# Fast and accurate ionization potentials and electron affinities with qsGW

In various areas of chemistry and material science, ranging from the design of photovoltaic devices to the understanding of photodamaged DNA, the precise calculation of electron attachment and removal energies using the GW method is of high interest. In typical implementations, the outcome of a GW calculation depends on the choice of an exchange-correlation (XC) functional as input parameter, which complicates its application and might deteriorate its accuracy.

The quasiparticle self-consistent GW (qsGW) method overcomes this shortcoming by solving the GW equations self-consistently, but so far its application has been limited to small molecules. The qsGW method has been recently implemented in the ADF engine of the Amsterdam Modelling Suite and unlocks routine applications to molecules with a few hundred of atoms. For example, the quasiparticle spectrum of a DNA oligomer with 1200 correlated electrons has been calculated on the TZ3P level (6300 Slater functions) in less than 4 days on a single compute node with 16 cores.(1)

In addition to its strict independence of an XC functional, qsGW is also among the most accurate GW methods. A recent benchmark for the ionization potentials and fundamental gaps of 24 organic acceptor molecules reveals excellent agreement with CCSD(T) reference values. The results clearly show that qsGW is much more accurate than G0W0 based on hybrid or range-separated hybrid starting points.(2)

qsGW and G3W2 will be available in the AMS2022 release, coming up soon!

### GW Recommendations:

• qsGW is preferred over qsGW0 and evGW(0). qsGW has little sensitivity to the xc functional
• use at least an all-electron TZ2P or larger for IP, Corr/TZ3P or larger for EA (see recommendations)
• for HOMO-LUMO gaps, Accuracy Normal usually suffices, for more accurate HOMO and LUMO energies, use Accuracy Good
• for QZ6P basis sets, use Accuracy VeryGeed
• reduce the Basis set and RIHF dependencies:
Dependency Bas=5e-3
RIHartreeFock DependencyThreshold=5e-3

In AMS2023 we will have qsGW-BSE available, check out Arno Forster’s web presentation!