The nuclear forensics community has long used traditional optical and scanning electron microscopy (SEM) techniques to observe morphological surface features. Recently, interest has increased in determining the sub-surface features of materials in three dimensions. High-energy X-ray 3D imaging techniques using a direct beam (micro computed tomography) or diffracted beam (high energy diffraction microscopy) are well suited to providing qualitative and quantitative sub-surface morphological information and have the added benefit of being non-destructive. This work describes the synthesis, conventional characterization, and synchrotron characterization of morphologically different samples of U₃O₈ as well as the development of machine learning algorithms to decrease the time required to analyze the produced synchrotron data.

Three separate samples of U₃O₈ were synthesized in gram batches from the same UO₃ starting material while varying the calcination temperature from 600 °C to 800 °C. Each of these was then split into several analytical samples for analysis using powder x-ray diffraction (p-XRD) and SEM imaging. The results from these analyses confirmed that the morphology of the samples differed depending on calcination temperature and indicated incomplete transformation of UO₃ to U₃O₈ at the lower temperatures. Subsequent efforts have concentrated on analyzing the samples using high-energy x-ray diffraction microscopy (HEDM) and micro computed tomography (µ-CT). While the sample synthesis produced powders that have grains too small to individually resolve using HEDM, the collected diffraction patterns can be used to produce a 3D reconstruction of crystal phase distributions. Due to the small beam size and relatively high spatial resolution in 3 dimensions, a significant amount of data is generated causing reconstructions to be slow. For example, a single 1.5 µm thick layer of sample takes a few hours. The use of machine learning algorithms and unique analysis methods in this work has resulted in a significant drop of that reconstruction time to ~3 minutes. This work presents initial sample analysis results along with a discussion of the HEDM technique and how it may be beneficial to nuclear forensic analysis.

Elizabeth Laudadio

Appearances