7 Discussion and conclusions

We reviewed the identification of lines in the spectrum of $\eta$ Car in the range 4040Å  to 10970Å. Our spectra have higher resolution than previous published data, for the same wavelengths, allowing the deblending of line components for the majority of spectral features. Based on precise radial velocities and variability from high to low state, we identified several new lines of Fe II, [Fe II], [Fe III], and Ca II and rejected a number of previous identifications.

Line parameters in high and low excitation states are presented for the narrow line components separated from the broad ones. We show that line strengths vary between these states in close relation to the excitation energy of the atomic transition. The higher the excitation level, the larger the fading of the line in low state. Broad components of several low excitation lines become stronger on low state, specially Fe II, [Fe II], and Ti II.

DCL suggested that $\eta$ Car is an eccentric binary system in which the secondary component is hotter than the primary. Radial velocity changes of the broad line components of He I, H I, and Fe II between the high and low states are compatible with Doppler motion of the stellar wind, following the primary star along the orbit. The low state occurs at the time of periastron passage. The main X-ray properties of $\eta$ Car like the temperature of the hard spectrum, the flux and column density variability are well accounted for by models of wind-wind collision in an eccentric binary system. Both the secondary star and the wind-wind collision zone produces high energy photons. In that model, the shocked zone becomes deeply enshrouded in the dense wind of the primary star at the periastron passage, screening the hard energy photons from reaching the outer layers of the stellar wind and the nebular region. The model fails just around the periastron passage, what indicates that something else than simple eclipses is playing an important role. It may be the a disk-like structure suggested by previous authors (see below).

The optical/NIR spectrum of $\eta$ Car shows a mixture of high and low excitation lines, requiring more than one excitation source. Permitted lines indicate that the main ionization source is cooler than the star P Cygni (T = 18000 K, Lamers et al. 1996), as this star shows Si IV, Si III, [N III], and Fe III lines compared to Si II, [N II], and Fe II in $\eta$ Car. Similar situation was found in the UV spectrum by Ebbets et al. (1997), who suggested a linear combination of spectral types B2 Ia and B8 Ia. In spite of the much denser wind in eta Car, Ebbets et al. (1997) showed that its spectrum is very similar to that of P Cygni. On the basis of our spectra we estimate the temperature of the primary star in the range 15000K$\leq$T$\leq$18000K. The spectral signatures of higher excitation comes from other(s) source(s) of energy in the system. Such temperature would place the primary star at the cooler side of the evolutionary track shown in Fig. 7 of DCL. A star at that position in the HR diagram is at the end of the H-burning and entering in the core He-burning stage. This implies that the LBV component will soon become a Wolf-Rayet star and the system will be similar to WR140, a WR+O colliding wind binary.

The huge variations observed in the strength of the high excitation lines are not accompanied by photometric variability at a significant level in the optical and UV energy ranges, indicating that the source of energetic photons must have a bolometric luminosity smaller than the primary one. The transition from high to low state spectrum occurs in time scale of 1 month. It is not clear if this behavior is due to an eclipse of the source of hard photons, as seen from the clouds emitting the narrow lines, or an intrinsic variation in the wind-wind colliding zone. In the case of eclipse, there must be some extended structure of high density gas around the primary star, in order to account for the long duration of the spectroscopic event that lasts for 2-8 months (depending on the energy level of the line). Both these mechanisms can explain why broad line components of He I lines fade when that of Fe II strengthens on low state. The lower excitation in the wind of the primary star, as seen from the Earth in low state, accounts for stronger P Cygni profiles and apparently unchanged velocities in the wind signatures. The exact mechanism that controls the excitation in $\eta$ Car lines, however, must be studied in light of additional data.

The complete picture of the $\eta$ Car system, however, is much more complicated than sketched above. The line profiles trace the existence of four main velocity structures:

• a high density, mildly excited stellar wind, accelerated up to rv $\sim$ -1350km s^-1, responsible for broad emission components of permitted lines and P Cygni profiles;
• a compact, slow moving gas layer (rv$\sim$-50km s^-1), probably an equatorial disc, responsible for the shell absorption in He I$\lambda$10830 and Balmer lines;
• a nebular region with a mixing of high and low excitation narrow lines, in a narrow range of velocities (fwhm$\sim$60km s^-1, rv$\sim$-40km s^-1), encompassing the Weigelt blobs around the Homunculus equator, responsible for the narrow line components;
• gas of intermediate velocity rv$\sim$-350km s^-1) emitting high excitation lines, probably ejected from the wind-wind collision zone. Such a velocity field and the region of high density suggested above leads to a picture of the primary star with an equatorial disk and fast polar winds. Non-spherical winds have been suggested by previous authors: Viotti et al. (1989) Davidson et al. (1995), Davidson & Humphreys (1997) pointed out a connection between eta Car and B[e] stars (Zickgraf et al. 1986).

The bipolar flow is composed of two thin expanding shells of gas and dust, enclosing the central stars. Such geometry has been attributed to the initial conditions of the eruption that occurred in the last century. However, we've detected high speed gas ($rv\,=\,-1350$km s^-1) that must be colliding against the internal walls of the Homunculus, causing an acceleratio by the continuous deposition of momentum from the stellar wind. In fact, the early $\eta$ Car spectra (Walborn & Liller 1977) show velocities much lower than presently seen in the Homunculus. A B[e]-like star, with wind speeds increasing from the equator toward the polar regions, could have shaped a bipolar flow like the presently observed Homunculus.

Acknowledgements

AD thanks Peter Conti and JILA for the opportunity of spending a sabbatical year in Boulder and to FAPESP and CNPq (Brazil) for financial support. AK's work was supported by the Deutsche Forschungsgemeinschaft (Wo 296/16-1,2). AK, OS, and BW want to thank all the FLASH/HEROS observers namely Th. Gäng, Th. Rivinius, C.A. Gummersbach, J. Peitz, J. Schweickhardt, J. Kovács, D. Schäfer, H. Mandel, Th. Szeifert, I. Jankovics, H. Lehmann, S. Stefl, D. Baade, U. Thiel, Th. Dumm, and W. Schmutz.