ARTICLE

Sulfides, native metals, and associated trace mineralsof the Skaergaard intrusion, Greenland: evidence for latehydrothermal fluids

Ben Wernette1& Peishu Li1 & Alan Boudreau1

Received: 11 November 2018 /Accepted: 9 September 2019# Springer-Verlag GmbH Germany, part of Springer Nature 2019

AbstractSulfide assemblages, precious metals, transition metal alloys, and associated accessory phases were characterized throughout theSkaergaard intrusion to better constrain the sulfide saturation history of the intrusion and the role of late magmatic volatiles inmodifying the Skaergaard metal budget and distribution. Sulfides in and below the Middle Zone of the Layered Series of theintrusion are readily replaced by low-Ti magnetite. The ratioΣ(low-Ti magnetite mode)/Σ(sulfide mode), indicating oxidation ofsulfides, reaches maximum values in the Lower Zone of the Layered Series. Sulfide assemblages below the Middle Zone aretypically accompanied by minor biotite, apatite, and rare calcite as well as trace compositionally distinctive clinopyroxene andorthopyroxene. The occurrence of Ag, Au, Pt, Cu, and metal alloys outside of the Middle Zone is further evidence of theSkaergaard intrusion parental magma being S-poor. Native Ag, commonly accompanied by trace amounts of Cl, occurs bothin and below the Middle Zone. Evidence of coexisting precious metal + brine assemblages exists where native metals areaccompanied by sylvite ± halite and Ag is accompanied by Ag halides. Ag occurrences in the Middle Zone are of irregularmorphology with trace Cl ± S ± calcite. Further evidence supportive of a metal + brine assemblage is observed where Ag + quartzis found in an apparent open clinopyroxene-hosted fluid inclusion consisting of Na, Si, Cl, Ca, K, and S. Ag is used to model thebehavior of precious and transition metals in the presence of an exsolving fluid phase. Numerical modeling suggests that, in asulfide-bearing system, residual Ag concentrations and concentrations in the exsolved fluid are most affected at the point wheresulfide is lost to a separating volatile fluid phase. It is suggested that, owing to the low-S nature of the Skaergaard system,fractional crystallization and early fluid saturation produced enrichment of Ag, with other precious and transition metals, in theinterstitial silicate liquid much higher than normal due to delayed sulfide saturation. As this interstitial liquid evolved, Agwas lostto an exsolved volatile phase of high salinity and migrated upward. A similar process likely occurred for Au and other elementswith high affinities for Cl.

Introduction

In layered mafic intrusions (LMIs), evidence for the existenceof volatile phases is preserved in fluid inclusions, apatite com-

position, and halogen geochemistry (Boudreau andMcCallum 1989; Hanley et al. 2008). Of debate, however, iswhether these fluids play a significant role in the processesresponsible for generating, or modifying, the strataboundplatinum-group element (PGE)-sulfide ore bodies for whichLMIs are known. Conventionally, these deposits are thoughtto form as the result of gravitationally driven downwardmovement of immiscible sulfide liquid droplets in the magmachamber. Large sulfide liquid/silicate liquid partition coeffi-cients (Mungall and Brenan 2014) and large silicate liquid/sulfide liquid mass ratios (R-factor, Campbell and Naldrett1979) allow for the extreme enrichment of transition and no-ble metals, as well as PGE, in immiscible sulfide droplets.Experimental results confirm that these are reasonable as-sumptions (Fleet et al. 1993).

Editorial handling: W. D. Maier

Electronic supplementary material The online version of this article(https://doi.org/10.1007/s00126-019-00924-1) contains supplementarymaterial, which is available to authorized users.

* Ben Wernettebenjamin.wernette@duke.edu

1 Division of Earth and Ocean Sciences, Duke University,Durham, NC 27705, USA

Mineralium Depositahttps://doi.org/10.1007/s00126-019-00924-1

Conversely, others have suggested that stratabound mineral-ization is the result of upward-moving exsolved fluid, rich inhalides, allowing for the efficient transport of transition andprecious metals vertically through an igneous body (Boudreauand McCallum 1992). Theoretical work (Shinohara 1994) sup-ports this model and experimental work performed on rhyoliticand granitic systems (Candela and Holland 1984; Simon et al.2008a, b; Frank et al. 2011) as well as theoretical studies ofLMIs (Boudreau and McCallum 1992; Meurer et al. 1999)suggest the importance of an exsolved halide-bearing volatilephase. Together, these endmember models are referred to as the“orthomagmatic” and “hydromagmatic” models, respectively.

Complicating our understanding of LMIs is that the largerintrusions (Bushveld and Stillwater) show evidence of multi-ple magma injections and prolonged cooling histories forwhich the preservation of original igneous textures and chem-istry is uncertain. It is generally agreed that the smallPaleogene Skaergaard intrusion of southeast Greenlandcooled and crystallized as a closed system (Holness et al.2007). Of interest is the fact that the Skaergaard intrusioncontains a zone enriched in Au and PGEs known as thePlatinova Reef (Andersen et al. 1998; Holwell and Keays2014; Godel et al. 2014; Nielsen et al. 2015). It is for thesereasons that the Skaergaard intrusion provides a unique op-portunity to investigate how metalliferous LMIs evolvethrough time. In this study, we examine sulfide assemblagesand their associated phases and characterize the occurrence ofprecious and transition metals outside of the Platinova Reef tobetter understand the role of exsolved volatiles in thedistribution and modification of metals in LMIs and thesulfide saturation history of the Skaergaard intrusion.

Geologic background

Skaergaard intrusion

For extensive reviews, readers are directed to the early studiesof Wager and Deer (1939) and the more recent summary ofNielsen (2004). Briefly, the Skaergaard Complex is a ~ 55Ma(Brooks andGleadow 1977; Hirschmann et al. 1997) intrusionlocated in southeast Greenland (Fig. 1). The intrusion formedduring the Paleogene rifting of Greenland from Eurasia(Nielsen 1975). The Complex is hosted uncomformably inArchean gneiss and, from bottom to top, is composed of theLayered and Upper Border Series. At its margin, the Complexis composed of the Marginal Border Series. The LayeredSeries is thought to have grown upward from the floor of theintrusion while the Upper Border and Marginal Border Seriescrystallized from the roof and the walls of the intrusion in-ward, respectively. The Layered Series is subdivided into theLower (LZ),Middle (MZ), and Upper Zone (UZ) according tothe presence of index minerals.

For the MZ, Wager et al. (1957) describe a Cu-rich sulfideassemblage consisting of chalcopyrite and bornite. The upper~ 100 m of the MZ is host to the Triple Group, a sequence oflayered gabbros (Nielsen et al. 2015) that contain zones ofanomalously high Au and PGE concentrations (Andersenet al. 1998). Individual layers within the Triple Group havebeen correlated across the intrusion (Andersen et al. 1998;Holwell and Keays 2014; Nielsen et al. 2015) and discretePd-, Au-, and Cu-rich subzones, or “offsets,” have been iden-tified (Bird et al. 1991; Andersen et al. 1998; Nielsen et al.2015; Holwell and Keays 2014). Collectively, these metallif-erous subzones are known as the Platinova Reef. Several pet-rogenetic models for the Platinova Reef have been proposedand broadly align with the orthomagmatic or hydromagmaticendmember models. Previous workers have suggested that thePlatinova Reef is the result of Rayleigh fractionation process-es acting to concentrate PGEs in sulfide (Holwell and Keays2014). In this model, large Dsulfide/silicate values act to concen-trate PGE in early formed layers (Pendergast 2000; Holwelland Keays 2014). To explain high PGE tenors in sulfide andthe observed Pd, Au, Cu offsets of the Platinova Reef, Holwelland Keays (2014) proposed a multistage sulfide saturationmodel whereby early formed sulfide droplets concentratechalcophile elements as they settle through the magma cham-ber. In their model, re-dissolution of this early sulfide as itmoves into hotter magma along the floor of the chamber actsto concentrate chalcophile elements in the magma. This isfollowed by a second local (or in situ) sulfide saturation eventthat generates low-volume high-tenor sulfides. The observedPd, Au, and Cu offsets are thought by Holwell and Keays(2014) to reflect variations in Dsulfide/silicate. Indeed, Godelet al. (2014) found textural evidence supportive of in situsulfide nucleation in the Platinova Reef. Andersen et al.(1998) suggested that compaction of cumulates forced sul-fides to migrate vertically generating the observed metal off-sets while potentially oxidizing existing sulfides. Analogousto the model of Andersen et al. (1998), Nielsen et al. (2015)proposed that the precious metal distribution observed in thePlatinova Reef can be explained by the upward migration ofFe-rich melts rich in volatiles and precious metals. Andersen(2006) proposed that late hydrothermal fluids transported pre-cious metals vertically along grain boundaries and fluid path-ways. In this model, redox barriers limit the PGE carryingcapacity of the hydrothermal fluid acting to separate the strat-igraphic occurrence of Pd from Au.

