Trace Element Zonation in Carlin-Type Pyrite: Tracking Ore-Forming Processes at the Nanoscale

Carlin-type gold deposits in north-central Nevada (Fig. 1) represent one of the world’s greatest accumulations of gold, but their origins are incompletely understood. This knowledge gap is due in part to the finely zoned ore minerals, which are not amenable to traditional study techniques (Richards, 2011). These deposits are characterized by “invisible” gold hosted within micron- to nanometer-scale hydrothermal arsenian pyrite overgrowths on precursor pyrite grains, predominantly in sedimentary rocks. Previous studies have found that ore-stage pyrite from Carlin-type deposits also contains elevated concentrations of Sb, Tl, Te, Hg, and Cu with variable W, Mo, Se, Ag, Pb, Ba, Cs, and Zn (Hofstra and Cline, 2000; Cline et al., 2005; Muntean et al., 2011). The gold has been proposed to come from the sedimentary host rock package itself (Hofstra, 1994; Hofstra et al., 1999; Ilchik and Barton, 1997; Emsbo et al., 1999; 2003; Hofstra and Cline, 2000; Large et al., 2011) or from calc-alkaline magmas emplaced in the Great Basin in the Eocene (Sillitoe and Bonham, 1990; Henry and Boden, 1998; Ressel et al., 2000; Johnston and Ressel, 2004; Seedorf and Barton, 2004; Muntean et al., 2011). Indeed, Carlin-type gold deposits have many similarities to sedimentary rock-hosted distal disseminated gold deposits in the Great Basin that formed in association with known causative plutons (Sillitoe and Bonham, 1990; Johnston, 2003; Johnston et al., 2008; Bonner, 2019; Holley et al., 2019). Most recently, Holley et al. (2022) showed that the δ³⁴S of Carlin-type pyrite varies with Au concentration at the nanoscale; the sulfur isotope signatures of the Au-rich zones are consistent with an Eocene magmatic origin, but meteoric fluids contributed during growth of low-Au zones.

High spatial-resolution analytical techniques have enabled detailed study of Carlin-type pyrite, paving the way for more complete understanding of Carlin-type gold deposit formation. For example, high-resolution transmission electron microscopy (HRTEM) and high-angular annular dark-field imaging-scanning transmission electron microscopy (HAADF-STEM) studies enabled imaging of Au nanoparticles in arsenian pyrite (Palenik et al., 2004) and identification of the As-dependent solubility limit of Au in pyrite (Reich et al., 2005), temperature and size limit of Au nanoparticles in pyrite (Reich et al., 2006), and Au concentrations in Carlin-type fluids (Liang et al., 2021). Laser ablation-inductively coupled plasma-mass spectrometry (LA-ICP-MS) studies revealed the contrasting trace element enrichments of the precursor pyrite and the hydrothermal overgrowths, as well as potential synsedimentary gold enrichment in the early pyrite (Large et al., 2009, 2011; Large and Maslennikov, 2020). However, the 25- to 40-μm spot sizes of quantitative laser-ablation analyses far exceed the scale of the zonation in Carlin-type pyrite, limiting the utility of the technique. Unstandardized, qualitative nanoscale secondary ionizing mass spectrometry (NanoSIMS) mapping of the pyrite has shown that the hydrothermal pyrite is zoned on the submicron scale, and the zones of high Au are also enriched in As, Sb, Cu, and in some instances Te (Barker et al., 2009). Atom-probe tomography of a single grain from Turquoise Ridge showed that the Au is lattice bound within the crystal structure of the pyrite, potentially facilitated by point defects caused by As (Gopon et al., 2019). Quantitative spot analyses are now achievable at a scale relevant for Carlin-type pyrite using NanoSIMS depth profiles (Holley et al., 2022).

Despite this progress, the mechanisms controlling trace element enrichments in Carlin-type gold remain unclear. Most studies have suggested that the Au-As zoning resulted from temperature changes or a changing fluid source. This idea is supported by stable isotope mixing models for the pyrite in Carlin-type gold deposits; the δ³⁴S signatures in the zoned hydrothermal pyrite require episodic mixing of magmatic and meteoric fluids at the local scale (Holley and Phillips, 2022). Thermochemical modeling shows that As concentrations increase in pyrite with increasing fluid-to-rock ratio, potentially leading to a decrease in pyrite volume and an increase in pyrite crystal porosity, which could enhance Au sequestration (Xing et al., 2019). Experimental studies and thermodynamic modeling show that Au partitions preferentially into pyrite as the As content of the pyrite increases (Kusebauch et al., 2019). Based on these observations, Kusebauch et al. (2019) suggested that gold is scavenged from the liquid during pyritization, and that sulfidation may become important later when the fluid is supersaturated with Au. However, these models are incompatible with the lack of correlation in As and Au enrichment observed in nanoscale zones (Holley et al., 2022).

In this contribution, we compare data from eight gold deposits in Nevada to shed light on the complex processes responsible for auriferous pyrite mineralization in the region. We present new high spatial-resolution analyses of Cu, As, Ag, Au, and Sb in hydrothermal pyrite overgrowths. We combine these data with the NanoSIMS δ³⁴S data published in Holley et al. (2022) and additional previously unpublished δ³⁴S analyses to examine the potential mechanisms that govern metal enrichment and ultimately the total Au endowment of Carlin-type gold deposits. We combine trace element data from NanoSIMS depth profile analyses and NanoSIMS mapping to characterize the trace element distributions in Carlin-type pyrite at a previously unachievable spatial resolution. We present results from Carlin, Deep Star, Beast, and unmineralized Eocene dikes at Betze Post and Deep Star on the Carlin trend, as well as data from Getchell and Turquoise Ridge on the Getchell trend. To assess the relationship between trace element enrichment and proximal intrusions, we compare these pyrites to new data from texturally similar hydrothermal pyrite overgrowths from two deposits in the nearby Battle Mountain district: the Lone Tree distal disseminated deposit and the Red Dot sedimentary rock-hosted gold deposit of unknown origins. Our data provide new insights on the genesis of these deposits, and our approach will enable similar studies in the numerous other mineralizing systems that host multiple generations of fine-grained ore minerals.

Most of Nevada’s Carlin-type gold deposits are hosted in Paleozoic sedimentary rocks that were deposited in carbonate shelf and slope sequences along a passive margin. Contractional deformation of the continental margin occurred from the Devonian to the Eocene. The Late Devonian to Early Mississippian Antler orogeny and the Late Permian to Early Triassic Sonoma orogeny reactivated ancient rift structures, thrusting deep-water siliciclastic rocks over carbonate shelf-slope rocks along the Roberts Mountain Thrust and Golconda Thrust, respectively. These events positioned extensively fractured carbonate facies over fault zones linked to basement rifts and overlain by comparatively unreactive siliciclastic caprocks (Hofstra and Cline, 2000; Muntean et al., 2011). The majority of Carlin-type gold is hosted beneath the Roberts Mountain allochthon. Carlin-type deposits are commonly localized along Eocene extensional faults, which may represent reactivation of earlier compressional structures. Beginning in the early Miocene, transtensional torsion of the continent led to the Basin and Range faulting, which dismembered many of the gold deposits.

The Eocene transition from shallow subduction to slab delamination and rollback is thought to have led to Carlin-type hydrothermal fluid circulation (Ressel and Henry, 2006; Muntean et al., 2011), and an Eocene mineralization age is generally assumed for Carlin-type gold. Geochronological evidence is based primarily on crosscutting relationships (numerous studies compiled in Cline et al., 2005; Ressel and Henry, 2006). However, direct dating has occurred at only a few deposits. Eocene ages were obtained at Getchell adjacent to Turquoise Ridge using Rb-Sr dating of late ore-stage galkhaite (a rare Hg-sulfosalt; Tretbar at al., 2000; Arehart et al., 2003) and ⁴⁰Ar/³⁹Ar dating of adularia at Twin Creeks on the Getchell trend (Groff, 1997; Hall et al., 2000). The adularia was sampled from vein mineralization atypical of Carlin-type mineralization, and the paragenetic relationship with gold is not entirely certain. Apatite fission track and apatite (U-Th)/He ages are reset by the temperature and duration of Carlin-type mineralization and have been used to obtain evidence for a pulse of late Eocene hydrothermal fluid flow on the Carlin trend (Arehart et al., 1993, 2003; Chakurian et al., 2003; Hickey et al., 2014), the Getchell trend (one sample; Hofstra et al., 1999), and at numerous Carlin-like sedimentary rock-hosted gold deposits in the Battle Mountain district (Huff et al., unpub. data, 2024).

The region experienced major periods of magmatism in the Jurassic, the Late Cretaceous, and the Eocene (du Bray, 2007), resulting in a metasomatized and potentially metal-rich subcontinental lithospheric mantle that could have contributed to mineralization (Muntean et al., 2011). Intrusive activity in the region was continuous but less voluminous from the Eocene to the Miocene (du Bray, 2007). Eocene intrusions are known near some Carlin-type gold deposits, whereas at others Eocene intrusions are lacking or remain unidentified. The presence of Eocene dikes and an airborne magnetic anomaly have been used to infer the presence of a ~12- × 50-km composite batholith underneath the Carlin trend (Ressel and Henry, 2006). On the Battle Mountain-Eureka Trend, Carlin-type gold mineralization is thought to have occurred prior to the formation of the giant late Eocene Caetano caldera, but cross-cutting relationships with dikes suggest that the magmatic system was already active at the time of mineralization (John et al., 2008; Colgan et al., 2011; Maroun et al., 2017; Henry et al., in press). In contrast, no Eocene intrusive rocks have been identified in the Gold Bar district in the southern part of the Battle Mountain-Eureka trend (Cline et al., 2005). The Au-rich zones of Carlin-type hydrothermal pyrite have magmatic δ³⁴S signatures at Carlin and Deep Star on the Carlin trend where Eocene magmatism is known to have occurred, as well as at Turquoise Ridge and Getchell where the nearest Eocene intrusions are 7 km away (Holley et al., 2022). In other areas, such as the Battle Mountain district, mineralization occurred more proximal to causative Eocene intrusions, including porphyry, skarn, and distal disseminated deposits (Theodore et al., 1973; Doebrich, 1995; Meinert, 2000; Kizis et al., 1997; Keeler, 2010; Reid et al., 2010; King, 2011, 2017; Holley et al., 2019). The district also hosts other gold deposits in sedimentary rocks without clear links to magmatism (Fithian et al., 2018; Huff et al.,unpub. data, 2024).

In this study we examine Au-bearing hydrothermal pyrite from Carlin-type and other sedimentary-hosted gold deposits, focusing specifically on the trace element deportments and zoning characteristics. We use these data to discuss the origins of these deposits, including the relationship to magmatism and potential precipitation mechanisms. In the following sections, we briefly summarize the geology of our study locations: Carlin, Deep Star, and Beast on the Carlin trend as well as dikes from Betze Post and Deep Star; Getchell and Turquoise Ridge on the Getchell trend; and Lone Tree and Red Dot in the Battle Mountain district.

