Correction of Residual Errors in Configuration Interaction Electronic Structure Calculations

TL;DR

This paper introduces a novel defect correction method for configuration interaction calculations that improves energy accuracy across various molecules and electronic states by applying atom-specific energy corrections.

Contribution

The paper presents a new defect correction approach that enhances CI energy calculations without detailed excitation analysis, applicable to diverse molecules and electronic states.

Findings

Corrected CI energies align closely with experimental data.

Method effectively reduces errors in truncated virtual space calculations.

Significantly improves transition energy predictions for transition metals.

Abstract

Methods for correcting residual energy errors of configuration interaction (CI) calculations of molecules and other electronic systems are discussed based on the assumption that the energy defect can be mapped onto atomic regions. The methods do not consider the detailed nature of excitations, but instead define a defect energy per electron that that is unique to a specific atom. Defect energy contributions are determined from calculations on diatomic and hydride molecules and then applied to other systems. Calculated energies are compared with experimental thermodynamic and spectroscopic data for a set of forty-one mainly organic molecules representing a wide range of bonding environments. The most stringent test is based on a severely truncated virtual space in which higher spherical harmonic basis functions are removed. The errors of the initial CI calculations are large, but in each case, including defect corrections brings calculated CI energies into agreement with experimental values. The method is also applied to a NIST compilation of coupled-cluster calculations that employ a larger basis set and no truncation of the virtual space. The corrections show excellent consistency with total energies in very good agreement with experimental values. An extension of the method is applied to dmsn states of Sc, Ti, V, Mn, Cr, Fe, Co, Ni and Cu, significantly improving the agreement of calculated transition energies with spectroscopic values.