School of Science Department of Physics 61 Department of Physics AI for Low Energy Electron Microscopy Supervisor: Michael Scott ALTMAN / PHYS Student: LEUNG Yu Fung / PHYS-IRE Course: UROP 1000, Summer This project explores the application of artificial intelligence (AI) to enhance the interpretation of Low Energy Electron Microscopy (LEEM) images of crystalline surfaces. LEEM imaging is critical for studying surface structures at atomic scales, but image analysis is complicated by defocus, aberrations, and noise. In the following, we employ Contrast Transfer Function (CTF) simulations implemented in MATLAB to generate synthetic LEEM datasets with controlled aberrations. Different object functions are tested to analyze their distinct imaging signatures. We will also discuss some shortcomings of the CTF simulation and how we can trade it off by changing parameter related to the aperture size. AI for Low Energy Electron Microscopy Supervisor: Michael Scott ALTMAN / PHYS Student: SARKAR Manjori / AE Course: UROP 1100, Summer This project investigates the use of artificial intelligence (AI) for the accurate analysis of crystalline sample surface images obtained via Low Energy Electron Microscopy (LEEM). The AI model is developed using carefully simulated LEEM images produced through Contrast Transfer Function (CTF) theory using various deep learning techniques, specifically through a Convolutional Neural Network (CNN) structured on the UNet design. In order to increase the accuracy of the model, an important advancement includes employing a multi-channel input method, where images taken at various defocus levels are tied to a single object function, or ground truth. This approach leverages the accurately labeled ground truth present in simulated data, allowing the AI to grasp the intricate relationships that control LEEM image phase contrast. This effort seeks to greatly improve and automate the evaluation of microscopy data which is beyond the resolution of the most advanced low energy electron microscopes currently available, effectively breaching a frontier in the field if our AI model is consistently accurate, speeding up scientific advancements in material characterization. The Structures and Motions of Novel Shaped Granular Particles Supervisor: HAN Yilong / PHYS Student: MA Jiachen / PHYS Course: UROP 1100, Summer Granular materials are macroscopic, discrete solid particles that exhibit unique properties due to their particulate nature. In this experimental study, we investigate the disintegration process of giant clusters composed of C-shaped particles when they are perturbed by hand-to-hand transfers. We observe a dramatic disintegration when the number of disturbances approaches a critical threshold. We further find that this critical threshold shifts slightly when the opening angle of the C-shaped particles is varied. In the future, we plan to conduct discrete element method (DEM) simulations to compare with these experimental results, aiming to derive more substantiated conclusions. This combined approach will provide deeper insights into the disintegration dynamics of granular assemblies under disturbances.
RkJQdWJsaXNoZXIy NDk5Njg=