Sulfides Under Upper Mantle Conditions - Experiments Under Controlled Sulfur Fugacities

Chris Ballhaus GEMOC Geology Department, A.N.U., ACT 0200, Australia

Sulfides and chalcophile/siderophile elements hold the key for a number of interesting problems in upper mantle geochemistry. For example, the chondritic relative PGE abundances of the primitive mantle become strongly fractionated during mantle partial melting, implying that some PGE are selectively retained in the mantle residue. Whether this can be achieved by a residual monosulfide liquid solution (mls) depends on a number of factors; (1) the relative sulfide/melt PGE partition coefficients and their dependence on sulfur fugacity (ŸS2), (2) sulfur abundance in the mantle, (3) sulfur solubility in basaltic melts at high pressure, and
(4) the mobility of a residual mls in the mantle matrix during partial melting. Another open question relates to the PGE abundances in the upper mantle. Absolute PGE concentrations are too high to reflect equilibrium with the metallic core, yet too low for equilibrium with the late accretionary component that brought the Ni-Co abundances and Ni/Co ratios to the mantle. One possiblity is that the PGE were selectively removed to the earth's core by segregation of a monosulfide liquid solution following accretion of the 'Ni-Co component', while the Ni/Co ratios and Ni-Co abundances were largely left untouched due to silicate/sulfide mass balance constraints.

To investigate such questions, we need to know the composition of a sulfide melt under upper mantle pressure, temperature, ŸS2, and ŸO2 conditions; however, such data are scarce and difficult to obtain. One reason is that in closed-system sulfide-silicate experiments at high pressure, it is difficult to buffer ŸS2. Pt, the capsule material most commonly used for experimentation at mantle conditions, limits ŸS2 to the equilibrium 2Pt + S2 = 2PtS even if the sulfide is not in direct contact with the Pt capsule material. Graphite, an alternative capsule material, limits the maximum oxygen fugacity (ŸO2) to the C-CO2-CO equilibrium (pure CO2-CO fluid), and this in turn constrains maximum ŸS2 via the dependent redox equilibrium 6FeS + 4O2 = 2Fe3O4 + 3S2. For the Pt-PtS equilibrium, the maximum ŸS2 achievable is 4 to 5 log-bars below the pyrrhotite-pyrite equilibrium, and usually lower in the presence of graphite. This is insufficient to study sulfide-silicate or sulfide-metal exchange equilibria sensitive to ŸS2.

To overcome this limitation a steady state experimental technique has been developed where ŸS2 is imposed by hydrogen fugacity (ŸH2). The sulfide-silicate assemblages plus 2 to 3 µl H2O are placed in an open sample container made of silica glass or San Carlos olivine (for mantle studies), or any material inert toward ŸS2 or ŸO2. Above the sample container is placed a welded Pt capsule filled with a strong H2 sink, e.g. Fe2O3, CuO, PtO2, or Ir2O3. As the H2 sink removes H2 from the system, ŸO2 rises according to the reaction 2H2 + O2 = 2H2O, and so does ŸS2, driven by the equilibrium 6FeS + 4O2 = 2Fe3O4 + 3S2. Any ŸS2 sensitive exchange vector that reaches equilibrium before the H2 sink is exhausted can be studied successfully with this technique, although equilibrium cannot be achieved with respect to all parameters.

The technique has been successfully applied to study the solubility of Pt and Pd in Fe1-xS between 450 and 950°C, over the entire range of metal/sulfide ratios in 1C pyrrhotite. It is shown that PGE solubility in crystalline Fe1-xS falls strongly with falling ŸS2, until in stoichiometric FeS Pt and Pd are nearly insoluble at all temperatures. The ŸS2-solubility profiles at given temperature suggest that PGE only replace Fe atoms in monosulfide solid solution (mss) that are surrounded by a minimum number of vacancies. The experimental technique is extended to study the partitioning of Ni and Co between mls and San Carlos olivine, and the solubility of PGE in mls under upper mantle conditions as a function of ŸS2. Preliminary data suggest that the solubilities of all PGE in mls fall strongly with rising metal/sulfur ratios, much like PGE in crystalline mss, and that all partition coefficients are strong functions of ŸS2.