatƒO2≥NNO+2(seeResults).Themost reduced experiments performed at Co-CoO buffer conditions differ fromotherrunsintheirproximitytothegraphite-C-H-Oequilibriumfield(UlmerandLuth1991) leading to extensive crystallization of graphite. In those runs the activity ofH2O was determined from graphite-fluid equilibrium diagrams of Woermann andRosenhauer (1985). However, the amount of CH4 and CO in these fluids is moredifficulttocalculate.2.4AnalyticalMethods2.4.1Electronmicroprobe(EPMA)

CompositionsofexperimentalrunproductsweredeterminedusingaCamecaSX-100electron microprobe. Glasses were analysed with 10 µm defocused 4 nA beam,operating at 15 kV accelerating voltage. Minerals were analysed with 1µm beamoperatingat15kVand10nA.AllelementsweremeasuredontheirKaemissionlineswith10secondsintegrationtimeforNa,Si,K,Al;20secondsforMgandCa,20or30secondsforTi,Fe,Cl,S.Theintegrationtimeforlowandhighbackgroundswassetathalfthetimeofthepeakvalue.Albite,Mg-richolivine,sanidine,wollastonite,BaSO4(sulphate),FeS2(sulphide),ilmeniteandNaClwereusedasstandards.ThecalculateddetectionlimitforSinglassesattheanalyticalconditionsisabout90ppm.Someof

10

thelowerSmeasurementsinreducedconditionexperimentswerecross-checkedat20 kV and 50 nA. For one particular NNO run at 1.0GPa, 900°C the S value wasconfirmedtobearoundthedetectionlimitof90ppm.2.4.2Secondaryionmassspectrometry(SIMS)Secondaryionmassspectrometry(SIMS)analysisofH2OandCO2

(as1Hand12C)inthe glasseswas carriedoutusing theNERCCameca IMS4-f ionmicroprobe at theUniversityofEdinburghwithaprimarybeamofnegative16O-ions,abeamcurrentof5nAandatotalimpactenergyof14.5keV.CO2wasmeasuredwithamassresolutionof1200todifferentiatebetween24Mg2+and12C+.Filteringofthesecondarybeamledto detection of only ions with 50±20 eV. H2O analyses on the same spots wereperformedseparatelyat lowermassresolution(500)withanenergywindowof75±20 eV.Working curves were established using calibrations on basaltic, andesiticandrhyoliticglassstandardswithvolatilecontentsupto10.5wt%H2Oand1.1wt%CO2.3.Results

Experimentalconditions,compositionoffluids,andphaseproportionsarepresentedinTable2.CompositionsofmineralphasesandquenchedglassesarelistedinTable3. The euhedral habit of solid phases, lack of observable zoning, and presence ofmetal+oxideinthebufferallsuggesttherunsreachedequilibriuminthe3-4dayrunduration(therequiredequilibriumtime isdiscussedat length inMatjuschkinetal.2015).Althoughnoreversalswereattempted,theclearshiftofthesamplefromtheloaded ƒO2 and S2-/S6+ ratio to both higher and lower values also gives confidencethatequilibriumwasachieved.3.1RunProducts3.1.1Textures.Representative images of run products are presented in figure 2a-c. All runs havecrystalsandglass(quenchedsilicatemelt),andarefluid-saturated.Thepresenceofexcess fluid is indicatedby vesicles (bubbles),whosedistribution in the capsule isnothomogeneous and showsa clear tendency tonucleatenear crystals or capsule

11

walls,sometimesformingvoids.Intheoxidized(HM)runsthevesiclesarebigger(upto50µm),whilstinthegraphite-bearingrunstheyarerarely>5µm.Withincreasingtemperaturemeltfractionincreasesandfluidfractiondecreases,asbestseeninFig.2d. The size of mineral grains varies from 5 to ~35 µm (Fig. 2a,b). Silicates andoxidesexhibit euhedral/subhedral crystal formsandarehomogeneous in chemicalcomposition.Sulphidesoccurasanhedral,roundedorsubangular1-3µmgrainsandoccasionallyasbigger5-20µmaggregates(Fig.2a).MostoftheEPMAanalysesmadeon small sulphide grains had to be screened for contamination by neighbouringcrystals or melt. Anhydrite forms ~10 µm long and 0.5-1 µm wide lath-shapedcrystals (Fig. 2b), that are difficult to analyse. Retaining anhydrite during samplepreparation requires special attention and it can easily be lost. The anhydritecomposition is not reported here but its presencewas confirmedby EDS analysesand Raman spectroscopy. Amphibole and plagioclase are present, sometimesforming mineral clots (Fig. 2a,c). In reduced runs, graphite forms platy, acicularcrystals that are difficult to polish and, in a similar fashion to fluids, tend to formaroundsilicatecrystalsespeciallynearthecapsulewall(Fig.2c).3.1.2PhasesRelations.

PhaserelationsareplottedasafunctionoftemperatureandƒO2at1.0GPainfigure3aandat1.5GPainfigure3b.Modalproportionsofphasesareshowninfigure4andarediscussedbelowforeachƒO2series.ThemostoxidizedexperimentswerebufferedbyHM(∆NNO≈+4.75)andonlyrunat1.0GPa.Thesearecharacterisedbyabundantamphibole,plagioclaseandmagnetite(Fig. 4). Anhydrite (CaSO4) is present, associatedwith sulphur-richmelt (seeMeltComposition). Melt fraction increases from 47 to 49 wt% with increasingtemperature,whilst the fluid fractionwas calculated to be~2.5wt% for all threeruns.Asingle0.5GParunwasperformedatRe-ReO2bufferedconditions(∆NNO≈+2)and900°C (not included in Fig. 3). The run product is similar to the HM series andconsistsofamphibole,plagioclase,magnetite,andanhydrite.However,thisrunhasahigher melt fraction (60 wt%) and higher exsolved fluid content (~3 wt%). Thesulphur content in the melt is the highest value observed in this study (see MeltComposition). It should be noted that the occurrence of anhydrite at this ƒO2 and

12

higher (HMruns) is consistentwithprevious studiesat similaror lowerpressures(Jugoetal.2010;Botcharnikovetal.2010).The higher pressure Re-ReO2 experiments produced surprising results. Thoseperformed at 1.0 GPa demonstrate the presence of amphibole, plagioclase andmagnetite,butaresaturatedwithpyrrhotite (FeS) insteadofanhydrite,whichwasunexpected.Withincreasingtemperaturefrom850to950°Cmeltfractionincreasesfrom34 to67wt%and fluid fractiondecreases from3.9 to2.2wt%.TheRe-ReO2runsat1.5GPahaveclinopyroxeneat950°C,garnetat900,950°C(andpresumablyat 850°C), but both plagioclase and amphibole are absent at 950°C. Magnetite isabsent at all temperatures owing to the presence of garnet; instead ilmenite wasfoundtobepresent,butonlyat850°C.Re-ReO2runsat1.5GPahavepyrrhotite,andlack anhydrite. Melt fractions are higher compared to 1.0 GPa runs and decreasefrom80 to47wt%withdecreasing temperature.Owing to increasing solubilityofwater and CO2 in the melt (e.g. Ghiorso and Gualda 2015), the calculated fluidfraction at 1.5 GPa is lower than at 1.0 GPa and varies from 1 to 2.5 wt% withdecreasingtemperature.The runs buffered at NNO (∆NNO ≈ 0) resemble the more oxidized runs in theirmineralassemblage,althoughmagnetite isabsentandTi-rich ilmeniteoccurs inall1.0GParuns.At1.0GPaand950°Cplagioclasecrystallisationissuppressedresultingin~14wt%highermeltfractioncomparedtomoreoxidizedrunsperformedatthesametemperatureandRe-ReO2.AfeatureoftheNNOexperimentsistheappearanceof titanite at 850°C and1.5GPa.AllNNO-buffered runshavepyrrhotite (sulphide-saturated), as expected from previously published experiments (e.g. Carroll andRutherford1987;Jugoetal.2010).Themostreducedexperimentsinthisstudy,bufferedatCo-CoO(∆NNO≈-1.7)differsignificantlyfromotherrunsintermsofmineralassemblage,crystallinityandfluidcomposition.AtthesepressuresthecalculatedCo-CoObuffershouldlieclosetothegraphite-C-O-H fluid equilibrium (C-CHO), although the latter is more pressuresensitive(seeMeltComposition).Duetothecloseproximityofthesetwobuffers,theentire suiteof1.5GPaexperimentscontainsaggregatesofaciculargraphite,butat1.0 GPa, graphite was identified only in one run performed at 900°C (Fig. 2c),suggestingother twoother runsmayhave experienced an imposed ƒO2 lying at orjustabovethegraphite-C-H-Oboundary.AfurtherdistinctivefeatureoftheCo-CoO

13

experimentsistheabsenceofplagioclaseatallpressureandtemperatureconditions,andanassociatedhighproportionofmelt,70to88wt%(Figure4).Thisisconsistentwith the observations of Freise et al. (2009) suggesting that the stability field ofplagioclase becomes more sensitive to melt water content with decreasing ƒO2.Hence, at a given pressure, temperature and melt composition, crystallisation ofplagioclase can be suppressed by reducing ƒO2. Titanite is present below 900°C atbothpressures.Ilmenitewasidentifiedfortheentiresuiteof1.0GParunsandinasinglerunat850°Cat1.5GPa.OwingtotheformationofgraphiteattheexpenseofCO2,theamountoffluidatNNO-1.7issignificantlylessthanthemoreoxidizedruns;estimatedtobe<1wt%(Fig.2e).3.1.3SolidPhaseCompositions

Amphibole and Plagioclase. Even though the experiments span a wide range oftemperaturesandƒO2,thecompositionofamphiboleappearstobesensitiveonlytopressure.Thus,Alcontentsincreasefrom1.76cationsperformulaunits(c.p.f.u.)at0.5GPa to2.25at1.0GPaand3.05at1.5GPa.With increasingpressureKandNacontents increase and Mg decreases. In comparison to 1.5 GPa runs, experimentsperformedat1.0GPademonstratelowerTiandhigherCavalues.Anorthite content of plagioclase varies from 26 to 42 mol% and increases withincreasing temperature. Temperatures calculated using the plagioclase-hornblendethermometer (Holland and Blundy 1994) are in excellent agreement withexperimental temperatures and deviate by ≤7°C with exception of one Re-ReO2experimentat1.5GPa,850°Cwheretemperatureisunderestimatedby26°C.Giventhe apparent high accuracy and precision, this thermometer was favoured to thethermocoupletemperatureinHM-bufferedrunswhere,asnotedabove,thehighƒO2conditionscausedpartialoxidationofthethermocouplewire.Clinopyroxene is limited to runsat1.5GPaand950°C.TheyareAl-Na-richaugiteswithMg#from0.64to0.75withaminimalrangeinNa,Ca,Mg,TiandFe.TotalAlcontentsrangefrom0.29to0.42c.p.f.u.andincreasewithincreasingƒO2.Garnetisonlyobservedat1.5GPa,whereitispresentinalmostallrunsapartfromthemelt-rich, higher temperature runs at the lowest ƒO2 (seeFig. 3). InNNO runs,

14

garnet is a significantphaseat950°Cwith14wt%. In contrast, at850°C it is onlypresentasaminorcomponent(<5wt%).AllgarnetshavehighFecontents(1.27to1.54 c.p.f.u.) increasingwithdecreasing temperature alongwithCa,whose contentranges from 0.6 to 1.12 c.p.f.u.. Mg contents vary from 0.52 to 1.24 c.p.f.u. andincreasewithincreasingtemperature.Oxides. At 1.0 GPa ilmenite is stable at reduced conditions (ƒO2≤NNO) and ischaracterizedbyhighTiO2/FeOratios,whilstmagnetiteisstableatƒO2≥NNO+2withamuchhigherFeOcontent.AtƒO2<NNO+2magnetiteandilmeniteareminorphasesand comprise less than 1wt% of run products. Only at NNO+4.7 is themagnetitefractionhigher,rangingfrom2.1to3.6wt%.Hematiteisnotobservedintheserunproducts as!!!! < 1 leads to a ƒO2 value slightlybelow theHMbuffer.At1.5GPailmenite is present at all ƒO2 conditions at 850°C, but not at higher temperatures,whereasmagnetiteisabsentduetothepresenceofgarnet.Similarto1.0GParuns,theFe-contentinilmeniteincreaseswithincreasingƒO2.3.1.4MeltComposition

MajorElements.Onatotalalkali-silica(TAS)diagram(LeBasetal.1986)thevolatile-free compositionof the startingmaterial is trachyandesite (57.2wt%SiO2),whilstthe compositions of quenched silicate glasses are trachydacitic, with SiO2concentrationsincreasingfrom60to69wt%withdecreasingtemperature(Fig.5).ThetotalconcentrationofNa2O+K2Orangesfrom7.4to9.3wt%andincreaseswithdecreasingtemperaturewiththeexceptionofoneCo-CoO-bufferedexperimentat1.0GPaand850°C,wherethetotalalkaliconcentrationdecreasesby0.5wt%relativetothe 900 °C run. Overall, irrespective of ƒO2, the concentrations of TiO2, Al2O3, FeO,MgO,CaOdecreasewithdecreasing temperatureowing tocrystallizationofsilicatephases (amphibole, plagioclase, garnet, clinopyroxene) and oxides (magnetite,ilmenite).AsplagioclasestabilityissensitivetoƒO2(Figs.3and4),theconcentrationsofAl2O3andCaO in themeltdemonstrateaclear tendency to followdistinctredoxtrends.At any given temperature, themost reduced runs exhibit thehighestAl2O3andCaOconcentrations,whereasinthemoreoxidizedrunsthesearelowerby1.3-2.3and0.5-1.7wt%respectively.ForrunsatintermediateƒO2(NNO)Al2O3andCaOconcentrationsliebetweentheseextremes(Fig.6).

