Artificial intelligence-driven platform for catalytic reactor discovery and optimization applied to carbon dioxide valorization

Digital technologies, including artificial intelligence and additive manufacturing advancements, have revolutionized all areas of chemistry and chemical engineering. In the field of reactor engineering, the development of novel geometries has enabled improved performance. However, until now the design has been largely relying on human design. This study introduces Reac-Discovery, a digital platform that integrates the generation, fabrication, and optimization of catalytic reactors based on periodic open-cell structures (POCs). It integrates the parametric design and analysis of advanced structures from mathematic models (Reac-Gen), the high-resolution 3D printing and functionalization of catalytic reactors (Reac-Fab) with an algorithm that validates the printability of reactor designs and a self-driving laboratory platform (Reac-Eval) capable of parallel multi-reactor evaluations featuring real-time NMR monitoring and machine learning (ML) simultaneous optimization of process parameters and topologic descriptors. Reac-Discovery can iteratively, dynamically, and rapidly refine reactor designs and reaction conditions, enabling the discovery of optimal geometries for industrially relevant processes. The heterogeneously catalyzed CO₂ cycloaddition at low pressure was chosen as a benchmark reaction. Reac-Discovery offered an improvement of ca. one order of magnitude over the state of the art. By seamlessly integrating digital design and machine learning, Reac-Discovery establishes a transformative framework for advancing complex chemical systems.