THE ASTROPHYSICAL JOURNAL, 486:L83–L86, 1997 September 10
©1997. The American Astronomical Society. All rights reserved. Printed in U.S.A.
Issue Table of Contents
Go to: Search Page | Previous Article | Next Article
Other formats: HTML (small files) | PDF (180 kb) | PostScript pages (261 kb)

X-Ray Observations and the Structure of Elliptical Galaxies 1


Received 1997 June 2; accepted 1997 June 26


     We compare optical and high-quality X-ray data for three bright elliptical galaxies in the Virgo Cluster, NGC 4472, 4649, and 4636. The distribution of total mass in NGC 4472 and 4649 determined from X-ray data is sensitive to the stellar mass over a considerable range in galactic radius extending to r ≈ re, the effective radius. The agreement of X-ray and optically determined stellar masses provides a unique verification of the stellar mass-to-light ratio, which is essentially constant over the range 0.1 ≲ r&solm0;re ≲ 1. However, for NGC 4636 the dark matter is important at all radii ≳0.35re. Evidently, the dark to stellar mass ratio varies in quite different ways in elliptical galaxies of comparable optical luminosity, implying that the radial structure of dark halos may not be universal. There is some evidence in NGC 4636 for additional support of the hot interstellar gas at r ≲ 0.35re; either a field B ∼ 10-4 G or a small (mechanically unstable) central region of high gas temperature (T ∼ 107 K) is required. The global temperature structure in the hot interstellar medium of many recently observed elliptical galaxies is very similar, reaching a maximum near 3re–4re. This feature, which may suggest a new structural scale in these galaxies, is inconsistent with current theoretical gasdynamical models.

Subject headings: cooling flows—galaxies: elliptical and lenticular, cD—galaxies: individual (NGC 4472, NGC 4636, NGC 4649)—X-rays: galaxies



     1 UCO&solm0;Lick Observatory Bulletin No. 1368.

     2 Dipartimento di Astronomia, Università di Bologna, via Zamboni 33, Bologna 40126, Italy;

     3 University of California Observatories&solm0;Lick Observatory, Board of Studies in Astronomy and Astrophysics, University of California, Santa Cruz, CA 95064;


     Elliptical galaxies are large, bright stellar systems with considerable structural regularity. The stellar light profiles are well fitted by the de Vaucouleurs law (Burkert 1993), and the stellar velocity dispersion, surface brightness, and effective radius re are constrained to a fundamental plane (Djorgovski & Davis 1987; Dressler et al. 1987). In contrast, the X-ray emission from hot interstellar gas in elliptical galaxies exhibits enormous variations in LX&solm0;LB (Eskridge, Fabbiano, & Kim 1995) and, to a lesser extent, in gas temperature (Davis & White 1996).

     We have reexamined optical and X-ray observations of elliptical galaxies in preparation for a new series of gasdynamical studies of the evolution of the interstellar gas. Although current X-ray observations are sparse, we have found that they reveal more useful information about elliptical galaxies than is currently realized. Gas temperature profiles T(R) for elliptical galaxies recently determined with ROSAT and ASCA reveal a surprising uniformity for radii r ≲ 10re. The total mass and density of all gravitating matter, Mtot(r) and &rgr;tot(r), can be found from the variation of temperature and density in the hot interstellar gas by assuming that the gas is in hydrostatic equilibrium in the galactic potential.

     We show here that the total mass determined from the X-ray gas for two bright Virgo elliptical galaxies (NGC 4472 and 4649) is in excellent agreement with a de Vaucouleurs profile for 0.1re ≲ r ≲ re using optically determined mass-to-light ratios. Remarkably, in another bright Virgo elliptical galaxy, NGC 4636, dark matter contributes substantially to the total mass within re, and the total mass found from X-ray observations is lower than the known stellar mass for r ≲ 0.35re, implying that thermal pressure is not the only support for the hot gas. These results suggest that AXAF X-ray observations of elliptical galaxies will provide powerful new constraints on both the interstellar physics and the nature of stellar populations in these galaxies.


