Annu. Rev. Astron. Astrophys. 2009. 47: 159-210 Copyright © 2009 by Annual Reviews. All rights reserved

5. ELLIPTICALS

5.1. General description and identification

Elliptical galaxies are recognizable by their smooth, symmetric, and deceptively simple-looking appearance. Figure 13 shows some typical elliptical galaxies drawn from the SDSS, using the classifications reported by NED. In this case, we had to reject about one-third of the classifications as obviously incorrect. Figure 12 shows the distribution of their broad-band properties along with the S0s. They are concentrated, uniformly red, and follow a reasonably tight size-luminosity relation.

Figure 12. Distribution of elliptical and lenticular galaxy types in optical broadband properties. Similar to Figure 8, but now showing S0 galaxies (orange), E galaxies (red), dE galaxies (black) and cD galaxies (magenta).

Figure 13. SDSS images of elliptical galaxies, selected according to classifications in NED. The images are sorted by absolute magnitude in the horizontal direction, ranging between Mr - 5log10 h ~ -18.5 and -22 from left to right, and concentration (r90 / r50) in the vertical direction, ranging between 2.2 and 3.8 from the bottom to the top. Thus, the brightest, most concentrated ellipticals are in the upper right. The galaxies shown were selected randomly, but roughly one-third were replaced because their NED classifications were clearly incorrect. We left cases that could be considered ambiguous in the figure.

Among luminous ellipticals there are two discernible classes, those with and without cores (Section 5.3; Kormendy et al. 2009 and references therein). The ellipticals with nuclear cores, relative to those without, tend to be more luminous, have b / a closer to unity, have boxier isophotes, less rotational support, and more signs of triaxiality.

Among lower luminosity ellipticals there are at least three classes: low surface brightness, exponential galaxies typed as dE or sometimes Sph (such as NGC 205); high surface brightness, concentrated galaxies usually typed as cE (such as M32); and ultra-compact dwarfs with r₅₀ as small as 10 pc (Drinkwater et al. 2004). Kormendy et al. (2009) reviews the distinction between the first two classes, arguing that the cE population is most similar to the giant elliptical population. Ferrarese et al. (2006) presents the alternative argument for dE galaxies.

Relative to S0s, the surface brightness profiles of ellipticals appear to have no "edge," as S0s do. However, in their other properties - sizes, profile shapes, colors, symmetry, and smoothness - Es and S0s are very similar. Thus, the recent literature on very large samples is plagued with "elliptical" or "early-type" samples that actually include a fair number of S0s as well as early-type or edge-on spirals. Interestingly, these interlopers do not appear to affect many of the scaling relations we discuss here; nevertheless, they can be important in some circumstances and we emphasize again the importance of separating these two populations.

5.2. Structural trends

Although many astrophysicists typically think of ellipticals as de Vaucouleurs profile galaxies, in fact they have a range of Sérsic indices, whose values depend rather strongly on luminosity (as has been recognized for a long time; Caon, Capaccioli & D'Onofrio 1993, Prugniel & Simien 1997). The first row of Figure 6 shows this distribution, which reveals that selecting ellipticals using a strict cut in Sérsic index will exclude a significant (and luminosity-dependent) fraction of viable elliptical galaxy candidates. Kormendy et al. (2009) demonstrate this trend most clearly with a careful analysis of elliptical surface brightness profiles.

These structural trends are mostly independent of environment, as Figure 6 shows (Blanton et al. 2005a, Park et al. 2007). With the exception of the most luminous cases, red galaxies of a given luminosity are no more or less likely to be concentrated when they are found in dense regions. This result suggests that any process that turns spiral galaxies into ellipticals acts similarly in dense and underdense regions.

One shortcoming of the 2D Sérsic profile is that when examined carefully few ellipticals actually have precisely elliptical isophotes. One way of quantifying this non-ellipticity is to consider the residuals of the actual isophotes relative to perfect ellipses. If one Fourier transforms these residuals, the cos4 term is of order 1% of the isophotal radius for many ellipticals (Bender & Moellenhoff 1987). The fractional amplitude of this term a₄ / a is defined such that positive values are "diskier" than ellipses while negative values are "boxier." Generally, boxy galaxies are more luminous than disky galaxies and more likely to exhibit a core (e.g., Kormendy & Djorgovski 1989; also see Section 5.4).

