Global efforts to mitigate CO₂ emissions are challenged by the high energy costs of conventional capture methods and the complexity of converting CO₂ into valuable products. Integrated carbon capture-utilization (ICCU) approach offers a promising way to reduce the overall cost of the carbon capture and utilization (CCU) processes. The challenge lies at the design and development of bifunctional adsorbent-catalyst materials (BFMs) with high adsorption and catalytic performances for use in ICCU processes. The advent of machine learning (ML), combined with density functional theory (DFT) computations, offers a promising route to rapidly discover and optimize advanced materials for CO₂ capture and conversion. Our research uses ML to identify and optimize BFMs that simultaneously capture CO₂ and catalyze its conversion to chemicals and fuels under mild conditions, advancing the ICCU technologies. We curated a comprehensive dataset from experimental literature, including material composition, structural descriptors (e.g., pore limiting diameter, surface area, void fraction), non-structural descriptors (e.g., electronic properties, synthesis conditions), CO₂ uptake, catalytic turnover frequencies, conversion rates, and selectivity metrics. ML models, including random forests and extra gradient-boosted trees, were trained to predict and rank BFMs for optimal CO₂ uptake capacity, conversion, and product selectivity. Preliminary results demonstrate that integrating structural and non-structural descriptors enhances prediction accuracy for both CO₂ capture and catalytic performance. This approach not only accelerates material discovery by reducing reliance on time-consuming trials but also improves energy efficiency in ICCU processes, offering a scalable strategy to reduce emissions, and highlights FECM RDD&D priorities in driving materials innovation and process optimization.
