Low operating voltage resistive random access memory based on graphene oxideepolyvinyl alcohol nanocomposite thin films

a co h Ng Viet Article history: Received 28 February 2020 Received in revised form 18 April 2020 Accepted 21 April 2020 Available online 29 April 2020 switching ratio of tching ratio, they as the polyvinyl p to 1.6  10 [3,9]. VAeGO/Al device t al. obtained the /PVA/Au structure ill operate at high witching voltages VRESET ¼ þ2.45 V) [9], whereas the ITO/PVAeGO/Al device has VSET ¼
0.75 V and VRESET ¼ þ3.0 V [10]. In particular, the multi- stack Au/PVA/PVA þ GO/PVA/Au device (VSET ¼ þ3.5 V and VRESET ¼
1.8 V) and the Au/PVA þ GO/Au device (VSET ¼ þ3.87 V and VRESET ¼
2.57 V) have different switching voltages but they have the same switching ratio [11]. Therefore, the research to * Corresponding author. Vietnam National University, Ho Chi Minh City, Viet Nam. ** Corresponding author. Faculty of Materials Science and Technology, University of Science, Ho Chi Minh City, Viet Nam. E-mail addresses: pbthang@inomar.edu.vn (B.T. Phan), phamkngoc@hcmus.edu. vn (K.N. Pham). Contents lists available at ScienceDirect Journal of Science: Advance journal homepage: www.el Journal of Science: Advanced Materials and Devices 5 (2020) 199e206Peer review under responsibility of Vietnam National University, Hanoi.switching ratio (i.e. the ION/IOFF ratio). Ray et al. reported resistive (VSET ¼
5.6 V and VRESET ¼ þ3.5 V) [9]. The ITO/AueRGOePVA/Al device has the lower switching voltages (VSET ¼
0.44 V and1. Introduction A high-performance non-volatile memory with low power consumption for long battery life and for large data storage is one of the key requirements of the wearable and other electronic Internet of Things (IoT) devices. Resistive Switching Random Access Mem- ories (RRAM) can meet well those requirements [1e3]. It is well known that Graphene oxide (GO, RGO) is reported as a switching medium in RRAM [4e11]. For the application in RRAM, switching materials should have a low conductivity in order to have a high switching of the ITO/RGO/Al device with the 2.5  102 [9]. In order to improve the swi dispersed RGO into an insulating matrix, such alcohol (PVA) then reached the switching ratio u In the same manner, Sun et al. reported a ITO/P with a switching ratio of 3  10 [4,10]. Kim e switching ratio of 104 from the Au/PVA/PVAþGO [11]. However, the structures mentioned above st voltages. The ITO/RGO/Al device has the high sKeywords: Graphene oxide Polyvinyl alcohol Resistive switchinghttps://doi.org/10.1016/j.jsamd.2020.04.008 2468-2179/© 2020 The Authors. Publishing services b ( high-performance non-volatile memory with low power consumption for long battery life and for large data storage is one of the key requirements of the wearable and other electronic Internet of Things (IoT) devices. In this study, we have fabricated and investigated the resistive switching behavior of an RRAM device using the nanocomposite of polyvinyl alcohol (PVA) and graphene oxide (GO) as the switching layer in a hybrid Ag/PVAeGO/FTO structure. The resistive switching behavior of the hybrid Ag/PVAeGO/ FTO device depends on the GO amount in the PVAeGO matrix. The optical analysis depicts the good interaction through the hydrogen bonds between the hydroxyl group (eOH) of PVA and C]O of GO which play an important role in lowering the power consumption (sweeping voltage
0.5 V to þ0.5 V, VSET ¼
