Natural virtual orbitals for the GW method in the random-phase approximation and beyond.

Area of Science:

• Computational Chemistry
• Quantum Chemistry
• Theoretical Molecular Science

Background:

• The GW method is increasingly utilized for determining vertical ionization energies in molecular systems.
• High computational cost, particularly steep scaling with system size and basis set, limits the applicability of standard GW methods.

Purpose of the Study:

• To implement and evaluate correlated natural virtual orbitals (NVOs) for improving the efficiency of GW calculations.
• To assess the accuracy and performance of truncated NVOs within various GW formalisms, including G0W0 and evGW0.
• To provide improved reference data for the GW100 test set.

Main Methods:

• Implementation of correlated NVOs based on second-order Møller-Plesset (MP2) perturbation theory.
• Testing NVOs and truncated NVOs in GW quasiparticle calculations, including G0W0, evGW0, and vertex corrections.
• Comparison of NVO-based GW results with coupled-cluster theory with singles, doubles, and noniterative triples [CCSD(T)] calculations.

Main Results:

• Correlated NVOs considerably improve computational efficiency for larger molecular systems and basis sets.
• Truncated NVOs achieve speedups of an order of magnitude with negligible loss in accuracy on the GW100 test set.
• The impact of basis set choice on accuracy is significantly greater than the effect of NVO truncation.

Conclusions:

• Correlated NVOs offer a computationally efficient alternative for GW calculations of vertical ionization energies.
• Truncated NVOs provide a practical strategy to achieve significant speedups without compromising accuracy.
• The NVO approach enables more accurate results at the same computational expense compared to standard methods.