Lining Wang_Poster
- 1. A Novel Automated Approach for Ronchigram Alignment of a Scanning Transmission Electron Microscope Lining Wang, Daniel Groom, Paulo Ferreira Materials Science and Engineering Program Cockrell School of Engineering The University of Texas at Austin, Texas, USA, 78712 Introduction • The Scanning Transmission Electron Microscope (STEM) is one of the most critical and unique scientific experimental tools today • STEMs can achieve a high spatial resolution of approximately 0.05nm • The alignment of STEMs will greatly affect the ultimate resolution • The most effective way to align a STEM is via the Electron Ronchigram Original electron ronchigram Canny Edge Detector Laplacian of Gaussians Edge Detector Trainable WEKA Segmentation • A Machine Learning type algorithm Creating classifier Training classifier Training classifier on a new electron ronchigram Training classifier based on the resultElectron Ronchigram • Electron Ronchigram or “shadow image” is generated when the intensity of a converged electron beam is formed at the microscope Fraunhofer diffraction plane of a specimen • Electron Ronchigram is very sensitive to Defocus, Two-fold Astigmatism and Coma and thus needs adjustment • A Ronchigram which shows a perfect circular contour without the absence of interference fringes and is maximized in terms of radius, represents the ideal alignment. Challenge: Finding the most accurate combination of defocus, astigmatism and coma to obtain the ideal alignment Edge Detection • Novel approach: image processing (Edge Detection) of the Electron Ronchigram • Most noise in Electron Ronchigram is reduced by using image filter before processed by edge detector • By edge detection, the contours of an Electron Ronchigram will be highlighted from its background. Subsequently, the shape of the contour will set up a basic criterion to determine the quality of the Electron Ronchigram. Objective • To develop a computer program which can automatically align the Ronchigram of a Scanning Transmission Electron Microscope Acknowledgements • This research is supported by the Office of the Vice President for Research and the Equal Opportunity in Engineering Program at The University of Texas at Austin Future Work • Create a database of around 100 electron ronchigrams to improve the classifier’s accuracy of segmentation • Explore new algorithms in image processing to improve the efficiency and accuracy of segmentation • Develop a software prototype to fully automate the alignment of STEMs Default Edge Detector without image filter