0.28 V, VRESET ¼ 0.34 V, switching ratio ION/IOFF ¼ 104) and switching mechanism of the hybrid Ag/PVAe0.5 wt% GO/FTO device compared to the Ag/PVA/FTO, Ag/GO/FTO and Ag/PVAe1.0 wt% GO/FTO devices. The electrical conduction mechanism is found dominant by the SCLC and the Ohm's law cor- responding to the high and low resistance states in which the combination of the trap filling and the delocalization of electrons within p bonding rings switch the device from the high to the low resistance state. © 2020 The Authors. Publishing services by Elsevier B.V. on behalf of Vietnam National University, Hanoi. This is an open access article under the CC BY license ( r t i c l e i n f o a b s t r a c tOriginal Article Low operating voltage resistive random graphene oxideepolyvinyl alcohol nano Huu Thoai Ngo a, b, Minh Trang Thi Nguyen b, c, Din Kieu Hanh Thi Ta a, b, c, Bach Thang Phan b, c, *, Kim a Faculty of Materials Science and Technology, University of Science, Ho Chi Minh City, b Vietnam National University, Ho Chi Minh City, Viet Nam c Center for Innovative Materials and Architectures, Ho Chi Minh City, Viet Namy Elsevier B.V. on behalf of Vietnamccess memory based on mposite thin films Phuc Do a, b, Kim My Tran a, b, oc Pham a, b, c, ** Nam d Materials and Devices sevier .com/locate/ jsamdNational University, Hanoi. This is an open access article under the CC BY license obtain a high switching ratio associated low switching voltages is still needed. Herein, we report the effect of the different GO concentrations (0, 0.5 and 1.0 wt% GO) dispersed into the PVA matrix on the switching performance of the hybrid PVAeGO composite films. Hybridization between PVA and GO and its influence on resistance switching of the PVAeGO composite films are discussed in detail. Among the investigated GO concentrations, the Ag/PVAe0.5 wt% GO/FTO device shows good performance with a high switching ratio of ~104 accompanied with low working voltages (VSET ¼
0.28 V and VRESET ¼ þ0.34 V). These values are good for devices with a high data storage and a low power consumption. that the particles show up with clear grain boundaries according to the FTO substrate in Fig. 3awhilst becomingmore opaque in Fig. 3b. This proves that the PVAeGO film has been successfully on the FTO substrate coated. Fig. 3c and d shows the TEM image of GO dispersed in DI solution at two distinct scales. At the scale bar of 1 mm, we can see some fader regions which indicates the well dispersion of GO in DI solution. In addition, the black spots may be due to the gathering of the GO sheets which are clearly seen in Fig. 3d. Fig. 4 depicts the currentevoltage (IeV) characteristics of 3 RRAM devices using PVA, GO and PVAeGO as a switching layer. The Ag/PVA/FTO structure exhibits no resistive switching behavior (Fig. 4a) whereas the Ag/GO/FTO shows a clear switching behavior H.T. Ngo et al. / Journal of Science: Advanced Materials and Devices 5 (2020) 199e2062002. Experimental The material and device preparation is schematically shown in Fig. 1. PVA solution (1.5 wt%) was prepared by dissolving PVA powder (99.99 in purity, Mumbai, India) in DI water while continuously being stirred in 2 h at 60 C. The GO solution of 0.5 wt % (GO powder 15e20 sheet, 4e10% edge-oxidized, Merck) was ultra-sonicated in 6 h before settled naturally in 2 h. The mixture of PVAeGO solution was prepared at the volume ratio of 1:1, ultra- sonicated and stirred in 30 min at 60 C afterward in order to ensure the uniformly dispersed substance. The 100-nm-thick PVAeGO films were deposited on a commercial 1 cm2 FTO sub- strate using the spin coating technique with the speed of 3000 rpm for 40 s. Then, 100-nm-thick Ag films were deposited as the top electrode using themagnetron DC sputtering technique (depositing power of P ¼ 20 W, target-substrate distance of d ¼ 10 cm and depositing pressure of p ¼ 2  10
3 torr). A mask was used for 1- mm-diameter Ag top electrode patterning. The film thickness was monitored using Dektak 6M stylus profiler. The morphology of the samples was observed by a field emission scanning electron mi- croscope (FESEM, S4800 Hitachi) and a transmission electron mi- croscope (TEM, JEOL JEM- 2100F). The FTIR spectrums were recorded using the Bruker Tensor 27 system. The UVeVis spectra were recorded by an Ultravioletevisible absorption spectrometer (UVevis) using the U2910 Hitachi (Japan). The Raman spectra were measured using a Labram 300 spectrometer. The real Ag/PVAeGO/ FTO device and electrical measurement configuration are shown in Fig. 2. Currentevoltage (IeV) measurement is carried out using Keithley 4200 SCS system with probe stations following the sequent bias (sweeping bias) 0 V/ þVmax/ 0 V/