     Figure 1 shows a plot of the radial variation of gas temperature with projected galactic radius R&solm0;re for six early-type galaxies. The remarkable feature that all galaxies share is a positive temperature gradient out to about 3re followed by a leveling off or gradual decrease toward larger radii. Data for Figure 1 are taken from the following sources: NGC 1399, Jones et al. (1997); NGC 5044, David et al. (1994); NGC 4636, Trinchieri et al. (1994); NGC 507, Kim & Fabbiano (1995); NGC 4472, Irwin & Sarazin (1996); NGC 4649, Trinchieri, Fabbiano, & Kim (1997). The temperature profile observed with the ASCA satellite by Mushotzky et al. (1994) for NGC 4636 is in excellent agreement with that of Trinchieri et al. (1994) for 20% solar abundance. Additional properties of these galaxies are listed in Table 1. It is remarkable that T(R) is so consistent in spite of the wide range of luminosities and cosmic environments among these galaxies. NGC 4636, 4649, and 4472 are in the Virgo Cluster, while NGC 1399, 5044, and 507 are the brightest galaxies in small groups or clusters. NGC 507 has an exponential surface brightness profile that is consistent with a very massive, face-on S0 galaxy (Magrelli, Bettoni, & Galletta 1992).

Fig. 1

     The temperature maximum observed in elliptical galaxies at about 3re suggests a new, previously unrecognized structural scale length. This cooling of gas within 3re–4re is not a natural result of galactic cooling flows, as many have suggested. In theoretical models, the radial gas velocity in the hot interstellar gas is very subsonic, since otherwise the gas replenishment time would be impossibly short and LX would be much too low. Straightforward spherical models of the hot interstellar gas (e.g., Fig. 3 of Brighenti & Mathews 1996) typically result in subsonic inflow with dT&solm0;dr < 0 for r ≳ 0.1re. Although the gas loses thermal energy by emitting the X-rays observed, it is heated by compression as it descends deeper into the galactic potential, i.e., dT&solm0;dr < 0. In this sense the term “galactic cooling flow” is a misnomer. In “mass dropout” models, in which gas is removed from the inner flow, the remaining gas is actually heated, since the Pdv work done by gravity must raise its temperature even higher to provide hydrostatic support for gas at larger radii. However, this rise in T(r) is fully compensated when radiation from the cooler, dropping-out gas is included (e.g., White & Sarazin 1987). In any case, mass dropout is usually invoked at small galactic radii, less than the scale of dT(r)&solm0;dr < 0 in Figure 1. The gas temperature is also sensitive to heating by supernova explosions, but this heating should be proportional to the stellar density, so dT&solm0;dr < 0 at r ≲ re would again be expected. To our knowledge, no theoretical cooling flow model can produce temperature profiles resembling those in Figure 1. Fortunately, regardless of the possibly complex thermal history of the interstellar gas, the observed variation of T and &rgr; can be used to determine the underlying mass distribution, as we now discuss.


     NGC 4472, 4649, and 4636 listed in Table 1 are all members of the Virgo Cluster. We adopt a distance of D = 17 Mpc for all three galaxies, similar to the Malmquist-corrected value D = 17.2 ± 1.9 Mpc of Gonzalez & Faber (1997). For any particular Virgo galaxy the distance is uncertain by about 20% because of the unknown location of the galaxy along the line of sight through the cluster.

     The total mass Mtot(r) in these galaxies can be found from the condition for hydrostatic equilibrium:

where the last term allows for the possibility of magnetic pressure Pm = B2&solm0;8&pgr;, and &phgr;m = Pm&solm0;P is the ratio of magnetic to gas pressure. The proton mass is mp, and we assume &mgr; = 0.63. The strong negative density gradient d log &rgr;&solm0;d log r is expected to be the largest of the three derivatives. Only the shape of &rgr;(r), not its absolute normalization, influences Mtot(r). The density distribution &rgr;(r) in the hot interstellar gas can be found from the X-ray surface brightness distribution. T(r) and possibly also &phgr;m(r) must be known to some precision in order to determine reliable total masses. The total mass density is &rgr;tot(r) = dMtot&solm0;4&pgr;r2dr.

