3 The data

The SED presented in this paper have been constructed using multifrequency data available in the literature or from our own observations, treated as consistently as possible, in order to produce an homogeneous data-set.

The UV data are taken from the FAUST (Lampton et al. 1990) and the FOCA (Milliard et al. 1991) experiments. In order to be consistent with our previous works, we transformed UV magnitudes taken at 1650 Å  by Deharveng et al. (1994) to 2000 Å  assuming a constant colour index UV(2000) = UV(1650) + 0.2 mag. This relation has been obtained by comparing the FAUST 1650 Å  with the SCAP (Donas et al. 1987) 2000 Å  UV magnitudes of 17 late-type galaxies in the Virgo cluster, observed by both experiments (Deharveng et al. 1994). FOCA magnitudes are from Deharveng et al. (2002), and Donas et al., in preparation. These are total magnitudes, determined by integrating the UV emission up to the weakest detectable isophote. The estimated error on the UV magnitude is 0.3 mag in general, but it ranges from 0.2 mag for bright galaxies to 0.5 mag for weak sources observed in frames with larger than average calibration uncertainties.

U, B and V photometry is generally derived from our own CCD measurements consistently with Gavazzi & Boselli (1996), as described in the appendix. When these are not available it is derived from aperture photometry taken from the literature. The (U,B,V) D₂₅ magnitudes, computed at the $\rm 25th~ mag~arcsec^{-2}$ isophotal B band diameter as in Gavazzi & Boselli (1996), have $\sim$10% uncertainty. They are on average 0.1 mag fainter than the total asymptotic magnitudes.

NIR data, from Nicmos3 observations, are taken mostly from Boselli et al. (1997) and Gavazzi et al. (2001). Magnitudes (J,H,K) are determined consistently with the optical magnitudes as in Gavazzi & Boselli (1996). The typical uncertainty in (J,H,K) is 10%. As for the visible magnitudes, they are on average 0.1 mag fainter than the total asymptotic magnitudes.

Mid-IR data, at 6.75 and 15 $\mu$m, are from Boselli et al. (2003). Flux densities have been extracted from ISOCAM images by integrating the emission until the weakest detectable isophote. Even if the mid-IR emission of these galaxies is less extended than in the visible and near-IR bands, ISOCAM data provide us with integrated flux denisties representative of the whole galaxy. The typic uncertainty on the ISOCAM data is $\sim$30%.

12, 25, 60 and 100 $\mu$m integrated flux densities from the IRAS survey are taken from different sources. The typical uncertainty in the IRAS data is $\sim$15%. Alternative Far-IR values at 60 and 100 $\mu$m from ISOPHOT, as well as 170 $\mu$m flux densities, are taken from Tuffs et al. (2002), with a typical nominal uncertainty of $\sim$10%. The comparison of ISO and IRAS data for the sample galaxies detected in both surveys reveals a systematic difference of $\rm ISO/IRAS=0.95$ and 0.82 at 60 and 100 $\mu$m respectively (Tuffs et al. 2002).

We collected radio continuum data at 2.8, 6.3, 12.6 and 21 cm from different sources. 21 cm radio continuum data, available for the whole sample, are mostly from the NVSS survey (Condon et al. 1998) (see Gavazzi & Boselli 1999). All radio continuum data are integrated fluxes. The typical uncertainty is $\sim$20%.

The photometric data for the whole sample are given in Table 3, arranged as follows:

• Column 1: VCC denomination.
• Column 2: UV magnitude at 2000 Å, uncorrected for galactic and internal extinction.
• Columns 3-8: U, B, V, J, H, K magnitudes in the Johnson system, determined as described in Gavazzi & Boselli (1996), uncorrected for galactic and internal extinction.
• Column 9: ISOCAM 6.75 $\mu$m flux density, in mJy.
• Column 10: IRAS 12 $\mu$m flux density, in mJy.
• Column 11: ISOCAM 15 $\mu$m flux density, in mJy.
• Column 12: IRAS 25 $\mu$m flux density, in mJy.
• Column 13: IRAS 60 $\mu$m flux density, in mJy.
• Column 14: ISOPHOT 60 $\mu$m flux density, in mJy.
• Column 15: IRAS 100 $\mu$m flux density, in mJy.
• Column 16: ISOPHOT 100 $\mu$m flux density, in mJy.
• Column 17: ISOPHOT 170 $\mu$m flux density, in mJy.
• Columns 18-21: radio continuum flux densities at 2.8, 6.3, 12.6 and 21 cm, in mJy.

All data given in Table 3 are observed quantities. The UV, optical and near-IR data are uncorrected for dust extinction, the mid-IR data for the contribution of the stellar component, the radio continuum data for the contribution of the nuclear emission.

