Annu. Rev. Astron. Astrophys. 1992. 30: 613-52 Copyright © 1992 by Annual Reviews. All rights reserved

4. CONCLUSIONS

Roughly a thousand spectroscopic redshifts have now been obtained for field galaxies at z > 0.1. The spectra reveal a distribution of redshifts at each magnitude that is not far from what is expected, at least for B < 21 (z_med ~ 0.2). Fainter than this, the predicted no-evolution median redshift falls short of that observed, which is expected if galaxies experienced higher rates of star formation at moderate redshifts. The observed samples at B > 20 contain some low- luminosity galaxies, L < 0.01 L*, but the fraction of such low-luminosity galaxies appears to be at least qualitatively consistent with what is known from brighter samples. Moreover, high-luminosity galaxies appear in about the expected numbers. Thus, the redshifts so far obtained in surveys of faint galaxies are broadly consistent with conservative expectations. More detailed conclusions cannot yet be made because it is clear that the inhomogeneous distribution of galaxies badly distorts the underlying redshift distribution, despite the combination of several fields. Large-scale structure is presumably also responsible for the significant differences between independently derived redshift distributions.

Bluer broad-band colors are more sensitive to the formation of massive stars, and provide a direct way to test whether the star formation rate has changed with redshift. Nothing especially dramatic is evident in the color-redshift diagram (Figure 3), but the distribution of points is potentially sensitive to observational selection (galaxies at the highest measured redshifts will be those with strong spectral features at conveniently observed wavelengths). A less direct index of star formation is the strength of emission lines such as [OII] 3727. Since there is evidence that emission-line statistics may depend on galaxy luminosity, it is important to compare samples at different redshifts that comprise a similar range of absolute magnitudes. When this is done, the data do not show a change in the [OII] properties with redshift. (It has also not been demonstrated that local samples of galaxies with measured W₃₇₂₇ have been selected in a comparable way to the samples at z > 0.1.)

Photometric measurements extend several magnitudes fainter than the spectroscopic surveys, but the conclusions drawn from them are necessarily more model-dependent. The visible-band color distribution shows a smooth and relatively mild trend to bluer colors with increasing magnitude, without any clear signature of an additional population of galaxies with unusual redshifts or intrinsic colors. Before the small number of very blue (or, indeed, very red) faint galaxies can be properly interpreted, the error distribution in the measured colors and reliable statistics of galaxies with extreme colors in local samples need to be established. The distribution of galaxies in multicolor diagrams has been successfully modelled with conventional assumptions about the galaxy population. The uncertainty in the spectral energy distributions in these models may explain the lack of detailed agreement, where such exists. The conclusions drawn from multicolor photometry in the visible band have not been shown to be inconsistent with colors that include measurements in the near-infrared.

The counts in visible passbands show an excess at faint magnitudes with respect to a model that assumes galaxies at high redshift are the same as they are locally. This effect has been known for a long time; newer counts extend the result to fainter magnitudes.

The K-band counts are better matched by the no-evolution model, although the significance of this may be lower because there are fewer surveys, each containing many fewer galaxies. The overall shape of the counts is well matched by a no-evolution model with = 0, and the normalization is consistent with the visible-band counts.

In this review, we have consolidated for the first time most of the existing data on B, R, and K galaxy counts; B - R colors; and redshift distributions. The bulk of these measurements are based on relatively complete samples of faint field galaxies that range roughly from B = 15 to B = 24 for redshifts, and from B = 15 to B = 28 for imaging photometry. Virtually all of these data have been acquired over the last decade and represent an enormous advance in the depth and quality of information relating to the evolution of distant galaxies and to cosmology. To account for part of these data, different groups have proposed modifications to the conventional picture of mild luminosity evolution for z < 1. Examples include adoption of a large cosmological constant; substantial merging at low redshifts; a vast increase in bursting activity at moderate look-back times; or entirely new populations of galaxies that were present at high redshift but absent today. We have instead taken a more conservative stance and asked whether all the data might be found to be consistent with a simpler picture in which the cosmological constant is zero, the number of galaxies is conserved over time, and the shape of the luminosity function for each galaxy class is constant. Notwithstanding the new redshift data, we have argued that the combined uncertainties in the models and in the data so far preclude the necessity of more exotic assumptions.

We find that a no-evolution model for galaxies which incorporates a standard cosmology is able to account for most of the qualitative trends. One main exception is the possible need for mild luminosity evolution, consistent with existing redshift data, to explain the excess counts at faint visible-band magnitudes. The other evolutionary effects displayed by field galaxies are even more subtle and have not yet been detected with confidence. This area of research is likely to remain active (and controversial) for years to come. ACKNOWLEDGEMENTS

We thank colleagues who sent preprints and data. Howard Yee and Arati Chokshi were most helpful in creating the simulations shown in Figure 8, and Craig Mackay and Len Cowie provided the data used in the comparison. P. Guhathakurta provided the faint color data in Figure 2. A number of other collaborators have assisted in the acquisition of the unpublished data presented here, for which we acknowledge telescope time at KPNO and at CTIO. Richard Ellis and Tony Tyson made a number of valuable comments on an earlier draft. The model discussed here relies on the work of Gustavo Bruzual. John Smetanka, Nancy Ellman, and Greg Wirth provided important assistance, and Richard Dreiser prepared the figures and provided editorial help. This work has been supported by AST-8858203 (NSF PYI) and by AST-8705517.