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

MM: Fachverband Metall- und Materialphysik

MM 19: Poster Session

MM 19.26: Poster

Dienstag, 10. März 2026, 18:00–20:00, P5

Machine-learned interatomic potentials for hydrogen-affected fracture in silica minerals — •Valentína Berecová^1,2, Martin Friák¹, and Jana Pavlů² — ¹Inst. Phys. Mater., Czech Acad. Sci., Brno, Czech Republic — ²Dept. Chem., Masaryk Uni., Brno, Czech Republic

Flint, a cryptocrystalline form of silica, is known for its ability to fracture into curved surfaces through conchoidal flaking. Although this behaviour has been used for millennia to produce cutting tools, the atomistic mechanisms that enable such edges remain poorly understood. To clarify these processes, we develop a machine-learned Si-O-H interatomic potential trained on high-accuracy quantum-mechanical data, providing near ab initio resolution at computational costs suitable for simulations. Particular attention is given to hydrogen-related interactions: natural flint contains hydration and hydroxylation that can influence fracture pathways and crack propagation. The training dataset includes hydrogen-containing species, surface terminations, strained configurations, amorphous structures and defect environments. The resulting potential is applied to atomistic simulations of fracture in defect-rich silica, offering new insight into conchoidal fracture at the atomic scale and supporting the design of brittle materials with controlled failure behaviour. Financial support from the Czech Academy of Sciences (Praemium Academiae of M.F. and the Strategy AV21 project "The power of objects: Materiality between past and future") is gratefully acknowledged. Computational resources were provided by e-INFRA CZ and IT4Innovations National Supercomputing Center.

Keywords: Machine-learning; Flint mineral; Mechanism of fracture; Interatomic potentials; Silica