Identification of crystal symmetry from noisy diffraction patterns by a shape analysis and deep learning

Abstract

The robust and automated determination of crystal symmetry is of utmost importance in material characterization and analysis. Recent studies have shown that deep learning (DL) methods can effectively reveal the correlations between X-ray or electron-beam diffraction patterns and crystal symmetry. Despite their promise, most of these studies have been limited to identifying relatively few classes into which a target material may be grouped. On the other hand, the DL-based identification of crystal symmetry suffers from a drastic drop in accuracy for problems involving classification into tens or hundreds of symmetry classes (e.g., up to 230 space groups), severely limiting its practical usage. Here, we demonstrate that a combined approach of shaping diffraction patterns and implementing them in a multistream DenseNet (MSDN) substantially improves the accuracy of classification. Even with an imbalanced dataset of 108,658 individual crystals sampled from 72 space groups, our model achieves 80.12 ± 0.09% space group classification accuracy, outperforming conventional benchmark models by 17–27 percentage points (%p). The enhancement can be largely attributed to the pattern shaping strategy, through which the subtle changes in patterns between symmetrically close crystal systems (e.g., monoclinic vs. orthorhombic or trigonal vs. hexagonal) are well differentiated. We additionally find that the MSDN architecture is advantageous for capturing patterns in a richer but less redundant manner relative to conventional convolutional neural networks. The proposed protocols in regard to both input descriptor processing and DL architecture enable accurate space group classification and thus improve the practical usage of the DL approach in crystal symmetry identification.

A Semi-Supervised Approach for Automatic Crystal Structure Classification

“…lattices and 144 space groups (we limit the number due to sparse coverage in crystal structure databases), which covers more crystal classes than other studies. Our models also drastically outperform current deep learning approaches for both space group and Bravais Lattice classification (Liu et al, 2019;Garcia-Cardona et al, 2019;Ryu et al, 2019;Ziletti et al, 2018;Tiong et al, 2020) using less training data.…”

“…This outperforms most current models that we are aware of: Liu et al (Liu et al, 2019) used machine learning with a pair-wise distribution function with a top-1 accuracy of 71% and a top-5 accuracy of 90% across 45 space groups. Tiong et al (Tiong et al, 2020) classified x-ray diffraction data into 8, 20, 49, and 72 space groups. Their accuracy decreased from 99% to 80% for 8 and 72 space groups respectively, implying that this accuracy would decrease further if their model was trained on more space groups.…”

Automated prediction of lattice parameters from X-ray powder diffraction patterns

“…In this work, we assume that the crystal system is provided beforehand. In doing so, we assume that in a real operando implementation this information could be obtained by leveraging the highly accurate 1D-CNN models recently developed for crystal-system classification (Park et al, 2017;Vecsei et al, 2019;Oviedo et al, 2019;Suzuki et al, 2020;Tiong et al, 2020). Here, we see our work as highly complementary to other ML-based approaches in the community.…”

“…In particular, the approach used a physics-based augmentation scheme in order to correct for strain and preferred orientation in thin films (Oviedo et al, 2019). More recent work has focused on developing extremely randomized trees for more interpretable ML predictions (Suzuki et al, 2020) and on methods to emphasize differences between patterns with closely related space groups (Tiong et al, 2020). Similar types of classification analysis have also occurred in the fields of electron diffraction (Aguiar et al, 2019), single-crystal X-ray diffraction (Souza et al, 2019) and neutron diffraction (Garcia-Cardona et al, 2019).…”

Disentangling multiple scattering with deep learning: application to strain mapping from electron diffraction patterns

“…Deep learning approaches have been been demonstrated to outperform classical algorithms in variety of computer vision problems in microscopy including classification and segmentation problems [40][41][42]. For instance, deep convolutional neural networks (CNNs) are implemented in the analysis of images collected with various microscopy techniques such as crystal phase classification from back-scattered diffraction patterns [43], structure measurement from electron diffraction and atomic-resolution STEM images [44] and from scanning tunneling microscopy [45], crystal symmetry identification from X-ray diffraction [46], defect analysis from atomic-resolution STEM images [47], crystal tilt and thickness detection from position averaged CBED patterns [48,49], and orientation and strain mapping from 4D-STEM diffraction datasets [50,51]. Li et al used manifold learning to directly classify different features in 4D-STEM data [50].…”