A nondestructive approach to study resistive switching mechanism in metal oxide based on defect photoluminescence mapping

Xiaohu Wang a, Bin Gao *a, Huaqiang Wu *a, Xinyi Li a, Deshun Hong b, Yuansha Chen b and He Qian a
aInstitute of Microelectronics, Tsinghua University, Beijing, China. E-mail: gaob1@tsinghua.edu.cn; wuhq@tsinghua.edu.cn
bState Key Laboratory of Magnetism, Institute of Physics, Chinese Academy of Sciences, Beijing, China

Received 22nd March 2017 , Accepted 9th June 2017

First published on 14th June 2017


The mechanism of resistive switching in metal oxides is a widely studied topic with interest in both fundamental physics and the practical need to improve device characteristics for memory based applications. Various experimental approaches were employed to reveal the different aspects of resistive switching; however, there is still a debate on the switching mechanism due to the lack of nondestructive microscopic characterization tools to monitor the oxygen vacancies. In this study, a novel approach using photoluminescence (PL) mapping was developed to study switching dynamics in metal oxides. By monitoring the emission properties with a confocal PL system, information regarding the switching mechanism can be obtained. The nondestructive nature of this approach allowed us to make comparisons between different switching conditions and endurance cycles. SrTiO3 based switching devices were used in the study. The distribution of oxygen vacancies can be positioned by mapping the integrated intensity of oxygen vacancy emission on a transparent top electrode, and both interface switching and filament switching can be distinguished. Moreover, the endurance study revealed a sudden rise in the emission intensity correlated with the device failure, which indicates an abrupt increase in the localized density of oxygen vacancies that results in an irreversible set process for the conductive filament.


Introduction

Non-volatile resistive switching (RS) in oxide-based metal–insulator–metal structures is an important research frontier owing to its application in emerging memory and huge potential in neuromorphic computing, and hardware security, etc.1–5 Tremendous research efforts have been placed on revealing the microscopic nature of RS, which can be roughly categorized as interface type, often associated with a Schottky contact,6–10 and filamentary type, which is based on the formation and rupture of conductive filaments (CF) consisting of oxygen vacancies.1,2,11–13 The mechanism of interface switching is quite complicated, and various experimental efforts have been made to understand the role of junction parameters and defect states when switching occurs. On the other hand, a filamentary switching device based on metal oxide has proven its potential in the demonstration of large resistive random access memory (RRAM) array,14 and thus understanding the detailed switching mechanism is vital for enhancing device performance and achieving real application. However, unlike the triumph in directly observing dynamics of CF that consist of heavy metal ions in conductive bridge random access memory (CBRAM) by SEM and TEM,15–17 the most challenging task with respect to oxide based RRAM devices is the lack of direct characterization techniques to study oxygen vacancy defects that constitute CF. To overcome this obstacle, various highly localized characterization approaches have been introduced over the years, which have substantially enriched our knowledge in oxide based RRAM. With nanoscale precision, these techniques provided insights in understanding the physical properties of CF, such as evidence of the filamentary nature using conductive-tip atomic force microscopy (CAFM),11 X-ray absorption spectroscopy (XAS) imaging of the oxygen vacancies diffusion during forming18 and X-ray photoemission electron microscopy (XPEEM), along with XAS revealing the cationic motion during switching.19,20 However, an important drawback in the above techniques is the requirement of the delamination of the top electrode before characterization. This process destroys the device totally, and hence the results for the investigations on issues related to reliability are difficult to achieve. To solve this issue, operando spectroscopy based on hard X-ray21 or using electron-transparent graphene electrodes22 was successfully demonstrated. However, these experiments often require a high energy X-ray beam from a synchrotron light source, which is sometimes difficult to access. Hence it is highly necessary to develop new characterization methods to study different aspects of RS.

