Other than the Milky Way, the GC systems of only two massive galaxies have been observed in detail, in the sense that a significant number of their GCs have been observed with high S/N spectroscopy and we can study their field stars directly. M31, our sister spiral galaxy in the Local Group, and NGC 5128, one of the nearest massive E galaxies, are both sufficiently well-studied that they can elucidate the role of GCs in tracing star formation in their parent galaxies.
Though the total mass of M31 is similar to that of the Galaxy (Evans & Wilkinson 2000), it has a GC system ~ 3 times as large, with ~ 400-450 GCs (van den Bergh 1999; Barmby & Huchra 2001). A comprehensive catalog of photometric and spectroscopic data for M31 GCs was assembled by Barmby et al. (2000) and used to show that, broadly speaking, the M31 GC system is quite similar to that of the Galaxy; M31 has an extended halo subpopulation and a more centrally-concentrated rotating bulge/disk subpopulation.
The kinematics of the M31 GC system may be complex. Perrett et al. (2002) found that the metal-rich GCs have a velocity dispersion similar to the bulge (~ 150 km/s) and a rotational velocity of 160 km/s. Surprisingly, they also found a large rotational velocity for the metal-poor GCs, ~ 130 km/s. By contrast, the metal-poor GCs in the Galaxy show little evidence for rotation. Visually, the spatial distribution of the Perrett et al. metal-poor GC sample appears to be the superposition of a centrally-concentrated spherical population and a relatively thin disk component. Indeed, Morrison et al. (2004) used the Perrett et al. velocities to argue that M31 possesses a rapidly rotating "thin disk" of old metal-poor GCs. Beasley et al. (2004) presented high-S/N spectra of a small number of these putative old metal-poor GCs and showed that they have spectra more similar to young (< 1 Gyr), approximately solar metallicity objects. Their luminosities suggest stellar masses comparable to massive Galactic open clusters or low-mass GCs. The presence of young GCs has also been claimed by Burstein et al. (2004) and Puzia, Perrett, & Bridges (2005) (both based in part on data from Beasley et al. 2004) and by Fusi-Pecci et al. (2005). Cohen et al. (2005) used Keck NIR/AO imaging to show that four of these very young clusters are actually "asterisms" - either chance groupings of bright stars or a few brighter stars superimposed upon sparse open clusters. A more detailed examination of archival HST imaging indicates that perhaps half of the candidate young clusters are real; previous HST/WFPC2 had confirmed the existence of four such clusters (Williams & Hodge 2001). A high-resolution imaging survey of the disk will be needed to assess the true number of young GCs. In any case, if these interloping objects (asterisms and/or YMCs) are removed from the metal-poor GC candidate list, much of the rotational signature observed by Perrett et al. would be erased. The apparent lack of an old metal-poor thin disk of GCs is important, since the existence of this disk would rule out any significant merger having taken place in the last ~ 10 Gyr. As discussed below, such a merger is the working theory for the existence of intermediate-age stars in M31's halo.
Due to selection effects, including reddening in the inner disk and the paucity of GC candidates at large radii, the overall fraction of metal-poor to metal-rich GCs is still poorly constrained in M31. From their kinematic and metallicity study, Perrett et al. (2002) found that only ~ 25% of a sample of ~ 300 GCs belong to the metal-rich (bulge) peak, compared to one third in the Galaxy. If, as discussed above, many of the candidate metal-poor GCs in M31 are interlopers, then the ratio of metal-poor to metal-rich GCs in M31 could be quite consistent with that found for the Galaxy (2:1). There are several ongoing studies of the M31 GC system with MMT/Hectospec that should provide a clearer picture of this situation.
Burstein et al. (1984)
found evidence for Balmer line anomalies in M31
GCs compared to GCs in the Milky Way. Their preferred explanation was
that the M31 GCs might be younger than their Milky Way counterparts.
Brodie & Huchra
(1991)
and Beasley et al. (2004)
compared integrated-light spectra of Milky Way and M31 GCs at the same
metallicity and found no evidence for enhanced Balmer lines in M31 GCs.
Beasley et al. (2005)
extended these results by fitting the Lick
indices measured from their sample GCs to stellar population models.
They employed the multi-index
2 minimization
method developed by
Proctor, Forbes, &
Beasley (2004)
to derive ages and metallicities for these
clusters. In addition to the normal old metal-poor and metal-rich GC
subpopulations, they found a group of six GCs with intermediate
metallicities and ages of 3-6 Gyr. A larger number of GCs with similar
ages were also reported by
Puzia et al. (2005).
