Nanoscale isotopic evidence resolves origins of giant Carlin-type ore deposits

The Carlin-type gold deposits in the Great Basin (western North America; Fig. 1) are the largest accumulations of gold and the least understood gold deposit type in the Northern Hemisphere, inspiring questions about the processes governing metal cycling and mineralization in the Earth's crust. More gold is produced annually from these deposits than from any other site in the world (Harper, 2020). Carlin-type deposits also represent potential resources of “critical minerals”, including arsenic and antimony (Goldfarb et al., 2016). Carlin-type ore occurs as disseminated hydrothermal replacement bodies, primarily hosted in structures crosscutting decarbonated silty limestones. The gold exists in solid solution or as nanoparticles within micron- to nanometer-thick rims of hydrothermal arsenian pyrite overgrowing older sedimentary and magmatic-hydrothermal pyrite grains that were present in the host rocks prior to gold mineralization. Although deposits with similar characteristics occur elsewhere on the planet, it has been suggested that the enormous gold endowment in the Great Basin represents a nonreplicable combination of geologic processes (Cline et al., 2005). The source of Carlin-type gold has eluded definition, echoing the debate waged between neptunists and plutonists at the birth of modern geology: Did meteoric fluids scavenge and then redeposit gold as they circulated through the sedimentary rocks (Ilchik and Barton, 1997; Emsbo et al., 2003; Large et al., 2011), or was the gold introduced by magmas (Sillitoe and Bonham, 1990; Ressel and Henry, 2006; Muntean et al., 2011)?

Carlin-type deposits are notoriously difficult to study because their mineralogy is not amenable to the traditional tools used to fingerprint the origins of metal enrichment (Richards, 2011). Fine-scale zonation in sulfide mineral geochemistry is common in many deposit styles, and the spatial resolution of traditional analytical techniques has been insufficient to differentiate among Carlin-type pyrite generations. Hydrogen and oxygen isotopes of the clays and fluid inclusions at Carlin-type deposits represent mixtures between syn-ore and pre-ore phases, consistent with precipitation from a meteoric fluid or mixing between meteoric and magmatic fluids (Hofstra and Cline, 2000; Cline et al., 2005, and references therein). Carlin-type mineralization ages are imperfectly constrained due to the paucity of dateable syn-ore minerals, but the timing of mineralization appears to track the late Eocene southwestern sweep of calc-alkaline magmatism through the region (Cline et al., 2005). Several Carlin-type gold districts in the Great Basin show no evidence of Eocene magmatism (Fig. 1), and the sedimentary host rocks are unusually enriched in metals including Au and As (Emsbo et al., 2003; Large et al., 2009). Similar smaller deposits elsewhere in the world lack evidence for syn-mineralization magmatism (Cline, 2018; Pinet et al., 2020).

Because reduced sulfur served as the principal ligand during gold transport (Cline et al., 2005), sulfur isotopes have long been considered the unattainable key to determine the origins of Carlin-type gold (gold itself has only one stable isotope). The term “δ³⁴S” refers to the isotope ratio ³⁴S/³²S (‰) relative to Vienna Cañon Diablo troilite. The δ³⁴S values of ore pyrite can be compared to the δ³⁴S of potential sulfur reservoirs. Elemental analyzer–isotope ratio mass spectrometry (EA-IRMS) of whole grains gives δ³⁴S values that average the older pyrite and hydrothermal pyrite overgrowths; results are permissive of either magmatic or sedimentary origins (Cline et al., 2005; Christiansen et al., 2011). Traditional secondary ion mass spectrometry (SIMS) studies reached varying conclusions based on only a few data points from spot sizes of 10–30 μm encompassing multiple compositional zones in the pyrite. Data are suggestive of a magmatic origin at Getchell and Betze-Post (Nevada, USA; Cline et al., 2003; Kesler et al., 2003; Henkelman, 2004; Kesler et al., 2005). Relative differences in ³⁴S/³²S between cores and rims were determined for three grains from Turquoise Ridge and West Banshee using qualitative nanoscale SIMS (NanoSIMS) mapping (Barker et al., 2009) and atom probe tomography (Gopon et al., 2019), but the data were not standardized so the origins of the gold-bearing fluid remained elusive. We paired NanoSIMS depth profiles and laser ablation–multicollector–inductively coupled plasma mass spectrometry (LA-MC-ICPMS) to resolve the insufficient spatial resolution and the potential for matrix effects inherent in previous methods.

