Parameterization of ReaxFF Potential of Mg/Al/Si/O Interaction and Investigation of Mechanical Properties for S-Glass

New ReaxFF parameters are developed for the description of Mg/Al/Si/O interaction for the Magnesium Aluminosilicate (MAS) glass structure. The training set contains energy curves from equation of state for various Mg/Al/Si/O crystals, valence angle and bond distance scan, and heat of formation for the Mg/Al/Si/O interactions. A semi-automated Genetic Algorithm assisted by Artificial Neural Network is applied for this parametrization. Validation efforts showed the current ReaxFF parameter set can describe the atomistic structure and property of tectosilicate MAS glass including S-glass. Estimated quasi-static modulus of S-glass structure matches well with experimental value. Analysis shows the key of high modulus of S-glass is numerous Mg-BO (Bridge Oxygen) interactions across the Mg-O-AlSi structure. In addition, atomistic origin of high ductility and progressive failure of S-glass is derived from the reconstruction of the atomic structure, forming Mg-BO-Si interactions that delays fracture formation.