Nongseo-dong, Giheung-gu, Yongin-si, Gyeonggi-do, 445-701, Republic of Korea ..... cell was optimized, IRESET slightly decreased compared to the previous ...
Copyright © 2012 American Scientific Publishers All rights reserved Printed in the United States of America
Journal of Nanoscience and Nanotechnology Vol. 12, 5263–5269, 2012
Kyung-Chang Ryoo1 2 ∗ , Jeong-Hoon Oh1 2 , Sunghun Jung1 , Hongsik Jeong2 , and Byung-Gook Park1 1
Inter-University Semiconductor Research Center and School of Electrical Engineering and Computer Science, Seoul National University, San 56-1, Sillim-dong, Gwanak-ku, Seoul 151-742, Republic of Korea 2 DRAM Process Architecture Team, Memory Division, Semiconductor Business, Samsung Electronics Co., Ltd., Nongseo-dong, Giheung-gu, Yongin-si, Gyeonggi-do, 445-701, Republic of Korea An interface-engineered resistive random access memory (RRAM) using bilayer transition metal oxide (TMO) is presented for improving unipolar resistive-switching characteristics. The experiment and simulation data show that better resistive switching characteristics and superb uniformity can be realized by inserting a thin AlOx insertion layer between the Ir/NiO interface. To elucidate the uniformity improvement of our bilayer structure, the conducting-defect effects in the resistive cell were also investigated using a random circuit breaker (RCB) simulation model. It has been verified that the forming and set characteristics are more effectively improved because the conductingdefect ratio in the insertion layer region is low, therefore making it more advantageous for a filament path controllability. Using the optimal oxygen contents in both the insertion layer and the resistive cell, it was confirmed that a significant reduction of up to 0.15 mA of the reset current (IRESET is Delivered by Publishing Technology to: Yonsei University possible compared to the conventional cell. These results indicate that new Al insertion has a large IP: 1.233.219.35 On: Thu, 07 May 2015 06:18:50 contribution to the reset and forming processes.
Copyright: American Scientific Publishers Keywords: RRAM, Unipolar, Resistive Switching, Reset Current, Forming Voltage, Low Power, Al Inserted.
1. INTRODUCTION In recent years, various studies on novel material-based new switching memories such as phase change RAM (PRAM), resistive RAM (RRAM), and magnetic RAM (MRAM) have been conducted as next-generation nonvolatile memories due to their numerous advantages, such as low writing power, fast writing time, excellent reliability, and possibility of outstanding scalability.1–10 Among these, RRAM, a new memory using resistance change, is considered extremely important as a next-generation memory. Despite its advantages of a simple structure, low cost, and CMOS process compatibility for mass production,11–13 however, it still has unsolved crucial issues, such as insufficient understanding of the switching mechanism and a high switching current level. Furthermore, it is very difficult to accomplish high-density RRAM due to the limitation of current photolithography issues.14–20 To address the problems of reliability and set voltage (VSET distribution issues, further research on the cell interface where resistive switching occurs is needed.8–9 Moreover, it is very difficult to reduce the forming voltage (VFORMING due to its ∗
Author to whom correspondence should be addressed.
J. Nanosci. Nanotechnol. 2012, Vol. 12, No. 7
difficult control of conducting-filament (CF) configuration in the unipolar resistive cell.6 In this paper, the improved switching behavior in Alinserted NiO-based unipolar RRAM was investigated. The relationships between the analytical parameters, such as the forming, set, and reset characteristics, which are very important in understanding the resistive-switching behavior of unipolar RRAM, were also examined in a cell using a random circuit breaker (RCB) simulation model.4 The effects of AlOx insertion at the Ir/NiO interface and the optimal oxygen contents are also discussed herein, which were determined using the fabricated AlOx /NiO bilayer unipolar RRAM cell structure. The motivation of this study is presented in Section 2. The results of the simulation study of the conducting defect of the bilayer structure and the resistive-switching characteristics of the AlOx /NiO bilayer cell structure are presented in Section 3. The conclusion is discussed in Section 4.
2. MOTIVATION Figure 1 shows the typical I–V and R–V curves based on the fabricated NiO unipolar RRAM structure.8 IRESET is
1533-4880/2012/12/5263/007
doi:10.1166/jnn.2012.6234
5263
RESEARCH ARTICLE
Interface-Modified Unipolar Resistive Random Access Memory (RRAM) Structure for Low-Power Application
Ryoo et al.
