Benjamin S. Murphy
a,b,∗, Jan Marten Huizenga
c,d,e, Paul A. Bedrosian
aaGeology,Geophysics,andGeochemistryScienceCenter,U.S.GeologicalSurvey,Denver,CO 80225,USA bGeomagnetismProgram,U.S.GeologicalSurvey,Golden,CO 80401,USA
cFacultyofEnvironmentalSciencesandNaturalResourceManagement,NorwegianUniversityofLifeSciences,P.O.Box5003,NO-1432,Ås,Norway dEconomicGeologyResearchCentre,CollegeofScienceandEngineering,JamesCookUniversity,Townsville,Queensland4811,Australia eDepartmentofGeology,UniversityofJohannesburg,AucklandPark2006,SouthAfrica
a rt i c l e i n f o a b s t r a c t
Articlehistory:
Received16June2021
Receivedinrevisedform23June2022 Accepted28June2022
Availableonlinexxxx Editor: R.Dasgupta
Keywords:
mineralsystems IOA-IOCG OlympicDam
SoutheastMissouriIronProvince magnetotellurics
graphite
Magnetotelluric (MT) imaging results from mineral provinces in Australia and in the United States showanapparentspatialrelationshipbetweencrustal-scaleelectricalconductivityanomaliesandmajor magmatic-hydrothermaliron oxide-apatite/ironoxide-copper-gold (IOA-IOCG) deposits.Although these observationshave drivensubstantial interest inthe use ofMTdata toimage ancientfluidpathways, the exact cause of these anomalies has been unclear. Here, we interpret the conductors to be the resultofgraphiteprecipitationfromCO2-richmagmaticfluidsduringcooling.Thesefluidswouldhave exsolvedfrommafic magmasatmid- tolower-crustal depths;salinemagmatic fluidsthat coulddrive mineralizationwerelikelyderivedfromrelated,moreevolvedintrusionsatshallowercrustallevels.Inour model,theconductivityanomaliesthenmarkzonesthatoncewerethedeeprootsofancientmagmatic- hydrothermalmineralsystems.
PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense (http://creativecommons.org/licenses/by/4.0/).
1. Introduction
The mineral systems approach to resource exploration has arisen as a novel method to vector towards ore deposits. This paradigmaimstointegratedisparategeoscientificdatainorderto formulateamodeloflithosphericarchitectureandhistorythatcan be usedtoidentifylocationsthat wereonceconducivetodeposit formation(e.g.,McCuaigandHronsky,2014).
Magnetotelluric (MT) imaging has emerged as a key tool in themineralsystemsframework,particularlyduetoobservedspa- tialrelationships betweenlithosphere-scaleelectrical conductivity anomaliesandironoxide-apatite(IOA)andironoxide-copper-gold (IOCG) deposits.The premier examples ofsuch correlations have come fromthe OlympicDam region (Heinsonet al., 2006, 2018) and theCloncurry District (Jianget al., 2019; Wang etal., 2018) inAustralia aswell asfromthe SoutheastMissouriIron Province (SMIP) in theUnited States (Fig. 1; e.g., McCafferty et al., 2019).
At the mineral-province scale, MT imaging reveals highly con- ductive (>10−1 S/m),steeply dippingpipe- or sheet-likeanoma-
*
Correspondingauthor.E-mailaddress:[email protected](B.S. Murphy).
liesatmid-lowercrustaldepths aswell asmoderatelyconductive (∼10−2 S/m)“fingers” that extendinto the uppercrust beneath individualdeposits(e.g.,Fig.1).
Presently intracratonic, Mesoproterozoic-aged IOA-IOCG de- positsin theOlympicDamregionandintheSMIPare viewedas magmatic-hydrothermalin origin.Genetic modelsinvokemantle- derived magmatism in an extensional setting (back-arc, orogenic collapse, or intraplaterift) to provide heat and fluids, aswell as possiblymetals,intothesystem(e.g.,Dayetal.,2016;McCafferty etal.,2019;Skirrowetal.,2018;seealsoFig.1C);mixingofthese magmaticfluidswithnear-surfacefluids(meteoricwatersorbasi- nal brines)in the upper crust is inferred to have driven deposit formation(e.g.,Barton,2014;Schlegeletal.,2020).Inthiscontext, theco-locatedelectricalconductivityanomalieshavelooselybeen interpreted asthe signature ofmetasomatism along crustal-scale magmatic fluid pathways (e.g., Heinson et al., 2018); however, a physically rigorous explanation of these apparently associated conductors is lacking. Here, we suggest that these conductivity anomaliesarespecificallytheresultofgraphiteprecipitationfrom CO2-richmagmaticfluids.Weproposeaquantitativegeneticmodel that links these crustal-scale anomalies to the formation of the IOA-IOCGdeposits,andindoingsoweunderscorethevalueofMT imaginginthemineralsystemsframework.
https://doi.org/10.1016/j.epsl.2022.117700
0012-821X/PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense(http://creativecommons.org/licenses/by/4.0/).