Anomalously high Cu/S ratios observed throughout theSkaergaard stratigraphy (Cu/S = 1–7, Keays and Tegner2016) have been attributed to early shallow-level degassingand S loss (Li and Boudreau 2017), late S loss to hydrothermalfluids (Andersen 2006), magmatic oxidation (Wohlgemuth-Ueberwasser et al. 2013), or anomalous Cu/S ratios in theSkaergaard parental magma source region (Keays andTegner 2016). Andersen (2006) concluded that sulfide texture

Miner Deposita

and mineralogy in the LZ is well explained by an exsolvedhydrothermal fluid oxidizing and replacing preexisting sul-fides with low-Ti magnetite and redistributing Cu, S, andprecious metals to different stratigraphic levels. Li andBoudreau (2017) arrived at a similar conclusion afterconducting a modal analysis of sulfides found in theLower Zone and Marginal Border Series. Conversely,Godel et al. (2014) note that LZ sulfides are generally notaccompanied by low-Ti magnetite, ultimately suggestingthat late hydrothermal fluids did not oxidize magmatic sul-fide assemblages.

There are a number of detailed studies of precious metal dis-tribution through the Platinova Reef (Bird et al. 1991; Godelet al. 2014; Nielsen et al. 2015; Holwell et al. 2015; Keays andTegner 2016). These studies focus on characterizing the volu-metrically significant Cu, Au, Pd, and Pt occurrences within theReef. In the Platinova Reef, researchers have observed the oc-currence of Ag in native form (Nielsen et al. 2015), electrum(Holwell et al. 2015), and Au-Cu-Ag alloys (Bird et al. 1991and Cabri et al. 2005), as well as Au-PGE alloys (Andersenet al. 1998). Additionally, Holwell et al. (2015) note Ag concen-trations of ~ 250 ppm in sulfides from the Triple Group. Reef

.

0 km0 km

Greenland

MBS

Glacial Ice

Water

Gneiss

Marginal Border Series

Lower Zone

Middle Zone

Upper Zone

Upper Border Series

Bassistopen Sheet

Basalt

Sandwich Horizon

Triple Group

0 km

N

Skaergaard Peninsula

Mellemo Is.

Ivnarmiut Is.

UttentalPlateau

Pukugagryggeen 1111’

12 13

22

41

4650

32

05

09

33’

Intrusion boundary

Sample Location

6808

’ N

31 45’ W

6813

’ N

31 35’ W

Basalt

UBSBS

BSUZbUZc

UZa

MZ

LZcUZa

LZb

LZa

MZ

UZb

Gneiss

Kraemer Is.

Uttental S

ound

2 km

Fig. 1 Simplified geologic map of the Skaergaard intrusion with the location of samples used in this study. Geologic map modified after McBirney(1989)

Miner Deposita

occurrences of Au include native Au (Bird et al. 1991), tetra-auricupride (AuCu, Cabri et al. 2005; Nielsen et al. 2015), andAu-Cu-Pd alloys (Nielsen et al. 2015). Pd and Pt form the pre-cious metal minerals (PMMs) skaergaardite and nielsenite whileCu is a constituent of PMMs or metal alloys. Away from thePlainova Reef, Andersen et al. (2017) note the occurrence of Agin Au-Cu-Ag alloys found in marginal basement schists thatformed as the Skaergaard magma interacted with its Archeanhost rock. In addition, they also found electrum (Au-Ag) andvarious Bi-bearing PMM.

Sampling and methods

A. Boudreau collected samples in 1990. Three samples,SK90-5, SK90-9, and SK90-13, were published in a previousstudy (Li and Boudreau 2017) and were re-examined andincluded in this study for completeness. In total, 11 polishedthin sections were systematically scanned using the CamecaCAMEBAX microprobe at Duke University. Sulfide assem-blages and accessory phases were characterized using energydispersive spectrometry (EDS, accelerating voltage = 20 KeV,beam current = 15 nA) and X-ray composite maps. The samemethods were used to identify and characterize the occur-rences of Ag, Au, Pt, Cu, and metal alloys. Sulfides, nativemetals, metal alloys, and accessory minerals were imaged andimported into the photo processing software IMAGE-J(Schneider et al. 2012) for area and perimeter measurements.Where reported, semi-quantitative plagioclase data were col-lected using EDS. Uncertainties are An# ± 4.

Results

Summary of stratigraphic trends

As summarized in Fig. 2 and discussed in detail below, strat-igraphic trends show a general decrease in chalcopyrite (and acorresponding increase in bornite) from LZ through MZ be-fore chalcopyrite and digenite join the sulfide assemblage inthe UZ. Similarly, the relative modal abundance of low-Timagnetite as a fraction of the sulfide assemblage is highestin LZ samples (Fig. 2). The abundance of late hydrous min-erals (biotite and amphibole) are typically higher in and belowthe MZ and broadly parallels the bulk rock loss on ignition(LOI) stratigraphic trends reported by Keays and Tegner(2016). The sulfide-magnetite assemblages occur at both grainboundaries and within primocryst silicate minerals. Further,the relative proportion of accessory minerals, such as calcite,biotite, apatite, and compositionally distinctive orthopyroxeneand clinopyroxene, as a function of a given sulfide assemblageis greatest in samples from the LZ.

Occurrences of native Ag, Au, Pt, and Cu are generallyrestricted to MZ and LZ samples and native Ag is more com-mon than Pt or Au in terms to total grains observed (12 grainsof Ag vs 3 for Au and Pt, respectively). Similarly, halite andsylvite were found only in those samples from and below theMZ. Native Ag is commonly associated with trace quantitiesof Cl. Although uncertain owing to the low total number ofgrains observed, there is a suggestion that the size and totalnumber of individual Ag grains per thin section broadly in-creases up stratigraphy. A brief description of the individualsulfide (n = 165) and metal and metal-alloy (n = 18) occur-rences are presented in Tables 1 and 2, respectively.Comprehensive descriptions of individual sulfide assem-blages can be found in the Electronic SupplementaryMaterial (ESM Table S1). Sulfides occurring in late crosscut-ting albite + titanite veins are not included in this dataset.

General description of sulfide assemblagesand associated trace minerals

LZa Sulfides, mainly chalcopyrite and bornite, are typicallyfound interstitial to olivine and plagioclase. Pentlandite is ob-served in limited quantities and generally comprises a smallproportion of individual assemblages (average of ~ 4 vol.%).Sulfides are commonly non-uniform in morphology and arecommonly replaced, in part, by low-Ti magnetite (Fig. 3a, b).Native Cu is observed where sulfides are extensively replacedby low-Ti magnetite (Fig. 3a). Hydrous minerals, such as bi-otite and amphibole, are observed truncating sulfides or man-tling sulfide assemblages (Fig. 3b, c). Minor calcite is foundmantling sulfides with no evidence of other low- or high-temperature silicate alteration. When observed, calcite isrounded and commonly no greater than 20 μm in size.Anhedral clino- and orthopyroxene are also found locallymantling sulfides. When adjacent, plagioclase is increasinglycalcic near sulfide assemblages (An# ~ 66, Fig. 3a) but lesscalcic away from sulfide assemblages (An# ~ 50).

LZb Chalcopyrite and bornite make up the entire sulfide as-semblage with near equal proportions of low-Ti magnetitereplacing the sulfides (Li and Boudreau 2017).

LZc Observed sulfide proportions by area average 9% chalco-pyrite and 91% bornite. Sulfides occur both within silicates(cores and margins) and at silicate grain boundaries. Low-Timagnetite is less abundant relative to LZa. Biotite truncatessilicate grain boundaries or crosscuts sulfides. Calcite wasobserved in a single sulfide assemblage. As for LZa, plagio-clase becomes increasingly calcic towards sulfide assem-blages where An# is similar to that observed in LZa.

MZ The sulfide assemblage below the Platinova Reef consistsentirely of bornite with variable amounts of low-Ti magnetite.

Miner Deposita

It is important to note that this differs from the MZ born-ite and chalcopyrite assemblage reported by Wager et al.(1957). Sulfides are interstitial to, or found entrainedwithin, silicates. Sulfide morphology is typically irregularand is positively correlated with the abundance of second-ary silicates or low-Ti magnetite. Biotite and amphiboleare observed in limited quantities mantling sulfides, andcalcite is only associated with native Ag or Ag halides. Atpyroxene grain boundaries, olivine is locally observedenclosing sulfide assemblages and, in some cases, low-Ti magnetite (Fig. 3d). When present, an increase in pla-gioclase anorthite content in adjacent plagioclase isobserved.

UZa The sulfide assemblage is composed entirely of bornite(82%), chalcopyrite (13%), and minor digenite (5%). Low-Timagnetite is observed replacing sulfides in quantities similar tothat observed in LZc. Apatite is regularly associated with sul-fides, typically surrounding the periphery of the mineral.Sulfides are notably non-uniform in composition with Cu-rich,S-poor areas (Fig. 3e). Anhydrite is observed in limited abun-dance where sulfides are in direct contact with plagioclase.Additionally, plagioclase becomes increasingly calcic towardssulfide assemblages. Near sulfide An# is as high as 72 (Fig. 3f).

