Astron. Astrophys. 353, 557-568 (2000)

4. Derived abundances

4.1. Molecular bands versus model atmospheres

In Fig. 1a we compare the partial pressures across the atmosphere, for CH, CN and C₂ using model atmospheres by Kurucz (1992), Gustafsson et al. (1975 and 1980, hereafter MARCS75) and the present models. Fig. 1b shows the reciprocal temperature as a function of the gas pressure, which traces the depth in the atmosphere.

Fig. 1. a Log scale variation of the partial pressure of molecules (Pmol/Pgas) versus the gas pressure in three model atmospheres: MARCS99 (C, N enriched, dotted lines ), MARCS75 (solid lines ) and Kurucz (1993, dashed lines ). The models were calculated for =4500 K, log g =1.0 and [Fe/H]=-3.0, and the dissociation equilibrium was calculated for [C/Fe]=+2. and [N/Fe]=+2. (The CH partial pressure was artificially multiplied by a factor 100 for the sake of clarity.) b  Temperature variation versus the gas pressure for the same three models.

The molecular concentration in the upper layers of the atmosphere is always lower in MARCS75 models due to a higher temperature, whereas in the Kurucz and Plez models, temperature tends to decrease towards the outer layers (Fig. 1b). As a consequence, the very strong lines, especially the molecular bandheads, are stronger in the Kurucz and MARCS99 models than in MARCS75 models.

On the other hand, intermediate and weak lines forming at log () 1.0 will be weaker with the MARCS99 models than with MARCS75 and Kurucz's models, since the former is systematically hotter in the inner layers (Fig. 1b). The difference will be particularly evident for CN and C₂, but less so for CH.

The shape of the computed molecular bands (from low to high vibrational transitions) will thus be very different when computed using a model atmosphere which explicitly takes into account the C and N overabundances or not (MARCS75, Kurucz). These changes are a result of the large differences in the temperature gradients inside the model stellar atmospheres. Deep in the atmosphere, the pressure inversion in the MARCS99 models is due to turbulent pressure in the convection zone.

4.1.1. Results given by different models

In Paper I we used classical MARCS75 models to represent these C-rich stars. It is clear from the discussion above that one should use appropriate models when possible. Using standard models in this case may introduce discrepancies between the strong and weak features of a single molecule, and also discrepancies between features of different molecules. For example, it might be expected that the abundance deduced from C₂ lines and CH lines (which are stronger) will disagree: the C₂ lines will be overestimated (leading to low C abundances) and the CH features will be underestimated (leading to high C abundances).

However, comparing the synthetic spectra calculated with the present models with the observed spectra of our stars, we found that the C₂ bands did not have exactly the right shape: the very strongest lines (the bandheads themselves) are slightly overestimated with respect to the medium to weak lines. This is most probably due to the fact that the models do not represent very well the atmosphere in the outermost layers, which could be due to the neglect of opacities from carbon-bearing polyatomic molecules. Therefore, in our analysis, we discarded the very strongest lines and relied on lines of moderate strength to determine the C and N abundances.

Some disagreement appears with our previous measurements (Paper I, where we used MARCS75 models), as well as between CH and C₂ bands for carbon. The C abundances derived from C₂ lines are about 0.5 dex higher than those derived from the CH G band (Table 8). This is possibly due to problems in the line lists, in addition to the model atmosphere problem outlined above - the same discrepancy has been found by Bonifacio et al. (1998). The most probable values of the C abundance are summarized in Table 8.

4.2. Heavy neutron-capture elements (Z35)

In view of the discrepancies pointed out in Sect. 4.1 concerning the intensity of CH versus C₂ lines, we adjusted the C and N abundances around each atomic line of interest.

Owing to the severe blending (mostly by molecular bands) effects on many of the atomic lines, the abundances of the neutron-capture elements were all determined by synthetic spectrum fitting. Figs. 2 to 5 show the typical fits obtained, and also indicate the effect of uncertainties on the C and N abundances. The line-by-line abundances are summarized in Table 7, where we report results only for the lines that were not too severely blended by strong molecular bands. In this table, a (:) sign denotes a very uncertain abundance, which was then not used to compute the mean abundance of the element reported in Table 8. Also reported in Table 8 are the 1- dispersion of the abundances derived from all the lines of a given element (when three or more lines could be measured). The abundance patterns of the heavy elements as a function of atomic number are illustrated in Fig. 7.

