A Proposal of Energy Efficient Ferroelectric PDSOI LIF Neuron for Spiking Neural Network Applications

Abstract

The proposed device is a partially depleted silicon on insulator (PD-SOI) with a ferroelectric material as a part of gate stack structure demonstrating the functions of leaky integrate and fire (LIF) neurons with a minimum energy of 9.375 pJ/spike and area of $0.25\ \mu \mathrm{m}^{2}$. A high-k ferroelectric (FE) dielectric in the gate stack improves the energy performance by reducing the subthreshold swing. The scaled down area of the device helps to integrate more LIF neurons in the network. The frequency of spiking increases with increase in input voltage which is also an important function in a biological neuron. Thus, the implementation of this device in neuromorphic systems reduces the power consumption by the electronic devices and improves the overall performance.