A multidisciplinary team at the Johns Hopkins Applied Physics Laboratory (APL) has discovered a new superconductor in just three months using artificial intelligence (AI).
The key to this success came from combining materials science expertise and real data into predictive AI models, which greatly accelerated the discovery timeline for specific materials.
Superconductors, materials that can conduct electricity without losing energy when cooled below a critical temperature, are used in many cutting-edge technologies. “One of the reasons we chose superconductors is that a new one could change the world,” explains Christopher Stiles, a senior researcher in computational materials in APL’s Exploration Research and Development Division.
The new superconductor is a zirconium–indium–nickel alloy with a superconducting transition temperature of about 9 K. Its discovery, which included the creation of several candidate materials, took only three months, clearly demonstrating the revolutionary potential of directed search. AI-enabled materials science. ,
The reason AI is particularly suited to this challenge is twofold. First, the sheer number of possible materials makes the task take an almost unimaginable amount of time for humans. Second, while both humans and computers reason from the known, computers can be trained to systematically sample the unknown. This second reason is a big part of why Stiles and his team chose superconductors as their test case: there are enough known superconductors, and enough information about their material composition, that there is a high potential for new superconductors to occur. Unused constructs can be identified and targeted. with AI.
The team took advantage of large, publicly available data sets of superconductors and other known materials to train their models. While useful, these data sets inevitably contain human bias, as scientists search for new superconductors by making small changes to existing ones rather than making big leaps, which can lead to costly failures.