The concentrations of major base metals in sulfide minerals were determined in polished, thick (∼120 μm) sections by wavelength-dispersive spectroscopy (WDS), electron-probe micro-analysis (EPMA) on a CAMECA SX100 instrument at the Research School of Earth Sciences of The Australian National University (ANU, Australia). Analyses were performed at an accelerating voltage of 15 kV and a sample current of 20 nA. Counting times were 5–15 s on background and 10 s (Si, S, and Cu), 30 s (Fe, Co, and Ni), and 60 s (As) on the peak. Matrix effects were corrected using Phi (r) z modeling available on Peak Sight^© software from CAMECA™. Errors on the measured concentrations were calculated as 3σ relative standard deviations (RSD), which typically give <4% for S, <2% for Fe, <7% for Co, <0.8% for Ni, and 0.5–20% for Cu and As. A set of standards (quartz, barite, troilite, metallic Co, metallic Ni, and metallic Cu) were analyzed as unknowns during each analytical session to estimate precision and drift, which was negligible. Major element maps were collected with the same instrument as for spot analyses but were performed at an accelerating voltage of 15 kV and an increased sample current of 40 nA. Some of the maps were collected using a Bruker energy-dispersive spectrometer coupled to the SX100 instrument.

Back-scattered electron (BSE) images and semi-quantitative phase analyses at high spatial resolution were acquired using a Tescan Mira II LMU field-emission scanning electron microscope (SEM) at the University of Lausanne (UNIL, Switzerland). In situ analyses were performed on this instrument by energy-dispersive spectrometry (EDS) using a Penta-FET 3x X-ray detector. BSE images were acquired at a working (sample) distance of 9 mm, an accelerating voltage of 20 kV, and a sample current of ∼0.5 nA, allowing a spatial resolution (spot size) of ∼5 nm. EDS analyses were performed at a working distance of 20–23 mm and an increased sample current of 0.9–1.3 nA to maximize count rates, leading to a spatial resolution of 6–7.5 nm. Acquisition parameters included an energy step of 20 eV, a process time of 5 s to increase signal to background ratio and spectral resolution, and an acquisition time of 1 mn per analysis. Data treatment was made using the Oxford Instruments AZtec software packaging.

As observed in previous studies, EDS analyses of low O contents in Fe-bearing sulfides suffer from large uncertainties, mainly because of the overlap of the Fe L lines onto the O K _α peak. This effect has been quantified and typically returns “false” O concentrations of 0.89±0.09 wt% in O-free standards (Fonseca et al., 2008). Therefore, O in sulfides was not quantified in this study. Instead, the O K _α peak was used as a tracer of the nature of inclusions, the analyses of which were not possible due to compositional overlap effects induced by the host sulfide.

The abundances of chalcophile and highly siderophile trace elements in sulfides were determined in thick sections by laser ablation inductively coupled plasma mass spectrometry (LA-ICPMS) at ANU. This system comprises a UV (λ=193 nm) Excimer laser (Lambda Physik CompEx 110) and an ANU-designed HelEx ablation cell coupled to an Agilent 7700x quadrupole ICPMS. Analyses were performed with a laser set at a 3 Hz pulse rate, 29.5 kV, 50 mJ, and using 28 μm beam diameters. For each laser ablation run, 25–30 s were counted on the carrier gas (background) followed by 35–40 s for signal. The NIST 610 glass standard (Sylvester and Eggins, 1997; Jochum et al., 2011; Kamenetsky et al., 2015) was used to calculate the concentrations of Co, Cu, Zn, Ag, Cd, Re, and Au, using ⁶²Ni as the internal standard. The CANMET po727 FeS standard (Fonseca et al., 2007; Kamenetsky et al., 2015) was used to calculate Ru, Rh, Pd, Os, Ir, and Pt (PGE) abundances, using ³⁴S as the internal standard. In both cases, internal standard concentrations were taken from EPMA data.

For all sulfide analyses, the isotopes used for data processing included ³⁴S, ⁵⁹Co, ^61,62Ni, ^63,65Cu, ⁶⁴Zn, ^99,101Ru, ¹⁰³Rh, ^105,106,108Pd, ¹⁰⁷Ag, ¹¹¹Cd, ¹⁸⁷Re, ^189,192Os, ¹⁹³Ir, ^195,196Pt, and ¹⁹⁷Au. The potential isobaric interferences produced by metal argides (⁴⁰Ar⁵⁹Co⁺ over ⁹⁹Ru, ⁴⁰Ar⁶¹Ni⁺ over ¹⁰¹Ru, ⁴⁰Ar⁶³Cu⁺ over ¹⁰³Rh, and ⁴⁰Ar⁶⁵Cu⁺ over ¹⁰⁵Pd) were monitored by calculating argide production rates during the analyses of PGE-free, metallic Co, Ni sulfide, and metallic Cu (C-430, Fisher Scientific™ Company), together with the natural samples (e.g., Guillong et al., 2011). The calculated production rates of metal argides during the analytical runs were small, with a range of 0.8–2.5×10^–4. Isobaric interferences were subtracted from the time-integrated signal, based on the measured Co, Ni, and Cu count rates in the sulfides and the argide production rates.

The accuracy and precision of the method were assessed by calculating 1σ RSD from replicate analyses of the NIST 610 glass standard during the runs, using ¹⁹⁷Au as the internal standard. Data were processed using an in-house Excel spreadsheet according to the method of Longerich et al. (1996).

