About this Abstract |
Meeting |
MS&T22: Materials Science & Technology
|
Symposium
|
2022 Undergraduate Student Poster Contest
|
Presentation Title |
Simulation-Trained Machine Learning for Segmenting Néel-Type Skyrmions |
Author(s) |
Alec Tyler Bender, Arthur McCray, Amanda Petford-Long, Charudatta Phatak |
On-Site Speaker (Planned) |
Alec Tyler Bender |
Abstract Scope |
Understanding the behavior of magnetic skyrmions is fundamental to the next generation of data storage devices. Skyrmion lattices can be studied using in situ Lorentz Transmission Electron Microscopy (LTEM), but the images are difficult to interpret. In this work, we have developed a machine learning technique to perform instant segmentation on LTEM images of skyrmion lattices and thus extract quantitative information. We used micromagnetics software to simulate 10,000 skyrmion lattices, from which we created ground truths of skyrmion sizes and positions and further simulated corresponding LTEM images. We then trained a convolutional neural network (CNN) on these simulated LTEM images and ground truths using supervised learning. Our results demonstrate that the CNN can accurately identify skyrmion locations and extent in both simulated and experimental data, providing a technique for quantitative analysis of skyrmion lattices going forward. |