Disentangling Coherent and Incoherent Effects in Superconductor Photoemission Spectra via Machine Learning

TL;DR

This paper uses machine learning, specifically CNNs, to analyze superconductor photoemission spectra, successfully identifying bilayer splitting across doping levels and challenging previous assumptions about its behavior.

Contribution

It introduces a CNN-based framework to disentangle spectral effects in superconductors, confirming bilayer splitting across all doping levels and questioning its dependence on doping.

Findings

Bilayer splitting is present across all doping levels.

The magnitude of splitting does not decrease with underdoping.

Machine learning effectively analyzes complex spectral data.

Abstract

Disentangling coherent and incoherent effects in the photoemission spectra of strongly correlated materials is generally a challenging problem due to the involvement of numerous parameters. In this study, we employ machine learning techniques, specifically Convolutional Neural Networks (CNNs), to address the long-standing issue of the bilayer splitting in superconducting cuprates. We demonstrate the effectiveness of CNN training on modeled spectra and confirm earlier findings that establish the presence of bilayer splitting across the entire doping range. Furthermore, we show that the magnitude of the splitting does not decrease with underdoping, contrary to expectations. This approach not only highlights the potential of machine learning in tackling complex physical problems but also provides a robust framework for advancing the analysis of electronic properties in correlated superconductors.