臺灣博碩士論文加值系統

在傳統燈絲型電阻轉換元件中，其電阻轉換行為通常具有突升式的寫入(SET)與 漸進式的抹除(RESET)，雖然可分別在SET與RESET過程中透過控制限制電流及RESET電壓達到不同的組態，但由於此類比式轉換的編程操作，會使用於類神經型 態運算的電路設計變得更複雜。
在本研究中的第一部份，吾人在主動層與上電極之間插入過度金屬氧化物製作出Ta/WOx/TaOx/Pt及Ta/ZrOx/TaOx/Pt元件，嘗試使電阻轉換元件中突升式的 SET 過程 轉變為漸進式並與Ta/TaOx/Pt做比較。根據它們的電子特性，發現 Ta/WOx/TaOx/Pt 元件具有漸進式的SET與RESET特性及較 小的操作電流，但 Ta/TaOx/Pt與Ta/ZrOx/TaOx/Pt 元件並沒有。在 XPS 分析中，當 WOx 疊在 TaOx 頂部時.由氧化還原反應所產生的氧空缺，其在靠近 WOx/TaOx 界面處的 TaOx 中會比 Ta/TaOx/Pt 元件中靠近 Ta/TaOx 界面處的 TaOx 還要少。吾人也透過調控 氧化鎢沈積時的氬氧流量比去了解氧化鎢是如何影響操作電流值，而其在不同流量比狀態下則展現了不同的操作電流。在阻抗圖譜中，其等效電路中的一個電阻於初始狀態時，在正偏壓下會與氧化鎢的氣體流量相關。根據這些結果，吾人可以推測Ta/WOx/TaOx/Pt元件的電阻轉換行為是由於氧空缺濃度及被捕捉電子數量在TaOx的改變所造成的，其導電機制與空間電荷限制相符。
在本研究中的第二部分，為了模擬神經型態運算中的生物突觸，吾人先透過重複 的電壓掃伏對有無氧化鎢的元件測試電壓及時間的相依性，而因為不具有氧化鎢的元 件需要達到一臨界電壓，故其呈現數位式、電壓相依性、時間不相依性的開關行為， 且其他的元件則呈現類比式、電壓及時間相依性的開關行為。然而，具有氧化鎢的元 件展現的類比式轉換行為是由於切換過程中較緩慢，故無法一次轉換完全。最後，吾 人在具有氧化鎢的元件上使用電壓脈衝量測以模擬生物系統。結果顯示我們的元件可 以成功地模仿生物突觸的行為，像是增益、抑制、雙重脈衝促進、經驗相關可塑性及在學習的行為並可應用於類神經運算。

In traditional filamentary resistive switching device, it usually had abrupt SET and gradual RESET RS behavior. Although it could achieve different resistive state by control compliance current and RESET voltage during SET and RESET process, respectively, it would make more complex for circuit design in neuromorphic computing because of programing operation for analog switching.
In the first part of this study, the transition metal oxides were inserted between TaOx active layer and Ta top electrode to fabricate Ta/WOx/TaOx/Pt and Ta/ZrOx/TaOx/Pt device in an attempt to let abrupt SET process of RS device change to gradually, and compare to the Ta/TaOx/Pt device. According to their electrical properties, we found out that the Ta/WOx/TaOx/Pt had gradual SET and RESET property and lower operation current, but the Ta/TaOx/Pt and Ta/ZrOx/TaOx/Pt devices did not. In X-ray photoelectron spectrum analysis, while WOx was stacked on top of TaOx, the oxygen vacancies formed by redox reaction at the TaOx near the WOx/TaOx interface in the Ta/WOx/TaOx/Pt device will less than the TaOx near the Ta/TaOx interface in the Ta/TaOx/Pt device. We also modulated Ar:O2 flow ratio of tungsten oxide during deposition in order to understand how the tungsten oxide affects the value of operation current, and it exhibited different operation current under different flow ratio condition. As shown in impedance spectrum, one of the resistances in the equivalent circuit under positive bias in pristine state was dependent on gas flow ratio of tungsten oxide. According to I-V curve fitting, we could speculate that the RS behavior in the Ta/WOx/TaOx/Pt device was caused by changing of oxygen vacancy concentration and the amount of trapped electron in the TaOx, because the conduction mechanism followed space charge limited conduction.
In the second part of this study, in order to simulate bio-synapse in neuromorphic computing, we test the voltage- and time-dependence of the devices with and without tungsten oxide through repeated voltage sweep first. The voltage was required to reach the threshold voltage for RS in the device without tungsten oxide, and it changed suddenly, which was shown as digital, voltage-dependence and time-independence switching behavior. However, the devices with tungsten oxide had analog, voltage- and time-dependence switching behavior, because it could not fully change in one time due to slower switching process. At last, we used voltage pulse measurement in the devices with tungsten oxide to mimic biological system. The results show our device could successfully mimic bio-synapse behavior for neuromorphic computing application, such as potentiation, depression, paired- pulse facilitation, experience dependent plasticity and relearning.

