The Relevance of Dispersion Interactions for the Stability of Oxide Phases

The total energies of TiO2 and Al2O3 allotropic forms computed using density functional theory (DFT) methods that include dispersion interaction effects are compared. For TiO2, adding energy terms of r(-6) form with coefficients derived from atomic polarizabilities leads to the correct result that rutile is more stable than brookite, anatase, and other forms, while previous DFT studies without this correction wrongly predicted rutile to be less stable. The magnitude of the correction is significant because of the high polarizability of the oxide ion, and produces energy differences between the titania phases that are reasonably close to available experimental values. The van der Waals density functional does not yield the correct result, but the error is significantly decreased. For Al2O3 the experimental energy difference between alpha and gamma forms is also approached better when including dispersion corrections in DFT calculations. The results show that dispersion interaction should not be ignored when computing with DFT energy differences between oxide structures and that for the latter the dispersion coefficients appropriate for neutral species should not be used. It is proposed that the correct reproduction of these differences be included in the benchmarks used for testing van der Waals-type functionals.