6. ORIGIN OF THE HUBBLE SEQUENCE

The availability of stellar masses for all types enables the construction of a powerful evolutionary plot, analogous to Figure 6, involving the stellar mass density, _stars(z, T), as a function of morphology T. Whilst the stellar mass density can grow by continued star formation, unlike the UV luminosity density, _UV, it is difficult to imagine how it can decline. As we saw earlier _UV can decline significantly in only 1-2 Gyr because of an abrupt truncation of activity. However, such a change would have very little effect on the infrared output as illustrated in Figure 13.

Brinchmann and Ellis (2000) secured K luminosities and optical-IR SEDS for over 300 galaxies in the CFRS/LDSS and Hawaii survey fields and derive _stars(z, T) (Figure 15). Estimating the integrated stellar mass density is prone to all of the difficulties reviewed earlier for the luminosity density and there is the added complication that the redshift surveys in question are optically-selected and thus must miss some (red) fraction of a true K-limited sample. Accordingly, the mass densities derived are lower limits to the true values.

Figure 15. Evolution of the stellar mass density stars(z, T) from the analysis of Brinchmann & Ellis (2000). A remarkable decline with time in stellar mass density is seen for the morphologically-peculiar class which argues against a truncation of their star formation activity as the primary cause for their demise. Brinchmann & Ellis argue that this population must be transforming, possibly via mergers, into the regular classes. A simple model which implements a likely redshift-dependent merger rate ([LeFevre et al 2000]) with elliptical products can broadly reproduce the trends observed (shaded area of the plot).

Remarkably, _stars(z, T) is a declining function for the intriguing population of morphologically-peculiar galaxies. Whereas the declining UV luminosity density could imply a fading population, such an explanation cannot be consistent with Figure 15 which argues, instead, that the objects are genuinely disappearing into other systems. The most logical explanation for their declining contribution to the stellar mass density is that morphologically-peculiar objects are being transformed, e.g. by mergers, into regular objects.

Figure 16. An increase in the merger fraction as a function of redshift from the HST analysis of LeFevre et al (2000). Galaxies of known redshift were examined for satellites brighter than a fixed rest-frame luminosity within a projected radius of 20 h-1 kpc and corrections made for unrelated line-of-sight contamination. This redshift-dependent merger rate was adopted by Brinchmann & Ellis (2000) in Figure 15.

Merging has been an attractive means for governing the evolution of galaxies for many years ([Toomre & Toomre 1972, Rocca-Volmerange & Guiderdoni 1989, Broadhurst et al 1992]) and of course is fundamental to the hierarchical formation picture. However it has been extremely difficult to determine the observed rate at intermediate redshift. The fundamental problem is that we observe galaxies at various look-back times via discrete `snapshots' without ever being able to prove two associated systems are destined to merge on a particular timescale. Using the CFRS/LDSS HST dataset referred to earlier, LeFevre et al (2000) undertook a quantitative survey of the fraction of luminous galaxies with satellites brighter than a fixed absolute magnitude within a 20 h^-1 kpc metric radius and, after allowance for projection effects, determined the merger fraction increases with redshift as (1 + z)^{3.4 ± 0.5} - a result consistent with earlier ground-based efforts. Sadly, it is not straightforward to convert the proportion of galaxies with associated sources into a physical merger rate or, as ideally required, a mass assembly rate without some indication of the dynamical timescale for each merger and the mass of each satellite. Moreover, there are several annoying biases that affect even the derived merger fraction.

Brinchmann & Ellis (2000) attempted to reconcile the decline of the morphologically-peculiar population, the redshift dependence of the LeFevre et al merger fraction and associated evidence for continued formation of ellipticals ([Menanteau et al 2000]) into a simple self-consistent picture. They transferred the dominant population of morphologically-irregular galaxies, via the z-dependent merger rate, into a growth in the regular galaxies (shaded area of Figure 15). This is clearly a simplistic view, but nonetheless, gives a crude empirical rate at which regular galaxies are assembling. If correct, how does this agree with mass assembly histories predicted, say in CDM?

