In this section, we present some properties of globular cluster systems and of young clusters in ellipticals, spirals and mergers. In the last section we mentioned what are the properties measured in globular cluster systems: The metallicity distribution can be obtained from photometry (colors) or spectroscopy (absorption line indices). The luminosity function of the clusters is computed from the measured magnitudes folded with any incompleteness or contamination function. The total number of clusters (and eventually number of metal-poor and metal-rich clusters) is obtained by extrapolating the observed counts over the luminosity function and eventually applying any geometrical completeness for the regions that are not covered. For the latter, one uses also knowledge about the spatial distribution (position angle, ellipticity) and radial density profile of the globular cluster system. For young star clusters, the color distribution no longer reflects the metallicity distribution, but a mix of ages and metallicities. More complex comparisons with population synthesis models and/or spectroscopy are needed to disentangle the two quantities. Of interest for young clusters are also the mass distribution (derived from the luminosity function) that helps to understand how many of the newly formed clusters will indeed evolve into massive "globular" clusters.
4.1. Globular clusters in early-type galaxies
Early-type galaxies have the best studied globular cluster systems. Spirals have the two systems studied in most details (i.e. the Milky Way and M31) due to our biased location in space, but a far larger sample now exists for early-type galaxies.
Despite looking remarkably similar in many respects (e.g. globular cluster luminosity function), globular cluster systems in early-type galaxies also show a large scatter in a number of properties. For example, the number of globular clusters normalized to the galaxy light (specific frequency, see Harris & van den Bergh 1981) appears to scatter by a factor of several, mainly driven by the very high specific frequencies of central giant ellipticals (and recently also observed in faint dwarf galaxies, see Durrel et al. 1996, Miller et al. 1998). Furthermore, the radial density profiles are very extended for large galaxies, while following the galaxy light in the case of intermediate ellipticals (e.g. Kissler-Patig 1997a).
In the early 90's, Zepf & Ashman (1993) discovered the presence of globular cluster sub-populations in several early-type galaxies. We will come back to the origin of the sub-populations in Sect. 5. Here, we will discuss the implications of sub-populations on our understanding of the globular cluster system properties.
Until the early 90's, properties were derived for the whole globular cluster system. Since then, it became clear that many properties need to take into account the existence of (at least two) different sub-populations, in order to be explained. Probably the first work to show this most clearly was the presentation of the properties of blue and red clusters in NGC 4472 by Geisler et al. (1996). Taking into account the existence and different spatial distribution of blue and red clusters, they explained two properties of whole systems at once. First, the color gradient observed in several systems could be explained by a varying ratio of blue to red clusters with radius (without any gradient in the individual sub-populations). Second, the mean color of the systems was previously thought to be systematically bluer than the diffuse galaxy light. It turns out that the color of the red sub-population matches the color of the galaxy, while it is the presence of the blue "halo" population that makes the color of the whole globular cluster system appear bluish.
It has not yet been demonstrated that the scatter in the specific frequency and in the slopes of the radial density profiles also originate from different mixes of blue to red sub-populations, but this could be the case. The few studies that investigated separately the morphological properties of blue and red clusters (Geisler et al. 1996, Kissler-Patig et al. 1997, Lee et al. 1998, Kundu & Whitmore 1998) found the metal-poor (blue) population to be more spherically distributed and extended than the metal-rich population that has a steeper density profile, tends to be more flattened and appears to follow the diffuse stellar light of the galaxy in ellipticity and position angle (cf. Fig. 5). Thus, a larger fraction of blue clusters in a galaxy would mimic a flatter density profile of the whole globular cluster system.
Figure 5. Left panel: The angular distribution of halo and bulge globular clusters around NGC 1380 in 30 degree sectors, after a point symmetry around the center of the galaxy. Note that the blue objects are spherically distributed, while the red objects have an elliptical distribution that peaks at the position angle of the diffuse stellar light. Right panel: Surface density profiles of red and blue globular clusters around NGC 1380, plotted once against the radius in arcseconds (upper panel and once against the semi-major axis (lower panel). Note how the blue objects have a much flatter density profile than the red ones, which are concentrated towards the center and follow a similar density profile as the stellar light. Both plots are taken from Kissler-Patig et al. (1997).
Furthermore, the specific frequency of the blue clusters (when related to the blue light) appears to be very high (> 30 see Harris 2000). This, by the way, could be explained if the latter came from small fragments similar to the dwarf ellipticals observed today, that also show high specific frequency values (although not as high, but in the range 10 to 20). Thus, an overabundance of blue clusters would imply a high specific frequency. Incidentally, the shallow density profiles are found in the galaxies with the highest specific frequencies (see Kissler-Patig 1997a). We can therefore speculate that the properties of the entire globular cluster systems of these massive (often central) giant ellipticals can be explained by a large overabundance of metal-poor globular clusters originating from small fragments. The scatter in the globular cluster system properties among ellipticals could then (at least partly) be explained by a varying fraction of metal-poor "halo" and metal-rich "bulge" globular clusters.
Observationally, this could be verified by determining the number ratios of metal-rich and metal-poor globular clusters in a sample of galaxies showing different globular cluster radial density profiles and specific frequencies. The number of studies investigating the properties of metal-poor and metal-rich populations needs to increase in order to confirm the general properties of these two groups. We end with a word of caution: the existence of such sub-populations has been observed in only ~ 50% of all early-type galaxies studied (e.g. Gebhardt & Kissler-Patig 1999), and still remains to be demonstrated in all cases. Furthermore, the exact formation process of these sub-populations is still unclear (see Sect. 5).
