Collaborative Research: Co-orchestrated autonomous experiments for discovery and optimization of novel ferroelectric and electrochemical materials

About This Grant

Non-Technical Summary As laboratory tools become increasingly automated and machine learning (ML) continues to transform scientific research, a new challenge has emerged: how can ML agents control and coordinate multiple instruments at different labs, share information between them, and make meaningful decisions to accelerate discovery? This project tackles that challenge by developing strategies for building smart experimental systems that allow different tools - such as microscopes, structural characterization, and synthesis instruments - to work together autonomously, and develop metrics that allow evaluation of the return on the investment. These systems are designed to not just automate tasks, but also to “learn” which experiments to run next based on previous results, optimizing both speed and insight. The research is focused on discovering new materials for energy storage and information technologies, such as batteries and next-generation electronics, where even small improvements in materials can have large technological and economic impacts. In addition to scientific breakthroughs, the project shares tools and training with students and researchers from a broad range of institutions, helping to build an innovation-driven workforce for deep tech industries and manufacturing. In doing so, this work supports NSF’s mission to promote the progress of science, support national prosperity and security, and prepare a skilled workforce. Technical Summary This research develops an autonomous experimental framework for materials discovery based on multi-instrument coordination, active learning, and reward-driven optimization. The central objective is to create machine learning agents that can operate multiple experimental tools - such as scanning probe microscopes, structural probes, and synthesis platforms - in parallel, sharing information and prioritizing experiments in real time. These systems are applied to the exploration of combinatorial materials libraries, particularly targeting ferroelectric and electrochemical functionalities relevant to energy storage and electronics. The proposed methods include the use of structured and deep Gaussian Processes, causal learning, and symbolic regression, all within a framework of reward function design - where experimental strategies are guided not only by accuracy but by their expected contribution to downstream functionality. In addition to combinatorial libraries, the same strategies can be applied to multiple identical synthesis tools exploring the same parameter space, enabling high-throughput autonomous experimentation across facilities. Emphasis is placed on multi-objective and multi-fidelity optimization, where low-cost proxy measurements are combined with high-resolution tools, and on building decision-making logic that can generalize concurrent decisions across dissimilar experimental platforms. The project contributes to reproducible AI-driven experimentation and provides shared tools and training to advance materials research infrastructure, aligning with NSF goals in innovation and national competitiveness. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.