The exchange-correlation dipole moment dispersion method

TL;DR

This paper introduces the XCDM model, an improved dispersion correction for DFT that enhances accuracy in molecular and solid-state systems, and proposes a new scheme for fair benchmark evaluation.

Contribution

The paper develops the XCDM model, adding dynamical correlation to XDM, and introduces the WTMAD-4 scheme for more balanced benchmark assessments.

Findings

XCDM improves accuracy for molecular $C_6$ dispersion coefficients.

XDM and MBD corrections are tested on GMTKN55 for the first time.

The WTMAD-4 scheme offers a fairer evaluation of functional performance.

Abstract

Density-functional theory (DFT) has become the workhorse of modern computational chemistry, with dispersion corrections such as the exchange-hole dipole moment (XDM) model playing a key role in high-accuracy modelling of large-scale systems. Here, we introduce a new physics-guided XDM variant, termed the exchange-correlation dipole moment (XCDM) model, which supplements XDM with same- and opposite-spin dynamical correlation terms, substantially improving accuracy for molecular $C_6$ dispersion coefficients. Both XDM and XCDM are implemented for use with the Becke-Johnson damping function based on atomic radii, as well as a one-parameter damping function based on atomic numbers, recently proposed by Becke. All four variants are benchmarked on the comprehensive GMTKN55 database using minimally empirical generalised-gradient-approximation, global hybrid, and range-separated hybrid functionals. This marks the first time that the XDM (and many-body dispersion, MBD) corrections have been tested for the GMTKN55 set. Five solid-state benchmarks spanning molecular crystals and layered materials are also considered. The B86bPBE0 hybrid functional, paired with any of the XDM variants, shows excellent performance for molecular systems. Finally, we identify a flaw in the weighted mean absolute deviation (WTMAD-2) scheme commonly used for the GMTKN55 set, which underweights some of its component benchmarks by orders of magnitude. We propose a new WTMAD-4 scheme based on typical errors observed for well-behaved functionals, ensuring fair treatment across all benchmarks.