Fig. 1.GeophysicalimagesfromtheSoutheastMissouriIronProvince(SMIP),USA.(A-B)Depthslice(at35kmdepth)andcrosssection(roughlyacrosstheaxisofextension intheSMIP;cf.DeLuciaetal.,2019)throughanewMT-derivedelectricalconductivitymodel(detailsoftheinversionmethodologyareprovidedintheSupplemental Materials).Invertedtrianglesdenoteironoxide-apatite(IOA)andironoxide-copper-gold(IOCG)deposits.Notethemoderatelyconductive(∼10−2S/m)“finger”thatextends intotheuppermostcrustbeneathoneofthesedepositsinthecrosssection.(C)CrosssectionthroughthemagneticsusceptibilitymodelofMcCaffertyetal.(2019).The highsusceptibilitybodyinthisimagehasbeenpreviouslyinterpretedasthesignatureofatrans-crustalintrusivesystem.(D)Discriminationdiagramthatdemonstratesthe spatialrelationshipsbetweenimagedconductivity(A-B)andsusceptibility(C).Valuesaresampledona2.5km×2.5 kmhorizontalgridacrosstheregionshownin(A)at1 kmintervalsfrom10-40kmdepth.
2. Conductivitymechanisms
Mostexplanations forcrustal electricalconductivityanomalies implyspecificphysicochemicalconditionsthatdonotmatchinfer- encesfortheOlympicDamregionandtheSMIP.Here,wedemon- stratethatgraphiteistheonlyreasonableexplanation,andindeed the best explanation,forthe crustal-scale conductivityanomalies observedintheseregions.
2.1. Freefluids
Although often invoked in tectonically active domains, deep saline fluidsare implausible in the ancient intracratonic settings inwhichthesedepositsarelocated.Fluidsassociatedwithoriginal tectonomagmatismwouldhavebeenconsumedby retrogradehy- drationreactionsonatimescaleof∼10 Ma(e.g.,Manning,2018;
Yardley andValley, 1997; see alsoYardley andBodnar,2014). As these regions have not experienced regionaltectonism in >1 Ga, there is no modern source fordeep fluid delivery into the mid- lowercrust,andweconsiderithighlyunlikelythatmeteoricfluids wouldpercolatetosuchgreatdepthsunderpresumablylithostatic pressuregradients.
2.2. Metallicsulfidesandoxides
Metallic sulfides (e.g., pyrite, pyrrhotite) are electrically con- ductive (∼103−105 S/m; Keller, 1966); however, we discount suchphasesgiventheavailablegeometricandtectonicconstraints.
Metasedimentary pyrite has been shown to be partly responsi- bleforconductivityanomalies inpaleo-subduction settings(Jones et al., 1997), but Olympic Dam and the SMIP are not located within suture zones. Subvertical, crustal-scale sulfide zones are problematic explanations in anyother tectonic setting, asunrea- sonablylargeamounts(>10 wt%;NelsonandVan Voorhis,1983) ofintrusion-relatedsulfidesthroughoutthemid-lowercrustwould be required to explain the observed anomalies. Although lower- crustal sulfide-rich intrusivebodies are recognized inthe surface rockrecord(e.g.,intheIvreaZone,northernItaly;Fiorentinietal.,
2018),onlylocallywithinthoseintrusionsdoesthesulfidecontent become high enough that a bulk conductivity anomaly (>10−1 S/m)wouldbeexpected(cf.Vukmanovicetal., 2019; Nelsonand VanVoorhis, 1983). Therefore,weconsiderit highlyunlikely that sulfidesalonecanexplaintheobservedhighconductivityvaluesin theOlympicDamregionandintheSMIP.
Metallic oxides (e.g., magnetite, ilmenite) are similarly elec- trically conductive (∼100−103 S/m; Keller, 1966), and labora- toryexperimentsindicatethattheycanbecomeinterconnectedat moderatevolumefractions(∼5%)toproduce abulk conductivity anomaly(Daietal.,2019).However,thesephasesalsooftenhave veryhighmagneticsusceptibilities(cf.Huntetal.,1995).Conduc- tive zonesbeneaththe SMIPare associatedwithdomains of low magnetic susceptibility (Fig. 1D), so such phases cannot be the causeoftheobservedhighconductivityvalues.
In both of these cases, relatively high volume fractions (>
10−2−10−1)oftheconductivephaseatspatialscalesrangingfrom 10sofmicrons(Daietal.,2019)to10sofmeters(NelsonandVan Voorhis, 1983) are needed to enhance bulk electrical conductiv- ity.This requirement isdue to the characteristicequant habit of metallicsulfidesandoxides.Onlyatsuchhighconcentrationscan thesemineralsformatexturally interconnectednetworkthat can produceaconductivityanomaly.
