A. RS characteristics

In this work, the average current–voltage (I–V) curves of the ZnO samples are measured in the same way as reported previously.^{Reference Herng, Kumar, Ong, Feng, Lu, Zeng and Ding16} As the method to obtain the average I–V curves is similar to what reported in the previous study,^{Reference Herng, Kumar, Ong, Feng, Lu, Zeng and Ding16} it is therefore not reported in this paper. These I–V curves are shown as Figs. S2 and S3 (Supplementary Information) for the ZnO samples deposited under the oxygen partial pressures of ${P_{{{\rm{O}}_2}}} = 1.2 \times {10^{ - 4}}$ and 1.6 × 10⁻² Torr, respectively. These I–V curves confirms the RS behavior in ZnO samples. It is found that the “set” voltage is approximately 3.5 V, whereas the “reset” voltage is about −3.5 V. Following this, PFS technique is used to determine the coercive voltage for polarization reversal (the detail of the experiments are not shown here). The results are shown in Fig. 2. The results show the coercive voltage for the polarization reversal is ±2.5 V.

FIG. 2.

(a) Phase loops measured by PFS measurements; and (b) piezoresponse hysteresis loops (calculated from PR(E) = A(E) * cos[ϕ(E)]) for ZnO thin films deposited under two different oxygen partial pressures: ${P_{{{\rm{O}}_2}}} = 1.2 \times {10^{ - 4}}$ Torr and ${P_{{{\rm{O}}_2}}} = 1.6 \times {10^{ - 2}}$ Torr, respectively.

After determination of the average I–V curves and coercive voltage for ZnO film, the RS of the undoped ZnO film (deposited under ${P_{{{\rm{O}}_2}}} = 1.2 \times {10^{ - 4}}$ Torr) is studied by DC poling experiments using the c-AFM technique. During the poling experiments, DC biases with different polarities are applied in sequence to the sample surface [Fig. 3(a)], i.e., a 3 × 3 μm² area is first poled with −10 V DC bias [Fig. 3(b)], and then the central 1 × 1 μm² area is poled with 10 V DC bias [Fig. 3(c)]. After these poling processes, a 5 × 5 μm² area which includes these poled areas is scanned with c-AFM under 0.5 V. It is found that this small DC bias, 0.5 V used for the c-AFM scanning, neither changes the film conductance as the “set” voltage is 3.5 V; nor perturbs the domain state as 0.5 V bias cannot switch the domain, as the scanning voltage is less than the voltage for polarization reversal (±2.5 V) (Fig. 2). Hereafter, the read-out voltage used in this work for c-AFM scanning is 0.5 V. To confirm whether the switching behavior in ZnO film can be maintained, the poling processes are conducted with two rates, 1 and 0.2 Hz [Figs. 2(d) and 2(e)]. It is interesting to find that, independent of the poling rate, when biases are on, the area poled by −10 V shows no current [Fig. 3(b)], whereas the area poled by 10 V shows high current [Fig. 3(c)]. Hence, during the poling processes, the area poled by −10 V has switched to HRS whereas the area poled by 10 V has switched to low resistance state (LRS). However, after the poling processes, when the area is scanned with 0.5 V by the c-AFM technique, the situation becomes very different. It is found that, when the areas are poled with 1 Hz, after the poling processes, the area poled by −10 V is maintained at HRS, but the area poled by 10 V is switched back to HRS [Fig. 3(e)], indicating that the LRS cannot be maintained if the poling rate is high. On the other hand, when the areas are poled under 0.2 Hz, both HRS and LRS are maintained after the poling processes [Fig. 3(d)]. This suggests that the ZnO films may have similar behavior as the nonvolatile RS. It also suggests that the switched states in ZnO can only be maintained at the slow poling rate. This may have important implication on how the ZnO film should be used for the data storage applications.

FIG. 3.