Vmin/ 0 V in which the top electrode Ag is biased and the bottom electrode FTO is grounded. 3. Results and discussion The morphology of pristine FTO substrate and PVAeGO/FTO as observed through SEM, TEM images, is shown in Fig. 3. We can seeFig. 1. A schematic diagram of the fabricatiowith a low On/Off ratio of 10 (Fig. 4d). In comparison, the switching behavior of the composite structures (0.5 wt%, 1.0 wt% GO is incorporated into PVA) are presented in Fig. 4b and c. The results reveal that composite structure of 0.5 wt% GO (Fig. 4b) shows good resistive switching performance with a very high On/Off ratio of ~104 obtained at low operating voltages (V < 1 V) while the com- posite structure of 1.0 wt% GO shows unclear switching behavior (Fig. 4c). The switching process in the Ag/PVAe0.5 wt% GO/FTO structure follows bipolar-type switching with VSET ¼
0.67 V (high resistance state e HRS to low resistance states e LRS) and VRESET ¼ þ0.4 V (LRS to HRS) (Fig. 4b). Furthermore, we measured the resistive behavior of the Ag/ PVAe0.5 wt% GO/FTO device in lower sweeping voltages (
0.5 V to þ0.5 V) and found that the resistive switching characteristic still remains with the high switching ratio of 104 at lower switching voltages (VSET ¼
0.28 V and VRESET ¼ þ0.34 V), as shown in Fig. 5a. An endurance test has been carried out at reading voltage of 0.2 V, as shown in Fig. 5b. The device presents a clear and stable reversible switching for over 50 cycles. Therefore, the GOePVA bonds remain steady after switching cycles. Fig. 6 shows a performance comparison between our composite thin film device and published results [7,9e19]. It is noted that our device has a high switching ratio of 104 along with very low operating switching voltages. In the composite structure, conju- gations between the component materials play an important role for the overall performance. Therefore, we seek to figure out the interaction between GO and PVA for understanding the low oper- ating voltages as well as the switching mechanism in our Ag/ PVAe0.5 wt% GO/FTO device through the analysis of FTIR, Raman and UVevis methods. It is well-known that the wavenumber shifts of the IR bands evidence the change in the potential energy distribution along the polymeric chain due to the conjugation [20]. Fig. 7 presents the FT- IR spectra of GO, PVAeGO composites and PVA samples. For the PVA sample (Fig. 7a), the absorption peaks at 3425 cm
1, 2924 cm
1, 1720 cm
1, 1539 cm
1, and 1257 cm
1 are attributed to the eOH stretching vibration of the alcohol and carboxylic groups,n process of an Ag/PVAeGO/FTO device. nd t H.T. Ngo et al. / Journal of Science: Advanced Materials and Devices 5 (2020) 199e206 201Fig. 2. The real Ag/PVAeGO/FTO device athe eCH2 asymmetric stretching vibration, the C]O stretching vibration of carbonyl group, the eCH2 in-plane bending and the CeO stretching of the vinyl ether group, respectively. The band at about 1100 cm
1 corresponds to the CeOeC stretching of the acetyl groups [21e24]. In the spectrum of the GO sample (Fig. 7d), the broad and intensive band at 3420 cm
1 is assigned to the eOH stretching vibration of the alcohol and carboxylic groups. The lower intensive band at 2924 cm
1 is attributed to the eCH2 asymmetric stretching vibration of Csp2 v
a Csp3. The presence of a band around 1643 cm
1 is due to the C]O stretching vibration of the carboxylic group. In addition, the peaks at 1540 cm
1 and 1100 cm
1 are assigned to the C]C stretching vibration and the CeOeC stretching of the acetyl groups [25e28]. In case of the PVAeGO composite samples (Fig. 7b and c), while the FTIR of the PVAe1.0 wt% GO sample presents almost those peaks observed in both the PVA and Fig. 3. The SEM images of (a) FTO substrate and (b) PVAeGO/FTO. Thehe electrical measurement configuration.GO samples, the FTIR of the PVAe0.5 wt% GO sample shows some noticeable changes as noted in dashed regions: (1) In comparison with the eOH stretching peak at 3425 cm