     Density profiles are available for all three elliptical galaxies from Einstein HRI observations (Trinchieri, Fabbiano & Canizares 1986) and from ROSAT PSPC (Trinchieri et al. 1994; Irwin & Sarazin 1996; Trinchieri et al. 1997). Fortunately, there is a substantial range of angular scale over which these data sets overlap, so values of &rgr;(r) from Einstein and ROSAT data can be renormalized to agree. Having done this, we fit n(r) = &rgr;(r)&solm0;mpwith a sum of functions n(r) = &Sgr;ni(r), where ni(r) = no(i){1 + [r&solm0;ro(i)]p(i)}-1, and the temperature is fitted with T(r) = 2Tm[rm&solm0;(r + rot) + (r&solm0;rm)q]-1. The temperature T(R) observed at any projected radius R is an average along the line of sight weighted by &rgr;2 and differs in principle from the temperature T(r) at physical radius r. However, &rgr;2 is a very steep function of galactic radius, and we find T(r) ≈ T(R) within 10%, sufficient for our purposes here, considering the observational uncertainties involved (Fig. 1). Figures 2a, 2d, and 2g show our fits to the X-ray data. The parameters for these fits are as follows: NGC 4472, no(i) = 0.095, 0.00597, -0.0004; ro(i) = 0.17, 0.95, 10; p(i) = 2.0, 1.14, 1.19; Tm = 0.75; rm = 0.5; rot = 0.75; q = 0; NGC 4636, no = 0.151; ro = 0.172; p = 1.57; Tm = 0.5375; rm = 0.475; rot = 0.6; q = 0.0; NGC 4649, no(i) = 0.1, 0.0014; ro(i) = 0.15, 3.0; p(i) = 1.8, 3.0; Tm = 0.9; rm = 4.0; rot = 4.0; q = 0.0, with radii in re, densities in cm-3, and temperatures in 107 K. Total masses Mtot(r) (Figs. 2b, 2e, and 2h) and the corresponding total mass densities &rgr;tot(r) (Figs. 2c, 2f, and 2i) are determined with &phgr;m(r) = 0. Our values for Mtot(r) in Figure 2b are in satisfactory agreement with those of Irwin & Sarazin (1996) (scaled to D = 17 Mpc), which are based only on ROSAT data; our Mtot(r) in Figure 2h is in excellent agreement with Mushotzky et al. (1994).

Fig. 2

     It is of particular interest to compare Mtot(r) and &rgr;tot(r) with corresponding stellar values. The total mass M*t is found from LB using “stellar” mass-to-light ratios of van der Marel (1991), determined by comparing stellar velocities measured out to robs ≈ 0.5re with solutions of the two-dimensional Jeans equations in slowly rotating model galaxies. This “stellar” M*t&solm0;LB ratio is sensitive to all mass within robs and may contain a component of dark matter that can be checked a posteriori (see below). In general, mass-to-light ratio values determined only in the core region (e.g., Faber et al. 1997) are larger; central black holes may account for this increase, but typical holes may not be sufficiently massive (Kormendy & Richstone 1995; Faber et al. 1997). M*(r) and &rgr;*(r) in Figure 2 are evaluated with de Vaucouleurs profiles (Young 1976).

     The Structure of NGC 4472 and 4649.—We find it quite remarkable that the total mass and density indicated by the X-ray observations agree almost exactly with stellar values in the range 0.1 ≲ r ≲ 1re, where the observational data are of high quality (Figs. 2b and 2e). Discounting a conspiracy of compensating errors, this superb agreement is possible only (1) if the stellar mass-to-light ratio values determined by van der Marel (1991) are essentially correct and constant for stars out to about 1re, and (2) if magnetic pressure or rotation contributes little or nothing to the total pressure support of gas in 0.1 ≲ r ≲ 1re. The total masses of dark and stellar matter are equal at about 2.5re, which includes robs, so van der Marel's M*&solm0;LB values are likely to be totally stellar. Dark matter dominates at r&solm0;re ≳ 2.5. At r = 10re the total mass-to-light ratio is about 78 (NGC 4472) and about 110 (NGC 4649), and Mtot&solm0;LB is undoubtedly higher at r&solm0;re ≳ 10. The stellar to dark halo transition is particularly striking in the density plots (Figs. 2c and 2f). None of the qualitative features in Figure 2 are changed if (M*&solm0;LB)core values are used from Faber et al. (1997), but the overall fit in 0.1 ≲ r&solm0;re ≲ 1 is of lower quality. The discrepancy Mtot < M* at r ≲ 0.1re may result from observational inaccuracies in this region, although Mtot(r) would be underestimated if a large magnetic field were present.

