DREAM—Data-Driven Reinvigorated Advanced Membrane Discovery Platform

Recent years have witnessed the transformative impact of Artificial Intelligence and Machine Learning (AI/ML) across academia and industry, leading to significant advancements in AI/ML algorithms and software frameworks. Block copolymers (BCPs), critical in materials science, have been widely applied in high-resolution etch masks, microelectronics, optics, and solar cells. However, the complex and time-consuming process of BCP synthesis restricts their data availability, which has limited the integration of AI/ML in their development. Supramolecular block copolymers (SBPs), which are BCPs assembled with non-covalent end-functional group pairs (e.g., hydrogen bonding), offer a promising alternative. The synthesis of SBPs with varying architectures, chain lengths, and chain length ratios can be achieved through the physical blending of two homopolymers. This method shows great potential for rapid and high-throughput synthesis. In this work, we developed a Data-driven Reinvigorated Advanced Membrane Discovery Platform (DREAM) based on SBPs. Using a modular synthesis approach, we rapidly synthesized a library of DAT- and Thy-functionalized homopolymers, including POEGMA-DAT, PEG-DAT, PMMA-DAT, PS-Thy, and PMMA-Thy, with different molecular weights. This enabled the fabrication of SBPs with diverse architectures and chain length ratios. We then utilized a liquid-handling robot for automatic, high-throughput synthesis of 273 different SBP copolymers, which were next spin-cast into thin films on Si wafers. Automated, high-throughput atomic force microscopy (AFM) was used to characterize the morphologies of these SBP thin films. The resulting data were analyzed using custom Python scripts to automate morphology analysis. Finally, machine learning algorithms were employed to train models based on the results, allowing for the prediction of morphologies and providing guidelines for the synthesis of new SBPs. This work establishes an automatic, high-throughput materials synthesis platform that accelerates material discovery and development through machine learning.