神经形态工程学
各向异性
人工神经网络
光电子学
计算机科学
材料科学
极化(电化学)
人工智能
可重构性
机器视觉
电子工程
图像质量
计算机视觉
预处理器
兴奋性突触后电位
集成电路
突触重量
物理
夜视
范德瓦尔斯力
图像传感器
铁电性
作者
Cheng Qi,Jinpeng Zhao,Weidong Dai,Xin Peng,Yunkang Tang,Shangqing Xu,Mengya Zhang,Yicheng Wang,Xing Xu,Yipeng Zhao,Honglai Li,Wei Zhou,Liang Ma,Xiaolin Wei
摘要
ABSTRACT Against the backdrop of rapid advances in big data and artificial intelligence, the exponential growth of visual information has driven increasingly demanding requirements for imaging quality and intelligent perception. However, inherent limitations in conventional image sensing—including restricted sensitivity, narrow spectral response, low integration density, and insufficient adaptability—impede their ability to meet these demands. Here, we report an anisotropic optoelectronic synaptic image sensor based on 2D non‐centrosymmetric van der Waals ferroelectric NbOCl 2 . Distinct from conventional centrosymmetric 2D semiconductors, NbOCl 2 possesses intrinsic spontaneous polarization and strong in‐plane anisotropy arising from its polar crystal structure, enabling built‐in polarization‐sensitive responses and anisotropic carrier transport. The device exhibits pronounced anisotropic properties, with electrical transport and photoresponse anisotropy factors reaching 1.68 and 1.76. Furthermore, temporally stretched excitatory postsynaptic current behavior enables angle‐resolved optical storage. By integrating the NbOCl 2 synaptic element with an image‐signal‐recognition system, polarization‐encoded preprocessing of the Modified National Institute of Standards and Technology handwritten digit dataset is achieved, and a three‐layer artificial neural network attains 95.1% recognition accuracy. Our findings establish 2D NbOCl 2 as a highly promising material platform for polarization‐sensitive optoelectronic synapses, providing a compact and energy‐efficient foundation for next‐generation neuromorphic vision and bioinspired perception systems.
科研通智能强力驱动
Strongly Powered by AbleSci AI