4. Element abundances
Table 2 compares the line abundances determined for unblended and only weakly blended Sm II lines for different cases. The first case corresponds to a 8.6 kG magnetic field and zero and model which is adopted for the final abundance analysis of HD 166473. In the two other cases we try to simulate the magnetic field broadening of lines with non-zero and , an approach which was frequently used in the past.
Table 2. Line abundances for unblended and only weakly blended Sm II lines for different models of magnetic broadening. Velocities are in km s-1.
It is interesting to note that with such a large field of 8.6 kG the magnetic intensification is not proportional to . Lines with a relatively large and a configuration close to a pseudo-triplet (e.g. 6426.62) split nearly completely in their components, which results in a low magnetic intensification. On the other hand, the 6604.53 line (see also Fig. 4) is an example for a significant magnetic broadening caused by a relatively small but complex Zeeman pattern. It is obvious that the more physical model results in a clearly smaller scatter of the line abundances. This figure illustrates also a large range of formal values (15 to 30 km s-1) needed to fit individual lines in low resolution spectra, if the magnetic field is ignored. Such an unrealistic situation of a variable is a direct indication for the presence of a strong magnetic field.
Table 3 summarizes the abundances of all elements investigated for HD 166473 with a model of = 7700 K, = 4.20, [M/H] = 0.5, = 0, = 0, and a magnetic field of 8.6 kG. For comparison we also give the abundances obtained by a standard analysis, approximating magnetic field effects with a increased to 1.0 km s-1 and a mean = 18 km s- 1 (fifth column). The microturbulence was estimated from Fe lines only.
Table 3. Elemental abundances for the roAp star HD 166473 ( = 7700 K, = 4.20, [M/H] = 0.5, = 0, = 0) normalized to the total number of atoms with formal error estimates in units of the last digit in parenthesis. Comparison with the solar abundances (Anders & Grevesse 1989), and with an abundance analysis based on an atmosphere without a magnetic field, but ranging from 15 to 40 km s-1 and = 1.0 km s- 1 which are needed to approximate magnetic field effects. Furthermore, VALD-I was used for the latter analysis to extract the atomic parameters, but VALD-II was used for the final analysis (second column). This difference is relevant in particular for C and O, because of the new values () (Wiese et al. 1996) which are now included in VALD-II . The last column lists abundances differences with and without magnetic field effects taken into account.
As expected, using a physically more realistic model by including effects of magnetic intensification gives lower abundances. The difference strongly depends on the adopted value of the microturbulence in the classical analysis (see Table 2). With a carefully chosen value we may decrease this difference to a minimum and come quite close to a full analysis. This is a comforting result, because within the error range of models and observations used, our previous analyses of moderate magnetic roAp stars do not have to be corrected. In some cases a nonmagnetic analysis provides even smaller abundances which may be explained with typically small Landé factors (Ce II lines) or differences in the oscillator strengths used (e.g., Dy II , Er II ).
As a last comment, the abundance pattern for the heavier elements almost perfectly fits the odd-even rule (Fig. 5).
4.1. Light elements
Similar to Cir, HD 203932 and Equ, the light elements C and O are underabundant. Na, Mg, Si, and S have nearly solar abundances, Al is overabundant as in Equ. But we observe solar Al abundance in roAp stars HD 203932 and HD 24712.
Some of the statements in the previous paragraph might have to be modified when better data become available, because they are based on only few and/or weak lines. New oscillator strength data for CNO elements were taken from the recent NIST compilation (Wiese et al. 1996).
The oxygen abundance is derived from the famous O I triplet which suffers from NLTE effects and usually provides an incorrect overabundance when treated in LTE. Nevertheless, due to the extreme weakness of this triplet in HD 166473 we do not apply any NLTE corrections.
The sodium abundance is based on three weak, low accuracy lines. The resonance lines, however, give an order of magnitude smaller Na abundance. NLTE corrections are about -0.1 dex for main sequence stars (Mashonkina et al. 1993) and hence cannot be responsible for this large discrepancy. A model atmosphere based on ODFs also does not provide agreement between resonance and weak lines. In addition, stratification effects may play a significant role in the atmosphere, as was demonstrated for the CP star 53 Cam by Babel (1992).
