Design and control of artificial spin ice

Abstract

Developing sustainable and robust data processing technology is an increasingly important global priority in an energy-scarce world where computational resources limit state-of-the-art applications. Power-hungry generative artificial intelligence systems consume vast amounts of energy and are finding important new use cases along with fast-paced technological progress. One strategy for surpassing the bottlenecks of development is to create new and improved materials with properties that help solve the technological challenges in a more energy-efficient manner. Metamaterials are one group of candidate material systems where an engineered micro- or nanostructure can give rise to tunable material properties beyond those available in common compound materials. These metamaterials can help with the posed challenges by offering computation as a property of the substrate itself; in-materio computation. This thesis concerns one specific group of metamaterials: artificial spin ices, which are magnetic systems composed of engineered configurations of nanomagnets. In artificial spin ices, the magnetic dipolar coupling between nanomagnets on a two-dimensional surface is defined by its geometric configurations. Large arrays of these nanomagnets with a vast number of interactions exhibit exotic emergent properties, such as tunable magnetic order and emergent chiralities, and might also exhibit emergent information-processing properties. The thesis aims to offer new methods of understanding, designing, and controlling these systems, and it builds on three papers. The first paper considers how to efficiently model artificial spin ice systems on a large scale, where emergent properties of the materials will manifest. A versatile and efficient simulator is introduced, the foundation of the rest of the work. In the second paper, the transition between two well-known artificial spin ice geometries is introduced, modeled, and investigated experimentally. The magnetization of the nanomagnets in the investigated artificial spin ices order into antiferromagnetic or ferromagnetic order, and we observe regions of metamaterial domains of particular magnetic orders. We find a controllable antiferromagnetic–ferromagnetic order transition by rotating the elements of a square lattice artificial spin ice. The transition is shown to have specific properties, continuity and coupling-dependence, that require a slight modification of the point-dipole simulator to model reliably. The second paper demonstrates experimentally the predictive power of the simulator. The third paper introduces a new method of controlling artificial spin ice. This method relies on details of the switching mechanism of the nanomagnets, represented by an astroid-like characteristic switching curve. Local control is achieved using globally applied magnetic fields that exploit the inherent dynamics present at domain wall boundaries. This technique represents a new way to interact with artificial spin ice systems that rely on the current microstate of the system and selectively evolves the state only at applied field pulses. The dynamics are governed by each field pulse in the sequence and we call the method astroid clocking. Through these three papers, the thesis introduces practical advances in designing and controlling artificial spin ices. These advances are interesting from a fundamental physics perspective and can help discover metamaterial properties useful for energy-efficient and scalable computation.