6. Advantages of the structural classification scheme
6.1. Low subjectivity
In the first step of the classification work, where we compute the various profiles, the only (very minor) subjective task is to select foreground or background objects to be masked because they contaminate the galaxy brightness.
The second step, the visual detection of the morphological components, is still partly subjective, in particular when we are classifying galaxies on the borderline between classes. But nothing prevents us from making this part of the analysis automatic, by introducing profile templates which mark the expected behavior for the morphological components, thus making it fully objective and reproducible. However, the complexity of Nature, which produces galaxies with disks, halos, bulges and bars with a variety of ellipticity and surface brightness profiles, and which mixes them and shows them to us projected on the sky under different angles, makes this automatic approach too complex just for determining the morphological type.
The third step of the classification method, assembling the structural components to derive the morphological type, is of course fully automatic.
The subjectivity arising from the visual detection of morphological components in our method is always lesser than in the traditional estimation of the types, since the detection of morphological components is easier on profiles than on direct images. Therefore, the number of difficult cases of classification is reduced. For example, deviations from de Vaucouleurs' law are easily visible on a surface brightness vs. profile, used by us, much more so than on direct galaxy images, used by classical morphologists. This allows us to detect disks in many almost face on S0s. In fact, the ellipticity distribution of Dressler's (1980) S0s in Coma shows a clear bias against round (face-on) galaxies (Michard 1996). Using our criteria to classify the galaxies, 1/3 of Dressler's (1980) Es move into the S0 class, thus strongly reducing the bias (Michard 1996).
The residual role played by the subjectivity can be estimated by comparing the rate of agreement among Hubble types estimated by different authors from the same galaxy profiles.
A more objective method should produce highly reproducible results. To check the reproducibility, and to understand the effect of any personal judgment on the morphological estimate, (at least) two authors classified a large subsample of galaxies in Coma independently, after a period of training. The galaxies to be classified had magnitudes in the range 12 to 17 B mag and their images had a typical seeing in the range 0.35 to 3 arcsec (see Andreon et al. 1996 for details). We had both CCD and digitized plate images available for classification. For this comparison, the images considered and all derived data (profiles of surface brightness, ellipticity, position angle, , etc.) were exactly the same.
A perfect agreement on the coarse Hubble type was found for all (more than 100), but two, galaxies. The two discrepant galaxies are peculiar: a very dusty galaxy (GMP 1646) with an r profile (elliptical or irregular?) and a dusty asymmetric lenticular (or spiral?) (GMP 1204).
The fact that, in the traditional morphological analysis, 15-20% of the galaxies have different morphological estimates introduces a scatter in the properties of the Hubble types precisely because of the highly subjective nature of the type estimate and this can mask real differences between the properties of the morphological types. With the reproducibility of 95% reached by our estimate of the type, only a small fraction of galaxies are not of the type assigned by us; this greatly reduces the scatter in the properties within classes and allows an easier comparison of the types in different locations in the Universe (cluster vs field, nearby vs distant, etc.) and greatly helps detecting previously unnoticed properties of the morphological types.
6.3. Stability with respect to observing conditions
To understand whether different telescope set-ups, seeing
conditions, filters, and other effects affect our morphological
classification, we observed the same subsample of galaxies in Coma
during different runs, at different telescopes, with different filters
(mainly Johnson V and Gunn r) and detectors (4 CCDs and one plate).
Different images of the same galaxy were classified by the same or by
different observers. Once the observing conditions are taken into
account, there is perfect agreement on the coarse Hubble type and on
the detailed type of the galaxies for all 54 comparisons done, but for
GMP 1300, classified as S from our plate and S0 from our CCD image.
Typical differences tied to observing conditions are due, for example,
to the limited field of view of observations at the 3.6m
Canada-France-Hawaii telescope in Hawaii (CFH), preventing one from
classifying galaxies larger than the CCD field of view, or to the
lower resolution of the prime focus plate used (FWHM=
) compared to the best quality CCD images. The
comparison of the CFH observations taken under excellent seeing
conditions (FWHM= ) with observations at the 2m
telescope of the Pic du Midi Observatory taken under good to fair
seeing conditions ( FWHM )
for common galaxies shows that Pic du Midi observations are good
enough for determining the galaxy type, since not one galaxy (out of
13) was classified in two different coarse classes from the two
observing materials (again, once the small field of view of CFH
observations is taken into account). The CFH images certainly allow
one to classify E galaxies more easily into the three families and to
detect small morphological components, such as dust lanes.
6.4. Ability to bring out new properties of galaxies in clusters
Our morphological scheme, whatever it traces, gives highly reproducible results and is not very sensitive to the observing conditions, provided the resolution is adapted to the difficulty of the galaxy classification. But this does not necessarily mean that our method brings galaxies into classes containing objects with similar physical properties better than the traditional classification scheme.
The relative quality of our classification method can only be judged on its results, i.e. on the ability of our scheme to separate galaxies of different physical types into different classes, and to bring out new properties of galaxies in clusters.
This ability is demonstrated by a series of recent results obtained
with this method. For example, we have detected a segregation of the
morphological types stronger than the usual clustercentric or
density segregations in the Perseus (Andreon 1994) and Coma (Andreon
1996) clusters and in the distant cluster Cl0939+4713 (Andreon,
Davoust & Heim 1996). Es have a fainter mean surface brightness
than S0s in Coma (Andreon 1996) and Cl0939+4713 (Andreon, Davoust,
Heim 1996). Furthermore, by using images of resolution adapted to the
difficulty of classification, we have found that the spiral fraction
rises by a factor 2 or 3 in the Perseus and Coma clusters (Andreon
1994, 1996) and, by indirect evidence, in most nearby clusters
(Andreon 1993), but not in the distant cluster Cl0939+4713; this
strongly reduces the evidence for a morphological evolution of
galaxies in clusters that many observations make unreasonable (see
Andreon 1993, and Andreon, Davoust & Heim 1996 for details).
© European Southern Observatory (ESO) 1997
Online publication: July 3, 1998