氧化物全光控突触研究进展  

Advances in All-optically Controlled Oxide Synapses

作　　者：单海 应宏微[3] 程培红 胡令祥[2] 王敬蕊 叶志镇[4,5] 诸葛飞[2,5,6,7] SHAN Hai;YING Hongwei;CHENG Peihong;HU Lingxiang;WANG Jingrui;YE Zhizhen;ZHUGE Fei(School of Materials Science and Chemical Engineering,Ningbo University,Ningbo 315211,China;Ningbo Institute of Materials Technology and Engineering,Chinese Academy of Sciences,Ningbo 315201,China;School of Electronic and Information Engineering,Ningbo University of Technology,Ningbo 315211,China;State Key Laboratory of Silicon and Advanced Semiconductor Materials,School of Materials Science and Engineering,Zhejiang University,Hangzhou 310027,China;Institute of Wenzhou,Zhejiang University,Wenzhou 325006,China;Center for Excellence in Brain Science and Intelligence Technology,Chinese Academy of Sciences,Shanghai 200072,China;School of Materials Science and Optoelectronic Technology,Chinese Academy of Sciences,Beijing 100029,China)

机构地区：[1]宁波大学材料科学与化学工程学院,浙江宁波315211 [2]中国科学院宁波材料技术与工程研究所,浙江宁波315201 [3]宁波工程学院电子与信息工程学院,浙江宁波315211 [4]浙江大学材料科学与工程学院,硅及先进半导体材料全国重点实验室,浙江杭州310027 [5]浙江大学温州研究院,浙江温州325006 [6]中国科学院脑科学与智能技术卓越创新中心,上海200031 [7]中国科学院大学材料科学与光电技术学院,北京100029

出　　处：《发光学报》2025年第2期245-259,共15页Chinese Journal of Luminescence

基　　金：国家自然科学基金(U20A20209,62304228);中国科学院战略性先导专项(XDB32050204);浙江省自然科学基金(LQ22F040003);宁波市自然科学基金(2023J356)。

摘　　要：类脑神经形态计算有望克服传统冯·诺依曼计算架构瓶颈,实现低功耗、高效信息处理,进而推动人工智能技术的发展。人工突触是构建神经形态系统的关键硬件,其中光电突触结合了电子学和光子学优势,具有光学感知、信息计算和存储等多种功能。新兴的全光控光电突触,其电导非易失性增大和降低均通过光信号实现,可有效避免电信号对器件微结构的破坏,改善工作稳定性,并且赋予突触器件新的功能。氧化物易制备,与CMOS工艺兼容性好,是使用最广泛的人工突触材料。本文梳理了具有长程可塑性的氧化物全光控突触器件研究进展,分别讨论了基于光波长和光功率调控的全光控突触,主要聚焦器件结构、材料选择和光电响应机制,最后分析了当前全光控突触发展面临的挑战。Brain-like neuromorphic computing is expected to overcome the bottleneck of traditional von Neumann computing architecture,achieving low power consumption and highly efficient information processing,thereby advancing artificial intelligence technology.Artificial synapses are key hardware for building neuromorphic systems,among which photoelectric synapses combine the advantages of electronics and photonics,offering multiple functions such as optical perception,information computation and storage.Emerging all-optically controlled photoelectric synapses,which allow nonvolatile increase and decrease in conductance by optical signals,can effectively prevent damage to the device microstructure caused by electrical signals,improving working stability,and endow synaptic devices with new functions.Oxide is the most widely used artificial synaptic material because of its ease of preparation and good compatibility with CMOS technology.This paper reviews the research progress of all-optically controlled(AOC)oxide synapses with long-term plasticity.The AOC synapses are discussed in terms of conductance modulation methods,including light wavelength and light power density modulation,focusing on device structure,material selection,and photoelectric response mechanism.Finally,we analyze the current challenges faced by all-optically controlled synapses.

关 键 词：神经形态计算 光电突触 全光控突触 长程可塑性 氧化物 

分 类 号：O472[理学—半导体物理]