The complete dataset supporting this study is reported in Supplementary Table S1. For simplicity in text and figures, I refer to the fringes and central parts of the veins by using the terms “rims” and “cores”, which are more conventionally used for single-grain characterization. Sulfides in the Type 1A orthopyroxenite veins from Kamchatka are distributed in a similar manner, whether the vein contain glass (sample Av55; Figures 2A–E) or not (sample Av20; Figure 2F). Small sulfides are abundant at vein rims (Figures 2A,B,F), whereas sparse, larger sulfides can be found in the vein cores (Figures 2C–F). A similar grain coarsening trend from vein rims to vein cores has been previously identified for opx (Bénard and Ionov, 2012; Bénard and Ionov, 2013; Bénard et al., 2022), which can also be accompanied by the presence of more abundant glassy areas toward the core. The latter glass enrichment trend is more obvious in the Type 1A orthopyroxenite vein from West Bismarck (sample 67-02E(3); Figures 2G,H), where sulfides are generally less abundant than in Kamchatka veins. While most sulfides appear as globules, which are frequently associated with small vugs or traces of glass, West Bismarck veins are characterized by the presence of sulfide-rich vesicles in glassy areas as well (Figure 2H). Alteration of sulfides is more pronounced in those vesicles, but secondary transformation of globules also occurs when they are associated with vugs (Figures 2B,D,H). EDS analysis reveals that the secondary assemblages formed from sulfides are essentially composed of Fe oxyhydroxide (FeO(OH)) with accessory magnetite (Supplementary Figures S1, S2 and Supplementary Table S1).

All sulfides in the Kamchatka veins appear as homogeneous phases from BSE images and element mapping, which also corroborate the common association of this phase with glass, even when it is included in opx (Figures 3A–G and Supplementary Figure S3). EDS characterization of the sulfides confirm their homogeneity at the sub-micron scale, even when this phase appears in opx-hosted MI (Supplementary Figures S3, S4). The only exception to this tendency is found in the Kamchatka glass-free vein, which has homogeneous rim sulfides associated with trails of minute, euhedral oxide grains but heterogeneous core sulfides with exsolutions (Figures 3H,I and Supplementary Figure S3). These exsolutions have already been identified in an earlier study of glass-free veins from Kamchatka, using three-dimensional imaging (Bénard et al., 2011). Here, EDS characterization reveals that these exsolutions are associated with higher Ni contents, which most likely trace pentlandite (Supplementary Figure S3).

Sulfides in the West Bismarck vein display more complex textural features and phase assemblages than those from Kamchatka. Both inclusion-bearing and homogeneous sulfides are found at vein rims (Figures 4A,B), while heterogeneous globules (i.e., unmixed into different sulfide phases) are found at veins rims and in the glassy areas of the vein core (Figures 4C–F). EDS characterization of the heterogeneous globules indicate that they are mostly composed of Fe- and Ni-rich phases in variable abundances, together with a minor Cu-rich phase (Figures 4C–F and Supplementary Figure S5). Element mapping confirms the relative abundances of the different phases in the heterogeneous globules (Figures 4G–J and Supplementary Figure S6). EDS characterization of inclusion-bearing rim sulfides allow identifying these inclusions as oxides (Figures 5A–D and Supplementary Figure S7), which are also found in clear association with the Cu-rich parts of heterogeneous globules (Figures 5E–G). Trails of minute, oxide grains were also found, albeit very rarely, in the interstitial glass surrounding some of the West Bismarck vein sulfides (Supplementary Figure S8).

Alloys and native metals were found in different parts of both Kamchatka and West Bismarck veins. Pt alloys were identified on the surface of nearly all the different types of sulfides described above, which include the homogeneous (Figures 6A–D and Supplementary Figure S9) and heterogeneous (Figures 6E–H) globules, as well as the sulfide-rich vesicles (Supplementary Figure S10). Some of these Pt alloys are located near the sulfides and can contain Rh at levels detectable by EDS (Supplementary Figure S10). Metallic Fe was found in curvilinear vugs, which are located in glassy areas and coated with altered sulfides (now found as Fe oxyhydroxide) and Cl-rich amphibole (Supplementary Figures S11, S12; Bénard et al., 2022). Wüstite was also clearly identified in those vugs, among different types of alloys, which are also abundantly found in these areas (Supplementary Figure S13).

All Kamchatka vein globules are monosulfide solid solutions (MSS; (Fe, Ni)_1-xS), which typically form at ca. 1,000–1,100°C (Figure 7A; Craig and Kullerud, 1969; Kullerud et al., 1969). In the Kamchatka glass-free veins, core MSS is dramatically enriched in Ni in comparison with rim MSS (Figure 7A). Bulk sulfide analysis of West Bismarck vein globules by EPMA show that these phases range in composition from high-T MSS (homogeneous globules) to Cu-enriched sulfides (heterogeneous globules), which presumably formed at least down to 900°C (Figure 7A; Craig and Kullerud, 1969; Kullerud et al., 1969).

Further EDS analyses reveal that the heterogeneous globules are made of two MSS phases (one being enriched in Fe and the other in Ni), in addition to the Cu-rich phase (Supplementary Table S1). A comparison with the sulfide crystallization experiments of Fleet et al. (1993) and Fleet and Pan (1994) shows that the compositional range of this Cu-rich phase is similar to that of a (Fe, Cu, Ni)S ternary solid solution ranging between MSS and intermediate solid solution (ISS; (Cu, Fe)S_1-x) (Figure 7A and Supplementary Table S1). This phase forms below 850°C and was named “xSS” by Fleet et al. (1993) (Figure 7A). By comparing EDS spectra, this Cu-rich phase is also identified as a component of inclusion-bearing sulfides, where it appears to be preferentially located around the oxide inclusions (Figures 5C,D,G, Supplementary Figure S7 and Supplementary Table S1).

The bulk compositions of some heterogeneous globules were calculated using EDS analyses of their constitutive, MSS and xSS, as well as phase proportions derived from BSE image processing (i.e., pixel counting with the software PHOTOSHOP from Adobe^® Company; Supplementary Table S1). Combining these calculations with all other phase analyses, the major base metal compositional range of Kamchatka and West Bismarck vein sulfides describes sharp Ni and Cu enrichment trends, consistent with a sulfur depletion relative to the sum of metals in their compositions (i.e., increasing atomic ∑metal/S; Figures 7B,C). Notably, sulfide compositions are consistent with decreasing temperatures of crystallization from vein rims to vein cores, and from the Kamchatka setting to the West Bismarck one (Figures 7A–C).

