4. Abundance analysis
In the investigation of elemental abundances we have employed both LTE and NLTE analyses. For complete information concerning the NLTE calculation for the C II and N II spectrum we refer the reader to Paper Iab, where a detailed description of the applied NLTE code (an updated implementation of Carlsson  MULTI code), atomic models, oscillator strengths, and photoionization and collision cross sections for all considered transitions are given. In principle, the LTE abundances could be also obtained with the MULTI code, but this appears to be a rather time consuming approach so we decided to use Kurucz's well-known WIDTH9 code for the LTE analysis. A test calculation showed that both MULTI and WIDTH9 yield identical LTE abundances.
Photoionization cross-sections were mainly taken from the Opacity Project (Yan et al., 1987). Our calculations maintain the detailed structure of their frequency dependence, including resonances.
In the present work, we have modified the Stark broadening constants used in the calculations relative to those used in Paper Iab. Their influence on the resulting abundances is quite significant. Therefore, we have paid special attention to this part of the analysis. To calculate the Stark parameters for the considered transitions, we used semiempirical formula provided by Dimitrijevi (1997) for the full width at the half maximum (FWHM):
Here is an effective principal quantum number and l is an angular momentum quantum number. Calculations using this formula were performed for T=20000 K.
It should be noted that recently obtained experimental data on Stark parameters are in excellent agreement with the predictions of Eq. 1 (see, for example, estimates made by Sarandaev & Salakhov  for the C II lines 6578 Å and 6583 Å and by Milosavljevi et al.  for the N II 4630 Å line).
4.2. Abundance analysis and results
The abundance analysis was carried out on the sample of carbon and nitrogen lines listed in Table 3. For the LTE analysis based on equivalent widths only unblended lines were selected. As the NLTE code enables the calculation of a synthetic spectrum, some blends (consisting of close lines of the same element) were also treated in the analysis. NLTE abundances were derived by fitting calculated and observed profiles as well as by using equivalent widths using the method described in Paper Iab. In Fig. 1 we show a comparison of calculated and observed profiles for a sample of lines. While fitting the profiles, projected rotational velocities were estimated (see Table 2).
Table 3. Equivalent widths of selected lines and their parameters
Abundances were calculated after the microturbulent velocity determination. The microturbulent velocity was determined by applying the usual condition that there should be no dependence between the calculated abundances from individual carbon or nitrogen lines and their equivalent width. Note, that there exists a difference between values determined under the LTE and NLTE approaches. As has often been found, the NLTE approach requires smaller values of the microturbulence parameter. This procedure has not been employed for all the stars. If a sufficient number of lines were not available, then = 3 (LTE case) or = 2 (NLTE case) was adopted. Average values for the carbon and nitrogen abundances are given in Table 4.
Table 4. Abundances of carbon and nitrogen
Let us compare the results on carbon and nitrogen abundances in HD 886 () obtained in the present study with those given in our previous work (Paper Iab). The previous result for carbon was , while present study yields . A similar comparison for nitrogen gives and respectively. The dominant reasons for the differences are differing stellar parameters, Stark broadening, and equivalent widths. In previous study we used the atmospheric parameters for HD 886 given by Gies & Lambert (1992), i.e., =22600 K and =4.0, while in present work we have used updated photometric calibrations for the temperature and gravity determination. As a result we obtained: =21600 K and =3.75. These latter atmospheric parameters we consider as being the more accurate. Stark broadening parameters greatly influence the derived abundances and in this study we recalculated the Stark broadening parameters using Dimitrijevi's updated formula which makes the present results more reliable. Finally, note that the previous study of was based on a single CCD spectrum. A comparison of the equivalent widths of carbon and nitrogen lines measured in that spectrum with those measured in spectra from the present analysis (as well as with the data of other authors, e.g. Gies & Lambert, 1992) shows that equivalent widths of some lines measured in Paper Iab were slightly overestimated (most probably due to an instrumental effect). Nevertheless, the detected abundance differences are not very pronounced, and indicate that the uncertainty in the abundances is at about the 0.1 dex or less level. Qualitatively, Paper Iab gives the same result as the present more reliable study.
© European Southern Observatory (ESO) 1999
Online publication: October 4, 1999