AI-powered microscope rivals human experts in analyzing 2D materials

Tuesday 28 October 2025 - 14:20

By: Dakir Madiha

AI-powered microscope rivals human experts in analyzing 2D materials

Researchers at Duke University have developed an artificial intelligence (AI)-powered microscopy system capable of analyzing two-dimensional (2D) materials with precision comparable to that of highly trained human experts. This breakthrough, named ATOMIC (Autonomous Technology for Optical Microscopy & Intelligent Characterization), represents a significant step forward in autonomous scientific research, achieving up to 99.4% accuracy in identifying material defects and layered structures.

Revolutionary integration of foundational AI models

The system’s development, published in ACS Nano on October 2, marks the first successful integration of publicly accessible foundational AI models, such as OpenAI’s ChatGPT and Meta’s Segment Anything Model (SAM), into autonomous laboratory instruments. Haozhe "Harry" Wang, the lead researcher from Duke’s Department of Electrical and Computer Engineering, explained that ATOMIC is designed to "understand" tasks rather than simply follow instructions.

"ATOMIC can autonomously evaluate a sample, make decisions, and produce results as effectively as a human expert," Wang noted. By connecting a standard optical microscope to these AI models, the system autonomously manages sample movement, image focusing, and lighting adjustments while simultaneously analyzing microscopic features.

Addressing critical research bottlenecks

This innovation tackles a longstanding bottleneck in materials science research: the characterization of 2D materials, which consist of crystals only a few atoms thick. These materials hold immense potential for next-generation semiconductors, sensors, and quantum devices, but their exceptional electrical properties can be undermined by manufacturing defects. Traditionally, mastering the analysis of such materials requires years of specialized training.

Jingyun "Jolene" Yang, a doctoral student and lead author of the study, highlighted that ATOMIC can detect grain boundaries at scales beyond human visibility. The system maintained exceptional accuracy even under suboptimal imaging conditions, such as overexposure, poor focus, or low lighting. In some cases, it identified imperfections that human observers could not detect.

Broader transformation in scientific research

ATOMIC exemplifies a growing trend in scientific research, where AI plays an increasingly central role in discovery processes. Recent studies in ACS Nano by teams from KAIST, Drexel University, and Northwestern University demonstrate how AI now facilitates everything from initial material discovery to optimization. Similarly, other advancements include the launch of autonomous lab platforms, such as AI-driven research factories by Lila Sciences, and systems capable of managing complete experimental workflows.

As OpenAI’s Sam Altman recently predicted, AI may achieve a groundbreaking scientific discovery within two years, underscoring its accelerating role in research. Wang’s team emphasized that while AI amplifies human expertise, researchers remain critical for interpreting results and determining their broader implications.