The GALAH survey: Improving chemical abundances using star clusters

TL;DR

This study uses stellar clusters as benchmarks to identify and correct systematic errors in spectroscopic stellar parameters and chemical abundances in the GALAH survey, enhancing the accuracy of large-scale stellar data.

Contribution

The paper introduces a method to use star clusters for validating and correcting systematic errors in spectroscopic parameters in the GALAH survey data release 4.

Findings

Trends in chemical abundances vary with temperature and differ between dwarfs and giants.

Corrected abundances for 24 and 31 elements are published after trend removal.

Spectral fitting degeneracies and linelist completeness significantly impact abundance measurements.

Abstract

Large spectroscopic surveys aim to consistently compute stellar parameters of very diverse stars while minimizing systematic errors. We explore the use of stellar clusters as benchmarks to verify the precision of spectroscopic parameters in the 4. data release (DR4) of the GALAH survey. We examine 58 open and globular clusters and associations to validate measurements of temperature, gravity, chemical abundances, and stellar ages. We focus on identifying systematic errors and understanding trends between stellar parameters, particularly temperature and chemical abundances. We identify trends by stacking measurements of chemical abundances against effective temperature and modelling them with splines. We also refit spectra in three clusters with the Spectroscopy Made Easy and Korg packages to reproduce the trends in DR4 and to search for their origin by varying temperature and gravity priors, linelists, and spectral continuum. Trends are consistent between clusters of different ages and metallicities, can reach amplitudes of ~0.5 dex and differ for dwarfs and giants. We use the derived trends to correct the DR4 abundances of 24 and 31 chemical elements for dwarfs and giants, and publish a detrended catalogue. While the origin of the trends could not be pinpointed, we found that: i) photometric priors affect derived abundances, ii) temperature, metallicity, and continuum levels are degenerate in spectral fitting, and it is hard to break the degeneracy even by using independent measurements, iii) the completeness of the linelist used in spectral synthesis is essential for cool stars, and iv) different spectral fitting codes produce significantly different iron abundances for stars of all temperatures. We conclude that clusters can be used to characterise the systematic errors of parameters produced in large surveys, but further research is needed to explain the origin of the trends.