In this section we come back to the presence of multiple sub-population of globular clusters around a number of giant galaxies. We will briefly review the different scenarios present in the literature that could explain the properties of such composite systems and discuss their pros and cons.
Sub-populations of globular clusters were first identified in the Milky Way (Morgan 1959, Kinman 1959, Zinn 1985), and associated with the halo (in the case of the metal-poor population) and the "disk" (in the case of the metal-rich population. The "disk" clusters are now better associated with the bulge (e.g. Minniti 1995, Côté 1999). The presence of multiple component populations in other giant galaxies was first detected by Zepf & Ashman (1993). Obviously the multiple sub-populations get associated with several distinct epochs or mechanisms of star/cluster formation.
The simple scenario of a disk-disk merger explaining the presence of two populations of globular clusters (Ashman & Zepf 1998) found a strong support in the community for 5-6 years, partly because of a lack of alternatives. It was backed up by the discovery of newly formed, young star clusters in interacting galaxies (Lutz 1991, Holtzman et al. 1992). Only recently, other scenarios explaining the presence of at least two distinct populations were presented and discussed.
5.1. The different scenarios for sub-populations
We will make a (somewhat artificial) separation in four scenarios and briefly outline them and their predictions.
The merger scenario
The fact that mergers could produce new globular clusters was mentioned in the literature early after Toomre (1977) proposed that ellipticals could form out of the merging of two spirals (see Harris 1981 and Schweizer 1987). But the first crude predictions of the spiral-spiral merger scenario go back to Ashman & Zepf (1992). They predicted two populations of globular clusters in the resulting galaxy: one old, metal-poor population from the progenitor spirals and one newly formed, young, metal-rich population. The metal-poor population would be more extended and would have been transfered some of the orbital angular momentum by the merger. The metal-rich globular clusters would be more concentrated towards the center and probably on more radial orbits.
In situ scenarios
In situ scenarios see all globular clusters forming within the entity that will become the final galaxy. In this scenario, globular clusters form in the collapse of the galaxy, which happens in two distinct phases (see Forbes et al. 1997, Harris et al. 1998, Harris et al. 1999). The first burst produces metal-poor globular clusters and stars (similar to Searle & Zinn 1978) and provokes its own end e.g. by ionizing the gas or expelling it (e.g. Harris et al. 1998). The second collapse happens shortly later (1-2 Gyr) and is at the origin of the metal-rich component. Both populations are linked with the initial galaxy.
Accretion scenarios were reconsidered in detail to explain the presence of the large populations of metal-poor globular clusters around early-type galaxies. In these scenarios, the metal-rich clusters belong to the seed galaxy, while the metal-poor clusters are accreted from or with dwarf galaxies (e.g. Richtler 1994). Côté et al. (1998) showed in extensive simulations that the color distributions could be reproduced. Hilker (1998) and Hilker et al. (1999) proposed the accretion of stellar as well as gas-rich dwarfs that would form new globular when accreted. In such scenarios, the metal-poor clusters would not be related to the final galaxies but rather have properties compatible with that of globular clusters in dwarf galaxies. Furthermore, this scenario is the only one that could easily explain metal-poor cluster that are younger than metal-rich ones. In a slightly differently scenario, Kissler-Patig et al. (1999b) mentioned the possibility that central giant ellipticals could have accreted both metal-poor and metal-rich clusters from surrounding medium-sized galaxies.
Pre-galactic scenarios were proposed long ago by Peebles & Dicke (1968), when the Jeans mass in the early universe was similar to globular cluster masses. Meanwhile, it was reconsidered in the frame of globular cluster systems (Kissler-Patig 1997b, Kissler-Patig et al. 1998b, Burgarella et al. 2000). The metal-poor globular clusters would have formed in fragments before the assembly of the galaxy, later-on building up the galaxy halos and feeding with gas the formation of the bulge. In that scenario too, the metal-poor globular clusters do not have properties dependent from the final galaxy, while the metal-rich clusters do. Also, metal-poor clusters are older than metal-rich clusters.
