Several large stellar spectroscopic surveys are producing overwhelming amounts of data that can be used for determining stellar atmospheric parameters and chemical abundances. Nonetheless, the accuracy achieved in the derived astrophysical parameters is still insufficient, mainly because of the paucity of adequate calibrators, particularly in the metal-poor regime ([Fe/H] $\leq -$1.0). Here, we introduce the Titans metal-poor reference stars: a sample of 41 dwarf and subgiant stars with accurate parameters. Effective temperatures (Teff) were derived by fitting observed H$\alpha$ profiles with synthetic lines computed using 3D hydrodynamic NLTE models. Surface gravities (logg) were computed using evolutionary tracks and parallaxes from Gaia EDR3. The same methods recover the Teff values of the Gaia benchmark stars, which are mostly based on interferometric measurements, with a 1$\sigma$ dispersion of $\pm 50$ K. We assume this to be the accuracy of the H$\alpha$ profiles computed from 3D non-LTE models for metal-poor dwarfs and subgiants. We achieved an internal precision typically between 30-40 K, these errors dominated by instrumental effects. The final total uncertainty for the Teff values of the Titans are thus estimated to be of the order of $1\%$. The typical error for logg is $\leq$ 0.04 dex. In addition, we identified a few members of Gaia-Enceladus, of Sequoia, and of the Helmi stream in our sample. These stars can pave the way for the accurate chemical characterization of these Galactic substructures. Using the Titans as reference, large stellar surveys will be able to improve the internal calibration of their astrophysical parameters. Ultimately, this sample will help users of data from Gaia and large surveys in reaching their goal of redefining our understanding of stars, stellar systems, and the Milky Way.
Comment: Accepted in A&A