Astron. Astrophys. 330, L17-L20 (1998)

3. Discussion

The stars observed so far at high resolution in our sample are listed in Table 1, along with their spectral class, period, distance and expansion velocity. Since no CO measurements are available for R Cen, the expansion velocity as derived from the SiO maser lines is given here. The parameter f is a measure of the symmetry of the light curve, i.e., the ratio of the time it takes to reach the maximum and the period. Values for this parameter and for the period are taken from measurements available from the American Association of Variable Star Observers (AAVSO, Mattei private communication). Unfortunately, for EP Aqr both are unknown. All stars in Table 1 have relatively low mass loss rates. They tend to be type I OH maser sources, i.e. the OH 1665/1667 MHz main-line maser is stronger than the 1612 MHz maser (e.g. Szymczak et al. 1995, Chapman et al. 1994, Slootmaker et al. 1985). The CO J=1-0 lines are weak or not detected (Loup et al. 1993 and references therein). For most of them H₂ O (Yates et al. 1995, Krocker and Hagen 1983, Lepine and Paes de Borres 1977) and SiO maser emission has been reported (Cho et al. 1996, Haikala et al. 1994).

Table 1. Stars observed at full-grating resolution.

We display the 7-16.5 µm section of the AOT01 SWS spectra for the stars in our sample in Fig. 1, including for comparison the spectrum of µCep which has no 13 µm dust emission feature. Note that the strength of the 13 µm dust feature varies from one object to another and that the 9.7 µm silicate feature also shows remarkable differences. Two emission features at 10.05 and 11.05 µm and an absorption dip at 9.35 µm are instrumental since they are present in the RSRF. The silicate profiles exhibit a shoulder on the red side compared to the "classical" silicate shape of µCep. Little and Little-Marenin (1990) studied a large number of IRAS/LRS spectra of Mira variables and noted variations in the shape of the silicate feature in many stars. It is possible that the profile is broadened by the presence of aluminum oxide which has a peak in its absorption efficiency around 11 µm (Eriksson et al. 1981, Begemann et al. 1997).

Fig. 1. The 7-16.5 µm section of the SWS/AOT01 spectra for the stars in our sample. Flux levels are in arbitrary units and spectra are shifted for clarity of presentation. Vertical tick marks indicate the positions of emission lines detected. The spectrum of µCep is plotted for comparison.

From Fig. 1 it is clear that all five objects show emission lines at 13.87 and 16.18 µm. RX Boo shows an additional emission line at 14.97 µm while W Hya and R Cas show absorption lines at the same wavelength. As a first step in the identification of these lines we note that the absorption line at 14.97 µm occurs at the wavelength of the fundamental ro-vibrational -band of gaseous CO₂, consistent with the fact that both W Hya and R Cas show a deep absorption line at 4.27 µm due to the fundamental ro-vibrational -band of CO₂ in our full grating scans (AOT1).

Fig. 2 shows the AOT06 spectra of W Hya and EP Aqr at a spectral resolution of 1500. In the spectrum of EP Aqr, we see two additional emission lines at 13.48 and 15.40 µm. On closer examination of the fast AOT01 grating spectra (Fig. 1) these two extra lines are also present in the spectrum of RX Boo. R Cen also exhibits the 13.48 µm line.

Fig. 2. High-resolution (AOT6) spectra of W Hya, showing two ro-vibrational bands of CO2 in emission and one in absorption, and of EP Aqr, showing five bands in emission.

The emission lines are quite strong and resolved at this resolution, suggestive of molecular bands. However, this region of the spectrum is heavily contaminated by fringes in the RSRF making it difficult to be more positive.

Very recently, we have been able to obtain a SWS Fabry-Perot scan of the 13.87 µm line in W Hya shown in Fig. 3. The spectrum shows a series of individual lines of the Q-branch of the - ro-vibrational transition of CO₂. This establishes beyond any doubt the identification of gaseous CO₂ as the carrier of these lines.

Fig. 3. Preliminary SWS Fabry-Perot spectrum of W Hya at 13.87 µm showing individual lines of the Q-branch of the - band of CO2. Note that the lines have not been corrected to the heliocentric rest frame.

The position of the 15.40 µm line is very close to the 03¹ 0 - 02² 0 band of ¹² CO₂ but the peak position is better matched by the 01¹ 0 - 00⁰ 0 band of ¹³ CO₂. In view of the strength of the 14.97 µm band (the equivalent band for ¹² CO₂) it is probable that the 15.40 µm line is due to ¹³ CO₂.

Upon further examination of the Hitran database (Rothman et al. 1986), we have been able to identify all observed emission lines with the Q-branches of ro-vibrational bands of CO₂. The energy level diagram which summarizes the observed transitions is shown in Fig. 4. The notation used is that from Herzberg (1966).

Fig. 4. Energy level diagram of CO2 indicating the ro-vibrational bands observed in our program stars (adapted from Fig. 84 of Herzberg (1966)).

Preliminary attempts to fit the observed emission lines of EP Aqr with emission from optically thin thermally populated CO₂ gas indicate that the emission lines are formed in a warm gas layer, probably located at a few stellar radii above the photosphere. The excitation temperature can be estimated by fitting the width of the lines: the higher the temperature, the broader the line due to the contribution from high-excitation lines. For all ¹² CO₂ lines, we obtain the same excitation temperature of 650 K. The existence of such a warm layer of molecular gas in AGB stars has recently been proposed by Tsuji et al. (1997).

For stars with relatively high mass loss rates in our sample, i.e. R Cas and W Hya, the 14.97 µm line appears in absorption. On close inspection of the spectra in Fig. 1 we note that the line at 14.97 µm when seen in emission is exactly coincident in wavelength with that expected from the lowest ro-vibrational transition in the -band of CO₂. However, when in absorption it is shifted to the blue by 0.02 µm, both in W Hya and in R Cas. This cannot be explained by a Doppler shift because this would require a velocity of the order of 400 km s^-1, not seen in any of the other lines. A more probable explanation is the contribution from hot -bands. We have been able to satisfactorily reproduce the shape of the absorption line by including three hotbands and assuming thermal equilibrium population at a temperature of 1250 K, again consistent with the existence of a warm molecular layer in the stellar envelope. More detailed modeling is obviously required.

From the data in Table 2 we conclude that the strengths of the CO₂ emission lines and of the 13 µm dust feature are well-correlated. This suggests that both are produced at the same location in the circumstellar envelope under similar physical conditions. For stars with substantial mass loss rate ( M yr^-1) the 13 µm dust feature is no longer detectable while the 10 µm silicate dust feature starts dominating the spectrum. This could be understood if the enhanced mass loss rate would stimulate the formation of silicate dust and/or prevent the formation of 13 µm dust and at the same time would quench the formation and/or excitation of CO₂ so that we do not observe the CO₂ emission lines in stars without the 13 µm dust feature. More extensive observational studies and detailed models are required to further study the implications of this correlation.

Table 2. Equivalent widths of the 13 µm dust feature and the detected emission lines.

© European Southern Observatory (ESO) 1998

Online publication: January 8, 1998
helpdesk.link@springer.de