Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

SYFD: Symposium Pushing the Boundaries of Fair Data Practices for Condensed Matter Insights: From Workflows to Machine Learning

SYFD 1: Pushing the Boundaries of Fair Data Practices for Condensed Matter Insights

SYFD 1.5: Hauptvortrag

Mittwoch, 19. März 2025, 11:45–12:15, H1

Machine Learning and FAIR Data in X-ray Surface Science — •Stefan Kowarik — Phys. Chemistry, Univ. of Graz, Austria

Synchrotrons are among the world's largest producers of scientific data, yet many experiments fail to contribute adequately to databases. Publishing raw data without comprehensive metadata fails to align with the "Findable" and "Reusable" principles of FAIR data, which are essential to unlocking the full potential of these datasets.ML not only benefits from large FAIR datasets but also facilitates their creation. Our recent work highlights live ML-based analysis of X-ray reflectometry (XRR) for thin-film characterization, enabling adaptive experimentation with a fourfold increase in speed. Additionally, we demonstrate automated crystal structure solutions from grazing-incidence X-ray diffraction (GIXD) of thin films. These advancements lay the foundation for self-driving laboratories, where integrated ML algorithms can control thin-film deposition processes, enhancing precision and throughput.Importantly, live ML analysis generates metadata, such as unit cell parameters in textured thin films, improving data findability and reusability. While XRR requires standardized structural model formats*efforts championed by groups like ORSO*GIXD leverages established crystallographic formats for database integration.*In the future, these advancements could culminate in expansive, standardized databases for surface science, encompassing thin-film crystal structures, surface reconstructions, and thin film material properties, analogous to established bulk crystallographic databases.