臺灣博碩士論文加值系統

由於傳統馮紐曼架構無法快速處理大量數據以符合人工智慧運算需求，因此利用非揮發性記憶體作為突觸以仿造大腦神經架構提升運算速度。電阻式記憶體為電子突觸應用之最具潛力元件，因此本論文利用銅為電極製作電阻式記憶體元件，探討其電子突觸特性，並藉由嵌入銅奈米粒(Nanoparticles，NP)於氧化層中改善其非理想特性。
第一部分探討Cu/SiO2/Pt和Cu/SiO2/Cu-NP/SiO2/Pt改善元件，透過循環掃描確認元件電阻切換機制，元件高電阻狀態的傳導機制為蕭特基傳導而低電阻狀態(LRS)為歐姆傳導。第二部分探討Cu/SiO2/Cu-NP/SiO2/Pt元件的電子突觸特性，經由脈衝波形量測元件突觸特性。Cu/SiO2/Cu-NP/SiO2/Pt元件相較於Cu/SiO2/Pt元件，直流電壓下表現出穩定的電阻切換特性，耐久度達103次，記憶時間於85°C下達104秒。元件的LRS電阻溫度係數為正，因此傳導路徑為金屬原子組成。突觸特性利用元件內部銅原子積累，經由調配後波形使元件電導透過脈衝電壓(0.35 V / -0.35 V , 500 s)逐漸增加或減少，增益(Potentiation)與抑制(depression)非線性度為1.85%與6.07%。耐久度脈衝參數經由調配為(0.2 V / -0.22 V , 250 us)，使電導於固定範圍內變化，元件耐久度可達27000次。長短期記憶擬合不同脈衝次數結果，鬆弛時間(relaxation time)τ於1、10和100次時分為225 s、290 s和380 s。成對脈衝促進擬合結果c1與c2為4.8與0.6，快速和緩慢鬆弛時間τ1與τ2分別為0.25 ms與0.5 s。間隔時間1 ms時影響5.24%，間隔時間0.2 s時影響僅剩0.28%。而後量測元件短期增強(Short Time Potentiation)特性，得知施加脈衝振幅相同時次數更多元件電導變化更大，強直後增強(Post-tetanic Potentiation)特性，經由不同頻率波形和振幅觀測出頻率高和振幅大時，元件電導變化更大，皆與生物突觸特性相似。

The traditional Von Neumann architecture cannot process large amounts of data quickly enough to meet the needs of artificial intelligence computing. Therefore, non-volatile memory is used as a synapse to imitate the neural architecture to improve the computing performance. Resistive memory (RRAM) is the most promising device for electronic synapse applications. Therefore, this thesis used copper as electrodes to fabricate RRAM devices, investigated its electronic synaptic properties, and improved its non-ideal properties by embedding copper nanoparticles in the oxide layer.
The first part studied Cu/SiO2/Pt and Cu/SiO2/Cu-NP/SiO2/Pt structures. These RRAM devices can be continuously switched their resistance with applied voltages. The conduction mechanisms of the high-resistance state (HRS) and low-resistance state (LRS) are dominated by Schottky emission and Ohmic conduction, respectively. The second part used voltage-pulse measurement to investigate the electronic synaptic properties of the Cu/SiO2/Cu-NP/SiO2/Pt device. Compared with the Cu/SiO2/Pt device, the Cu/SiO2/Cu-NP/SiO2/Pt device exhibited stable resistance switching characteristics with endurance of 103 times and retention of 104 seconds at 85 °C. The LRS temperature coefficient of resistance of the device was positive, so the conduction path were composed of metal atoms. The accumulation of copper atoms inside the device created synaptic properties. The device conductance was gradually increased or decreased by pulses (0.35 V / -0.35 V, 500 us). The nonlinearities of potentiation and depression operations were 1.85% and 6.07%, respectively. The conductance can be changed within a fixed range by pulse operation, and the operation number can be up to 27,000 times. According to the fitting results of long short-term memory, the relaxation time τ were 225 s, 290 s and 380 s for at 1, 10 and 100 pulses. The results of paired-pulse facilitation indicated that fast relaxation time (τ1) and slow relaxation time (τ2) were 0.25 ms and 0.5 s, respectively. The characteristics of short time potentiation showed that the more same applied pulses led to the greater conductance change. The characteristics of post-tetanic potentiation indicated that the higher frequency and larger amplitude of applied pulses caused greater conduction change, which is similar to the biological synapse.

