Screening Functions for Polarizable Force-Fields Via Force-Matching: A Promising Approach for Coarse-Graining Electrostatic Interactions

The most expensive part of classical molecular dynamics simulations is the evaluation of long range electrostatics, which is typically done using Ewald-like summation methods. While there are techniques to accelerate the computations of the reciprocal part by distributing the charges on a regular grid which allows fast NlogN algorithms, the large scale parallelization of Ewald and its applicability to large scale simulations still remain problematic.  In this talk we will discuss the approach to coarse-grain the electrostatic interactions by introducing a set of short-range inter-atomic forces that matches the forces calculated with the original full Ewald treatment of  long-range interactions. In condensed phases the distribution of ions or polar molecules often screens out electrostatic interactions to a significant extent and therefore interactions between charge caring force centers can be effectively approximated (or screened out) with short-range forces. We will show that it is possible to replace the full long-range electrostatic interactions with short-range screened out interactions computed via the force-matching approach. In our approach we use universal screening functions (represented by simple polynomial function) for each class of interaction (charge-charge, charge-induced dipole, and induced dipole-induced dipole) regardless of how many atom types and their distinct pair interactions are in the system.  We tested this approach in simulations using polarizable force-fields and for a large variety of systems, including  water, water/NaCl, NaCL melt, room temperature ionic liquids, and Li-salts in carbonate solvents.  We found an excellent agreement for structural and dynamic properties obtained from simulations using the developed screening functions for electrostatic interactions and the full Ewald approach.