基于氧化物基电解质栅控晶体管突触的关联学习

doi:10.15541/jim20220519

无机材料学报 ›› 2023, Vol. 38 ›› Issue (4): 399-405.DOI: 10.15541/jim20220519 CSTR: 32189.14.10.15541/jim20220519

所属专题：【信息功能】忆阻器材料与器件(202506)

• 专栏：神经形态材料与器件(特邀编辑：万青) • 上一篇下一篇

基于氧化物基电解质栅控晶体管突触的关联学习

方仁瑞¹^,²(), 任宽¹, 郭泽钰¹^,², 徐晗¹^,², 张握瑜¹^,², 王菲¹^,², 张培文¹^,², 李悦¹^,², 尚大山¹^,²()

1.中国科学院微电子研究所, 微电子器件与集成技术重点实验室, 北京 100029
2.中国科学院大学, 北京 100049

收稿日期:2022-09-04 修回日期:2022-09-30 出版日期:2023-04-20 网络出版日期:2022-10-28
通讯作者: 尚大山, 研究员. E-mail: shangdashan@ime.ac.cn
作者简介:方仁瑞(1994-), 男, 博士研究生. E-mail: fangrenrui@ime.ac.cn
基金资助:
国家重点基础研究发展计划(2018YFA0701500);国家自然科学基金(61874138)

Associative Learning with Oxide-based Electrolyte-gated Transistor Synapses

FANG Renrui¹^,²(), REN Kuan¹, GUO Zeyu¹^,², XU Han¹^,², ZHANG Woyu¹^,², WANG Fei¹^,², ZHANG Peiwen¹^,², LI Yue¹^,², SHANG Dashan¹^,²()

1. Key Laboratory of Microelectronic Devices and Integrated Technology, Institute of Microelectronics, Chinese Academy of Sciences, Beijing 100029, China
2. University of Chinese Academy of Sciences, Beijing 100049, China

Received:2022-09-04 Revised:2022-09-30 Published:2023-04-20 Online:2022-10-28
Contact: SHANG Dashan, professor, E-mail: shangdashan@ime.ac.cn
About author:FANG Renrui (1994-), male, PhD candidate. E-mail: fangrenrui@ime.ac.cn
Supported by:
National Key Basic Research Development Program of China(2018YFA0701500);National Natural Science Foundation of China(61874138)

摘要/Abstract

摘要：

电解质栅控晶体管(Electrolyte-gated transistors, EGTs)的沟道电导连续可调特性使其在构建神经形态计算系统中具有巨大应用潜力。本工作以非晶态Nb₂O₅作为沟道材料, Li_xSiO₂作为栅电解质材料, 制备了一种具备低沟道电导(~120 nS)的EGT器件。该器件利用Li⁺嵌入/脱出Nb₂O₅晶格导致的沟道电导连续可逆变化, 模拟了神经突触的短程可塑性(Short-term plasticity, STP)、长程可塑性(Long-term plasticity, LTP)以及STP向LTP的转变等功能。基于这种EGT突触特性, 本工作设计了关联学习电路, 实现了突触权重的负反馈调节, 并模拟了“巴普洛夫的狗”经典条件反射行为。这些结果展现出EGT作为神经突触器件的巨大潜力, 为实现神经形态计算硬件提供了器件参考。

关键词: 电解质栅控晶体管, 神经突触, 突触可塑性, 关联学习

Abstract:

The analog channel conductance modulation of electrolyte-gated transistors (EGTs) is a desirable property for the emulation of synaptic weight modulation and thus gives them great potential in neuromorphic computing systems. In this work, an all-solid-state electrochemical EGT was introduced with a low channel conductance (~120 nS) using amorphous Nb₂O₅ and Li-doped SiO₂ (Li_xSiO₂) as the channel and gate electrolyte materials, respectively. By adjusting the applied gate voltage pulse parameters, the reversable and nonvolatile modulation of channel conductance were achieved, which was ascribed to reversible intercalation/deintercalation of Li⁺ ions into/from the Nb₂O₅ lattice. Essential functionalities of synapses, such as the short-term plasticity (STP), long-term plasticity (LTP), and transformation from STP to LTP, were simulated successfully by conductive channel modulation of the EGTs. Based on these characteristics, a simple associative learning circuit was designed by parallel a resistor between the gate and the source terminals. The Pavlovian dog classical conditioning behavior was simulated based on associative learning circuit, where the resistor represented the unconditioned synapse and shared the gate voltage with EGT according to the proportion of its resistance, and the resistance between gate and source for negative feedback regulation of synaptic weights. These results demonstrate the potential of EGT for artificial synaptic devices and provide an insight into hardware implementation of neuromorphic computing systems.

