This content has been downloaded from IOPscience. Please scroll down to see the full text.
Download details:
IP Address: 140.113.38.11
This content was downloaded on 26/04/2014 at 07:13
Please note that terms and conditions apply.
Switching Mode and Mechanism in Binary Oxide Resistive Random Access Memory Using Ni
Electrode
View the table of contents for this issue, or go to the journal homepage for more 2013 Jpn. J. Appl. Phys. 52 031801
(http://iopscience.iop.org/1347-4065/52/3R/031801)
Received July 26, 2012; accepted December 14, 2012; published online February 13, 2013
Resistive-switching (RS) modes in different CMOS-compatible binary oxides have been shown to be governed by the interplay with the Ni top electrode. Unipolar RS and metallic low-resistance state in polycrystalline HfO2 and ZrO2 are distinct from the preferential bipolar RS and semiconductive low-resistance state in amorphous Al2O3and SiO2. Backside secondary ion mass spectrometry (SIMS) has shown the formation of Ni filaments in HfO2, in contrast to the formation of oxygen-vacancy filaments in Al2O3. The differences have been explained by strong dependence of Ni migration on the oxide crystallinity. Additionally, the RS mode can be further tailored using bilayer structures. The oxide layer next to the Si bottom electrode and its tendency of forming Ni filaments play significant roles in unipolar RS in the bilayer structures, in support of the conical-shape Ni filament model where the connecting and rupture of filaments for unipolar RS occur at the smallest diameter near the bottom electrodes. # 2013 The Japan Society of Applied Physics
1. Introduction
Resistive random access memory (RRAM) employs rever-sible resistive switching (RS) between low resistance state (LRS) and high resistance state (HRS) in simple metal-insulator-metal (MIM) structures for nonvolatile information storage.1,2) In redox-based RRAM, RS has been reported to occur at local conducting filaments comprising either oxygen vacancies or metal ions in insulators. By applying appropriate SET/RESET voltage (VSET=VRESET) between top and bottom electrodes, the connecting and rupture of conducting filaments are triggered by three major physical mechanisms and results in different bipolar RS (BRS) and unipolar RS (URS) modes, namely electrochemical metalli-zation of metal filaments for BRS, valance change effect of oxygen-vacancy filaments for BRS, and thermochemical effect for URS when Joule heating is predominant at RESET.2)
Among a wide range of materials showing RS phenom-ena, such as perovskite oxides, ferromagnets, chalcogenides, and binary metal oxides,3–9)the binary metal oxides such as HfO2 have been the most interesting because of their com-patibility with advanced CMOS technology.10,11) Further-more, the RS modes in RRAM had been shown to depend strongly on the metal electrodes using an identical binary oxide.12–14) In our earlier study, HfO2 RRAM exhibited BRS with TiN/Ti, Ta and Pt top electrodes (TEs) but different URS with electrochemically active Cu and Ni TEs.14)Among them, the HfO
2RRAM with Ni TEs showed excellent cycling variations, endurance, retention, and low RESET current of 50A.14,15)The forming and SET switch-ing were attributed to the Ni filament formation because of the electrochemical metallization effect, i.e., Ni cations oxidized at the Ni anode migrate through HfO2 before eventually reduced at the Si cathode. The RESET switching was attributed to the dissolution and rupture of the Ni filaments by Joule heating. However, the dependence of different binary oxides on the RS mode using an identical
metal TE was less discussed. Whether or not the same type of filaments forms in different binary oxide RRAMs using the same TE requires further investigations. Therefore, the interplay between metal electrodes and binary oxides is still not completely understood.
In this study, RS modes of CMOS-compatible binary oxides including SiO2, Al2O3, HfO2, and ZrO2 were sys-tematically investigated using Ni TEs. Preferential BRS in amorphous Al2O3 and SiO2 and URS in polycrystalline HfO2and ZrO2suggest a RS mechanism depending strongly on the binary oxide films. Backside secondary ion mass spectrometry (SIMS) confirmed substantial Ni migration in the polycrystalline HfO2 but not in the amorphous Al2O3 after forming, supporting that the formation of Ni conduct-ing filaments was responsible for the observed URS. The URS location where the connecting and rupture of conduct-ing filaments take place at VSET and VRESET was further examined using various bilayer structures. The oxide layer next to the bottom electrode (BE) was found to play a significant role in URS. The result was in good agreement with the conical-shape Ni filament model,15,16) where the URS location was postulated to be near the BE with a smaller filament diameter.
