Mainz 2026 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
MO: Fachverband Molekülphysik
MO 12: Poster – Novel Experimental Approaches and Novel Theoretical and Computational Approaches
MO 12.1: Poster
Tuesday, March 3, 2026, 17:00–19:00, Philo 1. OG
Analysis of ultrafast structural dynamics based on single-shot XUV FEL data through deep learning — •Viktoria Chatrchyan1, Alexander Magunia2, Thomas Pfeifer2, and Tiago Buckup1 — 1Physikalisch-Chemisches Institut, Ruprecht-Karls Universität Heidelberg, Germany — 2Max-Planck-Institut für Kernphysik, Heidelberg, Germany
Ultrafast time-resolved spectroscopy of core-electron excitation with extreme-ultraviolet (XUV) and x-ray free-electron lasers (FELs) is an effective approach to study structural dynamics and relaxation kinetics in atoms and molecules in real time. However, the kinetic analysis of such data is often accompanied by many challenges due to the complex, partially coherent nature of SASE FEL pulses. To compensate for the lack of control over exact machine parameters during a single experiment, the analysis is performed on averaged data from hundreds of single shot measurements described with statistical parameters. Thus, information about the direct impact of an individual laser pulse on a system is often lost. The goal of this work is to extract the exact correlation between single-shot laser pulse characteristics and the resulting changes in a molecular system. For this purpose, a deep neural network is being trained on transient absorption data and XUV FEL pulse spectra simulated using a partial-coherence method that predicts the structural dynamics of a system. This tool promises significant advantages for FEL science research allowing instantaneous and exact analysis of large amounts of data without omitting crucial information from single experiments.
Keywords: Machine Learning; XUV FEL spectroscopy