Quintic-scaling rank-reduced coupled cluster theory with single and double excitations (2109.08583v2)

Abstract: We consider the rank-reduced coupled-cluster theory with single and double excitations (RR-CCSD) introduced recently [Parrish \emph{et al.}, J. Chem. Phys. {\bf 150}, 164118 (2019)]. The main feature of this method is the decomposed form of the doubly-excited amplitudes which are expanded in the basis of largest magnitude eigenvectors of the MP2 or MP3 amplitudes. This approach enables a substantial compression of the amplitudes with only minor loss of accuracy. However, the formal scaling of the computational costs with the system size ($N$) is unaffected in comparison with the conventional CCSD theory ($\propto N^6$) due to presence of some terms quadratic in the amplitudes. We show how to solve this problem, exploiting the fact that their effective rank increases only linearly with the system size and reduce the scaling of the RR-CCSD iterations down to the level of $N^5$. This is combined with an iterative method of finding dominant eigenpairs of the MP2 or MP3 amplitudes which eliminates the necessity to perform the complete diagonalization. Next, we consider the evaluation of the perturbative corrections to the CCSD energies resulting from triply excited configurations. The triply-excited amplitudes present in the CCSD(T) method are decomposed to the Tucker-3 format using the higher-order orthogonal iteration (HOOI) procedure. This enables to compute the energy correction due to triple excitations non-iteratively with $N^6$ cost. The accuracy of the resulting rank-reduced CCSD(T) method is studied both for total and relative correlation energies of a diverse set of molecules. Accuracy levels better than 99.9\% can be achieved with a substantial reduction of the computational costs. Concerning the computational timings, break-even point between the rank-reduced and conventional CCSD implementations occurs for systems with about $30-40$ active electrons.