Apart from large-scale industrial glass production, glass is mainly produced using (semi-)empirical methods, including scientific research. This type of production is time-consuming and difficult to reproduce. Chemical compositions are usually based on empirical values, empirical correlations, or trial-and-error methods. Currently available glass databases such as SciGlass [1] (which became open source in 2019, containing about 420,000 types of glasses) and INTERGLAD [2] (a commercial database with around 380,000 types of glasses). While both databases have large gaps in glass properties, as they generally contain incoherent studies or data, they are nevertheless the two largest glass databases available.

To address this problem, the GlasDigital project [3] aims to develop an ontology-based digital infrastructure for high-throughput data-driven glass development in the Plattform MaterialDigital (PMD). By adopting the PMD core ontology, the GlasDigital ontology was developed for glass materials and particularly focuses on robot-assisted melting processes. It also provides glass terminology that aligns with the SciGlass database, facilitating the integration of SciGlass data into our digital infrastructure.

Since SciGlass has become open source, there are no longer any distributors in Germany. Given the importance of SciGlass to the glass community and the many publications on predicting glass properties based on machine learning, we have developed a standalone web version of SciGlass and look forward to collaborations to implement (general) glass models for predicting glass properties and thus accelerate the development of novel glass materials.

[1] https://github.com/epam/SciGlass

[2] https://www.newglass.jp/interglad_n/gaiyo/info_e.html

[3] https://www.materialdigital.de/project/4