Artificial Intelligence for Instability in Inorganic Perovskites: From Mechanism Discovery to Engineering Strategies

Abstract

Three-dimensional all-inorganic halide perovskites, represented by CsPbX$_3$ (X = Cl, Br, I), have attracted broad interest in photovoltaics, photodetectors, and light-emitting devices because of their outstanding optoelectronic properties. Their practical deployment, however, remains limited by instability under thermal, chemical, optical, and electrical stress. Conventional studies have established important experimental and theoretical foundations, but they still struggle with multimodal data, coupled degradation pathways, protocol dependence, sparse statistics, and uncertainty quantification. Artificial intelligence (AI) offers a practical route to address these limitations. This review summarizes recent progress in AI-assisted studies of instability in 3D CsPbX$_3$ and organizes the discussion around four linked tasks, including stability discrimination and diagnosis, microscopic mechanism analysis, consequence and reliability modeling, and engineering stability enhancement. We further discuss the main limitations of current methods, especially in data quality, protocol consistency, benchmark design, interpretability, and transferability across domains. Finally, we outline future directions for the field, including standardized data infrastructures, interpretable cross-scale models, and tighter integration of AI with automated experiments and physics-based modeling. The aim of this review is to provide a coherent and practically useful framework for researchers seeking to use AI to understand, predict, and mitigate instability in inorganic perovskites.