References to the photometric data are given in Table 4.

Additional emission line data are given in Table 5, arranged as follows:

• Column 1: VCC denomination.
• Column 2: H$\alpha$+[NII] equivalent width, in Å.
• Column 3: logarithm of the H$\alpha$+[NII] line, in erg cm^-2 s^-1.
• Column 4: reference to the H$\alpha$ data.
• Column 5: logarithm of the HI mass, in solar units, defined as MHI = 2.36  10⁵ D² SHI, where D is the distance of the galaxy in Mpc (from Table 2), and SHI is the integrated HI flux, in Jy km s^-1.
• Column 6: the HI-deficiency parameter, defined as the ratio of the HI mass to the average HI mass of isolated objects of similar morphological type and linear size (Haynes & Giovanelli 1984); it is used to discriminate between "normal'' galaxies and galaxies suffering for gas depletion due to ram pressure. Galaxies with an HI-deficiency parameter $\leq$0.3 can be treated as unperturbed, isolated galaxies.
• Column 7: the quality of the HI profile: 1 stands for high signal to noise, two-horns profiles, 2 for high signal to noise Gaussian profiles, 3 for average signal to noise profiles, 4 for poor quality data and 5 for those objects whose profile is not available in the literature.
• Column 8: HI line width, measured as the average value of the width at 20 and 50% of the peak, in km s^-1.
• Column 9: reference to the HI data.
• Column 10: logarithm of the H₂ mass, in solar units, defined as in Boselli et al. (2002b), determined assuming a luminosity dependent X conversion factor between the CO intensity and the H₂ surface density $\log X = -0.38 \log L_{\rm H} +24.23$ mol cm^-2 (K km s^-1)^-1. As shown in Boselli et al. (2002b), the use of a luminosity dependent rather than a metallicity dependent X conversion factor does not affect the uncertainty in the determination of the molecular hydrogen mass.
• Column 11: reference to the CO data.

3.1 The extinction correction

UV to near-IR data have been corrected for galactic extinction according to Burstein & Heiles (1982). The galactic extinction $A_{\rm g}(B)$, taken from NED and listed in Table 7, have been transformed to $A_{\rm g}(\lambda)$ assuming a standard galactic extinction law (see Table 6): $A_{\rm g}(\lambda)=c(\lambda)$ $A_{\rm g}(B)$, where $c(\lambda)=k(\lambda)/k(B)$.

 

 

Table 6: Galactic extinction law.

Filter	$\lambda$	c($\lambda$)
	Å
UV	2000	2.10
U	3650	1.15
B	4400	1.00
V	5500	0.75
J	12 500	0.21
H	16 500	0.14
K'	21 000	0.10

The observed stellar radiation of galaxies, from UV to near-IR wavelengths, is subject to internal extinction (absorption plus scattering) by the interstellar dust. In order to quantify the emission of the various stellar populations, UV, optical and, to a lesser amount, near-IR fluxes must be corrected for dust attenuation. Furthermore, since dust extinction varies from galaxy to galaxy (according to their geometrical parameters such as the inclination, their history of star formation and metallicity), corrections appropriate to each individual galaxy must be determined.

Estimating the dust extinction at different $\lambda$ in external galaxies is however very difficult (it has been done only for the Magellanic clouds). Buat et al. (2002) have shown that, for example, the Calzetti's law calibrated on the central part of starburst galaxies (Calzetti 2001) strongly overestimates the extinction in normal, late-type objects. This difficulty is mainly due to two reasons: a) the extinction strongly depends on the relative geometry of the emitting stars and of the absorbing dust within the disc of galaxies. The young stellar population are mostly located along the disc in a thin layer, while the old populations forms a thicker layer. This point is further complicated by the fact that different dust components (very small grains, big grains etc.), which have different opacities to the UV, visible or near-IR light, have themselves different geometrical distributions both on the large and small scales. b) it is still uncertain whether the Galactic extinction law is universal, or if it changes with metallicity and/or with the UV radiation field. Detailed observations of resolved stars in the Small Magellanic Cloud by Bouchet et al. (1985) indicate that the extinction law in the optical domain is not significantly different from the Galactic one in galaxies with a UV field $\sim$10 times higher and a metallicity $\sim$10 times lower than those of the Milky Way. A steeper UV rise and a weaker 2200 Å bump than in the Galactic extinction law have been however observed in the LMC and SMC (Mathis 1990).

While the adoption of the Galactic extinction law for external galaxies seems reasonable (even though it is questionable for low-luminosity galaxies), no simple analytic functions describing the geometrical distribution of emitting stars and absorbing dust, both on small and large scales, are yet available.

