Fe3+ has a significant effect on lower mantle properties such as electrical conductivity despite its low abundance relative to cations such as Fe2+, Mg and Si. It is difficult to determine the Fe3+ content of the lower mantle, however, and the partitioning of Fe3+ between lower mantle phases is unknown. We have studied the problem using two different approaches: 1) through synthesis of phases at high P,T and different fO2 to constrain the range of Fe3+ expected for lower mantle phases, and 2) by direct measurement of the Fe3+ content of natural samples believed to have come from the lower mantle.
The solubility of Fe3+ in (Mg,Fe)O at atmospheric pressure is a strong function of fO2, where the Fe3+ content varies from very small at low fO2 to very large at high fO2. To study the solubility of Fe3+ in (Mg,Fe)O at high pressure, we prepared two samples of Mg0.8Fe0.2O: one with high Fe3+ content and one containing minimum Fe3+. Both were run in a multianvil press at 18 GPa and 1000°C at two different oxygen fugacities: a) equilibrium with Fe (reduced); and
b) Re-ReO2 buffer (oxidised). In all high pressure experiments the amount of Fe3+ found in the quenched phases using both Mössbauer spectroscopy and X-ray diffraction was extremely small. One explanation for this behaviour is the exsolution of small amounts of spinel to accomodate excess oxygen. There are high-pressure phase transitions in both Fe3O4 and MgFe2O4 which stabilises oxygen in the spinel phase at high pressure, hence producing a more stoichiometric (Mg,Fe)O.
A similar study of (Mg,Fe)SiO3 perovskite was performed with the composition Mg0.95Fe0.05O3 run at
24-25 GPa and 1650°C in the multianvil press under both reducing and oxidising oxygen fugacities. At its low fO2 stability limit, Mg0.95Fe0.05O3 perovskite contains at least 5% Fe3+ (relative to total Fe), which increases to more than 10% Fe3+ at high fO2.
Some inclusions in diamonds from São Luiz, Brazil are believed to have come from the lower mantle, because they show (Mg,Fe)O together with (Mg,Fe)SiO3 and CaSiO3, which are inferred to represent perovskite phases (Harte et al., 1995 and references therein). An aluminous phase with the stoichiometry of a pyrope-almandine garnet also occurs. Four (Mg,Fe)O inclusions were studied using the recently developed Mössbauer milliprobe technique, and all four showed very small amounts of Fe3+. This is consistent with experimental results, which show low solubility of Fe3+ in (Mg,Fe)O at high P,T regardless of fO2. In contrast, a study of three silicate phases revealed extremely high amounts of Fe3+. For the sample with perovskite stoichiometry the large amount of Fe3+ corresponds to a high concentration of Al, and simple calculations of cation distribution based on stoichiometry and charge balance are consistent with the substitution Mg2+ + Si4+ Æ Fe3+ + Al3+.
The solubility of Fe3+ in the (Mg,Fe)SiO3 perovskite phase at high P,T is significantly higher than in (Mg,Fe)O, suggesting that Fe3+/ Fe values should be significantly higher in the lower mantle perovskite phase compared to (Mg,Fe)O. The total Fe3+ contents depend on both the partitioning of Fe2+ between lower mantle phases (which determines the total Fe in each phase) and the lower mantle fO2. Inclusions in diamonds believed to have come from the lower mantle show high Fe3+ contents in the silicate phases while (Mg,Fe)O contains little or no Fe3+. The high Fe3+ contents in the silicate phases suggests that the inclusions formed at conditions of high fO2. If fO2 conditions within the bulk of the lower mantle are similarly high, most of the Fe3+ in the lower mantle is likely to be in the perovskite phase, since it is believed to be the dominant phase in the lower mantle.
Harte, B., Hutchison, M.T. & Harris, J.W., Mineral. Mag. 58A, 386-387 (1995).