This section is intended to give a feeling for the observational methods used to study globular cluster systems. It addresses the problems still encountered in imaging and spectroscopy, as well as the improvements to be expected with future instruments.
To set the stage: we are trying to analyze the light of objects with typical half-light radii of 1 to 5 pc, at distances of 10 to 100 Mpc (i.e. half-light radii of 0.01" to 0.1" ), and absolute magnitudes ranging from MV ~ -10 to -4 (i.e. V > 20). The galaxies in the nearest galaxy clusters (Virgo, Fornax) have globular cluster luminosity functions that peak in magnitude around V ~ 24, and globular clusters have typical half-light radii of ~ 0.05" .
We intend to study both the properties of the individual clusters, as well as the ones of the whole cluster system. For the individual globular clusters, our goal is to derive their ages, chemical abundances, sizes and eventually masses. This requires spectral information (the crudest being just a color) and high angular resolution. For the whole system, our goal is to determine the total numbers, the globular cluster luminosity function, the spatial distributions (extent or density profile, ellipticity), and any radial dependencies of the cluster properties (e.g. metallicity gradients). These properties should also be measured for individual sub-populations, if they are present. The requirements for the systems are therefore deep, wide-field imaging, and the ability to distinguish potential sub-populations from each other.
3.1. Optical photometry
Globular clusters outside the Local Group were, for a long time, exclusively studied with optical photometry. Optical, ground-based photometry (reaching V > 24 in a field > 5' × 5') turns out to be sufficient to determine most morphological properties of the systems (see Sect. 4). The depth allows to reasonably sample the luminosity function, and a field of several arcminutes a side usually covers the vast majority of the system.
Problems with optical photometry arise when trying to determine ages and metallicities. It is well known that broad-band optical colors are degenerate in age and metallicity (e.g. Worthey 1994). A younger age is compensated by a higher metallicity in most broad-band, optical colors. To some extent the problem is solved by the fact that most globular clusters are older than several Gyr, and colors do not depend significantly on age in that range. This, of course, means that deriving ages from optical colors is hopeless, except for young clusters as seen e.g. in mergers.
For old clusters, the goal is to find a color that is as sensitive to metallicity as possible. The widely used V - I color is the least sensitive color to metallicity. B - V and B - I do better in the Johnson-Cousins system (e.g. Couture et al. 1991 for one of the first comparisons), but the mini break-through came with the use of Washington filters (e.g. Geisler & Forte 1990). These allowed the discovery of the first multi-modal globular cluster color distributions around galaxies (Zepf & Ashman 1993). However, the common use of the Johnson system, combined with large errors in the photometry at faint magnitudes, do still not allow a clean separation of individual globular clusters into sub-populations around most galaxies.
Another problem with ground-based imaging is that its resolution is by far insufficient to resolve globular clusters. This prohibits the unambiguous identification of globular clusters from foreground stars and background galaxies. In most cases the over-density of globular clusters around the host galaxy is sufficient to derive the general properties of the system. Control fields should be used (although often left out because of the lack of observing time), and statistical background subtraction performed. The individual identification of globular clusters became possible with WFPC2 on HST. Globular clusters appear barely extended in WFPC2 images, which allows on the one hand to reliably separate them from foreground stars and background galaxies, and on the other hand to systematically study for the first time globular cluster sizes outside the Local group (Kundu & Whitmore 1998, Puzia et al. 1999, Kundu et al. 1999). The disadvantage of WFPC2 observations is that the vast majority was carried out in V - I, the least performant system in terms of metallicity sensitivity. Furthermore, the WFPC2 has a small field of view which biases all the analysis towards the center of the galaxies, making it very hard to derive the global properties of a system without large extrapolations or multiple pointings.
In summary, ground-based photometry returns the general properties of the systems, and eventually of the sub-systems when high quality photometry is obtained. It suffers from confusion when identifying individual clusters, and is limited in age/metallicity determinations. Space photometry is currently as bad in deriving ages/metallicities, but allows to determine sizes of individual clusters. The current small fields, however, restrict the studies of whole systems.
Future progress is expected with the many wide-field imagers coming online, provided that deep enough photometry is obtained (errors < 0.05 mag at V = 24). These will provide a large number of targets for spectroscopic follow-up. In space, the ACS to be mounted on HST will superseed the WFPC2. The field of view remains modest, but the slightly higher resolution will support further size determinations, and the higher throughput will allow a more clever choice of filters, including U and B.
3.2. Near-infrared photometry
Since the introduction of 1024 × 1024 pixel arrays in the near infrared a couple of years ago, imaging at wavelength from 1.2 µm to 2.5 µm became competitive in terms of depths and field size with optical imaging (see Figure 2).
Figure 2. NGC 4365 observed in the V band with the WFPC2 on HST, and in the K band with ISAAC on the VLT. The field of views are similar; the resolution is ~ 0.1" in the HST images, ~ 0.4" in the K images; the depths are comparable. This illustrates that optical and near-infrared imaging are becoming more and more similar for purposes of studying globular cluster systems (images provided by T.H. Puzia).
