Our research group is at the forefront of integrating machine learning (ML) and artificial intelligence (AI) techniques into Scanning Transmission Electron Microscopy (STEM) workflows. We are developing cutting-edge automated STEM methodologies that leverage ML algorithms and autonomous agents for real-time decision-making and adaptive experimentation. Our approaches involve implementing active learning strategies, such as Bayesian optimization and reinforcement learning, to intelligently navigate vast parameter spaces and optimize data acquisition processes. Additionally, we are exploring instrument-level hyperlanguages to represent human operations, facilitating the transition from human-driven to automated experiments.
A key area of focus is the development of automated experiment flows that leverage cloud computing resources and distributed data processing capabilities. This enables efficient handling of high data rates generated by STEM, real-time analytics, and decision-making by ML agents during experiments. Furthermore, we are investigating the integration of automated STEM with other characterization techniques, such as spectroscopy and microscopy modalities, to enable multimodal and correlative analyses for a comprehensive understanding of material systems. Through our innovative efforts, we are driving transformative advances in materials science, nanotechnology, and related fields, accelerating scientific discovery and enabling more efficient and data-driven exploration of material systems.
Click here to watch the other demonstrations of the automated STEM capabilities