New AI tool to identify materials for plasma-facing components

Ames National Laboratory has announced a new tool that combines artificial intelligence and physics-based modeling to identify materials that can be used in fusion systems, where materials must withstand intense heat, radiation, and mechanical stress.

In a paper published in Acta Materialia, the researchers said the platform, named DuctGPT, "enables rapid and accurate prediction of ductility across a wide range of refractory multi-principal element alloys."

According to an Ames news release, the team began with AtomGPT, a model developed at the National Institute of Standards and Technology that uses chemical and structural text descriptions to predict material properties, including formation energies, electronic bandgaps, and superconducting transition temperatures. They made modifications to AtomGPT that fine-tunes it using material science data.

DuctGPT works like most GPT platforms, allowing researchers to ask questions and define parameters. The system can then search through a "very large number of element combinations in seconds."

"Now when you ask it, 'I want to design a material for fusion that has all x, y, z properties that are critical for use in fusion reactors. Tell me the combination of elements which satisfy the criteria,' it will give you those combinations of elements with properties," said Prashant Singh, a scientist at Ames and the project lead.

Tungsten's heat conductivity, strength, and high melting point make it a common choice for plasma-facing components in fusion machines, but it lacks low-temperature tensile ductility, which, Singh said, makes it difficult to form into complex shapes.

"With DuctGPT, we can now 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," he said.

An AI-based tool is well-suited for this type of search that depends on a complex interplay of many factors. DuctGPT includes both experimental and computational data on the density of states at the Fermi level, elastic constants, and valence electron concentration, aiming to "capture the fundamental mechanisms governing ductile versus brittle behavior," the researchers wrote.

"DuctGPT is particularly critical for [the] future of suitable RHEAs [refractory high-entropy alloys] search as these alloys are intrinsically brittle yet seen as most promising candidates for fusion due to their superior properties in extreme environment," said Singh in a LinkedIn post.

According to the paper, "Validation against experimental data confirms the model's ability to predict ductility with high fidelity and low uncertainty."