Description du projet
L’apprentissage automatique permet de concevoir des molécules dotées de propriétés magnétiques et vibratoires sur mesure
Le couplage spin-phonon, c’est-à-dire le couplage entre les spins électroniques et les vibrations du réseau, entraîne une relaxation de spin. La relaxation de spin constitue un événement particulièrement indésirable dans les dispositifs qui sont tributaires du spin pour le stockage d’informations et l’informatique quantique. La plupart des efforts déployés à ce jour ont permis de résoudre ce problème en maintenant des températures extrêmement basses. Cependant, cette solution ne sera manifestement pas applicable aux dispositifs de demain. Pour soutenir l’innovation dans le domaine du couplage spin-phonon et des molécules magnétiques, le projet AI-DEMON, financé par l’UE, développera un cadre de calcul exploitant l’apprentissage automatique et les premiers principes de la dynamique de spin. Une meilleure compréhension du mécanisme de relaxation spin-phonon permettra de concevoir des prototypes moléculaires dotés de propriétés magnétiques et vibrationnelles sur mesure.
Objectif
As technologies based on semiconductors and ferromagnets are reaching their limits in computational and memory-storage capabilities, new technologies based on spin are emerging as alternative. Magnetic molecules represent the ultimate small-scale magnetic unit that can be synthesized and processed into a device for spintronics and quantum computing applications but their use is confined to very low temperatures. The grand challenge of this proposal is to design magnetic molecules with long spin lifetime at ambient temperature by tuning the main microscopic interaction responsible for spin relaxation: the spin-phonon coupling. AI-DEMON will address this challenge by developing a novel first-principles and machine-learning computational framework able to cover all the essential aspects of the design of new coordination compounds with tailored properties. AI-DEMON has three main objectives, each one representing a major contribution to the field: i) I will unveil the mechanism of spin-phonon relaxation in magnetic molecules by developing a quantitative first-principles spin relaxation theory, ii) I will efficiently explore the chemical space of magnetic coordination compounds by developing a universal machine-learning model able to predict vibrational and magnetic properties, and iii) I will design molecular prototypes with tailored magnetic and vibrational properties by developing generative machine-learning methods. Preliminary results on spin relaxation theory and machine-learning applied to magnetic properties show great promise and set the cornerstone of the project. The use of novel methodologies, such as machine learning and first-principles spin dynamics, represent a strong disruption in the current approach to theoretical modelling and discovery of new magnetic molecules and will propel the field into a new and modern era. Significant impact beyond the field of molecular magnetism, e.g. bio-inorganic chemistry and solid-state qubits, can also be anticipated.
Champ scientifique
- natural sciencesphysical scienceselectromagnetism and electronicsspintronicsmolecular spintronics
- engineering and technologyelectrical engineering, electronic engineering, information engineeringelectronic engineeringcomputer hardwarequantum computers
- natural sciencescomputer and information sciencesartificial intelligencemachine learning
Mots‑clés
Programme(s)
Thème(s)
Régime de financement
ERC-STG - Starting GrantInstitution d’accueil
D02 CX56 Dublin
Irlande