Annu. Rev. Astron. Astrophys. 2006. 44: xxx-xxx Copyright © 2006 by Annual Reviews. All rights reserved

6. EPILOGUE

Almost thirty years ago Toomre (1977) remarked that star formation was observed only in galaxy disks, and further that the final state of merging spirals must be something resembling an elliptical galaxy. Thus, merging spirals to form ellipticals at relatively low redshifts became very popular, especially following the success of CDM theories in accounting for the growth of large scale structure from tiny initial perturbations.

However, in the intervening three decades an impressive body of evidence on galaxies at low as well as high redshift has accumulated, that at least in part contradicts Toomre's assumption. While in the local universe most of the star formation is indeed confined to disks; at z > 1-1.5 most of it appears to take place in starburst galaxies, such as ULIRGs, whose space density is orders of magnitude higher than in the local universe. Moreover, ~ 50% of all stars seem to have formed at z 1 (Dickinson et al. 2003), and to have occurred mostly in disks (Hammer et al. 2005), whereas, if the scenario shown in Figure 6 is basically correct, then the bulk of star formation in ETGs took place at much higher redshift. At the risk of some simplification, we can say that the era of ETG/spheroid/elliptical formation was largely finished by z ~ 1 (if not before), just when the major build up of disks was beginning (see also Papovich et al. 2005).

The evidence for the stellar populations in ETGs being old, and older in massive galaxies than in less massive ones, has been known for over ten years, along with the evidence for down-sizing and for the anticorrelation of mass and SFR. Theoretical models based on the CDM paradigm have recently incorporated these observational constraints, and have been tuned to successfully reproduce the down-sizing effect in star formation (e.g., De Lucia et al. 2006). In a hierarchical scenario, down-sizing in star formation is indeed natural. Star formation starts firsts in the highest density peaks, which in turn are destined to become the most massive galaxies later on. But until recently models predicted that star formation was continuing all the way to low redshift, as cooling flows were left uncontrasted, thus failing to even produce a red sequence. To get the old and dead massive galaxies we see in nature, such cooling flows (and the accompanying star formation) had to be suppressed in the models, which is now generally accomplished by invoking strong AGN feedback, as first incorporated in CDM simulations by Granato et al. (2001) ². Yet, the AGN responsibility in switching off star formation remains conjectural at this time, but we became aware that galaxies and supermassive black holes co-evolve, which means we must understand their formation as one and the same problem.

Baryon physics, including star formation, black hole formation and their feedbacks, is highly nonlinear, and it is no surprise if modelling of galaxy evolution relies heavily on many heuristic algorithms, their parameterization, and trials and errors. Dark matter physics, on the contrary, is extremely simple by comparison. Once DM halos are set into motion, there is nothing preventing them from merging with each other under the sole action of gravity, and growing bigger and bigger "galaxies" in an up-sizing process. Thus, the vindication of the CDM paradigm should be found in observations demonstrating that the biggest, most massive galaxies are the first to disappear when going to higher and higher redshifts. This is indeed what has not been seen yet, and actually there may be hints for the contrary.

ACKNOWLEDGMENTS

I thank Ralf Bender, Andrea Cimatti, Emanuele Daddi, Mauro Giavalisco, Laura Greggio, Silvia Pellegrini and Daniel Thomas for a critical reading of the manuscript and for their valuable suggestions. I am indebted to Mariangela Bernardi, Daniel Thomas, and Sperello di Serego Alighieri, for having provided respectively Table 1, Figure 2, and Figure 13, specifically for this paper. Finally, I am very grateful to my Annual Review tutoring editor John Kormendy for his guidance, to Doug Beckner for the final set up of all the figures, and to Roselyn Lowe-Webb for her patience in proof-editing the manuscript.

² See also Ciotti et al. (1991) and Ciotti & Ostriker (1997) for early attempts to suppress cooling flows in ETGs, either with Type Ia supernova feedback alone, or in combination with AGN feedback. Back.