Unveiling the metallogenic continuum of an Archean craton

Abstract

The synchronous formation between 2675–2655 Ma of hydrothermal gold mineralisation in the Kalgoorlie and Kurnalpi Terranes and of magmatic sulfide mineralisation enriched in nickel, copper and platinum-group elements in the South West Terrane of the Archean Yilgarn Craton of Western Australia offers key insights into first-order controls on the genesis of mineral systems. Here we show that hydrothermal and magmatic deposits formed on opposite sides of this craton share four key features: timing, enrichment in incompatible chalcophile elements, positive Δ³³S sulfur isotope signatures, and links to hydrous, metasomatised lithospheric mantle. These commonalities challenge conventional models that treat such mineralised systems as unrelated. Instead, they point to a shared origin: a fertile lithospheric mantle reservoir enriched by crustal recycling, which was subsequently tapped to generate magmas and fluids anomalously endowed in volatiles and metals that migrated through the overlying lithosphere, punctually triggering ore formation. These findings support a unified mineral system model in which mantle processes exert a first-order control over metal endowment of Archean cratons.

Introduction

Over the past two decades, the concept of exploring for ore deposits by identifying regions where trans-lithospheric structures intersected fertile mantle domains at specific geological times has gained traction¹. This unified mineral system model^2,3 proposes that, under certain geodynamic conditions, the lithospheric mantle may become anomalously enriched in volatiles and metals through a range of metasomatic processes^4,5,6. These enriched reservoirs may then supply metals to ascending magmas and hydrothermal fluids, which go on to form ore deposits at various depositional points along a mantle-to-upper crustal pathway⁷.

Although it is widely accepted that metals in magmatic and hydrothermal systems share a mantle origin, the deposits themselves are typically interpreted to form through distinct ore-forming mechanisms rather than as parts of a continuum linked by deeper processes in the mantle. Researchers investigating orthomagmatic systems (Ni-Cu-PGE) develop skills and datasets based on igneous petrology and geochemistry, whereas structural geologists and metamorphic petrologists apply their skills to the characterisation of hydrothermal and orogenic gold systems. This dichotomy has led to the perception that the differences between mineralised deposits far outweigh their commonalities, a conclusion that encourages forensic studies that are mainly carried out at the deposit scale.

Mineral system research has not been immune to this perception, with studies often defining criteria that focus on single deposit types, like magmatic Ni-Cu-PGE⁸ or orogenic gold⁹. However, recent work by Holwell et al.¹⁰ has established that different deposit types can form at different depositional points along a crustal pathway. Their work demonstrated that both mantle-derived magmas and upper crustal porphyry-epithermal ore systems inherit a common metallogenic signature, imparted by sharing the same enriched lithospheric mantle in post-subduction environments.

If this continuum model is correct, multiple deposit types could form simultaneously from the same mantle reservoir at different depositional points across the lithosphere. This challenges the assumption that a ‘system’ produces only one deposit type and aligns with holist ideas in systems theory, which suggest that the effects caused by interaction among the parts of a system (synergistic interaction) may be far more important to understanding how a system works than the nature of the parts alone, as discussed by Corning¹¹. In this context, cataloguing deposit-scale characteristics becomes a reductive exercise, focused on processes at one depositional point along the continuum, limiting our ability to target new occurrences of mineralisation. If various deposit types are expressions of the same interconnected system, recognising their commonalities, such as shared geochemical and/or isotopic signatures reflecting a common lithospheric mantle reservoir, becomes important, as Holwell et al.¹⁰ showed. By understanding these parts, their interactions and the effects of these interactions (i.e., the nature of the lithospheric reservoir, the geodynamic processes enriching it, the depositional points along a mantle-to-upper crustal pathway and the current preservation depth of the crust), we can better anticipate where mineralisation is most likely to develop, minimising the use of resources associated with exploration targeting and significantly reducing environment disturbance.

This study tests this idea by establishing whether such isotopic or geochemical commonalities exist between orthomagmatic and orogenic systems within the Archean Yilgarn Craton between 2675 and 2655 Ma. We use the most recent datasets and petrogenetic ideas for orogenic gold deposits in the Yilgarn Craton and compare them to original geochemical, mineralogical, and isotopic data for the Gonneville-Julimar PGE-Ni-Cu deposit, representing the first published material on this magmatic sulfide system. Discovered in 2020, Gonneville-Julimar is Australia’s largest PGE resource to date (13 Moz Pd, 2.9 Moz Pt, 0.5 Moz Au, 960 kt Ni, 540 kt Cu¹²), located 70 km northeast of Perth in the underexplored South West Terrane (Fig. 1). The host intrusion has a magmatic crystallisation age of 2668 ± 4 Ma¹³, coeval with the first phase of gold mineralisation in the Kalgoorlie and Kurnalpi Terranes (ca. 2675–2655 Ma¹⁴), located 550 km to the east in the Yilgarn Craton (Fig. 1), a mineralising event that generated some of Earth’s richest gold deposits (combined endowment >2400 t Au¹⁵).

Fig. 1: Map of the Yilgarn Craton showing Δ³³S values for 2.6–2.8 Ga granitoids and ore-related sulfides from 2675 to 2655 Ma orthomagmatic and orogenic mineralisation, adapted from Caruso et al.²⁹ and Duuring and González-Álvarez³⁰.

Data are compiled from this and previous studies^{24,25,27,28,29,30,75,76}. Data for Gonneville-Julimar are newly reported analyses, part of this work. Warm colours indicate positive Δ³³S values; cool colours indicate mantle-like or negative Δ³³S values.

Results

Gold deposits in the Kalgoorlie and Kurnalpi Terranes

The Kalgoorlie and Kurnalpi Terranes share a common magmatic, deformational, and mineralisation history (readers are referred to the work of Masurel and Thébaud¹⁶, Mole et al.¹⁷ and Witt et al.¹⁸ for detailed deposit descriptions and regional geological framework). Recent work by Sumail et al.¹⁴ splits the deposits into two major gold-forming episodes: an early syn-orogenic event (ca. 2675–2655 Ma), which occurred when mantle-derived hydrous magmas and hydrothermal fluids ascended along the margins of the Kalgoorlie and Kurnalpi Terranes, and a subsequent late-orogenic gold event (ca. 2640–2630 Ma), associated with granitic plutonism and peak metamorphism during the reactivation of major structural zones. Our study focuses on the first episode, which coincides with the formation of the Gonneville-Julimar PGE-Ni-Cu deposit. Gold deposits formed during this early episode share three key features: (i) a spatial and temporal link to ‘mafic-granite’ magmatism, (ii) Bi–Te enrichment, and (iii) a narrow range of positive Δ³³S isotopic values. These three key features define a metallogenic continuum within the Yilgarn Craton, discussed here for the first time, and are described briefly below for the orogenic gold systems of the Kalgoorlie and Kurnalpi Terranes.