     The Structure of NGC 4636.—Our total mass Mtot(r) for NGC 4636 (Fig. 2h) is in excellent agreement with the mass determined by Mushotzky et al. (1994), who also assumed D = 17 Mpc. A de Vaucouleurs profile is used to determine M*(r) and &rgr;*(r); we have verified that NGC 4636 satisfies a de Vaucouleurs profile by plotting the surface brightness data of Peletier et al. (1990) against R1&solm0;4.

     The most surprising result evident from Figures 2h and 2i is the relative dominance of dark matter in NGC 4636 as compared to NGC 4472 and 4649. The dark mass Mdark ≡ Mtot - M* in NGC 4636 becomes equal to the stellar mass M* at about r = 1.2re, and Mdark ≈ 0.83M* at r = re, where the total mass-to-light ratio is 19. This large amount of dark matter may be unusual, since it has been difficult until recently to find evidence of dark matter in elliptical galaxies from stellar velocities measured within re(e.g., Carollo et al. 1995). The “stellar” mass-to-light ratio M*&solm0;LB from van der Marel is determined from velocity observations only out to robs&solm0;re = 0.45, where M* ≫ Mdark (Fig. 2h). It is remarkable that there is no change in slope in Figure 2h as Mtot(r) crosses below M*(r) at r ≈ 0.33re, since all observations should be reliable at this radius. This failure of the X-ray determined mass to detect the stellar mass suggests large magnetic fields, unusually high gas temperatures, or rotational support in this region. Magnetic fields B ∼ 9 × 10-5 G would be required at r = 0.1re to support the hot gas against the stellar mass within (see Mathews & Brighenti 1997), or a curious central temperature inversion may exist: a mean temperature within 1 kpc (0&farcm;20) of 0.9 × 107 K is implied. This inversion is constrained so that the mean projected gas temperature within 1&arcmin;, 0.75 × 107 is within the observed limits 0.667 + 0.029&solm0;-0.041 × 107 K. However, the high-temperature interpretation for Mtot < M* in NGC 4636 is unlikely, since it is buoyantly unstable; dT&solm0;dr is superadiabatic for r ≲ 0.08re ≈ 0.7 kpc. Alternatively, M*&solm0;LB could decrease with galactic radius so that M* is lower than we think near r&solm0;re ≈ 0.1 in Figure 2h, but this seems unlikely in view of the uniform M*&solm0;LB values implied by Figures 2b and 2e. None of these conclusions are changed if the core value (M*&solm0;LB)core = 12.69 is used instead. The different distribution of dark matter at r≲re in NGC 4636 may provide evidence against a universal dark halo structure as proposed by Navarro, Frenk, & White (1996), but for r ≳ re the dark halos are all very similar, having slope &rgr;dark ∝ r-1.9 at r ∼ 10re. The total mass-to-light ratio near the outer limit of the X-ray observations shown in Figure 2g, r = 12.5re, is about 125.

