Comparative Morphological Signatures of Strike Ordered Uranium Oxides for Nuclear Forensics

Nuclear forensics remains critical in addressing threats associated with the illicit and malicious use of radioactive materials, necessitating continuous advancements in the characterization of uranium ore concentrates (UOCs). In this study, we investigate the influence of precipitation strike order on the morphological features of U₃O₈ derived from four synthetic routes: ammonium uranyl carbonate (AUC), ammonium diuranate (ADU), uranyl peroxide (UO₄), and magnesium diuranate (MDU). Machine learning classification was performed to differentiate between the synthetic route and strike order, capturing subtle variations in product morphology. Morphological signatures of scanning electron microscopy (SEM) images were analyzed using a convolutional neural network (CNN) architecture based on a ResNet34 backbone for transfer learning, with images processed into "spot" locations that combine multiple magnifications and detector modes to enhance feature extraction. While most synthetic pathways displayed marked differences in morphology between strike orders (p < 0.05), the magnesium diuranate (MDU) route exhibited only minor morphological differences, 42-72% discrimination accuracy, and no statistical significance, as determined by the CNN analysis. These findings indicate that the dependence of the final morphology of UOCs on strike order is dependent on the synthetic route, resulting in distinct signatures for AUC, ADU, and UO₄ but not MDU for the synthesis processes used. A distinct strike order fingerprint may have wider implications for forensic attribution and process monitoring.