Linkedin-in

Simulating transport properties of solid-state electrolytes via machine-learning models Dr Federico Grasselli

  • FIM-S3 SEMINAR
Speaker
Dr Federico Grasselli
Affiliation
École Polytechnique Fédérale de Lausanne (EPFL), Switzerland
Date
2023-07-24
Time
11:00
Venue
L1.3, Physics Building
Host
Marco Gibertini (FIM, UNIMORE) and Claudia Cardoso (S3, CNR-NANO)

ONLINE: Teams

Abstract:
Transport phenomena, like heat and charge conduction, are of paramount importance in materials science and technology, as they govern the efficiency of devices, fuel cells, and heat exchangers. Specifically, in the current quest for solid-state electrolytes (SSE) for the next generation of batteries, a tradeoff between a high flow of ionic charge - crucial for fast charging and large power needs - and a fast heat dissipation must be found to avoid overheating or explosions. In my talk, I will show how modelling transport phenomena in SSE strongly benefitted, in the last few years, from substantial theoretical advancements on the fundamentals of ab-initio heat and charge transport [1,2], as well as from machine-learning force fields, reaching ab-initio accuracy in the description of interatomic interactions, while keeping a linear scaling with system size [3]. I will showcase these approaches on typical SSE of particular interest thanks to the non-toxicity and large availability of their constituents, namely lithium chlorates and lithium thiophosphates [3,4]. I will conclude by commenting on finite-size effects in the ionic dynamics and thermal transport of superionic conductors [5], and by illustrating recent results on the robustness of the predictions of ML models [6].
 
References:
[1] Bertossa, R., Grasselli, F., Ercole, L., & Baroni, S. (2019). Theory and numerical simulation of heat transport in multicomponent systems. Physical Review Letters, 122(25), 255901.
[2] Grasselli, F., & Baroni, S. (2021). Invariance principles in the theory and computation of transport coefficients. The European Physical Journal B, 94(8), 1-14.
[3] Bartók, A. P., De, S., Poelking, C., Bernstein, N., Kermode, J. R., Csányi, G., & Ceriotti, M. (2017). Machine learning unifies the modelling of materials and molecules. Science advances, 3(12), e1701816.
[4] Pegolo, P., Baroni, S., & Grasselli, F. (2022). Temperature- and vacancy-concentration- dependence of heat transport in Li3ClO from multi-method numerical simulations. npj Computational Materials, 8(1), 1-9; Gigli, L., Tisi, D., Grasselli, F., Ceriotti, M., in preparation.
[5] Grasselli, F. (2022). Investigating finite-size effects in molecular dynamics simulations of ion diffusion, heat transport, and thermal motion in superionic materials. The Journal of Chemical Physics, 156(13), 134705.
[6] Chong, S., Grasselli, F., Mahmoud, C. B., Morrow, J. D., Deringer, V. L., & Ceriotti, M. (2023). Robustness of Local Predictions in Atomistic Machine Learning Models. arXiv preprint arXiv:2306.15638.