Correlation correction to configuration interaction singles from coupled cluster perturbation theory

TL;DR

This paper introduces a size-extensive correction to CIS excitation energies using coupled cluster perturbation theory, improving accuracy over CIS(D) and comparable to EOM-CC methods.

Contribution

A novel CIS-CCPT method derived from coupled cluster perturbation theory that enhances excitation energy calculations with infinite order effects.

Findings

CIS-CCPT2 achieves a standard deviation error of 0.18 eV for excitation energies.

The method outperforms CIS(D) in accuracy.

CIS-CCPT2 is comparable to EOM-CC in quality.

Abstract

A new state specific correlation correction to configuration interaction singles (CIS) excitation energies is preseted using coupled cluster perturbation theory (CCPT). General expressions for CIS-CCPT are derived and expanded explicitly to first order in the wavefunction and second order in the energy. By virtue of the nature of CCPT this method is {\it a priori} size extensive and incorporates infinite order effects into the wavefunction. This results in a balanced singles space excited state theory that at second order is an improvement over the ubiquitous CIS(D) method and comparable in quality to equation of motion coupled cluster (EOM-CC). A modest test set composed of the first four excited states from nine small organic molecules was used to quantify the accuracy and consistency of the CIS-CCPT2 excitation energies and density of states. We find that CIS-CCPT2 has a standard deviation error of 0.18 eV for excitation energies and 0.14 eV for density of states compared to EOM-CC, a factor of two better than CIS(D) with a significant reduction in the maximum deviation as well