Metal–silicate partitioning of Mo and W at high pressures and temperatures: Evidence for late accretion of sulphur to the Earth

Abstract

In order to place better constraints on the conditions of core formation on Earth and other planetary bodies we have performed experiments to determine the partitioning of Mo and W between liquid Fe-rich metal and liquid silicate at pressures of 1.5–24GPa and temperatures of 1803–2723K. Experiments performed in MgO capsules at 1.5GPa/1923K indicate that Mo is in the +4 oxidation state in the silicate at oxygen fugacities >2log units below the IW (Fe–FeO) buffer. In contrast W6+ is the dominant tungsten oxidation state in the silicate at 1.5GPa/1923K and 1.8–3.3log units below the IW buffer. When our 15 data for pressures between 6 and 24GPa are combined with those of Cottrell et al. (2009) we find evidence neither for a change in oxidation state of W above 6GPa nor for a change in pressure dependence of partitioning in the experimental fO2 range.Metal–silicate partitioning of both Mo and W shows strong dependence on silicate melt composition with both elements becoming more siderophile as the melt becomes more SiO2-rich. Although the trends in the partitioning data can be related to silicate melt composition in terms of the ratio of nonbridging oxygens to tetrahedral cations NBOT we find that use of a regular solution model for the silicate melt results in a significantly better fit to the data.We combined our results with those in the literature to obtain partitioning equations applicable to the Earth. In terms of weight partitioning we define Diwt and (KDi)wt as follows: (DMowt)=[Mo]met[Mo]sil;(DFewt)=[Fe]met[Fe]sil;(KDMo)wt=(DMowt)(DFewt)2;(KDW)wt=(DWwt)(DFewt)3The experimental data, when corrected for compositional effects, yield the following expressions for a pyrolite mantle:log(KDMo)wt=1.44-143T-167PT(0.19)log(KDW)wt=1.85-6728T-77PT(0.24)The value in brackets corresponds to 1 standard error of the fit. These expressions were combined with the continuous accretion model of Wade and Wood (2005) to investigate the constraints which they place on the accretionary process. We find, however, that, for accretionary paths consistent with the silicate Earth contents of Ni, Co, V, Cr and Nb, W should partition twice as strongly into the core as Mo. This is in stark contrast to the estimated core–mantle partition coefficients of ∼40 for W and 90–140 for Mo. Neither changes to the accretionary path nor the assumption of partial disequilibrium can readily alter this result. The answer appears to reside with the identity of one of the light elements in the core.We investigated the effect of S on our accretionary model by adding 2% of this element (consistent with cosmochemical estimates) to the core. If S is added at constant S/Fe ratio throughout accretion the net effect is negligible. If, however, S is added exclusively during the last 10–20% of accretion DMo and DW become consistent with the silicate Earth contents of these elements. Only small additional adjustments to the model are required to accommodate changes in partitioning of Ni, Co, V, Cr and Nb. We conclude that the Mo and W contents of the silicate Earth indicate that S (and other moderately volatile elements) was added to the Earth during core formation but only during the last ∼20% of accretion. This conclusion is the same as that reached by Schönbächler et al. (2010) from the Ag isotopic composition of silicate Earth.