Machine learning assisted nanobeam X-ray diffraction based analysis on hydride vapor-phase epitaxy GaN.

Area of Science:

• Materials Science
• Crystallography
• Data Science

Background:

• Nanobeam X-ray diffraction (nanoXRD) offers high spatial resolution and rapid data acquisition for in situ crystal structure analysis.
• Analyzing large nanoXRD datasets for defect recognition and structural feature discovery presents significant challenges.
• Machine learning (ML) methods show promise for efficiently analyzing large, complex datasets.

Purpose of the Study:

• To apply a machine learning algorithm, uniform manifold approximation and projection (UMAP), to improve nanoXRD data analysis.
• To enhance the categorization of crystal structures from high-dimensional nanoXRD data.
• To demonstrate the utility of UMAP in analyzing spectroscopic and diffraction data for crystal structure investigations.

Main Methods:

• Utilized uniform manifold approximation and projection (UMAP), a machine learning algorithm.
• Applied UMAP to analyze three-dimensional ω-2θ-φ diffraction patterns from a cross-sectional hydride vapor-phase epitaxy GaN wafer.
• Compared UMAP results with conventional fitting methods for crystal structure analysis.

Main Results:

• UMAP provided more precise categorization of crystal structures compared to conventional fitting methods.
• The high-dimensional data embedding property of UMAP effectively retained data structure, aiding nanoXRD profile analysis.
• Demonstrated UMAP's capability in analyzing other spectroscopic or diffraction datasets for crystal structure guidance.

Conclusions:

• UMAP is a valuable tool for enhancing crystal structure analysis from nanoXRD data.
• UMAP facilitates more accurate defect recognition and discovery of hidden structural features in complex materials.
• The application of UMAP extends to various spectroscopic and diffraction techniques for materials characterization.