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

CPP: Fachverband Chemische Physik und Polymerphysik

CPP 12: French-German Session: 2D Materials, Thin Films and Interfaces I

CPP 12.5: Vortrag

Montag, 9. März 2026, 16:15–16:30, ZEU/0255

Machine Learning for Grazing-Incidence Diffraction (GID): from Raw Data to Crystalline Structure — •Dmitry Lapkin, Ainur Abukaev, Ekaterina Kneschaurek, Mikhail Romodin, Constantin Völter, Vladimir Starostin, Alexander Hinderhofer, and Frank Schreiber — Universität Tübingen

The optimization of novel materials in the form of thin films for specific applications requires the appropriate structural characterization methods. X-ray and neutron scattering techniques, such as grazing incidence small- and wide-angle X-ray/neutron scattering (GISAXS/GIWAXS/GISANS) or grazing-incidence diffraction (GID), offer the ultimate spatial resolution and high surface sensitivity, making them indispensable tools for thin film studies. At the same time, advancements in X-ray and neutron sources, in conjunction with developments in area detector technologies, enable the recording of several terabytes of raw two-dimensional detector data within a single experiment. Conventional methods of analyzing the GID data are severely under-paced compared to the data generation rates, representing a significant bottleneck and leaving much of the measured data unanalyzed. To make the thin film studies more efficient and increase the amount of analysed and published GID data that can be reused, we are developing a comprehensive machine learning based data analysis pipeline that goes from raw detector patterns to the corresponding crystalline structures. This pipeline comprises several stages, including data preprocessing and reduction, Bragg peak identification and refinement, and crystal structure determination.

Keywords: X-ray scattering; Machine learning; Data analysis pipeline; GID; GIWAXS