First, the link to widespread ‘mafic-granite’ magmatism includes sanukitoids and lamprophyres emplaced between 2675 and 2655 Ma, peaking around 2665 Ma^14,19. These intrusions are sourced from a hydrous, volatile-rich mantle^20,21, and have a well-recognised spatial and temporal association with gold mineralisation across the Kalgoorlie and Kurnalpi Terranes²². Their timing of emplacement suggests a common hydrous mantle source for both ‘mafic-granite’ magmatism and gold mineralisation¹⁹.

Second, Bi–Te enrichment is characteristic of gold mineralisation in the Kalgoorlie and Kurnalpi Terranes, where Bi-Te minerals commonly appear either directly in ore or as pathfinder elements¹⁵. The only study that has systematically examined the modal abundance of Bi-Te minerals in the region focused on the Golden Mile world-class mineralised camp, which comprises the richest set of gold deposits in the area, where more than 25% of gold occurs as Au-tellurides²³.

Third, ore-related sulfides associated with gold mineralisation in the Kalgoorlie and Kurnalpi Terranes have a narrow range of Δ³³S values, with 89% of the values falling between 0 to +0.6‰^24,25 for multiple gold deposits over a strike distance of 400 km (Fig. 2C). The ∆³³S signature (∆³³S = δ³³S* − 0.515 δ³⁴S*) represents the deviation of the measured ³³S/³²S ratio from the expected value under mass-dependent sulfur isotope fractionation²⁶. A near-zero ∆³³S signifies no deviation from its expected value, i.e., mantle, whereas any other value, negative or positive ∆³³S, reflects mass-independent fractionation (MIF), a process caused by the photodissociation of volcanic SO₂ gas in the Archean oxygen-free atmosphere²⁶ (Fig. 2A).

Fig. 2: Multiple sulfur isotopes for 2675–2655 Ma ore-related sulfides and sulfide-bearing country rocks.

A Δ³³S vs. δ³⁴S plot of sulfur reservoirs caused by the breakdown of volcanic SO₂ gas in the Archean oxygen-free atmosphere (modified after Virnes et al.⁷¹). S₈ is the isotopic composition of reduced insoluble sulfur and H₂SO₄ oxidised, water-soluble sulfate. B Δ³³S vs. δ³⁴ for the Julimar greenstone belt country rocks and the Gonneville and Hann-Hooley-Dampier intrusions, plotted by intrusive stratigraphy. S¹ represents the approximate Δ³³S signature of the sulfur reservoir required before dilution by a juvenile magma to produce the Δ³³S signature in the Gonneville and Hann-Hooley-Dampier intrusions, MDF-S defined from LaFlamme et al.²⁵, error crosses represent the overall uncertainty and are calculated as the pooled 2σ standard deviation of the calibration standards for each analytical session. C Δ³³S vs. δ³⁴S for sulfides from gold deposits of the Kalgoorlie and Kurnalpi and South West Terranes (black circles and associated error crosses compiled from refs. ^{24,25,27,28,29,30,75,76}); results from the Tampia gold deposit are highlighted (orange circles), overlain by new data for magmatic sulfides from the Gonneville and Hann-Hooley-Dampier intrusions. D The Δ³³S-δ³⁴S data from the Tampia deposit³⁰ (orange diamonds) overlain on plot B; the Tampia Δ³³S signature presents a plausible reservoir that could form the Δ³³S signature at the Gonneville and Hann-Hooley-Dampier intrusions when assimilated and diluted by a juvenile mantle magma bearing a Δ³³S signature of 0‰; see text for details.

The observed narrow positive Δ³³S range that characterises gold mineralisation in the Kalgoorlie and Kurnalpi Terranes indicates that sulfur and gold were not scavenged by fluids from sediments local to each deposit, as the diverse range of sediments over 400 km would generate a more variable and scattered Δ³³S range across the different deposits²⁷. Instead, this narrow positive Δ³³S range is interpreted to primarily record the involvement of a lithospheric reservoir, which was modified away from 0‰ through the incorporation of Archean crust with Δ³³S ≠ 0‰. Given the large mass difference between recycled crustal material and the lithospheric mantle, such a combination likely resulted in a minor shift in the isotopic composition of the modified mantle²⁸. Incorporation of sulfide-rich crust (positive Δ³³S²⁶; Fig. 2A) would modify the reservoir toward slightly positive Δ³³S values (e.g. from 0‰ to values between 0.05‰–0.3‰²⁸), whereas incorporation of sulfate-rich crust (negative Δ³³S²⁶; Fig. 2A) is thought to modify the reservoir toward slightly negative values (e.g. from 0‰ to values between −0.05‰ and −0.3‰²⁸). The narrow range of positive Δ³³S values in the Kalgoorlie–Kurnalpi gold systems therefore suggests that sulfur was supplied by a lithospheric reservoir, which was modified by sulfide-rich Archean crust^24,25,27 prior to being tapped by magmas and/or fluids responsible for the gold mineralising event.

Gold deposits in the wider Yilgarn Craton

Caruso et al.²⁹ analysed 49 granitoids (between 2.8 and 2.6 Ga) across the Yilgarn Craton and identified the same distinct positive Δ³³S reservoir underlying both the Kalgoorlie and Kurnalpi Terranes as well as the South West Terrane (Fig. 1). In the Kalgoorlie and Kurnalpi Terranes, the authors confirmed matching Δ³³S signatures between granitoids and gold mineralisation, supporting a shared lower crustal or upper mantle sulfur source.

In the South West Terrane, work by Duuring and González-Álvarez³⁰ has also identified the occurrence of “early” gold mineralisation (e.g., Tampia) that displays positive ∆³³S, consistent with granite ∆³³S signatures reported for the South West Terrane by Caruso et al.²⁹ (Fig. 1). We emphasise that only “early” gold deposits from Duuring and González-Álvarez³⁰ are included for comparison in this study as they can be constrained as pre-2665 Ma through equilibrium textures, and are therefore relevant to the time period in this study. Conversely, “late” gold deposits display textural evidence for formation after peak metamorphism associated with the Corrigin Tectonic Zone between 2665 and 2635 Ma³¹.

