Defect detection in atomic-resolution images via unsupervised learning with translational invariance

Abstract

Crystallographic defects can now be routinely imaged at atomic resolution with aberration-corrected scanning transmission electron microscopy (STEM) at high speed, with the potential for vast volumes of data to be acquired in relatively short times or through autonomous experiments that can continue over very long periods. Automatic detection and classification of defects in the STEM images are needed in order to handle the data in an efficient way. However, like many other tasks related to object detection and identification in artificial intelligence, it is challenging to detect and identify defects from STEM images. Furthermore, it is difficult to deal with crystal structures that have many atoms and low symmetries. Previous methods used for defect detection and classification were based on supervised learning, which requires human-labeled data. In this work, we develop an approach for defect detection with unsupervised machine learning based on a one-class support vector machine (OCSVM). We introduce two schemes of image segmentation and data preprocessing, both of which involve taking the Patterson function of each segment as inputs. We demonstrate that this method can be applied to various defects, such as point and line defects in 2D materials and twin boundaries in 3D nanocrystals.

Insights into ZnO-based doped porous nanocrystal frameworks

“…Crystallographic defects detection based on ADF-STEM one-class support vector machine were proposed by Guo et al 102 ZrO 2 and bilayer (MoW)Te 2 grown from bulk crystal were used for their experimental detection of point, line (in 2D), and boundary (in 3D) defects. HAADF-STEM image-based dopant-segregation induced atomic-scale fractures arising within the Al 2 O 3 ceramic grain boundary core were also confirmed by Kondo et al 103 Loche et al optimized the maximum (7.5%) lanthanum doping in the CeO 2 matrix, which possesses an outstanding oxygen storage capacity.…”

Bringing into play automated electron microscopy data processing for understanding nanoparticulate electrocatalysts’ structure–property relationships

“…In a typical population of nanoparticles, different crystal defects such as twin boundaries, stacking faults, impurities and vacancies can be found, manifesting themselves in HAADF STEM images as local deviations in the contrast and lattice periodicity. A number of studies tackled the challenge of automatic defect detection, including those utilizing a support vector machine as an unsupervised machine learning method [40], geometric graph theory as shown in Figure 2c [35], a weakly supervised approach with a deep neural network [41], and a convolutional neural network [42].…”

Leveraging generative adversarial networks to create realistic scanning transmission electron microscopy images

“…Machine learning (ML) techniques have been widely applied in electron microscopy for applications such as atom localization [1][2][3] , defect identification [4][5][6] , image denoising [7][8][9] , determining crystal tilts and thickness [10][11][12] , classifying crystal structures 13,14 , optimizing convergence angles 15 , identifying Bragg disks 16 , visualizing material deformations 17 , automated microscope alignment 18 , and many others. Several recent reviews [19][20][21] provide an overview of new and emerging opportunities at the interface of electron microscopy and ML.…”