7.7. A Brief Comparison to Theories of Galaxy Formation
The Pritchet (1994) review on primeval galaxies speculates that progenitors of galaxies like our Milky Way should be very numerous, regardless of appearance. The data, in the form of images and spectra of distant z 3 Lyman-break systems and faint number counts, currently suggests that galaxies start as many smaller subclumps and halos so that the local density of L* galaxies (0.015 h503 Mpc-3) may substantially underestimate the comoving space density of small objects at z = 5.
The dominant paradigm for understanding the Lyman-break population is the "dark halo" model: galaxies quiescently form stars at the bottom of the potential wells of massive dark matter halos. Assuming rest-frame UV luminosities correlate with galaxy mass, the brightest Lyman-break galaxies should form first in regions where the density is highest. Since these regions are expected to be strongly clustered spatially, the high-redshift, large-scale structures discussed in Section 7.6 are explained naturally. Over time, the halos merge, forming the massive galaxies we see locally.
Baugh et al. (1998) present a semianalytic model of this hierarchical galaxy formation scenario, focusing on the properties of Lyman-break galaxies at z 3. With a "suitable" choice of parameters, they are able to reproduce the observed Lyman-break galaxy properties for cold dark matter (CDM) cosmologies with both 0 = 1 and 0 < 1. At high redshift, galaxies have very small bulges or no bulge at all: typical half-light radii are ~ 1 h50-1 kpc, in good agreement with the z ~ 3 results of Giavalisco et al. (1996) and the HDF images at z > 4. Baugh et al. (1998) also reproduce the strongly biased spatial distribution, with b 4 and a comoving correlation length r0 8 h50-1 Mpc at z 3. These models predict that the average L* galaxy today was in 4 subunits at z = 1 and 14 subunits at z = 5.
However, the Baugh et al. (1998) models fair less well with respect to star formation rates. They predict that at z 3, most galaxies are only forming a few solar masses of stars per year and only a very small fraction have star formation rates in excess of 40 h50-2 M yr-1. This is at odds with more recent estimates of the rest-frame UV extinction of the z 3 Lyman-break population, e.g., the near-infrared spectroscopic results of Pettini et al. (1998). The hierarchical models also predict that galaxies form the bulk of their stars at relatively low redshift (e.g., Baron & White 1987), with 50% of the stars formed since z 1. The Baugh et al. (1998) models predict that cosmic star formation history peaks around z = 1-2, in rough concordance with the first measurements of the comoving star formation history by Madau et al. (1996). More recent measurements, discussed in Section 7.2, still show the rapid evolution in comoving star formation rate from z = 0 to z ~ 1.5, but, with larger samples of high-redshift objects less vulnerable to cosmic variance and improved consideration of rest-frame UV extinction, the revised plots show a plateau in the comoving star formation density from z ~ 1.5 to z ~ 4 (Fig. 11).
An alternate view of the Lyman-break population maintains that these galaxies are primarily collision-induced galactic starbursts, triggered by small, gas-rich satellite galaxies (Lowenthal et al. 1997; Somerville, Primack, & Faber 1999; Kolatt et al. 1999). Using semianalytic models, Somerville et al. (1999) study the properties of individual galaxies in the "quiescent" dark halo scenario, similar to that addressed by Baugh et al. (1998), in comparison to the "collisional starburst" scenario. They argue that the high star formation rates, small nebular emission-line widths (~ 70 km s-1; Pettini et al. 1998), young ages (median age ~ 25 Myr; Sawicki & Yee 1998), and high surface densities are all better explained by the collisional starburst model. More recently, Kolatt et al. (1999) use high-resolution N-body simulations to address the clustering properties of Lyman-break sources in the collisional starburst model. They find that although most sources are relatively low mass in this scenario, they cluster around high-mass halos and therefore exhibit the observed strongly biased clustering.