The positive Δ³³S signature that characterises gold deposits in the Kalgoorlie, Kurnalpi and South West Terranes contrasts with the signature in the Youanmi Terrane, where both granites and gold mineralisation have negative Δ³³S (Fig. 1), implying that geodynamic processes may have preferentially recycled seawater sulfate as opposed to elemental sulfur into deep lithospheric reservoirs²⁹. Recognising the significance of this Δ³³S signature, Caruso et al.²⁹ proposed that granites might act as ‘crustal probes’, preserving the Δ³³S signature of lower crustal or upper mantle sulfur reservoirs with differences in the signature revealing the presence of a diverse range of reservoirs formed through different geodynamic events.

The Gonneville-Julimar deposit

The Gonneville-Julimar deposit comprises a series of variably mineralised meta-mafic-ultramafic intrusions with similar lithological and geochemical characteristics over a 10 km strike length in an area with extensive regolith and little outcrop. The southernmost intrusion, Gonneville, the focus of exploration and resource drilling, hosts the 17 Moz PGE resource¹². The other intrusions along strike are collectively known as the Hann-Hooley-Dampier (HHD) prospects. The intrusions share the same geological characteristics, fractionation sequence and PGE enrichment, but are less well defined by drilling. All these intrusions reside within the Julimar belt, a ~ 30 km by 8 km greenstone belt sequence composed of intercalated meta-sedimentary and meta-mafic-ultramafic rocks (Fig. 3).

Fig. 3: Gonneville-Julimar geology.

A BSE-SEM image of a typical PGM from the Gonneville intrusion showing both Pd-Bi-Te and Pd-Bi types. Further SEM images can be found in the Supplementary Materials Fig. 7. B BSE-SEM image of a Pd-Bi-Te grain from the Gonneville intrusion crosscut by serpentine (antigorite) laths. C Plan view of the Gonneville intrusion showing seven internal meta-cumulate layers, each defined by dominant mineralogy and MgO content (wt%), along with later dolerite dykes (green) and surrounding Julimar country rock sequence (grey). Lithogeochemistry is detailed further in Supplementary Materials 1.4. D Downhole plot for drill hole JD369W1 (drill hole trace shown on map in panel (C), which intersects most meta-cumulate layers from hanging wall to footwall. The x-axis shows 1 m assay intervals; MgO (wt%) is plotted on the top, total PGE (Pd + Pt + Au, ppm) on the bottom. Colours correspond to the seven layers shown in (C).

The Gonneville intrusion measures 1.8 km by 0.6 km and comprises meta-ultramafic-mafic cumulate rocks, predominantly serpentinised olivine peridotite (lizardite-antigorite-magnetite-amphibole-chromite) with lesser intervals of meta-orthopyroxenite (amphibole-chlorite), meta-norite and meta-leucogabbro (clinozoisite-amphibole). The intrusion exhibits distinct layering and a full magmatic fractionation sequence, transitioning upwards from primitive meta-harzburgite (>30% MgO) through meta-orthopyroxenite (20–30% MgO), to evolved meta-leucogabbro (<10% MgO; Fig. 3C and Fig. 3D). Deposit geology, petrology and lithogeochemistry are further detailed in Supplementary Material 1.1 to 1.5.

Both the Gonneville and HHD intrusions, along with their host Julimar sequence, have experienced extensive post-emplacement serpentinisation at upper-greenschist facies, which has entirely replaced the primary non-sulfide igneous minerals. The mineral and sulfide assemblages observed in the intrusions show no evidence for retrogressive re-equilibrium. Instead, metamorphic minerals pseudomorph primary igneous textures (see figures in Supplementary Material 1.2). The presence of an antigorite–actinolite assemblage, absence of metamorphic olivine, scarcity of talc, and the nature of the sulfide species present constrain the metamorphic conditions as being H₂O saturated, under reduced low CO₂ conditions, between 450 and 500 °C at 2–6 kbar. These conditions are consistent with upper greenschist facies metamorphism. Metamorphic constraints are detailed in Supplementary Material 1.2 and 1.3.

Whereas the average Pd + Pt grade in the deposit is 0.79 g/t, with an average Pd:Pt ratio of 4.5:1, higher grade lenses average 2.0 g/t Pd + Pt¹² and there is a maximum sample grade of 1799 g/t Pd. The PGEs occur within both platinum-group minerals (PGMs) and base metal sulfides (BMS), with PGMs present within sulfides, at sulfide boundaries and near sulfide but surrounded by silicate (Fig. 3A and figures in Supplementary Material 1.2). The PGMs and sulfides at Gonneville-Julimar are cross cut by antigorite (Fig. 3B), similar to textures described at Cape Smith³², Black Swan³³ and Six Mile³⁴. The observed texture is caused by antigorite growing into fractures on the edge of sulfides that open up due to expansion during serpentinisation^33,35. Five cm segments were selected from diamond drill cores of the Gonneville intrusion, the HHD prospects, and within the Julimar greenstone belt. Each segment was analysed for multiple sulfur isotopes, PGE deportment and trace element geochemistry (full details in ‘Methods’).

Multiple sulfur isotopes

Multiple sulfur isotopes were determined across mineralised zones in the Gonneville intrusion and the HHD prospects, as well as from sulfidic horizons in the Julimar greenstone belt sequence, by continuous-flow Elemental Analyser–Isotope Ratio Mass Spectrometry (EA-IRMS)³⁶. The isotopic values across the mineralised zones show a narrow positive range of δ³⁴S (+0.45‰ to +1.6‰) and Δ³³S (−0.07‰ and +0.9‰) over 10 km of strike, with an average Δ³³S of +0.5‰ (Fig. 2B). In contrast, sulfidic country rock samples, ranging from 5 m into the footwall to more than 5 km away, show broader δ³⁴S (−1.05‰ to +3.6‰) values and lower Δ³³S (−0.6‰ to +0.47‰), with an average Δ³³S across the sulfide zones of +0.1‰ (Fig. 2B). Pooled 2σ uncertainties varied from ± 0.14‰ to ±0.25‰ for δ³⁴S, from ±0.15‰ to ±0.18‰ for δ³³S, and from ±0.12‰ to ±0.18‰ for Δ³³S.

