Publication date: 11th March 2026
Mixed halide perovskites are highly versatile semiconductors, with applications in optoelectronic devices as diverse as photovoltaics, light-emitting diodes, photodetectors, and thermoelectrics. While halide perovskites are promising as high performing photovoltaics, they suffer from low stability, degrading rapidly both through transitions into a non-photoactive phase and halide segregation. The exact mechanisms of these photoinduced degradation processes are not fully understood. Addressing these unresolved degradation mechanisms requires a detailed understanding of the thermodynamic behavior of mixed perovskite systems, particularly the coupling between structural phase transitions and compositional stability. Investigating the thermodynamics, e.g., segregation and ordering, requires large simulation cells, which are not achievable with ab initio methods such as DFT. Here, we overcome this by training a machine-learned interatomic potential on DFT data, enabling molecular dynamics simulations of millions of atoms for nanosecond timescales. We use a combined Monte Carlo and molecular dynamics approach to investigate the phase diagram of CsPbX3 (X=Cl, Br, I) mixed perovskites. Our results demonstrate that all three binary perovskites have a miscibility gap, the extent of which is strongly correlated with halide ion size mismatch. Beyond the miscibility gap, all binary mixtures exhibit layered ordering tendencies. In CsPb(BrxI1-x)3 specifically, the ordering shifts the structural phase transitions, increasing the stability of the intermediate tetragonal phase and reproducing the experimental behavior of both the orthorhombic-to-tetragonal and tetragonal-to-cubic transitions. This demonstrates that halide ordering has significant influence on structural phase stability in mixed halide perovskites, suggesting that halide ordering may play a key role in photoinduced degradation of perovskite photovoltaics.
