Bio‐inspired inorganic/organic semiconductor heterostructure synapses with UV –visible– NIR wavelength‐dependent responses for image recognition

Abstract Neuromorphic vision systems (NVSs) that integrate perception and computing functions are important for edge intelligence. However, the existing implementations suffer from narrow spectral selectivity, the lack of wavelength‐programmable functionality, and the architectural separation between perception and processing. Herein, inspired by the broad spectral perception of the American bullfrog's retina‐pigment epithelium, we present a SnSe 2 /organic semiconductor heterostructure photonic synaptic transistor. The photonic synaptic transistor exhibits wavelength‐dependent duality, operating as a sensitive photodetector under ultraviolet (UV) illumination while exhibiting tunable synaptic plasticity under visible and near‐infrared (NIR) illumination. The wavelength‐dependent reconfigurability is enabled by engineered type‐II band alignment, material absorption characteristics, and interfacial physical mechanisms. Furthermore, the device achieves a low energy consumption of 17 fJ per synaptic event at a low operating voltage of 0.001 V. By exploiting the wavelength‐dependent synaptic characteristics of the device, reservoir computing (RC) is further employed to establish an integrated sensing‐computing NVS. Attributed to the preprocessing capability of the device, the NVS achieves a facial recognition accuracy of ~92.7%, higher than that obtained without preprocessing. Eliminating external optical filtering and preprocessing circuits, this study demonstrates the potential of efficient wavelength‐programmable neuromorphic vision for autonomous applications. image