2. Experimental Procedure
CMOS-compatible Ni/binary oxide/Si memory cells using SiO2, Al2O3, HfO2, and ZrO2 as monolayer or bilayer RS structures were fabricated on heavily doped pþ-Si substrates. After standard RCA clean and rapid thermal oxidation at 500C for 10 s in ambient O2, four different 10-nm monolayer films, HfO2/SiO2 (10 nm/2–5 nm), HfO2/Al2O3 (10 nm/2–5 nm), and Al2O3/HfO2(2–15 nm/10 nm) bilayer films were deposited. SiO2 was deposited by low-pressure chemical vapor deposition at 680C using Si(OC2H5)4 (TEOS) as the precursor. HfO2 and Al2O3 were deposited by metal organic chemical vapor deposition at 500C using Hf(OtBu)2(mmp)2/O2 and Al[OCH(CH3)2]3/O2 as the precursors, respectively. ZrO2 was deposited by atomic
layer deposition at 300C using Zr[N(CH3)(C2H5)]4 (TEMAZr) and O3 as the precursors, followed by post annealing at 400C. Finally, Ni TEs with a thickness of 100 nm and a diameter of 200m were deposited by DC sputtering and a shadow mask process. All devices were measured using an Agilent 4156B semiconductor parameter analyzer by applying voltage to the top electrodes while the pþ-silicon substrates were grounded.
3. Results and Discussion
Figure 1 illustrates the typical RS current–voltage (I–V) characteristics of the four binary oxides with Ni TEs using DC voltage sweep. To initialize RS, a positive-polarity forming process (not shown) was necessary using a fixed current compliance of 100A. In Fig. 1, the HfO2and ZrO2 devices showed URS with positive VSET and VRESET, whereas those with SiO2 and Al2O3 exhibited only BRS. The HfO2 and ZrO2 devices can also be reset at negative VRESET for BRS. The RS properties reported here were reproducible with at least several dozen cycles of continuous voltage sweeps. Figure 2 shows the statistical distributions of VSET andVRESET of different binary oxides, where both BRS and URS existed in the HfO2 and ZrO2 devices. The lack of URS in the SiO2and Al2O3devices was significantly different from the coexistence of URS and BRS in the HfO2 and ZrO2 devices. Furthermore, the temperature-dependent LRS resistances of the Ni/Al2O3/Si and Ni/HfO2/Si devices were measured, as shown in Figs. 3(a) and 3(b). Negative temperature coefficient of LRS in the Ni/Al2O3/Si device suggested a transport mechanism related to the charge hopping through semiconductive filaments. By con-trast, positive temperature coefficient of LRS in the Ni/ HfO2/Si device exhibited typical metallic characteristics. The results implied that different compositions of conductive filaments might exist in the Ni/Al2O3/Si and Ni/HfO2/Si devices and result in the different RS modes.
In our earlier study of the HfO2-based RRAM with various TEs,14) 50-nm HfO2 with TiN/Ti, Ta, and Pt TEs also showed preferential BRS. Oxygen-deficient filaments in HfO2 were confirmed by local energy dispersive X-ray spectroscopy (EDX)14) and believed to be responsible for BRS. Therefore, the BRS in Ni/Al2O3/Si and Ni/SiO2/Si might also correlate to oxygen vacancy-related filaments
in the bulk of binary oxides. This is further supported by the fact that the similar preferential BRS in Al2O3and SiO2was also observed using an inert Pt TE in Fig. 4, where filament compositions other than oxygen vacancy were less likely. Alternatively, the URS in Ni/HfO2/Si had been attributed to the formation of metallic Ni filaments, confirmed by excessive Ni diffusion by SIMS and high-resolution transmission electron microscopy (HRTEM) from several research groups.11,14,17)The resembling RS characteristics of Ni/ZrO2/Si suggested a similar mechanism.
Figure 5 compares the backside SIMS profiles of the Ni/ Al2O3/Si and Ni/HfO2/Si devices before and after positive forming. Backside SIMS profiling suppressed the artificial Ni knock-on effect from the Ni TEs in the conventional front-side SIMS, and allowed the investigation of the Ni filament formation in RRAMs.14) Some intermixing at the Ni/HfO2 and Ni/Al2O3 interfaces was attributed to the
Fig. 1. Typical RSI–V characteristics of 10-nm HfO2, ZrO2, SiO2, and Al2O3with Ni TEs.
Fig. 2. Statistical distribution of switching voltages of (a) BRS in HfO2, ZrO2, Al2O3, and SiO2devices, and (b) URS in HfO2and ZrO2devices. At least 30 switching cycles were measured for each device type.