The Carlin deposit is located approximately 30 km northwest of the town of Carlin, Nevada (Fig. 1). Carlin was the first Carlin-type gold deposit discovery and the first producing mine in the deposit type. The orebody was mined by open pit from 1965 to 1987, producing nearly 4 Moz Au (United States Geological Survey [USGS], 1994). Gold occurs in decarbonatized, silicified, and argillized carbonate rocks of the Roberts Mountain, Popovich, and Rodeo Creek formations. The deposit is located near the crest of the N-NW–striking Tuscarora anticline in association with numerous high-angle faults. The unoxidized ore occurs primarily within overgrowths on pyrite, in association with orpiment, realgar, cinnabar, stibnite, quartz, kaolinite, sericite, barite, and residual calcite, dolomite, illite, and carbonaceous material.

The high-grade Deep Star deposit is located northwest of Carlin (Fig. 1) and was mined by underground methods from the 1990s to 2008, producing approximately 1.9 Moz Au (Nevada Bureau of Mines and Geology, 2011). The orebody was about 300 m south of the Jurassic Goldstrike intrusive suite along the southern extension of the Gen-Post fault system. This fault system and other deep crustal structures throughout the Carlin trend related to Proterozoic rifting served as conduits for metal-rich fluids to reach overlying and adjacent favorable host facies (Heitt et al., 2003; Cline et al., 2005). The Deep Star orebody is hosted beneath the Roberts Mountain Thrust in brecciated silty and carbonaceous rocks of the Devonian Popovich Formation and partially within the Jurassic Goldstrike diorite. Gold mineralization is associated with quartz-kaolinite alteration and is hosted in arsenian rims on pyrite and marcasite, with post-ore calcite, siderite, and barite (Heitt et al., 2003). The entire Deep Star orebody lies within the hornfels contact metamorphic aureole of the Goldstrike stock and sill complex (Goldfarb et al., 2016). Gold mineralization at the deposit occurs in sedimentary rocks and to a lesser extent in the Goldstrike diorite. Aphyric rhyolite dikes intruded both syn- and post-ore, constraining the age of mineralization to roughly 38 Ma (Heitt et al., 2003). We examined pyrite grains in an unmineralized, detvitrified aphyric rhyolite dike from Deep Star (39.15 ± 0.26 Ma; Ressel and Henry, 2006).

The Beast deposit is located near the southern end of the Carlin trend, about 1.5 km south of Deep Star (Fig. 1). The deposit was mined by open pit from the mid-1990s to the 2000s. Approximately 0.2 Moz Au were hosted in the structurally controlled deposit, which follows the N-NW–striking Genesis fault system and is localized along the E-dipping dip-slip Beast fault splay (Ressel et al., 2000). Host rocks include silty carbonates and calcareous mudstones of the Roberts Mountain formation, as well as the Beast dike, which hosted about half of the deposit’s ore. The 55-m-wide Beast dike parallels the Beast fault and is part of a discontinuous series of porphyritic rhyolite and dacite dikes that stretches about 16 km in length from the deposit to Welches Canyon in the south (Ressel et al., 2000). At Beast, the rhyolite has low SiO₂ content and is coarsely porphyritic, characterized by phenocrysts of plagioclase, embayed quartz, biotite, hornblende, and sanidine; the latter are up to 2.5 cm long. Sanidine from this dike gave an ⁴⁰Ar/³⁹Ar age of 37.58 ± 0.06 Ma (Ressel and Henry, 2006).

We also examined pyrite from Eocene dikes at Betze-Post deposit to compare with the Carlin-type pyrite. The giant Betze-Post is located 3 km north of the Beast deposit and is mined via the Goldstrike open pit. Ore is hosted in the Popovich, Rodeo Creek, and Roberts Mountains formations (Ferdock et al., 1997; Peters et al., 1997). Finely porphyritic rhyolite dikes are exposed over a 9-km distance from Betze-Post to the south of Genesis, mostly along the Post fault (Ressel and Henry, 2006). The rhyolite is characterized by plagioclase and biotite ± quartz, and biotite from a dike in the Goldstrike pit gave an ⁴⁰Ar/³⁹Ar age of 39.32 ± 0.11 Ma (Ressel and Henry, 2006).

The Getchell mine is located about 70 km northeast of Winnemucca, Nevada (Fig. 1), and was first mined for gold in oxide ore in 1938, prior to recognition of Carlin-type gold deposits as as a deposit type. The mine has been operated intermittently since then, including as a producer of arsenic from roasting of sulfide ores during World War II. The Turquoise Ridge mine is located about 1 km east of the main Getchell pit and began producing oxide ore from an extension of the Getchell deposit in 1991. Current underground mining is by underhand cut-and-fill methods accessible by shaft, with 12 Moz Au reserves and resources as of 2019 (Nevada Gold Mines, 2019). The ore is hosted in Cambrian to lower Ordovician sedimentary rocks and sea-floor basalts, predominately in carbonate turbidites and debris flows of the Comus Formation in the footwall of the Getchell fault. The E-dipping Getchell range front fault bounds the eastern margin of the Osgood mountains. The gold-mineralized zones commonly occur in the sedimentary rocks near the contacts with granodiorite dikes of the Cretaceous Osgood stock (Muntean et al., 2009; Cox et al., 2018). The ore is discordant to stratigraphy, controlled by structural intersections, fracture zones, unit margins, fold margins, and calcareous rocks. Gold is associated with trace element-rich pyrite, kaolinite, illite, and jasperoid quartz, and late ore-stage minerals include drusy quartz, orpiment, fluorite, stibnite, realgar, and calcite (Cail and Cline, 2001; Cline, 2001).

The Lone Tree gold deposit is located at the northern end of the Battle Mountain district, approximately 50 km south of Getchell and Turquoise Ridge (Fig. 1). The deposit produced 4.2 Moz Au before it closed in 2006 with a reserve of 1.25 Moz (Newmont, unpub. data, 2004), and exploration recommenced in 2020 (i80 Gold Corp, 2021). Although the Battle Mountain district hosts several porphyry and skarn deposits, the gold at Lone Tree is mostly hosted in sedimentary rocks between the Roberts Mountain and Golconda thrusts. Host rocks include siliciclastic units of the Ordovician Valmy, Pennsylvanian-Permian Battle Mountain, and Edna Mountain formations; siliciclastic and carbonate rocks of the Mississippian-Permian Havallah Formation; and 40.95 ± 0.06 Ma rhyolite dikes. Like Carlin-type deposits, the gold is primarily hosted in fuzzy hydrothermal overgrowths on precursor pyrite, but the hydrothermal phases include both arsenian pyrite and arsenopyrite. Decarbonatization and Fe carbonate alteration are notable in the carbonate-bearing units, whereas in the dikes and siliciclastic rocks the gold is associated with sericitic and argillic alteration. In the dikes, the arsenopyrite-rimmed pyrite is intergrown with sericite with δ¹⁸O values of 1.6 to 9.5‰ and δD values of –105 to –145‰, consistent with felsic magmatic fluids that were minorly modified by mixing with meteoric water and wall rock exchange in the distal disseminated environment (Holley et al., 2019). The precursor pyrite has δ³⁴S values of 3.4 to 7.7‰, similar to the isotopic signature of magmatic sulfur in the district (Theodore et al., 1986; King, 2017; Holley et al., 2022), whereas the Au-bearing arsenopyrite rims are 5.3 to 6.5‰ higher than this range, based on lower spatial resolution laser ablation multicollector (MC)-ICP-MS data and whole-grain analyses. Holley et al. (2019) explored several mechanisms for the observed fractionation between pyrite cores and rims, concluding that the higher δ³⁴S in the rims resulted from increasing H₂S to SO₂ ratio as the magmatic-hydrothermal fluid interacted with sedimentary wall rocks prior to mineralization.

The Red Dot deposit is located about 10 km south of Lone Tree, within the Marigold mine area (Fig. 1). Marigold comprises an ~8-km N-S–trending cluster of gold enrichments that are mined by a series of open pits. The deposits share similar many geochemical, mineralogical, and structural traits in common with Carlin-type deposits (Fithian et al., 2018). However, ore at Marigold is primarily mined from oxidized siliciclastic rocks in the upper plate of the Roberts Mountain thrust, and little drilling has been conducted in the footwall rocks of the Comus-Preble Formation (Fithian et al., 2018). The oxidized nature of the Marigold ore zones has hampered geologic understanding of the ore-forming processes. The recently discovered Red Dot orebody hosts 1.7 Moz Au (SSR Mining Inc., 2022), mostly beneath the redox boundary. The geology of the deposit has been described in detail by Huff et al. (in press). The orebody is primarily hosted in quartzites and argillites of the Valmy Formation, which is locally up to 500 m thick and was extensively deformed prior to deposition of the overlying Antler sequence (Huff et al., in press). Mineralization at Red Dot is associated with silicification, argillization, and sulfidation. The gold occurs within arsenian pyrite overgrowths on precursor pyrite grains. The relatively unoxidized nature of the deposit provides a new opportunity to examine the hypogene ore, and in the present study we use NanoSIMS to compare the pyrite trace element geochemistry to Lone Tree and the classic Carlin-type deposits.

This study is based on data from 45 samples of high-grade (>5 g/t Au) ore from Carlin, Deep Star, Getchell, Turquoise Ridge, Lone Tree, and Red Dot, as well as dike-hosted pyrite from Beast, Deep Star, and Betze-Post. From each deposit, we targeted stratigraphic horizons known to be well mineralized, displaying characteristic Carlin-type or distal disseminated hydrothermal alteration styles. We selected samples based on the presence of abundant pyrite with hydrothermal rims. We targeted grains sufficiently coarse for in-situ NanoSIMS analysis (>5 µm), and we also documented the presence and characteristics of finer-grained ore minerals. The pyrite grains from Eocene dikes on the Carlin trend at Deep Star, Beast, and Betze Post were from units previously dated by Ressel and Henry (2006). There were relatively few grains in the dike samples from Deep Star and Betze-Post, so we selected the only grains of sufficient size. The dike samples from Beast contained more abundant pyrite, so we were able to use the selection practices described above.

Preliminary petrography was performed using a Carl Zeiss Axio Scope A1 polarizing microscope in the Colorado School of Mines Mining Geology Research Laboratory on polished 30-µm thin sections. Field emission-scanning electron microscopy (FE-SEM) images were collected at the Colorado School of Mines Mineral and Materials Characterization Facility using a TESCAN MIRA3 LMH Schottky FE-SEM equipped with a single-crystal YAG (yttrium, aluminum, garnet) backscattered electron detector (BSE) and a Bruker XFlash 6/30 silicon drift detector for energy dispersive X-ray (EDX) spectrometry. The BSE images were collected at an accelerating voltage of 20 kV, a 5-µm beam width, and a 10-mm working distance. Point analyses were also performed using EDS at the same conditions to help identify minerals based on semiquantitative element compositions and to identify As-rich pyrite, which commonly corresponds to Au enrichment (e.g., Reich et al., 2005; Kusebauch et al., 2019).

