What is it about?

Our paper explores how we can better understand the properties of glass-forming melts, such as viscosity, glass transition and liquidus temperatures, glass density or glass Raman spectra. We developed a powerful tool called i-Melt version 2 that allows us to calculate and predict these properties accurately. By using i-Melt, we were able to uncover fascinating insights into how different cations (positively charged ions) contribute to the density and heat capacity of these glasses and melts. This research has important implications for industries that rely on glass manufacturing and provides valuable knowledge for studying volcanic eruptions and developing new glass materials. Our findings open up exciting possibilities for future research in this field.

Featured Image

Why is it important?

Understanding the properties of glass-forming melts is crucial for several reasons. Firstly, it has significant implications for industries that rely on glass manufacturing, as properties like viscosity and glass transition temperature directly impact the production process and product quality. By accurately predicting these properties with tools like i-Melt version 2, we can optimize manufacturing conditions and improve the efficiency and quality of glass production. Secondly, studying the properties of glass-forming melts is valuable for investigating natural phenomena like volcanic eruptions. The viscosity of magma plays a critical role in determining eruption dynamics, including the explosiveness and flow behavior. By gaining insights into how different cations affect melt properties, we can enhance our understanding of volcanic processes and improve eruption forecasting and risk assessment. Lastly, this research contributes to the development of new glass materials. By comprehensively studying the relationship between cations and properties like density and heat capacity, we can design innovative glass compositions with tailored properties for specific applications. This opens up opportunities for advancements in various fields, including electronics, optics, and energy storage. In summary, this research on calculating the properties of glass-forming melts using i-Melt version 2 is important as it has practical implications for glass manufacturing, contributes to our understanding of volcanic eruptions, and enables the development of novel glass materials with tailored properties. It paves the way for improved manufacturing processes, volcanic hazard mitigation, and technological advancements in diverse industries.


This new version is very exciting. We are now predicting many different properties, for melts with compositions ranging from pure silica to complex alkali and alkaline-earth bearing aluminosilicates. The new model can help predicting properties for calcium aluminates used in the cement industry, silica optical fibers, window glass, and complex multicomponent lavas. Maybe most importantly, our work showcases how deep learning can be integrated with physical and thermodynamic equations to make physics-guided models that reach high accuracy in predictions. This paves the way to a new generation of models for many other applications!

Dr Charles Le Losq
Universite Sorbonne Paris Cite

Read the Original

This page is a summary of: Machine learning modeling of the atomic structure and physical properties of alkali and alkaline-earth aluminosilicate glasses and melts, Journal of Non-Crystalline Solids, October 2023, Elsevier,
DOI: 10.1016/j.jnoncrysol.2023.122481.
You can read the full text:




The following have contributed to this page