1995A&A...300..751A

Models of post-main sequence stellar evolution of VandenBerg & Bell (1985) have been applied to determine spectroscopic masses and distances for metal-poor stars. Careful consideration of the most important error sources published in more recent papers such as VandenBerg (1992) for the first time allow us to draw firm statistical conclusions. It is shown that the evolutionary calculations qualitatively fit to the observed stellar parameters whereas quantitatively they predict too high ages for metal-poor stars. As an important result we confirm that evolutionary sequences need to be calibrated with respect to their metal abundance in order to use their absolute predictions of temperature and luminosity. It turns out that this can be achieved by a simple shift of the evolutionary tracks and isochrones in effective temperature with values {DELTA}log T_eff≲0.03 which accounts for possible changes of the mixing-length and the O/Fe ratio with metallicity. The stellar luminosities and surface gravities obtained from evolutionary models are much more reliable than their effective temperatures. Therefore we conclude that the accuracy of the corresponding spectroscopic stellar gravities is systematically affected by deviations from LTE, in particular along the subgiant sequence where systematic errors less than {DELTA}log g ≃0.3 must be ascribed to the non-LTE ionization equilibrium of Fe II/Fe I. In our spectroscopic analyses the strong dependence between surface gravity and abundances determined from Fe I lines restricts the accuracy of Fe abundances in subgiants to 0.1 dex at best. The most remarkable result of our evolutionary and kinematic investigations of halo stars refers to the large fraction of slightly evolved subgiants among the so-called subdwarfs. Since conventional photometric approaches often assume that the great majority of metal-poor stars are dwarfs this results in distances that are systematically too low for their samples. Consequently, significant differences are found when comparing evolutionary and kinematic parameters obtained from either photometric or spectroscopic data. We demonstrate this by comparing the space velocities of the stars. It appears that stars with particularly high space velocities derived from spectroscopic distances show very often much lower velocities based on their main sequence parallaxes. We find that results refering to main sequence parallaxes are doubtful and can be used only with greatest care. An advantageous side-effect of the application of spectroscopic data to evolutionary calculations is the possibility to identify binary systems that are either standing out from the Toomre diagram with their unusually high space velocities, or from a log g - log T_eff diagram with apparently contradictory luminosities.