Electron probe microanalysis (EPMA) was used to identify minor and abundant trace element variations as well as major element concentrations in selected sulfide grains from Red Dot to achieve the relative sensitivity factor calibration described in the next section. Due to an analytical volume of several microns in diameter, EPMA data average several growth zones within many of the small hydrothermal pyrite grains. EPMA work was carried out at the University of Colorado Boulder Electron Microprobe Lab on a JEOL 8230 Superprobe equipped with five wavelength-dispersive spectrometers. The microprobe was operated with an accelerating voltage of 20 kV, beam current of 50 nA, 40˚ takeoff angle, and a focused beam. Elements analyzed include Ag, As, Au, Bi, Cd, Co, Cr, Cu, Fe, Hg, Ni, Pb, S, Sb, Te, Ti, U, V, and Zn.

Semiquantitative trace element maps and quantitative spot analysis depth profiles of ³²S^–, ³⁴S^–, ⁶³Cu^–, ⁷⁵As^–, ¹⁰⁷Ag^–, and ¹⁹⁷Au^– were acquired on the Stanford Nano Shared Facility CAMECA nanoscale secondary ionizing mass spectrometer (NanoSIMS) 50L instrument. In some analyses, ³⁶S^– was collected instead of ⁶³Cu^–, and in some mapping ¹²¹Sb^– was also collected. Prior to analysis, all samples were polished with 0.25-µm diamond grit and coated with 15 to 20 nm of carbon for conductivity. Thin section samples were cut down to 7- × 7-mm squares to fit in the NanoSIMS sample holders.

For mapping, all data were collected on electron multiplier detectors. The maps were generated using a ~52,000 pA Cs⁺ ion beam accelerated at 8 keV, a beam current of 3 to 4 pA, and a 100-nm beam diameter. The mass peak positions of ⁶³Cu^–, ⁷⁵As^–, ¹⁰⁷Ag^–, and ¹⁹⁷Au^– were confirmed with chalcopyrite, arsenopyrite, and Ag-bearing tennantite from our sample set (identified by SEM-BSE and EPMA), and gold coating working references, respectively. Trace element concentrations in sulfide mineral standards for SIMS have not been rigorously quantified at the nanoscale and are spatially heterogeneous in our experience. We quantified the trace element concentrations using the relative sensitivity factor (RSF) method described at the end of this section. The maps cover areas ranging from 20 × 20 µm to 30 × 30 µm at a sampling resolution of 256 × 256 pixels (78–117 nm per pixel), allowing the observation of changes on the single-micron scale. Prior to mapping, the region was presputtered in 50- × 50-µm rasters at 500 pA to remove the carbon coating. The maps comprise between 15 and 45 frames collected sequentially for each area. The NanoSIMS image files were reduced using the OpenMIMS plugin package for the software ImageJ. The frames were corrected for 44 ns deadtime, and prior to summing the frames for sulfur isotopes, the data were corrected for quasisimultaneous arrivals using a correction factor of 0.75 (Traxlmayr et al., 1984; Hillion et al., 2008).

For spot analyses, a ~350-pA Cs primary ion beam with 16-keV impact energy was used to sputter trace elements and S isotopes from the sample. The ³²S^– and ³⁴S^– data were collected on Faraday cups with 10¹¹ Ohm resistors to avoid the effects of quasisimultaneous arrival and fast electron multiplier aging (Zhang et al., 2014). The ³⁶S^–, ⁶³Cu^–, ⁷⁵As^–, ¹⁰⁷Ag^–, and ¹⁹⁷Au^– data were collected on electron multipliers, simultaneous with collection of ³²S^– and ³⁴S^–. Mass peak positions were confirmed as described above. Mass resolution ranged between ~6,300 and 7,400. For ³²S^–, ³⁴S^–, and ³⁶S^–, mass resolution was set to >6,300 to resolve isobaric interferences including ¹⁶O₂ (and ³¹P¹H^–) on ³²S^–, or ³³S¹H^– on ³⁴S^–, or ³⁵Cl¹H^– and ³⁶C₃ on ³⁶S^–. The entrance and aperture slits were used to yield flat-topped peaks with steep sides. Target areas were presputtered for up to 2 minutes to remove the carbon coat, remove surface contamination, and implant Cs⁺ in the surface of the sample. For each analysis, the primary beam was rastered over a 2- × 2-µm area divided into 64 × 64 pixels with a 245-µsec per pixel dwell time over 2,400 frames. Data were collected as counts per second, allowing for comparable time-resolved signals for different trace element masses during each depth profile analysis. Background and Faraday Cup corrections were conducted similar to Hauri et al. (2016). Using conservative estimates of the apparent thickness of hydrothermal pyrite rims in thin section, we determined that the spot analyses generated craters less than 2 µm deep. The individual data points in a depth profile represent depth intervals of less than 1 nanometer. Craters from several representative samples were also measured using the Colorado School of Mines Helios NanoLab 600i FIB-SEM, to confirm the estimate of crater depths.

We report the ³⁴S/³²S isotope ratios in parts per thousand (per mil or ‰) relative to Vienna Canyon Diablo Troilite (VCDT) in delta notation (δ) (Krouse and Coplen, 1997) using Equation (1):

The standard ratio for VCDT was taken as ³⁴S/³²S_VCDT = 0.04416259 (Ding et al., 2001).

We corrected for instrumental mass fractionation (IMF) using standard-sample-standard bracketing, wherein between one and five sample analyses were bracketed on each side by one, two, or three analyses of Elba pyrite (δ³⁴S_VCDT 9.37 ± 0.36; Holley et al., 2022) or Balmat pyrite (δ³⁴S_VCDT 16.39 ± 0.20; Kozdon et al., 2010). At the beginning of the first session of spot analyses, we used Cañon Diablo Troilite (CDT) as a secondary standard to bracket two to three analyses of the reference pyrite. During the second session of spot analyses, we analyzed CDT each day. We corrected for IMF using the normalization method, based on a time-dependent IMF correction. This method uses the same range of frames for the standard and the unknown to adjust for possible variation of IMF over the length of the analysis. Throughout our analytical sessions, the day-to-day standard reproducibility (1s) of Balmat pyrite varied from 0.4 to 0.7‰; the Elba pyrite varied from 0.7 to 1.2‰; and the Cañon Diablo Troilite varied from 0.5 to 1.1‰.

We confirmed that our δ³⁴S values were unaffected by spurious trace element-induced fractionation or “matrix effects,” using the procedure described in Holley et al. (2022). Laser-ablation MC-ICP-MS analyses have recently become possible at a 5-µm spot size. We used this approach to validate the NanoSIMS δ³⁴S data, confirming both the magnitude and direction of δ³⁴S fractionation data. These analyses were conducted using Teledyne CETAC Photon Machines G2 Excimer 193-nm laser ablation system coupled to a Nu Instruments HR^© MC-ICP-MS (Pribil et al., 2015) at the Geology, Geophysics and Geochemistry Science Center at the U.S. Geological Survey. The results of the validation and a representative subset of the ³²S and ³⁴S data are published in Holley et al. (2022).

We quantified our NanoSIMS elemental and isotopic compositions using plateaus from time- and depth-resolved spot analyses. To calculate plateau values, we identified regions of distinct isotopic and chemical signatures in each analysis and split the data into multiple points consisting of 101 to 401 frames based on inspection of the secondary ion intensity for each detector versus time. Some analyses penetrated multiple zones of varying trace element enrichment, such as depth profiles within hydrothermal pyrites and depth profiles through the transition between hydrothermal pyrite and precursor pyrite. For these plateau calculations, frames were selected on two criteria: flatness of the curve for the time-resolved ^34/32S ratio, and trace element data to demarcate the bounds of the flat segment. For analyses that contained multiple plateaus, we generated the corresponding number of plateau data points. To avoid statistical bias, we calculated the isotope ratios based on the sum of the total counts for each isotope, and not the mean of the ratios (Ogliore et al., 2011).

We present our NanoSIMS data as x-y plots of depth profiles, showing the variation in elemental or isotopic composition according to depth within the grain. Throughout the manuscript, we describe peaks and plateaus in the x-y plots. In LA-ICP-MS and NanoSIMS data, flat-topped and steep-sided peaks are commonly assumed to represent inclusions such as other sulfides, sulfosalts, and native metals in pyrite. In contrast, our NanoSIMS data show flat plateaus of homogeneous composition, interspersed with sharp-topped peaks showing gradational changes in composition. We interpret these gradational changes to represent compositional variation within the pyrite itself, rather than differences between discrete mineral phases. Depth profile intervals with numerous peaks and troughs represent transects through regions of pyrite that are geochemically zoned on a fine scale.

In addition to the x-y plots of depth profiles, we also report Pearson correlation coefficients for Au, Ag, Cu, and As in the depth profiles. These correlation coefficients show the extent to which the concentrations of two elements covary, within the region of the grain that was analyzed during spot analyses. We describe correlations as strongly positive (correlation coefficient >0.7), strongly negative (<–0.7), moderately positive (0.4–0.7), and moderately negative (–0.4 to –0.7); we also describe instances where no significant correlation is observed (–0.4 to 0.4). Since the correlation coefficients are each calculated using all the data from a depth profile (or region within a depth profile), they do not capture the subtle variation within and between zones that is best visualized using the x-y plot of the depth profile trace element variation.

Trace element concentrations were calculated using a single-point standard, based on methods from Zhang et al. (2017). This method was used instead of standardization because sulfide standards for minor and trace elements have yet to be rigorously characterized at the nanoscale. Calibration curves were not viable for RSF correction because most NanoSIMS depth profiles reported varying concentrations of trace elements that could not be matched with EPMA concentration due to the large EPMA sampling volumes. To create a single-point standard, RSFs were calculated from a uniform trace-element–rich spot analysis of a Red Dot pyrite and corresponding EPMA concentrations from the same zone. Red Dot was selected because the zonation is coarser than the Carlin-type pyrites analyzed in this study. All RSFs were calculated from Red Dot NanoSIMS spot #75 due to its consistent trace element concentrations throughout the analysis. Calculated RSFs for Cu, As, Ag, and Au were 0.57 ± 0.09, 0.034 ± 0.002, 4.2 ± 1.7, and 0.94 ± 0.09, respectively. Approximate trace and minor element concentrations were calibrated using the following equation:

where C_A and C_32S are the concentrations of A (a given element) and ³²S measured by EPMA, and A/³²S is the ratio of secondary ion intensity of element A to ³²S measured via NanoSIMS. The RSF values for Cu and As increase with depth in samples and standards. We were able to correct for the depth variation in RSF for arsenic by normalizing to the As/³²S curve of the bracketing standards. The standards contained little Cu, so the same approach could not be used to correct the depth variation in Cu RSF. Future studies should focus on characterizing the trace and minor elements in sulfide mineral standards alongside δ³⁴S.

We prepared transmission electron microscopy (TEM) foils of coarse-grained pyrite with thick hydrothermal rims from Deep Star and Red Dot. Standard focused ion beam (FIB) specimen preparation techniques were used to prepare TEM specimens from the same polished specimens used for NanoSIMS. An FEI Co. (Hillsboro, OR, USA) Helios NanoLab 600i scanning electron microscope (SEM)/FIB workstation with a Ga ion column was operated initially at 30 kV followed by 2 kV for ion beam damage removal. We examined the Red Dot sample foil on an FEI Co. Talos F200X 200 keV field mission scanning and transmission electron microscope (STEM). The instrument is equipped with four silicon drift detectors and has a STEM imaging resolution of 0.16 nm. Despite numerous attempts, extracted TEM specimens prepared from the Deep Star sample were not successful due to the porosity of the pyrite rim compromising the structural integrity of the thin specimen.

