Coarse-Grained Modeling in the Materials Science Suite

The Material Science Coarse-Grained Modeling facility is intended for molecular dynamics simulations of larger bulk systems over more extended period of time than all-atom (AA) models. It achieves this goal by mapping multiple atoms onto single interaction sites. Using fewer sites to represent the same chemical entities means that larger systems can be run with the same computer resources. Coarse grained (CG) simulations run faster than AA simulations for three main reasons:

• CG simulations have a lower density of interacting sites (D). As the computer time per time step roughly scales with D², a simulation with 1/10 the number of sites should use about 1/100 of the CPU time to calculate the forces and potential in the entire system.
• CG models can use time steps that are typically more than 3 times larger than AA models. Time steps are limited by the highest frequency modes in the system. CG sites are heavier and CG potentials tend to be softer, so for a CG model the fastest degrees of freedom have lower frequencies than the fastest degrees of freedom in the corresponding AA model.
• CG systems tend to evolve faster than AA systems. The removal of high frequency modes and the softer potentials result in inherently faster diffusion, i.e., the system changes more per unit of simulated time. The speedup is strongly dependent on the system and its characteristics, with speedup factors ranging from 1 to more than 10. This artificial speedup can alter the kinetics; however in many cases the resulting reduction in computer time is beneficial.

CG molecular dynamics simulations are performed using Desmond on GPGPUs.

The topics listed below provide general background information and advice on using the coarse-grained modeling capabilities of the Materials Science suite.