High-pressure phases in the Al 2 SiO 5 system and the problem of aluminous phase in the Earth's lower mantle: ab initio calculations

One of the main uncertainties in mineralogical models of the Earth's lower mantle is the nature of the aluminous mineral: it is not clear whether Al forms its own minerals or is mainly contained in (Mg,Fe)SiO3- perovskite. This question is very important, since it is known that if Al were mainly hosted by perovskite, it would radically change Fe/Mg-partitioning and phase equilibria between mantle minerals, and also alter many physical and chemical properties of perovskite, which is currently believed to comprise ca. 70% of the volume of the lower mantle. This, in turn, would require us to re- consider many of our geochemical and geophysical models for the lower mantle. This work considers the possibility of a V3O5-type structured modification of Al2SiO5 to be the main host of Al in the lower mantle, as proposed by previous workers. We report ab initio cal- culations, based on density functional theory within the generalised gradient approximation (GGA) with plane wave basis set and nonlocal pseudopotentials. We consider polymorphs of Al2SiO5 (kyanite, andalusite, sillimanite, and hypothetical V3O5-like and pseudo- brookite-like phases), SiO2 (stishovite, quartz) and Al2O3 (corundum). Computational conditions (e.g., plane-wave energy cutoA, Brillouin zone sampling) were carefully chosen in order to reproduce small energy changes associated with phase transitions between the Al2SiO5 polymorphs. Good agreement of crystal struc- tures, bulk moduli, atomisation energies and the phase diagram of Al2SiO5 with experimental data was found. Strong disagreement between the calculated lattice pa- rameters and density of V3O5-like phase of Al2SiO5 and experimental values, assigned to it by previous workers, suggests that a V3O5-structured phase of Al2SiO5 was never observed experimentally. In addition, we found that the most stable high-pressure assembly in Al2SiO5 system is corundum+stishovite, and the value of the transition pressure at T = O K (113 kbar) is in excellent agreement with experimental estimates (95-150 kbar). We explain the instability of octahedrally coordinated silicates of Al to decomposition on the basis of Pauling's second rule.