(a) Schematic diagram showing the poling processes (−10 and 10 V) for ZnO film sample; (b) current image from a 3 × 3 μm² area poled by −10 V bias; (c) current image from the middle 1 × 1 μm² area poled by 10 V bias; (d) current image of 5 × 5 μm² area by these poling processes with 0.2 Hz rate; and (e) current image of 5 × 5 μm² area by these poling processes with 1 Hz rate; the read out voltage is 0.5 V for all of the current images.

B. Correlation of RS behavior with polarization reversal

The areas poled by the c-AFM technique are then scanned using the PFM technique in the same SPM system. This experiment is similar to the work by Yang and colleagues on Ca-doped BiFeO₃ film.^{Reference Yang, Seidel, Kim, Rossen, Yu, Gajek, Chu, Martin, Holcomb and He24}Figures 4(a) and 4(b) are the current image and the vertical PFM phase image of the same area after the poling processes induced by c-AFM, respectively (the poling rate by c-AFM is 0.2 Hz on the ZnO film of ${P_{{{\rm{O}}_2}}} = 1.2 \times {10^{ - 4}}$ Torr). It is noticed that when the sample is switched to HRS by −10 V, the polarization direction is also switched by more than 150° (from yellow color to purple color in the red box), on the other hand, the 10 V poling has switched the sample from HRS to LRS, but cannot switch the polarization directions [green box in Fig. 4(b)], in addition, the line profile [Fig. 4(c)] also shows that there is no observable differences in the polarization angles between the red and green boxes. However, when poling is conducted at 1 Hz rate, as the LRS cannot be maintained and has been switched back to HRS after the poling bias is removed, it is found that the polarization within the green box has also been switched back to the angle similar to the unpoled region (from purple color to yellow color), this is also confirmed by the line profile [Fig. 4(f)]. These results have further confirmed that the polarization switching only occurs at HRS, whereas LRS cannot cause polarization switching.

FIG. 4.

(a) Current image after the DC poling processes (illustrated as in Fig. 3) (with the scan rate is 0.2 Hz) on ZnO film deposited under ${P_{{{\rm{O}}_2}}} = 1.2 \times {10^{ - 4}}$ Torr; (b) out-of-plane PFM phase image of the same area [shown in (a)] after the c-AFM measurement; (c) line profiles of the out-of-plane PFM phase image from the line section marked in (b); (d) c-AFM current image after the same DC poling processes (with the scan rate is 1 Hz) on the same film; (e) out-of-plane PFM phase image of the same area [as shown in (d)] after the c-AFM measurement; and (f) line profiles of the out-of-plane PFM phase image from the line section marked in (e).

