Computational Study of Dispersion Interactions through Local Orbital Analysis
by Axel Wuttke
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. Guido Clever
Files in this item
Name:graduate_thesis_Wuttke.pdf
Size:54.4Mb
Format:PDF
Abstract
English
Non-covalent interactions are responsible for molecular aggregation and thus indispensable in the study of condensed matter. In chemical research, such information plays an important role for example in creating new drugs with tailored properties through rational design. The tools of computational chemistry are invaluable to understand such events. Two popular methods are electrostatic potential maps (EPM) and noncovalent interactions (NCI) plots. They are both easy in application and give an intuitively interpretable graphical representation. However, in order to quantify and consistently study intermolecular interactions, higher level methods are required. In this thesis an extension to local correlation methods that allows to visualize dispersion interactions is presented. The so-called Dispersion Interaction Density (DID) is obtained by scaling the local molecular orbitals (LMOs) closed- shell densities with their corresponding contributions to the dispersion interaction. From the DID-matrix a 3-dimensional grid can be calculated and afterwards visualized in the form of contour plots, color projections on the electronic density or Voxel graphics. The latter proves to be the method with the greatest breadth of information. Studies on selected systems such as the well-known benzene dimer demonstrate their usefulness. Furthermore, it is also shown how intramolecular effects can be easily investigated by means of local orbital analysis. Having exhibited that the latter in conjunction with DIDs are well suited to study such effects, more complex structures have been examined. The Clever group has made important advances in metal mediated self assembly of supramolecular compounds. By different substitutions on a bis-monodentate pyridyl backbone two supramolecular structures which can serve as a host system for different guest molecules were obtained. The first host examined has adamatyl residues showing in the cages interior and is able to bind different ionic guest molecules. Local orbital analysis revealed that the main driving forces are given by dispersion and Pauli repulsion. Motivated by these results the Clever group is currently extending the scope of dispersion energy donors (DEDs) that can be implemented into the ligand backbones. The second host system discussed is an interpenetrated double cage structure. It is able to take up various neutral guest molecules in its middle pocket. Local correlation calculations revealed that this is mainly achieved by dispersion interactions. Further, a dependence between the dispersion contributions and experimental free binding energies was found. In order to build a bridge between theory and experiment a protocol to compute free binding energies was applied. For small guest molecules the scheme performed rather well while it could be shown that larger guest molecules require a relaxation of the pockets geometry. For this purpose three more sophisticated computing protocols are proposed. The study of non-covalent interactions is continued by the investigation of metallophilic contacts. The previous work of our group was extended by the investigation of experimental crystal structures and mixed metallophilic contacts. Through local orbital analysis, fragments were built and the interaction energies were decomposed. The outcome revealed that there are good reasons to highlight the strength of metallophilic contacts. However, mistaking such an interaction for the main driving force in molecular aggregation should be avoided. The calculations carried out for the ClAuR2bimy (R=Me, Et) complexes show that simple changes in the ligands can lead to crystal structures where even no aurophilic contacts are observed. The d10 cations can contribute in stabilizing a molecular crystal, but the ligand composition seems to be the dominating factor. In all previous calculations the DF-PAO-LMP2 code implemented in Molpro was used. Especially regarding the supramolecular host-guest structures it was difficult to make the calculations possible at all. This was at least achieved by applying approximations such as multipoles for the calculation of distant orbital pairs. Nevertheless, the calculations were very time intensive. By rediscovering PNO as virtual space the groups of Neese and Werner made great progress in the recent years. The computational cost of the latest generation local correlation methods scales linear with regard to the molecular size. Combined with massive parallelization over many processors large molecular clusters can nowadays be computed within minutes. To benefit from and inspired by recent development in local orbital analysis, an EDA scheme for PNO-LMP2 has been implemented in Molpro. It is shown that strict spatial localization of the PNOs and a reliable population analysis are necessary to obtain stable results, especially for the charge transfer terms. For the latter, a visualization related to DIDs is probed. The improvement of the EDAs stability is, however, also purchased with an increased computational cost of the method. In the future, calculations of large molecular systems must show whether a temporal advantage is obtained in comparison to the initial PAO-based methods.
Keywords: dispersion; energy decomposition analysis; local orbital analysis; dispersion interaction density; supramolecular host-guest chemistry