Analytic Energy Gradients and Spin Multiplicities for Orbital-Optimized Second-Order Perturbation Theory with Density-Fitting Approximation: An Efficient Implementation


Bozkaya U.

JOURNAL OF CHEMICAL THEORY AND COMPUTATION, vol.10, no.10, pp.4389-4399, 2014 (Peer-Reviewed Journal) identifier identifier identifier

  • Publication Type: Article / Article
  • Volume: 10 Issue: 10
  • Publication Date: 2014
  • Doi Number: 10.1021/ct500634s
  • Journal Name: JOURNAL OF CHEMICAL THEORY AND COMPUTATION
  • Journal Indexes: Science Citation Index Expanded, Scopus
  • Page Numbers: pp.4389-4399

Abstract

An efficient implementation of analytic energy gradients and spin multiplicities for the density-fitted orbital-optimized second-order perturbation theory (DF-OMP2) [Bozkaya, U. J. Chem. Theory Comput. 2014, 10, 2371-2378] is presented. The DF-OMP2 method is applied to a set of alkanes, conjugated dienes, and noncovalent interaction complexes to compare the cost of single point analytic gradient computations with the orbital-optimized MP2 with the resolution of the identity approach (OO-RI-MP2) [Neese, F.; Schwabe, T.; Kossmann, S.; Schirmer, B.; Grimme, S. J. Chem. Theory Comput. 2009, 5, 3060-3073]. Our results demonstrate that the DF-OMP2 method provides substantially lower computational costs for analytic gradients than OO-RI-MP2. On average, the cost of DF-OMP2 analytic gradients is 9-11 times lower than that of OO-RI-MP2 for systems considered. We also consider aromatic bond dissociation energies, for which MP2 provides poor reaction energies. The DF-OMP2 method exhibits a substantially better performance than MP2, providing a mean absolute error of 2.5 kcal mol(-1), which is more than 9 times lower than that of MP2 (22.6 kcal mol(-1)). Overall, the DF-OMP2 method appears very helpful for electronically challenging chemical systems such as free radicals or other cases where standard MP2 proves unreliable. For such problematic systems, we recommend using DF-OMP2 instead of the canonical MP2 as a more robust method with the same computational scaling.