In this study, a novel characterization technique based on photoluminescence (PL) mapping of oxygen vacancy emission was developed to study the switching phenomenon. Oxygen vacancy is the most common type of defect in metal oxides. It introduces a defect level inside the bandgap and often leads to defect emission when excited by laser radiation. Studies on oxygen vacancy emission were reported and investigated for a wide range of materials in which tantalum pentoxide (Ta2O5),23 hafnium oxide (HfO2),24 and strontium titanate (STO)25 are among the most important prototype RS materials. As illustrated in Fig. 1, using a high resolution confocal PL system, a mapping scan for oxygen vacancy emission can be performed on the RRAM device with a transparent top electrode (TE). Since CF in oxide RRAM can be regarded as a localized oxygen vacancy column, its location can be well positioned by its localized emission. Due to diffraction limitation, the resolution of confocal PL mapping is not comparable to that of AFM or XPEEM, etc. However, it has demonstrated great competence in characterizing single nanowire and 2D materials with atomic thickness.26–28 Moreover, being a nondestructive characterization technique, the PL spectrum at the CF site can be analyzed, which would facilitate comparison between different resistance states, retention duration and endurance cycles, and thus reveal information on switching and failure mechanisms in oxide RRAM.


image file: c7nr02023f-f1.tif
Fig. 1 Schematic of the PL mapping approach to investigate oxide resistive switching device.

Herein, we use STO as a model material to demonstrate the capability of the PL mapping approach.

As a prototype material for RS, STO is an ideal platform to investigate the switching physics owing to its high crystal quality and diversity in switching effects, of which both interface and filamentary switching have been extensively reported. Moreover, STO is also known for its strong blue luminescence at room temperature, which is related to oxygen vacancy emission.25 The possibility of optoelectronic applications of STO leads to a significant research interest in revealing its emission mechanism.25,29–32 Moreover, the ordering of oxygen vacancies in STO appears to be highly localized as atomic vacancy columns when examined by HRTEM.33 Therefore, based on the strong PL and localized nature of oxygen vacancy in STO, we studied the emission spectroscopy of oxygen vacancies inside the STO RS device with a transparent electrode on top. An emission intensity mapping regarding the planar distribution of oxygen vacancy is demonstrated with the STO device possessing different switching mechanisms. Furthermore, we exploited PL spectroscopy to monitor the intensity variance in the filament region with respect to the cycling endurance, and we observed an abrupt increase in emission intensity after endurance failure.

Results and discussion

Two types of STO based switching devices were fabricated with different oxygen vacancy contents, while chemical analysis of STO was performed by X-ray photoemission spectroscopy (see ‘Experimental’ section for details). The compositions of the STO thin films were obtained by comparing the ratio of the total intensities of Sr 3d, Ti 2p and titanate oxygen (Ti–O) peaks after subtraction of the background, based on the Shirley method. Due to the existence of surface contaminants such as hydroxyl and carbonate species,34 the measured O 1s spectra decomposed into multiple peaks as shown in ESI Fig. S. 2. The peak around 259.6 eV is attributed to the titanate oxygen (Ti–O)35 and during the chemical composition analysis, only the area for this peak is considered. After calculation, the compositions of the STO films under high (sample S1) and low oxygen growth pressure (sample S2) were found to be SrTi0.75O2.78 and SrTi0.66O1.92, respectively. Both films showed Sr enrichment, which was also observed for flame-grown single crystal STO.36 STO can be regarded as an intertwined sub-lattice of SrO and TiO2; thus, for high oxygen pressure growth STO film, the oxygen content is slightly greater than the stoichiometric composition (SrTi0.75O2.5), while the other sample shows oxygen deficiency with respect to the stoichiometric condition (SrTi0.66O2.32). Therefore the XPS results indicate a substantial difference in oxygen concentration for the STO thin films, leading to the diverse switching behavior, which is demonstrated in IV characteristics shown in Fig. 2(a) and (b). The S1 device with low oxygen vacancy concentration demonstrated a standard filamentary switching curve with abrupt changes in the IV characteristics, which represents the rupture and formation of the conductive filament. The ITO layer served as both a top electrode and an oxygen reservoir since the ITO layer exhibited high concentration of oxygen vacancy, while the increase in oxygen vacancy yielded a consequential increase in conductivity.37The forming voltage of S1 was 3 V and large on/off ratios under both positive and negative bias were observed. The inset of Fig. 2(a) represents a schematic drawing of the conductive filament formed with highly localized oxygen vacancies.33 Moreover, the current at LRS is in the order of 100 μA for S1, and according to our previous study,38 the relatively high LRS resistance is tentatively attributed to electron hopping conduction. On the other hand, device S2 with high oxygen vacancy content yielded distinctively different IV characteristics compared with S1. Its IV curve reveals clear rectifying characteristics, indicating the formation of Schottky contact at the LSMO and STO interface. Moreover, the hysteretic IV without forming process indicates an interface switching mechanism. The IV curve under low forward bias can be fitted by the following equation based on the thermionic emission model,9
 
