Adaptive Experiment Planning for Inverse Design and Understanding: Synergistic Interactions as Key to Optimized Multi-Promoter Formulations

Extending throughput capabilities and advanced sampling approaches are strongly accelerating catalyst discovery, increasingly performed within automated self-driving laboratories. The larger the tractable design spaces become though, the more questionable is the lasting value of individual optimal catalysts that are identified in black-box searches. Here, we demonstrate a sparse sampling approach that combines search efficiency with chemically interpretable insight into the topology of the design space. Applied to the nonoxidative propane dehydrogenation reaction, it readily finds pareto-optimal multipromoter formulations that exceed the present industry reference in both yield toward the desired commodity product propylene and catalyst longevity. At the same time, it explains this improved performance in terms of individual promoter effects and synergistic promoter interactions. The latter interactions are missed in prevalent empirical single-promoter studies and are shown here as a key element toward further performance gains expected upon insight-motivated future modifications of the design space.