Effect of sulfide equilibration on the oxidation state of the

The experimental results support a minimum partition coefficient for H par-titioning between a molten silicate and a liquid sulfide of ∼0.2, which implies that there will be some amount of hydrogen that is removed into the core. For H₂O to be an effective mantle oxidising agent, however, the resulting oxygen would have had to have been left behind in the silicate and not also partitioned into the core-forming phase. Hirschmann et al. (2012) have speculated on the likely effectiveness of oxidation by H₂O with respect to hydrogen partitioning into metallic iron. The authors point out that at the low oxygen fugacities and high pressures of metal-silicate equilibration, it is likely hydrogen would have already been present as H₂, therefore not necessarily associated with any oxy-gen to oxidize the mantle. Furthermore, reaction of H₂O with iron metal, at least at upper mantle conditions, can only produce more FeO in the mantle

and will not raise the ferric iron content and thus the oxygen fugacity of the upper mantle. However, loss of H₂ to the core would lower the magmatic H₂/ (H₂+H₂O) ratios, which would have driven the magma to more oxidised condi-tions. They conclude, however, that the presence of metallic iron buffers the system, resisting oxidation, and that therefore the effect of H₂ removal would have been relatively small (Hirschmann et al., 2012).

Even if metallic iron could have efficiently removed hydrogen to the core, the upper mantle is still several orders of magnitude more oxidised than the co-existence of metal permits; i.e., hydrogen removal must have been ongoing even at more oxidised conditions, allowing ferric iron to accumulate in the up-per mantle. In the case of a sulfide, the oxygen fugacity of the magma ocean would have already risen above the level at which metallic iron can be stable, and so it is possible that H was present in the melt mainly as H₂O, particu-larly at lower pressures (Hirschmann et al., 2012). The study of Buono and Walker (2015) indicates that H₂O can indeed disproportionate, with O bond-ing to Fe creatbond-ing ferropericlase and H enterbond-ing the FeS, constrained through a depression of the melting temperature.

Our data can be used to perform a similar calculation as presented above in section 5.5.1, to determine the degree of oxidation possible from water dispro-portionation and hydrogen segregation to the core. A peridotite melt at 1673 K with 1 wt % H₂O would be ∼1.8 mole percent H. Using our highest value forD_H of 0.4 (which can be taken as a minimum for reasons described above), a coexisting liquid sulfide would then be 0.7 mole % H. Assuming that this hydrogen derived from H₂O, and the oxygen is taken up by oxidising FeO to Fe₂O₃, the change in oxygen fugacity expected can be calculated as a function of amount of sulfur exsolved, and is plotted in figure 5.15. As we are un-certain that our quenched sulfide managed to capture the full complement of hydrogen that may have dissolved into the liquid during the experiment, also plotted in figure 5.15 is the amount of oxidation that would be expected if the sulfide became hydrogen saturated (Shibazaki et al., 2011). For comparison, the amount of reduction caused by the loss of oxygen (section 5.5.1) is also plotted.

Figure 5.15: Change in oxygen fugacity of a silciate melt caused by the loss of oxygen and hydrogen via dissolution into a core-forming sulfide melt. The blue line indicates the increase in oxygen fugacity that would be expected if our measurements are an accurate representation of the amount of dissolved hydrogen. The green line indicates the amount of possible oxidation if a sul-fide were to become hydrogen saturated, using the saturation limit reported by Shibazaki et al. (2011). Vertical dashed lines represent the approximate amount of exsolved sulfide predicted by Rubie et al. (2016).

As can be seen, the amount of oxidation caused by the removal of hydrogen is, at absolute maximum, negated by the reduction that is caused by the simul-taneous loss of oxygen. If the sulfide is at all undersaturated in hydrogen, the net effect on the oxidation state of a silicate magma ocean of sulfide exolution and removal would more than likely have been an overall reduction. Further-more, the effect is likely to have been very small. The vertical lines in figure 5.15 represent the amount of sulfide that is likely to have exsolved, estimated from the work of Rubie et al. (2016). Their model predicts that before sulfide segregation, the Earth’s mantle would have been overabundant in sulfide by a factor of ∼ 30. To bring the mantle concentration of sulfur to the present-day value of ∼ 200 ppm (McDonough and Sun, 1995) would have required about

∼0.8 - 1 mole % of sulfide to have exsolved. As shown in figure 5.15, this

amount of sulfide exolution could have, at most, changed the oxygen fugacity of the coexisting silicate by about 0.1 log unit. Changing the fO2 by even just 0.5 log units would have required the exolution of 5 mole % FeS, which is unlikely.

Futhermore, the effects of removing oxygen and hydrogen simultaneously act to cancel each other out, resulting in an even smaller net effect. It is even possible that H₂O does not disproportionate, and dissolves into the sulfide as a whole molecule, as there is evidence that H₂O is soluble in sulfide melts (Wykes and Mavrogenes, 2005). The low amount of hydrogen in our quenched sulfide melts inhibited attempts to determine the speciation of the hydrogen, but it is entirely possible that water may dissolve in sulfide without dispropor-tionating and therefore the exolution of a sulfide may not affect the oxidation state of the coexisting magma ocean at all.

In summary, it appears problematic for oxidation by H₂O to have raised the oxidation state of the mantle significantly above the level of iron metal sat-uration (Sharp et al., 2013). The hydrogen evolved would simply reduce the surrounding mantle material back to the original level and a separating sul-fide phase would be of insufficient volume to remove significant H. Only by releasing H at the very surface of the magma ocean might it be possible for oxidation to take place. Models of magma ocean dynamics would be necessary to determine if such a mechanism could lead to oxidation of the entire mantle by mixing. In contrast, the stabilization of ferric iron in a deep magma ocean in association with precipitation of iron metal would appear to be a much more effective mechanism for raising the redox state of Earth-sized planetary bodies.