Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
KFM: Fachverband Kristalline Festkörper und deren Mikrostruktur
KFM 24: New Methods and Developments: Spectroscopies, Diffraction and Others (joint session O/KFM)
KFM 24.5: Vortrag
Donnerstag, 8. September 2022, 11:45–12:00, H6
Unsupervised machine learning-assisted analysis of multidimensional ARPES data — •Steinn Ymir Agustsson1, Mohammad Ahsanul Haque2, Fatemeh Zardbani2, Davide Mottin2, Panagiotis Karras2, and Philip Hofmann1 — 1Institute of Physics and Astronomy, Aarhus University, Denmark — 2Institute of Computer Science, Aarhus University, Denmark
In recent years, the size and complexity of experimental data sets has been dramatically growing in many fields of science. For photoemission spectroscopy, the development of novel detectors and multi-dimensional measurement modes (e.g., including a time dependence or spatial dependence), has lead to orders of magnitude more data being produced. Even after a necessary upgrade of the data management system, it remains highly challenging to visualize and superficially interpret the data fast enough to feed back into decisions about what to measure in an ongoing experimental run. A promising approach to address this is the application of machine learning tools. These have shown promising results when applied to data reduction and feature detection tasks in many fields of science. We have developed an unsupervised clustering method which is able to distinguish differences between ARPES spectra obtained from different spatial locations in nanoARPES measurements. This enables quick and automatic identification and classification of regions with different spectral features, allowing to invest more time in the collection of significant data.