Predicting doped Fe-based superconductor critical temperature from structural and topological parameters using machine learning

Abstract

Recently, Fe-based superconductors have shown promising properties of high critical temperature and high upper critical fields, which are prerequisites for applications in highfield magnets. Critical temperature, Tc, is an important characteristic correlated with crystallographic and electronic structures. By doping with foreign ions in the crystal structure, Tc can be modified, which however requires significant manpower and resources for materials synthesis and characterizations. In this study, we develop the Gaussian process regression model to predict Tc of doped Febased superconductors based on structural and topological parameters, including the lattice constants, volume, and bonding parameter topological index H31. The model is stable and accurate, contributing to fast Tc estimations.