The Canada-France Redshift Survey (CFRS) (Lilly et al. 1995, and references therein) consists of 591 galaxies with secure redshifts (17.5 < I
< 22.5, 0.02 < z < 1.2). This long redshift baseline, a unique characteristic of the CFRS, makes it possible to study the evolution of the galaxy luminosity function without relying on local samples. Figure 
shows the CFRS galaxy luminosity function as a function of redshift with galaxies split according to color (blue 
bluer than Sbc). The local LF of Loveday et al. loveday92 is plotted in all the panels as a reference keeping in mind that is has not been color-split the way the CFRS sample has been.
Figure illustrates a number of important points. There is clear evidence for a population of faint galaxies (M
) in the lowest redshift bin with a significantly higher comoving number density than in the local LF of Loveday et al. loveday92. There is no evidence for evolutionary changes in the galaxy population between the 0.05 < z < 0.2 and 0.2 < z < 0.5 redshift bins, but there are significant differences with the Loveday LF. However, uncertainties in the local luminosity function (see ) make it hard to determine whether there is evolution back to z 
0.2. There is no change in the LF of redder galaxies over the entire CFRS redshift baseline. Therefore, there is no evidence for a substantial decrease with increasing redshift as expected if redder galaxies formed through mergers of massively star-forming sub-units, and the brightening is no more than a few tenths of magnitude as expected from the passive evolution of an old stellar population.
There is substantial evolution in the luminosity function of blue galaxies over the range 0.2 < z < 1.0 which could be viewed either as a luminosity brightening with look-back time or as an increase in the galaxy comoving density. For example, the blue 0.50 < z < 0.70 luminosity function could equally well be fitted by shifting the local LF to the left (luminosity brightening) or shifting the local LF upwards (number density increase). CFRS cannot distinguish between the two. Between 0.2 < z < 0.5 and 0.5 < z < 0.75, the luminosity function brightens by about 1 magnitude. There is no change at the bright end when going to 0.75 < z < 1.0, but there is an additional brightening of 1 mag around M
. It is important to note that this observed evolution does not depend on the local luminosity function at all.
Since the current [OII] kinematics survey covers blue galaxies with redshifts 0.25 < z < 0.45 (see section ), the blue CFRS 0.20 < z < 0.50 redshift bin is of particular interest here. It is interesting to note that, in this redshift bin, the bright end of the CFRS LF (M
) is significantly below the Loveday LF whereas the faint end (M > 
19) lies above it. Since the number of red and blue galaxies is nearly equal, it is reasonable to renormalize the Loveday LF by shifting it down by 0.3 dex. With this renormalization, the bright end of the CFRS LF now agrees with the Loveday LF. This renormalization also leads to much more evolution at M
than suggested by the initial layout of Figure .
The shape of the blue CFRS LF's as a function of redshift is also interesting because it raises the question of whether it can be adequately described by a Schechter function [\protect\astronciteSchechter1976]. The blue CFRS LF in the 0.05 < z < 0.2 and 0.2 < z < 0.5 redshift bins looks as though it could be fitted by a steep straight line, the 0.50 < z < 0.70 LF looks more Schechter-like, and the 0.75 < z < 1.0 LF again looks more like a steep straight line. This behavior is reflected in the values of 
. In the 0.50 < z < 0.70 redshift bin where the blue LF appears to be Schechter-like, 
has a value of 1.07 which is close to the Loveday value. In the 0.2 < z < 0.5 and 0.75 < z < 1.0 redshift bins, 
has much steeper values of 1.34 and 1.56. However, as stressed by the CFRS collaboration, there is a limited range of luminosities present in the survey at each epoch, and the parameter values of their Schechter segments were not intended as true determinations. If the luminosity function in the 0.2 < z < 0.5 redshift bin truly has a steeper slope and a higher normalization than the Loveday LF, then there is room for considerable evolution at M
19 to 18 at those redshifts.
