Robust automatic identification of lattice defects for in-situ TEM experiments
Submission Type
Synchronous Research Talk
Area of Study or Work
Physics
Zoom Meeting
Faculty Advisor
Lin Zhou
Expected Graduation Date
2022
Start Date
4-10-2021 11:00 AM
End Date
4-10-2021 11:20 AM
Abstract
Transmission electron microscopy (TEM) uses an electron beam to transmit through a specimen and form an image. Modern TEM can easily achieve atomic-resolution, making it a powerful instrument to study the crystal structure and defects, which is closely related to the materials’ properties. Moreover, the successful development of in-situ capabilities, by providing external stimuli (heating, biasing, etc) to the TEM sample, opens the opportunity for real-time observation of the formation and movement of defects under different conditions. These in-situ experiments have provided many new insights into conventional theories. Meanwhile, the massive amount of data generated by the in-situ TEM experiments, usually thousands of images, is always a challenge for analysis. Therefore, an automatic method, constructed on the fundamental knowledge of image processing and materials science, is crucial to identify and track the defects in these in-situ TEM experiments.
In partnership with the Department of Energy’s Ames Laboratory, we will investigate the materials’ behavior away from equilibrium by extracting structural data from in-situ TEM images. Based on our previous work on image analysis using Voronoi tesselleation, this project will develop automation filters, based on convolution and correlation, to identify the structural features from images with short exposure time. The results will provide a stable and robust way to identify lattice defects from high-noise images.
Robust automatic identification of lattice defects for in-situ TEM experiments
Transmission electron microscopy (TEM) uses an electron beam to transmit through a specimen and form an image. Modern TEM can easily achieve atomic-resolution, making it a powerful instrument to study the crystal structure and defects, which is closely related to the materials’ properties. Moreover, the successful development of in-situ capabilities, by providing external stimuli (heating, biasing, etc) to the TEM sample, opens the opportunity for real-time observation of the formation and movement of defects under different conditions. These in-situ experiments have provided many new insights into conventional theories. Meanwhile, the massive amount of data generated by the in-situ TEM experiments, usually thousands of images, is always a challenge for analysis. Therefore, an automatic method, constructed on the fundamental knowledge of image processing and materials science, is crucial to identify and track the defects in these in-situ TEM experiments.
In partnership with the Department of Energy’s Ames Laboratory, we will investigate the materials’ behavior away from equilibrium by extracting structural data from in-situ TEM images. Based on our previous work on image analysis using Voronoi tesselleation, this project will develop automation filters, based on convolution and correlation, to identify the structural features from images with short exposure time. The results will provide a stable and robust way to identify lattice defects from high-noise images.