Some Kamchatka MSS were large enough to perform replicate LA-ICPMS analyses and confirm their homogeneity (Figures 3D–G and Supplementary Figure S3), but most sulfides were relatively small so that the signal integrated for quantification was sometimes as short as ca. 20–30 s. However, the consistency of the data supports the fact that short integration times had an insignificant effect on data quality, likely because of very high signal/background ratios (Figure 8 and Figure 9). Kamchatka vein sulfides with relatively flat LA-ICPMS signals generally have similar, chondrite-normalized patterns (Figures 8A–C; Palme and O’Neill, 2014). Most of these patterns are characterized by a depletion of IPGE relative to PPGE, with a pronounced Ir negative anomaly relative to Os and Ru (Figure 8A). However, this Ir depletion is apparently absent from the rim sulfides of the glass-free veins from Kamchatka, which are inferred to be those formed at the highest temperatures in this study (Figure 7A and Figure 8A).

West Bismarck vein sulfides with relatively flat LA-ICPMS signals generally have chondrite-normalized patterns alike those from Kamchatka, though they differ in a few key points (Figures 8D,E). All West Bismarck vein sulfides are characterized by much lower Pd, with chondrite-normalized Pt/Pd systematically ranging above unity (Figure 8D). Some of the heterogeneous globules were analyzed using LA-ICPMS, which reveals that they have PGE and Re abundances like homogeneous MSS (Figure 8D). However, heterogeneous globules are characterized by higher Au, Ag, and Cu contents (Figure 8D). In the case of Cu, this observation is consistent with the major base metal dataset (Figures 7A,C and Supplementary Table S1).

Some of the sulfides analyzed by LA-ICPMS display compositional evidence for the contributions of Pt alloys, as revealed by positive spikes in the ablation signals and chondrite-normalized patterns (“nugget effect”: Figures 9A–D). In both Kamchatka and West Bismarck sulfides, Pt positive spikes are commonly coupled with higher levels of Rh and Au, and to varying extents, Os and Ir (Figures 9A,C). Some of the largest positive spikes in the ablation signals were used to derive semi-quantitative estimates of the Pt alloy compositions (i.e., on an Fe-free basis), which give 90–93 wt% Pt, 0.1–6 wt% Rh, 2–5 wt% Au, ∼1 wt% Os, and traces of Ir and Ru (Supplementary Table S1).

A wide compositional range in major base metals is preserved on a small scale in Kamchatka and West Bismarck vein sulfides (Figures 7A–C). Since all the sulfides are armored in orthopyroxenite veins, which were previously demonstrated to be melt channels (Bénard and Ionov, 2012; Bénard and Ionov, 2013; Bénard et al., 2018c; Bénard et al., 2022), these phases must have formed from a common, original sulfide liquid genetically linked to LCB. Below, I discuss the compositional relationships between the different vein sulfides to track the evolution trends of their parental liquids. I also use thermodynamic calculations to constrain the physicochemical conditions of their crystallization.

One of the key differences between the Kamchatka and the West Bismarck setting concerns the equilibration temperatures of the spinel harzburgites hosting the orthopyroxenite veins, which are significantly higher in the former (900–1,000°C; Bénard and Ionov, 2012; Bénard and Ionov, 2013; Bénard et al., 2022) than in the latter (750–800°C; Bénard et al., 2018c). This temperature difference is related to varying crustal thicknesses aong the two settings (Bénard et al., 2018a); it has further been confirmed in studies of vein-free harzburgite samples from both Kamchatka (900–1,000°C; Ionov, 2010) and West Bismarck (650–700°C; Bénard et al., 2017a). This translates into undercooling degrees (ΔT) of, respectively, 100–200°C and 300–450°C for the glass-forming, high-Mg# andesite melt when using mineral-glass thermometry data (∼1,100°C; Bénard et al., 2018c, 2022; Figure 10A). Since the equilibration conditions of the ambient sub-arc mantle virtually correspond to the lower-end temperatures for the LCB melt evolution path in the veins (e.g., andesitic glass in Kamchatka versus andesitic and dacitic glasses in West Bismarck; Figure 1A), it is also likely that these varying equilibration conditions have had contrasting effects on the sulfide liquid line of descent.

The undercooling degree of Kamchatka vein glass brackets the tight temperature interval spanned by the liquidus–solidus field for MSS in the Fe-Ni-Cu-S system (Figure 10A; Zhang and Hirschmann, 2016), whereas that of West Bismarck vein glass extends well below the solidus. The xSS and MSS phases in the heterogeneous globules from West Bismarck appear in chemical equilibrium with respect to their metal contents, when compared to experimental data in the Fe-Ni-Cu-S system (Figure 10B; Fleet et al., 1993; Fleet and Pan, 1994). Therefore, the calculated, bulk compositions of these heterogeneous globules most likely represent a precursor sulfide phase, which subsequently unmixed because of the lower temperature conditions characterizing the West Bismarck sub-arc mantle (Fleet, 2006). Therefore, it is unsurprising that these heterogeneous globules, as well their constitutive xSS and oxides, were not found in Kamchatka veins (Figure 3).

Collectively taken, sulfide compositions describe Ni and Cu enrichment trends at decreasing sulfur contents (Figures 10C,D). These trends are consistent with sharp increases in the atomic (Ni+Cu)/∑metal and ∑metal/S from the Fe-rich MSS at the rims of Kamchatka glass-free veins to the West Bismarck heterogeneous globules (Figures 7B,C). These two sulfide species represent the end-members formed at the highest and lowest temperatures, respectively (Figure 7A). First, increasing ∑metal/S in MSS is expected to depress the solidus of this phase further than what is shown in Figure 10A (Ballhaus et al., 2001; Ballhaus et al., 2006; Zhang and Hirschmann, 2016). Second, increasing (Ni+Cu)/∑metal and ∑metal/S in MSS provide evidence for the nature of the sulfide liquid line of descent (e.g., Ballhaus et al., 2001; Mungall et al., 2005).