UZb Bornite and minor digenite comprise the entire sulfideassemblage. Sulfides are commonly rounded and found at

HZ

LZa

LZb

LZc

MZ

UZa

UZb

UZc

UZc’

UZb’

UZa’

MZ’

LZc’

LZb’

LZa’

UZ

b*

UZ

a*

MZ

*

LZ

c*

LZ

b*

LZ

a*

***************

0

1.0

2.0

3.0

PR

Str

atigraphic

Heig

ht (km

)

Digenite

Bornite

Pentlandite

Chalcopyrite

Orthopyroxene

Clinopyroxene

Calcite

Biotite

LZa

LZc

Apatite

UZa

UZb

MZ

Low-Ti Magnetite

a

Other

pla

gio

cla

se

olivin

e

Ca-ric

h p

yroxene

pig

eonite

ilm

enite

magnetite

apatite

ferrobusta

mite

quartz

K-fe

ldspar

trid

ym

ite

cb

LZb

Legend

)83 = n(

)31 = n(

)7 = n(

)45 = n(

)63 = n(

)92 = n(

Sample

Location

Primocryst

Irregular

Primocryst

Postcumulus

Fig. 2 a Stratigraphy of the Skaergaard intrusion with the location ofsamples used in this study denoted by white stars. b Sulfide assemblageproportion oxides and sulfides (left column) and other accessory minerals(right column). Stratigraphic zone corresponds to the stratigraphic zone in

a and n is the number of individual sulfide assemblages studied for eachstratigraphic zone. c From Naslund (1984), the stratigraphic extent ofprimocryst minerals, irregular primocryst minerals, and latepostcumulus minerals

Table 1 Summary of sulfide assemblages reported in this study

Sulfide assemblage total area (μm2) Associated trace minerals (μm2)

Zone No. sulfideoccurrences

Chalcopyrite Pentlandite Bornite Digenite Ti-Poor Magnetite Biotite Apatite Calcite Cpx Opx

LZa 38 31,727 1594 6930 263 7467 6674 15 869 3489 –

LZb1 13 751 – 608 – 455 – – – – –

LZc 7 549 – 4768 – 233 4287 – 84 959 361

MZ 54 – – 537,079 – 100,690 5776 902 – 4257 –

UZa 36 59,997 – 374,590 28,443 11,710 2612 9259 – – –

UZb 29 – – 196,081 44,156 9432 1959 22,154 – – –

1 Data from Li and Boudreau 2017

Miner Deposita

silicate-silicate or silicate-oxide grain boundaries and lesscommonly within high-Ti magnetite. Additionally, sulfidesare found interstitial to modally abundant apatite. Low-Timagnetite is seen in minor abundance and biotite is observedin only one assemblage.

Occurrence of precious and transition metals

Below the MZ, native Ag occurs as discrete rounded grainsnear the cores of major minerals such as plagioclase, intersti-tial to late phases such as clinopyroxene and apatite (Fig. 4a),or at primocryst grain boundaries (Fig. 4b, c). In MZ samples,Ag occurs both as well-rounded grains and irregular shapedgrains in silicate minerals (Fig. 4d) or at silicate grain bound-aries. Those occurrences of irregular morphology (6 out of 12)are accompanied by trace amounts of Cl ± S ± calcite

(Fig. 4d, e) and Ag halides (Fig. 4d). Less common are Aggrains occurring with Au intergrowths and trace Cu with no-table trace quantities of Cl (Fig. 4e). Ag was not observedabove UZa. One occurrence of note (MZ), which is interpretedas a ruptured fluid inclusion, is native Ag found inclinopyroxene with quartz and trace quantities of Na, Si, Cl,Ca, K, and S (Fig. 5).

Grains of native Au or Pt are much less common thannative Ag. Native Au occurs as intergrowths with Ag in theMZ (Fig. 4e) and as singular occurrences with halite andsylvite in plagioclase in LZa (Fig. 6). In LZa, native Pt wasfound within high-Ti magnetite. Further, Pt-sulfides andPt-Au alloys are observed at primocryst grain boundarieswith trace Fe-Mg amphibole and apatite in MZ samples(Fig. 7a). Of note is one occurrence of Pt in plagioclasein UZb (Fig. 7b).

Table 2 Area and perimetermeasurements (microns) foroccurrences of native metalsdiscussed in this study

ID StratigraphicZone

Occurrence Area Perim Notes

SK90-5-B2 LZa Ag 21 17 In plagioclase with trace Cl-bearingphase (too small to determine)

SK90-5-H2 LZa Ag 392 148 In olivine with trace Na, K, and Cl

SK90-5-K2 LZa Ag 535 275 Olivine-plagioclase grain boundarywith trace Cl and associatedAg-halides (Fig. 4c)

SK90-5-O LZa Ag 15 15 Interstitial to cpx, apatite, andplagioclase (Fig. 4a)

SK90-5-L LZa Ag 57 9 In plagioclase near margin. No Clsignature observed (Fig. 4b)

SK90-5-S LZa Pt 133 70 In Ti-rich magnetite

SK90-5-Au LZa Au 64 40 In plagioclase with coexistinghalite and sylvite (Fig. 6)

SK90-13-Q LZc Ag 364 95 At pyroxene-pyroxene grainboundary with trace Cl and S

SK90-11-1 MZ Ag-Auintergrowth

679 239 Ag - Au intergrowth with Cu andcalcite (?) in plagioclase (Fig. 4e)

SK90-11-1b-J MZ Ag 493 177 In plagioclase with trace Cl andassociated Cu, Au

SK90-50-B-1 MZ Ag 371 207 In plagioclase with trace calcite,S, Cl (Fig. 4d)

SK90-50-B-2 MZ Ag 18 38 Mantling barite

SK90-50-0 MZ Ag-Zn alloy 4 8 Ag-Zn alloy and associated nativeCu in cpx with trace Cl (Fig. 4f)

SK90-50-E MZ Ag 35 7 In apparent fluid inclusion in cpx(Fig. 5)

SK90-50-0-1 MZ Cu 31 15 In cpx with notable Cl signature andassociated Ag-Zn alloy (Fig. 4f)

SK90-41b-BB MZ Cu 20 24 Cu in high-Ti magnetite withassociated trace silver

SK90-41b-GG MZ Pt-Au alloy 174 63 At plagioclase-cpx grain boundarywith trace apatite and Fe-Mgamphibole (Fig. 7a)

SK90-32-AA UZb Pt 92 50 In plagioclase near cpx and sulfideassemblage (Fig. 7b)

Miner Deposita

Native Cu is seen where significant replacement, or loss, ofS from sulfide minerals appears to have occurred (LZa,Fig. 3a). In some samples, however, Cu “colloids” are ob-served at sulfide grain boundaries without evidence of sulfurloss or silicate replacement in LZ and UZ samples (Fig. 3e). InMZ samples, Cu is observed in high-Ti magnetite and pyrox-ene with associated trace Ag. Finally, trace quantities of Cuare observed in pyroxenes from LZb and LZc and amphibolesfrom the MZ.

Approximately 70% of the occurrences of precious andtransition metals outside of the mineralized Platinova Reefare accompanied by trace Cl or are spatially associatedwith hydrous minerals such as biotite and amphibole.Furthermore, about 40% of the occurrences are either at,or close to, grain boundaries while the remaining 60% oc-cur in primocryst minerals. Finally, except for one instanceof Ag near sulfide (Fig. 3d), precious metals are not

observed near sulfide minerals, perhaps reflecting differentprecipitation processes.

Fluid inclusions

Fluid inclusions were not observed below the MZ; however,the former presence of a high-temperature fluid is inferredfrom the presence of sylvite inclusions in clinopyroxene andplagioclase primocrysts from LZa and LZc and halite crystalsat silicate grain boundaries in LZa. Within the MZ, fluid in-clusions were observed as singular occurrences in pyroxene oras inclusion trails in olivine (Fig. 8).

