Adaptive QM/MM simulations for reactions in solution

This thesis documents the development and testing of an adaptive QM/MM scheme suitable for modelling reactions in solution. A QM/MM scheme allows for the study of reactions in large systems, where a Quantum Mechanical (QM) model, required to model the small part of the system where the reaction takes place, is coupled to an empirical potential or Molecular Mechanics (MM) model that is capable of efficiently and sufficiently accurately describing the rest of the system.
Reactions in solution pose a further challenge, in that the QM region must adapt as solvent molecules move and diffuse in and out of the QM region. Existing QM/MM simulations are prone to inaccuracies caused by the QM–MM boundary which are particularly amplified by an adaptive QM region. The currently standard energy mixing QM/MM models are shown to be unsuitable for adaptive QM/MM simulation for this reason, and here we adopt a force mixing approach. The key novel feature of the adaptive QM/MM model developed in this work is the use of a buffer region to minimize inaccuracies at the QM–MM boundary; a feature taken from the Learn-On-The-Fly model for materials systems developed by Csanyi et al. and partially adapted to model pure water by Winfield.
After formulating the adaptive QM/MM model, it is benchmarked against the most advanced adaptive QM/MM model for solutions to date, the Hot Spot method, by modelling pure water, and the importance of the buffer region is demonstrated in order to recover the QM structural properties of water. Subsequently, the new adaptive QM/MM model is demonstrated to accurately reproduce the fully QM free energy curve of a simple reaction, the SN2 reaction of methyl chloride with a chloride anion. Finally, preliminary results of amino acid analogues’ adhesion are shown. These preliminary results involved automating the topology generation and associated MM modelling of amino acid analogues in solution in the neighbourhood of an interface, and are the prerequisites for future adaptive QM/MM simulations of such a system.