Cause of metallicity discrepancies for very carbon-rich stars between GALAH DR4 and Gaia XP-based predictions

Determine whether the systematic higher [Fe/H] reported by GALAH DR4 for extremely carbon-rich stars relative to the [Fe/H] predicted by the neural-network analysis of Gaia DR3 XP spectra arises from biases in the GALAH DR4 spectroscopic analysis or from limitations in the Gaia XP-based prediction pipeline, and quantify the conditions (e.g., carbon enhancement level, temperature, signal-to-noise) under which the discrepancy occurs.

Background

The paper compares metallicity and carbon-abundance predictions derived from Gaia DR3 XP spectra using a neural network against values from several high-resolution spectroscopic surveys. In the GALAH DR4 cross-match, the authors note that some of the most carbon-rich stars show higher [Fe/H] in GALAH than in their XP-based predictions.

Because very carbon-enhanced stars are extreme objects and challenging for both low-resolution spectro-photometry and high-resolution spectroscopy, identifying whether the discrepancy originates in the survey’s spectroscopic pipeline or in the XP-based method is important for establishing reliable metallicity scales for carbon-rich stars.