TITANS metal-poor reference stars II. Red giants and CEMP stars

Context. Representative samples of F-, G-, K-type stars located outside of the solar neighbourhood have started to become available in spectroscopic surveys. The fraction of metal-poor ([Fe/H] less than or similar to -0.8 dex) giants becomes increasingly relevant towards greater distances. In metal-poor stars, effective temperatures (T-eff) based on local thermodynamic equilibrium (LTE) spectroscopy and on former colour-T-eff relations - which are still widely used - have been reported to be inaccurate. It is necessary to recalibrate chemical abundances based on these Teff scales in the multiple available surveys in order to bring them to the same standard scale for their simultaneous use. To this end, a complete sample of standards is required, which until now has been restricted to a few stars with quasi-direct T-eff measurements. Aims. We aim to provide a legacy sample of metal-poor standards with proven accurate atmospheric parameters. We add 47 giants to the TITANS metal-poor reference stars.Methods. We derived T-eff using 3D non-LTE H alpha modelling, the accuracy of which was tested against interferometry and with the Infra Red Flux Method (IRFM). We derived surface gravity (log g) by fitting Mg I b triplet lines, and tested their accuracy against asteroseismology. Metallicity was derived using Fe II lines, and we find our results to be identical to the [Fe/H] derived from non-LTE spectral synthesis. Results. The T-eff that we find using 3D non-LTE H alpha is equivalent to interferometric and IRFM temperatures within a +/- 46 K uncertainty. We achieve precision of similar to 50 K for 34 stars with spectra with the highest signal-to-noise ratio (S/N). For log g, we achieve a total uncertainty of +/- 0.15 dex. For [Fe/H], we obtain a total uncertainty of +/- 0.09 dex. We find that the ionisation equilibrium of Fe lines under LTE is not valid in metal-poor giants. LTE leads to a small but significant metallicity underestimation of similar to 0.1 dex when derived from weak Fe I lines, and only provided accurate T-eff and log g. This bias totally disappears under non-LTE.