Predicting fracture in disordered network materials using the local intelligent stress threshold indicator

Abstract Network glass fracture occurs as a sequence of elementary events occurring at weak sites in the glass structure. Fracture is a highly complex process that occurs suddenly and without obvious structural or thermodynamic signs prior to the event’s occurrence. We show that a stress threshold value quantified by local mechanical probing highly correlates with nanoscale crack nucleation in a two-dimensional network glass. Subsequently, a neural network-based predictor, the local intelligent stress threshold indicator (LISTI), links the local stress threshold with the undeformed local structural topology. LISTI yields a reliable heatmap indicating soft spots that strongly correlate with the localized initiation and development of the fracture process. Finally, we show that LISTI can be used to find local zones prone to rearrangement in real-measured two-dimensional silica glass structures.