Astron. Astrophys. 349, 553-572 (1999)

7. Evidence for chemical peculiarities

As was outlined in the introduction, photospheric chemical peculiarities due to the presence of core-processed material are predicted by current evolutionary theories (the precise amount of mixing varying markedly between different models). CNO-cycled material would be expected to have increased helium and nitrogen abundances, accompanied by decreased carbon and oxygen abundances, with the changes in carbon and nitrogen being the largest. These patterns are observed in the linestrengths of our supergiant sample; Figs. 6 and 7 show very large ranges in observed equivalent widths due to C II and N II , with Fig. 8 showing only moderate ranges for O II , while no significant variations within the helium linestrengths are apparent.

7.1. Identification of chemically diverse groups

It will be useful for the discussions below, to identify those stars which may have contaminated photospheres. Fig. 9 shows the ratios of equivalent widths of the N II line at 4630 Å to those of the C II line at 4267 Å. These features have been selected as they have reliable linestrength measurements, while their general behaviour as a function of effective temperature also appears to be reasonably well modelled by our non-LTE computations. In producing this figure, we have not used the equivalent widths tabulated in Paper II, but rather have remeasured the spectral features from the original data in order to try and improve the accuracy and also to obtain formal measurement errors. Also shown are the theoretical values of this ratio for a normal nitrogen abundance, a carbon abundance [C/H] = 7.80 dex (see Sect. 6.3.3) and a microturbulence of 10 km s^-1. As the luminosity dependence of the spectral lines may be important, the figure includes three non-LTE loci - the central curve is the ratio of predicted linestrengths for the - relationship discussed in Sect. 6.1, while the dashed curves represent changes of 0.2 dex in . The relatively small displacements of these loci indicate that the luminosity effect cannot directly account for the observed linestrength patterns (note that some of the small structure in these loci may not be real but rather reflect the use of relatively simple interpolation techniques).

Fig. 9. Logarithmic ratios of equivalent widths for the lines at 4630 Å (N II ) and 4267 Å (C II ). The errors reflect uncertainties in the linestrength measurements. The non-LTE loci assume = 10 km s-1, a solar helium fraction and [N/H] = 7.69 & [C/H] = 7.80 dex. Three predicted non-LTE ratios are shown, viz. that for the - relationship shown in Fig. 2 (solid line) and two having the -scale displaced by 0.2 dex.

Three sub-groups can be tentatively identified in Fig. 9:

• supergiants that appear chemically near-normal (labelled `normal/moderate' and lying close to our theoretical ratios)

• supergiants that appear to have had their photospheres contaminated by the products of CN-cycle burning (labelled `highly processed' and lying away from the normal loci)

• supergiants which may have suffered from some contamination, but perhaps of a smaller magnitude (labelled `processed?').

This figure, however, does not include the possibility of varying line desaturation due to, for example, a luminosity-dependent microturbulence, or a strong stellar wind. In this context, it is interesting to note that all the `highly processed' supergiants have strong P Cygni H profiles, which would suggest that mass loss may play a role in the line formation. Nevertheless, whilst line desaturation could perhaps explain the increased N II linestrengths, it seems unlikely that such phenomena could explain the correlated weaknesses of some of the C II features.

The effective temperatures of the seven `highly processed' supergiants lie between 19000 to 23500 K and in this range, there are also seven `normal/moderate' supergiants. For the latter, the lines strengths of the C II line at 4267 Å and the N II line at 4630 Å are reasonably tightly bunched with mean value of mÅ and mÅ respectively. For the `highly processed' group, the mean line strengths are mÅ (C II ) and mÅ (N II ); the larger standard deviations, particularly for the nitrogen line, possibly implies that this is not a chemically homogeneous group. Although the observational data do not warrant an analysis of the individual stars, we have attempted to estimate the differences in the carbon and nitrogen abundances between the two groups of objects. We have assumed a representative effective temperature of 21000 K, with a logarithmic gravity of 2.5 dex (see Fig. 2) and a microturbulent velocity of 10 kms<sup>-1</sup>. The mean equivalent widths then imply that the `highly processed' group are enhanced in nitrogen by 1.4 dex and depleted in carbon by 0.5 dex.

However, we emphasize that these results must be treated with considerable scepticism as they are subject to large uncertainties. For example, if the microturbulent velocity is increased to 15 kms^-1, the nitrogen enhancement is decreased to 0.9 dex. Additionally as discussed above, the `highly processed' targets might have a larger microturbulence than the `normal/moderate' targets; arbitrarily adopting microturbulent velocities of 15 kms^-1 for the former and 10 kms^-1 for the latter would further reduce the nitrogen enhancement to 0.6 dex. Although these calculations should be considered as numerical experiments, they illustrate that any quantitative estimate of abundance variations are critically dependent on the theoretical assumptions. Hence we can only conclude that the `highly processed' targets probably exhibit a significant nitrogen enhancement coupled with a smaller carbon depletion (relative to the `normal/moderate' group) and that this is consistent with nuclear processed material being present in their atmospheres.

7.2. Evidence for evolutionary patterns in oxygen and helium

It has already been shown above that variations in the abundance patterns of O II and He I within the supergiant sample are not obvious. This is consistent with the predicted changes in these element abundances and also their sensitivity to other parameters, such as the microturbulence and gravity. However, having used the more sensitive lines of carbon and nitrogen to identify apparently normal and processed groups within the sample, it is worthwhile reconsidering the linestrength patterns of oxygen and helium in this context.

In Figs. 10 and 11 we reproduce, for representative lines of He I and O II , the linestrength plots of Sect. 6. Only those objects which have been assigned to one of the 3 sub-groups are included in these figures. For helium, no obvious correlation between assigned chemical sub-group and the strengths of the lines is apparent. In the case of oxygen, the highly processed supergiants (which tend to be the most luminous) seem to show stronger oxygen features than the chemically normal supergiants. This is opposite to the trend expected from evolutionary considerations and may be due to a positive dependence of microturbulence on luminosity, particularly as the value, derived from lines of O II , is very high for these luminous objects.

Fig. 10. Measured He I linestrengths for representative features at 4387 and 4471 Å. The symbols represent the CN status of the supergiants as follows: cross - `normal/moderate', open circle - `processed?', filled circle - `highly processed'.

Fig. 11. Measured O II linestrengths for representative features at 4075 and 4317 Å. The symbols are the same as in Fig. 10.

It is, therefore, not possible to derive quantitative information on the abundances of helium or oxygen from these plots. A more rigorous examination of the linestrength patterns, which allows for the effects of gravity and microturbulence, must be performed in order to elucidate any evolutionary patterns in these lines.

© European Southern Observatory (ESO) 1999

Online publication: September 2, 1999
helpdesk.link@springer.de