An Investigative Approach to Estimate the Critical Temperature of Superconductors Using Machine Learning

Abstract

Ever since the initial discovery of superconductivity, the fundamental concept and the complex relationship between critical temperature and superconductive materials have been subject to extensive investigation. However, identifying superconductors that exhibit such behavior at normal temperatures remains a significant challenge, and there are still significant gaps in our understanding of this unique phenomenon, particularly regarding the fundamental criteria used to estimate critical temperature. To address this knowledge gap, a plethora of machine learning techniques have been developed to model critical temperatures, given the inherent difficulty in predicting them using traditional methods. Additionally, the limitations of the standard empirical formula in determining the temperature range require the development of more advanced and viable methods. This thesis presents an advanced machine learning-based approach that utilizes the intricate properties of superconductive materials to accurately predict critical temperatures. The proposed model showcases impressive performance, as reflected by the Root Mean Squared Error (RMSE) of 9.68, R 2 score of 0.922, Mean Absolute Error (MAE) score of 5.383, and Mean Absolute Percentage Error (MAPE) score of 4.575, surpassing the performance of existing research works. The findings of this thesis shed new light on the effective implementation of a stacking ensemble method with hyper-parameter optimization, providing a promising avenue for accurate critical temperature estimation. The findings of this study have significant consequences for the decision-making involved in the synthesis of superconductors, as the viability of this complex and resource-intensive process significantly depends on the accuracy of the critical temperature estimation. This thesis contributes to the advancement of superconductivity research by proposing an approximation technique with remarkable precision for the key function of superconductors.