To study the effects of oxygen partial pressure to the polarization reversal, the same poling processes are also conducted on the film deposited at oxygen partial pressure of ${P_{{{\rm{O}}_2}}} = 1.6 \times {10^{ - 2}}$ Torr with the tip bias. Figures 5(a) and 5(b) are the vertical PFM phase images for the two films deposited at different oxygen partial pressures, i.e., a poling process with −10 V tip bias is first applied to a 3 × 3 µm² area (red square) and then a subsequent poling process with 10 V tip bias is applied to the central 1 × 1 μm² area (green square). It is observed that, under the tip-biased poling processes, the polarization switching occurs under HRS in both films. Figure 5(a) is the PFM phase image, which is obtained after the tip biased poling processes. It shows the similar results as that of the Fig. 4(b) which is obtained after poling processes with the sample bias, and hence this can exclude the effects of the different contact areas at the bottom-electrode/thin-film interface and tip/thin-film interface during the poling between the tip bias and sample bias. Under the same poling processes and conditions, the polarization reversal of ZnO film deposited under ${P_{{{\rm{O}}_2}}} = 1.2 \times {10^{ - 4}}$ Torr [Fig. 5(a)] is more clear and obvious than that of the ZnO film deposited under ${P_{{{\rm{O}}_2}}} = 1.6 \times {10^{ - 2}}$ Torr [Fig. 5(b)]. This can be clearly seen from the histogram of the phase angles of the two specimens [Figs. 5(c) and 5(d)]. Comparing Figs. 5(c) and 5(d), it shows that the polarization reversal is less than 180° for the film deposited at ${P_{{{\rm{O}}_2}}} = 1.6 \times {10^{ - 2}}$ Torr and it suggests that the polarization switching is uncompleted. In addition, the numbers of domain on both directions are less than those in the film deposited at ${P_{{{\rm{O}}_2}}} = 1.2 \times {10^{ - 4}}$ Torr. These results indicate that the oxygen partial pressure during the film deposition has significant effects on polarization reversal. The different oxygen partial pressures most likely result in different concentrations of the oxygen vacancy in these two samples, and the oxygen vacancy may be the domain inversed center and can assist the polarization reversal, as suggested by Herng et al.^{Reference Herng, Kumar, Ong, Feng, Lu, Zeng and Ding16} In addition, it was reported that the introduction of the oxygen concentration during deposition could reduce the amount of oxygen vacancies in the ZnO films.^{Reference Huang, Kang, Lai, He, Lin and Chueh35,Reference Ma, Du, Yang, Qiu, Zhang, Yang, Zhang, Zhao, Yang and Liu36} The oxygen vacancy concentration was reduced with increasing oxygen partial pressure when the films were deposited in the oxygen deficient atmosphere.^{Reference Herng, Qi, Berlijn, Yi, Yang, Dai, Feng, Santoso, Sánchez-Hanke, Gao, Wee, Ku, Ding and Rusydi37} Therefore, it is reasonable to believe that, for the undoped ZnO film samples, there are more oxygen vacancies in the film deposited under ${P_{{{\rm{O}}_2}}} = 1.2 \times {10^{ - 4}}$ Torr than that in the film deposited under ${P_{{{\rm{O}}_2}}} = 1.6 \times {10^{ - 2}}$ Torr. Here, the film deposited under ${P_{{{\rm{O}}_2}}} = 1.6 \times {10^{ - 2}}$ Torr shows a much more significant polarization switching behavior. Hence, the following discussions on possible RS mechanisms and the built-up schematic energy band diagrams as well as the Schottky barriers in Pt-tip/ZnO-film/Pt-bottom-electrode structures are all based on this assumption: i.e., there are more oxygen vacancies in the film deposited under low oxygen partial pressure. In fact, all of the observations can be well explained by this assumption. In this study, the polarization reversal are repeated at least five times on the same sample with the same tip but at different locations; it is also repeated by using the different tips on the same samples to eliminate the effects of wearing of the conductive coating of the tip or the changes of the tip geometry.

FIG. 5.

(a) and (b) are the out-of-plane phase images after the DC poling processes (illustrated as in Fig. 3) for ZnO films deposited under ${P_{{{\rm{O}}_2}}} = 1.2 \times {10^{ - 4}}$ Torr and ${P_{{{\rm{O}}_2}}} = 1.6 \times {10^{ - 2}}$ Torr, respectively; (c) phase histogram from the red marked squares (are poled by −10 V) for two samples; and (d) phase histogram from the green marked squares (area poled by 10 V) for two samples.

