npj Computational Materials (Jan 2021)
Towards fully automatized GW band structure calculations: What we can learn from 60.000 self-energy evaluations
Abstract
Abstract We analyze a data set comprising 370 GW band structures of two-dimensional (2D) materials covering 14 different crystal structures and 52 chemical elements. The band structures contain a total of 61716 quasiparticle (QP) energies obtained from plane-wave-based one-shot G0W0@PBE calculations with full frequency integration. We investigate the distribution of key quantities, like the QP self-energy corrections and QP weights, and explore their dependence on chemical composition and magnetic state. The linear QP approximation is identified as a significant error source and we propose schemes for controlling and drastically reducing this error at low computational cost. We analyze the reliability of the 1/N basis set extrapolation and find that is well-founded with a narrow distribution of coefficients of determination (r 2) peaked very close to 1. Finally, we explore the accuracy of the scissors operator approximation and conclude that its validity is very limited. Our work represents a step towards the development of automatized workflows for high-throughput G0W0 band structure calculations for solids.