Interpreting Dynamic Compression Experiments using Machine Learning

2D X-ray diffraction (XRD) patterns capture atomistic detail of the structure & orientation of dynamic compression experiments, encoding critical information about the mechanisms and kinetics of phase transformation. Even with fast, robust and accurate pattern simulation tools, analysis requires time-consuming forward simulation and comparison to experiment. Interpretation and analysis are made more challenging by the extreme paucity of experimental data and the fact that the data is often plagued with artifact from various sources. Therefore, there is a critical need for augmenting current analysis methodologies. In this work we circumvent this challenge with generative models to successfully generate an XRD pattern without background from a given experimentally obtained XRD. Furthermore, we integrate these models with top-of-the-line optimization frameworks to establish an accurate and computationally efficient tool capable of identifying the orientation of the crystal lattice from XRD data obtained from dynamic compression experiments.

SNL is managed and operated by NTESS under DOE NNSA contract DE-NA0003525. SAND No. SAND2025-01320A