Effects of self-consistency and plasmon-pole models on GW calculations for closed-shell molecules

TL;DR

This study compares different GW computational approaches for closed-shell molecules, highlighting how self-consistency and plasmon-pole models influence the accuracy of quasiparticle energy predictions.

Contribution

It systematically evaluates three GW methods, revealing the limitations of full-frequency $G_0W_0$ in predicting ionization potentials and electron affinities.

Findings

Full-frequency $G_0W_0$ can deviate by over 1 eV from experimental values.

Self-consistent $GW_0$ and plasmon-pole models yield better agreement with experiments.

Incorrect self-energy poles in FF-$G_0W_0$ cause significant inaccuracies.

Abstract

We present theoretical calculations of quasiparticle energies in closed-shell molecules using the GW method. We compare three different approaches: a full-frequency $G_0W_0$ (FF-$G_0W_0$) method with density functional theory (DFT-PBE) used as a starting mean field; a full-frequency $GW_0$ (FF-$GW_0$) method where the interacting Green's function is approximated by replacing the DFT energies with self-consistent quasiparticle energies or Hartree-Fock energies; and a $G_0W_0$ method with a Hybertsen-Louie generalized plasmon-pole model (HL GPP-$G_0W_0$). While the latter two methods lead to good agreement with experimental ionization potentials and electron affinities for methane, ozone, and beryllium oxide molecules, FF-$G_0W_0$ results can differ by more than one electron volt from experiment. We trace this failure of the FF-$G_0W_0$ method to the occurrence of incorrect self-energy poles describing shake-up processes in the vicinity of the quasiparticle energies.