I = SA*T2[thin space (1/6-em)]exp(−ϕ/kBT)[exp(−qV/nkBT) − 1](1)
where S is the contact area, A* is the effective Richardson constant, T is the absolute temperature, φ is the Schottky barrier height, kB is the Boltzmann constant, q is the electric charge and n is the ideality factor characterizing the degree from ideal thermionic emission. The two dashed lines in Fig. 2(b) are the fitting results in the voltage range between 0 to −1 V with φ and n as fitting parameters. The value for φ is 0.69 eV, which is comparable to the electron affinity of undoped STO and LMO of 3.9 eV and 4.4 eV, respectively.39,40 The ideality factor, n, for the low resistance state (LRS) is larger than that of the high resistance state (HRS), which is also observed for Pt/Nb:STO junctions.8 For the ideal Schottky junction, the forward bias is controlled by thermionic emission, in which case the ideality factor n is equal to 1. The deviation of n in our fitting from the ideal value indicates the existence of other conduction mechanism as illustrated in the inset of Fig. 2(b). Device S2 exhibited a high concentration of defect states acting as electron traps, which facilitated trap-assisted tunneling. We also observed the filamentary type of IV characteristics in device S2 when the sweeping voltage was extended to 5 V (ESI Fig. S. 2). In Fig. 2(c), filamentary switching IV is shown under positive bias, while the current at LRS is still 2 orders of magnitude lower than that of S1. Furthermore, the IV characteristics under negative bias are retained for S2, demonstrating a rectifying behavior. Thus, after high voltage forming, oxygen vacancies were partially connected and weak filaments were formed, while under negative bias, these weak connections were broken (inset of Fig. 2(c)). This transition of S2 is comparable to the recently reported coexistence switching in STO, where Muenstermann reported the change in switching mechanism from a strong filament to weak filaments distributed beneath the electrode,41 and Kubicek proposed that additional Schottky barrier affected the filamentary switching under a stronger electric field.42

image file: c7nr02023f-f2.tif
Fig. 2 (a) and (b), IV sweeping of STO devices, S1 and S2, with different oxygen growth pressures. (c) IV curve of S2 device after the forming process. The insets within the figures illustrate the schematic drawings of switching and interface physics models where the small open circles represent oxygen vacancies inside the STO thin film.

We also studied the switching response time under different pulse widths, and found that the response time for S1 was in the range of microseconds, while that of S2 extended to milliseconds. It is interesting to note that the high speed metal oxide RRAM demonstrated thus far are mainly based on filamentary mechanism and often listed as a merit compared to flash memory in switching speed.43–45 However, data regarding the pulse switch duration in interface switching is sometimes missed out in reports, while a report based on Ta2O5 interface switching presented a pulse width in order of 10 ms.46 Therefore, it is anticipated that the response time for interface switching should be longer than that for the formation and rupture of the filaments in filamentary switching, since the rupture of filaments requires ions to be drifted a few angstroms locally, which is intuitively faster than the interface area charging and discharge under external field. However, switching systems based on STO showed contradictory results, which added complexity to this issue. For example, in a report on interface switching in Ag/SrTiO3:Nb junction, a response time as short as 5 ns was measured;47 however, this fast response is somehow attributed to the existence of conductive filaments that tunnel through the depletion layer. Another study based on the STO system demonstrated a 10 ms switching pulse for the filamentary mechanism. Therefore, we believe that the large variance in switching speed for the two samples may well be a clue that the switching mechanism for the comparative sample S2 is different from that of the filamentary switching for S1. Moreover, considering its large concentration of oxygen vacancies and barrier potential, we tentatively believe that interface switching plays an important role in the switching mechanism of S2.42

