On the sensitivity of computed partial charges toward basis set and (exchange-)correlation treatment

TL;DR

This study investigates how computed partial charges depend on basis set and exchange-correlation treatment, revealing that certain hybrid functionals and basis sets yield reliable and converged charges across various definitions.

Contribution

It systematically analyzes the basis set and correlation treatment effects on partial charges for multiple definitions, highlighting the effectiveness of semi-empirical double hybrids and specific basis sets.

Findings

Range separation offers no clear advantage.

Global hybrids with 20-30% HF exchange perform best.

Augmented triple-zeta basis sets are sufficient for some methods.

Abstract

Partial charges are a central concept in general chemistry and chemical biology, yet dozens of different computational definitions exist. In prior work [M. Cho et al., \textit{ChemPhysChem} {\bf 21}, 688-696 (2020)], we showed that these can be reduced to at most three `principal components of ionicity'. The present study addressed the dependance on computed partial charges $q$ on 1-particle basis set and (for WFT methods) $n$-particle correlation treatment or (for DFT methods) exchange-correlation functional, for several representative partial charge definitions such as QTAIM, Hirshfeld, Hirshfeld-I, HLY (electrostatic), NPA, and APT. Our findings show that semi-empirical double hybrids can closely approach the CCSD(T) `gold standard' for this property. In fact, owing to an error compensation in MP2, CCSD partial charges are further away from CCSD(T) than is MP2. The non-local correlation is important, especially when there is a substantial amount of non-local exchange. Employing range separation provides no clear advantage, while global hybrids with 20-30\% Hartree-Fock exchange exhibit the best performance across all charge types. Basis set convergence analysis shows that an augmented triple-zeta haVTZ+d basis set is sufficient for Hirshfeld, Hirshfeld-I, HLY, and APT methods. In contrast, QTAIM and NPA display slower basis set convergence. It is noteworthy that for both NPA and QTAIM, HF exhibits the slowest basis set convergence when contrasted with the correlation components of MP2 and CCSD. Triples corrections in CCSD(T), denoted as CCSD(T)-CCSD, exhibit even faster basis set convergence.