To model large, complex systems, two approaches have been dominant in the past. Either classical force fields have been used to simulate dynamical features, or quantum mechanical calculations on a semiempirical level have been performed on frozen structures to access the electronic structure. Recent developments aim at an integration of these approaches: the important part is treated on a quantum mechanical level, e.g. to describe chemical reactions involving bond breaking and formation, or the charge carrier dynamics of the system. Coupled to this part is an environment described on a classical level. In contrast to established QM/MM procedures, we define the quantum part of the system not spatially, but by means of a sigma-pi-separation. In this manner, we avoid a number of technical problems and are able to study systems containing a large quantum mechanical part. We have implemented the procedure into the TINKER and the AMBER program packages, and have – to our knowledge – performed the first dynamic simulation of biological charge transfer in a realistic system, i.e. DNA oligomers. Similar simulations have been performed for the interface of an organic semiconductor to water and organic cations in solution (figure: triarylamine in dichloromethane).
Many macroscopic properties like the conductivity cannot be directly deduced from molecular calculations. To address this problem, we have developed a multi-scale procedure with the geometric basis of classical Monte Carlo simulations. On this, we define a quantum mechanical Hamiltonian that is carefully parameterized using ab initio computations, which leads to the parameters of Marcus’ theory via the variational approach to molecular charge transfer. Marcus’ theory and linear response concepts give local conductivities that enter a Kirchhoff network, the resulting conductivity values have to be averaged in a nontrivial way involving percolation arguments. We use similar approaches to describe nanoscopic conductivity setups at ambient conditions.
