Comment on “Generalized Gradient Approximation +U Study for Metallization Mechanism of Niobium-Doped Anatase Titanium Dioxide” [Jpn. J. Appl. Phys. 49 (2010) 055801

In a recent paper by Orita, 1 density functional calculations were performed on niobium-doped anatase TiO2, using the Perdew–Burke–Ernzerhof (PBE) generalized-gradient approximation (GGA) functional, with Dudarev +U terms applied to the titanium and niobium d states; GGA+U{Ti, Nb}. Therein it was predicted that excess electrons introduced by Nb-doping occupy the bottom of the conduction band in states hybridized between the Nb and many Ti d-orbitals, giving a metallic system. This contrasts with a previous GGA+U study of Morgan et al. where an excess electron was found to be strongly localized as a small polaron at a single Ti site adjacent to the Nb/Ti site. 2 In this earlier work the +U term was only applied to the Ti d states, since the excess electrons introduced by Nb-doping are expected to be transferred to lower lying Ti 3d states, as discussed by Osorio-Guillén et al. 3 Orita argued that the discrepancy between these GGA+U studies is a consequence of an absent UNb correction in the work of Morgan et al. A direct comparison between calculations with the same parameter sets and only the presence or absence of UNb as a distinction was however not made. Here we demonstrate this claim by Orita is incorrect: applying UNb has a negligible effect on the calculation. The metallic solutions obtained by Orita are explained by neglecting to allow for spin-polarization. Across a wide range of U values allowing for full spin relaxation achieves a more stable solution with a defect state split off from the conduction band, corresponding to a small polaron localized at a single Ti center.

article link | pdf