Unidirectional information flow in a nanomagnetic metamaterial

TL;DR

This paper introduces a new class of artificial spin ice metamaterials with inherent directionality, enabling unidirectional information flow and improved memory capabilities for magnetic neuromorphic computing.

Contribution

The authors develop a framework for non-reciprocal influence in nanomagnets, discovering directional geometries and experimentally demonstrating reconfigurable unidirectional domain growth.

Findings

Discovery of directional ASI geometries with inherent non-reciprocity

Experimental demonstration of reconfigurable unidirectional domain growth

Enhanced memory capabilities in reservoir computing using directional ASI

Abstract

Artificial spin ice (ASI) are metamaterials composed of interacting nanomagnets. Although ASI hold promise for low-power computing, the ability to transmit information through these two-dimensional systems has been limited. Inspired by non-reciprocal transport in nature, we develop a framework for non-reciprocal influence between nanomagnets. Using the framework we discover a family of ASI geometries with inherent directionality. Directional ASI have the property that, when driven by an external field protocol, domains grow and reverse in the same direction, illustrating an emergent non-reciprocity of the system. Combining growth and reversal results in unidirectional domain movement through the metamaterial. We focus on one member of the directional ASI family, and demonstrate unidirectional domain growth experimentally. Furthermore, we show that the direction of growth is reconfigurable by tuning the external field strengths. Finally, we demonstrate how the directionality of the system significantly improves memory capabilities in a reservoir computing framework. Our work is the first demonstration of an ASI with inherent directionality, offering a magnetic computing platform that combines memory and computation within a single neuromorphic substrate.