The fluid inclusions in pyroxene, inadvertently opened dur-ing thin section preparation, are no greater than 50 μm indiameter. They contain Ag, Ag-Zn alloys, Cu, and trace quan-tities of Cl, Na, Ca, and K. Fluid inclusions in olivine can becategorized into three principle groups that form inclusion

mtCu

ccpcpx

opxplag plag

ccpccp

bn

pnmt

bn

ccp ccp + bn

biotite

Cu

opx

high-Ti mt

cpx ol

olmt

Ag

bnCu-Fe-S oxide

bn

ccp

ol

Cu

a b

d e

40 µm 50 µm

50 µm 40 µm

high-Ti mt

ol

ol

cpx

opxplag mt

bn

bn

ol

biotite

plag

mt

opx

40 µm

40 µm

f

c

An# = 66

An# = 50

An# = 72

An# = 33

An# = 47

An# = 61

Mg S K Ca Fe AgCu

Fig. 3 a Sample SK90-5 (LZa) Sulfide assemblage readily altered tomagnetite. Note Cu occurring where much of the sulfide has been lostand that plagioclase immediately surrounding the assemblage is morecalcic (brighter orange, white dashed line) b Sample SK90-5 (LZa)Polysulfide assemblage cross-cut by magnetite and biotite. Note thatplagioclase immediately surrounding the assemblage is more calcic(brighter orange, white dashed line). c Sample SK90-5 (LZa) Sulfideassemblage truncated by biotite, orthopyroxene, and low-Ti magnetite.Note that the area immediately surrounding the sulfide is more calcic

plagioclase (brighter orange, white dashed line). d Sample SK90-50(MZ) Interstitial sulfide from the Middle Zone that is entirely oxidized.Note the minor Ag occurrence. e Sample SK90-32A (UZa) Roundedsulfide with areas of high and low-S concentrations with no sign of hostsilicate alteration. f SK90-32A (UZa) Interstitial bornite replaced by low-Ti magnetite. Note near sulfide high-Ca plagioclase (brighter orange,white dashed line). All figures are EDS false color composition maps.Elements and corresponding colors are given in d

Miner Deposita

trails, consisting of elongate or oblate shaped fluid inclusions~ 40 μm in diameter. One population of inclusions parallelsolivine fracture surfaces, while the other two are not associat-ed with any identifiable fracture surface and differ only in sizeand density of the inclusions (Fig. 8). Broadly, inclusions inolivine number in the dozens and are categorized as consistingof ~ 80% vapor fill with a small proportion of fluid(Fig. 8b, d). Where observed, solids make up ~ 10% of fluidinclusion volume (Fig. 8b). All olivine in sample SK90-50(MZ) host fluid inclusions.

Discussion

Sulfides and associated minerals

As noted above, the Ti-free magnetite has been previouslyinterpreted to be replacement of the original sulfide assem-blage as the result of S loss (Andersen et al. 1998; Andersen2006; Li and Boudreau 2017); our observations are consistentwith these interpretations. Specifically, sulfides in and below

UZa are commonly replaced by low-Ti magnetite as the resultof S loss via the ƒO2-neutral reactions:

CuFeS2 þ 1:15H2O→0:27Fe3O4 þ 0:19Cu5FeS4 þ 1:15H2Sþ 0:04SO2cpy vapor mt bn −−−−−− vapor−−−−−−

ð1Þ

and

CuFeS2 þ 1:11H2Oþ 2HClþ SO2→0:33Fe3O4 þ CuCl2 þ 2:11H2Scpy −−−−−−−vapor−−−−−− mt −−−− vapor−−−− ð2Þ

(Li and Boudreau 2017). Evidence of sulfur loss by thesereactions is ubiquitous from LZa through UZa.

Additionally, the abrupt decrease in secondary, or late hy-drous minerals above the MZ, is consistent with loss on igni-tion (LOI) data reported by Keays and Tegner (2016) wherevalues approach 1 wt.% below UZa but are closer to 0.2 wt.%in, and above, UZa (Fig. 9).

Similarly, the enrichment in the An content in plagioclasesuggests silica and Na loss to a silica-undersaturated fluid via:

2:0CaNaAlSi2:5O8 þ 4:0H2Oþ 4:0Hþ→CaAl2Si2O8 þ 3:0H4SiO4 þ Ca2þ þ 2:0Naþplagioclase −−−−vapor−−−− anorthite −−−−−−−−−vapor−−−−−−−−

ð3Þ

Support for this is the observation that Ca content increasestowards sulfide assemblages in LZa through UZa (Fig. 3a, b,c, and f). Further, silica loss can produce olivine at pyroxenegrain boundaries via the reaction:

2:0 Mg0:5Fe0:5ð ÞSiO3 þ 2:0H2O→ MgFeð Þ2SiO4 þ H4SiO4

orthopyroxene vapor olivine vapor

ð4Þ

Similar observations have been made at the 2.7 GaStillwater Complex (Meurer et al. 1997) where discor-dant troctolites were explained by silica loss to reactivefluids. Further work is needed to confirm that olivineobserved in MZ samples are metasomatic in naturethough fluid inclusions in olivine (Fig. 8) are consistentwith this hypothesis.

Evidence for volatiles in the Skaergaard intrusion

The reactants involved in Eqs. 1 through 4 require a sulfide-undersaturated volatile phase to produce the observed assem-blages. Direct evidence for a fluid includes (i) the fluid inclu-sions with associated Ag and quartz that retain trace quantitiesof Ca, Na, Cl, and S and (ii) MZ clinopyroxene-hosted Ag-Znalloys with Cu and trace Cl. Moreover, the association of Agwith Cl in MZ and LZ samples suggests their concentration

mechanism is the same. Further, native Au with proximalhalite and sylvite (Fig. 6) is supportive of a coexisting metal+ brine assemblage. Finally, the occurrence of Pt-Au atclinopyroxene-plagioclase grain boundaries with associatedFe-Mg amphiboles suggests that PGE interacted with a hy-drous fluid.

More broadly, Sonnenthal (1992) described Cl-bearing flu-id inclusions in late anorthositic pegmatites of the UZ whileLarsen et al. (1992) describe brine inclusions containing up to~ 22 wt.% NaCl in gabbroic pegmatites in and below the MZ.Sonnenthal (1992) determined that Cl-bearing fluid inclusionswere trapped between 500 and 800 °C while Larsen et al.(1992) estimate a much narrower range of 655–770 °C.These saline conditions and temperatures might reflect thoseappropriate for the transport of Ag by an aqueous NaAgCl2complex, as predicted by Yin and Zajacz (2018). Further,Larsen et al. (1992) suggest that fluid inclusions present with-in interstitial phases suggest a crystal + (evolved) silicate liq-uid + exsolved fluid assemblage, consistent with the presenceof apatite, biotite, and salts. This agrees with our observationof fluid inclusions in interstitial olivine (MZ) and plagioclasewith apparent silicate and iron-rich melt inclusions (MZ) asdescribed by Jakobsen et al. (2011). Na enrichment will occurduring reaction (3) where plagioclase is converted to anorthiteby a silica-undersaturated fluid. It is noted that, in some in-stances, plagioclase near sulfide assemblages preserve An# ~

Miner Deposita

62 near oxidized sulfides while An# of ~ 52 is preserved awayfrom sulfides (Fig. 3a, b, c). UZa plagioclase preserves largervariations, with one instance of near sulfide An# ~ 72 whileAn# away from the sulfide is ~ 33 (Fig. 3f), suggesting thisprocess becomes more efficient, and total Na+ increases, upstratigraphy.

Finally, evidence for volatiles in other LMIs includesaline fluid and halide melt inclusions (Hanley et al.2008) in the Archean Stillwater Complex, apatite halo-gen geochemistry of samples from the Archean Stillwaterand Bushveld Complexes (Boudreau et al. 1986), andH2O ± NaCl ± CO2 ± CH4 inclusions in quartz reportedfor the Bushveld Complex (Ballhaus and Stumpfl 1985).Of note is the similarity in fluid inclusion compositionsfrom the Bushveld Complex described by Ballhaus andStumpfl (1985) to those from the Skaergaard Complexdescribed by Larsen et al. (1992). The solutions de-scribed by Larsen et al. (1992) and Sonnenthal (1992)are saline in nature, so the observation of halite andsylvite at silicate grain boundaries reflects the former

presence of a high-temperature fluid rich in Cl. Thesefluids are of interest in LMIs because of their ability totransport precious and transition metals through an igne-ous body.

Precious metals in an exsolved volatile phase

Magmatic volatile phases are an important control on the dis-tribution of precious metals in epithermal systems (Heinrichet al. 2004) and Cu-Au porphyry systems (Candela andHolland 1984; Heinrich et al. 2004; Simon et al. 2008a). Inaddition, volatiles have also been argued to be important formetal distribution in LMIs (Boudreau and McCallum 1992;Meurer et al. 1999; Hanley et al. 2008). Previous studies of theSkaergaard intrusion have invoked magmatic volatile phasesto explain dendritic anorthosite and pegmatite found in UZa(Sonnenthal 1992), oxidized sulfides found throughout thestratigraphy (Andersen 2006), and unusually high Cu/S ratios(Andersen 2006). Further, Godel et al. (2014) assert that in thelowermost zone (Pd5) of the Platinova Reef, Au occurs at

10µm

Ag

cpxapatite

plag low- Ca plag

AgAg halide

Cl-rich inclusioncalcite

plag

plag

Ag

Au

Cu calcite

cpxvug

Ag-Zn alloy

CuCl

cpx

a

d e f

10µm

Ag

Ag halide

cpx

ol

10µm

c

10µm 10µm

10µm

b

low- Ca plag

Agplag

Mg SAl Ca Fe AgCuCl Au

Fig. 4 a Sample SK90-5 (LZa) Ag interstitial to plagioclase, apatite, andclinopyroxene. b Sample SK90-5 (LZa) Native Ag in plagioclase neargrain boundary. c Sample SK90-5 (LZa) Native Ag with associated Aghalides at olivine-clinopyroxene grain boundary. d Sample SK90-50(MZ) Native Ag in plagioclase with calcite and Cl-rich inclusion. e

Sample SK90-11-1b (MZ) Ag-Au intergrowth in plagioclase. Noteassociated Cu and calcite. f Sample SK90-50 (MZ) Ag-Zn alloy withassociated native Cu in a chlorine rich inclusion in clinopyroxene. Allfigures are EDS false color composition maps. Elements andcorresponding colors are given in d

Miner Deposita

silicate ground boundaries and is never enclosed by oxideminerals further implying that Au mineralization is not ofprimary magmatic origin.

