Oxygen in unevolved metal-poor stars from Keck ultraviolet hires spectra

The determination of the abundance of oxygen (O) is important in our unders tanding of mass-spectrum of previous generations of stars, the evolution of the Galaxy, stellar evolution, and the age-metallicity relation. We have m easured O in 24 unevolved stars with Keck HIRES observations of the OH line s in the ultraviolet spectral region at a spectral resolution of similar to 45,000. The spectra have high signal-to-noise ratios, typically 60-110, an d high dispersion, 0.022 Angstrom per pixel. Very special care has been tak en in determining the stellar parameters in a consistent way and we have do ne this for two different, plausible temperature scales. The O abundance fr om OH has been computed by spectrum synthesis techniques for all 24 stars p lus the Sun for which we have a Keck spectrum of the daytime sky. In additi on, we determined O abundances from the O I triplet with our stellar parame ters and the published equivalent widths of the three O I lines from six so urces. The comparison of data analyzed with the same, consistently determin ed, parameter sets show generally excellent agreement in the O abundances; differences in the origin of the models (not the parameters) may result in abundance differences of 0.07 to 0.11 dex. We show that the O abundances fr om OH and from Glare reliable and independent and average the two for the a dopted O. This averaging has the great benefit of neutralizing uncertaintie s in the parameters since OH and O I strengths depend on effective temperat ure and gravity in opposite directions. For these cool, unevolved stars we find that O is enhanced relative to Fe with a completely linear relation be tween [O/H] and [Fe/H] over 3 orders of magnitude with very little scatter; taking the errors into account in determining the fits, we find [O/H] = 0.66 (+/-0.02) [Fe/H] + 0.05 (+/-0.04). The O abundances from 76 disk stars of Edvardsson et al. have a measured slope of 0.66 (identical to our halo dwarf stars) and fit this relationship smoothly. The relation between [O/Fe ] and [Fe/H] is robustly linear and shows no sign of a break at metalliciti es between -1.0 and -2.0, as has been discussed by others. At low metallici ties, [Fe/H] < -3.0, [O/Fe] > + 1.0. The fit to this relationship (taking t he errors into account) is [O/Fe] = -0.35 (+/-0.03) [Fe/H] + 0.03 (+/-0.05) . The enrichment of O is probably still from massive stars and Type II supe rnovae; however, the absence of a break in [O/Fe] versus [Fe/H] runs counte r to traditional galactic evolution models, and the interplay of Type II an d Type Ia supernovae in the production of O and Fe should be reexamined. It appears that either Fe or O can be used as a chronometer in studies of gal actic evolution.