Translation. Region: Russian Federal
Source: Peter the Great St. Petersburg Polytechnic University –
An important disclaimer is at the bottom of this article.
Scientists from Peter the Great St. Petersburg Polytechnic University have created a unified database of chalcogenide glasses, which contains more than 20,000 records of their compositions and properties. The database is based on information published in scientific studies over the past 50 years, and includes most of the existing experimental results of studying the characteristics of chalcogenide glassy materials. Database chalcogenide glasses is patented and registered by the Federal Service for Intellectual Property (Rospatent). To work with the database, the Polytechnics developed a web interface that allows sorting, exporting and analyzing data by composition and properties. Thanks to this, new opportunities are opening up for the accelerated design of promising chalcogenide glasses – key materials for modern infrared optics, thermal imagers and night vision systems.
Chalcogenide glasses are amorphous inorganic materials in which oxygen atoms are replaced by sulfur, selenium or tellurium atoms. Chalcogenide glasses have attracted attention since their discovery due to their unique properties: due to the absence of oxygen in the structure, they have wide transparency in the infrared range, a high refractive index and a low softening point. The growth of practical interest in these materials in the last five years is associated with the development of thermal imaging systems and a more than threefold increase in the price of single-crystal germanium, the main material for the infrared spectral region up to 14 μm. Due to the great fragmentation and lack of systematization of published data, as well as the lack of a system for displaying the characteristics of glass compositions, the process of developing new compositions with the required set of properties has become more complex. Traditionally, it is based on the analysis of phase diagrams and the construction of local regression models.
To solve this problem, an interdisciplinary group of scientists from the Scientific and Educational Center "Nanotechnology and Coatings" of the Institute of Mechanical Engineering, Materials and Transport and the Higher School of Software Engineering of SPbPU carried out large-scale work to create a unified database. Using large language models (LLM – Large Language Model), scientists aggregated and structured information from more than 1000 scientific publications. In addition to the database itself, using artificial intelligence methods, models were developed to predict the properties of previously unknown glass compositions.
The main result of the work was not just a database, but an entire analytical platform. For the convenience of researchers, a specialized web interface has been developed that allows for prompt data analysis, comparison of results, and export of search results. To simulate the characteristics of glass before the stage of expensive laboratory synthesis, a model for predicting key glass parameters (density, softening temperature, refractive index) was developed based on machine learning models and neural networks. The proposed approach significantly reduces the time spent on developing promising compositions at the initial stage of research. In the future, it is planned to expand the scope of application of the predicted glass parameters, – said the project manager, PhD in Physics and Mathematics Victor Klinkov.
The software package can serve as a basis for the emergence of a new approach to the design of optical systems. The platform lays the foundation for a fundamentally different methodology: now it is possible to design "from the opposite" – from the characteristics required by the system to the targeted synthesis of material with the necessary parameter values. An important aspect of the project is its general availability. The platform creates a single field for scientific work, allowing both novice scientists and experienced specialists to quickly analyze their results in the context of global research practice and plan new projects.
The practical significance of the work lies in expanding the boundaries of understanding the nature of the glassy state using AI tools and in creating prerequisites for implementing these results in industrial optics design systems. Today, there are no direct analogues of the developed platform in Russia.
The work was carried out within the framework of the Blue Sky Research Digital Labs Campus project with the support of the St. Petersburg Foundation for the Support of Innovations and Youth Initiatives. Now scientists are improving the algorithmic support and expanding the functionality of the platform for the international scientific community.
Please note: This information is raw content obtained directly from the source of the information. It is an accurate report of what the source claims and does not necessarily reflect the position of MIL-OSI or its clients.
.