Can AI Solve the Rare Earths Problem? Chinese and U.S. Researchers Think So

metamorworks/Shutterstock.com

A research effort funded by China and the U.S. could speed up the discovery of new materials to use in electronics.

A joint U.S.-Chinese research team has shown that artificial intelligence can help find potent new combinations of materials to replace rare earth metals that are key to military technology. 

Rare earths materials drive today’s high-tech batteries and computer chips. It’s possible to engineer new compounds from common materials that can perform as well or better than the rare-earth-based ones found in common devices. But figuring out the right combinations of elements to design, say, new high-ion conductors or other materials useful for electronics, is an enormous task. If you’re looking to make a compound material with just four of the first 103 elements on the periodic table, you’re looking at ten-to-the-12th-power combinations. A tiny fraction of those would work for electronics. That’s where advanced forms of AI are proving themselves useful. 

A team of researchers at the University of South Carolina College of Engineering and Computing and Guizhou University, a research university located in Guiyang, China, with funding from the U.S. and the Chinese governments, have applied an advanced form of artificial intelligence to the task of finding new combinations of elements that could meet future needs for rare earth resources. 

“Considering the huge space of doped materials with different mixing ratios of elements and many applications such as high-temperature superconductors, where six to seven component materials are common, the number of potential materials is immense,” notes the paper published in the June issue of NPJ Computational Materials.

The researchers apply a Generative Adversarial Network, or GAN, to the problem. GANs work like conventional neural networks but with a twist. Whereas a conventional network might look at billions of pictures of, say, faces to differentiate a real face from a fake one, a generative adversarial network works that problem in reverse by pitting two neural networks against one another. So, applied to conductive compound discovery, it would work like this: while the first network tackles the problem of analyzing all the potential element combinations to find a good conducting material, the adversarial network works by taking that conclusion and reversing it, reasoning that ‘if this hypothetical material meets the description of being good for electronics, what combination of elements or other factors led to its existence?’ It infers the rules for combining materials to create new conductive compounds based on a hypothetical inorganic material that performs well. The researchers report that they are able to speed up material search by two orders of magnitude. 

The paper was supported by grants from the U.S. National Science Foundation as well as the Chinese government’s National Major Scientific and Technological Special Project of China. 

Chinese funding of U.S. academic research is facing increasing scrutiny. “For universities, China takes advantage of the commitment to intellectual freedom on campus, which strongly resists government scrutiny of the activities of foreign students in hard science programs and international academic cooperation,” notes a 2018 report from the Hoover Institution. 

In January, the head of Harvard’s chemistry and chemical biology lab was indicted for making false statements about receiving funding from the Chinese government. One of his students, a Chinese national, was indicted for attempting to take samples from the lab back to China. 

Those high-profile instances deal mainly with the failure of some researchers to disclose funding ties to the Chinese government, which is not the case here. (The paper’s corresponding author declined to answer questions about the National Major Scientific and Technological Special Project of China and its funding process.)  But Sen. Tom Cotton, R-Ark., has introduced legislation to basically prohibit Chinese graduate students from studying in any science or technology field in the United States. 

That could make discoveries like this one harder for the United States to realize in the future. Chinese students and researchers play a key role in the advancement of U.S. technological capabilities, Eric Schmidt, former Google CEO and the chair of the Defense Innovation Board, said at the Defense One Tech Summit in June.