Signature of Unconventional Superconductivity in the High Temperature Normal State Resistivity

TL;DR

This study uses machine learning to identify a strong correlation between normal-state resistivity and superconductivity in Fe-based superconductors, revealing predictive signals far above the superconducting transition temperature.

Contribution

It demonstrates that machine learning can uncover predictive features of superconductivity in normal-state resistivity over a broad temperature range, suggesting multiple scattering channels are involved.

Findings

Resistivity in 150-300 K range predicts superconductivity.

Signatures of superconductivity are distributed across multiple scattering channels.

Machine learning reveals correlations not evident through traditional analysis.

Abstract

Unconventional superconductivity remains one of the central unsolved problems in quantum materials, and revealing its connection to the normal state is widely believed to be key to uncovering the pairing mechanism. Previous efforts have largely focused on the temperature range immediately above the superconducting transition, where specific scattering channels-such as strange-metal transport-have been identified as sharing a possible microscopic origin with superconductivity. Here, using machine learning, we demonstrate a strong correlation between normal-state resistivity and superconductivity in Fe-based superconductors. Remarkably, the predictive information reside in a wide window of 150-300 K, far above $T_c$ of this family. We further show that the signatures of superconductivity are distributed across multiple scattering channels, which requires further theoretical investigation.