Fig. 3 shows the results of preliminary tests on the IV characteristics of the STO switching device PL spectra in different measuring regions for sample S1 (fewer oxygen vacancies). The confocal system with 40× UV objective allows a planar scanning resolution of hundreds of nanometers, while the lateral penetration depth of 325 nm laser is in the range of tens of nanometers, which guarantees that the excited emission signals are mainly generated within the thin film region. For direct measurement on STO thin film background or on the ITO transparent electrode, similar PL results were obtained (Fig. 3(b)). Despite the sharp peak at 490 nm, no visible emission was observed from the room temperature PL, indicating a scarce distribution of oxygen vacancy for S1. Moreover, the similar PL result for ITO background shows that the transparent ITO layer has little interference on the PL emission. On the other hand, as far as we know, the origin of the sharp emission peak at 490 nm has not been seen in previous reports. From the temperature dependent PL (ESI Fig. S. 3), this sharp peak becomes the dominate emission after 110 K, which correlates with the structural transition temperature of STO.48 Moreover, the green luminescence band was observed below 100 K in the as-grown STO thin film (Fig. 3(c)), but was soon quenched with rising temperature, which agrees with previous reports on STO.25,49,50 A 20 by 20 room temperature PL mapping was then performed on the ITO electrode. Again, uniform results were obtained with no visible emission, as shown in Fig. 4(a).


image file: c7nr02023f-f3.tif
Fig. 3 PL spectra of STO single crystal (a) and that of different regions on S1 (b) measured at 300 K, (c) low temperature PL at 80 K for as-grown thin film of S1, the inset shows a microscopic image representing the top view of a S1 switching device where the yellowish square is the ITO electrode of 20 μm.

image file: c7nr02023f-f4.tif
Fig. 4 Room temperature PL mapping of sample S1 with respect to the 425 nm emission peak before (a) and after (b) the forming process. (c) Room temperature mapping of the forming free sample S2 and (d) low temperature mapping at 80 K for S1 after forming.

After the forming process of device S1, the 20 by 20 PL mapping was repeated, where most of the spectra remained the same. Interestingly, there existed a small region showing weak visible emission whose PL spectrum was substantially different from other areas (Fig. 4(b)). As shown in Fig. 3(b), the spectrum consists of two bands of visible emission where the peak located around 420 nm has the energy level correlated to that of the oxygen vacancy emission reported previously.25 The nature of broad green luminescence around 540 nm is not clear. However, from the emission profile and energy level, it is certain that these emission bands originated from the defects emission as a result of the forming process. Since oxygen vacancy is the most common type of defect in STO single crystals,2 we further verified the result by measuring the PL for bulk STO single crystal for reference, where the peak around 420 nm was also observed. On the other hand, the IV characteristics of S1 showed a clear filamentary switching mechanism, which was experimentally proved to involve oxygen vacancies.18 Hence, by PL mapping of the top electrode with respect to the intensity of the blue emission peak, we could located the filament position for S1. Moreover, the mapping at 80 K (Fig. 4(d)) was in accordance with our finding. However, the low temperature scanning required the sample to be placed inside a cryostat, and thus UV objective (15×) with a long working distance was required, leading to a reduction of the scanning resolution.

For the comparative sample S2 grown under low oxygen pressure, the emission from the oxygen vacancy could be observed in a much broader region, as displayed in the mapping results (Fig. 4(c)); this indicated a higher content of oxygen vacancy with an interface switching mechanism, as discussed in the previous section. It is interesting to note the non-uniformity in mapping intensity of S2, which showed inhomogeneity of interface switching, was previously demonstrated by Wang et al. through CV and impedance spectroscopy.9 Moreover, after increasing the voltage sweeping range to 5 V when filamentary type of switching occurred, the PL mapping under this condition illustrated a slight increase in overall emission intensity, which might be due to the newly formed vacancy path. After comparing these two types of devices, it was found that it was more suitable to investigate the filamentary switching device by the PL mapping technique. Without the need to peel off the TE, nondestructive PL mapping allowed us to further investigate the reliability issues associated with the filamentary switching devices, which is difficult to achieve by means of conventional approaches. The scanning of consecutive switching from HRS to LRS (single cycle) did not yield noticeable differences from the mapping results, indicating that only small sections of the oxygen vacancies participate during switching. Therefore, we focused our attention on the cycling endurance where manifest changes in the integrated intensity can be anticipated. We studied the emission intensity variance of the filament in different endurance cycles, which was achieved by keeping the focal distance reading fixed along the Z-axis for each PL measurement, while normalizing the emission intensity of the background peak.

