Integrating Photovoltaic Detection and Synapse in MoS ₂ /Ta ₂ NiS ₅ /MoSe Jingjing Deng, Sina Li, Wenjie Chen, Qunrui Deng, Shankun Xu, Wei Gao, Tao Zheng, Yuan Pan, Nengjie Huo

ABSTRACT

Intense surface glare from the waxy cuticles of fruits frequently blinds conventional machine vision, causing severe data loss and target misidentification in unstructured agricultural environments. To address these limitations, a multifunctional optoelectronic neuromorphic transistor based on a MoS ₂ /Ta ₂ NiS ₅ /MoSe ₂ van der Waals heterojunction is demonstrated, enabling the seamless integration of photovoltaic sensing and synaptic functions within a single device. By leveraging rational band engineering and the strong intrinsic in‐plane anisotropy of Ta ₂ NiS ₅ , high‐performance polarization photodetection and bio‐inspired synaptic dynamics are synergistically coupled. The device yields a high responsivity of 0.48 A/W and a significant polarization ratio of 12 at 635 nm. Unlike conventional vision chips, this dual‐functional platform leverages polarization‐dependent synaptic plasticity driven by controllable charge trapping at atomic‐sharp interfaces to execute nonlinear hardware‐level feature encoding. This capability facilitates the construction of a polarization‐enhanced neuromorphic vision system that inherently suppresses specular reflections to enable robust, non‐destructive discrimination of agricultural products. A peak classification accuracy of 89.8% at a 0° polarization angle demonstrates that the integrated design of photovoltaic sensing and neuromorphic processing effectively reduces data redundancy and energy consumption. This dual‐mode anisotropic platform offers an efficient solution for intelligent vision in precision agriculture and complex target recognition.