Next Contents Previous


To find a PG, one must have some idea of what one is searching for. Here we discuss a few of the (more or less) model independent properties of PGs, and the relation of these properties to search strategies.

2.1 Surface Density

Suppose that PGs are found only at redshift zPG, and have a lifetime in a ``recognizable PG phase'' of DeltatPG. Further suppose that the present-day descendants of PGs possess number density n0 at the present epoch. The number of PGs visible in a solid angle Deltaomega can then (neglecting mergers) be written as

equation1 (1)

where dA is angular diameter distance. Now, the luminosity density of the Universe (in B) is ~ 2.4 x 108 hLBsmsun Mpc-3 (Felten 1985), and the characteristic luminosity of galaxies is L*B appeq 1.6 x 1010 h-2 LBsmsun (very similar to the luminosity of the Milky Way), where h = H0/100 and H0 is in km s-1 Mpc-1. Hence a reasonable estimate for the local number density of L* galaxies is n0 appeq 0.015 h3 Mpc-3. The surface density of such galaxies can therefore be written as

equation2 (2)

for Omega0 = 1 and a redshift of galaxy formation zPG = 5. (Here h50 is the Hubble constant expressed in units of 50 km s-1 Mpc-1. The choice of zPG and DeltatPG in this formula will be justified below.) The numbers are even larger in an open Universe. Thus the progenitors of galaxies like the Milky Way should be very numerous, regardless of the exact choice of parameters in Eq. (2).

2.2 Spectrum

Meier (1976) first demonstrated that the flux Fnu of an object with constant star formation rate was roughly flat between the Lyman and Balmer breaks, with Fnu proportional to the star formation rate in this spectral region. This result is usually expressed as (e.g., Meier 1976, Cowie 1988, White 1989)

equation3 (3)

in the wavelength range 1000Å-3600Å, for a Salpeter (1955) initial mass function (IMF) (2). There is some dependence of Eq. (3) on burst age and IMF, but in all cases the variation is less than a factor of three in Fnu (White 1989). We adopt Eq. (3) for the purposes of this paper, with the implicit understanding that this is probably a conservative estimate of Fnu.

Figure 1 Figure 1. Spectral energy distributions for stellar systems with a constant star formation rate of 1 Msmsun yr-1 (from the models of Bruzual and Charlot 1993). The ages of these systems range from 4 x 107 yr (bottom) to 1010 yr (top). The vertical axis is in units of erg s-1 Hz-1. The prominent absorption feature near 1216Å is Lyalpha. Note the relatively flat energy distribution from 1000Å to 3600Å; also note that the flux level over this wavelength range does not depend strongly on age. The total flux in the wavelength range 1000Å-3600Å varies between 75% (4 x 107 yr) and 40% (1010) of the total light beyond the Lyman discontinuity.

Some spectral energy distributions (SEDs) of stellar populations with a constant star formation rate of 1 Msmsun yr-1 are shown in Fig. 1 (from the models of Bruzual and Charlot 1993). The main contributors to the light at lambda < 3600Å are upper main sequence stars; at longer wavelengths there is also a contribution from supergiants, asymptotic giant branch, and finally giant stars as age progresses. However, the contribution of these populations never completely dominates the total energy output (at least for galaxy ages less than ~ 1010 yr), provided that the star formation rate remains roughly constant. Ignoring light below the Lyman discontinuity, the fraction of the total luminosity in the wavelength range 912Å-3600Å varies from 75% at an age of 4 x 107 yr, to 60% at 109 yr, to 40% at 1010 yr. (This ignores dust - see below.)

A typical PG will contain many individual star formation regions, and so the overall star formation rate should be smoothly varying on time scales longer than the age of the most massive stars and shorter than the collapse time tcoll (~ 109 yr). Thus it seems reasonable to characterize a PG by some mean star formation rate, from which a rough estimate of < Fnu > follows. In the Baron and White (1987, hereafter BW87) galaxy formation scenario (see below), the star formation rate is in fact predicted to be roughly constant over nearly one collapse time.

