Annu. Rev. Astron. Astrophys. 1989. 27: 87-138 Copyright © 1989 by . All rights reserved

4. ELLIPTICAL AND S0 GALAXIES

4.1. Discovery and Properties of the Hot Interstellar Medium

The absence of tracers of a cold interstellar medium (ISM) in most early-type galaxies (e.g. Sandage 1957, Gallagher et al. 1975, Faber & Gallagher 1976) has prompted the construction of models to explain the removal of the gas shed from stars during their evolution (Mathews & Baker 1971, Bregman 1978, White & Chevalier 1983), which at the present rates would amount to 10^9-10 M in a Hubble time in galaxies with an optical luminosity of 10^10-11 L (e.g. Faber & Gallagher 1976). Now, X-ray observations have revealed this long-sought interstellar medium.

A hot gaseous halo had been known for some time to be associated with the Virgo cluster and M87 (Kellogg et al. 1975, Mitchell et al. 1976, Malina et al. 1976, Serlemitsos et al. 1977, Gorenstein et al. 1977, Fabricant et al. 1978, Lea et al. 1979, Canizares et al. 1979), but the presence of a hot gaseous medium in more normal ellipticals was revealed for the first time by the Einstein survey of the Virgo cluster (Forman et al. 1979). Five early-type galaxies in Virgo were detected in X rays in this first paper, with X-ray luminosities of 5 - 70 × 10³⁹ erg s^-1, over a factor of ~ 100 less luminous than M87. The X-ray image of one of them, M86, which has a large velocity relative to the Virgo cluster mean, was particularly important for establishing the gaseous nature of the emission. The X-ray isophotes of this galaxy appear significantly asymmetric with respect to the optical image (see Figure 7 of the Annual Reviews paper of Forman & Jones 1982), which suggests the presence of a gaseous halo of ~ 6 × 10⁹ M that experiences ram pressure stripping in its approach to the cluster core (Forman et al. 1979; see also Fabian et al. 1980, Takeda et al. 1984, Forman et al. 1984b). Another galaxy in Virgo, NGC 4472 (Figure 5), was later found to have a similar X-ray morphology (Forman et al. 1985, Trinchieri et al. 1986). Forman et al. (1979) also concluded that the hot gas detected in the Virgo galaxies had to be indigenous, since the cooling time of the intracluster medium is too long to make accretion onto the galaxies possible.

Figure 5. The Einstein IPC X-ray map of NGC 4472. Notice the asymmetric halo.

Subsequent X-ray observations and data analysis led to the detection of many more early-type galaxies in Virgo, in poor clusters and groups, and in the field, showing the ubiquity of their X-ray emission [Kriss et al. 1980 , 1983, Biermann et al. 1982, Biermann & Kronberg 1983, Nulsen et al. 1984, Dressel & Wilson 1985, Forman et al. 1985, Trinchieri & Fabbiano 1985, Mason & Rosen 1985 (who report EXOSAT observations), Killeen et al. 1986, Trinchieri et al. 1986, Canizares et al. 1986, 1987, Killeen & Bicknell 1988a]. Typical X-ray luminosities in the Einstein band range from 10³⁹ to 10⁴² erg s^-1, except for group or cluster-dominant galaxies, which tend to be more luminous (in the 10⁴³ erg s^-1 range). The sources detected with better signal to noise are clearly extended, with radii of ~ 40-70 kpc, generally comparable with those of the stellar distribution. This spatial extent is not in itself proof of the gaseous nature of the emission, because none of these galaxies is close enough to allow the spatial resolution of single X-ray sources with the Einstein instruments, as in the case of nearby spirals. Moreover, the radial distributions of the X-ray and optical surface brightnesses in those galaxies for which both profiles are available tend to follow each other, at least if one excludes the outermost radii, where the uncertainties in the field background subtraction and possible environmental effects could affect the profiles and perhaps the inner core regions (Figure 6; Trinchieri 1986, Trinchieri et al. 1986, Killeen & Bicknell 1988a). However, other convincing indications were found for the presence of a hot gaseous component in at least the brightest galaxies.

Figure 6. Radial profile of the X-ray surface brightness of NGC 4472 (the points (Trinchieri et al. 1986)] together with the optical profile from King (1978)]. Note the north-south asymmetry at large radii.

One indication is given by the spectral characteristics of the X-ray emission, which in those galaxies bright enough to be so analyzed can be fitted with fairly low emission temperatures [kT ~ 0.5 - 2.0 keV (Forman et al. 1985, Trinchieri et al. 1986, Killeen & Bicknell 1988a)]. This is in contrast with the bright spiral galaxies, which have harder spectra, consistent with the presence of a population of binary X-ray sources and young supernova remnants (kT > 2 keV (Fabbiano & Trinchieri 1987)]. Another indication is given by a comparison of the global X-ray and optical properties of early- and late-type galaxies surveyed with the Einstein satellite. This comparison shows that the X-ray and optical luminosities of elliptical and SO galaxies follow a steeper correlation ( ~ 1.6) than does the linear one observed in spirals, which suggests that a different emission mechanism or at least an additional emission component is present in early-type galaxies (Fabbiano 1984, Forman et al. 1985, Trinchieri & Fabbiano 1985, Canizares et al. 1987). Those galaxies more luminous in X rays for which we have some spectral information are also those generally presenting excess emission relative to the relationship of the spirals (Figure 7). Therefore, this "excess" X-ray emission can be attributed to a hot gaseous component.

Figure 7. X-ray (in erg s-1) and optical (in solar units) luminosities of E and SO galaxies observed in X rays (from Canizares et al. 1987). The crosses are the galaxies in A1367 discussed by Bechtold et al. (1983). The solid line (Lspiral) is the best-fit line of early-type spirals from Fabbiano et al. (1988); the horizontal solid line delimits the maximum X-ray luminosity of spiral galaxies. The dashed line (LSN) is the estimate by Canizares et al. (1987) of supernova heating of the halos; the dotted line (FJT) is the estimate by Forman et al. (1985) (see Section 4.2); and the curves labeled 1, 2, 3, and 4 are the models of Sarazin & White (1988) for massive halos and supernova heating, supernova heating without massive halos, massive halos but no supernova heating, and no massive halos and no supernova heating, respectively.

