Accelerating GW calculations with optimal polarizability basis

Abstract

We present a method for accelerating GW quasi-particle (QP) calculations. This is achieved through the introduction of optimal basis sets for representing polarizability matrices. First the real-space products of Wannier like orbitals are constructed and then optimal basis sets are obtained through singular value decomposition. Our method is validated by calculating the vertical ionization energies of the benzene molecule and the band structure of crystalline silicon. Its potentialities are illustrated by calculating the QP spectrum of a model structure of vitreous silica. Finally, we apply our method for studying the electronic structure properties of a model of quasi-stoichiometric amorphous silicon nitride and of its point defects.

Quasiparticle GW calculations for solids, molecules, and two-dimensional materials

“…However, for a cutoff larger than 100 eV, the IP grows linearly with 1/E cut and this allows us extrapolate to the infinite cutoff (and number of empty bands) limit. 76,77 In this case the converged ionization potential is 12.1 eV, which is about 0.5 eV smaller than the experimental value. For all the molecules we have extrapolated the IP to infinite plane wave cutoff based on G 0 W 0 calculations at cutoff energies 200-400 eV.…”

Which electronic structure method for the study of defects: A commentary

“…Two examples represented in this compilation are the papers by Umari et al [131] and Giantomassi et al [112]. Umari et al [131,132] discuss the use of a separate small orthogonal basis set for expanding the polarizability operator.…”

First-principles investigation of organic photovoltaic materialsC60, C70, [C60]PCBM, and bis-Qian1, Umari2, Marzari3 2015Phys. Rev. B

“…As listed in Table I for C 60 , the quantitative description of quasiparticle energies at the G 0 W 0 level can be improved by solving the bottleneck of the sum-over-emptystates through the efficient Lanczos approach [69][70][71][72][73] We display in Figs. 2 and 3 the electronic density of states (DOS) of the five electron-acceptor molecules, calculated with G 0 W 0 @LDA and G 0 W 0 @GGA, respectively, and compare them with direct and inverse photoemission data without peak alignment, albeit neglecting any oscillator strength effect in the calculation.…”

“…The GW method has already been applied to fullerenes in some theoretical works [66][67][68] . Here, we use the recently developed many-body GW-Lanczos approach which is particularly effective in reaching numerical convergence [69][70][71][72][73] in large atomic structures without suffering from bottlenecks with respect to summing over a large number of empty Kohn-Sham orbitals. This allows us to calculate the electronic structure of the electron acceptors including C 60 , C 70 , [C 60 ]PCBM, and bis-[C 60 ]PCBM at both the DFT and GW level, and compare them with experimental photoemission results.…”