Figure 2 Figure 2. The effect of 1 mag of V-band absorption on the spectral energy distribution of a 109 yr old stellar population with constant star formation rate. The solid line shows the SED with no absorption, whereas the dotted line includes absorption. The absorption law is the Galactic extinction model of Fitzpatrick (1986) for lambda < 3600Å, with modifications to match the absorption measurements of Nandy et al. (1975) at longer wavelengths (see also Seaton 1979). The basic unabsorbed model is from Bruzual and Charlot (1993).

One effect that might cause a real PG to deviate from the idealization of Eq. (3) is dust absorption. One magnitude of absorption in the V band is roughly equivalent to 4.4 mag of absorption at 1000Å (e.g., Seaton 1979). (This is illustrated in Fig. 2) The absorbed UV radiation is re-emitted thermally at wavelengths of order 100µm, as in starburst galaxies. The relevance of this to PG searches will be discussed further in Section 6.

The flux predicted by Eq. (3) is from stellar photospheres alone. Real PG spectra will also show strong emission lines due to excitation of gas by UV photons from hot stars. A simple calculation based on Kennicutt's (1983) calibration of Halpha strength vs. star formation rate, coupled with an Halpha to Hbeta intensity ratio from Case B recombination theory, and with Fnu from Eq. (3), yields a Lyalpha equivalent width of ~ 200Å (e.g., Djorgovski 1992). This implies that roughly 3-6% of the bolometric luminosity from a PG should be emitted in Lyalpha. More detailed calculations by Charlot and Fall (1993) yield rest frame equivalent widths in the range 100-200Å (depending on the power law exponent of the initial mass function), for a constant star formation rate; these results are in quite good agreement with the simple empirical argument above. The intensity of other lines, such as [O II] 3727Å, Hbeta 4861Å, [O III] 5007Å, and Halpha 6563Å, is discussed in Thompson, Djorgovski, and Beckwith (1994); these lines are generally weaker than Halpha, which itself is about 8-11 x weaker than Lyalpha (Ferland and Osterbrock 1985) in the absence of dust.

2.3 Luminosity

The principal source of energy from PGs is the release of nuclear binding energy from fusion reactions in stars. Energy release from AGN's may be comparable, but the connection of the AGN phenomenon with galaxy formation is unclear, and the lifetimes of individual AGN sources are uncertain. Other sources of energy (such as gravitational binding energy release from galaxy formation or star formation, or energy release from supernovae) are smaller by 1-2 orders of magnitude (Djorgovski and Weir 1990, Djorgovski 1992).

A crude estimate of the luminosity of a PG can be obtained as follows. A PG will convert a fraction Z approx - Delta X of its hydrogen into metals; based on the metallicities of old disk stars, |Delta X| approx 0.01. Approximately 1% of this mass is converted into photons by release of binding energy in nuclear reactions. The net bolometric luminosity of such a source is then

equation4 (4)

where MPG is here the baryonic mass of the PG. The apparent brightness of such a source at z = 5 (h50 = 1, Omega0 = 1) is about 2.8 x 10-15 erg cm-2 s-1, or, in the AB magnitude system, mAB approx 24.5.

3 Figure 3. The AB magnitude of objects with a constant star formation rate of 100 Msmsun yr-1, as a function of redshift. (In this magnitude system mAB = 0 corresponds to fnu = 3.63 x 10-20 erg cm-2 s-1 Hz-1; see Oke [1974].) A Hubble constant of 50 km s-1 Mpc-1 is assumed for this diagram; the dashed line is for Omega0 = 0.2, whereas the solid line corresponds to Omega0 = 1. This diagram is based on Eq. (4), and hence assumes that the wavelength of observation corresponds to a restframe wavelength between the Lyman and Balmer discontinuitues.

