Improving metallicity estimates for very metal-poor stars in the Gaia DR3 GSP-Spec catalog

Context. In the latest Gaia Data Release (DR3), the GSP-Spec module has provided stellar parameters and chemical abundances measured from the RVS spectra alone. However, the GSP-Spec parameters - including metallicity - for very metal-poor (VMP; [Fe/H] < −2) stars suffer from parameter degeneracy due to a lack of information in their spectra, and are therefore affected by a large measurement uncertainty and systematic offset. Furthermore, the recommended quality cuts filter out the majority (~80%) of the VMP stars because some of them are confused with hot stars or with cool K- and M-type giants, for which the current pipeline is known to have problems.
Aims: We aim to provide more precise metallicity estimates for VMP stars analyzed by the GSP-Spec module by taking photometric information into account in the analysis and breaking the degeneracy.
Methods: We reanalyzed FGK-type stars in the GSP-Spec catalog by computing the Ca triplet equivalent widths from the published set of GSP-Spec stellar parameters. We compared these recovered equivalent widths with the values directly measured from public Gaia RVS spectra and investigated the precision of the recovered values and the parameter range within which the recovered values are reliable. We then converted the recovered equivalent widths to metallicities by adopting photometric temperatures and surface gravities that we derive based on Gaia and 2MASS catalogs.
Results: The recovered equivalent widths agree with the directly measured values with a scatter of 0.05 dex for the stars that pass the GSP-Spec quality cuts. Among the stars recommended for filtering out, we observe a similar scatter for FGK-type stars initially misidentified as hot stars. Contrarily, we find a poorer agreement, in general, for stars that the GSP-Spec identifies as cool K- and M-type giants, although we can still define subsets that show reasonable agreement. At the low-metallicity end ([Fe/H] < −1.5), our metallicity estimates have a typical uncertainty of 0.18 dex, which is about half of the quoted GSP-Spec metallicity uncertainty at the same metallicity. Our metallicities also show better agreement with the high-resolution literature values than the original GSP-Spec metallicities at low metallicity; the scatter in the comparison decreases from 0.36-0.46 dex to 0.17−0.29 dex for stars that satisfy the GSP-Spec quality cuts. While the GSP-Spec metallicities show increasing scatter when misidentified "hot" stars and the subsets of the "cool K- and M-type giants" are included (up to 1.06 dex), we can now identify them as FGK-type stars and provide metallicities that show a small scatter in the comparisons (up to 0.34 dex), which helps us to increase the number of VMP stars with reliable and precise metallicity.
Conclusions: The inclusion of photometric information greatly contributes to breaking parameter degeneracy, enabling precise metallicity estimates for VMP stars from Gaia RVS spectra. We produce a publicly available catalog of bright metal-poor stars suitable for high-resolution follow-up. The sample contains about 2345 VMP stars with an estimated contamination rate of 5%.

The catalog is available at the CDS via anonymous ftp to cdsarc.cds.unistra.fr (ftp://130.79.128.5) or via https://cdsarc.cds.unistra.fr/viz-bin/cat/J/A+A/685/A59