In Blanton et al. (2006) I argue that not many (< 10%) of the galaxies observed on the blue sequence at redshift z=1 can have been destroyed by redshift z=0.1 by any process, by comparing the number density of such objects in the Deep Evolutionary Extragalactic Probe 2 survey to that in the Sloan Digital Sky Survey. All of the figures are available in GIF format.
I build on our understanding of galaxy properties in the SDSS and use that sample to make "no evolution" predictions at higher redshift. I then apply the DEEP2 selection effects to the sample and compare the two results. I focus on the distribution of luminosities and colors. The top panel is the no evolution prediction, and the bottom panel are the DEEP2 results:
As one can see, both the blue sequence and red sequence are bluer at high redshift. In addition, the blue sequence is more highly populated relative to the red sequence at high redshift. We can take a better look at the demographics by splitting galaxies between red and blue and looking at the luminosity function of each:
First concentrate on the blue sequence. It is apparent that the luminosity function at high redshift and low redshift are consistent with a constant luminosity shift. In fact, a fading of about 0.9 mag (which as we will see is consistent with the change in colors of the blue galaxies) yields a best fit change in the number density of about 1.1 +/- 0.1. This means that it is (1) difficult for many galaxies to be destroyed on the blue sequence unless the population is somehow replenished and (2) it is therefore difficult to feed the red sequence by altering blue sequence galaxies.
The red sequence poses a harder problem. It is roughly constant in luminosity function shape and in overall luminosity density --- the population shifts in number (up) and in luminosity (down), however. These changes, along with its aging and reddening, imply a growth in its stellar mass somehow.
I explored stellar population synthesis models of these populations a bit, with some simple models:
The blue sequence models have initial bursts, followed by quiescent star-formation until z=1. The bursts vary in amplitude (which changes the color on the blue sequence) and the star-formation histories in general vary in metallicity (which changes the color on the red sequence). After z=0.1 there are three cases: a sharp cutoff in quiescent star-formation, a reduction of quiescent by a factor of three, and a continuation of quiescent star-formation at the same rate. The first case results in a very rapid reddening of the population up to the red sequence in < 1 Gyr. This rapidity explains the "bimodality" of galaxy colors --- any galaxies with a sharp cutoff in star-formation populate the red sequence almost immediately.
Let us look at these models in the context of the color-magnitude diagram for galaxies:
I have normalized each of the blue sequence models to land on the color-magnitude relation at z=1 (bottom panel). If the quiescent star-formation cuts off sharply they end up on the red sequence at z=0.1 (circles in top panel). If it continues unabated, they end up much bluer than low redshift blue sequence galaxies (triangles in top panel). If it slows by about a factor of three they land near the blue sequence at z=0.1. Such galaxies have faded about 0.9 mag.
Therefore, the number density and colors of blue galaxies at z=0.1 are consistent with a simple reduction in the average star-formation rate since z=1 without any reduction or increase in number density. Quantitatively, the limit is a change in number density of about 10%.
The red sequence is a more complex story. The stellar population synthesis models suggest that such galaxies should at most fade by about 0.8 mag between z=1 and z=0.1. If so, the most luminous galaxies at low redshift cannot be passively faded remnants of either luminous red or luminous blue galaxies at high redshift. To be such, the fading would have to be closer to 0.6 mag. Thus, absent errors in the models or the photometry at the 0.2 mag level, they must grow by mergers with smaller galaxies. I consider this weak but suggestive evidence --- and note that almost every other analysis of this and other data comes to the same conclusion.
Some of the less luminous galaxies on the red sequence can be produced by blue sequence galaxies that stop forming stars and fall onto the red sequence (after all, that happens quickly). But to produce all of them at M0.1g = 19 requires about 25% of the blue galaxies to stop forming stars, which is pushing the limits we found above. These less luminous galaxies on the red sequence may therefore also require merging in of smaller galaxies.
Therefore, in agreement with previous findings, there is evidence for growth of galaxies by mergers on the red sequence.