Using Data Analytics in the Search for New Superconductors



Journal Title

Journal ISSN

Volume Title



It is not easy to discover new superconductors. Scientists spent over 100 years to increase the record critical temperature of superconductors to above the temperature of the boiling point of liquid nitrogen. With the development of computer science, machine learning is becoming a useful tool to solve chemistry problems. In 2018, Dr. Stanev and his colleagues[1] used machine learning methods to build models of superconducting critical temperature. They built classifiers to separate high Tc superconductors and low Tc superconductors. Different regression models based on the two groups of superconductivity data were used to search for unknown superconductors in the Inorganic Crystallographic Structure Database (ICSD). Later, deep learning was used in the process by Dr. Konno and his colleagues.[2] The work herein included cluster analysis in the machine learning process in order to understand the possible patterns in the superconductors by using chemistry variables. Such patterns may guide us in the search for new superconductors. Inorganic data from the Pearson’s Crystallographic Database (PCD) were put into the machine learning models to look for candidate materials. Finally forty candidate materials were identified by the models as possible superconductors.