Forum Springer Astron. Astrophys.
Forum Whats New Search Orders

Astron. Astrophys. 332, 204-214 (1998)

Previous Section Next Section Title Page Table of Contents

2. Observational data for metal poor stars

Some of the data that we employ here have already been published. We concentrate on stars of modest metal deficiency. Five bright metal-deficient field stars are [FORMULA] Ori, [FORMULA] And, [FORMULA] Leo B, [FORMULA] Leo A and Arcturus. Their [FORMULA] values range from near +0.9 to -1.0 so they form an evolutionary sequence that starts from fainter absolute magnitude than can be reached with high spectral resolution in globular clusters. As a continuation of our "evolutionary sequence" we include two stars from the globular cluster 47 Tucanae whose properties indicate that they are first giant branch (i.e. not AGB) stars. The basic properties of our sample stars are shown in Table 1.


Table 1. Observed Stars a

For the field stars we derive the gravity values assuming a mass of 0.8 [FORMULA], the HIPPARCOS parallaxes, [FORMULA] 's derived from the [FORMULA] color- [FORMULA] relation (Ridgway et al. 1980, DiBenedetto 1993, Dyck et al. 1996), and the bolometric corrections of Buser and Kurucz (1992). This method is probably superior to spectroscopic determination of gravities using ionization balance; a test using the HIPPARCOS parallaxes indicates the spectroscopic gravities often are too low by a factor of two (Nissen & Hog 1997). The visual binary stars in [FORMULA] Leo do not have separate colors; instead, we adjust the spectroscopic [FORMULA] 's of Lambert & Ries (1981) downward by 210K, the mean offset between their temperature scale and the [FORMULA] scale. We note that the 47 Tuc stars are probably not on the same absolute magnitude scale as the HIPPARCOS results, since they depend upon a cluster distance modulus which probably needs revision in light of the HIPPARCOS results for the subdwarfs; see, e.g., Reid (1997). For our purposes, the precise value of the absolute magnitudes of the two globular cluster stars is not important: it matters here only that they are [FORMULA] mag more luminous than our most luminous field stars. 1

The abundances we use have been rederived with a number of improvements. For Arcturus and the [FORMULA] Leo pair we have combined the equivalent widths of Lambert & Ries (1981) and Ries (1981) with the empirical photometric [FORMULA] scale and the most recent and almost certainly most accurate value of the CN dissociation potential [FORMULA] = 7.65 eV (Bauschlicher et al. 1988). For 47 Tuc we have used the data of BWO with two changes. The analysis by Hesser et al. (1987) of the color-magnitude diagram of 47 Tuc indicate an iron abundance for 47 Tuc higher by about 0.2 dex as compared with the BWO value. Hence, we have accepted [Fe/H] = -0.65 rather than -0.85 for the 47 Tuc stars and have recalculated the CNO abundances using models with the new [Fe/H] values. For the [FORMULA] ratio in 47 Tuc star 4418, we have adopted the second solution in Table 2 of Bell et al. (1990; multiple solutions for [FORMULA] and carbon abundance are given for different values of the stellar microturbulence parameter) because the corresponding derived carbon abundance is closer to that of BWO (in which the microturbulence is derived as part of the analysis of the high resolution spectra in the CH region).

To extend our evolutionary sequence to lower luminosity on the giant branch we have obtained new high resolution spectra of two field stars [FORMULA] Ori and [FORMULA] And, whose metallicities are similar to those of the other stars. The new spectra were obtained at the Dominion Astrophysical Observatory with the 1.2-meter telescope, coude spectrograph, long camera and Reticon detector. This combination yields a resolving power of 35,000, and the exposures were timed for a signal-to-noise of about 200. Our spectra cover the C2 Swan band at 5635 Å, the [O I] line at 6300 Å, the Li line at 6700 Å and the 2-0 red system CN band at 8000 Å, thereby providing abundances of Li, C, N, O as well as iron and the ratio [FORMULA]. These stars were analyzed using the same models and gf-values as were used in the other analyses except that the C2 Swan bands replaced the violet CH bands as the source of the carbon abundance. To analyze the C2 Swan band at 5635 Å it was necessary to synthesize the spectrum over an interval from 5626-5640 Å. The spectroscopic data for the C2 Swan bands is that reviewed by Grevesse et al. (1991) in their re-evaluation of the solar C abundance.

Both [FORMULA] And and [FORMULA] Ori have been analyzed by others, most recently with modern model atmospheres and digital spectra by Cottrell & Sneden (1986) and subsequently by SSP93. These two stars provide an overlap with their large sample of stars of generally lower luminosity. We compare their data with ours in Table 2.


Table 2. Comparison of our abundances (upper analysis) with those of Cottrell and Sneden (1986) and Shetrone et al.(1993) (lower analysis) of [FORMULA] And and [FORMULA] Ori

Most of the differences between our abundances and those of the Texas group can be understood as due to differences in effective temperatures and gravities. Their lower effective temperatures are responsible for their lower iron abundances and nitrogen abundances because of lower opacities and the dissociation equilibrium of the CN molecule. The low masses derived by SSP93 (Table 6) and of Cottrell & Sneden (1986, Table 6) demonstrate that their gravities are too low by roughly a factor 3 which is the difference between their and our gravity values for the same stars.

Our final adopted abundances are shown in Table 3.


