5.1. More efficient star formation at high z
It is now widely recognized that starbursts were more frequent in the past, and galaxy imaging at high redshift with the Hubble Space Telescope has revealed considerable evolution. Although there are still many systematic biases in high-z studies, it appears that galaxies were more numerous, and in particular more perturbed and irregular. The Hubble classification is difficult to pursue at high z. Galaxies are knotty, have less organised structures, and much less bars (van den Bergh et al 2001). Their irregular appearance can be attributed to interactions, since there are more pairs and more mergers at high redshift (Lefevre et al 2000).
The higher star formation rate at high z is easy to explain :
In the frame of the hierarchical scenario, where large galaxies today have been formed by succesive mergers of smaller entities, the first haloes to form at high redshift have very small masses. But they are also denser, because they virialise from a much denser universe, due to expansion. The volumic density is going as (1 + z)3, and the dynamical time-scale inside these haloes is going as dyn ~ (1 + z)-3/2. Therefore, in addition to the larger fraction of mergers at z = 2, the efficiency for a given merger to form stars is even higher. The feedback mechanism, related to the life-time of OB stars, has no reason to vary with redshift, and the time-scale to accrete gas is shorter at high z.
Also, it is easy to predict, since galaxies accumulate mass in their bulge through secular evolution and galaxy interaction/merger, that galaxies in the past were more unstable, having a smaller bulge-to-disk ratio. Bar instability is then more violent, with more gas accretion, and bars are destroyed also more quickly. The fact that bars are transient might explain the observed lower bar frequency, although the present observations are still preliminary.
5.2. Relation between starburst and AGN
Starbursts and AGN compete for gas fuel. They relie on the same dynamical mechanisms to be feeded and active. The main consequence of radial gas flow due to bars and gravity torques is not only a nuclear starburst and an AGN, but also the bulge growth, and a massive black hole growth. However, the amount of gas required to grow the BH over Gyrs is small, ensuring that both can occur simultaneously, which is reflected in the observed correlation between the final masses: MBH = 0.1-0.2% Mbulge (Magorrian et al 1998, Ferrarese & Merritt 2000). The relation between starbursts and AGN is not only circumstancial, but there are effective regulation from one to the other and reciprocally. For instance, the central BH mass can modify the central dynamics, so as to favor gas accretion, or instead to destroy a bar, and stop accretion and star formation. Nuclear starbursts produce outflows (such as M82, N253) that regulate the BH grow, while the compact stellar clusters formed can provide fuel to the BH through stellar mass loss (e.g. Norman & Scoville 1988).
Although there is a massive black hole in almost every galaxy today, most of them are quiescent. According to quasars counts and luminosity as a function of redshift (e.g. Boyle et al 1991), QSOs were more numerous and more powerful in the past. This means that those black holes that were active were more massive, while at low redshift, only more modest black holes are entering their activity cycle (Haehnelt & Rees 1993).
We can deduce that the AGN-starburst connection at high redshift was a little different than today: composite objects were more dominated by their AGN, due to their greater black hole mass.
Another point comes from their lower bulge mass: the inner Lindblad resonance was less frequent, and in this case the gas can be accreted all the way down to the nucleus, since it is not stalled at ILR. Of course, the time-scale of gas accretion is longer when there is no resonance, but this might be compensated by the shorter dynamical time-scale. It is then likely that a black hole was easier to feed at high z. Besides, the accretion being easier, the regulating mechanism was operating faster, then destroying the bar after a shorter time-scale. All these phenomena have to be tackled in details, to determine their actual effect on evolution.