Annu. Rev. Astron. Astrophys. 2004. 42: 603-683 Copyright © 2004 by Annual Reviews. All rights reserved

4.5. Box-Shaped Bulges

Bulges with box-shaped isophotes (Figure 15) are well known (Burbidge & Burbidge 1959; Sandage 1961; de Vaucouleurs 1974b). Clear examples are seen in at about one-fifth of edge-on galaxies (Jarvis 1986; Shaw 1987; de Souza & dos Anjos 1987; Lütticke, Dettmar, & Pohlen 2000a). Numerical simulations universally show that bars heat themselves in the vertical direction; they suggest that box-shaped bulges are edge-on bars. If this is correct, then observing box-shaped isophotes is a sufficient criterion for identifying a pseudobulge. Probably indepently of this, boxy bulges also present us with a serious collision between simulations and observations. There are at least two problems. (1) Observations imply that bars are flat in the edge-on galaxies in which they can reliably be identified. (2) Bars and boxy bulges that are clearly distinct from each other occur together in several galaxies. In these galaxies, the major-axis radii of the boxy bulges are much shorter than the lengths of the bars.

That bars heat themselves in the axial direction was an immediate result of the first three-dimensional n-body simulations of unstable disks; it has been a robust theoretical prediction ever since (see Sellwood & Wilkinson 1993 for areview). Combes & Sanders (1981) were the first to point out that n-body bars look like boxy bulges (e.g., NGC 7332) when seen end-on and like peanut-shaped bulges (e.g., NGC 128) when seen side-on (both galaxies are illustrated in Sandage 1961 and in Sandage & Bedke 1994). Edge-on n-body bars looked boxy in some previous papers (e.g., Miller & Smith 1979), but these resulted from the collapse of spherical stellar systems, so it was not clear that their vertical structure was relevant to the evolution of disks. The Combes & Sanders (1981) results have been confirmed and extended by many authors (e.g., Combes et al. 1990; Pfenniger & Friedli 1991; Berentzen et al. 1998; Athanassoula & Misiriotis 2002; Athanassoula 2003). Early papers concluded that the orbits that contribute most to the boxy structure are in vertical ILR with the bar. With two vertical oscillations for each revolution, it is easy to arrange that a star be at its maximum height above the disk plane when it is near apocenter. It then contributes naturally to a box-shaped structure. The importance of vertical resonant heating was emphasized by Pfenniger (1984, 1985) and especially by Pfenniger & Norman (1990). From "sticky particle" simulations, Pfenniger & Norman (1990) found both the mass inflow discussed earlier and vertical heating that fed stars into a component with the scale height of a bulge. Timescales were short, on the order of one-tenth of a Hubble time.

In contrast, Raha et al. (1991) showed that buckling instabilities thicken bars in the axial direction. These are collective phenomena, so they are different from resonant heating. Raha et al. (1991) suggested that buckling instabilities also occurred in the above simulations; Pfenniger & Friedli (1991) acknowledged this possibility. Additional examples of buckling instabilities are in Sellwood (1993b), Kalnajs (1996), and Griv & Chiueh (1998). Further discussion is provided by Toomre (1966); Merritt & Sellwood (1994); Pfenniger (1996a); and Merrifield (1996).

However the heating happens, all of the simulators agree that bars and boxy bulges are connected. A few papers suggest only that disk stars are heated vertically and fed into the bulge (pre-existing or not), giving it a box-shaped appearance. But most authors advocate a stronger conclusion, namely that boxy bulges are nothing more nor less than bars seen edge-on.

What do the observations say? Persuasive observations show that boxy bulges occur in SB galaxies. However, they also suggest that box bulges are not identical to edge-on bars.

The obvious sanity check - that boxy bulges are seen in edge-on galaxies as frequently as well developed bars are seen in face-on galaxies - is passed with flying colors. References are in the first paragraph of this subsection.

