Analytical Gradient Theory for Resolvent-Fitted Second-Order Extended Multiconfiguration Perturbation Theory (XMCQDPT2)

TL;DR

This paper develops an analytical gradient algorithm for XMCQDPT2 with resolvent-fitting, enabling efficient optimization of molecular geometries and electronic states, especially for moderate active spaces, with demonstrated accuracy and parallel performance.

Contribution

It introduces an analytical gradient method for resolvent-fitted XMCQDPT2, enhancing geometry optimization and state analysis capabilities in multiconfigurational perturbation theory.

Findings

Resolvant-fitting approximations are accurate compared to canonical XMCQDPT2.

The algorithm effectively optimizes MECI geometries of retinal chromophore models.

Parallel implementation shows good computational efficiency.

Abstract

We present the formulation and implementation of an analytical gradient algorithm for extended multiconfiguration quasidegenerate perturbation theory (XMCQDPT2) with the resolvent-fitting approximation by Granovsky. This algorithm is powerful when optimizing molecular configurations with a moderate-sized active space and many electronic states. First, we present the powerfulness and accuracy of resolvent-fitting approximations compared to the canonical XMCQDPT2 theory. Then, we demonstrate the utility of the current algorithm in frequency analyses, optimizing the minimum energy conical intersection (MECI) geometries of the retinal chromophore model RPSB6, and evaluating nuclear gradients when there are many electronic states. Furthermore, we parallelize the algorithm using the OpenMP/MPI hybrid approach. Additionally, we report the computational cost and parallel efficiency of the program.