The detailed properties of individual extragalactic GCs - ages, metallicities, and chemical abundances - are important constraints on theories of GC and galaxy formation. Emphasis has been placed on accurate age-dating of GCs. To the extent that Es formed in recent gas-rich major mergers, this should be reflected in young age measurements for their metal-rich GCs, while both in situ and accretion scenarios predict old ages for both subpopulations.
In principle, integrated-light spectroscopy of individual GCs offers much stronger constraints on ages and elemental abundances than broadband photometry. In practice, even low-resolution spectroscopy is challenging. At the distance of Virgo (~ 17 Mpc), GCs are already faint, with the turnover of the GCLF occurring at V ~ 23.5. Except within the Milky Way and M31, it has so far been possible to assemble only small samples of high-quality GC data, and these have required significant commitments of 8 to 10-m telescope time. Historically, spectroscopic samples have been not only small but also biased to the brightest and/or reddest objects in a given galaxy, forcing a reliance on photometric studies for global conclusions about metallicity distributions. As discussed below, this combination of photometry and small-sample spectroscopy has been used to establish that the ubiquitous color bimodality is due primarily to metallicity differences between the subpopulations, without the need to invoke age differences. There are also biases in spatial sampling: the fields of view of the spectrographs used in most studies are too small to sample the outermost GCs, and the GCs in the central regions are generally undersampled due to the minimum length of slitlets and the high central concentration of most GC systems.
There is a fundamental difference between metallicity and age studies in terms of the conclusions that can be derived from small spectroscopic samples. Age substructure can be identified in biased samples of metal-rich GCs, since younger GCs (with a standard GC mass function) will be brighter than their older counterparts. Thus magnitude-limited samples set an upper limit on the proportion of younger GCs in the system. By contrast, to properly sample the metallicity distribution, it is necessary to obtain spectra over a color range representative of the entire system. There is no technical reason why large-sample spectroscopic studies with the new generation of highly-multiplexing spectrographs like Keck/DEIMOS and VLT/VIMOS are not feasible, and such work may proliferate in the near future. In Figure 6, we show representative spectra of three GCs in M31: an old metal-poor GC, an old metal-rich GC, and an intermediate-metallicity, ~ 2 Gyr GC (Beasley et al. 2005).
Figure 6. Representative fluxed M31 GC spectra from Keck/LRIS: from bottom to top, an old metal-poor GC (088-150), an old metal-rich GC (225-280), and an intermediate-age, intermediate-metallicity GC (NB67). The spectrum of NB67 is from Beasley et al. (2005), and has been smoothed with a 5-pixel boxcar. The former two spectra are unpublished data of the authors. NB67 has a metallicity ~ -1 and an age of ~ 2 Gyr; its combination of strong Balmer and metal lines distinguishes it from the old GCs.
Most spectroscopic work on extragalactic GCs has utilized Lick/IDS indices (Burstein et al. 1984; Worthey et al. 1994; Trager et al. 1998). These were developed to measure absorption features from ~ 4000-6400 Å in the spectra of early-type galaxies. Such galaxies typically have large velocity dispersions; thus the low resolution of the Lick system (~ 8-11 Å) and wide index bandpasses (tens of Å) required by the IDS (Robinson & Wampler 1972) were not severe impositions. Indices (essentially equivalent widths) are defined in terms of a central bandpass which contains the feature of interest and flanking bandpasses that set the local pseudocontinuum. Observations of distant systems are compared to simple stellar population (SSP) evolutionary models using a set of fitting functions (e.g., Worthey 1994) that predict Lick indices as a function of stellar parameters (e.g., Teff, log g, [Fe/H]). Indices measured on modern spectra must be "corrected" to the Lick/IDS system through observations of standard stars.
