Researchers at Skoltech and MIPT in Russia have developed a machine learning-driven approach to accelerate the search for high-performance metal alloys. This method, reported in npj Computational Materials, allows materials scientists to explore a wider range of alloy compositions, potentially leading to the discovery of new materials with superior properties for various industries.
Traditionally, alloy modeling has been computationally demanding, requiring materials scientists to make educated guesses about promising compositions. This new approach, however, leverages machine-learned potentials, which enable rapid calculations and allow for the exploration of all possible combinations up to a certain limit. This exhaustive search eliminates the risk of missing unexpected materials with exceptional characteristics.
The researchers validated their method on two systems: five high-melting-point metals (vanadium, molybdenum, niobium, tantalum, and tungsten) and five noble metals (gold, platinum, palladium, copper, and silver). Their algorithm identified 268 new alloys stable at zero temperature, many of which were not listed in a widely used industry database. For instance, in the niobium-molybdenum-tungsten system, the machine learning approach yielded 12 alloy candidates, while the database contained no three-component alloys of these elements.
The properties of these newly discovered alloys will be further investigated through simulations and experiments to determine their potential for practical applications. The researchers plan to expand their approach to include alloys with different compositions and crystal structures.