5.3. Nuclear properties

The structure of the central regions of ellipticals are now observable for large samples using HST imaging (Ferrarese et al. 2006, Lauer et al. 2007a; see in particular the review of Kormendy et al. 2009). Although these analyses have varied in their details, they all find that the central stellar mass surface density profiles can reasonably be modeled as a power law (r) r^-. As defined by Lauer et al. (1995), galaxies with ~ 0.0-0.3 are classified as "cores," and galaxies with ~ 0.5-1.0 are classified as "cusps," or sometimes "power-law," "extra-light," or just "coreless." There are of course galaxies in between the extremes (Rest et al. 2001).

The ellipticals with cores are the most luminous galaxies, with M_V - 5log₁₀ h < -21, and correspondingly have distinctly higher global Sérsic indices, boxier isophotes, slower rotation, and more triaxiality (Ferrarese et al. 2006, Lauer et al. 2007a, Kormendy et al. 2009). An oft-used model for the shapes of elliptical galaxy cores is the "Nuker" model introduced by Lauer et al. (1995), which is close to a broken power law. This model does not accurately reflect the surface brightness profiles of ellipticals at galactic radii larger than r₅₀, which tend to be closer to Sérsic profiles. Consequently, several researchers (e.g., Kormendy 1999, Trujillo et al. 2004, Ferrarese et al. 2006) have suggested the use of a Sérsic model altered to transition to a power law near the center. These cores are thought to be scoured by the gravitational effects of binary black holes during mergers (for a review see Kormendy et al. 2009).

For -21 < M_V - 5log₁₀ h < -18, ellipticals tend to be cuspy (Ferrarese et al. 2006). Relative to the Sérsic models, these coreless galaxies often actually have an "extra light" component. The central light typically has isophotes that deviate from ellipses, being "diskier" (a₄ / a > 0). If cuspy elliptical galaxies form in mergers, the diskiness of the central light components suggest that the last of these mergers should have been "wet," with dissipative processes triggering star-formation and raising the central stellar surface density, and swamping the effects of black hole scouring (Kormendy et al. 2009).

5.4. Fundamental plane and dynamics

The SDSS survey has provided an unprecedented sample of galaxies with reliable spectroscopic velocity dispersions, allowing a detailed analysis of the fundamental plane (FP; Djorgovski & Davis 1987, Faber et al. 1987) - the tight relationship among r₅₀, half-light surface brightness I₅₀, and velocity dispersion . Bernardi et al. (2007) selected nearby early-types according to concentration, color and emission line strength and found that for a relationship like r₅₀ I₅₀, the best fit is ( ~ 1.3, ~ -0.76), in only mild disagreement with the previous findings of Jørgensen, Franx & Kjaergaard (1996) based on more local samples ( ~ 1.25, ~ -0.82).

The FP differs from a naïve prediction based on the virial theorem. From dimensional analysis, M = c ²R / G, where c is a structural constant that depends on the orbital structure, the mass density profile, the surface brightness profile, and an appropriately weighted mass-to-light ratio, R is a characteristic radius, and G is the gravitational constant. Indeed, the rotation of these parameters into the -space of Bender, Burstein & Faber (1992) was motivated by this virial interpretation. In this case, for a constant mass-to-light ratio the FP parameters become ( = 2, = -1), rather different than the observed ones. Several possible explanations exist for the deviation from this relation: (a) a variation of the stellar mass to dynamical mass ratio; (b) a variation of the stellar mass-to-light ratio; or (c) "non-homology," that is, a variation in the nature of ellipticals that changes c. Most analyses now favor the first explanation, with a dynamical to stellar mass ratio that scales as (CREN PUT THIS EQUATION IN)

First, spectral synthesis modeling suggests that very little of that variation can be attributed to the stellar mass-to-light ratio (Padmanabhan et al. 2004, Bernardi et al. 2006, Trujillo, Burkert & Bell 2004, Proctor et al. 2008). Second, detailed modeling of the 2D dynamics using integral field measurements generally finds a variation of M_dyn / M_∗ that is quantitatively similar to the virial estimates that assume homology (Cappellari et al. 2006, Thomas et al. 2007). Third, direct constraints on mass density profiles with strong gravitational lensing indicate that the dynamical mass to light ratio is changing (Bolton et al. 2008).

