Astron. Astrophys. 333, 619-628 (1998)

3. Optical spectroscopy with FLAIR and data analysis

For our observations, we used the wide-field multi-object spectrograph FLAIR II (fiber-linked array-image reformatter; see Parker & Watson 1995) on the UK Schmidt Telescope of the Anglo-Australian Observatory. This instrument has 91 fibers which are terminated with right-angled prisms. Each fiber has to be positioned in exact alignment with the selected target star visible on a film copy of the corresponding Schmidt plate and glued onto the film. The core diameter of each fiber is 100 µm, corresponding to on the sky. The fibers guide the light into the floor-mounted spectrograph and finally to a CCD camera with pixels.

FLAIR allows to observe up to 91 objects simultaneously in the field of view of the Schmidt plate. However, since adjacent fibers result in a cross-talk between the spectra on the CCD, we decided to use every second fiber only, reducing the number of available fibers to 46. In order to be able to subtract the sky background from the object spectra, 6 of these fibers were placed onto blank patches of the sky close to the objects. Thus, we could observe 40 target stars per field.

While FLAIR is a powerful facility in terms of multiplex advantage and area coverage, it also imposes some observational restrictions which influence the selection of target stars. One factor is that the magnitude difference between the brightest and faintest stars observed simultaneously has to be kept relatively small ( mag). Another factor is that not all individual fibers can be positioned fully as desired, since each fiber covers up an area of several square arcminutes on the plate. Furthermore, there is a minimum distance of about between neighboring fibers on the plate. Thus, in those cases where more than one candidate star was visible within the X-ray error circle, we usually could observe only one star, and always preferred the brightest star. Finally, we did not observe candidates for which another star could be seen within about on the Schmidt plate, since the fixed fiber aperture of does not allow to exclude the light from such a companion.

It has recently been shown that the determination of the equivalent width of the 6708 Å Li line can be problematic if the spectral resolution is too low. Low resolution (4 - 8 Å) spectroscopy can lead to a serious overestimation of the Li line width (Covino et al. 1997). Intermediate resolution of at least 2 Å is necessary to measure the equivalent width reliably (cf. Neuhäuser et al. 1997). We therefore used a 1200 line/mm grating, the highest dispersion grating available for use with FLAIR, which gives a spectral resolution of Å. The wavelength range covered in this configuration is quite small (6050 - 6850 Å), making it very hard to determine spectral types. We therefore took additional low resolution spectra using a 250 line/mm grating, which covered the wavelength range 3900 - 7200 Å with Å resolution.

Our observing procedure was as follows: First we took bias and dome-flat frames and arc spectra with a Ne and a Hg-Cd calibration lamp. We started with the 1200 line/mm grating and took a series of min exposures. Then, we changed the grating and took a series of min exposures with the 250 line/mm grating.

During our first observing run in June 1996 we could only obtain the intermediate resolution spectra for the western-most field due to very bad weather conditions. During our second observing run in June 1997 we could obtain intermediate and low resolution spectra for the other 5 fields. During two nights up to three of the individual exposures contained only very low signal due to moving clouds. However, for each series of exposures we had at least 9 usable frames. In total we took spectra for 88 X-ray selected candidates, 136 proper motion candidates, and additionally for 13 known PMS stars from the list of W94 and K98 in order to use them as spectral standards.

Bias subtraction and flat-fielding was done in the usual way with the corresponding standard IRAF ¹ routines. Then the individual frames of each series were summed and cosmic rays removed. We used the IRAF task dohydra for subtraction of scattered light, extraction of the spectra, and for wavelength calibration. The dome-flat frames were used for the throughput correction during the sky subtraction. Nearly all spectra have at least 10 000 counts per pixel in the continuum around 6700 Å. As an example for our data we show parts of several intermediate resolution spectra in Fig. 2.

Fig. 2a and b. Examples of our FLAIR spectra. Each plot compares spectra with similar spectral type but different S/N ratio and Li line strength. The spectra are arbitrarily scaled, the uppermost spectrum has the highest S/N ratio. Important lines are marked by the dotted lines. The upper plot compares three G type stars, the lower plot three M type stars.

From our FLAIR data we could obtain usable spectra ( counts per pixel) for 78 X-ray selected candidates (69 RASS sources, 9 sources from pointed ROSAT observations) and 115 X-ray quiet proper motion candidates. From the normalized intermediate resolution spectra we determined the equivalent width of the 6708 Å Li line with the IRAF task splot. In order to avoid mis-identifications due to possible blending with other nearby lines, especially the 6705 Å and 6710 Å Fe lines, we measured the central wavelength of the Li line and computed the wavelength difference to the 6717 Å Ca line. The true wavelength difference between these lines is 9.7 Å, and if the measured difference deviated from this value by more than 1 Å , we used the measured Li equivalent width as upper limit only.

The equivalent width was measured in two ways, first by integrating over the line profile, and second by fitting a Gaussian to the line profile. The mean of both values was used as the final equivalent width. For nearly all spectra the two values agree to within and we assume this to be the uncertainty of our measurement. For the few spectra with slightly less than 10 000 counts per pixel the uncertainties are higher; they are marked with `:' in Tables 1 and 2. Spectra with a considerably smaller number of counts (mainly due to problems with the fiber transmission) were not analyzed.

For all stars with strong Li lines we determined the spectral type from the low resolution spectra by comparison with the standard stars. We assume our spectral types to be accurate to 1 subclass in most cases.

© European Southern Observatory (ESO) 1998

Online publication: April 20, 1998
helpdesk.link@springer.de