4.2. Globular clusters in late-type galaxies
The study of globular cluster systems of late-type spirals started with the work of Shapley (1918) on the Milky Way system. Despite a head-start of nearly 40 years compared to studies in early-type galaxies, the number of studied systems in spirals lags far behind the one in ellipticals. This is mainly due to the observational difficulties: globular clusters in spirals are difficult to identify on the inhomogeneous background of disks. Furthermore, internal extinction in the spiral galaxies make detection and completeness estimations difficult, and photometry further suffers from confusion by reddened HII regions, open clusters or star forming regions.
The best studied cases (Milky Way and M31) show sub-populations (e.g. Morgan 1959, Kinman 1959, Zinn 1985; Ashman & Bird 1993, Barmby et al. 1999) associated in our Galaxy with the halo and the bulge/thick disk (Minniti 1995, Côté 1999). Beyond the local group, spectroscopy is needed to separate potential sub-populations. Both abundances and kinematics are needed, while colors suffer too much from reddening to serve as useful metallicity tracers. Spectroscopic studies have been rare in the past, but are now becoming feasible (see Sect. 3.3 and 4.1). For example, a recent study of M81 allowed to identify a potential thick disk population beside halo and bulge populations (see Schroder et al. 2000 and references therein).
Some of the open questions are whether all spirals host halo and bulge clusters, and whether one or both populations are related to the metal-poor and metal-rich populations in early-type galaxies. The number of globular clusters as traced by the specific frequency appears roughly constant in spirals of all types independently of the presence of a bulge and/or thick disk (e.g. Kissler-Patig et al. 1999a). This would mean that spirals are dominated by metal-poor populations, with their globular cluster systems only marginally affected by the presence of a bulge/thick disk. If metal-poor globular clusters indeed formed in pre-galactic fragments, then one would expect the metal-poor populations in spirals and ellipticals to be the same. We know that the globular cluster luminosity functions are extremely similar, but the metallicity distributions and other properties remain to be derived and compared (see Burgarella et al. 2000 for a first attempt). Finally, a good understanding of the globular cluster systems in spirals will also help predicting the resulting globular cluster system of a spiral-spiral merger. Predictions can then be compared to the properties of systems of elliptical galaxies in order to constrain this mode of galaxy formation.
4.3. Star clusters in mergers and violent interactions
After some speculations and predictions that massive star clusters could/should form in mergers (Harris 1981, Schweizer 1987), these were finally discovered in the early 90's (Lutz 1991, Holtzman et al. 1992). Since then a number of studies focussed on the detection and properties of these massive young star clusters (see Schweizer 1997, and reviews cited in Sect. 1 for an overview).
The most intense debate around these young clusters focussed on whether or not their properties were compatible with a formation of early-type galaxies through spiral-spiral mergers. It was noticed early on (Harris 1981, van den Bergh 1982) that ellipticals appeared to host more clusters than spirals, and thus that mergers would have to produce a large number of globular clusters. Moreover, the specific frequency of ellipticals appeared higher than in spirals, i.e. mergers were supposed to form globular clusters extremely efficiently. In a second stage, a number of studies investigated whether or not these newly formed clusters would resemble globular clusters, and/or would survive as bound clusters at all.
The above questions are still open, except maybe for the last one. The young clusters studied to date show luminosities, sizes, and masses (when they can be measured) that are compatible with them being bound stellar clusters and able to survive the next several Gyr (see Schweizer 1997 for a summary of the studies and extensive references). Whether they will have the exact same properties as old globular clusters in our Milky Way is still controversial. First spectroscopic measurements found the young clusters in NGC 7252 compatible with a normal initial mass function (IMF) (Schweizer & Seitzer 1998), while in NGC 1275 the young clusters show anomalies and potentially have a flatter IMF (Brodie et al. 1998) which would compromise their evolution into old globular clusters, as we know them from the Galaxy.
The mass distribution of these young cluster was first found to be a power-law (e.g. Meurer 1995), as opposed to a log-normal distribution for old globular clusters. This result is likely to suffer from uncertainties in the conversion of luminosities into masses, when neglecting the significant age spread among the young clusters (see Fritze-von Alvensleben 1999). However, deeper data seem to rule out the possibility that the initial mass distribution has already the same shape as the one observed for old clusters (see Whitmore et al. 1999, Zepf et al. 1999). But the slope of the mass distributions could be affected during the evolution of the system by dynamical destruction at the low-mass end. Finally, Whitmore et al. (1999) recently found a break in the mass function of the young clusters of the Antennae galaxies, similar to the characteristic mass of the old clusters further supporting similar mass functions for young and old cluster populations (see also Sect. 7). Overall, the young clusters might or might not resemble old Galactic globular clusters, but some will survive as massive star clusters and could mimic a population of metal-rich globular clusters.
The most interesting point remains the number of clusters produced in mergers. Obviously, this will depend on the gas content (`fuel') that is provided by the merger (e.g. Kissler-Patig, et al. 1998b). Most gas-rich mergers form a large number of star clusters, but few of the latter have masses that would actually allow them to evolve into massive globular clusters as we observe them in distant ellipticals. Harris (2000) reviews comprehensively this issue and other problems related with a scenario in which all metal-rich globular clusters of ellipticals would have formed in mergers. The main problem with such a scenario is that the high specific frequency of ellipticals should be due to metal-rich clusters, which is usually not the case. Potential other problems, depending on the exact enrichment history, are that large ellipticals would be build up by a series of mergers that should probably produce an even broader metallicity distribution than observed; and that radial metallicity gradients might be expected to be steeper in high specific frequency ellipticals.
In summary, mergers are the best laboratories to study younger stellar populations and the formation of young stellar cluster, but how important they are in the building of globular cluster systems (and galaxies) remains uncertain. However, a good understanding of these clusters is crucial for the understanding of globular cluster systems in early-type galaxies, since merger events must have played a role at some stage.