Recent studies of distant field galaxies have produced surprising and seemingly inconsistent results. Perhaps one of the most striking observations is that the number counts in the blue show a steep rise with increasing blue magnitude, indicating an excess of blue faint galaxies (e.g. Tyson 1988). On the other hand, the number counts in the infrared show a shallower slope (see Koo and Kron 1992 for a compilation). The blue number counts disagree strongly with predictions based on simple models of galaxy evolution. These models can, however, reproduce the infrared number counts reasonably well.
One solution proposed to the problem of the excess blue galaxies, that they are a new population of distant galaxies (z > 2), turns out not to work. Most of the blue galaxies have redshifts well below 1, and the redshift distribution of the faint blue galaxies shows no evidence for strong evolution (Broadhurst et al. 1988, Colless et al. 1990). No satisfactory explanation of this paradox has been put forward. A number of possibilities have been discussed: non-standard cosmologies, mergers, bursts of star formation in dwarf galaxies, etc. (see Broadhurst et al. 1988, Lilly 1993).
One of the main difficulties with the interpretation of the observations is the proper modeling of the selection effects and biases in the observations of distant galaxies. Most models begin with a simplified representation of galaxies in the nearby universe. Typically, galaxies at low redshift are divided into a few morphological classes, each characterized by a luminosity function and single color. The star formation timescale in each class is then fixed to reproduce this color. These simplistic models may fail to adequately represent the galaxy population at low redshift, and may therefore be an important component in our inability to understand the distant field galaxy observations.
A paper by Koo et al. (1993) highlights this problem. Koo et al. turn the modelling process around: the number counts and redshift distributions are modelled without using any constraints from observations of nearby galaxies. After a successful model of the distant galaxies is found, they compare the predicted luminosity function of the local universe to the available observations. They find reasonable agreement between the predictions of a no-evolution model and the available observations. Koo et al. go on to produce color distributions as a function of luminosity. Unfortunately, no good nearby comparison sample exists.
Recently, Lilly et al. (1995) claim to have directly detected evolution of blue galaxies (bluer than present-day Sbc galaxies) at z > 0.5 equivalent to a brightening of 1 magnitude from present day luminosities. They find no evidence for evolution of the red galaxies. Lilly et al. construct luminosity functions for two color bins and four redshift bins (up to z ~ 1) from a survey of 591 I-band selected galaxies (the Canada-France Redshift Survey).
The results of Lilly et al. are valuable because they are model independent. However, if we are to understand the origins of the evolution they claim to find, and to confirm its detection, we must study the detailed properties of larger samples of galaxies as a funtion of redshift. Our field galaxy survey is an attempt to remedy the stuation noted by Koo et al.: one of the major problems with the interpretation of the properties of distant galaxies is the absence of a good nearby comparison sample!
The purpose of our study is to obtain an accurate description of the distribution of magnitude, structural parameters, color, and spectral type. The colors and spectral types are necessary to determine the probable evolutionary history of the galaxy. The magnitude and structural parameters (effective radius, surface brightness, rotational asymmetry and central concentration) are necessary to calculate the detection rates at increasing redshift. Furthermore, we can compare our images directly to images of high redshift galaxies obtained with HST (e.g. Schade et al 1995, Abraham et al 1996, van den Bergh et al 1996).