Fig. 3. (Color online) Temperature dependence of the LRS resistance for the (a) Ni/Al2O3/Si and (b) Ni/HfO2/Si devices. Negative and positive temperature coefficients of LRS were found in the Al2O3and HfO2devices, respectively.
K.-L. Lin et al. Jpn. J. Appl. Phys. 52 (2013) 031801
sputtering process of the Ni deposition. While substantial Ni migrated from the TE into the HfO2layer after forming with the amount decreased gradually toward the Si BE, the Ni migration into the Al2O3 layer after forming was prohibited and the peak oxygen intensity in Al2O3was reduced by 20% to become more oxygen deficient. Therefore, Ni filaments were believed to be formed only in HfO2 but not in Al2O3 and responsible for the distinct difference in the RS mode and the temperature coefficient of LRS. The HfO2and ZrO2 films used in this study were polycrystalline according to X-ray diffraction (XRD) analysis, as shown in Fig. 6. The oriented grain boundaries were suspected to allow rapid Ni migration across HfO2 and ZrO2. By contrast, the amor-phous SiO2 and Al2O3 lack of oriented grain boundaries
likely suppressed Ni migration. Note that amorphous Al2O3 had been explored as an excellent diffusion barrier for Cu/low- interconnects18) and the crystallinity is generally believed to have profound impacts on the effectiveness of Cu diffusion barriers.19)
Table I summarizes the interplay between TE and binary oxide materials and the corresponding RS modes in this work and in Ref.14. According to the aforementioned physical and electrical evidence, we summarize the plausible RS mechanisms as follows: The preferential BRS may be attributed to the formation of oxygen-deficient filaments. The connecting and rupture were driven by the migration of oxygen anions using bipolar electric field. URS existed only when using electrochemically active metals of Ni and Cu as anode electrodes, accompanied by polycrystalline binary oxides of HfO2 and ZrO2. Because the binary oxides acted as solid electrolytes in an electrochemical cell,5) dif-ferent crystallinities of binary oxides significantly affected the cation migration and thus the formation of metallic filaments. The URS RESET was enabled because of Joule heating and thermal dissolution of the Ni filaments.
Fig. 4. Typical BRS characteristics in SiO2and Al2O3devices with Pt TEs.
Fig. 5. (Color online) (a) Backside SIMS profiling of the Ni/Al2O3/Si device before and after positive forming. The profiling direction was from Si into Al2O3. Inset illustrates the oxygen profile in Al2O3in linear scale, showing decreasing oxygen intensity after forming. (b) The previous analysis on a Ni/50-nm HfO2/Si device14)using identical processes as this study is also replotted for easy comparison. Substantial Ni elements migrated into HfO2after forming whereas the Ni profile of the Ni/Al2O3/Si device remained unchanged.
Fig. 6. (Color online) XRD patterns of HfO2and ZrO2thin films deposited on the pþ-Si substrates. Both HfO2and ZrO2films were polycrystalline.
Table I. Summary of RS modes of different binary oxide RRAMs in Ref.14and this study.
TE Binary
oxide BE BRS URS Note 1 TiN/Ti HfO2 Si Ref.14
2 Ta HfO2 Si Ref.14 3 Pt HfO2 Si Ref.14 4 Cu HfO2 Si Ref.14 5 Ni HfO2 Si 6 Ni ZrO2 Si 7 Ni SiO2 Si 8 Ni Al2O3 Si 9 Pt SiO2 Si 10 Pt Al2O3 Si
The design of binary oxide bilayer structures can facilitate in-depth understanding on the URS location where the connecting and rupture of conducting filaments take place at VSET and VRESET. The Ni/3-nm Al2O3/10-nm HfO2/Si (NAHS) and Ni/10-nm HfO2/3-nm Al2O3/Si (NHAS) devices were fabricated using an identical process except the reversed deposition sequence of HfO2 and Al2O3. Both devices required a similar forming voltage of approximately 7.5 V. URS was observed in the NAHS device in contrast to preferential BRS in the NHAS device, as shown in Fig. 7. Because of the preferential BRS in Al2O3 and the coexistence of BRS and URS in HfO2, the RS mode in the bilayer devices is postulated to be governed by the binary oxide layer immediately next to the BE.
