## 2. Observations and spectra reductionThe observations of CrB were carried out in 1993-1995 with the coudé spectrograph of the 2.6 meter telescope of the Crimean Astrophysical Observatory, equipped with a CCD camera with red- sensitive GEC detector, pixel array. The spectral resolution is 45000. The typical S/N ratio is better than 200. The mean exposure time is about 20 - 30 minutes. Table 1 lists the observations of CrB.
The reduction of the spectra was made using the software SPE written by S. Sergeev at the Crimean Observatory. A detailed description of these operations is given in Sergeev (1996). For the lithium problem it is very important to determinate the wavelength scale with great accuracy. The majority of the observers of this spectral region make differential measurements of the position of the Li blend by using either some Fe I lines (Polosukhina & Lyubimkov, 1995) or the line of Ca I at 6717.685 Å (Smith et al., 1993) as reference lines. For metal-poor stars the method of differential measurements is correct, because the probability of blending is very low, and the accuracy obtained for the position of the Li line is high (about 0.003 Å). But in the case of CrB the spectrum is very rich of metallic lines and there are no Fe I lines free from blends. The accuracy of the determination of the Li line position becomes worse than 0.02 Å. Therefore, in order to obtain a larger precision for the wavelength scale, we used for the present analysis two comparison spectra of thorium (one above and one below the stellar spectrum). We used typically 8 to 12 comparison lines. The typical error for the position is equal to FWHM/(S/N). In our case, a line of the comparison spectrum has FWHM = 0.15 Å , S/N from 20 to 120, hence from 0.0075 Å to 0.0012 Å. The standard deviation of the dispersion curve from the polynomial fitting is 0.002 - 0.003 Å . Since we study a short spectral region, 31 Å, we used a first degree polynomial for fitting the dispersion curve. Hence we derived the wavelength scale from the comparison spectra for each stellar spectrum. After normalization of each spectrum to the continuum, we made the following operations: 1) All spectra of CrB were shifted to the position of the spectrum corresponding to phase .046 of the rotational period. This phase is near to the phase of crossover, and we use it as reference. The value of the shift was derived by cross-correlation function between each spectrum and that at phase .046. 2) We determined a mean spectrum from all the observations. We used the laboratory wavelength of Ca I 6717.685 Å as reference in the mean spectrum. The result of the first iteration was the value of the shift of the center of gravity of this Ca I line in the mean stellar spectrum to the position in the laboratory wavelength scale. Then each individual spectrum was shifted to the laboratory wavelength scale. The result of the next iteration using cross-correlation function for the individual spectra is the mean spectrum of CrB, with S/N = 1590 in the laboratory wavelength scale. The individual spectra are shown in Fig. 1.
The shifts of the individual spectra needed to obtain the mean spectrum were used for deriving the position and possible variations with the stellar rotation of the center of gravity of the Li feature and of some other lines. The mean spectrum is used for comparison with the synthetic spectrum and for the determination of the abundance of Li and other elements showing lines in this spectral region (Sect. 3). © European Southern Observatory (ESO) 1997 Online publication: July 3, 1998 |