Low-Power Artificial Neurons Based on Ag/TiN/HfAlOx/Pt Threshold Switching Memristor for Neuromorphic Computing

Lu, Yifan; Li, Yang; Li, Haoyang; Wan, Tianqing; Huang, Xiaodi; He, Yuhui; Miao, Xiangshui

doi:10.1109/led.2020.3006581

Cited by 73 publications

(34 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…We assume that the higher diffusion coefficient of Ag ions in V 2 C may enhance the diffusive process of Ag for threshold switching and emulate faster and more controllable neurons [45,56,57]. In conclusion, the superior strength-modulated spike frequency characteristic has been successfully implemented, which may strengthen the feasibility of MXene-based artificial neurons for neuromorphic systems [18,[58][59][60].…”

Section: Resultsmentioning

confidence: 74%

Artificial Neurons Based on Ag/V2C/W Threshold Switching Memristors

et al. 2021

View full text Add to dashboard Cite

Artificial synapses and neurons are two critical, fundamental bricks for constructing hardware neural networks. Owing to its high-density integration, outstanding nonlinearity, and modulated plasticity, memristors have attracted emerging attention on emulating biological synapses and neurons. However, fabricating a low-power and robust memristor-based artificial neuron without extra electrical components is still a challenge for brain-inspired systems. In this work, we demonstrate a single two-dimensional (2D) MXene(V2C)-based threshold switching (TS) memristor to emulate a leaky integrate-and-fire (LIF) neuron without auxiliary circuits, originating from the Ag diffusion-based filamentary mechanism. Moreover, our V2C-based artificial neurons faithfully achieve multiple neural functions including leaky integration, threshold-driven fire, self-relaxation, and linear strength-modulated spike frequency characteristics. This work demonstrates that three-atom-type MXene (e.g., V2C) memristors may provide an efficient method to construct the hardware neuromorphic computing systems.

show abstract

Section: Resultsmentioning

confidence: 74%

Artificial Neurons Based on Ag/V2C/W Threshold Switching Memristors

et al. 2021

View full text Add to dashboard Cite

show abstract

“…Similar results of TiO 2 buffer layer in an Ag/SiO 2 :Ag/TiO 2 /Si device was reported [ 42 ]. TiN layer could also slow down Ag diffusion into the dielectric and enhance the threshold switching properties [ 43 ].…”

Section: Resultsmentioning

confidence: 99%

Improved Stability and Controllability in ZrN-Based Resistive Memory Device by Inserting TiO2 Layer

Choi

Kim

2020

Micromachines

View full text Add to dashboard Cite

In this work, the enhanced resistive switching of ZrN-based resistive switching memory is demonstrated by embedding TiO2 layer between Ag top electrode and ZrN switching layer. The Ag/ZrN/n-Si device exhibits unstable resistive switching as a result of the uncontrollable Ag migration. Both unipolar and bipolar resistive switching with high RESET current were observed. Negative-SET behavior in the Ag/ZrN/n-Si device makes set-stuck, causing permanent resistive switching failure. On the other hand, the analogue switching in the Ag/TiO2/ZrN/n-Si device, which could be adopted for the multi-bit data storage applications, is obtained. The gradual switching in Ag/TiO2/ZrN/n-Si device is achieved, possibly due to the suppressed Ag diffusion caused by TiO2 inserting layer. The current–voltage (I–V) switching characteristics of Ag/ZrN/n-Si and Ag/TiO2/ZrN/n-Si devices can be well verified by pulse transient. Finally, we established that the Ag/TiO2/ZrN/n-Si device is suitable for neuromorphic application through a comparison study of conductance update. This paper paves the way for neuromorphic application in nitride-based memristor devices.