Table 3. Abundance Results

It is important to understand the uncertainties in the individual abundances in Table 3. The iron abundances are almost independent of the CNO abundances but not entirely so. Since many lines of Fe I and several lines of Fe II are easily observed the measured line strengths are not an important source of error. The uncertainty is surely dominated by the uncertainty in the atmospheric model, especially at small optical depth where mechanical energy input and backwarming, such as electron conduction from the chromosphere, can effect the temperature near the boundary, especially the temperature minimum (Kurucz 1996). Hence, we are dealing with an uncertain function, T([FORMULA]), not just an uncertain parameter. This problem was evaluated carefully by Leep et al. (1987) in connection with the metallicity of M71 and the derivation of solar f-values using either the Holweger-Müller (1974) model or the Bell et al. (1976) model. The conclusion of that paper was that no [Fe/H] value for a red giant can be more accurate than [FORMULA] 0.2 and we accept that here. The influence of CNO abundances on the [Fe/H] value comes in through the opacity of CO and CN which effects the atmospheric models. The infrared CO bands are important for the cooling and structure of the outermost layers of an atmosphere of a late-type star, but this effect is less important for stars as warm as those in this study. The CN lines can contribute substantially to the atmosphere structure of red giants because the CN red system blankets a rather large wavelength range near the flux peak, but CN diminishes in importance quickly as the metallicity of a star is decreased because the CN column density falls approximately as metallicity squared while the continuous opacity falls approximately as metallicity.

The uncertainty in the oxygen abundance is mostly determined by the actual measurement of the equivalent width of the only useful oxygen line, [FORMULA] 6300, in the globular cluster stars, and the second line at [FORMULA] 6363 which is barely detectable in the bright field stars. Since the oxygen line is formed over a wide range of optical depth the derived abundance is not very sensitive to the boundary temperature. The low carbon abundances in these stars means that the uncertainty in the C abundance has little effect on the solution for the O abundance despite the formation of CO in the atmosphere. For oxygen the uncertainties in O/Fe are surely [FORMULA] 0.1 dex or perhaps a little lower for Arcturus, for which the measurements are by far the most accurate. Hence, the somewhat higher O/Fe value for Arcturus, as compared with the other stars in Table 3, is probably real.

The formation of CO depletes carbon and an error of 0.10 dex in oxygen will cause an error of 0.05 dex in the carbon abundance. Fortunately the CH and C2 lines are not formed at a small optical depth (as they are in the sun) but rather below the levels at which carbon is depleted by CO formation, so the uncertainty in the boundary temperature is not important. Considering both the uncertainty in the oxygen abundance, the actual measurements of CH, and the presence of some saturation in the CH lines, an uncertainty of 0.15 dex is possible for the 12 C abundances.

The [FORMULA] ratio is determined largely by the clump of 13 CN lines near 8005 Å. This was illustrated in Fig. 3 of BWO: for that star, M13 I-48, a ratio of 6 was derived and the figure shows that a range from 5 to 8 is possible. Scaling that evaluation to Table 2 of this paper indicates that an observed ratio of 7 for three stars could lie between about 5.5 and 9; the pure equilibrium ratio of 3.5-4.0 is pretty well excluded as are values of 10 or above.

The nitrogen abundance, derived from CN lines, is less dependent on line measurement uncertainties because many ([FORMULA] 15-20) lines are available in our spectra. However, it does depend on the abundances of C and O through the dissociation equilibrium of CO, CH, and CN, as well as the formation of N2. Hence the uncertainties in the C and O abundances come into play as well as the uncertainty in the CN dissociation potential. For CN the difference between the theoretical value, 7.65 eV (Bauschlicher et al. 1988) and the recent experimental value of 7.77 [FORMULA].05 eV (Costes et al. 1990) is disturbing. We have used the 7.65 eV value. If the 7.77 eV value is correct our N abundances must be lowered by 0.2 dex, which lowers our C+N values by a somewhat smaller amount. The reduction in C+N+O introduced by such a reduction in N is .05 dex for Arcturus and the [FORMULA] Leo pair, and .08 dex for the 47 Tuc stars.

An error in either the dissociation potential of CN or the f -value of the CN red system 2-0 band would effect all of our nitrogen abundances equally. Both of these quantities are not as well determined as might be hoped: the dissociation energy of CN seems to be uncertain by about 0.2 eV, and different sources for the f-values disagree by as much as 30(see the discussion in Bauschlicher et al. 1988). The values used here for these quantities, 7.65 eV for [FORMULA] and [FORMULA] for [FORMULA], are those used in (for [FORMULA]) or derived from ([FORMULA]) the analysis of CN red system lines in the solar spectrum by Sneden & Lambert (1982), so by construction they yield the solar nitrogen abundance. A second source of uncertainty in the N abundance is introduced by the uncertainty in the C abundance. For measured CN lines any error in the C abundance introduces an error of comparable size and opposite sign in the N abundance. Since an error of as much as 0.15 dex in C is possible, an error of the same size with reversed sign is possible for N.

The [FORMULA] ratio also has some observational uncertainty, almost exclusively due to the difference in strength between the [FORMULA] and [FORMULA] lines, so that the microturbulent velocity enters into the isotope ratio solution. This uncertainty is aggravated as the [FORMULA] lines get stronger and as the isotope ratio increases. For the stars in this sample, the [FORMULA] lines are quite weak: for [FORMULA] Leo A, which has the strongest CN lines in the sample, the reduced equivalent widths are in the range [FORMULA], indicating that saturation (and hence microturbulence-related errors in the isotope ratio) are minimal. Our results are summarized in Figs. 1 and 3, and discussed below.

Previous Section Next Section Title Page Table of Contents

© European Southern Observatory (ESO) 1998

Online publication: March 10, 1998