15

DissolvedVolatiles.TheH2OandCO2contentsinsilicateglassesasanalysedbyionprobeareshowninfigure 7a for 1.0 and 1.5 GPa runs. The water content of the melt increases withdecreasing temperature owing to progressive crystallization of dominantlyanhydrous minerals, even when amphibole is stable. CO2 decreases as fluid XH2O

increases (see below). In general, oxidized melts exhibit somewhat lower H2Oconcentrationscomparedtomorereducedexperiments.ThisisrelatedtoformationofCaSO4and/orincreaseinferriciron,bothsequesteringoxygenfromthemelt,andliberating hydrogen. In reduced experiments increasing FeO content and thestabilizationofsulphidesandgraphiteleadtohigherH2Ovalues,consistentwiththeobserved destabilisation of plagioclase at lower ƒO2. As bulk water content is afunctionofƒO2.andphasecompositionitsdeviationfromruntoruncanbeestimatedbythefollowingmassbalance:

Fe2O3melt +H2 = 2FeO

melt +H2Omelt

CaSO4anhydrite +FeOmelt + 4H2

buffer = FeSsulphide +CaOmelt + 4H2Omelt

Ferric-ferrous ratios inmelt can be estimated usingKress and Carmichael (1991).Thus, thebulkH2O content in theNNO+4.8 suite of experiments is 0.5wt% lowercompared to NNO+2 and 1.45 wt% higher at NNO-1.7, assuming that for thesegraphite-saturatedrunsasubstantialpartofCO2isreducedtoC+H2Oviaequation4.The covariation of H2O with CO2 is illustrated in figure 7b and the oxidizedexperiments at 0.5 and 1.0 GPa fall approximately on isobars calculated from thesolubility model of Ghiorso and Gualda (2015). The 0.5 GPa runs yield calculatedpressuresof0.43GPa,suggestingthatthetruepressureoftheserunsislowerthantheirnominal0.5GPavalue(Pistoneetal.2016).Someofthemorereduced1.0GPaexperimentsatCo-CoOfallwellbelowtheH2O-CO2isobar.Someofthe1.5GPadatafallintheregionexpectedforhigherpressure,butagainthereducedrunsatCo-CoO(all graphite-saturated) fall well below the expected CO2 and H2O values. Thissuggestsaninsolublespecies(COorCH4)dilutesthesecomponentsinthefluid.The concentration of Cl inmelts increases slightlywith decreasing temperature inresponsetotheincreasingcrystalfraction(Fig.6).However,whentherelativemassofresidualmeltistakenintoaccount,asignificantamountofClappearstobelostto

16

thefluid.ItshouldbenotedthatwedidnotincludeP2O5inthestartingcomposition,so apatite cannot form to sequester the Cl, although some Cl is incorporated inamphiboleinconcentrationsfrombelowdetectionupto1400ppmwithdecreasingtemperature(table3).Sulphur contents of themelts (Figure6) are controlleddominantly by the ƒO2 andbuffering by solid S-bearing phases and temperature (at high ƒO2) rather than theproportion/compositionofmelt or fluid. The1.0GPa experiments atHM (∆NNO≈+4.75) have anhydrite as the stable S-bearing phase. This is associated with highdissolved sulphur contents inmelt; 390, 680 and600ppmat 841, 870 and875°Crespectively.Thesingleanhydrite-bearingRe-ReO2-bufferedrunat900°Cand0.5GPahasameltsulphurcontentof~1030ppm,which is thehighestvalueofall runs in thisstudy.TheRe-ReO2experimentsperformedat1.0GPa,andsaturatedwithpyrrhotite(FeS)insteadof the expectedanhydrite, havea sulphur content strongly increasingwithtemperaturefrom480ppmat850°Cto970ppmat950°C,suggestingsomeinfluenceofasulphatespeciesinthemelt.At1.5GPatheRe-ReO2.pyrrhotite-bearingrunshadmelts with lower sulphur contents restricted to a range from 460 to 340 ppmsuggestinglessinfluenceofanysulphatespecies.NNO-buffered experiments have the lowest S concentrations with an average of150±74 and 165±43ppm at 1.0 and 1.5 GPa respectively. This is approaching thedetection limit using EMPA and it is difficult to identify any pressure- ortemperature-relatedtrends.ExperimentalglassesattheCo-CoObufferhaveslightlyhighersulphurcontentsof217±56and315±75ppmat1.0and1.5GPa,respectively,withatendencytoincreasewithincreasingtemperature.3.1.5FluidcompositionMeasuring the H2O and CO2 contents of the melt allows an estimate of theexperimentalfluidcompositionsbymassbalance(seeMethods).ExcessC-O-Hfluidwasindicatedbythepresenceofvesiclesandvoidsinrunproducts,withsize,formandquantityvaryingwithpressure,temperatureandoxygenfugacity.Theestimatedmolar XH2O

rangesfrom0.70to0.85andincreaseswithdecreasingtemperaturedue

to crystallisationof anhydrousphases (Fig. 7a). Calculatedactivityofwater ( aH2O )

17

variesbetween0.75and0.87(seetable2);hence,theƒO2isreducedby-0.25to-0.12logunitsrelativetothebuffer.In contrast to the oxidized runs, experiments at NNO and Co-CoO approach or liewithin the graphite stability field respectively. Graphite formation is driven bycontinuous breakdown of CO2 owing to increasing ƒH2 (decreasing ƒO2) (see Luth1989).Graphitewasnot identified inanyNNOrunsor inNNO-1.7 runsat1.0GPaand850,950°C.However,itispresentintheentiresetofNNO-1.7experimentsat1.5GPa and in a single run at 900°C, 1.0 GPa. In those, the ƒO2 must lie below thegraphite-C-O-H fluid equilibrium (Ulmer andLuth1991).As carbonpartitions intographite, less is available for fluid species and XH2O

increases. Water content of

reduced fluids at NNO was reconstructed from XH2O− XCO2

solubility isopleths,

previouslydeterminedatRe-ReO2andHMbufferedconditions(ignoringanyminorCO or CH4). Fluid compositions of NNO runs are similar to those reported for∆NNO+2.At ƒO2 conditions corresponding to the Co-CoO buffer a different approach has

beenusedtodeterminethecompositionoffluid.Therunat1.0GPaand950°Clacksgraphite,which is perhapsnot surprising as the calculatedCo-CoObuffer only lies0.1±0.35log units below the graphite-COH fluid equilibrium. The lack of graphitemeanstheratioofwaterandCO2thatdominatethefluidcanbeestimatedusingthemethoddescribedforNNOrunsgiving XH2O

= 0.78 andasolubilitythatfallsonthe

1.0 GPa isobar in figure 7b. In contrast, at 900 and 850°C Co-CoO buffered runsshouldliefurtherintothegraphitestabilityfieldat0.3±0.35and0.4±0.35logunitsbelowgraphite-C-O-Hfluidequilibrium.At900°Cgraphiteisindeedobserved,whilstat850°C itwasnot identified,althoughthemeasuredamountofCO2 in themelt isless than at 950°C (see Fig.7a). Hence, the 1.0 GPa 900°C run lies in the graphitestabilityfield,whilstthe850°Crunmightlieslightlyaboveit.Theestimated aH2O at

850and900°Cis0.72(+0.18/-0.25)andwascalculatedfrom

aH2O =10(0.5∆ logƒO2 ) (Matjuschkinetal.2015),

where the ∆logƒO2 is the deviation of Co-CoO buffer from graphite-C-O-H fluidequilibrium. Overall, the measured C content in these lower temperature Co-CoObufferedmeltsismuchlowerincomparisonto950°C(Fig.7a)andwellbelowthe1.0GPa isobar in Figure 7b. This suggests that the excess fluid contains a significant

18

amountofaninsolublespecies,eitherCOorCH4,reducingtheCO2fugacity(Pawleyetal.1992;ThibaultandHolloway1994;Morizetetal.2010).The 1.5 GPa suite of experiments is 0.7 to 1 log units more reduced than the

graphite-COH buffer, contains graphite, and has an estimated water activity of~0.7±0.2.Again themeasuredC inmelt (expressedasCO2) is even lower than thevalueestimatedfromtheoxidized1.5GPaexperimentsinFig.7b,fallingwellbeloweven the 1.0GPa isopleths. This cannot be explainedby an evenhigher activity ofwaterandagainappearstoreflecthighactivitiesofinsolubleCOorCH4inthefluid.Unfortunately,itwasnotpossibletoestimatetheamountofsulphurenteringthe

fluidusingthemassbalancecalculationsduetothelargeamountofSincorporatedinsolidS-bearingphases.Thus,errors inestimatingthesmallamountsofsulphideorsulphatehavesignificanteffectonthemassbalance.However,wewereabletomakeanestimateofsulphurpartitioncoefficientbetweenfluidandmeltforasinglesuper-liquidusrun(Table1)at1.0GPa,~1120°CandNNO+4.7andwitharesultinghighScontent. This suggests a calculated Dfl/melt~0.1. Using this value for oxidizedconditions∆NNO≥±2themolfractionofSinthefluid(XS)isnegligible.Forreduced,Co-CoObufferedconditions the fluid/meltpartitioncoefficientof sulphurhasbeenreported as 468 at 0.2GPa and 850°C (Keppler 2010) and using this value ourreduced fluidSmaybeasmuchasXS<0.026mol%.Otherpartitioncoefficientsarereported in the literature (e.g. Scaillet et al. 1998), but using any of these stillsuggeststheamountofS-species inthefluidremains insignificant inthecontextofthis study. The systematics of S-partitioning in fluid as a function of ƒO2 at high-pressureconditionsrequiresfurtherexperimentalinvestigation.4.Discussion

Sulphur content and speciation in silicate melts are often used as monitors ofmagmaticƒO2conditions(e.g.Kress1997),butthis isonlyreallypossiblewhenthebulkSconcentrationareinthelowerppmrange.Asconcentrationsreachweightpercent levels, sulphur can itself act as a strong oxygen buffer rather than a sensorowing to itshighredoxcapacity, i.e. eightelectron transfer fromS2- toS6+. It is forthisreasonthatsubductedsulphatehasbeensuggestedasapotentialoxidantofthemantlewedge in sub-arc environments (Evans andTomkins2011). Bufferingwith

19

sulphuratmagmaticconditionsoperatesviavariousheterogeneousequilibriasuchas:

FeSsulphide +CaOmelt + 2.5O2 =CaSO4anhydrite +FeOmelt

3FeSsulphide +3CaSiO3wollastonite + 6.5O2 = 3CaSO4

anhydrite +Fe3O4magnetite +3SiO2

quartz

(CarrollandRutherford1987)

FeSsulphide +Fe2SiO4olivine + 2.5O2 =SO3

melt +Fe3O4spinel + SiO2

melt (Jugoetal.2005)

In this study, ƒO2 is set by an external buffer and sulphur speciation is forced toconform to the imposed ƒO2 conditions, although its concentration is in the rangeexpected for natural magmas (Wallace 2005). Redox conditions clearly affect thesolubility of sulphur (sulphate and sulphide species) in themelt, but it should benotedthatsolubilityisalsoafunctionofmeltcomposition,sulphurfugacity,activityofwaterandCO2inthemeltandpressure,temperatureconditions(MavrogenesandO’Neill1999;Liuetal.2007;Beermannetal.2011;MasottaandKeppler2015).Asaresult,ourexperimentsarenotdirectlyapplicabletoallnaturalmagmasystems,butarerelevanttosubductionzonemagmasand,moreimportantly,illustratesomekeyprinciples,asoutlinedbelow.Asallrunproductsaresaturatedeitherinpyrrhotiteoranhydrite;themeasuredsulphurcontentalsoestablishestheSulphurConcentrationatSulphideSaturation(SCSS)orAnhydriteSaturation(SCAS)foratrachydacitemeltgenerated from a basaltic trachyandesite parent by mid to lower crustaldifferentiation.4.1EffectofpressureandƒO2onsulphurspeciationandsolubility

Published experimental data on sulphur speciation in silicate melts at 0.2 GPasuggest the sulphate-only stability field is reached at∆NNO≥1.5 and the transitionfromsulphidetosulphatetakesplaceoveranarrowƒO2windowof2logunits(Jugoetal.2010;Botcharnikovetal.2010;MasottaandKeppler2015).Whilstour0.5GPaexperiment at NNO+2 (Re-ReO2 buffer) contains anhydrite consistent with thesepreviousstudies,ourhigherpressureresultsshowpyrrhotitepersistsaboveNNO+2.Evidently the sulphide stability field expands towards higher ƒO2 with increasingpressure.Inordertoreconstructthenewsulphide-sulphatestabilityfieldsat1.0and1.5GPa,knowledgeofsulphatecontentandƒO2ofthemeltforeachrunisrequired.Whereas

20

thelatterparameteriswellconstrained,ourexperimentalsamplesarenotamenabletodeterminationofsulphurspeciationusingEPMA(CarrollandRutherford1988),aslowsulphurcontentinthemeltleadstohighmeasurementuncertainties(Matthewset al. 1999)not tomentionotherproblemssuchasS-peakmigrationand sulphideoxidationwith a static beam. It also proved impossible to fit our data to the SCSSalgorithm of Mavrogenes and O’Neill (1999) and Liu et al. (2007) to calculatedissolved sulphate content by difference. The problem of this approach is theunexplained minimum in SCSS at ƒO2≈NNO±0.5, which was found in all theexperimentalseriesinthisstudyandisapparentfromthedatasetsofotherworkersin Fe-bearing systems (e.g. Katsura and Nagashima 1974; Carroll and Rutherford1985,1987;Jugoetal.2005;Clementeetal.2004;Botcharnikovetal.2010)andFe-free(e.g.NagashimaandKatsura,1973).Thisphenomenon isoftenascribedto thechanging Fe2+ content in melt, yet in our experiments (and others) it has littlecorrelation with silicate melt composition, Fe2+/Fe3+, ƒH2S or ƒS2 (Carroll andRutherford1985;Clementeetal.2004;NagashimaandKatsura1973).Onepossibleexplanation for theminimumintheSCSSvalue is the formationofan'intermediate' oxidation state sulphur-species existing in the ƒO2 range betweensulphide(S2-)andsulphate(S6+).Ifthisspecieshaslowsolubilityinthesilicatemeltandlacksanequivalentmagmaticsolidphasethenitwillpartitionstronglyintothefluidproducingaminimuminthesolubilitycurve.Moreover,ifthisspeciesispresentat low concentrations (low solubility) or is kinetically difficult to retain duringquenching,itmaybeundetectablebycommonspectroscopicmethods.Foralgebraicsimplicity in the following we consider this intermediate species to be S0 (i.e.elemental sulphur), but we stress that any relatively insoluble species with anintermediateoxidationstatebetweenS2-toS6+oxidationwillsuffice(S4+,S-orS3-,seebelowandtheexampleinFig.S3).ForthecaseofS0,equilibriumcanbedescribedbythefollowingreactionsandequilibriumconstants(K):

(A)

(B)

2H2S(m) +O2(m) = 2H2O(m) + 2S(m)

KA(P,T ) =ƒS2 ⋅ ƒH2O

2

ƒO2 ⋅ ƒH2S2 ≅

XS2 ⋅aH2O

2

ƒO2 ⋅XS2−2

2S(m) +3O2(m) = 2SO3(m)

21

whereS0(m)istheproposedintermediateoxidationstateS-species.

