4.3. Globular Clusters
Young star clusters are observed in star-forming galaxies like the LMC  and in intense starburst galaxies [80, 81]. These clusters have half-light radii of less than 5 pc, masses of 104 to 107 M, and metal abundances comparable to their parent starbursts. Their luminosity functions follow power laws with slopes of -1.6 to -2, intriguingly close to the mass function of giant molecular clouds . However, it's not entirely clear that cluster luminosity is a good indicator of mass since some range of cluster ages is usually present.
Evidence is accumulating that the globular cluster systems of field ellipticals are partly due to cluster formation in merger-induced starbursts:
Globulars in elliptical galaxies have bimodal color (metalicity) distributions.
These findings imply that metal-rich star clusters form during mergers and are gradually assimilated into existing globular cluster populations . However, the large populations of metal-poor globulars found in cluster ellipticals are not consistent with mergers of field spirals ; predicted specific frequencies of metal-poor clusters are SNP 1, while in fact SNP 4. This problem is even worse for cluster systems in cD galaxies, which have SNP 10; obviously, no amount of merging between metal-rich systems will produce metal-poor clusters!
The question of high-SN in cluster ellipticals boils down to this: fewer stars, or more globulars? One way to get fewer stars is to merge galaxies after their metal-poor globulars have formed but before they build up substantial disks. For example, the Milky Way as it was ~ 10 Gyr ago could serve as a building-block for cluster ellipticals; the halo of our galaxy, considered alone, has SNP 4. However, mergers of Milky Way halos (or dwarf elliptical galaxies ) still fall short of the high SNP values of cD galaxies. Another way to end up with fewer stars is to eject most of the gas after the initial epoch of cluster formation; the problem here is that the ejection efficiency must be higher in cD galaxies, which have the deeper potential wells and should be better at retaining gas .
Alternately, the production of globular clusters may have been more efficient in high-redshift starbursts. Even at low-z, about 20% of the UV emitted by starbursts comes from knots identified with young clusters ; if all these clusters survive, the specific frequency for a pure starburst population is SN 60. Moreover, these clusters are concentrated where the surface densities are highest; it's likely that net yields of star clusters increase rapidly with increasing surface density.
If so, then globular cluster systems reflect the starburst histories
of their parent galaxies: Large populations of metal-poor globulars
are due to efficient cluster production in early starbursts, while
predominantly metal-rich systems (eg., NGC 5846) formed in more recent
starbursts. Metallicity distributions for cluster systems support
this idea; giant elliptical galaxies have a range of distributions
with multiple peaks between [Fe/H] -1.2 and 0.2
Such variety seems hard to explain in a picture where
internal events determine the timing of cluster formation (eg.,
on the other hand, it's easy to imagine that different
distributions result from the different merging histories of
1 The specific frequency SN is defined as the number of globular clusters divided by the galaxy luminosity in units of MV = -15. Back.