Autonomous Thin Film Coating Enabled by AI/ML in Combination with Multi-Modal in Line/In Situ Diagnostics

Autonomous coating platforms equipped with inline sensing have the potential to become companion tools to thin film researchers that accelerate time-to-solution by 10X to 100X with the appropriate implementation of multimodal sensors, machine learning and artificial intelligence (ML/AI). Moreover, their implementation at the synchrotron will allow human-machine-AI teaming to address complex thin film problems in real time during the synchrotron beam time with the help of active learning, exploration and exploitation under uncertainty. In this presentation, we will present the RoboCoat AI, an autonomous spin-coater equipped with multi liquid dispensing, substrate cleaning and annealing, as well as multi-modal in-line optical sensors, real-time analytics and AI/ML. RoboCoat AI is shown to be compatible with synchrotron operation and has been successfully integrated at NSLS II's CMS beamline to utilize inline/<i>in situ</i> grazing incidence wide angle x-ray scattering (GIWAXS) measurements to incorporate into multi-objective optimization of thin film coatings. We will present an example of hybrid perovskite antisolvent processing to demonstrate how we have successfully addressed several key challenges, including (1) AI-guided mapping of perovskite film fabrication across a multi-dimensional parameter space navigated by AI, (2) autonomous development of optimal perovskite coating recipe using AI decision algorithms, (3) integration of multimodal inline diagnostics with synchrotron-based characterization to combine optimizations of coating quality and property with its microstructure, and (4) leveraging in situ metadata to develop interpretable coating knowledge.