RESEARCH ARTICLE
Interface-Modified Unipolar Resistive Random Access Memory (RRAM) Structure for Low-Power Application
Fig. 1. Typical I–V and R–V curves of the fabricated unipolar RRAM with a planar MIM structure, which showed forming, reset, and set states.
Fig. 2. Reset current as a function of the forming voltage for a singlelayer cell structure with various thicknesses, for which a random circuit breaker (RCB) simulation model was used.
−1 mA at reset voltage (VRESET = 06 V, and VSET = 12 V. IRESET and VFORMING in a single-layer cell structure, as The resistance values of the reset and set states are shown in Figure 2. Figure 3 shows the important resistive104 –105 and hundreds of ohms, respectively, so that switching characteristics of a single-layer cell structure for the reset/set resistance ratio is larger than 102 . In the different conducting defect ratios. Conducting defect can case of unipolar switching, there is no need to change be defined as charged particles such as oxygen vacancy the polarity. The unit cell size in the unipolar RRAM and metallic particles, which are sources of CF in the cell can be reduced because unipolar-switching RRAM unipolar RRAM cell. As can be seen in Figure 3, it is can implement a vertical-diode-type rectifying element, which has a smaller area than the transistor type. Due 8–9 Delivered by Publishing Technology to: Yonsei University to its structural advantages, the unipolar-type cell with 1 1.233.219.35 On: to Thu, 07 May 2015 06:18:50 a filamentary conduction pathIP:model was adopted Copyright:I American Scientific Publishers accomplish high-density RRAM application. RESET as a function of VFORMING for different cell sizes was investigated using an RCB simulation model, a dynamic percolation simulation model for unipolar resistive-switching research.4 This simulation model for unipolar switching was proposed by Chae et al. based on a random circuit breaker network. This model easily explains the resistive-switching behavior of unipolar RRAM with a relatively simple mechanism.4 Although the IRESET reduction mechanism has been reported by many researchers, it is still controversial.3–10 Previous researches have shown that there is no relationship between IRESET reduction and the resistive cell area. This result comes from the relatively large cell size of over-um level, which causes a smaller CF area than the total, resulting in an insignificant change in the set resistance. If the resistive cell area, however, becomes smaller than the sub-nm level, the area of the CF change becomes relatively larger than the total cell area, making it possible for the difference between the cell area and the set resistance to be large. Particularly in the RCB simulation model, the change in the CF area becomes relatively larger than the change in the total cell area because the simulation is extremely small, with 50 × 20 nm dimensions. Therefore, the effect of the CF change Fig. 3. Important resistive-switching parameters such as reset curcan be observed in the sub-nm cell size. As shown in rent (a), set voltage (b), standard deviation of set voltage (c), and forming Figure 2, the change in IRESET for different cell contact voltage (d) of the single-layer cell structure for different conductingdefect ratios. areas can be identified, but there is a trade-off between 5264
J. Nanosci. Nanotechnol. 12, 5263–5269, 2012
Ryoo et al.
Interface-Modified Unipolar Resistive Random Access Memory (RRAM) Structure for Low-Power Application (a)
3. RESULTS AND DISCUSSION 3.1. Simulation Study of the Conducting-Defect Effect in a Bilayer Structure The conducting-defect effects were investigated for a bilayer cell structure, using an RCB simulation model. Three different bilayer cell structures were prepared, as (b) shown in Figure 4. The first was a reference cell with a uniform conducting-defect ratio, the second a split 1 cell with a low conducting-defect ratio (interface) and a high conducting-defect ratio (bulk), and the third a split 2 cell with high and low conducting-defect ratios (interface and bulk, respectively). Figure 5 shows the simulated I–V characteristics of the (a) reference cell; (b) low-interface cell; and (c) high-interface cell. The bilayer cells show better resistive-switching characteristics than the reference cell, but in the case of the split 2 cell with high and low conducting-defect ratios (interface and bulk, respectively), it shows a worse VSET distribution compared to the other split cells in this I–V curve. To manifest this effect, staDelivered by Publishing Technology (c) to: Yonsei University tistical analyses were performed. IP: 1.233.219.35 On: Thu, 07 May 2015 06:18:50 Copyright: Figure 6 shows the statistical analysis of the (a)American forming Scientific Publishers and (b) set characteristics of various cell structures. The split bilayer cells show better VFORMING and VSET characteristics. Especially, the split 1 cell shows excellent VFORMING and VSET characteristics, as well as the best distribution characteristics compared to those of the other different split cells. When using a bilayer cell, CF is effectively formed at the switching interface. Especially, it has been verified that the forming and set characteristics are more effectively improved than in the case of the split 1 cell because the conducting-defect ratio in the bulk region, which forms the initial filament, is high, therefore making Fig. 5. Simulated I–V curves of the control cell (a), low-interface cell it more advantageous for a filament path to be formed. (b), and high-interface cell (c). The split 2 cell also shows better switching characteristics because its overall conducting-defect ratio is higher than that of the reference cell, but it contributes relatively less to forming the initial CF because it is limited to the switching interface. Also in the case of the split 1 cell, the relatively less conducting defect at the interface makes CF rupturing easier, making it possible to predict better set charactertistics as well. Thus, determining the optimal conducting-defect ratio is very important for improving the resistive-switching characteristics. Figure 7 shows the VFORMING and IRESET characteristics as functions of various split groups. Compared to the single-layer cell, both bilayer cells showed better Fig. 4. Various bilayer cell structures for understanding the effect of the conducting defects in the resistive cell. VFORMING and IRESET characteristics. Surprisingly, both split J. Nanosci. Nanotechnol. 12, 5263–5269, 2012
5265
RESEARCH ARTICLE
very difficult to reduce both IRESET and VFORMING due to the different needs of the conducting-defect condition in the single-layer cell structure. Therefore, more reliable modifications are necessary to improve both the reset and initial forming characteristics.