Several researchers (Simon et al. 2008a; Frank et al.2011; Zajacz et al. 2013; Yin and Zajacz 2018) havestudied the behavior of silver in S-poor granitic andrhyolitic systems. Zajacz et al. (2008) show that the Ag

fluid-silicate liquid partition coefficient (Dfluid=silicateAg ) in-

creases linearly with the molarity of Cl, mCl, in the

exsolved fluid phase such that Dfluid=silicateAg ≅ 4 × mCl.

Indeed, Cl has an affinity for the aqueous phase with

Dfluid=silicateCl values between 20 and 40 (Holland 1972)

or between .9 and 6 in molten basalts (Webster et al.

1999). A pertinent question is whether an exsolved Cl-bearing fluid phase can transport significant quantities ofAg as a neutral chloride complex:

AgO0:5 þ HCl→AgCl þ 0:5H2Omelt vapor ‐‐‐ vapor ‐‐‐

ð5Þ

(Simon et al. 2008b) and by inference, other Cl solublemetals such as Au, Cu, and Zn. Yin and Zajacz (2018) suggestthat at higher temperatures (~ 900 °C), the neutral NaAgCl2 isthe favored Ag species in Cl-bearing exsolved fluids via themodified H2-neutral reaction:

AgOþ NaClþ HClþ 0:5SO2→NaAgCl2 þ 0:5H2SO4

melt ‐‐‐‐‐‐‐‐‐‐vapor‐‐‐‐‐‐‐‐‐‐ ‐‐‐‐‐‐‐‐‐‐vapor‐‐‐‐‐‐‐‐‐‐ð6Þ

Agqtz

cpx

10 µm

Silver Map

a

c Ag

Ag

AgCl

1099

824

274

549

Cou

nts

X-Ray Energy (KeV)

1.79 2.49 3.07 3.71 4.35 4.99 5.63 6.27 6.91

d

503

1007

1511

2015

0.00 1.28 2.56 3.84 5.12 6.40

Cou

nts

Cl

SiNa

K

Ca

Fe

Ca

O

SCa

b

b

d

Mg

Al

Fig. 5 a Back-scattered electron image of a probable fluid inclusion inclinopyroxene containing native Ag with trace quantities of quartz. bSemi-quantitative X-Ray spectra for the field outlined in a. Note theabundance of Ca, Na, Si, Cl, and K. c EDS false color composition

map for Ag. Same field of view as a. d Semi-quantitative X-RaySpectra for the field outlined in c. Note the composition is almostentirely Ag with trace Cl

Miner Deposita

Na cations can come from reaction (3) suggesting that theconversion of plagioclase to anorthite by a silica-undersaturatedfluid is intimately related to the transport of Ag as illustrated byreaction (6).

Less well understood is the behavior of Ag in basalticsystems. Early studies investigated the concentration ofAg and other precious metals in MORBs to understandtheir potential as protoliths to greenstone Au deposits(Keays and Scott 1976). Recent experimental results have

allowed for the determination of Dsulfide=silicateAg of about

1000 (Li and Audetet 2012). Indeed, the investigation ofsulfide globules in MORB glasses leads to the determina-

tion of a similar Dsulfide=silicateAg of 1138 ± 245 (Patten et al.

2013). Workers have observed that MORB Cu/Ag valuesremain constant at ~ 3500 with decreasing Mg#, suggest-ing the two elements behave similarly throughout MORBfractionation (Jenner 2017). Additionally, similar Cu/Agvalues are observed in arc and back-arc environments pri-or to sulfide saturation (Jenner et al. 2010). Further sup-port for the associated nature of Cu and Ag is found inmantle peridotites where Cu/Ag values are nearly identi-cal to those observed in MORB (Wang and Becker 2015).Zajacz et al. (2013) determined Ag solubilities for basalticliquids to be ~ 11 and ~ 5.5 ppm for sulfide-bearing andsulfide-free compositions, respectively. Because of thelate S saturation in the S-poor Skaergaard system, it isappropriate to further investigate the partitioning of Aginto coevolving vapor in basaltic liquids. Doing so willallow for a better understanding of Ag behavior in S-poorintrusions.

Temperature and timing

Temperature constraints might allow for better under-standing of Ag (and other precious and transition metals)mobilization during the Skaergaard cooling history.Physical relations such as the interstitial nature of Ag inLZ samples and proximity to grain boundaries suggestthat Ag precipitation occurred late in the intrusion’scooling history but early enough to be locally entrainedduring overgrowth of primocrysts by crystallization ofinterstitial liquid. Further, ~ 40% of the precious metaloccurrences reported here occur at primocryst grainboundaries (see Table 2) and in some instances, theseoccurrences are in close proximity to hydrous minerals(Fig. 7a). Utilizing clinopyroxene solvus thermometry,Manning and Bird (1986) determined minimum tempera-ture of formation for late clinopyroxene rich veins to bebetween ~ 500 and 750 °C which is within the range oftemperatures reported for fluid inclusions by Sonnenthal(1992) and Larsen et al. (1992). It is suggested that this isa viable minimum temperature range for the preciousmetals reported here.

What remains unclear is why Ag is observed exclusive-ly in native form in lithologies below the MZ while Agoccurs as both alloys and native Ag and in sulfides in the

40 µm

10 µm

Pt-Au

plag

plag

Pt

cpx

bn

high-Ti mt

cpx

b

a

amph

ap

apamph

amph

Mg S Ca Fe Pt Au

Fig. 7 a Pt-Au alloy at plagioclase-clinopyroxene grain boundary withassociated Fe-Mg amphiboles and apatite from sample SK90-41b (MZ).b Native Pt in plagioclase in sample SK90-32 (UZb). Both figures areEDS false color composition maps. Elements and corresponding colorsare given in Fig. b

10 µm

halite

sylvite

halite

Au

plag

Na

K

Na KCl Au

Fig. 6 Sample SK90-5 (LZa). EDS false color composition map of nativeAu in plagioclase with associated halite and sylvite. Blue regionsrepresent areas of high K concentration while purple regions representareas of high Na concentration

Miner Deposita

MZ in general and in the Platinova Reef in particular (Birdet al. 1991; Andersen et al. 1998; Cabri et al. 2005;

Holwell et al. 2015). High Dsulfide=silicateAg of ~ 1000 (Li

and Audetat 2012; Patten et al. 2013) suggests that Agshould be retained in sulfide, if any is present. One possi-ble explanation is that the LZ magma was simply under-saturated in sulfide. Another scenario could be that LZ S

was lost to an exsolving volatile phase (e.g., Li andBoudreau 2017), leaving behind small amounts of Ag withlittle or no S. In that case, one would expect to see alloyswith Cu in approximate proportion to the metal concentra-tion of the original sulfide. That this is not generally seensuggests that metals have been effectively separated duringvolatile exsolution and transport.

Modeling silver solubility

To illustrate Ag enrichment during crystallization anddegassing, a simple partitioning model is calculated usingRayleigh fractionation (Shaw 1970):

Cliq ¼ Co*FD−1 ð7Þ

where Cliq and Co represent the concentration of Ag in thesilicate liquid and the initial concentration, respectively. Fcorresponds to the weight fraction of silicate liquid remaining;

a b

c d

b

d

vapor

solid

vapor

vapor

500 m

500 m

200 m

200 malterationFig. 8 a Plane polarized photomicrograph of fluid inclusions hosted inolivine from sample SK90-50 (MZ). b Plane polarized photomicrographof the field outlined in a. Note the crystalline solid phase present in onefluid inclusion and the ~ 90% vapor fill in the other. c Plane polarized

photomicrograph of a cluster of fluid inclusions in olivine from samplesSK90-50. d Plane polarized photomicrograph of the field outlined in c.The larger fluid inclusions are composed of a vapor + fluid. Note thepopulation of smaller elongate inclusions

LZa LZb LZc MZ UZa UZb

0.2

0.4

0.6

0.8

Lo

ss o

n I

gn

itio

n (

wt.

%)

0.0

50

.15

0.2

5

Fig. 9 Variation in loss on ignition data reported by Keays and Tegner(2016) as a function of lithologic unit. Also plotted in red is the Σ(low-Timagnetite mode)/Σ(sulfide mode)

Miner Deposita

D is the bulk partition coefficient of Ag in the system givenby:

D ¼ Dvap=liq*X vap þ ∑ Dxtal=liqi *X i

� �ð8Þ

Dvap/liq is the Ag vapor/liquid partition coefficient, Dxtal=liqi

corresponds to the Ag crystal/liquid partition coefficientfor each mineral phase i. Xi and Xvap are the weightfraction of crystal i and vapor separating from a silicateliquid, respectively. Because Ag does not readily parti-tion into common silicate minerals (Adam and Green

2006; Greaney et al. 2017), Dxtal=liqAg is equal to 0.

