AI-enhanced High Resolution Functional Imaging Reveals Trap States and Charge Carrier Recombination Pathways in Perovskite

Abstract: Understanding and controlling charge carrier recombination dynamics is essential for enhancing the performance of metal halide perovskite optoelectronic devices. In this study, we present a machine learning-assisted intensity-modulated two-photon photoluminescence microscopy (ML-IM2PM) method to quantitatively map recombination processes in MAPbBr3 perovskite microcrystalline films at micrometer-scale resolution. To improve model accuracy, we implemented a balanced classification sampling strategy during the machine learning optimization phase. The resulting regression chain model effectively predicts key physical parameters across a 576-pixel spatial map, including exciton generation rate (G), initial trap concentration (N_TR), and trap energy barrier (E_a). These extracted parameters were subsequently used to solve a system of coupled ordinary differential equations, enabling spatially resolved simulations of carrier populations and recombination dynamics under steady-state photoexcitation. The simulations reveal significant spatial heterogeneity in exciton, electron, hole, and trap populations, along with photoluminescence and nonradiative losses. Correlation analysis delineates three distinct recombination regimes: (i) a trap-filling regime dominated by nonradiative recombination, (ii) a transitional crossover regime, and (iii) a band-filling regime characterized by markedly enhanced radiative efficiency. A critical trap density threshold of approximately 10^17 cm^-3 marks the transition between these regimes. Overall, this work establishes ML-IM2PM as a robust framework for probing carrier dynamics and informing defect passivation strategies in perovskite materials.