Automated Data Analysis for Powder X-Ray Diffraction Using Machine Learning

Introduction Carbon neutrality and electrification of mobility are critical topics in today's industrial world as we increasingly focus on sustainable development goals. Since materials play a decisive role in solving this planet-scale challenge, rapid materials discovery through high-throughput materials synthesis and analysis is crucial to this task. Powder X-ray diffraction (XRD) patterns provide information on composition and crystal structure, which are essential characteristics of materials. Thus, XRD is one of the most fundamental analytical methods in materials research. Recent synchrotron radiation facilities can measure hundreds to thousands of XRD per day [1–3], and a large amount of XRD data is generated daily. Automated data analysis is being actively studied to cope with this tsunami of data [4, 5]. The data are being analyzed in a variety of ways. Especially in the past decade, automated data analysis methods using machine learning have made significant progress, backed by the large crystal structure databases and the dramatic development of machine learning techniques. These automated data analysis techniques enable fast and automated phase identification of XRD patterns, crystal structure analysis by Rietveld analysis, and prediction of material features from XRD patterns. This short commentary article briefly introduces recent advances in materials informatics (MI) on these topics.