Overall, the scenarios are discussed in the literature as different but do not differ by much. The first scenario explains the presence of the metal-rich population, as opposed to the last two that deal with the metal-poor population. These three scenarios are mutually not exclusive. Only in situ models connect the metal-rich and metal-poor components. For the metal-rich clusters, the question resumes to whether they formed during the collapse of the bulge/spheroid, or whether they formed in a violent interaction. Although an early, gas-rich merger event at the origin of the bulge/spheroid would qualify for both scenarios. In the case of metal-poor clusters, the difference between the last three scenarios is mostly semantics. They differ slightly on when the clusters formed, and models two and four might expect differences in whether or not the properties of the clusters are related to the final galaxy. But the bottom line is that the boarder-line between the scenarios is not very clear. Explaining the building up of globular cluster systems is probably a matter of finding the right mix of the above mechanisms, and this for every given galaxy.
5.2. Pros and cons of the scenarios
The predictions of the different scenarios are fairly fuzzy, and no scenario makes clear, unique predictions. Nevertheless, we can present the pros and cons to outline potential problems with any of them.
The merger scenario
Pros: we know that new star cluster form in mergers (e.g. above mentioned reviews, and see Schweizer 1997), and will populate the metal-rich sub-population of the resulting galaxy. Note also, that the merger scenario is the only one that predicted bimodal color distributions rather then explaining them after fact.
Cons: we do not know i) if all early-type galaxies formed in mergers, ii) if the star clusters formed in mergers will indeed evolve into globular clusters (e.g. Brodie et al. 1998), iii) if all mergers produce a large number of clusters (which depends on the gas content). Furthermore, we would then expect the metal-rich populations to be significantly younger in many galaxies (according e.g. to the merger histories predicted by hierarchical clustering models). There are still problems in explaining the specific frequencies and the right mix of blue and red clusters in early-type galaxies in the frame of the merger scenario (e.g. Forbes et al. 1997).
In situ scenarios
Pros: Searle & Zinn (1978) list the evidences for our Milky Way halo globular clusters to have formed in fragments building up the halo. The massive stars in such a population would quickly create a hold of the star/cluster formation for a Gyr or two.
Cons: if a correlation between metal-poor clusters and galaxy is expected, the scenario would be ruled out. A clear age sequence from metal-poor to metal-rich clusters is predicted but not yet verified. This scenarios is not in line with hierarchical clustering models for the formation of galaxies (Kauffmann et al. 1993, Cole et al. 1994), should the latter turn out to be the right model for galaxy formation.
Pros: dwarf galaxies are seen in great numbers around giant galaxies, and hierarchical clustering scenarios predict even more at early epochs. Dwarf galaxies do get accreted (e.g. Sagittarius in our Galaxy). We observe "free-floating" populations around central cluster galaxies (e.g. Hilker et al. 1999) and the color distributions of globular cluster systems can be reproduced (Côté et al. 1998).
Cons: we are missing detailed dynamical simulations of galaxy groups and clusters to test whether the predicted large number of dwarf galaxies gets indeed accreted (and when). We do not know whether the (dwarf) galaxy luminosity function was indeed as steep as required at early times to explain the large accretion rates needed. Also, the model does not provide a physical explanation for the metal-rich populations.
Pros: similar to the above, we observe a "free-floating", spatially extended populations of globular clusters around central galaxies. The properties of the metal-poor populations do not seem to correlate with the properties of their host galaxies (Burgarella et al. 2000). The metal-poor globular cluster are observed to be very old (e.g. Ortolani et al. 1995 for our Galaxy; Kissler-Patig et al. 1998a, Cohen et al. 1998, Puzia et al. 1999 for analogies in extragalactic systems).
Cons: galaxies and galaxy halos might not have formed by the agglomeration of independent fragments. No physical model exists, except a broad compatibility with hierarchical clustering models (see also Burgarella et al. 2000).
Some pros and cons are listed only under one scenario but apply obviously to others. It should be noted that these pros and cons apply to "normal" globular cluster systems. It has been noted that several galaxies host very curious mixes of metal-poor and metal-rich clusters (Gebhardt & Kissler-Patig 1999, Harris et al. 2000) that pose challenges to all scenarios. Fine difference will require a much more detailed abundance analysis of the individual clusters in sub-populations, as well as their dynamical properties and (at least relative) ages for the different globular cluster populations. These might allow to identify a unique prediction supporting the one or the other formation mode, or constrain the importance of each formation mechanism.