We now want to establish the relative calibration of the spectra in the various bands to be able to study the overall spectral energy distribution from the UV to the near-IR. It is not possible to derive these relative normalizations by simply using spectral observations because of several effects, such as slit losses, variability of the atmospheric transmission and uncertainties on the reference star colours. The usual method to normalize spectra taken in different bands is to use the broad-band colours of the target objects: spectra should be calibrated to reproduce the colours of the observed galaxies inside the spectroscopic aperture.
As this work is aimed at the construction of spectra representative of the average properties of the different classes of galaxies, we have calibrated the spectra by using the average colours of each class. This method guarantees that the final spectra closely match the average properties of the class and allows the reduction of the uncertainties on the colours as hundreds of galaxies can be used.
Of course, colours inside our rectangular spectroscopic aperture are not available. However, the dependence of near-IR colours and spectra with aperture is known to be very weak because the dominant stellar population at these wavelength is always the giant branch (with the exception of starburst galaxies, see for example, Maraston (1998)) and only the second-order metallicity and age gradients could be seen. Therefore we can use colours inside circular apertures as they are the only ones usually published, and the difference in shape between photometric and spectroscopic apertures is negligible if the two areas are chosen to be similar. The most similar circular aperture with enough published photometric data is the "effective" one, i.e., the aperture containing half of the light of the galaxy in the B band. Effective colours are available for a large number of galaxies and in many filters and is therefore possible to have large homogeneous data sets.
For the smallest objects, the effective aperture is similar to the spectroscopic one, for the largest galaxies the photometric aperture is larger. We have performed two checks to be sure that, even in the latter case, effective colours can reliably be used to calibrate the spectra: first we have compared spectra extracted from the central 10" of the slit with those of the outer part; second, we have compared the spectra of large and small galaxies of the same class (as NGC4350 having D0 = 2'.0 and NGC4382 with D0 = 8'.2) . In both cases no systematic differences are seen.
Once the aperture is chosen, we have also to select the luminosity of the galaxies used to derive average colours. As described in the previous section, the target objects are large, metal rich galaxies, and we want to use galaxies with the same properties to calibrate their spectra. In fact, smaller galaxies tend to be less metallic and therefore bluer in the optical colours (see, e.g., Fioc & Rocca-Volmerange, 1999), giving origin to the well-known colour-magnitude relation. We will use optical and near-IR colours appropriate for galaxies with MV < -21, corresponding to our sample.
Many photometry papers where used to derive these colours and check the results (Aaronson, 1978; Frogel et al., 1978; Persson et al., 1979; Griersmith et al., 1982; Glass, 1984; Giovanardi & Hunt, 1988; de Jong & van der Kruit, 1984; de Vaucouleurs et al., 1991; Buta et al., 1994, 1995a, 1995b; Prugniel & Hèraudeau, 1998, Fioc & Rocca-Volmerange, 1999). The optical colours are mainly based on the Prugniel & Hèraudeau (1998) catalog after correcting for Galactic dust using the RC3 value of the galactic extinction in B (de Vaucouleurs et al., 1991) and the Cardelli et al. (1989) extinction curve. The E and S0 optical-to-IR colours are based on Persson et al. (1979) and Frogel el al. (1978), while the colours of later spirals are mainly derived from Griersmith et al. (1982), de Jong & van der Kruit (1984) and Aaronson (1978). IR-to-IR colours are based on Glass (1984), Frogel et al. (1978) and Fioc & Rocca-Volmerange (1999). Note that the J-H and H-K colours of the early-spiral galaxies are redder than those of the ellipticals: a short discussion about this unexpected effect can be found in Fioc & Rocca-Volmerange (1999). All the colours were converted to Johnson's U, B and V, Cousins' R and I and CIT J, H and K. Great care was used in comparing different galaxy samples: whenever possible, the results from different samples were compared to each other to check for consistency and to discover any important selection effect. The results are listed in Table 2 together with the colour scatter and the number of galaxies used. The absolute magnitude threshold was reduced to MV = -20 for the Sd class (6 < T < 8) because these galaxies are on average intrinsically fainter than the other classes, and for the I class (T > 8) no magnitude restriction was used because of the large scatter observed. The colours of these latter two classes (not used for the templates and reported for completeness) are based on a smaller number of galaxies and are probably less accurate that the others.
|a: The J-H and H-K colours are based also on the results in Fioc & Rocca-Volmerange, 1999, where only average quantities are given. In these cases the scatter is not measured but estimated.|
|b: MV < -20|
|c: no magnitude selection.|
Earlier classes tend to be more homogeneous than the later ones. For the ellipticals, the spread of the colour distribution, when removing a few very discrepant objects, is comparable to the uncertainties on the single measure as quoted in the original papers, leaving little room to an intrinsic spread of the population due, for example, to different metallicity or star formation histories. On the contrary, for the later class the observed spread is several times larger than the measured uncertainties. As an example, for the ellipticals the V-K colour distribution has a standard deviation of 0.09 mag (and an error on the mean of the order of 1%), similar to the average estimated error on the single point of about 0.10 mag, while for Sc the observed spread is 0.24 mag.