Our samples from the Carlin mine are carbonaceous, orpiment- and realgar-bearing, decalcified, silty dolomitic limestone of the Silurian to Devonian Roberts Mountain Formation, described by Radtke (1985) and collected during the operating period of the mine. These rocks contain disseminated euhedral sedimentary pyrite grains that range from 5 to 20 µm in diameter at a modal abundance of approximately 1% (Fig. 2a-e). These pyrites are solid, competent grains with no internal variations visible in reflected light or BSE. The NanoSIMS element maps indicate that the sedimentary pyrites are trace element poor with little internal variation in As or Au, although one grain core shows cubic Ag-poor zones in a matrix of minorly elevated Ag (Fig. 2f-h). Quantitative NanoSIMS spot analyses gave concentrations up to 1.3 ppm Au, 4 ppm Ag, 12 ppm Cu, and 7 ppm As in the sedimentary pyrite. There is no correlation between Au and Ag or between Ag and As. However, minor enrichments of Au and As covary spatially, indicated by a correlation coefficient of 0.9 between these elements in the sedimentary pyrite from Carlin. The δ³⁴S composition of the sedimentary pyrite grains varies widely from 54.4 to –33.5‰ (Fig. 3a-d; App. 1).

All of the sedimentary pyrite grains in the Carlin deposit ore samples are coated by very thin arsenian pyrite rims. These rims are typically 1 to 2 µm wide where visible in reflected light (Fig. 2e), although rims less than a micron wide are visible in BSE imaging at high magnification. All rims have sharp contacts with the precursor pyrite, but within the rims the textures are fuzzy or porous. The exterior margins are wispy, and small tendrils extend into the rock matrix. There is also about 1% modal abundance of micron- to submicron-scale, single-stage arsenian pyrite or marcasite disseminated in the matrix, although these grains were too small for quantitative analysis. The NanoSIMS mapping of the arsenian pyrite rims identified elevated Au, As, and Ag relative to the grain cores. The NanoSIMS spot data show that the arsenian rims contain up to 689 ppm Au, 148 ppm Ag, 192 ppm Cu, and 1,990 ppm As (Table 1). Although none of the arsenian pyrite rims or disseminations show visible internal variation in reflected light or BSE, the NanoSIMS depth profiles show that the rims are internally heterogeneous, characterized by hundreds of nanoscale oscillations in trace element content across the total width of the rims in both concentric and patchy sectoral zones. The depth profiles show very similar peak positions and peak shapes for Cu, As, Ag, and Au, indicating that these elements are tightly correlated in the rims (Fig. 3a-d). However, the trace element ratios vary within individual zones, as evidenced by inconsistent trends in peak height between zones. Due to the small grain size and thin rims in the Carlin pyrites, all of the NanoSIMS depth profiles started in the precursor grains and penetrated into the underlying arsenian rims, allowing examination of the contacts between cores and rims. The contacts are marked by a sharp increase in trace metal concentrations for Cu, As, and Au, and in some cases Ag (Fig. 3a-c). However, some grains have higher Ag in the core than the rim, in which case the Ag concentration decreases sharply at the contact (Fig. 3d). The Au and As are strongly correlated within some hydrothermal rims (correlation coefficients of 0.8–0.9), whereas in other rims the correlation is weak (0.6) or absent (0.1–0.4). Correlations between Au and Ag are inconsistent with strong positive correlations in some grains and moderate negative correlations in others (–0.4 to 0.9), as is the case for Ag and As (–0.4 to 0.7). The Au is consistently correlated with Cu (0.8–0.9), and Cu is strongly correlated with As (1.0). The δ³⁴S values show a sharp transition at the contact between sedimentary pyrite grain cores and hydrothermal pyrite rims: low-δ³⁴S sedimentary pyrites are rimmed by hydrothermal pyrite with higher δ³⁴S than the pyrite grain core. High-δ³⁴S sedimentary pyrites are rimmed by hydrothermal pyrite with lower δ³⁴S than the pyrite grain core. The δ³⁴S values of the hydrothermal rims generally covary with Au content, approaching the magmatic δ³⁴S field where Au content is high, regardless of whether the precursor pyrite δ³⁴S is low or high (Fig. 3a-c).

Our samples from the Deep Star deposit are calcareous siltstone of the Devonian Popovich Formation, from a fault zone in the hornfels contact metamorphism adjacent to the Jurassic Goldstrike stock (Goldfarb et al., 2016). These ore samples contain disseminated, 20- to 200-µm, euhedral to subeuhedral pyrite grain cores (Fig. 4a-e) that have been attributed to Jurassic magmatic-hydrothermal skarn mineralization given their relatively coarse grain size and texture (Goldfarb et al., 2016). The δ³⁴S values of the pyrite grain cores are tightly clustered from 6.5 to 6.9‰ (Fig. 3e-h; App. 1). These pyrites have a modal abundance of about 1%. The majority are solid, competent grains with sharp margins and no apparent zonation of the grain core in BSE imaging (Fig. 4b, d, e, g). The trace element concentrations of these grain cores are low (Fig. 4f, h-p), although they contain minor Au up to 0.4 ppm, up to 24 ppm Ag, 0.3 ppm Cu, and 237 ppm As (Table 1). Depth profiles from individual NanoSIMS spot analyses provide further detail on the composition of the precursor magmatic-hydrothermal pyrites (Fig. 3e, f). The correlations between Ag and Au are negative to absent (–0.6 to 0.0), as are the correlations between Ag and As (–0.8 to 0.0). The correlations between Au and As vary depending on the sample (–0.4 to 0.8). There is no correlation between Cu and Au or between Cu and Ag, whereas Cu shows a moderate positive correlation with As (0.6). There are also aggregates of finer-grained pyrite on the scale of 2 to 10 µm at an abundance of less than 1%. These smaller grains have variable porosity and fracture patterns, and some exhibit subtle mottled textures in BSE.

Both the large and the small pyrite grains host 1- to 25-µm hydrothermal arsenian pyrite overgrowths that are visible in reflected light (Fig. 4a, c). The rims are brighter in BSE than the pyrite cores, but BSE imaging does not reveal internal zonation within the rims (Fig. 4b, d, e, g). The rims are porous with fuzzy tendrils extending into the host-rock matrix. The thickest rims surround the largest euhedral magmatic-hydrothermal precursor grains. The samples also contain 3 to 5% fine-grained (<5 µm), single-stage hydrothermal arsenian pyrite that is not associated with precursor pyrite grains. This phase can be disseminated or disproportionally concentrated in clusters near larger pyrite grains and along preexisting fractures.

The NanoSIMS trace element maps of the Deep Star hydrothermal arsenian pyrite rims show that Cu, As, Ag, Sb, and Au are elevated relative to the precursor magmatic-hydrothermal pyrite grain cores, and the rims are internally heterogeneous in Cu, As, Ag, Sb, and Au content (Fig. 4f, h-p). The zoning in the rims is primarily concentric. At the sharp contact with the precursor pyrite grains, some rims have a 1- to 2-µm zone (apparent thickness) that is enriched in Au and Cu and depleted in As and Ag relative to the rest of the rim (Fig. 4h, p). There is also some sectoral zoning, with patchy enrichment of As, Ag, and Sb preferentially in one region of a rim rather than equal distribution throughout a concentric zone. Quantitative NanoSIMS spot data show that the arsenian pyrite rims contain up to 1,960 ppm Au, 495 ppm Ag, 270 ppm Cu, and 3,770 ppm As. The NanoSIMS depth profiles through the rims allow investigation of intrarim composition in the z-direction (Fig. 3e-h). Although Cu, As, Ag, and Au are present together in the same zones, the relative concentrations of these elements vary. The correlations between Au and Ag are moderately positive in some hydrothermal rims and absent in others (correlation coefficients of 0.6 to –0.3). The Au and As are also inconsistently correlated (–0.7 to 0.9), as are Ag and As (–0.5 to 0.7) and Au and Cu (–0.7 to 0.9). Correlation between Cu and As is strongly positive (1.0), whereas Cu and Ag show moderate negative correlation (–0.4 to –0.5). One depth profile analysis commenced in the grain core and penetrated into the underlying rim, capturing the contact between the precursor Jurassic pyrite and the Carlin-type hydrothermal arsenian pyrite rim. This analysis shows the sharp decline in Ag at the contact from core to rim, coincident with the sharp increase in Au and Cu at the contact (Fig. 3f). The δ³⁴S compositions of the rims are fairly similar to the pyrite grain cores, although within the rims the δ³⁴S values are slightly lower where Au concentrations are higher.

The Betze Post finely porphyritic rhyolite contains less than 0.5% modal abundance of pyrite ranging from 20 to 150 µm in diameter (Fig. 5a-h). The pyrite is subhedral to anhedral with sharp grain margins at host rock contacts. Most individual grains are rounded. Some grains have portions of the margin that are rounded and portions that are straight with angular corners. Some grains have relatively clean interiors with few pits, and others are deeply pitted and fractured. The grains do not have visible rims in reflected light or BSE, and none have detectable As in EDS analyses. The NanoSIMS spot analyses show <1 ppm Au, As, and Ag (Table 1). No correlation was observed among these elements. The NanoSIMS depth profiles indicate compositional homogeneity with only minor variation in Ag (Fig. 3i), so we did not collect NanoSIMS maps. The δ³⁴S compositions of the Betze Post dike pyrites are 12.6 to 15.2‰ (App. 1).

The Deep Star aphyric rhyolite contains less than 0.5% modal abundance of pyrite. Some grains are lath shaped, ranging from 50 to 80 µm on the short axis and 125 to 250 µm on the long axis (Fig. 5g-k). The grain boundaries are sharp at the contact with the host rock. The grains host numerous fractures, and larger grains are typically accompanied by clusters of broken grain fragments. The grains do not have visible rims in reflected light or BSE, and As is not detectable in EDS analyses. However, NanoSIMS spot analyses show that the grains contain minor trace elements, up to 1.2 ppm Au, 33 ppm Ag, and 89 ppm As (Table 1). The NanoSIMS depth profiles indicate that the grains are chemically zoned (Fig. 3j). The correlations between Au and As are strongly positive (0.7–0.9). Although there is an Ag peak in the same location as the Au and As peak, there is also a zone of Ag enrichment where Au and As are low (Fig. 3j), leading to no correlation between Ag and these elements when the entire depth profile is factored into the correlation coefficient. The δ³⁴S compositions of the Deep Star aphyric rhyolite pyrites are 12.5 to 13.3‰ (App. 1).

The Beast coarsely porphyritic rhyolite contains 2 to 10% modal abundance of cubic pyrite and lath-shaped pyrite, preferentially concentrated in some clasts of the volcanic rock (Fig. 6). The lath-shaped grains may have originally been marcasite and are up to 50 µm on the short axis and 300 µm on the long axis. The cubic grains are typically 50 µm in diameter, although some clasts contain 1 to 3% modal abundance of disseminated <20-µm cubic pyrite. The pyrite grains show subtle variations in BSE in regularly shaped zones that may correspond to twinning or grain boundaries (Fig. 6G-I). The NanoSIMS spot analyses indicate that the Au concentrations of these grain cores are below detection. The Ag is up to 2 ppm, and the As is up to 23 ppm (Table 1), and there is no correlation among these elements. The δ³⁴S compositions of the Beast porphyritic rhyolite pyrite grain cores are 7.9 to 8.2‰ (App. 1).

