When low- and intermediate-mass stars evolve up the Asymptotic Giant Branch (AGB), they undergo substantial mass loss which determines their evolution in this phase. A complex chemistry - based on either carbon or oxygen preponderance - is initiated in the outflow leading to the formation of molecules and dust which are then dispersed in the interstellar medium. AGB stars are believed to be the main factories of interstellar dust and contribute significantly to the chemical enrichment of the gas of the galaxy. Although the analysis of observational data and the results of hydrodynamical model atmosphere calculations in recent years has greatly improved our knowledge on some topics, the mass loss mechanism is still not well understood. The way pulsation and dust formation affects the mass loss remains the subject of debate.
Observations with the ISO/SWS Kessler et al. 1996, de Graauw et al. 1996 have revealed the presence of numerous molecules in the outflows of AGB stars (Tsuji et al. 1997; Justtanont et al. 1998; Yamamura et al. 1999a,b). The excitation temperatures of these molecules and their typical distances to the central star suggest that the observed bands of different molecules are formed in slightly different regions : an "extended atmosphere".
As the "extended atmosphere" is extending from the stellar photosphere to the inner part of the dusty circumstellar material that is generally believed to drive the strong AGB mass loss, studying the physical properties of this region allows us to find the missing link between stellar pulsations and the physical processes of dust formation and wind acceleration in the outflow. The infrared molecular bands are excellent probes to study the physical conditions in the region of the most recent mass loss. Moreover, the very presence of these molecules may shed some light on the chemical pathways to dust formation.
Justtanont et al. (1998) and Ryde et al. (1998) independently reported the discovery with the ISO-SWS of ro-vibrational emission bands of CO2 between 13 and 17 µm, observed in the spectra of O-rich AGB stars that also exhibit the 13 µm dust feature.
By comparing the observations with optically thin LTE models, Justtanont et al. (1998) and Cami et al. (1997) concluded that the excitation temperature of the observed CO2 emission bands is of the order of 650 K and that the emitting CO2 layer is probably located at a few stellar radii. The optically thin LTE model is able to reproduce the observed Q-branch band profiles very well with a single temperature, but not the relative band intensities. Especially the fundamental bending mode at 15 µm is always weaker than expected from optically thin LTE models that fit the other CO2 bands; the 15 µm band is sometimes even seen in absorption.
Ryde et al. (1999) suggest that this is due to non-LTE effects. Estimating that the bands are optically thin, they argue that the density in the CO2 emitting region is probably well over the critical density for thermalizing the rotational levels - which determines the band width - but lower than the critical density for thermalizing the vibrational levels - which determines the relative strengths of the bands. This would explain why an optically thin LTE model can reproduce the individual bands with a single temperature but not the relative band intensities. However, even in an optically thin non-LTE situation, the relative intensities of the P-, Q- and R- branch bands originating from the same vibrational level are only determined by their relative Einstein A coefficients; therefore the relative intensities of these bands should be the same as for an LTE model - which is not the case for the observed bands. This suggests that the Q-branch bands may well be optically thick.
Gonzalez-Alfonso & Cernicharo (1999) present radiative transfer models of the pumping of CO2. Models neglecting the effect of dust produce strong emission in the 15 µm region; when dust is taken into account, absorption results. However, these models cannot reproduce the spectra of those O-rich AGB stars that show emission at 13.87 and 16.18 µm and absorption at 14.98 µm. Furthermore, these models predict absorption at 4.3 µm which is not observed.
The chemical pathway for the formation of CO2 in the extended atmosphere is recently explained by Duari et al. (1999). While thermal equilibrium calculations yield a rather low CO2 abundance, shock chemistry provides the solution to form CO2 in large amounts. At small radii the dominant pathway for the formation of CO2 is
where the OH is formed from the collisional destruction of water. At larger radii () the dominant pathway is
producing a significant increase in CO2 abundance (up to 6 10) although the reaction rate is quite uncertain.
In this paper we present a detailed analysis of the CO2 bands as observed by ISO/SWS in EP Aqr. The CO2 emission bands are most pronounced in the spectrum of this object; the excellent quality of the data - high resolution and high S/N - makes this object an ideal testcase for comparing data with models. We performed LTE model calculations including optical depth effects for a simple geometry with one CO2 layer in order to produce synthetic spectra that can be directly compared to the ISO/SWS observations.
Sect. 2 summarizes the most important background information on EP Aqr. In Sect. 3 we present the data and the necessary data reduction processes used to obtain a high resolution and high S/N spectrum. Sect. 4 describes in detail how our model is constructed, how the different parameters change the appearance of the output spectra and how a quantitative comparison is made between models and observations. Sect. 5 discusses the results of this comparison. These results are discussed in the framework of hydrodynamical model atmospheres, pulsations and dust formation in Sect. 6.
© European Southern Observatory (ESO) 2000
Online publication: August 17, 2000