Multilevel coupled cluster methods for large molecular systems

Abstract

The coupled cluster hierarchy of models can be used to accurately describe electronic excitations in molecular systems. However, the standard coupled cluster models are expensive and their cost severely limits the size of the molecular systems that can be handled. Multilevel coupled cluster theory can be used to obtain accurate excitation energies and transition strengths at significantly reduced cost. In the multilevel coupled cluster approach, a subset of the molecular orbitals are treated at a higher level of coupled cluster theory than the remaining orbitals. In this thesis, the multilevel coupled cluster hierarchy of models is further extended and optimized implementations are presented. Another challenge for large molecular systems lies in the handling of the electron repulsion integrals. The integral matrix is too large to store in memory and must either be calculated on-the-fly or compactly factorized, for instance by Cholesky decomposition. The thesis introduces a highly efficient Cholesky decomposition algorithm that is specifically designed for the integral matrix.