Most of the lath-shaped grains from Beast host very bright BSE arsenian pyrite rims that are up to 50 µm thick (Fig. 6e-k). These rims are porous, with fuzzy outer margins that form tendrils into the host rock, although the internal contacts with the lath-shaped pyrite grain cores are sharp and straight. Portions of the grains lacking these thick rims have thin, concentrically zoned 1- to 5-µm pyrite rims that are marginally brighter in BSE. The thick rims are rare on the cubic pyrite grains, but the narrower rims are universally present, and although the outer margins appear soft or diffuse, they lack visible tendrils. The thick rims are bright in BSE and contain up to ~25 wt % As in EDS analyses distributed in patchy sectoral zones, whereas the thinner, concentrically zoned rims have a maximum As concentration of 3 wt % determined by EDS (true arsenopyrite is 46 wt % As). The NanoSIMS spot analyses, indicate that the thicker rims contain up to 18 ppm Au, 2,050 ppm Ag, and 9,060 ppm As (Table 1). During some of the spot analyses the As and Ag detectors were malfunctioning, but the available data show correlation in peak positions for As, Ag, and Au in the depth profiles, suggesting that these elements are enriched in the same zones. In the available data, there is strong positive correlation between Au and Ag (0.7), Au and As (0.8), and Ag and As (0.7). The rims host numerous zones of Au enrichment that vary inversely with δ³⁴S composition (Fig. 3k, l).

Our Turquoise Ridge samples are friable, carbonaceous, orpiment- and realgar-bearing, decalcified limey mudstones of the Ordovician Comus Formation, similar to those described in Cail and Cline (2001). The rocks contain 1 to 5% modal abundance of pyrite and arsenian pyrite, disseminated and in clusters in a matrix of host rock, orpiment, and realgar. The sedimentary pyrite grains are rounded or anhedral (Fig. 7a-d), although euhedral to subhedral grains have been observed at Turquoise Ridge by Gopon et al. (2019). The sedimentary grain cores are commonly porous or sieve textured, and some are fractured. They host arsenian pyrite rims as well as texturally complex partial replacements of arsenian pyrite. The sedimentary pyrite is trace element poor, as seen in Turquoise Ridge pyrite BSE images by Gopon et al. (2019) and qualitative unstandardized NanoSIMS mapping by Barker et al. (2009). Unaltered sedimentary pyrite is limited in our samples, so we only obtained one quantitative NanoSIMS spot in this pyrite type from Turquoise Ridge, at 2 ppm Au, 0.4 ppm Ag, 1 ppm Cu, and 13 ppm As. The correlations among Au, Ag, Cu, and As are all strongly positive (0.8–1.0). The δ³⁴S composition of the sedimentary pyrite grains varies widely from 22.0 to –4.8‰ (Fig. 8a-d; App. 1).

The hydrothermal arsenian pyrite occurs as partial replacements and overgrowths on sedimentary pyrite as described above, and most commonly as texturally complex multistage hydrothermal pyrite without identifiable sedimentary grain cores (Fig. 7a-c). These multistage grains are irregular in shape, with a smooth, nonporous internal texture containing up to 20-µm-thick dark and light zones visible in BSE imaging (Fig. 7e, g). The zoning is primarily concentric, with layers of oscillating trace element content. There are also some sectoral zones in which the trace element distribution is patchy. NanoSIMS maps of these multistage grains show that all of the zones contain more Cu, As, Ag, Sb, and Au than the sedimentary pyrites, and the bright BSE zones contain higher concentrations (Fig. 7f, h-p).

All of the above-described types of pyrite host late-stage arsenian pyrite overgrowths that range from <1 to 15 µm thick at the outermost margins of the grains (Fig. 7f, h-p). These overgrowths are hardly visible in reflected light but appear as the brightest phase in BSE imaging. Their external margins are fuzzy, with tendrils extending into the host rock. The NanoSIMS maps show that the highest concentrations of Au and As occur in these overgrowths, regardless of whether they occur on sedimentary, partially replaced, or multistage hydrothermal grains.

It is difficult to distinguish between the multistage arsenian pyrite and the outermost arsenian overgrowths in NanoSIMS spot data, so quantitative results represent mixed analyses of the various hydrothermal arsenian pyrite stages. The hydrothermal ore-stage pyrite contains up to 1,470 ppm Au, 863 ppm Ag, 2,610 ppm Cu, and 17,300 ppm As (Table 1). Arsenic is strongly correlated with Au (0.7–0.9 in most analyses, 0.5 in one analysis) and Cu (0.9–1.0). The other elements show inconsistent moderate correlations or no correlation. For example, zones that contain high Au also have identifiable Cu, As, and Ag peaks. However, the peak heights vary; the highest Au peaks do not correspond to the highest Cu, As, or Ag peaks and vice versa, indicating that trace element ratios are inconsistent between zones of the hydrothermal pyrite. The δ³⁴S compositions of the hydrothermal pyrite covary with Au (Fig. 8a-d). In grains with isotopically light sedimentary cores, the δ³⁴S compositions of the hydrothermal pyrite are higher at high Au concentrations, whereas grains with high-δ³⁴S sedimentary cores have hydrothermal pyrite with lower δ³⁴S at high Au concentrations (Fig. 8a-d; App. 1).

Our Getchell samples are carbonaceous, orpiment- and realgar-bearing, decalcified limey mudstones of the Ordovician Comus Formation from the hanging wall of the Getchell fault (Cail and Cline, 2001). Samples from the Getchell mine contain 1 to 3% modal abundance of disseminated sedimentary pyrite grains that range from 10 to 100 µm in diameter (Fig. 9). The pyrites are euhedral to subhedral with minor porosity. Reflected-light and BSE imaging reveal no internal compositional variation within the sedimentary grain cores. Given the small size of the grains, we did not collect NanoSIMS maps, since the mapping can hamper the ability to conduct spot analyses. The NanoSIMS spot data indicate that the sedimentary pyrite contains <1 ppm Au and As and 5 ppm Ag (Table 1). There are moderate positive correlations between Cu and As, Ag, and Au, although there are no correlations among the other elements. The δ³⁴S compositions of the sedimentary pyrite range from 29.6 to 43.6‰ (App. 1).

The sedimentary pyrite grains in the Getchell ore samples are ubiquitously coated by porous, Au-bearing arsenian pyrite rims of irregular thickness ranging from 1 to 5 µm in apparent width (Fig. 9). Quantitative NanoSIMS spot analyses indicate that the rims contain up to 183 ppm Au, 1,460 ppm Ag, 772 ppm Cu, and 7,300 ppm As. Although internal variation is not visible in reflected-light or BSE imaging, the NanoSIMS depth profiles show that the rims are zoned with highly varied trace element content, mostly in concentric zones with some patchy sector zones (Fig. 8e-h). Within the depth profiles, the trace elements Cu, As, Ag, and Au show peaks at the same positions. All of these elements show strongly positive correlations with one another (0.7–1.0), although one analysis had a slightly lower correlation between Au and Cu and between Au and As (0.6). Similar to Deep Star, Carlin, and Turquoise Ridge, each trace element depth profile has different relative peak heights, indicating that the trace element ratios vary between zones. The δ³⁴S compositions of the hydrothermal pyrite are lower than the sedimentary pyrite grain cores and vary inversely with Au concentration within each hydrothermal rim (Fig. 8e-h).

A variety of gold mineralization styles occur at the Lone Tree deposit including quartz-sulfide ore, banded pyrite breccias, and oxide ore (Holley et al., 2019). The present study focused on samples of quartz-sulfide ore in quartzite of the Valmy Formation, characterized by 10- to 200-µm-diameter pyrite or marcasite and rare arsenopyrite intergrown with fine-grained quartz and sericite (Figs. 10, 11). We also examined sulfide minerals intergrown with sericite that replaces feldspars in phenocrysts of the rhyolite dikes. The sulfide minerals are mostly disseminated in the matrix, although some occur as aggregates or stringers of grains (e.g., the main cluster of grains in Figs. 10, 11). Sulfide mineral textures range from subhedral to ratty with semiporous grain cores, most of which host arsenopyrite rims. None of the grain cores show internal variation in reflected-light or BSE imaging, but EPMA (Holley et al., 2019) and NanoSIMS maps document the presence of trace Cu, As, Ag, Sb, and Au in the grain cores (Figs. 10, 11). New quantitative NanoSIMS spot analyses identified up to 2.3 ppm Au, 25 ppm Ag, 21 ppm Cu, and 60 ppm As in the grain cores. The only significant correlation in the pyrite grain cores was between Cu and As (0.9–1.0), and all other elements showed no correlation with one another. The host rock matrix also contains rare disseminated unrimmed, coarse pyrite and arsenopyrite that can reach up to 2 mm in diameter and contain no Au (Holley et al., 2019). The δ³⁴S compositions of the pyrite grain cores average 4.5‰ (App. 1).

Most of the precursor sulfide grains are overgrown by arsenopyrite, or less commonly by arsenian pyrite. The rims range from 1 to 25 µm wide, incompletely enclosing the grain cores (Figs. 10, 11). Their inner margins are sharp at the contact with the precursor pyrite or arsenopyrite, and their outer margins are fuzzy, with tendrils extending into the host rock matrix. The internal texture of the rims is semiporous, similar to those seen in classic Carlin-type gold deposits. Although the rims do not display internal zoning in reflected-light or BSE imaging, NanoSIMS maps reveal concentric zones of oscillating trace element concentrations, following the morphology of the outer rim margins (Holley et al., 2019). There are also some sectors of patchy trace element distribution within the rims. The rims contain up to ~2,100 ppm Au and up to several hundred ppm Cu, Sb, and Tl as determined by EPMA (Holley et al., 2019). The NanoSIMS maps show that the highest Au is concentrated in distinct zones that are only a few hundred nanometers across (Fig. 10t). The NanoSIMS spots in the rims gave Au values up to 2,020 ppm Au, 9,800 ppm Ag, 2,000 ppm Cu, and 70,100 ppm As (Table 1). The NanoSIMS depth profiles (Fig. 12a-d) show that Cu and As are positively correlated (0.5–1.0), whereas Au and As are strongly correlated in some of the rims and not correlated in others (1.0 to –0.3). Moderate correlation was observed between Ag and As in one analysis (0.5) and strong correlation between Au and Cu in one analysis (0.9), although most analyses showed no correlation among elements. The δ³⁴S compositions of the hydrothermal rims average 10.9‰ and do not show internal covariation with Au concentration (Fig. 12a-d; App. 1).