This system is not optimal for GCs, which have low velocity dispersions and metallicities compared to galaxies. Although the typical S/N of extragalactic GC spectra has not justified the use of narrower indices until very recently, the opportunity now exists to improve the placement of index and pseudocontinuum bandpasses and to define new indices in the feature-rich wavelength region below 4000 Å. A number of groups are currently defining new index systems on the basis of higher resolution (~ 1-3 Å) stellar libraries (e.g., the 2.3 Å MILES library; Sanchez-Blazquez et al. 2006), as well as pursuing the direct fitting of models to observed spectra, avoiding indices entirely (e.g., Wolf et al. 2005).
In the following subsections we discuss observational results on metallicities, ages, and individual elemental abundances of extragalactic GCs.
4.1. Metallicities and Ages
Before proceeding, it is worth noting that the term metallicity is not precisely defined in the context of GCs. Some estimates (e.g., composite PCA metallicities; Strader & Brodie 2004) are tied to Galactic GC metallicities on the Zinn & West (1984) scale. This scale measures a nonlinear combination of metals and is unlikely to reflect either [Fe/H] or "true" [Z/H]. The Carretta & Gratton (1997) and the Kraft & Ivans (2004) Fe scales have yet to be extended to the metal-rich regime and calibrations are limited to the metallicity distribution of the Milky Way GC system. This barely touches solar and is poorly matched to the metal-rich regime of GCs in gEs, which extends to solar (and perhaps beyond). The metal-rich GCs in the Galaxy are also few in number and generally highly reddened, as they are concentrated in the bulge. The alternative to direct calibration is to derive metallicities solely from models. A myriad of issues remain with this approach, including uncertain isochrones, meager stellar libraries, corrections to the Lick system, and adjustments for non-solar abundance ratios (see discussion in Section 4.2). Some authors have adopted the more agnostic [m/H] (with "m" for metallicity) to describe their values; others calculate a "corrected" [Fe/H] from [m/H] or [Z/H] by subtracting an assumed or derived [/Fe] (e.g., Tantalo, Chiosi, & Bressan 1998). Direct comparisons of metallicities derived from different methods can easily be uncertain at the ~0.2-0.3 dex level.
One method to estimate metallicities for GCs is to simultaneously use multiple lines indices; a weighted combination of six indices was used to derive metallicities for individual GCs in the Milky Way and M87 (Brodie 1981; Brodie & Hanes 1986). Brodie & Huchra (1990, 1991) applied a refined version of this approach to GCs in a variety of galaxies. Two other studies derived metallicity calibrations using principle components analysis of indices of Galactic GCs: Gregg (1994) used a large set of narrow indices, and Strader & Brodie (2004) used Galactic GC spectra from Schiavon et al. (2004) to derive a metallicity estimator that is a linear combination of 11 Lick indices. The PCA methods provide accurate metallicities on the Zinn & West (1984) scale in the range -1.7 [Fe/H] 0.0. The other principal method is to simultaneously derive metallicity and age through comparisons to SSP models (see discussion below).
The bulk of spectroscopic studies of extragalactic GCs have been conducted with two instruments, Keck/LRIS and VLT/FORS. Ages, metallicities, and [/Fe] ratios have been estimated predominantly by measuring Lick indices.
In a Keck/LRIS study, Kissler-Patig et al. (1998) inferred that the majority of GCs in the Fornax gE NGC 1399 were old, but that a small percentage might be as young as a few Gyr. However, the robustness of these conclusions was hampered by the low S/N of the spectra. In another Keck/LRIS program, Cohen, Blakeslee, & Ryzhov (1998) studied a sample of ~ 150 GCs in M87. Despite large errors on the absorption line indices measured in many of their spectra, the size of the sample allowed a good statistical comparison with spectra of Milky Way GCs. The M87 and Milky Way GCs were found to populate similar areas in Mgb vs. H; and Mgb vs. Fe5270 diagrams, suggesting that the M87 GCs have ages and [Mg/Fe] ratios comparable to those of the Galactic GCs. These ages and [/Fe] ratios are known, from detailed analysis of individual cluster stars, to be 10 Gyr and ~ +0.2-0.3, respectively (e.g., Carney 2001). Beasley et al. (2000) found old ages for coadded (on the basis of C - T1 color) WHT spectra of GCs in NGC 4472. Kuntschner et al. (2002) presented a study of the S0 NGC 3115 showing that GCs in both the metal-poor and metal-rich subpopulations are old.
