DFT-1/2 for ionic insulators: Impact of self-energy potential on band gap correction

Abstract

DFT-1/2 is an efficient band gap correction method for density functional theory under local density approximation or generalized gradient approximation. Nevertheless, it still predicts lower band gaps than experimental in highly ionic insulators, such as LiF and CsCl. In this study, we emphasize that the problem is more severe for halides. For example, the DFT-1/2 band gap for Li2O is much more satisfactory than that for LiF. Notwithstanding the traditional belief that high ionicity is the key, we ascribe the main reason to the fact that the self-energy potentials (SEPs) used for anions were derived based on neutral atoms. Extensive tests have shown that an SEP derived from an atom/ion with fewer electrons is stronger than that derived with more electrons. The degree of enhancement in the O SEP, based on O rather than O2–, could generally compensate for the loss of SEP in the trimming process. In the case of halides, such compensation is insufficient, resulting in small band gaps. For ionic nitrides involving N3–, the SEP is undesirably large if it is derived from a neutral N atom. Equivalent scaling factors are prescribed for the SEPs of F and N, with the former being greater and the latter being smaller than unity. With proper SEP scaling factors according to the charge carried by the anion, the DFT-1/2 band gaps are greatly improved for highly ionic insulators.