Assessing the Accuracy of Various Ab Initio Methods for Geometries and Excitation Energies of Retinal Chromophore Minimal Model by Comparison with CASPT3 Results.

Abstract

The effect of the quality of the ground-state geometry on excitation energies in the retinal chromophore minimal model (PSB3) was systematically investigated using various single- (within Møller-Plesset and coupled-cluster frameworks) and multiconfigurational [within complete active space self-consistent field (CASSCF) and CASSCF-based perturbative approaches: second-order CASPT2 and third-order CASPT3] methods. Among investigated methods, only CASPT3 provides geometry in nearly perfect agreement with the CCSD(T)-based equilibrium structure. The second goal of the present study was to assess the performance of the CASPT2 methodology, which is popular in computational spectroscopy of retinals, in describing the excitation energies of low-lying excited states of PSB3 relative to CASPT3 results. The resulting CASPT2 excitation energy error is up to 0.16 eV for the S0 → S1 transition but only up to 0.06 eV for the S0 → S2 transition. Furthermore, CASPT3 excitation energies practically do not depend on modification of the zeroth-order Hamiltonian (so-called IPEA shift parameter), which does dramatically and nonsystematically affect CASPT2 excitation energies.