Investigations of domain wall dynamics in ladder domain wall devices for neuromorphic computing

Abstract

Neuromorphic computing (NC) has drawn enormous attention from the research community for its efficacy in artificial intelligence. NC aims to emulate the neurobiological features of neurons and synapses, the building blocks of the human brain. To achieve this, researchers have studied various device designs based on spintronic domain wall (DW) devices. However, there is still a need for more efficient device design with optimized pinning strength. In this paper, we have proposed and studied the concept of ladder DW devices. The device resembles the conventional ladder, and the junction between two consecutive DW segments offers a torque originating from PLaplace on the DW, in a direction opposite to the torque due to the spin–orbit torque. We observed a variety of virtues such as DW pinning, damped DW oscillations, sustained DW oscillations, and DW depinning. Here, the pinned domain walls can effectively mimic the functionalities of neuromorphic synapses. Therefore, we studied the DW motion in ladder devices with eight pinning sites and demonstrated a maximum of 12 multilevel magnetization states, a prerequisite for the synaptic devices. Additionally, the DW oscillations (with a maximum of ∼0.25 GHz frequency) are potentially useful for realizing the neurons. We performed micromagnetic simulations and studied the above-mentioned functions for a wide range of parameters. This study paves the way for new device engineering to achieve multi-functional DW devices, useful for NC. Additionally, two-dimensional DW motion in our devices also opens the possibilities for DW-based logic devices.