The Ni and Cu enrichments in homogeneous MSS result of progressive cooling (presumably because of increasing defect concentrations in the solid solution), which is consistent with phase relations in the Fe-Ni-S system (Figure 7A; Kullerud et al., 1969; Ballhaus et al., 2001). For the ∑metal/S in MSS to increase as in West Bismarck veins (i.e., up to ∼1.15, whereas all sulfides in the glass-free Kamchatka vein range at 0.9±0.5; Figures 7B,C), temperature must have dropped well below 900°C. Furthermore, the greatest increases in (Ni+Cu)/∑metal in MSS are experimentally observed for parental sulfide liquids, which originally have the lowest ∑metal/S ranging from ∼0.9 to ∼1 (Ballhaus et al., 2001). Finally, these low-∑metal/S sulfide liquids will likely solidify in the stability field of ISS; that is, they will not produce eutectic, Cu-rich phase assemblages formed at even lower temperatures (Ballhaus et al., 2001).

From the considerations above and the low ∑metal/S of Fe-rich MSS formed at the highest temperatures in this study (0.85–0.9; Figures 7A,B), it can be implied that the compositional evolution of LCB vein sulfides reflects parental liquids with originally low ∑metal/S of ∼0.9. Such sulfide liquid is often referred to as “oxidized” since it is mainly O that substitutes for S (see for instance the experiments in the MS-12 system from Ballhaus et al., 2001). It is thus likely that such sulfide liquid will also fractionate some oxides, reflecting fO₂ conditions approaching the FMQ buffer during crystallization (e.g., Naldrett, 1969). Some of these oxides appear to have formed at an early stage, as observed near the high-T, Fe-rich MSS (Figure 3H). Oxides also occur in association with Ni-rich MSS and Cu-rich xSS, suggesting that they formed all along the sulfide liquid of descent with decreasing temperature (Figures 4B, 5F and Supplementary Figures S7, S8). These features are consistent with those of Ni-Cu-PGE ore samples, with magnetite occurring as a cotectic phase from MSS to ISS formation (Duran et al., 2020); although the oxides near LCB sulfides were generally too small to be analyzed, these are most likely magnetite. Magmatic chromite grains were also observed in all the veins in this study, but no in clear association with sulfides (e.g., Bénard et al., 2022), as it should be when crystallization occurs along a sulfide-spinel cotectic or during O loss from a sulfide liquid (Fonseca et al., 2008). The elevated Cr/(Cr+Al)≥0.75 of vein chromite also suggest that they rather formed at higher temperatures than MSS, most likely by direct precipitation from the silicate melt fraction in LCB magmas (Bénard and Ionov, 2012; Bénard and Ionov, 2013; Bénard et al., 2018c; Bénard et al., 2022).

Further constraints on the fO₂ and fS₂ conditions prevailing throughout the LCB sulfide liquid line of descent require independent thermodynamic calculations. All the glass-bearing veins in this study contain traces of magmatic olivine, which typically has higher Mg# than in the host spinel harzburgite (0.91–0.93; Bénard et al., 2018c; Bénard et al., 2022). The presence of this phase indicates that the thermodynamic activity of the fayalitic component was fixed during the formation of the orthopyroxenite veins, allowing the use of the olivine-opx+MSS equilibrium for fO₂ and fS₂ calculations. Using Equations 41 and 46 from the calibration of Eggler and Lorand (1993) for this equilibrium, I calculate fS₂ conditions for the LCB vein crystallization, assuming 1) −1≤ΔlogfO₂[FMQ]≤0 inferred from the sulfur contents at sulfide/sulfate saturation (SCSS) measured in the S-poor, vein interstitial glass (Figures 1A,B; Smythe et al., 2017); and 2) an Mg# varying from 0.84 to 0.93 in olivine, which matches the full compositional range of the coexisting vein opx (Bénard et al., 2018c; Bénard et al., 2022). Results show that at 0.5 GPa, which is likely the minimum pressure for the formation of the veins (Figure 1A; Bénard et al., 2018c; Bénard et al., 2022), the crystallization sequence from Fe-rich MSS (ca. 1,050–1,100°C) to xSS (≤850°C) will occur at relatively low fS₂ ranging from −3.5 to −1 bars in log units (Figure 10E). When compared with a series of univariant reaction curves, these fS₂ values place the vein sulfide crystallization conditions close to the Pt-PtS buffer (Figure 10E; Barin, 1995). The latter point is consistent with the widespread distribution of Pt alloys near all types of sulfides in the glass-bearing veins from both Kamchatka and West Bismarck (Figure 6 and Supplementary Figures S1, S9, S10).

From the inferred, low ∑metal/S of the parental sulfide liquids in LCB, it appears that heterogeneous globules made of MSS and xSS in West Bismarck veins are aggregates of different crystallization products from these liquids. Therefore, none of the phases analyzed in this study can be considered as a quenched sulfide liquid, the composition of which must be derived by calculations using experimental data.

In order to estimate the compositions of the parental liquids of vein sulfides, I use two main calculations combined in an iterative approach. First, I estimate the major base metal concentrations in the sulfide liquid in equilibrium with all sulfides analyzed in this study; that is, over temperatures ranging from ∼1,100°C down to ∼900°C using the compositions of homogeneous, Ni-rich MSS, and at ca. 750–850°C using the calculated compositions of bulk heterogeneous globules (i.e., MSS+xSS; Figure 7A). Thus, I hypothesize that these bulk heterogeneous globules constitute a cumulate mixture of different sulfide phases. For this first step, I use the metal partitioning data between MSS and sulfide liquid from the experiments of Fleet et al. (1993) and Fleet and Pan (1994) in the Fe-Ni-Cu-S system (Figure 11A). Second, I estimate O solubility in the calculated sulfide liquids using two relationships: 1) as a function of sulfide liquid composition, temperature, fO₂ and fS₂ using Equation 7 in Fonseca et al. (2008); and 2) as a function of silicate melt and sulfide liquid compositions using Equation 6 in Kiseeva and Wood (2015). In the first case, fO₂ and fS₂ are set using Equations 41 and 46 in Eggler and Lorand (1993), for −1≤ΔlogfO₂[FMQ]≤0 and a Mg# varying from 0.84 to 0.93 in olivine. In the second case, the lower- and upper-bound FeO_t contents of the silicate melt (respectively ∼4 and ∼9 wt%) are derived from the analyses of the high-Mg# andesite vein glass (Bénard et al., 2018c; Bénard et al., 2022).

