Band gaps in pseudopotential self-consistent GW calculations

TL;DR

This paper evaluates iterative GW calculation schemes for inorganic semiconductors, revealing that iterating only quasi-particle energies slightly overestimates band gaps, while including wave functions worsens agreement with experiments.

Contribution

It compares two iterative schemes for GW calculations in semiconductors, highlighting the effects of iterating energies versus wave functions on band gap accuracy.

Findings

Iterating only quasi-particle energies modestly overestimates band gaps.

Including wave functions causes poor agreement with experimental band gaps.

Iterative schemes' effectiveness depends on the specific approximation used.

Abstract

For materials which are incorrectly predicted by density functional theory to be metallic, an iterative procedure must be adopted in order to perform GW calculations. In this paper we test two iterative schemes based on the quasi-particle and pseudopotential approximations for a number of inorganic semiconductors whose electronic structures are well known from experiment. Iterating just the quasi-particle energies yields a systematic, but modest overestimate of the band gaps, confirming conclusions drawn earlier for CaB_6 and YH_3. Iterating the quasi-particle wave functions as well gives rise to an imbalance between the Hartree and Fock potentials and results in bandgaps in far poorer agreement with experiment.