Predicting electron-phonon coupling constants of superconducting elements by machine learning

Finding new properties of materials by machine learning is an active branch in materials research. Among the various materials groups, superconductivity is not well known despite numerous studies. In this work we have investigated the effect of fourteen important properties in superconductivity on the elements of periodic table and reported these features as priority in order. Then, one of the most important factors in superconductivity, i.e. electron-phonon coupling constant is investigated using machine learning algorithm. In this model, Debye and transition temperatures are as descriptors and the target value, electron-phonon coupling constant is predicted for 28 elements via cross-validation technique. Our predicted electron-phonon coupling constants results are in accordance with the available values by 88% accuracy. So, we are able to build up a model to predict the unavailable electron-phonon coupling constant of elements.