Decoding crystallography from high-resolution electron imaging and diffraction datasets with deep learning

Abstract

While machine learning has been making enormous strides in many technical areas, it is still massively underused in transmission electron microscopy. To address this, a convolutional neural network model was developed for reliable classification of crystal structures from small numbers of electron images and diffraction patterns with no preferred orientation. Diffraction data containing 571,340 individual crystals divided among seven families, 32 genera, and 230 space groups were used to train the network. Despite the highly imbalanced dataset, the network narrows down the space groups to the top two with over 70% confidence in the worst case and up to 95% in the common cases. As examples, we benchmarked against alloys to two-dimensional materials to cross-validate our deep-learning model against high-resolution transmission electron images and diffraction patterns. We present this result both as a research tool and deep-learning application for diffraction analysis.

Symmetry prediction and knowledge discovery from X-ray diffraction patterns using an interpretable machine learning approach

“…This idea to combine multiple ML models is called stacking 46 . Aguiar et al 32 , 33 reports that stacking improves the prediction accuracy for space group determination by electron diffraction. However, for our case, stacking did not improve the performance, i.e., the prediction performance was deteriorated.…”

Recent progress on discovery and properties prediction of energy materials: Simple machine learning meets complex quantum chemistry

“…To establish the strongly correlated connection between spin and orbit in solids, further to understand the structure-activity relationship of energy materials, it is absolutely necessary to use machine learning algorithms in quantum chemistry techniques. Besides decoding the fundamental relationships behinds structure and properties, the breakthrough of technologies, including synchrotron radiation [85], neutron diffraction [86], spherical aberra- https://engine.scichina.com/doi/10.1016/j.jechem.2020.05.044 tion corrected scanning transmission electron microscope [87], and so on, also promote the research and development of energy materials.…”

Advances in the electron diffraction characterization of atomic clusters and nanoparticles

“…18 represents a smaller multidimensional diffraction dataset, where applying the model developed by Aguiar et al which converts each pattern into crystal family prediction at each pixel. The strength associated with each prediction is reported without a-priori or ab-initio-based knowledge 83,84 . By processing each pattern, the probability and prominence between crystal families reflects the underlying structural changes at each hyper pixel, between the presence of a nanoparticle and underlying grid.…”