Iftheproportionsofallsulphurspecies,S6+,S2-orS0,arenowdefined,theratiosR6,R2andR0givetheiramountinrespecttothetotalsulphurcontentinthemeltandaredefinedasfollows:

Substitution of sulphide and sulphate parameters from equations A and B

and yield:

Since aH2O andƒO2areknownforeachrun,theratiosofS6+,S2-orS0arefunctionsof

KA and KB, which can be estimated from the experimentally obtained S-solubilitydata.CalculationofKAandKBusingthermodynamicdatabyRobieetal.(1978)andHollandandPowell(1998)yieldsconstantlogKB/logKA=1.78,withvaluesofbothKAandKBincreasingwiththedecreasingtemperature.

KB(P,T ) =ƒSO32

ƒO23 ⋅ ƒ

S02 ≅

XS6+2

ƒO23 ⋅X

S02

XS6+= KB

1/2 ƒO23/2X

S0

XS2−= KA

−1/2 ƒO2−1/2X

S0aH2O

R6 =1

1+ aH2O (KAKB )−1/2 ƒO2

−2 +KB−1/2 ƒO2

−3/2

R2 =aH2OKA

−1/2 ƒO2−1/2

1+ aH2OKA−1/2 ƒO2

−1/2 +KB1/2 ƒO2

3/2

R0 =1

1+KB1/2 ƒO2

3/2 + aH2OKA−1/2 ƒO2

−1/2

R6 =XS6+

XS6++ X

S2−+ X

S0

R2 =XS2−

XS6++ X

S2−+ X

S0

R0 =XS0

XS6++ X

S2−+ X

S0

22

Theresultingsulphurspeciationmodel isdisplayed in figure8aand illustrates thestability fields of sulphide, sulphate and intermediate sulphur species at differentpressure-temperatureconditions.ThemodelwasobtainedbyfittingbothKAandKBto measurements of S-contents in melt (Fig. 8b; for calculated KA,B values anduncertainty see supplementary materials Fig.S2). Note that calculations can beequally performed using other sulphur species with intermediate oxidation state,such asS!! or S4+. The algebraic treatment is similar and the calculated speciationcurves have broadly similar shape, although S4+ persists to a high ƒO2. OurpreliminarytestswithavarietyofS-speciessuggestthatS0givestheclosestfittoourexperimentally obtained data (for comparison S4+-fit see Fig.S3). However, otherintermediatesulphurspecies,orindeedmultipleotherspecies,couldbeinvoked.Oursulphurspeciationmodel(Fig.8c)illustratesthestrongeffectofpressureonthestabilityfieldsofsulphideandsulphate.Withincreasingpressurethestabilityfieldofsulphide expands towards oxidized conditions. Thus, the sulphate-only stabilityfields at 1.0 and 1.5 GPa now lie at NNO+3 and NNO+3.5, respectively, 1 and 1.5logƒO2unitshigherthanat0.2GPa(Jugoetal.2010,Botcharnikovetal.2010).4.2Evidenceforanintermediatesulphurspecies

Althoughthemodelpresentedtoexplaintheminimumindissolvedsulphurcontentat NNO±0.5 is based, for simplicity, on the presence of S0 as the intermediateoxidation state species, it is useful to explore other possible species. The mostobviouscandidatecurrentlyproposedintheliteratureissulphite(S4+),whichistheequivalentofSO2inthefluid,buthasnoobviousmagmaticsolidexpression,i.e.thereare no knownmagmatic sulphiteminerals. This species is predicted bymolecularsimulation (Machacek et al. 2010) and S4+ has been identified in quenched silicatemelts using XANES.Much of the initially identified S4+ inmelt inclusions has nowbeenascribed tobeam-induced reductionof S6+ (Backnaeset al. 2008;Wilkeet al.2008;Métrichetal.2009).HoweverS4+hasbeenunambiguouslyidentified,togetherwithS6+,insomeFe-freeexperimentalglassesinequilibriumwithCaSO4

at0.4to1.6GPa(Métrichetal.2009).AlthoughnotstrictlyperformedundercontrolledƒO2,theirresultssuggestthattheremaybeanƒO2

windownearNNOwhereS4+hasalow,butfinite, solubility in silicatemelts at high pressure, perhaps dissolved asmolecularSO2. S4+might be absent at higher ƒO2 (or in high-Fe samples) because only S6+ is

23

stable.AlternativelyitmayremainunidentifiedduetoanelectronexchangereactionwithFe3+(S4++2Fe3+=S6++2Fe2+)thatproceedsreadilyuponquenching(Berryetal.2006).Analternative intermediateoxidationstatesulphurspecies is thetrisulphur!!

!ion,which is known to be stable in aqueous solutions at temperatures ≥250°C andpressuresabove0.01GPabreakingdowntoS6+,S2-andnativeS0atlowerpressure-temperatureconditions(PokrovskiandDubrovinsky2011;Pokrovskietal.2015).Itremainsunclearifthetrisulphurioncanexistatmagmaticconditionsand,ifso,whatits solubility in silicatemeltsmight be. Experimental observations clearly indicate!!! coexistencewithpyrrhotite,and !!

!hasbeenshownto formanioniccomplexwithAuinaqueousfluids(Pokrovskietal.2015).InterestinglyarecentexperimentalstudybyBotcharnikov et al. (2010) reported a sharp increase in gold solubility insilicate melt over a narrow range of ƒO2 corresponding to the sulphide-sulphatetransitionat0.2GPa,whichmayindicatethestabilityofAuS3complexesinthemeltat ppb levels. Preliminarymeasurements for Au on the samples of this study alsosuggestasimilarmaximumcoincidingwithourSsolubilityminimum.ThetendencyfortrisulphurtobreakdowntootherS-speciesatlowtemperatureconditionsmightbe responsible for the formation of anhydrite-bearing,water-richmelts,which areoften found asmelt inclusions in epithermal gold deposits (e.g. Chamberfort et al.2008). Such deposits might represent a quench product of AuS3 complexes. Thepartitioning of trisulphur between fluid and silicate melt is unknown, so ourproposedeffectonSsolubilityinthemeltisspeculative.Italsoremainsunclearhowthis species might affect the composition of sulphides, which are often used toestimatesulphurfugacityatmagmaticP-Tconditions(e.g.FroeseandGunter1976;Clementeetal.2004).Notethatit isasimplemattertorecastequations(A)and(B)aboveintermsofS!!(orS4+)ratherthanS0; theessential formofFigure8willremainunchanged. If themodelled intermediate species in Fig. 8 is trisulphur this suggest a narrow ƒO2-windowofstability.4.3Implicationfortheformationofporphyrycopperdeposits

Experimental evidence for the shift of the sulphide stability field towardsoxidizedconditions with increasing pressure has consequences for the behaviour of

24

chalcophile elements in magmatic reservoirs at mid- to deep-crustal depths (≥15km). In particular, Cu is likely to be depleted even in the more oxidizedmagmasfound in arc systems, due to strongpartitioning into anydense immobile sulphidephase (sulphidemelt or solid pyrrhotite) that remains trapped in deep cumulates.ThepreventionofdeepsulphidesaturationviaoxidationalonewouldrequireeitherveryhighƒO2conditions∆NNO≥3.Thismaybeunrealisticevenforsubarcmagmas(Fig.1a,and9),althoughtheextremerequirementsofthisconditioncouldbelinkedtotherarityofPCDs.Analternativemeanstosuppresssulphideformationintherootsofvolcanicarcsisiftheparentalmagmas remain S-poor (<100ppmS). This requires that theprimarymagmas themselvesareS-poor,againanunusual situation forarcbasalts (Wallace2005)againperhapsexplainingfortherarityofPCDs.EvenifthemagmasstartwithlowS, the concentrationsare likely to increase throughextensivecrystallisation toeventually approach the critical SCSS level. A more plausible mechanism formaintaining low S contents is via degassing during early stages of magmaticdifferentiation, aswouldbe expected for arcmagmas that are initially very rich inothervolatilespecies,suchasH2Oand,particularly,CO2.Thesespecieswillexsolvetoproduceafluidphaseintowhichsulphurcanpartition(Munteanetal.2011).Theearlier amagmatic volatile phase separates during differentiation, the greater theensuingdepletioninS.Although,ourunderstandingofsulphurdegassingis limitedtoexperimentsatlowerpressures(<0.5GPa),itisknowntobeastrongfunctionofmelt composition and oxygen fugacity, with Ds fluid/melt increasing with increasingdegree of silicate melt polymerization and decreasing ƒO2 respectively (Keppler2010;WebsterandBotcharnikov2011;Zajacz2015).RegardlessoftheexactvalueofDStheamountofSquantitativelystrippedfromthemeltduringdegassingwillbeafunction of the total exsolved gas budget. Thus, unusually volatile-rich, evolvingmagmas that begin degassing at high pressures (perhaps at CO2 saturation) canmaintain dissolved S contents that are sufficiently low to preclude sulphidesaturation. The recent suggestion that arc magmas may have higher initial CO2contentsthanwouldbesuggestedonthebasisofmeltinclusionsalone(Blundyetal.2010)would favour early volatile exsolution. As is evident from our experimentalresults, preventing sulphide saturation is easier at elevated ƒO2 (>NNO), but theincreasedstabilityofsulphidespeciestohighƒO2athighpressuremeansthatitisthe

25

abundanceofothervolatiles,ratherthanƒO2,thatcouldplaythekeyroleinkeepingSbelowsulphidesaturation.We suggest that it is the combinationof veryhigh initial volatile contents coupledwiththeoxidizednatureofarcbasaltsthatprovidesthemostfavourableconditionsfor the formation of porphyry copper deposits through suppression of any earlysulphidesaturationthatwouldotherwiseremoveCu.Insomelocationsthesub-arcenvironmentfavoursboththeseconditionswithsubductionofsulphur-carbon-andwater-rich sediments releasing S and CO2-H2O fluids that dissolve in and laterexsolve from the ascending magmas buffering their ƒO2 above the graphite-C-H-Oreaction (Ulmer and Luth 1991; Connolly 2005; Fig.9). This mechanism at leastmaintainsthemeltsabovea‘minimum’ƒO2atthegraphite-CHObuffer(Fig.1a,9).Inother locations themantlemight bemore reduced and early (deep) exhaustion ofCO2mightresultinformationofgraphiteorthesubductedsedimentsarerelativelyCO2-poor.BothsituationsmightleadtotheformationofCu-poor‘barren’magmaticsystems.Ourfindingsalsoprovideapotentialsolutiontotheproblemofsulphideinclusionsbeingtrappedinearly-formedphenocrystsinmagmasthatappeartohaveahighƒO2,and sulphateas a stablemeltphase (e.g.MtPinatubo,Pallister et al. 1996;Hattori1993).Iftheearlyphenocrystsgrewathighpressure,themagmacouldhavehadaninitiallyhighƒO2(aboveNNO+2)andyetstillcrystallisesulphide(orformimmisciblesulphide melt). As the magma ascends to lower pressure anhydrite becomes thestablesulphurphasewithoutanychangeintheƒO2simplybecauseofthepressure-sensitivechangeinthesulphide-sulphateequilibrium.Inadditionanysulphidethatiscarriedbythemelt(dependingonblebsizeandascentrate)willbecomeunstableduringascentandstarttodissolveinthemelt,which,duetotheincreasingstabilityofS6+species in themelt,canalsoacceptmoretotalS.Consequently, therecordofearly,high-pressuresulphidesaturationwillonlybepreservedasisolatedinclusionsinearly-formedcrystals.5.Conclusions

High-pressurepistoncylinderexperiments reveal theeffectofpressureand ƒO2onphase relations and sulphur speciation in quenched silicate melts. At the mostreduced conditions imposed by the Co-CoO buffer (NNO -1.7) run products are

