ATOMIC integrates basic microscope controls through ChatGPT and utilizes SAM to identify discrete regions in material samples. To address challenges with overlapping layers, the team added a topological correction algorithm, enabling the system to separate single-layer regions from multi-layer stacks. The microscope then autonomously sorted regions based on their optical traits.
This platform matched or surpassed the accuracy of human analysis for 2D materials, detecting layer regions and subtle defects with up to 99.4 percent accuracy. ATOMIC maintained reliability with imperfect images, finding imperfections sometimes invisible to human observers.
"ATOMIC can assess a sample, make decisions on its own and produce results as well as a human expert," said Haozhe "Harry" Wang, the lab leader.
Jingyun "Jolene" Yang, lead author on the corresponding scientific paper, stated, "The model could detect grain boundaries at scales that humans can't easily see. When we zoom in, ATOMIC can see on a pixel-by-pixel level, making it a great tool for our lab."
By pinpointing microscopic defects, the platform assists in identifying pristine regions of 2D materials for further experiments, including applications in soft robotics and next-generation electronics.
Notably, ATOMIC required no specialized training data. Its "zero-shot" approach, based on broad foundation models, allowed rapid adaptation without custom image sets.
Research Report:Zero-Shot Autonomous Microscopy for Scalable and Intelligent Characterization of 2D Materials
Related Links
Pratt School of Engineering, Duke University
Space Technology News - Applications and Research
| Subscribe Free To Our Daily Newsletters |
| Subscribe Free To Our Daily Newsletters |
