3.3. The evolution of plausible merger remnants
From the earliest days of galaxy morphological classification, a population of galaxies whose light distribution is dominated by a smoothly distributed, spheroidal, centrally-concentrated light distribution was noticed. These early-type galaxies are largely supported by the random motions of their stars ([Davies et al. (1983)]). These properties are very naturally interpreted as being the result of violent relaxation in a rapidly-changing potential well ([Eggen, Lynden-Bell, & Sandage (1962)]). Therefore, in our present cosmological context, these galaxies are readily identified with the remnants of major galaxy mergers (e.g., [Toomre & Toomre (1972)]; [Barnes & Hernquist (1996)]). Detailed comparisons of simulated major merger remnants broadly supports this notion although some interesting discrepancies with observations remain, and are perhaps telling us about difficult-to-model gas-dynamical dissipative processes (e.g., [Bendo & Barnes (2000)]; [Meza et al. (2003)]; [Naab & Burkert (2003)]). There is strong observational support for this notion - the correlation between fine morphological structure and residuals from the color-magnitude correlation ([Schweizer & Seitzer (1992)]), the existence of kinematically-decoupled cores (e.g., [Bender (1988)]), and the similarity between the stellar dynamical properties of late-stage IR-luminous galaxy mergers and elliptical galaxies (e.g., [Genzel et al. (2001)]).
Therefore, study of the evolution of spheroid-dominated, early-type galaxies may be able to give important insight into the importance of galaxy merging through cosmic history. There are, as always, a variety of important complications and limitations to this approach. For example, a fraction of the galaxies becoming early-type during galaxy mergers will later re-accrete a gas disk, which will gradually transform into stars, making the galaxy into a later-type galaxy with a substantial bulge component (e.g., [Baugh, Cole, & Frenk (1996)]). Furthermore, not all galaxy mergers will result in a spheroid-dominated galaxy; some lower mass-ratio interactions will result in a disk-dominated galaxy (e.g., [Naab & Burkert (2003)]). In addition, it would be foolish to a priori ignore the possibility that an important fraction of early-type galaxies may be formed rapidly in mergers of very gas-rich progenitors at early epochs - a scenario reminiscent of the classical monolithic collapse picture ([Larson (1974)]; [Arimoto & Yoshii (1987)]). Yet, these interpretive complications, much as in all the cases discussed above, do not lessen the value of placing observational constraints on the phenomenology, with the confidence that our interpretation and understanding of the phenomenology will improve with time.
The rate of progress in this field has largely been determined by the availability of wide-format high-resolution imaging from HST. Ground-based resolution is insufficient to robustly distinguish disk-dominated and spheroid-dominated galaxies at cosmologically-interesting redshifts. In the local Universe, the vast majority of morphologically early-type galaxies occupy a relatively tight locus in color-magnitude space - the color-magnitude relation (e.g., [Sandage & Visvanathan (1978)]; [Bower, Lucey, & Ellis (1992)]; [Schweizer & Seitzer (1992)]; [Strateva et al. (2001)]). Therefore, many workers have focused on the evolution of the red galaxy population as an accessible alternative. The efficacy of this approach is only recently being tested. Accordingly, I explore the evolution of the red galaxy population first, turning subsequently to the evolution of the early-type population later.
3.3.1. The evolution of the total stellar mass in red-sequence galaxies
It has become apparent only in the last 5 years that the color distribution of galaxies is bimodal in both the local Universe (e.g., [Strateva et al. (2001)]; [Baldry et al. (2004)]) and out to z ~ 1 ([Bell et al. (2004b)]). This permits a model-independent definition of red galaxies - those that lie on the color-magnitude relation. A slight complication is that the color of the red sequence evolves with time as the stars in red-sequence galaxies age, necessitating an evolving cut between the red sequence and the `blue cloud'. The evolution of this cut means that some workers who explore the evolution of red galaxies using rather stringent color criteria - e.g., galaxies the color of local E/S0 galaxies - find much faster evolution in the red galaxy population than those who adopt less stringent or evolving cuts (e.g., [Wolf et al. (2003)]).
