The hidden transition from AGB to post-AGB as observed by AKARI

Felix Bunzel1, Dieter Engels1

Abstract: When stars leave the (AGB), they enter the post-AGB stage and are just about to become visible as a (PN). At this stage, their circumstellar shell has lost spherical symmetry and the spectrum of their dusty envelope has changed drastically. We investigate six sources selected as candidates for stars experiencing the hidden transition from AGB to post-AGB stars. All six sources exhibit -rich chemistry and were selected by their position in the IRAS two-colour diagram. Three of which still evolve on the AGB, while the other three have already entered the post-AGB stage. Observational data contain AKARI/IRC infrared spectra in the range of 2.1-25.5 µm with data gaps at 4.9-5.4 µm and 12.1-18.7 µm. The 1D radiation transfer code DUSTY was used to apply model fits. All spectra show a strong 10 µm absorption feature of amorphous silicates indicating the oxygen-rich nature of the sources. Features of crystalline silicates, however, are absent. Beside amorphous silicates DUSTY model calculations require a certain amount of amorphous dust to be present in the circumstellar shells of all investigated AGB as well as post-AGB sources. While the fraction of amorphous carbon does not exceed 15% for DUSTY AGB models, an inner shell composed of pure amorphous carbon dust surrounded by an outer shell of mixed chemistry had to be assumed for DUSTY post-AGB models. The transition of oxygen-rich and carbon-rich AGB stars to the post-AGB stage may no longer be considered to proceed separately, since oxygen-rich AGB stars can still be converted to carbon stars during the very final phase of AGB evolution.

Keywords: AGB and post-AGB stars, , circumstellar matter, infrared spectra

1. Introduction material from the core region to the stellar envelope via . Thus, a fraction of the lower-mass AGB popu- When intermediate mass stars experience their final nu- lation (M < 4-5 M ) is being converted into carbon stars, clear burning stage, they have reached the asymptotic since a radiative buffer zone located at the bottom of the giant branch (AGB) in the Hertzsprung-Russell diagram stellar envelope assures a separate evolution of the enve- (HRD). Here they lose a major part of its mass and exhibit lope and the core region of the . For more massive much higher than their main-sequence pro- AGB stars, the activation of hot bottom burning (HBB) pre- genitors. It is the mass-loss process, which leads to the vents the carbon (Habing & Olofsson, 2003). formation of stellar winds causing the total obscuration To investigate what is happening during the hidden tran- of the central star (at least for M > 3 M ) in the optical sition stage from AGB to post-AGB stars, spectroscopic wavelength range. For an astronomically short while, the observations of three AGB and three post-AGB stars were star is observable only in the infrared, since a circumstel- carried out by AKARI/IRC (Onaka et al., 2007). The lar envelope (CSE) consisting of the recently ejected matter IRAS two-colour diagram as well as IRAS variability in- absorbs all the optical radiation emerging from the star dices were used to distinguish AGB from post-AGB stars. and reemits it at longer wavelengths. The composition More information about our sample including the selec- of this recently ejected matter made up of gas and dust tion process is given in section 2. AKARI observations sensitively depends on the C/O ratio of the CSE. Origi- are described in section 3, while obtained spectroscopic nally, the abundance of oxygen in an AGB CSE exceeds the data and model calculations applied by the 1D radiation abundance of carbon, so that dust, which is able to form transfer code DUSTY (Ivezic et al., 1999) are presented in in this relatively cool and dense environment, primarily section 4. Finally, our results including implications on consists of amorphous silicates. This is due to the fact shell chemistry and the final phase of AGB evolution, are that, under the assumption of a local thermodynamical discussed in section 5. equilibrium (LTE), the less abundant species is completely locked in the CO molecule, since this atomic compound 2. Selection tools and sample sources has the highest binding energy in an LTE environment (e.g. Salpeter, 1974). When the star has lost its entire stellar envelope, the mass- As a star evolves along the AGB, it is supposed to move loss process stops and the star reappears in the optical. from the lower left to the upper right part of the IRAS It has now reached the post-AGB stage and is about to two-colour diagram (see Fig.1) as dynamical pulsations create one of the most fascinating phenomena of the sky, start and the star becomes variable. The meaning of the a planetary nebula. Post-AGB CSEs exhibit a variety of labelled regions is described by van der Veen & Habing axis- or point-symmetrical structures, while the mass out- (1988). Sources located in region IV are supposed to expe- flow on the AGB proceeds in a spherical symmetric way. rience the hidden transition from AGB to post-AGB stars. Additionally, the CSE chemistry may change from oxygen- Our sample (see Tab.1) consists of three oxygen-rich AGB rich to carbon-rich during the AGB evolution of the star, if stars, of which two (IRAS 11549-6225 and IRAS 18432- the third dredge-up transports enough freshly processed 0149) are located in region IV of the two-colour diagram, which is most probably the region, where AGB evolution 1Hamburger Sternwarte (http://www.hs.uni-hamburg.de) ends. Both of them show a high variability index indicat- Gojenbergsweg 112, 21029 Hamburg, Germany ing that dynamical pulsations have not stopped, yet. The E-Mail: [email protected] variability index of IRAS 14104-5819 is low, however, this

