Reset First Resistive Switching in Ni_1−xO Thin Films as Charge Transfer Insulator Deposited by Reactive RF Magnetron Sputtering

Abstract

Reset-first resistive random access memory (RRAM) devices were demonstrated for off-stoichiometric Ni_1−xO thin films deposited using reactive sputtering with a high oxygen partial pressure. The Ni_1−xO based RRAM devices exhibited both unipolar and bipolar resistive switching characteristics without an electroforming step. Auger electron spectroscopy showed nickel deficiency in the Ni_1−xO films, and X-ray photoemission spectroscopy showed that the Ni³⁺ valence state in the Ni_1−xO films increased with increasing oxygen partial pressure. Conductive atomic force microscopy showed that the conductivity of the Ni_1−xO films increased with increasing oxygen partial pressure during deposition, possibly contributing to the reset-first switching of the Ni_1−xO films.

1. Introduction

Resistive random access memory (RRAM) [1] has been widely studied as a candidate for next-generation non-volatile memory to overcome the limitations of conventional memories, such as flash memory and dynamic random access memory (DRAM). RRAM has a relatively low operation voltage with excellent program and erase speed [2]. In addition, the device could be fabricated in a simple metal–insulator–metal (MIM) [3] structure, enabling the high-density cell structure of a cross-bar array with 4F² [4,5]. It was reported that numerous transition metal oxides, including Al₂O₃ [6,7], HfO₂ [8,9,10], NiO_x [11,12,13,14], TiO_x [15,16], TaO_x [17,18], Nb₂O₅ [19,20], and Pr_1−xCa_xMnO₃ [21,22,23] show resistive switching (RS) characteristics. Moreover, various deposition techniques, such as sputtering [24,25,26,27,28], atomic layer deposition (ALD) [29] and pulsed laser deposition (PLD) [30] were used for the formation of such oxides. Notably, nickel oxide (NiO) film is one of the most widely studied oxides and is reported to have low operation power, a high on/off resistance ratio and is compatible with the CMOS fabrication process [31,32]. NiO has a rock salt structure composed of Ni²⁺ and O²⁻ and is a member of the strongly correlated 3d transition metal oxides that exhibit charge-transfer insulator behavior [33,34]. It is an insulating oxide with a wide bandgap (E_g ≈ 4.3 eV) due to the charge transfer gap caused by “Hubbard U” between the 2p and 3d states [34,35]. Therefore, the pristine state of NiO is typically the insulating state in RRAM [36,37]. The RS phenomenon in NiO has been mainly described as the formation and rupture of conductive filaments. This reversible resistance transition between the high-resistance state (HRS) and low-resistance state (LRS) is caused by applying electrical stress after an “electroforming” step [38]. It was suggested that oxygen atoms are migrated by the electric field, leaving oxygen vacancies (V_o²⁺) at the vacated sites during the electroforming step; the adjacent Ni²⁺ atoms are changed to Ni⁰ to compensate for the charge state, resulting in a Ni filament [39,40,41]. The electroforming process degrades the chemical and physical properties of devices of MIM structure, affecting their reliability. The characteristics of RS uniformity also deteriorate because of non-uniform filament formation among MIM devices [42]. Moreover, electroforming requires additional high-voltage circuits, significantly reducing the device density. Therefore, research on devices that can be operated without an electroforming step is essential for realizing RS memories [43,44,45].

This study investigated the RS characteristics of off-stoichiometric Ni_1−xO films for unipolar and bipolar RSs (URS and BRS, respectively). Particularly, it was demonstrated that nickel-deficient Ni_1−xO films deposited under excessive oxygen partial pressure exhibit a reset-first RS without an electroforming step. An RRAM device with a reset-first RS could be an alternative to overcome the limitations of RRAM requiring an electro-forming step.

