Modifying ring structures in lithium borate glasses under compression: MD simulations using a machine-learning potential

Machine-learning potentials (MLPs) are advantageous in modeling boron coordination and three-membered rings (3-rings) in borate glasses. In this study, the prominent ability of MLP enables us to study how boron coordination varies during deformation, and the change in microstructure circumvents the overload stress in lithium borate and lithium borosilicate glasses via molecular dynamics simulations. Under uniaxial and triaxial compressive deformations, some of the threefold coordinated boron, ³B, atoms were altered to be fourfold coordinated boron (⁴B) at a strain range beyond elastic deformations. According to the local deformation analysis, a microstructure at around ⁴B was determined to be rigid, specifically, that around ⁴B forming a 3-ring exhibited the least flexibility. Conversely, the local region at around ³B and oxygen atoms in nonring structures flexibly deformed following the entire deformation. As a result, nonbridging oxygen located in the vicinity of ³B was found to migrate to form ⁴B due to compression. Remarkably, most of the boron atoms became ⁴B, and they formed abundant three membered rings in the borate glasses at a certain strain owing to the triaxial compression. Consequently, it is inferred that such a drastic structural variation can be the origin of high damage resistance of borate glasses.