A novel framework for approximating resistance-temperature characteristics of a superconducting film based on artificial neural networks

Resistance versus temperature characteristics of superconducting films have been studied for decades, and are still considered an important subject of condensed matter physics. They have recently received increased attention, primarily motivated by electromagnetic metamaterial strategy, which has been used in the implementation of one-dimensional microwave transmission lines with high-temperature superconducting films. In some of the recent works, it has been argued that the physical measurement of these curves is a strenuous and costly process, which becomes tedious when incessantly performed for a wide range of parameters. Contemplating on their significance, in this work, we propose a resistance-temperature curves approximation framework using three different artificial neural networks architectures, and carry out a detailed comparison between the variants in terms of the accuracy they achieve. We demonstrate that the mean-squared error, between the approximated and the physically measured curves, is negligible, which justifies extrapolation of these curves over a wide range of parameters using the proposed framework.

Automated summary

The resistance versus temperature characteristics of superconducting films have been studied for decades and have recently gained increased attention due to the electromagnetic metamaterial strategy. A framework using artificial neural networks to approximate resistance-temperature curves was proposed, and a detailed comparison of the accuracy of different architectures was carried out. The study demonstrated that the mean-squared error between the approximated and physically measured curves is negligible, allowing for the extrapolation of these curves over a wide range of parameters using the proposed framework.