ICYA2009 proceedings 1 2 Felix Bunzel, Dieter Engels

Tab. 1 – The sample of oxygen-rich AGB and post-AGB stars observed to investigate the hidden transistion with AKARI

IRAS name Coordinates (J2000) Type Var RA DEC 11549−6225 11 57 30.8 -62 42 12 AGB 73 14104−5819 14 14 00.6 -58 33 57 AGB 15 18432−0149 18 45 52.4 -01 46 43 AGB 93 18450−0148 18 47 41.2 -01 45 11 post-AGB 19 20266+3856 20 28 30.7 +39 07 00 post-AGB – 19134+2131 19 15 35.1 +21 36 31 post-AGB 06

the spectra from the short exposures. If the calibration was Fig. 1 – IRAS two-colour diagram showing our sample considered unreliable, we removed data at the edges of the of oxygen-rich AGB (blue squares) and post-AGB dispersion element wavelength ranges. Finally we shifted (magenta squares) stars as well as oxygen-rich the flux calibration of the different wavelength ranges rela- AGB/post-AGB reference sources compiled by En- gels &Lewis (1996) (green triangles). Colours used tive to the SG1 spectrum. The resulting spectra are plotted in this plot were calculated by in Fig. 2 and 3. Note that the wavelength range 13-18 µm IRAS(25 − 12) = log(νFν(25)/νFν(12)) and was not covered by AKARI due to a malfunction of the IRAS(60 − 25) = log(νFν(60)/νFν(25)). LG1 dispersion element. source is located in region IIIb, where regular AGB stars are expected to be found. All three sample sources show 4. Infrared spectra and DUSTY models double-peaked OH maser emission and no optical coun- terpart. For all six sources of our sample we found a prominent The three post-AGB stars of our sample have moved a 10 µm absorption feature, indicating the oxygen-rich na- bit further along the proposed evolutionary track in the ture of the sources. While the three AGB stars show a two-colour diagram. For none of them an optical coun- certain amount of near-infrared flux, all three post-AGB terpart was detected, so that their low variability index stars are totally obscured in the near-infrared region. Their (if observed) indicates that dynamical pulsations have re- spectra could not be found on NP spectroscopic images. cently stopped. IRAS 18450-0148 and 19134+2131 belong The 10 µm absorption feature of the latter ones seems to to a small group of objects referred to as the “water foun- be located on the ascending blue part of the Planck curve, tains” (Likkel et al., 1992). High-velocity H2O masers while for the three regular AGB stars, the 10 µm absorp- of roughly 100 km/s, originating from an environment of tion feature is located close to the maximum. This is al- high temperatures and hydrogen densities, were detected ready an indication of the relatively cool dust around the from radio observations of these sources (Imai et al. (2002); post-AGB sources compared to the higher dust tempera- Imai et al. (2004)), implying that they must be located in tures at the inner CSE boundary of the regular AGB stars. the collimated bipolar outflows of a post-AGB star. For For applying DUSTY model calculations to fit the spectra IRAS 20266+3856 maser emission was detected only from of the regular AGB stars, the influence of all model pa- the OH molecule showing a regular double-peaked maser rameters was tested. Initial parameter values were taken spectrum. from the literature. These were varied until