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

Next Contents Previous


Fundamental new constraints on the structure of elliptical galaxies are emerging from studies of isophote shapes. This work may resolve a well-known shortcoming of the Hubble classification scheme: While the sequence SO-Im is one of changing physical properties, that from E0 to E6 is not a sequence of anything fundamental (Tremaine 1987). New observations suggest that ellipticals form a physical sequence that is continuous with SOs at one end. Along this sequence, rotation decreases in importance compared with anisotropic velocity dispersions. This subject is developing rapidly; we summarize it as of December 1988.

CCD photometry shows that the isophotes of elliptical galaxies are usually not perfect ellipses. Some are box shaped and others have disk-shaped distortions along their major axes (Carter 1979, 1987, Lauer 1985c, Jedrzejewski 1987a, b, Jedrzejewski et al. 1987, Michard & Simien 1987, 1988, Bender & Möllenhoff 1987, Bender et al. 1987, 1988a, Ebneter et al. 1988, Franx et al. 1988, Peletier et al. 1988a). It is convenient to parametrize these departures by the amplitude a(4) of the cos (4theta) term in a Fourier expansion of the isophote radius in polar coordinates [see Carter (1978) and the above papers]. Along the major axis, the fractional radial departures from ellipses are typically a(4) / a appeq 1%. Positive values of a(4) / a describe disky isophotes; negative values describe boxy isophotes.

Our discussion of a(4) / a measurements follows an excellent paper by Bender et al. (1988b; hereafter B+88). Figure 3 shows correlations of various parameters with a(4) / a. The upper-left panel shows that rotation is dynamically less important in boxy ellipticals than in disky ellipticals (Lauer 1985c, Carter 1987, Bender 1987, 1988a, Nieto et al. 1988, Wagner et al. 1988, B+88, Nieto & Bender 1988). All disky ellipticals show significant rotation, and many are consistent with isotropic models. Boxy ellipticals have a variety of (V / sigma)* values but include all of the galaxies with negligible rotation. They are also notable for showing minor-axis rotation (Davies & Birkinshaw 1986, Wagner et al. 1988). Bender, Nieto, and collaborators suggest that ellipticals can be divided into two groups: boxy, slowly rotating, anisotropic ellipticals; and rapidly rotating, disky galaxies that connect with the SO sequence.

Figure 3

Figure 3. Correlations of selected parameters with isophote shape. Here 100a(4) / a is the percent inward or outward perturbation of isophote radii along the major axis (B+88); negative values indicate boxy isophotes, and positive values indicate disky isophotes. Most of the galaxies are ellipticals, but a few galaxies with 100 a(4) / a gtapprox 2 are SOs. The upper-left panel shows the rotation parameter (V / sigma)* (Section 8.2). This panel is adapted from Bender (1988a), but with a(4) / a values from B+88 and with (V / sigma)* values added from Davies et al. (1983). The lower-left panel shows deviations of core mass-to-light ratios from the mean relation M / LV propto L0.20±0.04 found in K87; positive values imply that M/LV is larger than average for the galaxy's luminosity. The right panels (from B+88) show ellipticity epsilon and radio luminosity Lradio at 1.4 GHz (W Hz-1).

Such a division also segregates ellipticals by other physical properties. For example, the upper-right panel shows that a(4) / a correlates with ellipticity. The distribution of points is V-shaped. Galaxies that appear almost round are almost elliptical. More flattened galaxies tend to be either box- or disk-shaped (see also Jedrzejewski 1987a). Since ellipticals have a preferred shape of E3.8 (Sandage et al. 1970, Binney & de Vaucouleurs 1981), these flattened galaxies are close to edge-on, and most round galaxies are close to face-on. B+88 therefore suggest that essentially all ellipticals are either boxy or disky when seen edge-on. This again suggests a dichotomy (but see below).

