# Time-Dependent Multiconfigurational Short-Range Density Functional Theory with Generalized Valence Bond Wave Functions

## TL;DR

This paper introduces a new computational method for calculating molecular excitation energies and spin couplings using a combination of density functional theory and generalized valence bond wave functions.

## Contribution

The novel contribution is the development and implementation of TD-GVB-srDFT, which improves excitation energy accuracy compared to prior methods.

## Key findings

- TD-GVB-srDFT achieves excitation energy accuracy comparable to CAS-srDFT with deviations of 0.2 eV.
- The generalized Tamm-Dancoff approximation is essential for accurate triplet excitation calculations.
- GVB-srDFT accurately computes spin–spin coupling constants in organic molecules and transition metal complexes.

## Abstract

We present a theory and an efficient implementation of
TD-GVB-srDFT,
a time-dependent multiconfigurational range-separated density functional
theory based on generalized valence bond perfect-pairing (GVB-PP)
wave functions. In GVB-srDFT, dynamic correlation effects are incorporated
via range-separation of the Coulomb potential, using tailored Kohn–Sham
functionals of the density. The present implementation builds on our
earlier work on TD-GVB [Hapka et al. J. Chem. Phys.
2022, 156, 174102], which employs
direct Hessian techniques for both wave function optimization and
linear response. We benchmark the performance of TD-GVB-srDFT for
singlet and triplet excitation energies, as well as indirect spin–spin
coupling constants (SSCCs). Compared to the underlying GVB-PP model,
the method significantly improves excitation energies and achieves
accuracy comparable to the complete active space variant, CAS-srDFT,
with mean absolute deviations of 0.2 eV. The use of the generalized
Tamm-Dancoff approximation (gTDA) is mandatory for reliable treatment
of triplet excitations. For organic molecules, SSCCs computed with
GVB-srDFT closely match those from CAS-srDFT and HF-srDFT results,
whereas pure GVB-PP performs markedly worse than CASSCF for all coupling
terms. Both GVB-srDFT and CAS-srDFT accurately reproduce fluorine–metal
couplings in transition metal complexes, provided that gTDA is applied
to singlet contributions.