Reconfigurable VO₂ Mott Memristor for Neuromorphic Electronics

To develop memory-centric parallel computing architectures and artificial neural networks, it is essential to integrate numerous building blocks that offer a variety of switching characteristics. Incorporating a variety of materials and building blocks into a single device circuit, however, may result in severe issues, including incompatibility with fabrication processes and mismatching operating ranges. It is advisable to incorporate multiple switching modes into a single, simple device form, as this can enhance operating consistency and reduce the complexity of fabrication. Here, we have explored the potential of this approach through the VO₂-based mott memristor that can possess both analog nonvolatile and threshold volatile switching features. Basically, the VO₂-based mott memristor can easily reconfigure these threshold and nonvolatile switching modes by utilizing the varying voltage polarities. We speculated that the threshold switching is caused by a volatile transition from metal to insulator through phase change, while the nonvolatile switching is related to the conductive filament with a deficiency of oxygen in the oxide thin film. We conducted an analysis of the X-ray diffraction (XRD) patterns and depth-profiled X-ray spectroscopy (XPS) of the deposited VO₂ film. Our findings confirmed the presence of the monoclinic phase and the defective oxygen reservoir layer, which could support our switching mechanisms. Using a single VO₂-based mott memristor, we successfully demonstrated the neural and synaptic functions, such as long-term potentiation and depression, as well as the rate coding and leaky-integrated-and-fire (LIF) in various voltage schemes in the single device unit. The reconfigurable VO₂ mott memristor holds great potential for implementing memory-centric parallel computing architectures and artificial neural networks in a more efficient manner.