Bender et al. (1987) and B+88 also find that X-ray and radio luminosities of ellipticals show striking correlations with a(4) / a (e.g. Figure 3, bottom right). With few exceptions, only box-shaped ellipticals are strong radio or X-ray sources. Disky ellipticals have X-ray luminosities that are consistent with emission by compact sources only. Also, they are weak radio sources, like SO galaxies (Hummel & Kotanyi 1982). These results are remarkably clear-cut, at least in the B+88 sample. We do not know what they mean.

B+88 also note that average mass-to-light ratios are higher in boxy than in disky ellipticals. This is consistent with Heckman's (1983) finding that powerful radio galaxies have higher M / L values than do other ellipticals. B+88 calculate global M / L values using central velocity dispersions and effective radii. However, slow rotation demonstrates that velocity anisotropies are important in ellipticals, and these affect the calculations (e.g. Binney & Tremaine 1987, Merritt 1988). We therefore checked the B+88 results using core M / L values from K87. These are not immune from anisotropy problems but should be more secure. Only 24 objects with a(4) / a values quoted in B+88 are in both galaxy samples, but they are a fair sample of Bender's M/L values. We confirm Bender's result: The boxy and disky galaxies have mean M / LV values of 7.0 ± 0.6 and 4.2 ± 0.6, respectively. Since M / LV depends on L, this is not the best way to express the result (although these boxy and disky ellipticals happen to have the same mean luminosity). Rather, the bottom-left panel of Figure 3 suggests that a(4) / a may be a second parameter in the M / LV - L correlation. Galaxies with low mass-to-light ratios for their luminosities tend to have larger a(4) / a values.

We believe that this is new evidence for velocity anisotropy. Figure 3 (upper left) shows that boxy ellipticals are especially anisotropic. If the radial component of sigma is larger than the tangential component, then by ignoring this we overestimate M / LV. Similarly, we underestimate M / LV in disky ellipticals because we neglect rotational support. This suggests that there should be a correlation between delta log (M / LV) and (V / sigma)*, and one is observed, but it is not better than the one between delta log (M / LV) and a(4) / a. Therefore, anisotropy is not the whole story. A larger galaxy sample is needed to pursue these questions.

The distribution of points in Figure 3 and other similar correlations in B+88 suggest two alternative interpretations. First, it is possible that a(4) / a measures the distribution of ellipticals along a continuous (but not necessarily uniformly populated) sequence that connects smoothly with S0s at one end. As rotation decreases, galaxies become intrinsically more spherical and anisotropic. However, the most anisotropic galaxies must be flattened and turn out to be boxy. Alternatively, perhaps only the disky ellipticals are the continuation of the Im-S0 sequence, and boxy ellipticals are a separate group with a different origin.

It is clear that at least two kinds of boxy structure are seen (e.g. Bender 1988a, Nieto & Bender 1988), because boxy ellipticals include the slowest rotators, whereas box-shaped bulges of disk galaxies rotate rapidly (K82). Interestingly, the few box-shaped ellipticals that rotate rapidly are small companions of much larger galaxies (Nieto & Bender 1988; see also Jedrzejewski 1987a, Peletier et al. 1988a). This suggests to these authors that the boxy structure is related to interactions (May et al. 1985). Accretion events also seem capable of leaving behind an excess of box or tube orbits (ex-polar rings?) that could create slowly or rapidly rotating boxy structure, respectively (Binney & Petrou 1985, Whitmore & Bell 1988, Hernquist & Quinn 1988, Statler 1988). The extreme case IC 3370 may be an example of the latter (Jarvis 1987). Of course, it is also possible that one or both kinds of boxy structure are primordial.

These developments have great potential for clarifying our picture of galaxy formation and structure. However, it is still early. Also, we have ignored complications like dust, variations of a(4) / a with radius, and other Fourier components in the isophotes. The present discussion will undoubtedly prove inadequate; our main aim is to stimulate further work on these important issues.

Next Contents Previous