How much of the X-ray emission of early-type galaxies is due to the hot gaseous component and how much can instead be ascribed to a population of evolved galactic sources, similar to those found in late-type galaxies, is still a matter of debate. Forman et al. (1985; see also Jones 1987, Forman & Jones 1988) conclude that all of the emission of early-type galaxies with absolute blue magnitude M_B - 19 is due to gaseous halos. Trinchieri & Fabbiano (1985), Canizares et al. (1987), and Fabbiano et al. (1989) note, however, that the data require the presence of a gaseous component only in the more X-ray-luminous galaxies, whereas the emission of galaxies less luminous than ~ 10^40.5-41 erg s^-1 (these galaxies can be optically very bright, with M_B ~ - 22) can be easily explained with a population of accreting low-mass binary sources. These different conclusions are partly due to the different choices of benchmarks for the expected contributions of discrete X-ray sources, and they illustrate the uncertainties inherent in the present data. Forman et al. (1985) used the Einstein X-ray observations of Cen A [NGC 5128 (Feigelson et al. 1981)] as a benchmark. However, the complexity of the X-ray emission of Cen A makes the evaluation of its "binary" component uncertain. Trinchieri & Fabbiano (1985) and Canizares et al. (1987) based their estimate on an extrapolation of the X-ray properties of the bulge of M31, which is dominated by pointlike sources, and of its globular cluster system (Van Speybroeck et al. 1979, Van Speybroeck & Bechtold 1981, Long & Van Speybroeck 1983). This choice has been criticized by Forman & Jones (1988), who remark that the bulge of M31 could be peculiarly bright in X rays. However, the stellar content of this bulge is indistinguishable from that of well-studied elliptical galaxies such as NGC 4472 (Oke et al. 1981, Faber 1983, Bohlin et al. 1985), and therefore it represents our best nearby and well-studied example of such a system (Canizares et al. 1987). More recently, Fabbiano et al. (1989) have performed a direct comparison of the global X-ray and optical luminosities of ellipticals and SOs with those of early-type spirals and found that only in galaxies with X-ray luminosities greater than 10⁴¹ erg s^-1 is there a departure from a linear relationship between X-ray and optical luminosities, such as that observed in spirals, which then requires an additional, nonstellar emission component (see Figure 7).

An uncontroversial way to resolve this issue would be to perform a spectral analysis of early-type galaxies of different luminosities, since the spectral signatures of binary X-ray sources and gaseous halos differ, in particular, the X-ray spectrum of the bulge of M31 is typical of low-mass binary X-ray sources (Fabbiano et al. 1987b, Makishima et al. 1989) and differs from those of bright early-type galaxies. This analysis was attempted by Canizares et al. (1987) but proved inconclusive because of the poor statistics of the data. The resolution of this issue will have to wait for future, more sensitive X-ray observations.

There is a consensus, however, that the hot gaseous component dominates the X-ray emission of the more luminous galaxies. The X-ray data can then be used to derive the physical properties of this gas, such as central density, cooling time, and total mass. For the brightest galaxies, radial profiles of density and cooling time can also be derived. Two approaches have been followed (e.g. Fabian et al. 1981, Canizares et al. 1983, 1987, Fabricant & Gorenstein 1983, Stewart et al. 1984a, Forman et al. 1985, Thomas et al. 1986, Trinchieri et al. 1986, Killeen & Bicknell 1988a). One is to use an "onion skin" technique to deproject the X-ray surface brightness and to infer the gas density distribution directly from the data. The other is to assume a parameterization of the form

(1)

for the X-ray surface brightness profile (where r is the radial distance from the centroid of the X-ray surface brightness distribution, and a_X is an X-ray core radius) and then from here derive the deprojected electron density profile n_e(r) [and therefore the gas density _gas(r)] under the assumption that the gas is isothermal, i.e.

(2)

Fits to the X-ray data generally give ~ 0.4 - 0.6 [i.e. a radial dependence of the X-ray surface brightness of the type S_X(r) r^-(1.4-2.6) and of the electron density n_e(r) r^{- (1.2 - 1.8)}], although departures from simple power laws and from spherical symmetry are common (see Figure 6). The parameter a_X is less well defined but could be of the order of a few kiloparsecs or less.

Estimates of the gas parameters made by Canizares et al. (1987; see also Forman et al. 1985, Trinchieri et al. 1986, Thomas et al. 1986) give typical central densities of ~ 0.1 cm^-3, central cooling times ranging between 10⁶ and 10⁸ yr, and gas masses ranging between 10¹¹ and 10⁸ M for galaxies detected in X rays. The amount of hot gas detected in bright galaxies is therefore consistent with the amount expected assuming the present gas injection rates over a Hubble time (Forman et al. 1985), although this could be a coincidence, since the mass shed by the first generations of stars might have been larger than the present rate (Renzini & Buzzoni 1986, Jones 1987, Loewenstein & Mathews 1987a). However, upper limits on the gas masses can be as low as 10⁶ M, showing a deficiency of hot interstellar medium in the less luminous galaxies. It should also be noted that the density and mass estimates given above are likely to be upper limits for these galaxies, since the X-ray emission of these galaxies could contain a sizable contribution from the "discrete source" component.

4.2. Physical Status of the Halos

Many authors have tried to understand the present status and the evolution of hot gaseous halos by comparing the X-ray observations with different theoretical scenarios. The temperature of the halos suggests the need for heating in addition to that supplied to a static halo by the same gravitational field experienced by the stars (e.g. Fabbiano 1986a). The high X-ray luminosities and corresponding large gas densities of some galaxies rule out the existence of galactic winds (Mathews & Baker 1971, Faber & Gallagher 1976, MacDonald & Bailey 1981, White & Chevalier 1983) because the mass supply from normal stellar processes falls several orders of magnitude short of that required to replenish such a wind (Nulsen et al. 1984, Forman et al. 1985, Sarazin 1986, Loewenstein & Mathews 1987a, Sarazin & White 1988). Moreover, the radial dependence of the gas density (_gas r^-1.5 on average) is flatter than that expected of a wind [ _gas(wind) r^-2 (Forman et al. 1985)]. All of this points to the need for a confining mechanism.