     How do these Virgo elliptical galaxies differ in other respects? NGC 4636 is a relatively isolated galaxy quite far (≳3 Mpc) from the core of Virgo. Nevertheless, its X-ray image shows some azimuthal asymmetry at r ≳ 5re, but this emission may arise from a different source (Trinchieri et al. 1994). It also has a broad, faint stellar distribution characteristic of CD galaxies. NGC 4472 is optically the brightest galaxy in the Virgo Cluster, lies within a small subgroup, and is interacting with a nearby dwarf irregular galaxy, UGC 7636. At radii ≳3re the X-ray isophotes show a tail-like structure that may result from the motion of NGC 4772 through the Virgo Cluster medium (Forman, Jones, & Tucker 1985; Irwin & Sarazin 1996). NGC 4472 has a small kinematically decoupled (nonrotating) core within about 0.096re. NGC 4649 appears to be close to the center of Virgo. However, none of these galaxies shows evidence of unusual X-ray or optical structure at 0.1 ≲ r&solm0;re ≲ 3, where the difference in dark to stellar mass is so apparent in Figure 2.


     (1) A previously unrecognized spatial scale at about 3re is suggested by the similarity of gas temperature profiles in many recently observed elliptical galaxies. (2) The stellar mass-to-light ratios for NGC 4472 and 4649 have been verified by X-ray observations; the interstellar gas temperature is also correct. (3) The mass-to-light ratios in NGC 4472 and 4649 determined from stellar motions are constant over the range 0.1re ≲ r ≲ re. (4) In NGC 4472 and 4649 the hot interstellar gas is supported out to r ∼ re by thermal gas pressure; other means of support such as magnetic pressure or rotation are not evident. (5) Hot interstellar gas in the center of NGC 4636 may be supported by magnetic stresses (B ∼ 10-4 G), a transient unstable region of unexpectedly high temperatures (T ∼ 107 K), or rotation. (6) Elliptical galaxies with comparable luminosities can have quite different distributions of dark matter relative to stellar matter. (7) In NGC 4636 the mass of dark matter is comparable to that in stars at r ∼ re; in NGC 4472 and 4649 the mass of dark halo matter is negligible at r ∼ re. (8) The “stellar” mass-to-light ratios for these three elliptical galaxies as determined by van der Marel (1991) are unlikely to be contaminated by dark matter from the galactic halos. AXAF observations will be useful in interpreting the variation of M*&solm0;LB along the fundamental plane (Pahre & Djorgovski 1997) and in verifying central magnetically supported regions.


     Thanks to Michael Loewenstein and Sandra Faber for enlightenment and encouragement. Our work on the evolution of hot gas in elliptical galaxies is supported by NASA grant NAG 5-3060, for which we are very grateful. In addition, W. G. M. is supported by a UCSC Faculty Research grant and F. B. is supported in part by grant ASI-95-RS-152 from the Agenzia Spaziale Italiana.



Full image (21kb) | Discussion in text

     FIG. 1.—Radial dependence of gas temperature with projected radius in five early-type galaxies all showing maxima at about 3re.

Full image (79kb) | Discussion in text

     FIG. 2.—(a, d, g) Observations and our fits to gas temperature (top) and density (bottom) for Einstein HRI (filled circles) and ROSAT (open circles) data. (b, e, h) Solid lines, Mtot; points are values of Mtot from Irwin & Sarazin (1996) for 4472 and Mushotzky et al. (1994) for 4636; long-dashed lines, stellar mass M*; short-dashed lines, dark mass Mdark ≡ Mtot - M*, which loses accuracy when M* ≥ Mtot; dot-dashed lines, total mass of hot gas. (c, f, i) Solid lines, total mass density &rgr;tot; long-dashed lines, stellar mass density &rgr;* with a horizontal cut at the core or break radius rb; short-dashed lines, dark mass density.


NGCType aD b
LB c
(1010 LB⊙)
M*t&solm0;LB c dLX c
(1041 ergs s-1)
re c

a Type from RC3.
     b Distances NGC 507 and 5044 from Faber et al. 1989; NGC 1399 from Faber et al. 1997; NGC 4636 and 4472 from Gonzalez & Faber 1997.
     c Scaled to the distance in the second column.
     d Mass-to-light ratios from van der Marel 1991.
     e Break radii into core region from Faber et al. 1997.

Image of typeset table | Discussion in text
Up: Issue Table of Contents
Go to: Top of This Article | Search Page | Previous Article | Next Article