Chaotic Proliferation of Relativistic Domain Walls for Reservoir Computing

TL;DR

This paper demonstrates how chaotic patterns of multiple magnetic domain walls in antiferromagnets can be harnessed for ultrafast, energy-efficient reservoir computing, leveraging their complex dynamics and high short-term memory.

Contribution

It introduces a novel method to generate chaotic domain wall patterns for reservoir computing using high-speed driven antiferromagnetic domain walls.

Findings

Complex domain wall patterns form from a single seed wall at high speeds.

The resulting magnetic textures exhibit chaotic spatiotemporal dynamics.

The reservoir shows high short-term memory and nonlinearity for inputs.

Abstract

Magnetic domain walls in antiferromagnets have been proposed as key components for faster conventional information processing, thanks to their enhanced stability and ultrafast propagation. However, how non-conventional computing methods like reservoir computing might take advantage of these properties remains an open question. In this work, we show how complex domain wall patterns can form through the proliferation of multiple domain walls from the energy stored in a single seed domain wall driven to move at a high speed close to the relativistic limit. We demonstrate that the resulting magnetic texture, consisting of up to hundreds of domain walls with an overall conserved topological charge as the initial seed domain wall, can possess chaotic spatiotemporal dynamics depending on the strength of staggered spin-orbit field induced via applied current. These findings allow us to design a multiple-domain-wall reservoir with high short-term memory and nonlinearity with respect to spin-orbit field inputs, that is suitable for ultrafast, energy-efficient, non-conventional reservoir computing.