There are two indications that the halos are "hotter" than the stars. One is given by a comparison between the temperature inferred from the X-ray spectral fits [T ~ (5.8 - 23) × 10⁶ K, with best fits of ~ 1.2 × 10⁷ K] and that predicted under the assumption of thermal equilibrium with the stellar component. For typical line-of-sight velocity dispersions of ~ 150 - 300 km s^-1 (e.g. Whitmore et al. 1985), and under the assumption of isotropy, the latter would be given by T = µm_p ² / k ~ 1.4 - 5.4 × 10⁶ K, which is less than the best-fit temperature and probably near or below the lowest allowed by the data (Mathews & Baker 1971; see also Fabbiano 1986a, Killeen & Bicknell 1988a). The other indication is given by the radial dependence of the gas density as compared with that of the stellar light. Since the X-ray and optical surface brightness profiles tend to follow each other, and since the X-ray bremsstrahlung emission measure is a function of n_e², it is easy to see that _gas² ~ _stars. If the gas and the stars are two isothermal spheres experiencing the same gravitational potential, this implies that T_gas ~ 2T_stars (Cavaliere & Fusco-Femiano 1976 ; see Killeen & Bucknell 1988a). However, other heating mechanisms in addition to the motion of the mass-losing stars are readily available. They include supernova heating and gravitational heating. The latter would result from the dense gas cooling radiatively and then falling in the potential well in accreting "cooling flows" (Mathews & Baker 1971, MacDonald & Bailey 1981, Nulsen et al. 1984, White & Chevalier 1984, Sarazin 1986, 1987a, b, Thomas et al. 1986, Mathews & Loewenstein 1986, Canizares 1987, Fabian et al. 1987a, Canizares et al. 1987, Loewenstein & Mathews 1987a, Sarazin & White 1987, 1988, Umemura & Ikeuchi 1987, Vedder et al. 1988).

Nonthermal heating by the relativistic electrons in radio sources (e.g. M87) and heating by conduction from the surrounding intracluster medium have also been suggested (Tucker & Rosner 1983, Bertschinger & Meiksin 1986, Rosner & Tucker 1989). Once heated, the gaseous halos must be confined within their emitting volumes. This can be achieved by the pressure of an external medium (e.g. Fabian et al. 1980, Binney & Cowie 1981, Forman et al. 1984b, White & Chevalier 1984, Vedder et al. 1988), by radiative cooling (e.g. MacDonald & Bailey 1981, White & Chevalier 1984), and by gravity (e.g. Bahcall & Sarazin 1977, Mathews 1978, Nulsen et al. 1984, Forman et al. 1985; see Canizares et al. 1987).

Figure 7 shows different model predictions compared with the X-ray/optical scatter diagram of early-type galaxies. In first approximation the functional dependences of the models are easy to understand: In the supernova-dominated model the X-ray and optical luminosities would be linearly related because the supernova energy input is proportional to the stellar luminosity for a constant supernova rate (Tammann 1974); in the gravitational-cooling flow model the X-ray and optical luminosities would be related as L_X L_B^{1.5 - 1.7} because the X-ray luminosity will be proportional to the stellar mass loss rate and to the square of the stellar velocity dispersion ( L_X ²) and L_B, while the Faber-Jackson relation (1976; also Tonry 1981) gives L_B ^3-4 (White & Chevalier 1984, Nulsen et al. 1984, Sarazin 1986, 1987a, Canizares et al. 1987). Sarazin & White (1988), however, find that models can deviate somewhat from these simple approximations, because in parts of the models the temperature can be fairly low, so that the X-ray luminosity is not as simply related to cooling. If supernova heating is the dominant source of energy, the gas could be confined by a combination of radiative cooling and external pressure; if gravitational heating prevails, gravity would confine the gaseous halos (Canizares et al. 1987). In Figure 7 is also plotted the L_X L_B² line of Forman et al. (1985) for hydrostatic halos accumulated in the galaxy potential over a Hubble time. This model, however, does not take into account cooling and heating mechanisms, or at least assumes that they balance closely (Sarazin 1987a). Moreover, it assumes the same emitting volume for all the galaxies, since the volume also figures in the emission measure (Fabbiano 1986a).

One common characteristic of these predictions is that they can easily explain only the more X-ray-luminous galaxies. The supernova-dominated models, in particular, predict X-ray luminosities in agreement with the upper envelope of the scatter diagram. Therefore, for many galaxies either the supernova rate is significantly lower than Tammann's (1982) estimate or the bulk of supernova energy is dissipated or radiated outside the X-ray band. Gravitational-heating models also predict large X-ray luminosities.

Smaller luminosities could be obtained if the mean stellar mass loss rate were smaller than that used in the calculations, or if a considerable fraction of were not incorporated into the accretion flow or were removed from the flow by thermal instabilities before the flow had reached the center of the galaxy, although in the latter case the reduction is not dramatic (Thomas 1986, Thomas et al. 1986, Fabian et al. 1987a, Canizares et al. 1987, Sarazin & White 1988, Vedder et al. 1988, C.L. Sarazin & G.A. Ashe, submitted for publication, 1988). Moreover, the models cannot account for the observed scatter in the data. This scatter and the possibility that hot halos do not dominate the X-ray emission of the less X-ray-luminous galaxies (see the previous section) might be related to the effect of ram pressure stripping (Forman et al. 1984a, Sarazin & White 1988, Vedder et al. 1988) or to the presence of winds or partial winds (MacDonald & Bailey 1981, White & Chevalier 1984, Sarazin & White 1987; see also Loewenstein & Mathews 1987a, Umemura & Ikeuchi 1987) in some of the galaxies. A. Renzini (preprint, 1988) argues against cooling flows and in favor of winds for the low-X-ray-luminosity galaxies because the rate of stellar mass return is too high to reconcile the cooling flow model with the observed X-ray luminosities.

