Phase Mapping in EBSD Using Convolutional Neural Networks

Area of Science:

• Materials Science
• Crystallography
• Computational Materials Science

Background:

• Commercial electron backscatter diffraction (EBSD) equipment revolutionized materials characterization by providing orientation maps.
• Advancements in EBSD technology have improved data quality, detection rates, and analysis capabilities.
• Current EBSD methods often require significant human intervention for phase identification and mapping.

Purpose of the Study:

• To demonstrate a novel, high-throughput methodology for material phase mapping using machine learning and EBSD.
• To leverage diffraction pattern information directly for automated phase separation.
• To address key challenges in modern EBSD analysis, including accuracy and efficiency.

Main Methods:

• Collected diffraction patterns from diverse material samples.
• Trained a convolutional neural network (CNN), a machine learning algorithm, to recognize subtle differences in diffraction patterns.
• Utilized the trained CNN to autonomously output phase maps based on pattern recognition.

Main Results:

• The machine learning approach accurately separated material phases based on crystal symmetry, chemistry, and lattice parameters.
• The methodology demonstrated the capability to perform phase mapping with reduced human decision-making.
• This study represents the first successful demonstration of machine learning coupled with EBSD for autonomous phase mapping.

Conclusions:

• Machine learning techniques, specifically CNNs, can effectively interpret EBSD diffraction patterns for automated phase mapping.
• This approach offers a significant advancement towards high-throughput materials characterization.
• The developed methodology provides a scalable path for phase mapping as EBSD pattern databases grow.