Regensburg 2025 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
AKPIK: Arbeitskreis Physik, moderne Informationstechnologie und Künstliche Intelligenz
AKPIK 6: AI Methods for Materials Science
AKPIK 6.3: Talk
Thursday, March 20, 2025, 17:00–17:15, H5
Neural Networks in Surface Crystallography: A New Paradigm for GIWAXS Analysis — •Erwin Pfeiler1, Vladimir Starostin3, Alexander Hinderhofer3, Roland Resel2, Frank Schreiber3, and Stefan Kowarik1 — 1University of Graz, Austria — 2TU Graz, Austria — 3University of Tübingen, Germany
Thin film materials are essential to modern technology, with grazing incidence X-ray diffraction (GIWAXS) being the primary method for resolving their crystal structure. However, the current data analysis process is often slow and labor-intensive, frequently requiring more time and resources than the GIWAXS measurements themselves. Accelerating this bottleneck is critical for enabling automated materials discovery.
In this work, we present an AI-driven approach utilizing neural networks to streamline and enhance GIWAXS analysis. Our model predicts unit cell dimensions (a, b, c) and angles (α, β, γ) directly from the positions of Laue reflections. Additionally, it identifies the Miller indices of the contact plane, providing insights into the film texture.
Using both simulated GIWAXS data and real-world examples, we demonstrate the neural network’s potential to significantly accelerate the analysis process, delivering accurate structural predictions with sub 0.01 Å precision.
Keywords: x-ray; giwaxs; machine learning; neural networks; unit cell