Computational study of fracture in lithium aluminosilicate glass-ceramics

Crystals in glass-ceramics can induce crack deflection and diversion to impede crack propagation and consequently improve the fracture toughness compared to that of the parent glass. However, the structural details of these phenomena are not yet fully understood. Here, we simulate the fracture behavior of lithium aluminosilicate glass-ceramics with different crystallinity using molecular dynamics. We analyze the fracture data using a classification-based machine learning method to compute the so-called 'softness' metric. We find that the tendency of Al atoms to undergo bond switching events increases with the increase in the softness of Al atoms, which is manifested by frequent bond switching and rearrangements of Al ions during the fracture process. Cracks tend to expand along the path of the 'soft' Al atom distribution. These results can help guide the future design of optimized microstructures of glass-ceramics with high fracture resistant.