Electric Field-Induced Area Scalability toward the Multilevel Resistive Switching

Abstract

Nonvolatile memory devices based on resistive switching with fast switching speed and high density have driven extensive research that is potentially ideal for data-centric applications such as neuromorphic computing. However, due to uncontrolled filament formation, resistive random access memories (RRAM) still suffer from instability and reproducibility. In this study, NiO layer is added to the WO3-based RRAM layer to confine the conducting path for multilevel resistive switching. The current-voltage characteristics show the multilevel resistive switching, which is further confirmed by conductive atomic force microscopy measurements at nanoscale. At low voltages, few localized conducting channels are formed (in small area) within the oxide film while at the higher voltages they are distributed throughout the film and the active surface area increases from 4.5 to 75%. Kelvin probe force microscopy was employed to confirm nanoscale variations in surface potential, which are responsible for spatial current variation. Furthermore, it is found that the charge trapping/detrapping is the main governing mechanism and the conducting paths are formed gradually with increasing applied bias. We propose a strategy to achieve stable and reproducible electric field-induced multilevel memory storage which will be utilized for the development of ultrahigh density multilevel nonvolatile storage and neuromorphic computing.