The Red Dot deposit at Marigold hosts pyrite disseminated in quartzite host rock of the Valmy Formation, as well as in millimeter- to centimeter-scale hydrothermal quartz veinlets that resemble D and E veinlets in porphyry systems (Huff et al., in press). Pyrite is the primary sulfide mineral in the veinlets, where it occurs as finely disseminated grains and aggregates along the contact with the wall rock (Fig. 13). This pyrite can be intergrown with covellite, sphalerite, and tennantite (Fig. 13a). Pyrite also occurs as discontinuous veinlets unassociated with the quartz and as coarse isolated grains disseminated in the host rock. All of these pyrites serve as precursor grain cores for arsenian pyrite rims, which are described separately in the subsequent paragraph. The veinlets host pyrite grains up to 2 mm in diameter that are subhedral and heavily fractured with sieve textures and elongated holes. The coarse isolated pyrite grains are disseminated within the host quartzite matrix and not associated with any fractures or veinlets. They are subhedral to euhedral and reach up to 2 mm in diameter. The fine-grained disseminated pyrite grains are much smaller than other textural varieties, ranging from submicron-scale up to 50 µm across. These fine-grained pyrites occur as euhedral hexagonal isolated grains, loose clusters of grains, and closely packed aggregates of grains concentrated at the contacts between wall rock and quartz veinlets. Although none of the precursor pyrite types are visibly zoned in reflected-light, BSE, or NanoSIMS maps, these pyrite types contain up to 6.1% As (average 1.2 wt %), 1,700 ppm Cu, 3,210 ppm Hg, and 1,080 ppm Ni (EPMA data from Huff et al., in press). The fine-grained disseminated grains and aggregates in the quartz veinlets contain up to 3 ppm Ag and 38 ppm Au. The precursor pyrite grain cores have δ³⁴S values ranging from 7.9 to 9.6‰ (Fig. 12e-g). These grain cores have strongly negative correlations between Cu and As (–0.7) and Cu and Ag (–0.9); strongly positive correlation between Ag and As (0.8–1.0); and moderate positive correlation between Au and As (0.4).

At Red Dot, hydrothermal arsenian pyrite coats all of the types of earlier pyrite described above. These rims occur most prominently on the small, disseminated pyrite and pyrite aggregates and on larger pyrite within the discontinuous pyrite veinlets. In some locations the precursor pyrite grains with sharp grain boundaries occur in a matrix of spongy, pitted pyrite with ~1 wt % As (Fig. 13a, d, j). Hydrothermal pyrite with higher As concentrations and crisp grain boundaries overgrows the spongy pyrite and also directly coats the precursor pyrite grains where the spongy pyrite is absent. The hydrothermal pyrite rims range from 1 to 25 µm thick. In BSE and NanoSIMS maps, multiple stages of concentrically zoned hydrothermal arsenian pyrite are visible within the rims (Fig. 13). The first generation is a thin (1- to 3-µm) layer at the contact with the precursor pyrite. This generation appears light gray in BSE images (Fig. 13m), present in most but not all rims. The NanoSIMS maps indicate that these innermost rims contain elevated Au, Ag, As, and Sb, with NanoSIMS spot data up to 944 ppm Au; 1,320 ppm Ag; 26,700 ppm Cu; and 6,570 ppm As (Table 1). The early arsenian pyrite is encapsulated by another ~5-µm-thick layer of arsenian pyrite that appears medium gray in BSE images (Fig. 13m). The NanoSIMS maps show subtle oscillations between low and moderate trace element concentrations within this zone. Individual oscillations are too small to resolve from NanoSIMS spot data, but the entire zone contains up to 753 ppm Au and 413 ppm Ag (Table 1). The outermost growth zone of arsenian pyrite is 2 to 5 µm thick and appears light gray in BSE images. The exterior contacts with the host rock are fuzzy and diffuse (Fig. 13). The NanoSIMS maps show that this outer layer contains the highest concentrations of Cu, As, Ag, Sb, and Au of all the arsenian pyrite generations (Fig. 13). A submicron layer of supergene covellite coats the edges of most Red Dot pyrite grains and does not appear to be related to the preceding ore-stage hydrothermal arsenian pyrite growth.

The NanoSIMS depth profiles through the Red Dot pyrites show a sharp contact between the weakly mineralized cores and the strongly mineralized rims, and each of the above-described layers of arsenian pyrite contains countless nanoscale compositional zones (Fig. 12e-g). The contacts between cores and the innermost rims are marked by a prominent double peak in Au concentration (Fig. 12e-g). The peak positions typically correlate for Cu, As, Ag, and Ag, showing that some zones are consistently enriched in trace elements relative to other zones. However, the middle and outermost layers of arsenian pyrite have some zones in which Ag and Au are inversely correlated, and other zones in which Cu and As are inversely correlated (Fig. 12g-h). Most analyses showed a strong positive correlation between Ag and As (0.7–0.9), although one analysis showed no correlation. Correlations among the other elements are inconsistent between depth profiles in our data set, likely since most depth profiles transected multiple types of hydrothermal pyrite (Table 2). The δ³⁴S compositions of the pyrite show a sharp change at the contact between cores and rims. The rims are several per mil lower than the cores, and within the rims the δ³⁴S compositions vary inversely with Au concentrations (Fig. 12g-h).

An overview bright-field TEM image of the Red Dot TEM specimen is shown in Figure 14a. The interface between the pyrite grain and the host rock matrix is easily observable. Based upon the NanoSIMS results, Au and As concentrations are highest at the outer rim of the pyrite, and we focused our TEM investigations on that region of the pyrite. Dark field imaging of the outer pyrite rim contains abundant ~10-nm nanoparticles (Fig. 14b). The STEM-EDS spectral images (Fig. 14c-f) confirm the S, Fe, As, and Au distributions observed in the NanoSIMS maps high-resolution TEM imaging of the nanoparticles (Fig. 14g) were completed on the <321> pyrite zone axis, as seen in the indexed selected area diffraction pattern (Fig. 14i). The nanoparticles are observed to be darker in contrast compared with the pyrite, potentially implying a higher atomic number. A Fourier transform of the high-resolution TEM image (Fig. 14h) indicates a change in lattice parameter between the (111) pyrite planes and the nanoparticles. Measuring the change in radius of the rings in the Fourier transform, the nanoparticles are found to have a lattice parameter 1.1 times larger than the pyrite, approximately 0.5 nm for a cubic lattice, likely indicative of Au.

All of the pyrite types examined in this study contain at least minor Au, Ag, As, and Cu, as well as Sb where analyzed (Fig. 15), although there are key differences in concentrations, trace element correlations (Tables 1, 2), and δ³⁴S compositions (Figs. 3, 8, 12). Geochemical zoning occurs at the nanoscale in the hydrothermal Carlin-type pyrite, distal disseminated hydrothermal arsenopyrite at Lone Tree, and hydrothermal pyrite at Red Dot. The number, sequence, and thickness of the zones are inconsistent throughout the data set, even between grains from the same hand sample. For example, the total number, position, and width of the peaks vary among the four analyses from Turquoise Ridge (Fig. 8a-d) or the four analyses from Carlin (Fig. 3a-d), and growth zones cannot be correlated across these grains. Our high spatial-resolution data set makes it clear that it is an oversimplification to correlate growth zones from multiple pyrite grains across a deposit, and that individual pyrite growth zones cannot be easily attributed to deposit-wide events (cf. Arehart et al., 1993; Simon et al., 1999a; Barker et al., 2009; Large et al., 2009; de Almeida et al., 2010). The following discussion demonstrates the strengths and limitations of using nanoscale trace element and isotopic depth profiles to evaluate the potential source, transport, and pyrite growth mechanisms of the studied pyrite types.

We suggest that Carlin-type Au may have come from an Eocene magmatic source, based on the observed relationships between Au and δ³⁴S in our depth profiles. Within individual hydrothermal pyrite rims, Au and δ³⁴S covary at the nanoscale (e.g., Figs. 3, 8, 12). Hydrothermal pyrite zones with low Au concentrations have δ³⁴S values similar to the pyrite grain core, whereas zones with the highest Au have δ³⁴S values that could have been derived from Eocene magmatic fluids. The mean δ³⁴S of Great Basin granitoid magmas was 7.1‰ in the Tertiary (Arehart et al., 2013). The fluids that produced Eocene Great Basin porphyry and skarn pyrite had δ³⁴S values from –1.8 to 7‰ (Theodore et al., 1986; King, 2017; Holley et al., 2022). Fluid-mineral δ³⁴S fractionation is affected by pH, temperature, and oxygen fugacity to varying degrees. Within the pyrite stability field, changes in pH have relatively minor effects on δ³⁴S (Ohmoto, 1972). Sulfur isotopic fractionation between H₂S and pyrite increases at lower temperatures, although the differences are relatively minor. As an example, pyrite that forms at 400°C will be 0.9‰ higher than the mineralizing fluid, and pyrite that forms at 200°C will be 1.8‰ higher than the fluid (Ohmoto and Rye, 1979). Based on the simplest scenario that only accounts for temperature-related fractionation, the Eocene magmatic-hydrothermal fluids that produced porphyry and skarn sulfide minerals in northern Nevada would precipitate pyrite with δ³⁴S values from 0 to 8.8‰ at 200°C (pink shaded region in Figs. 3, 8, 12). Changes in H₂S/SO₂ ratios in magmatic gases can also lead to a significant effect on δ³⁴S. For example, a 20% increase in H₂S/SO₂ leads to pyrite that is 6‰ higher (Rye and Ohmoto, 1974). Due to the combined effects of temperature and oxygen fugacity, the Eocene magmatic-hydrothermal fluids in the region would precipitate Carlin-type pyrites in carbonaceous rocks in the δ³⁴S range of 0 to 8.8‰ or slightly higher.

We interpret the δ³⁴S compositions of our hydrothermal pyrite rims as the product of mixing between sulfur in that individual grain core and sulfur in Eocene magmatic-hydrothermal fluids. We discuss how mixing influences the δ³⁴S of hydrothermal pyrite overgrowing high-δ³⁴S sedimentary pyrite, low-δ³⁴S sedimentary pyrite, and magmatic pyrite with moderate δ³⁴S. In the following paragraphs, we also assess the relationships between δ³⁴S and Au.

The sedimentary pyrite grain cores that we examined typically have high δ³⁴S, up to 54.4‰ (e.g., Figs. 3a, 8a). These values are consistent with formation during diagenesis of carbonaceous, iron-poor sediments beneath a marine basin that may have been closed. The Carlin-type hydrothermal overgrowths on these grains have lower δ³⁴S, which we attribute to the input of magmatic-hydrothermal fluids. The depth profiles show a negative correlation between Au and δ³⁴S: –0.9 at Turquoise Ridge (grain in Fig. 8a), –0.7 to –0.9 at Getchell (grains in Figs. 8e-h), and –0.6 at Carlin (grain in Fig. 3a). The lowest δ³⁴S (most magmatic) values occur in the zones of greatest Au enrichment (Figs. 3b-c 8a).

Several grains at Turquoise Ridge and Carlin have sedimentary pyrite cores with δ³⁴S values ranging from –1.3 to –33.5‰ (Figs. 3b-c, 8b-d). These values are typical for diagenetic pyrite in sediments corresponding chemically to average marine shale. The hydrothermal rims on these grains contain higher δ³⁴S than the sedimentary grain cores, which we attribute to the input of magmatic-hydrothermal fluids. At Carlin where the grain cores contain low δ³⁴S (Fig. 3b-c), the depth profiles show a positive correlation between δ³⁴S and Au (0.7–0.8). The highest δ³⁴S occurs in regions that are enriched in Au.

