The O abundance is determined from the 6300.31Å line of O i, which is available in our spectra with a S/N ratio above 100 only for the stars of NGC 2447 and for one of the stars in NGC 2360. This line is slightly blended with a line of Sc ii at 6300.68Å, but both lines are fairly well separated in our spectra (see Fig. 8). Fortunately, no significant telluric line is spoiling the O and Sc lines, though some are present 1 Å away or more. Using the equivalent width of the 6300.31Å line, the O abundance is found to be slightly deficient with respect to iron, with [O/Fe] to -0.3 dex (see Table 4).
The O i line at 6300.31Å, however, may be blended with another line at 6300.33Å from Ni i. The of the Ni i line is still poorly known. If we take the only data available in the recent literature, i.e. and eV proposed by Kurucz (1994), and if we assume [Ni/Fe]=0.00, then the oxygen abundance predicted by synthetic spectra reaches -0.6 dex on average (cf. [O/Fe]* in Table 4). A smaller value of is however suggested from the analysis of the solar spectrum. Lambert (1978) estimates the width of the Ni i Å line in the solar spectrum to be below 0.1 - 0.5 mÅ. The oscillator strength derived from the solar Ni abundance would be ten times smaller than Kurucz's even with mÅ. The true of this Ni i line is thus very probably smaller than -2.74 dex. Therefore, the blend cannot affect the oxygen abundance determination by more than a few hundredths of dex.
The surface O underabundance predictions after the first dredge-up are shown in Fig. 7 as a function of stellar mass. Models with (open circles) and without (filled circles) core overshooting during MS both predict very little variation of the surface O abundance in red giants (depletion by less than 0.04 dex). This results from the fact that only the deepest stellar layers are affected by the N-O cycle of H-burning. As a result, the dilution factor of the envelope's oxygen is rather small when the envelope extends down to the inner regions.
The predicted surface O underabundances fall much below the observed ones (by 0.1 - 0.3 dex). This discrepancy is unexplained so far.
Such a deficiency of O in red giants has already been reported in the literature (e.g. Luck 1994). Interestingly, Venn (1999) argues that the Sun presents an overabundance of oxygen of about 0.3 dex relative to the galactic A supergiants, to B stars in the solar neighborhood and in Orion, and to nebular abundances in the Orion nebula. The conclusion that the solar oxygen abundance is higher than what should be expected by galactic chemical evolution is also reached, at least qualitatively, from the sample of G dwarfs of Edvardsson et al. (1993), which show an average oxygen abundance of [O/Fe]-0.1 at [Fe/H]-0.1. If this empirical fact is confirmed, it would explain in a simple way the oxygen underabundances found in our giants, at least if we neglect the contribution of the Ni i line in the synthetic spectra.
© European Southern Observatory (ESO) 2000
Online publication: August 17, 2000