Abstract
The study investigates oxide thin films with mixed electron-ion conductivity, which display neuromorphic behavior when metal is incorporated into a cross-point structure. This device, comprising metal oxide-metal multilayers, forms a Schottky barrier between the electrode and the oxide layer, creating a localized conductive filament between the electrodes. Neuromorphic resistive switching memorization is achieved through the electron and oxygen vacancy migrations at the Schottky barrier that correspond to changes in synaptic potential. The change in current value caused by the application of pulsed voltage emulates the memory mechanism of the brain, and memory can be controlled by changing the learning interval [1]. Previous studies have shown that the Pt/Ti0.99Sc0.01O2- δ (Rutile)/Pt multilayer device with electron-proton mixed conduction does not form the conductive filaments and exhibits neuromorphic learning and forgetting functions by local proton migration near the Schottky barrier [2]. However, the drawback of Rutile-type TiO2 is in the low resistive state, which has a high amount of oxygen vacancy and high electron conductivity with deposition temperature. On the other hand, Anatase-type TiO2, a wide-gap semiconductor and a metastable phase of TiO2, can be deposited at low temperatures. Therefore, it is possible to control electronic conductivity and provide mixed electron-ion conductivity. In this study, we deposited Rutile-type and Anatase-type TiO2 thin films to replicate neuromorphic function for electrical properties and applied pulsed voltage to Pt/Ti0.99Sc0.01O2- δ (Rutile/Anatase)/Pt multilayers. These films were deposited on Al2O3 (0001) substrates using RF magnetron sputtering, with Rutile type deposited at 600℃ and Anatase type at 300℃. We measured crystal structure, temperature, and oxygen partial pressure dependence of electrical conductivity, pulse voltage response, and electronic structure. The XRD pattern results confirmed that monolayer Rutile and Anatase thin film could be created on an Al2O3 (0001). The oxygen partial pressure dependence indicates mixed electron-ion conductivity, suggesting a strong relation between brain-type property and thin film's electron-ion mixed conductivity.Figures (a) and (b) show the current curves with time for the Pt/Ti0.99Sc0.01O2- δ (Rutile, Anatase)/Pt multilayer when electrical pulses of 0.8 V were applied for 4 s at the interval time of 80 s and 13 s, respectively. An applied voltage during the interval time was 0.1 V. Each upper panel is the input pulse voltage. Each lower panel is the relative output current versus time. In Figure (a), the output current after the pulse input rapidly decays, and the minimum current remains unchanged, indicating STM. In Figure (b), the relative output current after the pulse input slightly increases with the input pulse, indicating the learning of LTM. The current increase was 0.36% for the Rutile-type TiO2 and 6.45% for the Anatase-type, indicating that learning of LTM in Anatase-type TiO2 is about 18 times faster than in Rutile-type. According to these results, the relaxation time (τ) was obtained from the current decay curve of LTM functions after pulse application by the Debye relaxation: 1.55~2.18s for the Rutile-type TiO2 and 0.25~0.3s for the Anatase-type. For the Pt/WO3/Pt multilayer device with conducting filament created by redox reaction, τ has been reported to be about ∼0.4 s [3].The above results suggest that the Anatase-type Pt/Ti0.99Sc0.01O2-δ/Pt multilayer reproduces neuromorphic function higher than in Rutile-type and forms an electrical double layer near the Schottky barrier of the interface, it can be processed faster than a multilayer device with conducting filament. Based on the basic properties of Pt/Ti0.99Sc0.01O2- δ/Pt multilayers and the results of various structural and electrical properties, we will report on the mechanism of brain-type properties at the Rutile and Anatase-type interfaces on the day. Reference [1] K. Kawamura, T. Tsuchiya, M. Takayanagi, K. Terabe, and T. Higuchi, Jpn. J. Appl. Phys. 56, 06GH01 (2017).[2] M. Taniguchi, T. Takada, K. Tomiyoshi, T. Wada, D. Nishioka and T. Higuchi, Jpn. J. Appl. Phys. 62, SG1022 (2023).[3] R. Yang, K. Terabe, Y. Yao, T. Tsuruoka, T. Hasegawa, J. Gimzewski and M. Aono, Nanotechnology 24, 384003 (2013). Figure 1
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call