There is no correlation between δ³⁴S and Au in grains where the δ³⁴S composition of the grain core overlaps with that of the hydrothermal rim, as in some Turquoise Ridge analyses (Fig. 8b-d) or at Deep Star and Beast where grain cores and rims both plot within the δ³⁴S range of magmatic-hydrothermal pyrite (Fig. 3f-h, k-l). In those cases, quantification of the relationship between δ³⁴S and Au requires a more sophisticated statistical approach than can be achieved by visual assessment of the NanoSIMS depth profiles.

The relationships between δ³⁴S and Au have been examined statistically by Holley and Phillips (2022), who used Bayesian isotopic mixing models to study the source contributions to Carlin-type hydrothermal pyrite. The modeling treated the hydrothermal pyrite as a mixture with several possible sources of sulfur that may have contributed to the “mixture,” including Eocene magmas, Jurassic and Cretaceous magmas, local sedimentary pyrite near the analyzed samples, and sedimentary pyrite from important stratigraphic units hosting Carlin-type deposits. The modeling showed that all of these sources likely contributed to Carlin-type pyrite, underscoring the importance of fluid mixing during mineralization. The Au concentration of the pyrite covaried with the contribution of sulfur from Eocene magmas: analyses with higher Au concentrations had higher proportional source contributions from Eocene magmas. In the present study, we attribute the covariation of Au and δ³⁴S to episodic mixing as modeled by Holley and Phillips (2022). The Au-rich magmatic-hydrothermal fluids likely encountered Au-poor meteoric fluids with δ³⁴S values close to that of the host rock, resulting in oscillatory zoning of high-Au zones with magmatic-like δ³⁴S, and low-Au zones with host rock-like δ³⁴S. Although mixing likely occurred at multiple scales, a component of very local mixing is evidenced by neighboring grains in a single sample that display different nanoscale patterns of δ³⁴S and Au (e.g., Fig. 3a-d).

Proponents of a meteoric origin for Carlin-type gold have suggested that deeply circulating meteoric fluids may have become enriched in Au by encountering metal-rich diagenetic pyrite (Large et al., 2011). Organic-rich muds on the floor of ocean basins contain Au, As, Ag, Cu, and other metals in organometallic complexes, which could have been incorporated into diagenetic pyrite (Large et al., 2011). We infer that the meteoric fluids did not carry much of the Au that precipitated in Carlin-type pyrite, since deeply circulating meteoric fluids would achieve δ³⁴S signatures averaging the δ³⁴S compositions of the stratigraphy rather than the observed covariation of δ³⁴S and Au. If the As, Cu, and Ag were carried in the meteoric fluid, we would expect these elements to display the inverse of the relationship observed between Au and δ³⁴S (e.g., As, Ag, and Cu enrichment in zones of amagmatic δ³⁴S). Although As, Ag, and Cu are commonly present in the Carlin-type pyrite, we did not observe consistent trends with δ³⁴S.

Magmatic-hydrothermal processes such as partitioning during phase separation and differential transport by ligands have been invoked to explain the relative enrichments of Au, As, Ag, and Cu in Carlin-type pyrite. Muntean et al. (2011) suggested that Au, As, and Cu would ascend in the vapor phase, whereas the Ag would partition into the hypersaline liquid. This interpretation was based on faulty fluid inclusion data which suggested that Cu would partition into the vapor phase alongside Au (cf. Seo and Heinrich, 2013), leading Muntean et al. (2011) to invoke monosulfide solid solution at greater crustal depths to sequester Cu and explain the low Cu/Au ratios in Carlin-type pyrite. Experimental studies by Hurtig and Williams-Jones (2014) and numerical simulations by Hurtig et al. (2021) suggest that vapor-like magmatic-hydrothermal fluids can carry sufficient concentrations of Cu and Ag to form ore deposits. The low Cu/Au and Ag/Au ratios in Carlin-type deposits may instead reflect the preferential transport of Au to shallow environments based on its temperature-dependent solubility (e.g., Murakami et al., 2010). Gold reaches maximum solubility at 340° to 510°C, whereas Cu and Ag are most soluble at higher temperatures (Hurtig et al., 2021). Furthermore, at the relatively low temperatures of Carlin-type mineralization, Au travels as a bisulfide complex, whereas Cu and Ag are transported by the Cl₂^– ligand (Pokrovski et al., 2014). We observed that As, Cu, and Ag are commonly present with Au in the Carlin-type pyrite, but the correlations among these elements are inconsistent, suggesting that pyrite growth mechanisms may also play a role (Table 2).

Numerous studies have considered the potential mechanisms of Carlin-type pyrite growth in efforts to explain the trace and minor element compositions and distributions. In the following paragraphs we discuss our observations in the context of proposed models and evaluate the extent to which our data constrain the possibilities.

The residence of Au as Au⁺¹ or Au⁽⁰⁾ can reveal information about the processes of pyrite growth and Au incorporation. Reich et al. (2005) inferred that most Carlin-type gold occurs in solid solution as Au+1, based on annular dark-field and high-resolution TEM images showing an absence of nanoparticles and a distortion of the pyrite structure. Gopon et al. (2019) identified Au⁺¹ at Turquoise Ridge using atom probe tomography. Simon et al. (1999b) used X-ray absorption near edge structure-extended X-ray absorption fine-structure spectroscopy (XANES-EXAFS) to identify nanoparticles of Au⁽⁰⁾ and Au⁺¹ at Twin Creeks, noting that Au⁺¹ can be incorporated linearly with two sulfide ligands, or in fourfold coordination with Au⁺¹ in a vacancy position, as an unknown Au-As-S compound, or in a distorted Fe²⁺ octahedral site. Although we were not able to obtain TEM data from the Carlin-type hydrothermal pyrite, we can make inferences based on the compositions measured during NanoSIMS spot analyses using comparisons to published data. Reich et al. (2005) determined the solubility limit for solid solution of Au in arsenian pyrite as a function of the As content: arsenian pyrite with Au/As molar ratios above about 0.2 will host gold as native Au nanoparticles; below this limit, gold occurs in solid solution as Au⁺¹. Our depth profile data give new insights on the residence of Au by transecting numerous zones across the hydrothermal pyrite. The majority of the Au/As molar ratios from our depth profiles plot below the solubility limit, indicative of Au in solid solution, but about a quarter of the data points plot above the limit and are likely nanoparticles of Au⁽⁰⁾ (Fig. 15). Several depth profile analyses cross the solubility limit line, indicating that gold can occur in solid solution and as nanoparticles in different zones of the same grain. The presence of Au as both Au⁺¹ and Au⁽⁰⁾ could indicate that the saturation state of Au in the fluids varied during pyrite growth. Nanoparticles are thought to be more common at higher temperatures since Au becomes increasingly soluble within pyrite as temperature declines (Reich et al., 2005; Steadman et al., 2021), perhaps providing some indication of temperature change during hydrothermal pyrite growth. Alternatively, the nanoparticles may not be primary.

If the Au nanoparticles are primary, they could indicate electrochemical precipitation of Au. Variation in the trace element content of the pyrite can lead to electric potential differences that allow for electron transfer and reduction of Au⁺¹ to Au⁽⁰⁾ (Gaboury and Sanchez, 2020; Wang et al., 2021). Pyrites containing As are P-type semiconductors, whereas those containing Cu are N-type semiconductors. A mix of N- and P-type pyrites leads to an electrical potential difference which can cause precipitation of native Au in pyrite. When Cu was synthetically added to As-bearing P-type pyrites, the pyrite changed to an N-type semiconductor (Wang et al., 2021). Our data show that the Carlin-type pyrite contains both Cu and As, and that these elements are commonly correlated, so we infer that the pyrite is mostly N-type. However, zones with low Cu may have enabled electron transfer and precipitation of Au⁽⁰⁾.

Experiments by Kusebauch et al. (2019) showed that the partitioning of Au into pyrite increases with As content of the pyrite during pyritization, a phenomenon that those authors used to explain the relatively constant pyrite Au/As ratios reported in other studies (e.g., Reich et al., 2005; Deditius et al., 2014). Kusebauch et al. (2019) inferred that Au deposition by sulfidation only occurs when the fluid becomes supersaturated in Au, although they validated their model based on observations of Carlin-like pyrite from Lannigou, China (Yan et al., 2018), in which the highest Au concentrations are in the earliest overgrowths, which is not consistent with our observations. Kusebauch et al. (2019) also suggest that in As-bearing systems, the reduction of Au could be prohibited by As^–1 substituting on the surface for S^–2, potentially explaining the presence of Au in solid solution. Those authors suggest that in the absence of As, the Au⁺¹ could be reduced to Au⁽⁰⁾. Alternatively, dissolution of arsenian pyrite could lead to locally reduced conditions at the mineral surface that would facilitate precipitation of Au¹⁺ complexes (Pokrovski et al., 2014). Complexation at the surface of the pyrite crystal (e.g., Chouinard et al., 2005a) is known to be an important control on mineralization in nonboiling fluids that are undersaturated with respect to gold, such as Carlin-type mineralizing fluids. Pokrovoski et al. (2021) suggested that Au-As redox reactions are the controlling mechanism for solid solution of Au in arsenian pyrite, with Au hosted in the octahedral Fe site as [AuAs_1-3S_5-3], [AuAs₃S₃…AuAs₆], and [AuAs₆].

Examination of the nanoscale depth profiles allows us to make several inferences on the role of As during incorporation of Au in Carlin-type pyrite. We observed a wider range of Au/As ratios than typically reported in Carlin-type pyrite (Fig. 16), as well as inconsistent correlations between As and Au (Table 2), likely due to the high spatial resolution of the data. The correlation between Au and As varied from strongly negative (–0.7) to no correlation (0.0) to strongly positive (1.0), even within grains from the same sample (e.g., Deep Star; Table 2). Detailed examination of the depth profiles shows that As is consistently present in zones of Au enrichment, even though the ratios vary. Therefore we cannot attribute the reduction of Au⁺¹ to Au⁽⁰⁾ to a lack of As (cf. Kusebauch et al., 2019). We also observe that Au is not present in all zones of As enrichment. The reactions described by Pokrovoski et al. (2021) can only take place if Au is available, further underscoring the idea that Au was inconsistently available during Carlin-type pyrite growth. Temporal variation in Au availability at a local scale may also explain why neighboring grains in the same deposit can have the highest Au enrichment in different zones (Fig. 4). Our Sb data are limited, but Sb can exist in the same oxidation states as As and may have played a similar role in the incorporation of Au into the pyrite.

