High-dimensional neural network potential for borophene on metallic surfaces
Résumé
Single layer materials have drawn a lot of attention due to their peculiar physical properties (opto-electronic properties, high conductivity, flexibility…). In particular, it has been predicted that boron could exist as a single atomic layer in distinctive crystallographic configurations (allotropes), called borophene -in reference to the carbon equivalent, graphene. Borophene is one of the only 2D material with metallic behaviour, among other interesting properties [1]. Recent studies have focused on the synthesis of such material under various allotropic forms, the obtained allotrope depending on the substrate used and experimental parameters such as synthesis temperature [2-5]. However, the link between the various synthesis parameters and the obtained allotrope is still unclear. To be able to control the synthesis of allotropes selected for their wanted properties, one needs a good understanding of the growth mechanisms and phase transitions at stake in this system. Therefore, a strong theoretical support is needed, with accurate reactive simulations of large systems. However, while ab initio simulations are accurate, they are slow and do not allow studying large systems -and classical Molecular Dynamics require a force field that is not available for this system.
In this work [6], we have developed a new atomic potential using a machine learning approach [7-9], which allows us to explore multiple structural arrangements of borophene allotropes on metal substrates. The developed potential presents the advantage of performing fast simulations with a level of accuracy comparable to ab initio calculations [10]. Here, we will present the methodology to develop this machine learning potential as well as the various borophene allotropes that have been simulated on Ag surfaces, and present our first results on the growth and phase transitions in this system using our machine learned potential.
Domaines
| Origine | Fichiers produits par l'(les) auteur(s) |
|---|---|
| Licence |