A somewhat more accurate approach to calculating the continuum brightness of PGs is simply to use the relation between star formation rate and Fnu in Eq. (3). The result of such a calculation is shown in Fig. 3 for two cosmologies and a star formation rate of 100 Msmsun yr-1 [implicit in Eq. (4) above]. It can be seen that the results of this calculation are in reasonable agreement with the simple arguments above.

Thus it appears that PGs may be visible in continuum light in the optical-near infared bands, but of course only provided that the wavelength of observation lies redward of the Lyman discontinuity - i.e., lambda gtapprox (1 + z) 912Å. For example, for observations in the I band, PGs may be visible provided that z ltapprox 8 (although the limiting magnitude of typical I band surveys probably precludes observation beyond z approx 4). The situation is similar for ground-based observations in the 1-2.4 µm window: the sky brightness in the K band limits observations to K < 23 or mAB < 25) (e.g., Gardner, Cowie, and Wainscoat 1993), so some PGs may be present in K band continuum surveys.

It is of interest to consider the total surface brightness of PGs on the sky. One could approximate this quantity by simply multiplying the characteristic surface density of PGs (Eq. (1)) by their brightness (from Eq. (4)). A more exact, more elegant, and physically more revealing approach is to consider the production of metals (Lilly and Cowie 1987, Cowie 1988). The basic point of this approach is that energy production and metal production are closely linked processes. From arguments similar to those leading to Eq. (4), it is straightforward to show that a production rate of metals of 1 Msmsun y-1 leads to a flux Fnu = 2 x 1029 erg s-1 Hz-1, with very little dependence on the choice of initial mass function. [Assuming a mean metal yield ~ 1% then leads to Eq. (1), with, however, some IMF sensitivity.] From this can be derived a surface brightness due to metal formation of

Equation 5 (5)

where rhoZ is the present day density of metals in the Universe. With the indicated choice of rhoZ (corresponding roughly to 1/3 solar metallicity - e.g., Cowie 1988), this works out to a surface brightness µAB approx 31 mag arcsec-2. Note that this is not in conflict with actual measurements of the background light; for example, Dube, Wickes, and Wilkinson (1979) derive an upper limit to the V band background nearly 100X brighter. [Other upper limits to background fluxes can be found in Fig. 2 of Cowie (1988) and Fig. 3 of De Jager, Stecker, and Salamon (1994).]

The limit from eqn. (5) (µAB approx 31 mag arcsec-2) can be compared with with background levels from known sources. From the number counts of Lilly, Cowie, and Gardner (1991), it can be seen that the surface brightness in galaxies (for a 1 mag bin) is µAB approx 30.9 mag arcsec-2 (at BAB = 25) and 30.0 mag arcsec-2 (at IAB = 24). (These levels do not depend strongly on magnitude because the logarithmic slope of galaxy counts is fairly close to 0.4.) It is therefore possible (although not proven) that a small fraction (< 10%) of galaxies in faint galaxy surveys could be PG's.

2.4 Size and Morphology

The visual appearance of a PG has been subject to a great deal of speculation. If the observable PG phase were the end product of the monolitic collapse of a massive cloud of gas (as proposed by, for example, Eggen, Lynden-Bell, and Sandage 1962; see also Larson 1974), then the bulk of its star formation might occur in a small region kiloparsecs in size (where 1 kpc appeq 0".2-0".4, depending on z and cosmology, although not with great sensitivity); such an object would be close to being unresolved with a ground-based telescope. An extreme example of such a model, in which most star formation occurs in a very small region, is model A of Lin and Murray (1992).