Nevertheless, as Cappellari et al. (2006) notes, it is curious that the simple virial estimate yields the same mass-to-light ratio trends as more complex and direct modelling. After all, the assumptions of homology must be wrong in detail, as Trujillo et al. 2004 point out. Cappellari et al. (2006) further caution that the virial estimate only works if the de Vaucouleurs estimate of r₅₀ is used - notwithstanding the fact that the de Vaucouleurs model is incorrect for many of these galaxies!

The FP is a weak function of environment, as with all other scaling relationships. For example, Park et al. (2007) shows that the Faber-Jackson relationship does not vary significantly. The FP analysis of Bernardi et al. (2006) detected a 0.1 mag surface brightness shift between the highest and lowest density subsamples, which they estimated was consistent with an age difference of 1 Gyr between cluster and field ellipticals (see Section 5.7).

The apparent simplicity of the FP relationship masks a good deal of variability and complexity in elliptical dynamics. For example, ellipticals are known to occasionally contain multiple kinematically distinct components (Efstathiou, Ellis & Carter 1982), often in the form of rotating disks around the core (Bender 1988). In addition, the kinematics is often misaligned with the photometry or twists as a function of radius, often interpreted as a sign of triaxiality (e.g., Binney 1978, Schechter & Gunn 1979). The recent analysis of SAURON data has greatly increased our understanding of the variety and incidence of kinematic substructures and other features (e.g., Krajnovic et al. 2008).

The FP also masks the importance of rotation to the global dynamics of many ellipticals. Traditionally, this importance has been quantified by the ratio of rotation speed to central velocity dispersion, V/ (originally due to Illingworth 1977, but see the latest discussion of Binney 2005). Using SAURON data, Cappellari et al. 2007 argue for a different estimator _r = <R|V|> / <R (V² + ²)^1/2>, where the averages are weighted by flux. This measurement requires 2D velocity maps but is apparently a better tracer of the angular momentum content than V / , which is more significantly affected by kinematically decoupled cores as well as projection effects.

Figure 14 shows the distribution of _r from Emsellem et al. (2007), for the SAURON sample of Es and S0s. The upper left panel shows the relationship between angular momentum content and virial mass M_vir. The red points are the "slow rotators" according to the SAURON definition (_r < 0.1). For comparison, in the bottom two panels we also show the distribution of the isophotal shape parameter a₄ / a (Section 5.2) as a function of both M_vir and _r. The upper right panel of Figure 14 shows _r versus the observed ellipticity within r₅₀. Overplotted is the approximate curve for an oblate rotator with an isotropic velocity dispersion, seen edge-on.

Figure 14. Distribution of E/S0 galaxy properties from the SAURON sample of Emsellem et al. (2007). Black symbols indicate "fast rotators," while red points indicate "slow rotators". Filled points indicate optically classified elliptical galaxies, while open square symbols indicate S0s. Mvir indicates the virial mass; r indicates the angular momentum content as defined by Emsellem et al. (2007) and described in Section 5.4; is the ellipticity within r50; and a4 / a is the boxy/disky parameter (positive is disky, negative is boxy). The solid grey curve corresponds to an oblate rotator with an isotropic velocity dispersion observed edge-on.

The most massive systems have a strong tendency to be the slowest rotators, the closest to perfect ellipsoids, and the most axisymmetric. Relative to the "fast rotators," these slow rotators tend to have higher masses, flat or falling _r profiles, less cuspy centers (Section 5.2), more boxy isophotes, more kinematically decoupled cores, and greater kinematic misalignments (indicating triaxiality). It is difficult to tell with the available data whether there are some independent relationships among these properties that are responsible for the others. However, the overall trend is suggestive of the importance of major dry mergers in the formation of these systems (e.g., Hernquist 1993), but perhaps not minor dry mergers (Burkert et al. 2008).