This group of GCs forms
a coherent chemical and kinematic group of objects in M31, consistent
with the accretion of an SMC-type galaxy within the last several Gyr (in
the Galaxy, there are several several comparable GCs associated with the
Sgr and CMaj dSphs; Section 10).
Ashman & Bird (1993)
had previously analyzed M31 GC kinematics and found evidence for ~ 7
distinct groups of GCs. The clustering analysis of
Perrett et al. (2003)
found ~ 8 such
groups, but Monte Carlo simulations indicated that statistically half or
more of these should be chance groupings. In an addition to the rapidly
growing zoo of stellar clusters,
Huxor et al. (2005)
have identified 3 metal-poor objects in the halo of M31 that have luminosities typical of
GCs, but sizes from 25-35 pc. They suggest that these could be stripped
nuclei of dwarfs, analogous to
Cen in the Galaxy
(Majewski et al. 2000).
Brown et al. (2004) obtained extremely deep HST/ACS photometry of the M31 GC 379-312 ([Fe/H ~ -0.6) and through isochrone fitting derived a formal age of 10-1+2.5 Gyr, perhaps ~ 1 Gyr younger than Galactic metal-rich GCs. Rich et al. (2005) carried out photometry deeper than the horizontal branch for a sample of 12 M31 GCs. They found the classic second parameter effect, but with an offset that might be explained if the GCs were ~ 1-2 Gyr younger than their Galactic counterparts. These results hint at an interesting relative age difference between the two GC systems, but hard evidence will be difficult to come by. It would require a UV/optical space telescope with an aperture larger than HST to derive turnoff ages for significant numbers of M31 GCs.
Burstein et al. (1984) also discovered CN enhancements in M31 GCs (with respect to Galactic GCs). This result was later confirmed by Brodie & Huchra (1991), who pointed out that CH at 4300 Å did not appear to be anomalous, and argued that the observed CN enhancement might be due to an overabundance of N. Beasley et al. (2004) did not see the enhancement clearly in the Lick CN2 index, which measures the weaker 4215 Å bandhead, but it is obvious in a different index sensitive to the 3883 Å transition. That N is the culprit in CN variations was confirmed directly by near-UV spectra of Galactic and M31 GCs around the NH band at 3360 Å (Burstein et al. 2004). Observations in this region of the spectrum have been limited by the low fluxes of GCs and the low UV efficiencies of telescopes, instruments, and detectors. As a result, the SSP modeling effort has proceeded slowly, and quantitative [N/Fe] values do not yet exist. Thomas, Maraston, & Bender (2003) found that [N/Fe] ~ +0.8 was required to match CN in some Galactic GCs; in M31 GCs it must be even higher. Since NH (in the UV) is dominated by light from stars around the main sequence turnoff, the N enhancements are probably primordial in origin and cannot be attributed to non-canonical mixing processes on the red giant branch.
M31's stellar halo, studied in fields from ~ 10-30 projected kpc, appears predominately metal-rich ([m/H] ~ -0.5), a full 1 dex higher than the Milky Way (e.g., Mould & Kristian 1986; Durrell et al. 1994). In one M31 halo field, Brown et al. (2003) found that ~ 30% of the stars were of intermediate-age, again unlike the Milky Way. Given the large disk and bulge of M31, it has been suggested that these results may be explained by the contamination of putative halo samples by disk/bulge stars (Worthey et al. 2005). However, this interpretation does not appear to be consistent with recent kinematic studies of a subset of the intermediate-age stars (Rich et al. 2006). In any case, M31 also appears to have a metal-poor projected r-2 stellar halo, like the Galaxy (Guhathakurta et al. 2005; Kalirai et al. 2006). Numerical simulations of M31 indicate that a relatively minor accretion event (involving an LMC-sized galaxy) would suffice to redistribute the observed number of metal-rich, intermediate-aged stars from the disk into the halo (Font et al. 2006).
Before discussing how field star properties relate to GCs, it is helpful to define the quantity Tn, which is the Zepf & Ashman (1993) T parameter normalized to the mass of a particular stellar population (vs. the original definition, which normalizes a single GC subpopulation to the total stellar mass in the galaxy). For spheroids with few (or no) metal-poor stars, Tredn is the same as Tred, modulo uncertainties in the adopted M/L.
The Galaxy has ~ 100 metal-poor GCs and a halo mass of ~ 109
M
(Carney, Latham, &
Laird 1990),
resulting in a metal-poor
Tbluen ~ 100. By contrast, the
corresponding metal-rich value is Tredn ~ 5
(there are only ~ 50 metal-rich GCs and bulge mass ~ 1010
M;
Kent 1992).