Abstract II
誌謝 IV
Contents VI
Figure Contents VIII
Table Contents XIII
Chapter 1 Introduction 1
1-1 Motivation 1
1-2 Characteristics and background of synapse 1
1-2-1 Hebbian theory 2
1-2-2 Synaptic behavior 2
1-3 Resistive switching mechanism 7
1-4 Brain-inspired computing 10
Chapter 2 Resistive switching characteristic 12
2-1 Fabrication of resistive switching device 12
2-2 Material analysis of the thin film 15
2-2-1 Transmission electron microscope 15
2-2-2 X-ray Photoelectron Spectroscope 18
2-3 Resistive Switching Characteristics 27
2-3-1 Ta/TaOx/Pt 27
2-3-2 Ta/WOx/TaOx/Pt 27
2-3-3 Ta/ZrOx/TaOx/Pt 28
2-4 Influence of tungsten oxide under different deposition gas flow ratio 30
2-5 Impedance analysis 32
2-5-1 Forming process 32
2-5-2 SET process 32
2-5-3 RESET process 33
2-6 Switching mechanism 37
Chapter 3 Analog resistive switching for synaptic device 41
3-1 Effect of different write and erase voltages on analog switching 41
3-1-1 Ta/TaOx/Pt 41
3-1-2 Ta/WOx/TaOx/Pt 41
3-1-3 Ta/WOx/WOy/TaOx/Pt 42
3-2 Effect of different write time on analog switching 50
3-2-1 Ta/WOx/TaOx/Pt 50
3-2-2 Ta/WOx/WOy/TaOx/Pt 50
3-3 Potentiation and depression behavior 56
3-3-1 Potentiation 56
3-3-2 Depression 56
3-4 Other Synaptic characteristics 61
3-4-1 Paired-pulsed facilitation (PPF) 61
3-4-2 Short- and long-term transition 61
3-4-3 Experience dependent plasticity (EDP) 62
3-4-4 Relearning process 63
Conclusions 69
References 70

[1]P. Chi, S. Li, C. Xu, T. Zhang, J. Zhao, Y. Liu, Y. Wang and Y. Xie, Prime: A novel processing-in-memory architecture for neural network computation in reram-based main memory, presented at the ACM SIGARCH Computer Architecture News, 2016.
[2]Q. T. Zhang, H. Q. Wu, P. Yao, W. Q. Zhang, B. Gao, N. Deng and H. Qian, Sign backpropagation: An on-chip learning algorithm for analog RRAM neuromorphic computing systems, Neural Netw. 108, 217-223 (2018).
[3]F. R. Meng, Y. Xue and C. M. Yang, Power- and Endurance-Aware Neural Network Training in NVM-Based Platforms, IEEE Trans. Comput-Aided Des. Integr. Circuits Syst. 37, 2709-2719 (2018).
[4]J. Woo, K. Moon, J. Song, M. Kwak, J. Park and H. Hwang, Optimized Programming Scheme Enabling Linear Potentiation in Filamentary HfO2 RRAM Synapse for Neuromorphic Systems, IEEE Trans. Electron Devices 63, 5064-5067 (2016).
[5]J. C. Liu, T. Y. Wu and T. H. Hou, Optimizing Incremental Step Pulse Programming for RRAM Through Device-Circuit Co-Design, IEEE Trans. Circuits Syst. II-Express Briefs 65, 617-621 (2018).
[6]C. Sung, S. Lim, H. Kim, T. Kim, K. Moon, J. Song, J. J. Kim and H. Hwang, Effect of conductance linearity and multi-level cell characteristics of TaOx-based synapse device on pattern recognition accuracy of neuromorphic system, Nanotechnology 29, 5 (2018).
[7]C. H. Wang, W. He, Y. Tong and R. Zhao, Investigation and Manipulation of Different Analog Behaviors of Memristor as Electronic Synapse for Neuromorphic Applications, Sci Rep 6, 9 (2016).
[8]J. Woo, K. Moon, J. Song, S. Lee, M. Kwak, J. Park and H. Hwang, Improved Synaptic Behavior Under Identical Pulses Using AlOx/HfO2 Bilayer RRAM Array for Neuromorphic Systems, IEEE Electron Device Lett. 37, 994-997 (2016).
[9]W. Wu, H. Q. Wu, B. Gao, N. Deng, S. M. Yu and H. Qian, Improving Analog Switching in HfOx-Based Resistive Memory With a Thermal Enhanced Layer, IEEE Electron Device Lett. 38, 1019-1022 (2017).
[10]D. O. Hebb, The organization of behavior, Annee Psychol. 51, 493-494 (1949).
[11]J. W. Jang, S. Park, G. W. Burr, H. Hwang and Y. H. Jeong, Optimization of Conductance Change in Pr1-xCaxMnO3-Based Synaptic Devices for Neuromorphic Systems, IEEE Electron Device Lett. 36, 457-459 (2015).
