4 Discussion

With regard to the color distribution of the globular cluster candidates, it is interesting to note that we do not find a single population, the metal-poor one, as seems to be the common case for dwarf galaxies (Durrell et al. 1996b; Miller et al. 1998a,b; Miller 1999), but an extended distribution, which appears to be bimodal though we cannot prove it statistically due to the small sample involved. In addition, if we take into account the specific frequency estimated for dwarf galaxies (Miller et al. 1998a; Elmegreen 1999) and the luminosity function of the globular cluster candidates, we should have found significantly fewer globular cluster candidates than we actually do.

According to the projected density of the potential globular clusters, they show no concentration towards the dwarfs while they do show concentration with respect to the center of the cluster. These results led us to speculate that the globular cluster candidates may not be associated to the dwarf galaxies themselves. We are then left with three possibilities: first, that these globular cluster candidates belong to the globular cluster system of NGC 1399; second, they may be moving freely throughout the potential well of the cluster, without being bound to any galaxy in particular; or third, that they are a mix of both. In the first case, we should be accepting that the globular cluster system of NGC 1399 is much more extended and numerous than previously thought. According to this hypothesis, Fig. 6 suggests that there should be clusters up to at least an intermediate angular distance of 80$^\prime$ (a projected distance of about 430 kpc) from the central galaxy; the CCD study over the largest area around NGC 1399 was performed by Dirsch et al. (2002b), which extends up to 22$^\prime$, and showed that the globular cluster system extends over a radial projected distance of more than 100 kpc.

The total number (background corrected) of globular clusters within the area covered by our observations can be roughly estimated as about 550 clusters, by extrapolating their luminosity function over the whole range of T₁ magnitudes. Thus, the number of clusters that should be distributed within a circular area of radius 80$^\prime$ around NGC 1399 may be calculated, just taking into account the ratio of the areas, as several 10⁴ clusters. For comparison, the total number of globular clusters associated with galaxies in the Fornax Cluster may also be roughly estimated as follows. The blue magnitude of all the galaxies included in the Ferguson (1989) catalogue is B = 8.3 mag; adopting for them a mean color index $B-V \approx 0.8$ mag and the distance modulus mentioned above, we obtain an absolute visual magnitude M_V = -23.8. If we assume a "typical'' specific frequency $S_{\rm N}=5$ we conclude that about $1.6\times 10^4$ globular clusters should be associated with galaxies in Fornax. It is interesting to note that the number of globular clusters that we found within a circular area of radius 80$^\prime$ around NGC 1399 is of the same order o larger than the estimated number of globulars associated to galaxies in the whole Fornax Cluster.

It is also likely that some globular clusters might have escaped from its parent galaxies and, after that, remained within the potential well of the Fornax Cluster as a whole (see, for instance, Kissler-Patig et al. 1999). White (1987) proposed that the distribution of the stripped globular clusters within the cluster will follow the same density profile as the galaxies and they might form a kind of envelope around the central galaxy. Alternative origins for the intraclusters are mentioned by West et al. (1995), who speculate that they might have formed "in situ'', without a parent galaxy, or during mergers of sub-systems with a high gas content.

Deeper images are required to clarify this picture and a new survey in the Fornax Cluster is in progress. The true nature of these candidates might be confirmed by means of spectra.

Acknowledgements

We wish to thank the referee, Dr. M. Kissler-Patig, for his comments which helped to improve the present paper. We are also grateful to S. D. Abal, M. C. Fanjul and R. E. Martínez for technical assistance. This work was partially supported by grants from CONICET and Fundación Antorchas, Argentina.