Elastic Modulus Prediction in Bioactive Glasses
The development of bioactive glasses with improved mechanical performance is often limited by the high cost and time associated with experimental synthesis and mechanical testing. In particular, establishing reliable structure–property relationships for elastic properties remains a major challenge due to the vast compositional space of these materials.
To address this limitation, we combined molecular dynamics simulations and machine learning techniques to predict the elastic modulus of bioactive glasses in the SiO₂–Na₂O–CaO–P₂O₅ system, focusing on compositions around the bioactivity envelopes of the classical 45S5 bioglass and its variants with 4, 5, 6, and 7 wt% P₂O₅. High-throughput molecular simulations were used to generate a consistent and physically grounded database of elastic modulus values, which served as the foundation for training and validating machine learning models, as reported in Journal of Non-Crystalline Solids (DOI: 10.1016/j.jnoncrysol.2023.122507).
Several machine learning algorithms were evaluated to capture the nonlinear relationship between glass composition and elastic modulus. Among them, ensemble-based models and neural networks showed excellent predictive performance, with coefficients of determination (R²) above 0.91, demonstrating their suitability for large-scale screening of bioactive glass formulations beyond those explicitly simulated.
To facilitate reproducibility and promote reuse of these results, we provide the following resources for download:
These resources enable researchers to reproduce the published results, explore new bioglass compositions, and directly integrate the trained models into data-driven workflows for the design and optimization of bioactive glasses with enhanced mechanical performance.