5 Line strengths in high and low states

 

Figure 3: Index of variability var as defined by Eq. (1). Panel a) shows the high excitation lines, panel b) intermediate excitation lines and panel c) the low excitation lines. Large symbols are for broad line components, small symbols for narrow components and intermediate symbols for lines in which the components were measured together (as high excitation lines, for example)

Line flux variability in the $\eta$ Car spectrum seems complicated, at first glance. However, it follows a simple general pattern: the higher the excitation level of the transition, the higher the fading of the line in low state. Some of the lowest excitation lines are several times stronger during low state than in high state. For intermediately excited lines, the narrow component suffers strong fading in low state but the broad one fades by a lower extent. We will describe with more detail the strongest and less blended lines of each atomic species. Variations in the line flux are plotted in Fig. 3. In that figure, the x-axis represents the line wavelength and the y-axis the var parameter defined in Eq. (1) and displayed in Column 15 of Table 1. In Fig. 3, large symbols label broad components, small symbols the narrow components, and intermediate size symbols label the lines without separation between the components.

5.1 High excitation forbidden lines

The lines of [Ne III]$\lambda$3868; [Fe III]$\lambda$4658, and $\lambda$4701; [N II] $\lambda$5755, $\lambda$6548 and $\lambda$6583; [S III]$\lambda$6312; and [Ar III]$\lambda$7135 have the largest excitation levels in the optical/NIR wavelength range. On high state, the flux ratio is broad/narrow $\sim$2.3 and both components disappear completely during the low state (Fig. 3a). Lines of [N II]$\lambda$5755, $\lambda$6548, and $\lambda$6583 do not show a complete fading on the low state, because of contamination by blends. We plotted Mn II lines together with the high excitation lines because Mn II are excited by Ly$\alpha$fluorescence (Johansson et al. 1995).

5.2 Neutral helium lines

The He I lines in high state show line profiles similar to that of the highest excitation lines. They have smoother blue shoulders and a little larger flux ratio: broad/narrow $\sim$2.6. On the low state, the narrow component disappears completely and the broad components suffer strong fading (Fig. 3b). He I$\lambda$10830 show a spectacular behavior, ranging from a strength similar to that of H$\alpha$ in high state to almost disappearance in low state. The luminosity in this single atomic transition varies from 3000 $L_{\odot}$ to $\sim$15 $L_{\odot}$ from high to low state. The real figures would be larger, if redening was taken into account. This line is fed by UV continuum and, in other LBV stars, a decrease in the level of line excitation is correlated with a reddening in the color indices and, consequently, a brightening of the star in the optical range. The behavior of $\eta$Caris completely different: the luminosity in the continuum and color indices remain almost constant when lines vary by a factor of 14000% in He I$\lambda$10830, showing that the spectroscopic events are not due to S Doradus variability. This line is a privileged one for timing the 5.52 year cycle and tracing its physical nature. Although it is easily detectable by normal CCDs at high spectral resolution, it hasn't received much attention from observers until now.

5.3 Neutral hydrogen lines

The H I lines have intermediate excitation levels and show a rich pattern of variability. Paschen lines exhibit the typical broad and narrow components with a flux ratio broad/narrow $\sim$ 3.8 in high state. The strength of the narrow components increases toward longer wavelengths. On low state, the flux ratio jumps to broad/narrow $\geq$50, mainly due to the fading of the narrow component as the strength of the broad one changes a slightly (Fig. 3b). For Balmer lines, an absorption component is superimposed on the narrow emission seen in Paschen lines, as described in the next section.

5.4 Fe II and [Fe II] lines

Many Fe⁺ lines show very strong narrow components relative to the broad ones. On high state, the flux ratio is broad/narrow $\sim$1.3. On low state, this ratio grows to broad/narrow $\sim$4. This is due to a combined effect by which the broad components become stronger in low state and the narrow components become fainter. The fading of the narrow component in Fe II (var =$80\pm 20$) is more pronounced than that of [Fe II] (var =$36\pm 12$). The strengthening of the broad components changes in the opposite way: var =-47 $\pm$ 20 for Fe II and var =-116 $\pm$ 93 for [Fe II], as can be seen in Fig. 3c.

