Machine-learning Surrogate Model Leads to Discovery of new Iridium-Oxide Catalysts

A newly developed Active Learning (AL)  algorithm (AL_IrO3_Anim (1).html) was used to search the bulk-polymorphic space of the Ir-Oxides. Using this approach we are able to find 196 IrO2 polymorphs and 75 synthesizable IrO3 polymorphs and report a previously unknown FeF3-type structure as the most stable,
termed α-IrO3. This project was funded under TRI and SUNCAT-FWP. Please follow this link to see the algorithm in motion: