| About this Abstract |
| Meeting |
MS&T23: Materials Science & Technology
|
| Symposium
|
Materials Informatics for Images and Multi-dimensional Datasets
|
| Presentation Title |
Informing Autonomous Processing via STEM-EELS Using Variational Autoencoders for Classification and Decision |
| Author(s) |
Jonathan Hollenbach, Mitra L Taheri |
| On-Site Speaker (Planned) |
Jonathan Hollenbach |
| Abstract Scope |
In-situ Electron Energy Loss Spectroscopy (EELS) is an instrumental technique in understanding how processing materials changes local structure and composition, but acting on the transient changes occurring at high temporal resolutions requires new artificial intelligence frameworks for characterization and analysis to match the data rate of the detector. We demonstrate a machine learning framework for rapid assessment and characterization of operando EELS data, informing decision frameworks and updating process changes. Variational Autoencoders (VAEs) have been proven ideal for encoding the complex relations of chemical changes to latent representations. These tools are paired with automation tools controlling the instrument and specialty in-situ stages to enact precise driving forces on the sample. We explore how latent encodings are used as basis for classification against theoretical structures and as inputs for decision frameworks to control autonomous experiments within the Transmission Electron Microscope (TEM). |