Bipolar resistive switching characteristics of amorphous SrTiO3 thin films prepared by the sol-gel process

ABSTRACT Amorphous SrTiO3 thin films grown on fluorine-doped tin oxide (FTO) glass were fabricated via the sol-gel route and coating process. The composition and chemical state of the thin films were studied by X-ray photoelectron spectroscopy. A high switching ratio and good endurance were demonstrated in the Au/amorphous SrTiO3/FTO/glass memory cells, with an ability to achieve a ratio of high and low resistance (Roff/Ron) of 102. A stable switching voltage and uniform resistance states could be identified, moreover, using standard Weibull distribution. The results showed that Ohmic and space charge limited conduction mechanisms coexisted in the amorphous SrTiO3 thin films. Ohmic conduction dominated in the initial high- and low-resistance state, but the space charge limited conduction mechanism was dominant in the later high-resistance field. The resistance switching effect in the device was explained by the formation and rupture of oxygen vacancies interrelated with filaments. These amorphous SrTiO3 films have potential resistive memory applications.


Introduction
In recent decades, continuous optimization of computer technologies marking the rapid development of modern information technology has progressively changed people's lifestyles. Memory is an indispensable carrier of information technology and is regarded as one of the most important technologies in the field of integrated circuits [1]. Semiconductor memory has been widely applied in various fields, such as information technology, social security, aerospace, defense and military [2]. Compared with volatile memory, however, non-volatile memory has great superiority in the field of mobile storage media owing to its ability, to maintain its internal storage properties even after a power failure. With the burgeoning of technologies, various types of new electronic products have emerged in an endless stream. These electronic products also have more stringent requirements for memory performance, such as high reading and writing speeds, high storage density, low power consumption, long life, greater thinness and smaller size [3,4]. The flash memory devices that play significant roles in the current electronic market still suffer from many disadvantages, such as the low operation speed, poor endurance and high write voltage problems. Eventually, miniaturization limits will be a critical issue for flash memory in future application [5]. As a result, most of today's electronic technology research is focussing on new types of memory devices nowadays. There are many kinds of memory devices based on different mechanisms and materials that are highly likely to replace non-volatile memory, such as resistive random access memory (ReRAM), ferroelectric RAM (FeRAM), magnetic RAM (MRAM), and phase change memory (PCM) [1][2][3][4][5][6]. Among these new types of memory, resistive switching (RS) memory has been widely studied due to such advantages as its simple preparation process, low energy consumption, predominant memory density, small size, and good compatibility with the conventional CMOS process [7,8]. The ReRAM memory device is a novel non-volatile memory based on the principle of resistance change of thin film materials [9,10]. The ReRAM's metal/insulator/metal (MIM) structure is composed of a thin film functioning as an insulating layer, with conductive materials as bottom and top electrodes [11,12]. This development has attracted widespread attention in the industry and academia, and many new investigations of resistive memory have been initiated by researchers [13,14]. The RS effect of the insulator layer has been discovered in various kinds of amorphous metal oxides, such as ZnO, HfO 2 , TiO 2 , MgO, Al 2 O 3 , and Y 2 O 3 [15][16][17][18][19] [20][21][22][23][24][25][26], etc. Among these, the amorphous strontium titanate-based memory structure uses the following electrodes: Pt, Ti, and indium tin oxide (ITO) [23][24][25][26]. These form Pt/amorphous-SrTiO 3 (a-STO)/Pt [23], Pt/Ti/a-STO/Pt [24], Pt/Ti/a-Nb:STO x /Pt [25] and ITO/Ti/Ti 2 O 3 /a-STO/ITO [26] memory cells, respectively.
The electrode affects the resistive switching characteristics of memristor devices, and memristors with different structures are formed using different electrodes and CONTACT Xin-Gui Tang xgtang@gdut.edu.cn Supplemental data for this article can be accessed here.
offering different resistance characteristics [27]. The resistivity characteristics of amorphous strontium titanate memory with fluorine doped tin oxide (FTO) and Au electrodes have not yet been reported. In this work, FTO was used as the electrode; we report here the RS characteristic of an Au/amorphous SrTiO 3 /FTO/Glass memristor fabricated by the sol-gel method. The RS mechanism is also discussed.

