ELEMENTS: py-MCMD: software for hybrid Monte Carlo/molecular dynamics simulations

Wayne State University, Detroit MI

Investigators

Abstract

Physics-based computer simulations are an essential tool for understanding the relationship between atomic-level interactions and physically observable properties of materials. It is from knowledge of structure-property relationships that new materials may be designed, with properties specifically tailored to address the problem of interest. The effectiveness of computer simulation, however, depends primarily on two things: the accuracy of the models used to describe interactions between molecules, and the ability to sample the relevant molecular configurations and conformations for the system of interest. It is the latter problem, improving sampling of phase space in computer simulations, that is addressed in this work. New software is developed that combines two widely used methodologies, molecular dynamics and Monte Carlo, in a single simulation. The strengths of each method are used to overcome barriers to accurate sampling of phase space. The project provides training for graduate and undergraduate students in computer simulation, algorithm design, parallel computation and software development. Participation in computational science is broadened through the creation and distribution of video tutorials and corresponding Python workflows. A new software tool, known as py-MCMD, is used to link an existing molecular dynamics simulation software (NAMD), with the Monte Carlo software GOMC. py-MCMD is updated to significantly reduce latency in Monte Carlo/molecular dynamics (MC/MD) cycles through improved memory and disk usage. Support for multi-scale simulations with on-the-fly changes in the resolution of the system is added using the Nested Monte Carlo Chain approach. Configurational-bias Monte Carlo sampling algorithms are revised to improve their performance on many-core architectures. The resulting open-source software and Python workflows are distributed via GitHub, while video tutorials are shared on YouTube. The work satisfies a need of the research community by providing a hybrid MC/MD software package that is simultaneously high performance, rigorously validated, integrated with existing software, open-source, and user and developer friendly. This award by the NSF Office of Advanced Cyberinfrastructure is jointly supported by the Division of Chemistry. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.