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

CPP: Fachverband Chemische Physik und Polymerphysik

CPP 15: Emerging Topics in Chemical and Polymer Physics, New Instruments and Methods II

CPP 15.1: Vortrag

Montag, 9. März 2026, 17:15–17:30, ZEU/0255

Deep Learning-Driven Grazing Incidence Small-Angle X-ray Scattering Data Processing for Nanostructure Characterization — •Yufeng Zhai¹, Jungui Zhou¹, Shachar Dan¹, Julian Heger², Benedikt Sochor^1,3, Arno Jeromin⁴, Wenbo Wang⁵, Wolfgang Parak⁵, Sarathlal Koyiloth Vayalil^1,6, Thomas Keller^4,5, Andreas Sitierle^4,5, Alexander Hexemer³, Peter Müller-Buschbaum², and Stephan V. Roth^1,7 — ¹DESY, Hamburg, Germany — ²TUM, Garching, Germany — ³ALS/LBNL, California, United States — ⁴CXNS, Hamburg, Germany — ⁵UHH, Hamburg, Germany — ⁶UPES, Dehradun, India — ⁷KTH Stockholm, Sweden

Nanostructured thin films formed via nanoparticle deposition or self-assembly exhibit diverse morphologies that are crucial for their functional properties. Grazing-incidence small-angle X-ray scattering (GISAXS) provides detailed structural information on such systems, but conventional model-based fitting remains limited by simplified assumptions and convergence difficulties. We employ the distorted wave Born approximation (DWBA) to simulate a wide range of two-dimensional GISAXS patterns, which are used to train convolutional neural networks (CNNs) for predicting nanoparticle size distributions. The trained models demonstrate robust performance on both simulated and experimental data, providing a faster and more flexible alternative to traditional fitting. In addition, we have developed a graphical user interface (GUI) that integrates conventional fitting routines with our deep learning framework, providing a user-friendly platform for rapid GISAXS analysis.

Keywords: GISAXS; Deep learning; Convolutional neural network (CNN); Nanostructure characterization; Inverse scattering problem