There are several existing models for which we see no textural or geochemical evidence in our analyses of Carlin-type pyrite: solid-state refinement, recrystallization, secondary dissolution-reprecipitation, and crystallographic structural control. Solid-state refinement has been suggested as an alternative process that could lead to Au-As zoning, based on simulations by Xing et al. (2019) showing direct precipitation of the rims would require fluids with impractically high As (e.g., compositions that cannot be achieved by leaching or by the metamorphic conversion of pyrite to pyrrhotite proposed by Large et al., 2011). Xing et al. (2019) do not explain exactly how this mechanism would produce the textures commonly observed in Carlin-type pyrite. Lattice-bound elements are thought to be less mobile than native metals during solid-state processes (Mumin et al., 1994; Large et al., 2009), and solid-state diffusion of Au in pyrite is not expected below temperatures of 370°C (Reich et al., 2006). Therefore, we assume that refinement did not play a significant role in creating the observed zonation in our Carlin-type pyrites. We also note that refinement is unlikely to have produced the observed correlation between Au and δ³⁴S. Fougerouse et al. (2016) studied the trace element characteristics of arsenopyrite from an orogenic gold deposit, finding textural and crystallographic evidence for gold remobilization by dissolution and reprecipitation. The reprecipitated arsenopyrite is Au poor and has sharp grain boundaries where it crosscuts the earlier oscillatory zoned arsenopyrite, and the remobilized Au is concentrated in microfractures. The studied Carlin-type pyrite shows no evidence for crosscutting phases. Instead, the NanoSIMS maps of the Carlin-type hydrothermal pyrite show that distribution of Au and the thickness of the Au-rich zones are relatively consistent on all sides of the grain and are mostly concentric rather than sectoral. The same is true for Ag, As, Cu, and Sb where measured. These textures suggest that the distribution of these elements was controlled primarily by elemental availability, as opposed to remobilization, surface enrichment, or crystallographic structural controls exerted by the hydrothermal pyrite (e.g., Watson and Liang, 1995; Watson, 1996; Chouinard et al., 2005b; Barker et al., 2009).

Eocene magmatic pyrite in the Beast, Deep Star, and Betze Post dikes is compositionally distinct from the Carlin-type hydrothermal overgrowths, containing only minor Ag, As, and Au. The pyrite in the grain cores may have originally formed as pyrrhotite at subsolidus temperatures during dike emplacement, since most felsic magmas have insufficient f_S2 to crystallize magmatic pyrite (Audétat et al., 2011). The aphyric rhyolite that we studied from Deep Star is widely distributed at Deep Star, Genesis, Richmond Mountain, and Welches Canyon, and a related extrusive body occurs at Welches Canyon. It is locally mineralized elsewhere on the northern Carlin trend. Dunbar (2001) and Heitt et al. (2003) used crosscutting relationships to suggest that mineralization was closely bracketed by two episodes of aphyric rhyolite dike emplacement at ~39.1 Ma. The Betze Post dike is thought to have been emplaced at 39.32 ± 0.11 Ma, prior to mineralization, and was likely sourced from a different pluton than the aphyric rhyolite based on composition (Ressel and Henry, 2006). At Beast, Carlin-type hydrothermal pyrite did overgrow the dike pyrite after the dike was emplaced at 37.58 ± 0.06 Ma (Ressel and Henry, 2006). Mineralization is thought to have been asynchronous in the region, younging to the southwest along with the magmatism (Cline et al., 2005; Ressel and Henry, 2006; John et al., 2008; Muntean et al., 2011). Our data point toward further temporal complexity, since the mineralization occurred as a later phase after formation of the sulfide grain cores in the dikes.

The pyrites at Lone Tree are distinct from the dike pyrites on the Carlin trend, and also distinct from Carlin-type hydrothermal overgrowths. The magmatic pyrite grain cores at Lone Tree have relatively high trace element concentrations (8.9 ppm Au, 25 ppm Ag) and internal zonation in Ag, Cu, and As (Fig. 12a, c, d), even in analyses with no Au (Fig. 12c). The Lone Tree magmatic pyrite grain core δ³⁴S values average 4.5‰. The hydrothermal arsenopyrite at Lone Tree is texturally similar to the overgrowths on the Beast dike pyrite, occurring along a sharp margin with the precursor pyrite and protruding into the host rock in a comb texture comprising arrays of relatively straight fuzzy tendrils. The As content is up to 7 wt % and the Ag concentrations are up to 2,022 ppm Ag at Lone Tree, in the range of values at Beast, but the Au and Cu at Lone Tree are much higher (Table 1). In contrast with the hydrothermal overgrowths at Beast and the other Carlin-type deposits, the hydrothermal arsenopyrite overgrowths at Lone Tree do not show nanoscale correlations between Au and δ³⁴S: the δ³⁴S depth profile trends are fairly flat even where Au concentrations fluctuate widely (Fig. 12a-b). The δ³⁴S signatures of the Lone Tree hydrothermal rims are higher than the Eocene magmatic grain cores and higher than the range of Eocene magmatic sulfur in the region, which led Holley et al. (2019) to infer that the fractionation between cores and rims was due to increased H₂S/SO₂ from wall-rock interaction of the magmatic-hydrothermal fluids in the distal disseminated environment. Sectoral zoning in the arsenopyrite rims may reflect crystallographic control on incorporation of As, Sb, and Ag (e.g., Watson and Liang, 1995; Watson, 1996; Chouinard et al., 2005b).

There are more textural varieties of hydrothermal pyrite overgrowths at Red Dot than at Lone Tree or any of the Carlin-type deposits examined in this study. The early porphyry-style pyrite may be Cretaceous or Eocene, since there are porphyry deposits of both ages in the district. The δ³⁴S signatures of these grains are similar to porphyry-related sulfide minerals at nearby deposits (Holley et al., 2022). The spongy, pitted pyrite is similar to spongy pyrites described in some Carlin-type deposits (e.g., Betze; Peters et al., 1998). At Red Dot, this spongy texture appears to have formed by hydrothermal overgrowth on porphyry or sedimentary pyrite grain cores, followed by dissolution or alteration of the hydrothermal overgrowth. The porphyry pyrite and the spongy pyrite are both overgrown by highly zoned multistage rims with crisp zone boundaries and exterior margins, similar to those described in the Cove South Deep zone of the Cove deposit in the Fish Creek mountains 25 km south of Red Dot (Bonner, 2019). At Cove, this pyrite type can be up to 1.5 wt % Ag. In our EPMA analyses from Red Dot, the high As analyses are accompanied by lower S concentrations, although there are some slightly lower Fe concentrations. Most atomic-scale studies indicate that As is most likely to substitute for S in arsenian pyrite (Reich et al., 2005; Blanchard et al., 2007), although it is possible that the As may also substitute for Fe at Red Dot based on the EPMA data (e.g., Chouinard et al., 2005b; Deditius et al., 2008). The δ³⁴S values of the multistage rims at Red Dot mostly range from 4.0 to 7.5‰, but Au-rich zones can be as low as 0‰. This range of values is lower than the porphyry pyrite cores, suggesting that the Au in the rims was derived from a fluid compositionally distinct from that which formed the porphyry pyrite. This fluid may have encountered sedimentary pyrite with low δ³⁴S, or it could have been sourced from a nearly contemporaneous magmatic event in a telescoped system, or a later magmatic event coincident with the Eocene hydrothermal fluid flow recorded by apatite fission track and (U-Th)/He ages of in the district (Huff et al., unpub. data, 2024).

Nanoscale zones of Carlin-type hydrothermal pyrite with high Au have δ³⁴S values matching the expected composition of hydrothermal pyrite from Eocene magmatic fluids. At lower Au concentrations, the δ³⁴S values in the hydrothermal pyrite are more similar to the host pyrite. These data lead us to suggest that Carlin-type gold came from Eocene magmas (e.g., Ressel and Henry, 2006; Muntean et al., 2011; Holley et al., 2022). Local mixing with meteoric fluids led to covariation in Au and δ³⁴S values in successive pyrite growth zones. The contained Au was iteratively upgraded as new Au-rich pyrite layers grew. This local-scale process resulted in the enormous Au endowment of the deposits, as well as local variation in Au, Ag, As, and Cu correlation coefficients and a lack of deposit-scale correlation in pyrite growth zones. Arsenic is present wherever Au is present, but the reverse is not true. The As likely assisted with incorporation of Au into the hydrothermal pyrite, perhaps by Au-As redox reactions at the surface of the growing pyrite (e.g., Pokrovski et al., 2014; Kusebauch et al., 2019) when Au was available during fluid mixing. While Au was being incorporated into the growing pyrite, variations in saturation state of Au in the fluid may have led to the occurrence of Au⁺¹ and Au⁽⁰⁾ in different zones of the same grain. Alternatively, variations in As and Cu content could have helped create electrical potential differences that allowed for reduction of Au⁺¹ to Au⁽⁰⁾.

Despite both being sourced from Eocene magmas, Carlin-type hydrothermal pyrite and Eocene dike pyrite are distinctly different, requiring different growth mechanisms or compositional evolution of the causative fluids between precipitation of the dike pyrite and the Carlin-type pyrite, potentially during ascent through the crust. Carlin-type hydrothermal pyrite shares many characteristics in common with distal disseminated hydrothermal arsenopyrite from Lone Tree, including textures and nanoscale trends in trace element zonation. However, the lack of nanoscale Au-δ³⁴S correlations at Lone Tree suggests that the metal sources or processes of fluid evolution in the distal disseminated environment are to some degree different from those that formed Carlin-type deposits. Hybrid characteristics are present in hydrothermal arsenian pyrite rims at Red Dot, suggesting that the Marigold deposits may represent the continuum between distal disseminated and Carlin-type mineralization.

Given the occurrence of finely zoned ore minerals in numerous other geologic settings such as porphyry, epithermal, intrusion-related, orogenic, and Carlin-like deposits (e.g., Fig. 17), investigations of nanoscale elemental and isotopic zonation are likely to reveal new insights on ore-forming processes. If zoned ore minerals are present, caution should be taken when applying genetic models that are unspecific to the observed zoning, or models that were developed based on data that are lower spatial resolution than the zoning. We recommend that future work should leverage nanoscale techniques to characterize trace element and isotopic zoning in a wider array of ore mineral and ore-forming systems, since the causative processes may be more complex than previously surmised. Rich information can be obtained by assessing NanoSIMS depth profiles both qualitatively and quantitatively. Our contribution serves as a roadmap for NanoSIMS studies of zoned minerals and opens an exciting new chapter in the study of ore deposits.

This work was funded by NSF Career Award EAR-1752756 (EAH). The Stanford Nano Shared Facilities are supported by NSF ECCS-2026822. We thank Jean Cline, Matt Fithian, Phillip Gopon, Al Hofstra, Dante Huff, Nicole Hurtig, Celeste Mercer, Mike Ressel, and Patrick Sack for discussion and samples. Jae Erickson, Kelsey Livingston, Katharina Pfaff, and Sage Langston-Stewart assisted with sample preparation and analyses at Colorado School of Mines, as did Heather Lowers at the U.S. Geological Survey and Aaron Bell at the University of Colorado Boulder. Prior collaborations with Craig Johnson and Michael Pribil at the U.S. Geological Survey helped frame the investigation. Reviews by Artur Deditius and Dennis Sugiono improved the manuscript.

Elizabeth Holley is an associate professor at Colorado School of Mines, specializing in mineral exploration and mining geology. This work was conducted under a National Science Foundation (NSF) Career Award on the role of magmatism in sedimentary-hosted gold mineralization. Dr. Holley also leads the NSF-funded Responsible Critical Minerals project and is the site director for the NSF Center to Advance the Science of Exploration to Reclamation in Mining. She is a fellow of the Payne Institute of Public Policy and a Fellow of the SEG.