We examined ore pyrites in 40 samples from five well-studied Carlin-type deposits in Nevada—Carlin, Deep Star, Beast, Turquoise Ridge, and Getchell—as well as northern Carlin-trend Eocene dikes (Fig. 1; see the Supplemental Material¹). From petrography and scanning electron microscopy of thousands of pyrite grains, we selected 64 locations in representative grains for in situ sulfur isotopic and trace element study. We made NanoSIMS maps of the grains by collecting ⁶³Cu, ⁷⁵As, ¹⁰⁷Ag, ¹¹⁷Sb, and ¹⁹⁷Au data on electron multipliers and calibrated the data with relative sensitivity factors using an electron microprobe. Figure 2 shows representative examples of the target locations and NanoSIMS maps.

Standardized, quantitative analyses of sulfur isotopes in sulfide minerals were previously only possible at a spatial resolution of 1–15 μm (Zhang et al., 2014, 2017; Hauri et al., 2016). The methods ignored compositional heterogeneity in the Z-direction, averaging NanoSIMS data over the length of an entire analytical run. We improved the method to record nanoscale compositional variation by producing a depth profile at each of our 64 spots, gathering 2400 individual data points per analytical run as the beam penetrated through successive heterogeneous geochemical zones (see the Supplemental Material). Each depth-profile data point represents a depth interval of <1 nm. Figure 3 shows six representative depth profiles. To quantify trends in a manageably reduced number of data points, we calculated plateau averages from zones of consistent composition within each depth profile (see the Supplemental Material). In Figures 4A–4D and in Table S1 (in the Supplemental Material), we present data for 89 NanoSIMS plateaus. The average δ³⁴S ratio error for plateaus was ±0.86 (one standard deviation) (Table S1). We validated our NanoSIMS results from coarse areas using 5 μm LA-MC-ICPMS spots in 23 locations. The δ³⁴S data are consistent between the two methods (Table S2), indicating that our observed core-rim fractionations are not due to matrix effects.

The pre-mineralization sedimentary pyrite at Carlin, Getchell, and Turquoise Ridge contains little Au or As (Figs. 2 and 3A–3E). Sedimentary pyrite δ³⁴S varies widely between locations and stratigraphic horizons, and most of our samples are isotopically heavy (Figs. 3A–3C and 3E versus Fig. 3D; Table S1). Many of our depth profiles through sedimentary pyrite grain cores generated smooth plateaus (representative depth profiles in Figures 3A–3C, resulting in plateau data points shown in Figure 4). Several showed heterogeneity in δ³⁴S (Figs. 3D and 3E), perhaps due to fluctuations in microbial activity during sedimentary pyrite formation.

The unmineralized Jurassic magmatic-hydrothermal pyrite grain cores from Deep Star contain minor Au and δ³⁴S values of 6.5‰–6.9‰ (Fig. 4D), close to the mean δ³⁴S of Jurassic magmatic sulfur in the Great Basin (Arehart et al., 2013). The Eocene magmatic pyrites at Betze-Post, Deep Star, and Beast contain minor Au with δ³⁴S values (Fig. 4D) within the range of Tertiary magmatic sulfur in the Great Basin, which is itself isotopically variable due to variable host-rock interaction (Fig. 4F).

The NanoSIMS maps and depth profiles show a sharp contact between the precursor pyrite cores and the Au-As–rich hydrothermal rims (Figs. 2 and 3). The Au concentrations vary within the rims at a finer scale than previously surmised (Cline et al., 2005; Barker et al., 2009; Muntean et al., 2011; Large and Maslennikov, 2020) and also vary widely between samples (Figs. 3 and 4). In most samples, the depth profiles also show a dramatic change in δ³⁴S at the contact between precursor cores and hydrothermal rims (Fig. 3). Within the rims, the δ³⁴S values commonly vary inversely with Au (Figs. 3 and 4A–4C) but lack correlation with As. The lowest plateau values from the rims come from Au-rich zones: 1.7‰ δ³⁴S at Getchell, 2.5‰ at Carlin, 1.2‰ at Turquoise Ridge, 4.2‰ at Deep Star, and 2.1‰ at Beast (Figs. 4A–4D).

At each deposit, the δ³⁴S plateau values from the rims plot on a mixing line between two end members (Figs. 4A–4E): (1) an Au-poor sulfur source isotopically similar to local host pyrite cores, and (2) an Au-rich sulfur source with δ³⁴S values similar to those of mineralizing Eocene magmatic-hydrothermal fluids in the nearby Battle Mountain district (−1.8‰ to 7‰; Fig. 4F) and similar to the mean δ³⁴S of Great Basin Tertiary granitoid magmas (7.1‰; Arehart et al., 2013). Our microanalytical evidence for two-component mixing is supported by whole-rock geochemical data from Betze-Post's Screamer ore body (Fig. 4D; Christiansen et al., 2011): isotopically variable host pyrite in the Popovich Formation controls the δ³⁴S values in samples without detectable gold, and a sulfur source near 0‰ contributes substantially at high ore grades. The depth profiles show nanoscale zonation resulting from variation in relative contributions of the two sources over time (Fig. 3).