Interface-Modified Unipolar Resistive Random Access Memory (RRAM) Structure for Low-Power Application
Ryoo et al.
3.2. Resistive-Switching Characteristics of the AlOx /NiO Bilayer Cell Structure
RESEARCH ARTICLE
(a)
(b)
Figure 8 summarizes IRESET and VFORMING as functions of different split groups. As can be seen in the figure, the split bilayer cells showed better resistive-switching characteristics compared to the reference cell. The improvements of both the IRESET and forming characteristics, however, are slightly different from the conducting-defect conditions in each bilayer cell. Thus, the optimal process condition with a deep understanding of the cell configuration is also needed to achieve better resistive-switching characteristics for commercial memory applications. To clarify the effect of the bilayer cell structure, an Ir/AlOx /NiO/Ir bilayer RRAM cell was fabricated. Figure 9 illustrates the proposed AlOx -layer-inserted bilayer RRAM cell structure. This bilayer cell was built between two intersecting metal lines, for which a resistive cell material and reading/writing data were selected alternatively. Using an additional AlOx layer at the switching interface, joule-heating effect enhancement occurred with a power of I 2 RT during the reset state, leading to uniform resistive switching and IRESET reduction. This interface-engineered structure resulted in an increased number of oxygen vacancies at the cell interface, leading to metallic Ni movements and making the thermally activated diffusion process easier. Moreover, due to the lower electronegativity of Al compared
Delivered by Publishing Technology to: Yonsei University IP: 1.233.219.35 On: Thu, 07 May 2015 06:18:50 Copyright: American Scientific Publishers Statistical analysis of the forming (a) and set (b) characteristics
Fig. 6. of the bilayer cell structure.
bilayer cells showed better VFORMING and IRESET characteristics, as opposed to the reference cell. This means that the optimal process condition plays an important role in resistive switching, and a bilayer with a different conducting-defect ratio can be the solution for resistiveswitching improvement.
(a) Fig. 8. Change in the reset current and forming voltage with different split groups.
(b)
Fig. 7. Forming voltage (a) and reset current (b) characteristics as a function of various split groups.
5266
Fig. 9. Illustration of the proposed Al-layer-inserted bilayer cell structure.
J. Nanosci. Nanotechnol. 12, 5263–5269, 2012
Ryoo et al.
Interface-Modified Unipolar Resistive Random Access Memory (RRAM) Structure for Low-Power Application
Fig. 11.
Illustration of the resistive-switching mechanism without an AlOx (a) and with an AlOx (b) layer at the Ir/NiO interface.