On the other hand, some models of galaxy formation (e.g., BW87) predict that PGs will be lumpy in structure and extended in size, primarily due to the fact that in these models galaxy formation is a hierachical process. The BW87 models predict PGs that are ~ 5-10" in size, with individual knots that may be bright enough to be visible. An extreme model would be the case in which there are a very large number of star formation sites spread throughout a large, more or less uncollapsed halo; the individual star forming subunits would not be visible as discrete entities. Such an object would have very low surface brightness (~ 30 mag arcsec-2), a size of perhaps 10-20", and would be extremely difficult to detect using conventional techniques. (Nevertheless, it seems plausible that at least some of the gas in such an extended configuration would find its way into a nuclear starburst by dissipation.)

2.5 Redshift of Galaxy Formation

The redshift of galaxy formation, zGF, is of great importance in deciding on a PG search strategy, since zGF determines both how bright a PG will be, and in what wavelength range a search should optimally be conducted.

In this paper we will concentrate, as do many other authors (BW87, White 1989), on the formation of spheroidal systems. There is in fact very good evidence that the formation of disks is more quiescent than that of spheroids (e.g., Gunn 1982), and that disks were never much brighter than they are today (3). For example, Larson (1992) summarizes evidence that the mean SFR in the disks of nearby spirals (including the Milky Way - e.g., discussion in Majewski 1993) was never much more than twice the present value. Recent estimates of the gas consumption timescales in disks yield mean values ~ 7 Gyr (cf. Kennicutt 1983, Sandage 1986, Donas et al. 1987; see also discussion in Larson 1992). This is not too different from the ages of disks, suggesting that, with a modest amount of infall, the mean SFR need not have been too different from the present value. This is all supported by the observations of Cowie, Songaila, and Hu (1993), who present direct evidence of the gradual growth of disks. From their observations it appears that disks may have grown by a factor of ~ 2-4X in mass since z appeq1.

There is also some evidence that the bulk of disk formation followed well after spheroid formation for the Milky Way (e.g., review by Norris 1989). The most direct information on this point comes from the work of Winget et al. (1987) on the low luminosity cutoff in the white dwarf luminosity function. From this work it appears that the age of the disk is most likely appeq 10 Gyr, which is significantly younger than the 13-17 Gyr ages associated with (most) globular clusters (cf. Bergbusch and VandenBerg 1992; Lee, Demarque, and Zinn 1990).

If the age of the Universe T0 > 14 Gyr and the age of the disk is appeq 10 Gyr, then this suggests that the bulk of disk formation began in earnest at z ltapprox 1. Now, if most disk formation took place in a short burst of activity at z ltapprox 1, then the objects involved should stand out in existing faint object redshift surveys (e.g., Lilly 1993, Colless et al. 1990, 1993), since they would have mAB ltapprox 22-23. The fact that they do not appear in these surveys strongly suggests that these disks formed in a quiescent fashion over a much longer interval of time.

Thus in what follows we will consider the epoch of galaxy formation to be the time at which the spheroid forms. In this picture the buildup of the disk might be thought of as a manifestation of galaxy evolution rather than galaxy formation.

2.5.1 Some Simple Model Independent Arguments

There are several arguments for the redshift of galaxy formation that are more or less independent of an assumed model for galaxy formation (Peebles 1989; see also Peebles 1993, Section 25). Most provide upper limits to the redshift of galaxy formation.

(1) Omega argument. For a Universe with Omega0 < 1, growth of structure begins to freeze out at a redshift marking the transition between Einstein-de Sitter expansion and linear expansion - i.e. at

Equation 6 (6)

(e.g., Peebles 1980, Section 11). Thus at least some galaxy formation must occur at quite large redshift in a low Omega Universe, although infall continues right up to the present epoch in such a cosmological model.

(2) Overlap argument. Consider the precursors of galaxies as spheres. At what redshift were these spheres just touching? The separation of galaxies at this epoch was ~ 2 fc R, where R is the radius of the halo of the Milky Way (~ 100 kpc), and fc is the amount by which the proto-halo collapsed before forming stars. (Dissipationless collapse provides a lower limit on fc of 2.) The present-day separation of L* galaxies is ~ n0-1/3, where n0 is the present-day number density ofL* galaxies (e.g., Section 2.1). Then

Equation 7 (7)

Clearly this is a reasonable upper limit to the redshift of galaxy formation for L* galaxies.