The lower mass systems have a stronger tendency to be "disky" and are faster rotators. They tend to have disky isophotes as well as aligned kinematics and photometry without twists (consistent with the importance of rotation; Cappellari et al. 2007). Detailed analysis of their 2D dynamics suggests that they tend to have anisotropic velocity dispersions, as their positions in the upper right panel of Figure 14 suggests (Cappellari et al. 2007). Interestingly, the S0s, while always fast rotators, do not appear to be particularly distinct from fast rotator ellipticals in their dynamics (though the SAURON analysis is limited to radii r₅₀).

Thus, though the FP is simple and appears to remain relatively constant with environment, there are still only small samples available with truly detailed 2D dynamics ( 100 nearby ellipticals total). Thus, nobody has explored whether these more detailed properties vary with environment even as the FP remains relatively constant; such a detection would be an important constraint on formation mechanisms.

In this section (as in this review as a whole) we have focused mainly on the properties of the luminous galaxies. Just as dwarf disk galaxies deviate from the Tully-Fisher relation, the dE galaxies deviate from extrapolations of the FP (Geha, Guhathakurta & van der Marel 2003, van Zee, Skillman & Haynes 2004), while cE galaxies do not (Kormendy et al. 2009). However, low luminosity ellipticals do follow a more general, but still regular, relationship described by Zaritsky, Gonzalez & Zabludoff (2006), a "fundamental manifold" for spheroids.

5.5. Brightest cluster galaxies and cD galaxies

Among ellipticals, brightest cluster galaxies (BCGs) and cD galaxies form a special class (Morgan & Lesh 1965, Sandage 1972). BCGs are usually defined as the highest optical luminosity galaxy in any reasonably massive cluster (> 10¹⁴ M). A large fraction of BCGs have a larger associated distribution of stars that extends out into the host cluster, called the "cD" envelope for historical reasons, but sometimes referred to as the "intracluster light" (ICL).

In Figure 12, BCGs and galaxies with cD envelopes are shown as the magenta dots. BCGs are luminous and massive, with a log-normal distribution of stellar mass with a mean of M_∗ ~ 2 × 10¹¹ h^-2 M and dispersion _lnM ~ 0.4 (Lin & Mohr 2004, Hansen et al. 2009, Yang, Mo & van den Bosch 2008). These luminosities are a function of the host cluster mass, with L_K M_h^0.2-0.3 for massive clusters (Lin et al. 2004, Hansen et al. 2009) and a steeper relationship at lower masses (Yang et al. 2007, Brough et al. 2008). Thus, as the total halo mass and luminosity rises, the fractional contribution of the BCG decreases.

Tremaine & Richstone (1977) found hints that BCGs were not merely the brightest members of a randomly sampled luminosity function for each cluster. In particular, the second brightest galaxy is normally at least 0.8 mag fainter than the first. In contrast, if the luminosities randomly sampled almost any conceivable distribution, then the expected magnitude "gap" between the first and second ranked cluster galaxies would be about the same or less than _{ln M}, or about 0.4 mag (Vale & Ostriker 2008, Loh & Strauss 2006). This "gap" in magnitude suggests an anti-correlation among galaxy luminosities within the same cluster.

The dry merger, or cannibalism, scenario for the growth of BCGs is consistent with the last two results. First, the dynamical friction time scales for galaxies tends to increase as the host cluster mass increases, explaining why a smaller fraction of the cluster luminosity is accreted onto the central galaxy as mass increases (Cooray & Cen 2005). Second, if the second ranked galaxy is close in mass to the BCG, it is drawn preferentially to the BCG and will merge with it. This process will open a gap between the masses and luminosities of the first and second ranked galaxies (for recent analyses see Loh & Strauss 2006, Milosavljevic et al. 2006).

Other hints that BCGs are special come from comparing their fundamental plane relation to that of other ellipticals. This comparison is complicated by the fact that BCGs are much brighter than the typical elliptical. Oegerle & Hoessel (1991) reported that BCGs roughly followed the fundamental plane defined by lower luminosity galaxies. However, above M_r - 5log₁₀ h ~ -22.3 and ~ 280 km s^-1, the velocity dispersion of BCGs becomes a weaker function of luminosity, causing a deviation from the extrapolation of the fundamental plane. This weakening has been verified in modern data sets (Lauer et al. 2007b, von der Linden et al. 2007, Bernardi et al. 2007, Desroches et al. 2007). However, at luminosities where BCGs and elliptical populations overlap, the differences in their velocity dispersions are < 5% (von der Linden et al. 2007, Desroches et al. 2007).

