High consistency VO2 memristor for artificial auditory neuron

Abstract

With scalability and diverse device behavior, memristors present potential for executing neuromorphic computation with lower hardware cost and power consumption. However, Low consistency currently limited the application of memristors. Herein, we report a VO2-based memristor fabricated through magnetron sputtering and multiple annealing processes, exhibiting extremely low in both cycle-to-cycle (C2C) and device-to-device (D2D) variations. Further, a Hodgkin-Huxley model neuron circuit based on the extremely high consistency of the devices is established, which achieves auditory neuron perception simulation through spatiotemporal processing of spike signals, allowing for clear differentiation of sound source direction and displaying patterns akin to biological behavior. This easily implemented and highly consistent artificial neuron offers a promising approach for the development of next-generation artificial auditory systems.