A bi-functional three-terminal memristor applicable as an artificial synapse and neuron

Area of Science:

• Materials Science
• Neuroscience
• Computer Science

Background:

• Spiking neural networks (SNNs) mimic the brain for efficient spatiotemporal data processing.
• Two-terminal memristors offer synaptic and neuronal functions but face operational challenges.
• Three-terminal memristors (3TMs) present a potential solution for integrated neuromorphic computing.

Purpose of the Study:

• To develop and characterize a three-terminal memristor (3TM) capable of performing both synaptic and neuronal functions.
• To demonstrate the 3TM's ability to emulate short-term plasticity and learning behaviors.
• To integrate synaptic and neuronal components for efficient SNN hardware implementation.

Main Methods:

• Fabrication of a 3TM based on oxygen ion migration.
• Experimental demonstration of synaptic functions like pair-pulse facilitation and dynamic filtering.
• Emulation of the leaky-integrate-and-fire neuronal model using the 3TM's intrinsic properties.

Main Results:

• The 3TM successfully exhibited short-term plasticity and 'learning-forgetting-relearning' behavior.
• Relearning required less power than the initial learning phase.
• The 3TM emulated neuronal leakage without external components, simplifying SNN hardware.

Conclusions:

• The developed bi-functional 3TM offers a promising platform for SNN hardware.
• This device enhances process compatibility for integrating synaptic and neuronal functions.
• The 3TM contributes to more energy-efficient and time-efficient neuromorphic computing.