Dresden 2026 – wissenschaftliches Programm

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

O: Fachverband Oberflächenphysik

O 5: Focus Session: Mineral-water interfaces I

O 5.4: Hauptvortrag

Montag, 9. März 2026, 11:30–12:00, HSZ/0403

Machine learning exploration of water binding and ice nucleation at silicate and carbon surfaces — •Mie Andersen — Department of Physics and Astronomy, Aarhus University, Denmark

Mineral-water interfaces play a central role in environments ranging from the interstellar medium to Earth’s surface and subsurface. In astrochemistry, the chemical evolution of star-forming regions is controlled by gas-grain interactions on dust grains that may be partially or fully covered by water ice. A key quantity underlying these processes is the binding energy (BE) of adsorbates, which determines desorption and diffusion rates and controls chemical kinetics.

In this contribution, we explore computational strategies to quantify BE distributions at complex, partially ice-covered mineral surfaces. Using graphene and forsterite as model grain substrates, we combine atomistic simulations with machine-learning interatomic potentials to generate realistic surface structures and to efficiently sample BEs. The modelled ice structures range from amorphous structures generated by low-temperature molecular dynamics simulations to highly stable crystalline structures generated by global structure optimization. The resulting BE distributions are analyzed in terms of surface heterogeneity, ice coverage and hydrogen bonding.

While motivated by astrochemical applications, the presented methodology is broadly transferable to mineral-water interfaces under terrestrial conditions. The approach offers a general framework for linking molecular-scale interfacial structure to macroscopic transport and reaction models across disciplines.

Keywords: Machine learning interatomic potential; Astrochemistry; Dust grain; Interstellar ices; Binding energy