2.3.Silicateminerals
Electricalconductioninsilicatemineralsisthermallyactivated, sothesephasesareincapableofproducingamajorelectricalcon- ductivityanomalyatthecoldcrustaltemperaturesinthesestable Precambriandomains.Lower-crustaltemperaturesare<600◦Cat present in the SMIP (e.g., Schutt et al., 2018); at these temper- atures, intrinsicsemiconduction in potentially volumetrically sig- nificant silicates(e.g., feldspars,amphiboles, epidote) wouldpro- duce bulk conductivity values <10−2 S/m (e.g.,Hu etal., 2013, 2017,2018).Extrinsic,volatile-mediatedsemiconductionineither micas (e.g., phlogopite) or nominally anhydrous minerals (e.g., feldspars,pyroxenes) would similarly only produce bulk conduc- tivity values of ∼10−2 S/m at these cold temperatures (e.g.,
grainsizes (e.g.,Hanetal.,2021;Pommieretal., 2018).However, the degree of conductivity enhancement depends upon temper- ature. Crustal conductivity values >10−1 S/m at grain sizes of 1−100 μm(as would be expectedfor typical mid-lower crustal mylonitic shearzones; e.g., Okudairaet al., 2017; Waters-Tormey and Tikoff, 2007) generally require temperatures >800◦C (e.g., Han etal., 2021),whichisunreasonablefortheOlympicDamre- gionandtheSMIP(asdiscussedinSection2.3).
2.5. Graphite
Given the above considerations, graphite isthe only plausible causeoftheseconductivityanomalies,asthisphaseishighlycon- ductive(∼106S/m;Keller,1966)andcanbecomereadilyintercon- nected at very smallvolume fractions dueto its sheet-like habit (e.g., Glover, 1996) in order to produce a major bulk conductiv- ityanomaly.Graphitehas infactbeenwidely invokedina range of settings to explain observed electrical conductivity anomalies (e.g.,Evans,2012;GloverandÁdám,2008;Haaketal.,1997).For example, graphite is often used to explain crustal conductors in paleo-subductionsettings,wherebiogeniccarbonhasbeenrecrys- tallized tointerconnected graphite indeformed metasedimentary rocks(e.g.,Boerneretal.,1996).However,thisgeneticmodelisin- compatible withtectonic inferences for the Olympic Dam region and the SMIP. Instead,we view graphite here as having precipi- tated from a CO2-rich magmatically derived fluid. Such an inter- pretation has been qualitatively discussed previously to various extents (e.g.,Monteiro Santos etal., 2002;Selway, 2014),includ- ing in the contextof mineralsystems (e.g.,Heinson et al., 2006;
WannamakerandDoerner,2002);however,itsimportanceasasig- natureoftectonomagmatismhasnotbeenwidelyappreciated,and arigorousquantitativejustificationforthisinterpretationhasbeen criticallylacking.
There are several mechanismsby which graphite mayprecip- itate from magmatic fluids. In cooling rocks that contain titano- magnetite,the oxyexsolutionreaction 6Fe2TiO4 + CO2 = 2Fe3O4
+6FeTiO3+C(Frostetal.,1989),wheretheCO2isderivedfroma magmaticfluid,isonepossiblemeans toformgraphite.However, conductivity anomalies in the SMIP are spatially associated with domainsoflowmagneticsusceptibility(Fig.1D);thisrelationship suggeststhatthereislittlemagnetiteintheconductivezonesthat could havefacilitatedthisreaction.Fluid-rock hydrationreactions canalsodrivegraphiteprecipitation(e.g.,Rumble,2014).Suchre- actions,however,areoftenself-limitingbecausethewallrockgen- erallycannotserveasaninfinitesinkforwater(e.g.,Luque etal., 2014). Instead,we considergraphiteprecipitation via carbonsat- uration inacoolingCO2-rich fluidtobe themostlikelyscenario, as thismechanism doesnot impose anypetrologic requirements uponthesystemandassuchreactionscanfacilitatetheprecipita- tionoflargeamountsofgraphite(e.g.,Huizenga,2011).Thismode ofgraphitedepositionhasbeenwellstudied(e.g.,Huizenga,2011;
Luque etal., 2014; andreferencestherein) andisinfact thought
Fig. 2.Conceptualmodelforhydrothermalgraphiteprecipitationbeneathmagmatic- hydrothermalIOA-IOCGdepositsinanextensionalsetting(possiblywithsomecom- ponentoflateraltranslation;e.g.,DeLuciaetal.,2019).Mantle-derivedmeltsstall inthe lowercrustandexsolveaCO2-richfluid;asthismagmaticfluidcools,it precipitatesgraphite.Athighercrustallevels,magmaticfluidsderivedfrommore evolvedmelts,producedbycrustalassimilationandfractionalcrystallization(AFC) andcrustalanatexis,drivemineralizationviamixingwithanear-surfacefluid(e.g., Barton,2014).
tobecommoninthecrust(Rumble,2014).Consequently,wecon- siderhydrothermalgraphiteprecipitationfromacoolingCO2-rich magmaticfluid toin factbe a naturalexplanationforthesecon- ductivityanomalies.
3. Modelforgraphiteprecipitation
Inourconceptualmodel(Fig.2),graphiteprecipitatesfrommi- gratingCO2-rich fluidsthat are carried intothe crust by mantle- derived melts in an extensional setting. Although carbon could bescavenged fromlocalmetasedimentaryrocks, we considerthe mantleasthekeyquantifiablesourceforcarbon,asgeneticmodels fortheseIOA-IOCGdepositsinvokemantle-derivedmeltstoestab- lishthe magmatic-hydrothermal system(McCafferty et al., 2019;
Skirrowetal.,2018)andascarbonisotopesfromtheSMIPsupport theinput of mantle-derivedcarbon (Johnsonetal., 2016). Inour model,graphitelargelyprecipitatesatmid- tolower-crustallevels fromfluidsthat areexsolved fromstalled mafic melts;atupper- crustallevels,fluidsderivedfromrelated,moreevolvedmeltsdrive IOA-IOCGmineralizationviamixingwithnear-surfacefluids.
