A Water-Resistant Bi-Based Optoelectronic Synapse for Underwater Neuromorphic Vision

The exploration of marine environments demands intelligent sensing systems capable of stable operation underwater while performing real-time visual information processing. Conventional neuromorphic devices, though promising for brain-inspired computing, suffer from severe degradation in aqueous environments, limiting their application in underwater scenarios. Here, we develop a water-resistant optoelectronic synapse based on a Rubrene/FA₃Bi₂I₉/ZnO heterojunction, where the hydrophobic Rubrene layer effectively suppresses water infiltration, preserving the microstructure and optoelectronic performance of the lead-free perovskite even after prolonged water immersion. The device exhibits various synaptic plasticities and can achieve controllable transitions from short-term memory to long-term memory by adjusting the parameters of the light pulses. Due to its excellent multi-spectral response characteristics, we demonstrate the ability to filter redundant information and extract target features in complex light environments. Moreover, a 5 × 5 pixel array exhibits signal‑to‑noise separation, while a 256 × 144 resolution input frames integrated with a fully connected neural network achieves high‑precision classification of aquatic organism movement trajectories with 95.89% accuracy. This work provides a proof-of-concept technical path for the development of intelligent visual perception systems and neuromorphic computing hardware suitable for toward future complex underwater environments.