ML-Assisted Fully Autonomous Robotics for High-Throughput Performance Optimization of Organic Mixed Ionic–Electronic Conductors

Machine learning (ML) has revolutionized the way materials research is performed nowadays by intelligently navigating to vast experimental design space and identifying promising material candidates with desired properties, thereby reducing the time and cost associated with traditional trial-and-error approaches. Recently, a novel category of materials: organic mixed ionic–electronic conductors (OMIECs) attract considerable attention for simultaneously facilitating highly efficient electron/hole and ion transport, and thereby are widely adopted in signal processing, sensing, and energy storage devices. Among these, the organic electrochemical transistor (OECT) is of particular interest due to its high transconductance (<i>G</i>_m), which is essential for enabling biosensing with high sensitivity. To gain the high <i>G</i>_m, high [<i>μC</i>*] needs to be achieved, which is determined by the intrinsic morphology and energy profile of OMIECs that can be tunned by vast processing conditions and molecular design strategies, resulting in a large searching space. Therefore, ML can play an important role in not only accelerating the OMIEC exploration but also uncovering the underlying knowledge. However, to realize this, it requires built-in fully autonomous robotics for acquiring high-fidelity source data, and a well-designed ML algorithm and workflow for enhancing exploration efficiency. Herein, we build the fully autonomous Polybot system to carry out the closed-loop workflow for screening out the best processing conditions of OMIECs. The entire system is monitored in real-time by a built-in live streaming platform providing imaging and data analytics tools. During the ML-guided experimentation, we identify the optimal processing conditions of our model OMIEC with improved [<i>μC</i>*] by over 2 folds. Interpretability techniques were also employed during the process to facilitate a deeper understanding of the materials’ mixed conducting behaviors and unveil fundamental relationships between [<i>μC</i>*] and processing conditions.