Rank-Reduced Second-Order Approximate Coupled Cluster Singles and Doubles
Otto Fajen, Stanford University
Second-order approximate coupled cluster theory (CC2) is an O(N5) scaling method that can provide excitation energies with reasonable errors to higher order coupled cluster methods for many types of excitations. Building on recent work on rank-reduced CCSD (RR-CCSD) and rank-reduced equation-of-motion CCSD (RR-EOM-CCSD), which employ a compression of the doubles amplitudes to solve the CC equations in a reduced space, we apply rank-reduction to CC2 and EOM-CC2, introducing RR-CC2 and RR-EOM-CC2, for the low-rank determination of CC2 ground and excited state energies. Because of the efficient factorization of the CC2 doubles amplitudes, we are able to directly compress the CC2 doubles amplitudes, instead of an approximate array, such as the MP2 or MP3 doubles amplitudes. We show that RR-CC2 can reproduce the canonical CC2 ground state energies to within 0.1 kcal/mol while achieving compression to less than 5% of the original noccnvir space, while RR-EOM-CC2 can reproduce EOM-CC2 excitation energies to accuracies of 0.001 eV while compressing the noccnvir space to less than 1% of its original size. The RR-CC2 and RR-EOM-CC2 methods offer efficient and black-box opportunities to extend the regime of systems for which CC2 and EOM-CC2 energies are available.
Abstract Author(s): O. Jonathan Fajen, Edward G. Hohenstein, Todd J. Martinez