Spintronic Memtransistor Leaky Integrate and Fire Neuron for Spiking Neural Networks

Abstract

Spintronic devices based on DWss and skyrmions have shown significant potential for applications in energy-efficient data storage and beyond CMOS computing architectures. Based on the ferromagnetic multilayer spintronic devices, a magnetic field-gated and current-controlled spintronic Leaky Integrate-and-Fire (LIF) neuron with memtransistor properties is showcased. The memtransistor property allows for tuning firing characteristics through external magnetic fields and current pulses. A LIF neuron model is developed based on measured characteristics to integrate the device into system-level Spiking Neural Networks (SNNs). The scalability of the neuron device is confirmed with the micromagnetic simulations in a domain-wall magnetic tunnel junction device. When integrated into SNN and convolutional SNN frameworks, the device achieves classification precision above 96%. The study highlights the device's potential as a neuron in hardware SNN architecture-based neuromorphic computing applications, combining memtransistor properties of the device and high pattern classification accuracy. The results demonstrate a promising path toward developing energy-efficient and scalable neural networks.