Freiburg 2024 – wissenschaftliches Programm

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

A: Fachverband Atomphysik

A 17: Poster III

A 17.7: Poster

Dienstag, 12. März 2024, 17:00–19:00, Tent C

Classifying single-shot diffraction images utilizing machine learning — •Hendrik Tackenberg, Paul Tuemmler, Christian Peltz, and Thomas Fennel — Institute for Physics, University of Rostock, Albert-Einstein-Str. 23-24, D-18059 Rostock, Germany

Single-shot coherent diffractive imaging (CDI) at X-ray free-electron lasers (FELs) has evolved into a well-established method for the structural characterization of unsupported nano-objects with targets ranging from superfluid helium droplets to large biomolecules. Expanding the corresponding experimental setup by additional excitation options, such as short pulse lasers, opens up new routes to study structural dynamics on the femtosecond time and nanometer spatial scale. However, in most scenarios, the dynamics of interest significantly depend on parameters varying on a shot-to-shot basis, such as the objects’ orientations, sizes, or positions in the FEL focus. A rigorous quantitative analysis, therefore, critically depends on the evaluation of a sufficiently large data set to sample the relevant parameter space. Recording millions of scattering images in a single experiment is not unusual nowadays and calls for advanced analysis strategies like model-based forward fitting and automated data set classification.

Here, we present a machine-learning-based classification approach that we successfully applied to characterize a recent experiment studying the strong-field induced anisotropic nanoplasma expansion of laser-driven SiO₂ nanospheres at the European XFEL.

Keywords: coherent diffractive imaging (CDI); X-ray scattering; neural networks; machine learning; European X-FEL (EU-XFEL)