C. Correlation of RS behavior with the local charge distribution

KPFM technique is used to study the local charge characters in this work. First, the surface potential changes after a poling process (by tip bias) is measured by KPFM technique on the ZnO film deposited under ${P_{{{\rm{O}}_2}}} = 1.2 \times {10^{ - 4}}$ Torr. Figure 6(a) is the surface potential image after poling by 10 and −10 V tip biases in the adjacent two areas of 6 × 3 μm² within a 10 × 10 μm² scanned area. The scan rate is 0.5 Hz. It is generally known that the surface potential in the biased region included three types of charges, i.e., injection charge, screen charges, and polarization charges.^{Reference Kim, Bae, Ryu, Ko, Kim, Hong and Shin32,Reference Wong, Herng, Zhang, Zeng and Ding34} To quantitatively compare the KPFM results from different regions, the surface potential value of the unbiased region is reset to zero as in the previous studies^{Reference Kumar, Herng, Zeng and Ding21,Reference Kim, Bae, Ryu, Ko, Kim, Hong and Shin32,Reference Wong, Herng, Zhang, Zeng and Ding34} [Figs. 6(a)–6(f)]. If the region exhibits a larger surface potential value than that of the un-poled region, it means that the poled region is positive charged and vice versa. Figure 6(a) shows that the bright contrast in the region after poling by tip bias of 10 V, which indicates that the positive charges are stored in this region. Whereas the dark contrast in the region poled by tip bias of −10 V suggests that the surface potential value is smaller than that of the un-poled region and this means that the negative charges are stored in this region. As the injection charge and screen charge are related to the tip bias and the environments, such as adsorbed water molecules at the surface, ground tip scanning is conducted after the initial KPFM measurement. Figure 6(b) is the surface potential image after two times grounded-tip scans. It is obvious that the surface potential values in the two regions are changed significantly. The surface potential in the region poled by −10 V tip bias increases to the value similar to that of the un-poled region. This shows that the stored charges after poling by negative tip bias can be removed by grounded tip scanning and this suggests that the increased surface potential is most likely due to the surface screen charges rather than the polarization charges. In addition, as the grounded tip scanning may not be able to completely rule out the effects of adsorbed water molecules when the KPFM technique is conducted in the ambient air, the KPFM measurements are also conducted in the slow-flowing Ar environment after the sample was placed in a gas cell with Ar flowing for 2 h [Figs. 6(c) and 6(d)]. It is clear that the surface potential in the 10 V biased region (tip bias) is significantly reduced. This shows the increased surface potential after the positive biased in air is mainly due to the adsorbed water molecules. Whereas the surface potential in the negative biased region is also somewhat screened out by the water molecules. Next, as the c-AFM technique was conducted with sample bias (tip grounded), the effects of sample bias on the KPFM measurement is therefore studied, i.e., the poling process is conducted in air by using sample bias, and then the sample was placed in the gas cell with Ar flowing for 2 h and followed by the KPFM scanning in the slow-flowing Ar environments [Figs. 6(e) and 6(f)]. It is found that, unlike the case of poling process by tip bias, the effects of adsorbed water molecules to the poling by sample bias is much weaker, the KPFM measurements in Ar after 2 h still show higher surface potential on the region poled by −10 V sample bias (corresponding to 10 V tip bias). This result suggests that the effect of adsorbed water molecules during the KPFM measurements for the poling process by sample bias is very different from that of the tip bias. For the tip bias poling processes, especially under the positive bias, the increase of the surface potential is mainly due to the environmental effect, whereas for the sample poled by the negative sample bias, the effects of environmental is less significant. The surface potential is still higher after scanned in Ar environment.

FIG. 6.

Surface potential images for ZnO film deposited under ${P_{{{\rm{O}}_2}}} = 1.2 \times {10^{ - 4}}$ Torr: (a) after poling by 10 and −10 V tip biases in the two adjacent areas of 6 × 3 μm² within a 10 × 10 μm² scanning area in air; (b) the same region after two times grounded tip scanning in air; (c) after the same poling process as in (a) at another location in air; (d) after exposing sample to a flowing-Ar for 2 h; (e) after poling by 10, 0, and −10 V sample biases in three adjacent areas of 3 × 1 μm² within a 5 × 5 μm² scanning area in air; (f) the same region after exposing sample to slow flowing-Ar for 2 h. The surface potential values of the un-biased region in all of the images are reset to zero for comparison purpose.

