Decoding Crystallography from High-Resolution Electron Imaging and Diffraction Datasets with Deep Learning

J. A. Aguiar, M.L Gong, T. Tasdizen
Idaho National Laboratory, Idaho, United States

Keywords: Deep learning, Microscopy, Diffraction

While machine learning has been making enormous strides in many technical areas, it is still massively underutilized in electron microscopy and diffraction. To address this, we will present on a convolutional neural network model for reliable classification of crystal structures from small numbers of electron images and diffraction patterns. Diffraction data containing 571,340 individual crystals divided amongst seven families, 32 genera, and 230 space groups was used to train the network. Even in the worst cases, the system can narrow down the set of space groups to the top two likely options with over 70% confidence. As examples, we benchmarked against a series of materials including alloys to two-dimensional materials to cross-validate our deep-learning model against results from both high-resolution transmission-electron images and diffraction patterns. We present this result both as a research tool and as an example of the power of machine learning concepts for the future of research in engineering and materials research.