2 The method

Different methods have been used to find extragalactic AGB stars (e.g. Richer 1989; Groenewegen 1999): (i) low dispersion objective prism observations (Westerlund 1978); (ii) optical follow-up spectroscopy on stars identified in colour-magnitude diagrams (Cannon et al. 1981; Aaronson & Mould 1980); (iii) grism surveys (Westerlund 1979); and finally (iv) narrow band photometry, described in more detail below. While (i) and (iii) are normally impossible in crowded regions, (ii) is a perfect follow-up method, but with few exceptions inefficient or not feasible in larger systems (often a few 10³interesting objects). Since photometry in crowded regions is still possible, method (iv) turns out to be quite powerful and plays an important part for statistical studies and the pre-selection for spectroscopic follow-up programmes in external galaxies.

Wing & Stock (1973) first proposed, that a subset of the eight-filter system of Wing (1971) could be very useful in a search for and chemical characterization of AGB stars, and it has proven to be quite successful. A description can be found e.g. in Palmer & Wing (1982) or Cook & Aaronson (1989). Later the method was developed and refined for CCD use by two groups (led by H. Richer and M. Aaronson). It uses conventional V and I filters as temperature indicators, whereas two narrow band Wing-filters around 800nm provide low-resolution spectral information for discriminating between O- and C-rich stars.

Figure 1 shows the filter curves of the narrow band filters (we bought, to start our search for AGB stars in Local Group galaxies), together with typical AGB star spectra demonstrating the principle of the selection method. The first filter is centered on a TiO band, which is prominent in M-stars and on a relatively clean piece of "continuum'' in C-stars, while the second one is centered on a CN band, which is prominent in C-stars and on the "continuum'' in M-stars. This gives a strong contrast between the measures taken in these two filters, and the colour index (TiO-CN) can efficiently separate the two spectral types.

Figure 1: Transmission curves of the two narrow band filters, and typical spectra (Schultheis 1998) of a C3.2 III (upper) and a M6.3 III star (lower)

The synthetic colours in Fig. 2 are calculated by folding the measured response curves of our new narrow band filters with the CCD QE curve (taken from the NOT web-documentation) and spectroscopic data from the literature: synthetic M-giant spectra from Fluks et al. (1994), synthetic hydrostatic C-star spectra from Loidl (priv. communication; C11026 e.g. means C-star with ${\rm C/O}=1.1$ and $T_{\rm eff}=2600$K), and the remaining from the spectrophotometric atlas of Pickles (1998). While the features of TiO/CN do not have significant strengths for stars bluer than $(V-i)\approx 1.5^{\rm mag}$ [and therefore (TiO-CN)$\approx$0 $^{\rm mag}$], they become prominent for later stars. Their strengths increase strongly with the spectral type, and so does the colour index (TiO-CN). In the colour-colour diagram (Fig. 2) a clear bifurcation between M- and C-stars can be seen towards the red end of the sequence. It can be used to identify and characterize them.

This method works well, as was already shown e.g. by Brewer et al.(1995, 1996). We plan to use it to investigate the AGB populations of Local Group galaxies.

Figure 2: Colour-colour diagram with synthetic colours for different spectral types and chemistries. Early stars have no TiO/CN-features, but late-type stars can be clearly separated into O- and C-rich by photometric means