2018MNRAS.478.2812C

Intrinsically luminous giant stars in the Milky Way are the only potential volume-complete tracers of the distant disc, bulge, and halo. The chemical abundances of metal-poor giants also reflect the compositions of the earliest star-forming regions, providing the initial conditions for the chemical evolution of the Galaxy. However, the intrinsic rarity of metal-poor giants combined with the difficulty of efficiently identifying them with broad-band optical photometry has made it difficult to exploit them for studies of the Milky Way. One long-standing problem is that photometric selections for giant and/or metal-poor stars frequently include a large fraction of metal-rich dwarf contaminants. We re-derive a giant star photometric selection using existing public g-band and narrow-band DDO₅₁ photometry obtained in the Kepler field. Our selection is simple and yields a contamination rate of main-sequence stars of <=1 per cent and a completeness of about 80 per cent for giant stars with T_eff <= 5250 K - subject to the selection function of the spectroscopic surveys used to estimate these rates, and the magnitude range considered (11 <= g <= 15). While the DDO₅₁ filter is known to be sensitive to stellar surface gravity, we further show that the mid-infrared colours of DDO₅₁-selected giants are strongly correlated with spectroscopic metallicity. This extends the infrared metal-poor selection developed by Schlaufman & Casey, demonstrating that the principal contaminants in their selection can be efficiently removed by the photometric separation of dwarfs and giants. This implies that any similarly efficient dwarf/giant discriminant (e.g. Gaia parallaxes) can be used in conjunction with wise colours to select samples of giant stars with high completeness and low contamination. We employ our photometric selection to identify three metal-poor giant candidates in the Kepler field with global asteroseismic parameters and find that masses inferred for these three stars using standard asteroseismic scaling relations are systematically overestimated by 20-175 per cent. Taken at face value, this small sample size implies that standard asteroseismic scaling relations overpredict stellar masses for metal-poor giant stars.