The Au-poor sulfur source may represent dissolution of sulfur-bearing minerals and organosulfur complexes during meteoric fluid circulation through the sedimentary host-rock package. The Au-rich sulfur requires an alternate source to explain the strong correlations between δ³⁴S and Au. A meteoric fluid convecting deeply through sedimentary rock would achieve δ³⁴S compositions representing regional or local averages of the stratigraphy. Although such a fluid could become Au rich by interacting with large volumes of rock containing trace metals or during passage through metalliferous sedimentary horizons (e.g., Large et al., 2011), initial correlations between δ³⁴S and metal content at a mutual point of origin would be lost during fluid circulation due to interaction with isotopically varied sulfur elsewhere in the rock package.

The Au-rich sulfur in Carlin-type ore was most likely derived from Eocene magmas. Magmatic-hydrothermal sulfide minerals from the nearest Eocene porphyry and other pluton-proximal deposits in the Battle Mountain district have δ³⁴S values ranging from −1.0‰ to 6.6‰ (this study; Theodore et al., 1986; King, 2017; Holley et al., 2019). Using a range of realistic precipitation temperatures, the δ³⁴S of the causative Eocene magmatic fluids can be constrained to −1.8‰ to 7.0‰ (Table S3). Temperature-induced fractionation would cause those fluids to precipitate Carlin-type pyrite with a δ³⁴S of 0.0‰–8.8‰ at 200 °C (a reasonable temperature estimate for Carlin-type mineralization; Cline et al., 2005). Because the Au-rich zones of our pyrites gave δ³⁴S values in this range (Figs. 4A–4D), we attribute their origins to the Eocene magmatic fluid.

The causative magmas were isotopically similar to those that generated the Beast dike (δ³⁴S depth profile values of 2.1‰–8.5‰). Eocene magmas of similar compositions either stalled out at depth beneath the Getchell trend or remain unrecognized. During magma cooling, Au and As would have become enriched in the exsolving fluids. Circulation of these fluids in the magmatic-hydrothermal environment led to variable interaction with Au-poor meteoric fluids and other sulfur sources, including isotopically heavy and light sedimentary sulfur minerals, as well as older magmatic and magmatic-hydrothermal sulfur and metals. Upon reaching favorable lithologic horizons and hydrologic or structural traps, these mixing fluids encountered preexisting pyrite. Sulfidation led to hydrothermal pyrite precipitation, and temporal fluctuations in the relative contribution of Au-rich magmatic and Au-poor meteoric fluids led to sequential nanoscale zones with covarying Au and δ³⁴S.

Carlin-type pyrite provides insights into the processes driving the formation of giant ore deposits. Fluid mixing led to fluctuations in metal precipitation, although the time scales over which the relative fluid contributions varied are unknown. These repeated oscillations were essential in upgrading metal concentrations at the mineral scale, ultimately leading to the formation of world-class ore bodies. In the absence of nanoscale data, previously developed models for metal enrichment in these deposits were overly simplistic, and such models require reevaluation (e.g., Sillitoe and Bonham, 1990; Ilchik and Barton, 1997; Emsbo et al., 2003; Ressel and Henry, 2006; Large et al., 2011; Muntean et al., 2011; Kusebauch et al., 2019; Xing et al., 2019). Low-spatial-resolution analytical methods have been applied to ore deposits for decades, even where micron- to submicron-scale trace element zonation or mineral intergrowths are visible in reflected light microscopy or scanning electron backscatter imaging. Such textures give intriguing hints that fluid mixing played a key role during mineralization in numerous geological settings, and our study highlights how high-spatial-resolution observations can elucidate the underlying geological processes.

This study was funded by U.S. National Science Foundation (NSF) Career Award EAR-1752756 (E.A. Holley). The Stanford Nano Shared Facilities are supported by NSF award ECCS-2026822. We thank Jean Cline, Phillip Gopon, Al Hofstra, Mike Ressel, and Patrick Sack for samples and discussion; Aaron Bell, Nigel Kelly, and Katharina Pfaff for analyses; Jae Erickson, Kelsey Livingston, Sage Langston-Stewart, and Heather Lowers for sample preparation; and Chris Henry, Celestine Mercer, Adam Simon, and an anonymous reviewer for comments. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government.