A low-power artificial spiking neuron based on ionic memristor for modulated frequency coding

Abstract

Neurons encode information through firing spikes with rich spatiotemporal dynamics. Using artificial neuron hardware based on memristors to emulate neuronal firing is of great significance for advancing the development of brain-like computing and artificial intelligence. However, it is still challenging to achieve low power frequency coding in memristive artificial neurons. Here, a low-power ionic memristor based on Pt/HfO2/Ag is reported for artificial spiking neurons. The device is driven by a low bias current and the filament dynamically ruptures and forms, producing oscillated voltage spikes that resemble neuronal spikes. The oscillation frequency increases from 0.5 Hz to ∼2.18 Hz with the stimulation current increasing from 1 nA to 5 nA, enabling the emulation of neuronal frequency-coding function. The low power consumption of ∼70 pJ per pulse indicates that the device is promising for energy-efficient neuromorphic computing applications. In addition, the device is found to be capable of simulating the phasic,adaptive, and burst firing modes of neurons.