Untitled Document

Partitioning of iron between magnesian silicate perovskite and magnesiowüstite at about 1 Mbar

S.E. Kesson, J.D. Fitz Gerald, H.StC. O’Neill and J.M.G. Shelley

The two major phases in the Earth’s lower mantle are magnesian silicate perovskite (Mg,Fe)SiO3 and magnesiowüstite (= ferropericlase) (Mg,Fe)O. Allocation of ferrous iron and magnesium between them is described by a parameter called a "distribution coefficient" (or "partition coefficient") KD such that

where X means mole fraction. Over the past two decades, substantial effort has been directed towards experimental measurements of perovskite-magnesiowüstite distribution coefficients, with the objective of understanding the effects of pressure, temperature and system chemistry. These goals are important because that distribution relationship describes the essential phase chemistry of the lower mantle, and also offers a valuable independent constraint on geophysical and geochemical models of the Earth’s interior. However a review of the scientific literature soon reveals that, whilst it is agreed that iron is preferentially-concentrated in magnesiowüstite, preferred values for the distribution coefficient range from as low as 0.1 to near unity!

Some of the dispersion amongst distribution coefficient values arises from the inadvertent – often unavoidable – oxidation of some ferrous iron to the trivalent state during experiments. Unfortunately in many cases this process has been unrecognised or unacknowledged. And this ferric iron, we now know, is concentrated in magnesian silicate perovskite in preference to magnesiowüstite, which results in distribution coefficients with seemingly elevated values. There has also been considerable debate as to whether distribution coefficients measured in the widely-studied experimental system MgO-FeO-SiO2 are applicable to the real Earth. The argument hinges on the role of Al2O3, which under lower mantle conditions is accommodated in magnesian silicate perovskite. The results of some classes of high-pressure experiments imply that incorporation of Al3+ in perovskite requires coupled substitution with ferric iron: Mg2+ + Si4+ = Fe3+ + Al3+. And the presence of iron in both valence states leads to predictable confusion over what thereby constitutes an appropriate value for the distribution coefficient.

Our objective has been to determine Mg2+-Fe2+ distribution coefficients for perovskite-magnesiowüstite pairs produced by disproportionation of olivine Fo90, at pressures relevant to the deeper part of the lower mantle (~ 1 Mbar), for which data are minimal. We then examined our data critically from the perspective of the various issues raised above, and from thermodynamic fundamentals.

The disproportionation products of olivine Fo90 in our diamond anvil cell experiments at pressures of about 1 Mbar are found to be magnesian silicate perovskite Mg# ~94 and a series of magnesiowüstites Mg# 85—90, yielding a recommended value of 0.4 ± 0.1 for the distribution coefficient that defines the exchange of Mg2+ and Fe2+ between the two phases (KD). The distinctive compositional trends which would signify that new ferric iron had been stabilised during experiments are lacking. We cannot measure the temperature of our experiments, so Wood’s multi-anvil study (Earth Planet. Sci. Lett. (2000) 174, p.341) was used to link our KD values to temperature. Our recommended value of 0.4 ± 0.1 is thereby equated with temperatures of 2000 ± 300 K. We compared our data with the results of an analogous experimental campaign that we had earlier conducted at much lower pressures of 25—50 GPa, and found them to be much the same. However, the implication that pressure accordingly has minimal effect on distribution behaviour is not in accordance with our thermodynamic calculations, which predict that KD should fall by about 40% if pressure increases from 50 GPa to 1 Mbar. An explanation for this discrepancy is required. The most likely culprit is the molar volume for the hypothetical end-member perovskite FeSiO3. If this were in error by only -0.5%, the volume change for the exchange reaction, and hence the pressure-dependency of KD, would be effectively zero.

We were also able to demonstrate that the presence of Al3+ in magnesian silicate perovskite in amounts appropriate for the lower mantle (about 5 mole %) does not modify distribution behaviour, provided the oxygen content of the system remains essentially constant – a situation that would necessarily prevail in the real Earth. The above considerations allow us to predict that magnesian silicate perovskite Mg# 93.0 ± 0.6 would coexist with magnesiowüstite Mg# 82.6 ± 0.9 just beyond the 660 km discontinuity (1800 K), whilst at the core-mantle boundary (2650 K) the corresponding values would be 91.9 ± 0.4 and 84.9 ± 0.8.