Fig. 5 shows the endurance data of one S1 device along with 6 sequences of PL mapping, and the hollow squares represent the integrated PL intensity at the filament region with respect to the vacancy emission peak (ESI Fig. S. 4). Data errors for each PL integration data point including random noise in the PL measurement, uncertainties in the focal distance and excitation power have been considered. After around 80 cycles of switching, the device reaches a low resistance state and is unable to switch back to the high resistance level. The inset of Fig. 5 shows the IV characteristics before and after endurance failure. For the measured vacancy emission intensity, we observed that before failure, the emission intensity (first 4 data points) was rather stable at the filament region. The small intensity of the first data point compared to the other three points is likely due to the unstable filament right after forming. However, when endurance failure took place, we observed a noticeable raise in the emission intensity, which was confirmed by the last data point, where the final 20 cycles of IV sweeping were performed after failure. The 30% rise in emission intensity before and after endurance failure suggests an evident increase in oxygen vacancy concentration. From this observation, it can be concluded that the failure in endurance is more of an abrupt change in vacancy concentration rather than an accumulative effect during the switching cycles. However, at the current stage, we are unable to give quantitative information regarding this change in vacancy concentration due to endurance failure since defect emission involves complex physics and is very sensitive to composition and density. Changes of defect parameters would not only affect the emission intensity, but also could affect other spectral features such as peak position, full width at half maximum, etc.51,52 Moreover, detailed spectroscopic investigations involving temperature dependent and time resolved spectroscopy will be conducted.


image file: c7nr02023f-f5.tif
Fig. 5 Endurance data of the S1 device (solid squares and dots) along with the integrated PL intensity at the 20 cycle interval (open squares), the inset shows the IV characteristics before and after endurance failure.

Conclusion

In this study, we have demonstrated the capability of PL mapping in the investigation of an STO based RS device. Two types of resistive switching were obtained by controlling the oxygen partial pressure during STO growth. We found that a weak conductive path can be formed in the interface switching device at a high sweeping voltage, and the distribution of oxygen vacancies over the interface is not uniform. After locating the filament region, nondestructive PL spectroscopy offered insights into the endurance failure mechanism. We showed that the variation of oxygen vacancy during the endurance test was more of an abrupt process in several cycles rather than a cumulative effect from the beginning to failure. With new advances in PL imaging resolution being available, we believe that this non-destructive technique could further be applied to other switching materials such as HfO2 and Ta2O5 and would continue to enrich our understandings on the physical origin of resistive switching, particularly targeted towards investigating the retention and endurance failure mechanism.

Experimental

The lateral device structure can be found in ESI Fig. S. 1. In order to facilitate PL measurements, indium tin oxide (ITO) was chosen to be the transparent top electrode (TE). Lanthanum strontium manganite La2/3Sr1/3MnO3 (LSMO) was used as the bottom electrode (BE), replacing the common niobium doped STO in order to avoid PL signal interference. Moreover, to serve this purpose, lanthanum aluminate (LAO) was chosen to be the substrate material instead of STO single crystals, while both LAO and LSMO exhibited perovskite crystal structure that allowed the growth of high quality STO thin film on top. The STO thin films were grown by conventional pulsed laser deposition at 700 °C. For comparison, two set of STO films named S1 and S2 were fabricated with different oxygen growth pressure of 20 Pa and 0.01 Pa, respectively. For S1, an ex situ annealing of 300 °C in oxygen was conducted to further reduce oxygen vacancies, and ITO top electrodes were deposited on both sets of samples by conventional physical vapor deposition and photolithography.

After device fabrication, the chemical composition of STO thin films was investigated by X-ray photoelectron spectroscopy (XPS) performed on a PHI Quantera SXM. The X-ray source was Al Kα radiation with a beam diameter of 100 μm. Using C 1s line at 284.8 eV as reference, high resolution energy spectra of O 1s, Ti 2p and Sr 3d were obtained. The IV characterizations on switching properties were performed on the Agilent B1500 semiconductor device analyzer. Optical characterization was carried out in a Horiba LabRAM HR Evolution system with a He–Cd laser (325 nm) as the excitation source. The magnifications of UV objective used for room temperature and 80 K studies were 40× and 15×, respectively; for the low temperature measurements, the sample was placed inside a Linkam cryostat cooled using liquid nitrogen.

