Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
CPP: Fachverband Chemische Physik und Polymerphysik
CPP 41: Charged Soft Matter, Polyelectrolytes and Ionic Liquids I
CPP 41.6: Vortrag
Freitag, 21. März 2025, 11:00–11:15, H38
Machine learning potentials for redox chemistry in solution — •Redouan El Haouari1,2, Emir Kocer1,2, and Jörg Behler1,2 — 1Theoretische Chemie II, Ruhr-Universität Bochum, Germany — 2Research Center Chemical Sciences and Sustainability, Research Alliance Ruhr, Germany
Machine-Learning Potentials (MLPs), which can offer the accuracy of quantum mechanics at a fraction of the costs, have been applied with great success in atomistic simulations of many systems. Still, most MLPs rely on environment-dependent atomic energies, and are thus unable to distinguish different oxidation states of simple ions in solution. Here, we show for the example of ferrous (Fe2+) and ferric (Fe3+) chloride in aqueous solution that this limitation can be overcome with 4th-Generation High-Dimensional Neural Network Potentials (4G-HDNNPs), in which the local atomic energies are complemented with global charges from a charge equilibration scheme. We find that the iron oxidation states match the total number of chloride ions in the system irrespective of their positions. Furthermore, the model captures charge transfers between ferrous and ferric ions, enabling the general simulation of redox chemistry in solution involving different oxidation states.
Keywords: Machine-Learning Potentials; High-Dimensional Neural Network Potentials; 4G-HDNNPs; Charge Transfer; Redox Chemistry in Solution