Mapping mesoscopic phase evolution during E-beam induced transformations via deep learning of atomically resolved images

作     者：Rama K.Vasudevan Nouamane Laanait Erik M.Ferragut Kai Wang David B.Geohegan Kai Xiao Maxim Ziatdinov Stephen Jesse Ondrej Dyck Sergei V.Kalinin 

作者机构：Center for Nanophase Materials SciencesOak Ridge National LaboratoryOak RidgeTN 37831USA Institute for Functional Imaging of MaterialsOak Ridge National LaboratoryOak RidgeTN 37831USA Computational Sciences and Engineering DivisionOak Ridge National LaboratoryOak RidgeTN 37831USA Quantum Computing InstituteOak Ridge National LaboratoryOak RidgeTN 37831USA UnitedHealth GroupPO Box 1459Minneapolis 55440 MNUSA 

出 版 物：《npj Computational Materials》 (计算材料学（英文）)

年 卷 期：2018年第4卷第1期

页      面：396-404页

核心收录：

学科分类：0711[理学-系统科学] 07[理学] 08[工学] 081101[工学-控制理论与控制工程] 0811[工学-控制科学与工程] 071102[理学-系统分析与集成] 081103[工学-系统工程] 

基　　金：This research used resources of the Compute and Data Environment for Science(CADES)at the Oak Ridge National Laboratory which is supported by the Office of Science of the U.S.Department of Energy under Contract No.DE-AC05-00OR22725 

主　　题：beam evolution atomic 

摘      要：Understanding transformations under electron beam irradiation requires mapping the structural phases and their evolution in real *** date,this has mostly been a manual endeavor comprising difficult frame-by-frame analysis that is simultaneously tedious and prone to ***,we turn toward the use of deep convolutional neural networks(DCNN)to automatically determine the Bravais lattice symmetry present in atomically resolved images.A DCNN is trained to identify the Bravais lattice class given a 2D fast Fourier transform of the input ***-Carlo dropout is used for determining the prediction probability,and results are shown for both simulated and real atomically resolved images from scanning tunneling microscopy and scanning transmission electron microscopy.A reduced representation of the final layer output allows to visualize the separation of classes in the DCNN and agrees with physical *** then apply the trained network to electron beam-induced transformations in WS2,which allows tracking and determination of growth rate of *** highlight two key aspects of these results:(1)it shows that DCNNs can be trained to recognize diffraction patterns,which is markedly different from the typical“real imagecases and(2)it provides a method with inbuilt uncertainty quantification,allowing the real-time analysis of phases present in atomically resolved images.