Ultimately, the concentration of Ag in the vapor ascrystallization proceeds is given by:

Cvapor ¼ Co*FD−1*Dvap=liq ð9Þ

Simon et al. (2008b) determine that Dvap=liqAg in rhyolitic

systems is 32 and Dbrine=liqAg is ~ 1000. For the purposes of this

calculation, a conservative Dvap=liqAg of 10 is used. Because the

initial Ag concentration of the Skaergaard silicate liquidis unknown, endmember calculations are carried out for amid-ocean ridge basalt (MORB) (CAg = 0.0718 ppm,sample: T3-72D254-16-4, Keays and Scott 1976) and aHawaiian basalt from the Kilauea Iki lava lake (CAg=0.09 ppm, sample 67-3-6.8, Greaney et al. 2017) undersulfide-free conditions. In this example, Kilauea Iki isthought to be analogous to a sulfide-undersaturateddegassing cumulate pile.

As illustrated in Fig. 10, Ag increases in both the volatileand liquid phase as the weight fraction of silicate liquid de-creases. Also plotted is a dashed line corresponding to Ag

saturation in a S-free basaltic liquid (~ 5.5 ppm, Zajacz et al.2013). For MORB liquid, extreme fractionation (F =0.002) is required before MORB liquid reaches Ag sat-

uration values of ~ 5.5 ppm. Because Dvap=liqAg is 10,

coevolving MORB vapor saturates in Ag much soonerduring the evolution of the liquid (F = 0.02). The evolu-tion of the Hawaiian basalt is identical in behavior ex-cept that saturation of Ag in the liquid and vapor occursearlier in the evolution of the liquid (F = 0.015 and 0.15,respectively). In the absence of sulfide, Ag (and othermetals with a similar affinity for Cl) will preferentiallyconcentrate in the coevolving vapor phase. This obser-vation is confirmed by a number of studies that haveinvestigated gas condensates from volcanic fumeroles(Yudovskaya et al. 2006; Yudovskaya et al. 2008;Chaplygin et al. 2015).

In the following, the partitioning of Ag into a volatile phasein both sulfide-free (case 1) and sulfide-bearing (case 2) sys-tems is examined. We follow the methods of Boudreau andMcCallum (1992) where fractional vaporization from an in-terstitial liquid is assumed like fractional melting and is de-scribed by Eq. (10);

Cr

Co¼ 1–Fvap

� � 1D−1ð Þ ð10Þ

Where Cr = the residual concentration of the element ofinterest, Co = the initial concentration of the element of inter-est, Fvap = the weight fraction of vapor exsolving from thesilicate liquid and D = bulk partition coefficient given byEq. (11);

D ¼ X liq

Dliq

� �þ X sulf *

Dsulf

D f

� �ð11Þ

Where Xliq = the weight fraction of silicate liquid re-maining in the system, Xsulf = the weight fraction of sulfidein the system, Df = the fluid/silicate liquid partition coef-ficient while Dsulf = the sulfide/silicate liquid partition co-efficient for the element of interest. Finally, the concentra-tion of the element of interest in the separated fluid at anyinterval is given by;

C f

Co¼ 1

D*Cr

Coð12Þ

where, Cf = the concentration in the exsolved fluid.

Case 1 The partitioning of Ag into an exsolved fluid in asulfide-free system (Fig. 11a, b). Here, it is assumed that75% silicate liquid has crystallized to form silicate mineralswhile 25% interstitial silicate liquid remains. The liquid is

0.01

0.1

1

10

100

1000

0.01

0.1

1

10

100

1000

0.0010.010.11

Weight fraction of melt (F) in the system

[Ag] in

silic

ate

liq

uid

(ppm

)

[Ag] i n

vo

lat ile

(p

pm

)

Liquid

Vapor

MORB

Hawaii

Hawaii

MORB

Ag saturation = 5.5 ppm

Fig. 10 Numerical calculation of Ag enrichment in endmember basalticliquids and evolving vapor as a function of fraction of silicate liquidremaining. Ag concentrations for the liquid and coexisting vapor areshown for Hawaiian basalt and MORB. 5.5 ppm Ag saturation in a S-free basaltic liquid is from Zajacz et al. (2013)

Miner Deposita

undersaturated in volatiles. The interstitial liquid is consideredto saturate in Cl-bearing vapor at 5% such that for every 1%silicate liquid crystallized, 5% of that is Cl-bearing vapor.Initial Cl concentrations of 1, 5, 10, and 20 mol are assumedcorresponding to modest values of 12.5, 62.5, 125, and

250 ppm (ρ = 2.8 g/cm3) or Dfluid=silicateAg values of 4, 20, 40,

and 80, respectively (Zajacz et al. 2008). Cl is perfectly in-compatible in the solids prior to vapor saturation so that con-centrations increase exponentially as crystallization takesplace. As shown in Fig. 11a and b, residual concentrationsare most affected by larger Ag fluid/silicate liquid partitioncoefficients where residual Ag concentrations are half theoriginal value after ~ 20% silicate liquid has crystallizedand the bulk weight fraction H2O has evolved to ~ 0.012.Similarly, Ag in the exsolved fluid is highest in the early

stages of fluid evolution before values diminish due todilution (Fig. 11b).

Case 2 Figure 11c and d show a similar calculation for asulfide-bearing system composed of 74.9% crystals,25% silicate liquid, and .1% sulfide liquid. Cl molaritiesremain unchanged from case 1. It is assumed that theliquid is sulfide saturated at 400 ppm S. Each fractionof water evolved contains 5% S which is subtractedfrom the starting S content before each subsequent cal-culation. The Ag sulfide/silicate liquid partition coeffi-cient is taken to be 1000 (Jenner 2017) while Ag fluid/silicate liquid partition coefficient values are as in theS-free calculation. Calculations suggest that minimalchange is observed in residual concentration prior to

0

0.2

0.4

0.6

0.8

1

1.2

0 0.002 0.004 0.006 0.008 0.01 0.012

Cr/C

o

Fraction Water Evolved

0.1

1

10

100

1000

0 0.002 0.004 0.006 0.008 0.01 0.012

Cf/C

o

Fraction Water Evolved

0

0.2

0.4

0.6

0.8

1

1.2

0 0.002 0.004 0.006 0.008 0.01 0.012

Cr/C

o

Fraction Water Evolved

0.1

1

10

100

1000

0 0.002 0.004 0.006 0.008 0.01 0.012

Cf/C

o

Fraction Water Evolved

Dfluid

= Dsulfide

su

lfid

e o

ut

m Cl = 1

m Cl = 5

m Cl = 10

m Cl = 20

a

c

b

d

Fig. 11 Numerical calculations simulating the dissolution of Ag in sulfide-free (a, b) and sulfide-bearing (c, d) systems

Miner Deposita

sulfide being completely exhausted (vertical blue line,Fig. 11c, d). This is in large due to the high Agsulfide/silicate liquid partition coefficient values suggest-ing the partitioning of silver into a fluid in a sulfide-bearing system is controlled by the amount of sulfide

present. Indeed, for each Dfluid=silicateAg value, concentra-

tion in the exsolved fluid is maximized after sulfide isentirely exhausted (Fig. 11).

In either scenario, calculations suggest that a volatilephase efficiently removes Ag from the residual silicateliquid. In the sulfide-free system, the earliest exsolvedfluid (Xwater < .002) experiences the highest enrichmentin Ag that decreases as Ag is removed from the system.The residual silicate liquid experiences a gradual reduc-tion in Ag content as vapor is lost. In the sulfide-bearing system, Ag in the residual silicate liquid is mostaffected after sulfide is totally resorbed during S loss tothe vapor (at Xwater ~ .0094). This suggests that, if sul-fide is present, Ag will be preferentially partitioned intosulfide unless large fractions of volatiles evolve from

the interstitial silicate liquid or Dfluid=silicateAg values are

sufficiently high.

Evaluating existing genetic modelsfor the Skaergaard intrusion metal budget

Although this contribution is based on a limited numberof samples, reconciling observations presented here withthe late-stage evolution of the Skaergaard and existinggenetic models for the Platinova Reef are worthwhile.Godel et al. (2014) and Nielsen et al. (2015) suggestthat discordant Au mineralization in the Platinova Reefis explained by the late-stage evolution of hydrothermalfluids. Textural observations presented in this study con-firm the existence of metal + brine assemblages andsupport the ideas of Godel et al. (2014) and Nielsenet al. (2015). The limited number of PGE outside ofthe Reef with the observed absence of Pd might reflectdifferences in Dsulfide/silicate where Dsulfide/silicate Pd > Pt>> Au > Ag ≈ Cu (Mungall and Brenan 2014). It ispossible that existing sulfide would scavenge PGE pref-erentially based on Dsulfide/silicate values. Limited sulfideavailability would prevent the partitioning of lowerDsulfide/silicate elements (e.g., Au, Ag, Cu) into sulfideallowing for concentration in the evolving silicate liquidwhere exsolved hydrothermal fluids might ultimatelytransport these metals along grain boundaries. One ob-vious complication to this hypothesis is the observationthat a distinct Cu-sulfide zone exists in the PlatinovaReef and this occurrence is commonly attributed toorthomagmatic processes (Holwell and Keays 2014;Keays and Tegner 2016). Further, native Cu reported

in this study is spatially related to sulfide minerals andis likely the result of oxidation processes.

