Abstract
TeO₂-Na₂O glass and glass-ceramic materials have received interest thanks to their interesting properties, such as their low glass transition temperature and their thermal and chemical stability. These advantageous properties make TeO₂-Na₂O glassy materials promising for several technological applications. In additions, when V₂O₅ is added to the system, it leads to a high electrochemical performance of the cathode material that shows a mixed ionic and electronic conductivity suitable for Na cathode materials. Furthermore, the glassy nature of the system minimizes the issues of grain boundaries when the cathode is interfaced with an electrolyte. Nevertheless, while there are only few reports in the literature on binary TeO₂-Na₂O and TeO₂V₂O₅ glasses, there is no study on (Na₂O)_x-(V₂O₅)y-(TeO₂)_1-x-y glassy systems. In this work, we resort to first-principles molecular dynamics (FPMD) to model binary (TeO₂)_1-x -(Na₂O) glasses with 0.1 ≤ x ≤ 0.4 and ternary (Na₂O)_x (V₂O₅)_y-(TeO₂)_1-x-y glassy systems with x = 0.10 and 0.40 ≤ y ≤ 0.60. Our models are validated against experimental results obtained at IRCER lab. In addition, we complement our investigation by resorting to machine learning techniques to produce interatomic potentials based on the FPMD data. Specifically, Gaussian approximation potential (GAP) scheme is adopted to fit interatomic potentials for both binary and ternary systems. The obtained potentials allow us to access extended space and time scales while retaining, to a large extent, the first-principles accuracy. Our results provide a comprehensive description of the glassy structures in close relation to the Na dynamical properties.

Acknowledgments
This work was supported by the French ANR via the AMSES project (ANR-20-CE08-0021) and by région Nouvelle Aquitaine via the CANaMIAS project AAPR2021-2020-11779110. Calculations were performed by using resources from Grand Equipement National de Calcul Intensif (GENCI, project No. 100926). We used computational resources provided by the computing facilities Mésocentre de Calcul Intensif Aquitain (MCIA) of the Université de Bordeaux and of the Université de Pau et des Pays de l’Adour.