Benchmark of GW approaches for the GW100 testset

TL;DR

This paper systematically evaluates various GW computational methods for molecular ionization energies on the GW100 test set, comparing them to high-level CCSD(T) reference data to identify the most accurate approaches.

Contribution

It provides a comprehensive benchmark of GW methods, highlighting the accuracy of scGW and qsGW, and introduces DSLE-minimized GW for improved ionization energy predictions.

Findings

scGW and qsGW agree well with CCSD(T) within 0.3 eV and 0.2 eV respectively.

Starting point dependence affects scGW0 and G0W0 accuracy.

DSLE minimization systematically improves ionization energy predictions.

Abstract

For the recent GW100 test set of molecular ionization energies, we present a comprehensive assessment of different GW methodologies: fully self-consistent GW (scGW), quasiparticle self-consistent GW (qsGW), partially self-consistent GW0 (scGW0), perturbative GW (G0W0) and optimized G0W0 based on the minimization of the deviation from the straight-line error (DSLE-minimized GW). We compare our GW calculations to coupled-cluster singles, doubles, and perturbative triples [CCSD(T)] reference data for GW100. We find scGW and qsGW ionization energies in excellent agreement with CCSD(T), with discrepancies typically smaller than 0.3 eV (scGW) respectively 0.2 eV (qsGW). For scGW0 and G0W0 the deviation from CCSD(T) is strongly dependent on the starting point. We further relate the discrepancy between the GW ionization energies and CCSD(T) to the deviation from straight line error (DSLE). In DSLE-minimized GW calculations, the DSLE is significantly reduced, yielding a systematic improvement in the description of the ionization energies.