Palladium deportment

In the Gonneville intrusion, PGEs occur within both PGMs and BMS. Elemental deportment between these phases was determined by a mass balance calculation using in situ LA-ICP-MS analysis of sulfides, trace element geochemistry, and automated mineralogy via a TESCAN TIMA SEM³⁷. The results show that PGMs host over 90% of Pd and over 98% of Pt, with less than 10% of Pd occurring in solid solution within BMS, predominantly pentlandite (Fig. 4A). Bright-phase scans were used to identify and quantify these PGM phases, with palladium-bismuthotellurides (Pd-Bi-Te) identified as the dominant PGMs, constituting 75% of the total PGMs (Fig. 4B). The performance, repeatability and representativeness of the bright phase search in capturing the true PGM population is detailed in Supplementary Material 2.

Fig. 4: Palladium deportment and trace element geochemistry of the Gonneville-Julimar PGE-Ni-Cu deposit.

A Elemental deportment into base metal sulfides (BMS). B Total PGM types by area (µm²), normalised to Pd assay per sample and summed across all samples. The “Other” category comprises PGM types that each contribute <2% of the normalised area, including Pd–Pb, Rh–As, Pt unassigned, Ag–Te, and Au. C Chalcophile element plot, normalised to primitive mantle⁷⁷ for samples with <2 %S (0.1–1.73 wt% S). D Rare earth element plot, normalised to chondrite⁷⁸ for samples with <2 %S (0.1–1.73 wt% S). E λ₀ vs. λ₂ plot, compares the REE pattern of the Gonneville SP1 and SP2 cumulates to a compilation of rock types including: Boninites from the Izu-Bonin arc⁷⁹, olivine-rich cumulates from selected orthomagmatic PGE deposits (Stillwater⁸⁰ and Bushveld⁴³), average Munro, Yilgarn and Barberton komatiite⁸¹ and MORB³⁹. λ coefficient plots are a way to visualise and compare the smoothness and shape of chondrite-normalised REE patterns. λ₀ describes average REE for each sample and λ₂ the quadratic curvature of the normalised REE pattern. A high λ₂ is attributed to strong u-shaped REE pattern, a pattern often used to describe boninite or parental orthomagmatic magma REE patterns, whereas a λ₂ near 0 represents a flat REE pattern, and a negative λ₂ represents LREE depleted rocks. Results follow methodology of O’Neill³⁹ and are calculated in BLambdaR⁸². Only samples with a chi-squared <15 are plotted. F Chondrite-normalised plot for Gonneville SP1 cumulates (light grey with circles) overlain by the calculated shape component (orthogonal polynomial function) for the SP1 normalised plot shape (dark grey), ignoring Eu. Individual λ coefficients used to describe this orthogonal polynomial, λ₀, λ₁ and λ₂, are plotted to help visualise what the different λ values describe in plot (E).

Trace element geochemistry

An additional dataset of low-sulfur samples (0.1–1.73 wt% S) from across the Gonneville intrusion cumulate stratigraphy was analysed for whole rock major oxides and trace elements at ALS Laboratories, Perth. These samples were selected to best represent the silicate cumulate minerology and to minimise the influence of sulfides on the fractionation and concentrations of chalcophile and siderophile elements across the intrusions. Trace element patterns for these low-sulfur samples show flat to slightly depleted HREE profiles (average normalised Gd/Yb of 0.83) and LREE enrichment (average normalised La/Sm of 2.17) across all stratigraphic levels, independent of cumulate layering (Fig. 4D).

These inflected or u-shaped trace element patterns closely resemble those described for boninites or high magnesian andesite parental magmas and their olivine-rich cumulates, which are common parent magmas for other orthomagmatic PGE deposits³⁸. Figure 4E plots lambda coefficients λ₂ and λ_0, which describe the orthogonal polynomial functions (shape components) of these trace element patterns³⁹ (Fig. 4F). Such signatures can form either through (i) extensive crustal assimilation by melts from high degrees of melting of anhydrous peridotite^40,41, or (ii) from melts incorporating a lithospheric component that has been fluxed by fluids rich in ‘crustal’ components, produced during the devolatilisation of recycled crust^42,43. These low-sulfur samples also show that the Gonneville intrusion is enriched in incompatible chalcophile elements (Rh-Pt-Pd-Au-Cu-Te) relative to compatible components (Ni-Co-Os-Ir), with notably high Pd/Ir ratios (average 141; Fig. 4C), which exceed typical values for anhydrous mantle melts (Pd/Ir = 5–30⁴⁴) and closely resemble global boninite signatures (Pd/Ir ~120⁴⁵).

Discussion

This study proposes a new process-driven mineral system model for ore genesis that accounts for the observed craton-wide spatial distribution of Δ³³S signatures between 2675 and 2655 Ma. Central to this model is crustal recycling, a process that introduced positive Δ³³S and volatiles (including water) into the lithospheric mantle beneath the Yilgarn Craton. It is argued that this modified enriched reservoir was subsequently tapped by rising melts and hydrothermal fluids, which formed a continuum of deposits, including orthomagmatic PGE-Ni-Cu and orogenic Au mineralisation. Accordingly, it is proposed that these mineralising systems are genetically related, sourced from a shared enriched lithospheric mantle reservoir that underlay the Yilgarn Craton between 2675 and 2655 Ma.

Hartnady et al.⁴⁶ modelled the fluid budgets of mafic protoliths (basaltic through to komatiitic) and established that ultrahigh-MgO metabasite (komatiite) can retain significant amounts of H₂O in chlorite until temperature exceeds 700 °C. Their modelling demonstrated that the burial of Archaean mafic–ultramafic crust along typical mid to late Archean geotherms results in dehydration of hydrated high-MgO ultramafic rocks within the lithospheric mantle, with estimates of at least 6, and up to 16 vol.% mineral-bound H₂O⁴⁷ being released. This process is viable in either subduction or delamination ‘sagduction’ scenarios via greenstone drip. Importantly, in the Yilgarn Craton, any crust that is recycled prior to and between 2675 and 2655 Ma would consist of mafic-ultramafic rocks that were emplaced into submarine rifts during plume-induced crustal thickening events between 2710 and 2690 Ma^17,48. Therefore, the recycled material would be composed of hydrated mafic crust from a submarine environment, interlayered with sediments that carry a MIF-S isotopic signature (Δ³³S ≠ 0‰)^49,50. We suggest, using work completed by Hartnady et al.⁴⁶, that the dehydration of the ultramafic component in this recycled crust would release substantial volatiles into the lithospheric mantle, providing a mechanism not only capable of triggering the emplacement of the sanukitoids and lamprophyres between 2675 and 2655 Ma^14,19, but also a process for enriching the mantle lithosphere in both metals and a MIF-S isotopic signature.