Pressure confinement has been invoked (Forman et al. 1985) to explain the presence of the very bright coronae in Abell 1367 reported by Bechtold et al. (1983), which (as shown by Figure 7) would be over two orders of magnitude more luminous than early-type galaxies of comparable optical luminosity in the field and in Virgo. Canizares et al. (1987) disagreed with this explanation on the basis that the density of the cluster medium of A1367 is similar to that of Virgo (Forman et al. 1979, Bechtold et al. 1983). More recently, Vedder et al. (1988) developed steady-state cooling flow models for early-type galaxies and concluded that similar galaxies with similar gas content should have similar X-ray luminosities regardless of their location, since the external pressure around a galaxy would not affect the luminosity of the gas within the galaxy, but rather its temperature, and it would only affect the X-ray surface brightness near the outside of the galaxy. However, if conduction is important, heat could flow from the cluster to the galaxy and increase the X-ray luminosity (C. Sarazin, private communication, 1988).

The X-ray/optical correlation does not allow one to discriminate between the different models. However, there are two other observational constraints that need to be satisfied: the radial distribution of the halos, and their emission temperature. Comparison with models of the observed surface brightness profiles have been attempted by various authors, usually by using "average" profiles, which disregard the asymmetries and departures from power laws that are sometimes observed at large radii (see Trinchieri et al. 1986). Since these events could be due to interaction with the environment, this approach might appropriately describe an undisturbed early-type galaxy. Sarazin (1986; see also Sarazin 1987a, b, Loewenstein & Mathews 1987a, Sarazin & White 1988) pointed out that if heating and cooling balance locally in a halo, and if the heating of the gas per unit mass is independent of position, then _gas² _stars, and thus the X-ray and optical surface brightness profiles would then follow each other, as observed (Trinchieri et al. 1986, Killeen & Bicknell 1988a). Cooling flow calculations, however, in the absence of conduction heating and under the simplest assumption that the gas is a one-phase medium, generate models that typically have total luminosities that are too large (see above) and have a central peak of the X-ray emission that is not observed (Thomas 1986, Thomas et al. 1986, Loewenstein & Mathews 1987a, Sarazin & White 1988, Vedder et al. 1988). Modifying the models by allowing "sinks" of mass from the flow to prevent accretion at the cores; or even cooling outflows, produces profiles in closer agreement with the observations (Thomas 1986, Thomas et al. 1986, White & Sarazin 1987a, b, c, Vedder et al. 1988, C.L. Sarazin & G.A. Ashe, submitted for publication, 1988). It is also possible that many different phases are present in the cooling gas (Nulsen 1986, Thomas et al. 1987). At present the data do not allow us to discriminate readily between different models with different inputs of supernova energy, dark massive halos, and external pressure. Different models, however, produce different temperature profiles (e.g. Sarazin & White 1987, Vedder et al. 1988), and future X-ray observations will be able to give us an answer. I discuss in the next sections the constraints on the mass of the galaxies imposed by this type of model fitting.

4.3. Cooling Flows

As discussed above, with the exclusion of the static halo scenario of Forman et al. (1985), most models for the gaseous halos involve cooling flows to the galaxy cores. The presence and the amount of these flows have been objects of debate, chiefly because observational evidence at wavelengths other than the X ray is not clear-cut. Although a full discussion of this subject is beyond the scope of this review, I summarize here the aspects that are more relevant for the halos of elliptical galaxies. These flows have been suggested because the cooling times implied by the X-ray data for the central regions of most observed galaxies, and in some cases for most of the galaxy volume, are considerably shorter than the Hubble time (Nulsen et al. 1984, Trinchieri et al. 1986, Sarazin 1987a, Canizares et al. 1987). In the absence of other sources of heat, therefore, the gas in a volume element will cool and accrete subsonically in pressure-driven flows toward the center of the galaxy potential (e.g. Silk 1976, Fabian & Nulsen 1977, Cowie & Binney 1977, Mathews & Bregman 1978). Derived mass accretion rates range from a few to ~ 1000 M yr^-1 for dominant cluster galaxies (Fabian et al. 1981, Mushotzky et al. 1981, Canizares et al. 1982a, Lea et al. 1982, Stewart et al. 1984a, b, Arnaud et al. 1984, Matilsky et al. 1985, Canizares et al. 1988) to ~ 1M yr^-1 for normal ellipticals (Nulsen et al. 1984, Trinchieri et al. 1986, Canizares et al. 1987).

Although there is evidence for cooler gas in optical filaments and emission-line regions associated with cluster cooling flows and with the central regions of some early-type galaxies [e.g. Fabian et al. 1984b (and references therein), 1985, 1987b, Demoulin-Ulrich et al. 1984, Hu et al. 1985, Phillips et al. 1986, Crawford et al 1987], the main problem of this scenario has been with finding incontrovertible evidence of star formation, which should be both substantial (given the mass cooling rates) and happening at all radii (Stewart et al. 1984a, Nulsen et al. 1984, Fabian et al. 1984b, 1987a, Nulsen & Carter 1987, White & Sarazin 1987b, c, 1988). Star formation with a normal IMF might be taking place in some galaxies (Silk et al. 1986, Romanishin 1987, Johnstone et al. 1987, Burstein et al. 1988, Bertola 1988); however, in most cases the galaxies' colors and 2.3-µm CO absorption index suggest that only low, or very low, mass stars may be forming (Fabian et al. 1982, Sarazin & O'Connell 1983, Fabian et al. 1984a, Arnaud & Gilmore 1986, Johnstone et al. 1987, O'Connell & McNamara 1988). This unusual IMF has been justified on the grounds that the large pressure in the flows will decrease the Jeans mass (Jura 1977, Fabian et al. 1982, Sarazin & O'Connell 1983), but the inconsistency of this explanation with the formation of massive stars in dense molecular clouds has led some authors to invoke additional sources of heat in the halos (e.g. Silk et al. 1986). Heating by conduction, in particular, has been suggested for galaxies in clusters (Tucker & Rosner 1983, Bertschinger & Meiksin 1986, Rosner & Tucker 1989), but it might only apply for a very limited range of physical parameters (e.g. Bregman & David 1988). This is very much an open field of investigation, and there is no doubt that it will continue to be pursued from both an observational and a theoretical point of view.