It was reported that the filament morphology of the Ni/HfO2/Si URS devices may be consistently explained by a conical-shape filament model, where the connecting and rupture of Ni filaments occurred locally near the BE with a smaller diameter.15)The numerical simulation of the temperature profile at RESET in a cylindrical and a conical-shape filament are shown in Fig. 8. The RESET temperature of 600 K was chosen for illustration purpose only, but agreed with the fact that the Ni filaments may be ruptured above
573 K.14) The temperature rise because of Joule heating was the highest at the middle of the cylindrical filament but near the BE of the conical-shape filament. Therefore, the tendency of Ni filament formation in the oxide layer next to the BE played a significant role in URS.
The thickness effect of Al2O3 in the NAHS and NHAS bilayer structures was also examined. Another Ni/10-nm HfO2/SiO2/Si (NHSS) bilayer structure with various SiO2 thicknesses was also fabricated for comparison. All struc-tures demonstrated high-yield BRS. Additionally, Fig. 9 shows the device yield of URS in all bilayer structures. The yield device was defined allowing at least five consecutive URS cycles, and total 50 devices were measured for each structure. For the NAHS structures, the yield for URS degraded as the Al2O3 thickness increased. This may be explained by suppressing Ni migration into HfO2 because of the increasing thickness of the Al2O3 diffusion barrier between the Ni TE and HfO2. On the other hand, all NHAS devices exhibited only BRS regardless of the Al2O3 thick-ness investigated in this study. In contrast to the NAHS structures where abundant Ni was supplied from the TE, Ni migration was filtered first by HfO2 before much less amount of Ni can reach the BE. The result suggests that a very thin Al2O3 layer on top of the BE was sufficient to prevent the formation of complete Ni filaments. By contrast, reproducible URS was observed in the NHSS devices with thinner SiO2 layers whereas the yield of URS degraded to nearly zero with a SiO2 layer of 5 nm. SiO2 is known to be an inferior diffusion barrier as compared with Al2O3,18)and
Fig. 7. Typical URS characteristics in the Ni/3-nm Al2O3/10-nm HfO2/ Si (NAHS) device and typical BRS characteristics in the Ni/10-nm HfO2/ 3-nm Al2O3/Si (NHAS) device.
Fig. 9. Device yield of unipolar RS in (a) Ni/Al2O3/HfO2/Si, (b) Ni/ HfO2/Al2O3/Si, and (c) Ni/HfO2/SiO2/Si structures with various Al2O3
and SiO2thickness. The thickness of HfO2for all samples was 10 nm. The yield device was defined allowing at least five consecutive unipolar RS cycles, and total 50 devices were measured for each structure.
Fig. 8. (Color online) Temperature profile simulation at RESET in (a) a cylindrical and (b) a conical-shape filament. Joule heating is enhanced near the bottom electrode with a smaller diameter in the conical-shaped filament because of the current-crowding effect. The RESET condition was defined when the maximum temperature reached 600 K.
K.-L. Lin et al. Jpn. J. Appl. Phys. 52 (2013) 031801
thin SiO2 might allow the formation of Ni filaments and URS. The strong dependence of the SiO2 thickness between the Si BE and HfO2 also explains stable URS in Ni/HfO2/ Si, where a thin interfacial layer of SiOx (about 1.5 nm) existed between HfO2 and Si,14) whereas only BRS existed in the aforementioned 10-nm SiO2 monolayer.
Figure 10(a) depicts a RS model where the composition of the filament in bilayer structures depends on the metal cation diffusivity in binary oxides. For the NHAS device, Ni ions diffuse into polycrystalline HfO2 through grain boundaries before eventually blocked by the amorphous Al2O3 barrier during the forming and SET process. Hence, the dominant composition of conducting filaments was Ni in HfO2 and oxygen vacancy in Al2O3. The preferential bipolar RESET was consistent with that in Ni/Al2O3/Si and may be attrib-uted to the rupture of the oxygen-vacancy filaments in Al2O3. However, further investigations would be necessary to exclude the possibility of rupturing the Ni filaments in HfO2 under bipolar RESET driven by electric field. For the NAHS device with a thinner Al2O3layer, sufficient Ni ions were able to diffuse through thin Al2O3 into HfO2 because of the abundant source, and formed conical-shape Ni filaments necessary for URS. However, the filament size regulated by the top Al2O3layer is expected to be smaller in the NAHS device as compared with that in the Ni/HfO2/Si monolayer device. This is further supported by the tighter distribution of HRS resistance (RHRS), LRS resistance (RLRS), VSET, and VRESET of the NAHS bilayer device, as shown in Figs. 10(b) and 10(c). The higherRLRSand lower RHRS are the signature of the reduced filament size because of the increase of filament resistance at LRS and residual small filaments at HRS that were unable to be completely ruptured because of insufficient RESET power.15)
4. Conclusions
The RS modes of binary oxides, including SiO2, HfO2,
occurred locally near the Si BE for URS, in support of the conical-shape Ni filament model. Additionally, the thickness effect of the Al2O3and SiO2 interfacial layers in the bilayer structures can be consistently explained by the Ni migration. The results suggest new possibilities of designing the RS mode using engineered interfacial layers between binary oxides and metal electrodes.
