Atomistic simulations of glasses made FAIR
Date:
In this talk I presented amorphouspy, an open Python package for the automated, reproducible atomistic simulation of glasses following FAIR principles. [Documentation]
Abstract: Atomistic simulations are integral to the knowledge and design of glasses, but are nonetheless challenging due to slow structural dynamics, complex workflows, and issues of data management. With advances in high-throughput modeling and machine-learned interatomic potentials, it is worth increasing the Findability, Accessibility, Interoperability, and Reusability (FAIR) of these simulations. Here we present a Python package for the automated setup, execution, and analysis of atomistic simulations of glasses. The software package has semi-to-fully automated modules for the setup of molecular dynamics simulations, preparation of glassy systems, calculation of properties (elastic moduli, viscosity), and structural analysis of the resulting atomic configurations across short- and medium-range order. The computational and workflow management capability is based on the pyiron framework. Showcases of the automated glass preparation, property calculation, and analysis will be presented. The package includes a dedicated web application programming interface for use by large language models. We therefore provide an extensible platform for standardized, FAIR, AI-assisted atomistic simulation of glasses.