J. Nanosci. Nanotechnol. 12, 5263–5269, 2012
5267
RESEARCH ARTICLE
The formation of metallic Ni and oxygen vacancies as sources of CF was generated, which resulted in easy switching. Further, because of the Al insertion, the atomicdiffusion rate and the thermally activated diffusion process were enhanced during the reset process. The Al atoms prefer oxidation due to a lower electronegativity and higher binding energy with oxygen during the reset process. Such formed AlOx assists localized switching, as shown in Figure 11, improving the irregular switching behavior and reducing IRESET . The total area of the conducting filament (ACF associated with the reset state can also be reduced by the formation of AlOx associated with the CF localization at the NiO/TE interface, resulting in IRESET reduction. Then the optimal AlOx thickness and the optimal initial cell condition are crucial for low IRESET . As shown in Figure 10, I RESET decreased when optimal cell oxidation and the optiFig. 10. Typical I–V curves for the reset process in the reference (NiO mal inserted Al layer were applied. In Figure 10, IRESET single layer) and bilayer (AlOx /NiO bilayer) structures. efficiently decreased from 1.6 mA at VRESET ≡ 08 V to 0.15 mA at VRESET ≡ 06 V by more than one order comto that of Ni, it was expected that filament disconnection pared to that of the reference cell. Resistive switching was would become easier due to the interfacial AlOx formaclearly observed with a resistance change of 1.5 orders of tion at the reset state. Figure 10 shows the typical I–V magnitude. The set current slightly decreased compared to curves for the reset process in the NiO (single-layer) and the previous reference cell. The VSET difference was about AlOx /NiO (bilayer) devices. 0.3 V, but in the case of the oxygen content optimizaWhen an additional AlOx layer is inserted in the Ir/NiO tion, the difference is miniscule. In other words, the initial interface, IRESET decreases compared to the single-layer resistance can be controlled through oxygen content opticell, showing that the AlOx layer is effective for IRESET mization Al insertion, and as a result, good switching Delivered by Publishing Technology to:and Yonsei University reduction, acting as a buffer layer. Below is the mechabehavior can be obtained. Figure 12 shows the I–V curves IP: 1.233.219.35 On: Thu, 07 May 2015 06:18:50 nism in which the inserted Al layerCopyright: contributes American to IRESET Scientific Publishers of the forming (a) and reset (b) processes in the reference reduction. cell and oxygen-content-optimized cell, respectively. It is considered that the Al insertion into the NiO cell Lee et al. reported that the “hard-breakdown-like” pheleads to the formation of defects at the interface, creating nomenon happens if oxygen atoms will be impeded by the oxygen vacancies formed by the Al atom.3 enhanced oxidation layer, which acts as a barrier of ionic movement.17 It has an adverse impact on the resistiveAl − + switching process, which means that when the insertion NiO −−→ NiAl + Ooxygen + Voxygen
Interface-Modified Unipolar Resistive Random Access Memory (RRAM) Structure for Low-Power Application
Ryoo et al.
4. CONCLUSION
(a)
RESEARCH ARTICLE
To implement a low-switching-current RRAM, the relationships between the conducting defect and various related analytical parameters were investigated. The experiment and simulation data obtained showed that better resistive switching can be realized by a bilayer cell structure, which is inserted as an AlOx buffer layer between the Ir/NiO interfaces. A sufficient, thermally activated diffusion process enhanced by joule heating at the reset state, an oxygen vacancy, and AlOx formation due to the lower electronegativity of Al sufficiently enabled the IRESET to reach a value below 0.15 mA. The optimal formation of an AlOx interface with the optimal oxygen content in the resistive cell led to uniform forming with a low IRESET . Through its adoption in a cross-point structure, RRAM can have the highest feasibility in the high-density emerging non-volatile memory (NVM).
(b)
Acknowledgment: This work was supported by BK21 Program, Inter-University Semiconductor Research Center (ISRC), and Nano-Systems Institute (NSI-NCRC) Program, sponsored by Korea Science and Engineering Foundation (KOSEF).
References andUniversity Notes Delivered by Publishing Technology to: Yonsei IP: 1.233.219.35 On: Thu, 07 May 2015 06:18:50 1. J. F. Gibbons and W. E. Beadle, Solid-State Electron. 7, 785 (1964). Copyright: American Scientific Publishers 2. A. Sawa, T. Fujii, M. Kawasaki, and Y. Tokura, Appl. Phys. Lett. 85, Fig. 12. I–V curves for the forming (a) and reset (b) process of the reference cell and the oxygen-content-optimized cell.
