Direct growth of TiO₂ nanotubes on transparent substrates and their resistive switching characteristics

Significant progress has been made in the application of various TiO₂ nanotube-based electronic, optoelectronic and sensor devices [1–4]. There are three general approaches to the synthesis of TiO₂ nanotubes, namely chemical synthesis using template [5], electrochemical approaches by anodizing of Ti foils [6, 7] and the alkaline hydrothermal method [8]. Each method has its advantages and drawbacks. For instance, anodization of Ti foil represents one of the simplest methods to obtain highly ordered TiO₂ nanotubes, while Ti foils are not suitable for the fabrication of optoelectronic devices that usually require transparency. Thus, the pursuit of novel fabrication methods for nanotubes continues to be driven by an ever increasing world of practical device applications.

Recently, the resistive switching behaviour in TiO₂ has drawn attention due to its application to resistive random access memory (RRAM) devices [9–12]. RRAM has been regarded as next-generation nonvolatile memory devices due to its simple structure and compatibility with complementary metal oxide semiconductor technology [9, 10, 13–15]. In comparison with traditional nonvolatile memories (flash), RRAM has unique advantages, such as much faster writing rate, smaller bit cell size and lower operating voltages. So far, the resistive switching behaviours have been only reported in TiO₂ thin films and nanowires, but not in TiO₂ nanotubes.

In this work, TiO₂ nanotubes were directly grown on fluorine-doped tin oxide (FTO) substrate and their resistive switching characteristics were investigated. The distinct geometry of TiO₂ nanotubes leads to an excellent nonvolatile behaviour with a narrow dispersion of ON/OFF ratio, and this study investigated the potential of TiO₂ nanotubes in next-generation memory devices.

The growth of TiO₂ nanotubes on FTO was carried out using an electrochemical deposition process. Electrodeposition was carried out using an aqueous solution containing (NH₄)₂TiF₆ and HBO₃ (99% purity, Wako) by constant potential deposition (−0.8 V) at 90 °C, 30 min, using an Autolab 302N Potentiostat. A standard three-electrode setup in an undivided cell was used. FTO (9.3–9.7 Ω, Asahi Glass Corporation, Japan, 1.1 mm × 26 mm × 30 mm) was used as the working electrode while platinum foil (0.2 mm × 10 mm × 20 mm) was used as the counter electrode. The distance between the two electrodes was 30 mm. The reference electrode was an Ag/AgCl electrode in 4M KCl solution, against which all the potentials reported herein were measured. The morphologies of the samples were observed by field-emission scanning electron microscopy (FESEM; JSM-6335FM, JEOL, with an accelerating voltage of 5 kV) and transmission electron microscopy (TEM, JEOL 2010). To measure the electrical property of the films, Au top electrodes were patterned and deposited by sputtering using a metal shadow mask. Voltage–current curves of the films were measured using an Autolab 302N electrochemical workstation controlled with Nova software. During measurement, the working electrode and sensor electrode were connected to the top Au electrode, and the reference and counter electrode were connected to the FTO substrate.

FESEM and high-resolution TEM (HRTEM) images provided insight into the structure and morphology of the TiO₂ nanotubes. As shown in figure 1, TiO₂ nanotube arrays exhibit high orientation perpendicular to the substrates, with average external diameter of 150 nm and wall thickness of 30 nm. Moreover, a dense layer with small nanoparticles on the surface of the substrate can be found, which indicates that the TiO₂ nanotubes were grown from this dense layer. HRTEM images in the inset of figure 1 further confirm the tube walls are comprised of crystalline titania, with a measured lattice spcaing of 0.189 nm, assigned to be the (1 0 1) plane of the anatase phase.

Zoom In Zoom Out Reset image size

Figure 2(a) shows the typical I–V characteristics of Au/TiO₂/FTO capacitor, measured by sweeping voltage, at a speed of 1 V s⁻¹, in the sequence of 0 → 3 → 0 → −3 → 0 V. During the measurements, the bias voltages were applied on the top electrode with bottom electrode grounded, and neither a forming process nor a current compliance was necessary for activating the memory effort. When a sweep voltage from zero to positive was applied (the arrow denoted '1' in the figure), a sudden resistance decrease from high-resistance status (HRS) to low-resistance status (LRS) was observed at about 2.5 V. As the applied voltage swept from positive to zero, the LRS was maintained. Sweeping from negative to zero (the arrow denoted '4' in the figure), an increase in resistance was observed at about −1.3 V, and the difference of set/reset might be caused by the asymmetric structure of bottom and top electrodes. The endurance properties of the devices were investigated by applying electrical pluses. During the switching process, a 0.2 ms voltage pulse of 4 V switches the device to the on-state (LRS), while a 0.2 ms voltage pulse of −4 V is used to turn the device state back to the off-state (HRS). The evolution of resistance of the two well-resolved states in 7000 cycles is shown in figure 2(b). Under the pulse condition, both the HRS and LRS states are stable and the resistance fluctuations are very small within the range of experimental error. The resistance ratios of HRS to LRS are about 9 times, without degradation up to 7000 cycles.

