We have argued that bulges are oblate spheroids, but it has long been known that many bulges show a boxy or peanut-shaped (B/PS) morphology (e.g. Shaw 1987). The vast majority of these are probably bars seen edge-on. N-body models show that shortly after a bar forms, it buckles, thickens, and appears almost round, peanut, or boxy when seen end-on, side-on, or at an intermediate angle. The evolution strongly depends on the (dark) halo-to-disk mass ratio, but simulations always result in a B/PS bulge with an exponential vertical light profile (e.g. Combes et al. 1990). The thickening is probably due to vertical heating of the disk through resonant scattering of orbits by the bar (vertical inner Lindblad resonance (ILR)): Ωp = Ω − νz / 2, where Ωp is the bar pattern speed and Ω and νz the stellar rotation and vertical oscillation frequencies. The vertical and horizontal ILRs also converge, so that κ ≈ νz where the maximum thickening occurs (κ is the epicyclic frequency), and the peanut shape is sustained by orbits trapped around the 3D generalization of the (2D) x1 family.
B/PS bulges are not found preferentially in groups or clusters, but they do show an increase of nearby companions. Although accretion (soft merging) can lead, in principle, to B/PS bulges (Binney & Petrou 1985), it probably accounts only for a minor fraction, perhaps related to the “thick boxy bulges” of Lütticke & Dettmar (1999). Hybrid scenarios, where the formation of a bar is triggered by an interaction, are also possible (Mihos et al. 1995). N-body models show cylindrical rotation in the inner parts of B/PS bulges, as suggested by the few observations available (e.g. NGC4565; Kormendy & Illingworth 1982), the fraction of B/PS bulges and barred disks are consistent (Lütticke, Dettmar, & Pohlen 2000), and B/PS bulges show plateaus in their light profiles. However, to prove that B/PS bulges are related to bars, and are thus triaxial, one really wants to probe the potential, requiring kinematics in the bulge region.
Periodic orbits provide a zeroth order view of stellar kinematics. Because of the non-homogeneous distribution of the orbits (see, e.g., Contopoulos & Grøsbol 1989), clear signatures of non-axisymmetry are seen in the position-velocity diagrams (PVDs) of edge-on barred disks (Bureau & Athanassoula 1999). But stars can move on trapped or chaotic orbits, washing out PVD substructures, and more realistic N-body models indeed reveal subtler signatures (Bureau & Athanassoula, in preparation). Gas, however, responds very strongly to a non-axisymmetric potential. Shocks along the bar cause inflow, deplete the gas in the outer bar regions, and lead to characteristic gaps in the PVDs (if a nuclear spiral is formed, requiring an ILR; Kuijken & Merrifield 1995, Athanassoula & Bureau 1999). Line-ratios can also help identify bars (shock versus photoionization). Merrifield & Kuijken (1999) and Bureau & Freeman (1999) applied these diagnostics and showed an almost one-to-one correspondence between B/PS bulges and large-scale bars (Fig. 4), although a few cases may be due to accretion. The strength of the bar also correlates with the boxiness of the isophotes. Thus, contrary to ellipticals, where it is caused by anisotropic velocity dispersions, triaxiality in bulges is due to high rotation (bar instability).
Figure 4. Bar signature in B/PS bulges. Image and ionized gas PVD (on the same scale and along the major-axis) for two B/PS bulges. Left: NGC5746, probably seen side-on. Right: IC5096, probably seen end-on. Adapted from Bureau & Freeman (1999) with permission.
As face-on galaxies often show photometrically distinct bars and (rounder) bulges, it is still unclear whether the above thick bars are truly one with the bulge, or whether a more axisymmetric bulge is simply buried within them. A complete 3D picture of barred galaxies (and bulges) is thus still missing.