目錄
中文摘要 I
英文摘要 III
誌謝 V
目錄 VI
圖目錄 VIII
表目錄 XII
第一章 緒論 1
1-1前言 1
1-2研究動機與目的 1
1-3論文架構 2
第二章 文獻回顧與基礎理論 3
2-1電阻式記憶體發展 3
2-2電阻式記憶體和電子突觸比較 9
2-3電子突觸操作特性介紹 17
2-4電流機制 19
2-5電子突觸記憶特性探討 21
第三章 實驗步驟 26
3-1實驗流程說明 26
3-2實驗設備 27
3-3元件製備 31
3-4電性分析 33
第四章 結果與討論 42
4-1 Cu/SiO2/Pt元件電阻切換特性與傳導機制探討 42
4-1-1元件製備流程 42
4-1-2元件電阻切換機制 43
4-1-3元件耐久度分析 46
4-1-4元件記憶時間分析 47
4-1-5元件電阻溫度係數分析 49
4-2 Cu/SiO2/Cu-NP/SiO2/Pt元件之突觸特性分析 50
4-2-1增益與抑制分析 50
4-2-2耐久度分析 55
4-2-3長短期記憶分析 57
4-2-4成對脈衝促進分析 58
4-2-5短期增強和抑制特性分析 59
4-2-6強直後增強特性分析 62
4-2-7 Cu/SiO2/Cu-NP/SiO2/Pt元件絲狀路徑形成與斷裂 64
第五章 結果與未來展望 65
5-1結論 65
5-1-1電導變化 65
5-1-2非線性度 65
5-1-3長短期記憶 66
5-1-4成對脈衝促進 66
5-2未來展望 67
參考文獻 68
論文比對 72

[1]F. Anzalone, "Statistical models of program/verify algorithm in resistive memory arrays for neural network accelerators", Politesi, 2020.
[2]Z. Shen, C. Zhao, and Y. Qi et al., "Advances of RRAM devices: resistive switching mechanisms, materials and bionic synaptic application", Nanomaterials 10, no 8: 1437, Jul. 2020.
[3]H. S. P. Wong, S. Raoux, and S. B. Kim et al., "Phase change memory", IEEE, vol. 98, no. 12, pp. 2201-2227, Dec. 2010.
[4]X. Zhang, L. Xu, and H. Zhang et al., "Effect of joule heating on resistive switching characteristic in AlOx cells made by thermal oxidation formation", Nanoscale Research Letters, 15(1), pp.1-8, Jan 2020.
[5]S. Munjal, and N. Khare, "Valence change bipolar resistive switching accompanied with magnetization switching in CoFe2O4 thin film", Scientific reports, 7(1), pp.1-10, Sep 2017.
[6]W. S. Choi, J. T. Jang, and D. Kim et al., "Influence of Al2O3 layer on InGaZnO memristor crossbar array for neuromorphic applications", Chaos, Solitons & Fractals, vol. 156, p.111813, Mar 2022.
[7]J. Y. Mao, L. Zhou, and Y. Ren et al., "A bio-inspired electronic synapse using solution processable organic small molecule", Journal of Materials Chemistry C, 7(6), pp.1491-1501, Dec 2018.
[8]J. Park, "Neuromorphic computing using emerging synaptic devices: a retrospective summary and an outlook", Electronics, vol. 9(9), p.1414, Sep 2020.
[9]H. U. Dike, Y. Zhou, and K. K. Deveerasetty et al., "Unsupervised learning based on artificial neural network: a review", In 2018 IEEE International Conference on Cyborg and Bionic Systems (pp. 322-327), Oct 2018.
[10]M. Ismail, C. Mahata, and H. Abbas et al., "Bipolar, complementary resistive switching and synaptic properties of sputtering deposited ZnSnO-based devices for electronic synapses", Journal of Alloys and Compounds, vol. 862, p.158416, May 2021.
