Astron. Astrophys. 353, 666-690 (2000)

3. Observations and reduction problems

The first step in an investigation of the Ca II IRT lines in stellar spectra is to consider the solar reference case. This is the case for which the best observations are available and the physics of which is best known. Many stellar spectroscopic analyses are carried out differentially with respect to a solar spectrum recorded under the same observational conditions as the program stellar spectrum. In this first part devoted to the solar case, the computed synthetic profiles are compared with both the best quality solar flux spectrum available (the Solar Flux Atlas of Kurucz et al. 1984) and a reflected sunlight spectrum recorded under "stellar" conditions with a typical high resolution stellar spectrograph. G. Cayrel de Strobel has accumulated a large number of reference solar (moonlight) spectra of the region of the first two Ca II IRT lines, which she kindly allowed me to use. Among these, some were recorded with the coudé spectrograph of the Canada-France-Hawaii (CFH) telescope atop Maunea Kea, Hawaii, a very dry site at 4200m altitude, and are almost devoid of water vapour absorption. One of them, recorded in 1983, was selected as the reference spectrum for this study.

Many of the investigations of the Ca II IRT in stellar spectra have ignored the third line at 8662.14 Å, probably for convenience because the two other lines at 8498.02 and 8542.09 Å fit on a single frame in modern high-resolution spectrographs. Moreover, as shown by Smith & Drake (1988) for the solar spectrum, this third line is the only one for which a measurable asymmetry gives some evidence of a small but significant perturbation by one of the hydrogen Paschen lines (here, P13) which fall in the red wings of the three Ca II lines. Often, stellar investigations concentrate exclusively on the strongest of the three Ca II IRT lines at which is the least blended and conveys all the information which can be extracted from the triplet.

3.1. Water vapour contamination

Contrary to what is claimed in a number of papers, the whole wavelength region is in fact appreciably polluted by telluric water vapour absorption. In Fig. 1, the selected CFH solar spectrum (solid line) is compared with a spectrum recorded also by G. Cayrel de Strobel but with the Aurélie coudé spectrograph of the Observatoire de Haute Provence (OHP) 1.52m telescope in 1990 (dotted line) during a period of exceptionally high water vapour absorption. The bottom box of Fig. 1 is just a close-up of the same data centered on the most important Ca II line. These figures illustrate how important it is to try and minimize the water vapour contamination and show which regions of the continuum and line profiles stay unperturbed.

Fig. 1. Telluric water vapour lines in the region of the Ca II IRT lines. Top: the solid line shows a spectrum recorded under very dry conditions and the dotted line shows a spectrum recorded under conditions of exceptionally high water absorption. Bottom: same data enlarged for the line region.

3.2. Definition of the continuum location

An accurate definition of the location of the local continuum is essential if we want to carry out precise fits of the observations by synthetic calculations. On their highest quality and resolution solar intensity spectra of the 8450-8702 Å region obtained with the Fourier Transform Spectrometer at the Mac Math Solar Telescope (Kitt Peak), Smith & Drake (1988) identify ten continuum points where the depression is less than 0.5%, with a huge gap of 105 Å (8473.0-8577.5 Å) where no real continuum point is found. In the same way as for the intensity data, on the Solar Flux Atlas the continuum has been carefully set from the consideration of very long wavelength scans. This cannot equally carefully be done on typical stellar spectrograms. Depending on the resolution mode, a single frame recorded with a typical stellar high resolution spectrograph in current use covers a spectral region of width varying between 75 and 200 Å. Even after applying the usual "flat-fielding" techniques, the spectrograms are generally far from being flat and a reasonable approximation of the continuum requires at least a third order polynomial or sliding polynomials or splines defined by at least 5 to 6 reference points. This implies average distances between reference points of 15 to 30 Å. This lack of real continuum window in the vicinity of the first two Ca II IRT lines is further illustrated by the solar synthetic spectra in Fig. 2. The situation gets more and more difficult as one goes towards cooler stars. Fig. 1 of Zhou (1991) shows how hazardous the continuum normalization can become for stars of the late-K and M types.

Fig. 2. Continuum location. Dotted line: observed CFH spectrum normalized "classically" through local higher intensity points. Thin solid line: synthetic spectrum with only the Ca II IRT lines. Thick solid line: same synthetic spectrum, but taking into account the absorption by hydrogen Paschen lines.

On these short "stellar" wavelength scans, it seems indeed almost impossible to accurately define the run of the continuum without recourse to some kind of theoretical prediction. Therefore, as in usual reduction procedures, I first identified windows of higher local intensity, but then normalized the observed spectrum by assigning to them the residual depths predicted by a "best fit" synthetic spectrum. Formally, the predicted synthetic depths depend on the model atmosphere and line data used in calculating the synthetic spectrum; the continuum normalization is thus an iterative process. The sensitivity of the local pseudo-continuum points to the choice of the model atmosphere turns out to be weak, so that the normalization process converges very quickly and needs not be repeated for relatively small changes of the parameters of the synthetic spectrum calculation. We shall see in Sect. 5 that explicitly taking into account the absorption by the hydrogen Paschen lines in the region results mainly, in the solar case, in appreciably lowering the apparent continuum level (Fig. 2). Thus, for consistency, when it is deemed necessary to take into account the contribution of Paschen lines, this must be done also for the continuum normalization of stellar observations. Experience shows that the constraints introduced by the fit to the calculated pseudo-continuum windows does not lower the diagnosis capabilities of the Ca II IRT profile fitting.

For stars of the late-K and M types, detailed synthetic calculations of the contribution of the molecular bands, such as those reported by Erdelyi-Mendez & Barbuy (1991), should be necessary.

© European Southern Observatory (ESO) 2000

Online publication: December 17, 1999
helpdesk.link@springer.de