The solubility of ruthenium in sulfide liquid: implications for platinum group mineral stability and sulfide melt–silicate melt partitioning

To more completely assess the primary magmatic origin of Ru–Os–Ir (IPGE) alloys, we conducted experiments to evaluate the effects of T, fO2, fS2 and melt composition on the solubility of Ru in molten Fe–Ni–sulfide. Fe–Ni–S melt+Ru were held in olivine crucibles, and experiments were done in a vertical-tube gas-mixing furnace at 1200–1400 °C for 1–5 days. At constant fO2 and fS2, Ru solubility increases with T, and a similar result is obtained if fO2 is varied parallel to the fayalite–magnetite–quartz buffer (FMQ), with fS2 levels to maintain sulfide liquid saturation. At log fS2 of −1.7, Ru solubility decreases from ∼11 wt.% at log fO2 of −10.8, to ∼0.3 wt.% at log fO2 of −8.1. At a log fO2 of −8.6, a similar reduction in Ru solubility occurred as log fS2 decreased from −0.9 to −3.0. Substitution of Ni for Fe in the sulfide results in an increase in Ru solubility, with values ranging from ∼3 wt.% at Fe/Ni of 36 to ∼10 wt.% at Fe/Ni of 6 (log fO2, fS2 of −9.1, −1.7, respectively). Dilution of Ru with a 1:1 mix of Os+Ir results in a 4- and 10-fold decrease in melt Ru content for alloys with ∼60 and 35 mol% Ru, respectively. For the fO2–fS2 conditions required for sulfide liquid saturation in natural basaltic magmas, pure Ru solubility in molten sulfide is expected to exceed 10 wt.%, and dilution by Os+Ir is still likely to require wt.% levels of Ru for IPGE alloy saturation. Since Ru abundances of ore-grade massive sulfide is <50 ppm, our results would preclude IPGE alloy saturation in the presence of immiscible sulfide liquid. Activity–composition relations determined for Ru in ternary Ru–Os–Ir alloy suggest, however, that the concentration of Ru in molten silicate required for alloy saturation is at or below levels in natural, high Mg igneous rocks, implying such alloys could form in sulfide-undersaturated systems. gative fO2 dependence of Ru solubility in sulfide liquid is opposite that for Ru (and other PGEs) in silicate melt, suggesting that a decrease in fO2 will favor the partitioning of Ru into the sulfide liquid. However, it is not clear how much this effect will be offset by the concomitant reduction in fS2 required to maintain saturation in immiscible sulfide liquid. Comparison of our Ru solubility data with values determined in silicate melt yields apparent sulfide–silicate melt partition coefficients that exceed 107, which is more than 1000× larger than direct measurements on coexisting sulfide–silicate compositions. Possible reasons for this discrepancy are that (1) measured sulfide–silicate partition coefficients are inaccurate due to incomplete phase separation, (2) non-Henryian activity–composition relations at high Ru concentrations, (3) Ru solubilities in silicate melt measured at both high fO2 and in the absence of sulfur cannot be accurately extrapolated to the low fO2 and sulfur-bearing conditions of previous two-liquid partitioning experiments.