Strength of 2D Glasses Explored by Machine-Learning Force Fields

\begin{abstract}Strength and toughness of glasses are strongly tied to their atomic-level inhomogeneity.The statistical physics of this correlation is often explored by atomistic simulations for the lack of spatiotemporal resolution in experimental studies.However, theoretical exploration is limited by the mysterious glassy structures and the fidelity of interatomic potentials used in simulations.2D silica with all structural units exposed to the environment allows direct observation in experiments.Their failure mechanisms are studied here by developing a neural network force field for fracture (NN-F$^{3}$) based on the deep potential-smooth edition (DeepPot-SE) framework.Representative atomic-level structures from crystals, nanocrystalline, paracrystalline glasses, and continuous random network glasses are investigated.We find that the virials or bond lengths control the initialization of bond-breaking events, creating nanoscale voids in the vitreous network.However, the voids do not necessarily lead to crack propagation due to the disorder-trapping effect, which is much stronger than the lattice-trapping effect in the crystalline lattice.Fracture can be initialized by the coalescence of voids and proceed by bridging, leaving atomistically smooth facets in the crystalline domains and amorphous edges with atomic-level roughness in the glassy phase.These understandings of 2D crystals and glasses sharing the same SiO$_{2}$ chemistry elucidate the critical role of atomic-level structures in determining the kinetics and path selection in material fracture.