近年來，電阻式記憶體由於具備突觸特性而被受到廣泛的注目。然而，在眾多提出的元件中，找到真正合適的突觸元件仍然是一個很大的問題；這個突觸元件必需具有高密度堆疊、低功耗等特性，才能實現與大腦相比之類神經網路。
此論文致力於開發能實現高密度、低能耗的突觸元件。以 Ti/Pt/TiO2/Pt/HfO2/Ti/Al 堆疊結構電阻式記憶體，成功將氧化鉿基電阻式記憶體與氧化鈦基電阻式記憶體結合在同個金屬-絕緣層-金屬堆疊的高密度元件之中，並模仿突觸權重的增益效應與抑制效應，更透過寫入波型的最佳化，調整權重可塑性之非線性型，使兩種電阻式記憶體分別達到了0.08與0.15的優異非線性度，成功改善元件之非線性並提升仿生運算系統的訓練準確性，並使兩種記憶體達到最高為98% 以上的圖形辨識模擬準確率，最後在反覆讀寫能力上，最高來到了10^7以上的操作次數。相較於其他已發表之文章，本篇論文在密度、非線性度、模擬準確度與反覆讀寫能力上，都具有明顯的優勢，證明了該元件對於未來深度學習的硬體應用有相當的可行性。

Recently, resistive switching memory (RRAM) has attracted considerable attention because of its intriguing synaptic features. However, among the various RRAM devices currently available, finding a truly appropriate synaptic device is still challenging because high-density integration and low power consumption are required to construct a neural network comparable to that of the human brain.
This thesis thus focuses on the development of synapses that can achieve high integration density and low-power consumption. We successfully demonstrated Ti/Pt/TiO2/Pt/HfO2/Ti/Al RRAMs, which the HfO2-based RRAM and TiO2-based RRAM are combined into one MIMIM high density device, and mimic the basic characteristics of a synapse, potentiation and depression well. We also adjusted the nonlinearity of the weight update using input waveform optimization. Both of the RRAMs can achieve the extraordinary nonlinearity of 0.08 and 0.15,respectively. The improvement of the nonlinearity successfully increases the accuracy of the pattern recognition simulation up to 98%. For the endurance, over 10^7 switching cycles could be observed. Comparing with other published papers, the advantages of the density, nonlinearity, accuracy of pattern recognition and endurance is obvious, indicating the excellent feasibility for future deep learning hardware applications.

[1]P.-Y. Chen and S. Yu, “Compact modeling of RRAM devices and its applications in 1T1R and 1S1R array design,” IEEE Transactions on Electron Devices, vol. 62, no. 12, pp. 4022–4028, 2015.
[2]S. Yu, B. Lee, and H.-S. P. Wong, “Metal oxide resistive switching memory,” Functional Metal Oxide Nanostructures, Springer, New York, NY. pp. 303–335, 2011.
[3]H.-S. P. Wong, H.-Y. Lee, S. Yu, Y.-S. Chen, Y. Wu, P.-S. Chen, B. Lee, F. T. Chen, and M.-J. Tsai, “Metal–Oxide RRAM,” Proceedings of the IEEE, vol. 100, no. 6, pp. 1951–1970, 2012.
[4]A. Sawa, “Resistive switching in transition metal oxides,” Materials Today, vol. 11, no. 6, pp. 28–36, 2008.
[5]F.-C. Chiu, “A review on conduction mechanisms in dielectric films,” Advances in Materials Science and Engineering, vol. 2014, pp. 1–18, 2014.
[6]E. Lim and R. Ismail, “Conduction mechanism of valence change resistive switching memory: A survey,” Electronics, vol. 4, no. 3, pp. 586–613, 2015.
[7]Y. Q. Zhu, H. Qian, L. F. Wang, L. Wang, and J. Y. Tang, “Measurement and analysis of substrate leakage current of rf mems capacitive switches,” Microelectronics Reliability, vol. 54, no. 1, pp. 152–159, 2014.
[8]B. Majkusiak, P. Palestri, A. Schenk, A. S. Spinelli, C. M. Compagnoni, and M. Luisier, “Modeling and simulation approaches for gate current computation,” Nanoscale CMOS, ISTE, London, UK, pp. 213–257, 2013.
