4. Spectroscopic observations
High resolution spectra of the thirty-one candidates were obtained either with the Kitt Peak coudé feed spectrograph or with the ESO-CAT spectrograph at La Silla, Chile; six stars were observed at both observatories. The journals of the observations are given in Table 2 and Table 3 for Kitt Peak and ESO respectively. These tables contain the coordinates (Columns 2 and 3) and V magnitude (Column 4) of each star. The UT date, starting time and duration of each integration is given in Columns 5, 6 and 7. The S/N of each spectrum (Column 8) was determined by using the IRAF splot task which determined the () near the 4481 Mg II line in each spectrum. The measured heliocentric radial velocities and their rms errors are given in Columns 9 and 11 and the number of lines used is in Column 10. The agreement between the two sets of observations for the stars in common is satisfactory if we consider the number of lines that were available and also that three of these stars (HD 16456, HD 202759 and possibly HD 139961) are variable. The spectra of BD +00 0145 and HD 014829 have a significantly poorer quality than the others and were not used for a complete abundance analysis. We were able, however, to measure the equivalent width of the 4481 Mg II line in these spectra and so derive an approximate [Fe/H] for these stars as explained in Sect. 8.2.
Table 2. Journal of KPNO spectra of BHB star candidates.
Table 3. Journal of ESO-CAT spectra of BHB star candidates.
4.1. KPNO observations
The spectroscopic observations of the northern BHB candidates were made by Kinman and Harmer using the Kitt Peak 0.9 m coudé feed spectrograph. The long collimator (F/31.2; focal length 10.11 m) and camera 5 (F/3.6; focal length 108.0 cm) were used with grating A (632 grooves/mm) in the second order with a Corning 4-96 blocking filter. This gives a 300 Å bandpass covering 4260-4560 which includes both H, the Mg II 4481-line and a selection of Fe I , Fe II and Ti II lines. The detector was a Ford 3KB chip (30721024 pixels) with a pixel size of 15 microns. This gives a 3-pixel resolution of approximately 0.3Å. The nominal resolution at 4 500Å is therefore 15 000. Biases were taken at the start of each night and a series of flat field quartz calibration exposures were taken at the start and end of each night. ThAr arc lamp spectra for wavelength calibration were made at the start, end and at frequent intervals during each night. The spectra were reduced using standard IRAF proceedures of bias subtraction, flat field correction and the extraction of the [1-d] spectrum. The wavelength calibration was made using the ThAr arc spectrum that was closest in time to the program spectrum.
The spectra were normalized to the continuum level interactively by using an updated version of the NORMA code (Bonifacio 1989; Castelli & Bonifacio 1990). These normalized spectra were used to derive stellar parameters from the H-profile and for the comparison with the synthetic spectra.
4.2. ESO-CAT observations
The southern BHB candidates were observed by Bragaglia with the CAT + CES (Coudé Auxiliary Telescope, 1.4 m diameter + Coudé Echelle Spectrograph) combination at La Silla, Chile, during April and September 1995. This equipment gives a single echelle order which was observed with two different instrumental configurations. In April we used an RCA CCD (ESO #9), 1024 pixels long, covering about 40 Å at a resolution of 0.14 Å (or R 30 000), while in September the detector was a Loral CCD (ESO #38), 2688512 pixels, covering about 50 Å at a resolution of 0.11 Å (or R 40 000). In both cases the spectra were centered on the 4481 Mg II line. Integration times ranged from 10 to 70 minutes; the faintest stars were observed twice.
The ESO-CAT spectra also were reduced with standard IRAF proceedures. The extraction of the [1-d] spectra from the [2-d] images was performed weighting the pixels according to the variance and without automatic cleaning from cosmic rays. The wavelength calibration also was computed from a series of Thorium arc-spectra and is estimated to be accurate to a few hundredths of an Å. IRAF tasks were used to clean the spectra from cosmic rays and defects, for flattening and for normalization.
4.3. The measurement of the Kitt Peak (KPNO) and ESO-CAT spectra
In order to be able to compare the spectra with the models, they were transformed to zero velocity using the IRAF dopcor routine. Line positions and equivalent widths were obtained from the reduced spectra using the IRAF splot routine, approximating (or deblending, if necessary) lines with gaussians functions. When either two KPNO or two ESO-CAT spectra were available for the same star, they were measured independently and the values of the equivalent widths were averaged. The comparison with the synthetic spectra was made, however, with the spectrum of highest quality in order to compare H profiles, to derive stellar parameters, to test abundances (derived from the averaged equivalent widths), to test the microturbulent velocity and to derive the rotational velocities. Our measured equivalent widths, together with the derived abundances (Sect. 8), are given in Table 4 and Table 5 for the KPNO and ESO-CAT spectra respectively. The wavelengths and multiplet numbers in these tables are taken from Moore (1945).
Fig. 2 gives a comparison of the equivalent widths that we and other observers obtained from the spectrum of HD 161817. Fig. 2 (a) compares the equivalent widths obtained from the 1994 Kitt Peak spectrum of HD 161817 with those obtained from a spectrum that was taken with the same equipment at the end of the night of 1995 May 03 UT and which has a significantly poorer focus than any of our other program spectra; even in this case, the effect on the equivalent widths appears to be minor. The comparison in (b) is with the early photographic observations of Kodaira (1964) which were made with the Palomar coudé spectrograph (10Å/mm) and shows a fairly large scatter (presumably because of the low S/N of single photographic exposures) but the systematic differences are small. The comparison in (c) with the more recent photographic observations of Klochkova & Panchuk (1990), however, shows substantial differences in the sense that the equivalent widths of these authors are systematically too large with respect to the present measurements. On the other hand, the systematic agreement of our data for this star with those of Adelman et al. (1987) shown in Fig. 2 (d) is quite good. The Adelman et al. spectrum was derived from 12 co-added photographic spectra (6.5Å/mm) and has a resolution of about 25 000; a 100Å-section of this spectrum is shown in Fig. 3 (above) together with the KPNO spectrum of 1994 Sept 06 UT (below) 7. The noise in the KPNO CCD spectrum is such that some of the fainter lines (equivalent widths less than about 30 mÅ) can be quite distorted. Such lines can generally be recognized and omitted from our analysis.
Fig. 4 compares the equivalent widths for HD 74721, HD 86986 and HD 93329 from the KPNO spectra with those measured by Adelman & Philip (1994, 1996a) using the same equipment. The agreement is satisfactory except for the Fe II 4555 and Ti II 4563 lines in HD 74721. The KPNO equivalent widths give abundances that are in agreement with the other lines of these species and were preferred. Otherwise, these various comparisons give no evidence for significant systematic differences between our equivalent widths and those given by Adelman & Philip.
Fig. 5 compares the KPNO equivalent widths of HD 31943 and HD 180903 with those obtained from the higher resolution ESO-CAT spectra; the agreement is very satisfactory. The 28 spectra of the stars in our sample identified by us as BHB stars are shown for the spectral region 4 475 to 4 490 Å in Fig. 6; they are numbered as in Table 1.
The determination of the chemical composition of the BHB stars requires a knowledge of the parameters that govern the physical conditions in their atmospheres such as the effective temperature (), the surface gravity (), the microturbulent velocity () and also assumptions about convection. These parameters are determined from both spectroscopic and photometric data. The latter require correction for interstellar reddening and this can be determined in several ways. These different methods and the extent to which they agree are discussed in the next section.
© European Southern Observatory (ESO) 2000
Online publication: December 15, 2000