Extracting causality from spectroscopy.

Area of Science:

• Materials Science
• Condensed Matter Physics
• Data Science

Background:

• Causality is key to understanding physical phenomena but difficult to extract from complex data.
• Machine learning offers potential for causal inference, yet lacks a guiding principle.
• Spectroscopy data analysis is crucial for materials characterization.

Purpose of the Study:

• To propose and validate a protocol for causal inference in spectroscopy data using DirectLiNGAM.
• To uncover causal relationships in the kagome superconductor CsV₃Sb₅.
• To demonstrate a novel approach for discovering physical laws from complex datasets.

Main Methods:

• Applied DirectLiNGAM, a statistical causal inference method, to analyze spectroscopy data.
• Utilized spatially resolved core-level photoemission spectroscopy measurements.
• Developed a protocol for analyzing complex spectral data to identify causal links.

Main Results:

• Identified causal relationships between Cs surface coverage, core-level intensity/position, and spectral background in CsV₃Sb₅.
• Provided an explanation for the polar surface formation in CsV₃Sb₅.
• Discovered an unexpected causal link in the intensity of spin-orbit satellite peaks.

Conclusions:

• The DirectLiNGAM protocol offers a powerful tool for causal inference in spectroscopy.
• This method can reveal new physical laws not easily found with traditional techniques.
• The findings advance the understanding of causality in complex material systems like CsV₃Sb₅.