2000AJ....120.1072S

We derive a B-V color-temperature relation for stars in the least model-dependent way employing the best modern data. The fit we obtained with the form T_eff=T_eff{cubl0}(B-V)₀, [Fe/H], logg{cubr0} covers stars in the range F0-K5 with metallicity [Fe/H]=-1.5 to +0.3 for both dwarfs and giants. The fit is well constrained, and the residual temperature of the fit is 62 K, which is consistent with what is expected from the quality of the input data. Metallicity and surface gravity effects are well separated from the color dependence. Dwarfs and giants match in a single family of the fit, differing only in logg. The fit also detects the interstellar extinction for nearby stars with the amount E(B-V)=0.235±0.03 mag.kpc^–1. Taking our newly obtained relation as a reference, we examine a number of B-V color-temperature relations and atmosphere models available in the literature. We find with the Kurucz atmosphere a systematic error of 0.07 mag in B-V in the color-temperature relation across G-K5 dwarfs. On the other hand, the Bell-Gustafsson atmosphere gives colors in agreement with our empirical relation from F to G stars; for late-K stars, however, it gives colors that are too blue by 0.05 mag. We also argue for errors in the temperature scale adopted in popularly used stellar population synthesis models; synthetic colors from these models, based on the temperature calibration of Ridgway et al., may be too blue for aged elliptical galaxies. Finally, we derive the color index of the Sun to be (B-V)_☉=0.626±0.018 and discuss that redder colors (e.g., 0.65-0.67) often quoted in the literature are incompatible with the color-temperature relation for normal stars.