Inordertoevaluatewhetherourconceptualmodelcanexplain theMTobservations, we calculatetheamplitudeofthe electrical conductivityanomalythat wouldbeproduced giventheavailable carbonbudgetandphysicochemicalconstraints.Wesummarizethe three mainsteps ofthis calculation below;additional details are providedintheSupplementalMaterials.
3.1. Carbonbudget
First,we estimate theabsolute amountof carbonthat ismo- bilized from the mantle during magmatism. Estimates of carbon content inthe uppermantle are ∼30−1000 ppmwC (e.g., Lee etal., 2019); weconservatively assume an averagecontent of30
Fig. 3.CalculatedrelativeamountofCthatcanbedissolvedinthefluidatCsaturationforafixed fO2,expressedaslogunitoffsetfromtheFMQbuffer.Themaximum relativeamountofCthatcanbedissolvedinthefluidbeforegraphitesaturationis∼33.3 mol%,correspondingtonearlypureCO2.DecreasingCcontentindicatesthat graphiteprecipitatesoutofthefluid.ThearrowshowsthehypotheticalP/Tpath(25◦C/km)discussedinthetext.
ppmw C. Because CO2 is highly incompatible, we make a sim- plifying assumption that this full amount of C will be removed fromthemantle,regardless ofthedegree ofpartialmelting(e.g., Moore and Bodnar,2019). Geodynamic modeling studies suggest an∼1000−4000 km3 mantlesourcevolumeperkilometerofrift length (e.g., Schmeling, 2010). Taking the lower bound, we esti- matethat8.24×1012 molCismobilizedperkilometer alongthe axisofextensionduringmantlemelting.Wefurtherconservatively assumethat10%ofthisavailableC(8.24×1011molC)willbere- leased intothe mid-lowercrust via expulsionfromstalled melts, while the remainder is transported with melt into near-surface volcanicsystems.
3.2. Graphiteprecipitation
Next, we calculate the relative amount of carbon that would precipitate fromthe magmatic fluid underchanging pressure (P) andtemperature(T)conditions.Wemodelthisfluidwiththesys- temC-O-H.Althoughtheexsolvedmagmaticfluidwouldlikelybe saline,immiscibilityathighP/Tyields twoseparatefluids, aCO2- rich fluid and a hypersaline brine, that would evolve separately (e.g.,Manning, 2018; YardleyandBodnar,2014). TheC-O-Hfluid isthereforeausefulapproximationforourpurposeshere.Weuse the GFluid spreadsheet (Zhang andDuan,2010), modified to use the fayalite-magnetite-quartz (FMQ) oxygen fugacity (fO2) buffer ofBallhausetal.(1991),tocalculatethemaximumrelativeamount ofCatsaturation(i.e.,fluidCactivityisunity)thatcanberetained inthefluid asa functionofP/T.Inthesecalculations, weassume thatthewallrockservesasasinkforO2sothatthefluidisheldat constant fO2 duringgraphiteprecipitation.(IfO2 insteadremains inthefluid,thenfluid fO2 willincreaseduring graphiteprecipita- tion,andthenetgraphiteprecipitationpotentialwillthenbeless than withconstant fluid fO2; Huizenga,2011. However, thewall rock isexpected tobuffer the fluid to constant fO2; Selverstone, 2005.)
The resultsof thesecalculationsare shownin Fig.3. Inthese plots, the fractional difference betweenany two points gives the relativeamountofCthatisprecipitatedout of(ordissolved into) the fluid underchangingP/T conditions.The fluidbecomes satu- rated in C with decreasing T and will then precipitate graphite, which leads to the C decreasein the fluid shownin theseplots.
We assume a hypothetical P/T path (arrow in Fig. 3) represent- ing fluid migration out of the lower crust to higher crustal lev- els, withthe fluid cooling along an approximate average crustal geotherm during tectonomagmatism. (Although magmatic intru- sions willmodulate thecrustal thermalfield,we expectthat this path will generallycapture the average P/T changes experienced
during fluidflow. Furthermore,althoughmigrating fluidswillad- vectheat to some extent,they generallyshed much ofthat heat to reachthermal equilibriumwiththe rocks through which they pass;e.g.,Ague,2014.)Wefurtherassume thatthemantlesource region sets fluid fO2 to ∼FMQ; this value is reasonable for the uppermantle,eveninsubduction-modifiedsettings(FrostandMc- Cammon,2008).Igneouszircon-hostediron oxideinclusionsfrom theSMIPalsoprovidesupportingevidencethatredoxconditionsof ore-stagemagmaswerenearFMQ(WattsandMercer,2020).(The IOA-IOCGdeposits themselves are moreoxidized due, at leastin part,tofluidmixing.)ForthisassumedP/Tpath,96% oftheCin themagmaticfluidisprecipitatedasgraphite.
