Speaker
Description
Flint, a cryptocrystalline form of silica, is known for its characteristic conchoidal fracture, which produces smooth, curved surfaces and exceptionally sharp edges. Although this behaviour has been exploited for millennia in the production of cutting tools, the atomistic mechanisms governing crack initiation and propagation in flint remain poorly understood. Direct quantum-mechanical simulations can provide the required atomistic accuracy but are limited to relatively small systems and short time scales, making realistic simulations of fracture computationally prohibitive.
To overcome these limitations, we develop a machine-learned Si–O–H interatomic potential trained on high-accuracy quantum-mechanical data, aiming to retain near-ab initio accuracy while enabling simulations at substantially larger length and time scales. Particular attention is given to hydrogen-containing environments, since natural flint contains both molecular water and hydroxyl groups that may modify local bonding and fracture behaviour. The training dataset therefore spans a broad range of relevant structural environments, including hydroxylated and hydrated configurations, different surface terminations, clusters, strained structures, amorphous silica and defects.
The potential will be systematically validated against quantum-mechanical reference data before being employed in large-scale molecular dynamics (MD) simulations. These simulations will be used to investigate crack initiation, bond-breaking processes and crack propagation in chemically heterogeneous silica. The resulting model should provide an atomistic framework for understanding the characteristic fracture behaviour of flint and, more generally, the role of hydrogen-containing environments in brittle fracture of silica-based materials.