Development and Application of Automated Quantum Chemical Workflows for the Computation of Non-Covalent Interactions and Mass Spectra

Abstract

The topic of this thesis is to advance the automated and efficient quantum chemical description of chemical systems, with a particular focus on non-covalent interactions (NCIs) and mass spectrometry (MS). In Chapter 1, both topics are introduced, and the associated challenges and opportunities in their computation are outlined. An overview of the relevant quantum chemical methods and theoretical concepts to accurately describe these systems is given in Chapter 2. The typically large system size of NCI complexes requires the use of efficient approximate low-cost methods, such as density functional theory (DFT), semiempirical quantum mechanical (SQM) methods, or force field (FF) methods, whose accuracy has to be benchmarked against more accurate and robust reference methods. This is addressed in the LNCI16 benchmark study, which assesses the performance of low-cost DFT, SQM, and FF methods for calculating gas-phase interaction energies of 16 very large, non-covalently bound complexes. With system sizes of up to approximately 2000 atoms, this benchmark provides a unique testing ground for evaluating the robustness of computational methods applied to large molecular systems. To describe the binding behavior of supramolecular complexes, their lowest-energy geometry must be determined, taking into account their conformational flexibility, as well as thermal and solvation effects. These factors are explored through a dedicated benchmark study named HS13L, which employs different methods in a multilevel workflow for direct comparison to experimental binding constants. Herein, excellent agreement with the experimental reference values was achieved. <br/> In the second part of this thesis, the computation of MS is investigated. The challenges in computing mass spectra arise from the high energies involved in the experiments, which result in a large number of possible fragmentation reactions that must be computed efficiently and in an automated manner. Since a generally applicable and sufficiently accurate quantum chemical approach for this task is still lacking, a new program, QCxMS2, was developed. Whereas existing quantum chemical approaches, such as QCxMS, are based on molecular dynamics simulations, QCxMS2 follows a novel strategy based on automated reaction discovery. Herein, the development of QCxMS2 is presented, along with its superior agreement with experimental electron ionization (EI) mass spectra compared to its predecessor and main competitor, QCxMS. Furthermore, the extension of QCxMS2 to enable the calculation of collision-induced dissociation (CID) mass spectra is described. As for EI-MS, a significant improvement over QCxMS was demonstrated through comparison with experimental spectra. In conclusion, the compiled benchmark sets in this thesis yield useful information on which method to employ for the efficient modeling of supramolecular complexes. Furthermore, the newly developed open-source software QCxMS2 provides a valuable tool that can be integrated into automated structure elucidation workflows for the identification of unknown compounds.