6 Velocity field

There are three main kinds of features in the $\eta$ Car spectrum: the narrow components, with radial velocity $rv = -40\pm5$km s^-1 and fwhm $\sim$ 60km s^-1; the broad components with fwhm $\sim$-480km s^-1 and a large range of values in radial velocity; and P Cygni components with rv = -330 to -630km s^-1 and fwhm $\sim$ 60 to 600km s^-1. Figure 4 displays the fwhm of narrow and broad line components as a function of the ionization potential of the ions. There is a trend for lines belonging to ions of ionization potential lower than 30 eV to show lower fwhm. The effect is more clear in narrow than broad line components.

 

Figure 4: Full width at half maximum (fwhm) of broad and narrow line components as a function of ionization potential. It seems to be a trend for lines belonging to ions of ionization potential lower than 30 eV to be narrower than for ions of higher ionization potential

Narrow lines do not show measurable changes in radial velocity or fwhm from high to low state. The small dispersion in radial velocity indicates that all narrow lines are formed in the same region, in spite of the large ranges in energy levels. Davidson et al. (1997), using HST high spatialy resolved spectra, showed that narrow lines are formed in the nebular region $\sim$0$.\!\!^{\prime\prime}$3 from the central star, containing the Weigelt blobs (Hofmann & Weigelt 1988). These authors also showed that broad emission lines are restricted to the central knot. Broad permitted lines are formed in the stellar wind, but it's not clear if the broad forbidden lines can be formed in the wind too. The P Cygni components are formed in regions where the stellar wind apparently didn't yet reach terminal velocity. Several absorption structures, ranging from -130 to -1350km s^-1, appear in different lines, indicating that the velocity field around the star is quite complex. P Cygni components of H I, Fe II, Si II, and N I lines show large increase in radial velocity from high to low states (Fig. 5). This does not necessarily imply variation of the wind flow parameters, as discussed later.

 

Figure 5: Radial velocities of P Cygni components. Velocities are larger in low state (filled symbols) than in high state spectrum (open symbols) for H I, Fe II, Si II, and N I lines. Higher members of the Balmer and Paschen series are formed in a zone of rapid acceleration

High excitation forbidden lines are on average a little broader than others. The narrow components have fwhm $\sim$ 70km s^-1 and the broad ones fwhm $\sim$ 510km s^-1. The radial velocity of the narrow components is similar to all other narrow lines, but the broad component has velocity rv$\sim$-150km s^-1 that is blueshifted compared to the same feature in lines of lower excitation. The broad components of the high excitation lines have a remarkable line profile (Fig. 6). By subtracting the contribution of the narrow component, what remains is an asymmetric line with a distinctive emission at the blue side, rv$\sim$-330km s^-1 (labeled bshd in Table 1) and a fainter shoulder at the red side. Rather than extinction by dust, as suggested by Hamann et al. (1994), this kind of line profile seems to be produced by true anisotropy in the velocity field. The higher the ionization level of the ionic species producing the line, the broader are the line components. The broad component in high excitation lines doesn't change in radial velocity or line profile from high to low state, contrary to H, He I, and some Fe II lines. This indicates that the broad components of high excitation lines are unlikely to be formed in the stellar wind. Present data are insufficient to decide if this third emitting region is bound to the central knot or to the surrounding nebular region. Spectroscopy at radio wavelengths (Duncan et al. 1997) revealed high speed ionized gas in the region containing the Weigelt blobs. Davidson et al. (1997) didn't analyze high excitation forbidden lines and so, presently, we can't exclude that broad components of this type of lines can be formed outside the central object.

 

Figure 6: Line profile of typical high excitation transitions. The broad components are centered at $\sim$-150km s-1, compared to the narrow ones, which are centered at -40 km s-1

 

Figure 7: Line profile of the He I $\lambda$10830 in high (1995) and low excitation phases (1992). The absorption features indicated as NDC (-1061km s-1), P Cyg (-604km s-1), and shell (-50km s-1) remain stationary from high to low state. The structure of the wind flow seem to be unchanged from high to low state, and only the emissivity of the gas at different velocities seems to change

There are different line profiles for different He I lines. All He I lines show radial velocity variations in the broad component from high to low state. The amplitude of radial velocity variations are larger than in any other lines, but the separation between the broad and narrow component is difficult to measure in high state. The P Cygni components are stronger on low than high state, but without changing radial velocity. The He I$\lambda$10830 line has a unique line profile through all the $\eta$ Car spectrum (Fig. 7). He I$\lambda$10830 shows an absorption feature at rv = -50 km s^-1, a little more negative than the narrow emission component. It is actually a shell absorption, as it is narrow, fwhm=30km s^-1, and the line center drops below the continuum level on low state. No radial velocity nor fwhm variation was seen from high to low state in the shell component. This feature indicates the existence of high density, low velocity gas in front of the star, in the line-of-sight towards the Earth. The high velocity components are formed in the wind. During low state, He I$\lambda$10830 shows a remarkable P Cygni profile, centered at $rv\,=\,-630$km s^-1 and extending to more than -1400km s^-1. By adopting the criterium that terminal velocity is measured at $I/I_{\rm c} = 0.95$ (Eenens & Williams 1994; Crowter et al. 1997), $v_{\rm term}$=-1350km s^-1 in low state. A narrow displaced component (NDC), resembling that frequently observed in the UV resonance lines of other hot stars, is seen at $rv\,=\,-1061$km s^-1. On high state, emission fills up the high velocity regions of the P Cygni profile, except for the NDC, which becomes even more distinctive. This component seems to mark a region of critical transition in the wind flow. There is no clear sign that the wind flow properties are changing from high to low state. The high velocity structures in the He I$\lambda$10830 do not change in radial velocity. Variations in the profile of this line seem due to changes in the emissivity of the gas at different radial velocities, caused by an external source.

