1 Introduction

The Fornax Cluster is an excellent target for studying globular clusters: it is very rich, contains different types of galaxies and the globular cluster candidates can be detected at least as unresolved sources. There are numerous photometric studies of globular cluster systems in selected Fornax galaxies, particularly the central one NGC 1399 as well as other early-type galaxies. Most of these studies show that the color distribution of the globular clusters is bimodal due to the presence of two globular cluster populations, "red'' and "blue''; these integrated colors are mainly driven by metallicity in objects as old as these ones. With regard to the globular cluster system of NGC 1399, the Washington photometry by Ostrov et al. (1993) confirmed the existence of a color gradient that had been detected earlier by Bridges et al. (1991), and suggested that the color distribution was multimodal, as was also supported by the V, I photometry by Kissler-Patig et al. (1997). The HST imaging by Forbes et al. (1998) and Grillmair et al. (1999), and a refined analysis of their previous Washington photometry by Ostrov et al. (1998), pointed to the bimodal character of the color distribution. More recently, the wide-field study by Dirsch et al. (2002a,b) showed that globular custers with a broad metallicity distribution - that cannot be fitted with a single Gaussian - extend further than 100 kpc from NGC 1399 center.

There are fewer investigations of globular cluster systems in dwarf galaxies that are not in the Local Group. Miller and collaborators (Miller et al. 1998a,b; Miller 1999; Lotz et al. 2001) carried out a survey with images from the Wide Field Planetary Camera 2 of the Hubble Space Telescope (FOV $\approx$ 6 arcmin²) to analyze the properties of globular clusters and nuclei of dwarf elliptical galaxies (dEs) in the Fornax and Virgo Clusters and the Leo Group. They include about 20 dEs from the Fornax Cluster but none of them are in common with the present work. They show that the globular cluster specific frequency $S_{\rm N}$ (number of globular clusters with respect to the parent galaxy's luminosity) of the dEs is in the range 2-6; the luminosity function of the globular cluster candidates is consistent with a Gaussian with peak at $M_V^0 \approx - 7$ mag and dispersion $\sigma_V \approx~1.4$ mag (assuming a distance modulus of 31.4 for the Fornax Cluster); and most of the globular cluster (V-I) colors are similar to those of the metal-poor globular cluster population (also Ashman & Bird 1993). The globular cluster system of the luminous dE, N galaxy NGC 3115 DW1 was studied by Durrell et al. (1996a) and Puzia et al. (2000) who obtained mean metallicities $\rm [Fe/H]= -1.2$ and -1, respectively, and they both agreed on a specific frequency $S_{\rm N}$ = 4.9. Durrell et al. (1996b) obtained, on the basis of Washington photometry, a low mean metallicity, $\rm [Fe/H]= -1.45$, for the globular cluster systems of two dE galaxies in the Virgo Cluster and they suggested that the dwarf galaxies globular cluster systems present a similar range of metallicities ( $\rm [Fe/H] = -2$ to -1) as globular clusters in the halos of spiral galaxies, in agreement with Ashman & Bird (1993). Turning to the Local Group, Minniti et al. (1996) constructed a master-dE galaxy combining the data from globular cluster systems of several dEs in this Group; they found an old and metal-poor globular cluster population whose metallicity distribution matched the one of the Milky Way halo globulars.

