Machine learning doped MgB2 superconductor critical temperature from topological indices

Abstract

Abstract Due to the absence of weak-links in grain boundaries, less anisotropy, and high availabilities at reasonable cost, magnesium boride, MgB2, has been studied extensively in the past decade. It has relatively high critical temperature, which is correlated to crystallographic and electronic structures. Two topological indices, the electric connectivity index and valence energy level connectivity, are characteristics of compound branching. We develop the Gaussian process regression (GPR) model to shed light on the relationship between topological descriptors and superconducting transition temperature for doped MgB2 superconductors. The model is highly accurate and stable, which contributes to fast predictions of superconducting transition temperature.