Benchmarking total energies with Hund's J terms in Hubbard-corrected spin-crossover chemistry

TL;DR

This study evaluates the impact of Hund's J terms in DFT+U+J methods on spin-crossover Fe(II) molecules, revealing limitations in current approaches for accurately capturing static correlation effects in covalent systems.

Contribution

It systematically assesses Hund's J parameters in DFT+U+J corrections and highlights their limitations, proposing directions for improving static correlation modeling in spin-crossover chemistry.

Findings

Hund's J terms often fail to improve energy predictions in covalent systems.

Simplified rotationally-invariant Hubbard functionals can approximate adiabatic energy differences.

Canonical Hund's J terms may not be suitable for strongly covalent spin-crossover molecules.

Abstract

The effect of the Hund's J terms in various DFT+U+J corrections to semi-local spin-density functional theory is assessed for a series of four octahedrally coordinated Fe(II) spin-crossover molecules spanning the covalent end of the ligand field spectrum. We report values and analyze trends for the Hubbard U and Hund's J parameters determined via minimum-tracking linear response for all valence atomic subspaces and relevant spin states in these molecules. We then methodically apply them via simplified rotationally-invariant Hubbard functionals in search of the simplest combination to yield reliable adiabatic energy differences with respect to those obtained using CASPT2/CC. The observed failure of canonical, positively-signed Hund's J terms in furthering the already robust capacity of DFT+U to obtain accurate energetics prompts an evaluation of their limitations when seeking to account for the static correlation phenomena in such strongly covalent systems and suggests directions for their improvement.