Bearing this in mind, I show the evolution in the total stellar mass in red galaxies in Fig. 6. Although many surveys could in principle address this question (e.g., the surveys discussed in Section 2.2), very few have split their stellar mass evolution by color, and to date most surveys have only published evolution of luminosity densities in red galaxies. The z = 0 stellar mass density in red galaxies is from SDSS and 2MASS ([Bell et al. (2003)]). The COMBO-17 datapoints for jB evolution ([Bell et al. (2004b)]) are converted to stellar mass using color-dependent stellar M/Ls as defined by [Bell & de Jong (2001)]; extrapolation to total stellar mass density adopts a faint-end slope = - 0.6, as found by [Bell et al. (2003), Bell et al. (2004b)] for red-sequence galaxies at z 1. Error bars account for stellar mass uncertainties and cosmic variance, defined by the field-to-field scatter in stellar mass densities from the 3 COMBO-17 fields. Stellar masses for the LCIRS sample of red galaxies were estimated using the rest-frame R-band luminosity density presented by [Chen et al. (2003)], accounting for the mildest possible passive luminosity evolution (to account for evolution in stellar population color and luminosity in the most conservative way, so as to minimize any stellar mass evolution), and using a stellar M/L for early-type galaxies from [Bell & de Jong (2001)] using a [Kroupa (2001)] IMF, and adopting a color of B - R = 1.5 for early-type galaxies as a z = 0 baseline. Error bars for include stellar mass estimation uncertainties and estimated cosmic variance, following [Somerville et al. (2004)]. The K20 data point at z ~ 1.1, derived from ERO `old' galaxy space densities from [Cimatti et al. (2002)], is very roughly calculated using a number of doubtless poor assumptions. A (rather large) stellar mass of ~ 1011 M is attached to each galaxy, and the densities are multiplied by 2 to account for the star-forming EROs (the split was roughly 50:50). This stellar mass density was multiplied by 2 again to account for fainter, undetected galaxies. Error bars of ± 0.2 dex and ± 0.3 dex, combined in quadrature, account for cosmic variance following [Somerville et al. (2004)] plus our poor modeling assumptions. Owing to their use of discordant cosmologies, I do not show the inferred stellar mass evolution of the CFRS red galaxies in Fig. 6; however, like [Bell et al. (2004b)] they infer no evolution in the rest-frame B-band luminosity density to within their errors ([Lilly et al. (1995)]), meaning that their stellar mass evolution would fall into excellent agreement with those of [Bell et al. (2004b)] or [Chen et al. (2003)].
Figure 6. The evolution of the stellar mass density in red-sequence galaxies. Stellar masses assume a [Kroupa (2001)] IMF and H0 = 70 km s-1 Mpc-1. The local data point is taken from [Bell (2003)], and data for 0.2 < z < 1.3 is taken from [Bell et al. (2004b)], [Chen et al. (2003)], and [Cimatti et al. (2002)]. The solid line shows a rough fit to the total stellar mass density in red-sequence galaxies in the [Cole et al. (2000)] semi-analytic galaxy formation model. The dotted line shows the expected result if red-sequence galaxies were formed at z >> 1.5 and simply aged to the present day.
The results are shown in Fig. 6. To first order, the luminosity density in the optical in red galaxies is constant out to z ~ 1; this is confirmed by a number of surveys (e.g., [Lilly et al. (1995)], [Chen et al. (2003)], or [Bell et al. (2004b)]). Coupled with the passive ageing of the stellar populations of these red galaxies (as is indicated by their steady reddening with cosmic time, and is confirmed by study of dynamically-derived M/Ls and absorption line ratios; e.g., [Wuyts et al 2004]; [Kelson et al. (2001)]), this implies a steadily increasing stellar mass density in red galaxies since z ~ 1.2. To date, to the best of our knowledge, there are no published determinations of red galaxy stellar mass or luminosity density which contradict this picture.
The implications of this result are rather far-reaching. Bearing in mind that at z 1 that the red galaxy population may be significantly contaminated and/or dominated by dusty star-forming galaxies (e.g., [Yan & Thompson (2003)], [Moustakas et al. (2004)]), this evolution may well represent a strong upper limit to the stellar mass in early-type galaxies since z ~ 1.2, unless large populations of very bright blue morphologically early-type galaxies are found. I address this question next, by exploring the stellar mass evolution in morphologically early-type galaxies since z ~ 1.
3.3.2. The evolution of the total stellar mass in early-type galaxies
Owing to the lack of extensive wide-area HST-resolution imaging data, there are even weaker constraints on the evolution of stellar mass in morphologically early-type galaxies. There is only one published determination of the luminosity density evolution in morphologically early-type galaxies ([Im et al.(2002)]); I supplement it with a preliminary analysis of galaxies from GEMS and COMBO-17, which is presented with the permission of the GEMS and COMBO-17 teams.
[Im et al.(2002)] present a thorough study of the luminosity density evolution of E/S0 galaxies from the DEEP Groth Strip Survey, defined using bulge-disk decompositions and placing a limit on residual substructure in the model-subtracted images, supplemented with HST V and I-band imaging for 118 square arcminutes. A final sample of 145 E/S0 galaxies with 0.05 < z < 1.2 are isolated. The fit of average early-type galaxy stellar M/L as a function of redshift from COMBO-17/GEMS; log10M / LB = 0.34 - 0.27z, was used to transform the published B-band luminosity densities into stellar mass densities for the purposes of Fig. 7 3.
