The Switching Characteristics in Bilayer ZnO/HfO₂ Resistive Random-Access Memory, Depending on the Top Electrode

Abstract

In this study, the bipolar switching behaviors in ZnO/HfO₂ bilayer resistive random-access memory (RRAM), depending on different metal top electrodes (TE), are analyzed. For this purpose, devices with two types of TE–TiN/Ti and Pd, which have varying oxygen affinities, are fabricated. X-ray diffraction (XRD) analysis shows that ZnO has a hexagonal wurtzite structure, and HfO₂ exhibits both monoclinic and orthorhombic phases. The average grain sizes are 10.9 nm for ZnO and 1.55 nm for HfO₂. In regards to the electrical characteristics, the I–V curve, cycling test, and voltage stress are measured. The measurement results indicate that devices with TiN/Ti TE exhibit lower set and higher reset voltage and stable bipolar switching behavior. However, a device with Pd TE demonstrates higher set and lower reset voltage. This phenomenon can be explained by the Gibbs free energy of formation (∆G_f°). Additionally, the Pd TE device shows unstable bipolar switching characteristics, where unipolar switching occurs simultaneously during the cycling test. This instability in devices with Pd TE could potentially lead to soft errors in operation. For guaranteeing stable bipolar switching, the oxygen affinity of material for TE should be considered in regards to ZnO/HfO₂ bilayer RRAM.

1. Introduction

The resistive random-access memory (RRAM) includes two-terminal devices with a metal–insulator–metal (MIM) structure. Since the operation of RRAM depends only on the applied voltage, the oxide is crucial for the switching characteristics. Various emerging memristor materials have recently been studies, including metal oxides, organic molecules, and 1D, 2D, and 0D materials [1,2,3]. Among these materials, the metal oxide-based bilayer has attracted attention due to its bipolar switching stability, along with various structures of bilayer RRAM, which are typically composed of a buffer layer and a switching layer, such as AlO_x/HfO₂ [4], HfO₂/AlO_x [5], Fe₃O₄/Ta₂O₅ [6], WO_x/NbO_x [7], and TaO_x/TiO₂ [8]. The mechanism of operation stability is the fact that filament formation and dissolution occur relatively stably at the interface between different oxides. L. Zou [9] reports that filament dissolution easily occurs at the TiO₂/Co₃O₄ interface. The high resistance layer of Co₃O₄, where filaments are constantly formed, acts as a virtual electrode, where filament formation and dissolution occur at the TiO₂/Co₃O₄ interface. Additionally, the stable bipolar operation in ZnO- and HfO₂-based bilayer RRAM has been studied [10,11,12]. HfO₂, known for its stability as an oxide material and as a high-k material with fewer oxygen vacancies and a high resistance layer, is used as the switching material. ZnO is used as a buffer layer due to its large oxygen vacancies, low binding energy, and higher Gibbs free energy.

However, the stable bipolar operation is influenced, not only by the oxide resistance layer, but also by the top electrode (TE). In particular, bipolar operation tends to exhibit longer endurance compared to unipolar operation due to its different reset mechanism [13], highlighting the critical role of TE. In the simulated results, both the unipolar and bipolar switching mechanisms in RRAM, depending on the TE’s varying oxygen affinity, have been reported [14]. In unipolar switching, Joule heating is a primary factor in the reset operation. The Joule heating generated by the current thermally activates the diffusion of oxygen ions, which typically migrate from the interface or the area surrounding the filaments [15]. Previous research reported that the local temperature around the filaments would rise to several hundred Kelvin due to the large current flow, as determined by electrothermal calculations [16]. The unipolar operation occurs when TE is made up of inert materials such as Pd or Pt. In contrast, bipolar switching occurs when the TE is oxidizable. The TE is made up of oxidizable materials such as Ti or TiN, and bipolar operation occurs [17]. M. Uenuma [18] has analyzed Joule heating in the NiO RRAM, reporting that bipolar operation generates less Joule heating, whereas unipolar operation produces significantly more heat. Considering these factors, the thermal situation and stability within the ZnO/HfO₂ bilayer oxide are interconnected and different, depending on the TE. Nevertheless, research on the stability of bipolar switching in the ZnO/HfO₂ bilayer RRAM remains insufficient.

Therefore, in this study, a ZnO/HfO₂ bilayer RRAM is fabricated, using HfO₂ as the switching layer and ZnO as a buffer layer, integrating TiN/Ti and Pd TE, with varying oxygen affinities. Its electrical characteristics are analyzed to understand the effect of TE on switching stability. The TE effect on ZnO/HfO₂ RRAM is explained from Gibbs free energy (∆G_f°) to understand its switching characteristics with stability.