Another way of testing some of the cooling flow theories is to try to find evidence of a colder interstellar medium in early-type galaxies and then compare these observations with the X-ray data. Canizares et al. (1987) searched for possible correlations between the X-ray emission and either the optical emission line (e.g. Phillips et al. 1986, Veron-Cetty & Veron 1986; see also Sadler 1988) or the H I emission (e.g. Knapp et al. 1985, Wardle & Knapp 1986), that could suggest a direct connection between the hot and the cooler interstellar medium, but they found none. However, the subsets of galaxies considered in each case are rather small and, at least for the H I, are dominated by nondetections. This lack of correlation would be consistent with the suggestion that the H I could have been accreted as a result of close encounters between galaxies (e.g. Knapp et al. 1985). However, Bregman et al. (1988) report the discovery of 1.5 × 10⁸ M of neutral hydrogen in NGC 4406 that they believe might be created within a cooling flow, and a similar conclusion has been reached by Huchtmeier et al. (1988) from their detection of CO emission from NGC 4472. The neutral hydrogen in NGC 4406 is centrally peaked, and there is no evidence for rotational support, as is often found in elliptical galaxies, where the H I tends to lie in rotationally supported extensive rings that are thought to be of external origin (e.g. van Gorkom et al. 1986). For completeness, it should be mentioned that Nulsen et al. (1984) have suggested that rotationally supported H I rings, not necessarily aligned with the galaxy's rotation axis, could be excreted from a cooling flow as a way of dissipating angular momentum.

Work on the IRAS data has shown that cool dust is present in early-type galaxies (Jura 1986, Wrobel et al. 1986, Thronson & Bally 1987). A possible, although not strong, correlation was reported between the far-infrared and the X-ray emission (Jura 1986, Knapp 1988, Kim 1988) that suggested a link between the hot and the cold interstellar medium in these galaxies, possibly through a common origin for the dust and the hot gas from stellar ejection. In particular, Jura (1986) concluded that the cold far-infrared-emitting dust should exist in pockets separated from the hot X-ray-emitting gas; otherwise, the dust would be destroyed by thermal sputtering in a short time. This might be evidence for the existence of a multiphase interstellar medium, which has been suggested in the frame-work of cooling flow modeling of the X-ray halos (e.g. Thomas et al. 1987).

In a recent series of papers, Mathews (1988a, b, c) has explored the constraints provided by the observational mass-to-light ratios to the mass of optically dark stars, originating from cooling flows in the cores of elliptical galaxies, and from these to the IMF and the history of winds and cooling inflows in the interstellar medium. He concludes that a component of dark stars in the cores is unlikely to be more massive than 30 times the core mass of luminous stars (Mathews 1988c) and suggests that the IMF of stars in elliptical galaxies is significantly flatter than the Salpeter IMF. In this scenario strong winds have expelled from the galaxies the mass shed by stars in early times, creating the intracluster medium, while present-day ellipticals are likely to be closed systems (Mathews 1988a). A. Renzini (preprint, 1988) also points out the presence of winds in young galaxies, given the enhanced supernova rate, and suggests that these winds might still be prevalent in the less X-ray-luminous ellipticals.

4.4. The Mass of Early-Type Galaxies

X-ray observations have been used to measure the mass of early-type galaxies (previous reviews on this subject include Canizares 1987, Sarazin 1987b, Fabian & Thomas 1987, Fall 1987, Trimble 1987). These measurements are based on the assumption that the X-ray emission is thermal bremsstrahlung from hot gaseous halos, and that the gas traces the potential but does not contribute to it significantly. This is a reasonable assumption (Forman et al. 1985), since M_gas / M_stars < 7% if one assumes that M_stars / L_stars = 6 (Faber & Gallagher 1979). Three approaches have been used: The first has been to measure the binding mass within a certain radius, under the assumptions of hydrostatic equilibrium and spherical symmetry. This method was first applied to M87 and has more recently been used to estimate the mass of less X-ray-luminous galaxies. The second approach has made use of more detailed modeling to match theoretical predictions to the observed X-ray surface brightness profiles. Finally, Fabian et al. (1986a) have devised a method to estimate lower limits to the total mass of a galaxy. In the following I discuss the results of these three types of analyses, with an eye to examining the assumptions made and the limitations of the current data.

The discovery of the extended thermal X-ray source in the Virgo cluster centered on M87 prompted the first attempts at estimating the binding mass of an early-type galaxy within radii well away from the optical core, to which optical measurements have been restricted. The first estimates, based on the assumption of the gas being in hydrostatic equilibrium in the galaxy potential at a temperature T ~ 3 × 10⁷ K, suggested very large binding masses, between 10¹³ and 10¹⁴ M (Bahcall & Sarazin 1977, Mathews 1978). However, Binney & Cowie (1981) showed that the X-ray data then available could also be fitted with a model of pressure-confined cooling atmosphere surrounding a low-mass galaxy. The observations of M87 with the Einstein IPC, which produced an image and spatially resolved spectral information, have led to more accurate measurements of the binding mass and have confirmed the earlier suggestions of the presence of a massive dark halo (Fabricant et al. 1980, Fabricant & Gorenstein 1983). In particular, Fabricant & Gorenstein (1983) could exclude a radial dependence of the temperature of the gaseous halo consistent with Binney & Cowie's model and found that the integral mass-to-light ratio (M/L) of M87 in solar units must increase from 5-15 at a radius of 1" [~ 4.4 kpc (from the optical data of Sargent et al. 1978)] to over 180 at 20" (~ 87 kpc), which implies that a dark massive halo is present. They derived a total mass M ~ 3 - 6 × 10¹³ M within 60" (~ 260 kpc) from the core. Similar parameters were found for other central galaxies in clusters, including NGC 1275 in Perseus (Fabian et al. 1981, Branduardi-Raymont et al. 1981) and NGC 4696 (Matilsky et al. 1985), the dominant galaxy of the Centaurus cluster. An apparently isolated galaxy, identified as the counterpart of a serendipitous X-ray source [lE0116.3-0116 (Maccagni et al. 1987)], and central galaxies in poor clusters (Kriss et al. 1983, Canizares et al. 1983, Biermann & Kronberg 1983, Malumuth & Kriss 1986) have also been reported to have similar X-ray luminosities (a few 10⁴³ erg s^-1) and M/L ratios.

The equation used by Fabricant et al. (1980) to estimate the binding mass of M87 was derived from the equation of hydrostatic equilibrium in combination with the ideal gas law, under the assumption of spherical symmetry, and is given below:

(3)

where G is the gravitational constant, and µm_H is the mean gas particle mass (µ is taken to be 0.6, and m_H is the mass of the hydrogen atom).

