$\begin{figure}\par\hbox{ \psfig{figure=De142_f4.eps,height=4.3cm} \hspace{1.4cm} \psfig{figure=De142_f5.ps,width=4.3cm} } \par\end{figure}$

Figure 4: Simple geometrical model. The left panel sketches the position of the rings viewed from above the plane defined by the main axis of the homunculus and the line through the centers of the rings. There are two solutions because in the projection seen from Earth it is unknown which ring is in front. The right panel shows the projected rings overlaid on the 11.9 $\mu$ m image.

PL00 show that the inner homunculus structure seen at 20 $\mu$ m represents regions of increased column density. If indeed the rings trace a density enhancement, then they could be denser rings in a bipolar nebula, similar to the rings in SN1987A (Burrows et al. 1995). We also note that this geometry shows strong resemblance to that seen in PNe such as He2-113 (Sahai et al. 2000) and Hb 12 (Welch et al. 1999) including the two rings, the misalignment between the bipolar structure and the rings, and the offset of the central star with respect to the ring structure. It seems reasonable to conclude that the physical mechanism causing these structures is generic and acts in high mass as well as in low mass objects.

A number of mechanisms to produce double rings in bipolar nebulae have been proposed for PNe (Icke 1988) and SN1987A (Crotts & Heathcote 2000), but which one applies where has not been established. In all cases the rings are perpendicular to the major axis of the nebula, which for $\eta$ Car implies that the major axis of this inner nebula is at a significant angle (37 or 58 degrees) to the major axis of the homunculus.

It seems difficult to avoid the conclusion that there must have been a change in the orientation of the outflow between the moment of production of the homunculus and the creation of the double ringed structure. This strongly favours the binary model for the $\eta$ Car system. The shredded appearance of the skirt in the HST images and the proper motion of the condensations indicate that the equatorial regions were highly perturbed by the great eruption. It is therefore likely that the rings were produced after the great eruption. 1) The change of orientation could result from an asymmetry in the mass loss during the great eruption. 2) It could be due to tidal interaction of the eccentric binary with material in its environment. The required mass for such a process can be estimated in the following crude way. A gravitational perturbation can act most easily in the apocenter. In a Keplerian motion about a mass M_* with eccentricity e and semi-major axis a, the apocenter distance and velocity are given by r=a(1+e) and $v^2=\frac{GM_{*}(1-e)}{a(1+e)}$ . The required acceleration to change the orbital inclination by about 1 radian is of the order of v²/a(1+e). Since the great eruption of 1840, about $N=160/5.5 \approx 29$ orbital periods have passed. In order to produce the required total change in N steps, a disturbing mass $\tilde{M}$ at distance R would have to fulfil the condition $\frac{\tilde{M}}{M_{*}}(\frac{a}{R})^2 = \frac{1}{N}\frac{1-e}{(1+e)^2} \approx 0.0054$ , with M_* the reduced mass of the binary. Therefore with e=0.6 and $R\approx a$ , then $\tilde{M} \approx 5~\times10^{-3} \,M_{*}$ . A moderate amount of mass close to the binary could already be sufficient to explain the observed change in the system orientation.

4 Discussion