Improving excited state potential energy surfaces via optimal orbital shapes

TL;DR

This paper shows that optimizing molecular orbitals for each excited state improves the accuracy of potential energy surface predictions without extra computational cost, avoiding complex weighting schemes.

Contribution

It introduces a single-state orbital optimization method that enhances excited-state surface predictions, reducing reliance on expensive dynamic correlation methods.

Findings

State-specific orbitals provide qualitative improvements.

Method is effective across diverse chemical systems.

Avoids need for state-averaging and dynamic weighting.

Abstract

We demonstrate that, rather than resorting to high-cost dynamic correlation methods, qualitative failures in excited-state potential energy surface predictions can often be remedied at no additional cost by ensuring that optimal molecular orbitals are used for each individual excited state. This approach also avoids the weighting choices required by state-averaging and dynamic weighting and obviates their need for expensive wave function response calculations when relaxing excited state geometries. Although multi-state approaches are of course preferred near conical intersections, other features of excited-state potential energy surfaces can benefit significantly from our single state approach. In three different systems, including a double bond dissociation, a biologically relevant amino hydrogen dissociation, and an amino-to-ring intramolecular charge transfer, we show that state-specific orbitals offer qualitative improvements over the state-averaged status quo.