# Spin–Orbit Torque-Driven Perpendicular Magnetization Switching for Artificial Synapses in Co/Ho Multilayer Systems

## TL;DR

This paper explores using Co/Ho multilayer systems for artificial synapses, leveraging spin-orbit torque to enable efficient and stable magnetization switching for neuromorphic computing.

## Contribution

The study introduces a Co/Ho multilayer system with high SOT efficiency and stable perpendicular magnetization for artificial synapses.

## Key findings

- Co/Ho multilayer structures maintain high perpendicular magnetic anisotropy in thick layers.
- Antiferromagnetic coupling between Co and Ho enhances spin-orbit torque efficiency with a spin Hall angle of 0.22.
- Multistate magnetization switching is demonstrated, suitable for simulating synaptic weight updates in neural networks.

## Abstract

Spin–orbit torque (SOT)-based spintronic devices have emerged as a preferred candidate for next-generation artificial synaptic devices due to their advantages of non-volatility, high speed, and low power consumption. The development of high-performance SOT-based artificial synaptic devices relies on the breakthrough in SOT-driven magnetization switching, wherein the performance regulation and structural design of the magnetic layer are the core critical factors. In this work, the Co/Ho multilayer system is employed as the magnetic layer to investigate its SOT-driven magnetization switching characteristics and application potential in artificial synapses. By regulating the periodic parameters of the Co/Ho multilayer structure, high perpendicular magnetic anisotropy (PMA) can be stably maintained in devices with relatively thick ferrimagnetic layers. Moreover, we elucidate the role of the antiferromagnetic coupling interface between Co and Ho in the multilayer structure in enhancing SOT efficiency and demonstrate the achievement of a high spin Hall angle of up to 0.22. The high SOT efficiency of the system enables it to drive the 8.4 nm-thick magnetic layer to achieve highly stable magnetization switching. Multistate magnetization switching behavior is observed, which can be used to simulate synaptic weight updates in neuromorphic networks, demonstrating the broad application prospects of this system in the field of artificial neural networks.