Figure 7. The evolution of the stellar mass density in early-type galaxies. Stellar masses assume a [Kroupa (2001)] IMF and H0 = 70 km s-1 Mpc-1. The left-hand panel shows the stellar mass densities from GEMS only; solid circles denote early-type galaxies and open circles red-sequence galaxies. The ratio of stellar mass in early-type galaxies to the red sequence is shown in the lower left panel; the asterisk is the local value taken from [Bell (2003)], and the grey line shows a linear fit to the GEMS data only (RMS ~ 15%). The right-hand panel shows the resulting corrected early-type galaxy stellar mass density evolution, where again the local data are taken from [Bell (2003)], the solid circles denote the GEMS+COMBO-17 result, and the diamonds show results from [Im et al.(2002)].
The preliminary GEMS/COMBO-17 early-type galaxy depict the evolution of total stellar mass in galaxies with Sérsic indices n > 2.5. In GEMS, [Bell et al. (2004a)] found that galaxies with n > 2.5 included ~ 80% of the visually-classified E/S0/Sa galaxy population at z ~ 0.7, with ~ 20% contamination from later galaxy types (recall, high Sérsic indices indicate concentrated light profiles). Here, in this very preliminary exploration of the issue, a n = 2.5 cut is adopted irrespective of redshift, ignoring the important issue of morphological k-correction. To recap, stellar masses are calculated using color-dependent stellar M/Ls from [Bell & de Jong (2001)], again extrapolating to total using a faint-end slope = - 0.6.
The resulting n > 2.5 stellar mass evolution for the 30' × 30' GEMS field is shown in the left panel of Fig. 7. One can clearly see strong variation in the stellar mass density of early-type galaxies, resulting from large scale structure along the line of sight. From such data, it is clearly not possible to place any but the most rudimentary and uninteresting limits on the evolution of early-type galaxies over the last 9 Gyr. Yet, noting that the bulk of early-type galaxies are in the red sequence at z ~ 0 (e.g., [Strateva et al. (2001)]) and at z ~ 0.7 (e.g., [Bell et al. (2004a)]), and that the stellar mass density of red-sequence galaxies undergo very similar fluctuations, it becomes interesting to ask if the ratio of stellar mass in red-sequence galaxies to early types varies more smoothly with redshift. This is plotted in the lower left panel of Fig. 7. There is a weak trend in early-type galaxy to red-sequence galaxy mass density, caused by an increasingly important population of blue galaxies with n > 2.5 towards higher and higher redshift. Importantly, however, there is only a ~ 15% scatter around this trend despite the nearly order of magnitude variation in galaxy density, arguing against strong environmental dependence in the early type to red galaxy ratio.
This relatively slow variation in early-type to red-sequence ratio (modeled using a simple linear fit for the purposes of this paper) is used to convert COMBO-17's stellar mass evolution in red galaxies from Fig. 6, which is much less sensitive to large scale structure, into the evolution of stellar mass density in morphologically early-type galaxies, which is shown in the right panel of Fig. 7. It is clear that there is a strongly increasing stellar mass density in morphologically early-type galaxies since z ~ 1.2 4. While there are recent indications that lower-luminosity early-type galaxies are largely absent at z 0.8 (e.g., [Kodama et al. (2004)], [de Lucia et al. (2004)]), the total stellar mass density is strongly dominated by ~ L* galaxies, and there is no room to avoid the conclusion that there has been a substantial build-up in the total number of ~ L* early-type galaxies since z ~ 1.2. Like the results presented in sections 3.1 and 3.2, these results suggest an important role for z 1 galaxy mergers in shaping the present-day galaxy population 5.
3 Again, a [Kroupa (2001)] IMF was used. Back.
4 [Im et al.(2002)] report a low stellar mass density in early-type galaxies at 0.05 < z < 0.6. We would attribute this deficiency in stellar mass to incompleteness at bright apparent magnitudes, leading to a deficit of nearby E/S0s with large stellar masses (as argued by Im et al. themselves), and perhaps to small number statistics and cosmic variance (as the volume probed by this study at z 0.5 is rather small). Further work is required to explore further this discrepancy. Back.
5 Disk re-accretion and fading work in opposite directions in this context; disk accretion following a major merger takes galaxies away from the early-type class, whereas fading of disks which formed at earlier times will increase the relative prominence of spheroidal bulge components and add galaxies to the early-type class. A thorough investigation of these issues, involving bulge-disk decompositions of rest-frame B-band images of the GEMS sample and drawing on galaxy evolution models to build intuition of the importance and effects of different physical processes, is in preparation (Häußler et al., in prep.). Back.