Performance enhancement in NbO2/HfO2 bilayer threshold switching memristor as an oscillatory neuron

Abstract

NbO2-based threshold switching (TS) devices show great potential for artificial neurons in neuromorphic computing. Here, by inserting an HfO2 layer with a thickness of 10 nm, an NbO2/HfO2 bilayer memristor achieved excellent volatile performance with a low threshold voltage (Vth) of 1 V, a small Vth/Vh coefficient of variation, and high selectivity over 102. From the temperature dependence I–V curves of NbO2/HfO2 memristors, it is observed that Vth decreases and OFF current (IOFF) increases with the increase in the test temperature, which agrees well with the behavior of the NbO2 TS memristor originating from the Mott transition mechanism. Furthermore, a leaky integrate-and-fire (LIF) neuron circuit containing an NbO2/HfO2 memristor is built to simulate biological neuron characteristics, and the oscillation frequencies for the neuron can achieve a high value over 4 MHz, which belongs to the fastest ones among TS devices and shows that the design of the Ti/NbO2/HfO2/Pt memristor is feasible for enhancing the response speed and information processing capability for application in neuromorphic computing as the main component of an artificial neuron system.