The target galaxies were selected in the morphological classes E, S0, Sa, Sb and Sc. In each class we selected large nearby galaxies without indication for peculiar activity. An absolute magnitude threshold of MV < -21 was used to ensure a homogeneous metallicity and to allow a sample as close as possible to that in K96. We also mainly observed galaxies whose optical spectra were available in the literature to be also able to study single objects. Therefore we selected galaxies from the samples by K96 and Kennicutt (1992a, 1992b), adding one more elliptical galaxy from the list given in Goudfrooij (1994). The resulting 28 targets are listed in Table 1, together with the morphological informations derived either from the LEDA catalog (http://cisr.univ-lyon1.fr/~leda/) or from the reference listed in the table. The objects are sorted in the table and divided into 5 classes according to the morphological index T (see, for example, Buta et al., 1994).
|NGC||R.A. (1950) DEC.||Type||Ta||Ref.b||Bandsc|
|E (-5 T < -3)|
|S0 (-3 T < 0)|
|Sa (0 T < 2)|
|Sb (2 T < 4)|
|Sc (4 T < 6)|
|a: morphology stage index (e.g., Buta et al., 1994)|
|b: references. 1: Kennicutt (1992b); 2: Goudfroij (1994); 3: K96.|
|c: bands used for the templates|
We planned to have 5 galaxies for each class between E and Sc with good S/N ratio (SNR) to have an estimate of the intrinsic spread in the spectra of each class and to average out peculiar characteristics. This number was not reached for the Sa class because NGC2798 turned out to be peculiar (it is part of an interacting system) and with emission lines. This galaxy is listed in the "Others" section of Table 1 together with 3 more observed objects of later stages not used for the templates.
All the spectra were obtained at TIRGO, an IR-optimized 1.5m telescope on the Swiss Alps. We used the long-slit spectrometer LonGSp (Vanzi et al., 1997) which is based on a NICMOS3 256 × 256 pixels Rockwell array and is sensitive between 0.9 and 2.5 µm.
Each galaxy was observed through the largest available slit, 7" wide, in order to have a good coverage of the galaxy. Spectra were obtained in the J, H and K bands for integration times between 15 and 30 minutes for each band. The resulting spectral resolution is about 300 in J, 400 in H and 500 in K. The telescope was chopped every minute between the galaxy and the nearby sky to have a good sky subtraction. Each time the galaxy was placed on a different position of the slit to minimize the effect of bad pixels and distortions of the flat-field. For each galaxy, stars of spectral type between F8 and G8 were observed in order to measure the atmospheric transmission and the variation of the instrumental efficiency with the wavelength, as discussed in Maiolino & Rieke (1996). The total number of acquired images is about 7000.
All the spectra were sky subtracted using the average of the adjacent spectra, and divided by differential flat-fields obtained by imaging the dome. The resulting frames were then rectified, co-aligned and stacked together using a clipping algorithm. A one-dimensional spectrum was then extracted from the central 53" in order to maximize the observed fraction of the galaxies. The resulting aperture of 7" × 53" is similar to the aperture of 10" × 20" used by K96 to define their templates and allows us to merge the two data sets. At the average distance of the galaxies of our sample, 27 h-150 Mpc (h50 = H0 / 50 km/sec/Mpc), the spectroscopic aperture corresponds to about 1 × 7 Kpc. We estimated the fraction of the galaxy light collected by the slit by using the existing aperture photometry in the V band (Prugniel and Hèraudeau, 1998): on average, our spectra contain about 20% of the galaxy light, ranging from 7% for the closest and largest galaxies to about 35% for the smallest ones. Therefore our results should not be considered "total" spectra, even if the difference is expected to be negligible in the near-IR (see discussion below). This is more important when also the K96 spectra are considered as in the optical the radial gradients are expected to be larger (but see also the discussion in K96).
The "true" spectra of the reference stars were modeled by the solar spectrum (Livingston & Wallace, 1991) corrected for the slightly different temperature of the stars. The derived transmission curve was then applied to the galaxy spectra. The final spectrum of each galaxy in each band is then corrected for Galactic extinction, normalized, shifted to zero redshift, and averaged with the other galaxies of the same class. A clipping algorithm was applied to remove a few residual discrepant value due to cosmic rays, erratic bad pixels not present in the mask and uncorrected features of the reference stars. In Figures 1, 2 and 3 we show the average spectra and the relative standard deviations in the three bands.
Figure 1. Rest frame average spectra of each class of galaxies in the J band. The thick line is the average of the observed spectra, the thin lines show the ranges within 1 standard deviation. Arbitrary offsets were added to the spectra for clarity.
Figure 2. As Figure 1, H band.
Figure 3. As Figure 1, K band.
The differences between the spectra of the galaxies of the same class is due to both all the observational uncertainties (noise, residual sky contribution, distortion on the flat field, non perfect correction of the instrumental and atmospheric transmission) and the intrinsic differences between the observed galaxies. We can therefore have a robust upper limit of the total uncertainties of our spectra by looking at the resulting RMS among the galaxies of the same type. This quantity is between 0.7% and 1.4% in K, between 1.7% and 2.4% in H and between 2.4% and 3.6% in J, depending on the morphological class. The tendency to have larger spreads towards bluer wavelengths is probably dominated by larger intrinsic differences, and the value in K, about 1%, can be assumed as the real uncertainty in the single spectrum. Because of the different redshift of the galaxies, less galaxies contribute to the edges of the spectra and therefore the uncertainties tend to be larger in the external part (about 5%) of the bands. The parts of the spectra without indication of standard deviation in figures 1, 2 and 3 have been derived from only a galaxy and should be treated with caution.
If the normal galaxies show a high degree of uniformity, on the contrary large differences are found for the few galaxies with peculiar properties or of later types. The K spectra of 3 of these galaxies (NGC1569, NGC2798 and NGC4449) are shown in Figure 4, while for NGC3432 the SNR is too low. The galaxies in the figure show prominent emission lines whose brightness and ratios change much from one object to the other. In this case no attempt was made to create a template.