Machine Learning and Electronic Structure Methods in the Prediction of the Superconducting Critical Temperature of Metals

We apply the techniques of Machine Learning (ML) to build ML models that can predict the superconducting critical temperatures $T_c$ of metals. The models

were built by employing a database of superconducting materials together with assumed properties of the compounds as obtained from those of their elements, and the density of

states from electronic structure methods which approximately characterizes the material as whether it is a band insulator or not. The materials with the highest values of $T_c$ were then studied using electronic structure methods. Assuming electron-phonon coupling is responsible for superconductivity in these materials, their $T_c$ values were estimated from calculated values of the electron-phonon coupling ($\lambda$) obtained through the use of Density Functional Perturbation Theory (DFPT).