2. Experimental

MIM devices with Pt/NiO/Pt and Pt/NiO/TiN stacks were fabricated for electrical characterization. First, Ti/TiN adhesive layers with thicknesses of 10–50 nm were deposited onto SiO₂ on a Si substrate using DC magnetron sputtering. Pt or TiN films were then deposited as bottom electrodes (BE). BE with various areas of 0.18~4.0 µm² were formed to investigate the area-dependence of the electrical characteristics. After BE formation, off-stoichiometric Ni_1−xO films with a thickness of 10 nm were deposited via reactive RF magnetron sputtering using a Ni target under various O₂ partial pressures. During sputtering, the base and working pressures were less than 3 × 10⁻³ and 3 mTorr, respectively. During deposition, the RF power and temperature of the substrate were main-tained at 100 W and 400 °C, respectively. The fraction of the O₂ partial pressure in the mixture of Ar and O₂ varied from 10% to 50% for deposition. Finally, Pt top electrodes (TEs) with a thickness of 100 nm were formed using DC magnetron sputtering and a lift-off process. The electrical characteristics of the device were characterized using a Keysight B1500A analyzer at 21~23 °C. RS under DC bias was measured with a com-pliance current of 10 mA to avoid hard breakdown of the Ni_1-xO films. The spatial distribution of conductivity in the pristine state was investigated using conductive atomic force microscopy (C-AFM) (Park Systems, XE-100) with a measurement bias of 3 V [46,47]. Grazing incidence X-ray diffraction (GI-XRD, Rigaku SmartLab), Auger electron spectroscopy (AES, PHI-700, ULVAC-PHI), and X-ray photoelectron spectroscopy (XPS, K-alpha, Thermo U. K.) analyses were conducted to investigate the crystallinity, composition, and valence states of Ni in the Ni_1−xO films, respectively.

3. Results and Discussion

XRD analysis was conducted to investigate the crystallinity of Ni_1−xO films. The XRD patterns of Ni_1−xO films deposited under various O₂ fractions are illustrated in Figure 1a. The peaks of NiO (111), NiO (200), NiO (220), and NiO (311) imply a polycrystalline structure [48]. NiO films, deposited with an O₂ partial pressure fraction of 50% showed lower intensity with a more comprehensive full-width half maximum (FWHM), implying poorer crystallinity of NiO films. The XRD peak of the (111) plane shifted to lower diffraction with increasing O₂ partial pressure, indicating an increase in the lattice constant with increasing O₂ partial pressure, as shown in Figure 1b. The increase in the lattice constant could be ascribed to the increased strain effect as Ni vacancies increase with excessive O₂ partial pressure [48,49,50]. Figure 1c shows the composition of Ni and O, estimated from AES analysis of the Ni_1−xO films with various O₂ partial pressures during deposition. The volume of Ni is gradually reduced with increasing O₂ partial pressure, resulting in a Ni-deficient Ni_1−xO film. The compositions of nickel oxide at 10% and 50% O₂ partial pressures were estimated to be Ni_0.89O and Ni_0.86O, respectively.

Figure 2a shows the typical behavior of Pt/Ni_1−xO/Pt stacks. The pristine Ni_1−xO films deposited under an O₂ partial pressure fraction of 10% offered an initial high resistivity [51] at an applied voltage of 1.77 V (1.4 MV/cm) on the TE. The film resistance changed from HRS to LRS during the forming step. The resistance state was changed back to HRS at 0.64 V (0.5 MV/cm) during the subsequent bias application, exhibiting reversible switching for the positive bias on TE. The difference between the forming voltage (V_form) and set voltage (V_set) was approximately 0.57 V. In contrast, pristine Ni_1−xO films deposited under the 30% or 50% O₂ ratio showed low resistance in the pristine state without the electroforming step and reset-first RS behavior, where the initial LRS state was changed to the HRS state, as shown in Figure 2b,c. While V_set is similar to that of Ni_1−xO films for the O₂ partial pressure fraction of 10%, the I_HRS/I_LRS ratio decreased because of the overall high current level in the HRS state. In particular, the I_HRS between these oxygen partial pressure fractions showed that the 50% O₂ ratio was 10 times higher than that of 30% O₂. The I-V curves of TiN/Ni_1−xO/Pt stacks are plotted in Figure 2d–f. The Ni_1−xO film deposited under a 10% O₂ partial pressure fraction show BRS [52] characteristics, as shown in Figure 2d. The pristine Ni_1−xO film showed high resistivity, and the resistance state changed to LRS after the electroforming step with a negative bias on TE. The difference between V_form (−4.0 V) and V_set (−0.7 V) was approximately 3.3 V. On the contrary, the Ni_1−xO film deposited under the 30% or 50% O₂ partial pressure fraction showed reset-first BRS behavior for a positive voltage on the TE, as shown in Figure 2e,f.