3.3.Resultingconductivityanomaly
Finally,weevaluatethemagnitudeoftheconductivityanomaly that wouldbe produced by graphiteprecipitation fromthemag- maticfluid.CombiningourabsoluteCbudgetestimate(Section3.1) withourcalculatedrelativeamountofCprecipitatedalongtheas- sumed P/Tpath (Section3.2), we concludethat 7.89×1011 mol Cisprecipitatedasgraphiteduringmagmaticfluidmigration.We assume that the graphite is distributed over an area of 100 km across the axis of extension (based on the region of maximum surface deformation during extension fromgeodynamicmodeling studies; e.g., Schmeling, 2010) by 10km indepth (reflectingthe arrow in Fig. 3), per kilometer along the axis of extension. This would produce a bulk graphite volume fraction of 4.16×10−6. Although the precipitated graphite would likely be concentrated along discrete permeablefluid pathways (either tectonically con- trolledorcreatedbyfluidoverpressurization),thisvaluereflectsan overallbulk-averagevolumefractionacrossthismodeldomain.In ordertoevaluatetheexpectedmagnitudeoftheresultingelectrical conductivityanomaly,wethencomparethisvaluetovarioustwo- componentmixinglawsbetweenelectricallyresistiverock-forming silicatemineralsandelectricallyconductivegraphiteinFig.4.
ThekeycurvesinFig.4aretheHashin-Shtrikman(HS)bounds, whichprovidethemostgeneralupperandlowerlimitsonthebulk electrical conductivity of a two-component mixture (Hashin and Shtrikman,1962;seealsoSchmeling,1986).Theupperconductiv- ityboundreflectsasysteminwhichthemoreconductivephase(in thiscase,graphite)istexturallywellconnected,whereasthelower conductivityboundreflectsageometryinwhichthemoreconduc- tive phase is poorly connected. The other mixinglaws shownin Fig.4reflect specifictextural geometries ofthetwo phases, with curves plotting near the upper HS conductivity bound reflecting geometriesinwhichthemoreconductivephaseisgenerallyinter- connected.In comparing our calculated volume fraction to these
Fig. 4. Calculatedelectrical conductivity as afunction of graphitevolume frac- tionforvarioustwo-componentmixinglaws(Schmeling,1986;graphite:106S/m, Keller,1966;crystallinesilicatecrust:10−4S/m).Verticaldotted/dashedlinesde- notegraphitevolumefractionsdiscussedinthetext.Mixinglawsneartheupper Hashin-Shtrikman(HS)conductivityboundreflectsystemsinwhichtheconductive phaseiswellinterconnected;mixinglawsnearthelowerHSconductivitybound reflectsystemsinwhichtheconductivephaseispoorlyinterconnected.
mixinglaws, thedegree oftextural interconnection isthe crucial factor that dictateswhether such smallamounts of graphitewill enhancebulkelectricalconductivity.AsFig.4demonstrates,ifthe precipitatedgraphiteinourmodelisindeedwell connected,such thatitischaracterizedbytheupperconductivitybound,thenour calculated bulk-average graphite volume fraction of 4.16×10−6 (andeven bulkvolume fractionsan orderofmagnitudelessthan this value) will produce a conductivity anomaly comparable to whatisobservedbeneaththeseIOA-IOCGdeposits(∼100 S/m).
The graphite will indeed likely be well connected when pre- cipitatedfromamobilefluidalongpermeablepathways.Although CO2-rich fluidsthemselves arenot expectedto be readilymobile due to poor grain-boundary wetting (e.g., Manning, 2018), their mobility will be greatly enhanced by the additional presence of salinefluids(e.g.,NewtonandTsunogae,2014;Touretetal.,2019) that are commonin magmaticsystems (e.g.,Pirajno, 2021; Yard- ley andBodnar, 2014). Therefore, the related exsolved magmatic salinefluidinourmodelthat resultsfromfluidimmiscibility(see Section 3.2) willpromotemobilityoftheCO2-richfluid and,con- sequently,interconnectionoftheprecipitatedgraphite.(Regardless ofthecoexistenceofsalinefluids,observationsofpassiveCO2de- gassing inextensional settingsindicates that such CO2-richfluids are nevertheless mobile via some mechanism through the crust;
Tamburello et al., 2018.) Syngenetic shearing oftectonically con- trolledfluidpathwayscouldacttofurtherpromoteinterconnection (e.g.,Jödickeetal.,2004).
4. Discussion
4.1. Implicationsforregionalmineralexploration
Ourpurposefullyconservativecalculationsherevalidatethein- ference that electrical conductivityanomalies, in the correct tec- tonicsetting,maydemarcatezonesofcrustal-scalemagmaticfluid flow. However, there are many ways to produce a conductivity anomaly(asdiscussedinSection 2),andnotallanomalieswillbe linkedtooremineralization.Complementarygeological,geochemi- cal,andgeophysicalobservationsareneededtodeterminewhether hydrothermalgraphiteprecipitationmaybeareasonableinterpre- tationforagivenfeatureinanMTimagethatcouldthenprovidea
sameasthosethatdriveoremineralization;formationofmineral depositsatuppercrustal depthsislikely theresultofmixingbe- tweenmore evolvedmagmaticfluids, derivedfrommore evolved melts, and near-surface fluids (Fig. 2; Barton, 2014; Schlegel et al., 2020). Consequently, inourmodel, graphiteprecipitation and mineral deposit formation are two independent components of a single mineral system. When combined withother datasets to constraintectonicsetting,MT-imagedconductivityanomaliesmay then beindicative ofa paleo-tectonomagmatic systemthat could havedrivenoremineralizationatuppercrustallevels.
