4.1
Some Broadband Searches
The first PG searches were motivated by the models of
Partridge and
Peebles (1967), which suggested that PGs would undergo their first
burst of star formation near the maximum expansion phase of a
perturbation (z 
10). Thus PGs were predicted to be large
(~ 10"), red objects. Such sources were searched for by
Partridge
(1974); his 7 candidate PGs all appear to be normal galaxies at
z 
0.3 (Koo
1986). Davis
and Wilkinson (1974) developed an
innovative large aperture (~ 1000 arcsec2) photometer to search
for fluctuations in the background that might have been due to a
high surface density population of diffuse sources. No evidence was
found for such a population, down to a flux level of 0.5% of the red
(
7500Å) night sky brightness
for a 100 arcsec2
source. Comparable flux limits were achieved in a search for
fluctuations in the infrared K band ( 2.2
µm) by
Boughn et
al. (1986); this survey is noteworthy insofar as it extended
the search for PGs to much higher redshifts (possibly 10).
Further constraints on PG's and high redshift galaxies using broadband surveys may be found in Koo (1986), Tyson (1988), and Guhathakurta, Tyson, and Majewski (1990); see also Smith, Thompson, and Djorgovski (1993).
4.2
Quasars and PG Searches
Models of galaxy formation that included dissipation (Larson 1974) completely changed the search strategy for PGs. PGs were now predicted to be quite compact, and the question naturally arose as to whether there were any PGs lurking in surveys of quasars (e.g., review by Koo 1986; Koo and Kron 1988). The answer appears to be no: for example, none of the ~ 250 QSO candidates in the CFHT survey (Crampton, Cowley, and Hartwick 1990) have spectra which could be placed in the PG category (Hartwick, private communication).
The question nevertheless remains as to whether QSO's themselves could be PGs. This conjecture has a long history that has been reviewed by Koo (1986 and references therein; Djorgovski 1994). Recently Terlevich and collaborators have revived this idea (e.g., Terlevich and Boyle 1993); they propose that quasars could be the luminous, star forming cores of very young elliptical galaxies undergoing their first burst of star formation (but see Heckman 1991, Filippenko 1992). Perhaps a more plausible scenario is one in which the formation of a luminous AGN (powered by conventional means - e.g., infall onto a massive collapsed object) is both associated with and contemporaneous with galaxy formation (Djorgovski 1994): in this case some (though not necessarily all) quasars may be the beacons that signal galaxy formation.
4.3
Redshift Surveys
Deep redshift surveys are generically related to broadband searches for
PGs, because the objects for which redshifts are obtained are first
identified in broadband images. A number of such redshift surveys have
been completed (e.g.,
Broadhurst, Ellis
and Shanks 1988;
Colless et al. 1990;
Lilly et al. 1991;
Colless et al. 1993;
Lilly 1993;
Tresse et
al. 1993), and several others are nearing completion. The limiting
magnitudes for these surveys are typically B 
22.5
(Colless et
al. 1990,
1993), or
Ilim 22.5
(Lilly 1993)
or 22.1
(Tresse et
al. 1993). Many thousands of objects now have redshifts, and
few if any have characteristics that could ambiguously lead to
classification as a PG.
4.4
Searches for Redshifted 21 cm Radiation from
Neutral Hydrogen
Gunn and Peterson
(1965) first noted that a diffuse neutral
intergalactic medium would produce an absorption trough at wavelengths
shorter than Ly
in the spectra
of high-redshift quasars. The fact
that no such trough is seen (e.g.,
Steidel and
Sargent 1987) provides an
extraordinarily sensitive limit to the cosmological mass density in a
diffuse neutral IGM, 
H
I < 10-5: either the IGM is
ionized (e.g.,
Miralda-Escudé and Ostriker 1990), or it is strongly
clumped.
In the latter case it is in principle possible to search for redshifted
21 cm radiation from the expected structures
(Hogan and Rees
1979).
This topic is nevertheless somewhat orthogonal to the rest of this
review, for several reasons. First, a typical (~ mJy) sensitivity
limit in redshifted 21 cm corresponds to a neutral hydrogen mass ~
1014 M
at z 3 - i.e., a protocluster rather than a
protogalaxy. Second, objects observed in redshifted 21 cm are likely to
be in an evolutionary stage preceeding the epoch at which stars start
to form (in contrast to the other search strategies that we discuss).
