CDS&E: AI-Enabled Analysis of Grain-Boundary-Mediated Hydrogen Diffusion Kinetics within Polycrystals

ABSTRACT The primary objective of this proposal is to develop an artificial intelligence (AI)-enabled computational framework to elucidate the coupled interactions between hydriding phase transformations (HPTs) and grain boundary (GB) deformation across diverse GB types. Grains form the fundamental building blocks of polycrystalline materials, and their boundaries govern bulk properties such as electrical conductivity, diffusivity, and ductility. GB effects on solute- induced phase transformations are particularly critical for energy storage applications, including electrical energy storage in batteries and chemical hydrogen (H) storage in metals. Despite their importance, the structure and evolution of GBs in nanostructured materials, as well as their role in reaction and phase transformation kinetics, remain poorly understood. To address these gaps, this project will integrate AI techniques with a long-time-scale atomistic approach, Diffusive Molecular Dynamics (DMD), to predict HPTs in polycrystals with varied GBs. The study will investigate how GBs influence H diffusion kinetics and pathways, and conversely, how HPTs modify GB structures through the large structural changes associated with hydride formation. Voronoi tessellation will be employed to identify potential H trapping sites within arbitrary GBs, while the Nudged Elastic Band (NEB) method will be coupled with DMD to calculate energy barriers between trapping sites as determined by complex local atomic environments. The resulting DMD-NEB datasets will then serve as training data for machine learning (ML) models capable of rapidly predicting energy barriers on-the-fly during DMD simulations, thereby enabling efficient prediction of local diffusivity and diffusion pathways within GBs. Finally, bicrystals and polycrystals containing both special coincidence site lattice (CSL) GBs and random arbitrary GBs will be constructed and simulated to demonstrate how these interfaces mediate HPTs.