Nano Express | Open | Published:
Low-Power Resistive Switching Characteristic in HfO2/TiOx Bi-Layer Resistive Random-Access Memory
Nanoscale Research Lettersvolume 14, Article number: 157 (2019)
Resistive random-access memory devices with atomic layer deposition HfO2 and radio frequency sputtering TiOx as resistive switching layers were fabricated successfully. Low-power characteristic with 1.52 μW set power (1 μA@1.52 V) and 1.12 μW reset power (1 μA@1.12 V) was obtained in the HfO2/TiOx resistive random-access memory (RRAM) devices by controlling the oxygen content of the TiOx layer. Besides, the influence of oxygen content during the TiOx sputtering process on the resistive switching properties would be discussed in detail. The investigations indicated that “soft breakdown” occurred easily during the electrical forming/set process in the HfO2/TiOx RRAM devices with high oxygen content of the TiOx layer, resulting in high resistive switching power. Low-power characteristic was obtained in HfO2/TiOx RRAM devices with appropriately high oxygen vacancy density of TiOx layer, suggesting that the appropriate oxygen vacancy density in the TiOx layer could avoid “soft breakdown” through the whole dielectric layers during forming/set process, thus limiting the current flowing through the RRAM device and decreasing operating power consumption.
Resistive random-access memory (RRAM) provides a promising solution for scaling down beyond traditional charge-based memory due to simple cell structure, non-volatile storage, high-speed operation, and high on/off ratio [1,2,3,4,5,6,7,8,9,10]. Recently, One-transistor one-resistor (1T1R) is a widely accepted structure to prevent inaccurate resistance measurements caused by a sneak path current in 1R array [11, 12]. Besides, the emerging memory, especially oxide-based RRAM, has been proposed for plastic synaptic devices due to the gradual conductance modulation with pulse number, which can mimic biological synaptic behavior to receive signals from pre- and postsynaptic neuron [13,14,15,16,17]. However, high resistive switching current is the main limitation for low-power and high-density application [18,19,20]. The 1T1R array also faces scaling challenges if the operation current of RRAM cannot scale accordingly. For example, when the CMOS technology is scaling down to 27 nm, the drive current will decrease to 40 μA . Therefore, reducing operation current of RRAM device down to 10 μA by optimizing structure and material is necessary to continue 1T1R scaling . In addition, biological synapses consume around 1 ~ 10 fJ per event in the complex human brain, thus, reducing the energy consumption of electrical synaptic devices as little as biological synapses is important for the development of neuromorphic artificial neural networks (ANNs) [23,24,25]. Therefore, limiting the device current and reducing the power consumption will benefit the practical process for data storage and neuromorphic computing application.
In this work, Pt/HfO2/TiOx/Pt devices with a different oxygen content of TiOx film were fabricated, and low-power characteristic in low oxygen content was demonstrated. The memory device achieved 1.52 μW set power and 1.12 μW reset power through decreasing oxygen content of the TiOx film during the sputtering process. The conductive mechanism for low-power characteristic was analyzed further to provide insights into oxide RRAM design.
The Pt/HfO2/TiOx/Pt device structure and fabrication process are shown in Fig. 1a and b. At first, on Si/SiO2/Ti substrate, a 100-nm Pt bottom electrode (BE) was prepared by direct current (DC) sputtering at room temperature. Next, 3 nm HfO2 was deposited by atomic layer deposition (ALD) (Picosun R200) technique at 300 °C using TEMAH and H2O as precursors. Subsequently, 30 nm TiOx was deposited with different oxygen content by radio frequency sputtering. During TiOx film sputtering process, fixing the total gas flow of argon (Ar) and oxygen (O2) as 20 sccm and changing the oxygen partial pressure with 9%, 11%, and 13%, three sample devices (S1, S2, and S3) were obtained to investigate the influence of oxygen content of TiOx film on the resistive switching performance. Following that, a 70-nm Pt top electrode (TE) was deposited by DC sputtering and patterned with lithography. Finally, 100 μm × 100 μm square-shape devices were formed by reactive ion etching (RIE). Bias voltage was applied on the TE, and the BE was connected with the ground. The high-resolution transmission electron microscope (HRTEM) images of the cross-section of the Pt/HfO2/TiOx/Pt are shown in Fig. 2. The electrical characteristics of the devices were measured with Agilent B1500A semiconductor parameter analyzer. The chemical states of atoms in the TiOx films were investigated by X-ray photoelectron spectroscopy (XPS, Axis Ultra).
