High-throughput computational materials screening of perovskite oxides for chemical looping air separation

Abstract High-throughput atomistic simulations were performed to identify novel and sustainable high-performance materials for chemical looping air separation. Nitrogen gas is an important commercial product and intermediate. Conventional methods for the production of pure nitrogen by separating air into its components remain energy-intensive and suffer from a high process complexity. Chemical looping air separation (CLAS) is a promising alternative for large-scale nitrogen gas production due to its high energy efficiency and easy integration with industrial processes. A major obstacle for the large-scale deployment of CLAS remains the lack of available high-performance functional materials needed to carry out the related thermochemical reactions. In this study, high-throughput atomistic simulations were used to identify perovskite oxides consisting of abundant elements with low criticality capable of a reversible absorption of large amounts of oxygen from air. A set of hierarchical filters was applied to pre-select plausible perovskite structures from existing first-principles data. Subsequently, a machine-learned interatomic force field was used to calculate their redox thermodynamics to identify materials with the best CLAS oxygen sorbent properties. The suggested computational workflow as well as the shortlisted materials can serve as starting points for future theoretical and experimental studies of perovskite oxide materials for CLAS as well as chemical looping processes in general. Discussion Thermochemical air separation with complex perovskites using critical, toxic, or expensive raw materials may not realistically compete with cryogenic air separation given material and synthesis costs. For a meaningful theoretical comparison of oxygen sorbent materials undergoing non-stoichiometric redox reactions during chemical looping air separation, oxygen vacancy formation energies need to be determined in small increments and over a wide range of non-stoichiometry. The unfavorable scaling of computational cost with system size in electronic structure calculations is prohibitive for the high-throughput simulation of reactions involving small oxygen vacancy concentration changes. Machine-learned interatomic force fields might be better suited for high-throughput calculations of non-stoichiometric perovskite oxide redox thermodynamics than electronic structure calculations. Graphical abstract