Exponential Escape Rate of Filamentary Incubation in Mott Spiking Neurons

Abstract

The resistive switching induced in Mott materials by a strong applied voltage allows for artificial spiking neurons with great potential. However, controlling the resistive collapse is a challenge, as we lack physical understanding of the phenomenon. Here numerical simulations and experiments reveal how thermal and electronic effects jointly contribute to the phenomenon. The resistive collapse is intrinsically stochastic, and more surprisingly it is a Poissonian process with an exponential escape rate, just like the firing of actual biological neurons. This result provides an unexpectedly realistic aspect to the implementation of tomorrow's energy-efficient neurocomputing hardware.