[3J1-OS-15a] OS-15

10:15 AM - 10:30 AM JST(1:15 AM - 1:30 AM UTC)

[3J1-OS-15a-06] Space-Group Identification from Multi-Phase Powder XRD via Latent Phase Modeling

〇Tomoya Murata¹ (1. Toyota Motor Corporation)

Keywords:

Machine Learning,Materials infomatics,X-ray difffraction

Space-group identification from powder X-ray diffraction (XRD) is a critical early step in crystal structure analysis; however, most existing methods assume single-phase samples or mixtures of known phases, making them difficult to apply to multiphase samples containing unknown phases. In this work, we propose a new learning framework based on latent phase modeling for space-group prediction from multiphase powder XRD profiles. By representing the mixed diffraction signal as a set of latent phase components and extracting phase-specific symmetry information through supervised learning, our approach enables space-group identification without explicitly separating individual diffraction patterns. Evaluation on a large-scale simulated multiphase XRD dataset constructed from crystallographic databases demonstrates that the proposed method is effective even in the presence of out-of-library phases.