Subsequent studies undertaken with Keck/LRIS include: Ellipticals - NGC 1399 (Forbes et al. 2001); NGC 4472 (Cohen, Blakeslee, & Côté 2003); NGC 4365 (Larsen et al. 2003; Brodie et al. 2005); NGC 3610 (Strader et al. 2003; Strader, Brodie, & Forbes 2004b); NGC 1052 (Pierce et al. 2005); NGC 1407 (Cenarro et al. 2006); S0s - NGC 1023 (Brodie & Larsen 2002); NGC 524 (Beasley et al. 2004); Spirals - M81 (Schroder et al. 2002); NGC 4594 (Larsen et al. 2002). In all cases it was concluded that at most a small fraction of the observed samples of GCs were young or of intermediate age ( 5-6 Gyr). In the case studies of the intermediate-age merger remnant E NGC 3610, Strader et al. (2003, 2004b) found that only two out of ten prime candidate intermediate-age GCs had ages consistent with formation in the merger. Figure 7 shows a typical index-index plot used to derive metallicities and ages for GCs, in this case, from NGC 1407 (Cenarro et al. 2006). Here the GCs studied appear to have ages consistent with Galactic GCs, but extend to higher metallicities.
Figure 7. HA and H; vs. [MgFe]' age-metallicity index-index plots for GCs in NGC 1407 (circles and triangles; Cenarro et al. 2006) and Galactic GCs (stars, from Schiavon et al. 2004). The filled circles are "normal" GCs, the open circles have anomalously high [Mg/Fe] and [C/Fe], and the open triangles are GCs with enhanced Balmer lines (probably due to blue horizontal branches, but younger ages are also a possibility). The large star is a central re/8 aperture for NGC 1407 itself. The overplotted grids are Thomas, Maraston, & Korn (2004) models with [/Fe] = +0.3 and the ages and metallicities indicated. The model grid lines cross at old ages and low metallicities, making exact age estimates impossible.
In an effort to use more of the information available in a spectrum than is contained in a traditional index-index plot, Proctor, Forbes, & Beasley (2004; see also Proctor & Sansom 2002) developed a 2 minimization routine that simultaneously fit all of the Lick indices to SSP models, exploiting the different sensitivities of each of the indices to age, metallicity, and [/Fe]. Metallicities derived in this manner are complementary to ones derived using the PCA methods described above, and, for GCs in the range over which the PCA metallicities are calibrated, the agreement is excellent (e.g., Cenarro et al. 2006). Multi-index approaches offer the advantage of minimizing random and systematic problems in any individual index, such as may arise in low S/N data. Moreover, where the GC background is an emission line region (from either sky or the host galaxy), or when individual element abundance anomalies or extreme horizontal branch morphologies may be present in the GC, such precautions are essential (see below).
The best of the Keck data were compiled in a meta-analysis by Strader et al. (2005). This work synthesized more than a decade's worth of effort on the GCs systems of a heterogeneous sample of galaxies, spanning the range from dwarfs to gEs (all but NGC 4594 were of early-type). A typical galaxy had only 10-20 spectra of sufficiently high S/N to be included in the analysis. Lick indices measured from these spectra were combined, and ages were derived through direct differential comparison to indices of Galactic GCs, in part to avoid SSP model uncertainties. This study showed that both the metal-poor and metal-rich GC subpopulations had mean ages as old as (or older than) Galactic GCs. In Figure 8, from Strader et al. (2005), PCA metallicity is plotted against a composite Balmer line index of H; and HA (H was excluded because the strong G band lies in its blue pseudocontinuum bandpass). The implication from their sample of galaxies is that GC formation, and by extrapolation, the bulk of star formation in spheroids, took place at early times (z 2). As discussed in the Annual Review of Renzini (2006, this volume), fossil evidence from the local universe and in situ studies at high redshift lead to similar conclusions: most of the stars in massive Es formed at z ~ 2-5. Of course, studies of larger samples of GCs in a wide range of galaxies are needed to further test this result.
