Hybrid Simulation Methods for Systems in Condensed Phase

Abstract

Reactions in solution are important in the chemical and pharmaceutical industry as well as in biochemical contexts. In fact, the solvent effects are stronger than many other factors and can slow down or speed up a reaction by many orders of magnitude. However, the complexity of chemistry in solution poses a serious challenge for the computational chemistry community. Therefore, methods have been developed starting from different approximations which are well suited for some aspects but are forced to neglect others. Two aspects have to be considered in order to model a system in solution. The first aspect is the description of the potential energy, which have to be described accurately. Methods ranging from a purely classical description to a complete quantum mechanical one are available. The second aspect concerns temperature effects. Molecular simulations are commonly used although continuum solvation models pose an alternative since they include these effects implicitly. In this work I will present a novel hybrid quantum mechanics/molecular mechanics approach. The solute is described with quantum mechanical methods allowing to account for reactivity and polarisation effects while the solvent is described by molecular mechanics. This strikes a good compromise between accuracy and computational costs for the energetics. The simulations are carried out with the Metropolis Monte Carlo method. High efficiency is achieved by three key approaches: (1) computation of the electrostatic coupling between solute and solvent with 1st order perturbation theory, (2) efficient evaluation of the long-range electrostatics with a shifted force operator, (3) efficient evaluation of the interactions by an numerical integration implemented for graphical processing units. The influence of the parameters inherent to our approach has been thoroughly investigated on a number of benchmark systems. Empirical guidelines have been established along the way which have been used for subsequent applications to biochemically relevant systems e.g. the mutagenic properties of halogenated uracil bases. Properties like solvent structures and electronic spectra as well as relative free energies can be computed efficiently by the here presented approach.