Key words: electrolyte-gated transistor, synapse, synaptic plasticity, associative learning

中图分类号:

TQ174

方仁瑞, 任宽, 郭泽钰, 徐晗, 张握瑜, 王菲, 张培文, 李悦, 尚大山. 基于氧化物基电解质栅控晶体管突触的关联学习[J]. 无机材料学报, 2023, 38(4): 399-405.

FANG Renrui, REN Kuan, GUO Zeyu, XU Han, ZHANG Woyu, WANG Fei, ZHANG Peiwen, LI Yue, SHANG Dashan. Associative Learning with Oxide-based Electrolyte-gated Transistor Synapses[J]. Journal of Inorganic Materials, 2023, 38(4): 399-405.

图/表 5

图1 经历不同栅压范围扫描后EGT的输出特性曲线(插图为器件的光学图像)(a), 转移特性曲线(b), 栅极泄漏电流曲线(c)及其工作机理(d~f)

Fig. 1 Output characteristic curves under different gate voltages with inset showing the optical pattern (a), transfer characteristic curve (b), gate leakage current curve (c), and working mechanism (d-f) of EGT

图2 EGT和生物突触的类比(a), (3 V, 9 s)脉冲刺激下的EPSC(b), 一对间隔0.6 s的栅极电压脉冲刺激下的PPF行为(c), PPF指数随脉冲间隔的变化(d)

Fig. 2 Analogy between EGT and biological synapse (a), EPSC under (3 V, 9 s) pulse (b), PPF stimulated by a pair of gate voltage pulses spaced 0.6 s apart (c), PPF index as a function of pulse interval (d)

图3 EGT对不同的脉冲幅值(a)、脉冲宽度(b)和脉冲频率(c)的栅压脉冲的响应, 与脉冲幅值(d)和脉冲宽度(e)依赖的电导变化比

Fig. 3 Response of EGT to a series of gate voltage pulses of different pulses amplitude (a), pulses width (b), and pulses frequency (c), dependence of conductance change ratio on pulses amplitude (d) and pulses width (e)Colorful figures are available on website

图4 长期增强突触权重对神经形态计算功能的影响

Fig. 4 Effect of long-term enhanced synaptic weights on neuromorphic computational function (a) Long-term potentiation induced by 64 positive gate voltage pulses (3.6 V, 10 ms); (b) long-term depression induced by 64 negative gate voltage pulses (-3.4 V, 10 ms) and taking points 10 s after pulses; (c) cyclic change results for 100 cycles of alternately applied 64 pulses (+3.6 V/-3.4 V, 10 ms) and taking points 10 s after pulses

图5 利用EGT实现经典条件反射示意图(a)和对应的等效电路(b), 生物经典条件反射模拟(c)

Fig. 5 (a) Schematic diagram of classical conditioning using EGT and (b) corresponding equivalent circuit, and (c) simulation of biological classical conditioning