Figure 8. Combined Balmer index vs. metallicity for extragalactic GC subpopulations and Galactic GCs (Strader et al. 2005). The extragalactic subpopulations are plotted as open triangles (metal-poor), open circles (intermediate-metallicity), and open squares (metal-rich), with the subpopulations defined from broadband photometry of the GC systems. Individual Galactic GCs are plotted as filled triangles (data from Gregg 1994) and filled squares (data from Puzia et al. 2002b). At all metallicities, the extragalactic subpopulations appear coeval with (or older than) the comparison Galactic GCs. This suggests mean ages > 10-13 Gyr for the extragalactic GCs. 14 Gyr model lines from Thomas, Maraston, & Korn (2004) are superposed ([Z/H] = -2.25 to 0, with [/Fe] = 0 on bottom and +0.3 on top). These include a blue horizontal branch below [Z/H] = -1.35. At fixed metallicity, older ages lie at weaker Balmer line strength. The data are not fully calibrated to the models, so cannot be directly compared, and the offset between the data and the model lines is not significant.
This conclusion is still consistent with the existence of young early-type galaxies in the local universe (Trager et al. 2000), because galaxy age estimates are luminosity-weighted. A small "frosting" of recent star formation can easily result in the measurement of young spectroscopic ages for galaxies whose stars are predominately old, and there is an inverse relationship between the age and required burst strength. In addition, radial age gradients in early-type galaxies are often measured. Younger ages are derived from spectra extracted from small (1/8 of the effective radius; re) apertures than from larger (re / 2) ones. In any case, a small amount of star formation is expected to produce a correspondingly small subpopulation of young GCs, which could easily have escaped detection in existing samples. High quality spectra of > 100 GCs per galaxy, now obtainable with highly multiplexing instruments such as Keck/DEIMOS and VLT/VIMOS, would be necessary to probe these low-level star-forming events.
Puzia et al. (2005; see also Puzia 2003) presented a spectroscopic analysis of a more homogeneous sample of galaxies, comprising 5 Es and 2 S0s, all with -19.2 > MB > -21.2, and mostly in small groups. The low end of this luminosity range is near L*, so their sample, like that of Strader et al. (2005), was dominated by early-type galaxies brighter than the knee of the galaxy LF. They included a detailed discussion of the advantages of the different Balmer lines in terms of dynamic range, age-metallicity degeneracy, and ability to correct to the Lick system. Using a combined Balmer line index and the -insensitive metallicity proxy [MgFe]' (Thomas, Maraston, & Bender 2003), Puzia et al. found 4 GCs (out of their total of 17 high-quality spectra) with index measurements consistent with ages less than 9 Gyr (their Figure 7). While this sample is not sufficient to address the proportion of young GCs that might be present in any individual galaxy, the presence of some younger GCs in these typically group galaxies would be consistent with "downsizing", whereby lower-mass galaxies in lower-density environments tend to have younger mean ages (Cowie et al. 1996). There remains disagreement on the proportion of young GCs present in galaxies of moderate luminosity.
