In 1972, Toomre & Toomre put forth the bold hypothesis that most giant ellipticals might be the remnants of major disk-disk mergers. Toomre (1977) elaborated on this idea, proposing a sequence of 11 increasingly merged disk pairs and refining the argument that from the current merger rate one could expect between 1/3 and all local ellipticals to be remnants of ancient mergers. Much evidence has since accumulated to support this hypothesis.
Yet, beginning with the 1996 release of the Hubble Deep Field data, a new generation of astronomers has begun to study galaxy formation directly at high redshifts, often with remarkable success, but too often also making claims about elliptical formation that run afoul of the merger hypothesis and its strong supporting evidence in the local universe. For example, claims about (1) an "E formation epoch" ending around z 2 and (2) constant comoving space densities of ellipticals since then abound, but are clearly mistaken.
Few astronomers would contest that disk-disk mergers are occurring locally (z 0) and forming remnants remarkably similar to young ellipticals. Evidence that some field Es contain intermediate-age stellar populations is also increasingly being accepted. What remains controversial is how most older ellipticals formed, say the majority that formed during the first half of the Hubble time and now appear uniformly old, crammed as they are into a small, 0.3-dex logarithmic-age interval. Did they form by major disk mergers as well, or did they form by a process more akin to "monolithic collapse"?
First, the similarities between recent, 1 Gyr-old merger remnants like NGC 3921 or NGC 7252 and giant Es (e.g., Toomre 1977; Schweizer 1982, 1996; Barnes 1998) are worth re-emphasizing. The above two remnants currently have luminosities of ~ 2.8 LV* and will still shine with ~ 1.0 LV* after 10-12 Gyr of evolution. They feature r1/4-type light distributions, power-law cores, UBVI color gradients, and velocity dispersions typical of Es. Both also possess many young, metal-rich halo globular clusters. They show integrated "E+A" spectra indicative of b 10% starbursts (Fritze-von Alvensleben & Gerhardt 1994), as do many other similar young merger remnants in the local universe (Zabludoff et al. 1996). Hence, claiming that E formation ceased around z 2 is as mistaken as would be any claim that star formation ceased then. Local starbursts and merger remnants tell a different story.
Second, although the age distribution of local E and S0 galaxies is clearly weighted toward old ages, it does show a tail of youngish galaxies, especially in the field, with luminosity-weighted mean population ages of ~ 1.5-5 Gyr (Gonzalez 1993; Trager et al. 2000; Kuntschner et al. 2002). Hence, in the field E+S0 formation has clearly not ceased yet.
Third and to astronomers' surprise, massive disk galaxies not unlike the Milky Way have been discovered at 1.4 z 3.0 (Labbé et al. 2003) and thus were available for major mergers at the epoch of peak QSO formation. These galaxies seem to represent ~ half of all galaxies with LV 3LV* at those redshifts. Complementing such IR-optical observations, Genzel et al. (2003) have found a large disk galaxy at z = 2.8 whose rapidly rotating CO disk indicates a dynamical mass of 3 × 1011 M. Even more surprising is a massive old disk galaxy at z = 2.48 that shows a pure exponential disk ( 1.7 kpc) and no bulge, has a luminosity of ~ 2 LV*, and has not formed stars for the past ~ 2 Gyr (Stockton et al. 2004). This galaxy indicates that massive Milky-Way-size disks were available for E formation through major mergers even at z > 3.
With this high-z availability of disks and the above evidence that disk mergers continue to form E-like remnants to the present epoch, it is instructive to revisit Toomre's (1977) argument that most ellipticals may be merger remnants. Figure 4 shows, to the left, his original sketch of the merger rate (t) as a function of time t and, to the right, a modern version of it, in which I have plotted the rate and computed number of remnants vs. (1 + z). From the ~ 10 ongoing disk-disk mergers among ~ 4000+ NGC galaxies and their mean age of ~ 0.5 Gyr, Toomre argued that there should be at least 250 remnants among these NGC galaxies, had the rate stayed constant, and more likely ~ 750 remnants if the merger rate declined as t-5/3 (assuming a flat distribution of binding energies for binary galaxies). The latter number being close to the number of Es in the catalog, he suggested that most such galaxies may be old merger remnants.
Figure 4. Merger rates and numbers of merger remnants as functions of cosmic epoch. (Left) Toomre's (1977) original sketch and (right) a modern version of it. For details, see text.
The top panel of the modern diagram shows the same rate, t-5/3, plotted vs. (1 + z) assuming that major disk merging began 1 Gyr (dotted lines) or 2 Gyr (solid) after the big bang, with the corresponding numbers of remnants labeled N1 and N2 in the bottom panel. Dashed lines mark the case of constant (t) for comparison, and epochs for the standard CDM cosmology are given at the top. Notice that major disk mergers beginning at 1 Gyr (z 5.6) would produce more remnants than needed to explain all Es among present-day NGC galaxies, while such mergers beginning at 2 Gyr (z = 3.15) would produce just about the right number. Interestingly, half of the N2 remnants would already have formed at z = 1.64 (dot on N2 curve), which may explain why observers are having a hard time deciding whether the comoving space density of Es changes from z 1.5 to 0 or not.
In summary, with massive disk galaxies already present at z 3, major disk mergers must have contributed to a growing population of elliptical galaxies ever since. Like star formation, E formation through major mergers is an ongoing process in which gaseous dissipation and starbursts play crucial roles.
Acknowledgments I thank A. Baldi, J.E. Barnes, G. Fabbiano, A. Toomre, and B.C. Whitmore for permission to reproduce figures, and gratefully acknowledge support from the NSF through Grant AST-0205994.