Towards high-throughput superconductor discovery via machine learning

2021/04/22 by Stephen R. Xie, Y. Quan, Ajinkya Hire +5 · 3 citations
Physics and Astronomy · #cond-mat.supr-con

paper · pdf · doi:10.1088/1361-648x/ac2864

Submitted to J. Phys. Cond. Matt. as part of a ROADMAP article on "Designing Room Temperature Superconductors"

arxiv created 2021/04/22 · arxiv updated 2021/10/25

Abstract

Even though superconductivity has been studied intensively for more than a century, the vast majority of superconductivity research today is carried out in nearly the same manner as decades ago. That is, each study tends to focus on only a single material or small subset of materials, and discoveries are made more or less serendipitously. Recent increases in computing power, novel machine learning algorithms, and improved experimental capabilities offer new opportunities to revolutionize superconductor discovery. These will enable the rapid prediction of structures and properties of novel materials in an automated, high-throughput fashion and the efficient experimental testing of these predictions. Here, we review efforts to use machine learning to attain this goal.

Cited by

• Emergence of the logarithmic average phonon frequency in the superconducting critical temperature formula
• The compliance of the molecular hydride superconductor BiH₄ with the Migdal's theorem
• First-principles calculation of coherence length and penetration depth based on density functional theory for superconductors