Results and Discussion
Figure 3a, b, and c show the XPS O 1s core-level spectra of TiOx films. To clarify the chemical bond of oxygen in the films, the asymmetric O 1s peaks are divided into two peaks, which are generally ascribed to the O2− bonded by metal ions and O2− in the oxygen-deficient region . Oxygen partial pressure during TiOx film sputtering process was set as 9%, 11%, and 13%, respectively, and the corresponding oxygen-deficient content in three samples is about 28.23%, 24.06%, and 23.63%, indicating that the content of non-lattice oxygen ions and oxygen vacancies decreases with increasing oxygen partial pressure.
For the fresh devices, the original state is in high resistance state (HRS). As shown in Fig. 4, current forming (CF) is applied to initiate the formation of the conductive filament and change the device state to low resistance state (LRS) . When applying 1 μA of current compliance, a conductive path is formed in S1 and the stable set/reset process can be achieved in the subsequent operation. For S2 and S3, reset operation is not successful from the middle state of the device during CF process until the current compliance is up to 20 mA.
In order to test the electrical performance of the RRAM device, DC measurements under voltage sweep are carried out. Positive bias voltage in forming and set process is applied on TE to complete the conductive filament, and negative bias voltage in reset process is to break the filament. When the bias is swept back and forth, 100 cycles of bipolar switching current-voltage (I-V) curves of three samples are shown in Fig. 5. The S1 devices achieve stable resistive switching performance under 10 μA current compliance, but the operation current is up to 10 mA for the other two samples. The low-power characteristic of S1 could be attributed to high oxygen vacancy content preexisting in TiOx film, which limits the current effectively during forming/set process.
Figures 6 and 7 depict the cycle-to-cycle and device-to-device variation (relative standard deviation, (σ/μ)) of three samples, and the statistics are summarized in Tables 1 and 2. For S1, weak hopping current causes modest resistance distribution, and the strong conductive filaments formed in S2 and S3 guarantee the relatively stable resistance distribution. Although there is a little degradation for S3 after dozens of cycles, the on/off ratio is still over 100.
As shown in Fig. 8, the set power (Pset) 1.52 μW and the reset power (Preset) 1.12 μW are reached under a low compliance current of 1 μA. The power consumption of the other two samples is tens of milliwatt due to 10 mA of operation current. Besides, the resistance states of S1 can keep retention characteristics over 104 s under 85 °C with approximate 100 on/off ratio, which is stable for data storage application.
To elucidate the conductive mechanism of low-power characteristic, we carried out temperature measurements for S1 and S3 with different operation current and investigated the corresponding mechanism, as shown in Figs. 9 and 10. From 25 °C to 125 °C, the resistance of S1 decreases with temperature, but the resistance of S3 almost does not change. The experimental conductance and temperature are fitted with Mott’s variable range hopping model , as shown as Fig. 11, which indicates that the main conductive mechanism of S1 is electrons hopping via localized oxygen vacancy defects in dielectric oxide . When decreasing the oxygen partial pressure during the TiOx sputtering process, as shown in S1, the oxygen vacancy content in initial TiOx layer increases and the film resistance decreases . The voltage on the TE is applied mainly on the HfO2 layer and a mass of oxygen vacancies are motivated to form the conductive filament. After that, new oxygen vacancies are also motivated in the TiOx layer, but the connection among oxygen vacancies is weak. Therefore, electrons hopping conduction in TiOx is dominant, which ensures low 1-μA resistive switching current.
