Effective core potentials as a pathway to self-interaction error correction: a proof-of-concept study on one-electron systems

TL;DR

This study introduces a simple correction method using effective core potentials to reduce self-interaction errors in density functional theory, demonstrated on one-electron systems and hydrogen transfer reactions.

Contribution

The paper proposes a novel self-interaction correction method (SIP) using effective core potentials, easily implementable in existing quantum chemistry codes, to improve accuracy in one-electron systems.

Findings

SIP reduces self-interaction errors in test systems.

Improves accuracy for systems with functional- or density-driven errors.

Proof-of-concept demonstrates potential for future development.

Abstract

In all applications of Density Functional Theory there is always a degree of one-electron self-interaction error (SIE). Here, we propose a simple self-interaction correction by applying an effective core potential (ECP) that replaces no electrons: we dub this the self-interaction potential (SIP). ECPs are already implemented in all major quantum chemistry codes and so there is minimal effort required by developers and users to access our correction. The goal of SIPs is to reduce the overly severe SIE - commonly manifesting as the unphysical delocalization of an electron over two or more nuclei, or even over an entire chemical system. We propose two first generation SIPs (optimized SIPs and subtraction SIPs) that can reduce the SIE in various one-electron test systems and a hydrogen transfer reaction. Our tests show improvements for systems that suffer from predominantly functional- or density-driven errors, pointing at the potential robustness of the approach. Herein, the viability of SIPs is demonstrated in a proof-of-concept study and several avenues for improvement are identified that will help with the construction of future generations of SIPs.