Acknowledgments
The experimental work was performed at the Nano Facility Center (NFC) of National Chiao-Tung University and National Nano Device Laboratories (NDL). Financial support was partially provided by National Science Council of the Republic of China under contract No. NSC 97-2218-E-009-039-MY3.
1) I. G. Baek, M. S. Lee, S. Seo, M. J. Lee, D. H. Seo, D. S. Suh, J. C. Park, S. O. Park, H. S. Kim, I. K. Yoo, U.-I. Chung, and J. T. Moon:IEDM Tech. Dig., 2004, p. 587.
2) R. Waser, R. Dittmann, G. Staikov, and K. Szot:Adv. Mater.21 (2009) 2632.
3) C. C. Lin, C. Y. Lin, M. H. Lin, C. H. Lin, and T. Y. Tseng:Appl. Phys. Lett.93 (2008) 042115.
4) D. Hsu, J. G. Lin, and W. F. Wu:Appl. Phys. Lett.88 (2006) 222507.
5) M. N. Kozicki, M. Park, and M. Mitkova:IEEE Trans. Nanotechnol.4 (2005) 331.
6) 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 (2004) 5655.
7) W. Guan, S. Long, Q. Liu, M. Liu, and W. Wang:IEEE Electron Device Lett.29 (2008) 434.
8) W. Y. Chang, C. A. Lin, J. H. He, and T. B. Wu:Appl. Phys. Lett.96 (2010) 242109.
9) T. M. Pan and C. H. Lu:Appl. Phys. Lett.99 (2011) 113509.
10) Y. S. Chen, H. Y. Lee, P. S. Chen, P. Y. Gu, C. W. Chen, W. P. Lin, W. H. Liu, Y. Y. Hsu, S. S. Sheu, P. C. Chiang, W. S. Chen, F. T. Chen, C. H. Lien, and M.-J. Tsai:IEDM Tech. Dig., 2009, p. 105.
11) X. Li, W. H. Liu, N. Raghavan, M. Bosman, and K. L. Pey:Appl. Phys. Lett.97 (2010) 202904.
12) C. Y. Lin, C. Y. Wu, C. Y. Wu, T. C. Lee, F. L. Yang, C. Hu, and T. Y. Tseng:IEEE Electron Device Lett.28 (2007) 366.
13) C. B. Lee, B. S. Kang, A. Benayad, M. J. Lee, S. E. Ahn, K. H. Kim, G. Stefanovich, Y. Park, and I. K. Yoo:Appl. Phys. Lett.93 (2008) 042115.
14) K. L. Lin, T. H. Hou, J. Shieh, J. H. Lin, C. T. Chou, and Y. J. Lee:J. Appl. Phys.109 (2011) 084104.
15) T. H. Hou, K. L. Lin, J. Shieh, J. H. Lin, C. T. Chou, and Y. J. Lee:Appl. Phys. Lett.98 (2011) 103511.
16) K. M. Kim and C. S. Hwang:Appl. Phys. Lett.94 (2009) 122109.
17) Y. Y. Chen, G. Pourtois, C. Adelmann, L. Goux, B. Govoreanu, R. Degreave, M. Jurczak, J. A. Kittl, G. Groeseneken, and D. J. Wouters:
Appl. Phys. Lett.100 (2012) 113513.
18) D. C. Perng, J. B. Yeh, K. C. Hsu, and S. W. Tsai:Thin Solid Films518 (2010) 1648.
19) Z. H. Cao, K. Hu, and X. K. Meng:J. Appl. Phys.106 (2009) 113513. Fig. 10. (Color online) (a) Schematic of different composition of
conducting filaments in Ni/HfO2/Al2O3/Si and Ni/Al2O3/HfO2/Si devices at LRS. Cumulative probability of (b)RHRSandRLRSand (c)VSET
andVRESETfor Ni/HfO2/Si and Ni/3-nm Al2O3/HfO2/Si devices under