1 for the PVA sample (Fig. 7d), the peak of the eOH stretching of the PVAe0.5 wt% GO sample (Fig. 7c) shifts to a higher wavenumber with the lower in- tensity; (2) In addition, the peak at 1720 cm
1 (C]O) of the PVAe0.5 wt% GO sample develops to be a shoulder while that peak of the PVAe1.0 wt% GO sample still remains a clear peak with a lower intensity; (3) the intensity of the CeOeC stretching of the acetyl groups at 1100 cm
1 decreased. It is well-known that the PVA matrix contains abundant hy- droxyl groups attached to its carbon chain to allow other materials to be incorporated via the hydrogen bonding. The three observed changes in the PVAe0.5% wt GO sample can be explained in terms of a good dispersion of the low GO concentration (0.5 wt%) in the TEM images of (c) GO at resolutions of 1.0 mm and (d) at 200 nm. cedH.T. Ngo et al. / Journal of Science: Advan202PVA matrix and also a good conjugation through the formation of the hydrogen bonds between the oxygen-containing groups of the GO sheets and the hydroxyl groups of the PVA matrix. Our results are consistent to the other works [29e31]. Raman spectroscopy has been widely employed to determine the conjugations in hybrid structures [32e39]. The typical Raman spectrum of carbon materials contains bands marked as D, G and 2D [32e36]. The D band located around 1330 cm
1 results from the presence of vacancies or dislocations in the graphene layer and at the edge of this layer. The G band located around 1582 cm
1 is related to the in-plane vibration of the sp2 hybridized carbon atoms which is bound with the symmetry and crystallization of carbon. The 2D band around 2670 cm
1 is related to the number of gra- phene layers [32e36]. In addition, the ID/IG ratio is related to the amount of defects present in the material, while the I2D/IG ratio is related to the number of graphene layers therein [37e39]. Fig. 4. The IeV plots of (a) Ag/PVA/FTO, (b) Ag/GO/FTO, (c) Ag Fig. 5. (a) The IeV plot of the Ag/PVAe0.5 wt% GO/FTO device in the sweeping voMaterials and Devices 5 (2020) 199e206As shown in Fig. 8, the ID/IG ratio of the PVAe0.5 wt% GO sample is the largest indicating the highest level of defects in this sample compared to the other investigated ones. This result shows that the conjugation of GO with PVA alters the structure of GO through the formation of more sp3 carbons within the sp2 carbon network of graphene, resulting in the largest ID/IG. The similar results were observed in the GOeAPTES composites [40,41]. In addition to ID/IG ratio, the mean crystallite size La decreases following LaðnmÞ ¼ ½ð2:4  10
10ÞðlLÞ4=
ID IG  , in which La is the average crystallite size of the sp2 domains and lL is the input laser energy (632.8 nm), resulting in smaller graphic domains in the PVAe0.5 wt% GO sample compared to those in the GO sample [42]. The I2D/IG ratio of the PVAe0.5 wt% GO sample is the largest compared to that of other investigated samples indicating a decrease of the number of GO in the composites or that 0.5 wt% /PVAe0.5% wt GO/FTO and (d) Ag/PVAe1.0% wt GO/FTO. ltage range of
0.5 to þ0.5 V and (b) The endurance of the memory device. H.T. Ngo et al. / Journal of Science: Advanced Materials and Devices 5 (2020) 199e206 203GO dispersed well in the PVA matrix. These results again confirm that the 0.5 wt% GO is successfully incorporated into the PVA matrix. Interestingly, due to a proper amount of 1.0% GO in the composite mixture, the GO sheets stack together rather than bond to PVA, resulting in the largest La value obtained in the PVAe1.0 wt% GO composites compared to those in the pristine GO sample and PVAe0.5 wt% GO sample. Fig. 6. Comparison of the switching performanc Fig. 7. The FTIR