Annu. Rev. Astron. Astrophys. 1989. 27: 235-277 Copyright © 1989 by Annual Reviews. All rights reserved

4. DUST IN ELLIPTICAL GALAXIES

According to classical definition (Hubble 1926, de Vaucouleurs 1959, Sandage 1961), elliptical galaxies contain no dust. Galaxies with dust have usually been given So or later-type classifications. Now sensitive searches are finding that even the remaining, classical ellipticals often contain dust. This section summarizes its properties. Other recent reviews have been given by Schweizer (1987), Bertola (1987), and Nieto (1988; hereafter N88). A catalog of dusty ellipticals has been published by Ebneter & Balick (1985).

Progress in this subject has depended critically on the ability to detect subtle absorption features superposed on steep brightness gradients. CCD surveys are especially powerful: Their dynamic range is large, and the data can easily be subjected to digital "unsharp masking" (Sandage & Miller 1964, Malin 1977). This is done by dividing the image by a model of the overall brightness distribution without the fine structure. The model can be a smoothed version of the original image, or a synthetic galaxy image with the best-fitting elliptical isophotes, or an image taken in a redder bandpass. (In the last case, the ratio is a color image.) These techniques show that 50% of bulges and elliptical galaxies contain dust.

4.1. Frequency of Occurrence of Dust

A few dusty ellipticals have been known for years. They received little systematic attention until Bertola & Galletta (1978) pointed out that several ellipticals have dust lanes along their minor axes and therefore may be prolate. This had immediate impact because of the recent discovery (Bertola & Capaccioli 1975, Illingworth 1977) that most bright ellipticals are dynamically supported not by rotation but by velocity dispersion anisotropy, which suggests that they are triaxial (Binney 1976, 1978a, b, 1982a, b).

Systematic surveys for dust followed, and detection rates increased as search techniques improved. Hawarden et al. (1981) examined carefully chosen diskless galaxies on the ESO/SRC IIIa-J and Palomar sky surveys and found a substantial number (40) with dust. Sadler & Gerhard (1985a, b) found dust in 23 ± 7% of ellipticals with mean diameters of at least 2' on the ESO B survey. Like all such estimates, this is a lower limit. The dust is usually in well-defined, nearly edge-on disks; this implies that many face-on dust distributions are going undetected. Sadler & Gerhard estimated that the true fraction of ellipticals with dust is at least 40%. A CCD survey by Sparks et al. (1985) led to similar conclusions. More recently, Djorgovski & Ebneter (1986) and Ebneter et al. (1988) have detected dust in 36% of the 116 ellipticals they studied. Finally, CCD photometry with the CFHT (Kormendy & Stauffer 1987; J. Kormendy, to be published) shows a still higher detection frequency, because of the excellent seeing on Mauna Kea. Dust distributions are often so small that they are barely detected even with the CFHT. Many more may await discovery with the Space Telescope.

This dust was also found by the Infrared Astronomical Satellite (IRAS). Detection frequencies in co-added IRAS survey data on bright, nearby ellipticals are comparable to or larger than those seen optically (Jura et al. 1987). Optical and IRAS photometry both imply that typical dust masses are ~ 10⁵ - 10⁶ M; for canonical gas-to-dust ratios, this corresponds to ~ 10⁷ - 10⁸ M of cold gas (e.g. Sadler & Gerhard 1985b, Sparks et al. 1985, Jura 1986, Jura et al. 1987, Véron-Cetty & Véron 1988).

It is now clear that dust in elliptical galaxies is not rare. This is one more piece of evidence that ellipticals contain substantial amounts of interstellar matter [see Schweizer (1987) for a review). Some gas is acquired by accretion (see the next section), and some is expected from mass loss during stellar evolution (Sandage 1957, Faber & Gallagher 1976). With the discovery that ellipticals generally contain 10⁹ - 10¹⁰ M of X-ray-emitting gas (e.g. Forman et al. 1985), the idea that they are surprisingly free of interstellar matter has disappeared.

Although the precise frequency is uncertain because of classification bias, the above surveys show that bulges contain dust still more often than ellipticals. Even prototypical bulges can be riddled with dust (e.g. M31; Johnson & Hanna 1972, Kent 1983, McElroy 1983), as well as ionized (Ciardullo et al. 1988) and other gas.