Finally, and perhaps most important, these searches implicitly assume
as their starting point that massive objects have collapsed at early
epochs - i.e., that structure forms in a ``top-down'' hierarchy, as
proposed by, for example, Sunyaev and Zel'dovich (1975). In contrast,
we have for the most part assumed in this review a ``bottom-up''
hierarchy of structure formation - a view that is supported by many
lines of reasoning (e.g., Ostriker 1993). Nevertheless, a discussion of such
objects is clearly relevant to any discussion of galaxy formation.
Searches for redshifted 21 cm radiation from H I in protoclusters have
been reported by a number of groups
(Davies, Pedlar,
and Mirabel 1978;
Bebbington 1986;
Hardy and Noreau
1987;
Noreau and Hardy
1988), at
redshifts ranging from 3.3 to 8.4. Perhaps the most sensitive survey at
the low end of this redshift range is that of
Uson, Bagri and
Cornwell (1991a), who reach a detection limit of around 1014
M in
neutral hydrogen. The null results that these authors report are not
surprising if structure forms in a bottom-up manner - i.e., as expected
in Universes dominated by CDM (but see Uson et
al. 1991b).
4.5
Searches for UV-Optical Emission Lines
A PG can be thought of as a giant H II region that has been
photoionized by massive stars from the first burst of star formation.
Hence it is expected that PGs should possess a rich emission line
spectrum; even in the earliest (very low metallicity) phases of PG
evolution one might expect strong Balmer and Lyman emission lines. One
of the most intensively applied methods for detecting PGs is in fact
searching for redshifted Ly
1216Å
(Meier 1976;
Hogan and Rees
1979), which we discuss extensively in this section of the
paper. Other emission lines have been proposed for PG searches in the
near-infrared (1-2.5µm); these include [O II] 3727Å
(Thompson et
al. 1994), and H
(Pritchet and
Hartwick 1994), which is less
affected by dust absorption than Ly.
There are three principal search techniques for faint emission line objects, illustrated in Figure 6. The performance of these techniques is mainly distinguished by volume surveyed and by limiting brightness.
Figure 6. Search strategies for emission line objects. (a) Slitless spectroscopy; (b) long-slit spectroscopy; (c) narrow-band imaging. Although the volume surveyed by slitless spectroscopy is largest of the three techniques, the limiting flux is brighter because each object is superimposed on a night-sky background from all wavelengths passed by the optics.
(1) Slitless spectroscopy [e.g., Fig. 6(a)] allows the sampling of a very large volume of space, but results in a relatively bright limiting threshold, because the sky brightness from all wavelengths passed by the bandpass of the system is superimposed on any spectral feature. This technique is commonly used in QSO grism/grens surveys (e.g., Crampton et al. 1987, 1990), and two different PG searches using this technique have been reported by Koo and Kron (1980). These latter searches reach limiting fluxes ~ 10-14-10-15 erg cm-2 s-1 and cover solid angles ~ 1 deg2.
(2) Long-slit spectroscopy [Fig. 6(b)] generally
results in
the best limiting flux for detecting (unresolved) emission lines,
because the spectral resolution, and hence sky background per
resolution element, is lowest. However, the volume surveyed is usually
much smaller than that achieved with other techniques. The principal
long-slit surveys are those of Cowie and Hu (see Cowie 1988),
Lowenthal et
al. (1990), and Djorgovski and collaborators
(Thompson,
Djorgovski, and Trauger 1992;
Djorgovski,
Thompson, and Smith 1993). Typical
limiting fluxes are 10-16-10-17 erg
cm-2 s-1, and
areas surveyed are
10-4 deg2. Anecdotal evidence
suggests that many other groups involved in faint object spectroscopy
have searched for serendipitous emission-line objects that happen to
lie along their slits, without success. Pritchet and Hartwick
(unpublished) have detected several such objects in long-slit
spectroscopy with the CTIO 4 meter telescope and RC spectrograph, but
they are almost certainly low luminosity [O II] 3727Å emitters -
i.e. ``normal'' intermediate redshift galaxies.
