Tailoring the A-site composition enhances the stability of mixed-cation perovskites such as Cs1–xFAxPbI3, positioning them for next-generation photovoltaics. However, their intrinsic tendency toward cation phase segregation induces structural heterogeneity that remains a critical bottleneck limiting performance. In this work, we employ machine learning potential (MLP)-based molecular dynamics simulations to investigate the structural heterogeneity induced by different A-site cation configurations and track the structural evolution of Cs1–xFAxPbI3. By validating against electron microscopy measurements, we introduce a new structural descriptor, heterogeneity coefficient, which captures medium-range structural inhomogeneity by quantifying spatial cation distribution arising from A-site connectivity. This descriptor not only accurately predicts both the organic cation rotation and octahedral distortion, two key factors governing the electronic properties of Cs0.5FA0.5PbI3, but also quantitatively captures the degree of A-site phase segregation and identifies charge localization arising from segregation driven by A-site cation migration as a key contributor to photovoltaic performance degradation.
Du et al. (Mon,) studied this question.