Hyperspace mapping and iterative dimensionality reduction reveal radical cascade networks controlled by perovskite-type catalysts

Organic reactions are often simplified as linear equations, obscuring the complexity of competing pathways. This work advances a hyperspace framework that reconceptualizes reactions as dynamic networks embedded in multidimensional condition "hyperspaces." Using robot-assisted experimentation and AI-guided dimensionality reduction, we explore radical cascades involving 2-vinylpyridine and tert -butyl hydroperoxide and catalyzed by previously untested perovskite-like materials. Our pipeline reveals that distinct catalyst compositions unlock up to ten discrete products, including several synthetically valuable ones. The study highlights three key insights: the balance between broad hyperspace exploration and practical dimensionality reduction, the link between hyperspace composition and network structure, and the use of catalysts to steer network-wide reactivity.