Furthermore,asgraphiteprecipitationandmineraldepositfor- mationoccurindisparatecrustalintervalswithintheoverallmin- eral system, there should be no expectation that the graphite thatcausesthesemid-lowercrustalconductivityanomalieswillbe significantly exposed in the upper crust, at the deposit or even regional district scale. Although this mismatch in crustal levels presentsa challenge in validatingour model,examination of ex- posedlower crustalsectionsmayprovide valuableinsights inas- sessingtherole ofhydrothermalgraphite precipitationinmineral systems(seeSection4.4).
4.2.Redoxrequirementsandimplications
Althoughwehavepurposefullymadeconservativeestimatesin ourCbudgetcalculations,ourresultsaredependentuponthere- dox conditions of the magmatic fluid. The C mol% isopleths in Fig.3changesignificantlywhenmovingawayfromtheFMQbuffer.
The distribution ofthese isoplethsdetermines both thedepth at whichgraphitewill precipitate fromthe fluid(based on P/Tcon- ditions through the crust) and how much cooling is required to precipitate a fixed amount ofgraphite (more reducedconditions requiremorecoolingforthesameCprecipitation).However,ifthe fluid fO2 weretodeviate significantlyfromFMQ,evenprecipitat- ing just 3% of the available C (1 mol% of fluid C) still results in asignificantbulkconductivityanomaly(∼10−1 S/m,comparedto
∼10−4 S/mcrystallinecrustal background;Fig.4), aslongasthe graphiteis texturally interconnected.(Similarly, iffluid fO2 were to increase during graphite precipitation due to violation of our assumptions in Section 3.2, an order of magnitudeless graphite wouldstill yieldabulkconductivityanomaly comparabletowhat isobservedinFig.1.)
Thehighestamplitudeportions oftheobservedanomalies be- neath Olympic Dam and in the SMIP generally appear to be re- strictedtoacertaindepthrange(e.g.,Fig.1;Heinsonetal.,2018).
Althoughithasbeensuggestedthatthe topsoftheseconductors reflectthebrittle-ductiletransitionatthetime ofmineralsystem formation (e.g.,Heinson etal., 2018), we considerit more likely thatthispatternreflectsthespatiallocationofP/Tconditionsthat weremostconduciveforgraphiteprecipitationfromthemagmatic fluid—i.e., in Fig. 3, the P/T field over which the fluid C content drops rapidlyfrom∼30 to ∼1 mol%. Ifthe redox conditionsof themagmaticfluidcanbeestimated,thenthedepthrangeofthese
conductorsmayprovideconstraintsonpaleoP/Tconditionsduring fluid flowwithimplicationsformineralexplorationefforts.Alter- natively,ifpaleoP/Tconditionscanbeindependentlyconstrained, then thedepthextentofthesefeatures mayprovideinsights into magmaticredoxconditions.
4.3. Graphitevolumefractionsandthescaleofinterconnection
Inordertoproduceabulkelectricalconductivityanomalycom- parabletoobservations,thesmall(4.16×10−6)bulk-averagevol- umefractionofgraphitethatwecalculateheremustbeintercon- nectedacrossourmodeldomain.Laboratoryexperimentsindicate that graphite volume fractions ashighas ∼10−2−10−1 are in- sufficient to enhance bulk electrical conductivitydue to poorin- terconnection ofthe conductive phase (e.g., Hashim et al., 2013;
Jödickeetal.,2007;Wangetal., 2013;ZhangandYoshino, 2017).
However, such experiments have often focused on biogenic car- bon in metasedimentary rocks and graphite in synthetic crystal aggregates; few laboratory observations have been madespecifi- callyonrockscontainingfluid-precipitatedgraphite.Wherelabora- toryexperimentsindicatelowelectricalconductivity(<10−3 S/m) for rocks that do contain hydrothermal graphite, the low val- ues are suggested to be an artifact due to disruption of the graphite electrical connection during rock exhumation (e.g.,Kat- sube andMareschal, 1993). Laboratory observations ofincreasing electrical conductivity with increasing pressure in metamorphic rockscontainingsmallamounts(<10−2 volumefraction)offluid- precipitatedgraphite(Mathezetal.,1995;Shanklandetal., 1997) suggest that small volumefractions ofthis conductive phasecan indeedcontributetoenhancedelectricalconductivityunderrealis- ticcrustalphysicochemicalconditions.
A key complication in comparing laboratory experiments to geophysical imaging results, however, is the intrinsic mismatch betweenobservation scales. Laboratory experimentsatthe hand- sample scale (∼10s of cm) struggle to capture the bulk crustal propertiesthat wouldbe expectedwhenaveragingoverthe large scales (∼10s of m to several km) atwhich geophysical imaging techniques such as MT aggregate information (e.g., Bahr, 1997).
