Computational Studies of Many-body effects in Molecular Crystals
by Thorsten Lennart Teuteberg
Date of Examination:2019-01-25
Date of issue:2019-04-12
Advisor:Prof. Dr. Ricardo A. Mata
Referee:Prof. Dr. Ricardo A. Mata
Referee:Prof. Dr. Dietmar Stalke
Files in this item
Name:thesis_linked_linearized.pdf
Size:5.99Mb
Format:PDF
Abstract
English
A method for multiscale modeling of molecular crystals is presented - the additive crystal quantum mechanics/molecular mechanics model (ac-QM/MM). At the one-body level, a single molecule is chosen as the QM region. The surrounding MM region consists of a finite cluster of explicit MM atoms, represented by point charges and Lennard-Jones potentials, with additional background charges to mimic periodic electrostatics. QM-derived charges are obtained self-consistently to enable polarisation of the environment. An expansion with two-body corrections to the interaction of the central molecule with its nearest neighbours is also included. Fully analytical gradient expressions for both the molecular geometry and the cell parameters have been developed, hence, allowing for an unrestricted optimisation. Both the lattice energy and, in principle, all molecular properties can be calculated within this scheme. Benchmarking the approach with the X23 reference set yields reasonable geometries and cohesive energies for many systems, but hydrogen bonded molecules pose a significant challenge. The inclusion of two-body corrections greatly improves the results. Differences to experimental findings are significant, but similar deviations are obtained across the whole X23 set. A comparison to plane-wave DFT reveals a quite systematic overestimation of cohesive energies by 6-16 kj/mol proving a consistent behaviour of the ac-QM/MM model. The primary objective is to offer an inexpensive and flexible method to account for the crystal environment, but it can be converged to chemical accuracy if correlated wave function methods are applied. Due to the requirement of numerical gradients within the QM part these are only applicable to the smallest systems, but allow for impressive comparisons to experiment. Similar calculations based on DFT optimised geometries unfortunately do not provide overall improved results among the X23 set. The ac-QM/MM model was applied to an interesting crystal structure, which shows an uncommon bending of electron-poor hydrogen atoms towards a cation. From electrostatics, a repulsion between hydrogen atoms (positive partial charge) and the cations would be expected. Gas phase calculations show this expected inclination of hydrogen atoms away from the cation. However, the ac-QM/MM scheme qualitatively reproduces the experimental bending. The investigation of an experimentally challenging X-ray structure of a diselenium compound, providing distinct charge accumulations aside the selenium atoms, could also be aided. An ac-QM/MM study helped to determine a triplet state to be the most probable source of the charge concentrations. The respective structure was found to be mostly similar to the ground state geometry, but a broken Se-C bond leads to a distinctly different orientation of the Se atoms. These are likely responsible for the charge accumulations observed experimentally.
Keywords: molecular crystals; computational chemistry; QM/MM; geometry optimisation; many-body; embedding