(3) Density argument. An interesting constraint on zGF for L* galaxies is obtained by finding the redshift at which the mean density of galaxies at the present epoch was just equal to the density of the Universe. This is an upper limit on zGF for any reasonable model of galaxy formation. For a galaxy with total (i.e., dark plus baryonic) mass 1012 Msmsun, with total radius 100 kpc, the mean density is ~ 2 x 1026 g cm-3, or a factor of f3c smaller before any collapse that took place. The matter density of the Universe is

Equation 8 (8)

from which we find

Equation 9 (9)

One could repeat the above considering baryonic matter only (M approx 1011 Msmsun within R appeq 10 kpc for the Milky Way). From nucleosynthesis constraints (Walker et al. 1991) Omegab < 0.1, and fc gtapprox 10 by noting that the initial radius of the perturbation must have been gtapprox 100 kpc. The result is very similar: zGF ltapprox 10.

(4) Collapse argument. Perhaps the most physically insightful of arguments is derived from consideration of an expanding perturbation in the early Universe. It is straightforward to show that the collapse time (time for a perturbation to expand with the Hubble flow, turn around, and collapse completely) is just

Equation 10 (10)

where Rmax is the radius of the perturbation at turnaround (Rmax = fc R gtapprox 200 kpc for the Milky Way). To convert to a redshift requires a choice of cosmology that gives a reasonable T0: for example, if T0 = 13 Gyr, then zcoll lies between 1.8 (h50 = 1, Omega0 = 1) and 3.5 (h50 = 1.5, Omega0 = 0) for galaxies like the Milky Way.

2.5.2 The Empirical Evidence

There is a strong body of empirical evidence that can be marshalled to place constraints on zGF, and also on the nature of PGs.

(1) Redshift surveys. A number of redshift surveys at faint magnitudes have been conducted over the past few years (e.g., Broadhurst, Ellis and Shanks 1988; Colless et al. 1990, 1993; Lilly 1993; Tresse et al. 1993). From these surveys, it is clear that widespread galaxy formation activity is not occurring at z ltapprox 1; otherwise PGs, given their brightness at z approx 1 (e.g., Fig. 3) and their surface density [Eq. (2)], would appear prominently in current redshift surveys. Put another way, the Universe at z approx 1 is not too dissimilar to the Universe at the present epoch (Peebles 1989).

That is not to say that some galaxy formation activity is not still underway at the present epoch (4). Individual low luminosity objects undergoing an intense burst of star formation are seen nearby (e.g., H II region galaxies - Terlevich 1988; blue compact dwarf galaxies - Thuan 1987). A debate continues as to whether these objects really are undergoing their first burst of star formation, or whether there is an underlying old population (Thuan 1987; Keel 1988; Meurer, Mackie, and Carignan 1994). Nevertheless, these objects comprise a tiny fraction of all nearby galaxies. It should also be noted that a significant fraction of the metals in present-day galaxies appears to be produced by the faint blue galaxy population at z < 1 (Cowie, Lilly, Gardner, and McLean 1988), although if these galaxies are low mass then their metal-enriched gas may be expelled. But these objects are not, we believe, the long sought after primeval galaxies; they represent, rather, relatively recent galaxy evolution, possibly driven by mergers, of a low mass galaxy population (e.g., Cowie, Songaila, and Hu 1993).

(2) Cluster ellipticals. The small scatter in the colors of cluster ellipticals in the Virgo and Coma clusters provides an interesting constraint on the redshift of formation of ellipticals (Bower, Lucey, and Ellis 1992). If galaxy ages differ typically by an amount of order their collapse time (i.e., Delta TGF approx TGF, where TGF corresponds to zGF), then zGF gtapprox 2 for luminous ellipticals. This lower limit on zGF could be larger if any of the scatter in colors originated in intermediate redshift starbursts.

