Physics-informed Machine Learning Analysis for Nanoscale Grain Mapping by Synchrotron Laue Microdiffraction

TL;DR

This paper introduces a physics-informed machine learning approach combining CNNs and filtering algorithms to achieve nanoscale grain mapping in nanocrystals using synchrotron Laue microdiffraction, surpassing traditional resolution limits.

Contribution

The novel PIML method integrates physics-based filtering with deep learning to resolve nanoscale grain features from diffraction data, improving spatial resolution beyond beam size constraints.

Findings

Successfully mapped grain morphology and orientation of Au nanocrystals

Achieved resolution comparable to electron backscatter diffraction

Applicable to other diffraction-based nanoscale characterization techniques

Abstract

Understanding the grain morphology, orientation distribution, and crystal structure of nanocrystals is essential for optimizing the mechanical and physical properties of functional materials. Synchrotron X-ray Laue microdiffraction is a powerful technique for characterizing crystal structures and orientation mapping using focused X-rays. However, when grain sizes are smaller than the beam size, mixed peaks in the Laue pattern from neighboring grains limit the resolution of grain morphology mapping. We propose a physics-informed machine learning (PIML) approach that combines a CNN feature extractor with a physics-informed filtering algorithm to overcome the spatial resolution limits of X-rays, achieving nanoscale resolution for grain mapping. Our PIML method successfully resolves the grain size, orientation distribution, and morphology of Au nanocrystals through synchrotron microdiffraction scans, showing good agreement with electron backscatter diffraction results. This PIML-assisted synchrotron microdiffraction analysis can be generalized to other diffraction-based probes, enabling the characterization of nanosized structures with micron-sized probes.