2005ApJ...626..446R

The infrared flux method (IRFM) has been applied to a sample of 135 dwarf and 36 giant stars covering the following regions of the atmospheric parameter space: (1) the metal-rich ([Fe/H]≳0) end (consisting mostly of planet-hosting stars), (2) the cool (T_eff≲5000 K) metal-poor (-1≲[Fe/H]≲-3) dwarf region, and (3) the very metal-poor ([Fe/H]≲-2.5) end. These stars were especially selected to cover gaps in previous works on T_eff versus color relations, particularly the IRFM T_effscale of A. Alonso and collaborators. Our IRFM implementation was largely based on the Alonso et al. study (absolute infrared flux calibration, bolometric flux calibration, etc.) with the aim of extending the ranges of applicability of their T_effversus color calibrations. In addition, in order to improve the internal accuracy of the IRFM T_effscale, we recomputed the temperatures of almost all stars from the Alonso et al. work using updated input data. The updated temperatures do not significantly differ from the original ones, with few exceptions, leaving the T_effscale of Alonso et al. mostly unchanged. Including the stars with updated temperatures, a large sample of 580 dwarf and 470 giant stars (in the field and in clusters), which cover the ranges 3600K≲T_eff≲8000K and -4.0≲[Fe/H]≲+0.5, have T_effhomogeneously determined with the IRFM. The mean uncertainty of the temperatures derived is 75 K for dwarfs and 60 K for giants, which is about 1.3% at solar temperature and 4500 K, respectively. It is shown that the IRFM temperatures are reliable in an absolute scale given the consistency of the angular diameters resulting from the IRFM with those measured by long baseline interferometry, lunar occultation, and transit observations. Using the measured angular diameters and bolometric fluxes, a comparison is made between IRFM and direct temperatures, which shows excellent agreement, with the mean difference being less than 10 K for giants and about 20 K for dwarf stars (the IRFM temperatures being larger in both cases). This result was obtained for giants in the ranges 3800K<T_eff<5000K and -0.7<[Fe/H]<0.2 and dwarfs in the ranges 4000K<T_eff<6500K and -0.55<[Fe/H]<0.25; thus, the zero point of the IRFM T_effscale is essentially the absolute one (that derived from angular diameters and bolometric fluxes) within these limits. The influence of the bolometric flux calibration adopted is explored and it is shown that its effect on the T_effscale, although systematic, is conservatively no larger than 50 K. Finally, a comparison with temperatures derived with other techniques is made. Agreement is found with the temperatures from Balmer line profile fitting and the surface brightness technique. The temperatures derived from the spectroscopic equilibrium of Fe I lines are differentially consistent with the IRFM, but a systematic difference of about 100 and 65 K (the IRFM temperatures being lower) is observed in the metal-rich dwarf and metal-poor giant T_effscales, respectively.