Gregory D. Bothun

Over the last 20 years, rather compelling evidence has been gathered which suggests that there is more mass contained in galaxies than can be accounted for by the amount of light that they emit. This blurs our concept of what actually constitutes a galaxy and suggests that the luminous stars that form from the gas in galaxies may simply be tracers of some large concentration of nonluminous matter, whose nature is unknown. This complicates our concept of galaxy formation because it is not clear whether it fundamentally refers to the problem of the formation of dark matter potentials or to the problem of the condensation of baryonic material within these dark potentials. The formation of the luminous component of galaxies, will be discussed here. In this context, the term protogalaxy refers to a galaxy that is experiencing its first generation of star formation, having consisted totally of hydrogen gas trapped in some dark matter potential prior to this point.

Although nearby galaxies presently exist in a wide variety of shapes and forms, there are three general features of the galaxy population that must be explained by any theory:

  1. Galaxies come in two basic kinds, namely, elliptical galaxies that are supported by the internal velocity dispersion of the stars (isotropic orbits) and spiral galaxies that are flattened disk systems in which the stellar orbits are circular.

  2. Most spiral galaxies contain a mini-elliptical galaxy at their center (i.e., a bulge).

  3. Galaxies are not randomly distributed in space but are highly clustered.

Galaxies may be conveniently parameterized by their bulge-to-disk (B/D) ratio, where the bulge component is a spheroidal distribution of stars with isotropic orbits and the disk component is a highly flattened distribution of stars and gas with circular orbits. Interestingly, the B/D ratio of a given galaxy seems to be dependent on its environment in that galaxies located in the cores of rich clusters are preferentially bulge dominated, whereas those located in the lower-density regions of the universe are preferentially disk dominated. This indicates that protogalaxies rarely form in isolation and hence interactions with other protogalaxies at the time of formation may greatly determine the evolutionary course of a given galaxy.


The formation of large-scale structure (i.e., clusters of galaxies, superclusters of clusters) is intimately related to the formation of individual galaxies. Currently, there are two competing scenarios for the formation of structure in the universe and current observations are incapable of distinguishing between them. Clearly, galaxies represent density enhancements in the universe which makes their existence somewhat difficult to understand in terms of the hot big bang theory for the origin of the universe. In particular, it is known from observations of the cosmic microwave background (CMB) that the early universe was very homogeneous on large scales. Furthermore, when the universe was less than approximately 300,000 years old, the energy density contained in this radiation field far exceeded that which was contained in matter. In this physical situation, gravity, in effect, was nonexistent and the distribution of matter was governed by the distribution of radiation, which we now measure to be quite homogeneous. Hence it is quite paradoxical that any inhomogeneities would form and so our understanding of the existence of galaxies is challenged at a very fundamental level.

Because of this paradox, it is common practice to use the present-day distribution of galaxies as a fossilized imprint of what the original spectrum of density perturbations must have been after matter and radiation decoupled and gravity became important. This leads to two possible scenarios for the formation of galaxies and clusters of galaxies. In the fragmentation picture, only very massive perturbations (mass up to 106 Msmsun (solar masses); a typical galaxy has a mass of 2-5 x 1011 Msmsun) survived the radiation-dominated era. These large-mass perturbations are identified today as the largest known superclusters of galaxies. Subsequent cooling and fragmentation within the overall density perturbation then produces smaller-scale clusters of galaxies (of mass 1014-1015 Msmsun). Further fragmentation within those individual clusters then produced individual galaxies. In this scenario, virtually all galaxies that formed should be members of clusters and/or superclusters. To a large degree, this seems to be verified by current observations. However, this scenario also suggests that, because galaxies are forming via the process of fragmentation and collapse within a much larger cloud of gas, protogalaxies originally started out as objects that were about 10 times larger than galaxies today. Because protogalaxies are already clustered by this point, their environment fosters strong interactions between them. Moreover, because the disk component forms over a significantly longer time scale than the bulge component, these interactions do not favor the production of disk galaxies due to the tidal disruption of the gas destined to collapse into a disk. Hence, whereas this scenario does qualitatively predict the correlation between B/D and local density, it is not at all clear that any disk-dominated galaxies should have formed.

The alternative to the cooling and fragmentation picture is the idea of hierarchical clustering of small-mass units. In this picture, the first objects to form in the universe were 105-106 Msmsun. objects that gravitationally coalesced into larger-mass units (i.e., galaxies). The process of gravitational coalescence then continues with galaxies forming clusters of galaxies and then with clusters forming superclusters (and this process may still be continuing). Like the earlier scenario, the gravitational clustering idea also predicts that virtually all galaxies should be members of a cluster or a supercluster. However, in this picture, galaxies are allowed to originally form in isolation before becoming part of the cluster environment and this would tend to favor the production of disk galaxies. Moreover, this scenario also qualitatively predicts the correlation between B/D ratio and local density in the sense that bulge-dominated systems originally formed in denser areas than disk-dominated systems. Hence gravitational clustering around bulge-dominated galaxies will occur more quickly and disk-dominated systems will then infall at some later time into these newly formed clusters. This process of disk galaxy infall is something that is observed in the case of the Virgo cluster.


