SI2-SSE: Highly Efficient and Scalable Software for Coarse-Grained Molecular Dynamics

点击翻译按钮获取中文摘要

Molecular simulation provides a powerful complement to conventional experimental techniques, offering both high-resolution information and unusual levels of control over the experimental conditions. While atomic-resolution molecular simulations are well established and widely used, it is also possible to remove extraneous detail from molecular representations to create highly efficient "coarse-grained" (CG) models. CG approaches can expand the potential applications of molecular simulations far beyond atomic-resolution models: the computational efficiency of CG models allows the scientist to investigate not only significantly larger systems, but also phenomena that require significantly longer time scales. These CG approaches are of particular interest in the study of systems where key aspects of various processes emerge from interactions between large numbers of molecules over relatively long distances. CG models can therefore provide crucial insight into the molecular basis of such systems, e.g., new materials. However, CG models can require significant scientific understanding to create and use effectively. To bring cutting-edge CG methodologies into a wider degree of use, this project will implement key algorithmic advances and associated CG functionalities into the widely-used LAMMPS molecular dynamics simulation code. Furthermore, the project will implement a publically-accessible repository for CG model parameters and input files to accelerate the dissemination of exemplar CG models throughout the scientific community.The project will integrate key functionalities for very large-scale and dynamic CG models into the LAMMPS molecular dynamics package. These functionalities include not only sparse memory optimizations (e.g., template molecular topology descriptions and spatial data structures for link cell algorithms) but also user-defined transition information for the propagation of "ultra-coarse-grained" (UCG) models; parameterization of the latter can be achieved by using the integrated multi-scale coarse-grained force matching code (MSCGFM). Furthermore, direct incorporation of experimental data into CG models will be assisted by implementations of the "experiment directed metadynamics" (EDM) and "experiment directed simulation" (EDS) algorithms. Taken together, these enhancements will provide cutting-edge CG model generation and simulation techniques to a wide user community. To complement the extended functionality of the LAMMPS code, a user-driven data and metadata repository for CG models will be provided to assist with efficient dissemination of model parameters and simulation/validation data to the scientific community.