«Globular Clusters as Fossils of Galaxy Formation Richard B. Larson Yale Astronomy Department, Box 208101, New Haven, CT 06520-8101 Abstract. The ...»
Formation of the Galactic Halo.... Inside and Out
ASP Conference Series, Vol. 92, 1996
Heather Morrison and Ata Sarajedini, eds.
Globular Clusters as Fossils of Galaxy Formation
Richard B. Larson
Yale Astronomy Department, Box 208101, New Haven, CT 06520-8101
The globular clusters in the halos of large galaxies like our own are
almost certainly fossil remnants of the early star-forming subsystems from
which these galaxies were built. The ages of the halo clusters in our Galaxy indicate a prolonged period of galaxy building lasting at least several Gyr, and their masses indicate that they were formed in very massive star-forming complexes in protogalactic subsystems that may have resembled the present ‘blue compact dwarf’ galaxies. The surviving descendants of these subsystems are probably among the present dwarf spheroidal or nucleated dwarf galaxies, and the recently discovered Sagit- tarius dwarf is probably an example of such an object just now being accreted by our Galaxy and depositing into its halo four globular clusters including the second most luminous one in our Galaxy, M54.
1. Introduction: What Can the Globular Clusters Tell Us?
As the oldest known subunits of our Galaxy, the globular clusters are our primary fossils from its early evolution, and they may hold the key to understanding its formation. What kinds of information might they provide about galaxy formation? To help focus this question, it will be useful to recall the major scenarios that have been proposed for the origin of the globular clusters. Two broad types of possibilities have been considered: one is that the globular clusters were the ﬁrst condensed systems to form in the early universe, and the second is that they originated in larger star-forming systems that later merged to form the present galaxies. The ﬁrst possibility includes the suggestion of Peebles & Dicke (1968) that the globular clusters were formed by Jeans fragmentation in the early universe, and the suggestion of Fall & Rees (1985, 1988) that they were formed by thermal instability in early hot gaseous halos. The Fall & Rees hypothesis has been popular among theorists who have used it to predict the characteristic properties of globular clusters, although it has proven diﬃcult to justify the assumed thermal behavior of the cluster-forming gas clouds (Palla & Zinnecker 1988). The second type of scenario has in any case gained increasing attention in recent years, partly because it is more consistent with current cosmological models which predict that cosmic structures including galaxies are built up by the bottom-up merging of smaller units into larger ones (for reviews, see Larson 1990b, 1992b). In the ﬁrst type of scenario, the globular clusters might tell us something about fragmentation processes in the early universe, while in the bottom-up picture o
represent the densest surviving parts or ‘bones’ of the early subsystems that merged to form the present galaxies, and they might then tell us something about the chronology of the galaxy building process and something about the nature of the primitive galactic building blocks.
Galaxies are observed to be clustered hierarchically, and they are also distributed in a network of ﬁlaments and cellular structures on a wide range of scales (Maddox et al. 1990; de Lapparent, Geller, & Huchra 1991; Giovanelli & Haynes 1991); clearly, any adequate understanding of their origin must account for these clustering characteristics. Numerical simulations of the growth of structure in a dark-matter-dominated universe have successfully reproduced the basic clustering properties of galaxies by postulating an initial spectrum of density ﬂuctuations that is approximately a scale-free power law on the relevant range of scales; since these density ﬂuctuations have their largest amplitudes on the smallest scales, small systems tend to form ﬁrst and larger ones are then built up by the progressive merging of smaller structures into larger ones (Frenk et al. 1988; Zurek, Quinn, & Salmon 1988; Carlberg & Couchman 1989; White & Frenk 1991; Evrard, Summers, & Davis 1994). In a high-density universe, such mergers are predicted to continue at a signiﬁcant rate up to the present time and beyond. Simulations that include the physics of the gas show that the gas condenses strongly at the centers of the dark-matter ‘halos’ that form, and that if the feedback eﬀects of star formation are not included, the gas becomes highly clumped and loses much of its angular momentum to the dark matter, becoming much more centrally concentrated than in real galaxies (Navarro & Benz 1991;
Navarro & White 1994). Moreover, too many small objects are formed for the results to be consistent with the observed galaxy luminosity function (White & Frenk 1991; Cole 1991; Cole et al. 1994). Eﬀorts are under way by several groups to produce more realistic simulations that properly include the eﬀects of star formation, but the existing results suggest that the ﬁrst star formation may occur in gas that has become highly condensed at the centers of the ﬁrst dark-matter halos to form; even though such systems may not closely resemble most present-day galaxies, they may still be of interest as the possible birth sites of globular clusters, as will be discussed further below.
