Advances in artificial intelligence-based approaches to enhance dark field X-ray microscopy analysis

Abstract Dark field X-ray microscopy (DFXM) has emerged as a powerful technique for characterizing dislocations in bulk crystalline materials, whose high penetration depth and non-destructive evaluation offer unique advantages over traditional electron microscopy methods. The interpretation and analysis of the DFXM data presents significant challenges that have limited its broader adoption. Here we review our recent advances using artificial intelligence (AI) methods to enhance DFXM analysis, particularly focusing on dislocation characterization. We discuss the development of physics-informed AI approaches that combine theoretical understanding with data science techniques to enable both time-resolved dislocation dynamics studies and dimensional reduction of complex DFXM datasets. Our work demonstrates how semi-automated workflows, guided by dislocation theory and employing techniques such as wavelet transforms and Bayesian inference, can effectively track and analyze dislocation behavior across multiple time scales. These methodologies have been successfully applied to various materials science challenges, from studying thermally activated dislocation motion to characterizing dislocation networks. By presenting our works that integrate physics-based modeling into AI capabilities, we demonstrate how our and other works can extract new important quantitative dislocation data from DFXM measurements. Graphical abstract