Development and application of embedded methods to strongly and weakly correlated systems

Abstract

Coupled cluster (CC) theory is a popular and reliable tool in quantum chemistry due to its improvable hierarchy of methods able to rapidly converge to the full configuration interaction (FCI) limit in weakly correlated systems. Although it represents one of the most efficient single reference methods to treat many-body correlations with high accuracy and reliable outcomes, it yields qualitatively erroneous results when applied to strongly correlated systems. Within this thesis, the Distinguishable Cluster (DC) method (i.e., a small modification of CC amplitude equations able to qualitatively describe strongly correlated systems), is combined with FCI Quantum Monte Carlo (FCIQMC) in order to present a new tailored approach, the tailored DC (TDC), which is more accurate than the corresponding tailored CC and the pure DC. To demonstrate this, the method is first benchmarked with a variety of test cases and then further evaluated with computation of spin-state splittings in a few Fe(II) complexes. The systematic improvability of the TDC method is shown as the active space is increased. In the last part of the thesis, a further embedding scheme to treat strong correlation effects is evaluated. Specifically, the development and application of a screened Coulomb formalism is discussed. This simple approach inspired by Random Phase approximation (RPA) shows to be extremely efficient in the dissociation of one- and two-dimensional hydrogen systems.