Figure 3 shows the electric currents at 0.64 V of the Pt/Ni_1−xO/TiN stacks in the LRS and HRS states, where Ni_1−xO films were deposited at various O₂ partial pressures. The mean values of I_HRS and I_LRS (red line) increased with the O₂ ratio, suggesting that the Ni_1−xO film conductivity depends on the O₂ partial pressure, as shown in Figure 3a. The Ni_1−xO films with a 10% O₂ fraction required electroforming for resistive switching, but the Ni_1−xO films with a 30% O₂ fraction or higher showed reset-first RS behavior without electroforming. Figure 3b shows the electrical currents at 0.64 V in the LRS states, which has a similar tendency to the I_HRS with O₂ partial pressure, but the slope was lower than that of the I_HRS state. The I_HRS and I_LRS showed the highest values for Ni_1−xO films deposited under the 50% O₂ partial pressure fraction.

To understand the nature of resistance switching, HRS and LRS resistances were measured from devices with BE of 0.18, 0.38, 2.00, and 3.69 μm² at a bias of ±0.48 V. Figure 4a shows the area dependent resistance for BRS device with Ni_1−xO films deposited by 10% O₂ partial pressure fraction. The resistance of the HRS remained almost constant with decreasing geometric device area, while that of the LRS is almost independent of the device area. These area-independent characteristics imply that resistance switching through the device occurs in local regions, such as filament paths, rather than homogeneously distributed switching paths [53,54,55,56,57]. Meanwhile, the resistances of reset-first RS devices with Ni_1−xO films deposited at 50% O₂ partial pressure showed increased dependence on the device area, as shown in Figure 4b. Because the area dependence of the LRS for Ni_1−xO films with 50% O₂ partial pressure is close to that of Ni_1−xO films with 10% O₂ partial pressure, the nature of the RS is filamentary in the local area. The significant dependence of HRS on the Ni_1−xO films with 50% O₂ partial pressure is attributed to the reduced resistance of the Ni_1−xO films, as shown in Figure 4b.

The DC, and AC endurance characteristics of the Ni_1−xO device are shown in Figure S1. DC endurance in Figure S1a was measured at a read voltage (V_read) of ±0.25 V under a compliance current of 10 mA. The measured I_HRS/I_LRS ratio is higher than 10¹ even after 10³ cycles. Figure S1b shows the AC endurance under pulse, which is measured with a set pulse of −0.95 V with 180 ns, a reset pulse of 1.2 V with 180 ns, and a V_read of 0.3 V conditions. The device has a uniform I_HRS/I_LRS ratio even after 10⁵ cycles, which results in a stable RS property.

