Quantum Monte Carlo calculations of electronic excitation energies: the case of the singlet $n \to \pi^*$ (CO) transition in acrolein

TL;DR

This study uses quantum Monte Carlo methods to accurately calculate the singlet n→π* excitation energy in acrolein, examining the impact of basis sets and wave function optimization techniques on the results.

Contribution

It demonstrates that Slater basis sets and certain CAS wave functions yield accurate excitation energies without reoptimization, advancing computational approaches in excited state calculations.

Findings

Slater basis sets improve excitation energy accuracy.

State-average and state-specific CAS(6,5) wave functions perform well.

Reoptimization is necessary for smaller CAS(2,2) wave functions.

Abstract

We report state-of-the-art quantum Monte Carlo calculations of the singlet $n \to \pi^*$ (CO) vertical excitation energy in the acrolein molecule, extending the recent study of Bouab\c{c}a {\it et al.} [J. Chem. Phys. {\bf 130}, 114107 (2009)]. We investigate the effect of using a Slater basis set instead of a Gaussian basis set, and of using state-average versus state-specific complete-active-space (CAS) wave functions, with or without reoptimization of the coefficients of the configuration state functions (CSFs) and of the orbitals in variational Monte Carlo (VMC). It is found that, with the Slater basis set used here, both state-average and state-specific CAS(6,5) wave functions give an accurate excitation energy in diffusion Monte Carlo (DMC), with or without reoptimization of the CSF and orbital coefficients in the presence of the Jastrow factor. In contrast, the CAS(2,2) wave functions require reoptimization of the CSF and orbital coefficients to give a good DMC excitation energy. Our best estimates of the vertical excitation energy are between 3.86 and 3.89 eV.