However, as shown in Fig. 10, after increasing oxygen content during TiOx sputtering process, HRS and LRS remain almost unchanged with temperature, which is most likely associated with the metallic-like transport mechanism, induced by electrons transport through conductive filament consisted of concentrated oxygen vacancies. Compared with the RRAM device of S1, fewer oxygen vacancies in initial TiOx layer of S3 are not enough for electrons hopping conduction. Besides, due to the increased resistance of TiOx film, the voltage bias is applied on both HfO2 layer and TiOx layer at the same time. Electric initialization leads to plenty of oxygen vacancies motivated in HfO2 and TiOx layers. These oxygen vacancies form a strong conductive filament in both two dielectric layers, and the abundant extended electrons flow through the two adjacent oxygen vacancies, which causes high operation current during resistive switching process.
In principle, it is possible to control oxygen content carefully to achieve low-power performance in other oxide resistive memories (OxRRAM) related to the oxygen vacancy. The requirement for the oxide layer is that there should be enough oxygen vacancies in the initial state for electrical hopping conduction in case of the device breakdown. However, the excessive oxygen vacancies will cause unstable endurance characteristic and deteriorate the device performance. Therefore, the appropriate oxygen vacancies are necessary to limit operation current and decrease power consumption.
Table 3 compares some of the key parameters of the Pt/HfO2/TiOx/Pt device with other recent reports. The device has important merits of low 1.12 μW resistive switching power and over 100 HRS/LRS ratio among various RRAM devices.
In this work, 1-μA resistive switching current was demonstrated in Pt/HfO2/TiOx/Pt structure device. For the conductive mechanism, electrons hopping conduction is dominant in low oxygen content of the TiOx layer, which limits operation current and decreases power consumption. Metallic-like transport is dominant in high oxygen content of the TiOx layer, and “soft-breakdown” of two dielectric layers causes high operation current and high power consumption. The appropriate oxygen content of TiOx film can limit the RRAM current and contribute to low-power characteristic, which provides a solution for large operation current and high-power issues and shows the promise for future embedded non-volatile memories and the Internet of things (IoT) applications.
Atomic layer deposition
Artificial neural networks
High resistance state
High-resolution transmission electron microscope
Internet of things
Low resistance state
- On/off ratio:
Oxide resistive memory
Reactive ion etching
Resistive random-access memory
X-ray photoelectron spectroscopy
- σ/μ :
Relative standard deviation
Wong HSP, Lee HY, Yu S et al (2012) Metal–oxide RRAM. Proc IEEE 100(6):1951–1970
Kim YB, Lee SR, Lee D et al (2011) Bi-layered RRAM with unlimited endurance and extremely uniform switching. In: VLSI Technology, 2011 Symposium on. IEEE, pp 52–53
Li KS, Ho CH, Lee MT et al (2014) Utilizing sub-5 nm sidewall electrode technology for atomic-scale resistive memory fabrication. In: VLSI Technology: Digest of Technical Papers, 2014 Symposium on. IEEE, pp 1–2
Govoreanu B, Kar GS, Chen YY et al (2011) 10× 10nm 2 Hf/HfOx crossbar resistive RAM with excellent performance, reliability and low-energy operation. In: Electron Devices Meeting (IEDM), 2011 IEEE International. IEEE, pp 31.6. 1–31.6. 4
Schönhals A, Rosário CMM, Hoffmann-Eifert S et al (2018) ReRAM: role of the electrode material on the RESET limitation in oxide ReRAM devices. Adv Electron Mater 4(2):1870011
Li H, Li KS, Lin CH et al (2016) Four-layer 3D vertical RRAM integrated with FinFET as a versatile computing unit for brain-inspired cognitive information processing. In: VLSI Technology, 2016 Symposium on. IEEE, pp 1–2
Rehman S, Khan MF, Aftab S et al (2019) Thickness-dependent resistive switching in black phosphorus CBRAM. J Mater Chem C 7(3):725
