Institute of Solid State Physics


SS23WS23SS24WS24SS25      Guidelines for Master Students

Neural Networks for GIXD Analysis
Erwin Pfeiler
Institute of Solid State Physics
https://tugraz.webex.com/tugraz/j.php?MTID=mec3e2bd1f220b2195c690b475879ad48
11:15 - 12:15 Wednesday 20 November 2024 

Thin film materials are a cornerstone of modern technology, and grazing incidence X-ray diffraction is the main technique of resolving their crystal structure. At the moment, the slow and manual data analysis has been restricted to experts and this phase of the process often requires more time and human resources than the GIXD measurements themselves. Therefore, accelerating this bottleneck step can open avenues to automatic materials discovery.

Here I present an AI-driven approach using neural networks to streamline and enhance GIXD analysis. The neural network predicts unit cell dimensions (a, b, c) and angles (alpha, beta, gamma) from the positions of the Laue reflexes. Those predicted cell parameters are then improved with a least squares fit to obtain high precision results.

Through tests on simulated GIXD data and real-world examples, the neural network's potential to accelerate the analysis process is demonstrated, delivering accurate structural predictions while also opening the possibility of meta-analysis of GIXD experiments, like the effect of missing reflexes or additional reflexes from the sample environment.