Detecting Thermodynamic Phase Transition via Explainable Machine Learning of Photoemission Spectroscopy

TL;DR

This paper introduces an explainable machine learning approach using photoemission spectroscopy data to accurately identify thermodynamic phase transitions in low-dimensional materials, overcoming data scarcity and revealing key spectral features.

Contribution

The study develops a domain-adversarial neural network that detects phase transitions from spectra, effectively compensating for limited experimental data with simulated data and providing interpretability.

Findings

Achieved 97.6% accuracy in identifying superconducting transitions.

Demonstrated the importance of in-gap spectral weight in phase detection.

Enabled spectroscopic identification of fluctuating orders in low-dimensional systems.

Abstract

Identifying thermodynamic signatures of electronic phases, such as superconductivity, is challenging in low-dimensional materials due to strong fluctuations and low probing volume. Spectroscopic methods are often used to identify new bulk phases, but their main measurable quantity -- electronic energy gaps -- is no longer an effective order parameter in low-dimensional and fluctuating systems. Combining angle-resolved photoemission with a domain-adversarial neural network, we report a data-driven method to identify thermodynamic phase transitions solely based on single-particle spectra. We demonstrate 97.6$\%$ accuracy in cuprate superconductor Bi$_2$Sr$_2$CaCu$_2$O$_{8+\delta}$ with strong superconducting fluctuations. This model notably compensates for the scarcity of experimental data by leveraging virtually inexhaustible simulated data. Further, its explainability reveals the crucial role of in-gap spectral weight in detecting phase fluctuations and thermodynamic transitions. Our work pinpoints the spectroscopic signatures of fluctuating orders and enables using spectroscopy for machine-learning-assisted material discovery for low-dimensional and strong coupling systems.