The most realistic simulations of the formation of a spiral galaxy like our own have been made by Katz (1992) and Steinmetz & M¨ller (1994, 1995), u who have simulated the evolution of a spherical galaxy-sized piece of a standard ‘cold dark matter’ universe which is arbitrarily given the appropriate amount of angular momentum. The results show an initial chaotic stage during which mergers between clumps build up a dark halo and a stellar spheroid, followed by a period during which the remaining gas organizes itself into a disk. During the initial chaotic stage several small satellites are formed by the condensation of gas in peripheral dark-matter clumps, and these satellites may survive for a few orbits before being disrupted and merged into the forming galaxy. It is plausible that such small satellites could be the birth sites of globular clusters, as in the hypothesis of Searle (1977) and Searle & Zinn (1978) that the globular clusters in the outer halo of our Galaxy were formed in protogalactic ‘fragments’ which survived for a time as independent star-forming systems before being merged to build up the halo (see also Larson 1988, 1990b, 1992b; Freeman 1990, 1993, 1996).
The Stellar Initial Mass Function 3
2. Globular Clusters and Galactic Chronology
If the globular clusters in the Galactic halo were formed in protogalactic fragments or satellite systems that were later merged to build the halo, the ages of these clusters might provide some information about the chronology of the galaxy building process. However, the cluster ages themselves do not necessarily directly record the merger history of the halo, since the halo clusters could have been formed in satellite systems long before these satellites were merged into the halo. Thus, the discovery of very old clusters in the Galactic halo would not strongly constrain possible merger histories, but the discovery of relatively young halo clusters would be a more signiﬁcant ﬁnding because the progenitor systems of these young clusters could only have been disrupted and merged with the halo after the clusters had formed, and such mergers could have occurred relatively recently in Galactic history.
In several earlier reviews of globular cluster chronology, it was concluded that the globular clusters in our Galaxy are not all coeval and that age diﬀerences of several Gyr exist in at least a few well-studied cases (e.g., Larson 1990a, 1992b). This conclusion still appears to be valid, and it is supported by the most recent discussion of this subject by Chaboyer, Demarque, & Sarajedini (1996), which is based on new age estimates for 43 globular clusters. Although the uncertainties remain too large for strong statements to be made about age diﬀerences in most individual cases, Chaboyer et al. (1996) argue that the sample of clusters studied has a statistically signiﬁcant age spread of at least 5 Gyr, and that some of the age diﬀerences are much larger than their uncertainties.
These results also support Zinn’s (1993) division of the halo clusters into two age groups on the basis of horizontal-branch morphology, since they show that Zinn’s ‘younger halo’ group has a signiﬁcantly smaller average age than his ‘old halo’ group by about 2 to 3 Gyr. Since the younger group has a larger average distance from the Galactic center, this ﬁnding also supports the suggestion of Searle & Zinn (1978) that the outer halo is on the average a few Gyr younger than the inner halo. The most striking feature of the age distribution of the Galactic globular clusters, which has now been known for some years and is not disputed by any researchers in the ﬁeld, is that the outer Galactic halo contains a group of mostly relatively small clusters including Pal 12, Rup 106, Arp 2, and Ter 7 whose ages of the order of 10 to 12 Gyr make them substantially younger than the bulk of the halo clusters.
