Forum Springer Astron. Astrophys.
Forum Whats New Search Orders

Astron. Astrophys. 339, 537-544 (1998)

Previous Section Next Section Title Page Table of Contents

4. Abundance analysis

The determination of elemental abundances is interlocked with the microturbulent velocity [FORMULA], which can be derived if a sufficient number of lines of one ion can be measured over a wide range of line strengths. In the optical O II lines are most suitable for this purpose as they are most frequent. For ROA 5701 we could measure equivalent widths of 22 O II lines, which yield a value of 20 km/s for [FORMULA]. Gies & Lambert (1992) note in their analysis of B-type supergiants that the high microturbulent velocities of about 20 km/s obtained from LTE analyses decrease to about 10 km/s if NLTE effects are taken into account. The microturbulent velocity of Barnard 29 , [FORMULA]=10 km/s, has been determined by Conlon et al. (1994). 2

4.1. Methods

Abundances have been derived using the classical curve-of-growth technique as well as a spectrum synthesis technique. In both cases we computed model atmospheres for the appropriate values of effective temperature, surface gravity, and cluster metallicity and used the LINFOR spectrum synthesis package for the further analysis.

curve of growth analysis
We calculated curves of growth for the elements of interest, from which abundances were derived. We took into account that in many cases more than one line contributed to the measured equivalent widths by treating those lines as blends in LINFOR. In addition we tried to avoid lines with significant contributions from other elements or ions.
spectrum synthesis
In a second trial we fitted the whole spectrum at once. In this mode the LINFOR package tries to fit the line profiles of the metal lines using a [FORMULA] test by adjusting the abundance of the element(s) that are fitted. We used the same line lists as for the curve-of-growth analysis.

4.2. Analysis of the optical spectrum of ROA 5701

For the analysis of the CASPEC data of ROA 5701 we used atomic data for C II (Yan et al., 1987), N II (Becker & Butler, 1989), O II (Bell et al., 1994) and Si III (Becker & Butler, 1990). Table 2 lists the results for individual lines. For C II we could only derive an upper limit, assuming an equivalent width of 10 mÅ for the C II line at 4267 Å . The spectrum synthesis resulted in abundances higher than those from the classical curve-of-growth analysis by about 0.1 dex for N II, O II, and Si III (cf. Table 5). Part of this offset is due to the fact that the spectrum synthesis has to use a "global continuum" for its fit (see also Sect. 2.3).


Table 2. Equivalent widths and abundances for C II (upper limit only), N II, O II, and Si III as derived from the CASPEC spectra of ROA 5701 .

4.3. Iron abundances

Our main aim is to determine the iron abundances of both stars from the Fe III lines in the UV. We used the atomic data given by Ekberg (1993). The measured equivalent widths and the resulting abundances for both stars are listed in Table 3. The spectrum synthesis yields iron abundances about 0.2 dex lower than those obtained from the equivalent widths.


Table 3. Equivalent widths and abundances as derived from the GHRS spectra of Barnard 29 and ROA 5701 . We list the equivalent widths that were measured for a global continuum. Multiplet numbers are from Ekberg (1993).

4.4. Error estimates

The iron abundances derived for ROA 5701 and Barnard 29 from different equivalent width measurements differ by up to 0.1 dex and 0.05 dex, respectively. The effects of differences in effective temperature, surface gravity, and microturbulent velocity are given in Table 4. To check the effects of using different line lists we also derived abundances using the line lists of Kurucz (1991, priv. comm., observed lines only). The derived mean abundances differed by about 0.05 dex from those given in Table 5. The errors given in Table 5 include those of Table 4 plus 0.05 dex (to account for possible errors in the line lists) and 0.1 dex resp. 0.05 dex in the iron abundances derived from [FORMULA] (to account for errors in the equivalent widths measurements).


Table 4. Error estimates


Table 5. Abundances derived for ROA 5701 and Barnard 29 . [FORMULA] gives the number abundance of the respective element with [FORMULA] = [FORMULA].
1) Conlon et al. (1994)
2) Dixon & Hurwitz (1998)

The stars we analyse are in a temperature-gravity range where NLTE effects start to play a rôle. Gies & Lambert (1992) and Kilian (1994) find that the NLTE abundances of C, N, O, and Si for B-type main sequence and supergiant stars are on average about 0.1 - 0.2 dex lower than the corresponding LTE abundances. As the NLTE corrections shift the abundances of all elements in the same direction the relative abundance trends remain unchanged within our error bars.

Prompted by a remark from the referee Dr. P. Dufton we investigated whether the low iron abundances that we derived from UV lines of Fe III might be due to systematic errors. For this check we obtained 6 high resolution UV spectra of the normal main sequence B star [FORMULA] Peg from the IUE final archive, which we coadded in order to increase the S/N. As line blending, continuum placement and incompleteness of atomic line lists are much more severe problems for solar metallicity B stars than for our (metal-poor) programme stars, we synthesized the spectral range in question using the entire Kurucz line list and varied the iron abundance. From the IUE data we get an iron abundance of [FORMULA] 3. This is at variance with the near solar value ([FORMULA], Pintado & Adelman, 1993) derived from the analysis of optical Fe III lines for this star, which we confirm from ESO CASPEC spectra. A similar discrepancy has been found by Grigsby et al. (1996) for the normal main sequence B star [FORMULA] Her, for which the iron abundance from ultraviolet Fe II and Fe III lines was found to be more than 0.47 dex below the near solar value obtained from the analysis of optical lines. Grigsby et al. argue that the lower iron abundance derived from the UV lines is the correct one, as the abundances for both ions agree, whereas the previous optical analyses found differences between the two ions of up to 1.0 dex.

In summary the UV spectral analysis of normal B stars is currently so severely hampered by line crowding, continuum definition and incompleteness of atomic data that it is impossible to draw any clear cut conclusions on systematic abundance offsets for our metal poor programme stars, which are much less affected by these problems. We therefore do not apply any offsets to the iron abundances derived from the GHRS spectra.

Previous Section Next Section Title Page Table of Contents

© European Southern Observatory (ESO) 1998

Online publication: October 21, 1998