The substantial gains in computational power over recent years have been the result of increased processor parallelization. For atomistic simulation, parallelization of spatial dimensions is a viable approach, leading to large length scale simulation. Parallelization of time, on the other hand, can be achieved by averaging of fast time scale motions. Coarse-Grained (CG) models accomplish this by replacing common groups of atoms with 0-dimension points (beads) and defining potentials between the beads. The challenge then is to choose functional forms and parameters for these potentials that accurately reproduce the behavior of equivalent atomistic and experimental systems. In this work, we consider the fundamental unit of water to be a bead representing four water molecules. The beads are made polarizable by the addition of a variable angular term between three charged points on each charge neutral site. Ionic beads are defined as equivalent to four water molecules solvating one atomistic ion. Further, building blocks of linear alkanes are beads of either 3 or 4 carbon units. Non-bonded, non-Coulomb interactions between beads are described by a modified Morse potential which addresses the complexity of multi-atomic beads by decoupling the short-range and long-range behavior. Instead of using a potential of relatively simple form, the comparatively complex potential is computed by lookup table. Optimization of force-field parameters is performed using FFOpt, an in-house software package. The method explores parameter space using Nelder-Mead, a simplex optimization algorithm. The error function for optimization is defined by comparison of CG simulation results to either those of atomistic simulation or experimental values. Analysis used is optimization includes density profile, diffusion coefficient, dielectric constant, and solvation free energies.
Read full abstract