[9]J.-J. Huang, Y.-M. Tseng, W.-C. Luo, C.-W. Hsu, and T.-H. Hou, “One selector one resistor (1S1R) crossbar array for high-density flexible memory applications,” 2011 International Electron Devices Meeting, pp. 31.7.1-31.7.4, 2011.
[10]A. Flocke and T. G. Noll, “Fundamental analysis of resistive nano-crossbars for the use in hybrid nano/cmos-memory,” ESSCIRC 2007 - 33rd European Solid-State Circuits Conference, pp. 328-331, 2007.
[11]J. Liang and H.-S. P. Wong, “Cross-Point memory array without cell selectors—device characteristics and data storage pattern dependencies,” IEEE Transactions on Electron Devices, vol. 57, no. 10, pp. 2531–2538, 2010.
[12]B. Hudec, C.-W. Hsu, I.-T. Wang, W.-L. Lai, C.-C. Chang, T. Wang, K. Fröhlich, C.-H. Ho, C.-H. Lin, and T.-H. Hou, “3D resistive ram cell design for high-density storage class memory—a review,” Science China Information Sciences, vol. 59, no. 6, pp. 21, 2016.
[13]H.-Y. Chen, S. Brivio, C.-C. Chang, J. Frascaroli, T.-H. Hou, B. Hudec, M. Liu, H. Lv, G. Molas, J. Sohn, S. Spiga, V. M. Teja, E. Vianello, and H.-S. P. Wong, “Resistive random access memory (rram) technology: From material, device, selector, 3D integration to bottom-up fabrication,” Journal of Electroceramics, vol. 39, no. 1-4, pp. 21–38, 2017.
[14]Y. Deng, H.-Y. Chen, B. Gao, S. Yu, S.-C. Wu, L. Zhao, B. Chen, Z. Jiang, X. Liu, T.-H. Hou, Y. Nishi, J. Kang, and H.-S. P. Wong, “Design and optimization methodology for 3D RRAM arrays,” 2013 IEEE International Electron Devices Meeting, pp. 25.7.1-25.7.4, 2013.
[15]S. Yu, Y. Wu, R. Jeyasingh, D. Kuzum, and H.-S. P. Wong, “An electronic synapse device based on metal oxide resistive switching memory for neuromorphic computation,” IEEE Transactions on Electron Devices, vol. 58, no. 8, pp. 2729–2737, 2011.
[16]A. Schwartz, “Drives, affects, behavior-and learning: Approaches to a psychobiology of emotion and to an integration of psychoanalytic and neurobiology thought,” Journal of the American Psychoanalytic Association, vol. 35, no. 2, pp. 467–506, 1987.
[17]K. Moon, S. Lim, J. Park, C. Sung, S. Oh, J. Woo, J. Lee, and H. Hwang, “RRAM-based synapse devices for neuromorphic systems,” Faraday Discussions, vol. 213, pp. 421–451, 2019.
[18]D. Kuzum, S. Yu, and H.-S. Philip Wong, “Synaptic electronics: Materials, devices and applications,” Nanotechnology, vol. 24, no. 38, p. 382001, 2013.
[19]H. Markram, “A history of spike-timing-dependent plasticity,” Frontiers in Synaptic Neuroscience, vol. 3, pp. 4.2, 2011.
[20]G.-qiang Bi and M.-ming Poo, “Synaptic modifications in cultured hippocampal neurons: dependence on spike timing, synaptic strength, and postsynaptic cell type,” The Journal of Neuroscience, vol. 18, no. 24, pp. 10464–10472, 1998.
[21]S. H. Jo, T. Chang, I. Ebong, B. B. Bhadviya, P. Mazumder, and W. Lu, “Nanoscale memristor device as synapse in neuromorphic systems,” Nano Letters, vol. 10, no. 4, pp. 1297–1301, 2010.
[22]T. Chang, P. Sheridan, and W. Lu, “Modeling and implementation of oxide memristors for neuromorphic applications,” 2012 13th International Workshop on Cellular Nanoscale Networks and their Applications, 2012.