Equation 3 shows that the measurement of the binding mass depends on four variables: the radius within which the mass is estimated, the gas temperature at this radius, and the logarithmic gradients of both temperature and gas density at the same radius. Each of these variables is a potential source of uncertainty. While the observations of M87 yielded enough signal to noise to allow a relatively well-constrained mass estimate, the same unfortunately is not true when this method is applied to early-type galaxies not at the center of clusters. These galaxies typically are ~ 100 times less luminous than and at least as distant as M87, and therefore their radial profiles and especially temperature information are not as well constrained. Forman et al. (1985) pointed out that the assumption of hydrostatic equilibrium for these galaxies is reasonable even in the presence of cooling flows because these flows would be largely subsonic. They derived M/L ratios ranging from ~ 10 to 90 for a sample of 13 galaxies by assuming that the X-ray emission is due to an isothermal gas at T = 1.2 × 10⁷ K, adopting as r_gas the outermost radius at which X-ray emission could be detected, and using an average fitted value for the density gradient. Taken at face value, this result would suggest that dark extended halos are a common feature of early-type galaxies. A supporting argument given by these authors for the widespread presence of massive dark halos is the suppression of galactic winds (see the previous section). However, radiative wind suppression in the inner regions and an external pressure from a surrounding intracluster or intergalactic medium (e.g. Fabian et al. 1980, White & Chevalier 1984) could also be responsible for the retention of the hot interstellar medium (Canizares 1987, Canizares et al. 1987). The temperature in this case should increase at the outer radii. Temperature profiles are needed to exclude this possibility (Sarazin & White 1987, Vedder et al. 1988). Forman et al. (1985) performed a spectral fit of the X-ray image of NGC 4472 in three radial annuli and concluded that the data are consistent with isothermality. However, a sharp increase of the temperature at the outer radii, as required by the pressure-confined models, is also possible, given the large uncertainties of the fit. Moreover, Trinchieri et al. (1986) suggest a complex spectrum for this X-ray source, which would thus increase the uncertainties on the fitted temperature.

Subsequent work (Trinchieri et al. 1986; see also Fabbiano 1985b, 1986a, b, Canizares 1987) has shown that the uncertainties of this type of mass measurement can be large. In particular, detailed analysis of the Einstein data of six elliptical and SO galaxies, which are among those detected with higher signal to noise in the sample of Forman et al. (1985), shows both departures from spherical symmetry, which can be ascribed either to interlopers in the field or to the interaction with the surrounding medium, and peculiar variations of the gas density gradient at the outermost radii. Even ignoring some of these effects, the uncertainties of the gas temperature and of its gradient are such that mass estimates differing by up to a factor of 10 are possible for a single galaxy (Figure 8). Jones (1987) and Forman & Jones (1988) have pointed out that the assumption of isothermality is justified in galaxies with a gravitationally bound hot halo (e.g. Norman & Silk 1979; however, see the discussion above about nongravitational confinement). The uncertainties in the IPC spectral fits, however, are such that the range of possible temperatures is still very large, even if one were to assume isothermality (Trinchieri et al. 1986; see Figure 8). Within these uncertainties, the X-ray measurements are generally consistent with the mass estimates from optical velocity dispersion. Therefore, large dark halos are allowed by these data but not required by this analysis. The use of less luminous galaxies or of Sa galaxies will introduce an additional uncertainty that cannot be resolved with the present data. This is the possibility that most of the X-ray emission is due to the evolved stellar component and not to a hot gaseous medium (Fabbiano & Trinchieri 1985, Trinchieri & Fabbiano 1985, Canizares et al. 1987, Fabbiano et al. 1989), thus invalidating the whole approach. In this regard, as remarked by Knapp (1987; see also Fall 1987), it may be significant that the masses of the Sombrero galaxy (an Sa) and of NGC 5128 (Cen A), estimated from HI rotation curves (Bajaja et al. 1984, van Gorkom 1987), are significantly smaller than the estimates of Forman et al. (1985). Although the HI measurements stop at radii smaller than those used for the X-ray measurements, the discrepancy would imply in both cases that the rotation curves, which are flat in the H I measurements, would then rise steeply between the H I and X-ray radius. This effect might be real [a radial increase of the circular velocity occurs in M87 if the halo is gravitationally confined (Binney & Cowie 1981, Sarazin 1987b)], but it could also indicate some problem with the X-ray estimate, since rotation curves generally tend to flatten out at large radii (Knapp 1987).

Figure 8. Mass estimates and related uncertainties for five early-type galaxies (adapted from Trinchieri et al. 1986). The shaded areas are the allowed regions from Equation 3, as in the above paper; the central solid line is the estimate for an isothermal halo; and the dashed and dot-dashed lines are the estimates for halos with moderate radial dependences of the temperature (T ~ r0.5 and T ~ r-0.5, respectively). The vertical lines restrict the temperature ranges to the values obtained by the spectral fits of the X-ray data (at the 90% confidence level). The diagonal short-dashed lines are isothermal estimates of the binding mass for more conservative choices of the halo outermost radii, which exclude regions where the radial surface brightness profiles depart from a power law (see Trinchieri et al. 1986). The horizontal dashed and dotted lines are mass-to-light ratio determinations from optical velocity dispersions (see Trinchieri et al. 1986, and references therein). The diagonal lines with upward-pointing arrows are the lower limits to the total binding mass (and mass-to-light ratios) calculated following Fabian et al. (1986a, Equation 4 herein) for the less conservative estimates of the outer radii (larger rmax); if one uses the smaller rmax values, these estimates will be displaced downward.

