Predicting critical temperature of superconductors via machine learning methods

Datum

15.09.2025

Zeit

14:00 - 15:00

Sprecher

Ihor Svynarskyi

Zugehörigkeit

Kyiv Academic University, Kyiv

Sprache

en

Hauptthema

Materialien

Host

Grit Rötzer

Beschreibung

Superconductivity has been studied for over a century already, yet understanding the nature of high-temperature superconductivity remains rather challenging. Using the 3DSC database, which incorporates information on crystal structure of superconductors through special descriptors, and integrating a fingerprint of electronic density of states into it, we address this problem with machine learning techniques. We explore dimensionality reduction via Principal Component Analysis, unsupervised clustering of superconductors and finally prediction of the critical temperature with regression models, namely K-nearest neighbours, Random Forest, and Gradient Boosting. Despite the limited data (fewer than 4000 compounds, including around 900 HTSC), our models achieve competitive results, leading the way to the search for new superconductors. .

Links

• Seminars at the IFW

Letztmalig verändert: 15.09.2025, 07:38:07