Nanotechnology

Elucidating material reactions at the nanoscale using AI-based technology

Illustration of a workflow dedicated to detecting and segmenting defects in Cu metal during in-situ ion irradiation of TEM images. Credit: APL Machine Learning (2024). DOI: 10.1063/5.0186046

Corey Burns, a professor at the University of Virginia School of Engineering and Applied Sciences (UVA), is a materials science researcher who uses artificial intelligence to improve the characterization of materials. In an APL Machine Learning paper, he and his collaborators representing multiple universities and national laboratories detail an innovative new technique to study how to better determine the nanoscale effects of radiation on materials. Explained.

UVA is collaborating with Oak Ridge National Laboratory, which co-sponsors Burns’ research. This study features the largest labeled dataset of its kind and provides insight into how materials behave not only under irradiation conditions, but also potentially under other types of extreme conditions. It is hoped that this will advance the understanding of

Industries such as renewable energy, space exploration, and advanced electronics can benefit from improved materials that can withstand harsh environments.

For everyday consumers, this breakthrough could mean longer battery life, more reliable electronics, and safer medical devices.

“Nanoscale radiation-induced defects can have a significant impact on performance and structural lifetime,” Professor Burns said. Professor Burns joined the Department of Materials Science and Engineering as a rising researcher in 2022 before becoming an assistant professor in August. “By examining the fundamental interactions within materials, we can devise better strategies to extend their lifetime.”

small and fast changes

Transmission electron microscopy (TEM) is an imaging technique that uses a beam of electrons to pass through very thin samples (often called thin films because they are very flat).

TEM can reveal atomic-level, nanoscale details in a sample that cannot be seen with an optical microscope. This can include small changes caused by crystal structure or surface interactions, making TEM an essential tool in materials science.

Scientists can also use convolutional neural networks (CNNs) to study changes over time. Unlike traditional models, CNNs learn from large groups of data at once.

Burns’ team combined the two approaches and compared the CNN results with conventional TEM images to assess the model’s effectiveness in capturing nanoscale interactions.

“Our model reduces human error, accelerates analysis, and quantifies rapid responses,” Burns said. “However, accurate results depend on proper data preparation and fine-tuning of model settings.”

Defects in metals are different

Using advanced transmission electron microscopy time-series imaging techniques, the team compiled more than 1,000 images capturing more than 250,000 defects formed during ion bombardment. These defects include helium bubbles and planar defects known as “dislocation loops.”

Key findings from the study highlight the complexity of defect classification. The study revealed that defects in materials such as copper and gold behave differently than defects in palladium. This difference highlights the need for specialized analytical models to accurately study these materials under radiation.

One of the big challenges the researchers faced was “drift,” where images shift and can become inaccurate due to changes in the experimental environment. To address this, the team proposed the use of advanced techniques such as denoising autoencoders, which clean up images and improve data reliability.

Burns collaborated on the research with engineers and other experts from the University of California, Berkeley, Sandia National Laboratories, the Massachusetts Institute of Technology, Los Alamos National Laboratory, the University of Florida, the University of Michigan, Lawrence Berkeley National Laboratory, and the University of Tennessee. did. Knoxville.

More information: Kory Burns et al, Deep learning-enabled probes of irradiation-induced defects in time-series micrographs, APL Machine Learning (2024). DOI: 10.1063/5.0186046

Provided by University of Virginia

Citation: AI-enhanced technology illuminates materials reactions at the nanoscale (October 24, 2024) from https://phys.org/news/2024-10-ai-technique-illuminates-materials-reactions.html 10/2024 Retrieved on March 27th

This document is subject to copyright. No part may be reproduced without written permission, except in fair dealing for personal study or research purposes. Content is provided for informational purposes only.

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button