Speeding Up Intrinsically Slow Collective Processes in Particle Simulations by Concurrent Coupling to a Continuum Description

Abstract

The difficulty to study intrinsically slow collective processes by computer simulation of particle models stems from multiple disparate time scales (e.g., stiff bonded interactions versus soft nonbonded interactions). Continuum models, which describe the system by collective variables rather than the coordinates of the individual molecular constituents, often do not suffer from this time-scale problem because the stiff microscopic degrees of freedom have been integrated out. We propose to concurrently couple these two descriptions by a heterogeneous multiscale method. We illustrate the technique by studying the Lifshitz-Slyozov coarsening mechanism in a binary polymer blend using a soft coarse-grained particle model and a Landau-Ginzburg-de Gennes free energy functional, respectively. A speedup of up to two orders of magnitudes is achieved.