(3) Milky Way globular clusters. The ages of globular clusters in the Milky Way are now know to span a range of at least ~ 3 Gyr (cf. VandenBerg, Bolte, and Stetson 1990; Sarajedini and Demarque 1990). A reasonable assumption is that globular cluster formation starts no earlier than the epoch of maximum expansion of a perturbation (tcoll / 2). For a collapse time of gtapprox 109 yr, the midpoint of the PG phase will then lie gtapprox 2 Gyr after the big bang, so that zGF approx 2.5-4 (0.2 leq Omega0 leq 1, 13 leq T0 leq 15 Gyr). In this picture the end of the formation phase would be 3.5 Gyr after the big bang, or at redshifts 1.5-2.2.

The above argument is probably one of the more powerful empirical arguments available on the epoch of galaxy formation, but it is not without problems. It is unclear whether the formation of globular clusters coincides with the formation of the halo/spheroid of the Milky Way, or, for that matter, whether there is a similar spread in ages found in field halo stars that would indicate a prolonged period of halo formation, independently of the globular clusters.

4 Figure 4. The fraction of the critical density in damped Lyalpha absorbers, Omegagas, as a function of redshift (from Lanzetta 1993, Lanzetta et al. 1994). The dashed line indicates the present-day density of luminous material (stars) in galaxies. It can be seen that most galaxy formation has taken place since z = 3.5. This diagram is for H0 = 100 km s-1 Mpc-1 and Omega0 = 1; both the data points and dashed line scale as h-1.

(4) Damped Lyalpha absorbers. Lanzetta and collaborators (Lanzetta 1993; Lanzetta, Wolfe, and Turnshek 1994) have studied the amount of gas in damped Lyalpha absorbers over a range of redshifts. Figure 4 plots their results for Omegagas, the density in damped Lyalpha absorbers divided by the critical density, as a function of redshift; the present-day density of luminous material in galaxies and disks is shown for comparison. It can be seen that Omegagas is very similar to the value of Omega computed for stars in galaxies at the present epoch. This result suggests that at least some damped Lyalpha systems originate in proto-spheroids; this is supported both by the short time scale of gas consumption in Fig. 4 (ltapprox 1 Gyr), and also by the low mean metallicity of these clouds (see Lanzetta 1993 and Lanzetta et al. 1994 for further details).

If the damped Lyalpha absorbers really are the precursors of all modern-day galaxies (5), as is suggested by Fig. 4, then it follows that much of the stellar mass in galaxies formed between z = 3.5 and 1.5, and that perhaps half of these stars formed between z = 2-3.

(5) High redshift radio galaxies. The SEDs of luminous, high redshift radio galaxies in principle offer an opportunity to determine their redshift of formation. There seem to be two distinct interpretations of the spectra of these radio galaxies (e.g., McCarthy 1993a); (i) they form at high redshift (zGF > 5), but are viewed at an epoch at which they are undergoing a small (in fractional mass) burst of star formation (Lilly 1988); or (ii) they are viewed shortly after their formation epoch (Chambers and Charlot 1990), a view that leads to estimates of zGF more like 3-4. The remarkable tightness of the K-band Hubble diagram (Lilly 1989; McCarthy 1993a, Section 6.5) seems to argue for a relatively high zGF 5; but this assumes no evolution in the masses of radio galaxies with redshift (see also discussion in Eales et al. 1993).

The most recent work on radio galaxies at high redshift indicates that the high z radio galaxies may well be a ``mixed bag'' in terms of origin. Very blue objects such as 0902+34 (Eisenhardt and Dickinson 1992) and the bluest objects in the survey of McCarthy 1993b) may have formation redshifts as small as 3-4 (see also Eales and Rawlings 1993). On the other hand, the reddest objects observed by McCarthy seem to require much larger formation redshifts. Nevertheless, these objects could have spurious colors due to non-stellar light or emission lines (Eales and Rawlings 1993). In addition, it appears that the colors of these objects are not, in any case, well-matched by models (e.g., Fig. 2 of McCarthy 1993b). Given all of the above, plus the fact that radio galaxies are not typical of the general population of galaxies, we view constraints on galaxy formation redshift from these objects with suspicion.