D. Correlations between nonvolatile RS and polarization switching behavior as well as the charge distribution

After the confirmation that the way of biasing (tip or sample bias) has little effect on the polarization switching measurements, whereas there is a significant effect on the surface potential measurements, the sample bias is used in the subsequence studies. First, the c-AFM and KPFM techniques are applied to the same location (Fig. 7) to study the correlation between the RS behavior and local charge storage. Figures 7(a) and 7(b) are the topography and current images obtained by c-AFM technique. For this analysis, the sample surface is first poled with the sample biases of 10, 0, and −10 V on three adjacent areas of 3 × 1 μm² (enclosed by red lines), followed by c-AFM scan over a 5 × 5 μm² area with the sample bias of 0.5 V and the scan rate of 1 Hz. The same area is then scanned by KPFM mode. Figure 7(b) shows that the region poled by the −10 V sample bias has negligible current and hence it is HRS (blue color). On the other hand, the region poled by the 10 V sample bias is still at LRS, similar to that of the middle region (scanned with 0 V). Figure 7(c) is the corresponding surface potential images obtained by KPFM technique on the same area as that in Fig. 7(b). The region poled the −10 V sample bias exhibits a higher surface potential (bright contrast). The higher surface potential region has been extended beyond the boundary of the poling (−10 V) into the nearby region (poled by 0 V and unpoled region), however, there is no significant change on the surface potential in the region poled by the 10 V sample bias, which is consistent with the result shown in Fig. 6(e). Comparing Figs. 7(b) and 7(c), it is clear that the HRS region (poled by −10 V sample bias) shows a higher surface potential, however, there is no observable change in surface potential in the LRS region. As discussed before, the measured charges distribution also includes the polarization charges. Hence, the PFM measurement is then conducted on the same region to understand the function of the polarization charges during this process. Figures 7(d) and 7(e) are respectively the out-of-plane (vertical) PFM phase and amplitude images obtained on the same area after the c-AFM and KPFM measurements. It is clear that the PFM phase angle after poled by −10 V has switched from ∼180° to approximately 0° in the HRS region [Fig. 7(f), obtained from the line profile (red) in Fig. 7(d)]. The spontaneous polarizations are switched to the downward direction toward the bottom electrode. In addition, it is also noticed that the PFM amplitude has increased significantly at the same region [Fig. 7(e)]. These results suggest that the region poled by −10 V sample bias has switched to HRS, at the same time, the same region also shows a higher surface potential and a clear ∼180° out-of-plane polarization switching. However, for the regions poled by sample biases of 0 or 10 V (LRS), there is no clear polarization-switching process. This result unambiguously confirms that the polarization switching behavior is related to the RS characteristics, and only HRS can show the polarization switching behavior in the un-doped ZnO thin films; this micro-scale observation also confirms the previous findings on global scale reported by Herng and colleagues.^{Reference Herng, Kumar, Ong, Feng, Lu, Zeng and Ding16} In addition, it also suggests that the RS phenomena can be related to the charge storage capacity.

FIG. 7.

Correlation between c-AFM, KPFM, and PFM images on undoped ZnO films deposited under ${P_{{{\rm{O}}_2}}} = 1.2 \times {10^{ - 4}}$ Torr: (a) topography; (b) current image after applying −10, 0, −10 V biases on the three adjacent areas of 3 × 1 μm² (enclosed by red lines); (c) corresponding surface potential image; (d) and (e) corresponding out-of-plane PFM phase and amplitude images; and (f) the line profiles of the out-of-plane PFM phase changes from the line section in (d).