Results show that O solubility in the calculated sulfide liquids are predicted in a consistent way by both the parameterizations of Fonseca et al. (2008) and Kiseeva and Wood (2015) (Figure 11B). Because of the well-known effect of the presence of Ni on lowering O solubility in sulfide liquids (see Fonseca et al., 2008 and references therein), the calculated concentrations of the latter element drop below ∼2 wt% once Ni exceeds ∼5 wt%, which is the case for LCB sulfide liquids (Figure 11B). The upper O concentrations predicted for sulfide liquids formed in equilibrium with Ni-rich MSS in LCB generally do not exceed ∼1 wt% when inferred from the parameterization of Kiseeva and Wood (2015), which I use for the calculations reported here (Figure 11B and Supplementary Table S1). The low O contents calculated for the sulfide liquids formed from LCB are consistent with the limited amounts and small sizes of primary oxides in clear association with sulfides in the veins (Figures 3H, 4B, 5F and Supplementary Figures S7, S8).

As their daughter MSS, the estimated sulfide liquids are particularly rich in Ni (from ∼25 to ∼35 wt%; Supplementary Table S1), while they experience a pronounced increase in their ∑metal/S with decreasing temperature (≥1.2 at ≤1,000°C; Figure 11C and Supplementary Table S1). The latter trend was not reproduced in differentiation experiments of sulfide liquids with initially low ∑metal/S (0.9–1), as this parameter was only reaching ca. 1.1–1.2 at 950°C with progressive MSS fractionation (Ballhaus et al., 2001). It should be noted, however, that the experimental design in Ballhaus et al. (2001) included: 1) original charge compositions calculated to simulate equilibrium with a tholeiitic basalt (i.e., fertile mantle sources); 2) fO₂ left unbuffered during the experimental runs, though it was probably kept to a minimum, and fS₂ intrinsically defined by the compositions of sulfides; and 3) no run temperatures below 880°C. In experiments where fO₂ and fS₂ were respectively constrained at the FMQ and Pt-PtS buffers (Mungall et al., 2005), elevated ∑metal/S like those calculated in this study were encountered (Figure 11C). In this Fe-Ni-Cu-S-O system, Ni enrichment can further occur in sulfide liquids (i.e., this element is even more incompatible), when fS₂ conditions range from the Pt-PtS buffer to even lower partial pressures (Craig and Kullerud, 1969; Ebel and Naldrett, 1996; Mungall et al., 2005).

The estimated sulfide liquid line of descent in LCB is therefore consistent with the evolution of an originally Ni-rich and Cu-poor sulfide liquid at fO₂ and fS₂ conditions, which respectively range near and below the FMQ and Pt-PtS buffers. This conclusion is further supported by independent thermodynamic calculations using the olivine-opx+MSS equilibrium (Figure 10E; Eggler and Lorand, 1993). The sulfide liquid line of descent in LCB is driven by the crystallization of MSS, which typically ranges from Fe-rich to Ni-rich with decreasing temperature (e.g., Craig and Kullerud, 1969; Fleet et al., 1993; Fleet and Pan, 1994; Ebel and Naldrett, 1996). Metal enrichment and S depletion characterize the sulfide liquid line of descent in LCB, where the final crystallization products are likely xSS or ISS (e.g., Ballhaus et al., 2001).

As a comparative test of the Ni-rich and Cu-poor compositions of sulfide liquids inferred from the orthopyroxenite vein sulfides, I use data for erupted LCB from Cape Vogel (Papua New Guinea; König et al., 2010), which contain 60±8 ppm Co, 345±94 ppm Ni, and 38±12 ppm Cu, for 10.2±0.2 wt% FeO_t and 16±3 wt% MgO with a Mg# of 0.73±0.03 (all 1σ, the cumulate rock “PNG14” is excluded). An average sulfide liquid composition is calculated in equilibrium with the Cape Vogel LCB using sulfide liquid/silicate melt partition coefficients for Fe, Co, Ni, and Cu (Gaetani and Grove, 1997) and the O solubility model in sulfide liquid of Kiseeva and Wood (2015), for fO₂ and fS₂ conditions inferred at atmospheric pressure, 1,300–1,340°C and −1≤ΔlogfO₂[FMQ]≤0 (Eggler and Lorand, 1993; Mungall and Brenan, 2014). This calculation provides an average sulfide liquid composition of 30.54±0.05 wt% Fe, 0.2211±0.0005 wt% Co, 29.32±0.01 wt% Ni, 1.17±0.02 wt% Cu, and 0.35±0.01 wt% O (all 1σ) with S fixed at 38.40 wt% (average S data from (Fleet et al., 1993) and (Fleet and Pan, 1994)). This estimate agrees with earlier calculations based on orthopyroxenite vein sulfides, although the latter are more imprecise in nature given the wide compositional range of MSS therein (Figure 11C). Both calculations imply Ni-rich and Cu-poor compositions for sulfide liquids formed from magmas derived from spinel harzburgite sources, for instance in comparison with those calculated from lherzolite-derived primitive MORB, which contain only ~25 wt% Ni but up to ~8 wt% Cu (Kiseeva and Wood, 2015).

Note that it is important to find similar compositions for sulfide liquids calculated from MSS in orthopyroxenite veins, on the one hand, and from erupted LCB, on the other hand. Indeed, there is a priori no guarantee that the parental sulfide liquids formed in the LCB veins did not experience one or several immiscibility gaps before MSS crystallization, even those preserved as homogeneous globules. Liquid immiscibility in the Fe-Ni-Cu-S-O system has been suggested in pioneering field studies (e.g., Distler et al., 1986). However, more recent experimental works have shown that this process was extending below the MSS liquidus, but only for sulfide liquids with originally high ∑metal/S>1, and with the main involvement of a fractionating, Cu-rich phase (Ballhaus et al., 2001). For instance, there is no evidence for such liquid immiscibility gap in the ∑metal/S of homogeneous vein MSS, which stays in a narrow range of ca. 0.9–1 (Figure 7B; Ballhaus et al., 2001). Therefore, calculations of sulfide liquid compositions in equilibrium with Cape Vogel LCB confirm that immiscibility gaps unlikely altered the vein MSS compositions formed at least down to ∼950°C (Supplementary Table S1).

