A Machine Learning Potential for the in silico Design of Functional Organic Molecular Crystals

The design and discovery of functional molecular materials can be greatly accelerated through in silico approaches. Machine learning (ML) models, in particular, demonstrate significant promise for the rapid analysis and manipulation (and re-analysis) of both molecular and crystal structures, essential steps to hasten materials design and discovery. A critical element of these approaches, however, is that robust ML potentials be developed that allow for accurate predictions of energies and forces acting on atoms within a given crystal. Here, we present an ML potential trained on 15,000 molecular crystal structures and more than 2.5 million crystal geometries. The ML potential can predict the energies and forces on atoms in a crystal structure comprised of π-conjugated organic molecules with a mean average error (MAE) of 0.022 eV/atom for energy and 0.056 eV/Å for force when compared with density functional theory (DFT) calculations. To enable the use of the ML potential for in silico discovery of functional molecular crystals, we develop workflows for analyzing crystal surfaces and morphologies, and manipulating and relaxing crystal structures. Furthermore, we deploy these workflows on an open-access web-based tool, OCELOT XtalTransform, that bridges the gap between sophisticated simulations and user-friendly interfaces. By offering these capabilities, OCELOT XtalTransform aims to democratize access to advanced ML potentials for functional molecular crystals, thereby providing new mechanisms for materials design and discovery to the broader scientific community.