material is used for improving the resistive-switching characteristics, finding an optimal process condition is crucial to avoid these effects. Especially, the determination of the oxygen contents of the resistive cell and the insertion layer have important roles in resistive switching. The oxygen content’s effect on the resistive cell structure was investigated. When the oxygen content of the cell was optimized, IRESET slightly decreased compared to the previous reference cell, but VFORMING drastically decreased from 2.5 to 1.7 V, as shown in Figure 12(a), which means that the optimal oxygen content allows both the effective reset and forming processes. Even though the optimal oxygen content can aid the forming process, the oxygen content difference is not the sole solution of IRESET improvement. Moreover, if the AlOx is sufficiently thick, the IRESET characteristics will not be improved. The Al thickness for the experiment in this study was about 8 Å. If the Al thickness is over the nm level, IRESET will not be reduced. Thus, combining a bilayer structure with the optimal oxygen content and the AlOx thickness are needed to improve both the forming and reset characteristics. 5268
4073 (2004). 3. S. Seo, M. J. Lee, D. H. Seo, E. J. Jeoung, D.-S. Suh, Y. S. Joung, I. K. Yoo, I. R. Hwang, S. H. Kim, I. S. Byun, J.-S. Kim, J. S. Choi, and B. H. Park, Appl. Phys. Lett. 85, 5655 (2004). 4. S. C. Chae, J. S. Lee, S. Kim, S. B. Lee, S. H. Chang, C. Liu, B. Kahng, H. Shin, D.-W. Kim, C. U. Jung, S. Seo, M.-J. Lee, and T. W. Noh, Adv. Mater. 20, 1154 (2008). 5. U. Russo, D. Ielmini, C. Cagli, and A. L. Lacaita, Electron Devices, IEEE Transactions 56, 186 (2009). 6. C. H. Cheng, A. Chin, and F. S. Yeh, Electron Device Letters, IEEE, 31, 1020 (2010). 7. J.-B. Yun, S. Kim, S. Seo, M.-J. Lee, D.-C. Kim, S.-E. Ahn, Y. Park, J. Kim, and H. Shin, Phys. Stat. Sol. 1, 280 (2007). 8. K.-C. Ryoo, J.-H. Oh, S. H. Jung, and B.-G. Park, IEEE Nanotechnology Materials and Devices Conference (NMDC) 12, 356 (2010). 9. M.-J. Lee, Y. Park, B.-S. Kang, S.-E. Ahn, C. Lee, K. Kim, W. Xianyu, G. Stefanovich, J.-H. Lee, S.-J. Chung, Y.-H. Kim, C.-S. Lee, J.-B. Park, I.-G. Baek, and I.-K. Yoo, Int. Electron Devices Meeting (IEDM) Tech. Dig. 771 (2007). 10. K. M. Kim, D. S. Jeong, and C. S. Hwang, Nanotechnology 22, 254002 (2011). 11. K. M. Kim and C. S. Hwang, Appl. Phys. Lett. 94, 122109 (2009). 12. R. Waser, R. Dittmann, G. Staikov, and K. Szot, Adv. Mater. 21, 2632 (2009). 13. D. C. Kim, M.-J. Lee, S. E. Ahn, S. Seo, J. C. Park, I. K. Yoo, I. G. Baek, H. J. Kim, E. K. Yim, J. E. Lee, S. O. Park, H. S. Kim, U.-I. Chung, J. T. Moon, and B. I. Ryu, Appl. Phys. Lett. 88, 232106 (2006). 14. I. H. Inoue, S. Yasuda, H. Akinaga, and H. Takagi, Phys. Rev. B 77, 035105 (2008).
J. Nanosci. Nanotechnol. 12, 5263–5269, 2012
Ryoo et al.
Interface-Modified Unipolar Resistive Random Access Memory (RRAM) Structure for Low-Power Application 18. K. Fujiwara, T. Nemoto, M. J. Rozenberg, Y. Nakamura, and H. Takagi, Japanese Journal of Applied Physics 47, 6266 (2008). 19. H. D. Lee and Y. Nishi, Appl. Phys. Lett. 97, 252107 (2010). 20. W. Guan, S. Long, Y. Hu, Q. Liu, Q. Wang, and M. Liu, J. Nanosci. Nanotechnol. 9, 723 (2009).
Received: 1 August 2011. Accepted: 23 December 2011.
Delivered by Publishing Technology to: Yonsei University IP: 1.233.219.35 On: Thu, 07 May 2015 06:18:50 Copyright: American Scientific Publishers
J. Nanosci. Nanotechnol. 12, 5263–5269, 2012
5269
RESEARCH ARTICLE
15. L. Nodo, L. Yves, and J. Yun Hee, J. Nanosci. Nanotechnol. 11, 339 (2011). 16. Y. Sato, K. Kinoshita, M. Aoki, and Y. Sugiyama, Appl. Phys. Lett, 90, 242906 (2007). 17. B. H. Lee, H. J. Hwang, C. H. Cho, S. K. Lim, S. Y. Lee, and H. Hwang, J. Nanosci. Nanotechnol. 11, 256 (2011).