Forexample, inthe MTimages ofthe SMIPshown inFig.1, the nominal horizontal cell size in the inversion is 2.5 km by 2.5 km (seethe Supplemental Materials).In thiscase, electrical con- ductivityinformationisinherentlybulkaveraged overthatlength scale (at a minimum; regularization, survey design, and variable electromagnetic diffusion further increase this averaging length scale).Consequently,withinsuch largeaveragingvolumes,theac- tual distributionofgraphiteisunconstrained.Fig. 5demonstrates theequivalencebetweenhomogeneouslylowandheterogeneously highgraphiteconcentrationstoyieldthesameobservedbulkelec- trical conductivityvalue withina givenaveragingvolume.Forex- ample,the samebulk-averageelectricalconductivity(∼2.7×100 S/m) wouldarise foreitherthe entireaveraging volume contain- ing a graphite volume fraction of4.16×10−6 or fora fraction of 4.16×10−6ofthetotalaveragingvolumecontainingveinsofpure (volumefraction100)graphite(theseareendmembersofthethick dotted line in Fig. 5). Near-surface electrical observations of the spatialdistributionofgraphiteinexposedmid-lowercrustalrocks inNorway(Engviketal., 2021)andSriLanka(Wickramasingheet al.,2018)confirmthat zonesofgraphiteenrichment(>10−1 vol- umefraction, withelectricalconductivity >10−1 S/m) arehighly heterogeneouslydistributed,butvariablylaterallyconnected,atthe scaleof100sofmetersto10sofkilometers.
Inourmodel,we envisiongraphitedeposition occurringalong discretepermeablefluid-flowpathways.Sectionsofthecrustalong thesepathways wouldthenbe enriched ingraphite compared to ourcalculatedoverallbulk-averagegraphitevolumefraction(likely by ordersofmagnitude),whereassectionsofthecrust outsideof
Fig. 5.Bulkmixingrelationshipbetweenthefractionofthetotalgeophysicalaver- agingvolumethatisenrichedintheconductivephase(inthiscase,graphite)and theconcentrationoftheconductivephasewithinthoseenrichmentzones.Diagonal linesofconstantbulk-averagevolumefractiondemonstratetheequivalenceofvari- ousdistributionsoftheconductivephasewithinthegeophysicalaveragingvolume.
Thickdottedanddashedlinesdenotethebulkvolumefractionscalculatedanddis- cussedinSections3.3and4.2(andalsoshowninFig.4).Thinsolidlinesarelabeled withtheequivalentelectricalconductivityvaluethatwouldbeexpectedforthere- sultingconstantbulk-averagedgraphitevolumefraction,assumingthegraphiteis wellconnectedandcanbedescribedbytheupperHashin-Shtrikman(HS)conduc- tivitybound.
these pathways would be largely devoid of graphite. Large-scale electricalinterconnectionofdiscrete,graphite-enriched permeable pathwayswouldlikelybe characterizedby theupperHS conduc- tivity bound, as we expect these pathways to be texturally well connectedatlargescalesgiventhattheyfacilitatefluidflow.Con- sequently, despite the aforementioned ambiguities in laboratory observations, we consider it likely that the small overall bulk- averagevolumefractionoffluid-precipitatedgraphitethatwecal- culateinSection3.3willtexturallybecharacterizedby theupper HS conductivity bound at the large scales at which geophysical observationsaggregate information.Wethen expect theresulting bulk-averageconductivityvaluestomatchMTobservations.
4.4.Potentialanalogsinthesurfacerockrecord
Along the permeable fluid pathways in our model, we envi- siongraphite precipitatingeitheras grain-boundaryfilms or ina crustal-scale network of graphite veins. Grain-boundary graphite films havelong been recognized inmid-lower crustal rocks (e.g., Frostetal.,1989;Mareschaletal.,1992);however,laboratoryex- periments are inconsistent in demonstrating the effects of such filmsinenhancingbulkelectricalconductivity.Olderlaboratoryex- perimentshaveprovidedevidencethatgraphite filmsdoenhance bulk conductivitytosome extent whenheld atmid-crustal pres- suresthat wouldkeep graphitenetworksinterconnected(Mathez etal.,1995;Shanklandetal., 1997),andexperimentswithvapor- depositedgraphitefilmssimilarlysupportanenhancementofcon- ductivityatlow volume fractions(Roberts etal., 1999). However, newerlaboratoryexperimentsindicatethatthin(<1 μm)graphite films are unstable at high temperatures (>700◦C; Yoshino and Noritake, 2011) and over long time scales (>10s of ka). Given these competing laboratory results, we conclude that cooling
(e.g., Touret et al., 2019) may provide an analog for what lies beneath the Olympic Dam region and the SMIP. The Sri Lankan graphite is inferred tohave formed frommantle-derived carbon- richfluids, similartowhatweenvisionherewithin thesemineral systems.TheSriLankandepositscontainlargevolumesofgraphitic carbon;basedonlocalgeologicmapping(Erdosh,1970)andnear- surfaceelectricalobservations(Wickramasingheetal.,2018),volu- metricratiosofgraphitetocountryrockare∼10−4inthevicinity ofveindeposits,roughlytwoordersofmagnitudegreaterthanthe volumefractionwecalculateinourmodel.However,graphiteveins are irregularlydistributed along the∼500 km corridorin which they arefoundinSriLanka (e.g.,Touret etal.,2019).Whenaver- agedoverlargerareas,thebulkvolumefractionofgraphitemaybe comparable tothatpredictedbyourmodel,particularlygiventhe conservativeassumptions underlyingourcalculations. Giventhese considerations,weviewtheSriLankangraphiteveinsasapromis- ing analog for the occurrence of graphite beneath Olympic Dam andtheSMIP.
