The seismic velocities of ferric iron-enriched bridgmanite under core-mantle boundary (CMB) conditions were calculated using GGA + U ab initio molecular dynamics to probe whether ferric iron enriched mantle can explain the properties of ultra-low velocity zones (ULVZ). Under these conditions ferric iron demonstrated some unusual properties.

The effect of ferric iron on the 0 K transition pressure of the bridgmanite (bdg) to postperovskite (ppv) transition with GGA + U has some dependence upon the value of U-eff that is set for B site ferric iron which is in contrast to most substances where U-eff has little effect on the properties. We find that ferric iron can both stabilise and destabilise the bdg phase relative to the ppv phase depending upon the value of U-eff. This is due to the spin state transition of B-site ferric iron and was not seen in ferrous or aluminous ferric bridgmanite which lack such a transition at mantle conditions. Due to the similar energies and pressure derivatives of the bdg and ppv phases very subtle energy changes (such as that of iron clustering or different theoretical or experimental setups) can have large effects on the relative phase stabilities at different pressures.

The spin state of ferric iron in bdg demonstrates some novel behaviours. With large iron enrichment the spin state of the B site has a non-linear dependence on concentration and depends upon the local arrangement of iron. At typical lower mantle pressures, ferric iron is expected to be high spin in the A site and low spin at the B site but at the temperatures of the CMB thermal spreading of the electrons causes ferric iron at the B site to become high spin.

This is not a typical "spin transition" however as it involves no structural rearrangement of the unit cell and is predicted to have no effect on the elasticity and thus the seismicity of the crystal.

Our elasticity results show that at the CMB ferric iron-enriched bridgmanite has nearly identical V-p and V-s to ferrous iron-bearing bridgmanite, and thus bridgmanite containing large amounts of Fe2+ or Fe3+ can both explain ULVZ properties when mixed with ferropericlase. We find, however, that mixing of both ferric and ferrous iron in bridgmanite causes a large (up to 1 km/s) non-ideal decrease in V-p when there is middling amounts of ferric iron. This non-linear mixing should have significant effects throughout the lower mantle whenever such amounts of ferric iron are present though such a ferrous:ferric ratio would typically require an increase of the natural ferrous:ferric ratio (for example through very high oxidation fugacities or chemical doping). In the case of ULVZs this non-linear mixing significantly impairs the fitting to some ULVZ data sets (those with a dlnV(s)/dlnV(p) similar to 3), but greatly improves the fitting to other ULVZ data sets (those with a dlnV(s)/dlnV(p) similar to 1).

The oxidation state of highly enriched iron in bridgmanite at CMB conditions has, therefore, a large effect on the ability of mantle mixtures that are highly enriched in iron to replicate ULVZ properties.