An Artificial Neuron with a Leaky Fin‐Shaped Field‐Effect Transistor for a Highly Scalable Capacitive Neural Network

Abstract

A capacitive neural network with a capacitive crossbar array that can replace a traditional resistive crossbar array can drastically lower static power consumption during reading operations because a capacitor consumes only dynamic power. Herein, a leaky fin‐shaped field‐effect transistor (L‐FinFET) neuron is fabricated and then applied for use in a highly scalable capacitive neural network with leaky integrate‐and‐fire (LIF) operations that are attributed to a leaky charge trap layer in a gate stack. An additional circuit such as a voltage‐to‐current converter (V–I converter) is no longer required when the L‐FinFET is applied to the capacitive neural network, as the L‐FinFET can directly accept a voltage signal from capacitive synapses. Furthermore, a reset circuit is not necessary given the ability to spontaneously restore to the initial state owing to the leaky charge trap layer. A highly scalable capacitive neural network is realizable due to the size‐reduction ability of the L‐FinFET and the simplified circuit. Finally, an entirely hardware‐based capacitive neural network with the L‐FinFET is demonstrated for the recognition of a simple pattern.