The method of choice for the title study, Chemistry in Water, is the Car-Parrinello molecular dynamics (CPMD) simulation technique. CPMD embodies, in an efficient algorithm, the evaluation of the classical Newtonian equations of motion for the molecular dynamics with, simultaneously, the quantum mechanical calculation of the electronic structure, using density functional theory (DFT). The atomic interactions during the molecular dynamics are obtained from the electronic structure calculation, instead of from empirical forcefields, so that the CPMD method is commonly seen as an ab initio molecular dynamics method (AIMD).
The advantages of AIMD for our application are numerous, of which most importantly: 1) the molecular forcefield is obtained from first principles which makes it possible to simulate chemical reactions within molecular dynamics; 2) therefore, we have access to the electronic structure during the molecular dynamics, from which the chemistry can be understood; 3) the solvent effect on chemical reactions is included from the dynamics of explicit solvent molecules (in contrast to e.g. Langevin equation methods and polarizable continuum models), so that solvent behavior, such as changes in the hydrogen bonded network, during chemistry can be studied; 4) additional molecular dynamics techniques, such as periodic boundary conditions and thermostats, realistically mimic the bulk solvent, which carries the inclusion of solvent effects beyond those of microsolvation and cluster studies.
However, models are by definition not perfect. Important approximations in our model originate in the first place from the accuracy of the molecular interactions, which is dictated by the limits of DFT. Particularly, current electron exchange and correlation functionals are still under development and several cases are known where DFT fails. We will leave the construction of improved functionals to others, and use the established Becke-Perdew gradient corrected functional throughout our studies, as it has proven to be very reliable for predicting energies and geometries of molecules, condensed phase systems and transition metal complexes. Secondly, the computational demands of AIMD are very high, which limits the system size to 10-1000 atoms or equivalently, 30-3000 valence electrons and the trajectory lengths to 10-100 picoseconds. This is just enough to include the first and second solvation shells of solutes in water, but approximations have to be made on the long-range interactions. Other approximation of minor significance are 3) the neglect of quantum mechanical effects in the classical molecular dynamics, such as tunneling and zero-point motions; 4) the frozen core approximation; 5) neglect of relativistic effects; 6) the Born Oppenheimer approximation (i.e. the electron subsystem follows the nuclei adiabatically and always remains in the ground level) and 7) the neglect of pressure effects as the simulations are performed in the canonical ensemble. A large part of the work has therefore been dedicated to severe testing and critical comparison of our model to other methods and literature; only a small part of which is found back in this thesis. Nevertheless, AIMD is hands down the best method for the job, and this thesis gives a clear impression of the state-of-the-art on computer simulations of chemical reactions in the condensed phase.