[23]S. Yu, B. Gao, Z. Fang, H. Yu, J. Kang, and H.-S. P. Wong, “A neuromorphic visual system using rram synaptic devices with sub-pj energy and tolerance to variability: Experimental characterization and large-scale modeling,” 2012 International Electron Devices Meeting, pp. 10.4.1-10.4.4, 2012.
[24]S. M. Sadaf, X. Liu, M. Son, S. Park, S. H. Choudhury, E. Cha, M. Siddik, J. Shin, and H. Hwang, “Highly uniform and reliable resistance switching properties in bilayer WOx /NbOx RRAM devices,” Physica Status Solidi(A), vol. 209, no. 6, pp. 1179–1183, 2012.
[25]J. Liu, H. Yang, Y. Ji, Z. Ma, K. Chen, X. Zhang, H. Zhang, Y. Sun, X. Huang, and S. Oda, “An electronic synaptic device based on HfO2 /TiOX bilayer structure memristor with self-compliance and deep-RESET characteristics,” Nanotechnology, vol. 29, no. 41, p. 415205, 2018.
[26]P.-Y. Chen, B. Lin, I.-T. Wang, T.-H. Hou, J. Ye, S. Vrudhula, J.-sun Seo, Y. Cao, and S. Yu, “Mitigating effects of non-ideal synaptic device characteristics for on-chip learning,” 2015 IEEE/ACM International Conference on Computer-Aided Design (ICCAD), pp. 194-199, 2015.
[27]D. Piccinotti, B. Gholipour, J. Yao, K. F. MacDonald, B. E. Hayden, and N. I. Zheludev, “Photonic metamaterials: Optical response of nanohole arrays filled with chalcogenide low-epsilon media,” Advanced Optical Materials, vol. 6, no. 22, p. 1870088, 2018.
[28]J. Gierak, “Focused ion beam nano-patterning from traditional applications to single ion implantation perspectives,” Nanofabrication, vol. 1, no. 1, pp. 35-52, 2014.
[29]P. Goodhew, “General introduction to transmission electron Microscopy,” Aberration-Corrected Analytical Transmission Electron Microscopy, pp. 1-19, 2011.
[30]G. Kaur, Biomedical, therapeutic and clinical applications of bioactive glasses. Oxford: Woodhead Publishing, an imprint of Elsevier, pp. 313-329, 2019.
[31]J. Burger and C. Teuscher, “Volatile memristive devices as short-term memory in a neuromorphic learning architecture,” 2014 IEEE/ACM International Symposium on Nanoscale Architectures, pp. 104-109, 2014.
[32]G. W. Burr, R. M. Shelby, C. di Nolfo, J. W. Jang, R. S. Shenoy, P. Narayanan, K. Virwani, E. U. Giacometti, B. Kurdi, and H. Hwang, “Experimental demonstration and tolerancing of a large-scale neural network (165,000 synapses), using phase-change memory as the synaptic weight element,” 2014 IEEE International Electron Devices Meeting, pp. 29.5.1-29.5.4, 2014.
[33]I.-T. Wang, C.-C. Chang, L.-W. Chiu, T. Chou, and T.-H. Hou, “3D Ta/TaOx/TiO2/Ti synaptic array and linearity tuning of weight update for hardware neural network applications,” Nanotechnology, vol. 27, no. 36, p. 365204, 2016.
[34]S. Chandrasekaran, F. M. Simanjuntak, R. Saminathan, D. Panda, and T.-Y. Tseng, “Improving linearity by introducing al in HfO2 as a memristor synapse device,” Nanotechnology, vol. 30, no. 44, p. 445205, 2019.
[35]S. Cho, J. Jung, S. Kim, and J. J. Pak, “Conduction mechanism and synaptic behaviour of interfacial switching AlOσ-based RRAM,” Semiconductor Science and Technology, vol. 35, no. 8, p. 085006, 2020.
[36]W. Wu, H. Wu, B. Gao, P. Yao, X. Zhang, X. Peng, S. Yu, and H. Qian, “A methodology to Improve linearity of analog RRAM for neuromorphic computing,” 2018 IEEE Symposium on VLSI Technology, pp. 103-104, 2018.