Kormendy and Illingworth (1982) found that in the box-shaped bulge of NGC 4565, rotation velocities remained almost constant with increasing height z above the disk plane to z 30" 2.7h-1 kpc (Hubble constant H0 = 50h km s-1 Mpc-1). In contrast, rotation velocity decreased rapidly with increasing z in three galaxies with elliptical or disky-distorted bulge isophotes.
Cylindrical rotation is seen in all box-shaped bulges that have been observed. This includes the prototype, NGC 128 (Jarvis 1990; Bertola and Capaccioli 1977), which contains one of the most peanut-shaped bulges known (Sandage 1961; Jarvis 1990). In it, V(z) is constant up to z = 20" = 8.5h-1 kpc. Three other box-shaped bulges have been measured and show cylindrical rotation, NGC 3079 (Shaw, Wilkinson, and Carter 1993), and NGC 1381 and NGC 7332 (Davies, Illingworth, and Kormendy 1993). Thus boxy bulges are particularly rapid rotators.
They also are more than just a curiosity. In a survey of all large, normal edge-on spiral and lenticular galaxies in the Second Reference Catalog (de Vaucouleurs, de Vaucouleurs, and Corwin 1976), Shaw (1987) found boxy bulges in 20 ± 4% of the objects. Therefore it is important to understand their origin.
Two mechanisms have been suggested. Most relevant to this paper is the suggestion that bars manufacture boxy bulges by vertical heating of the disk through resonant scattering of stellar orbits by the bar (see section 6). An alternative mechanism based on accretion has been suggested by Binney and Petrou (1985), Jarvis (1987), Whitmore and Bell (1988), Hernquist and Quinn (1988), and Statler (1988). If a bulge or elliptical galaxy accretes a dynamically cold object at an oblique angle, then differential precession will phase-mix the orbits until they produce a boxy or X-shaped structure embedded in the bulge. This must happen. But in general we cannot tell whether a particular boxy bulge originated in this or some other way. An accretion origin has been proposed for the boxy distortions in IC 4767 (Whitmore and Bell 1988) and IC 3370 (Jarvis 1987). IC 3370 is an elliptical; it clearly shows cylindrical rotation.
Therefore it is likely that at least two different processes make boxy bulges. Since boxy ellipticals generally do not rotate significantly, they are made in still a third way. Only a few, mostly low-luminosity boxy ellipticals appear related to boxy bulges. Three mechanisms of origin are an inconvenience. However, it is easy to identify anisotropic ellipticals by their slow rotation, and boxy bulges can be built by a bar only if there is a substantial disk. Therefore the critical question is: how common are boxy distortions made by accretion?