神经形态工程学
MNIST数据库
材料科学
记忆电阻器
光电子学
突触
计算机科学
调制(音乐)
纳米棒
人工神经网络
信号(编程语言)
量子点
信号处理
电子工程
突触重量
电压
特征(语言学)
布线(电子设计自动化)
图像处理
光开关
突触可塑性
作者
Wenxiao Zhao,Zexi Lin,Liyan Zhang,Sheng Xu,Enguo Chen,Tailiang Guo,Yun Ye
出处
期刊:ACS Photonics
[American Chemical Society]
日期:2025-11-29
卷期号:12 (12): 6935-6945
标识
DOI:10.1021/acsphotonics.5c02156
摘要
In novel neuromorphic computations, it is essential to achieve bidirectional modulation of synaptic plasticity by external signal stimulation. However, conventional optoelectronic co-modulation leads to the complexity of external devices as well as operational drudgery. Therefore, it is significant that the modulation of synaptic plasticity in neuromorphic devices is realized entirely by optical signal stimulation. Thus, an all-optical memristor based on CdS nanorods/PbS quantum dots PN junctions and a PMMA defect capture layer is proposed. In particular, the CdS nanorods can be modulated in size by controlling the reaction conditions. The size of the nanorods affects the UV response current of the CdS-based memristor. The all-optical synaptic device exhibits EPSC under UV light (365 nm) irradiation and IPSC under red light (635 nm) irradiation. Transient absorption spectroscopy results indicate that the presence of a PN junction and defect trapping are responsible for bidirectional photomodulation. The “AND” and “XOR” logic operations are realized by selecting the input of light signals of different wavelengths. In addition, the feature enhancement operation of the image is realized by using the bidirectional photoresponse of the device. Finally, an artificial neural network is constructed for image recognition on the MNIST data set with a recognition accuracy of 90%. This work enriches the choice of materials for optoelectronic synaptic memristors and the design of device structures for neuromorphic computing.
科研通智能强力驱动
Strongly Powered by AbleSci AI