Projected Coupled Cluster Amplitudes from a Different Basis Set As Initial Guess.

Abstract

Many model chemistry schemes require a sequence of high level calculations, often at the coupled cluster (CC) level, with increasingly larger basis sets. The CC equations are solved iteratively, and the rate of convergence strongly depends on the quality of the initial guess. Here, we propose a strategy to define a better guess from a previous CC calculation with a different basis set by employing the concept of corresponding orbitals. (1, 2) Only the part of the converged amplitudes from the previous calculation that has a large correspondence to the space spanned by the new basis set is projected and used as a new starting point. The computational time for the projection is negligible and significantly reduces the number of cycles necessary for convergence in comparison to the standard guess based on the first order wave function. Numerical results are presented for ground and excited state calculations with the CC singles and doubles (CCSD) and the equation of motion CCSD (EOM-CCSD) methods with the restricted and unrestricted Hartree-Fock (HF) reference functions. However, this approach is more general and can be applied to any truncation of the cluster expansion and reference function.