Building an Active Learning System to Find New Alloys
On July 16, a team led by Yu Zou unveiled an active learning system designed to discover novel alloys. This innovative platform can independently select candidates, fabricate them, run tests, and plan experiments. Within just a few weeks, it identified six alloy formulas, including compositions based on nickel, cobalt, and chromium. For context, this approach represents a significant leap in using artificial intelligence to streamline materials discovery, which traditionally relies on extensive trial and error.
Alloy Compositions Uncovered
Among the formulas found, one standout alloy contains:
- 12% nickel,
- 62% cobalt,
- 26% chromium.
This material retains its hardness at temperatures up to 600°C, beating the industry benchmark Inconel 625 by 4.5% in lab tests. Notably, Inconel 625 contains more than ten elements.
Another discovered formula consists of:
- 36% nickel,
- 14% cobalt,
- 50% chromium.
This alloy proved 85% more resistant to oxidation at temperatures reaching 1000°C. The team plans to increase the complexity of their alloys, aiming to incorporate 10 to 12 elements in future iterations.
Ajay Talbot, a graduate student in Zou's lab, noted: 'The challenge you often face when trying to apply AI to materials design is that most machine learning models need a lot of data on material properties to train.'
Talbot added: 'But if you're working in a part of the design space that hasn't been explored yet, that data doesn't exist, so you're essentially flying blind.'
Talbot highlighted that the team found a material that 'resists oxidation extremely well at such high temperatures,' and they aim to push temperatures up to 1200°C in the future. The findings were published in the journal Advanced Manufacturing.
Yu Zou's research marks a major advance in materials science, particularly in creating alloys that can endure extreme conditions. With their promise of reaching 1200°C, these new alloys could unlock new applications in aerospace, energy, and other industries where material reliability is critical. These achievements also underscore the potential of using artificial intelligence in scientific research, which could accelerate the discovery of new materials.