In-situ Monitoring of Automated Defect Formation in STEM

One of the very base forms of dynamic behavioral studies is the in situ examination of defect structure formation within materials.In this effort, through utilizing unconventional beam control and automated data analysis, insights into the formation of unique defect structures as a function of electron (e -) beam irradiation are revealed and the fabrication of different types of defects can be achieved.To accomplish this, the sub- sized e -beam of a scanning transmission electron microscope (STEM), following unique beam pathways, is used as a manufacturing tool for atomic-scale defect formation.Previous works have demonstrated that it is possible to fabricate unique nanostructures within 2D materials using similar techniques by precisely placing the e -beam [1][2][3][4].In this study the formation of different defect structures within the transition metal dichalcogenide, MoS 2 , are studied, with the Mo 6 S 6 nanowire-type (MoS-NW) edge structures of particular interest [5-8].These metallic defects have been shown to alter functional properties within the semiconducting monolayer and, while the formation of these structures can be achieved stochastically with simple heating, more localized control of the formation of these structures can is needed to reliably form functionalized structures within 2D materials.In this work, through the combined use of machine learning techniques for rapid data analysis and an FPGA for high speed scan coil control, an in-line Python-enabled STEM feedback-controlled workflow is utilized to form pre-selected atomic defects within 2D materials [9][10][11][12][13][14][15].The experimental workflow, seen in Figure 1, demonstrates a single cycle of the automated feedbackcontrolled fabrication process to form various MoS 2 defect structures.Using the ensemble learning iterative training (ELIT) approach [11], robust and fast models were used to predict atomic coordinates which were then further classified [12][13].Desired milling sites and fabrication steps were then determined and sent to the FPGA scan controller to precisely monitor the movement of atomic species to ensure removal.This process is then repeated using an automated step-by-step method to fabricate MoS-NW defect structures.With this autonomous electron microscopy platform, pre-determined desired defect formation can be monitored as a function of e -beam irradiation in situ to a greater extent than previously possible [17].