The abovementioned bimodality in the color distribution of globular clusters in elliptical galaxies, is closely related to the formation of the globular clusters and a variety of scenarios have been proposed (for reviews on the subject see, e.g., Ashman & Zepf 1998 or Harris 2001). Ashman & Zepf (1992) and Zepf & Ashman (1993) predicted this bimodal metallicity distribution of globulars in elliptical galaxies as a result of gas-rich mergers; they proposed that the blue population originally belonged to the progenitor galaxies and the red population formed during the merger. The numerical simulations from Bekki et al. (2002) showed that dissipative mergers create new globular clusters but they were not able to reproduce properly the bimodal metallicity distribution observed in elliptical galaxies. A different point of view was exposed, e.g., by Forbes et al. (1997) who found a correlation between the mean metallicity of the metal-rich globular clusters and the parent galaxy luminosity, which suggested that they share the same chemical enrichment process, while the mean metallicity of the metal-poor ones seems to be independent of the galaxy luminosity. They proposed that the bimodality originated in two phases of globular cluster formation from gas with different metallicities, and that most of them formed "in situ''. The HST study of 17 early-type galaxies by Larsen et al. (2001) showed a correlation between the colors of both, blue and red globular clusters populations, with the B-luminosity and central velocity dispersion of the host galaxy, and concluded that their observations support globular cluster formation "in situ'', in the protogalaxy potential well. Alternatively, Forbes & Forte (2001) analyzed the relation between the mean color of blue and red globulars with the galaxy velocity dispersion and suggested that red globular clusters share a common origin with the host galaxy and blue ones seem to have formed quite independently; according to Côté et al. (1998) these blue globular clusters may have been captured from other galaxies by merger processes or tidal stripping. The idea of the accretion of dwarf galaxies into the cD halo of NGC 1399, the stripping of their gas and globular clusters and the formation of new clusters from this gas poses a different origin for the red globular clusters (Hilker et al. 1999). Burgarella et al. (2001) analyzed the blue globular cluster populations from 47 galaxies and found no correlation between their mean metallicity, which is very similar for all these systems, and the galaxy properties (luminosity, velocity dispersion, etc.); they proposed that the metal-poor globular clusters may have formed from gas fragments of similar metallicity, as already suggested by Ashman & Bird (1993), and located within the dark halo of the galaxy. More recently, the semianalytic model by Beasley et al. (2002) assumed that the metal-poor globular clusters formed in protogalactic fragments and the metal-rich ones originated in the gas-rich mergers of such fragments that occurred later.

Assuming the presence of globular clusters inside clusters of galaxies, an alternative scenario is proposed by White (1987) and West et al. (1995), who pointed to the possible existence of a population of globular clusters that are not bound to individual galaxies; instead, they are supposed to move freely in the central regions of the galaxy clusters. These intracluster globular clusters may be the result of interactions or mergers between the galaxies, or they may have formed precisely in the environment of a galaxy cluster without any parent galaxy. The kinematic analysis by Minniti et al. (1998) and by Kissler-Patig et al. (1999) also suggested that some globular clusters may be associated with the gravitational potential of the galaxy cluster and not solely with NGC 1399. On the other hand, Grillmair et al. (1999) found no evidence of intergalactic globular clusters in an HST/WFPC image at a radial distance of about $1~\hbox{$.\!\!^\circ$ }4$ from NGC 1399, but due to the small field of view, they were not able to rule out their existence. Several objections against the intraclusters were raised by Harris et al. (1998) who tried to explain by means of their existence the supposed high specific frequency of M87, the central Virgo galaxy; but the latest values of $S_{\rm N}$ obtained for NGC 1399 by Ostrov et al. (1998) and Dirsch et al. (2002b) showed that it is not so high ( $S_{\rm N} = 5.6$ and 5.1, respectively). In favor of the existence of intergalactic material, Theuns & Warren (1997), Mendez et al. (1997), and Ciardullo et al. (1998) presented evidence for the presence of intergalactic planetary nebulae within the Fornax and Virgo Clusters, while Ferguson et al. (1998) reported several hundreds of intracluster red giants in Virgo. Some globular clusters may have been stripped with them from other cluster galaxies if this is their origin (Harris 2001).

In this paper, we analyze the characteristics of globular cluster candidates found near dwarf galaxies in the Fornax Cluster and its connection with the above mentioned scenarios. It is organized as follows: Sect. 2 describes the observations and the adopted criteria for the globular cluster candidates' selection. In Sect. 3 we analyze the color distribution, luminosity function and spatial distribution of the candidates. Finally, a summary of the results and a discussion on their implications are provided in Sect. 4. Preliminary results of this work have been presented by Bassino et al. (2002).