A different approach to mass measurements relies on a comparison of the observed radial profile of the X-ray surface brightness with model predictions. The method generally followed is to assume a potential that describes the stellar radial distribution and the observed stellar velocity dispersion and then to generate, under various assumptions, the X-ray (gas) surface brightness profile (and/or temperature), to be then compared with the data. This approach was used on M87 and poor clusters (e.g. Mathews 1978, Binney & Cowie 1981, Stewart et al. 1984a, Canizares et al. 1983). In particular, Stewart et al., using the Einstein spectral and spatial data of M87, chose a family of models, which describe the radial mass distribution, in agreement with the optical data and with the X-ray data of Fabricant & Gorenstein (1983). Model fitting of less X-ray-luminous, normal early-type galaxies gives less clear results. Using ad hoc models, Trinchieri et al. (1986) showed that the observed gas distributions do not necessarily require the presence of dark halos; Canizares et al. (1987) concluded similarly that the observed correlation between X-ray and optical luminosities does not put strong constraints on the existence of massive halos; and Vedder et al. (1988) pointed out that the only visible effect of massive halos would be in the temperature profiles, which cannot be measured with the Einstein data. Sarazin & White (1987, 1988) instead conclude that cooling flow models, possibly without supernova heating but with massive halos, better reproduce the data, although they stress the large uncertainties of their result. A similar conclusion had been reached by Thomas (1986), who applied a cooling flow model to NGC 4472. However, the X-ray isophotes of this galaxy are not circular at large radii, which suggests interaction with the intracluster medium. The binding mass within 40 kpc, which is the radius up to which the "undisturbed" gas distribution should extend in this asymmetric halo (see Trinchieri et al. 1986), is in very good agreement both with the estimate of Trinchieri et al. (1986) and with optical estimates. Killeen & Bicknell (1988a) have estimated the binding mass of NGC 1399, using both model comparison and Equation 3, and conclude that the determination is rather uncertain and does not exclude M/L values in agreement with optical measurements: Thus the presence of a large massive halo, although allowed, is not required. A different approach has been followed by Mathews & Loewenstein (1986), Loewenstein & Mathews (1987a; see also Loewenstein & Mathews 1987b), and Umemura & Ikeuchi (1987), who have studied the time history of gaseous halos in early-type galaxies and find luminosities in agreement with the data only in the presence of dominant dark halos.

Instead of relying on a measure of the binding mass within a certain observed radius, Fabian et al. (1986a) devised a method for estimating a lower limit to the total binding mass. Their method is based upon three assumptions: (a) that the gas within the observed radius is confined by a hydrostatic outer atmosphere, (b) that the halo is convectively stable, and (c) that the pressure gradient is always negative. With these assumptions they obtain the equation

(4)

where r₀ is the outermost detected radius of the halo, P₀ and T₀ are the gas pressure and temperature at this radius, and r and P are the outer radius of the halo and the gas pressure at this radius, respectively. From this equation, and using the outermost radii (r₀) and best-fit values of temperatures from the literature, they derive mass-to-light ratios for a sample of early-type galaxies, which lead them to conclude that there is overwhelming evidence for large massive halos in these galaxies. However, the same uncertainties apply here as in the calculation of the binding mass from Equation 3. Using the five galaxies for which the binding mass had been measured by Trinchieri et al. (1986) and temperature estimates that take into account the uncertainties of the spectral fits, together with the values of r₀ derived in that paper, I have calculated lower limits to the mass with Equation 4. They are shown in Figure 8. For two of the galaxies (NGC 4649 and NGC 4472) these limits do not require particularly large M/L ratios in excess of those inferred from optical measurements. For the other three galaxies (NGC 4636, NGC 1395, and NGC 720), however, values of M/L > 20 are clearly required. Therefore, there is evidence for large dark halos in some normal elliptical galaxies at least, unless the three basic assumptions of Fabian et al. do not apply. This might be the case for galaxies experiencing considerable external forces, as, for example, is the case for M86, which could be ram pressure confined (Fabian et al. 1980; see also Fabian et al. 1986a). Also, as remarked by the latter authors, if there is a significant confining pressure due to intracluster or intragroup gas, the lower mass limit may not apply to the galaxy but to the group as a whole.

Considering these results, I believe that, purely on observational grounds, there is still insufficient evidence to prove that dark massive halos are a general feature of early-type galaxies, although there is evidence of their existence in cluster dominant galaxies and perhaps in some of the most X-ray-luminous normal ellipticals. The use of X-ray data to measure the mass of these galaxies is very promising, but future more sensitive X-ray observations are needed to get a more definite answer.

Assuming that dark halos are a common feature of early-type galaxies, one may ask the question if all galaxies have similar dark halos. Forman et al. (1985) suggested that this may be the case, and moreover that the ratio between total and luminous mass in early-type galaxies could be the same as those of clusters and groups (Blumenthal et al. 1984). This would indicate that the dark matter of clusters might just be the superposition of single galaxy halos. It is also possible that normal ellipticals might be less massive than group or cluster dominant galaxies. In particular, Mathews (1978) remarked that the presence of an extended luminous X-ray halo in M87 and the absence of comparable X-ray emission from NGC 4472, which is also in the Virgo cluster and is optically more luminous than M87, argued for an exceptional massive component in the latter (see also Forman et al. 1984a).

One can investigate further this issue, insofar as the data will allow it, by plotting the mass-to-light ratios of different galaxies (including dominant galaxies in clusters and groups, and normal early-type galaxies) as a function of the ratio between their X-ray and optical luminosities, which is a measure of their gaseous content. This is shown in Figure 9a. There is a suggestion of a possible correlation in this plot (if one disregards the large uncertainties on each point) in the sense that cD galaxies, which retain a larger amount of hot gas, may be more massive than normal ellipticals, and therefore different galaxies might have different amounts of dark matter relative to their luminous matter. It is interesting that there is no correlation of the central stellar velocity dispersions with the X-ray to optical ratio, in the same range of ratios (Figure 9b), consistent with the dark matter being in extended halos so as not to disturb the stellar component. However, galaxies with X-ray to optical ratios consistent with those of spirals (L_X / L_B < 10³⁰ erg s^-1 L^-1, this corresponds to a monochromatic flux ratio of 10^-6.5), which therefore might not be able to retain gaseous halos, tend to have smaller central velocity dispersions. Given the paucity of points in Figure 9a, however, and the large error bars, the possibility that all early-type galaxies have similar mass-to-light ratios cannot be discounted either. Also, these ratios depend on the measured extent of the X-ray emission and therefore are technically lower limits (see Forman et al. 1985). What has to be explained in this case is why there is a systematic difference between the X-ray luminosities of different types of galaxies, some of which are at the same distance (in Virgo) and were observed with comparable sensitivity. In particular, why can dominant galaxies in clusters retain far larger gaseous halos than those in poor groups, which are still more successful in this endeavor than general field or nondominant cluster galaxies? And, even in the non-dominant cluster galaxies, why can galaxies with similar optical luminosity or central velocity dispersion have such a large range of X-ray luminosities (Canizares et al. 1987, Sarazin & White 1988)? Although other interpretations, invoking dynamical stripping of galaxies orbiting the cluster, cannibalism, or past enhanced accretion, are certainly possible (e.g. Fabian et al. 1981, Takeda et al. 1984, Stewart et al. 1984b), it is possible that different amounts of dark matter in otherwise similar galaxies might ab initio be responsible for these results.

