Title:Searching High Temperature Superconductors with the assistance of Graph Neural Networks

Abstract:Predicting high temperature superconductors has long been a great challenge. A major difficulty is how to predict the transition temperature Tc of superconductors. Recently, progress in material informatics has led to a number of machine learning models predicting Tc, which greatly improves the efficiency of prediction. Unfortunately, prevailing models have not shown adequate physical rationality and generalization ability to find new high temperature superconductors, yet. In this work, in order to give a trustable prediction on the unexplored materials, we built a bond-sensitive graph neural network (BSGNN), which is optimized to process the information of chemical bond and electron interaction in the crystal lattice, to predict the Tc maximum of each type of superconducting materials. On the basis of the domain knowledge considered in the data preparation and algorithm design, our model revealed a relevance between the Tc-Tc maximum and chemical bonds. The results indicate that shorter bond length is favored by high Tc, which is in accordance with previous human experience. Moreover, it also shows that some specific chemical elements are favored by high Tc, which is beyond what human experts already knew. It gives a convenient guidance for searching high temperature superconductors in materials database, by ruling out the materials that could never have high Tc.