[11]H. Cho, and S. Kim, "Short-term memory dynamics of TiN/Ti/TiO2/SiOx/Si resistive random access memory", Nanomaterials, vol. 10(9), p.1821, Sep 2020.
[12]N. C. Das, M. Kim, and J. R. Rani et al., "Low-temperature characteristics of magnesium fluoride based bipolar RRAM devices", Nanoscale, 14(10), pp.3738-3747, Feb 2022.
[13]Y. Sun, H. Xu, and C. Wang et al., "A Ti/AlOx/TaOx/Pt analog synapse for memristive neural network", IEEE Electron Device Letters, 39(9), pp.1298-1301, Sep 2018.
[14]S. Chandrasekaran, F. M. Simanjuntak, and D. Panda et al., "Enhanced synaptic linearity in ZnO-based invisible memristive synapse by introducing double pulsing scheme", IEEE Transactions on Electron Devices, 66(11), pp.4722-4726, Nov 2019.
[15]P. Y. Chen, X. Peng, and S. Yu et al., "NeuroSim: a circuit-level macro model for benchmarking neuro-inspired architectures in online learning", IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, vol. 37(12), pp.3067-3080, Dec 2018.
[16]Y. Xi, B. Gao, and J. Tang et al., "In-memory learning with analog resistive switching memory: a review and perspective", Proceedings of the IEEE, 109(1), pp.14-42, Jan 2020.
[17]N. Ilyas, D. Li, and C. Li et al., "Analog switching and artificial synaptic behavior of Ag/SiOx:Ag/TiOx/p++-Si memristor device", Nanoscale research letters, 15(1), pp.1-11, Jan 2020.
[18]G. González-Cordero, M. Pedro, and J. Martin-Martinez et al., "Analysis of resistive switching processes in TiN/Ti/HfO2/W devices to mimic electronic synapses in neuromorphic circuits", Solid-State Electronics, 157, pp.25-33, Jul 2019.
[19]K. C. Chuang, C. Y. Chu, and H. X. Zhang et al., "Impact of the stacking order of HfOx and AlOx dielectric films on RRAM switching mechanisms to behave digital resistive switching and synaptic characteristics", IEEE Journal of the Electron Devices Society, vol. 7, pp.589-595, May 2019.
[20]D. Lee, A. S. Sokolov, and B. Ku et al., "Improved switching and synapse characteristics using PEALD SiO2 thin film in Cu/SiO2/ZrO2/Pt device", Applied Surface Science, 547, p.149140, May 2021.
[21]Y. Sun, H. Xu, and S. Liu et al., "Short-term and long-term plasticity mimicked in low-voltage Ag/GeSe/TiN electronic synapse", IEEE Electron Device Letters, 39(4), pp.492-495, Apr 2018.
[22]K. Zarudnyi, A. Mehonic, and L. Montesi et al., "Spike-timing dependent plasticity in unipolar silicon oxide RRAM devices", Frontiers in Neuroscience, 12, p.57, Feb 2018.
[23]W. Chen, H. J. Barnaby, and M. N. Kozicki., "Volatile and non-volatile switching in Cu-SiO2 programmable metallization cells", IEEE Electron Device Letters, 37(5), pp.580-583, May 2016.
[24]C. L. Hsu, A. Saleem, and A. Singh et al., "Enhanced linearity in CBRAM synapse by post oxide deposition annealing for neuromorphic computing applications", IEEE Transactions on Electron Devices, 68(11), pp.5578-5584, Nov 2021.
[25]M. K. Kim, and J. S. Lee., "Synergistic improvement of long‐term plasticity in photonic synapses using ferroelectric polarization in hafnia‐based oxide‐semiconductor transistors", Advanced Materials, 32(12), p.1907826, Feb 2020.
[26]A. Ali, H. Abbas, and M. Hussain et al., "Thickness-dependent monochalcogenide GeSe-based CBRAM for memory and artificial electronic synapses", Nano Research, 15(3), pp.2263-2277, Sep 2021.