E. Discussions of RS mechanisms

We now discuss the possible RS mechanisms. The RS behavior may be due to several reasons: First, under external electric field, the formation of depletion region due to the Schottky barriers at the interfaces of Pt-tip/ZnO-film/Pt-bottom-electrode because of the different work functions of the contacting materials.^{Reference Kumar, Herng, Zeng and Ding21} Second, as the uncompensated polarization may also affect the electronic conductance in the ferroelectric materials^{Reference Jiang, Wang, Jin, Liu, Scott, Hwang, Tang, Lu and Yang25,Reference Yuan and Wang38,Reference Maksymovych, Jesse, Yu, Ramesh, Baddorf and Kalinin39} ; when the thin film thickness is relatively thin, the uncompensated polarization would affect the Schottky barrier since this uncompensated polarization can lead to a built-in filed, which may correspond to the order of a few MV/cm of the applied electric field and may increase the barrier's height,^{Reference Maksymovych, Jesse, Yu, Ramesh, Baddorf and Kalinin39,Reference Gruverman, Wu, Lu, Wang, Jang, Folkman, Zhuravlev, Felker, Rzchowski, Eom and Tsymbal40} and this may switch the sample to HRS. A recent work by Maksymovych and colleagues has confirmed that Fowerlor–Nordheim tunneling conductance was significantly reduced when the polarization vector was antiparallel to the applied electric field and this could lead the sample to HRS.^{Reference Maksymovych, Jesse, Yu, Ramesh, Baddorf and Kalinin39} On the other hand, the Schottky barrier height might be decreased and the sample was then switched to LRS as the screening of the polarization charges.^{Reference Maksymovych, Jesse, Yu, Ramesh, Baddorf and Kalinin39} Finally, the surface states of the samples or other environmental factors may also contribute to RS behavior as the experiment is conducted in the ambient condition.^{Reference Yuan and Wang38,Reference Qin, Yao and Liang41} Figures 8(a)–8(d) are the schematic energy band diagrams as well as the changes of the Schottky barriers at Pt-tip/ZnO-film/Pt-bottom-electrode structure due to the work function difference, depolarization electric field caused by uncompensated polarization, and the external electric field. Figure 8(a) is the energy band diagram in an ideal thermal equilibrium state of the Pt-tip/ZnO-film/Pt-bottom-electrode structure without the spontaneous polarization in ZnO film and external electric field. The symmetrical depletion region is due to the band bending resulting from the different work functions at the Pt/ZnO interfaces. Figure 8(b) is the modified energy band diagram with most upward spontaneous polarization in as-grown ZnO film and without external electric field. When ZnO film is deposited under deficient oxygen partial pressure, the bands will bend because of positively charged oxygen vacancies and the released electrons by oxygen vacancies.^{Reference Chang, Lai, Wu, Wang, Chen and Tsai17,Reference Yuan and Wang38} Figure 8(b) shows that the released free electrons can compensate the positive polarization charges, +Q _s, and hence the band is bent downward, which can reduce the Schottky barrier height. On the other hand, the negative polarization charges, −Q _s, can be compensated by the positively charged oxygen vacancies. Hence the uncompensated polarization (P _E) can lead to a built-in field, E _built-in, oriented from positive polarization charges toward negative polarization charges. The band can be bended upward and hence the depletion region is affected by the concentration of V _O and the built-in field E _built-in (Ref. Reference Qin, Yao and Liang41). Figure 8(c) is the energy band diagram when poling with negative sample bias (e.g., −10 V) which is larger than the negative coercive bias and the polarization is switched to downward direction toward the bottom electrode. Both the migration/formation of positively charged oxygen vacancies and the switched downward polarization toward the bottom electrode interface can increase the width of the depletion region. These two factors together can switch the sample to HRS. In this case, the sample may behave similar to a capacitor. Hence, the positive charges can be stored in the sample, i.e., the formed V _O. Figure 8(d) is the energy band diagram in the structure when poling with positive sample bias (e.g., 10 V) and the polarization remains the same as that in the as-grown film. The forward bias at the Pt-bottom-electrode/ZnO interface may lead to Ohmic-like contact at this interface. In addition, there may be a large number of electrons injection at the Pt-tip/ZnO interface due to the reverse bias in this interface, this may screen the polarization effects, hence the polarization does not arouse a built in field antiparallel to the external electric field. Therefore, the sample is in LRS and significant current can be observed in this region (poled by positive bias).