2. Materials and Methods

The SiO₂/Si substrate is prepared. Then, an adhesion layer of 20 nm of Ti and 100 nm of Pd bottom electrode (BE) is deposited using RF-sputtering (KVS-2000 series, Korea Vacuum Tech, Gimpo-si, Republic of Korea). This sputtering system operates at a base pressure of $2\times {10}^{-6}$ Torr, with an argon gas flow of 10 sccm and a working pressure of 2 mTorr. The targets used are Ti (>99.95%) and Pd (>99.95%), with a sputtering power of 100 W. Subsequently, layers of 3 nm of HfO₂ and 3 nm of ZnO are deposited by atomic layer deposition (ALD). Tetrakis (dimethylamino) hafnium (TDMAHf) and H₂O are used as the precursor for HfO₂ at 200 °C, and diethylzinc (DEZ) and H₂O are used for the ZnO layer at 80 °C. The crystal phase and grain size are analyzed using X-ray diffraction (XRD, SmartLab 9 kW, Rigaku Corporation, Tokyo, Japan), which is equipped with a Cu-Kα radiation instrument (λ = 1.54 Å) in the range of 2θ = 10–90°. To analyze the variation in TE, two types of TE are deposited: one device with a 10 nm Ti layer and a 50 nm TiN layer, as shown in Figure 1a, and another with a 100 nm Pd, as depicted in Figure 1b, both using RF sputtering with targets of TiN (>99.95%), Ti (>99.95%), and Pd (>99.95%), under the same conditions as those previously mentioned. Figure 1c presents a top-view optical microscopic image of the fabricated RRAM. To investigate the switching characteristics, the DC I–V electrical characteristics and voltage stress are measured using a semiconductor analyzer (HP4155B, HP, Englewood, CO, USA) equipped with a static source-measure unit (SMU). Resistance measurements at various temperatures are performed on a hot stage (Linkam LTS 420, Linkam Scientific Instruments Ltd., Tadworth, Surrey, UK). During the measurements, bias is applied to the TE, while the BE is grounded.

3. Results and Discussion

3.1. XRD Analysis of the ZnO/HfO₂ Layer

The analysis of XRD is conducted using the 2-theta method. The XRD spectrum of the ZnO, used in the fabricated RRAM, is shown in Figure 2a. Compared to the ZnO ICDD card (no. 01-070-8072) in Figure 2c, the peaks at (100), (002), (101), (101), (110), (103), and (112) are observed. The ZnO exhibits a hexagonal wurtzite crystal structure. Regarding the XRD of the HfO₂ layer shown in Figure 2b, when compared with the ICDD card (no. 01-075-6426) in Figure 2d, the peaks at (011), (-111), (111), (002), (220), (202), and (013) are identified. The HfO₂ exhibits a crystalline structure, with monoclinic and orthorhombic phases [19]. The average grain size of each layer, calculated using Scherrer’s equation [20], is 10.9 nm for ZnO and 1.55 nm for the HfO₂ layer.

3.2. Electrical Properties of the TiN/Ti TE and Pd TE Device

Figure 3a displays the I–V curve of the TiN/Ti TE RRAM device, showing bipolar switching characteristics. The set voltage, which transitions the device from a high resistance state (HRS) to a low resistance state (LRS), ranges from 1.1 V to 2 V. The reset voltage, at approximately −5.1 V, transitions the device from LRS back to HRS. Additionally, the high oxygen reactivity of the TiN/Ti TE is attributed to an increase in oxygen vacancies, resulting in a relatively low resistance value in HRS. This leads to an increase in the leakage current and a decrease in the on/off ratio. In the case of single oxide layer devices, such as ZnO [21] and HfO₂ [22] with Ti TE, similar results are observed, with the resistance value in HRS being close to 10⁵ Ω. Furthermore, the TiN/Ti TE device maintains stable HRS below −6 V. Therefore, the TiN/Ti TE device shows stable bipolar switching in the ZnO/HfO₂ RRAM with the TiN/Ti TE. In the case of the device featuring Pd TE, the I–V curve also exhibits bipolar switching behavior, as shown by the arrow sequence in Figure 3b. The set voltage is fixed at 3.5 V, and the reset voltage varies from −1.2 to −2.5 V, with the compliance current set at 50 μA. The Pd TE device, which has lower oxygen affinity, exhibits a higher resistance value in HRS compared to that of the TiN/Ti TE device. This is because the leakage current is reduced due to the small oxygen vacancy in the oxide [23].

