神经形态工程学
光探测
材料科学
光电子学
异质结
计算机科学
冯·诺依曼建筑
偏压
记忆电阻器
电子工程
卷积(计算机科学)
逻辑门
CMOS芯片
晶体管
光学计算
电压
纳米线
突触
光电探测器
联轴节(管道)
卷积神经网络
人工神经网络
光电效应
光伏系统
光开关
纳米技术
图像传感器
光功率
匹配(统计)
水准点(测量)
功能(生物学)
作者
Kai Cao,Zhengming Lv,Fengzhen Huang,Zeyu Xiao,Biwei Shen,Yuanyuan Zhang,Xueli Hu,Jiaqi Lian,Shuhao Jiang,Jinsong Zhu,Xiaomei Lü
出处
期刊:Small
[Wiley]
日期:2026-01-23
卷期号:22 (17): e13679-e13679
被引量:7
标识
DOI:10.1002/smll.202513679
摘要
ABSTRACT Constructing multifunctional neuromorphic optoelectronic devices that integrate sensing, memory and computing is an important approach to overcome the computational bottlenecks of the traditional von Neumann architecture and improve the performance of artificial visual systems. In this work, a Bi 4 . 15 Nd 0 . 85 Ti 3 FeO 15 (BNTF)/ZnO ferroelectric‐semiconductor heterojunction device was designed and fabricated via the sol‐gel method, realizing the integration of self‐powered photodetection and low‐power optical synapse in a single two‐terminal device. Under zero bias, the device exhibited polarization‐tunable photoresponsivity ( R up /R down = 4.1) with high response speed, and thus excellent imaging capability and photoelectric coupling logic operation functions. Under small bias voltage (≤0.5 V), it could act as an artificial optical synapse device and achieve multiple types of synaptic plasticity. Notably, by the self‐denoising function of the optical synapse, the recognition accuracy of digit images improved from 87.5% to 96.8%. The in‐sensor reservoir computing system based on BNTF/ZnO synapse array achieved a high accuracy of 97.3% in moving target recognition. The device also featured an ultralow power consumption of 4.6 fJ, matching or even lower than that of most biological synapses. This study revealed the unique advantages and application prospects of ferroelectric‐semiconductor heterojunction devices in constructing low‐power integrated imaging‐memory‐computing vision systems.
科研通智能强力驱动
Strongly Powered by AbleSci AI