Computational Mining of Meso-Scale Physics From High-Energy X-Ray Data Sets

Abstract

Migration of grain boundaries in an idealized, well-ordered polycrystalline sample is an often-studied phenomenon in microstructural materials science. It is the subject of many imaging experiments in two dimensions and simulations in two and three dimensions, the collective knowledge of which has given us many insights into behavior of specific materials. This thesis describes the characterization of the grain boundaries in a specially prepared polycrystalline aggregate of high-purity iron with a bodycentered cubic lattice whose microstructure was imaged non-destructively in three dimensions with near-field high-energy diffraction microscopy (nf-HEDM) which uses synchrotron X-rays as a probe. The sample was imaged using nf-HEDM before and after a cycle of annealing in order to activate boundary migration for a short interval of time. Also described are the development of computational techniques for denoising grain boundary images in three dimensions as well as a scheme to solve the inverse problem of computing the dynamical parameters influencing boundary migration from the observed boundary geometries and transport. The two snapshots of the full three-dimensional grain boundary network were used to quantify the geometry and transport of individual grains and track their progress through the annealing. A study of the influence of grain boundary curvature on boundary velocity revealed a weak correlation for a large fraction of the boundaries.