4.2. Origin of Dust: Further Evidence for Galaxy Mergers

There is strong evidence that many large-scale dust and H I gas distributions are accreted. The most convincing evidence is kinematic: The gas and dust are usually in disks rotating at random orientations with respect to the optical major axis [H I (e.g. Gallagher et al. 1977); H II (Schweizer 1980, 1981a, 1982, Davies & Illingworth 1986, Caldwell et al. 1986); H II associated with dust (Burbidge & Burbidge 1959, Graham 1979, Marcelin et al. 1982, Möllenhoff 1982, Sharples et al. 1983, Caldwell 1984, Bertola et al. 1985, Möllenhoff & Marenbach 1986, Wilkinson et al. 1986, Bland et al. 1987, Varnas et al. 1987, Galletta 1987, Bertola & Bettoni 1988, Bertola et al. 1988a, b, Möllenhoff & Bender 1988)]. Minor-axis dust lanes rotate at right angles to the stars. Sometimes dust lanes and stars even rotate in opposite directions. This gas cannot come from internal mass loss. Accretion is also suggested by the morphology (although dust is not correlated with the presence of ripples and shells; Schweizer & Ford 1985). At large radii, dust lanes often show S-shaped warps or transitions from regular disks to irregular distributions. Such behavior is expected for material just settling into equilibrium, since orbital clocks run slower at larger radii. Note that accretion does not require cannibalism; gas can be donated by a galaxy that gets away (Schweizer 1987).

Small dust lanes are more common than large-scale dust distributions. Whether or not they have the same origin is not clear. They are usually well-defined rings or disks near the center, often oriented parallel to the major axis. Many resemble the inner dust lanes commonly seen in SO and spiral galaxies (Sandage 1961). Some or even most may have an internal origin. However, a folklore is developing, perhaps prematurely, that dust in ellipticals is always accreted. Kinematic constraints are badly needed on the fraction of inner dust disks that are accreted. Is the fraction of counterrotating cases near 50% (as it is for large-scale dust lanes; Bertola et al. 1988a, b), or is it much smaller? At stake is a better understanding of how much secular evolution results from mergers and how much from internal processes.

4.3. Three-Dimensional Shapes of Ellipticals Containing Dust

Bertola & Galletta's (1978) pioneering paper raised the hope that dust-lane geometry could be used to measure galaxy shapes. However, the large number of free parameters make this complicated. The results provide further evidence that ellipticals are triaxial, and they sometimes favor an oblate or a prolate configuration, but tbey have not securely told us the shape of any individual galaxy.

This subject is reviewed in detail by Merritt & de Zeeuw (1983) and will also be reviewed in the next volume of this series by de Zeeuw (1990). Thus our summary of the predictions is brief. Gas in a spheroidal or triaxial potential settles into certain preferred planes through differential precession and dissipation (Kahn & Woltjer 1959, Gunn 1979, Lake & Norman 1983). Consider first the simplest case, in which the shape of the potential does not rotate. Then the gas settles into one of two planes, i.e. perpendicular to the shortest or to the longest axis (e.g. Heiligman & Schwarzschild 1979, Tohline et al. 1982, Steiman-Cameron & Durisen 1982). In a spheroidal galaxy, only the equatorial orbits are stable; polar orbits gradually tip over into the equatorial plane. If we knew that ellipticals are spheroids, then those with minor-axis dust lanes would be prolate and those with major-axis dust lanes would be oblate. But ellipticals can be triaxial. Then, for some infall angles and galaxy shapes, gas is captured into polar orbits. Therefore, a dust lane along a particular axis is consistent with either oblate or prolate structure. Already there is no unique relationship between galaxy shape and dust-lane geometry.

The next complication is that the figure can tumble (angular velocity _p 0). However, the angular velocity we measure is that of the stars, and they stream through the figure (as they do through bars and spiral arms). Therefore, we do not know _p; in fact, we are as interested in estimating _p as in finding the shape of the galaxy. Tumbling elliptical galaxies allow additional equilibrium orbits, as summarized in Figure 1.

Figure 1. Stable orbits of gas in a rotating triaxial galaxy (adapted from Merritt & de Zeeuw 1983). As illustrated, the figure tumbles in the direction of stellar rotation (p > 0); if p < 0, the sense of gas rotation is reversed. Assume that the figure rotates about its shortest or longest axis (left). The second column gives the kind of orbit, and the third sketches resulting dust lanes seen edge-on. Anomalous orbits have different orientations at different radii (van Albada et al. 1982). They are the analogues of polar orbits in a stationary potential; at small radii, where p is unimportant, they are polar. At large radii, the figure rotates several times during an orbit and so is effectively oblate-spheroidal; then the orbit is equatorial (Simonson 1982). In between, the orbits have skew orientations determined by the Coriolis force. The schematic illustrations of dust lanes show the directions of stellar and gas motion; indicates approach, and indicates recession. The right column states the kinematic signature, i.e. the sense of rotation of the dust lane with respect to the stars.