To estimate the oxygen movement during the set and reset operation in the fabricated RRAM, the standard Gibbs free energy of formation (∆G_f°) is generally used [24,25]. The ∆G_f° value for materials used in the TiN/Ti TE device is shown in Figure 4a and for the Pd TE device in Figure 4b. From a thermodynamic perspective, based on ∆G_f° theory, HfO₂, with the lowest value of −1088.2 kJ/mol [26], forms stable oxygen bonds in the device. Thus, the main switching layer, where resistance change occurs, is HfO₂ for both devices. In the device, ZnO, with a value of −320 kJ/mol [11], acts as a buffer layer, providing oxygen ions or states of oxygen vacancy to other layers. When analyzing the impact of TE for both devices, the ∆G_f° of TEs becomes crucial. In the TiN/Ti TE device, ZnO supplies oxygen ions to TiO₂, which has a low ∆G_f° value of −888.8 kJ/mol [27], thereby generating many states of oxygen vacancies in ZnO. The oxygen ions of HfO₂ can easily move to ZnO, leading to the formation of filaments. This process requires lower energy for the formation of filaments, resulting in a lower set voltage. However, when a negative voltage is applied to TE, oxygen ions are forced to move to BE. Filling the generated oxygen vacancies in ZnO and HfO₂ necessitates significant energy; thus, a higher reset voltage is required. In this process, TiO₂, which exhibits a high irreversibility, does not return all oxygen, causing a significant flow of leakage current in the HRS. In contrast, in the Pd TE device, Pd has a ∆G_f° value of −169 kJ/mol [28], inhibiting the movement of oxygen ions in ZnO to the Pd and preventing the generation of states of oxygen vacancy in ZnO. Generating sufficient oxygen vacancies requires more energy, leading to a higher set voltage. Conversely, the low ∆G_f° value of Pd suggests a low irreversibility, making it relatively easier to fill oxygen vacancies. Thus, a lower reset voltage is needed. Wei Zhang et al. [29] also reported the same phenomenon when comparing the ZnO/HfO₂ stack and the HfO₂/ZnO stack with either Ti or Pt TE.

For the cycling test in the TiN/Ti TE device, the bipolar switching is maintained over 110 cycles, as depicted in Figure 5a. In the cycling test for the Pd TE device, depicted in Figure 5b, reset failure occurs after three cycles. This is caused in the Pd TE device by the unexpected unipolar switching under negative bias, as indicated by the arrow sequence in Figure 5c. This behavior is typical of unipolar switching, resulting from inert TE [30,31].

To examine the stress stability of the device, resistance measurements are conducted, depending on temperatures and voltage stress. When tested at a read voltage (V_read) of 0.5 V, the TiN/Ti TE device demonstrates stable HRS at 25 °C, 85 °C, 140 °C, and 150 °C, as shown in Figure 6a. However, the Pd TE device in Figure 6b exhibits failure after 6 × 10³ s at 150 °C. This is because the high temperature accelerates the diffusion of oxygen ions, thus reducing the stability of the HRS. It also suggests that the small number of filaments in the Pd device makes it more susceptible, probabilistically, to transitioning to the LRS due to diffusion. In the voltage stress test, the TiN/Ti TE device endures up to 10⁴ s under negative voltage stress, depicted in Figure 7a. In contrast, the Pd TE device fails after 7 × 10² s at −0.7 V, and after 2 × 10³ s at −1 V, as shown in Figure 7b. The current induced by voltage stress within the oxide layer increases the temperature, leading to changes in the filaments and ultimately degrading stability. This phenomenon indicates that Pd devices are more susceptible to Joule heating, resulting in unstable unipolar operation. Conversely, TiN/Ti TE devices demonstrate better stability and reliability and are less sensitive to temperature fluctuations. This shows that TiN/Ti TE can maintain stable bipolar switching.

Considering the previous result, the switching mechanism is suggested for the TE. In ZnO/HfO₂ bilayer RRAMs, filament formation and dissolution are dependent on the oxygen affinity of the TE. In structures with inert Pd metals for both BE and TE, as shown in Figure 8a, the stability of the filaments decreases due to the Joule heating effect. Also, the HRS of the Pd device is vulnerable to Joule heating when the negative bias is applied for a long period of time. Compared with the operational characteristics of the previously fabricated Pd TE device, this device exhibits lower stability, as observed in the mechanism shown in Figure 8a. Using a TiN/Ti TE with high oxygen affinity, as shown in Figure 8b, enables the formation of a stable filament and ensures stable operation. Therefore, the TiN/Ti TE device demonstrates relatively higher stability.

4. Conclusions

The ZnO/HfO₂ bilayer RRAM devices using TiN/Ti TE and Pd TE are analyzed to evaluated their switching properties. The I–V curve indicates that the device with TiN/Ti TE exhibits a lower set voltage and a higher reset voltage. In contrast, the device with Pd TE demonstrates a higher set voltage and a lower reset voltage. This difference is due to the Gibbs free energy of the formation (∆G_f°) of TE, which influences the amount of energy required for oxygen movement, resulting in changes to the operating voltage. Additionally, cycling tests and retention measurements in Pd TE devices show low stability, which potentially leads to worse performance. This instability can be attributed to the effect of TE, which accelerates Joule heating and results in unipolar switching. To guarantee stable bipolar switching, the oxygen affinity of the material for TE should be considered. Overall, the choice of electrode material significantly affects the stability and performance of bilayer oxide RRAM devices.