The release of water and volatiles into the lithospheric mantle is known to reduce the solidus temperature by at least 300 °C at depths greater than 90 km⁵¹. This process induces partial melting of both the recycled material and surrounding lithospheric mantle, causing the incongruent melting of both lithospheric mantle and recycled sulfide, enriching reservoirs in sulfur, incompatible trace elements and chalcophile metals, which include both semi-metals (Bi-Te) and ore-metals (Pt-Pd-Au-Cu)^52,53,54. Alternatively, if the released water and volatiles are oxidised, they may leach metals from the surrounding lithospheric mantle and recycled sulfide, through coupled dissolution and reprecipitation mechanisms, by hydrating wall rocks in exchange for metals and mobilisation of sulfur to form fusible enriched reservoirs of incompatible trace elements and chalcophile metals^6,55. It is argued that the composition of the recycled material is likely to influence which of these mechanisms is favoured. Regardless of whether one or both operated, these reservoirs would be anomalously sulfur-rich, giving them an enhanced capacity to carry chalcophile metals^56,57. Furthermore, these processes would impart on the reservoirs a mixed signature comprising both mantle Δ³³S (0‰) and recycled Δ³³S ( ≠ 0‰), because of the incorporation of MIF-S carrying sulfur from the recycled interlayered sedimentary horizons.

Further enrichment of these lithospheric reservoirs in incompatible metals can be caused by later low-degree partial melting events trigged by a series of processes including pressure reduction from rifting, renewed volatile input from recycled material, or a thermal pulse from a mantle plume. The metals from this enriched lithospheric mantle must then have been transferred into the crust between 2675 and 2655 Ma, although the exact geodynamic environment that caused this transfer is unclear. Possible mechanisms for mantle-crust transfer include an additional heat source⁷, hydration-induced melting⁵⁸, fluid devolatisation⁵⁹ or further low degree partial melting⁵⁵. Regardless, it is argued that subsequent mantle melts or hydrothermal fluids tapped these enriched reservoirs to form a continuum of deposits at different crustal depths, spatially localised along structures active between 2675 and 2655 Ma⁶⁰.

The mechanisms that incorporated metals into mantle melts and hydrothermal fluids, and the local conditions present at the depositional points across the lithosphere determine the attributes and metallogenic assemblage of any specific deposit style along the continuum. These factors include: degree of mantle melting, redox state, fluid temperature, ligand availability and crustal depth. For example, for the formation of a PGE deposit like Gonneville-Julimar, it is likely that higher degrees of partial melting and deeper deposition within the crust favoured Pd-rich mineralisation hosted within ultramafic cumulates. In contrast, shallower crustal conditions and possibly lower degrees of melting promoted Au-focused mineralisation within lower-temperature silica-rich shear zones and veins in the Kalgoorlie and Kurnalpi Terranes. Despite these differences, if the same enriched reservoir supplied metals to ascending magmas and hydrothermal fluids, the mineralisation will retain three defining features: (i) positive Δ³³S signatures, caused by the introduction of sulfides that had been recycled from the Archean atmosphere-hydrosphere into the upper lithosphere, (ii) evidence of a hydrous lithospheric mantle source, caused by the dehydration and devolatilisation of the recycled crust, and (iii) an enrichment in incompatible chalcophile metals caused by incongruent melting.

We test this prediction by examining evidence for each defining feature in the mineralisation hosted both at Gonneville-Julimar and in the Kalgoorlie–Kurnalpi Terranes, starting with the enrichment of incompatible chalcophile metals, focusing on Bi and Te. At the Gonneville-Julimar deposit, 90% of PGEs partition into PGMs rather than BMS (Fig. 4A). Such a high partition of PGEs into PGMs is rarely reported in economic deposits, with most studies reporting a higher proportion of Pd in BMS⁶¹. The reason for a high PGM deportment at Gonneville-Julimar is likely driven by the direct crystallisation of PGMs from semi-metal droplets, due to higher semi-metal budget in the original silicate magma. Similar observations are reported in other occurrences of PGM-rich mineralisation^62,63 and has support from a number of experimental studies^64,65.

First, sulfide liquid saturates and separates from silicate magma, capturing chalcophile elements⁶⁶. At high temperatures (~1000 °C), a monosulfide solid solution (MSS) and an immiscible sulfide liquid (ISS) form, with IPGEs (Ir, Os, Ru, Rh) favouring MSS⁶⁷. As the temperature drops (~900 °C), these liquids crystallise, with Pt and Pd forming either PGMs or substitute into BMS⁶². Helmy et al.⁶⁴ investigated the causes of this difference, with their experiments showing that when sufficient semi-metals are available during MSS-ISS segregation, both Pt and Pd combine with the semi-metals to form droplets, with PGM phases crystallising directly from these droplets as the melt cools. Their experiments also showed that when insufficient semi-metals were available, Pt and Pd substitute into the MSS and ISS, staying in solid solution to be included in BMS as a cation, with Pd preferentially concentrating in pentlandite during subsolidus transformation of MSS and ISS.

Other mechanisms for high PGM deportment at Gonneville-Julimar can be ruled out: (i) the maximum metamorphic temperature did not exceed 500 °C, a temperature that is too low to cause the dissolution and remobilisation of pre-existing TABS⁶⁸. (ii) There is no petrographic evidence for PGM addition by a metamorphic fluid: PGMs do not concentrate in veins and have no relationship with external contacts or faults; furthermore, antigorite crosscuts both PGMs and sulfide (Fig. 3B), implying that the PGMs were already present on the sulfide-silicate boundary prior to serpentinisation and fluid ingress. In contrast, the PGM-enriched horizons follow the same stratigraphic dip and strike as those defined by the primary magmatic stratigraphy (Fig. 3C). And (iii) if sulfur loss was driving the Pd deportment in PGMs, sulfides would retain a core of Pd and show a rim of Pd depletion, marking shrinkage during exsolution as described by Mansur et al.⁶⁹. However, LA-ICP-MS maps (Supplementary Material Fig. 9) of Gonneville-Julimar sulfides show no such pattern. While minor localised versions of these processes may occur, they cannot explain the bulk of PGM mineralisation at Gonneville–Julimar. The most plausible explanation for the high PGM deportment is that the primary magma had sufficient semi-metals (Bi-Te) to cause Pd to partition into PGMs rather than into BMS.

