基于ZnO忆阻器的神经突触仿生电子器件  

Synaptic Devices Based on ZnO Memristors

作　　者：潘若冰[1,2] 胡丽娟[1] 曹鸿涛[2] 竺立强[2] 李俊[2] 李康[2] 梁凌燕[2] 张洪亮[2] 高俊华[2] 诸葛飞[2] PAN Ruobing HU Lijuan CAO Hongtao ZHU Liqiang LI Jun LI Kang LIANG Lingyan ZHANG Hongliang GAO Junhua ZHUGE Fei(Institute of Materials Science, School of Materials Science and Engineering, Shanghai University, Shanghai 200072, China Ningbo Institute of Materials Technology and Engineering, Chinese Academy of Sciences, Ningbo 315201, China)

机构地区：[1]上海大学材料科学与工程学院,上海200072 [2]中国科学院宁波材料技术与工程研究所,浙江宁波315201

出　　处：《材料科学与工程学报》2017年第2期232-236,共5页Journal of Materials Science and Engineering

基　　金：国家自然科学基金资助项目(51272261和61474127)

摘　　要：本文采用ZnO忆阻器模拟了生物神经突触的记忆和学习功能。ZnO突触器件表现出典型的随时间指数衰减的突触后兴奋电流(EPSC),以及EPSC的双脉冲增强行为。在此基础上,实现了学习-遗忘-再学习的经验式学习行为,以及四种不同种类的电脉冲时刻依赖可塑性学习规则。ZnO突触器件实现了超低能耗操作,单次突触行为能耗最低为1.6pJ,表明其可以用来构筑未来的人工神经网络硬件系统,最终开发出与人脑结构类似的认知型计算机以及类人机器人。ZnO memristive devices have been employed to emulate synaptic memory and learning functions.ZnO synaptic devices show a typical excitatory post-synaptic current（EPSC）,i.e.exponentially decreasing with time,and pair-pulse facilitation behavior of EPSC.Furthermore,the learning-forgettingrelearning empirical behavior and four types of spike-timing-dependent-plasticity learning rules have been implemented.Ultra-low energy consumption operation has been realized in ZnO synaptic devices showing a minimum energy consumption of 1.6pJ for a single synaptic behavior.The results indicate that ZnO synaptic devices can be potentially used to construct the future artificial neural networks in hardware and ultimately develop cognitive computers operating like human brains and humanoid robots.

关 键 词：忆阻器 神经突触器件 人工神经网络 ZNO 

分 类 号：TB43[一般工业技术]