GW100: Benchmarking G0W0 for Molecular Systems.

Area of Science:

• Computational Chemistry
• Quantum Chemistry
• Materials Science

Background:

• Accurate prediction of molecular electronic properties is crucial for materials design and chemical reaction understanding.
• The GW method is a powerful tool for calculating quasi-particle energies, but its implementation across different codes can lead to variations.

Purpose of the Study:

• To establish a benchmark dataset (GW100) for ionization potentials and electron affinities of 100 molecules.
• To compare the performance of three independent GW codes (TURBOMOLE, FHI-aims, BerkeleyGW) using various methodologies.
• To identify best practices and quantify error bars for common approximations in GW calculations.

Main Methods:

• Calculation of highest-occupied molecular orbital (HOMO) and lowest-unoccupied molecular orbital (LUMO) energies using the G0W0@PBE level of theory.
• Utilizing three distinct software packages: TURBOMOLE, FHI-aims, and BerkeleyGW.
• Investigating variations in basis sets, treatment of unoccupied states, core/valence electrons, frequency dependence, and quasi-particle equation solvers.

Main Results:

• Generation of the GW100 benchmark set, providing reference quasi-particle energies for 100 molecules.
• Quantitative comparison of results obtained from different GW codes and methodologies.
• Assessment of the impact of various computational choices (basis sets, pseudopotentials, etc.) on accuracy.

Conclusions:

• The GW100 set serves as a valuable resource for validating and improving GW methods.
• Understanding the discrepancies between different GW implementations is key to achieving reliable predictions.
• This work provides guidance for future high-throughput calculations and method development in electronic structure theory.