Figure 17 shows a recent prediction of the assembly history of spheroids and disks (Frenk, private communication). Although there are some discrepancies between this and its equivalent prediction from Kauffmann & Charlot (1998, Figure 3), the trends are clear. The strongest evolutionary signal is expected in terms of a recent assembly of massive spheroids; the equivalent growth rate in stellar disks is more modest. To the extent it is currently possible to test this picture, the qualitative trend is supported by the data. Field ellipticals are certainly still assembling ([Menanteau et al 2000]) but perhaps more slowly than expected according to Figure 17; unfortunately deeper samples with redshifts are needed for a precise statement. Brinchmann (in prep.) has examined the stellar mass growth rate in disks using the infrared-based method over 0 < z < 1 and finds only modest changes. This is very much a developing area and one that would benefit from significantly enlarged HST datasets chosen to overlap the growing faint redshift survey databases.

Figure 17. Predicted evolution in stellar mass functions for disk and spheroidal populations in a CDM hierarchical model (Frenk, priv. comm). The curves define mass functions as a function of redshift (z = 0,0.5,1,2, from right to left). Modest growth over 0 < z < 2 is expected for disk galaxies but significant growth is predicted for massive spheroidals.

50% of local spirals have bars which are thought to originate through dynamical instabilities in well-established differentially-rotating stellar disks. If we could determine the epoch at which bars begin appearing, conceivably this would shed some light on how recently mature spirals came to be. Via careful simulations based on local examples, Abraham et al (1999) showed that face-on barred galaxies should be recognisable to z 1 in the HDF exposures. In fact, many are seen (Figure 18) but tantalisingly the barred fraction of face-on spirals appears to drop beyond a redshift z 0.6. The effect is marginal but illustrative of a powerful future use of morphological imaging with the Advanced Camera for Surveys.

Figure 18. Face-on barred spirals of known redshift in the Hubble Deep Field. Abraham et al (1999) claim that all such systems should easily be recognised to z 1 in HDF-quality data whereas, beyond z 0.6, there appears to be a marginal paucity of such systems compared to their non-barred counterparts. If supported by further data, this could indicate an epoch corresponding to the `dynamical maturing' of stellar disks.

The story with bulges is also unclear, although potentially equally exciting. Traditionally, bulges were thought to represent miniature ellipticals which formed monolithically at high redshift ([Eggen, Lynden-Bell & Sandage 1962]). Detailed studies of local examples, including the Galactic bulge, have shown a considerable diversity in properties, both in integrated color and even in their photometric structure ([Wyse 1999]). There is some evidence of a bimodality in the population; prominent bulges in early type spirals share surface brightness characteristics of ellipticals, whereas those in late-type spirals are closer to exponential disks. This might indicate two formation mechanisms, one primordial (as in the traditional picture), the other related perhaps to the merging assembly history or via disk instabilities through what is termed `secular' evolution.

Figure 19. The remarkable diversity of intermediate redshift spiral bulges in the Hubble Deep Fields as revealed in the analysis of Ellis et al (2000). (Top) Selected face-on spirals in the HDF with pixel-by-pixel BVI color distributions. The marked points represent various aperture selections which serve to define the mean bulge color; in each case the bulge remains the reddest part of the spiral galaxies. (bottom) V - I aperture color for bulges (open circles) and integrated color for ellipticals (filled circles) versus redshift. Bulges are generally more diverse with a mean color bluer than their elliptical counterparts. Curves illustrate that a continued infall of 5% by mass over 1-2 Gyr timescales could explain the observed trends.

Taking advantage of the HDF images, including those from NICMOS, Ellis et al (2000) have examined the color distribution for a large sample of spirals bulges of known redshift and compared these colors with their integrated equivalent for the HDF ellipticals. If bulges are miniature ellipticals formed at high redshift, one would expect similar trends. Interestingly, in the hierarchical picture, one expects bulges to be older and presumably redder than ellipticals (since the latter predominantly form from merged disk systems which most likely contain bulges as early merger remnants). Ellis et al (2000) find intermediate redshift bulges are the reddest part of a typical spiral but, surprisingly, they are often bluer than their elliptical counterparts and far less homogeneous as a population. Contamination from disk light is an obvious concern though simulations suggest only modest bias arises to redshifts where these trends become prominent. What could be responsible for this puzzling behavior? Evolutionary synthesis modelling suggest only a modest amount of star formation corresponding to continued infall of 5% by mass would be needed to explain the bluing.