Crystal symmetry determination in electron diffraction using machine learning

Area of Science:

• Materials Science
• Crystallography
• Machine Learning

Background:

• Electron backscatter diffraction (EBSD) is crucial for crystal structure determination but requires manual input for phase identification.
• Current EBSD methods are not optimized for high-throughput analysis due to reliance on human interpretation.

Purpose of the Study:

• To develop a general methodology for rapid and autonomous identification of crystal symmetry from EBSD patterns.
• To enable automated phase identification, transforming EBSD into a high-throughput technique.

Main Methods:

• Utilized a machine learning-based approach for analyzing EBSD patterns.
• Developed a neural network algorithm for autonomous crystal symmetry identification.

Main Results:

• The algorithm successfully identified crystal symmetry from EBSD patterns.
• Evaluated algorithm performance on diffraction patterns not included in the training set.
• The neural network prioritized symmetry features consistent with crystallographer analysis.

Conclusions:

• The developed machine learning methodology facilitates autonomous EBSD phase identification.
• This approach significantly enhances the potential of EBSD as a high-throughput characterization technique.