The limited number of PGE and PGM observed outsideof the Reef might be the result of the limited number ofsamples examined in this study. While little can be saidabout the petrogenesis of the Platinova Reef, it is importantto emphasize that our results indicate that late magmatichydrothermal fluids are, in some instances, directly linkedto precious metals, particularly Ag and Au. For both Agand Au, evidence suggestive of a coexisting metal + Cl-bearing phase is observed. If fluids contemporaneouslyoxidized magmatic sulfides, the S lost from below thePlatinova Reef and its transport upward could give rise tosulfide saturation in the upper parts of the crystal pile. Inthis case, the exsolved fluid would consist of metals com-plexed as Cl and S species. Variable redox sensitivity (Cl+

vs HS−) would act to separate metals transported as Clcomplexes (Ag) from those transported as S complexes(Au) (Li and Boudreau 2019). In this scenario, the netresult would not be unlike the Pd, Au, and Cu offsets de-scribed by others (Bird et al. 1991; Andersen et al. 1998;Nielsen 2004; Holwell and Keays 2014).

Genetic model

A model is proposed where Ag, with other Cl solubleelements having low Dmnrl/melt partition coefficients, Auin particular, becomes increasingly concentrated in theresidual liquid during crystallization. At low liquid/silicate mineral mass fraction (F), Ag is concentrated inthe interstitial liquid. At this point, or shortly thereafter,enough silicate liquid volume has crystallized to allowfor the separation of H2O, CO2, Cl, and other volatilecomponents with low Dmnrl/meltpartition coefficients.Owing to its relatively high Dfluid/silicate partition coeffi-cient, Ag partitions into the exsolved fluid as Ag chlo-ride or perhaps Na-Ag-chloride. The exsolved fluidtransporting Ag will travel vertically along grain bound-ary networks, and fluid pathways, interacting with otherinterstitial phases. Heating of the fluid as it moves intohotter regions can lead to S and SiO2 loss from the hostassemblage while leaving behind a small portion of theAg-bearing volatile phase. Eventual cooling of the fluidcan cause Ag prec ip i t a t ion or, when a f lu id -undersaturated interstitial liquid is encountered, re-dissolution of the fluid initiates the precipitation of Agand perhaps an increased volatile content for that sectionof the intrusion. Minor amounts of metal + volatile aretrapped in the interstitial liquid and are presumed to havegenerated textures seen in Figs. 4a and 7. The cumula-tive effect of this process most affected the MZ wherethe silicate liquid became saturated in fluid + vapor early

Miner Deposita

enough to preserve fluid inclusions in olivine,clinopyroxene, and plagioclase.

Conclusions

Evidence for S loss is present throughout the Skaergaardstratigraphy, with the relative proportion of low-Ti magne-tite replacing sulfide being most extensive in the LZ.Native Ag is regularly associated with trace quantities ofCl in and below the MZ, consistent with a coexisting metal+ brine assemblage. Ag and other Cl soluble metals can bereadily mobilized by an exsolved volatile Cl-bearing phasein both sulfide-free and sulfide-bearing liquids. This con-tribution has implications for the origins of the PlatinovaReef. In particular, our work supports the assertions ofGodel et al. (2014) and Nielsen et al. (2015) who attributeAu mineralization to late hydrothermal fluids. Ultimately,it is suggested that a late fluid rich in halides transportedAg and perhaps other precious metals vertically throughthe Skaergaard cumulate pile. Numerical modeling sug-gests that the Kilaeua Iki lava lake may be analogous to aS-free degassing cumulate pile. Indeed, it is possible thatAg and Au are readily mobile provided the magma is S-poor or experienced an early sulfide saturation event.Finally, this contribution presents new evidence supportiveof late hydrothermal processes in LMIs and suggests thatsimilar work may clarify late-stage evolution of other in-trusions, particularly those intrusions where evidence ofhydrothermal activity has been reported (e.g., Stillwater,Bushveld).

Acknowledgments Discussions with J. Cann, E. Klein, and T.L.Nielsen were helpful and improved the quality of this manuscript.Alex Hammerstrom is acknowledged for his work compiling anddigitizing sample location information. Thorough and constructivereviews by David Holwell and Katie McFall greatly improved thequality and presentation of this manuscript. Finally, editorial han-dling by Wolfgang Maier and Georges Beaudoin is acknowledgedand much appreciated.

References

Adam J, Green T (2006) Trace element partitioning betweenmica- and amphibole-bearing garnet lherzolite and hydrousbasanitic melts: 1. Experimental results and the investigationof controls on partitioning behavior. Contrib Mineral Petrol152:1–17

Andersen JCØ, Rasmussen H, Nielsen TFD, Ronsbo JG (1998) TheTriple Group and the Platinova gold and palladium reefs in theSkaergaard Intrusion; stratigraphic and petrographic relations.Econ Geol 93(4):488–509

Andersen JCØ, Rollinson GK, McDonald I, Tegner C, Lesher CE(2017) Platinum-group mineralization at the margin of theSkaergaard intrusion, East Greenland. Mineral Deposita52(6):929–942

Andersen JCØ (2006) Postmagmatic Sulphur loss in the Skaergaard in-trusion: implications for the formation of the Platinova Reef. Lithos92(1):198–221

Ballhaus CG, Stumpfl EF (1985) Occurrence and petrological signifi-cance of graphite in the Upper Critical Zone, western BushveldComplex, South Africa. Earth Planet Sci Lett 74(1):58–68

Bird DK, Brooks CK, Gannicott RA, Turner PA (1991) A gold-bearinghorizon in the Skaergaard Intrusion, East Greenland. Econ Geol 86:1083–1092

Boudreau AE, McCallum IS (1989) Investigations of the StillwaterComplex: part V. Apatites as indicators of evolving fluid composi-tion. Contrib Mineral Petrol 102(2):138–153

Boudreau AE, McCallum IS (1992) Concentration of platinum-groupelements by magmatic fluids in layered intrusions. Econ Geol87(7):1830–1848

Boudreau AE, Mathez EA, McCallum IS (1986) Halogen geochemistryof the Stillwater and Bushveld Complexes: evidence for transport ofthe platinum-group elements by Cl-rich fluids. J Petrol 27(4):967–986

Brooks CK, GleadowAJW (1977) A fission-track age for the Skaergaardintrusion and the age of the East Greenland basalts. Geology 5(9):539–540

Cabri LJ, Beattie M, Rudashevsky NS, Rudashevsky VN (2005) Processmineralogy of Au, Pd and Pt ores from the Skaergaard intrusion,Greenland, using new technology. Miner Eng 18(8):887–897

Campbell IH, Naldrett AJ (1979) The influence of silicate: sulfide ratioson the geochemistry of magmatic sulfides. Econ Geol 74(6):1503–1506

Candela PA, Holland HD (1984) The partitioning of copper and molyb-denum between silicate melts and aqueous fluids. GeochimCosmochim Acta 48:373–380

Chaplygin I, Yudovskaya M, Vergasova L, Mokhov A (2015)Native gold from volcanic gases at Tolbachik 1975-76 and2012-13 fissure eruptions, Kamchatka. J Volcanol GeothermRes 307:200–209

Fleet ME, Chryssoulis SL, Stone WE, Weisener CG (1993) Partitioningof platinum-group elements and Au in the Fe-Ni-Cu-S system: ex-periments on the fractional crystallization of sulfide melt. ContribMineral Petrol 115:36–44

Frank MR, Simon AC, Pettke T, Candela PA, Piccoli PM (2011)Gold and copper partitioning in magmatic-hydrothermal sys-tems at 800°C and 100 MPa. Geochim Cosmochim Acta75(9):2470–2482

Godel B, Rudashevsky NS, Nie lsen TFD, Barnes SJ ,Rudashevsky VN (2014) New constraints on the origin ofthe Skaergaard intrusion Cu-Pd-Au mineralization: insightsfrom high-resolution X-ray computed tomography. Lithos190-191:27–36

Greaney A, Rudnick RL, Helz RT, Gaschnig RM, Piccoli PM, Ash RD(2017) The behavior of chalcophile elements during magmatic dif-ferentiation as observed in Kilauea Iki lava lake, Hawaii. GeochimCosmochim Acta 210:71–96

Hanley JJ, Mungall JE, Pettke T, Spooner ETC, Bray CJ (2008) Fluid andhalide melt inclusions of magmatic origin in the Ultramafic and