Estimates of the P-T-fO₂-fS₂ conditions and compositional range of the sulfide liquid line of descent in LCB has implications for the evolution of sulfur and precious metals in the spinel harzburgite sources of these magmas. First, it could be expected that these sources will not contain more than ∼50 ppm S and will record relatively low fS₂, given the observations in studies of spinel harzburgites (Luguet et al., 2007; Lorand et al., 2013). This stands, however, only when assuming that no metasomatism of the spinel harzburgite sources occurred, which can possibly involve oxidized, S-rich fluids or melts in a subduction zone setting (e.g., Bénard et al., 2018b).

Spinel harzburgites from Kamchatka and West Bismarck, which are the putative sources and/or residues formed in equilibrium with LCB, record −0.5≤ΔlogfO₂[FMQ]≤+1.5 (Bénard et al., 2018a; Bénard et al., 2018b; Bénard et al., 2018c). High-T fS₂ conditions during LCB and orthopyroxenite formation can be calculated at 0.5–1 GPa and 1,000–1,400°C, using Equation 6 in Mungall and Brenan (2014) and the (Fe₂O₃)_t contents in silicate melts derived from pMELTS (Ghiorso et al., 2002) calculations for LCB differentiation (Figure 1A; Bénard et al., 2018c; Bénard et al., 2022). The results indicate that fS₂ at sulfide/sulfate saturation ranges from 1.6 to 3.9 bars at ΔlogfO₂[FMQ]=+1.5 and from −1.3 to 1.1 bars (both in log units) at ΔlogfO₂[FMQ]=−1. The estimated fS₂ range at relatively low fO₂ overlaps with the results based on the olivine-opx+MSS equilibrium for the veins (Figure 11E). In the first part of this study, it was shown that the parental LCB melts of the veins were originally equilibrated at ΔlogfO₂[FMQ]∼+1.5, allowing them to carry up to ∼2,600 ppm S (opx-hosted MI data; Figures 1A,B; Bénard et al., 2022). Assuming that these relatively high fO₂ conditions were also prevalent during the partial melting events forming LCB, it is more likely that sulfides, alloys and platinum-group minerals (PGM) in their spinel harzburgite sources were partially consumed by the formation of these melts (see Mungall and Brenan, 2014 and references therein). Consequently, these calculations also imply that a dramatic drop in fS₂ conditions must have occurred during vein formation, which involved LCB melt fractionation and local fO₂ reduction during S⁶⁺-Fe²⁺ redox exchange with the host mantle (Figure 1B; Bénard et al., 2022).

An originally sulfide-undersaturated state has been proposed for the formation of “second-stage” melts such as boninites (Hamlyn and Keays, 1986; Keays, 1995; Valetich et al., 2019). In the case of the orthopyroxenite veins in this study, a comparison of S abundances in opx-hosted MI and in the interstitial glass is indeed consistent with an originally sulfide-undersaturated state of the LCB melt intruding the sub-arc mantle (Figure 1A and Figure 12A; Bénard et al., 2018c; Bénard et al., 2022). Intrusion and cooling caused saturation in S-bearing, hydrothermal fluids in the veins, shortly after (or concurrently with) the onset of magmatic sulfide formation in the derivative high-Mg# andesite and dacite (Figure 12A; Bénard et al., 2018c; Bénard et al., 2022). Caution must be taken, however, when extrapolating the original sulfide-undersaturated state of vein-forming LCB in the shallow sub-arc mantle (≤1.5 GPa; Bénard et al., 2017a; Bénard et al., 2018a) to source conditions because of the strong effect of pressure on S solubility in silicate melts (e.g., Mavrogenes and O’Neill, 1999; Moretti and Baker, 2008). However, much of the data collected on the Kamchatka and West Bismarck veins point towards original LCB melts produced by partial melting at relatively low-P conditions of ca. 1–1.5 GPa (Figure 10A; Bénard and Ionov, 2012; Bénard and Ionov, 2013; Bénard et al., 2018c; Bénard et al., 2022), suggesting at least that the decompression effect on S solubility between the sources and the final emplacement sites was minimal.

An elevated oxidation state (ΔlogfO₂[FMQ]∼+1.5) and sulfide undersaturation during partial melting of spinel harzburgite at low pressure, that is, in the theoretical stability field of Pt alloys (Mungall and Brenan, 2014), provides key conditions for the transport and concentration of sulfur and precious metals. The low degrees of melting inferred for spinel harzburgite melting (F≤5% at 1 Gpa and ≤1,380°C; Bénard et al., 2017a; Bénard et al., 2018c; Bénard et al., 2022) ensure a natural concentration process for S in primary LCB. On the basis of differentiation models for these magmas calculated by Bénard et al. (2018c), ∼700 ppm S can be estimated in the primary melt (Figure 12A). This translates into ca. 7–35 ppm S in the spinel harzburgite sources when assuming F=1–5%. Note that for all these calculations, sulfur is assumed to be purely incompatible, which is consistent with the fact that this element is mainly dissolved as S⁶⁺-bearing species in the original silicate melts (Figure 1B; Bénard et al., 2022). Sulfide-poor spinel harzburgites contain alloys and PGM (e.g., Lherz peridotites; Luguet et al., 2007), which can be partially destabilized at relatively high fO₂ conditions to be transported by the primary LCB melt (Mungall and Brenan, 2014). The presence of minor S²⁻ in this melt, even at ΔlogfO₂[FMQ]∼+1.5 (Figure 1B), will also allow the formation of dissolved PGE-S²⁻ complexes, as it has been inferred for Pd or Ru (Laurenz et al., 2013; Mungall and Brenan, 2014). After being extracted from their spinel harzburgite sources, further dissolution of sulfides, alloys and PGM, either encapsulated in mantle minerals or located at grain boundaries, can occur in LCB during reactive melt percolation (Bénard and Ionov, 2012; Bénard and Ionov, 2013), for instance when these melts percolate through rocks with which they are out of equilibrium (e.g., lherzolites).