The evidence that most GCs in massive galaxies are quite old is relevant to the distinction between "core" and "power-law" E galaxies seen in the local universe. These designations stem from the surface brightness profiles in the very inner parts of the galaxy. The core galaxies tend to be bright (MV -21.5), have boxy isophotes, and do not rotate; the power-law galaxies are fainter, have disky isophotes, and are at least partially rotationally supported (Kormendy & Bender 1996; Faber et al. 1997). The working hypothesis for the creation of central cores is scouring by merging binary black holes. Kormendy et al. (2006) have unified this scenario by demonstrating that power-law galaxies have excess light in their central parts, above an extrapolation of a best-fit Sersic profile in the outer parts. Core galaxies have no such excess light. They argue that the distinction between core and power-law galaxies represents Es whose last major merger was dry or wet, respectively (i.e., without or with star formation). In power-law galaxies, black hole scouring may still have occurred, but it is swamped by the light from the stars in the center that formed in the merger. The old GC ages inferred from observations tentatively indicate that, even in power-law galaxies, the wet merger that formed the E either occurred long ago, or produced only a relatively small amount ( 10-20%) of the galaxy's current stellar mass. However, much larger spectroscopic samples of GCs in power-law galaxies are needed to better constrain the fraction of mass produced in wet mergers.
4.1.1 UNCERTAINTIES An important uncertainty in age determinations is the level of contribution from hot stars to the integrated spectra of GCs. These can "artificially" enhance the Balmer line strengths beyond the values set by the main sequence turnoff and can lead to an underestimation of the GC age. Blue stragglers and blue horizontal branch stars are the primary offenders - hotter stars, including extreme HB, AGB-manque, and post-AGB stars have little flux in the optical region where ages are usually derived. At least in the Milky Way, blue stragglers have quite a small effect on a GC's optical integrated spectrum (e.g., Schiavon et al. 2002). It is worth noting that this may not necessarily be the case in extragalactic GCs that have high binary fractions or are very compact.
This leaves blue HBs stars as the main systematic source of uncertainty in GC age estimates. HB morphologies are primarily set by GC metallicity, with Balmer line strengths peaking at [Fe/H] ~ -1.3 and decreasing toward lower metallicities as the stars become hotter and H is ionized. The classic "second-parameter problem" - that GCs of a given metallicity can have quite different HB morphologies - is currently unresolved, despite intensive research efforts over several decades. It has frequently been suggested that age is the second parameter (e.g., Chaboyer, Demarque, & Sarajedini 1996). However, this does not appear to be true in the few cases where direct turnoff age comparisons can be made by normalizing the main sequence luminosity functions (e.g, Stetson, Vandenbergh, & Bolte 1996). While age differences certainly will produce changes in HB morphology, it is dangerous to use the HB as an age indicator. As an example, several metal-poor GCs in M33 were studied with HST/WPFC2 by Sarajedini et al. (2000) and found to have unusually red HBs for their metallicities. Sarajedini et al. suggested intermediate ages might be the cause. However, Larsen et al. (2002) derived dynamical masses for these GCs and instead found that they had M/L ratios typical of old GCs. At present there is no known a priori predictor of GC HB morphology. Consequently, various HB diagnostics have been suggested. These include (i) the Ca H+K line strength inversion; enhanced H from hot stars can increase the Ca H line equivalent width compared to the Ca K line (Rose 1985), (ii) Near-ultraviolet (NUV - V) colors, and (iii) a tendency for progressively younger ages to be inferred from increasingly higher-order Balmer lines because of an increasing contribution from putative blue HB stars (Schiavon et al. 2004). The effect of BHB stars on integrated Balmer line strength can be substantial: Puzia et al. (2005) estimated that H may be increased by up to 0.4 Å (based on their analysis of Galactic GCs), while Maraston (2005) inferred an even greater effect from her SSP models at high metallicities (see also Thomas, Maraston, & Korn 2004).
