Susanne Sauer

• Within this work, an additive and a subtractive QM/MM interface were implemented into CAST. The interactions between QM and MM system are described via electrostatic embedding. Link atoms are used to saturate dangling bonds originating from the separation of QM and MM system. Available energy evaluation methods to be combined include force fields (OPLSAA and AMBER), semi-empirical programs (Mopac and DFTB+), and quantum-chemical methods (from Gaussian, Orca, and Psi4). Both the additive and the subtractive interface can deal with periodicWithin this work, an additive and a subtractive QM/MM interface were implemented into CAST. The interactions between QM and MM system are described via electrostatic embedding. Link atoms are used to saturate dangling bonds originating from the separation of QM and MM system. Available energy evaluation methods to be combined include force fields (OPLSAA and AMBER), semi-empirical programs (Mopac and DFTB+), and quantum-chemical methods (from Gaussian, Orca, and Psi4). Both the additive and the subtractive interface can deal with periodic boundary conditions. The subtractive scheme was extended to enable QM/QM, three-layer, and multi-center calculations. Another feature only available within the subtractive interface is the microiteration procedure for local optimizations. The novel QM/MM methods were applied to the investigation of the reaction path for the complex formation between rhodesain and K11777. Benchmark calculations show a very good agreement with results from Gaussian-ONIOM. When comparing the relative energies obtained with different options to a computation where the whole system was treated with the “QM method” DFTB3, the electrostatic embedding scheme with option “delM3” gives the best results. “delM3” means that atoms with up to three bonds distance to the QM region are ignored when creating the external charges. This is done in order to avoid a double counting of Coulomb interactions between QM and MM system. The embedding scheme for the inner system in a three-layer calculation, however, does not have a significant influence on the energies. The same is true for the choice of the coupling scheme: Whether the additive or the subtractive QM/MM interface is applied does not alter the results significantly. The choice of the QM region, though, proved to be an important factor. As can be seen from the comparison of two QM systems of different size, bigger is not always better here. Instead, one has to make sure not to separate important (polar) interactions by the QM/MM border. After this benchmark study with singlepoint calculations, the various possibilities of CAST were used to approximate the solution of a remaining problem: The predicted reaction energy for the formation of the rhodesain-K11777 complex differs significantly depending on the starting point of the reaction path. The reason for this is assumed to be an inadequate adjustment of the environment during the scans, which leads to a better stabilization of the starting structure in comparison to the final structure. The first approach to improve this adjustment was performing the relaxed scan with a bigger QM region instead of the minimal QM system used before. While the paths starting from the covalent complex do not change significantly, those starting from the non-covalent complex become more exothermic, leading to a higher similarity of the two paths. Nevertheless, the difference of the reaction energy is still around 15 kcal/mol, which is far from a perfect agreement. For this reason, Umbrella Samplings were run. Here, the adjustment of the environment is not done by local optimizations like in the scans, but by MD simulations. This has the advantage that the system can cross barriers and reach different local minima. The relative free energies obtained by Umbrella Samplings with suitable QM regions are nearly identical, independently of the starting point of the calculation. Thus, \(\Delta A\) evaluated by these computations can be assumed to reproduce the real energy change best. An MD simulation that was started from the transition state in order to mimic a “real-time” reaction indicates a very fast adjustment of the environment during the formation of the complex. This confirms that Umbrella Sampling is probably better suitable to describe the reaction path than a scan, where the environment can never move strong enough to leave the current local minimum.…
• In dieser Arbeit wurden ein additives und ein subtraktives QM/MM-Interface in CAST implementiert. Die Wechselwirkungen zwischen QM- und MM-System werden durch elektrostatische Einbettung beschrieben. Link-Atome dienen dazu, lose Bindungen abzusättigen, die durch die Trennung von QM- und MM-System entstehen. Als Methoden zur Energieberechnung, die kombiniert werden können, stehen Kraftfelder (OPLSAA und AM- BER), semiempirische Programme (Mopac und DFTB+) und quantenchemische Verfahren (aus Gaussian, Orca und Psi4) zur Verfügung. Sowohl dasIn dieser Arbeit wurden ein additives und ein subtraktives QM/MM-Interface in CAST implementiert. Die Wechselwirkungen zwischen QM- und MM-System werden durch elektrostatische Einbettung beschrieben. Link-Atome dienen dazu, lose Bindungen abzusättigen, die durch die Trennung von QM- und MM-System entstehen. Als Methoden zur Energieberechnung, die kombiniert werden können, stehen Kraftfelder (OPLSAA und AM- BER), semiempirische Programme (Mopac und DFTB+) und quantenchemische Verfahren (aus Gaussian, Orca und Psi4) zur Verfügung. Sowohl das additive als auch das subtraktive Interface können mit periodischen Randbedingungen verwendet werden. Erweiterungen des subtraktiven Schemas ermöglichen Berechnungen mit QM/QM, drei Schichten o der mehreren QM-Zentren. Ebenfalls nur im subtraktiven Interface verfügbar ist die lokale Optimierung mittels Mikroiterationsschema. ...…