Sonde S, Chakrabarti B, Liu Y et al (2018) Silicon compatible Sn-based resistive switching memory. Nanoscale 10:9441
Yoonho A, Wook SH, Hoon LT et al (2018) Effects of Nb nanopin electrode on resistive random-access memory switching characteristics of NiO thin films. Nanoscale. https://doi.org/10.1039/C8NR02986E
Gao S, Liu G, Chen Q et al (2018) Improving unipolar resistive switching uniformity with cone-shape conducting filaments and its logic-in-memory application. ACS Appl Mater Interfaces. https://doi.org/10.1021/acsami.7b19586
Su YT, Liu HW, Chen PH et al (2018) A method to reduce forming voltage without degrading device performance in hafnium oxide-based 1T1R resistive random access memory. IEEE J Electron Devices Soc 6(1):341–345
Wu MC, Lin YW, Jang WY et al (2011) Low-power and highly reliable multilevel operation in ZrO2 1T1R RRAM. IEEE Electron Device Lett 32(8):1026–1028
Yu S (2018) Neuro-inspired computing with emerging nonvolatile memorys. Proc IEEE 106(2):260–285
Kim CH, Lim S, Woo SY et al (2018) Emerging memory technologies for neuromorphic computing. Nanotechnology 30(3):032001
Snider GS (2008) Spike-timing-dependent learning in memristive nanodevices. In: Proceedings of the 2008 IEEE International Symposium on Nanoscale Architectures. IEEE Computer Society, pp 85–92
Kuzum D, Yu S, Wong HSP (2013) Synaptic electronics: materials, devices and applications. Nanotechnology 24(38):382001
Markram H, Gerstner W, Sjöström PJ (2011) A history of spike-timing-dependent plasticity. Front Synaptic Neurosci 3:4
Cheng CH, Chin A, Yeh FS (2010) Novel Ultra-low power RRAM with good endurance and retention. In: VLSI Technology. IEEE, pp 85–86
Lee MJ, Lee CB, Lee D et al (2011) A fast, high-endurance and scalable non-volatile memory device made from asymmetric Ta2O5−x/TaO2−x bilayer structures. Nat Mater 10(8):625
Mandal S, El-Amin A, Alexander K et al (2014) Novel synaptic memory device for neuromorphic computing. Sci Rep 4:5333
Zahurak J, Miyata K, Fischer M et al (2014) Process integration of a 27nm, 16Gb Cu ReRAM. In: Electron Devices Meeting (IEDM). IEEE
Yu S, Chen PY (2016) Emerging memory technologies: recent trends and prospects. IEEE Solid-State Circuits Mag 8(2):43–56
Yu S, Gao B, Fang Z et al (2013) A low energy oxide-based electronic synaptic device for neuromorphic visual systems with tolerance to device variation. Adv Mater 25(12):1774–1779
Drachman DA (2005) Do we have brain to spare? Neurology 64(12):2004–2005
Guizzo E (2011) IBM’s Watson Jeopardy computer shuts down humans in final game. IEEE Spectr 17. https://spectrum.ieee.org/automaton/robotics/artificial-intelligence/ibm-watson-jeopardy-computer-shuts-down-humans
Veith GM, Lupini AR, Dudney NJ (2009) Role of pH in the formation of structurally stable and catalytically active TiO2-supported gold catalysts. J Phys Chem C 113(1):269–280
Gao B, Liu L, Liu X et al (2012) Resistive switching characteristics in HfOx layer by using current sweep mode. Microelectron Eng 94(none):14–17
Lin GR, Chen WC, Chang CS et al (1994) Electrical characterization of arsenic-ion-implanted semi-insulating GaAs by current-voltage measurement. Appl Phys Lett 65(25):3272–3274
Gao B, Kang JF, Chen YS et al (2011) Oxide-based RRAM: unified microscopic principle for both unipolar and bipolar switching. In: Electron Devices Meeting. IEEE
Bally AR, Hones P, Sanjinés R et al (1998) Mechanical and electrical properties of fcc TiO1+x thin films prepared by r.f. reactive sputtering. Surf Coat Technol s 108–109(98):166–170
This work was supported in part by the National Key Research and Development Program (2017YFB0405600), the National Natural Science Foundation of China (61421005, 61874006, 61334007 and 61834001) and the 111 Project (B18001).
Availability of Data and Materials
All data generated or analysed during this study are included in this published article.
The authors declare that they have no competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.