Experimental problems with the measurement of the metal-sulfide-silicate partition coefficients of the Highly Siderophile Elements

The Highly Siderophile Elements (HSEs: the six PGEs plus Re and Au) are defined by their extreme metal/silicate and sulfide/silicate partition coefficients. Accurate values of these partition coefficients as well as of liquid/solid silicate, sulfide and metal partitioning are needed to understand the origin of magmatic sulfide ore deposits, and to constrain models of of core formation in planetary bodies. Partitioning between metal or sulfide and silicate is expected from thermodynamic theory to depend on temperature, pressure, fO2 and fS2, as well as the compositions of all phases, and, unlike the geochemically more familiar partitioning of lithophile trace elements between crystals and silicate melts, these parameters, especially fO2 and fS2, may cause the partition coefficients to vary by orders of magnitude. This large variation makes it difficult to compare studies done under different conditions, so it is essential that the thermodynamics controlling the partitioning be understood. Another problem is when partition coefficients are high, the slightest contamination of the low-HSE phase by the high- HSE phase can obscure the true partitioning behaviour completely. A particular problem has been the presence of `micronuggets' in quenched run products, which may often be due to contamination, for example, in silicate melts due to a persistent suspension of HSE particles at the ppm level, but in other situations may result from exsolution during quenching. In the former case the micronuggets need to be subtracted out to obtain true partition coefficients, in the latter they should be included. It is therefore not always clear whether great differences in reported values of experimentally determined partition coefficients are due to different conditions or micronugget contamination. While advances in microanalytical techniques, particularly laser-ablation ICP-MS, have certainly been helpful, the solution lies in designing experiments that provide an independent test of the results, for example by varying parameters such as fugacities or activities systematically and seeing if the results conform with relations expected from simpler experiments or theory. Conventional reversals are not particularly useful. Some examples will be presented and discussed.