神经形态工程学
材料科学
光电子学
纳米线
铁电性
调制(音乐)
记忆电阻器
光电导性
功率(物理)
纳米技术
电子工程
人工神经网络
计算机科学
电介质
人工智能
哲学
工程类
物理
美学
量子力学
作者
Pengshan Xie,Yulong Huang,Wei Wang,You Meng,Zhengxun Lai,Fei Wang,SenPo Yip,Xiuming Bu,Weijun Wang,Dengji Li,Jia Sun,Johnny C. Ho
出处
期刊:Nano Energy
[Elsevier]
日期:2022-01-01
卷期号:91: 106654-106654
被引量:41
标识
DOI:10.1016/j.nanoen.2021.106654
摘要
The gallop of artificial intelligence ignites urgent demand on information processing systems with ultralow power consumption, reliable multi-parameter control and high operation efficiency. Here, the poly(vinylidene fluoride-trifluoroethylene) (P(VDF-TrFE)) wrapped InGaAs nanowire (NW) artificial synapses capable to operate with record-low subfemtojoule power consumption are presented. The essential synaptic behaviors are mimicked and modulated effectively by adjusting the thickness of top P(VDF-TrFE) films. Moreover, the long-term depression is realized by applying visible light (450 nm) because of the negative photoconductivity of InGaAs nanowires. Combined with optimal P(VDF-TrFE) films, the synaptic devices have the more linear long-term potentiation/depression characteristics and the faster supervised learning process simulated by hardware neural networks. The Pavlovian conditioning is also performed by combining electrical and infrared stimuli. Evidently, these ultralow-operating-power synapses are demonstrated with the brain-like behaviors, effective function modulation, and more importantly, the synergistic photoelectric modulation, which illustrates the promising potentials for neuromorphic computing systems.
科研通智能强力驱动
Strongly Powered by AbleSci AI