参考文献 28

[1]	MURAT O, NICOLAS E, WANG B M, et al. Nanosecond protonic programmable resistors for analog deep learning. Science, 2022, 377: 539.
[2]	LIANG F X, WANG T, HOU T H. Progress and benchmark of spiking neuron devices and circuits. Advanced Intelligent Systems, 2021, 3(8):2100007. DOI URL
[3]	ABU S, MANUEL L G, RIDUAN K A, et al. Memory devices and applications for in-memory computing. Nature Nanotechnology, 2020, 15(7):529. DOI PMID
[4]	YANG K, YANG J J, HUANG R, et al. Nonlinearity in memristors for neuromorphic dynamic systems. Small Science, 2021, 2: 2100049.
[5]	ZHU J D, ZHANG T, YANG Y C, et al. A comprehensive review on emerging artificial neuromorphic devices. Applied Physics Reviews, 2020, 7: 011312.
[6]	NGUYEN N A, SCHNEEGANS O, SALOT R, et al. An ultralow power Li_xTiO₂-based synaptic transistor for scalable neuromorphic computing. Advanced Electronic Materials, 2022, 8(12):2200607. DOI URL
[7]	ZHANG W Q, GAO B, TANG J S, et al. Neuro-inspired computing chips. Nature Electronics, 2020, 3: 371.
[8]	XU H, LU J K, LI Y, et al. Improvement of weight stability in Li-ion-based electrolyte-gated transistor synapse by silica protective process. Applied Physics Letters, 2022, 121(11):113505. DOI URL
[9]	LEE H, RYU D G, LEE G, et al. Vertical metal-oxide electrochemical memory for high-density synaptic array based high-performance neuromorphic computing. Advanced Electronic Materials, 2022, 8(8):2200378. DOI URL
[10]	NAYEON K, HEEBUM K, HYUN W K, et al. Understanding synaptic characteristics of nonvolatile analog redox transistor based on mobile ion-modulated-electrolyte thickness model for neuromorphic applications. Applied Physics Letters, 2022, 121(7):072105. DOI URL
[11]	LEE J, NIKAM R D, KWAK M, et al. Improved synaptic characteristics of oxide-based electrochemical random access memory at elevated temperatures using integrated micro-heater. IEEE Transactions on Electron Devices, 2022, 69: 2218. DOI URL
[12]	REVANNATH D N, LEE J, CHOI W, et al. Ionic sieving through one-atom-thick 2D material enables analog nonvolatile memory for neuromorphic computing. Small, 2021, 17(44):2103543. DOI URL
[13]	FENG G, JIANG J, ZHAO Y H, et al. A sub-10 nm vertical organic/inorganic hybrid transistor for pain-perceptual and sensitization-regulated nociceptor emulation. Advanced Materials, 2020, 32(6):1906171. DOI URL
[14]	LEE J, NIKAM R D, KWAK M, et al. Strategies to improve the synaptic characteristics of oxygen-based electrochemical random-access memory based on material parameters optimization. ACS Applied Materials & Interfaces, 2022, 14(11):13450.
[15]	CHENG Y C, LI H, LIU B, et al. Vertical 0D-perovskite/2D-MoS₂ van der Waals heterojunction phototransistor for emulating photoelectric-synergistically classical pavlovian conditioning and neural coding dynamics. Small, 2020, 16(45):2005217. DOI URL
[16]	LI Y, LU J K, SHANG D S, et al. Oxide-based electrolyte-gated transistors for spatiotemporal information processing. Advanced Materials, 2020, 32(47):2003018. DOI URL
[17]	LI Y, XUAN Z H, LU J K, et al. One transistor one electrolyte-gated transistor based spiking neural network for power-efficient neuromorphic computing system. Advanced Functional Materials, 2021, 31(26):2100042. DOI URL
[18]	LI Y, XU H, LU J K, et al. Electrolyte-gated transistors with good retention for neuromorphic computing. Applied Physics Letters, 2022, 120(2):021901. DOI URL
[19]	AUGUSTYN V, COME J, LOWE M A, et al. High-rate electrochemical energy storage through Li⁺ intercalation pseudocapacitance. Nature Materials, 2013, 12(6):518. DOI
[20]	GRIFFITH K J, FORSE A C, GRIFFIN J M, et al. High-rate intercalation without nanostructuring in metastable Nb₂O₅ bronze phases. Journal of the American Chemical Society, 2016, 138(28):8888. DOI URL
[21]	PRADEP P A, WADE G R. Determinants of the time course of facilitation at the granule cell to Purkinje cell synapse. The Journal of Neuroscience, 1996, 16(18):5661. DOI URL
[22]	ROBERT S Z, REGEHR W G. Short-term synaptic plasticity. Annual Review of Physiology, 2002, 64: 355.
[23]	WANG I T, CHANG C C, CHIU L W, et al. 3D Ta/TaO_x/TiO₂/Ti synaptic array and linearity tuning of weight update for hardware neural network applications. Nanotechnology, 2016, 27(36):365204. DOI URL
[24]	JANG J W, PARK S, BURR G W, et al. Optimization of conductance change in Pr_1-_xCa_xMnO₃-based synaptic devices for neuromorphic systems. IEEE Electron Device Letters, 2015, 36(5):457. DOI URL
[25]	MCGANN J P. Associative learning and sensory neuroplasticity: how does it happen and what is it good for? Learning & Memory, 2015, 22(11):567.
[26]	TRAXLER J, MADDEN V J, MOSELEY G L, et al. Modulating pain thresholds through classical conditioning. PeerJ, 2019, 7: 6486.
[27]	MAURICIO R P, BITTERMAN M E. The role of contingency in classical conditioning. Psychological Review, 1990, 97(3):396. PMID
[28]	YU F, ZHU L Q, XIAO H, et al. Restickable oxide neuromorphic transistors with spike-timing-dependent plasticity and pavlovian associative learning activities. Advanced Functional Materials, 2018, 28(44):1804025. DOI URL

基于氧化物基电解质栅控晶体管突触的关联学习

Associative Learning with Oxide-based Electrolyte-gated Transistor Synapses

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 5

参考文献 28

相关文章 1

编辑推荐

Metrics

本文评价