show abstract

“…(e) Circuit of the LIF neuron based on diffusive memristors. (f) Electrical behavior of the LIF neuron based on diffusive memristors [54] 为了进一步探索忆阻神经元的功能，本课题组 [43] 制作了一种具有快速易失特性和电池效应的 W/WO 3 /PEDOT:PSS/Pt 忆阻器；具有电池效应的忆阻器内部存在多种载流子的传输过程，与定制电阻串联后可以模拟神经元中细胞膜的离子迁移过程。基于两个 W/WO 3 /PEDOT:PSS/Pt 忆阻器搭建了神经元电路，包括连接模块、LGP 模块、比较模块和类生物脉冲发放模块，并成功实现了类生物脉冲发放和时空信息整合功能。基于同样的原理和方式，深圳大学韩素婷课题组 [79] 制作了 Au/MAPbI 然后逐渐上升到静息电位，更具生物合理性。南加州大学杨建华课题组 [54] 利用 TS 忆阻器和电容并联后再串联一个电阻组成了 LIF 神经元(图 4(e) ) ，并联电容上的电压表示 LGP，当有脉冲时，电荷在电容中累积，发生整合；当没有脉冲时，电容发生自发的漏电；一旦电容电压高于忆阻器的阈值时，该器件切换到 LRS，此时电容的电压无法保持，开始发放电流脉冲(图 4(f) ) 。为了进一步降低人工神经元电路复杂性，华中科技大学缪向水课题组 [82] 直接利用 Pt/Ag/TiN/HfAlO x /Pt 器件和电阻串联搭建了 LIF 神经元电路，模拟了漏电整合发放行为。研究人员除了利用各种模型实现 LIF 神经元和类生物发放功能外，也在致力于探索更多神经元功能的模拟 [83] 。生物神经元在受到刺激产生兴奋时，会对自身和相近的神经元产生抑制作用，也就是在一定时间内，即使再给自身或相近的神经元刺激时，也不会再产生兴奋，前者称为不应期(Refractory period, RP) ，后者称为侧向抑制 (Lateral inhibition) ，侧向抑制可以避免信息过载，支持神经网络进行竞争性学习。中国科学院微电子研究所刘明课题组 [84] 提出了基于忆阻器-CMOS 混合神经元电路数据的分析与处理上颇具优势 [85] 。图 6 神经网络的结构 [86] . (a)前馈式神经网络； (b)RNN；( c)SNN Figure 6 Architectures of neural networks [86] .…”

Section: Hodgkin-huxley(hh)模型unclassified

基于离子型忆阻器的脑启发式人工智能

Wen¹,

Huang²,

Wang³

et al. 2022

Chin. Sci. Bull.

View full text Add to dashboard Cite

Human society is now moving from informatization to intelligence, in which artificial intelligence (AI) is the key. In many fields, AI has demonstrated the ability to surpass the human brain, due to the continuous innovation of AI algorithms and rapidly increasing computing powers of super-computers. Considering the current data volume, it is particularly important to improve the computing power. In this aspect, innovative AI chips are critical.In the pursuit of high performance and low power consumption, AI chips face two bottlenecks. (1) The memory wall: The access speed of memories cannot keep up with the speed of data consuming by computing units, which perplexes the computer architecture for a long time. (2) The failure of the Moore's law, due to the limitation of basic physical principles. It becomes more and more difficult to improve the integration density and reduce the power consumption of chips by reducing the transistor size.A memristor is a kind of circuit device with memory effect. Under the action of external electric fields, the device can change reversibly between high and low conductance states.The different conductance states are controlled by the migration of various ionic defects, e.g., oxygen vacancies, proton defects, cations, etc. Memristors have the characteristics of simple sandwich structure, fast operation speed, low energy consumption, and rich performances. In addition, memristor arrays with crossbar structure can achieve the integration of the data storage and computing by the vector-matrix multiplication. Therefore, memristor-based neuromorphic computing, with its ability to perform computation where data are generated and the advantages of convenient realization of brain-inspired algorithms, is very promising to crack the memory wall and the failure of the Moore's law. Ionic defects are electrically

show abstract

Low-Power Artificial Neurons Based on Ag/TiN/HfAlOx/Pt Threshold Switching Memristor for Neuromorphic Computing

Cited by 73 publications

References 28 publications

Artificial Neurons Based on Ag/V2C/W Threshold Switching Memristors

Artificial Neurons Based on Ag/V2C/W Threshold Switching Memristors

Improved Stability and Controllability in ZrN-Based Resistive Memory Device by Inserting TiO2 Layer

基于离子型忆阻器的脑启发式人工智能

Contact Info

Product

Resources

About