Composition-engineered intrinsic stability in 2D perovskite optoelectronic synapses for neuromorphic vision

Abstract Operational instability remains a key obstacle for halide perovskites in neuromorphic optoelectronics, where sustained and reproducible photoresponses are required for in-sensor computing. Here we report vapor-phase growth of composition-tunable two-dimensional (2D) inorganic perovskites, CsxRb1−xPbBr3 (0 ≤ x ≤ 1), and show that A-site cation alloying provides an effective route to simultaneously improve optical performance and structural robustness. By mapping the full Cs-Rb compositional space, we identify a narrow intermediate regime (x ≈ 0.8) where lattice strain is minimized and defect formation is suppressed. Perovskites in this regime exhibit substantially enhanced resistance to moisture, thermal stress, and ion migration compared with either end member. These stability gains translate directly to device operation, enabling broadband photodetectors with fast response, high reproducibility, and long-term durability. Using the optimized Cs0.81Rb0.19PbBr3 composition, we further demonstrate optoelectronic synaptic devices that reproduce key synaptic functions, including paired-pulse facilitation (PPF) and light-programmable synaptic weights. Integrated into a spiking neural network (SNN), the devices enable real-time fire detection with 85% accuracy while maintaining stable performance over seven days. This work establishes composition-engineered 2D inorganic perovskites as a robust platform for neuromorphic vision and energy-efficient in-sensor computing.