A Novel Approach to Implementing Artificial Thalamic Neurons with Ferroelectric Transistors

Abstract

Artificial thalamic neurons offer significant potential for medical treatment and neuromorphic computing applications. Their implementation with CMOS technology typically requires a large number of transistors and capacitors, leading to increased power consumption and reduced integration density. This work presents an artificial thalamic relay neuron using only five identical ferroelectric Schottky barrier field-effect transistors (Fe-SBFETs) based on silicon CMOS technology, forming a double inverter and a sensing transistor. The ambipolar switching behavior of Fe-SBFETs not only supports both excitatory and inhibitory synapses with a single device but also allows for the construction of inverters with just two identical transistors. The fabricated thalamic neuron exhibits leaky integrate-and-fire-or-burst (LIFB) functionality with self-resetting capabilities. The artificial neuron enables the device to generate spikes with a reset time of 10 µs, a spike frequency of 8.3 kHz, and an average energy loss of 40 pJ spike⁻¹. The successful implementation of artificial neurons is able to develop low-power, compact neural networks with relatively high operating frequencies.