interaction is required to explain the correlations (Rosario and Lutz [2013]). A supply of cold gas drives both star formation and black hole growth; the energy released by the AGN may trigger new star formation; it may also deplete the supply of cold gas, but on a timescale which has only a marginal impact on starburst events.
To summarise: most star formation, and most black hole accretion, appear to take place via secular, smooth, main-sequence, processes. Major mergers play a minor role, although they are seen in some dramatic cases. Star formation and black hole accretion co-evolve, but the extent and direction of feedback between each is still controversial; the earlier view of AGNs quenching star formation seems unlikely to be a dominant process, and perhaps the inflow and availability of cool gas flows, which power both processes, accounts for the similarity in their evolution. However, the existence of large numbers of quiescent galaxies by z ∼ 3 (Man et al. [2014]) still remains to be explained successfully.
2.6
The galaxy main sequence and starbursts
In recent years, in analogy with the ‘main sequence’ of stars (the Hertzsprung- Russell diagram), a main sequence for galaxies (the ‘star-forming main sequence’) has been proposed, showing that the star formation rate (SFR) of normal galaxies (i.e. not quasars or starbursts) is correlated with stellar mass (Noeske et al. [2007]; Elbaz et al. [2007]; Daddi et al. [2007]; see Figure 2.6):
SFR∝ Mα
∗ (2.1)
with α ∼ 0.9 for a sample of galaxies at z ∼ 2 in Daddi et al. [2007].
Outliers above the main sequence are thought to be starburst galaxies (i.e. with enhanced SFRs, up to ∼ 103 M
⊙ yr−1 or higher). As shown in Figure 2.6,
the correlation is fairly broad, and the line above which starbursts are defined is somewhat arbitrary. The galaxy main sequence varies with redshift. Unlike the Hertzsprung-Russell diagram for stars, variations in the galaxy main sequence for different samples of galaxies cannot be used to find an evolutionary path for galaxies.
2.6 The galaxy main sequence and starbursts
Figure 2.6: The galaxy main sequence: the stellar mass - SFR relation at 1.5 < z < 2.5
(figure from Rodighiero et al. [2011]). The different coloured symbols show four different surveys: red circles = shallow PACS-COSMOS; cyan squares = deeper PACS-GOODS- S; black circles = deep BzK-GOODS-S; black dots = BzK-COSMOS. The solid black line shows the main sequence for star-forming galaxies defined by Daddi et al. [2007]. The bottom-right region is not well sampled. Inset is the SSFR-M∗ relation.
The stellar mass of a galaxy reflects past star formation, and current star formation comes from cold gas and dust, so the ‘main sequence’ relation implies a correlation between stellar mass and the mass of cold gas and dust, which is difficult to measure directly. In effect, the galaxy main sequence is a reformulation of the Kennicutt-Schmidt law which relates the star formation rate density to the surface density of gas (Schmidt [1959]; Kennicutt [1998a]).
This relationship of the SFR with stellar mass is sometimes normalised by the stellar mass (referred to as the specific SFR: SSFR= ˙M∗/M∗) which was defined in Brinchmann et al. [2004] who suggested that the correlation between stellar mass and SFR normalised by mass may make relationships with other physical parameters of galaxies clearer. Star formation has two modes: a “normal” star formation mode (or main-sequence mode), which has a typical SSFR at each redshift, and an accelerated (starburst) mode, probably caused by a merger of two galaxies, with a higher SSFR.
2.6 The galaxy main sequence and starbursts
COSMOS and GOODS-S fields at 1.5 < z < 2.5 to study the SFR - M∗ relation. Two of the samples were from PACS observations as part of Herschel-PEP (with fairly high SFRs), and two were star-forming galaxy (SFG) samples selected using UV and BzK selections from existing data, in order to probe to lower SFRs (it reached down to a few M⊙ yr−1, not quite reaching down to the star formation
level of the Milky Way). The result (see Figure 2.6) is different for each sample, but broadly confirms a general correlation. Starbursts are suggested above lines at either 4× or 10× main sequence level. Rodighiero et al. [2014] found that galaxies selected from far-infrared Herschel-PACS data did not conform to the previously found slope for SFR-M∗ but stacking BzK-selected galaxies on the PACS maps did find overall agreement with previous results. They concluded that far-infrared-selected galaxies were outliers from the main sequence.
Whitaker et al. [2012] analysed a sample of 22,816 star-forming galaxies in the range 0 < z < 2.5. They found that star-forming galaxies could be classified into three groups: dusty blue galaxies with high SFRs; actively star-forming galaxies on the main sequence and red star-forming galaxies with low SFR (possibly on their way to quiescence). Excluding the red star-forming galaxies gave a linear main-sequence relation, whereas including all galaxies did not. Whitaker et al. [2014] extended this with a study of 39,106 low-mass star-forming galaxies at 0.5 < z < 2.5, finding the slope of the SFR-stellar mass relation is steeper for low mass (α ∼ 1) than for high-mass galaxies. Leja et al. [2015] found this variation in the slope made possible a reconciliation with observed stellar mass functions, which was not possible for a single α∼ 0.9 slope.
Speagle et al. [2014] combined 25 studies to investigate the evolution of the main sequence for a redshift range out to z ∼ 6, finding a clear consensus among observations. They found the width of the scatter was ∼0.2 dex throughout this period, with an evolution in the slope of the main sequence. BzK-selected galaxies were well below galaxies selected by other methods. Steinhardt et al. [2014] found the main sequence relation extended out to z ∼ 6 in a study of 3,398 star forming galaxies. Bauer et al. [2013] found that an analysis of the GAMA survey for ∼73,000 low-mass galaxies at 0.05 < z < 0.32 showed a surprisingly high SSFR. Atek et al. [2014] found that starbursts may be more common than thought, from a study of dwarf galaxies at high redshift (1≥ z ≥ 2). Pannella et al. [2015] in a