SKM 2023 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
TT: Fachverband Tiefe Temperaturen
TT 63: Superconductivity: Theory
TT 63.10: Talk
Friday, March 31, 2023, 12:00–12:15, HSZ 103
A low-dimensional ML Surrogate Model for Critical Temperature Prediction of Superconductors — •Angel Diaz Carral1, Martin Roitegui Alonso1, and Maria Fyta2 — 1Institute for Computational Physics, Universität Stuttgart, Allmandring 3, 70569, Stuttgart, Germany — 2Computational Biotechnology, RWTH-Aachen University, Worringerweg 3, 52074, Aachen, Germany
A general theory of superconductivity has been the focus of research over the last decades. Machine learning (ML) approaches based on chemical and structural features have been developed in order to predict both the critical temperature (Tc) and potential novel superconducting structures. Nevertheless, the applied ML models lack interpretability; either the feature matrix is reduced via SVD/PCA transformations or augmented with statistical feature generation. Here, we introduce a ML model based only on electronic descriptors derived from the composition formula and the individual elements. We reach a very high accuracy (R2 over 91%) using a considerably reduced descriptor dimensionality, while retaining the physical meaning of the feature space. Our active learning model is efficiently tested in predicting critical temperatures and links to the discovery of new superconducting structures.