Experimental
In this work, amorphous SrTiO 3 (abbreviation a-STO) thin film was prepared by the sol-gel method and coating process [28]. Sr(NO 3 ) 2 and C 16 O 4 Ti was dissolved in 2,4-Pentanedione at room temperature. In order to ensure that the solute was distributed completely in the solvent, the reaction flasks and the experiment environment was kept clean and dry. Secondly, these two solutions were mixed under continuous stirring at 60°C for 2 hours. Stirring of the mixed solution was then continued. Finally, 2,4-Pentanedione solvent was used to adjust the final mixed solution to 0.25 mol/L. No deposition was added to the final mixed solution after it was left standing for 3 days, enabling it to be used for fabrication. Prior to depositing the film, the precursor solution was filtered with filter paper to avoid particulate contamination. The precursor solution was then deposited smoothly on a commercially available FTO-coated glass substrate. The solution spreads evenly on FTO/glass substrates under 3000 rpm speed rotation. Then coated wet film was then placed on a stable heating platform to promote thermal decomposition. The first process of thermal decomposition was conducted at about 300°C for 15 min. The above steps were duplicated three times to achieve a film thickness of about 200 nm. Finally, the film was annealed at 400°C for 20 min in the air by the RTP-1000D4 Rapid Thermal Annealing method (RTA, Hefei Kejing Materials Technology Co., LTD, China) at a heating rate of 3°C/s. A cross-sectional image of a-STO thin film was observed using a scanning electron microscope (SEM, Hitachi S-3400N-II). X-ray photoelectron spectroscopy (XPS, Escalab 250Xi) was applied to analyze the surface composition, chemical states and precise positioning of the thin film with a beam spot diameter of 650 mm. The room temperature current-voltage properties of the Au/a-STO/FTO device were characterized by a two-probe method using a Keithley 2400 programmable electrometer (Tektronix Company, USA).

Results and discussion
3.1. Microstructure Figure 1(a) displays a schematic illustration of Au/a-STO/ FTO heterostructure devices for measurement. Glass was selected as the substrate, and FTO and Au were used for the bottom and top electrodes, respectively. An SEM image of a cross-section of the a-STO thin film is shown in Figure 1(b). The thickness of the a-STO thin film can be seen to be about 200 nm. But no obvious grain can be observed in the film, which has a different morphology from FTO layers with columnar grains. In addition, the a-STO film shows a smooth appearance with no holes or cracks, and the thickness of the film is about 200 nm. It is indicated that the connector between the top and bottom electrode layers is well congested with a compact microstructure. In addition, the GIXRD pattern of STO/ FTO/Glass with the intensity increased 10 times and the XRD pattern of FTO/Glass films are shown in Fig. S. The XRD profile of STO/FTO/Glass shows a wide-peaked pattern at around 2θ = 10-35°, indicating that STO film is amorphous. It can be observed; moreover, that the three main peaks at (110), (200) and (211) appear in the XRD pattern of STO film, implying a budding polycrystalline perovskite structure in the STO film. These three weak peaks and the clear wide peak at around 2θ = 10-35°F demonstrate that an amorphous structure is dominant in the STO film. In a previous study, it was found that SrTiO 3 film retained amorphous sections when the annealing temperature was kept to 700°C [29,30]. In this work, the annealing temperature is 400°C, which is not high enough to promote good SrTiO 3 crystallization. The crystallization of the STO film depends on the hightemperature annealing process of RTA.