Conflict of interest

The authors declare no competing financial interest.

Acknowledgements

This study is supported by the China Key Research and Development Program under Grant 2016YFA0201801 and in part by the Beijing Municipal Science and Technology Project (D161100001716002) and the National Natural Science Foundation of China under Grant 61076115, Grant 61674089, Grant 11374339, and Grant 61674087.

References

  1. 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, Proc. IEEE, 2012, 100(6), 1951–1970 CrossRef CAS .
  2. R. Waser, R. Dittmann, G. Staikov and K. Szot, Adv. Mater., 2009, 21(25–26), 2632–2663 CrossRef CAS .
  3. J. J. Yang, D. B. Strukov and D. R. Stewart, Nat. Nanotechnol., 2013, 8(1), 13–24 CrossRef CAS PubMed .
  4. M. Prezioso, F. Merrikh-Bayat, B. D. Hoskins, G. C. Adam, K. K. Likharev and D. B. Strukov, Nature, 2015, 521(7550), 61–64 CrossRef CAS PubMed .
  5. A. Chen, Comprehensive Assessment of RRAM-based PUF for Hardware Security Applications, In IEDM Tech. Dig., 2015, pp. 265–268 Search PubMed .
  6. T. Fujii, M. Kawasaki, A. Sawa, H. Akoh, Y. Kawazoe and Y. Tokura, Appl. Phys. Lett., 2005, 86(1), 012107 CrossRef .
  7. C. Park, Y. Seo, J. Jung and D.-W. Kim, J. Appl. Phys., 2008, 103(5), 054106 CrossRef .
  8. E. Mikheev, B. D. Hoskins, D. B. Strukov and S. Stemmer, Nat. Commun., 2014, 5 Search PubMed .
  9. Y. H. Wang, K. H. Zhao, X. L. Shi, G. L. Xie, S. Y. Huang and L. W. Zhang, Appl. Phys. Lett., 2013, 103(3), 031601 CrossRef .
  10. R. Buzio, A. Gerbi, A. Gadaleta, L. Anghinolfi, F. Bisio, E. Bellingeri, A. S. Siri and D. Marrè, Appl. Phys. Lett., 2012, 101(24), 243505 CrossRef .
  11. K. Szot, R. Dittmann, W. Speier and R. Waser, Phys. Status Solidi RRL, 2007, 1(2), R86–R88 CrossRef CAS .
  12. C. Baeumer, N. Raab, T. Menke, C. Schmitz, R. Rosezin, P. Müller, M. Andrä, V. Feyer, R. Bruchhaus and F. Gunkel, Nanoscale, 2016, 8(29), 13967–13975 RSC .
  13. K. Szot, W. Speier, G. Bihlmayer and R. Waser, Nat. Mater., 2006, 5(4), 312–320 CrossRef CAS PubMed .
  14. T. y. Liu, T. H. Yan, R. Scheuerlein, Y. Chen, J. K. Lee, G. Balakrishnan, G. Yee, H. Zhang, A. Yap, J. Ouyang, T. Sasaki, A. Al-Shamma, C. Chen, M. Gupta, G. Hilton, A. Kathuria, V. Lai, M. Matsumoto, A. Nigam, A. Pai, J. Pakhale, C. H. Siau, X. Wu, Y. Yin, N. Nagel, Y. Tanaka, M. Higashitani, T. Minvielle, C. Gorla, T. Tsukamoto, T. Yamaguchi, M. Okajima, T. Okamura, S. Takase, H. Inoue and L. Fasoli, IEEE J. Solid-State Circuits, 2014, 49(1), 140–153 CrossRef .
  15. Y. Yang, P. Gao, S. Gaba, T. Chang, X. Pan and W. Lu, Nat. Commun., 2012, 3, 732 CrossRef PubMed .
  16. Z. Xu, Y. Bando, W. Wang, X. Bai and D. Golberg, ACS Nano, 2010, 4(5), 2515–2522 CrossRef CAS PubMed .