Several large studies based on SDSS also indicate that the most luminous BCGs typically are larger than the radius-luminosity relation for typical ellipticals would predict (r₅₀ L, with ~ 0.6). Either BCGs and other ellipticals lie on different loci in r₅₀ - L space, or the locus they both live on is curved such that r₅₀ is a stronger function of L at high luminosity. In fact, there is evidence that both effects exist. At higher luminosities, the non-BCG ellipticals appear to deviate from a power-law relation (Desroches et al. 2007, von der Linden et al. 2007), with a local power law varying from 0.5 at M_r - 5log₁₀ h ~ -20 to 0.7 at M_r - 5log₁₀ h ~ -24. Simultaneously, BCGs appear to be larger than non-BCGs at a given magnitude, but the literature has not converged on the details. von der Linden et al. (2007) find that the BCGs are all 10% larger on average than non-BCG ellipticals, constant with luminosity, defining a slightly different fundamental plane. Desroches et al. (2007), meanwhile, find that the BCGs define a steeper relationship than the ellipticals, and that at M_r - 5 log₁₀ h ~ -22 the typical BCG is smaller than the typical elliptical. Bernardi et al. (2007) also find a steeper slope for BCGs, but that the relationships converge near M_r - 5log₁₀ h ~ -22.

These results suffer from two major systematic effects First, there are some significant ambiguities in defining a "BCG", particularly for cluster catalogs (e.g., Miller et al. 2005) based on an incomplete redshift survey like the SDSS (even 5%-10% incompleteness matters; see von der Linden et al. 2007). Second, BCGs are large objects on the sky, and their photometry is not handled correctly by the SDSS (Section 3; Lauer et al. 2007b). All of the work cited here either reanalyzes the images (Bernardi et al. 2007, Desroches et al. 2007) or corrects the catalog parameters in an ad hoc way (von der Linden et al. 2007). No cross-comparison of their analyses has been published.

Many BCGs have an extended distribution of stars that is inconsistent with a single de Vaucouleurs profile extrapolated from small radii, or even with a more general Sérsic profile, known as a cD envelope or the ICL. Mihos et al. (2005) found that the Virgo Cluster has a particularly dramatic ICL component, with visible streams and other features that might suggest a tidal stripping or merging scenario for its formation. For a large sample of clusters, Gonzalez, Zabludoff & Zaritsky (2005) recently showed that detected ICL components are often discernably separate entities from the host BCGs, with well defined transitions in the surface brightness profile, axis ratio, and position angle. Zibetti et al. (2005) stacked images of many different SDSS clusters and statistically detected the ICL component. Zibetti et al. (2005) and a study of individual detections by Krick & Bernstein (2007) both conclude that of order 5%-20% of the total cluster optical light within 500 kpc or so comes from the ICL; the results of Gonzalez, Zabludoff & Zaritsky (2005) imply even more. If so, the stellar mass in the ICL is comparable to that in the BCG itself (or possibly 4-5 times the BCG mass if the analysis of Gonzalez, Zabludoff & Zaritsky (2005) is correct). Its existence appears consistent with the merger hypothesis for BCGs themselves.

The dynamics of the ICL component is poorly known in most cases, and it is unclear to what degree is represents the dynamics of the cluster as a whole. Some BCGs show the rising velocity dispersion profile at large radius expected from the cluster dark matter distribution (most notably those found in Dressler 1979, Kelson et al. 2002). However, the large studies of Fisher, Illingworth & Franx (1995) and Loubser et al. (2008) find that only a minority of cD envelopes have such a profile. Interestingly, detectable rotation is seen in a number of cD envelopes, with V / ~ 0.3.