These results suggest that the halo of our Galaxy was built up over a period of at least a few Gyr, perhaps by an inside-out accretion process, and that this process continued at least until the youngest outer halo clusters were formed perhaps 10 Gyr ago. The fact that the oldest stars and clusters in the local Galactic disk are about as old as the youngest halo clusters suggests that the local thin disk may have formed only after cluster formation in the halo was completed (Larson 1990a,b, 1992b). Although our Galaxy has evidently not experienced any further major accretion events capable of disrupting the disk, it is possible that minor accretion events aﬀecting only the halo and not the disk have continued to occur (Navarro, Frenk, & White 1994); indeed, several contributors to this meeting have already noted that our Galaxy is apparently just now accreting the recently discovered Sagittarius dwarf galaxy, which contains four already known outer-halo globular clusters including Arp 2 and Ter 7 (see 4 Richard B. Larson Section 6). This discovery illustrates the fact that, even though the formation of the Galactic globular clusters may have ceased about 10 Gyr ago, the merging of the cluster-forming subsystems into the halo could have continued until much more recently, and even up to the present time.
3. Lifetimes and Luminosity Functions: Are the Globular Clusters Special?
What can we say about the nature of the cluster-forming protogalactic fragments or satellite systems from which the Galactic halo was built? Clearly, these systems must have been capable of forming clusters much more massive than the smaller open clusters that have formed more recently in the disks of our Galaxy and other spiral galaxies. Did this require a special cluster formation mechanism that was qualitatively diﬀerent from the processes that have occurred more recently in our Galaxy and others, or was only a quantitative diﬀerence in the scale of cluster formation processes required?
An important related question is whether the globular clusters are a unique class of objects that diﬀer in some fundamental way from the open clusters, or whether there is a continuity in properties between globular clusters and open clusters, whose formation can be studied directly at the present time. A diﬀerence that has sometimes been emphasized is that the luminosity function of the globular clusters is sharply peaked and narrower than the luminosity function of the open clusters, which is not peaked but increases monotonically toward smaller luminosities (Harris 1993; van den Bergh 1993). Many authors have assumed, following Fall & Rees (1985, 1988), that the peaked luminosity function of the globular clusters requires a special formation mechanism that produces objects with a preferred mass. However, the fact that the globular clusters have a peaked luminosity function at the present time does not necessarily imply that they were formed with such a peaked luminosity function, since any clusters that might initially have been formed with masses much smaller than those of the observed clusters would have had shorter lifetimes and so might not have survived to the present time; the present peaked luminosity function could then have resulted just from cluster destruction processes.
The most important destruction process for most globular clusters is evaporation due to two-body relaxation (Aguilar, Hut, & Ostriker 1988), and the most realistic models of cluster evolution predict that a cluster evaporates completely after about 20 half-mass relaxation times (Spitzer 1987; Larson 1992a). The halfmass relaxation time is given approximately by 0.01 times the number of stars in the cluster times the crossing time at the half-mass radius, and the median value of this relaxation time for the observed globular clusters in our Galaxy is about 1 Gyr. The median evaporation time of these clusters is therefore predicted to be about 20 Gyr, which is comparable to the Hubble time and also to the typical ages of these clusters. Since the relaxation time is proportional to the cluster mass for a given half-mass density, the predicted evaporation times are generally shorter for the less massive clusters. The fact that the median cluster lifetime is comparable to the Hubble time is therefore almost certainly not an accident, but suggests that many clusters that were initially formed with smaller masses and shorter lifetimes have since disappeared, leaving only those clusters that The Stellar Initial Mass Function 5 had suﬃciently large masses and long lifetimes to survive to the present (Surdin 1979; Larson 1988, 1992a). The surviving clusters would then have a peaked luminosity function even if cluster formation processes in the Galactic halo had initially produced objects with a mass function increasing monotonically toward smaller masses, like the mass function of the open clusters.