26

graphite-saturatedandexhibit lowamountsof fluid(<0.8wt%),buthighdissolvedH2OcontentsowingtobreakdownofCO2toCwhichincreasesthewateractivityofthefluid.SilicatemeltsintheseexperimentsexhibithighCaOandAl2O3contentsduetothesuppressionofplagioclasecrystallization.Thisalsoresultsinan8-30%highermelt fraction compared to more oxidized runs. Co-CoO buffered runs arecharacterisedbytheoccurrenceofTi-richphasessuchastitaniteandilmenite.AtNNObuffered conditions fluids areCO2-saturated, although they lie close to thegraphitestabilityfield.Plagioclaseisstableattemperatures≤900°Canditspresenceyields lowerAl2O3andCaOcontents inthemelt foragiventemperaturerelativetoCo-CoO(NNO-1.7).Ti-richilmeniteisthetypicaloxidephaseforreducedconditions.The seriesofRe-ReO2 experiments (NNO+2)differ inmany respects from those atlower ƒO2 including thepresenceofCO2-rich fluids, formationofFe-richmagnetite,andabsenceofTi-rich ilmeniteandtitanite.Plagioclase isstable inalmostall runs,suggesting that its stability field increases with increasing ƒO2 and becomes lesssensitivetowatercontentinthemelt.AsaresulttheAl2O3andCaOcontentsinthemeltarelowerincomparisontoNNO.The most oxidized experiments buffered at hematite-magnetite (NNO+4.75) aresimilartothosebufferedbyRe-ReO2withthemaindistinguishingfeaturebeingthelowTiandveryhighFecontentsofmagnetite.Themostsurprisingresultisrelatedtothestabilityofthesulphur-bearingphasesinthese experiments. At 0.5 GPa the anhydrite stability field was encountered atNNO+2 in agreement with other low-pressure studies (e.g. Masotta and Keppler2015).Incontrast,the1.0and1.5GPaNNO+2runscontainedpyrrhotitedefining,forthefirst time,expansionof thesulphidestability fieldtohigherƒO2with increasingpressure.At1.0and1.5GPa,thetransitiontoanhydritestabilityisshifted~1and1.5logƒO2unitshigherthanat0.5GPa.Implication for the formation of PCDs. Porphyry copper deposits are widelyconsidered to be the result of copper enrichment in arc magmas duringdifferentiation.This ismost readily achievedwith the suppressionof any sulphidephase,whichwouldstronglysequesterCu(partitioningvalues intheregionof500for solid sulphideor1000 for sulphidemelt;Leeetal.2012;LiandAudétat2012;Zajaczetal.2013).Weshowthatsuppressionofsulphideismostreadilyachievedbydeep S-degassing, which in turn requires the presence of substantial dissolved

27

volatiles(H2OandCO2)inprimarymagmas.InthiscaseSpartitionsintotheexsolvedfluidphase,maintainingScontentsbelowsulphidesaturation(SCSS).ElevatedƒO2isalso favourable to preventing sulphide saturation, but may be subordinate to theeffectofhighinitialvolatilecontents(Fig.9).Weagreewithmanypreviousauthorsthat it is the volatile-rich nature of arc magmas makes them conducive to PCDformation,ratherthansimplytheirelevatedƒO2,asclaimedbyothers(e.g.Sunetal.2014). Althoughbothare typical featuresofarcmagmas(Humphreysetal.,2015)wehighlighttheimportanceofvolatileprocesseseveninthedeeparccrust.Acknowledgments

This projectwas funded by a research Grant from BHP Billiton.Wewould like toacknowledge Charles Clapham and Donovan Hawley for maintaining the high-pressure laboratory inworkingconditionandcontributingtothedevelopmentofapiston cylinder cell for the controlled ƒO2 experiments used in this study.We alsowould like to thank JohnMavrogenes, Brian Tattitch, Simon Kohn,Marion Louvel,Kevin Klimm, and Amy Gilmer for fruitful discussion and helpful advice. StephanSchuth is gratefully acknowledged for providing S- and C- analyses of startingmaterials,BenBuseandStuartKearns forhelpwithEMPAandRichardHinton forhelpwithionprobeanalysis.

28

Citedreferences

Armstrong LS, Hirschmann MM, Stanley BD, Falksen EG, Jacobsen ST (2015)SpeciationandsolubilityofreducedC-O-H-Nvolatilesinmaficmelt:implicationsforvolcanism,atmosphericevolution,anddeepvolatilecyclesintheterrestrialplanets.GeochimCosmochimAc171:283-302Arai S, Ishimaru S, Okrugin VM (2003)Metasomatized harzburgite xenoliths fromAvachavolcanoasfragmentsofmantlewedgeoftheKamchatkaarc:Implicationsforthemetasomaticagent.IslArc12:233–246BacknaesL,StellingJ,BehrensH,GoettlicherJ,MangoldS,VerheijenO,BeerkensRGC,DeubnerJ(2008)Dissolutionmechanismsoftetravalentsulphurinsilicatemelts:evidencesfromsulphurKedgeXANESstudiesonglasses.JAmCeramSoc91:721–727BallhausC,BerryR,GreenD (1991)Highpressure experimental calibrationof theolivine–orthopyroxene–spinel oxygen geobarometer: implications for the oxidationstateoftheuppermantle.ContribMineralPetrol107(1):27–40Ballhaus (1993) Redox states of lithospheric and asthenospheric upper mantle.ContribMineralPetrol114:331-348Berry A J, O’Neill HStC, Scott DR, Foran GJ, Shelley J MG (2006) The effect ofcompositiononCr2+/Cr3+insilicatemelts.AmMineral91:1901–1908BeermannO,BotcharnikovRE,Holtz F,DiedrichO,NowakM (2011)Temperaturedependence of sulfide and sulfate solubility in olivine-saturated basaltic magmas.GeochimCosmochimAc75:7612-7631Bernard A, DeMaiffe, D, Mattielli N, Punongbayan RS (1991) Anhydrite-bearingpumice from Mount Pinatubo:Further evidence for the existance of sulphur-richmagmas.Nature354:139-140

29

Blatter, DL, Carmichael ISE (1998) Hornblende peridotite xenoliths from centralMexicorevealthehighlyoxidizednatureofsubarcmantle.Geology26:1035-1038BlundyJ,CashmanK,RustA,WithamF(2010)AcaseforCO2-richarcmagmas.EarthPlanetScLett290:289-301Botcharnikov RE, Linnen RL,WilkeM, Holtz F, Jugo PJ, Berndt J (2010) High goldconcentrationsinsulphide-bearingmagmaunderoxidizingconditions.NatureGeo4:112-115Brandon AD, Draper DS (1996) Constraints on the origin of the oxidation state ofmantle overlying subduction zones: An example from Simcoe, Washington, USA.GeochimCosmochimAc60:1739-1749BrookerR,HollowayJR,HervigRL(1998).Reductioninpistoncylinderexperiments:Thedetectionofcarboninfiltrationintoplatinumcapsules.AmMineral83:985-994Camus F, Dilles JH (2001) A special issue devoted to porphyry copper deposits ofNorthernChile.EconGeol96:233-237Carmichael ISE(1991)Theredoxstatesofbasicandsilicicmagmas:areflectionoftheirsourceregions?ContribMineralPetrol106:129-141CarmichaelISE,GhiorsoMS(1986)Oxidation-reductionrelationsinbasicmagma:acaseforhomogeneousequilibria.EarthPlanetScLett78:200-210CarrollMR, RutherfordMJ (1985) Sulfide and sufate saturation in hydrous silicatemelts.JGeophysRes90:C601-C612CarrollMR,RutherfordMJ(1987)Resultsand implication forsulphurbehaviour inthe1982ElChichonTrachyandesiteandotherevolvedmagmas.JPetrol28(5):781-801

30

CarrollMR,RutherfordMJ(1988)Sulfurspeciationinhydrousexperimentalglassesof varying oxidation state: Results frommeasuredwavelength shifts of sulphur X-rays.AmMineral73:843-849Candela PA, Holland HD (1984) The partitioning of copper and molybdenumbetweensilicatemeltsandaqueousfluids.GeochimCosmochimAc48:373-380ChamberfortI,DillesJH,KentAJR(2008)Anhydrite-bearingandesiteanddaciteasasource for sulfur in magmatic-hydrothermal mineral deposits. Geology 36(9):719-722ChurakovSV,GottschalkM (2003)Perturbation theorybasedequationof state forpolarmolecularfluids:II.Fluidmixtures.GeochimCosmochimActa67:2415-2425Clemente B, Scaillet, Pichavant M (2004) The solubility of sulphur in hydrousrhyoliticmelts.JPetrol45:2171-2196Comodi P, Nazzareni S, Zanazzi PF, Speziale S (2008) High-pressure behaviour ofgypsum:Asingle-crystalX-raystudy.AmMineral93:1530-1537ConnollyJAD(2005)Computationofphaseequilibriabylinearprogramming:Atoolfor geodynamic modelling and its application to subduction zone decarbonation.EarthPlanetSciLett236:524-541Crabtree SM, Lange RA (2012) An evaluation of the effect of degassing on theoxidation state of hydrous andesite and dacitemagmas: a comparison of pre- andpost-eruptiveFe2+concentrations.ContribMineralPetrol163:209-224Dauphas N, Craddock PR, Asimow PD, Bennett VC, Nutman AP, Ohnenstetter D(2009) Iron isotopes may reveal the redox conditions of mantle melting fromArcheantoPresent.EarthPlanetScLett288:255-267.

31

Evans KA, Tomkins AG (2011) The relationship between subduction zone redoxbudgetandarcmagmafertility.EarthPlanetScLett308:401-409Evans KA, Elburg MA, Kamentsky VS (2012) Oxidation state of subarc mantle.Geology40(9):783-786FoleySF(2011)AreappraisalofredoxmeltingintheEarth’sMantleasafunctionoftectonicsettingsandtime.JPetrol52:1363-1391Frey FA, GerlachDC,HickeyRL, Lopez-Escobar L,Munizaga-Villavicencio F (1984)PetrogenesisoftheLagunadelMaulevolcaniccomplex,Chile(36°S)ContribMineralPetrol88:133-149.doi10.1016/j.epsl.2014.05.038Freise M, Holtz F, Nowak M, Scoates JS, Strauss H (2009) Differentiation andcrystallisation conditions of basalts from theKerguelen large igneous province: anexperimentalstudy.ContribMineralPetrol158:505-527Froese E, Gunter AE (1976) A note on the pyrrhotite-sulphur vapour equilibrium.EconGeol71:1589-1594GhiorsoMS,GualdaGAR(2015)AnH2O-CO2mixedfluidsaturationmodelcompatiblewithrhyolite-MELTS.ContribMineralPetrol169:53Gill R (2010) Igneous rocks and processes: a practical guide. Wiley-Blackwell,Malaysia.HattoriK (1993)High-sulfurmagma, a product of fluid discharge fromunderlyingmaficmagma:EvidencefromMountPinatubo,Philippines.Geology21:1083-1086Herrmann W, Berry RF (2002) MINSQ – a least squares spreadsheet method forcalculatingmineralproportions fromwholerockmajorelementanalyses.GeochemExplorEnvironAnal2:361-368

32

Holland T, Blundy J (1994) Non-ideal interactions in calcic amphiboles and theirbearingonamphibole-plagioclasethermometry.ContribMineralPetrol116:433-447HollandTJB, PowellR (1998)An internally consistent thermodynamicdata set forphasesofpetrologicalinterest.JMetamorphGeol16:309-343HumphreysMCS,Blundy JD, SparksRSJ (2006)Magmaevolutionandopen-systemprocessesatShiveluchVolcano: Insights fromphenocrystzoning. JPetrol47:2303-2334Humphreys MCS, Brooker RA, Fraser DG, Burgisser A, Mangan M, McCammon C(2015)Coupled interactionbetween volatile activity andFe oxidation state duringarccrustalprocesses.JPetrol56:795-814JennerFE,O’NeillHStC,ArculusRJ,Mavrogenes, JA (2010)Themagnetitecrisis intheevolutionofarc-relatedmagmasandtheinitialconcentrationofAu,AgandCu.JPetrol51:2445-2464Johnson KE, Davis AM, Bryndzia LT (1996) Contrasting styles of hydrousmetasomatism in the upper mantle: An ion microprobe investigation. GeochimCosmochimAc60:1367-1385JugoPJ,LuthRW,RichardsJP(2005)Anexperimentalstudyofthesulphurcontentinbasaltic melts saturated with immiscible sulfide or sulfate liquids at 1300°C and1.0GPa.JPetrol46(4):783-798JugoPJ(2009)Sulphurcontentatsulphidesaturationinoxidizedmagmas.Geology37(5):415-418JugoPJ,WilkeM,BotcharnikovRE (2010)SulfurK-edgeXANESanalysisofnaturaland synthetic basaltic glasses: implications for S speciation and S content as afunctionofoxygenfugacity.GeochimCosmochimAc74:5926-5938

33

KatsuraT,NagashimaS(1974)Solubilityofsulfurinsomemagmasat1atmosphere.GeochimCosmochimAc38:517-531Kelley KA, Cottrell E (2009) Water and the oxidation state of subducted zonemagmas.Science325:605-607Kelley KA, Cottrell E (2012) The influence of magmatic differentiation on theoxidationstateofFeinabasalticarcmagma.EarthPlanetScLett329-330,109-121Keppler H (2010) The distribution of sulfur between haplogranitic melts andaqueousfluids.GeochimCosmochimAc74:645-660KingHE, Prewitt CT (1982)High-pressure andhigh-temperaturepolymorphismofironsulfide(FeS).ActaCrystB38:1877-1887KressV(1997)MagmamixingasasourceforPinatubosulphur.Nature389:591-593Kress VC, Carmichael ISE (1991) The compressibility of silicate liquids containingFe2O3and theeffectof composition, temperature,oxygen fugacityandpressureontheirredoxstates.ContribMineralPetrol108:82-92LeBasMJ,LeMaitreRW,StreckeisenA,ZanettinB(1986)Achemicalclassificationofvolcanicrocksbasedonthetotalalkali-silicadiagram.JPetrol27:745-750LeeC-TA,LuffiP,LeRouxV,DasguptaR,AlbaredeF,LeemanW(2010)TheredoxstateofarcmantleusingZn/Fesystematics.Nature468:681-685LeeC-TA,LuffiP,ChinEJ,BouchetR,DasguptaR,MortonDM,LeRouxV,YinQ,JinD(2012) Copper systematics in arc magmas and implications for crust-mantledifferentiation.Science336:64-68