Half of the configurations shown in Figure 1 may be uncommon. If a galaxy tumbles about its long axis, stellar rotation velocities will be large along the minor axis and zero along the major axis. This has been observed in only a few galaxies (e.g. NGC 4261; Davies & Birkinshaw 1986, Wagner et al. 1988). We assume that ellipticals usually tumble about their short axes. Then stable major-axis dust lanes should be prograde. Minor-axis dust lanes should be perpendicular at small radii and should twist at large radii and show retrograde rotation.

What do we observe? Major-axis dust lanes counterrotate in two of the four ellipticals studied (Bertola et al. 1988a); retrograde gas velocities are also seen in the SB0 galaxy NGC 4546 (Galletta 1987). Of seven minor-axis dust lanes measured so far, three show retrograde-rotating twists [NGC 1316 (), NGC 5363 (Sharples et al. 1983, Bertola et al. 1985), and A0609-33 (Möllenhoff & Marenbach 1986)] and four show prograde twists (NGC 4589 (Möllenhoff & Bender 1988), NGC 5128 (e.g. Davies et al. 1984, Bertola et al. 1985, Wilkinson et al. 1986, Bland et al. 1987), NGC 5266 (Caldwell 1984, Möllenhoff & Marenbach 1986, Varnas et al. 1987), and A0151-49 (Sharples et al. 1983, Bertola et al. 1985)].

The hypothesis that these dust lanes are in equilibrium can be saved if _p < 0 (e.g. Varnas et al. 1987). Although it is difficult (Vietri 1986), Vietri (1988) has succeeded in constructing at least one realistic dynamical model in which retrograde figure rotation is slow enough, and prograde stellar streaming large enough, so that the sum (i.e. the observed galaxy rotation velocity) is opposite to the tumbling direction at some radii (see also Freeman 1966). On the other hand, N-body models that collapse and become bar-unstable have always resulted in _p > 0 [see van Albada (1987) for a review]. It is not clear whether retrograde tumbling is a viable interpretation.

Therefore the observations suggest that many dust-lane warps are transient - that gas has settled to a preferred plane at small radii but still remembers the merger geometry in the warp (Tubbs 1980, Simonson 1982, Bertola et al. 1985, Wilkinson et al. 1986, Schweizer 1987, Schwarzschild 1987, Möllenhoff & Bender 1988). This possibility has existed from the beginning; it was resisted mainly because, warps then tell us less about galaxy shapes. But the fact that dust lanes are often regular at small radii and irregular farther out (e.g. NGC 1316; Schweizer 1980) should already have convinced us that settling into principal planes is not always complete.

We dwell on this subject because it has seemed to be the most rigorous new method to measure the shapes of individual ellipticals. It remains promising. But even with photometry and kinematic data, it is difficult to unravel the many unknowns: the amount of triaxiality, the orientation of the galaxy, the pattern rotation speed, and the question of whether dust has settled into equilibrium. There are other uncertainties that we have not discussed. For example, a slowly rotating elliptical may not be exclusively oblate or prolate; it may at some radius change from one to the other. And some conclusions summarized in Figure 1 may be violated in special potentials. Simple deductions seem reasonably secure: (a) Ellipticals are generally triaxial; (b) some are prolate and others are oblate; and (c) some warps imply that _p 0 and others imply nonstationary structure. But more detailed progress has been elusive.

We know of no simple remedies. As Merritt & de Zeeuw (1983) point out, better statistics would help. We may be basing far-reaching conclusions on configurations that turn out to be rare. New discoveries of systematic behavior may reduce the available parameter space. But it appears that the implications of dust-lane geometry and kinematics become statistical. Unless further work sheds new light, we still do not know how to measure the shapes of individual ellipticals.

4.4. Dust and the Distinction Between E and SO Galaxies

The presence of dust contributes to a blurring of the distinction between E and S0 galaxies. This is partly just a practical problem of classification. When ellipticals were dust free by definition, dusty galaxies were easy to classify. If we now adopt as the main classification criterion the presence (S0) or absence (E) of a disk, then it is difficult to distinguish ellipticals from S0s with faint disks [see K82 and Capaccioli (1987) for reviews]. A significant number of galaxies must be misclassified in the literature. If the E-S0 sequence is continuous, this makes little difference for an individual object (Schweizer 1987). But it can systematically affect galaxy samples selected for physical studies.

There is also a more difficult problem of principle. Since dust lanes and gas can form stars, a galaxy can change our perception of its morphological type. For example, a slowly rotating, bright elliptical may, through judicious cannibalism, grow a disk and come to look like an SO. This would contribute noise to correlations between physical properties and type. For example, even if real bulges rotate rapidly, there would be apparent exceptions because some SOs started life as ellipticals. This is an example of how secular evolution can obscure a physical correlation that was set up during an earlier phase of galaxy formation. Since far-reaching conclusions are often based on a few galaxies with surprising behavior, we need to be careful to understand and allow for secular evolution.