Understanding the composition dependence of glass hardness is of critical importance for both advanced glass applications and for revealing underlying fracture mechanisms. We present a topological approach for quantitative prediction of hardness in multicomponent glassy systems. We show that hardness is governed by the number of network constraints at room temperature and that a critical number of constraints is required for a material to display mechanical resistance. Applied to a series of soda lime borate glasses, the predicted values of hardness are in excellent agreement with experimental measurements. Our approach is generally applicable to any network glass and demonstrates the importance of accounting for the temperature dependence of the network constraints.

Addition of alumina to sodium silicate glasses considerably improves the mechanical properties and chemical durability and changes other properties such as ionic conductivity and melt viscosity. As a result, aluminosilicate glasses find wide industrial and technological applications including the recent Corning® Gorilla® Glass. In this paper, the structures of sodium aluminosilicate glasses with a wide range of Al/Na ratios (from 1.5 to 0.6) have been studied using classical molecular dynamics simulations in a system containing around 3000 atoms, with the aim to understand the structural role of aluminum as a function of chemical composition in these glasses. The short- and medium-range structures such as aluminum coordination, bond angle distribution around cations, Qn distribution (n bridging oxygen per network forming tetrahedron), and ring size distribution have been systematically studied. In addition, the mechanical properties including bulk, shear, and Young's moduli have been calculated and compared with experimental data. It is found that aluminum ions are mainly four-fold coordinated in peralkaline compositions (Al/Na < 1) and form an integral part of the rigid silicon-oxygen glass network. In peraluminous compositions (Al/Na > 1), small amounts of five-fold coordinated aluminum ions are present while the concentration of six-fold coordinated aluminum is negligible. Oxygen triclusters are also found to be present in peraluminous compositions, and their concentration increases with increasing Al/Na ratio. The calculated bulk, shear, and Young's moduli were found to increase with increasing Al/Na ratio, in good agreement with experimental data.

The application of machine learning to predict materials’ properties usually requires a large number of consistent data for training. However, experimental datasets of high quality are not always available or self-consistent. Here, as an alternative route, we combine machine learning with high-throughput molecular dynamics simulations to predict the Young’s modulus of silicate glasses. We demonstrate that this combined approach offers good and reliable predictions over the entire compositional domain. By comparing the performances of select machine learning algorithms, we discuss the nature of the balance between accuracy, simplicity, and interpretability in machine learning.