The enrichment of the original magma in Bi-Te can be caused by the assimilation of local Bi-Te rich crust or the incorporation of a Bi-Te rich mantle reservoir as a mantle-derived melt ascends. Local crustal assimilation is unlikely: sulfur isotopes can serve as a proxy for evaluating the extent of crustal assimilation in ore formation in magmatic systems^50,70. If there was extensive assimilation of local crust during formation, sulfides within the intrusion should reflect the sulfur isotopic signature of the local crust⁷¹. Two observations argue against significant local crustal sulfur input. First, samples from the magmatic intrusion show a tight isotopic cluster for both δ³⁴S and Δ³³S across 10 km of strike extent, whereas local crustal samples exhibit high variability for both δ³⁴S and Δ³³S (Fig. 2B). Significant crustal assimilation across 10 km should reflect this variability in the isotopic signature of the intrusion, which it does not. Second, Δ³³S is an indelible tracer: once incorporated into a sulfur source it can only be diluted by mantle sulfur toward 0‰²⁴. Therefore, to generate the Δ³³S values at Gonneville-Julimar, a reservoir with a higher Δ³³S must be diluted toward mantle values by the addition of mantle-derived magma (represented by the S¹ reservoir in Fig. 2B). The Δ³³S signatures of the local crust are either at, or below those of the intrusive samples (Fig. 2B), suggesting that assimilation of the local crust cannot fully explain the Δ³³S in the magmatic sulfides, leaving the reservoir required to generate these values unaccounted for.

There is evidence for this missing reservoir in the sulfur isotope data collected for “early” gold mineralisation in the South West Terrane³⁰. The Tampia gold deposit in the South West Terrane has a narrow range of high Δ³³S values between 1.01 and 1.24‰ (Fig. 2D). It is therefore plausible that the same lithospheric reservoir that supplied Tampia with its positive Δ³³S signature was assimilated and mixed into the Gonneville–Julimar magma, causing a dilution of the Δ³³S signature towards mantle and resulting in the observed Δ³³S values. Together, these observations suggest a lithospheric mantle-derived source for both sulfur and metals, including semi-metals Bi-Te, at the Gonneville-Julimar deposit. This conclusion matches the interpretation for the Kalgoorlie–Kurnalpi deposits, where Bi-Te enrichment is widespread and the presence of a narrow, positive Δ³³S signature (Fig. 2C) establishes the semi-metals as derived from the lithospheric mantle.

The positive Δ³³S signature that characterises mineralisation across both the Kalgoorlie–Kurnalpi and Gonneville–Julimar deposits indicates that the sulfur was sourced from a lithospheric mantle reservoir modified by deep crustal recycling (Fig. 2C). This sulfur isotopic signature is also present in the Caruso et al.²⁹. Yilgarn granitoid dataset. Granites in both the South West and Kalgoorlie–Kurnalpi Terranes contain positive Δ³³S values that match those of the associated Au and PGE-Ni-Cu mineralisation, respectively (Fig. 1). This observation suggests that both granites and ore-forming fluids tapped the same fertile lithospheric mantle reservoir. As such, granites with this isotopic signature not only provide an independent line of evidence for the presence of a craton-wide metasomatised lithospheric mantle between 2675 and 2655 Ma, but also map the extent of this mantle reservoir beneath the Yilgarn Craton. Figure 5 presents a conceptual cross section that illustrates how the relative spatial relationship between mineral systems and granites is linked to inheritance of their lithospheric mantle-derived geochemical and isotopic signatures through emplacement along trans-lithospheric structures at 2675–2655 Ma.

Fig. 5: Time-specific mantle fertility beneath the Yilgarn Craton 2675–2655 Ma.

A conceptual and simplified cross section (not to scale) illustrating a shared fertile mantle beneath both the South West Terrane and the Kalgoorlie–Kurnalpi Terranes. Commonalities across different types of mineralisation and granitoids in these terranes indicate a shared origin from the same craton-wide fertile mantle, characterised by (i) Bi–Te enrichment, (ii) positive Δ³³S values, and (iii) evidence of a hydrous lithospheric mantle. Formed through crustal recycling, this enriched lithospheric mantle contributed metals, volatiles, and sulfur to a metallogenic continuum active between 2675 and 2655 Ma. See main text for discussion of the apparent separation of this fertile domain by a broad zone of negative Δ³³S in the Youanmi Terrane. Plan view and colours adapted from Fig. 1.

The positive Δ³³S domains in the Kalgoorlie–Kurnalpi and South West Terranes are separated by a broad zone of negative Δ³³S centred on the Youanmi Terrane (Fig. 1 and Fig. 5), suggesting that different sources of sulfur may have been recycled into the mantle lithosphere. The origin of this difference remains unresolved. It may reflect the same sulfide-rich melts being funnelled to the thinner cratonic margins on the South West and Kalgoorlie–Kurnalpi Terranes^57,72, or the occurrence of analogous geodynamic processes, such as slab or drip delamination tectonics¹⁷, acting simultaneously on both sides of this lithospheric block. An alternative scenario involves subduction along the eastern margin of the Yilgarn Craton and the involvement of a possible subduction-like setting in the South West Terrane⁷³.

Dehydration of the recycled material would cause the third key feature tested in this study: a link to petrogenesis from, or the incorporation of, a hydrous lithospheric mantle. The proposed model predicts that crustal recycling prior to and between 2675 and 2655 Ma produced a volatile-rich mantle lithosphere capable of generating hydrous magmas. In the Kalgoorlie and Kurnalpi Terranes, the connection is direct: gold mineralisation shows a clear spatial and temporal association with sanukitoid and lamprophyre magmatism, a connection which has been extensively discussed in the literature^14,19,22. At Gonneville-Julimar, the link is more cryptic, but still evident. While no direct spatial link exists between the Gonneville or HHD intrusions and mafic-granite magmatism, the LREE patterns (Fig. 4E) combined with enrichment of incompatible chalcophile over compatible chalcophile elements (a high Pd/Ir ratios of 141; Fig. 4C) are consistent with partial melting of a depleted mantle source that incorporates a lithospheric component that has been fluxed by fluids rich in ‘crustal’ elements: incompatible elements (LREE) and incompatible chalcophile metals (Rh-Pt-Pd-Au-Cu-Te) through small-degree incongruent melting of recycled crust and crustal sulfide material^52,53,55.

