Y. Liu, M. Checa, S. Jesse, R.K. Vasudevan
Oak Ridge National Laboratory,
United States
Keywords: scanning probe microscopy, AI, autonomous experiments
Summary:
In this talk, I will present our recent development of AEcroscopy (Automated Experiments in Microscopy), a cross-platform Python API that enables automated, high-throughput, and intelligent microscopy experiments. AEcroscopy empowers users to design and execute complex scanning probe microscopy (SPM) workflows with enhanced efficiency, reproducibility, and scalability. By integrating artificial intelligence (AI) models into SPM workflows, AEcroscopy advances beyond traditional automation toward AI-driven autonomous experimentation. In particular, large language models (LLMs) can translate expert scientific intent into executable Python workflows, assist in data interpretation, and accelerate experimental design. Building upon this foundation, we have further developed a gated active learning framework that incorporates prior scientific knowledge and human insight into active learning–based decision-making loops. This transforms conventional autonomous discovery pipelines into adaptive decision-making processes effectively integrated human guidance. Together, AEcroscopy is reshaping how microscopy experiments are performed, enabling researchers to explore materials landscapes more efficiently, uncover complex phenomena more rapidly, and accelerate the pace of scientific discovery. Acknowledgments: This research was supported by the Center for Nanophase Materials Sciences (CNMS), which is a US Department of Energy, Office of Science User Facility at Oak Ridge National Laboratory.