Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
AKPIK: Arbeitskreis Physik, moderne Informationstechnologie und Künstliche Intelligenz
AKPIK 6: AI Methods for Physics and Materials Science
AKPIK 6.6: Vortrag
Donnerstag, 12. März 2026, 18:00–18:15, BEY/0127
Towards machine-learning-based on-the-fly analysis of neutron reflectometry — •Anne Rentzsch1, Valentin Munteanu1, Oliver Anyanor2, Shreya Shah1, Philipp Gutfreund3, Rémi Perenon3, Anthony Higgins2, Vladimir Starostin4, Alexander Hinderhofer1, Dmitry Lapkin1, and Frank Schreiber1 — 1Institut für Angewandte Physik, Universistät Tübingen, 720726 Tügingen — 2School of Engineering and Applied Science, Swansea University, Swansea SA1 8EN, Wales, United Kingdom — 3Institut Laue-Langevin, 38000 Grenoble, France — 4Cluster of Excellence ’Machine learning - new perspectives for science’, Universität Tübingen, Maria-von-Linden-Straße 6, 72076 Tübingen, Germany
We present a real-time data analysis pipeline for neutron reflectometry that integrates reflectorch, a machine-learning-based software for reflectometry data analysis, into the IT infrastructure at the Institut Laue-Langevin. The workflow was tested during an experiment on the mixing behavior of bilayer thin films. Measured data are automatically reduced and analyzed, and the predicted sample parameters are returned to the instrument control system. The automated analysis can be triggered as frequently as every 10 seconds, enabling parameters and their uncertainties to be tracked with high temporal resolution and supporting continuous monitoring and data-driven adjustments. Compared to conventional software, reflectorch is up to two orders of magnitude faster. Together, these results pave the way for closed-loop experiments and demonstrate the potential of machine learning to enhance the efficiency of neutron reflectometry experiments.
Keywords: Neutron Reflectometry; Machine-learning-based analysis; On-line analysis