Key signatures of harzburgite sources can be found in the major and trace metal compositions of vein sulfides, in particular their elevated Ni contents and sometimes very low Pd abundances with characteristic chondrite-normalized Pt/Pd>1 (Figures 8D, 9C, 10C). It has long been known that elevated Ni contents characterize sulfides formed from mafic and ultramafic magmas, with a negative relationship observed between Cu/(Cu+Ni) in sulfides and the MgO contents of the parental rocks (e.g., Naldrett and Cabri, 1976). At a given set of P-T conditions, the lherzolite-harzburgite source transition is not only accompanied by a drastic decrease in melt productivity (Hirschmann et al., 1999; Bénard et al., 2018c; Bénard et al., 2022), but also by a pronounced shift in primary melt compositions with increasing SiO₂ and MgO contents (Falloon and Danyushevsky, 2000; Bénard et al., 2018c; Bénard et al., 2022). Ni and Cu being respectively compatible and incompatible during mantle melting, the former element remains in the residues but not the latter, so that spinel harzburgites typically host higher Ni and lower Cu contents than spinel lherzolites (e.g., Lorand et al., 2013; Bénard et al., 2021). Furthermore, MgO-rich LCB are expected to be enriched in Ni in comparison with basalts, owing to the decrease in Ni partition coefficients between olivine and silicate melt with increasing MgO contents of the melt (e.g., Wang and Gaetani, 2008). These considerations explain why sulfide liquids derived from LCB are so enriched in Ni but not in Cu (Supplementary Table S1).

Low Pd contents at constant Pt levels, together with characteristic chondrite-normalized Pt/Pd>1 in some LCB sulfides must also be inherited from their spinel harzburgite sources (Figure 8D and Figure 9C). Knowing that Pd can concentrate in Cu-rich phases such as xSS or ISS (e.g., Liu and Brenan, 2015), it is important to note here that the variability in Pt/Pd is clearly unrelated to the presence of such phase, since homogeneous and heterogeneous globules in West Bismarck veins were analyzed in this study and both display this compositional feature (Figure 8D and Figure 9C). Instead, chondrite-normalized Pt/Pd>1 appears to be a characteristic signature of the West Bismarck vein sulfides (Figure 8 and Figure 9). So far, no MSS/sulfide liquid equilibration experiment has inferred any significant fractionation between Pd and Pt, with PPGE being always similarly incompatible in MSS (Fleet et al., 1993; Li et al., 1996; Ballhaus et al., 2001; Bockrath et al., 2004; Mungall et al., 2005; Ballhaus et al., 2006; Liu and Brenan, 2015). Therefore, variable Pt/Pd observed among Kamchatka and West Bismarck vein sulfides must reflect the PGE signatures of their parental sulfide liquids, since Pd is not significantly hosted by any of the Pt alloys identified in this study (Supplementary Table S1). In addition, the sulfide liquid/silicate melt partition coefficients of Pt and Pd are similar, laying both near 10⁵ (Mungall and Brenan, 2014), whereas the co-precipitation of Pd-free, Pt alloys can only lower Pt/Pd in coexisting MSS.

From the points discussed above, chondrite-normalized Pt/Pd ranging at ca. 10–30 in the West Bismarck vein sulfides must result from the equilibration of the original LCB silicate melts with Pd-depleted mantle rocks (Figure 8D). Such Pd depletion is a characteristic feature of many spinel harzburgites, either sampled as magma-hosted xenoliths or exhumed rocks in peridotite massifs (see Lorand et al., 2013 and references therein). Multiple melt extraction events forming spinel harzburgite residues are responsible for this Pd depletion, since this element is much more soluble than other PGE in silicate melts for a given set of P-T-f conditions (see Mungall and Brenan, 2014 and references therein). Though chondrite-normalized Pt/Pd>1 in magmatic MSS could thus be a key feature for formally identifying spinel harzburgite sources, fluctuations of this ratio among Kamchatka and West Bismarck LCB sulfides must also reflect the variable compositions of these sources, as can be expected from the wide, PGE compositional range of spinel harzburgites (see Lorand et al., 2013 and references therein). A possible origin for these Pt/Pd variations are Pd-rich sulfides armored in mantle-derived silicate or oxide grains, which do not react during the melting events forming spinel harzburgites, such as is the case for olivine under low-P and/or hydrous conditions at subduction zones (Gaetani and Grove, 1998).

The Ni-rich nature of sulfide melts formed from LCB has important implications for the evolution of precious metals when these magmas differentiate, for instance through crystal fractionation. Regarding phase relations, the first effect of the Ni-rich nature of the LCB sulfide liquids is to shift the cotectics close to the Fe(±Ni, Cu)-S join in the triangular Fe(±Ni, Cu)-S-O plane, for instance in comparison with the Fe-S-O system (Figure 12B; Naldrett, 1969; Mungall et al., 2005). A second effect is that the low O solubility resulting from high Ni contents places the estimated LCB sulfide liquids on the MSS side of the magnetite-MSS join (Figure 12B; Mungall et al., 2005; Fonseca et al., 2008), consistently with the paucity of primary oxides clearly associated with sulfides in the veins. It also appears that the estimated, sulfide liquid line of descent in LCB leads the final, xSS-forming products close to the metallic Fe-MSS join (Figure 12B). This is consistent with the identification of metallic Fe in vein vugs coated with altered, hydrothermal sulfides, while the presence of wüstite in the same fluid-derived spaces suggests that the eutectic of the Fe(±Ni, Cu)-S-O system was also reached (Figure 12B and Supplementary Figures S11, S12, S13). Finally, the trajectory of the sulfide liquid line of descent is parallel to the Fe(±Ni, Cu)-S join, which suggests that decreasing fO₂ and fS₂ during partial crystallization and fractionation, assuming a closed system, were mainly driven by cooling upon LCB intrusion in the sub-arc mantle.

