Demonstration of Low Power and Multifunctional Neurons with Only a Few Ferroelectric Transistors

Abstract:
Neuromorphic computing relies on both synapses and neurons. Developing energy-efficient, highly scalable, and compact devices or circuits is therefore in high demand. This work demonstrates the implementation of both artificial synapses and neurons using a single type of ferroelectric Schottky barrier FET (Fe-SBFET). The Fe-SBFET exhibits synaptic properties with short-term memory effects. Notably, we achieve an efficient thalamic neuron using only five Fe-SBFETs, demonstrating leakage-integration-fire and burst (LIFB) functions with an energy consumption of 2.6 pJ per spike at 5.6 kHz. Additionally, a leakage-integration-fire (LIF) neuron is realized with just two Fe-SBFETs, achieving an energy consumption of 0.67 pJ per spike.