FIG. 8.

Schematic energy band diagrams and Schottky barriers for Pt-tip/ZnO-film/Pt-bottom-electrode system: (a) in an ideal thermal equilibrium condition without the spontaneous polarization in ZnO film and external applied field; (b) modified band diagram with most upward spontaneous polarization in as-grown ZnO film and without external applied field; (c) when poling with negative sample bias (e.g., −10 V) and the polarization is switched to downward direction toward the bottom electrode; and (d) when poling with positive sample bias (e.g., 10 V) and the polarization remains the same as that in the as-grown film. In all of the figures, W _d indicates the width of the depletion region.

This proposed RS mechanism is based on the c-AFM, KPFM, and PFM measurements on the same location. There are two possible sources of the surface charges in the KPFM measurement. The first is the migration/formation of the positively charged oxygen vacancies at the surface of the ZnO film, and/or holes injection into the ZnO film from the tip when the sample is under negative bias. Therefore, the surface potential by KPFM measurement shows a larger value. The second is that the negative charges in ZnO film can be attracted to and the positive charges can be repelled from the tip contacted area when the sample under negative bias.^{Reference Lee and Hwang8} In this case, the surface potential should show a smaller value. The results in this study have confirmed that after the poling process by the sample bias of −10 V, the samples are switched to HRS and the domains are switched to downward direction [Figs. 7(b) and 7(d)]. If there is no screen effect by the surface charge, the surface potential should be lower than that from the un-poled region. However, the KPFM measurements show the surface potential in this region is actually higher than that of the un-poled region [Figs. 6(e) and 7(c)], hence the polarization charges are likely screened by the positive charges; and therefore the first source is most likely dominated. Furthermore, comparing with the c-AFM current image [Fig. 7(b)], it can be confirmed that oxygen vacancy migration/formation is dominant over the holes injection effect. Since the ZnO thin film is in HRS within the negative poled region (poled by −10 V sample bias). If the effect of holes injection from tip is dominated, it should incur current flow. On the contrary, the migration/formation of the oxygen vacancies on the film surface, which may be the centers for domains reversal, can also assist polarization-switching process. In addition, as the downward polarization vector is antiparallel to the applied electric field and it can increase the Schottky barrier height, this can also switch the poled region to HRS [Fig. 8(c)]. Therefore, among the possible switching mechanisms, this oxygen-vacancy-related RS can simultaneously satisfy the c-AFM, KPFM, and PFM results. In the LRS, the electrons injection effect may be dominate. This is proved by the abruptly current changes [Fig. 7(b)]; in addition, there is no change in the polarization orientation when the sample is under 10 V sample bias [Fig. 7(d)].

The c-AFM, KPFM, and PFM measurements are also conducted on the same location on the undoped ZnO film deposited under ${P_{{{\rm{O}}_2}}} = 1.6 \times {10^{ - 2}}$ Torr. The poling processes are the same as before. The results are shown in Fig. S4. It is clear that the region poled by −10 V sample bias has also been switched to HRS, and the surface potential is also higher than that in the other regions. In addition, the domains in this region are also switched to downward direction in relative to the bias direction. However, it is also obvious that those changes are not as clear as those in the undoped ZnO film deposited at ${P_{{{\rm{O}}_2}}} = 1.2 \times {10^{ - 4}}$ Torr. Under the same condition, there are still some upward domains at the surface and have not been switched to downward direction. It is expected that the ZnO film deposited under higher oxygen partial pressure will have less amount of oxygen vacancies, therefore, the result reveals the oxygen vacancy plays an important role in the RS and polarization switching behavior in the ZnO film. Furthermore, the results also suggest that the ZnO film deposited under lower oxygen partial pressure can show a better RS and polarization switching behaviors.