34

LiY,AudétatA(2012)PartitioningofV,Mn,Co,Ni,Cu,Zn,As,Mo,Ag,Sn,Sb,W,Au,Pb and Bi between sulfide phases and hydrous basanite melt at upper mantleconditions.EarthPlanetScLett355-356:327-340LiuY,SamahaN-T,BakerDR(2007)Sulfurconcentrationatsulfidesaturation(SCSS)inmagmaticsilicatemelts.GeochimCosmochimAc71:1783-1799LowensternJB(2001)Carbondioxideinmagmasandimplicationsforhydrothermalsystems.MinerDeposita36:490-502LuhrJF(1990)Experimentalphaserelationsofwater-saturatedandsulfur-saturatedarcmagmasandthe1982eruptionsofElChichonvolcano.JPetrol31:1071–1114LuthR(1989)Naturalversusexperimentalcontrolofoxidationstate:EffectsonthecompositionandspeciationofC-O-Hfluids.AmMineral74:50-57LuthRW (1989)Natural versus experimental control of oxidation state: effects onthecompositionandspeciationofC-O-Hfluids.AmMineral74:50-57Machacek J,GedeonO,LiskaM,MarhoulF (2010)Molecular simulationsof silicatemeltsdopedwithsulphurandnitrogen.J.Non-Cryst.Solids356:2458-2464MallmanG,O’NeillH(2009)Thecrystal/meltpartitioningofVduringmantlemeltingasafunctionofoxygenfugacitycomparedwithsomeotherelements(Al,P,Ca,Sc,Ti,Cr,Fe,Ga,Y,ZrandNb).JPetrol50:1765-1794MasottaM,KepplerH(2015)Anhydritesolubilityindifferentiatedarcmagmas.GeochimCosmochimAc158:79-102Matjuschkin V, Brooker RA, Tattitch BT, Blundy J, Stamper CC (2015) Control andmonitoringofoxygenfugacityinpistoncylinderexperiments.ContribMineralPetrol169:9.doi10.1007/s00410-015-1105-z

35

MatthewsSJ,MoncrieffDHS,CarrollMR(1999)Empiricalcalibrationofthesulphurvalence oxygen barometer from natural and experimental glasses: method andapplications.MineralMag63:421-431MavrogenesJA,O’NeillHStC(1999)Therelativeeffectsofpressure,temperatureandoxygenfugacityonthesolubilityofsulfideinmaficmagmas.GeochimCosmochimAc63:1173-1180McDade P, Wood BJ, Van Westrenen W, Brooker R, Gudmundsson G, Soulard H,NajorkaJ,BlundyJ(2002)Pressurecorrectionforaselectionofpiston-cylindercellassemblies.MineralMag66(6):1021-1028MétrichN, Berry, AJ, O'NeillH StC, Susini J (2009)The oxidation state of sulfur insyntheticandnaturalglassesdeterminedbyX-rayabsorptionspectroscopy.GeochimCosmochimAc73:2382–2399Moretti R, Papale P (2004) On the oxidation state and volatile behaviour inmulticomponentgas-meltequilibria.ChemGeol213:265-280Morizet,Y,ParisM,GaillardF,ScailletB(2010)C–O–Hfluidsolubilityinhaplobasaltunderreducingconditions:Anexperimentalstudy.ChemGeol279:1–16MungallJE(2002)Roastingthemantle:slabmeltingandthegenesisofmajorAuandAu-richCudeposits.Geology30(10):915-918Muntean JL, Cline JS, Simon AC, Longo A (2011)Magmatic-hydrothermal origin ofNevada’sCarlin-typegolddeposits.NatureGeosci4:122-127Nagashima S, Katsura T (1973) The solubility of sulfur in Na2O-SiO2 melts undervariousoxygenpartialpressuresat1100°C,1250°C,and1300°C.BChemSocJpn46:3099-3103

36

NewtonRC,ManningCE (2008)ThermodynamicsofSiO2-H2O fluidnear theuppercritical end point from quartz solubilitymeasurements at 10kbar. Earth Planet ScLett274:241-249Pallister JS, Hoblitt RP, Meeker GP, Knight RJ, Siems, DF (1996)Magmamixing atMount Pinatubo: Petrographic and chemical evidence from the 1991 deposits. In:FireandMud:EruptionsandLaharsofMountPinatubo,PhilippinesParkinson IJ,ArculusRJ (1999)Theredoxstateof subductionzones: insights fromarc-peridotites.ChemGeol160:409-423ParkinsonIJ,PearceJA(1998)PeridotitesfromtheIzu-Bonin-MarianaForearc(ODPLeg 125): Evidence for mantle melting and melt-mantle interaction in a supra-subductionzonesetting.JPetrol39:1577-1618.ParkinsonIJ,ArculusRJ,EgginsSM(2003)PeridotitexenolithsfromGrenada,LesserAntillesIslandArc.ContribMineralPetrol146(2):241-262Pistone M, Blundy JD, Brooker RA, (2016) Textural consequences of interactionbetweenhydrousmaficand felsicmagmas:anexperimental study.ContribMineralPetrol171:8,doi10.1007/s00410-015-1218-4Pokrovski GS, Dubrovinsky LS (2011) The S3- ion is stable in geological fluids atelevatedtemperaturesandpressures.Science331:1052-1054Pawley AR, Holloway JR,McMillan PF (1992) The effect of oxygen fugacity on thesolubilityofcarbon-oxygenfluidsinbasalticmelt.EarthPlanetScLett110,213-225PokrovskiGS,KokhMA,GuillaumeD,BorisovaAY,GisquetP,HazemannJ-L,LaheraE, Del NetW, ProuxO, Testemale D, Haigis V, Jonchiere R, Seitsonen AP, Ferlat G,VuilleumierR,SaittaAM,BoironM-C,Dubessy J (2015)Sulfur radical species formgold deposits on Earth. Pnatl Acad Sci USA 112(44) 13484-13489.doi:10.1073/pnas.1506378112

37

RobieRA,HemingwayBS,Fisher JR(1978)Thermodynamicpropertiesofmineralsandrelatedsubstancesat298.15Kand1bar(105pascals)pressureandathighertemperatures. U.S. Geological Survey; For sale by the Supt. of Docs., U.S. G.P.O.,Bulletin1452,iii,p456.http://pubs.usgs.gov/bul/1452/report.pdfSamara GA and Giardini AA (1965) Compressibility and electrical conductivity ofcadmiumsulfideathighpressures.PhysRev140(1A):A388-A396SatoK(2012)Sedimentarycrustandmetallogenyofgranitoidaffinity: implicationsfrom the geotectonic histories of the Circum-Japan sea region, central Andes andSoutheasternAustralia.ResourGeol62(4):325-351Scaillet B, Clemente B, Evans BW, Pichavant M (1998) Redox control of sulphurdegassinginsilicicmagmas.JGeophysRes103(B10):23937-23949SchaafP,StimacJ,SiebeC,MaciasJL(2005).Geochemicalevidenceformantleoriginand crustal processes in volcanic rocks from Popocatépetl and surroundingmonogeneticvolcanoes,CentralMexico.JPetrol46:1243-1282Scott JAJ, PyleDM,Mather TA, RoseWI (2013)Geochemistry and evolution of theSantiaguitovolcanicdomecomplex,Guatemala.JVolanolGeothRes252:92–107.Sillitoe RH (2012) Copper provinces. Society of Economic Geologists, SpecialPublications16:1-18SissonTW,RatajeskiK,HankinsWB,GlaznerAF(2005)Voluminousgraniticmagmasfromcommonbasalticsources.ContribMineralPetrol148:635–661SunW,ArculusRJ,KamenetskyVS,BinnsRA(2004)Releaseofgold-bearingfluidsinconvergentmarginmagmaspromptedbymagnetitecrystallization.Nature431:975-978

38

SunW-d,LiangH-y,LingM-x,ZhanM-z,DingX,ZhangHZ,YangX-y,LiY-l.IrelandTR,WeiQ-r,FanW-m(2013)Thelinkbetweenreducedporphyrycopperdepositsandoxidizedmagmas.GeochimCosmochimAc103:263-275SverjenskyDA,HarrisonB,AzzoliniD(2014)WaterinthedeepEarth:thedielectricconstantandthesolubilitiesofquartzandcorundumto60kband1200°C.GeochimCosmochimAc129:125-145ThibaultY,HollowayJR(1994)SolubilityofCO2

inaCa-richleucitite:effectsofpressure,temperatureandoxygenfugacity.ContribMineralPetrol116:216–224.UedaA,ItayaT(1981)MicrophenocrysticpyrrhotitefromdaciterocksofSatsuma-Iwojima, Southwest Kyushu, Japan and the solubility of sulfur in dacite magma.ContribMineralPetrol78:21–26UlmerP,LuthRW(1991)Thegraphite-COHfluidequilibriuminP,T, ƒO2space.Anexperimentaldeterminationto30kbarand1600°C.ContribMineralPetrol106:265-272WallacePJ (2005)Volatiles in subduction zonemagmas: concentrationsand fluxesbasedonmeltinclusionandvolcanicgasdata.JVolcanolGoethRes140:217-240WebsterJD,BotcharnikovRE(2011)DistributionofsulfurbetweenmeltandfluidinS-O-H-C-Cl-bearing magmatic systems at shallow crustal pressures andtemperatures.RevMineralGeochem73:247-283WilkeM,JugoPJ,KlimmK,SusiniJ,BotcharnikovR,KohnSC,JanouschM(2008)TheoriginofS4+detectedinsilicateglassesbyXANES.AmMineral93:235–240WoermannE,RosenhauerM(1985)FluidphasesandtheredoxstateoftheEarth’smantle.Explorationsbasedonexperimental,phase-theoreticalandpetrologicaldata.FortschrMineral63(2):263-349

39

Wood BJ, Virgo D (1989) Upper mantle oxidation state: Ferric iron contents oflherzolite spinel by 57Fe Mössbauer spectroscopy and resultant oxygen fugacities.GeochimComochimAc53:1277-1291WoodBJ,BryndziaLT,JohnsonKE(1990)Mantleoxidationstateanditsrelationshiptotectonicenvironmentandfluidspeciation.Science248:337-345Zajacz Z, Candela PA, Piccoli PM, Sanchez-Valle C, Wälle M (2013) Solubility andpartitioningbehaviourofAu,Cu,AgandreducedSinmagmas.GeochimCosmochimAc112:288-304ZajaczZ(2015)Theeffectofmeltcompositiononthepartitioningofoxidizedsulfurbetweensilicatemeltsandmagmaticvolatiles.GeochimCosmochimAc158:223-224ZellmerFG,EdmondsM,StraubSM(2014)Volatilesinsubductionzonemagmatism.GeologicalSocietymLondon,SpecialPublications.doi10.1144/SP410.13

40

FigureCaptions

Figure1:Theoxidationstatesofmagmas,theirmantlesourcesandsomesolid-statebuffers.In(a)theƒO2isshownforavarietyoftectonicsettings.ThemainhistogramsforshallowmantleperidotitesarefromthedatasetofFoley(2011)withadditionalindividual samples fromAvachinsky,Araietal., (2003); Japan-Ichinomegata,WoodandVirgo(1989); Izu-Bonin,ParkinsonandPearce (1998);Cascades,BrandonandDraper, (1996); Grenada, Parkinson et al., (2003); Shiveluch, Bryant et al., (2007);Mexico, Blatter and Carmichael (1998). Redox state of the arc mantle, as impliedfrom geochemical features of lavas, from Mallmann and O’Neill (2009), Lee et al.(2005, 2010) and Evans et al. (2012). Arc basalt range from Carmichael (1991),Eggins (1993), Heath et al. (1998), Macdonald et al. (2000), Hoog et al. (2004),Crabtree and Lange (2011). Specific examples include:Mexico, Carmichael (1991);Shiveluch, Humphreys et al (2006); Grenada, Parkinson et al. (2003). Sulphate-bearing andesites and dacites as listed in Scaillet et al. (1998). MORB range fromChristieetal.(1986),Ballhausetal.(1991),CottrellandKelley(2011).In(b)logƒO2rangeofvarioussolid-statebuffersasa functionof temperatureandpressure.Theupperedgeofeachrangerepresents2.0GPa, the loweredge0.1MPa(1bar).Thisillustratesthesmalleffectofdepthfromthetoptobottomofthecrust.ThelogunitvariationfromtheNi-NiObufferisexpressedas∆NNOandherethe1.0GPalinesareshown.Atypicalchange fromsulphide(S2-) tosulphate(S6+)asa functionof ƒO2 isshown for thecondition0.2GPaand950°C (Botcharnikovetal.2010) to illustratetherelativescaleofthiseffect.