Zoom In Zoom Out Reset image size

The mechanisms of resistive switching in TiO₂ are interesting but still controversial. The filamentary mechanism has been reported to explain the resistive switching behaviour in TiO₂ films [16]. According to the filament theory, resistive switching effect originates from the formation and the rupture of conductive filaments, which may be realized by the electrical migration and thermal diffusion of defects, such as oxygen vacancies [17]. Oxygen migration model could be important for understanding the origin of the resistive switching effect, although the direct experimental evidence is still needed. According to this model, the oxygen vacancies are the active agents for the resistance switching effect. The mobility of oxygen vacancies is enhanced and thereby their pile-up near metal electrodes is caused under the applied electric field during a switching operation. Thus, a local change in oxygen vacancy concentration near the electrodes results in the resistive switching. The LRS can be imagined as a conductive path between the two electrodes that turns on the devices. The HRS may be described as the destruction of this conductive path as the devices are turned off. Accordingly, it is expected that the TiO₂/FTO junction plays a dominant role in the hysteretic I–V behaviour. In our experiment, we think the conducting filaments which consist of oxygen deficient TiO₂ nanotubes formed or ruptured during switching, also extra oxygen within FTO electrode (bottom electrode) provides bigger chance to enhance the endurance phenomenon. Thus the switching properties of the device might be described as the formation and rupture of the filamentary pats via generation and recovery of oxygen vacancies and oxygen ions. Furthermore, polycrystalline phase of FTO might act as the reservoir of defects such as oxygen vacancies and oxygen ions simultaneously. In these studies, the conducting filament phenomenon and role of FTO interface are suggested to present their roles in switching properties. Based on the experimental results, it should be taken into consideration that the oxygen vacancies can introduce current–voltage hysteresis among the TiO₂/FTO interface.

To investigate the existence of the defects, the x-ray photoelectron spectroscopy (XPS) were employed, as shown in figure 3. The binding energy for Ti 2p_3/2 and 2p_1/2 was observed at 458.9 and 464.7 eV, which were typical XPS spectra of Ti⁴⁺ in TiO₂ [18]. Moreover, no trace of Ti³⁺ or Ti²⁺ was found, whose Ti 2p_3/2 locate at 457.6 eV and 456.4 eV, respectively. As to the O1s spectrum, it can be fitted by three Gaussian components at 530.0, 531.3 and 532.9 eV. The binding energy at 530.0 eV can be assigned to oxygen bound to Ti⁴⁺ ions in TiO₂, while binding energy locating at 531.3 eV can be attributed to oxygen deficiencies of TiO₂. The small shoulder at 532.9 eV implies that the surface is partially covered with hydroxide OH groups. It seems to be obvious from XPS studies that there are local oxygen vacancies in TiO₂ nanotubes. Therefore, conducting filaments in TiO₂ nanotubes might be created by the alignment of the oxygen vacancies under an applied electric field. As the mobility of defects on the surfaces is much higher than that in the crystal, oxygen vacancies can easily condense to form tiny filaments on the wall of the TiO₂ nanotubes. The gathering of these tiny filaments gives rise to the formation of straight and extensible conducting filaments along the direction of vertically aligned TiO₂ nanotubes. As current conduction in straight filaments of nanotubes is more stable than that in branched thin films, a superior stability in the resistive switching behaviour can be achieved in TiO₂ nanotube arrays.

Zoom In Zoom Out Reset image size

In summary, oriented TiO₂ nanotubes were successfully fabricated on FTO substrates and their bipolar resistive switching characteristics were observed. The distinct geometry of TiO₂ leads to the formation of straight and extensible conducting filaments along the walls of TiO₂ nanotubes, resulting in stable resistive switching behaviour with narrow dispersion in the resistance states. This work demonstrates that TiO₂ nanotubes prepared by this method have the potential for next-generation nonvolatile memory applications.

This research was sponsored by the Australian Research Council (ARC) Discovery Project (DP110102391).