(3) Narrow-band imaging [Fig. 6(c)] imposes a
restriction on
z, the range of
redshifts that is surveyed, and so results
in a smaller surveyed volume. Most of this work consists of
optical searches for Ly at
redshifts of order 2-6 using CCDs and
narrowband interference filters
(e.g., Cowie
1988;
Pritchet and
Hartwick 1987,
1990;
Rhee, Webb, and
Katgert 1989;
Smith et al. 1989;
Wolfe et al. 1992;
Djorgovski and
Thompson 1992;
Djorgovski et
al. 1993;
De Propris et
al. 1993;
Møller and
Warren 1993;
Macchetto et
al. 1993),
or with a Fabry-Perot system
(Djorgovski et
al. 1993;
Thompson et
al. 1992).
The limiting flux for these surveys is much improved over slitless
spectroscopy - typical flux limits are between 10-16
and 10-17 erg cm-2 s-1, and typical CCD fields
now approach 0.01-0.1 deg2 (or larger for CCD mosaics)
on 4 meter class telescopes.
More recently, narrow-band imaging surveys have been extended to the
infrared. Pritchet and Hartwick (1994) have completed a CCD
survey for
redshifted Ly at z 7 (~ 9100Å), and have achieved a
remarkably faint limiting flux of ~ 10-17 erg cm-2
s-1 by employing narrow band filters designed to avoid the
strongest OH emission lines (which are the dominant source of sky
brightness in the region 7000Å-2.3µm). Using infrared arrays,
Parkes, Collins,
and Joseph (1994) have pushed the search for
redshifted Ly out to z
9, using a similar ``low OH''
technique in the infrared J band; furthermore, Thompson et
al. (1994)
have searched for redshifted [O II] 3727Å emission from PGs
associated with QSO's at z > 4. These latter two infrared array
surveys reach limiting fluxes of order 10-15 erg cm-2
s-1, over relatively small solid angles of
10-4-10-3
deg2 (the small sizes of the surveys being dictated by the small
areas of infrared arrays, a limitation that is expected to improve in the
future).
The result of all of these surveys is identical: no emission line PGs have been found.
Figure 7 shows the constraints on surface
density and limiting flux
for observations of redshifted Ly. (Similar diagrams have been
presented by a number of authors - e.g.,
Koo 1986,
Pritchet and
Hartwick 1990,
Djorgovski and
Thompson 1992.) We have assumed that the upper
limit on surface density is one object per sampled solid angle. The
plotted points should be interpreted as excluding the region to the
upper left of each point (higher fluxes and higher surface
densities). The stippled area roughly combines the limits from
different groups; PGs should possess lower fluxes and/or surface
densities than the shaded area, and the integral PG luminosity function
should not pass through this shaded region. (However it should be noted
that the survey limits plotted in Fig. 7 are
independent, and
hence the combined limit of all the surveys should be even stronger
than the shaded area.)
Figure 7. Limits that have been placed on the surface density and apparent Ly
The limits plotted in Fig. 7 are illustrative
only. Different
groups use different criteria for defining their limiting fluxes (in
terms of both the number of noise 
for a real detection, and
also the size of the image), and it is difficult to
correct the published information to a uniform system.
The solid line in Fig. 7 represents a rough model
computed as in Fig. 5. Insofar as
the model cuts through the
region excluded by the limits, this model appears to be in disagreement
with the observations. However, this model has been computed for
z = 3,
z = 0.1, whereas the
observations cover a range of redshifts (z
2-5) and redshift intervals.
To rectify this problem, we
convert the observed surface density and flux limits into
corresponding volume density and absolute luminosity limits
(Fig. 8). In this diagram the limits can be
compared with one model
independently of zGF; the model in this figure is that
described in Section 3 and used (in
observational form) in Fig. 5. The
meaning of the reference numbers is the same as in
Fig. 7.
Figure 8. Limits on PG comoving densities and absolute Ly
Again it can be seen that primeval galaxies as predicted by our simple dust-free model should have been easily detected if they were unresolved. The discrepancy between model and observations exceeds a factor of 100X in surface density, or 10X in luminosity, if galaxies formed in the redshift interval 2-5. The limits at higher redshift are not as strong at the present time.