Integrating Density Functional Theory, Machine Learning, and Device Simulation for Perovskite Photovoltaics: A Review

Abstract Metal halide and lead-free double perovskites are promising materials for next-generation photovoltaics. But it is a challenge to navigate their huge compositional space and to simultaneously control long-term stability and non-radiative recombination. This review provides a critical synthesis of the cross-scale computational paradigm that brings together Density Functional Theory (DFT), machine learning (ML) and macroscopic device simulations like SCAPS-1D. We discuss how to transform first-principles atomistic parameters such as band edge alignments, dielectric constants and effective masses into meaningful device parameters without physical oversimplification. This work does not see parameter transfer as a "plug-and-play" process, but rather explains the bridging mechanisms that must be developed between microscopic energetics and macroscopic transport equations. We critically assess functional dependencies, active learning efficiencies, device-level limitations, and uncertainties related to ion migration and transient dynamics. This review provides actionable roadmap towards high-throughput physics-informed computational discovery of robust perovskite solar cells.