Stability and its mechanism in Ag/CoOx/Ag interface-type resistive switching device

Stability is an important issue for the application of resistive switching (RS) devices. In this work, the endurance and retention properties of Ag/CoOx/Ag interface-type RS device were investigated. This device exhibits rectifying I–V curve, multilevel storage states and retention decay behavior, which are all related to the Schottky barrier at the interface. The device can switch for thousands of cycles without endurance failure and shows narrow resistance distributions with relatively low fluctuation. However, both the high and low resistance states spontaneously decay to an intermediate resistance state during the retention test. This retention decay phenomenon is due to the short lifetime τ (τ = 0.5 s) of the metastable pinning effect caused by the interface states. The data analysis indicated that the pinning effect is dependent on the depth and density of the interface state energy levels, which determine the retention stability and the switching ratio, respectively. This suggests that an appropriate interface structure can improve the stability of the interface-type RS device


I. Composition analysis
Figures S1(a) and (b) show the X-ray diffraction (XRD) patterns and X-ray photoelectron spectra (XPS) of the CoO x film, respectively. The deconvolution of Co 2p electron region XPS profile is done with CasaXPS software. The background has been subtracted by the software automatically. As shown in Figure S1(b), Co in the surface of our sample is composed of Co 2+ and Co 3+ with a roughly ratio of 4:1 (mole 2 ratio). 1 By considering the contribution of Co 3 O 4 for the Co 2+ , a proportion (mole ratio) about 3:1 for the CoO and Co 3 O 4 can be obtained from the XPS-peak-resolving analysis. We have analyzed the XPS data from three different points on the CoO x surface and found that the proportion (mole ratio) of CoO and Co 3 O 4 is almost a constant of 3:1 which shows the composition uniformity in some degree. Figure  As demonstrated in Fig. S2(a), various diameters of 400μm，500μm and 700μm of Ag electrodes have been fabricated by using masks. The resistances show a dependence on electrode areas. This indicates that the Ag/CoO x /Ag RS device is interface-type. 2 In order to provide the evidence of electrical homogeneity on nanoscale, we collected the conductivity maps of the device by conductive atomic force microscope (c-AFM). The results are shown in Fig. S2(b)~(e). Generally, the conductive filaments 3 appear after voltage switching, namely forming process. Before scanning the conductivity maps, a forming process was realized by applying +5V or -5V voltage pulses, and then the silver paint electrodes were removed by acetone. The area covered by the electrodes (have been removed) was scanned by a +5V biased probe tip to search conductive filaments. The results are shown in Fig. S2(b) and (c). As a comparison, a pristine area without switching process was scanned with the probe tip biases of +5V and -5V, respectively, which have been shown in Fig. S2 In order to create asymmetric electrode interfaces, we broke one of the interface barriers by applying a high voltage. As illustrated in the inset of Fig. S3(a), a 10 V voltage was applied between electrode pairs ′ and ′ to break the barriers at the electrode interfaces. Then we measured the I-V curve between A and B. The nearly linear I-V curve is shown in Fig. S3(a), which indicates that the interface barriers have been destroyed. This result also confirmed that the hysteresis loop is originated from the interface barriers instead of CoO x itself. After this, we measured the I-V curve between the electrodes B and C, which has a single Schottky barrier in the circuit. The result is shown in Fig. S3(b). It is a typical rectifying I-V curve with a little hysteresis loop. This asymmetric I-V curve is corresponding to the asymmetric barrier of electrodes and the resistance of reverse biased barrier is more than one order larger than that of the forward biased one. It supports the assumption that the voltage drops mainly on the reverse biased one when there are both reverse and forward biased barriers in the circuit. So it is easy to understand that the I-V curve will be dominated by the property of reverse biased barrier in low voltage range.

II. Electrode area dependent resistance and conductivity maps on nanoscale
From the inset of Fig. S3(b), it can be seen that the negative branch of I-V curve is also nonlinear and hysteresis. We analyze the branch 3 of I-V curve and find that ln (| |) has a linear relationship with | | 1/4 and | | 1/2 in the voltage ranges of 0~0.8V and 0.8~2V, respectively. This is in agreement with the analysis in the article. temperatures.
To explore the stability under different ambient conditions, we measured the electrical properties under various temperatures of 250K, 300K and 350K and the I-V curves have been shown in Fig. S5(a). We can see that the currents are intensely increasing with temperature.
Moreover, the I-V curves in atmospheric pressure and 0.1 Pa of oxygen in Fig.   S5(b) show that there are no obvious differences under various oxygen pressures. This result rules out the possibility of extensive redox.
Also, we repeated the measurement of I-V curves at 300K to verify the stability ( Fig. S5(c)). It can be seen that ten I-V curves overlap together, which shows a highly repeatability under the same condition.
To extend this model to other temperatures, we partially fitted the relationship of ln (| |) versus | | 1/4 for the I-V curves under 250K, 300K and 350K. As shown in Fig. S5(d), the slopes of the fitting curves decrease with the increasing of temperature, which is in agreement with that the slope is proportional to  Fig. S7(a). On the other hand, we can roughly verify the fitting parameter  by reading out the lifetime from the Fig. 4(a).
To explore the effect of different switching voltages on the decay process, we measured the decay processes with the switching voltages of 5  V and 4  V, respectively. Results in Fig. S7(b) indicate that the values of HRS and LRS are dependent on switching voltage, while the decay constant is independent of it. This property allows multi-bit storage in a single memory unit.