Achieving accurate ionization potential of semiconductors by the efficient Kohn-Sham scheme of density functional theory

TL;DR

This paper demonstrates that an approximate optimized effective potential within density functional theory can accurately predict semiconductor ionization potentials at a computational cost similar to standard DFT methods, rivaling more complex approaches.

Contribution

The study introduces a surface technique using Becke-Johnson'06 exchange within DFT to accurately compute semiconductor ionization potentials, matching GW approximation accuracy at lower computational cost.

Findings

Achieved semiconductor IP accuracy comparable to GW approximation.

Used approximate OEP with LDA correlation for efficient calculations.

Highlighted the importance of long-range potential effects on IP.

Abstract

Despite of its huge successes in vast amount of applications, the Kohn-Sham scheme of density functional theory (DFT-Kohn-Sham) has not been able to get reliable ionization potentials (IP) for semiconductors, due to self-interaction error in the local density approximation (LDA) and generalized gradient approximations (GGA), and the difficulty of using asymptotically long-ranged potentials for surface calculations. An approximate optimize effective potential (OEP), the Becke-Johnson'06 exchange, is used to explore the capability of OEP to calculate semiconductor IP with a surface technique suitable for both short- and long-ranged potentials. Combined with the LDA correlation, the approximate OEP has achieved an IP accuracy for 17 semiconductors which is similar to the much more sophisticated $GW$ approximation (GWA), with the computational cost of only LDA/GGA. For the first time, this opens the way for systematic calculation of semiconductor IP's by the DFT-Kohn-Sham scheme. The long-range part of the $v_{xc}$ is not as important for solid surface as for atoms and molecules since the wave functions are mainly bound to the slab while less exposed to the asymptotic region. Nevertheless, it contributes significantly to the total IP and therefore shall not be simply ignored. Without including its effects, it is unlikely that the same GWA procedure could achieve universally accurate band gaps and accurate IP's at the same time, as is indeed seen in practical calculations.