(376e) Automating Transferable Coarse-Model Generation and Back-Mapping with an Open Source Toolchain

Coarse-grained models have enabled access to length scales and time scales in molecular simulations that permit structure-property relationships to be probed in complex soft materials. For example, in previous work, we found that simulation volumes with 35nm edge lengths are necessary for to avoid finite size effects in certain toughened epoxy thermoset systems, and 2.1nm periodic features in organic photovoltaic melts require (10nm)^3 volumes to resolve. When coarse-grained models are derived from more detailed structural information, these techniques can simultaneously provide accurate views of morphology while accelerating equilibration dynamics. Reconstructing atomistic representations is often needed for precise prediction of mechanical or electrical properties. As examples, the elastic modulus of simulated epoxies and the charge transport properties of organic photovoltaic morphologies are far more reliable from atomistically-detailed representations of structure.

In this work, we present coarse-graining infrastructure built using open source molecular modeling tools. We utilize SMARTS strings to define chemistries and coarse-grained mappings. We show how this functionality enables facilities rapid prototyping of coarse models as well as high-throughput screening of multiple chemistries with varying degrees of resolution. In sum, we describe techniques for quickly generating and evaluating models that enable trade-offs between speed and accuracy to be rationally informed. We demonstrate its utility for the multiscale modeling of complex polymer systems. We conclude by discussing examples of how this toolchain can be used to cross-validate simulation predictions, enhance reproducibility, and extensibility.