A link between n-body bars and boxy bulges is the observation in both of cylindrical rotation to substantial heights above the equatorial plane (see Bertola & Capaccioli 1977; Kormendy & Illingworth 1982; Jarvis 1990; Shaw, Wilkinson, & Carter 1993a; Shaw 1993a; Bettoni & Galletta 1994; Fisher, Illingworth, & Franx 1994; D'Onofrio et al. 1999; and Falcón-Barroso et al. 2004 for the observations and Combes et al. 1990; Sellwood 1993a; Athanassoula & Misiriotis 2002 for simulations). Classical bulges and ellipticals do not rotate cylindrically, as evident from early long-slit spectroscopy (Illingworth & Schechter 1982; Kormendy & Illingworth 1982; Binney, Davies, & Illingworth 1990) and now beautifully shown by integral-field spectroscopy (de Zeeuw et al. 2002; Verolme et al. 2002, Bacon et al. 2002; Copin, Cretton, & Emsellem 2004; Falcón-Barroso et al. 2004; Krajnovic et al. 2004).

Kuijken & Merrifield (1995) and Merrifield (1996) suggest that a kinematic signature of edge-on bars is a splitting in the gas velocities just interior to corotation because the gas there is depleted by radial transport. They observe such velocity splitting in NGC 5746 and NGC 5965 and argue that both galaxies are barred. Merrifield & Kuijken (1999) and Bureau & Freeman (1999) show additional examples. NGC 5746 from the latter paper is shown in Figure 15. The "figure 8" pattern in the emission line is the bar signature. The rapidly rotating gas is identified with a nuclear disk of x₂ orbits, and the slowly rotating component shows the line-of-sight velocities in the disk beyond the end of the bar. The lobes of the "figure 8" are empty because an annulus between the nuclear disk and the end of the bar contains little gas. The idea is that the missing gas has been transported to the center or to an inner ring at the end of the bar. This is an interpretation: an axisymmetric disk containing an annulus devoid of gas would also show the "figure 8". The connection with bars is indirect: (1) in face-on galaxies, gasless annuli are seen only in mature SB(r) galaxies, and (2) [N II] 6584Å emission is much stronger than H in the steep-rotation-curve central disk; this is a possible diagnostic of the shocks expected in the inner parts of the bar (Bureau & Freeman 1999). On the other hand, we noted in section 2.1 that mature SB(r) galaxies - the ones in which an annulus interior to the inner ring has been cleared of gas - do not have the radial dust lanes that are characteristic of shocks. Despite these uncertainties, the almost universal detection of figure-8-like (or at least, X-shaped) line splitting in boxy bulges and - equally important - the lack of such splitting in elliptical bulges argues that the former are found in barred galaxies.

Figure 15. (Top) NGC 5746 (Sb) has a prominently box-shaped bulge (see also Sandage & Bedke 1994). (Bottom) Position-velocity diagram of the [N II] 6584Å emission line along the major axis registered in position with the image. The "figure 8" pattern is interpreted as the signature of a barred galaxy by Bureau & Freeman (1999, who kindly supplied this figure) and by Kuijken & Merrifield (1995).

A third observation that connects boxy bulges with bars is the detection in the disks of a few edge-on examples of density enhancements that plausibly are inner rings (Aronica et al. 2003).

Galaxy mergers probably create a minority of boxy bulges (Jarvis 1987). Also, Patsis et al. (2002) illustrate a simulation that makes a boxy-bulge-like structure in the absence of a bar. However, the conclusion that galaxies with boxy bulges generally contain bars seems reasonably secure.

This is not a proof that they are the same things. There are two problems with the simple, well motivated, and almost universally accepted notion that boxy bulges are edge-on bars.

First is the observation that at least some edge-on bars are flat. The "Rosetta stone" object for this subject is NGC 4762. It is studied in an important paper by Wakamatsu & Hamabe (1984) and is illustrated in Figure 16.

Figure 16. From Wakamatsu & Hamabe (1984), (top) brightness cuts parallel to the major axis of NGC 4762 and displaced from it by z along the minor axis; (bottom) assumed viewing geometry: face-on (upper diagram) and as seen by us (middle sketch and major-axis brightness cut).

