Bioinspired Tunable‐Tunneling Heterostructure for Dynamic Grayscale Perception and Spike Encoding

作者：Xuhao Fan, Shian Mi, Xiaohan Wei, Fengyi Zhu, Sheng Ni, Chongyu Li, 茹文君, Yuhang Ma, Changyi Pan, Haibiao Guan, Weiwei Tang, Songyuan Ding, Haibo Shu, Yi Zhou, Guanhai Li, Changlong Liu, Xiaoshuang Chen · 发表于：Advanced Functional Materials · 年份：2026 · DOI：10.1002/adfm.75868 · 被引用次数：1 · 研究领域：Neural Networks and Reservoir Computing、Advanced Memory and Neural Computing、Ferroelectric and Negative Capacitance Devices

ABSTRACT Bionic vision systems employ spike‐based neural encoding—inspired by biological vision—to achieve high perceptual efficiency and adaptability. However, most current implementations commonly rely on carrier relaxation, charge trapping/detrapping, or ion migration processes along with peripheral circuits to emulate neural dynamics, which limit response speed, increase energy consumption, and hinder efficient device‐level spike encoding. Here, we demonstrate a bioinspired tunneling photodetector using an SnSe 2 /MLG/WS 2 junction, in which an electrically triggered switch from direct tunneling (DT) to Fowler–Nordheim tunneling (FNT) enables optical‐power‐dependent grayscale resolution and spiking‐encoded output within a single structure. By leveraging tunable tunneling transports for efficient photocarrier collection, the device achieves a responsivity of ≈233 A/W and a response time of ≈97 µs at 638 nm. We further demonstrate the device's capability for light‐intensity encoding and dynamic perception through grayscale image transmission. Specifically, spike signals encoded via the nonlinear photocurrent response allow a trained spiking neural network (SNN) to achieve 96.5% accuracy in grayscale letter recognition. Our results demonstrate that the tunable tunneling‐encoding strategy provides a practical pathway toward compact bionic visual sensors capable of grayscale imaging in neuromorphic vision front‐ends.