Artificial Spin Ice for Rewritable Magnonics
Lead Research Organisation:
Imperial College London
Department Name: Physics
Abstract
The key physical concept of this project is that magnetic spin-waves, or their quanta magnons, can act as information carriers and be manipulated for information processing & computation. Conventional computers rely on physically moving particles (electrons), and vast amounts of energy are wasted by ohmic loss and heating induced by electronic transit, both within the logic devices and particularly between the separate logic and storage media. If current trends continue, computation will consume one third of global energy production by 2040, and consequently increasing computational energy efficiency is a critical challenge. Because magnets can transfer information from one device to the next without the exchange of any physical particles and have intrinsic passive data storage, 'magnonics' is in principle orders of magnitude more energy efficient than standard electronics & a promising route to aiding the global energy crisis.
Magnets are used in memory devices as they passively retain information written into them (non-volatile). This project will enable creation of coupled arrays of nanomagnets that can be viewed as both memory and processor where novel circuits can be written and reprogrammed at will. Our ability to accomplish this exploits a technique which we have developed called All-Optical Magnetic Switching (AOMS), allowing controlled writing of any individual nanomagnet in the array with a low-power laser like a Blu-Ray player, plus world-leading expertise harnessing nanomagnetic arrays for spin-wave information processing - including world-first demonstration of magnonic neuromorphic computation in an array of interacting nanomagnets.
Each ferromagnetic nanoisland stores a fixed average magnetization, but the magnetic moment is not completely static, instead precessing around the average direction at characteristic resonant frequencies in the microwave (GHz) range. For a single nanomagnet, the frequency is controlled by its size and shape in the same way that shortening a guitar string changes the note. Coupled arrays of nanomagnets have distinct spectral fingerprints and these can be used for readout of states. The magnonic resonances are also highly sensitive to the magnetic texture of each island, and one of our recent breakthroughs exploits this to prepare bistable vortex & macrospin islands exhibiting far greater functional magnonic flexibility versus conventional all-macrospin systems.
It is already well established from simulations that the exact microstate of the array controls the resonant frequency of the magnons and that we can realise switches and transistor type devices for logic functions where the magnetic state controls whether magnons of a specific frequency can pass through or not.
This project aims to integrate different functional elements and explore prototype magnonic components and circuits. It is highly adventurous, and there are many experimental challenges to overcome to realise fully magnonic computation. For example, a process called damping causes travelling spin waves to attenuate rapidly with both time and distance. This presents a challenge in terms of completing the full computation before information is lost, as well as representing a source of energy inefficiency - though it can be avoided using resonant 'standing wave' magnons with which our scheme also functions. Although it is straightforward to measure these 'standing wave' magnons in a large array, detecting travelling magnons in nanoscale device structures is at the edge of state-of-the-art capabilities. In this project we aim to develop and expand these capabilities, building on our expertise and establish fundamental understanding of the physics of coupling, synchronization, transmission, and loss between different magnonic crystal states, and deliver a fruitful playground to explore novel computation architectures.
Magnets are used in memory devices as they passively retain information written into them (non-volatile). This project will enable creation of coupled arrays of nanomagnets that can be viewed as both memory and processor where novel circuits can be written and reprogrammed at will. Our ability to accomplish this exploits a technique which we have developed called All-Optical Magnetic Switching (AOMS), allowing controlled writing of any individual nanomagnet in the array with a low-power laser like a Blu-Ray player, plus world-leading expertise harnessing nanomagnetic arrays for spin-wave information processing - including world-first demonstration of magnonic neuromorphic computation in an array of interacting nanomagnets.
Each ferromagnetic nanoisland stores a fixed average magnetization, but the magnetic moment is not completely static, instead precessing around the average direction at characteristic resonant frequencies in the microwave (GHz) range. For a single nanomagnet, the frequency is controlled by its size and shape in the same way that shortening a guitar string changes the note. Coupled arrays of nanomagnets have distinct spectral fingerprints and these can be used for readout of states. The magnonic resonances are also highly sensitive to the magnetic texture of each island, and one of our recent breakthroughs exploits this to prepare bistable vortex & macrospin islands exhibiting far greater functional magnonic flexibility versus conventional all-macrospin systems.
It is already well established from simulations that the exact microstate of the array controls the resonant frequency of the magnons and that we can realise switches and transistor type devices for logic functions where the magnetic state controls whether magnons of a specific frequency can pass through or not.
This project aims to integrate different functional elements and explore prototype magnonic components and circuits. It is highly adventurous, and there are many experimental challenges to overcome to realise fully magnonic computation. For example, a process called damping causes travelling spin waves to attenuate rapidly with both time and distance. This presents a challenge in terms of completing the full computation before information is lost, as well as representing a source of energy inefficiency - though it can be avoided using resonant 'standing wave' magnons with which our scheme also functions. Although it is straightforward to measure these 'standing wave' magnons in a large array, detecting travelling magnons in nanoscale device structures is at the edge of state-of-the-art capabilities. In this project we aim to develop and expand these capabilities, building on our expertise and establish fundamental understanding of the physics of coupling, synchronization, transmission, and loss between different magnonic crystal states, and deliver a fruitful playground to explore novel computation architectures.
Publications

Alatteili G
(2023)
Gænice: a general model for magnon band structure of artificial spin ices

Alatteili G
(2024)
Gænice: A general model for magnon band structure of artificial spin ices
in Journal of Magnetism and Magnetic Materials



Dion T
(2024)
Ultrastrong magnon-magnon coupling and chiral spin-texture control in a dipolar 3D multilayered artificial spin-vortex ice.
in Nature communications

Lee O
(2024)
Task-adaptive physical reservoir computing.
in Nature materials

Lee O
(2023)
Perspective on unconventional computing using magnetic skyrmions
in Applied Physics Letters

Mondal A
(2024)
Brillouin light scattering spectral fingerprinting of magnetic microstates in artificial spin ice
in Nano Today

Stenning KD
(2024)
Neuromorphic overparameterisation and few-shot learning in multilayer physical neural networks.
in Nature communications
Description | Coherent Soft X-ray Imaging and Diffraction (CSXID) |
Organisation | Diamond Light Source |
Country | United Kingdom |
Sector | Private |
PI Contribution | Prepared samples and manned beamline experiments |
Collaborator Contribution | Provided 1 week beamtime on I17 beamline (Coherent Soft X-ray Imaging and Diffraction (CSXID)). Also worked with us in advance of beamtime to advise on sample preparation |
Impact | Data collection finished 05/03/204 |
Start Year | 2023 |
Description | Skyrmion Magnonics |
Organisation | RIKEN |
Country | Japan |
Sector | Public |
PI Contribution | Measurement of magnonic properties and development of neuromorphic computing in the samples |
Collaborator Contribution | Providing samples of Skyrmion crystal material |
Impact | https://doi.org/10.1038/s41563-023-01698-8 |
Start Year | 2023 |
Description | Skyrmion Magnonics |
Organisation | Technical University of Munich |
Country | Germany |
Sector | Academic/University |
PI Contribution | Measurement of magnonic properties and development of neuromorphic computing in the samples |
Collaborator Contribution | Providing samples of Skyrmion crystal material |
Impact | https://doi.org/10.1038/s41563-023-01698-8 |
Start Year | 2023 |
Description | Skyrmion Magnonics |
Organisation | University of Tokyo |
Country | Japan |
Sector | Academic/University |
PI Contribution | Measurement of magnonic properties and development of neuromorphic computing in the samples |
Collaborator Contribution | Providing samples of Skyrmion crystal material |
Impact | https://doi.org/10.1038/s41563-023-01698-8 |
Start Year | 2023 |