  17. S. J. Choi, G. S. Park, K. H. Kim, S. Cho, W. Y. Yang, X. S. Li, J. H. Moon, K. J. Lee and K. Kim, Adv. Mater., 2011, 23(29), 3272–3277 CrossRef CAS PubMed .
  18. B. P. Andreasson, M. Janousch, U. Staub and G. I. Meijer, Appl. Phys. Lett., 2009, 94(1), 013513 CrossRef .
  19. C. Lenser, M. Patt, S. Menzel, A. Köhl, C. Wiemann, C. M. Schneider, R. Waser and R. Dittmann, Adv. Funct. Mater., 2014, 24(28), 4466–4472 CrossRef CAS .
  20. C. Lenser, A. Koehl, I. Slipukhina, H. Du, M. Patt, V. Feyer, C. M. Schneider, M. Lezaic, R. Waser and R. Dittmann, Adv. Funct. Mater., 2015, 25(40), 6360–6368 CrossRef CAS .
  21. T. Bertaud, M. Sowinska, D. Walczyk, S. Thiess, A. Gloskovskii, C. Walczyk and T. Schroeder, Appl. Phys. Lett., 2012, 101(14), 143501 CrossRef .
  22. C. Baeumer, C. Schmitz, A. Marchewka, D. N. Mueller, R. Valenta, J. Hackl, N. Raab, S. P. Rogers, M. I. Khan and S. Nemsak, Nat. Commun., 2016, 7 Search PubMed .
  23. R. S. Devan, W.-D. Ho, C.-H. Chen, H.-W. Shiu, C.-H. Ho, C.-L. Cheng, S. Y. Wu, Y. Liou and Y.-R. Ma, Nanotechnology, 2009, 20(44), 445708 CrossRef PubMed .
  24. J. Ni, Q. Zhou, Z. Li and Z. Zhang, Appl. Phys. Lett., 2008, 93(1), 011905 CrossRef .
  25. D. Kan, T. Terashima, R. Kanda, A. Masuno, K. Tanaka, S. Chu, H. Kan, A. Ishizumi, Y. Kanemitsu and Y. Shimakawa, Nat. Mater., 2005, 4(11), 816–819 CrossRef CAS .
  26. J. Wang, M. S. Gudiksen, X. Duan, Y. Cui and C. M. Lieber, Science, 2001, 293(5534), 1455–1457 CrossRef CAS PubMed .
  27. M. Heiss, S. Conesa-Boj, J. Ren, H.-H. Tseng, A. Gali, A. Rudolph, E. Uccelli, F. Peiró, J. R. Morante and D. Schuh, Phys. Rev. B: Condens. Matter, 2011, 83(4), 045303 CrossRef .
  28. X. Wang, J. Ning, C. Zheng, B. Zhu, L. Xie, H. Wu and S. Xu, J. Mater. Chem. C, 2015, 3(11), 2589–2592 RSC .
  29. D. Kan, R. Kanda, Y. Kanemitsu, Y. Shimakawa, M. Takano, T. Terashima and A. Ishizumi, Appl. Phys. Lett., 2006, 88(19), 191916 CrossRef .
  30. A. Rubano, D. Paparo, F. Miletto, U. S. di Uccio and L. Marrucci, Phys. Rev. B: Condens. Matter, 2007, 76(12), 125115 CrossRef .
  31. Y. Yamada and Y. Kanemitsu, Phys. Rev. B: Condens. Matter, 2010, 82(12), 121103 CrossRef .
  32. W. Xu, J. Yang, W. Bai, K. Tang, Y. Zhang and X. Tang, J. Appl. Phys., 2013, 114(15), 154106 CrossRef .
  33. C. Jia, M. Lentzen and K. Urban, Science, 2003, 299(5608), 870–873 CrossRef CAS PubMed .
  34. J. Hanzig, B. Abendroth, F. Hanzig, H. Stöcker, R. Strohmeyer, D. C. Meyer, S. Lindner, M. Grobosch, M. Knupfer and C. Himcinschi, J. Appl. Phys., 2011, 110(6), 064107 CrossRef .
  35. V. V. Atuchin, V. G. Kesler, N. V. Pervukhina and Z. Zhang, J. Electron Spectrosc. Relat. Phenom., 2006, 152(1), 18–24 CrossRef CAS .
  36. P. Nagarkar, P. Searson and F. Gealy III, J. Appl. Phys., 1991, 69(1), 459–462 CrossRef CAS .