5.6. Deviations from smooth profiles

In addition to boxiness and diskiness, elliptical galaxies often show other deviations from smooth elliptical isophotes, at least at very faint levels (Kormendy & Djorgovski 1989). The most common deviation from smoothness is due to dust features; as we outline in Section 5.8, many elliptical galaxies have a small but detectable amount of cold gas and associated dust. For example, Colbert, Mulchaey & Zabludoff (2001) find that about 75% of ellipticals have detectable dust extinction in the optical, irrespective of environment.

Elliptical galaxy profiles also often reveal shells or ripples, and other signs of recent interaction (Athanassoula & Bosma 1985). While Malin & Carter (1983) found that about 10% of all ellipticals had detectable features, perhaps unsurprisingly it appears that deeper observations reveal structure in a larger fraction. van Dokkum (2005) find that 70% of ellipticals have detectable tidal features down to 27-28 mag arcsec^-2. For about 20% of those detections, a secondary object that appears responsible for the features is observed. These observations suggest that some growth of elliptical galaxies occurs through merging; however, precisely how much is difficult to infer from such observations. Colbert, Mulchaey & Zabludoff (2001) also searched their sample for tidal features, finding that the fraction of disturbed ellipticals is a strong function of environment, more common in isolated regions than in groups and clusters. Finally, ellipticals with more fine structure tend to be slightly bluer than those with less (Schweizer & Seitzer 1992) and perhaps lie on the bright side of the fundamental plane (Michard & Prugniel 2004).

A final deviation from a purely elliptical profile is that of isophotal twists. These features, which can be caused by triaxiality or by tidal effects, appear to be rare: a survey of Virgo ellipticals using HST / ACS reported only seven cases of isophotal twisting (some of these very marginal or related to central disks) out of 100 observed early-type galaxies (Ferrarese et al. 2006).

5.7. Stellar populations

The spectra of elliptical galaxies are dominated by emission from the surfaces of stars, typically K giants but comprising some mixture of stellar types depending on the age, metallicity, and metal abundances of the stellar population. For this reason ellipticals all have nearly the same optical broad-band color, with a weak dependence of color on galaxy luminosity (and equivalently, stellar mass or velocity dispersion). This dependence is due to both age and metallicity trends as a function of mass, as detailed spectroscopic analyses reveal. Renzini (2006) review the subject in detail. However, in brief, Balmer absorption lines such as H, H and H tend to trace age in old stellar populations, while metal-line indices such as <Fe>; and Mg b yield information about the metallicity and abundances in the stellar atmospheres.

Based on high signal-to-noise optical spectroscopy of morphologically selected E and S0 galaxies, Thomas et al. (2005) show that the metallicity (or iron abundance) and [ / Fe] ratio are both correlated with velocity dispersion (or mass). As found previously, the Mg b indicator is much more tightly correlated to mass than <Fe> (e.g. Worthey 1998 and references therein). Similar results have also been found with large samples of (lower signal-to-noise) SDSS galaxies (e.g., Eisenstein et al. 2003, Bernardi et al. 2006, Gallazzi et al. 2006, Jimenez et al. 2007).

Many ellipticals show evidence for recent star formation in their optical spectra, and the incidence is higher in lower-mass galaxies (Trager et al. 2000, Thomas et al. 2005). These conclusions have been strengthened by recent observations with GALEX. Yi et al. (2005), Kaviraj et al. (2007), and Schawinski et al. (2007) have all found evidence for recent ( 1 Gyr) star formation in 15%-30% of morphologically selected elliptical galaxies, which accounts for 1%-3% of the stellar mass of the galaxy.

There appears to be some variation of these star-formation histories with environment. Figure 15 shows H, <Fe> and Mg b as a function of velocity dispersion for E/S0 galaxies from Thomas et al. (2005). Filled symbols correspond to galaxies in dense regions, while unfilled symbols are for galaxies in underdense regions. They appear to lie on somewhat different loci. According to Thomas et al. (2005), based on these results the field early-type galaxies are on average ~ 2 Gyr younger and slightly more metal-rich, while both populations show comparable [ / Fe] ratios. Bernardi et al. (2006) found qualitatively similar results, though with a rather different analysis technique. Schawinski et al. (2007) further found that early-types with recent star formation were more prevalent in low density environments, even after controlling for the fact that lower-mass (massive) galaxies are found preferentially in underdense (dense) regions.

