Simulated X-ray Diffraction and Machine Learning for Interpretation of Dynamic Compression Experiments

X-ray diffraction data collected under dynamic compression conditions are extremely challenging to interpret due to non-ideal sources and geometries, in addition to an extreme paucity of data. We resolve these issues with a robust simulated X-ray diffraction tool based within the LAMMPS molecular dynamics code that enables the generation of accurate and realistic diffraction patterns in a computationally efficient manner. Subsequently, we leverage this enhanced simulation tool to generate a diverse dataset on which we train machine learning models that are capable of removing noise/background and extracting lattice symmetry and orientation from experimental diffraction experiments. Specifically, we have developed two ML-based tools. The first tool is a convolutional neural network (CNN) capable of identifying the lattice and orientations of the underlying structure from multicomponent experimental diffraction patterns. The second tool identifies and separates X-ray diffraction contributions from pulsed power specific background, including reflection and detector noise using a CNN-based Auto-Encoder/Decoder architecture.



Sandia National Laboratories is a multi-mission laboratory managed and operated by National Technology and Engineering Solutions of Sandia, LLC, a wholly owned subsidiary of Honeywell International, Inc., for the U.S. Department of Energy’s National Nuclear Security Administration under contract DE-NA-0003525. Sand no. SAND2022-2819 A