AI speeds up the search for materials for safe nuclear waste storage

# AI Accelerates the Search for Materials for Safe Nuclear Waste Storage

Russian scientists have found a way to solve one of the most complex problems in nuclear energy — finding materials for reliable nuclear waste storage. The secret lies in an unusual alliance: classical chemistry combined with machine learning. A team of researchers from Skoltech, AIRI, and Sber AI built an accurate phase diagram of technetium carbides using graph neural networks instead of traditional quantum mechanical calculations. The results were published in the prestigious journal Acta Materialia. This is not just a scientific result — it is a demonstration of how AI can fundamentally change the pace of development in materials science.

The task was truly challenging. Technetium carbides are critically important for safe nuclear waste disposal, but studying their properties requires enormous computational resources. The traditional approach relies on quantum mechanical calculations — methods that model the behavior of electrons and nuclei according to physical laws. These calculations are incredibly accurate, but they require days or weeks of computation even on powerful supercomputers. When you need to investigate thousands of potential materials and their combinations, it becomes an almost impossible task. Scientists were stuck in a deadlock: without a complete understanding of material structure, you cannot guarantee its reliability for storing hazardous substances, yet calculating this structure takes an excessively long time.

This is where graph neural networks came into play — models that represent molecules as graphs, where atoms become nodes and chemical bonds become edges. This architecture allows the network to understand spatial relationships between atoms and predict their behavior. Researchers trained the neural network on the results of classical quantum mechanical calculations, allowing it to "learn" the patterns those calculations revealed. After training, the model could instantly predict the properties of new compounds, something that would have taken weeks before. It's like training an experienced craftsman and then asking him to write down a summary of his knowledge — then other people can use that summary instead of undergoing lengthy apprenticeships.

The results are impressive. Thanks to the AI model, researchers built a detailed phase diagram of technetium carbides — a map showing which crystalline structures of the material are stable at various temperatures and pressures. Without losing accuracy, the process accelerated by thousands of times. This means that scientists were able to investigate in a few weeks the volume of materials that would have previously required years. The work proves: machine learning in materials science works not merely as a demonstration toy, but as a full-fledged tool for real scientific tasks.

The significance of this breakthrough extends far beyond technetium. The nuclear industry needs materials capable of withstanding extreme conditions — high temperatures, radiation, corrosion. All of this requires a deep understanding of material structure, and traditional methods work slowly. The application of AI opens the way for systematic search for new compounds with predetermined properties. This could accelerate the development of safer waste disposal systems, and perhaps even the emergence of new generations of nuclear reactors.

The demonstration of how neural networks handle chemistry and materials science signals profound changes in science. AI does not replace scientists, but it ceases to be merely a tool for data analysis. Machine learning becomes a partner in the creative process of finding solutions. By shortening the path from theory to practice, such approaches accelerate not just research, but the very pace of scientific progress in fields where every day matters.