Redshift surveys are vastly superior to number count studies because the intrinsic luminosity of excess galaxies can be determined by using redshift as a distance indicator. Knowing their intrinsic luminosity helps in deciding what role selection effects may or may not play in the discrepancy between local and intermediate redshift samples. For example, galaxy surface brightness dimming goes as (1+z) and such a strong function of redshift will work against high redshift galaxies. As discussed in Phillipps and Driver phil95 and references therein, there is also a local bias against low surface brightness galaxies. Distant galaxies are frequently surveyed with extremely low isophotal thresholds around = 25 mag/arcsec corresponding to an intrinsic surface brightness 27.5 mag/arcsec, but very little is known about local galaxies with surface brightness below 25 mag/arcsec.
The heart of the faint galaxy problem lies with the fact that the same no-evolution models which fail to explain the number counts discussed in the previous section seem to fit the redshift distributions N(z) of faint galaxies. The contradiction between N(m) and N(z) poses a basic problem to understanding intermediate redshift galaxy evolution.
The Durham/Anglo-Australian Telescope faint galaxy survey [\protect\astronciteBroadhurst et al.1988] studied over 200 field galaxies selected in apparent magnitude slices in the range 20.0 < b < 21.5 in five high Galactic latitude fields. The wavelength coverage sampled distinctive spectral features such as [OII] 3727 Åand Ca II H and K 3968,3933 Åover the redshift range 0 < z < 0.6. The mean redshift was 0.25. No high redshift galaxies were found. All redshifts were below 0.47. This was a surprising result. Luminosity evolution in L galaxies should have made them visible at higher redshifts. Their no-evolution model based on the DARS results [\protect\astroncitePeterson et al.1986] provided a good fit to N(z) while falling short (by a factor of 1.5 at b = 21.5) of reproducing blue number counts. The observed [OII] equivalent width distribution N(W) showed an excess of [OII] strong (W > 20 Å) objects. The slope of the counts changed as a function of W, going from = 0.18 for W < 20 Åto = 0.61 for W > 20 Å. This excess matched the count excess at b 2121.5, and, since [OII] strong objects are usually very blue, they concluded that the excess population seen in the counts could be identified with the star-forming strong emission-line galaxies.
Better constraints were placed on the evolution of galaxies with luminosities greater than L by the LDSS survey [\protect\astronciteColless et al.1993][\protect\astronciteColless et al.1990]. This survey looked at 149 objects with magnitudes in the range 21 b 22.5. The LDSS redshift distribution is reproduced in Figure . The LDSS survey found that no more than 2 (90 confidence level) or 4 (99 confidence level) of galaxies brighter than b = 22.5 were at redshifts higher than 0.5. The 90 level upper limit on the number of high-redshift galaxies was not consistent with any evolution of the most luminous galaxies, and the 99 limit was consistent with no more than 1.01.2 magnitudes of brightening by z = 1. These limits led to the luminosity-dependent luminosity evolution hypothesis (discussed in section ) to reconcile N(m) and N(z). The br colors as a function of redshift in the LDSS survey spanned the full range of colors expected from the reddest spectral energy distributions of E/S0 and the spectral energy distribution of the bluest local galaxy, NGC 4449. In fact, some LDSS galaxies were bluer than NGC 4449 would appear at those redshifts.
The deepest B band redshift survey is the LDSS-2 [\protect\astronciteGlazebrook et al.1995a]. It produced 73 redshifts for objects in the magnitude range 22.5 < B < 24. The median redshift was z = 0.46. The survey showed a large excess of galaxies at z 0.4 with respect to luminosity evolution models of the form with b=0 being the no-evolution case. There was an increase in the space density of galaxies with L L(z = 0) which could arise through mergers or luminosity-dependent luminosity evolution. The survey also showed no trace of an excess of z < 0.2 galaxies predicted by models trying to explain number counts solely on the basis of a steep faint end to the local luminosity function.
A K-band selected redshift survey of 124 galaxies down to K 17.3 [\protect\astronciteGlazebrook et al.1995c] showed no evidence for evolution of the K-band luminosity function below z = 0.5, but the luminosity function required a high normalization of = 0.026 h Mpc. Beyond z = 0.5, M increased by 0.75 mag at z = 1. This result was opposite to expectations from simple merger-dominated models in which the masses of galaxies should decrease with redshift.