2026 CAP Congress / Congrès de l'ACP 2026

Numerical study of Li ion diffusion in Li-M alloys using machine-learning assisted canonical sampling

Not scheduled

15m

U. Ottawa - Learning Crossroads (CRX) Building

Oral not-in-competition (Graduate Student) / Orale non-compétitive (Étudiant(e) du 2e ou 3e cycle) Condensed Matter and Materials Physics / Physique de la matière condensée et matériaux (DCMMP-DPMCM) (DCMMP) T1-4 | (DPMCM)

Speaker

Hoan Tran Van (Institut de recherche sur l’hydrogène, Université du Québec à Trois-Rivières, 3351, boulevard des Forges, Trois-Rivières (Québec) G9A 5H7, Canada)

Description

The incorporation of binary Li-M alloys into anode architectures for solid-state batteries has attracted considerable attention due to their high capacities and effectiveness in suppressing dendrite formation. Numerical simulations of Li-ion diffusion in bulk structures, on surfaces, and at interfaces supports the design of suitable anode architectures with fast Li-ion transport. In this prospect, we investigate the Li-ion migration in Li-rich Li(1-x)Mx alloys (M = Mg, Zn, and Sn; x ≤ 0.5) from first principles. We quantify the vacancy formation energies, and compute the diffusion barriers using nudged elastic band (NEB) calculations. A transition path sampling scheme based on machine-learning-assisted canonical sampling (MLACS) is employed to accelerate the NEB calculations. We show that this scheme allows to reduce computational costs by two orders of magnitude compared to conventional DFT-NEB calculations for supercells containing more than 100 atoms, thus enabling the screening of promising alloy structures for anode applications.