3 Description of the spectrum

The remarkable characteristic of the optical spectrum of HD 101584 is the fact that different spectral regions resemble different spectral types. The spectrum in the UV region is similar to that of $\alpha$ Lep which is an F-supergiant (Bakker 1994). The optical spectrum in the range 3600 Å - 5400 Å is dominated by absorption lines. Most of them are due to neutral and single ionized lines of Ti, Cr and Fe. The CaII H and K absorption lines are strong. The strength of the absorption lines are similar to that observed in an A2 supergiant. In the yellow and red spectral regions, most of the lines are in emission (Fig. 1). The emission lines show complex line profiles. The absorption lines of NI, OI, CII and SiII are broad. The Paschen lines are in absorption. Some of these absorption lines are blended with emission lines and many have asymmetric profiles. The OI lines at 6156 Å are blended with emission lines of FeI. The NI lines are strong and show asymmetric line profiles. The blue wing is shallow compared to the red wing. The CII lines at 6578 Å and 6582 Å are weak. The Na D lines, KI 7700  Å (Fig. 2), the CaII IR triplet lines (Fig. 3), [OI], [CI] and MgI 6318.7 Å lines are found in emission. The OI triplet lines (Fig. 2) are very strong indicating an extended atmosphere and NLTE effects.

  

Figure 1: High resolution spectra of HD 101584 obtained with the ESO CAT-CES

 

Figure 1: continued

 

Figure 1: continued

  

Figure 2: The spectrum in the upper panel shows several nitrogen lines and emission lines of Fe. The lower panel shows the KI 7699 Å in emission and the strong absorption due to OI triplet at 7777 Å

  

Figure 3: CaII IR triplet lines showing P-Cygni emission.This spectrum is of 2.5 Å resolution, obtained from VBO, Kavalur

3.1 P-Cygni profiles

The H$\alpha$ line has a very strong P-Cygni profile indicating an outflow. The profile looks very complex. It shows at least 6 velocity components. The FeII line at 6383 Å is in emission and the profile is very similar to that of H$\alpha$ (Fig. 4). Similar behaviour of the 6383 Å FeII line and H$\alpha$ line is also noticed in the post-AGB F supergiant IRAS 10215-5916 (García-Lario et al. 1994). The H$\alpha$ and the FeII 6383 Å line show an outflow velocity of 100 $\pm$ 10 km s^-1. The H$\beta$ line also shows a P-Cygni profile. It has a broad emission wing at the red end. This indicates that the line forming region is extended. The H$\beta$, NaI D1, D2 and the CaII IR triplet lines (Fig. 3) show an outflow velocity of $75\pm20$ km s^-1. The velocity structure seen in these P-Cygni profiles could be due to emission from different shells formed during the episodic mass-loss events.

  

Figure 4: P-Cygni profile of H$\alpha$ and FeII(6383 Å) lines showing similar velocity structures

3.2 FeI and FeII emission lines

The presence of numerous emission lines of FeI and FeII makes it possible to derive the physical conditions of the line forming region. From the curve of growth analysis of the FeI and FeII emission lines (Viotti 1969), we have derived $T_{\rm exi}=6300\pm 1000$ K and $5550\pm 1700$ K respectively (Fig. 5). The scatter found could be due to the fact that the lines are not optically thin. On the other hand, there are only few emission lines of FeII present in the spectra and thus the estimate from FeII might not be accurate. In order to determine whether the large scatter observed in Fig. 4 is reflecting optical thickness effects we have done self-absorption curve (SAC) analysis (Friedjung & Muratorio 1987) for the FeI emission lines.

  

Figure 5: Curve of growth analysis of Fe emission lines. + represents the FeI lines and $\diamondsuit$ represents FeII lines. The slope gives $T_{\rm exi}=6300\pm 1000$ K for FeI lines and $T_{\rm exi}=5550\pm 1700$ K for the FeII lines. The large dispersion is because the lines are optically thick. The errors bar show the error in the least square fit

SAC is a kind of curve of growth applied to emission lines, but it has certain advantages as compared to the classical emission line curve of growth analysis. This method of analysis is valid also for optically thick lines. It deals with each transition separately, so that it is possible to get the population of different levels without assuming a Boltzmann distribution. In this curve, a function of the line flux emitted in the different transition of a given multiplet is taken in such a way that it is constant for a optically thin uniform medium. As the optical thickness increases the curve will move towards a straight line inclined at $-45^{\rm o}$. The shape of the SAC in Fig. 6a shows the lines are optically thick. The shape of the SAC is obtained by shifting all the multiplets with respect to a reference multiplet. Here we have taken multiplet 207 as reference. The X and Y shifts of each multiplet gives the relative population of the lower and upper level with respect to the reference multiplet. Figures 6b,c shows the Y and X shifts versus the upper and lower excitation potential from which we derive the $T_{\rm exi}=6100\pm 200$ K.

  

Figure 6: a) The plot shows the shape of the SAC. The + sign indicates multiplets 167, 168, 169, 204, 207, 208 of Fe I having similar excitation potential. $\diamondsuit$ indicates multiplets 1002, 1005, 1014, 1077, 1105, 1140, 1153, 1220, 1229, 1277 of FeI. The fit was obtained after shifting higher multiplets 1002, 1005, 1014, 1077, 1105, 1140, 1153, 1220, 1229, 1277 w.r. to the lower multiplets 167, 168, 169, 204, 207, 208

 

Figure 6: b) The fit shows the distribution of upper level population of different multiplets of FeI with respect to the multiplet 207, versus the upper excitation potential

 

Figure 6: c) The fit shows the distribution of lower level population of different multiplets of FeI with respect to the multiplet 207, versus the lower excitation potential

3.3 Forbidden lines

The forbidden emission lines at 5577 Å, 6300 Å and 6363 Å of neutral oxygen are present in the spectra. The forbidden line of neutral carbon at 8727 Å is also seen. The 6300 Å line is blended with ScII line and the 5577 Å line is very weak. We have calculated the I(6300)+I(6363)/I(5577) to be 13.3. From the flux ratio we can calculate $T_{\rm e}$ (Osterbrock 1989). This flux ratio is not very accurate because of very the weak 5577 Å line and poor signal to noise spectrum. For the flux ratio of 13.3 we derived a function depending on the electron density $N_{\rm e}$ and temperature $T_{\rm e}$. Figure 7 shows the $N_{\rm e}$ and $T_{\rm e}$ contours for different values of flux ratio around 13.3. Since we do not see any other forbidden lines which are sensitive to the electron density, we could not fix both $N_{\rm e}$ and $T_{\rm e}$ uniquely. But assuming a temperature derived from the Fe emission lines, an electron density of $1 \ 10^{7}$ is obtained. For this value of electron density and temperature the C/O = 0.5 $\pm$ 0.2 has been obtained.

  

Figure 7: Plot of electron density $N_{\rm e}$ and electron temperature $T_{\rm e}$. The dotted line is the contour for the observed ratio (13.3) of [O]I lines 5577 Å, 6300 Å and 6363 Å. Each contour in the plot is for a change in the flux ratio of 0.5