Universally characterizing atomistic strain via simulation, statistics, and machine learning: Low-angle grain boundaries

Abstract

When applied to catalysis and related materials phenomena, grain boundary (GB) engineering optimizes over many currently disparately defined properties. Such properties include GB mobility, solute diffusivity, and catalytic footprints correlating current density with dislocation-induced strain. A recent universalizing framework has systematically classified low-Σ GBs in relation to analogous high-angle references, distinguishing them using footprints formed from the directional straining needed to reversibly yield bicrystals from their separated grains. Correlating the elastic work profiles derived from this thermodynamic process with matching changes in GB dislocations, strain footprints can comprehensively link formerly disparate catalytic properties and materials phenomena. This research investigates such structure-energy correlations to evaluate differences between low-angle (LAGBs) and high-angle (HAGBs) GBs, systematically delineating LAGB-HAGB transitions, explaining their origins, and connecting transitions to materials phenomena. A hierarchical statistical model, nesting GB degrees of freedom within one another, systematically detects such transitions via simplified strain footprints without failure of a single unique GB structure and material combination. A more comprehensive analysis of footprint directional components and discontinuities links transitions to catalytically relevant materials phenomena, describing thermal grooving, shear coupling, complexions, and defect migration under a single universal atomistic framework. With machine learning and spatially generalized strain footprints, this framework reconciles such phenomena via more comprehensive geometry-energy correlations.