6.1. Irregular and compact objects
The first classification that can be carried out for our sample is between compact, isolated objects, and galaxies that are clearly undergoing an interaction or exhibit an irregular, diffuse shape. The easiest way to do this is by visual inspection, since all our objects are well resolved. We find 35 compact galaxies out of 41, or 85% of the sample. The ERO HR10 (Hu & Ridgway 1994; Graham & Dey 1996) is included in the subsample of irregular objects. We have attempted to recognize in each of these irregular objects a brighter component, such as could be expected in a merging system, and obtain a fit of its brightness distribution after a proper masking of the surrounding areas. This was not possible for object 21, object 5, and object 30 (HR 10): in the case of object 21, as we mentioned in Sect. 3, the surface brightness distribution is too diffuse to isolate a major component; object 5 is compact, but its nucleus has a rather irregular shape, probably due to the superposition of two or more close and equally bright components; in object 30, a brighter component can be easily recognized, but we did not obtain a satisfactory fit to its brightness distribution. A radial brightness profile was anyway extracted for object 5 and object 30, as shown in Fig. 2. A few more galaxies are close to other objects that, however, are not disturbing their morphology: in these cases a proper masking was also applied, as indicated in Table 3, to avoid any problem with the fitting procedure.
6.2. Structural parameters
The best-fit profiles are plotted in Fig. 2, the best-fit parameters for each object are shown in Table 3, while Fig. 7 shows the estimated location in the - plane of all our sample galaxies with different symbols for each instrumental configuration. The encircled symbols correspond to the 3 galaxies classified as irregular/interacting for which we could obtain a fit, after masking the lower flux companion: since for these objects we consider only a fraction of the total flux, they are typically placed in the lower part of the plot. Effective radii and surface brightnesses are plotted in instrumental units (pixels and counts/pixel), to allow a direct comparison with Fig. 6. The y coordinates of the data points, however, are not exactly the values determined by the fit, since each galaxy has been scaled to the noise level adopted for the simulations by applying a proper shift along the brightness axis. The uncertainties listed in Table 3 are evaluated by interpolating the results from the simulations at the locations of the real galaxies in the parameters' space; we report 1- errors for all the parameters except n: when its integer value is retrived correctly, we assign to this quantity a formal error of 0.5, otherwise the integer value reported corresponds to the largest possible error. As we mentioned in the previous section, we have checked - using a theoretical approach - that the estimated errors for n roughly correspond to a 90% confidence level. In a few cases (7 out of 38, 2 of which classified as irregular) the thoretical estimate exceeds the one derived from the simulations; for these galaxies the uncertainties reported are the thoretical ones.
The 4 HDFS galaxies (38-41) were studied by Benítez et al. (1999) with a technique very similar to ours (a best fit to the brightness profiles with a de Vaucouleurs' law), so that their and our results can be easily compared. We find that a de Vaucouleurs law is the best fit to the data for two of these galaxies (), the other two being best represented by an and profile respectively. For what concerns the integrated fluxes, the average difference is 0.07 0.05 magnitudes, whereas our effective radii are 0.84 0.25 times the ones by Benítez et al., on average. We conclude that the differences between the two works are not relevant, and characterized by only a modest scatter in the measured quantities. Excellent agreement is then found, in the case of object 39, with the results by Stiavelli et al. (1998): although our best fit for this galaxy is with , both our effective radius and our total flux are coincident with the Stiavelli et al. values, derived adopting a de Vaucouleurs' distribution.
Fig. 8 shows again the - plane in standard units (arcsec, mag arcsec-2), with a different symbol for each filter; the dotted line represents the slope of constant flux, at fixed shape index n. The plot illustrates the limits in size and surface brightness of the sample in the HST filters.
6.3. The shape index n and the fraction of ellipticals
We evaluated previously, through visual inspection, the fraction of irregular objects, concluding that they constitute only a minority of our ERO sample. For what concerns the shape of the best-fit distributions, we performed two types of classifications.
A first order classification was performed by comparing the results assuming that each galaxy can be properly described by either an exponential distribution () or a de Vaucouleurs one (). To do that, we considered only the simulations belonging to these two classes, as we have seen in the previous section that the true n value can be retrieved for all the galaxies in the sample. The resulting number of de Vaucouleurs distributions is then 21 out of 41 (51%).
A more detailed classification was made leaving n free to vary among the integer values . Using this approach, the relative abundance of non-exponentials () is slightly higher if we choose the best n value for each galaxy from the whole set of fits performed: 25/41 (61%). In particular, four galaxies previously catalogued as exponentials are now fitted better by an distribution, whereas the other 10 objects with have their best fits confirmed. As we discussed previously, for the galaxies placed above the dotted line in Fig. 7 we can reliably distinguish between and , whereas at fainter fluxes distributions may be mistaken for exponentials. This "high-signal" subsample, therefore, provides a particularly accurate estimate of the fraction of likely ellipticals which, in this case, is even larger, amounting to 81% (21 out of 26).
As mentioned in Sect. 2, for , the WFPC2 and NICMOS images cover the rest-frame UV and the optical spectral regions respectively. As it is well known that the galaxy morphology depends strongly on the wavelength (e.g. Kuchinski et al. 2000and references therein), one may argue if this can have effects on our results. In this respect, we can envisage three cases. First, if a galaxy is a passively evolving elliptical, then its morphology does not depend on (e.g. Kuchinski et al. 2000) and it would be classified as elliptical both in WFPC2 and in NICMOS images. Second, if a galaxy is irregular, then it would be reliably classified as such both in WFPC2 and in NICMOS images (e.g. HR10; Graham & Dey 1996; Dey et al. 1999). Finally, there could be cases of elliptical galaxies with a disk component having if observed in the optical (WFPC2) and if observed in the near-IR (e.g. spheroidal galaxies with a disk component becoming more prominent in the rest-frame UV). Our analysis does not allow to investigate if such latter cases are present in our sample because no NICMOS images are available for the 9 objects with observed with WFPC2.
To summarize, we conclude that the galaxies that can be reliably classified as ellipticals amount to 5080% of the total sample of 41 EROs. Although our sample is incomplete, we note that our results are in good agreement with those of Stiavelli & Treu (2000) based on a complete sample of NICMOS-selected EROs.
Such a high fraction of ellipticals strengthens the scenario proposed by Daddi et al. (2000) who suggested that, because of their strong clustering, EROs are likely to be dominated by ellipticals rather than dusty starbursts.
© European Southern Observatory (ESO) 2000
Online publication: December 15, 2000