Designing New Metallic Catalysts by Transversing The ‘Hidden’ Compositional Terrain

To discover materials with enhanced catalytic behavior, we convert the complex elemental feature space into a connected network, representing the similarities and underlying physics of catalytic alloys. A challenge in the design is due to limited number of property measurements, thereby making typical data driven modeling inaccurate and uncetain. To address this challenge, we apply a random walk approach to capture the reaction behaviors and link the kinetic and chemical behaviors of the alloys. The results lead to a more comprehensive understanding of metallic catalysts design with the most promising chemistries identified across a variety of gas reactions. Through this work, the design space is reduced and the necessary experiments are defined.