6.2.2. Deep Galaxy Counts: The Mystery of the Faint Blue Galaxies
The second method of estimating the space density of galaxies can be
done by doing deep galaxy counts and plotting the surface density
of galaxies as a function of their apparent flux. In principle, if
all galaxies had the same luminosity as a function of redshift
(or you had an independent way
of picking galaxies of the same luminosity, say through some kind
of filter selection) this diagram is very sensitive to the geometry
of the universe and hence 
. Realistically, the stellar
populations of galaxies do evolve and hence one expects moderate to
strong luminosity evolution for galaxies.
The history of using deep galaxy counts as a cosmological probe is
that of inconsistent, ambiguous and confusing results. The naive
expectation is that 1) faint galaxies should, in general be red (due
to the effects of redshift) and 2) have number counts that decline
significantly more rapidly than (1 + z)3 since the
volume per unit area
decreases with distance in cosmologies based on the Robertson-Walker
metric. When the first deep counts came in, based on long exposure
4-m photographic plates, it quickly became apparent that neither
of these two effects was seen in the data (e.g., Koo and Kron 1982).
Most noticeable was a significant population of rather faint but
fairly blue galaxies. These galaxies are known as Faint Blue
Galaxies (FBGs). Several other surveys verified the existence of
the FBG population. This led to a widely accepted model in which
these galaxies were experiencing a phase of significantly enhanced
star formation. The high number density of FBGs could best be explained
if these galaxies were located at redshifts z = 1-3 and thus a very
large volume was being sampled.
However, by the late 80's the advent
of multi-object fiber spectroscopy meant that a deep field containing
several FBGs could be exposed for several hours, thus returning spectra
of sufficient quality for redshift measurements. In general, these
redshift surveys showed that the FBGs were primarily a low redshift
population (z 
0.7)
(Broadhurst et al. 1988, Colless et al. 1993, Glazebrook et al. 1995).
In general, blue galaxies imply significant star fromation. In the
basic CDM scenario, relatively late formation of galaxies is possible.
Hence, the FBGs may represent galaxies undergoing their initial
bursts of star formation, several billion years after the Universe
was formed. The exact conditions that would cause this delayed star
formation are not well understood. One possibility is that the gas
in these potentials was ionized early on and took a few billion years
to cool. However, in this scenario would has to 1) identify the sources
of ionization and 2) explain why only this population was effected.
The latter can be partially explained if the FBGs are relatively low mass
and low density galaxies.
Still, high-redshift FBGs have been detected. Cowie and Lilly (1988) were
among the first to identify a high redshift (z
2.5) population
of FBGs. This high redshift population has been confirmed by
a number of others, most recently Steidel et al. (1996) who present
indisputable evidence that star forming galaxies exist out to at
least z 
3.5. These
observations demonstrate that FBGs can
be found over a wide redshift range and hence are a very heterogeneous
population.
The nature of the FBGs seen in the deep count data, therefore continues to be elusive. If the FBGs however, are a significant population at moderate redshifts then they must somehow disappear by z = 0. This indicates either strong luminosity or strong density evolution of the GLF. In particular if galaxy merging is occurring as a result of galaxy-galaxy interactions, then galaxy number is not conserved and number density of galaxies increases with redshift. This is a serious complication for structure formation theories that attempt to predict the GLF since it implies that the normalization of the GLF changes with redshift. Patton et al. (1996) derive a merging rate, for a sample of intermediate redshift galaxies, of (1 + z)2.9 ± 0.9. This exponent is close to the expected (1 + z)3 volume evolution, although the observed error bar is too large to confirm this. The other possibility of strong luminosity evolution suggests that the FBGs have a star formation history that allows them to rapidly fade so that by z = 0 they are extremely LSB and faint galaxies. This latter prediction is also testable to some extent (see below).
In the last 4 years, advances in detector technology have paved the
way for performing deep galaxy counts at near-IR wavelengths.
Figure 6-4 summarizes the current status of
the deep count data. Note that
in some cases there does not appear to be a significant turnover
in the counts even down to very low flux levels. Whether this is
a result of accessing a huge volume (i.e. the faintest galaxies
are at the highest redshifts) or a reflection of a very steep faint
end slope of the GLF (i.e. the faintest galaxies represent a
very numerous population of low mass objects) is unclear. In addition,
there appears to be a major inconsistency between the optical and
the near-IR counts. Compared to the simple no evolution (NE) models,
the optical counts exceed the predicted counts by approximately
a factor of 10 down to an apparent magnitude limit of B = 25.
In contrast, the near-IR counts are significantly lower and are
fully consistent with the NE model. Recent redshift surveys have
shown that the distribution in redshift space of faint galaxies
is also quite consistent with the NE model. In particular, the
mild luminosity evolution models, popular in the mid 80's, clearly
would predict a tail of z 1
objects in the redshift distribution
which is not observed. In addition, the
FBGs generally have emission lines (indicative of star formation)
and are more weakly clustered than redder galaxies of the same
apparent flux.
|
Figure 6-4: Summary of recent deep count data in various wavebands provided by Dave Koo and Caryl Grownwall. |
There are several possible explanations for the FBGs some of which are quite relevant to the question of where the baryons are at z = 0. We list and comment on the most popular below:
The FBGs are a population of
star bursting dwarf galaxies
located at modest redshift. This suggestion takes advantage of the
fact that in any GLF with 
-1, low mass dwarf galaxies
dominate the space density. To produce the FBGs, however, these
dwarf galaxies have to be at least an order of magnitude brighter
at these modest redshifts which requires a fairly significant star
formation rate. Subsequent heating of the ISM by massive stars
and supernova should be sufficient to heat it beyond the escape
velocity of these low mass systems (see Wyse and Silk 1985). These
galaxies would have a significant phase of baryonic blowout after
which they fade to very low absolute luminosities and are hard to
detect at z = 0. This mechanism effectively gives the Universe a
channel for making baryons "disappear" with time.
