Materials structure–property factorization for identification of synergistic phase interactions in complex solar fuels photoanodes

Abstract Properties can be tailored by tuning composition in high-order spaces. For spaces with complex phase behavior, modeling the properties as a function of and distribution remains formidable challenge. We present materials structure–property factorization (MSPF) an approach to automate such data identify synergistic interactions. MSPF is interpretable machine learning algorithm that couples mapping via Deep Reasoning Networks (DRNets) matrix factorization-based representative each dataset. demonstrated for Bi–Cu–V oxide photoanodes solar fuel generation, which contains 25 different combinations correspondingly exhibits composition-structure-photoactivity relationships. Comparing measured photoactivity learned model non-interacting phases, interactions are identified guide further optimization understanding. identifies BiVO 4 -like both Cu 2 V O 7 CuV 6 creating avenues understanding photoelectrocatalysts.