Cloud Hosting

Superconductors, known for their ability to exhibit zero electrical resistance when cooled below a critical temperature, have tremendous potential for applications in energy, transportation, and cutting-edge electronics. Researchers from Georgia Tech and Hanoi University of Science and Technology have taken the first step towards incorporating atomic-level information into machine learning pathways to discover new conventional superconductors.

To overcome the barrier of lacking atomic level information, the researchers curated a dataset of 584 atomic structures with over 1100 computed values of and log at different pressures. Machine learning models were developed for and log and used to screen over 80,000 entries in the Materials Project database. Through first-principles computations, the researchers identified two materials that may exhibit superconductivity at a critical temperature of approximately 10^-15K and ambient pressure.

The researchers used the machine learning models to predict superconducting properties for 35 candidates, with six of them having the highest predicted critical temperatures. Further stabilization calculations were required for some candidates. After verifying the stability of two remaining candidates, CrH and CrH2, the researchers calculated their superconducting properties using first-principles calculations. The accuracy of the predictions was validated within 2-3% of the reported values through additional calculations using the local-density approximation (LDA) XC functional.

Additionally, the researchers investigated the synthesizability of the superconductors by tracing their origin in the Inorganic Crystalline Structure Database (ICSD). They found that these materials had been experimentally synthesized in the past, providing hope for future tests to confirm their predicted superconductivity.

In future research, the researchers plan to enhance their machine learning approach by expanding and diversifying the dataset, employing deep learning techniques, and integrating an inverse design strategy for more efficient exploration of materials. They also aim to collaborate with experimental experts for real-world testing and synthesis of high critical temperature superconductors.

See the original post here:
Machine Learning Used to Discover New Superconductors - Fagen wasanni