[Paper Review] Paradigm shift in electron-based crystallography via machine learning

Kevin Kaufmann, Chaoyi Zhu+5 more|arXiv (Cornell University)|Feb 10, 2019

X-ray Diffraction in CrystallographyMaterials Science50 references43 citations

TL;DR

The paper presents a deep learning approach that autonomously identifies crystal structures from EBSD patterns without prior knowledge, outperforming traditional Hough-based EBSD and revealing learned crystallographic features.

ABSTRACT

Accurately determining the crystallographic structure of a material, organic or inorganic, is a critical primary step in material development and analysis. The most common practices involve analysis of diffraction patterns produced in laboratory XRD, TEM, and synchrotron X-ray sources. However, these techniques are slow, require careful sample preparation, can be difficult to access, and are prone to human error during analysis. This paper presents a newly developed methodology that represents a paradigm change in electron diffraction-based structure analysis techniques, with the potential to revolutionize multiple crystallography-related fields. A machine learning-based approach for rapid and autonomous identification of the crystal structure of metals and alloys, ceramics, and geological specimens, without any prior knowledge of the sample, is presented and demonstrated utilizing the electron backscatter diffraction (EBSD) technique. Electron backscatter diffraction patterns are collected from materials with well-known crystal structures, then a deep neural network model is constructed for classification to a specific Bravais lattice or point group. The applicability of this approach is evaluated on diffraction patterns from samples unknown to the computer without any human input or data filtering. This is in comparison to traditional Hough transform EBSD, which requires that you have already determined the phases present in your sample. The internal operations of the neural network are elucidated through visualizing the symmetry features learned by the convolutional neural network. It is determined that the model looks for the same features a crystallographer would use, even though it is not explicitly programmed to do so. This study opens the door to fully automated, high-throughput determination of crystal structures via several electron-based diffraction techniques.

Motivation & Objective

• Develop a machine learning workflow to rapidly and autonomously identify crystal structures from EBSD patterns of metals, alloys, ceramics, and geological specimens.
• Evaluate performance on patterns from samples unknown to the model with no human input or data filtering.
• Compare the ML approach to traditional Hough transform EBSD that requires prior phase determination.
• Visualize and interpret the internal features learned by the neural network to assess alignment with crystallographer intuition.
• Demonstrate the potential for fully automated, high-throughput crystal-structure determination via electron-based diffraction techniques.

Proposed method

• Construct a deep neural network model to classify EBSD diffraction patterns to a specific Bravais lattice or point group.
• Train the model on diffraction patterns from materials with known crystal structures.
• Evaluate the model on samples unknown to the computer with no human input or data filtering.
• Visualize the symmetry features learned by the convolutional neural network to interpret its decision basis.
• Compare performance to traditional Hough-transform EBSD that requires prior phase knowledge.
• Demonstrate applicability to multiple electron-based diffraction modalities through EBSD.

Experimental results

Research questions

• RQ1 Can a deep neural network classify EBSD patterns to the correct Bravais lattice or point group without prior knowledge of the sample?
• RQ2 How does the ML approach compare to traditional Hough transform EBSD in requiring prior phase information?
• RQ3 Do the learned features in the CNN align with crystallographer-recognized symmetry features?
• RQ4 Is the method robust when presented with unknown samples and no data filtering?
• RQ5 What is the potential for automated, high-throughput crystal-structure determination using this approach?

Key findings

• The model classifies EBSD patterns to a Bravais lattice or point group for metals, alloys, ceramics, and geological specimens.
• The approach works on samples unknown to the computer without human input or data filtering.
• Compared to Hough-transform EBSD, the ML method does not require prior phase determination.
• Visualization shows the network learning symmetry features akin to those used by crystallographers.
• The study suggests a pathway to fully automated, high-throughput crystal-structure determination via electron-based diffraction.