Mechanism Analysis and Highly Scaled Aluminum Nitride-Based Self-Rectifying Memristors

Storage-class memory and advanced neuromorphic-related applications place an urgent requirement on large-scale, highly compacted memristor array. However, crosstalk issue constrains the addressing accuracy and effective data scale of current array-level memristor, which is detrimental to the industrial realization. Herein, scalability and physical processes of Al/AlN/W self-rectifying memristor are investigated. The devices exhibit improved rectification ratio from 94 to 2600, switching ratio from 286 to 6099, higher uniformity, reliability, and nonvolatility by scaling the size from 10 to 5 mu m. The application of aluminum nitride as active layer material not only makes the electrical properties more sensitive to the external electric field, but also dominates the switching process by constructing nitrogen vacancy conducting filaments, as demonstrated by a combination of scalability and mechanistic analysis. The high-scaled AlN-based self-rectifying memristor exhibits a maximum effective array size of N = 14 372, which is 53 times enhancement compared to 10 mu m size cells. The Al/AlN/W self-rectifying memristor is confirmed to have a strong ability to suppress sneak currents and has the potential to serve large-scale array-level applications.