A hybrid machine learning approach for high-throughput screening of thermodynamically stable double perovskites with optimal band gap

IZ-DPscreen is a hybrid machine learning framework developed for the high-throughput screening of double perovskites. It employs a stacking regressor approach to accurately predict key material properties, including band gap, formation energy, and thermodynamic stability. By integrating multiple learning algorithms, the model enhances prediction reliability and robustness. The performance of IZ-DPscreen is validated through benchmarking against both theoretical (DFT) and available experimental data, demonstrating its capability for reliable materials screening. This approach enables the rapid identification of thermodynamically stable double perovskites with optimal band gaps, thereby accelerating the discovery of promising candidates for optoelectronic and energy applications.