XPS analysis
The accurate composition and chemical state of a-STO were studied by XPS. The atomic ratio of Sr/Ti was measured as 1.002, which is exactly the same as the theoretical stoichiometric value of SrTiO 3 . To investigate the valence states and precise positions of the C 1s, O 1s, Ti 2p and Sr 3d electronic levels in the a-STO thin film, a narrow scan spectrum was fitted using the Advantage data analysis system, as shown in Figure 2. Figure 2(a) clearly shows that the C 1s peak at 284.6 eV was used as a standard to correct the binding energy of the XPS spectrum [31]. There is also a weak peak that can be attributed to C-O bonding at 286.0 eV, which is considered to synthesis by the air-atmosphere annealing process [32]. In Figure 2(b), only two peaks are observed in the XPS spectrum related to Sr 3d. And the spin-orbit splitting between the Sr 3d 3/2 and the Sr 3d 5/2 peaks located at 134.9 and 133.25 eV is 1.65 eV, as expected [33]. Analysis of the value of the binding energy of Sr 3d shows that Sr ions have a chemical state of 2 + . The Ti 2p spectrum (in Figure 2(c)) also presents a doublet structure. Ti 2p 1/2 and Ti 2p 3/2 are clearly separated at 463.85 and 458.30 eV, respectively. The spin-orbit splitting between them is 5.55 eV [31,32]. The valence of Ti ions is 4+, which is obtained from binding energy. As shown in Figure 2(d), the fitted O 1s narrow-scan spectrum is also a doublet structure. The O 1s spectra on the film surface can be decomposed into low binding energy (O LBE ) and high binding energy (O HBE ) at 529.75 and 531.63 eV, respectively. The O LBE peak and the O HBE peak are caused by oxide and hydroxide/absorbed oxygen [31][32][33], respectively. current in the stable range is extremely low. It can be determined from Figure 3 that Au/a-STO/FTO devices display a remarkable diode feature with the open-circuit voltage reaching +2.1 V. And the weak discrepancies between the 1 st , 50 th and 100 th test results indicate that the diode rectification characteristics of the device are fatigued. Figure 3(b) displays the I-V hysteresis loops with the sweeping direction of the applied voltage (from 0 to 2 V, then from 2 V to −2 V, and finally from −2 to 0). As shown in the 1 st test curve, the memory is initially in a high resistance state (HRS) with a slow current increase when operating at a positive voltage from 0 to 1.00 V. The slight increases can be clearly observed from the semilogarithmic of the I-V hysteresis loops (Figure 3(c)). If the operating bias voltage exceeds the threshold voltage of 1.00 V (V set ), however, the current increases suddenly from 6.02 × 10 −5 to 0.0043 A, suggesting that memory device is transformed from HRS to LRS. Furthermore, the current values decrease gradually after the uninterrupted cycling test. In paths 2 (+2 V→0) and 3 (0→-2 V), the LRS is stably maintained when the applied voltage sweeping is reversed. In path 4, when the operating bias voltage is over the threshold voltage of −0.95 V (V reset ), the current sharply decreases from 0.01 to 7.98 × 10 −5 A. It is worth noting that the second RS from LRS to HRS is as violent as the first switch. In the 50 th cycle, the first transition of the resistance state from LRS to HRS takes place at 1.05 V, which is higher than in the 1 st cycle. But for the 100 th cycle, there is no obvious transition of the resistance state in the positive voltage range, while the resistance state mode is similar for the 1 st and 50 th cycles in the negative voltage range. To analyze the switching behavior, the ratio of off-state and on-state (R off /R on ) was calculated from the positive voltage part of the 1st test results. It was found that the value of R off /R on is about 10 2 in the 0 V→-1.00 V range. The ratio of R off /R on in the first test results is close to that in the fiftieth results, moreover, which means the effect of RS is greater endurance and retention characteristics.
The statistical distributions of the V set and V reset and R on and R off were measured at 100 cycles to clarify the uniformity of the device. The standard deviation (Δ), average (μ) and coefficient of variation (Δ/μ) of V set and V reset and R on and R off are shown in Table 1. As displayed in Table 1, the Δ/μ of V set and V reset and R on and R off are 0.1769, 0.2327, 0.1662 and 0.3969, respectively. To further analyze the stability of RS effect in the a-STO film, the Weibull distributions of V set and V reset and R on and R off were studied, and the results are shown in Figure 4. The Weibull distribution is expressed as follows [34,35]: where F is the cumulative probability, x is the scale factor, the parameter x 0.63 is the scale factor which is the value of the statistical variable at F ≈ 63%, and β is the Weibull slope representing the statistical dispersion. The straight-fitting line reveals that the scale factor distributions are in accordance with the Weibull mode, and the values for x 0.63 and β are given by the Weibull plots and their fitting lines. All the parameters of standard Weibull distribution for V set and V reset and R on and R off are displayed in Table 1, with the β found to be 5.5175, 4.0237, 6.6758 and 15.4641, respectively. Higher values for β in the Weibull distribution and lower values for Δ/μ in the normal distribution indicate the stability and uniformity of the RS properties in the a-STO film [35][36][37]. Figure 5 shows the I-V characteristic curve of Au/ a-STO/FTO devices. The related fitting of HRS and LRS in the 0→+2 V→0 range is also drafted, and the plots can be well fitted by linear segments with different slopes. When the voltage was lower than the 1.00 V (V set ) in the 1 st test of Au/a-STO/FTO devices in Figure 5(a), the slope was 1.04, approximating 1 and indicating that the initial HRS state is Ohmic conduction. After this HRS state, the slope of the logarithmic I-V curve becomes 1.86, submitting to Child's Law (Eq. 3). The current then increases steeply to the LRS state, and the slope value is 1.10, showing an Ohmic conduction mechanism [38,39]. In Figure 5(b), a similar phenomenon can also be observed in the 50 th cycling test results, where the log|I|-log|V| curve in the HRS state is initially composed of Ohmic conduction (I~V 0.99 ) and a higher slope segment (I~V 1.28 ) and then shows Ohmic conduction (I~V 1.10 ) in the LRS state. It is well known that there are two leakage current mechanisms, namely Ohmic and space-charge-limited conduction (SCLC, Child's Law) [40][41][42]. The I-V characteristics can be expressed as [40,41]:

Conduction mechanism
where the parameters J, q, V, n 0 , ε, μ and d are the current density, elementary charge, applied voltage, concentration of free charge carriers in thermal equilibrium, static dielectric constant, electron mobility, and thickness of the thin film, respectively. The slope of I~V is 1, indicating Ohmic conduction. While the slope of I~V is larger than 1, the current increases sharply with the applied voltage, which is similar to the phenomenon seen in the later HRS region. With increases in the applied voltage, more and more electrons are injected into the insulating layer, leading to a disproportion of the space charge. Consequently, the SCLC mechanism becomes dominant. This same physical conduction mechanism is found in many amorphous films [26,43]. To sum up, HRS in the later stage of the prepared memory is mainly controlled by SCLC, while Ohmic conduction is dominant in the initial HRS and LRS regions.

Filament principle
It is reported that the significant physical conduction mechanisms include conductive filament growth and rupture, charge trapping and release, and electrothermal chemical transformation. The formation of conductive filaments is related, moreover, to oxygen vacancies or metal ions. According to the previous analysis of the RS mechanism, we refer to the oxygen vacancies (OVs)-dependent filament mode to explain the RS mechanism in Au/a-STO/FTO memory devices [44][45][46][47]. As displayed in Figure 6, the small dots represent oxygen atoms in the filament model. Initially, OVs are randomly dispersed in the a-STO representing the front of the HRS range. As the applied voltage increases, filaments are formed by the composed OVs after the setting process, and the device is transferred from HRS to LRS. Then, Ohmic conduction plays a dominant role in the LRS range. The thermal effects due to electronic motion and memory switching from LRS to HRS then eliminate the oxygen vacancies, causing filament breakdown. The primary conduction mechanism of HRS is SCLC, indicating that some oxygen defects are still present in the film.

Conclusions
In summary, the RS effect of amorphous a-STO-based memristive devices, which were prepared by the traditional chemical solution deposition route, has been demonstrated. The most significant finding is that the largest R off /R on ratio of the a-STO/FTO heterostructure devices reaches 10 2 . The results for Weibull distribution and normal distribution indicate stable switching voltages (V set and V reset ) and uniform resistance states (R on and R off ) as RS characteristics in a-STO films. The conduction mechanism was also investigated, revealing that the Ohmic conduction is dominant in the HRS state, and that the SCLC model dominates the conduction in the second part of HRS. Then, after HRS is transformed to the LRS state, the Ohmic conduction becomes dominant again. A reasonable filament theory of formation and rupture of oxygen vacancies was established, moreover, to explain the transitions of the HRS and LRS states. This study provides a memristive device with excellent performance and low-temperature preparation which can promote applications in the ReRAM field.

Disclosure statement
No potential conflict of interest was reported by the authors.