X-ray strain microscopy, Coherent Diffraction Imaging and Machine Learning at the ESRF synchrotron

Abstract: 

X-ray diffraction strain imaging has emerged as a powerful technique for investigating in situ materials evolution, including nanoparticles under catalytic conditions, battery materials during electrochemical cycling, and crystallographic phase transitions. The advent of fourth-generation synchrotron sources, such as the ESRF–EBS in Grenoble, together with the increase in coherent photon flux, now enables the ID01 beamline to reach strain sensitivities better than Δa/a = 10⁻⁵ with spatial resolutions below 10 nm.

I will first show how the use of sub-micron X-ray beams (< 1 μm) in Scanning X-ray Diffraction Microscopy (SXDM) has provided new insights into the deformation mechanisms of charge density wave (CDW) materials under applied current, revealing a previously unobserved transverse deformation induced by surface pinning effect. I will then introduce Bragg Coherent Diffraction Imaging (BCDI), which enables three-dimensional strain mapping of individual nanoparticles smaller than 1 μm. By combining BCDI with complementary techniques available at the ESRF, I will show how this approach helps our understanding biomineralization processes as examples.

Finally, I will discuss how the recent rise of GPU-accelerated Machine and Deep Learning (DL) methods, together with user-friendly Python packages such as scikit-learn, TensorFlow, and PyTorch, is opening new opportunities for synchrotron data analysis. I will illustrate how these tools can be applied to BCDI data pre-processing—such as clustering-based artifact removal, deep-learning-based detector gap inpainting—as well as to data analysis, including deep-learning-enhanced phase retrieval and crystallographic defect classification.

Bibliography:
[1] Machine learning assisted masking of parasitic signals in Bragg coherent diffraction imaging
Ewen Bellec et al.
J. Synchrotron Radiation 33.2 (2026).
[2] Phase Retrieval of Highly Strained Bragg Coherent Diffraction Patterns using Supervised Convolutional Neural Network
Matteo Masto, Ewen Bellec et al.
arXiv:2507.06644 (accepted)
[3] Patching-based deep-learning model for the inpainting of Bragg coherent diffraction patterns affected by detector gaps
Matteo Masto, Ewen Bellec et al.
Journal of Applied Crystallography (2024)
[4] A convolutional neural network for defect classification in Bragg coherent X-ray diffraction
Bruce Lim, Ewen Bellec et al.
npj Computational Materials 7.1 (2021)
[5] Evidence of charge density wave transverse pinning by x-ray microdiffraction
Ewen Bellec et al.
Physical Review B (2020)