Figure2: Secondary (SE)andbackscattered (BSE)electron imagesof runproductsproduced at various pressure, temperature and ƒO2 conditions. (a) Experiment atNNO, 1.5GPa and 950°C with garnet (grt) as a main phase and sulphides (sulph)presentas<3µmsphericalgrainsinsilicatemeltandalsoasinclusionsingarnetandamphibole(amph).(b)AnexampleofhighlyoxidizedexperimentatNNO+4.7,1GPaand 841°C showing the presence of anhydrite (anh) and absence of sulphides. (c)GraphiteflakesaroundamphiboleagglomeratesinreducedNNO-1.5runat1.0GPaand 900°C. Like the COH-fluid, graphite tends to nucleate near crystals or capsule

41

walls. (d) The decrease in fluid and increase inmelt proportions with decreasingtemperature for a fixed ƒO2 and pressure (NNO+2, 1.5GPa). (e) Effect of oxygenfugacityonfluidspeciation.CO2-richfluidsarestableatƒO2≥NNOandproducelargevesicles≤50µmindiameter. Incontrast,withgraphitepresentatNNO-1.5vesiclesare~3µm(notechangeinscale).Figure 3: Phase relations as a function of oxygen fugacity and temperature forexperimentsperformedat(a)1.0and(b)1.5GPa.Allrunscontainfluidandsilicatemelt. Red line illustrates the transition from CO2-H2O to graphite-COH fluidcalculatedusingUlmerandLuth(1991).Notethatat1.0and1.5GPaandNNO+2runproductsaresaturatedwithpyrrhotiteatalltemperatures.Figure 4: Graphic representation of phase proportions demonstrating the effect oftemperatureandoxygenfugacityonthephaserelationsinhydrousmagmas.Overall,the melt fraction increases with decreasing ƒO2 owing to the suppression ofplagioclase crystallisation and/or the increasing stability of H2O due to redoxreactions.Figure 5: Bulk starting composition and the range of experimental silicate meltsplottedonatotalalkali-silicadiagram(LeBasetal.1986).Figure 6: Compositional trends for selected components that show distinctivevariation with experimental ƒO2. Data are presented on an anhydrous basis andpresented in wt% and ppm concentrations with ±1σ uncertainty. Plots of otherelementsareprovidedinsupplementarymaterials(Fig.S1). Figure 7: Ion-probemeasurements ofwater andCO2 contents in quenched silicatemelts as a function of temperature. (a) With decreasing temperature there is a

42

general increase indissolvedwateranddecrease inCO2.ReducedCo-CoO-bufferedrunsexhibithighestwaterand lowestCO2 contentsowing tobreakdownofCO2 tographite+H2O(±CH4±CO).(b)Acomparisonofthevolatilecontentsoffluid-saturatedmelts in this study and isobars for oxidized (CO2-H2O) fluid-saturated meltscalculated using themodel of Ghiorso and Gualda (2015) at 0.5 and 1.0 GPa. Thepositionofthe1.5GPareduceddataisdiscussedinthetext.Figure 8: (a) Sulphur speciation in melt as a function of oxygen fugacity. ThemodelledspeciationincludesanintermediateoxidationS-speciesbetweenS2-andS6+transitionandislabelledasS0.(b)Experimental(symbols)andmodelled(greylines)S-contents inmelt for theentire rangeof studied ƒO2.Note that temperaturehasastrong effect on the Sulphur Content at Anhydride Saturation (SCAS) but only aminor effect on Sulphur Content at Sulphide Saturation (SCSS). The red symbolsrepresent experimentswithin a narrow temperature range from841 to 875°Cbutwith unreliable thermocouple readings (see section 2.2). (c) Illustration of thepressureeffectonsulphurspeciationinFe-bearingsilicatemeltsexpressedas!!!!

The0.2GPalineisbasedonthebasalticdataofJugoetal.(2010)andBotcharnikovetal.(2011).The0.5GPalineisestimatedfromasingleruninthisstudy;theothertrachydacitedataat1.0and1.5GPalinesareconstructedfromthedatain(a).Withincreasing pressure the stability field of sulphide shifts significantly towardsmoreoxidizedconditions.Itshouldbenotedthatthe0.2GPa/800-1000°CdataofMasottaand Keppler (2015) suggest there is a minor effect of melt composition on theestimatedpositionoftheS6+-S2-betweenbasalt,trachyandesiteanddacite.Thisshiftismuchsmallerthanthepressureeffectillustratedhere.Figure 9: Possible pressure and ƒO2 controls onCu-enrichment in silicatemagmas.Greyareadelimitsthefieldofgraphitesaturationatwhich>80%ofCO2breaksdowntoC+H2O(±CH4±CO)leadingtosignificantreductionintheamountofexsolvedfluid.At ƒO2 above the graphite stability field themagmatic fluid is composedmainly ofH2O+CO2. In relatively 'dry', water under-saturated magmatic systems Cu-enrichment in the absence of sulphide requires very high ƒO2 values ∆NNO≥2 at

43

0.5GPa(midcrust)butanextremelyhighvalue∆NNO≥3.5(seeFig1a)towardsthedeepercrust(oruppermantle)atpressuresapproaching1.5GPa.Incontrast,fluid-saturatedmagmaswould facilitate Cu-enrichment atmore realistic, ƒO2 as sulphurpartitioning into exsolving magmatic fluids limits the potential for sulphidesaturationandCuremoval.FigureS1:Compositionofsilicatemeltsinexperimentsperformedat1.0and1.5GPa.Thevaluesarecalculatedonanhydrousmeltbasisandrepresentmeanvaluesof7-21analyses;errorbarsreferto±1σreproducibility.FigureS2:CalculatedvaluesforKAandKBconstantsat1.0and1.5GPa.WhilstKBmostlycontrolsthepositionof!!!!+,KAdeterminesthepositionof!!!! and!!!!lines.FigureS3:AlternativemodeltoFig.8,with(a)S-speciationand(b)S-contentsinsilicatemeltsat1.0GPaassumingthepresenceofS4+-intermediatespeciesinsteadofS0.NotethemaximumforS4+isclosertotheoxidizedfieldcomparedtoS0(Fig.8).ThiswouldlimitthestabilityfieldofS2-tobelowNNO+1whichcontradictstheexperimentalobservations.

1

Table 1 Composition of the starting material (in wt.%). 1 2 SiO2 TiO2 Al2O3 FeO Fe2O3 MgO CaO Na2O K2O S2- SO3 Cl H2O CO2 Total 52.63 1.18 17.13 4.52 2.06 2.71 5.74 4.38 1.74 0.14 0.27 0.15 6.00 1.36 100

3 Table 2 Experimental conditions, composition of fluids and proportions of run 4 products. 5 6 Pressure, GPa

Temperature, °C

XH2O in fluid

aH2O logƒO2∆NNO Run products

Hematite-magnetite buffer

1.0 841 0.81(2) 0.83 4.77 gl (43.1), fl (2.9), hbl (21.8), plag (27.2), mag (3.6), anh(1.4)

1.0 870 0.78(3) 0.81 4.76 gl (47), fl (2.5), hbl (23.9), plag (23.3), mag (2.1), anh(1.1)

1.0 895 0.80(2) 0.82 4.77 gl (49), fl (2.7), hbl (20.3), plag (23.1), mag (3.5), anh(1.3)

Rhenium-rhenium oxide buffer

0.5 900 0.79(2) 0.81 1.95 gl (65.2), fl (3), hbl (11.4), plag (17), mag (3.2), anh(0.3)

1.0 850 0.85(2) 0.87 2.21 gl (34.4), fl (3.9), hbl (28.5), plag (31.9), mag (0.6), sulph (0.4)

1.0 900 0.76(2) 0.80 2.06 gl (55.7), fl (2.4), hbl (23.9), plag (16.1), mag (0.7), sulph (0.6)

1.0 950 0.76(2) 0.79 1.98 gl (67.3), fl (2.2), hbl (20.5), plag (9.4), mag (>0.1), sulph (0.6)

1.5 850 0.80(2) 0.83 2.30 gl (46.6), fl (2.5), hbl (30.6), plag (20.3), mag (>0.1), sulph (<0.1)

1.5 900 0.78(4) 0.81 2.19 gl (59.3), fl (2.1), hbl (26), plag (9.8), grt (2.8), sulph (<0.1)

1.5 950 0.70(4) 0.75 2.05 gl (80.3), fl (1), grt (11), cpx (11), sulph (<0.1)

Nickel-nickel oxide buffer

1.0 850 0.80(5) 0.83 -0.16 gl (49.8), fl (2.8), hbl (26.4), plag (19.4), ilm (0.8), sulph (0.7)

1.0 900 0.75(5) 0.79 -0.21 gl (64.4), fl (1.7), hbl (23.9), plag (8), ilm (0.2), sulph (1)

1.0 950 0.73(5) 0.77 -0.23 gl (83.8), fl (0.8), hbl (13.4), ilm (0.1), sulph (1.9)

1.5 850 0.80(5) 0.83 -0.15 gl (44.4), fl (2.9), hbl (26.6), plag (20.9), grt (4.2), ilm (>0.1), sulph (>0.1), ttn(1)

1.5 900 0.80(5) 0.82 -0.17 gl (67.7), fl (1.3), hbl (21.5), plag (2.6), grt (6.8), sulph (0.1)

1.5 950 0.75(5) 0.79 -0.29 gl (77.3), fl (1), hbl (<0.1), grt (13.9), cpx (7.1), sulph (0.6)

Cobalt-cobalt oxide buffer

1.0 850 - 0.90 -1.74 gl (75.6), fl (0.7), hbl (21), ilm (0.2), sulph (1), ttn (1.1)

1.0 900 - 0.91 -1.74 gl (74.6), fl (0.7), hbl (21.3), ilm (0.1), sulph (1.6), ttn (1.4), gr

1.0 950 0.75(5) 0.78 -1.53 gl (84), fl (0.2), hbl (15), ilm (0.6), sulph (1)

1.5 850 - 0.91 -1.74 gl (69.7), fl (0.3), hbl (25.5), grt (2.6), ilm (>0.1), sulph (>0.1), ttn(1.3), gr

1.5 900 - 0.91 -1.74 gl (74.5), fl (0.1), hbl (24), sulph (1), ttn(0.4), gr

1.5 950 - 0.91 -1.75 gl (87.9), fl (0.2), cpx (7.7), sulph (3.9), gr

7 Quenched glass (gl), fluid (fl), hornblende (hbl), plagioclase (plag), garnet (grt), 8 clinopyroxene (cpx), magnetite (mag), Fe-sulphide (sulph), titanite (ttn), 9 anhydrite (anh).10

2

Table 3 Composition of run products in wt.%, except where indicated.

Phases (#)

SiO2 TiO2 Al2O3 FeOt or Fe (in

sulphide)

MgO CaO Na2O K2O S Cl Total Mg# H2O CO2 S, ppm

Hematite-magnetite buffer (NNO+4.75)

1.0GPa, 841°C (+anh)

Gl (10) 68.58(37) 0.16(2) 17.39(26) 1.63(10) 0.54(6) 2.06(9) 6.11(25) 3.19(11) 0.04(1) 0.26(2) 100.00 0.37 7.22 0.48(12) 387±127

Hbl(8) 40.19(53) 2.70(40) 13.94(23) 13.69(52) 11.75(72) 10.94(13) 3.16(10) 0.72(4) b.d.l. b.d.l. 97.09 0.6 7.22 0.48(12)

Plag(8) 60.31(63) 0.04(3) 24.52(38) 0.58(6) b.d.l. 6.33(42) 8.27(22) 0.44(3) b.d.l. b.d.l. 100.49

Mag(5) 0.08(3) 15.40(50) 1.18(2) 73.14(56) 1.15(10) 0.11(5) b.d.l. b.d.l. b.d.l. b.d.l. 91.06

1.0GPa, 870°C (+anh)

Gl(10) 66.87(39) 0.28(3) 18.13(22) 2.42(12) 0.55(4) 2.48(4) 5.95(13) 2.98(6) 0.07(1) 0.19(1) 100.00 0.29 7.47(16) 0.53(1) 680±146

Hbl(10) 40.45(69) 2.34(28) 12.87(57) 16.71(42) 10.38(28) 10.49(14) 3.05(8) 0.81(5) b.d.l. 0.08(1) 97.18 0.53

Plag(12) 60.29(53) 0.03(1) 25.14(44) 0.50(9) b.d.l. 6.47(51) 7.93(28) 0.46(6) b.d.l. b.d.l. 100.82

Mag(9) 0.13(3) 24.34(27) 1.05(6) 65.13(34) 1.01(3) 0.15(3) b.d.l. b.d.l. b.d.l. b.d.l. 91.81

1.0GPa, 875°C (+anh)

Gl(13) 67.71(40) 0.21(2) 17.68(19) 2.07(9) 0.70(3) 2.45(6) 5.99(23) 2.88(7) 0.06(1) 0.20(2) 100.00 0,38 6.92(11) 0.52 599±134

Hbl(9) 40.33(66) 2.86(23) 13.90(56) 13.68(69) 11.68(42) 10.89(8) 3.16(8) 0.72(5) b.d.l. 0.06(1) 97.28 0.6

Plag(9) 58.83(35) 0.03(2) 25.75(23) 0.55(8) b.d.l. 7.47(12) 7.57(15) 0.35(2) b.d.l. b.d.l. 100.55

Mag(10) 0.07(2) 14.28(44) 1.36(8) 73.95(38) 1.08(4) 0.09(2) b.d.l. b.d.l. b.d.l. b.d.l. 90.83

Rhenium-rhenium oxide buffer (NNO+2)

0.5GPa, 900°C (+anh)

Gl(10) 64.00(21) 0.75(3) 18.09(19) 3.70(10) 1.32(6) 3.58(8) 5.85(6) 2.33(8) 0.10(1) 0.18(2) 100.00 0.39 5,87(23) 0.13(1) 1032±137

3

Hbl(14) 42.78(83) 2.63(40) 10.62(58) 11.91(42) 14.24(56) 11.18(49) 2.67(13) 0.48(4) b.d.l. b.d.l. 96.51 0.68

Plag(4) 51.86(53) 0.29(7) 27.40(91) 1.96(4) 0.57(9) 11.72(49) 2.93(18) 0.69(10) b.d.l. 0.07(2) 97.49

Mag(3) 0.55(11) 6.00(12) 4.68(37) 77.77(63) 3.33(19) 0.17(2) b.d.l. b.d.l. b.d.l. b.d.l. 92.50

1.0GPa, 850°C

Gl(8) 67.86(35) 0.31(3) 17.60(9) 2.31(10) 0.36(4) 2.51(8) 5.77(17) 2.97(9) 0.05(1) 0.22(2) 100.00 0.22 7.88(11) 0.47(2) 478±112

Hbl(9) 40.67(41) 2.36(13) 12.67(29) 17.96(43) 9.42(49) 10.57(16) 3.16(10) 0.88(6) b.d.l. 0.11(1) 97.8 0.48

Plag(10) 61.61(33) 0.04(1) 24.03(41) 0.47(7) 0.03(1) 5.56(37) 8.62(21) 0.56(9) b.d.l. b.d.l. 100.92

Mag(2) 0.44(40) 29.36(14) 0.77(7) 62.32(55) 0.84(2) 0.1(2) b.d.l. 0.06(2) b.d.l. b.d.l. 93.89

Sulph(2) 0.21(2) 0.06(1) 0.03(1) 59.91(11) b.d.l. 0.11(1) b.d.l. b.d.l. 37.64(25) b.d.l. 97.96