Although the two scenarios discussed previously can both give rise to galaxies that are distributed in a highly clustered manner, each makes different Predictions regarding the redshift at which galaxy and cluster formation should occur. In the cooling and fragmentation picture, it is possible that virialized clusters were present at redshifts as large as 2 (80% of the age of the universe). The best signature of a virialized cluster is the emission of x-rays from the hot intracluster medium. The strength of the x-ray emission is directly proportional to the mass of the cluster because the gas is heated by the cluster gravitational potential. Current x-ray surveys are only sensitive to x-ray emitting clusters out to a redshift of z approx 0.7 (55% of the age of the universe) and hence can place no stringent limits on the epoch of cluster virialization. ROSAT and future x-ray satellites such as AXAF may prove invaluable in this regard.

One of the main difficulties with the hierarchical clustering idea is the amount of time it takes for the process to begin. Specifically, gravitational amalgamation of subunits can only occur efficiently when the subunits themselves are reasonably close to one another. Because the universe is expanding then, at a given mass density, there will be an epoch past which this process can never get started because the average spacing between subunits is too great. At present, there is at least an order of magnitude range in our estimate of the mass density of the universe. Various theoretical arguments favor a mass density that actually closes the universe. Observations of large-scale deviations from Hubble flow, however, indicate that the mass density is between 10 and 20% of the closure density. In such a low-density universe, the process of hierarchical clustering must start at redshifts of 50-100 (0.1% of the age of the universe) and the physical details regarding the merger of subunits at such an early time remain obscure.


Of course, speculation regarding the formation of protogalaxies would cease upon their discovery. In fact, one can make a simple argument that suggests that protogalaxies might easily be detectable. A typical elliptical galaxy has 1011 Msmsun of stars. Characteristics of the stellar population in present-day ellipticals suggest that the bulk of this star formation occurred very early on, perhaps in the phase of protogalactic collapse. The collapse or free-fall time of any cloud is given by [sqrt(Grho)]-1, which is 4 x 108 yr for a typical elliptical with an initial radius of 50 kpc. (G is the gravitational constant and rho is the density.) In this case the predicted star formation rate would be approximately 250 Msmsun yr-1, which is a factor of 50 larger than the current star formation rate in nearby spiral galaxies. A star formation rate this high produces a great deal of luminosity in hot young stars as well as a supernova rate of about 1 per year! Most of this luminosity comes out at 1000 Å in the rest frame of the galaxy and so the redshift of this hypothesized protogalaxy would have to be 3 geq z geq 10 in order to be accessible to ground-based observations.

Before discussing possible reasons why this population of galaxies has yet to be detected, it is worth pointing out that protogalaxies with initial densities that are rather low will take a correspondingly longer time to form. In general, their formation process will be interrupted by the tidal shearing forces of neighboring galaxies. However, if these objects are in relative isolation, then some may be collapsing now, particularly in the case of flattened disk galaxies. Interestingly, a number of very low surface mass density disk galaxies have been discovered over the last two years. Although they are not protogalaxies by our adopted definition, their discovery in nevertheless important because these galaxies represent examples of galaxy formation that has been quiescent and that extended over a considerable period of time.

So the task of discussing protogalaxies now becomes one of suggesting reasons why they have escaped detection. We close with the following four possibilities. The most obvious possibility is that the first phase of vigorous star formation was essentially complete prior to z = 10 (which, coincidentally, is when the universe is approximately 1 galactic free-fall time old). However, in this case the radiation from the hot young stars is redshifted into the near-infrared part of the electromagnetic spectrum (e.g., 1-5 µm). Searches for this population with newly developed near-infrared imaging arrays are presently underway. Second, because the ultraviolet radiation from young stars is strongly absorbed by dust, it is possible that the early universe is not transparent to this radiation. This provides a rather effective screen that can potentially hide the process of galaxy formation from us. In this case, the absorbed ultraviolet photons should heat the dust to a temperature of 50-100 K. The radiation from this heated dust will be redshifted to the submillimeter portion of the electromagnetic spectrum. The energy density of these hypothetical sources is thought to be detectable with the SIRTF mission to be launched sometime around the year 2000. Third, it is well known that the peak in the redshift distribution of quasars is at z approx 3. Because quasars are now known to be in the nuclei of galaxies and the energy source is thought to be the infall of gas onto a massive black hole, it is possible that this infall process is facilitated by protogalactic collapse. In that case the radiation from the quasar effectively overwhelms that from the rest of the galaxy, imposing a sort of cosmic censorship against observing galaxies in the act of formation. Finally, our expectation that protogalaxies should be observable stems from a relatively naive physical argument. Perhaps, galaxy formation is a far more gentle and quiescent process than we have assumed and there simply is no ultraluminous phase which marks the birth of a galaxy. Future advances in telescopes and instrumentation will hopefully yield firm detections of protogalaxies, from which we can finally study their formation in detail.

Additional Reading
  1. Hensler, G. and Burkert, A. (1989). The initial conditions of protogalaxies - constraints from galactic evolution. In Progress Report on Cosmology and Gravitational Lensing, G. Borner, T. Buchert, and P. Schneider, eds. Max Planck Institute for Physics and Astrophysics, Garching, p. 203.
  2. Meier, D.L. and Sunyaev, R.A. (1979). Primeval galaxies. Scientific American 241 (No. 5) 130.
  3. Peebles, J. (1984) Origin of galaxies and clusters of galaxies. Science 224 1385.
  4. Rees, M.J. and Silk, J. (1970). The origin of galaxies. Scientific American 222 (No. 6) 26.
  5. Silk, J., Szalay, A.S., and Zeld'ovich, Y.B. (1983). The large-scale structure of the universe. Scientific American 249 (No. 4) 72.
  6. See also Cosmology, Galaxy Formation; Galaxies, Formation; Galaxies, High Redshift.