STAR CLUSTERS, GLOBULAR, EXTRAGALACTIC WILLIAM E. HARRIS The extremely old star clusters that inhabit the halo or spheroidal region of a galaxy are rather commonly found in galaxies other than our own Milky Way. A globular cluster system (GCS) is the ensemble of all such clusters belonging to one galaxy, and can be treated as a dynamically distinct subsystem of the halo (see Fig. 1). Typically, the GCS comprises only about 1% of the total halo light (or mass in the form of visible stars), and so it must represent a special subset of all the matter found in the halo: Either the globular clusters we see now are merely the ones that have survived a long series of erosive processes since their epoch of formation in the early Galaxy, or else they were formed under initial conditions unlike those for the vast majority of halo stars. Because the globular clusters are among the oldest visible objects anywhere, comparing the GCSs in different galaxies is a way to find out how different (or how similar) the first processes of star formation were in different protogalactic environments. The intrinsic luminosity of an average globular cluster is about 10* L*(solar luminosity units) and the largest ones attain almost 2x10* L*. They can therefore be detected in relatively distant galaxies (with CCD imaging on large ground-based telescopes, the brightest globular clusters are visible around large galaxies as remote as *100 Mpc), but the detail in which they can be studied falls dramatically with distance. For galaxies more distant than d*2 Mpc, the individual stars in the globular clusters are no longer directly resolvable; for d*5 Mpc, the clusters themselves have angular sizes *1 arcsec and so appear star-like on photographs or digital images (these distance limits should be multiplied by *5 for the higher angular resolutions that are possible by imaging from space). Thus for most galaxies, the GCS is detectable primarily as an excess population of star-like images concentrated around the central spheroid and halo. The existence and structure of the GCS in our own Milky Way was first recognized in the early part of this century by Harlow Shapley; in his classic work of 1918 he used the space distribution of the globular clusters to derive the distance of the Sun from our galactic center for the first time. In the early 1930s, Edwin P. Hubble discovered the GCS around the nearest large galaxy, M31 (the "Andromeda nebula"); in the 1950s and early 1960s, especially with the work of William A. Baum, Allan R. Sandage, and Rene Racine, GCSs were found around many of the large elliptical galaxies in Virgo. In M31, which contains roughly twice as many globular clusters as does the Milky Way, it is possible to study the integrated spectral properties of the clusters in much detail, and even to measure color-magnitude diagrams for many individual clusters. These measurements reveal certain differences from the Milky Way globular clusters that may be due to slightly different ages or abundance ratios, but in general they demonstrate that the globular clusters in these Local Group galaxies are basically the same objects. For clusters in much more distant systems, low-dispersion spectra and photometric indices must be used as more approximate indicators of composition, but the pattern of first-order similarity is maintained to a remarkable degree. When other global characteristics of the CGSs are added (see below), the existing evidence suggests that the clusters we see in the halos of spiral galaxies, disk galaxies, dwarf ellipticals, and giant ellipticals are all generically similar; that is, they are old, populous star clusters with heavy-element abundance ratios ranging from near-solar down to about 1/100 solar. The presence of GCS therefore appears to represent a common theme in the very earliest evolutionary stages of large galaxies. Obtaining accurate observations for globular clusters in galaxies much beyond the Local Group was a formidable task when photographic emulsions were the principal detectors for astronomical imaging and spectroscopy. The advent of CCD detectors in the mid-1980s, with their enormously higher quantum efficiency and linearity, began to turn this field into a mature observational subject and brought a far wider range of galaxies into reach. Nevertheless, the large-scale properties of a GCS must, in most cases, be characterized by just a few simple quantities: The total number of clusters present in the system. Their spatial distribution around the parent galaxy. The internal dynamics of the system (the average motions of the clusters in the halo). The number of clusters at any given brightness (i.e., their luminosity function or LF). The spectroscopic properties of the clusters or (more crudely) their photometric colors, indicative of their chemical composition. GCSs have now been observed in several dozen galaxies. In general, bigger galaxies have more clusters, the number increasing roughly in direct proportion to the total spheroid luminosity (i.e., the amount of old stellar population present) of the parent galaxy. For elliptical and SO galaxies, the "spheroid" luminosity in this sense means essentially that of the entire galaxy. For disk galaxies of type Sa to Sb, the halo is relatively less prominent, but virtually all the star clusters in the halo and central spheroid region still resemble the classic old globular type. For galaxies still further along the Hubble sequence (types, Sc, Sd, and irregular) the stellar component is almost completely dominated by the younger Population I, the visible halo is nearly negligible, and very few star clusters can be found that are plainly classifiable as old globular. The prototypes of this