Low-scaling correlated methods for intermolecular interactions

In this work, I develop theories, and their implementations, for the high-accuracy study of large molecular systems. Intermolecular inter- actions are fundamental in nature, in particular in materials chemistry and biological systems. They are difficult to study theoretically due to the small energy differences and vast numbers of molecules involved. Present high-accuracy methods can typically only deal with at most a few molecules in a single calculation, limiting applications to the gas phase. Using absolutely localised molecular orbitals (ALMOs), I develop a new correlated method for intermolecular interactions that is linear scaling in the number of molecules, with accuracy sim- ilar to coupled-cluster with single and double excitations (CCSD). I give details of an implementation that minimises the memory imprint and processor time, as well as allowing for extensive parallelisation. Results over benchmark databases of non-covalent interactions show consistent agreement within 0.5 kcal/mol of the CCSD result, with timings two orders of magnitude smaller. Subsequently, I derive an- alytical derivatives for the total energy, allowing for rapid geometry optimisations on large scale systems. Again, geometrical parameters and vibrational frequencies are shown to agree well with CCSD results. Finally, I extend the ALMO approach to multi-configurational systems, and in particular excited states, demonstrating how it can then be used as an embedding-type method, where different subsys- tems are treated to different levels of accuracy. This culminates in the reproduction and elucidation of experimentally observed shifts in the photoelectron spectrum of phenol when in water, compared to the gas phase. The methods developed herein thus allow for the high- accuracy treatment of much larger condensed-matter systems than has previously been possible.