(6) Cosmic microwave background radiation. At a given epoch, small amplitude perturbations will take longer to grow into galaxies than large amplitude perturbations. Hence it follows that perturbations at z approx 1000, and hence fluctuations in the cosmic microwave background (CMB), must be larger if zGF is increased. In order not to violate the observed amplitude of CMB fluctuations, zGF < 3-4 for a CDM model (Kashlinsky 1993, Nusser and Silk 1993).

(7) Quasars. There are several ways of using quasars as indicators of the redshift of galaxy formation. The first is to note that the peak in the space density of quasars lies around z = 2-3 (e.g., Crampton, Cowley, and Hartwick 1987; Schmidt, Schneider, and Gunn 1991). If quasars are fueled by mergers (Carlberg 1990), and if mergers play an important role in galaxy formation (e.g., White and Frenk 1991), then it seems reasonable to suppose that the peak of the ``galaxy-building'' epoch lies at or around z = 2. Note however that this argument associates no particular importance to quasars in relation to galaxy formation, other than as an indicator of the merger rate (but see Djorgovski 1994).

2.5.3 Redshift of Galaxy Formation - A Summary

As noted above, the two most interesting pieces of empirical evidence constraining the epoch of galaxy formation are (i) the age spread of Milky Way globular clusters, and (ii) a direct estimate of gas depletion (presumably due to star formation) from observations of damped Lyalpha absorbers towards QSOs. These two approaches yield characteristic epochs for the first burst of star formation for most galaxies around z approx 2-3. Somewhat remarkably, this is in reasonable agreement with the predictions of the model independent ``collapse argument'' above (for a collapse factor of 2), and with the existence of various ``primeval galaxy candidates'' discussed below. Furthermore, virtually all theories of galaxy formation in a CDM-dominated Universe predict galaxy formation at relatively low redshift (e.g., BW87, Carlberg 1988).

This moderately self-consistent picture is, however, disturbed by two recent papers by Turner (1991a, b), who convincingly demonstrates that the properties of quasars at high redshift demand at least some galaxy formation activity at z gtapprox 5. For example, the very existence of quasars at z = 5 implies structure formation and collapse at much higher redshift. The fueling of QSO's at high redshift requires the existence of high overdensity perturbations in the accretion flow; otherwise the rate at which fuel can be supplied is too low. The high abundance of heavy elements in quasars at z gtapprox 4 requires at least one (and possibly 2 for nitrogen) generations of stars to preceed the quasar, and hence a redshift of galaxy formation certainly greater than 5 (see also Hamann and Ferland 1993).

A related argument is that the presence of heavy element absorption lines in the spectra of quasars implies that significant star formation and chemical enrichment has occurred along random lines of sight in the Universe significantly earlier than z appeq 4. Finally, Hu and Ridgway (1994) have reported on the discovery of two extremely red [(I - K) > 6] galaxies. The simplest explanation of the SEDs of these objects is that they are ellipticals with an age gtapprox 3 Gyr at a redshift z appeq 2.4. This requires (conservatively) a formation redshift > 6.

It seems most reasonable to conclude, then, that, even if the bulk of galaxy formation activity takes place at z ltapprox 3.5, there must still have been some galaxy formation activity at higher redshifts. In fact this is not at all unreasonable, since the spectrum of perturbations in the early Universe encompasses a wide range of amplitudes and mass scales.

2.6 Primeval Galaxy Candidates

As was discussed in Section 1, the principal focus of this review is on the detection of a widespread, high surface density population of PG's that should exist in all directions in the sky. Nevertheless, it is of interest to briefly mention some objects or classes of objects which have been proposed as candidate PG's.