Several Fe⁺ lines, however, show a pronounced narrow line component, relative to the broad one, and much larger variations from high to low state. For example, Fe II$\lambda$8490, has a very strong narrow component in high state that disappears almost completely in low state. This line is thought to be excited by fluorescence from Ly$\alpha$ photons (Johansson & Hamann 1993). The Fe II$\lambda$$\lambda$2508-10 lines are also excited by the same mechanism and disclosed large variations in the 1981 spectroscopic event (Viotti et al. 1989). This suggests that other Fe⁺ lines that follow the same kind of pattern could also be powered by fluorescence. The Fe II$\lambda$9997 line does not show a particularly strong nor variable narrow component, in spite of being thought to be excited by Ly$\beta$ fluorescence. This suggests that fluorescence is not operative for Fe II$\lambda$9997 or that complicate radiation transfer effects are involved.

5.5 Other lines

The Si II$\lambda$6347 line behaves like hydrogen Paschen lines (Fig. 3b). In high state, the flux ratio is broad/narrow $\sim$4 and in low state broad/narrow $\geq$50. The N I and [N I] lines resemble Fe IIand [Fe II] lines (Fig. 3c). The broad/narrow flux ratio in neutral nitrongen lines, however, is smaller than that in typical Fe II lines. The N I narrow line fluxes in low state are $\sim$50% lower than the flux in high state. The [N I] lines, however, remain almost at the same level and both N I and [N I] have enhanced broad components in low state.

The Ni II$\lambda$6666 line is representative of the other Ni II lines. The total energy flux in this lines decreases a little from high to low state in the same way as Fe⁺ broad lines (Fig. 3c). The broad component of Ni II is difficult to separate from the narrow one and the flux ratio is broad/narrow $\sim$1.2 in high state and broad/narrow $\sim$1.6 in low state. The narrow component is fainter in low than in high state, which indicates that the broad component may be constant or become a little stronger in low state, as Fe II lines (Fig. 3c).

The Ti II lines are faint in $\eta$ Car spectrum but strengthen remarkably in low state, as can be seen in Fig. 3c (var = $-100\pm 70$). The broad line components are very faint in our spectra. However, based on the fact that in all other lines the narrow component is fainter in low than in high state, we infer that the Ti II broad components strengthen in low state. This kind of variability reminds us of that seen in Fe II lines of multiplets 42, 48, 49, 74 etc., but with larger amplitude from high to low state. It is intriguing that many old papers described Ti II line profiles in some detail while it was painful to get even equivalent widths from our spectra. Are Ti II lines fainter today than several decades ago? This seems not to be the case, as the strengths of several other lines (He I$\lambda$4471, He I$\lambda$5876, [N II]$\lambda$5755 and Fe II) in high state seems to have been unchanged through all this century. It is not expected that only Ti II lines have varied. Our guess is that the S/N ratio in our CCD spectra is lower than were in photographic spectra in the blue spectral region.

Only the narrow component is seen in Cr II lines and they show smaller line strengths in low than high state, similar to the narrow components of Fe II lines.

The Na I$\lambda$$\lambda$5890/5 lines do not show the narrow component, a rarity in $\eta$ Car's spectrum. The line flux is difficult to measure because of blending with HeI 5876Å, but almost no variation is seen when superimposing the high and low state spectra.

The only oxygen line with firm identification in our spectra is O I$\lambda$8446 (O I$\lambda$7776 was not observed in 1992 and will not be discussed here). Only the narrow component can be seen in the blend involving H I, O I, and Fe II lines. This line suffers a huge fading in low state. It might even disappear completely in the central phase of the event. The [O I]$\lambda$6300 is not a reliable identification for the line at 6299.49Å, as discussed previously.

The Ca II lines are also difficult to measure because of severe blends. The H and K lines in the blue spectral region show only P Cygni components. The narrow components in two members of the NIR triplet are definitively present and they seem to be constant or possibly a little enhanced in low state (Fig. 3c).