(736e) A General Algorithm for Efficient Reverse-Mapping of Coarse-Grained Configurations to the Atomistic Scale

Nowak, C., Cornell University
Escobedo, F., Cornell University
Misra, M., Cornell University
Despite the ever-growing computational resources, atomistic models (AM) are often intractable for the length and time-scales required for molecular simulations. Coarse-grained models (CG), that integrated out many degrees of freedom, have been successful at “unlocking” the length and time scales needed to perform meaningful computer experiments. The main drawback to CG models is that some quantities of interest like charge transfer in electronically conducting polymers, require atomistic resolution. As such, CG models are used to quickly achieve equilibrium at coarser length scales, and then the molecues “reverse-mapped” (or reverse coarse-grained, RCG) into AM so that the molecules can be relaxed at atomistic length scales. This RCG step is often hard-coded and can be difficult to modify for arbitrary chemistries. With the prevalence of generalized software for various computational techniques such as LAMMPS (molecular dynamics), Cassandra (Monte Carlo), and ABINIT (density functional theory), we present a generalized software that requires only two initial inputs, an atomistic input file containing the molecular topology (e.g., pdb, lammps data file), and a corresponding input file containing the CG configuration and topology. The initial RCG procedure self-consistently determines the coarse-graining scheme, and creates coarse-grain bead files, which can be used in subsequent iterations of the program to reduce program runtime. We will present the current implementation of the program, sample applications, as well as propose potential future extensions.