Figure 9. (a) Mass-to-light ratio (in solar units) versus the ratio of the X-ray (in erg s-1) to the optical (in L) luminosity for "normal" elliptical galaxies and for group and cluster dominant galaxies (see text for references). MSS is the galaxy found by Maccagni et al. (1987) in the serendipitous Einstein survey. (b) Square of the central stellar velocity dispersion versus the X-ray to optical ratio. The vertical dashed line represents the average ratio for spiral galaxies. The M87 velocity dispersion is from Sargent et al. (1978), and the other velocity dispersions are from Whitmore et al. (1985). X-ray data are from the compilations of Canizares et al. (1987) and Matilsky et al. (1985).

And finally, getting onto even more speculative ground, we can ask, what would the nature of this dark matter be? One possibility is that of exotic particles, as in the cold dark matter scenario (e.g. Blumenthal et al. 1984; see also Forman et al. 1985). Another possibility is that the dark matter is composed of ordinary stellar remnants or low-mass concentrations, as suggested by Fabian et al. (1986b, 1987a) in the framework of the cooling flow scenario [but see White & Sarazin (1987) as a dissenting voice: the distribution of the matter deposition from cooling flows may be more similar to the light distribution than to extended massive halos].

4.5. Radio Sources and Gaseous Halos

The importance of hot gaseous halos for both fueling nuclear, radio sources through accreting cooling flows and confining extended radio lobes has long been recognized (e.g. Cowie & Binney 1977, MacDonald & Bailey 1981, Norman & Silk 1979). A correlation found between radio power and X-ray emission of central cluster galaxies gave empirical evidence of the first phenomenon; the effectiveness of the intracluster gas in confining the radio sources was demonstrated by detailed comparisons of radio and X-ray data (Burns et al. 1981, Forman & Jones 1982, Valentijn & Bijleveld 1983, Jones & Forman 1984, Harris et al. 1984, Feretti et al. 1984a, Morganti et al. 1988, Valentijn 1988). Similar comparisons show that the hot interstellar medium of more isolated early-type galaxies is also effective in confining the radio sources and is likely to play a role in their origin (Biermann & Kronberg 1983, Stanger et al. 1984, Dressel & Wilson 1985, Stanger & Warwick 1986, Thomas et al. 1986, Fabbiano et al. 1987a, 1989, Killeen et al. 1986, 1988). The extended hot gaseous halos can also be responsible for the depolarization observed at long radio wavelengths (e.g. 49 cm) in the bridges of double-lobed powerful radio galaxies (Strom & Jagers 1988).

If one excludes the more "active" specimens (e.g. 3CR galaxies, Cen A, For A), early-type galaxies tend to have relatively faint radio emission confined within their cores. These radio sources in some cases have clearly the same morphology as the more powerful radio galaxies and could therefore be considered physically similar to them and related to nuclear activity (e.g. Ekers & Ekers 1973, Bieging & Biermann 1977, Condon & Dressel 1978, Ekers & Kotanyi 1978, Hummel et al. 1983, Birkinshaw & Davies 1985, Fabbiano et al. 1987a, 1989). Correlations between the radio power and the X-ray luminosity of small samples of relatively radio-faint early-type galaxies suggested a possible link between the hot interstellar medium and the nuclear activity, since the X-ray emission of these galaxies is typically dominated by the extended thermal component (Dressel & Wilson 1985, Fabbiano et al. 1987a; see also Trinchieri 1988). More recently, Fabbiano et al. (1989) have compared the X-ray and radio continuum properties of the larger sample of early-type galaxies studied by Canizares et al. (1987) and report a correlation between the radio "core" power and the X-ray to optical ratio. The latter is a measure of the excess X-ray emission over the linear correlation observed in spiral systems and thus can then be considered as a direct indicator of the gaseous component (see previous discussion). This correlation therefore reinforces the connection between the hot interstellar medium and the nuclear radio sources and points to accreting cooling flows as the fuel.

The hot interstellar medium can also play an important role in the formation of extended radio structures. Fabbiano et al. (1989) notice that in their sample of early-type galaxies, very extended radio lobes, such as those of Cen A (NGC 5128) and For A (NGC 1316), tend to be associated with galaxies with relatively small X-ray to optical ratios. By contrast NGC 1399, which has a comparable core component, and therefore a comparably powerful central engine, but a larger X-ray to optical ratio, has the radio lobes well contained within its optical body. This result suggests that the gaseous halos also play a fundamental role in disrupting the radio jets and confining the extended radio structures. This conclusion had been previously reached by de Ruiter & Parma (1984), who observed significant distortions of the less extended sources in maps of radio galaxies of low to moderate power, suggesting interaction and bending of the jets by the interstellar medium. It has also been addressed in theoretical papers by Soker & Sarazin (1988) and Norman et al. (1988; see also Killeen & Bicknell 1988b). Fabbiano et al. (1989), in particular, demonstrate that the equation of Soker & Sarazin (1988) for the critical luminosity of the radio source, below which the jet is likely to be disrupted by shocks at the sonic radius of a galaxy cooling flow, correctly predicts the typical radio power of a few 10²⁹ erg s^-1 Hz^-1 at 5 GHz below which extended lobes are not found (Colla et al. 1975, Jenkins 1982, Feretti et al. 1984b, Fabbiano et al. 1989). Once the jets have been disrupted, the thermal pressure of the hot gas is typically effective in confining the less powerful radio lobes well within the galaxies (see also Stanger & Warwick 1986, Killeen et al. 1988).