Transport properties of solid-state electrolytes via machine learning potentials

Lithium ortho-thiophosphate (Li3PS4) is a promising candidate for solid-state-electrolyte batteries.
In this talk, I will show how we build a machine learning potentials (MLP) targeting state-of-the-art DFT references (PBEsol, SCAN, and PBE0), to study the microscopic mechanism of electrical and thermal conductivity of all the known phases of Li3PS4 (ɑ, β and γ), for large system sizes and timescales.
I will discuss the microscopic origin of the observed superionic behaviour of Li3PS4, ruling out the effects of any paddle-wheel effects: the activation of PS4 flipping drives a structural phase transition to a highly conductive phase, characterised by an enhancement of Li-site availability and by a drastic reduction in the activation energy of Li-ion diffusion.
I will show the effect of the phase transition on both the electrical and thermal conductivity temperature behaviour.
Our results show a dependence on the target DFT reference, with PBE0 yielding the best quantitative agreement with experimental measurements.

References:
- L. Gigli, D. Tisi, F. Grasselli, and M. Ceriotti. Chemistry of Materials 36, 1482-1496 (2024).
- D. Tisi, F. Grasselli, L. Gigli, and M. Ceriotti. arXiv preprint arXiv:2401.12936 (2024).