Aside from the classic second-parameter effect, there is some evidence that the HB morphologies of GCs in other galaxies can be quite different from those in the Milky Way. For example, in a HST/WFPC2 study by Rich et al. (2005), a sample of 12 M31 GCs appeared offset from the Milky Way HB-metallicity relation. One potential explanation is that these M31 GCs are ~ 1-2 Gyr younger than their Galactic counterparts. More startling is the finding by Sohn et al. (2005), from an HST/STIS FUV imaging study of bright M87 GCs, that all the GCs have much bluer FUV - V colors than Galactic GCs of the same metallicity. Increasing the age of the M87 GCs relative to Galactic GCs is one potential solution, but it would require implausibly large age differences of 2-4 Gyr to bring the two samples into line. Fuel consumption arguments, and the fact that spectroscopic ages of these GCs are old (Cohen et al. 1998), suggest that the hot stars in M87 GCs are a redistribution of stars blueward of a normal BHB, and not simply an extension of it. A caveat to this finding is the UV-flux limited nature of their sample, which tends to select the most extreme GCs.
A night of 8-10-m telescope time can generate spectra of GCs at the distance of Virgo (~ 17 Mpc) from which H; line index strengths can be measured to an accuracy of ~ 0.1Å for GCs with V ~ 21.5 (for fainter GCs, the errors are correspondingly larger). As Figure 7 illustrates, this translates into an error of ~ 1-2 Gyr at old ages. Since the isochrones are more widely separated at younger ages, the ability to discriminate fine age differences improves with decreasing age. In these SSP model grids, BHB effects actually cause the isochrones to cross at low metallicities and old ages, setting a fundamental upper limit on our age estimates of ~ 10 Gyr at these metallicities. It is important to remember that the actual ages of the majority of GCs in the universe could well be older than this.
It is widely recognized that measurements of the [/Fe] ratio of a stellar population can provide valuable insight into its star formation history, particularly by constraining the timing and duration of a starburst. Most -elements (e.g., Mg, Ti, Ca, Si) and 1/3 of the Fe (for the solar mixture) are generally thought to form in Type II SNe. Since these come from the explosions of massive stars, they closely trace the star formation rate, and begin to ignite within tens of Myr of the onset of a starburst. The remainder of the Fe is produced in Type Ia SNe, which typically occur over Gyr timescales. Enhanced [/Fe] ratios would thus indicate rapid star formation, occurring before there was time for significant enrichment of the interstellar medium by Type Ia SNe. This is generally the case in early-type galaxies (e.g., Worthey, Faber, & Gonzalez 1992; Matteucci 1994; Trager et al. 2000; Thomas et al. 2005). A complication is that the timescale over which fresh ejecta become incorporated into interstellar gas is not necessarily known. Other nucleosynthetic sites, operating on a variety of time scales, also contribute to the cosmic mix (e.g., AGB stars produce a significant amount of s-process and light elements).
The naive expectation for metal-poor GCs in external galaxies is that they will have supersolar [/Fe], since they formed early in the universe, before there was substantial metal enrichment. The situation for metal-rich GCs is less clear, as it will depend upon the exact formation mechanism. If they formed along with most of the field stars in massive Es, they should share the [/Fe] ~ +0.2-0.4 of their parent galaxies. They could perhaps reach even higher values if they formed preferentially early in the starburst. Metal-rich GCs formed at lower redshift from ~ solar metallicity gas in major disk-disk mergers might be expected to have [/Fe] closer to 0, since it is more difficult to enhance [/Fe] when starting from a high metallicity (although metallicity gradients in spiral disks must also be considered). The results from extragalactic GCs do not generally agree with these expectations, though much of this may be due to the observational and technical difficulties in accurately estimating [/Fe], as discussed below.
In theory, it should be very easy to determine [/Fe] from low-resolution GC spectra. Mg is the element of choice because of the strong Mgb triplet and MgH bandhead in the optical. Trager et al. (2000) described a method for estimating [Mg/Fe] using models by Worthey (1994) and corrections from Trippico & Bell (1995). However, the widespread estimation of this quantity in extragalactic GCs did not occur until Thomas, Maraston & Bender (2003) published the first models to explicitly include the effects of -enhancement.