Fig. 2. CS 22948-27: YII 4883.69 and LaII 4086.68 Å lines computed with respectively [Y/Fe] = +0.8, +1.2 dex and [La/Fe]=+2.5, +2.7 dex. The observed spectrum is displayed as dots. In the right panel, the dashed lines show the effect of increasing the C abundance by 0.2 dex, showing that the blending CN line introduces an uncertainty of 0.3 dex on the derived La abundance.

Fig. 3. CS 29497-34: BaII 5853.69 and 6141.73 Å lines computed with [Ba/Fe] = +1.9, +2.1.

Fig. 4. CS 29497-34: NdII 4012.03 and SrII 4077.7 Å lines computed with, respectively, [Nd/Fe] = +1.8, +2.1 and [Sr/Fe] = +1.0, +1.2 (solid lines). The observed spectrum is displayed as dots. In the right panel, the dashed line shows the effect of increasing the C abundance by 0.2 dex.

Fig. 5. CS 22948-27: EuII 6645.13 Å computed for [Eu/Fe] = 0.0 and +2.1 (solid lines). The dashed line corresponds to a change in the intensity of CN bands, obtained by decreasing the C abundance by 0.2 dex.

Table 7. Line-by-line elemental abundances. (X) is the logarithmic value = log (X/H) + 12.

Table 8. Mean abundances for Li, C, Na, Mg and heavy elements .
Notes:
a) from moderate strength lines around 4000 Å
b) from the G band (very strong lines)

The Th II 4019.13 Å line which has been observed in CS 22892-52 (interestingly, another C-enhanced star) has been discussed at length by many other authors (Sneden et al. 1996; Norris et al. 1997b; Cowan et al. 1999). The interest in the derivation of an accurate Th abundance is clear, since this radioactive element can serve as a chronometer to measure the age of a star (provided suitable comparison r-process elements are also observed). Unfortunately, at the resolution of our spectra this is not possible because there is a strong ¹³CH line falling on top of it (as was cautioned by Norris et al. 1997b). For the two stars under study here, although an absorption line is clearly visible at 4019.13 Å, the uncertainty on the carbon abundance and ratio does not allow us to determine the real thorium contribution to the blend. We therefore prefer not to make use of this chronometer based on the present data.

Europium is a very critical element to measure since it is thought to be produced almost solely by the r process. The main lines of this element are Eu II 4129.7, 4205.05, 6437.64 and 6645.13 Å. Unfortunately, both the 4129.7 and 4205.05 Å lines are too severely blended by CN lines to determine abundances from them. The lines measured were thus the two reddest ones. The Eu II 6645.13 Å is illustrated in Fig. 5.

4.3. Lithium abundance

The Li I 6707 Å line was also measured in both CS 22948-27 and CS 29497-34, and the derived lithium abundances are given in Table 8. Both stars exhibit lithium abundances which are lower than the estimated primordial abundance, as expected for giant stars, which are capable of diluting their initial lithium by internal mixing processes. The rather low ratio found in these stars (14 and 12 respectively for CS 22948-27 and CS 29497-27, derived from the CN lines at 8003-8004 Å - see Paper I), is similar to the ratio found in typical field halo giants (Sneden et al. 1986, Pilachowski et al. 1997 ¹), and in these stars, such a low value is interpreted as due to mixing. On the other hand, the observed Li abundance might be considered "high", since lithium is not often detected in metal-poor giants. This high Li abundance is an argument in favor of an additional supply of Li by mass transfer from an evolved companion (AGB) providing both Li and elements. Other metal-poor C-rich stars are known in the literature with various Li abundances (e.g., Norris et al. 1997a) - the dwarfs and subgiants having seemingly larger Li abundances than the giants, in agreement with the effect of mixing in giants. Variations on these ideas could be considered, such as a common envelope evolution.

© European Southern Observatory (ESO) 2000

Online publication: December 17, 1999
helpdesk.link@springer.de