Russian scientists have developed a machine learning system for the accurate calculation of the physical characteristics of molten salts used in molten salt reactors. The system allows determining their density, viscosity, thermal conductivity, and other properties without the need for complex experiments.
The method of targeted search for salts with the desired properties helps to quickly find optimal materials for coolants, minimizing resource-intensive experiments. The effectiveness of the approach was confirmed: calculations for the FLiNaK mixture (lithium, sodium, and potassium fluorides) showed agreement with experimental data.
Molten salt reactors offer to replace the traditional coolant water with molten salts. This allows operating at high temperatures and low pressures, increasing efficiency and safety. However, selecting a suitable combination of salts is difficult due to their corrosiveness and the need to operate at extreme temperatures. To solve this problem, virtual experiments on supercomputers are often used.
Usually, such calculations require complex quantum-mechanical methods. Researchers from Skoltech and the Yekaterinburg Institute of High-Temperature Electrochemistry of the Ural Branch of the Russian Academy of Sciences have simplified the process by applying machine-learned interatomic potentials and molecular dynamics modeling. The new approach provides high accuracy at lower computational costs, opening up opportunities for further development of materials for reactors.
Read materials on the topic:
Rosatom Completes Testing of Unique Stand for Low-Carbon Hydrogen Production
Russian Scientists Create Zinc Nanosheets for Water Purification Systems
Russian Scientists Propose New Method for Determining Atomic Charges in Materials
