Emerging memristive artificial neuron and synapse devices for the neuromorphic electronics era

Area of Science:

• Materials Science
• Computer Engineering
• Neuroscience

Background:

• The exponential growth of data necessitates energy-efficient computing solutions.
• The 'von Neumann bottleneck' limits traditional computing architectures.
• Neuromorphic electronics, mimicking biological systems, offer in-memory computing capabilities.

Purpose of the Study:

• To review recent advancements in emerging memristive devices for artificial neuron and synapse applications.
• To discuss the physics and characteristics of various memristive switching mechanisms.
• To propose a universal benchmark for evaluating artificial neuron and synapse devices.

Main Methods:

• Review of recent literature on memristive devices for neuromorphic computing.
• Discussion of device physics including valence changing, electrochemical metallization, phase changing, interfaced-controlling, charge-trapping, ferroelectric tunnelling, and spin-transfer torquing.
• Proposal of a benchmark focusing on spiking energy consumption, standby power consumption, and spike timing.

Main Results:

• Identified key memristive device types and their underlying physics.
• Established a benchmark for evaluating artificial neuron and synapse performance.
• Addressed challenges and provided design guidelines for device optimization.

Conclusions:

• Memristive devices show significant promise for energy-efficient artificial neurons and synapses.
• The proposed benchmark facilitates standardized evaluation and comparison of neuromorphic devices.
• Future research should focus on intra-device and inter-device design for enhanced neuromorphic applications.