C-AFM measurements investigated the two-dimensional (2D) variation of the Ni_1−xO film conductivity. Figure 5a illustrates the scheme of the C-AFM measurement. NiO/Pt and NiO/SiO₂/Pt stacks were simultaneously formed on a sample to compare the differences during the current image mapping. Cross-sectional TEM images of the Ni_1−xO films for C-AFM measurements are shown in Figure 5b. The sample-to-sample variation in the Ni_1−xO thickness on the SiO₂/Pt stacks was estimated to be within 15%. Therefore, we ignore the difference in conductivity due to thickness variation. Figure 5c–e show the current mapping images at a bias of 3 V from Ni_1−xO films deposited under various O₂ partial pressures. The left region of each mapping image represents the reference of the insulating SiO₂ between the BEs and Ni_1−xO films. The regions on the right represent the Ni_1−xO films on the Pt BEs in their pristine state. Similar to the I-V characteristics of MIM devices, C-AFM showed an increased current through the Ni_1−xO films with increasing O₂ partial pressure. The conductive regions in the Ni_1−xO film regions increased with increasing O₂ partial pressure fraction, as shown in Figure 5d,e. In particular, the current distribution is relatively uniform in Ni_1−xO film with a 50% O₂ fraction. In contrast, films deposited under 10% O₂ partial pressure fraction showed improved resistivity, as shown in Figure 5c.

The effect of the O₂ partial pressure on the chemical bonding states in the Ni_1−xO films is investigated through XPS analysis. Figure 6a–c show the Ni 2p_3/2 peaks of Ni_1−xO films deposited with various O₂ partial pressures. Ni⁰, Ni²⁺ and Ni³⁺ states with binding energies of 852.5, 853.7, and 855.5 eV, respectively, are used for deconvolution of Ni 2p_3/2 peaks [58,59].

The proportion of the Ni³⁺ state was estimated from the ratio of the Ni³⁺ peak area to the Ni²⁺ peak area. The Ni³⁺ valence state increased while the fraction of Ni²⁺ ions decreased with increasing O₂ partial pressure (Figure 6a–c). The Ni³⁺ ratio in the film grown under 10% and 50% O₂ partial pressure was estimated at 14.0% and 23.9%, respectively. Meanwhile, the Ni⁰ state at the 852.5 eV peak was not observed in our Ni 2p_2/3 peak analysis, although it was considered a conductive path in previous studies [39,40,41]. Conventionally, Ni vacancies form in Ni-deficient NiO films with relatively excessive oxygen. It was reported that nickel deficiency could promote the further oxidation of Ni²⁺ ions, which can be expressed with Kröger–Vink notation, as follows [48,49]:

$$2N{i}_{Ni}^x+\frac{1}{2}{O}_2\left(g\right) \to 2N{i}_{Ni}^{\bullet }+O_o^x+V_{Ni}^{”},$$

(1)

where $N{i}_{Ni}^x$, $N{i}_{Ni}^{\bullet }$, $O_o^x$, $V_{Ni}^{”}$ represent Ni²⁺, Ni³⁺, O²⁻, and ionized Ni vacancies, respectively. Ni²⁺ ions react with oxygen to generate ionized nickel vacancies and two Ni³⁺ ions, which affect the conductivity of the nickel oxide films. Therefore, it is shown that the increase in Ni³⁺ in Ni_1−xO films is related to the increase in the current in the HRS state of MIM devices and C-AFM. It is expected that Ni deficiency in Ni_1−xO films grown under high O₂ partial pressure causes a high Ni³⁺ concentration, leading to a highly conductive state and possibly the reset-first RS behavior with reinforced localized conductive paths [39,60,61]. Further investigation is required to understand how excess Ni³⁺ ions produce the reset-first resistive switching behavior in Ni_1−xO films.

4. Conclusions

In this study, the reset-first RS characteristics of off-stoichiometric Ni_1−xO films were investigated. The RS behavior without the electroforming step was observed in the unipolar and bipolar off-stoichiometric Ni_1−xO films. Ni³⁺ distribution contributes significantly to the conductivity of the pristine Ni_1−xO films. The conductivity and Ni deficiency of pristine Ni_1−xO films increased as the O₂ partial pressure increased during a deposition as revealed by the C-AFM and AES results. Moreover, Ni²⁺ was further oxidized to Ni³⁺ as the O₂ partial pressure increased, as revealed by the XPS results.

The Ni₂O₃ bonding by Ni³⁺ ions is related to the reset-first RS behavior without the electroforming step. This is advantageous in terms of device scale-down, making Ni_1−xO films promising candidates for memory applications by overcoming the limitations of the electroforming step in RRAM.