On the Reduction of Molecular Degrees of Freedom in Computer Simulations

Abstract

Molecular simulations, based on atomistic force fields are a standard theoretical tool in materials, polymers and biosciences. While various methods, with quantum chemistry incorporated, have been developed for condensed phase simulations during the last decade, there is another line of development with the purpose to bridge the time and length scales based on coarse-graining. This is expected to lead to some very interesting breakthroughs in the near future. In this lecture we will first give some background to common atomistic force fields. After that, we review a few common simple techniques for reducing the number of motional degrees of freedom to speed up the simulations. Finally, we present a powerful method for reducing uninteresting degrees of freedom. This is done by solving the “Inverse Problem” to obtain the interaction potentials. More precisely, we make use of the radial distribution functions, and by using the method of Inverse Monte Carlo [Lyubartsev & Laaksonen, Phys. Rev. E. 52, 3730 (1995)], we can construct effective potentials which are consistent with the original RDFs. This makes it possible to simulate much larger system than would have been possible by using atomistic force fields. We present many examples: How to simulate aqueous electrolyte solutions without any water molecules but still having the hydration structure around the ions – at the speed of a primitive electrolyte model calculation. We demonstrate how a coarse-grained model can be constructed for a double-helix DNA and how it can be used. It is accurate enough to reproduce the experimental results for ion condensation around DNA for several different counterions. We also show how we can construct site-site potentials for large-scale atomistic classical simulations of arbitrary liquids from smaller scale ab initio simulations. This methodology allows us to start from a simulation with the electrons and atomic nuclei, to construct a set of atomistic effective interaction potentials, and to use them in classical simulations. As a next step we can construct a new set of potentials beyond the atomistic description and carry out mesoscopic simulations, for example by using Dissipative Particle Dynamics. In this way we can tie together three different levels of description. The Dissipative Particle Dynamics method appears as a very promising tool to use with our coarse-grained potentials.