At the time of writing the authors are not aware of any systematic, detailed, low detection PGE assays for the ore fluids in the Kalgoorlie–Kurnalpi deposits; however, Choi et al.²¹ assessed PGE, whole rock and trace elements in 2684–2640 Ma lamprophyres across the Kalgoorlie–Kurnalpi Terrane to propose that observed enrichments in Pt and Au in the lamprophyres likely reflect a lithospheric source modified by slab-related fluid fluxing. As a direct sample of the lithospheric mantle and a proxy for the gold mineralising fluids, these lamprophyres suggest that the PGE patterns at Gonneville-Julimar may be tracing the same enriched lithospheric mantle. Together with widespread ca. 2.65 Ga mafic–granite magmatism across the Yilgarn Craton, this piece of evidence supports the presence of a large-scale, hydrous, fertile mantle domain beneath the craton at the time.

Subtle differences remain between the South West and Kalgoorlie-Kunalpi Terranes. For example, the two terranes remain separated by the negative Δ³³S signature in the Youanmi Terrane. Furthermore, a higher thermal input may be required at Gonneville-Julimar to generate asthenosphere-derived melts capable of incorporating lithospheric mantle⁵⁷, relative to the type of magmatism and hydrothermal activity recorded at the same time in the Kalgoorlie and Kurnalpi Terranes. These variations can be explained by differences in the nature of crustal material being recycled, or by spatial difference in thermal input, reflecting variations in heat distribution across the lithosphere associated with plume- or delamination-derived thermal anomalies, leading to different mechanisms of metal transfer from the enriched lithosphere into the crust. Alternatively, these differences may relate to variable preservation, with Gonneville and the HHD intrusions possibly representing a portion of the crust located deeper than the Kalgoorlie and Kurnalpi Terranes, although this hypothesis remains speculative as no geobarometric analyses have been completed for the Julimar greenstone belt. Regardless, our focus here is both process-led and practical: to better predict the location of mineralised environments, we must first determine when and where metasomatism occurred. Rather than debate tectonic interpretations, we advocate for the use and development of geochemical and isotopic tools, such as Δ³³S mapping of granites, as direct measurements of metasomatised lithosphere in space and time.

By examining coeval mineralisation of different types, we shift focus from their differences to their shared features. This perspective reveals that mineral deposits formed on opposite sides of the Yilgarn Craton between 2675 and 2655 Ma retain the same four defining features of their lithospheric mantle source, regardless of deposit type. These commonalities demonstrate the presence of a metallogenic continuum, controlled by a metasomatised lithospheric mantle, across the Yilgarn Craton. This approach invites a re-evaluation of how craton-scale mineral systems form and evolve, encouraging geoscientists to consider time-specific lithospheric mantle fertility as the underlying control on mineralisation. An idea that shifts the focus away from thinking of deposits as distinct ore-forming mechanisms and instead recognises them as products of a single interconnected lithospheric system, where multiple deposit types can form simultaneously from the same lithospheric mantle. The aim of systems science is to be predictive, an aim inextricably linked to mineral exploration. Studies, including this one, continue to emphasise the importance of a fertile lithospheric reservoir, a component that can be directly detected in time. Like the presence of lamprophyres and sanukitoids, Δ³³S in granitoids offers a promising proxy to map these reservoirs. If constrained with small age gaps, this isotopic signal could trace the presence and evolution of fertile reservoirs through time, across different crustal levels, and form the basis of pre-competitive datasets to guide exploration in underexplored Archean domains of the lithosphere. Although this work focuses on Ni, Cu and precious metals, we argue that it is a practical thought model that could enhance exploration targeting regardless of commodity.

Methods

Sampling

5 cm segments were selected from diamond drill cores (n = 51) of the Gonneville intrusion (n = 29/51), the HHD intrusions (n = 11/51), and the Julimar country rock sequence (n = 11/51). To ensure datasets can be compared, the same physical 5 cm segments of each sample were used across all analytical workflows, sulfur isotope analysis, automated mineralogy, in situ sulfide analysis, and whole-rock geochemistry. Samples include mineralised ore from different horizons within the intrusions and sulfide-rich, unmineralised country rock located 5 m to 5 km from the intrusion footwall. An additional dataset of low-sulfur (0.1–1.73 wt%) samples was analysed for whole-rock geochemistry only (n = 72). This dataset sampled all stratigraphic levels within the Gonneville intrusion: Serpentine 1 (n = 11/72), Serpentine 2 (n = 14/72), Pyroxene 1 (n = 10/72), Pyroxene 2 (n = 8/72), Internal Gabbro (n = 10/72), Leucogabbro (n = 10/72) and Low Nickel Pyroxenite (n = 9/72).

Sulfur isotope analysis

Sample blocks were prepared for each sample and analysed for δ³⁴S and δ³³S using a continuous-flow Elemental Analyser–Isotope Ratio Mass Spectrometer (EA-IRMS) at the Stable Isotope Geochemistry Laboratory at the University of Queensland, following a procedure modified from Baublys et al.³⁶. Sulfide grains were microdrilled to obtain 0.25–0.4 mg aliquots. Samples were combusted at 1150 °C in an Elementar Vario Isotope Cube in sulfur-only mode, with resulting gases analysed by a PrecisION IRMS. Mass fragments 50/48 and 49/48 were used to calculate δ³⁴S and δ³³S, respectively. Isotope values are reported relative to Vienna-Canyon Diablo Troilite (V-CDT), with δ values defined as: δ^xS = [1000[(^xS/³²S) sample/(^xS/³²S) V-CDT − 1], where ^x is ³³ and ³⁴. ∆³³S value (∆³³S = δ³³S* − 0.515 δ³⁴S*), indicating mass-independent sulfur isotope fractionation, was calculated from δ³³S* and δ³⁴S* values, where δ^xS* = 1000 ln[(δ ^xS/1000) + 1], with ^x representing ³³ and ³⁴. Each sample was analysed in duplicate and repeated if the results differed by >±0.4‰. Measurement precision of δ³⁴S, δ³³S and ∆³³S values were calculated as the pooled 2σ standard deviation of the calibration standards (and check standard, if no drift correction was applied) for each analytical session (following Virnes et al.⁷¹). Pooled 2σ uncertainties varied from ±0.14‰ to ±0.25‰ for δ³⁴S, from ±0.15‰ to ±0.18‰ for δ³³S, and from ±0.12‰ to ±0.18‰ for Δ³³S.