Miner Deposita

Lower Banded Series, Stillwater Complex, Montana, USA. J Petrol49(6):1133–1160

Heinrich CA, Driesner T, Stefansson A, Seward TM (2004) Magmaticvapor contraction and the transport of gold from the porphyry envi-ronment to epithermal ore deposits. Geology 32(9):761–764

Hirschmann MM, Renne PR, McBirney AR (1997) 40Ar/39Ar dating ofthe Skaergaard intrusion. Earth Planet Sci Lett 146:645–658

Holland HD (1972) Granites, solutions, and base metal deposits. EconGeol 67:281–301

Holness MB, Nielsen TFD, Tegner C (2007) Textural maturity of cumu-lates: a record of chamber filling, liquidus assemblage, cooling rateand large scale convection in mafic layered intrusions. J Petrol 48:141–157

Holwell DA, Keays RR (2014) The formation of low-volume, high-tenormagmatic PGE-au sulfide mineralization in closed systems: evi-dence from precious and base metal geochemistry of the PlatinovaReef, Skaergaard Intrusion, East Greenland. Econ Geol 109(2):387–406

Holwell DA, Keays RR, McDonald I, Williams MR (2015) Extremeenrichment of Se, Te, PGE, and Au in Cu sulfide microdroplets:evidence from LA-ICP-MS analysis of sulfides in the SkaergaardIntrusion, east Greenland. Contrib Mineral Petrol 170:53

Jakobsen JK, Veksler IV, Tegner C, Brooks CK (2011) Crystallization ofthe Skaergaard intrusion from an emulsion of immiscible Iron- andsilica-rich liquids: evidence from melt inclusions in plagioclase. JPetrol 52(2):345–373

Jenner FE (2017) Cumulate causes for the low contents of sulfide-lovingelements in the continental crust. Nat Geosci 10:524–530

Jenner FE, O’Neill HSC, Arculus RJ, Mavrogenes JA (2010) The mag-netite crisis in the evolution of arc-related magmas and the initialconcentration of Au, Ag, and Cu. J Petrol 51:2445–2464

Keays RR, Scott RB (1976) Precious metal in ocean-ridge basalts; impli-cations for basalts as source rocks for gold mineralization. EconGeol 71(4):705–720

Keays RR, Tegner C (2016) Magma chamber processes in the formationof the low-sulfide magmatic Au-PGE mineralization of thePlatinova Reef in the Skaergaard Intrustion, East Greenland. JPetrol 56(12):2319–2340

Larsen RB, Brooks CK, Bird DK (1992) Methane-bearing, aqueous,saline solutions in the Skaergaard intrusion, east Greenland.Contrib Mineral Petrol 112:428–437

Li Y, Audetat A (2012) Partitioning of V, Mn, Co, Ni, Cu, Zn, As, Mo,Ag, Sn, Sb, W, Au, Pb, and Bi between sulfide phases and hydrousbasanite melt at upper mantle conditions. Earth Planet Sci Lett 355:327–340

Li C, Boudreau AE (2017) The origin of high-Cu/S sulfides by shallowlevel degassing in the Skaergaard intrusion, East Greenland.Geology 45(12):1075–1078

Li P, Boudreau AE (2019) Vapor transport of silver and gold in basalticlava flows. Geology 47:877–880

Manning CE, Bird DK (1986) Hydrothermal clinopyroxenes of theSkaergaard intrusion. Contrib Mineral Petrol 92:437–447

McBirney AR (1989) The Skaergaard layered series: I. structure andaverage compositions. J Petrol 30(2):363–397

Meurer WP, Klaber S, Boudreau AE (1997) Discordant bodies fromolivine-bearing zones III and IV of the Stillwater Complex,Montana – evidence for postcumulus fluid migration and reactionin layered intrusions. Cotrib Min Pet 130(1):81–92

Meurer WP, Willmore CC, Boudreau AE (1999) Metal redistributionduring fluid exsolution and migration in the Middle Banded seriesof the Stillwater Complex, Montana. Lithos 47:143–156

Mungall JE, Brenan JM (2014) Partitioning of platinum-group elementsand Au between sulfide liquid and basalt and the origins of mantle-

crust fractionation of the chalcophile elements. GeochimCosmochim Acta 125:265–289

Naslund HR (1984) Petrology of the Upper Border Series of theSkaergaard intrusion. J Petrol 25(1):185–212

Nielsen TFD (1975) Possible mechanism of continental breakup in theNorth Atlantic. Nature 253:182–184

Nielsen TFD (2004) The shape and volume of the Skaergaard intrusion,Greenland: implications for mass balance and bulk composition. JPetrol 45(3):507–530

Nielsen TFD, Andersen JCO, Holness MB, Keiding JK,Rudashevsky NS, Rudashevsky VN, Salmonsen LP, TegnerC, Veksler IV (2015) The Skaergaard PGE and gold deposit:the result of in situ fractionation, sulphide saturation, andmagma chamber-scale precious metal redistribution by immis-cible Fe-rich melt. J Petrol 56(8):1643–1676

Patten C, Barnes S-J, Mathez EA, Jenner FE (2013) Partitioncoefficients of chalcophile elements between sulfide and sil-icate melts and the early crystallization history of sulfideliquid: LA-ICP-MS analysis of MORB sulfide droplets.Chem Geol 358:170–188

Pendergast MD (2000) Layering and precious metals mineralization inthe Rincon del Tigre Complex, Eastern Bolivia. Econ Geol 95(1):113–130

Schneider CA, RasnadWS, Eliceiri KW (2012) NIH image to ImageJ: 25years of image analysis. Nat Methods 9(7):671–675

Shaw DM (1970) Trace element fractionation during anataxis. GeochimCosmochim Acta 34:237–242

Shinohara H (1994) Exsolution of immiscible vapor and liquid from acrystallizing silicate melt: Imlications for chlorine and metal trans-port. Geochim Cosmochim Acta 58(23):5215–5221

Simon AC, Candela PA, Piccoli PM, Mengason M, Englander L (2008a)The effect of crystal-melt partitioning on the budgets of Cu, Au, andAg. Am Miner 93(8–9):1437–1448

Simon AC, Pettke T, Candela PA, Piccoli PM (2008b) The partitioningbehavior of silver in a vapor-brine-rhyolite melt assemblage.Geochim Cosmochim Acta 72(6):1638–1659

Sonnenthal EL (1992) Geochemistry of dendritic anorthosites and asso-ciated pegmatites in the Skaergaard Intrusion, East Greenland: evi-dence for metasomatism by a chlorine-rich fluid. J VolcanolGeotherm Res 52(1–3):209–230

Wager LR, Deer WA (1939) Geological investigations in East Greenland,par t I I I . The pet rology of the Skaergaard intrus ion,Kangerdlugssuaq, East Greenland, Meddelelser om Grønland, 105,pp 1–352

Wager LR, Vincent EA, Smales AA (1957) Sulphides in the Skaergaardintrusion, east Greenland. Econ Geol 52:855–903

Wang Z, Becker H (2015) Abundances of Ag and Cu in mantle perido-tites and the implications for the behavior of chalcophile elements inthe mantle. Geochim Cosmochim Acta 160(1):209–226

Webster JD, Kinzler RJ, Mathez EA (1999) Chloride and water solubilityin basalt and andesite melts and implications for magmaticdegassing. Geochim Cosmochim Acta 65:729–738

Wohlgemuth-Ueberwasser CC, Fonseca ROC, Ballhaus C, Berndt J(2013) Sulfide oxidation as process for the formation of copper-rich magmatic sulfides. Mineral Deposita 48(1):115–127

Yin Y, Zajacz Z (2018) The solubility of silver in magmaticfluids: implications for silver transfer to the magmatic-hyd ro the rma l o re - fo rming env i ronmen t . GeochemCosmochem Acta 238:235–251

Yudovskaya MA, Distler VV, Chaplygin IV, Mokhov AV, Trubkin NV,Gorbacheva SA (2006) Gaseous transport and deposition of gold inmagmatic fluid: evidence from the active Kudryavy volcano, KurileIslands. Mineral Deposita 40:828–848

Miner Deposita

Yudovskaya MA, Tessalina S, Distler VV, Chaplygin IV, Chugaev AV,Dikov YP (2008) Behavior of highly-siderophile elements duringmagma degassing: a case study at the Kudryavy volcano. ChemGeol 248:318–341

Zajacz Z, HalterWE, Pettke T, GuillongM (2008) Determination of fluid/melt partition coefficients by LA-ICPMS analysis of co-existingfluid and silicate melt inclusions: controls on element partitioning.Geochem Cosmochem Acta 72(8):2169–2197

Zajacz Z, Candela PA, Piccoli PM, Sanchez-Valle C, Walle M (2013)Solubility and partitioning behavior of Au, Cu, Ag and reduced Sin magmas. Geochim Cosmochim Acta 112:288–304

Publisher’s note Springer Nature remains neutral with regard tojurisdictional claims in published maps and institutional affiliations.

Miner Deposita