The radiative transfer models of Witt & Gordon (2000) have however shown that the FIR to UV flux ratio, being mostly independent of the geometry, of the star formation history (the two radiations are produced by similar stellar populations) and of the adopted extinction law, is a robust estimator of the dust extinction at UV wavelengths. Here we will use this method to estimate the extinction correction in the UV, the wavelength most affected by dust.

We propose an internal extinction correction prescription similar to that described in Gavazzi et al. (2002a).

Our semi-empirical determination of A(UV) takes into account the scattered light. Following Buat et al. (1999), we estimate A_i(UV) from the relation:

$$A_i({\rm UV}) 0.466 + {\rm Log}{\rm (FIR/UV)} +0.433 \times \left(\rm Log(FIR/UV)\right)^2 ~~~~~~~~~~~~~~~~~~~{\rm [{mag}]}$$ (1)

where

$${\rm FIR} 1.26 \times (2.58 \times 10^{12} \times F_{60} + 10^{12} \times F_{100}) \times 10^{-26} ~~~~~~\left[{\rm Wm^{-2}}\right]$$ (2)

F₆₀ and F₁₀₀ are the IRAS FIR fluxes (in Jy) and

$${\rm UV}=10^{-3} \times 2000*10^{({\rm UV}_{\rm mag}+21.175)/-2.5} ~~~~~\left[{\rm Wm^{-2}}\right].$$ (3)

$A_i(\lambda)$ can be derived from A_i(UV) once an extinction law and a geometry for the dust and star distribution are assumed. We adopt the sandwitch model, where a thin layer of dust of thickness $\zeta$is embedded in a thick layer of stars:

$$A_i(\lambda)=-2.5 \cdot \log\left(\left[\frac{1-\zeta(\lambda)}{2... ...}^{-\tau(\lambda) \cdot {\rm sec}(i)}\right)\right) ~~~~~~~~~~~~~~~~[{\rm mag}]$$ (4)

where the dust to stars scale height ratio $\zeta(\lambda)$ depends on $\lambda$ (in units of Å) as:

$$\zeta(\lambda)=1.0867{-}5.501 \times 10^{-5} \cdot \lambda.$$ (5)

Relation (5) has been calibrated adopting the average between the optically thin and optically thick cases with $\lambda$ dependent dust to star scale height ratios given by Boselli & Gavazzi (1994). Observations of some edge-on nearby galaxies show that it is still unclear whether $\zeta$ depends or not on $\lambda$ (Xilouris et al. 1999). As shown in Gavazzi et al. (2002a), however, similar values of $A_i(\lambda)$are obtained in the case of a sandwitch model and of the extreme case of a slab model ($\zeta=1$), meaning that the high uncertainty on $\zeta$ is not reflected on $A_i(\lambda)$.

In the case of the UV band ( $\lambda=2000$ Å), $\zeta=1$, and Eq. (4) reduces to a simple slab model. In this case $\tau$(UV) can be derived by inverting Eq. (4):

$$\tau({\rm UV})=[1/{\rm sec}(i)] \cdot \big( 0.0259+1.2002 \times A_i({\rm UV})$$

$$\left.+1.5543 \times A_i({\rm UV})^2-0.7409 \times A_i({\rm UV})^3 +0.2246 \times A_i({\rm UV})^4\right)$$ (6)

using the galactic extinction law $k(\lambda)$ (Savage & Mathis 1979), we than derive:

$$\tau(\lambda) = \tau({\rm UV}) \cdot k(\lambda) / k({\rm UV})$$ (7)

and we compute the complete set of $A_i(\lambda)$ using Eq. (4).

FIR/UV is available for 44 objects. If FIR or UV measurements are unavailable we assume the average values $A_i({\rm UV}) = 1.28$; 0.85; 0.68 mag for Sa-Sbc; Sc-Scd; Sd-Im-BCD galaxies respectively, as determined when FIR and UV measurements are available.

Once corrected adopting the aformentioned prescription, we checked empirically that the SED do not contain a residual dependence on galaxy inclination. The corrected SEDs of 32 Sc galaxies, binned in 4 intervals of inclination, and their fit parameters were found very consistent one another. The galactic and internal extinction correction (in magnitude) for the observed galaxies are given in Table 7.

This empirical attenuation law gives a zeroth order estimate of the attenuation in the UV regime, the most affected by dust. We stress however that the shape of the corrected spectrum, in particular at UV wavelengths, is still uncertain. This is due not only to the lack of observational constraints other than the 2000 Å  flux, but also to the large uncertainties on the relative geometrical distributions of dust and stars and on the extinction law, which might significantely depend on the UV field and metallicity in this wavelength regime.