 

Figure 8: Central portion of H$\alpha$ line profile in high (dotted line) and low state (continuous line). The small velocity of the narrow component (-10km s-1) compared to -40km s-1 in other lines seems to be due to an absorption component centered at -50km s-1

The absence of the narrow component in Balmer H lines seems to be due to the same mechanism that produces a shell component in the He I$\lambda$10830 line. In Figure 8, a faint absorption is seen at -50km s^-1, the blue side of the H$\alpha$narrow component. This component makes the narrow component of H$\alpha$ asymmetric and shifts its peak from the expected -40km s^-1 to -10km s^-1. The strength of this absorption component grows toward the head of the Balmer series, making the narrow component disappear at H 2-6, as seen in Fig. 9. The radial velocity of this absorption component is exactly the same as that of shell component in He I$\lambda$10830. The discovery of a true absorption, in place of the double-peaked profile referred to by previous authors, has important consequences for understanding the reddening and velocity field around $\eta$ Car. The faintness of the narrow components in Balmer lines compared to Paschen series had been attributed to extinction effects. If that was the case, the broad components would suffer from extinction much more than the narrow components, as broad lines are formed in the inner stellar wind and narrow components in the surrounding gas (Davidson et al. 1997). The effect of extinction would cause Balmer lines to have profiles similar to those of the Paschen series, but with smaller broad/narrow flux ratio, contrary to the observations. The observed differences between Paschen and Balmer line series are more likely due to line opacity than extinction effects. The region that forms the absorption component at $rv\,=\,-50$km s^-1 must be very dense in order to produce the remarkable absorption and probably has small spatial extent as the fwhm is only 30 km s^-1, the smallest we measured in our spectra. Could it be produced by an equatorial disk?

 

Figure 9: Line profile of Paschen-$\gamma$ compared with a Balmer-$\delta$ (same upper levels) in high state. H 2-6 (continuous line) shows a strong absorption component at -50km s-1 that shifts the peak of the narrow component to lower velocity than H 3-6 (dashed line). HeI$\lambda$10830 (dotted line) on low state is displayed to reinforce the reality of this absorption feature

The P Cygni components in H lines strengthen toward the upper members of each series, but with radial velocity increasing in the opposite sense. Radial velocities of the P Cygni components are higher on low state than in high state spectrum. The radial velocity curve as a function of wavelength flattens around -500km s^-1 on low state and -470 km s^-1 in high state. This is not, however, the terminal velocity of the wind, as the P Cygni component in the He I$\lambda$10830 line is more negative than in the other lines. A transition in the velocity regime occurs for quantum numbers $n\,=\,5$ in the Balmer series and $n\,=\,11$ in the Paschen series. There is a rapid increase in radial velocity for higher energy levels, which indicates that these features are formed in a region of fast acceleration. It is important to note that the broad components of H lines are shifted toward the blue when P Cygni components are stronger. If this was due to intrinsic variability in the wind, the strengthening of the P Cygni absorption would occur at the expense of the emission component, shifting the emission component toward longer wavelengths. Based on this fact, DCL attributed radial velocity variations in the barycenter of broad H line components to Doppler effect due to orbital motion in a binary system.

The Fe⁺ lines have broad line components with fwhm unchanged from high to low state. Radial velocity variations in the broad component is not evident in [Fe II] lines but is easily seen in many Fe II lines, especially from multiplets 27, 38, 42, 49, and 74. Some of these lines show permanent P Cygni components, which strengthen in low state. The P Cygni features in Fe II lines are more blueshifted in low than in high state, as also seen in H I, Si II, and N I lines (Fig. 5). We have not found the P Cygni components in [Fe II] lines reported by Aller & Dunham (1966). The Fe⁺ broad line components show a discrepancy from the trend between fwhm and ionization potential (Fig. 4). Both narrow and broad components of Fe⁺ show a large scatter in fwhm, compared to other lines. The Fe II and [Fe II] lines show the similar fwhm values, contrary to what was found in narrow line components by Hamann et al. (1994).

Si II $\lambda$6347 shows a very prominent P Cygni component for a brief time interval at the low state. This short-lived component (about 1 month) is a good tracer of the central phases of the spectroscopic event.

Na I$\lambda$$\lambda$5890-5 show seven absorption structures, superimposed on a broad emission. No substantial difference is seen in these components from high to low states. The changes seen in the blue extremity are likely due to the wings of He I$\lambda$5876 line. Besides the interstellar component and the strong P Cygni profile at $rv\,=\,-474$km s^-1, the other five components at $rv\,=\,-103$, -210, -380, -440, and -506km s^-1 are too faint to deserve special attention. The broad emission and the P Cygni components have velocity characteristics similar to that of other low excitation lines, indicating that they also are formed in the stellar wind.

Ca II H & K presents only P Cygni profiles at rv = -470 km s^-1, but emission components cannot be excluded, as these lines are blended at the red side.