spectra of (a) PVA, (b) PVAe0.5% wt GO, (c) PVAe1.0% wt GO and (d) GO samples.Fig. 9 presents the UVevis spectra of the GO, PVAeGO com- posites and PVA samples, respectively. For the GO sample, the intensive absorption bands observed at 278 nm and 331 nm correspond to the p/ p* transition of the aromatic CeC bonds and the p/ p* transition of the C]O bonds, respectively [43]. These fragments above 350 nm may have originated from the crushing process and the ultrasonic treatment, which have been also observed and reported by other groups [43]. Absorbance is an indication for the amount of the aromatic CeC and C]O bonds present. The formation of the aromatic type CeC bonds indicates the relative abundance of the CeC bonds in GO. A higher absorption was observed for GO sample than that for the composite samples. Therefore, GO sample contains comparatively a higher number of the aromatic CeC groups and carbonyl bonds than GO in the composite samples. In addition, the degree of bonding in the GOmaterial can be expressed in terms of absorption wavelength where the higher wavelength values or lower energy ones indicate that the bonding of the system is well established. When comparing the absorption wavelength values for GO and composite samples, it is an obvious fact that bonding of GO in GO sample and in PVAe1.0 wt% GO samples are better than that of GO in the PVAe0.5 wt% GO sample. This is consistent to the largest La value in the PVAe1.0 wt% GO samples mentioned above. In other words, the conjugation between the PVA and GO is better for the PVAe0.5 wt% GO sample. This reconfirms the above discussions based on the FTIR and RAMAN analysis. As a result from the above analysis, it might suggest that conjugation between the oxygen-containing groups of the GO sheets and the hydroxyl groups of the PVAmatrix controls pathway e (VSET, VRESET and switching ratio ION/IOFF). cedH.T. Ngo et al. / Journal of Science: Advan204for carrier transport through the structure. This result in different IeV characteristics mentioned in Fig. 4 above. For example, the weak conjugation between the oxygen-containing groups of the GO sheets and the hydroxyl groups of the PVA matrix in the PVAe1.0% GO sample leads to limited pathway or more barriers for carrier transport through the structure. As a results, the Ag/PVAe1.0% GO/ FTO device shows IeV curve with unclear switching behavior (Fig. 4c) while the Ag/PVAe0.5% GO/FTO device has clear and stable switching behavior (Fig. 4b) due to its good conjugation between the oxygen-containing groups of the GO sheets and the hydroxyl Fig. 8. (a) The RAMAN spectra of (i) PVA, (ii) PVAe0.5% wt GO, (iii) PVAe1.0% wt GO Fig. 9. The UVevis spectra of a) PVA, b) PVAeGO and c) GO samples.and (iv) GO samples; (b) ID/IG and I2D/IG ratio and (c) The mean crystallite size La. Materials and Devices 5 (2020) 199e206groups of the PVAmatrix. Recently, some groups reported the effect of conjugation between GO and PVA that influences significantly those materials properties which much depend on GO concentra- tion. For example, the 0.7 wt% GO doping of PVA resulted in the enhancement of the tensile strength and Young's modulus by 62% and 76%, respectively [29]; The 0.5 wt% GO-doped PVA composite showed a high specific capacitance of 400 F/g [30]; The 0.5 wt% GO doping in PVA film increased its crystallinity (23.7% of PVA to 44.4%), tensile strength (25.6 MPa of PVA to 43 MPa), Young's modulus (144 MPa of PVA to 404 MPa) and decreased its band gap (5.31 eV of PVA to 4.64 eV) [31]. Gogoi et al. also concluded that the set voltages decrease with an increase in GO nanofillers, whereas for a fixed compliance current, the ION/IOFF ratio is found to increase with a decrease in the thickness as well as loading of GO nanofillers in the poly(methyl methacrylate) (PMMA) thin films [44]. In order to get insights in the operation process of the Ag/ PVAe0.5 wt% GO/FTO device, the IeV plot was fitted corresponding