Density-Based Formulation of Multi-Level Hartree-Fock Theory

Abstract

A scheme for multi-level (embedded) Hartree-Fock theory is developed. The goal is to reduce computational costs by treating different parts of an electronic system with different degrees of accuracy. In this thesis, a two-level scheme is considered, where one part is optimized through SCF iterations, and the other kept constant throughout optimization. Cholesky decomposition is used to partition the start guess density matrix. Test runs on some simple systems are presented, and show reasonably good agreement with otherwise equivalent non-embedded calculations, although we cannot at the present make any conclusions as to whether the method does in fact lower computational costs.