Annu. Rev. Astron. Astrophys. 1997. 35: 389-443 Copyright © 1997 by Annual Reviews. All rights reserved

4.5. Evidence for Mergers?

An increase with redshift in the rate of galaxy merging is an attractive way to satisfy the changing shape of the LF. Such an explanation is a natural consequence of hierarchical pictures (Carlberg 1992) and features prominently in many of the ab initio models discussed earlier (Baugh et al 1996). The importance of merging in the faint counts has a long but somewhat inconclusive history (see Carlberg 1996 for a recent summary). The optical LF data has been particularly difficult to interpret in this picture, primarily because the rest-frame light indicates the history of star formation, whereas the merger predictions are based on the evolution of the mass. Rocca-Volmerange & Guiderdoni (1990) introduced a self-similar, mass-conserving evolutionary model in which the comoving number density is required to increase as (1 + z)^1.5 in a = 1 universe. Broadhurst et al (1990), analyzing redshift and number count data in the context of a low local normalization, required a mass growth rate such that a typical galaxy became 4-6 fragments by z = 1. Eales (1993) incorporated a more physically based model and predicted changes in LF shape not dissimilar to those seen in the recent redshift surveys. To overcome uncertainties in relating mass and light, Broadhurst et al (1990) advocated conducting deep K-limited surveys to test the merger hypothesis and predicted a turnover in the mean redshift at faint limits consistent with the absence of large mass objects at high z. Detailed predictions in this context have been made by Carlberg & Charlot (1992) and the latest available K-selected data provide some support for such a picture (Cowie et al 1996); over K = 18-20, the mean population redshift hardly increases with apparent magnitude. However, incompleteness in the z > 1.5 range remains a concern.

A more direct approach might be to estimate the interaction rate by searching for close pairs. Barnes & Hernquist (1995) discussed the importance of "major" and "minor" mergers with respect to the morphology of the host galaxy, and energy arguments suggest a strong redshift dependence in the interaction rate (Carlberg 1996). However, even the local interaction rate remains uncertain because not all of the diagnostic features of a merger are expected to be easily visible (Mihos 1995). One of the earliest observational studies attempting to define the merger rate at large look-back times was that of Zepf & Koo (1989), who found 20 close pairs in a faint photographic sample limited at B = 22 and concluded that the pair merger rate increases with redshift as (1 + z)^m where m = 2-4. The difficulty here lies in correcting for a large number of biases that may artificially raise the apparent interaction rate. Such biases include the dissimilar tactics in analyzing low and high redshift data, a possible boosting in luminosity of satellite galaxies whose star formation is triggered by a merger, and the k-correction, which leads to an increase in the number of late-type spirals at high redshift that often have peripheral H II regions; the latter are rendered more visible in the rest-frame UV. A recent analysis using ground-based data is discussed by Woods et al (1995).

The arrival of HST images has led to renewed interest in this area. Neuschaefer et al (1997) presented a comprehensive analysis of the number of close pairs (< 3 arcsec) to I = 23.5 in 56 MDS fields and analyzed their results in the context of HST and ground-based angular correlation functions. In hierarchical merging, one need not expect to find a significant excess in the angular correlation function on small scales. Neuschaefer et al's data supercedes the earlier MDS study of Burkey et al (1994) [which claimed an increase in the merger rate comparable to Zepf & Koo (1989)] and gave m = 1.2 ± 0.4, as did Woods et al (1995). No convincing excess of pairs with separations less than 3 arcsec was found in comparison with an extrapolation of the angular correlation function.

Greater progress will be possible when redshifts are available. The physical scale around each host galaxy can then be defined and, more importantly, satellites can be constrained to include only those above a fixed luminosity limit (providing some estimate of the k-correction is made). Patton et al (1997) use the field galaxies located via the CNOC1 cluster galaxy survey (Carlberg et al 1996) to define a sample of close pairs within 20 h^-1 kpc. Redshifts are available for half of the secondary images and demonstrate that a sizable fraction are true physical associations. By z = 0.33, 4.7 ± 0.9% of the faint population are claimed to be merging, and comparison with local data suggests m = 2.8 ± 0.9. A number of corrections are required in this analysis to allow for the idiosyncrasies of the observing strategy that produced the primary sample and contamination from cluster galaxies in the secondary sample. LeFevre et al (1996a) had both the advantage of HST images and a very wide redshift range, which means that the evolutionary trend can be examined without reliance to any local data. Moreover, LeFevre et al sampled to much fainter limits around each primary galaxy. More than half of the major mergers in the LeFevre et al sample have z > 0.8; indeed no strong trend is seen until the redshift is quite large, in qualitative agreement with the predictions of hierarchical clustering (Baugh et al 1996). However, the absolute rate remains uncertain when determined by pair counts, and so whether merging is the dominant process driving the evolution of the LF remains unclear.

The above studies demonstrate the difficulties of verifying the merger hypothesis quantitatively. However, some important points can be made. First, following the upward revision in the local LF (Section 4.1), there is less need for rapid number evolution at low z, and this considerably reduces the difficulties concerning the abundance of recent merger products (Dalcanton 1993). Second, notwithstanding the uncertainties, there is growing observational evidence from HST images of galaxies of known redshift and the modest depth of the faintest K-limited surveys that merging is of increasing importance at high redshift.