Automated minerology

100-µm-thick carbon-coated polished thin sections were prepared from each sample block and analysed using a TESCAN TIMA SEM with four EDAX 30 mm² EDS detectors at the University of Western Australia. In Phase 1, high-resolution back-scattered electron (BSE) images and mineralogical maps were collected with operating conditions of 25 kV, 112 μm aperture, and 10 nA beam current, with a 15 µm pixel step and minimum 1000 cps. In Phase 2, bright-phase scanning used a BSE contrast threshold >60% and mineral EDS analysis collected at 25 kV, 112 μm aperture, and 10 nA beam current, with a 0.5 µm pixel step and minimum 1500 cps. Repeatability of the method was 97.1%, see Supplementary Table 1 for detail. Only PGMs ≥2 µm² were included. Mineral phases were manually reviewed, and ambiguous Pd- or Pt-bearing grains were labelled as “Pd unknown” or “Pt unknown,” and comprise <1.9% of the dataset, Supplementary Material 4 gives an overview of this classification method.

In situ sulfide analysis

Sulfide phases in the same thin sections were analysed by LA-ICP-MS using a Photonmachines, ATLex 300si-x Excite 193 nm Excimer ArF laser connected to an Agilent 7700 ICP-MS located in the National Geosequestration Laboratory at CSIRO in Perth. The instrument was calibrated on NIST610 and NIST612 silicate glasses. The plasma conditions were optimized daily, to obtain highest counts with oxide production (²⁴⁸ThO/²³²Th) remaining below 0.4%. Sulfides were measured by analysing the following 35 isotopes: Al²⁷, Si²⁹, S³³, S³⁴, Mn⁵⁵, Fe⁵⁷, Co⁵⁹, Ni⁶⁰, Ni⁶¹, Cu⁶³, Cu⁶⁵, Zn⁶⁶, Zn⁶⁸, As⁷⁴, Se⁷⁵, Se⁷⁶, Se⁷⁷, Mo⁹⁵, Ru¹⁰¹, Rh¹⁰³, Pd¹⁰⁵, Ag¹⁰⁷, Pd¹⁰⁸, Cd¹¹¹, Sn¹¹⁸, Sb¹²¹, Te¹²⁵, Re¹⁸⁵, Os¹⁸⁹, Ir¹⁹³, Pt¹⁹⁴, Pt¹⁹⁵, Au¹⁹⁷, Pb²⁰⁸, Bi²⁰⁹. The spot sizes were changed between 25 µm and 50 µm diameter spots depending on the target phase. The primary reference material used was UQAC FeS-1, with the secondary reference material FeS-5 used to check the precision and accuracy. The internal standard was set to Fe and the stoichiometric value for the mineral was used. The data reduction was carried out in Iolite⁷⁴.

Whole rock geochemistry

The remainder of each sample block was crushed to <2 mm at ALS, Perth, with a 70% pass rate. A Boyd rotary splitter was used for subsampling, following ALS standard procedures and all samples were pulverised using a tungsten bowl (85% pass rate <75 µm). Whole-rock major oxides and trace elements were analysed at ALS, Perth. Major oxides (SiO₂, Al₂O₃, Fe₂O₃, CaO, MgO, Na₂O, K₂O, Cr₂O₅, TiO₂, MnO, P₂O₅, SrO, BaO) and loss-on-ignition were determined by X-ray fluorescence (ALS technique ME-XRF26) from fused disks. S and C were determined by induction furnace and infrared detection (ALS technique ME-IR08). Trace elements (Ba, Ce, Cr, Cs, Dy, Er, Eu, Ga, Gd, Ge, Hf, Ho, La, Lu, Nb, Nd, Pr, Rb, Sm, Sn, Sr, Ta, Tb, Th, Tm, U, V, W, Y, Yb, Zr) were analysed by ICP mass spectrometry from fused disks (ALS technique ME-MS81). Base metals (As, Cd, Co, Cu, Li, Mo, Ni, Pb, Sc, Zn) were analysed using a four-acid digest followed by and ICP mass spectrometry (ALS technique ME-4ACD81), while semi-metals (As, Bi, Hg, In, Re, Sb, Se, Te, Tl) were measured by aqua regia digestion with ICP mass spectrometry (ALS technique ME-MS42L). Precious metal (Pd, Pt, Ir, Os, Ru, Rh and Au) assays were completed at Intertek-Genalysis, Perth. These were analysed by a 25 g fire assay using a Ni sulphide collector and an ICP mass spectrometry finish (Intertek technique NS25L/MS). A full list of detection limits is provided in Supplementary Datafile.

Mass balance

Mass balance was calculated using the equation: (F_a × Cⁱ_a/C_wr) × 100, where F_a is the weight fraction of phase a, Cⁱ_a is the concentration of element i in phase a, and C_wr is the concentration of element i in the whole rock. Weight fractions (F_a) for pentlandite, violarite, chalcopyrite, pyrrhotite, and pyrite were assigned using the automated mineralogy, based on the total mass percent of each phase in the thin section. The F_a value for pentlandite was cross-checked against the formula (Ni_wr – 0.025)/(Ni_pn), where Ni_wr is the whole-rock Ni concentration, and Ni_pn is the average Ni concentration in pentlandite, see Supplementary Material 3 for detail. Element concentrations (Cⁱ_a) were assigned from in situ sulfide LA-ICP-MS analyses for Ni⁶⁰, Co⁵⁹, Pd¹⁰⁸, Pt¹⁹⁵, Au¹⁹⁷, Te¹²⁵, As⁷⁴, Bi²⁰⁹, Sb¹²¹ and Se⁷⁵. Corresponding whole-rock concentrations (C_wr) were taken from the whole rock geochemistry analyses.