Superconductors, which are used in MRI machines, nuclear fusion reactors, and magnetic-levitation trains, have the unique property of conducting electricity with no resistance at extremely low temperatures near absolute zero, or -459.67°F.
Efforts to discover a conventional superconductor that operates at room temperature have been ongoing for about a century, but recent advances in machine learning using supercomputers, like Expanse at the San Diego Supercomputer Center (SDSC) at UC San Diego, have accelerated the research significantly.
Senior research scientist Huan Tran from the Georgia Institute of Technology and Professor Tuoc Vu from Hanoi University of Science and Technology have collaborated on Expanse to develop an artificial intelligence/machine learning (AI/ML) approach for identifying potential superconductors more efficiently.
In their recent publication titled “Machine-learning approach for discovery of conventional superconductors” in the journal Physical Review Materials, they focus on chromium hydride (CrH and CrH2) as potential superconductors.
Their research aims to reliably predict superconductivity, especially at zero pressure, where these materials can have significant impacts on human life. Previous AI/ML approaches lacked atomic-level information, which is crucial for accurate predictions.
The researchers built a database with essential atomic-level information about materials, and they “buried” the effects of high pressure on this data to make the database large and diverse. By combining quantum-mechanical computational methods and ML models trained on their database, they accelerated the search for possible superconductors.
The toolkit they developed allowed them to search the Materials Project database, containing nearly 100,000 materials, and discover potential superconductors at zero pressure, including CrH and CrH2. Quantum-mechanical computations verified these findings.
Although the predicted superconductivity for CrH and CrH2 is at relatively low temperatures (about -260°C or -436°F), it is a promising sign for the viability of their program.
Expanse provided them with the computational power they needed, and SDSC’s computational team supported them on the software side, making their work fast and efficient.
Their next steps involve expanding their database of superconductors, including computational and experimental data, to create a top-notch ML platform for discovering superconductors that perform well at ambient pressure and temperatures. Collaboration with experimental experts will be essential to synthesize and test their discoveries.
Finding room-temperature superconductors at ambient conditions could lead to transformative technologies, such as ultra-efficient electricity grids, energy-efficient computer chips, and powerful magnets for levitating trains and controlling fusion reactors.