L.C. Johnson, F.R. Phelan Jr.
National Institute of Standards and Technology,
United States
Keywords: coarse-grained molecular dynamics, polymer melt, friction coefficient
Summary:
Coarse-grained (CG) models of polymers, in which a number of atoms in an all-atom (AA) representation are subsumed into a single larger simulation site, aim to reduce computational effort yet retain the hierarchy of length and time scales crucial to macromolecules. Two primary classes of coarse-graining methods focus on either preserving chemical specificity, “bottom-up” methods, or reproducing physical properties, “top-down” methods. Here, we combine a bottom-up coarse-grained model with a dissipative potential with the goal of obtaining a chemically specific, thermodynamically consistent, and dynamically correct model. We generate the conservative part of the force field using the iterative Boltzmann inversion (IBI) method which seeks to recover the AA structure and thus thermodynamics of AA simulations. In our IBI code, we employ machine learning and filtering techniques to handle noisy distributions and low sampling regions in an automated manner to enable rapid convergence to smooth force profiles. We develop a similar approach for parameterization of the dissipative potential to correct the dynamics of the IBI-generated force field which suffers from unphysically fast dynamics compared to the AA representation due to a smoother potential energy landscape. In this method, we utilize Langevin dynamics, which introduces friction and random forces set by a scalar friction parameter, as a means to recover the dynamics of the AA representation of the molecular systems. We demonstrate this method for oligomers in the melt state and test the recovery of various measurements of dynamics for consistency with the AA model. Recent efforts include a study of the parameterization of the friction factor to material properties. Development of these methods into complementary automated packages is also presented.