The number density of
galaxies is not conserved and the
FBGs merge with other galaxies. It is difficult to support this
hypothesis because 1) the FGBs are already weakly clustered and
2) the required merging rate is significantly higher than the
rate measured at modest redshift by Patton et al. (1996). The
merger idea works best if the FBGs are predominately at higher
redshift, where the merger rate is higher owing to the much smaller
volume of the Universe.
Over the redshift interval
which contains most of the FBGs,
the volume is larger due to a positive cosmological constant. Non-zero
Universes have larger
volumes per unit redshift interval compared
to = 0 models.
As in the case of fits to the power spectrum, non-zero
models
also fit the deep count data rather well, although if the FBGs are
primarily at low redshift (z
0.7), the volume effect is less
pronounced.
The FGBs represent an
entirely new population of galaxies -
one defined by a star formation history and or initial mass function
that allows only a limited window of visibility before the galaxies
fade to extremely low surface brightness levels by z = 0. In general,
its always dangerous to introduce a new population of objects in
the Universe without strongly considering the possibilities of
detecting the relic population (see below).
The apparently high number
density of the FBGs is an artifact of uncertainties in the determination
of the local GLF (see Gronwall and Koo 1995).
In particular, the faint end slope of the GLF has
been seriously underestimated from nearby samples (Sprayberry et al. 1997).
This possibility remains highly viable (see below) and in fact,
incorporating a steeper faint end slope can remove much of the apparent
excess.
The local normalization
(
(0)) of the GLF is too low. This could
result if, for instance, deep surveys were more efficient at selecting
LSB galaxies than nearby surveys. While the evidence presented below
strongly supports this idea, the effect of increasing the space density
at z = 0 can only partially offset the excess FGB counts. A much larger
lever arm is provide by increasing .
v
Very recently, Lilly et al. (1995) (see also Eillis et al. 1996)
have presented a redshift survey of
500 faint galaxies. Their
sample has excellent quality control
and is fairly free from selection effects and is primarily aimed
at determining the GLF up to a
1. Their results have helped
clear some of the confusion cited above. Their principle result is that,
for blue galaxies, there is a change in the GLF by approximately
one magnitude between z
0.38 and z
0.62 and then another
magnitude between z
0.62 and z
0.85. Moreover, many of
these galaxies have been observed with HST in order to measure
characteristic surface brightnesses. Schade et al. 1995 find that
for these blue galaxies, their disks are
1 magnitude higher in
surface brightness at z = 0.8 than z = 0.3. Taken together this
consitutes rather strong evidence for luminosity evolution in the FBGs.
For a 15 Gyr universe, there is approximately 3.3 billion years
between z = 0.85 and z = 0.38. The data indicate that a typical
FGB would decline in luminosity by a factor of 6 over this time
period. This is a modest decline that is quite consistent with
standard population synthesis models involving a normal IMF.
The decline in luminosity is primarily a reflection of the disappearance
of the upper main sequence. At this rate, by z = 0, these galaxies
will certainly not have faded to levels that preclude their detection, although
many would be of LSB.
In contrast to the blue galaxies, the LF for the red galaxies appears
to show very little change back to z
1. As its these objects
which should dominate the near-IR counts, the lack of evolution seen
in those counts is not surprising. Still caution should be exercised
in the interpretation of "blue" vs "red" as the color distinction
is based on the colors of local galaxies (e.g., those of class Sbc)
and its unclear if the Lilly et al. division really allows one to be
comparing the same galaxies at high redshift to those nearby. For
instance, one could get LF evolution with redshift for the "blue galaxies"
because galaxy types that are in their "blue sample" at high redshift
are in fact one's which would be in their definition of a "red sample"
at low redshift due to natural evolution of their stellar populations.
Adding to this confusion is the study of Im et al (1996) whose morphologically
selected sample of E and SO galaxies does exhibit LF evolution between
z = 0.5 and z = 1 of
1 mag. The weakness of that
study is that,
unlike the Lilly et al. sample, the Im et al. sample uses photometric
redshifts, which are probably highly uncertain.
In sum, it seems likely that there are two main populations of galaxies in the Universe - those that are evolving very slowly and those that are showing mild to perhaps rapid luminosity evolution. By z = 0, these two populations should evolve to a population of galaxies which exhibits a wide range of surface brightnesses. If this is the case, then their could be a population of sufficiently diffuse galaxies that have escaped our detection. In turn, this would give rise to the "missing" baryon problem as well as providing the illusion that there are more galaxies at high redshift than at low redshift. For many years, this wide range of surface brightness was not seen in the data. However, once the effects of surface brightness selection of galaxies became understood, these "missing" galaxies were found, and found in large numbers. What follows is the story of that particular scientific journey.