[12]R. S. Zucker and W. G. Regehr, Short-term synaptic plasticity, Annu. Rev. Physiol. 64, 355-405 (2002).
[13]D. Purves, R. Cabeza, S. A. Huettel, K. S. LaBar, M. L. Platt, M. G. Woldorff and E. M. Brannon, Cognitive neuroscience. (Sunderland: Sinauer Associates, Inc, 2008).
[14]G. Q. Bi and M. M. Poo, Synaptic modification by correlated activity: Hebb's postulate revisited, Annu. Rev. Neurosci. 24, 139-166 (2001).
[15]T. V. P. Bliss and G. L. Collingridge, A synaptic model of memory: long-term potentiation in the hippocampus, Nature 361, 31-39 (1993).
[16]S. F. Cooke and T. V. P. Bliss, Plasticity in the human central nervous system, Brain 129, 1659-1673 (2006).
[17]M. A. Alam, B. E. Weir and P. J. Silverman, A study of soft and hard breakdown - Part II: Principles of area, thickness, and voltage scaling, IEEE Trans. Electron Devices 49, 239-246 (2002).
[18]M. A. Alam, B. E. Weir and P. J. Silverman, A study of soft and hard breakdown - Part I: Analysis of statistical percolation conductance, IEEE Trans. Electron Devices 49, 232-238 (2002).
[19]J. W. McPherson and H. C. Mogul, Underlying physics of the thermochemical E model in describing low-field time-dependent dielectric breakdown in SiO2 thin films, J. Appl. Phys. 84, 1513-1523 (1998).
[20]J. J. Yang, M. D. Pickett, X. M. Li, D. A. A. Ohlberg, D. R. Stewart and R. S. Williams, Memristive switching mechanism for metal/oxide/metal nanodevices, Nat. Nanotechnol. 3, 429-433 (2008).
[21]P. C. Wang, P. G. Li, Y. S. Zhi, D. Y. Guo, A. Q. Pan, J. M. Zhan, H. Liu, J. Q. Shen and W. H. Tang, Bias tuning charge-releasing leading to negative differential resistance in amorphous gallium oxide/Nb: SrTiO3 heterostructure, Appl. Phys. Lett. 107, 4 (2015).
[22]K. P. Biju, X. Liu, E. Bourim, I. Kim, S. Jung, M. Siddik, J. Lee and H. Hwang, Asymmetric bipolar resistive switching in solution-processed Pt/TiO2/W devices, J. Phys. D-Appl. Phys. 43, 5 (2010).
[23]T. J. Chu, T. M. Tsai, T. C. Chang, K. C. Chang, C. H. Pan, K. H. Chen, J. H. Chen, H. L. Chen, H. C. Huang, C. C. Shih, Y. E. Syu, J. C. Zheng and S. M. Sze, Ultra-high resistive switching mechanism induced by oxygen ion accumulation on nitrogen-doped resistive random access memory, Appl. Phys. Lett. 105, 4 (2014).
[24]Z. Fan, H. Fan, L. Yang, P. L. Li, Z. X. Lu, G. Tian, Z. F. Huang, Z. W. Li, J. X. Yao, Q. Y. Luo, C. Chen, D. Y. Chen, Z. B. Yan, M. Zeng, X. B. Lu, X. S. Gao and J. M. Liu, Resistive switching induced by charge trapping/detrapping: a unified mechanism for colossal electroresistance in certain Nb:SrTiO3-based heterojunctions, J. Mater. Chem. C 5, 7317-7327 (2017).
[25]E. Mikheev, B. D. Hoskins, D. B. Strukov and S. Stemmer, Resistive switching and its suppression in Pt/Nb:SrTiO3 junctions, Nat. Commun. 5, 8 (2014).
[26]W. S. McCulloch and W. Pitts, A logical calculus of the ideas immanent in nervous activity, Bull. Math. Biol. 52, 99-115 (1990).
[27]A. L. Samuel, Some studies in machine learning using the game of checkers, IBM J. Res. Dev. 3, 211-＆ (1959).
[28]J. F. Moulder, W. F. Stickle, P. E. Sobol and K. D. Bomben, Handbook of X-ray photoelectron spectroscopy. (Minnesota, 1995).
[29]H. Demiryont, J. R. Sites and K. Geib, Effects of oxygen content on the optical properties of tantalum oxide films deposited by ion-beam sputtering, Appl. Optics 24, 490-495 (1985).
[30]R. Sohal, C. Walczyk, P. Zaumseil, D. Wolansky, A. Fox, B. Tillack, H. J. Mussig and T. Schroeder, Thermal oxidation of chemical vapour deposited tungsten layers on silicon substrates for embedded non-volatile memory application, Thin Solid Films 517, 4534-4539 (2009).
[31]K. Senthil and K. Yong, Growth and characterization of stoichiometric tungsten oxide nanorods by thermal evaporation and subsequent annealing, Nanotechnology 18, 7 (2007).