Thus the metal-poor GCs formed with an efficiency twenty times
higher than the metal-rich GCs with respect to field stars.
In the Galaxy, the association of the metal-poor GCs with halo stars is clear; they have similar peak metallicities, spatial distributions, and kinematics (though the metallicity dispersion of the field stars is much larger than that of the GCs). The situation for the metal-rich GCs is somewhat less well-defined but, in general, their spatial distribution and kinematics are more similar to the bulge than the thick disk (see Section 7.2). However, the metallicity distribution of the metal-rich GCs does not match that of the bulge. Fulbright et al. (2006) used Keck/HIRES spectra to recalibrate the metallicity distribution of red clump stars in Baade's Window (Sadler et al. 1996). They found a peak at [Fe/H] ~ -0.15, roughly 0.3-0.4 dex higher than the peak of the metal-rich GCs. Baade's Window is located beyond one bulge effective radius, so if the bulge has a negative metallicity gradient the mean offset could be even larger.
The metallicity of the M31 bulge is similar to the bulge of the Galaxy (Sarajedini & Jablonka 2005). The GC color distributions in the Milky Way and M31 show no significant differences, suggesting a similar metallicity offset in M31 between the metal-rich GCs and the bulge (as borne out by spectroscopy of individual GCs in M31). The metallicity of the metal-poor component of the stellar halo in M31 is not well-constrained, so no direct comparison can yet be made with the GCs.
As one of the nearest (albeit disturbed) giant ellipticals (~ 4 Mpc), NGC 5128 holds special significance for stellar population studies. W. and G. Harris and coworkers have taken the lead in studying the relationship between GCs and field stars in this galaxy.
Their HST/WFPC2 photometry of red giants in fields from ~ 8 to
31 kpc (projected radius) give metallicity distributions that peak at
[m/H] ~ -0.2 to ~ -0.4, with a slight negative metallicity gradient with
galactocentric radius
(Durrell, Harris &
Pritchet 2001;
Harris & Harris
2002).
Such metallicities are consistent with those of the
metal-rich GCs, modulo uncertainties in the relative zeropoints of the
metallicity scales. However, the presence of a radial metallicity
gradient in the field stars complicates any comparison with the flat
metallicity distribution of the GCs. At what radius should the two be
compared? The field star metallicity distribution is reasonably well-fit
with an accreting box model like the one needed to solve the Galactic
G-dwarf problem
(Lynden-Bell 1975).
Rejkuba et al. (2005)
used an HST/ACS outer halo (38 kpc) pointing to derive a
metallicity distribution for NGC 5128 field stars very similar to that
of Harris et al., except that they found a larger number of metal-rich
stars (
solar). These stars were missed by Harris et al. because of large
bolometric corrections in V - I at high metallicities. An
age analysis of the Rejkuba et al. field, based primarily on the
locations of the ABG bump and the red clump, yielded a mean age of ~ 8
Gyr. However, the data are also consistent with a two-component model,
comprising an older base population plus a small percentage of 5-7 Gyr
stars. Based on the number of AGB stars
(Soria et al. 1996)
and Mira variables
(Rejkuba et al. 2003),
a figure of ~ 10% has been suggested
for the fraction of the total stellar mass in intermediate-age
stars. This is consistent with the 7% fraction of intermediate-age GCs
in the galaxy (see below).
NGC 5128 possesses only a relatively minor tail of metal-poor "halo" stars, even in the outermost pointing, but (as in M31) it is unclear whether this offers a constraint on the presence or absence of a Milky Way-like metal-poor projected ~ r-2 halo. Harris & Harris (2002) inferred from the paucity of metal-poor stars that Tbluen must be much larger than Tredn. They argued for a scenario in which the metal-poor GCs formed at the beginning of starbursts in small potential wells. These could be efficiently evacuated by supernova feedback before many accompanying metal-poor stars had a chance to form. In principle, other truncation mechanisms (including reionization) could produce a similar result (see Section 11 and 12).
There are good theoretical reasons to believe that GCs might form in the initial stages of major star forming events. This may be when the gas pressure is highest, favoring the creation of massive compact clusters (e.g., Elmegreen & Efremov 1997; Ashman & Zepf 2001). By contrast, the metal-rich GCs are expected to form during a disspational starburst in a more fully-assembled potential well. Here feedback is less effective, and many field stars form. The scenario in which GCs form in the initial stages of a starburst could also qualitatively reproduce a metallicity offset between metal-rich GCs and field stars. Such an offset is observed between the zeropoints of the metallicity-galaxy luminosity relations for metal-rich GCs and field stars (Forbes et al. 1997). However, a difficulty with any such comparison is the existence of radial metallicity gradients in galaxies, since metal-rich GCs show no such gradient. If both metal-poor GC and field star formation are simultaneously truncated at high redshift, their metallicity distributions may show no offset, as is the case in the Galaxy.
