This review has been rather narrowly focused on the internal evolution of isolated disk galaxies. The environment surely does play a substantial role in galaxy evolution; it is probably responsible for warps, lop-sidedness, tidal bridges and tails, and a whole host of phenomena related to halo substructure, halo streams, galaxy transformations, dry mergers, etc., but broadening this review to include all, or even some, of these topics would have necessitated either a shallower treatment or a greatly increased length.
The internally driven evolution of galaxy disks would scarcely be of any interest if the disk were composed of stars alone. Spiral activity would heat the disk on the time-scale of a few disk rotations, causing later spiral episodes to be progressively weaker and less distinct. The extent to which the overall distribution of angular momentum among the stars could be rearranged on large scales is strongly limited, since redistributive changes necessarily increase random motion. The fractional change in angular momentum of a distribution of stars (Section 3.5) with radial velocity dispersion R and typical radial excursion a ~ R / is bounded by
where Vc is the circular orbit speed at radius R, and m 2 is the angular periodicity of the spiral patterns. Thus the small value of both factors on the right-hand side provides a very tight constraint on the extent to which the distribution of angular momentum among the stars of a galaxy disk can have changed since their birth.
However, this constraint does not apply to individual stars, which can migrate radially for large distances within the disk through interactions near the corotation resonance of spirals (Section 3.3). Since gains by some stars are roughly matched by losses by others in every diffusive step, these changes alter only the distribution of metals in the disk with almost no change to its dynamical structure. In particular, they neither lead to increased random motion, nor do they cause the disk to spread.
Note also that Eq. (30) does not limit the possible angular momentum changes of the gas component. The random motions of gas clouds, which experience similar radial accelerations from nonaxisymmetric disturbances as do the stars, are quickly damped through dissipative collisions with other clouds. Furthermore, the low velocity dispersion of the clouds makes them highly responsive to nonaxisymmetric disturbances, allowing them to exchange angular momentum with the driving potential to a greater extent than for the stars. Thus secular evolution in galaxies is greatly accelerated by the gas component. Since gas is consumed by star formation, it requires constant replenishment, as is expected in hierarchical structure formation models (e.g. Gunn 1982).
The rising velocity dispersion of disk stars with age is now thought (Section 3.2.6) to be driven by the combined influence of deflections away from circular orbits by scattering at the resonances of spiral patterns, with the resulting in-plane peculiar motions being efficiently redirected into the third dimension by encounters with massive gas cloud complexes. No other combination of heating and scattering can account for both the high dispersion of the older disk stars and the fact that the velocity ellipsoid maintains a roughly constant shape as it grows in size. This combination of factors has not been tested in fully self-consistent simulations because particle masses in most simulations are too large to mimic the two processes separately. The vertical heating that has been reported in some simulations is probably due to collisional relaxation (Sellwood 2013b).
Bars are another important agent of secular evolution. The formation of a bar causes the largest change in the distribution of angular momentum among the stars of a disk, and further evolution occurs only through the influence of the outer disk, halo, and/or gas component. Bars can continue to grow, losing angular momentum to the outer disk, or to the halo, and the fact that bars are usually surrounded by an extensive disk suggests that halos cannot be dense enough to cause them to grow excessively (Section 5.3).
Bars slow, as well as grow, through dynamical friction from the halo (Section 6). The loss of angular momentum by this mechanism also causes the disk mass to contract slightly, which actually deepens the gravitational potential, overwhelming any tendency for halo density to decrease as a result of its energy gain from the disk. While the central density rises, bars also grow in length as they slow, and the fact that corotation of most bars today appears to lie just beyond the bar end requires that the inner DM halos have lower densities than is predicted by CDM models of galaxy formation (Section 6.2).
Bars also drive gas in towards their centers, causing the build up of gas-rich nuclear rings (Section 5.5) where stars are seen to form at a high rate (Section 7.1). The integrated inflow over the lifetime of a galaxy can lead to the build up of concentrations of stars and gas in the center that may be able to destroy the bar and to form a pseudobulge (Section 8).
Substantial evolutionary changes to the structure of disks could also occur through outside intervention, although the degree to which minor mergers could be important is again strongly constrained by data (Section 4). The infrequency of classical bulges (Kormendy et al. 2010) places strong constraints on past merging activity, as does both the thinness of the main disks, and the absence of young stars in thick disks.
The realization that secular evolution is capable of rearranging the structure of disk galaxies from their initially endowed properties has been gradual. The topic was perhaps begun by Kormendy 1979, and it has gradually gained credence, largely through his constant advocacy. Despite the enormous progress described in this review, there are many areas where more work, such as the shaping of rotation curves (Section 3.4) and the weakening of bars by spirals (Section 5.6), or even new ideas, such as to account for the observed fraction of galaxies that host bars (Section 5.7) or the formation of double bars (Section 5.1) or of lens components (Section 8), are needed. Above all, we need better algorithms, with low numerical viscosity (Section 5.5), to capture the role of gas in more realistic manner – a need that is also recognized in galaxy formation.
I thank Tad Pryor, Michael Solway, and Ortwin Gerhard for helpful conversations, and the editor for his patience. Comments by an anonymous referee, Rok Roskar, Victor Debattista, James Binney, and especially by John Kormendy were extremely valuable. This work was supported in part by NSF Grants AST-1108977 and AST-1211793.