Forum Springer Astron. Astrophys.
Forum Whats New Search Orders

Astron. Astrophys. 332, 459-478 (1998)

Previous Section Next Section Title Page Table of Contents

7. Discussion

7.1. Aperture bias

When performing a spectral classification, one must keep in mind that, in general, galaxy slit spectroscopy provides only a partial sampling of the objects. Here we comment on some of the biases which could be present in the spectra, and could produce an erroneous interpretation of the results. Two different phenomena produce what is called an "aperture bias". A first bias originates from the fact that the gathered light from galaxies at different distances, sampled with a slit of fixed width (in our case [FORMULA] to [FORMULA]), correspond to different spatial extensions (and stellar populations) in the corresponding galaxies (this is called "aperture bias" hereafter). A second bias (called "orientation bias" hereafter) comes from the varying spatial orientation of the slit with respect to the observed galaxies due to observational constraints (see Fig. 16). In order to study the aperture bias, Zaritsky et al. (1995) have made some simulations for characterizing the influence of the size of an optical fibre onto their spectral classification for galaxies with [FORMULA]. Following their conclusions, our slits of [FORMULA] to [FORMULA] should not cause any significant aperture bias for galaxies with [FORMULA]. Moreover, the slits used for the ESS observations contain [FORMULA] of the bulge and disc emission for a typical face-on spiral of [FORMULA] kpc in diameter at [FORMULA] (for spirals with other orientations, the fraction of the bulge and disk sampled is even larger). Therefore the results presented here are weakly affected by the aperture bias. For elliptical galaxies, the "orientation bias" is null, because the stellar constituents generally have uniform distributions along the galaxy extent. On the other hand, for spiral galaxies, the early and late stellar populations are not equally distributed along the galaxy profile, and slits oriented along the minor axis will over-sample the bulge of the galaxy with respect to the disc. In this case, and assuming a standard morphological-spectral relationship, the young stellar populations are under-sampled and therefore the spectral type would appear earlier than the morphological type, or at least, earlier than the real integrated spectrum. This effect could have important consequences in the fraction of types found as elliptical and spiral. Observing our CCD images, along with the slit orientations, allows us to conclude that an orientation bias could exist for spiral galaxies which have [FORMULA] (where only a fraction of the galaxy profile is inside the slit). However, less than 4% of the ESS galaxies have [FORMULA]. We emphasize that only a rigorous study of the data from spectro-imaging surveys can quantify both the aperture and orientation bias present in existing and future redshift surveys (see Hickson et al. 1994).

7.2. Comparison with other surveys

The fraction of different galaxy types found in the ESS sample may be compared to those found in other redshift surveys. Table 9 shows the fractions of the different morphological types found in the 6 other redshift surveys which provide the adequate information for comparison with the ESS data. For the ESS sample, we assimilate the spectral type with the corresponding morphological type from the Kennicutt average templates used for the classification. When examining Table 9, the reader should be aware of the different classification criteria used in the various surveys. Moreover, some morphological differences are subtle and can only be distinguished from high-quality imaging. The uncertainties in a given morphological type are in general not provided. However, some studies (Naim et al. 1995, and references therein) have shown that, when comparing the morphological classification of nearby galaxies by 6 different experienced astronomers, the r.m.s. scatter between two observers is typically 2-T units (de Vaucouleurs 1959). In the case of deep surveys, the errors in the classification are complicated by the appearance of new morphological types as in the case of HST images (see van den Bergh 1997), and/or by the image quality (Dalcanton & Shectman 1996). For the ESS, we are able to check and quantify the different error sources. The major error originates from the flux calibration, whose measured uncertainty leads to an absolute error in the type fractions of [FORMULA] (see Sect. 6.1).


Table 9. Different morphological mix obtained by other surveys. The ESS classifications are summarized in the last 2 rows. All the type fractions are percentages.

We first examine the non-uniform binning. Given our classification errors for elliptical and spirals ([FORMULA] 5% of the fraction in each class), the fraction of E/S0 is comparable to that given by RSAC and by the faint DWG (references in the Table 9). Clearly, we found more ellipticals than Griffiths et al. and than in the HDF, as expected from the fainter limiting magnitude of the 2 surveys. In these cases the large number of spirals is due to the combination of evolution and "morphological K-corrections": in the case of the HDF, the survey is sampling the UV spectral bands, which trace the peculiar morphologies related to the star-forming regions within the galaxies, (see van den Bergh 1997). On the other hand, the fraction of E/S0 in the ESS is smaller than the fraction found in the CfA1 and CfA2 surveys. For the CfA2 survey, the fraction of ellipticals is larger probably due to the fact that the eyeball classification is performed from photographic plates, and some proportion of spirals is likely classified as ellipticals due to the disappearance of the spiral arms when the bulges are saturated (Huchra et al. 1995). The fraction of spirals (Sa/Sb/Sc) in the ESS sample is quite similar to that for the RSAC but is significantly larger that in all other surveys. The number of Sa/Sb/Sc for the ESS sample probably includes some fraction of Sd galaxies, because this morphological type is absent from our classification (see Table 4 and Sect. 6.1).

It is interesting to note that if we consider the classification provided by the uniform binning in the [FORMULA] parameter (see § 6.1 and last row of Table 9), our fractions of E/S0 and Sa/Sb/Sc agree quite well with the respective fractions for the CfA1 survey. However, we emphasize that such a uniform binning system is not realistic, in the sense that we obtain a poor correlation between the spectral properties and the morphological type. This leads to the important issue of the discreteness of type definition. It has been shown (Morgan & Mayall 1957, Aaronson 1978, Abraham et al. 1994), and we confirm in this paper, that the spectral and morphological properties span non uniformly but continuously over most sets of parameters used for classifying galaxies. The assignment of a type in the Hubble system suffers from an artificial discretization which can lead to significant differences depending on the classification procedure which is adopted. This is the case for the E/S0/Sa types, for which the spectral properties show small variations due to the small changes in the stellar populations among these types (this is confirmed by the small range in [FORMULA] describing these types; see Fig. 8).

The ESS tends to have a similar type distribution to that observed in local or intermediate redshift surveys, as the CfA1, the RSAC, and the DWG survey. This seems to indicate that the galaxy distribution as a function of type is rather stable up to [FORMULA]. Other surveys indeed indicate that up to this redshift, only a marginal galaxy evolution is present (Hammer et al. 1995, Hammer et al. 1997, Lilly et al. 1996): the [OII] equivalent width does not increase significantly and galaxy types are roughly uniform. The comparison of the ESS spectral classification with other surveys constitutes not only a test of our classification procedure, but also provides a quantitative insight into the origin of the observed differences, which can be related to galaxy evolution and the well-known morphology-density relation (Postman & Geller 1984). These effects will be further investigated in forthcoming papers.

Previous Section Next Section Title Page Table of Contents

© European Southern Observatory (ESO) 1998

Online publication: March 23, 1998