Proceedings of Asia-Pacific International Conference on Perovskite, Organic Photovoltaics and Optoelectronics (IPEROP27)
Publication date: 22nd October 2026
Understanding how grain-boundary structure and thermal disorder influence electronic properties is important for interpreting the behaviour of polycrystalline metal-halide perovskites in photovoltaic and light-emitting devices. However, first-principles modelling of both structural diversity and finite-temperature motion remains computationally demanding. Here, we combine machine-learning interatomic potentials with efficient electronic-structure methods to investigate these effects in CsPbBr3 grain boundaries.
A MACE foundation potential [1] is sequentially refined using bulk and grain-boundary reference data. Refinement improves predictions of grain-boundary forces and relative energies, with complementary validation against bulk reference data. A screened library spanning different boundary orientations and local structural environments, together with configurations sampled by molecular dynamics, provides a basis for evaluating electronic models. Density-functional tight-binding [2], including spin–orbit coupling (DFTB+SOC) reproduces trends in static band gaps across the library, with a root-mean-square deviation of 0.40 eV from r²SCAN DFT reference calculations without SOC [3]. Along a grain-boundary MD trajectory, its band-edge fluctuations correlate with those obtained from reference DFT calculations, although systematic differences remain in gap values and fluctuation amplitudes. Using deep learning tight-binding [4] and equivariant Hamiltonian learning [5], we explore electronic-model transferability across bulk, interfacial, and thermally disordered environments.
These results provide a basis for connecting grain-boundary structural diversity and thermal motion with the local electronic properties of halide perovskites. By combining atomistic sampling with separately benchmarked electronic models, the approach supports larger-scale investigations of structure–property relationships relevant to perovskite optoelectronics.
This work was supported by UK Research and Innovation (UKRI) under project number UKRI1487.
