Prediction of critical temperature and new superconducting materials
- PDF / 912,120 Bytes
- 10 Pages / 595.276 x 790.866 pts Page_size
- 17 Downloads / 153 Views
Prediction of critical temperature and new superconducting materials Anton Matasov1,2 · Varvara Krasavina3 Received: 16 April 2020 / Accepted: 27 July 2020 © Springer Nature Switzerland AG 2020
Abstract Models for predicting the critical temperature are constructed on the widest base of superconductors. Based on the selection of features, the most important physical parameters were obtained for determining the critical temperature. Models of the best quality were built on the divided data on the number of chemical elements. The resulting models are used to model new superconducting materials. The resulting materials exceed the known used superconductors at a critical temperature. Keywords Superconductivity · Superconductor · Machine learning · Critical temperature
1 Introduction At present, there is no complete theory of superconductivity. What follows is the impossibility of predicting the critical temperature of most known superconducting materials. The critical temperature is the most important parameter that determines the superconducting state, the economic feasibility of using superconducting materials. Also, due to the lack of a general theory, the search for new superconducting materials with a higher critical temperature is mostly intuitive. The generally accepted theory of superconductivity is the Bardeen–Cooper–Schrieffer (BCS) theory [1, 2]. The critical temperature in this theory in the weak coupling limit depends on the Debye temperature, the electron–phonon interaction potential, and the density of electronic states at the Fermi level. These parameters cannot always be accurately measured, and it is also shown that the electron–phonon interaction is insufficient for the appearance of superconducting properties in high-temperature superconductors [3, 4]. Another general theoretical approach to determining the critical temperature is the effect of zeropoint oscillations on the formation of superconducting
particles [5–7]. But this theory works only for ordinary metals and at the moment does not allow predicting the critical temperature of more complex materials. Another alternative approach in predicting the critical temperature may be the use of machine learning methods. There are several works in this direction, where statistical methods are considered and models are constructed to determine the critical temperature of high quality [8–12]. Thus, in the present work, machine learning methods will be used to solve the problem of determining the critical temperature and searching for new superconducting materials. The work [12] uses the most comprehensive data on superconducting materials that contain all classes of superconductors. The disadvantage of [12] is that when creating the database, the influence on the quality of the model of the presence in the data of materials with the same name, input features, but different critical temperatures was not taken into account. This, apparently, is due to the fact that when creating the database, the authors did not take into account the different oxyge
Data Loading...