1.0GPa, 900°C

Gl(9) 64.51(28) 0.47(5) 18.65(19) 3.40(10) 0.65(4) 3.59(6) 5.84(19) 2.58(8) 0.06(1) 0.18(1) 100.00 0.25 7.47(29) 0.59(4) 622±115

Hbl(9) 40.26(70) 2.61(17) 13.71(15) 16.41(44) 10.14(40) 10.71(21) 3.19(8) 0.79(4) b.d.l. 0.06(1) 97.43 0.52

Plag(10) 57.72(44) 0.05(2) 26.52(34) 0.54(11) 0.03(3) 8.48(22) 7.06(19) 0.37(9) b.d.l. b.d.l. 100.77

Mag(6) 0.12(8) 30.15(48) 0.99(4) 61.66(58) 1.22(4) 0.15(4) b.d.l. b.d.l. b.d.l. b.d.l. 94.29

Sulph(2) 0.22(0) 0.09(1) 0.08(6) 59.95(12) b.d.l. 0.15(6) b.d.l. b.d.l. 37.65(20) b.d.l. 98.14

1.0GPa, 950°C

Gl(10) 62.46(39) 0.63(4) 18.67(16) 4.57(11) 1.08(3) 4.11(7) 5.81(23) 2.30(7) 0.10(1) 0.19 100.00 0.30 6.39(22) 0.56(1) 971±88

Hbl(10) 39.61(60) 2.99(9) 14.86(42) 15.51(37) 10.18(34) 10.63(16) 3.13(6) 0.79(4) 0.03(1) 0.07(1) 97.80 0.54

Plag(10) 56.97(38) 0.05(4) 26.71(48) 0.59(15) 0.05(5) 8.87(32) 6.84(16) 0.35(9) b.d.l. b.d.l. 100.43

Mag(4) 0.15(5) 28.94(43) 1.23(2) 61.10(62) 1.45(5) 0.17(4) b.d.l. 0.04(2) b.d.l. b.d.l. 93.08

Sulph(1) 0.18 b.d.l. 0.04 59.78 b.d.l. 0.1 b.d.l. b.d.l. 37.72 b.d.l. 97.82

1.5GPa, 850°C

4

Gl(10) 66.78(18) 0.44(5) 18.10(18) 2.59(13) 0.44(3) 3.01(5) 5.70(8) 2.67(8) 0.05(1) 0.17 100.00 0.24 8.05(7) 0.77(2) 463±93

Hbl(9) 39.32(30) 2.55(32) 15.01(45) 17.97(73) 8.00(34) 9.52(18) 3.00(10) 1.04(4) 0.03(1) 0.10(2) 96.54 0.44

Plag(11) 60.69(88) 0.03(2) 24.36(59) 0.40(8) b.d.l. 5.82(62) 7.62(41) 0.44(5) b.d.l. b.d.l. 99.36

Ilm(5) 0.06(2) 40.84(5) 0.67(4) 57.22(12) 1.78(2) 0.09(3) b.d.l. b.d.l. b.d.l. b.d.l. 100.66

Sulph(3) 0.15(16) 0.04(1) b.d.l. 60.04(98) b.d.l. 0.08(3) b.d.l. b.d.l. 39.61 b.d.l. 99.92

1.5GPa, 900°C

Gl(10) 65.60(30) 0.55(7) 18.35(17) 3.04(10) 0.66(4) 3.07(5) 6.03(25) 2.47(8) 0.03(1) 0.16(2) 100.00 0.28 7.34(1) 0.96(2) 352±82

Hbl(12) 38.83(22) 3.28(8) 15.62(22) 16.91(34) 8.28(13) 9.44(16) 3.04(7) 0.89(3) 0.04(1) b.d.l. 96.33 0.47

Plag(8) 59.52(59) 0.03(1) 25.07(31) 0.4(5) b.d.l. 6.53(33) 7.29(18) 0.36(2) b.d.l. b.d.l. 99.20

Grt(8) 37.94(27) 1.25(5) 19.92(13) 23.02(17) 6.45(8) 9.89(29) 0.11(2) b.d.l. b.d.l. b.d.l. 98.58 0.33

Sulph(3) 0.13(12) 0.04(2) b.d.l. 60.28(21) b.d.l. 0.1(2) b.d.l. b.d.l. 39.36 b.d.l. 99.91

1.5GPa, 950°C

Gl(10) 59.95(38) 1.22(12) 18.61(23) 5.77(13) 1.75(6) 4.53(10) 5.77(20) 2.15(6) 0.03(1) 0.17(2) 100.00 0.35 6.81(13) 0.99(10) 338±112

Grt(10) 39.23(21) 1.03(16) 20.91(19) 19.92(31) 10.92(31) 7.35(14) 0.04(2) b.d.l. b.d.l. b.d.l. 99.40 0.49

Cpx(5) 47.36(35) 1.22(10) 9.41(24) 10.26(53) 10.82(18) 17.38(48) 1.69(4) 0.04(1) b.d.l. b.d.l. 98.18 0.65

Sulph(1) 0.18 0.06 0.04 59.52 b.d.l. 0.19 b.d.l. b.d.l. 38.2 b.d.l. 98.19

Nickel-nickel oxide buffer (NNO)

1.0GPa, 850°C

Gl(8) 65.74(13) 0.50(2) 18.50(8) 2.63(11) 0.51(2) 3.12(1) 6.07(21) 2.73(9) 0.02(1) 0.17(1) 100.00 0.26 7.76(18) 0.58(1) 164±56

Hbl(10) 41.28(61) 1.94(53) 13.34(97) 16.33(85) 10.34(47) 10.11(30) 3.23(15) 0.73(4) b.d.l. 0.07(1) 97.37 0.53

Plag(8) 59.23(82) 0.03(2) 25,52(58) 0.31(6) b.d.l. 7.08(57) 8.03(25) 0.38(6) b.d.l. b.d.l. 100.58

Ilm(5) 0.51(42) 48.66(31) 0.40(15) 44.29(40) 2.89(8) 0.25(5) b.d.l. 0.06(2) b.d.l. b.d.l. 97.06

Sulph(4) 0.65(20) 0.13(4) 0.22(14) 62.31(18) b.d.l. 0.23(9) b.d.l. b.d.l. 38.69(11) b.d.l. 102.05

5

1.0GPa, 900°C

Gl(7) 63.78(29) 0.63(2) 19.11(18) 3.33(9) 0.73(5) 3.94(10) 5.87(19) 2.40(5) b.d.l. 0.19 100.00 0.28 7.79(19) 0.55(1) (b.d.l.)

Hbl(10) 40.47(83) 3.01(31) 14.45(37) 15.26(83) 10.08(60) 10.70(26) 3.06(10) 0.81(7) b.d.l. 0.06(1) 97.9 0,54

Plag(10) 57.13(66) 0.07(9) 27.02(55) 0.43(15) 0.04(2) 9.01(50) 6.56(37) 0.31(6) b.d.l. b.d.l. 100.57

Ilm(1) 0.16 49.55 0.35 44.28 3.2 0.21 b.d.l. 0.05 b.d.l. b.d.l. 97.80

Sulph(5) 0.25(23) 0.05(2) b.d.l. 60.88(99) b.d.l. 0.11(5) b.d.l. b.d.l. 38.18(114)

b.d.l. 99.47

1.0GPa, 950°C

Gl(21) 60.59(37) 1.11(4) 19.30(18) 4.24(10) 1.76(7) 5,00(10) 5.66(18) 2.11(6) 0.02(1) 0.19(1) 100.00 0.42 7.40(8) 0.64 217±97

Hbl(11) 41.39(67) 2.79(53) 14.69(66) 11.26(70) 12.94(70) 10.76(36) 3.05(10) 0.72(6) b.d.l. 0.05(1) 97.65 0.67

Ilm(1) 0.48 50.68 0.56 39.4 6.19 0.56 0.11 0.04 b.d.l. b.d.l. 98.02

Sulph(4) 0.28(2) 0.09(1) 0.09(6) 60.95(51) b.d.l. 0.17(3) b.d.l. b.d.l. 36.72(19) b.d.l. 98.30

1.5GPa, 850°C

Gl(13) 68.57(25) 0.35(2) 17.80(16) 1.99(10) 0.31(2) 2.42(5) 5.63(14) 2.71(10) 0.01(1) 0.21(2) 100.00 0.22 8.26(11) 0.67(2) 116±59

Hbl(9) 39.71(56) 2.62(38) 14.81(61) 18.79(76) 7.26(53) 9.41(36) 3.07(13) 0.96(8) b.d.l. 0.14(2) 97.77 0.41

Plag(9) 61.06(50) 0.04(3) 24.06(60) 0.32(6) b.d.l. 5.42(54) 8.02(32) 0.44(9) b.d.l. b.d.l. 99.36

Grt(3) 37.57(53) 1.24(26) 20.39(38) 23.36(55) 4.16(14) 12.06(59) 0.14(1) b.d.l. b.d.l. b.d.l. 98.92 0.24

Ilm(1) 0.21 51.4 0.34 42.02 1.97 0.27 b.d.l. b.d.l. b.d.l. b.d.l. 96,21

Sulph(1) 0.29 0.04 0.08 61.48 b.d.l. 0.08 b.d.l. b.d.l. 38.21 b.d.l. 100.18

Ttn(3) 30.19(5) 35.98(19) 2.33(20) 0.97(8) b.d.l. 28.00(21) b.d.l. 0.06(2) b.d.l. b.d.l. 97.53

1.5GPa, 900°C

Gl(10) 64.74(!6) 0.71(3) 18.99(23) 2.75(10) 0.74(4) 3.43(6) 6.02(10) 2.41(5) 0.02(1) 0.18(2) 100.00 0.32 7.42(38) 0.92(8) 188±135

Hbl(5) 39.17(64) 3.35(34) 16.31(53) 15.76(76) 8.48(65) 10.00(21) 3.06(14) 0.93(7) b.d.l. 0.09(1) 97.15 0.49

6

Plag(5) 58.24(30) b.d.l. 25.78(35) 0.35(10) b.d.l. 7.43(23) 6.76(10) 0.34(9) b.d.l. b.d.l. 98.90

Grt(9) 37.93(29) 1.52(13) 20.57(23) 21.90(52) 6.26(27) 10.74(43) 0.12(2) b.d.l. b.d.l. b.d.l. 99.04 0.34

Sulph(1) 0.57 0.16 0.35 60.9 b.d.l. 0.15 b.d.l. 0.06 38.42 b.d.l. 100.61

1.5GPa, 950°C

Gl(10) 62.18(25) 1.12(4) 19.16(21) 3.61(10) 1.34(5) 4.18(8) 5.99(11) 2.21(10) 0.02(1) 0.18(1) 100.00 0.4 7.43(8) 1.07(9) 191±112

Hbl(4) 39.78(42) 3.44(39) 16.21(76) 12.91(55) 10.67(65) 9.83(17) 3.05(22) 0.81(4) b.d.l. 0.08(0) 96.87 0.6

Grt(17) 38.53(26) 1.39(17) 21.13(28) 20.65(33) 8.66(30) 9.21(22) 0.09(4) b.d.l. b.d.l. b.d.l. 99.66 0.43

Cpx(5) 47.91(53) 1.16(17) 8.62(63) 10.73(99) 10.90(90) 17.58(90) 1.93(17) 0.05(5) b.d.l. b.d.l. 98.88 0.64

Sulph(1) 0.25 0.11 b.d.l. 61.38 b.d.l. 0.3 b.d.l. b.d.l. 38.31 b.d.l. 100.35

Cobalt-cobalt oxide buffer (NNO-1.7)

1.0GPa, 850°C

Gl(9) 64.52(24) 0.49(2) 19.94(13) 2.40(12) 0.62(5) 4.25(7) 5.53(12) 2.25(8) 0.03(1) 0.16(2) 100.00 0.28 9.48(43) 0.33(9) 277±122

Hbl(8) 42.58(84) 1.91(24) 13.84(75) 13.48(80) 12.38(49) 10.56(42) 3.02(10) 0.63(2) b.d.l. 0.03(2) 98.43 0.62

Ilm(7) 0.18(7) 50.31(84) 0.23(3) 43.24(27) 3.01(9) 0.18(5) b.d.l. b.d.l. b.d.l. b.d.l. 97.15

Sulph(4) 0.45(33) 0.08(1) 0.16(11) 62.24(29) b.d.l. 0.21(10) b.d.l. b.d.l. 37.76(30) b.d.l. 100.9

Ttn(1) 31.6 36.7 2.6 0.9 0.1 27.6 b.d.l. b.d.l. b.d.l. b.d.l. 99.5

1.0GPa, 900°C (+gr)

Gl(8) 63.81(28) 0.62(3) 20.09(29) 2.17(12) 0.74(2) 4.30(8) 5.79(14) 2.27(6) 0.02(1) 0.17(2) 100.00 0.34 8.99(14) 0.23(1) 208±109

Hbl(6) 42.29(82) 2.36(92) 13.89(33) 13.05(92) 12.44(93) 10.26(68) 3.20(8) 0.66(7) b.d.l. 0.05(1) 98.29 0.63

Ilm(1) 0.48 51.06 0.39 40.62 4.64 0.23 0.1 0.07 b.d.l. b.d.l. 97.59

Sulph(3) 0.07(6) b.d.l. b.d.l. 63.17(22) b.d.l. 0.08 0.06(1) b.d.l. 38.29(2) b.d.l. 101.67

Ttn(5) 30.69(75) 36.10(57) 2.37(27) 0.50(70) 0.09(2) 27.55(16) 0.05(2) 0.05(2) b.d.l. b.d.l. 97.35

7

1.0GPa, 950°C

Gl(11) 60.85(23) 0.91(2) 20.52(20) 3.42(13) 0.98(4) 5.11(9) 5.77(14) 2.21(7) 0.02(1) 0.20(1) 100.00 0.34 8.44(14) 0.57(2) 166±100

Hbl(7) 40.46(40) 3.01(64) 15.97(55) 10.54(32) 12.72(31) 11.00(24) 3.17(4) 0.78(3) b.d.l. 0.04(1) 97.69 0.68

