Abstract:
In the last decades ab initio methods have become a standard tool in chemistry,
biochemistry, and physics. Nevertheless, the size of treatable systems was limited to the hundred atoms region, even for the less demanding Hartree-Fock (HF) and Kohn-Sham density functional theory (KS-DFT), because of their at least cubic scaling behavior with system size. Ab initio methods of course profit by the fast evolution of computer technology, but their application to larger systems is primarily hampered by the unfavorable scaling behavior. Thus the development of linear scaling methods is a focus of quantum chemistry in order to expand the spectrum of treatable molecular systems.
The aim of the present work is the development of new linear scaling methods for the calculation of molecular properties at the HF and KS-DFT level of theory and quantum Monte Carlo energies by reformulations in terms of electron density matrices.
The first part of this work is concerned with the development of density matrix-based coupled-perturbed self-consistent field methods to determine the response of the system due to static (D-CPSCF) and dynamic (D-TDSCF) perturbations. The linear-scaling D-CPSCF method is applied to the calculation of nuclear magnetic resonance (NMR) chemical shielding tensors, where first results for solid- and solution-state systems are presented. Here, the calculation of systems containing 1000 and more atoms at the Hartree-Fock and Kohn-Sham density functional theory level of theory, respectively, become
possible for the first time on conventional single-processor computer. The D-TDSCF method was applied to the calculation of dynamic polarizability and first hyperpolarizability tensors. Furthermore, the Wigner (2n+1) rule was
employed to determine the first hyperpolarizability by using first order
transition densities only. First exemplary applications are presented to illustrate the accuracy and linear scaling behavior of the new method.
The final part of this work is focused on the development of a linear scaling
method for the local energy in variational and fixed-node diffusion quantum
Monte Carlo, which both combine high accuracy and a favorable scaling
behavior as compared to the more difficult "Post Hartree-Fock" methods like
for example perturbation theory or the coupled cluster approximation. In this thesis a quantum Monte Carlo method with linear scaling effort in the computation of the local energy has been developed by a reformulation of the
corresponding equations in the basis of the N-particle density matrix. Apart
from the derivation and implementation of the new equations, first tests of
their performance are shown.