The phase relations for the sulfide liquid line of descent in LCB described above provide further insights into the origin of oxides in the Kamchatka and West Bismarck veins. As mentioned above, some of the minute oxide grains found in association with late-stage xSS likely formed by cotectic precipitation, while those in the glass near the rims of MSS could be related to O diffusion out of the crystallizing sulfide liquid into the surrounding silicate melt (Figures 3H, 4B, 5F and Supplementary Figures S7, S8; Naldrett, 1969; Fonseca et al., 2008). The effects of the low O and high Ni contents in the sulfide liquids formed from LCB will be reinforced by the potential loss of O to the surrounding, high-Mg# andesite melts. Such rapid O loss will further contribute to moving the line of descent of these sulfide liquids on the MSS side of the MSS-magnetite join. Some of the magnetite grains identified on the rims of sulfides, however, could also be related to late-stage oxyhydration and FeO(OH) formation (Supplementary Figures S1, S2; Lorand, 1990). The common association of these alteration products with vugs and sulfide-rich vesicles (Figure 2H and Supplementary Figure S2; Bénard et al., 2022) suggests that they formed through a deuteric process involving high-T, hydrothermal fluids directly formed from LCB, most likely from ∼1,000°C and shortly after (or concurrently with) the onset of magmatic MSS formation (Figure 10A and Figure 11C; Bénard et al., 2022).

During the ascent and emplacement of LCB as opx-rich dykes or sills, the decrease of fS₂ resulting from decompression and cooling leads to the precipitation of alloys together with sulfides, as observed in the orthopyroxenite veins (Figures 6, 9 and Supplementary Figures S9, S10; Mungall and Brenan, 2014). As calculations based on the fS₂ parameterization of Mungall and Brenan (2014) suggest (see above), decreasing temperature and Fe oxide contents of the silicate melts through fractionation are key factors controlling fS₂ at sulfide/sulfate saturation. Indeed, this parameter decreases by ca. 1.6–1.9 log units throughout the LCB liquid line of descent when assuming a fixed internal fO₂ buffer (e.g., either ΔlogfO₂[FMQ]=−1 or ΔlogfO₂[FMQ]=+1.5). This indicates that the “low-Fe” fractionation trend of LCB (see Arculus, 2003 and references therein) exerts a first-order control on decreasing fS₂ during cooling of these magmas. In the specific case of the Kamchatka and West Bismarck veins, it appears that fO₂ variations are also crucial in triggering sulfide and alloy precipitation, since the originally S⁶⁺-bearing LCB melt was locally converted into S²⁻-bearing, high-Mg# andesite and dacite derivatives through redox reactions with the surrounding, Fe²⁺-bearing mantle minerals (Figures 1A,B; Bénard et al., 2018c; Bénard et al., 2022). In the case of a local decrease in the LCB melt oxidation state from ΔlogfO₂[FMQ]=+1.5 to ΔlogfO₂[FMQ]=−1 during intrusion, the resulting drop in fS₂ through cooling and differentiation could be as high as ca. 4.4–4.8 log units.

From the decreases of fO₂ and fS₂ during the formation of sulfides in LCB dykes and sills, it can be further inferred that PGE-bearing alloys must form from the onset of sulfide crystallization, or even at greater temperatures. This appears to be consistent with the widespread dissemination of alloys in the veins (Figures 6, 9 and Supplementary Figures S1, S9, S10), but also the variable Ir contents in vein sulfides; the greatest levels in this element appear in Fe-rich MSS formed at the highest temperatures along the sulfide liquid line of descent (Figure 8A). The orthopyroxenite vein sulfides are otherwise more enriched in Os and Ru than those found in high-Mg basaltic lavas from Kamchatka (Figure 8; Zelenski et al., 2017). Taken collectively, these pieces of evidence suggest that the original sulfide liquids were certainly more enriched in IPGE than what can be expected from the highly siderophile trace element patterns of most sulfide grains analyzed in this study. Even though direct formation of Os-Ir-Ru alloys from boninites has been suggested (Peck et al., 1992), the results in this study are also consistent with the precipitation of IPGE-bearing, Pt alloys before and during sulfide formation (Supplementary Table S1). The early saturation of LCB in such Pt alloys agrees with the nature of these phases when they are found in spinel harzburgites, which can be otherwise free of base metal sulfides (e.g., Pt-Ir±Os alloys; Luguet et al., 2007).

Other precious metals, such as Au and Ag, appear at relatively low levels in LCB, for instance in comparison with high-Mg basaltic lavas from Kamchatka (Figure 8; Zelenski et al., 2017). As in the case of Cu, this presumably results from the originally low abundances of Au and Ag in the depleted, spinel harzburgite sources of LCB (Figure 8A). These metals are expected to become enriched only in the very late-stage LCB sulfides (i.e., xSS or ISS), according to their partitioning behavior (Liu and Brenan, 2015); this point is supported by LA-ICPMS analyses of xSS-bearing, heterogeneous globules in this study (West Bismarck veins; Figure 8D).

Finally, it can also be expected from phase relations and the inferred fS₂ drop during LCB differentiation that a variety of metallic alloys and native metals will form down to the solidus temperatures of sulfide liquids derived from these magmas, and potentially further below (Figure 5, Figure 6 and Supplementary Figure S7). In addition, the separation of high-T, hydrothermal fluids from ∼1,000°C, proven to be enriched in the volatile elements S and Cl (Bénard et al., 2018c; Bénard et al., 2022), can lead to the efficient solubilization of precious metals under the form of metal-Cl⁻ or metal-S²⁻ complexes to concentrate them in vugs or sulfide-rich vesicles (Supplementary Figures S10, S11, S12; Sullivan et al., 2022a; Sullivan et al., 2022b). Overall, the abrupt sulfide and alloy precipitation processes from low-S solubility, high-Mg# andesite and dacite melts formed by the differentiation of primary LCB, which chronologically match the formation of hydrothermal fluids (Figure 1A), allow the efficient capture of precious metals in localized, opx-rich zones.