Historically, the first near-infrared measurements of extragalactic globular clusters were carried out in M31 (Frogel, Persson & Cohen 1980) and the Large Magellanic Cloud (Persson et al. 1983). Why extend the wavelength range to the near infrared? For old globular clusters, V - K is a measure of the temperature of the red giant branch that is directly dependent on metallicity but hardly on age. V - K is even more sensitive to metallicity than the Washington C - T1 index. The combination of optical and near-infrared colors is therefore superior to optical imaging alone, both for deriving metallicities, and for a clean separation of cluster sub-populations (see Figure 3). It is also used to detect potential sub-populations were optical colors failed to reveal any.
Figure 3. Globular cluster colors shown in the (V-I) - (V-K) plane. For this plot, data for NGC 3115 from ISAAC on the VLT and WFPC2 onboard HST were combined to allow a better separation of the two sub-populations (taken from Puzia et al. 2000, in preparation). Density contours show the color peaks for two main sub-populations.
In young populations, V - K is most sensitive to the asymptotic giant branch which dominates the light of populations that are 0.2 to 1 Gyr old. The combination of optical and near infrared colors can be used to derive ages (and metallicities) of these populations (e.g. Maraston, Kissler-Patig, & Brodie 2000).
The disadvantages of complementing optical with near-infrared colors is the need for a second instrument (usually a second observing run) in addition to the optical one. Near-infrared observations will continue fighting against the high sky background in addition to the background light of the galaxy which requires blank sky observations. Overall, obtaining near-infrared data is still very time consuming when compared to optical studies. For example, a deep K image of a galaxy will require a full night of observations. Currently both depth and field size do not allow the near infrared to fully replace optical colors for the study of morphological properties or the globular cluster luminosity function. But this might happen in the future whth the NGST.
The immediate future looks bright, with a number of "wide-field" imagers being available, such as ISAAC on the VLT, SOFI on the 3.5m NTT, the Omega systems on the 3.5m Calar Alto, etc.. and 2k × 2k infrared arrays coming soon. The ideal future instrument would have a dichroic which would allow to observe simultaneously in the near-infrared and the optical.
3.3. Multi-object spectroscopy
Spectroscopy is the only way to unambiguously associate a globular cluster with its host galaxy by matching their radial velocities. Also, it is the most precise way to determine the metallicity of single objects, and the only way to determine individual ages. Obviously, it is also the only way to get radial velocities. Ideally, one would like a spectrum of each globular cluster identified from imaging.
In practice, good spectra are still hard to obtain. Early attempts with 4m-class telescopes succeeded in obtaining radial velocities, but mostly failed to determine reliable chemical abundances (see Sect. 6). With the arrival of 10m-class telescopes, it became feasible to obtain spectra with high enough signal-to-noise to derive chemical abundances (Kissler-Patig et al. 1998a, Cohen, Blakeslee & Ryzhov 1998). Such studies are still limited to relatively bright objects (V < 23) and remain time consuming (~ 3h integration time for low-resolution spectroscopy). Figure 4 shows a few examples of globular clusters in NGC 1399. High-resolution spectroscopy in order to measure internal velocity dispersions of individual clusters is still out of reach for old clusters, and was only carried out for two nearby super star clusters (Ho & Fillipenko 1996a, b), in addition to several clusters within the Local Group. Even low-resolution spectroscopy is currently still limited to follow-ups on photometric studies, targeting a number of selected, representative clusters, rather than building up own spectroscopic samples.
Figure 4. Three representative spectra of globular clusters are shown, ranging from blue, over red, to very red color. While the H gets slightly weaker from the blue to the red object, the metal lines (Mg, Fe, Na) become much stronger (taken from Kissler-Patig et al. 1998a).
Current problems are the low signal-to-noise, even with 10m-telescopes, that prohibit very accurate age or metallicity determinations for individual clusters. The multiplexity of the existing instruments (FORS1 & 2 on the VLT, LRIS on Keck) is low and only allows to spectroscopy a limited number of selected targets. Finally, the absorption indices that are being measured on the spectra in order to determine the various element abundances are not optimally defined. These indices lie in the region 3800Å to 6000Å and were designed for spectra with 8Å to 9Å resolution. They often include a number of absorption lines in the bandpass (or pseudo-continuum) other than the element to be measured. This introduces an additional dependence e.g. of the Balmer indices on metallicity, etc... Using a slightly higher resolution might help defining better indices.
The immediate future of spectroscopy are instruments such as VIMOS on the VLT or DEIMOS on Keck that will allow a multiplexity of 100 to 150. These will allow to increase the exposure times and slightly the spectral resolution to solve a number of the problems mentioned above. These will also allow to obtain several hundred radial velocities of globular clusters around a given galaxy in a single night, improving significantly the potential of kinematical studies (see Sect. 6).