Ultralow Energy Domain Wall Device for Spin-Based Neuromorphic Computing

Area of Science:

• Materials Science
• Spintronics
• Neuromorphic Computing

Background:

• Neuromorphic computing (NC) aims for low-power intelligent devices.
• Spintronic devices offer high endurance for synthetic neurons and synapses.
• Domain wall (DW) devices are favored for low-energy operation in NC.

Purpose of the Study:

• To demonstrate ultralow-energy domain wall (DW) motion for neuromorphic computing.
• To engineer beta-tungsten (β-W) materials to reduce energy consumption in DW devices.

Main Methods:

• Engineered β-W spin-orbit coupling (SOC) material.
• Demonstrated DW motion at current densities as low as 10^6 A/m^2.
• Utilized a meander DW device configuration for controlled DW motion.

Main Results:

• Achieved ultralow pinning fields and reduced current density by 10^4.
• Measured energy consumption of 0.4 fJ for 18.6 μm DW motion.
• Attained an energy consumption of 27 aJ/bit for a 1 μm bit length.

Conclusions:

• Demonstrated a viable path towards ultralow energy consumption in spin-based neuromorphic elements.
• Controlled DW motion in a meander device configuration is suitable for synapse applications.
• Engineered β-W materials show significant potential for energy-efficient neuromorphic devices.