Although our calculations in Section 3 apply most directly to high conductivity anomalies in the mid-lower crust, observed moderately conductive “fingers” in the upper crust (e.g., Fig. 1;
Heinsonetal.,2018)canperhapsalsobeexplainedbyhydrother- malgraphiteprecipitation.TheBorrowdalegraphitedepositinthe U.K. (e.g., Luque et al., 2014) may be a reasonable analog for these conductive features. At Borrowdale, upper-crustal magmas assimilatedlarge volumesof crustalcarbon. Magmaticexsolution andconsequent overpressurization of carbon-richfluids, with re- doxconditionsatorjustabovetheFMQbuffer,causedbrecciation ofthehostrock;graphitewas thenprecipitatedfromthesefluids during pervasivewall-rock hydration reactions. Although there is no evidenceforthemagmaticremobilizationofcrustal carbonin eithertheOlympicDamregionortheSMIP,evolvedupper-crustal magmatic fluidsin these regions could havestill carriedenough mantle-derivedcarbonattheappropriateredoxconditionstopre- cipitate graphite via hydration reactions at shallow depths (i.e., duringpropyliticalterationofthecountryrockimmediatelybelow the IOA-IOCGdeposits) andto thereby yield a moderateconduc- tivity (∼10−2 S/m)anomaly at such depths.These upper-crustal
“fingers” may be further enhanced or alternatively explained by sulfide vein mineralization orhydrothermalclays associatedwith wall-rockalteration(e.g.,Delayreetal.,2020).
4.5. Applicationtoothermineralsystems
The viability of using electrically conductive hydrothermal graphite to identify the roots of magmatic-hydrothermal mineral systemsmorebroadlywillbe dictatedbytheredox requirements for graphite precipitation. Although graphite will readily precipi- tate froma fluidwithredox conditionsnear(withinonelogunit of) orbelowtheFMQ buffer,graphite isnot expectedto precipi- tatefromafluidatmorethantwologunitsabovetheFMQbuffer (e.g., Huizenga, 2011). Consequently, hydrothermal graphite pre-
2018).Thatdepositlikelyformed,atleastpartially,inan orogenic setting (e.g., Barton, 2014), so underthrust conductive metasedi- mentaryunits (e.g.,Boerner etal., 1996; Jones et al., 1997) may provideanalternativeexplanationfortheobservedanomaly.
Although we have specifically argued for mantle-derived car- bon inour model, it is nevertheless possible that remobilization andreprecipitationofcrustalcarbonthrough mechanismssimilar tothosethatwehavearguedforhere(e.g.,Luqueetal.,2014)may alsoproducea regional-scaleconductivityanomaly.Magmaticas- similationofcarbonaceous materialandsubsequentexsolutionof carbon-richfluidsmayultimatelydrivehydrothermalgraphitepre- cipitationthatcouldyieldaconductivityanomalywithinamineral system.Additionally, becausecrustal carboncanact asa reducing agentduringmagmaticassimilation,hydrothermalgraphiteprecip- itation fromfluidsassociated withsuch hybridizedmagmas may beausefulgeophysicaltracerforstudyingsystemsinwhichmag- maticredoxprocessesarekeyformineralsystemevolution(e.g.,in certainreducedgoldsystems,suchasgoldporphyriesandpossibly Carlin-type/-likesystems;Johnsonetal.,2020).
5. Conclusions
We have quantitatively demonstrated that graphite precipita- tionfromCO2-rich magmatically derived fluidsisthebest expla- nation forthe crustal-scale electrical conductivity anomalies ob- served beneath the Olympic Dam region and the SMIP. Further- more, hydrothermal graphite is a natural explanation for these anomalies,asgraphitewillreadilyprecipitatefromcarbon-bearing fluidsthroughoutthecrustal columneitherviacooling orviahy- dration reactions, under reasonable fluid redox constraints. The physicochemical principles upon which our model is based can consequentlybereadilyextendedtoothermagmatic-hydrothermal mineralsystems,providedthatthereisasourceofcarbon(either fromthe mantle orfromthe crust) andthat themagmatic fluid redox conditions are appropriate for graphite precipitation. Hy- drothermal graphitemay then be a usefulelectrically conductive indicatorofancientcrustal-scalefluidflowandavaluablecrustal- scalevectorinexploringformagmatic-hydrothermaloredeposits.
CRediTauthorshipcontributionstatement
BenjaminS.Murphy:Conceptualization,Investigation,Method- ology, Writing – original draft. Jan MartenHuizenga: Investiga- tion,Methodology,Validation,Writing –review& editing.PaulA.
Bedrosian:Investigation,Writing–review&editing.
Declarationofcompetinginterest
Theauthorsdeclarethattheyhavenoknowncompetingfinan- cialinterestsorpersonalrelationshipsthatcouldhaveappearedto influencetheworkreportedinthispaper.