NGC 4762 is unique among edge-on galaxies studied so far because it has, in addition to a bulge, three clearcut shelves in its major-axis brightness distribution. All three shelves are visible in the Hubble Atlas images (Sandage 1961), which also show that the bulge is slightly boxy. More face-on galaxies show us that three shelves in the surface brightness profile are common in early-type galaxies that contain a bar, a lens, and an outer ring (see NGC 1291 in Figure 2; NGC 3945 in Figure 5; NGC 2217 and NGC 2859 in the Hubble Atlas). Lenses and outer rings have shallow brightness gradients interior to a sharp outer edge; their two nested ovals are exactly analogous to those in later-type oval galaxies (Figure 9). Kormendy (1979b) emphasizes that the bar almost always fills the lens in its longest dimension. Since SB(lens)0 galaxies are common and since they are the only S0s with three prominent shelves in the brightness profile, interpreting NGC 4762 is reasonably straightforward. Wakamatsu & Hamabe (1984) suggest that the outer shelf is an outer ring, that the middle shelf is a lens, and that the inner shelf is a bar. Since the inner shelf has a smaller radius than the middle shelf, the bar must be seen at a skew orientation (Figure 16). Wakamatsu and Hamabe point out that their interpretation is supported by four observations: (1) The deprojected profile of the outer shelf is that of a ring: it has a minimum interior to an outer maximum. (2) The radius of the outer shelf satisfies the correlation between outer ring radii and galaxy luminosity; (3) the radius of the inner shelf satisfies the correlation between lens radii and galaxy luminosity; both correlations are from Kormendy (1979b). (4) The ratio of the radius of the outer shelf to the radius of the inner shelf is 2.4 ± 0.2, consistent with the average ratio of outer ring to lens radii, 2.21 ± 0.12 (Kormendy 1979b; Buta & Combes 1996 and references therein).

We belabor these points because it is critically important to know that the inner shelf is the bar. The reason is illustrated in the top panel of Figure 16. Wakamatsu & Hamabe (1984) show convincingly that the bar is flat. In the series of brightness cuts parallel to the disk major axis and displaced from it by z = 0", 1", 2", ... 10", the bar disappears as a feature distinct from the lens by z 5". That is, its scale height is less than that of the lens and much less than that of the bulge. The bar is the flattest component in the galaxy.

Also, the bar and the bulge are photometrically distinct. The boxy outer part of the bulge (which is not evident in the brightness cuts in Figure 16) has a radius about half as big as the projected radius of the bar. If the bar fills the lens, then this is about one-fifth of the true radius of the bar.

Similar evidence for flat bars is presented in de Carvalho & da Costa (1987); Lütticke, Dettmar, & Pohlen (2000b); and Quillen et al. (1997).

The second problem with the assumption that boxy bulges are edge-on bars is the observation that both occur together but are distinct from each other in NGC 7582 (Quillen et al. 1997). We see this galaxy at an inclination i 65° that is close enough to edge-on so that the boxy bulge is visible in the infrared but far enough from edge-on so that the bar can be recognized (Sandage & Bedke 1994). In fact, the galaxy has the morphology of a typical oval disk with the bar filling the inner oval along its apparent major axis. Therefore the bar is seen essentially side-on. However, the bar is very flat, the boxy bulge is clearly distinct from it, and the maximum radius of the boxy structure along the disk major axis is about one-third of the radius of the bar.

These observations suggest that boxy bulges and edge-on bars are not exactly equivalent. Interestingly:

Observations and theory are consistent with the hypothesis that at least some and possibly most box-shaped bulges are edge-on nuclear bars. E.g., the two nested triaxial components in our Galaxy proposed by Blitz & Spergel (1991, their Figure 1) are similar to the bar-within-bar structure in Section 4.4. If the inner bar has a radius of 1 - 2 kpc (Binney et al. 1991; Blitz & Spergel 1991; Binney & Gerhard 1993; Sellwood 1993b; cf. Dwek et al. 1995), then it is more nearly the length of typical nuclear bars than of typical main bars. (Scaling our Galaxy to other Sbcs, a normal bar should be ~ 3.5 kpc in radius.) It is the inner bar that looks boxy in COBE images (Weiland et al. 1994; Dwek et al. 1995). We may live in a weakly barred or oval galaxy with a boxy nuclear bar. However, only one-quarter of strongly barred galaxies contain nuclear bars. There may be too few of them to account for all boxy bulges.

Another solution may be the indication in Figure 1.1 (b) of Shen & Sellwood (2004) that the boxy part of their n-body bar is smaller than the bar as a whole. We are indebted to Jerry Sellwood for pointing this out. Athanassoula (private communication) emphasizes the same point.

The safest conclusion - and one sufficient for our purposes - is that boxy bulges are connected with bars and owe their origin to them. All mechanisms under discussion build the box structure out of disk material. We therefore conclude that detection of boxy bulge isophotes is sufficient for the identification of a pseudobulge. However, the disagreement between the bar simulations and the above observations needs attention.