category are the Magellanic Clouds, in which only a handful of star clusters have the unambiguous defining marks of extreme age (halo-type space motions, low metallicity, and color-magnitude diagrams with well developed red giant and horizontal branches or RR Lyrae variables). The ratio of total cluster population to spheroid (halo) luminosity is called the specific frequency S of the GCS. This ratio is a convenient index of comparison for galaxies of different types and sizes. Elliptical galaxies generally have 2-3 times higher S-values (that is, relatively more globulars) than disk or spiral galaxies. The surrounding environment of the parent galaxy may, however, have an equally important effect: E galaxies in rich groups such as the Virgo cluster tend to have higher specific frequencies than those in sparse groups, and some giant E galaxies that are sitting at the centers of rich clusters have enormously larger globular cluster populations, by about 3 times the normal numbers. Because it is a ratio, S is relatively unaffected by interactions between galaxies (except for the most extreme total mergers), because encounters between galaxies exchange or eject both halo stars and clusters in similar relative numbers. Thus differences in specific frequency from galaxy to galaxy suggest that the formation rate of globular clusters in protogalaxies varied depending on the surrounding density and type of other protogalaxies. The true spatial extent of a GCS can be enormous (reaching to galactocentric radii larger than 100 kpc in the biggest giant ellipticals). An important feature of their spatial distribution in large ellipticals is that the GCS is often less centrally concentrated than the halo of the galaxy itself. If their space distribution (number per unit volume) is approximated by a power law *****, then within the same galaxy the exponent n(GCS) may be smaller than n(halo) by *0.5. However, in other galaxies, such as smaller ellipticals or disk systems, the GCS space distribution seems to follow the spheroid light distribution more closely. A common phenomenon for large galaxies is that the innermost "core" (r*2 kpc) of the GCS has a flat, near-uniform density; either the clusters did not form there as easily, or (perhaps more likely) they were depopulated by the destructive processes that become extremely effective very close the nucleus of the galaxy, such as dynamical friction and tidal shocking. The internal dynamics of a GCS may be studied by measuring the radial velocities of the individual clusters. Data of this type have been obtained for only a few large galaxies (such as the Andromeda galaxy M31, the large nearby elliptical NGC 5128, and the Virgo ellipticals M87 and M49). However, these data confirm the overall pattern seen in the Milky Way halo clusters, that is, that the GCS is a dynamically "hot" system in which the clusters follow a wide range of randomly oriented, elliptical orbits with little or no overall rotation around the galactic center. At any galactocentric radius, the velocity dispersion (the scatter of the individual cluster velocities around the mean) can be employed as a direct estimate of the mass of the galaxy M(r) contained within radius r, through the generalized viral expression ************************, where ** is the measured radial velocity of each cluster relative to the mean of the entire system and * is a constant determined by the mean orbital characteristics of the clusters (**0.1 for an isotropic velocity distribution). Results of this type have been used for selected galaxies (notably the giant ellipticals M87 and NGC 5128, as well as the Milky Way) as important confirmation that large amounts of "dark matter' dominate the mass distribution of galaxies in their outer regions [the velocity dispersion ** is found to be approximately the same at any radius r, so M(r) must increase in direct proportion to r]. The luminosity function (LF) for globular clusters is likely to be largely determined by their original mass spectrum of formation, because most clusters lie in the low-density halo regions of their parent galaxy where tidal disruption and other destructive effects are minor. Thus another important clue to the apparent near-universality of their formation process is that the LF has a characteristic shape that seems to be reproduced from one galaxy to another: Clusters with sizes near 10** L* are the most common, with ones at either higher or lower luminosity being less frequent. The number of clusters at a given magnitude (i.e., the logarithmic luminosity) is empirically well matched by a simple gaussian function. Thus to first order, the LF is described by only three parameters: the total sample population N, the magnitude m* of the peak frequency, and the dispersion *(m) of the distribution. Existing data for several galaxies of a wide range of types and sizes indicate that *(m) is consistently in the range 1.3*0.2 mag, and that m* is roughly constant (to within *0.3 mag or *30%), but not enough is yet known about the LF parameters to establish in detail their systematic variation with galaxy type. If these features can be fully calibrated and better understood theoretically, they will become attractive "standard candles" for calibrating the extragalactic distance scale, because they are such luminous objects and can be found readily in quite distant galaxies. Additional Reading Grindlay, J.E. and Philip, A.G.D., eds.(1988). Globular Cluster Systems in Galaxies. IAU Symposium 126. Kluwer Academic Publishers, Dordrecht. Harris, W.E.(1991). Globular clusters in distant galaxies. Sky and Telescope 81 148. Harris, W.E. and Racine, R.(1979). Globular clusters in galaxies. Ann. Rev. Astron. Ap. 17 241.