Researchers have used machine learning to find new superconductors. A superconductor is a material that carries electric current with zero resistance. These materials usually only work at extremely low temperatures. They are used in quantum computers, brain scanning devices, and magnetic trains. Finding new superconductors is very hard because almost any combination of chemical elements could be one, but very few actually are. Scientists have found over 7,000 superconductors mostly by accident.
A research group called SuperC wants to find a superconductor that works at room temperature. This would greatly reduce global energy use. To do this, the researchers combined machine learning with quantum geometry (see this paper for background information on quantum geometry). Then the researchers used an algorithm to quickly filter many possible materials. The algorithm looked for combinations where electrons form flat bands in a kagome lattice. A kagome lattice is a pattern of interlocking triangles, similar to traditional Japanese basket weaving.
Creating the new materials
After the computer narrowed down the choices, the researchers did detailed calculations on the most promising ones. This process identified two new superconductors. Collaborators at a university then synthesized these materials, meaning they chemically combined raw elements to create the new compounds. They tested the materials and confirmed that they are indeed superconductors.
Normally, predicting new superconductors takes massive computing power, which is why so few have been predicted. By using machine learning to do the first screening, researchers can now process billions of material combinations. This method is expected to greatly speed up the discovery of new superconductors. Finding a room-temperature superconductor would forever change how we consume energy by reducing the heat footprint of the technology sector. With this new method, scientists have a much better starting point to achieve this goal.
The researchers have described the methods and results of this study in a paper published in Physical Review Research.