Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

CPP: Fachverband Chemische Physik und Polymerphysik

CPP 15: Poster Session I

CPP 15.2: Poster

Montag, 17. März 2025, 19:00–21:00, P4

Comparing Machine Learning Force Fields for Proton Transfer Dynamics in Solid Acids — •Jules Oumard, Aaron Flötotto, and Christian Dreßler — Technische Universität Ilmenau, Fakultät für Mathematik und Naturwissenschaften, Institut für Physik, Fachgebiet Theoretische Festkörperphysik, Weimarer Straße 32, 98693, Ilmenau

Solid acids are excellent water-free proton conductors and can be used in fuel cells. [1]. The rarity of long-range proton transfer events in ab initio molecular dynamic simulations makes the calculation of converged diffusion coefficients challenging. This can be overcome by accelerating these simulations with machine learning force fields (MLFF). We compare two MLFF approaches: Gaussian Approximation Potentials (GAP) with on-the-fly learning [2] and equivariant graph neural networks [3]. A protocol for fine-tuning GAP models is presented. We evaluate the calculated diffusion coefficients and explain trends in terms of jump rate functions and anion rotation rates.

[1] Mohammad, N. et al. (2016). Journal of Power Sources, 322, 77-92. doi:10.1016/j.jpowsour.2016.05.021

[2] Jinnouchi, R. et al. (2019). Physical Review B, 100(1), 014105. doi:10.1103/PhysRevB.100.014105

[3] Batatia, I., et al. (2022). Advances in Neural Information Processing Systems. https://openreview.net/forum?id=YPpSngE-ZU

Keywords: Machine learning force field; Gaussian Approximation Potential; Message Passing Neural Network; Solid acids; Proton Transfer