High-throughput screening of SrxA1−xFeyB1−yO3 perovskites for low-temperature chemical looping air separation using graph neural networks

• • The thermodynamic threshold for CLAS oxygen carriers is Eovf < 2.3 eV, ensuring feasible oxygen release at low temperatures (400–650 °C). • • The ALIGNN graph neural network predicts Eovf with a mean absolute error (MAE) of 0.26 eV, enabling reliable high-throughput screening. • • Screening of 3,649 SrxA1−xFeyB1−yO3 compositions identified that Co doping at the B-site and Ba/Ca doping at the A-site effectively lower Eovf, guiding rational material design. • • The model rediscovered known low-temperature CLAS oxygen carriers, validating the screening methodology and accelerating material discovery.