A lower Mg abundance is obtained from 3 strong Mg I lines (5167, 5172, 5183 Å), but two other Mg I lines (5528 and 5711 Å) give abundances which are close to those obtained from Mg II lines. We can slightly decrease this discrepancy by applying a NLTE correction to the Mg I lines of dex (Mashonkina et al. 1996).
The sulphur abundance is based on four groups of S I lines centered on 6173.6, 6743.5, 6748.7, and 6757.0 Å. The first group is resolved only in the high resolution spectrum. Although all groups provide the same sulphur abundance, the last three groups show a shift of +0.13 Å relative to the rest of the spectrum, which is yet unexplained.
4.2. Iron peak elements
The oscillator strengths for the iron peak elements were taken from the new release of the Vienna Atomic Line Database, VALD-2 , (Kupka et al. 1999, Ryabchikova et al. 1999a). The main new data sources are Lawler & Dakin (1989) for Sc II , Biémont et al. (1989) for V II , O'Brian et al. (1991), Bard et al. (1991), Bard & Kock (1994) for Fe I , and Blackwell et al. (1989), Wickliffe & Lawler (1997a) for Ni I .
We find the abundance pattern of the iron peak elements to be very similar to that of the other stars investigated by us so far: Ca, Ti, Fe, and Ni are the least enhanced elements and their abundances are close to solar. Cr and Mn are more abundant, V and Co are strongly overabundant. Furthermore, the prominent overabundance of Co, determined already in all other roAp stars, is confirmed. Sc is overabundant in HD 166473 in contrast with the underabundance of this element in the other stars in our sample.
The total abundance of all iron peak elements, from Ca to Ni, results in a metallicity [M/H] = +0.30 which is closer to the chosen model atmosphere with [M/H] = +0.50 than to the value estimated previously from photometry, which was [M/H] = +1.00.
4.3. Sr, Y, Zr, Ba, and rare-earth elements
Sr, Y, and Zr are overabundant in HD 166473 with Y being the most anomalous element in the whole group. In contrast, Ba is deficient - as is also the case for Cir.
The large overabundance of the rare earth elements (REE) allowed for the first time the derivation of abundances for 13 out of 14 stable REE. In the atmosphere of cool Ap stars lines of the doubly ionized REE are dominant, and therefore potentially the perferable source for an abundance determination. Unfortunately, transition probabilities have been available only since 1997. Bord et al. (1997), Wyart et al. (1997), Cowley & Bord (1997) provided calculations for the transition probabilities of Ce III , Er III , and Nd III . For Pr III we used calculations kindly provided by D. Bord (private communication). For the first ions of the REE we used oscillator strength data extracted from VALD-2 . They are the same as in VALD-1 (Piskunov et al. 1995) with the exception of Dy II (Biémont & Lowe 1993), Tm II (Wickliffe & Lawler 1997b), and Lu II (Bord et al. 1998, Den Hartog et al. 1998). Abundances of all REE obtained from the lines of the first ions exceed on average the solar abundances by +2.8 dex. There is a marginal tendency for the lighter REE (La to Nd) being less overabundant than the heavier REE (Sm to Er).
The most striking result obtained from our REE analysis is the abundance from the second ions. They exceed the values obtained from the lines of the first ions by +1.2 dex (Pr), +1.5 dex (Nd), and +0.7 dex (Er), respectively. The same relative overabundance of Nd was obtained by Cowley & Bord (1998) for another roAp star, Equ. This observational fact cannot be explained by a wrong absolute scale of the oscillator strengths, because a similar analysis for the non-pulsating star CrB did not indicate a significant deviation from the ionization equilibrium between the first and the second ions of the REE (Ryabchikova et al. 2000) which also means that hyperfine structure and isotopic shifts cannot be responsible for the observed differences. Pr has one stable isotope, Nd has 6 (and one with -yr half-life), Er has 6 stable istotopes and all three elements show similar relative overabundances of doubly ionized REE. The accuracy of the oscillator strength calculations is 0.15 (Cowley & Bord, 1997). The analysis of Pr and Nd lines in spectra of roAp stars definitely gives evidence for an unexpected strength of the lines of Pr III and Nd III exclusively relative to non-roAp stars. In addition, Malanushenko et al. (1998) found for Equ that maximum pulsation radial velocity amplitudes are observed for these 3rd spectrum REE lines.
© European Southern Observatory (ESO) 2000
Online publication: March 28, 2000