Ilm(2) 0.78(40) 49.3(40) 0.6(14) 42.09(13) 4.10(27) 0.43(6) 0.09(3) 0.07 b.d.l. b.d.l. 97.46

Sulph(4) 0.80(73) 0.10(3) 0.32(31) 61.50(44) b.d.l. 0.24(10) b.d.l. b.d.l. 37.47(52) b.d.l. 100.43

1.5GPa, 850°C (+gr)

Gl(10) 66.09(25) 0.41(3) 19.39(13) 1.85(11) 0.45(4) 3.37(5) 5.88(16) 2.32(5) 0.03(1) 0.18(2) 100.00 0.3 9.31(3) 0.10 258±113

Hbl(7) 39.76(49) 2.95(11) 15.87(52) 17.28(88) 7.34(35) 9.86(23) 3.16(12) 0.97(5) b.d.l. 0.13(2) 97.32 0.43

Grt(8) 38.00(18) 1.53(16) 20.63(42) 20.96(79) 4.44(29) 13.42(53) 0.12(3) b.d.l. b.d.l. b.d.l. 99.10 0.27

Ilm(2) b.d.l. 50.93(38) 0.39(6) 41.91(3) 2.76 0.97(13) b.d.l. b.d.l. b.d.l. b.d.l. 96.96

Sulph(3) 0.12(6) 0.06(3) 0.05(3) 62.03(30) b.d.l. 0.11(3) b.d.l. b.d.l. 38.13 b.d.l. 100.50

Ttn(4) 30.46(20) 35.73(44) 2.67(27) 0.96(5) 0.05(3) 27.92(9) 0.04(3) 0.07(3) b.d.l. b.d.l. 97.92

1.5GPa, 900°C (+gr)

Gl(10) 62.24(19) 0.73(2) 19.81(25) 2.32(10) 0.80(7) 3.98(12) 5.65(10) 2.25(7) 0.03(1) 0.17(1) 100.00 0.38 9.26(32) 0.14(1) 287±138

Hbl(9) 39.16(62) 3.44(19) 17.56(33) 13.95(99) 9.19(63) 10.58(22) 3.09(9) 1.03(4) 0.04(1) 0.08(1) 98.12 0.54

Sulph(6) 0.34(10) 0.08(3) 0.09(2) 61.87(19) b.d.l. 0.18(3) b.d.l. b.d.l. 38.26(31) b.d.l. 100.82

Ttn(1) 31.54 35.62 3.03 0.84 0.1 27 0.28 0.16 b.d.l. b.d.l. 98.57

1.5GPa, 950°C (+gr)

Gl(10) 61.83(29) 1.28(3) 20.01(18) 1.85(5) 2.36(8) 5.00(9) 5.40(16) 2.03(7) 0.04(1) 0.15(2) 100.00 0.70 8.49(28) 0.21(1) 399±61

Cpx(9) 50.45(32) 0.9(12) 6.55(69) 7.86(89) 13.49(49) 18.83(59) 1.32(16) 0.05(2) b.d.l. b.d.l. 99.45 0.75

Sulph(4) 0.22(5) 0.07(2) 0.05(1) 61.66(24) b.d.l. 0.17(4) b.d.l. b.d.l. 38.39(12) b.d.l. 100.56

8

Composition of quenched glasses (gl) is presented as anhydrous normed to 100wt.%. Anhydrite analyses are not present due to interferences caused by neighboured crystals and melt during analyses (see text). Detailed concentrations of S in glass in ppm are given on the right side.

(a) (b)

Temperature, °C

log ƒ

O2

800700 900 1000 1100 1200 1300

-16

-14

-12

-10

-16

-8

-6

-4

-2

∆NNO=0

+4

+3

+2

+1

-1

-2

Re-ReO2

Ni-NiO

Co-CoO

H-M

S6+

S2-

ƒO2(∆NNO)

20

40

60

0

-6 -4 -2 0 +2 +4 +6

Implied Arc

Mantle (from lavas)

Range for Lavas from

Arcs (orange)

and MORB settings

Cumulative Measurements

for Shallow Mantle

Peridotites

Black - Abyssal Oceanic

Grey - Continental

Orange - Supra-Arc

Sulphate bearing lavas;

EL= El ChiconNR = Navada del RuizP = Pinatubo

-6 -4 -2 0 +2 +4 +6

EL NR P

Arc Basalt Measured Range

Mexico

Shiveluch

GrenadaMORB

M&O’N 09 & L 05;10Evans 12

Avachinsky

Japan

Cascades

Grenada

Shiveluch

Mexico

graphite-COH

graphite-COH

CO2-H2O

CO2-H2O

-2

-1

0

1

2

3

4

5lo

gƒO

2∆NNO

Temperature, °C800 850 900 950 1000

1.0 GPa+ silicate melt + COH-fluid+ amphibole

magnetiteilmenite

anhydritepyrrhotite

plagioclase in

titanite in

-2

-1

0

1

2

3

4

5

logƒ

O2∆N

NO

Temperature, °C800 850 900 950 1000

clinopyroxeneamphiboleplagioclase

titanite

garnetanhydritepyrrhotite

magnetiteilmenite

1.5 GPa+ silicate melt + COH-fluid+ pyrrhotite

plagio

clase

in

titanite ingarnet incl

inop

yrox

ene

in

ilmen

ite in

amph

ibol

e in

(a)

(b)

100%

80%

60%

40%

20%

0%875 870 841 950 900 850

Re-ReO2

100%

80%

60%

40%

20%

0%

950 900 850

Re-ReO2

100%

80%

60%

40%

20%

0%950 900 850

Ni-NiO100%

80%

60%

40%

20%

0%950 900 850

Co-CoO100%

80%

60%

40%

20%

0%

950 900 850

Hematite-Magnetite Ni-NiO100%

80%

60%

40%

20%

0%

glass

fluid amphibole

plagioclase magnetite/ilmeniteanhydrite

pyrrhotitegarnet

clinopyroxene

titanite

950 900 850

Co-CoO100%

80%

60%

40%

20%

0%

Run temperature, °C

Mod

al p

ropo

rtion

, wt%

Run temperature, °C

Mod

al p

ropo

rtion

, wt%

950 900 850

Re-ReO2

100%

80%

60%

40%

20%

0%

Run temperature, °C

0.5GPa 1.5GPa

1.0GPa

Picro-

basalt

BasaltBasalticandesite

AndesiteDacite

Rhyolite

Trachyte

iayBasaltictrachy-andesite

Trachy-basalt

Tephrite or

Basanite

Phono-Tephrite

Tephri-phonolite

Phonolite

Foidite

35 40 45 50 55 60 65 70 75

0

2

4

6

8

10

12

14

16

Trachy-andesite

Trachydacite

Na

2O

+K

2O

in m

elt, w

t%

SiO2 in melt, wt%

NNO+2 NNO NNO-1.51.5GPa:

NNO+4.75 NNO+2 NNO NNO-1.51.0GPa:

Starting composition

0.10

0.15

0.20

0.25

0.30

Cl in

melt, w

t%

800 850 900 950 1000

0.10

0.15

0.20

0.25

0.30

Cl in

melt, w

t%

800 850 900 950 1000

800 850 900 950 1000

2.0

2.5

3.0

3.5

K2O

in m

elt,

wt%

800 850 900 950 1000

2.0

2.5

3.0

3.5

K2O

in m

elt,

wt%

800 850 900 950 1000

800 850 900 950 1000

800 850 900 950 1000

17

18

19

20

21

Al 2O

3 in

me

lt,

wt%

2

3

4

5

6

CaO

in m

elt,

wt%

S in

melt, ppm

0

200

400

800

600

1000

800 850 900 950 1000

800 850 900 950 1000

800 850 900 950 1000

17

18

19

20

21

Al 2O

3 in

me

lt,

wt%

2

3

4

5

6

CaO

in m

elt,

wt%

S in m

elt, ppm

0

200

400

800

600

1000

Temperature, °C Temperature, °C

NNO+2 NNO NNO-1.7

1.5GPa:NNO+4.75

NNO+2 NNO NNO-1.7

1.0GPa:

H2O in melt, wt%

CO

2 in

mel

t, w

t%

0

0.4

1.2

0.8

0.2

0.6

1.0

40 12 16 208

900°C

850°C900°C950°C

Re-ReO2 Hematite-Magnetite Re-ReO2

Ni-NiO Co-CoO

Re-ReO2

Ni-NiO Co-CoO

Ghiorso and Gualda2015: 0.5GPa 1.0GPa 1.5GPa

0.5GPa

1.0GPa

This study:

800 850 900 950 1000

H2O

in m

elt,

wt%

5

6

7

8

9

10

800 850 900 950 1000

H2O

in m

elt,

wt%

5

6

7

8

9

10

800 850 900 950 1000Temperature, °C

CO

2 in

mel

t, w

t%

0

0.2

0.4

0.6

0.8

1.2

1.0

800 850 900 950 1000Temperature, °C

CO

2 in

mel

t, w

t%

0

0.2

0.4

0.6

0.8

1.0

1.2

(b)(a) 1.5GPa1.0GPa

1.5GPa1.0GPa

S in m

elt,

ppm

logƒO2

∆NNO

-3 -2 -1 0 +1 +2 +3 +4 +5

0

200

400

600

800

1000

1200

14001.0GPa

900°C

850°C

950°C

900°C

850°C

950°C

S6+S2-

S0

S6+S2-

S0

XS6+in glass

S in m

elt,

ppm

logƒO2

∆NNO

-3 -2 -1 0 +1 +2 +3 +4 +5

0

200

400

600

800

1000

1200

14001.5GPa

logƒO2

∆NNO

-3 -2 -1 0 +1 +2 +3 +4 +50

1.0

0.8

0.6

0.4

0.2

1.0GPaX

i

1.5GPa

Xi

(a)

(b)logƒO

2

∆NNO

-3 -2 -1 0 +1 +2 +3 +4 +50

1.0

0.8

0.6

0.4

0.2

logƒO2

∆NNO

-3 -2 -1 0 +1 +2 +3 +4 +50

1.0

0.8

0.6

0.4

0.2

S6

+

(c)

1.0

GP

a

1.5

GP

a

0.2

GP

a(†

)

0.5

GP

a

NNONNO NNO 1

1.5GPa:

NNO 5 NNONNO NNO 1

1.0GPa:

† basalt Jugo et al. (2010)

0.2GPa/1050°C and

andesite Botcharnikov et al. (2011)

0.2GPa/1050°C

FeS CaSO4

FeS CaSO4

950°C

Cu-enrichment

in fluid saturated

and dry melts

1.5

Pre

ssure

, G

Pa

o gen ugacit ∆NNO-5 -4 -3 -2 -1 0 +1 +2 +3 +4 +5

1.0

0.5

0

~D

epth

, km

0

50

40

30

20

10

gra hite CO luid e uilibriuUlmer and Luth (1991)

entrance of “sulphate only”

stabilit ield this studentrance of “sulphate only”

stabilit ield Botcharnikov et al. (2011)

graphite saturated fluids

no Cu/Au-enrichement

Cu-enrichment

in fluid saturated melts

Experimental data:

FeS

stable

CaSO4

stable

Figure S1: Composition of silicate melts in experiments performed at 1.0 and

1.5GPa.Thevaluesarecalculatedonanhydrousmeltbasisandrepresentmean

valuesof7-21analyses;errorbarsreferto±1σreproducibility.

59

61

63

65

67

69

1

2

3

4

5

6

5.0

5.5

6.0

6.5

0.10

0.15

0.20

0.25

0.30

Cl i

n m

elt,

wt%

Temperature, °C800 850 900 950 1000 800 850 900 950 1000 800 850 900 950 1000

Temperature, °C Temperature, °C

800 850 900 950 1000

800 850 900 950 1000

800 850 900 950 1000 800 850 900 950 1000

800 850 900 950 1000

800 850 900 950 1000 800 850 900 950 1000

800 850 900 950 1000

800 850 900 950 1000

Na 2O

in m

elt,

wt%

FeO

* in

mel

t, w

t%Si

O2 i

n m

elt,

wt%

TiO

2 in

mel

t, w

t%

2.0

1.5

1.0

0.5

0M

gO in

mel

t, w

t%

2.0

2.5

3.0

3.5

K 2O in

mel

t, w

t%H

2O in

mel

t, w

t%17

18

19

20

21

Al2O

3 in

mel

t, w

t%

2

3

4

5

6

CaO

in m

elt,

wt%

S in

mel

t, pp

m

0

200

400

800

600

1000

CO

2 in

mel

t, w

t%

1

3

2

1

0

5

6

7

8

9

10

0

0.2

0.4

0.6

0.8

∆NNO+2 ∆NNO±0 ∆NNO-1.5

1.0

1.2

1.5GPa

FigureS2:CalculatedvaluesforKAandKBconstantsat1.0and1.5GPa.WhilstKB

mostlycontrolsthepositionof!!!!+,KAdeterminesthepositionof!!!! and!!!!lines.

23

25

logK

B

logKA

27

29

31

33

35

8 9 10 11 12 13 13 15

1.0GPa

1.5GPa

incre

asing

tempe

ratur

e

FigureS3:AlternativemodeltoFig.8,with(a)S-speciationand(b)S-contentsin

silicatemeltsat1.0GPaassumingthepresenceofS4+-intermediatespecies

insteadofS0.NotethemaximumforS4+isclosertotheoxidizedfieldcompared

toS0(Fig.8).ThiswouldlimitthestabilityfieldofS2-tobelowNNO+1which

contradictstheexperimentalobservations.

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

-3 -2 -1 0 +1 +2 +3 +4 +5

S2-

S4+S6+

Xi

logƒO2∆NNO

-3 -2 -1 0 +1 +2 +3 +4 +5

logƒO2∆NNO

0

200

400

600

800

1000

1200

1400

S in

mel

t, pp

m

950°C

900°C

850°C

FeS CaSO40

1.0 GPa(a)

(b)