Astron. Astrophys. 363, L33-L36 (2000)

3. Boron abundance analysis

A standard LTE abundance analysis was performed using a MARCS model atmosphere (Gustafsson et al., 1975; Asplund et al., 1997) with parameters K, , [Fe/H] , and km s^-1 adopted from Edvardsson et al. (1994). The increased relative abundances of elements in metal-poor stars was accounted for by adopting [, where even-atomic-number elements from C to Ti were considered to be elements.

Basic atomic line data (wavelengths, excitation energies, oscillator strengths, radiation damping parameters and energy level designations) for atomic and singly ionized lines in the wavelength region 2083 - 2092 Å were obtained from the VALD data base (Piskunov et al. 1995; Ryabchikova et al. 1999; Kupka et al. 1999; and a large number of references therein). The treatment and parameters for "van der Waals" broadening was for most strong lines obtained from the publications of O'Mara and colleagues, see Barklem & O'Mara (1998, and references therein). For the remaining lines correction factors were applied to the classically derived parameters as detailed in Edvardsson et al. (1993). The line oscillator strengths were then modified to give a good general fit of the line absorption in the wavelength region.

Simultaneously with the line data fitting, the absolute wavelength scale of the observed spectrum was determined by fitting to the synthetic spectrum. The final agreement for most of the strong lines is very good, and deficiencies in the line list must be blamed for less god fits. We estimate that the resulting absolute wavelength scale is better than mÅ. We have not fitted all lines rigorously, e.g. the line at 2091.7 Å. Fig. 1 shows the fit of the wavelength-adjusted observed spectrum to the synthetic spectrum.

Wavelengths and oscillator strengths for B I lines, including isotopic shifts, were adopted from Johansson et al. (1993). As an exercise and check we repeated the LTE analysis of the 2496.7 Å B I line from Edvardsson et al. (1994), with identical result: (LTE). Corrections for the simplifying but erroneous assumption of LTE for boron for both the 2496.7 Å line (+0.52 dex) and the 2089.6 Å line (+0.61 dex) were adopted from Kiselman & Carlsson (1996, Table 3). Synthetic spectra with (LTE) are compared with the observations in Fig. 2. This corresponds to a corrected abundance of (NLTE). It should be noted that the differential NLTE correction of 0.09 dex between the two boron lines is quite insensitive to uncertainties in the NLTE analysis (Kiselman, 2000).

Fig. 2. Top: The observed spectrum of HD 140283 in the 2089.6 Å region (solid line ). The three dotted lines show our synthetic spectra, one without any boron, and the two other ones with the B I abundance adopted from the 2497 Å line by Edvardsson et al. (1994): the bluer one with only 11B, and the more red one with only 10B. (The NLTE corrections of Kiselman & Carlsson (1996) are consistently taken into account.) The two lower spectra show the two possible two-pixel binned versions corresponding to the nominal resolution of the observations and a synthetic spectrum with as expected from high-energy spallation reactions. We can only determine an upper limit for the boron abundance and are proposing further observations to ascertain detection of the boron line

As seen in Fig. 2 we can not claim a detection of the boron feature. One might suspect a weak absorption near 2089.60 Å which would be best fit by a pure ¹⁰B line (the 25 mÅ isotope shift is indicated at the top of Fig. 2). We do not know, however, of any process which would be expected to produce predominantly ¹⁰B, and we consider the unexpected line position to be most probably due to a statistical fluctuation. We try instead to estimate an upper limit for the boron abundance: The width over which the expected boron feature should mainly depress the continuum is about 90 mÅ or 5 resolution elements (10 pixels). This region is indicated by the thick solid lines in Fig. 2. The integrated absorption equivalent width in this interval of the un-binned observed spectrum is 2.3 mÅ. With a per resolution element, we have for the line, which gives a equivalent width uncertainty of mÅ. From the discussion of the continuum level determination procedure and Fig. 1 and Fig. 2, we judge that the continuum level may at the most be wrong by 1.5%, which would correspond to 1.4 mÅ to the line. Adding these two uncertainties in quadrature makes an estimated equivalent width uncertainty of 1.6 mÅ. From these considerations we estimate that a upper limit of the equivalent width is mÅ. This value is just barely consistent with the abundance derived from the 2496.7 Å line by Edvardsson et al. (1994). Note, however, that the center-of-gravity of the possibly present feature is shifted by 25 mÅ from where it would be expected.

A simple Monte-Carlo simulation with 500 realizations of Gaussian noise added to a synthetic spectrum confirms the equivalent width uncertainty estimate above. It also indicates that a wavelength shift due to noise should be 30 mÅ.

In conclusion: The possible feature is about weaker and a little less than shifted in wavelength compared to the bona fide expected 4.0 mÅ equivalent width and position.

© European Southern Observatory (ESO) 2000

Online publication: December 5, 2000
helpdesk@link.springer.de