  37. H. Kim, C. Gilmore, A. Pique, J. Horwitz, H. Mattoussi, H. Murata, Z. Kafafi and D. Chrisey, J. Appl. Phys., 1999, 86(11), 6451–6461 CrossRef CAS .
  38. Y. Zhang, H. Wu, Y. Bai, A. Chen, Z. Yu, J. Zhang and H. Qian, Appl. Phys. Lett., 2013, 102(23), 233502 CrossRef .
  39. J. Robertson and C. W. Chen, Appl. Phys. Lett., 1999, 74(8), 1168–1170 CrossRef CAS .
  40. P. Jha, T. D. Sands, L. Cassels, P. Jackson, T. Favaloro, B. Kirk, J. Zide, X. Xu and A. Shakouri, J. Appl. Phys., 2012, 112(6), 063714 CrossRef .
  41. R. Muenstermann, T. Menke, R. Dittmann and R. Waser, Adv. Mater., 2010, 22(43), 4819–4822 CrossRef CAS PubMed .
  42. M. Kubicek, R. Schmitt, F. Messerschmitt and J. L. M. Rupp, ACS Nano, 2015, 9(11), 10737–10748 CrossRef CAS PubMed .
  43. P. Shrestha, D. Nminibapiel, J. Kim, J. Campbell, K. Cheung, S. Deora, G. Bersuker and H. Baumgart, in Energy control paradigm for compliance-free reliable operation of RRAM, Reliability Physics Symposium, 2014 IEEE International, IEEE, 2014, pp. MY.10.1–MY.10.4 Search PubMed .
  44. L. Goux, A. Fantini, A. Redolfi, C. Chen, F. Shi, R. Degraeve, Y. Y. Chen, T. Witters, G. Groeseneken and M. Jurczak, in Role of the Ta scavenger electrode in the excellent switching control and reliability of a scalable low-current operated TiN\Ta2O5\Ta RRAM device, VLSI Technology (VLSI-Technology): Digest of Technical Papers, 2014 Symposium on, IEEE, 2014, pp. 1–2 Search PubMed .
  45. H. Lee, Y. Chen, P. Chen, P. Gu, Y. Hsu, S. Wang, W. Liu, C. Tsai, S. Sheu and P. Chiang, in Evidence and solution of over-RESET problem for HfOx based resistive memory with sub-ns switching speed and high endurance, Electron Devices Meeting (IEDM), 2010 IEEE International, IEEE, 2010, pp. 19.7.1–19.7.4 Search PubMed .
  46. L. Gao, I.-T. Wang, P.-Y. Chen, S. Vrudhula, J.-s. Seo, Y. Cao, T.-H. Hou and S. Yu, Nanotechnology, 2015, 26(45), 455204 CrossRef PubMed .
  47. X. Zhang, Q. Yu, Y. Yao and X. Li, Appl. Phys. Lett., 2010, 97(22), 222117 CrossRef .
  48. L. Rimai and G. A. deMars, Phys. Rev., 1962, 127(3), 702–710 CrossRef CAS .
  49. L. Grabner, Phys. Rev., 1969, 177(3), 1315–1323 CrossRef CAS .
  50. R. Leonelli and J. L. Brebner, Phys. Rev. B: Condens. Matter, 1986, 33(12), 8649–8656 CrossRef CAS .
  51. F. C. Ling, Z. Wang, L. P. Ho, M. Younas, W. Anwand, A. Wagner, S. Su and C. Shan, Phys. B, 2016, 480, 2–6 CrossRef CAS .
  52. U. Kaufmann, M. Kunzer, M. Maier, H. Obloh, A. Ramakrishnan, B. Santic and P. Schlotter, Appl. Phys. Lett., 1998, 72(11), 1326–1328 CrossRef CAS .

Footnote

Electronic supplementary information (ESI) available. See DOI: 10.1039/c7nr02023f

This journal is © The Royal Society of Chemistry 2017