Parts | Days | Selection | Search | Updates | Downloads | Help
MO: Fachverband Molekülphysik
MO 2: X-ray FELs (joint session MO/A)
MO 2.4: Talk
Monday, March 14, 2022, 14:45–15:00, MO-H5
Universal Reconstruction of Nanoclusters from Wide-Angle X-Ray Diffraction Patterns with Physics-Informed Neural Networks — •Thomas Stielow and Stefan Scheel — Institut für Physik, Universität Rostock, Albert-Einstein-Straße 23, 18059 Rostock
Single-shot diffraction imaging by soft X-ray laser pulses is a valuable tool for structural analysis of unsupported and short-lived nanosystems, while the exact inversion of the scattering patterns still proves challenging [1]. Deep learning, on the other hand, is widely used in data sciences for the extraction of information from images and has recently been used to accelerate parameter reconstructions from wide-angle scattering patterns [2]. Here, we show how a deep neural network can be used to reconstruct complete three-dimensional object models of uniform, convex particles from single two-dimensional wide-angle scattering patterns. Through physics-informed training the reconstructions achieve unprecedented levels of detail on real-world experimental data [3].
[1] I. Barke et al. Nat. Commun. 6, 6187 (2015).
[2] T. Stielow et al. Mach. Learn.: Sci. Technol. 1, 045007 (2020).
[3] T. Stielow and S. Scheel, Phys. Rev. E 103, 053312 (2021).