Figure 15. Dependence of E/S0 spectral indices on and environment from Thomas et al. (2005). Filled blue circles are low density points; open red squares are high density points. Top panel is the Mg b indicator ( abundance), middle panel is the <Fe> indicator (iron abundance), and bottom panel is H (an age indicator). Typical uncertainties in the measurements are shown in the lower right corners.

The nature of elliptical galaxy stellar populations is an old, storied, and quite controversial one, and the limited space here cannot do it justice. However, we do note several investigations that have reached conclusions in conflict with those described above. First, recent Spitzer observations of the mid-IR (9-12 µ) spectra of ellipticals in nearby clusters have found that the vast majority are consistent with being purely passively evolving systems (Bressan et al. 2006, Bregman, Temi & Bregman 2006). Second, Trager, Faber & Dressler (2008) find that early-type galaxies in the Coma cluster show the same mean age as field ellipticals (though less scatter). The exact nature of the recent star formation in ellipticals, and its variation with environment, therefore appears to be in some doubt.

5.8. Cold gas content

Elliptical galaxies contain a small amount of cold atomic and molecular interstellar gas (Faber & Gallagher 1976). When detected, the gas comprises 1% of the total mass of the system (Knapp, Turner & Cunniffe 1985), so it is not a dominant baryonic component.

Early observations showed that the H i detection rates in ellipticals are several times higher in objects with morphological fine structure such as visible dust lanes, shells, and ripples (Bregman, Hogg & Roberts 1992, van Gorkom & Schiminovich 1997), which themselves occur more frequently in field ellipticals (Malin & Carter 1983, Schweizer et al. 1990). This result indicates a close connection between the gas properties of a galaxy and visible signs of interaction.

The morphology of the H i gas in early-type galaxies generally falls into two categories: most have disk- or ring-like structures with regular kinematics, extending up to ~ 200 kpc in diameter, while in others the H i appears in an irregular, tail-like structure or in individual clouds that are offset from the center of the galaxy (Morganti et al. 2006, Oosterloo et al. 2007). Morganti et al. (2006) surveyed a representative subset of the SAURON galaxies and detected H i emission in 70% of the sample, with gas masses ranging from 10⁶ to 10⁹ M. They found that galaxies with H i disks had the most ionized gas emission (see also Serra et al. 2008), but that the H i properties were uncorrelated with either the age of the stellar population or the stellar kinematics (fast vs. slow rotators). Thus, most early-types seem to accumulate at least some gas, irrespective of their evolutionary past.

Surveys of molecular gas in elliptical galaxies have also rapidly advanced, although the samples remain relatively sparse. Combes, Young & Bureau (2007) reported that 28% of the SAURON sample has detectable CO. Similarly, Sage, Welch & Young (2007) detected CO emission in 33% of a volume-limited sample of field ellipticals; their survey was designed to detect at least 1% of the gas expected to have been returned to the interstellar medium by evolved stars in a Hubble time. From interferometric mapping, the molecular gas is typically in a rotationally supported disk 2-12 kpc in diameter (Young 2002, Young 2005). In many cases these gas disks contain higher specific angular momentum than the stars, or are counter-rotating with respect to the stars, suggesting an external origin (Young, Bureau & Cappellari 2008).

The analysis by Combes, Young & Bureau (2007) of CO-rich early-type galaxies also found that their sample obeyed the Kennicutt-Schmidt relation (Section 3.5; Kennicutt (1998b)) for disk and starburst galaxies, but at gas and star formation rate surface densities two orders of magnitudes lower. Further evidence for ongoing star formation in early-type galaxies was presented by Young, Bendo & Lucero (2008), who studied a sample of CO-rich E/S0 galaxies and found good spatial correspondence between the CO, 24 µm, and radio continuum emission, from which they concluded that the 24µm emission is predominantly due to star formation (rather than AGN or circumstellar in origin).

The nature of the gas in ellipticals is mysterious. Stellar recycling arguments suggest that they should contain about ten times the atomic hydrogen mass that they actually do. Additionally, the gas mass is not correlated with the stellar mass, which it would be if recycling were an important source (Ciotti et al. 1991, Sage, Welch & Young 2007). Where the recycled gas goes is an unsolved problem.