Artificial LIF Neuron With Bursting Behavior Based on Threshold Switching Device

Abstract

Bio-inspired computing architecture based on artificial neurons and synapses is attracting intensive attention in the field of power-efficient artificial intelligence. Artificial neurons with bio-inspired firing patterns can significantly improve the computing power of neural networks. However, the reported neuron circuits with burst firing lack the changing of interspike interval (ISI) under different amplitudes of input pulses. Here, a compact leaky integrate-and-fire (LIF) neuron circuit based on the Ag/Ti/HfSe2-xOy/Pt threshold switching (TS) device with a feedback module is designed to achieve bio-inspired burst firing. By adjusting the oxidation conditions and embedding an ultrathin Ti layer, the stability, switching speed, and power consumption of the device have been significantly improved, and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${V}_{\text {th}}$ </tex-math></inline-formula> and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${V}_{\text {hold}}$ </tex-math></inline-formula> have been reduced. These improvements are attributed to the synergistic effect of diffusion barrier and point contact caused by the naturally formed uneven TiOx layer at the interface through the embedding Ti layer. Taking advantage of the low <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${V}_{\text {hold}}$ </tex-math></inline-formula> value of the Ag/Ti/HfSe2-xOy/Pt device, the bionic bursting functions have been realized in the LIF neuron, and the spiking regularity of the burst firing can be modulated through the adjustment of the input pulse.