Evidence for a sulfur-undersaturated lunar interior from the solubility of sulfur in lunar melts and sulfide-silicate partitioning of siderophile elements

Abstract

Sulfur concentrations at sulfide saturation (SCSS) were determined for a range of low- to high-Ti lunar melt compositions (synthetic equivalents of Apollo 14 black and yellow glass, Apollo 15 green glass, Apollo 17 orange glass and a late-stage lunar magma ocean melt, containing between 0.2 and 25 wt.% TiO2) as a function of pressure (1–2.5 GPa) and temperature (1683–1883 K). For the same experiments, sulfide-silicate partition coefficients were derived for elements V, Cr, Mn, Co, Cu, Zn, Ga, Ge, As, Se, Mo, Sn, Sb, Te, W and Pb. The SCSS is a strong function of silicate melt composition, most notably FeO content. An increase in temperature increases the SCSS and an increase in pressure decreases the SCSS, both in agreement with previous work on terrestrial, lunar and martian compositions. Previously reported SCSS values for high-FeO melts were combined with the experimental data reported here to obtain a new predictive equation to calculate the SCSS for high-FeO lunar melt compositions. Calculated SCSS values, combined with previously estimated S contents of lunar low-Ti basalts and primitive pyroclastic glasses, suggest their source regions were not sulfide saturated. Even when correcting for the currently inferred maximum extent of S degassing during or after eruption, sample S abundances are still > 700 ppm lower than the calculated SCSS values for these compositions. To achieve sulfide saturation in the source regions of low-Ti basalts and lunar pyroclastic glasses, the extent of degassing of S in lunar magma would have to be orders of magnitude higher than currently thought, inconsistent with S isotopic and core-to-rim S diffusion profile data. The only lunar samples that could have experienced sulfide saturation are some of the more evolved A17 high-Ti basalts, if sulfides are Ni- and/or Cu rich.Sulfide saturation in the source regions of lunar melts is also inconsistent with the sulfide-silicate partitioning systematics of Ni, Co and Cu. Segregation of significant quantities of (non)-stoichiometric sulfides during fractional crystallization would result in far larger depletions of Ni, Co and Cu than observed, whereas trends in their abundances are more likely explained by olivine fractionation. The sulfide exhaustion of the lunar magma source regions agrees with previously proposed low S abundances in the lunar core and mantle, and by extension with relatively minor degassing of S during the Moon-forming event. Our results support the hypothesis that refractory chalcophile and highly siderophile element systematics of low-Ti basalts and pyroclastic glasses reflect the geochemical characteristics of their source regions, instead of indicating the presence of residual sulfides in the lunar interior.