GCs in NGC 5128 appear to fall on the structural fundamental plane defined by Galactic and M31 GCs (Harris et al. 2002), although the mean cluster ellipticity in the current (small) sample is higher than that of Galactic GCs and more similar to the distribution in the old GCs of the LMC. Harris et al. also identified six massive GCs that had evidence of extratidal light. This could suggest that these objects are stripped nuclei of dEs, or simply that they are normal GCs, but that King models provide a suboptimal fit to their surface brightness profiles. Wide-field photometry in the Washington system by Harris et al. (2004), although complicated by the proximity of NGC 5128 and its low galactic latitude, revealed a total population of ~ 1000 GCs (SN = 1.4), evenly divided between metal-poor and metal-rich GCs. Peng et al. (2004) presented a large photometric and spectroscopic survey of more than 200 GCs in NGC 5128. 20-25 of the GCs had high enough S/N to determine spectroscopic ages and metallicities, and they found a wide range of ages for the metal-rich GCs, with a mean age of ~ 5 Gyr. Through direct comparison with integrated spectra of Galactic GCs from Cohen et al. (1998), Peng et al. found some evidence for [Mg/Fe] > 0, although perhaps not as high as in the Galaxy.
Beasley et al. (2005) estimated ages and metallicities for ~ 200 GCs in NGC 5128 from deep 2dF spectroscopy (with some overlap with the Peng et al. sample). Using the multi-line fitting method of Proctor et al. (2004), they deduced that the bulk of the GCs in this galaxy can be assigned to one of three old subpopulations: the usual metal-poor and metal-rich subpopulations, plus an intermediate-metallicity subpopulation reminiscent of one discovered in the Virgo gE NGC 4365 (Brodie et al. 2005). Evidence for trimodality is also presented in Woodley, Harris, & Harris (2005). In Figure 9, histograms of the GC B - I and metallicity distributions show the trimodality in the metallicity distribution. As is readily apparent, the nonlinear color-metallicity relationship combines the metal-poor and intermediate-metallicity peaks into a single peak in the B - I histogram. The presence of old intermediate-metallicity GCs is unusual among massive Es. If these GCs have an accompanying subpopulation of field stars, the field star metallicity distribution could be a combination (albeit not necessarily well-mixed) of intermediate-metallicity and metal-rich stars. However, this would be difficult to detect in present data. In addition to the old subpopulations, about 7% of the GCs in the Beasley et al. sample appear to be metal-rich with intermediate ages (2-8 Gyr), contrary to previous results. Candidate intermediate-age GCs were specially targeted, so this fraction is probably an upper limit.
 |
Figure 9. Smoothed kernel density histograms in V - I and [m/H] for GCs in NGC 5128. The kernels used are 0.03 mag and 0.1 dex, respectively. The metallicity histogram shows evidence for three distinct subpopulations of GCs, and a comparison of the two histograms indicates the clear nonlinearity of the color-metallicity relation (Beasley et al. 2006). |
The kinematics of the GCs are similar to those observed in the Galaxy: significant rotation is seen in the metal-rich GCs (as defined in Peng et al. 2004), but only weak rotation is apparent in the metal-poor GCs. This result is contrary to that expected to result from a disk-disk merger. In a merger remnant, the inner parts of the galaxy should display very little rotation because angular momentum is transferred to the outer regions. Of course, the remnant properties will depend on the details of the merger, and a statistical sample of Es that might plausibly have formed in recent major mergers does not yet exist. See Section 7 for addition discussion of GC kinematics.
An alternative approach to studying the relationship between GCs and field stars has been taken by Forte, Faifer, & Geisler (2005), who assume that the field star subpopulations follow the radial and color distributions of the GC subpopulations and use galaxy surface photometry to decompose the integrated light into metal-poor and metal-rich constituents. For NGC 1399, they find Tbluen ~ 25 and Tredn ~ 4. This metal-rich value is virtually identical to that of the Galaxy and, as before, the metal-poor value is much larger than the metal-rich one. While this approach is potentially a powerful tool for studying field stars in galaxies with unresolved stellar populations, substantial future work will be required to establish the veracity of the key underlying assumption: that the radial and metallicity distributions of GCs and their associated field stars are the same.