Interface-engineered TiO2/MoO3 bilayer memristor with MoN electrode for stable and low-voltage neuromorphic applications

Artificial synapses based on transition metal oxides (TMO) are increasingly seen as promising for brain-inspired computing because their adjustable defect chemistry allows for synaptic data processing and nonvolatile memory storage functionality. Nevertheless, the inherent defect density in monolayer metal oxides frequently falls short of enabling stable and controllable synaptic functionality. In this study, a TiO 2 and MoO 3 bilayer memristor with a thickness approximately 65 nm was fabricated using radio-frequency (RF) magnetron sputtering conducted at ambient temperature. The device demonstrates stable nonvolatile resistive switching (RS) with promising spatial uniformity over 500th DC I-V cycles, coupled with a low read voltage (+0.7 V), extended retention (>10 4 s), and dependable performance at temperatures up to 150 °C. The conduction process is regulated by space-charge-limited current (SCLC), which is influenced by the managed oxygen vacancy levels in the TiO 2 /MoO 3 bilayer. Furthermore, the device effectively replicates essential synaptic functions, including long-term potentiation/depression (LTP/LTD), paired-pulse facilitation/depression (PPF/PPD), and learning-forgetting behavior, thereby closely simulating biological synapses. A device-aware deep learning model further exhibits a classification accuracy of approximately 98% on the MNIST dataset. In summary, the proposed TiO 2 /MoO 3 bilayer memristor with a MoN bottom electrode demonstrates stable resistive switching and reliable synaptic characteristics, highlighting its potential for future neuromorphic computing applications