We have already mentioned (Section 2.5) the possibility that QSO's or powerful radio galaxies might have been PG's at some stage of their evolution. There is also evidence that the somewhat weaker (~ 0.1 Jy) radio source 53W002 may be close to its initial star burst. This object was discovered in the Leiden-Berkeley Deep Survey, and was found to be narrow-line galaxy with redshift z = 2.39 (Windhorst et al. 1991). It appears to have started forming stars only ~ 0.5 Gyr earlier than its epoch of observation, placing the redshift of galaxy formation zGF appeq 2.7-4.2, depending on cosmology (Windhorst, Mathis, and Keel 1992).

A class of superluminous, IR-bright galaxies has been identified that may be related to the PG phenomenon. The best-known member of this class is the z = 2.286 IRAS galaxy F10214+4724, which, with a bolometric luminosity of > 1014 Lsmsun, is one of the most powerful known sources in the Universe (Rowan-Robinson et al. 1991). Two other objects, though not as extreme, are now known to resemble F10214+4724: P09104+4109 at z = 0.44 (Kleinmann and Keel 1987), and F15307+3252 at z = 0.93 (Cutri et al. 1994). The z = 3.8 radio source 4C41.17 (Chambers, Miley, and van Breugel 1990) is somewhat similar to the extent that it is extremely luminous (L > 1013 Lsmsun), appears to be undergoing an intense burst of star formation, and contains an enormous quantity of dust (Dunlop et al. 1994) - comparable to that observed in F10214+4724. All of these sources can be interpreted as being powered either by starbursts (Brown and Vanden Bout 1991; Solomon, Downes, and Radford 1992; Rowan-Robinson et al. 1993), or by AGNs (Hines and Wills 1993; Elston et al. 1994). Regardless of the dominant source of luminosity, it seems plausible to suppose that these objects are massive elliptical galaxies in a very early stage of evolution (Kormendy and Saunders 1992); the observations of Elston et al. (1994) do not rule out substantial star formation, and are not in conflict with this interpretation. However, it is clear that not all PG's could be like these objects; otherwise a huge sub-mm background would result.

Damped Lyalpha absorbers may also be related to PG's, as discussed in the previous section. However, direct evidence of star formation in these objects is lacking, with the exception of ~ 3 absorbers that possess emission lines (Elston et al. 1991, Wolfe et al. 1992, Møller and Warren 1993). However, several Lyalpha sources have been discovered that are clustered around damped Lyalpha absorbers (e.g., Lowenthal et al. 1991; Macchetto et al. 1993; Møller and Warren 1993; see also Wolfe 1993). The Macchetto et al. object is a radio-quiet galaxy at z = 3.428 with strong Lyalpha (rest frame equivalent width gtapprox160Å), and an inferred star formation rate of 18 Msmsun yr-1. Both this object and the Lowenthal et al. object, which it resembles, are candidates for PG's.

2 The models of Bruzual and Charlot (1993) give about 2-3 times more flux than predicted by Eq. (3), most likely because the upper main sequence models that they used included mild convective overshooting, which lengthens main sequence lifetimes (S. Charlot, private communication). Back.

3 A clear counterexample, however, is the Large Margellanic Cloud. Back.

4 Nevertheless, galaxy formation activity at z = 0 must be nearly finished, because most (> 90%) of the baryonic material locally appears not to be in the form of gas. Back.

5 Songaila et al. (1994) have recently determined the deuterium to hydrogen ratio in a z = 3.3 cloud towards a QSO. Their result implies a primordial Omegab h2 = 0.005, where Omegab is Omega in baryons; if this result is confirmed, then the damped Lyalpha absorbers must be the precursors of galaxies, because Omegab (damped Lyalpha at z = 3.5) approx Omegab (galaxies at z = 0) approx Omegab (primordial). Back.

Next Contents Previous