In practice, the interpretation of the observations using these models has been complicated. The GCs in the metal-rich subpopulations in a variety of massive galaxies have been found to have [/Fe] ranging from 0 to ~ +0.3 (e.g., Kuntschner et al. 2002; Beasley et al. 2005) or higher (Puzia et al. 2005 find ~ +0.45 but with large scatter). By contrast, the metal-poor GCs frequently appear to have [/Fe] ~ 0 or even lower (e.g., Olsen et al. 2004; Pierce et al. 2005), although the convergence of SSP model lines at low metallicities increases the errors on these determinations of [/Fe]. If [/Fe] for metal-poor GCs were indeed this low, it would be quite surprising. In the simple view of nucleosynthesis outlined above, supersolar [/Fe] results from star formation over "short" (< 1 Gyr) timescales while solar or subsolar [/Fe] indicates very extended star formation histories. Thus we expect the metal-poor GCs to have high [/Fe].
Galactic GCs from both subpopulations have [/Fe] ~ +0.3 (e.g., Carney 1996) or perhaps a little lower for some metal-rich bulge GCs. These values are quite closely reproduced by the popular Thomas et al. (2003) SSP models for the Galactic GC data of Puzia et al. (2002), although the results from applications to extragalactic GCs are mixed (as described above). Current stellar libraries and the Worthey (1994) fitting functions utilize few stars with metallicities in the range of metal-poor GCs, and SSP models may be more uncertain in this regime.
Schiavon (2006) gave a good summary of the issues involved in creating non-solar [/Fe] models. These include (i) the importance of employing the proper isochrones and luminosity functions, and (ii) the need to correct the index predictions of the fitting functions. Item (ii) is usually carried out differentially, using stellar models for stars in representative parts of the CMD. A preliminary attack on this problem by Trippico & Bell (1995), using just three stars, produced broadly similar results to the more thorough treatment of Korn et al. (2005). Korn et al. also discussed in detail the elemental sensitivities of the individual Lick indices.
A potentially confusing difference among models is the treatment of [/Fe] vs. [Fe/H]. Thomas et al. (2003) calculated their models at fixed total metallicity ([Z/H]) and produced supersolar [/Fe] by lowering [Fe/H], since the dominant component of Z is the -element O. By contrast, Schiavon (2006) calculated models at fixed [Fe/H]. Schiavon's approach has the benefit of a direct link to the measurable quantity [Fe/H].
These SSP uncertainties, when combined with the wide Lick index bandpasses, which always admit contributions from other elements in addition to the targeted feature, make most current estimates of [/Fe] untrustworthy. We suggest intensive study of nearby GC systems (e.g., M31), where potentially more accurate results from high-resolution integrated spectroscopy of GCs can be directly compared to low-resolution spectra.
4.3. Abundance Anomalies
In addition to the -elements, GCs show a variety of abundance anomalies with respect to the solar mixture. This is most clear in the Galaxy, where detailed study of individual stars in GCs is possible (see the Annual Review of Gratton, Sneden, & Carretta 2004). In extragalactic GCs, the most obvious anomaly is CN-enhancement. This was first reported in M31 GCs by Burstein et al. (1984), and, at least in this galaxy, appears to be due to an excess of N above even the levels in Galactic GCs (see Section 6.1), which themselves are enhanced in N over the solar mixture. Trager (2004) has argued that a similar CN anomaly (presumably due to N) is present in many GC systems, including NGC 3115 (Kuntschner et al. 2002), the old GCs in NGC 3610 (Strader et al. 2003a), and the Fornax dSph (Strader et al. 2003b). It is also present in GCs in the gE NGC 1407 (Cenarro et al. 2006). We must consider very high N abundances to be a generic feature of the early chemical evolution of GC systems (and perhaps their host galaxies). Interestingly, of the two confirmed metal-rich intermediate-age GCs in NGC 3610, one shows the CN anomaly, and one does not.