DuctGPT cuts AI fusion materials discovery time from months to hours

DuctGPT’s workflow moves from atomic-structure input and training data through computational simulation and property prediction to experimental testing, compressing AI fusion materials discovery from months to hours

(Image courtesy of Ames National Laboratory)

Scientists at Ames National Laboratory have developed an AI tool that reduces the timeline for finding materials needed for next-generation fusion energy systems. Named DuctGPT, the system combines advanced AI with physics-based modeling to help researchers predict which alloy compositions can function in the intense heat, radiation, and mechanical stress inside fusion reactors. The tool runs on a normal desktop computer and, according to the team, can reduce materials discovery time from months to days or hours.

AI fusion materials discovery built on an existing foundation

The project was led by Ames Lab scientist Prashant Singh. Rather than building from scratch, the team took AtomGPT, an existing AI model developed by the National Institute of Standards and Technology, and modified and fine-tuned it using existing materials science data. The result is a platform that lets researchers pose their questions and define parameters in conversational text, a capability the primary source describes as integral to how GPT-based AI programs work.

Researchers can ask DuctGPT to identify element combinations that satisfy specific fusion material criteria and receive candidate compositions in return. Queries run on a normal desktop computer rather than requiring costly supercomputer calculations, which is how the tool reduces discovery time so substantially.

Tungsten’s ductility problem and what DuctGPT addresses

Tungsten sits at the centre of this research. The primary source describes it as one of the best materials for withstanding the very high heat generated during nuclear fusion. It also has a relatively short cooling period and remains radioactive for the shortest amount of time after exposure to nuclear fusion. Its core limitation, as Singh stated, is a lack of low-temperature tensile ductility, which makes it difficult to form into complex shapes.

DuctGPT targets this gap directly. Singh noted the tool allows researchers to query compositions within a desired space, such as tungsten-titanium-zirconium-hafnium, to identify alloys that maintain tungsten’s strength and high melting temperature while improving ductility. Ames Lab has confirmed it holds unique capabilities to synthesise and test predicted materials, verifying that they exhibit the competing properties fusion applications require.

Expanding the platform toward operational conditions

The team is expanding the platform to better predict how materials behave during operation by integrating new data and models. This effort, supported by the ARPA-E CHADWICK program and laboratory investments, aligns with the broader goals of the DOE’s Genesis mission to accelerate the discovery and deployment of advanced materials for future energy technologies.

The research is published in Acta Materialia, authored by Sai Pranav Reddy Guduru, Mkpe O. Kekung, Ryan T. Ott, Sougata Roy, and Prashant Singh.

Stay ahead in the fusion revolution explore more breakthroughs from leading innovators in clean energy technology.