Relativistic Two-component Double Ionization Potential Equation-of-Motion Coupled Cluster with the Dirac--Coulomb--Breit Hamiltonian

TL;DR

This paper develops and compares relativistic double ionization potential equations-of-motion coupled cluster methods using various Hamiltonian frameworks, demonstrating their accuracy and limitations for atomic and molecular systems.

Contribution

The authors implement relativistic DIP-EOMCC methods within multiple X2C frameworks and evaluate their accuracy against full 4-component calculations and experimental data.

Findings

DCG-X2C-DIP-EOMCCSD agrees well with 4c calculations.

The DCB-X2C-DIP-EOMCCSD overestimates double IPs by >0.25 eV with complete basis.

A composite scheme achieves near-experimental accuracy for atomic systems.

Abstract

We have implemented relativistic formulations of DIP-EOMCCSD and DIP-EOMCCSDT within the 1eX2C and DC-, DCG-, and DCB-X2C frameworks. Direct comparisons against full 4c-DIP-EOMCCSD calculations show excellent agreement with DC(G)-X2C-DIP-EOMCCSD, suggesting, at least for the systems studied herein, two-electron relativistic effects are well-described by the mean-field treatment in mmfX2C, and remaining relativistic two-electron and electron-positron correlation effects are negligible. A subsequent basis set study on vertical double IPs for noble gas and diatomic species has shown that DCB-X2C-DIP-EOMCCSD tends to overestimate double IP values in the limit of a complete one-electron basis, by more than 0.25 eV, on average. For atomic systems, we were able to demonstrate that a composite scheme whereby the dominant correlation effects are captured by large-basis DCB-X2C-DIP-EOMCCSD and remaining high-order correlation effects are approximately modeled via small-basis DCB-X2C-DIP-EOMCCSDT brings the double IP values into excellent agreement with experiment; for Xe atom, for example, absolute errors in double IP values from this approach are less than 0.02 eV. However, we found the ANO-RCC family of basis sets used in our composite approach to have poor convergence behavior in terms of DCB-X2C-DIP-EOMCC calculations, as the estimates computed using the non-relativistic DIP-EOMCC approach at the large basis set limit indicates a larger 0.1--0.2 eV error relative to experimental data.