a ’standard model’ was found, in which all parameters with a rela- tively small influence on the model spectrum were fixed 3. AKARI observations to values, which provide the best fit for all three sources. As a result of our obtained standard model, the temper- The observations with the AKARI satellite (Murakami ature of the central star was set to T∗ = 2500 K, the dust et al., 2007) were made between May 10, 2006 and May 17, temperature on the inner boundary of the dust shell was 2007. We used the Infrared Camera (IRC) (Onaka et al., set to Tdust = 1000 K, a single dust grain size distribution 2007) with the spectroscopic observation mode AOT04. of a = 0.27 µm was used and the density profile was calcu- Long and short exposed spectroscopic observations were lated via an analytical radiatively driven winds solution, obtained with the dispersion elements NP (1.8-5.2 µm), which is implemented in DUSTY and applies a full hydro- SG1 (5.4-8.4 µm), SG2 (7.5-12.9 µm), and in two cases LG2 dynamics calculation in a spherical symmetry. (17.5-25.7 µm). The dispersion varied from 0.06 µm/pix at The only variable parameters in our AGB models are the 3.5 µm to 0.175 µm/pix at 25 µm. The data reduction was optical depth at 10 µm and the composition of the dust. made with the IRC Toolkit Version 20080528 The optical depth was varied until a good fit for the 10 µm (Ohyama et al., 2007). For all sources except for IRAS absorption feature was obtained. However, with a pure- 19134+2131 and IRAS 20266+3856, where spectroscopic silicate dust composition no good fit could be obtained. data could be obtained only from the MIR-S channel, we Thus, an admixture of the most common dust for C-rich encountered saturation effects in the long exposures of one AGB CSEs, amorphous carbon dust, was added and in- or more dispersion elements. In such cases we extracted creased until a good fit was obtained. AKARI/IRC spectra The hidden transition from AGB to post-AGB stars 3

1e-09 1e-09

20266+3856 (* 100) 14104-5819 (* 100) 1e-10 1e-10

19134+2131 (* 10) 18432-0149 (* 10)

1e-11 1e-11 ] ] -2 -2

18450-0148 [W m [W m λ λ

F 11549-6225 F λ 1e-12 λ 1e-12

1e-13 1e-13

1e-14 1e-14 2 3 4 5 6 7 8 9 10 20 30 2 3 4 5 6 7 8 9 10 20 30 wavelength [µm] wavelength [µm]

Fig. 2 – AKARI/IRC spectra (in black) of all three regular Fig. 3 – Spectra of all three post-AGB stars of our sample AGB stars of our sample are shown. DUSTY models (denotations as in Fig.2) (in red) are overplotted.If available, the Spitzer continuum spectra (in blue) are used to bridge the data gap in the 13-18 µm region. Tab. 3 – Derived DUSTY model parameters, containing the optical depth at 10 µm for the inner (τin) and outer out shell (τ ) as well as the dust temperature (Tout) Tab. 2 – Derived DUSTY model parameters, containing the plus abundances of amorphous silicates (ηSil) and optical depth τ at 10 µm as well as abundances of amorphous carbon dust grains (ηC) of the outer amorphous silicates (ηSil) and amorphous carbon shell dust grains (ηC) in out IRAS name τ10µm τ10µm Tout ηSil ηC IRAS name τ10µm ηSil [%] ηC [%] [K] [%] [%] 14104−5819 9 90 10 20266+3856 0.3 4 160 26 74 11549−6225 16 86 14 19134+2131 1 5 200 37 63 18432−0149 18 97 03 18450−0148 5 5 140 40 60

as well as obtained final DUSTY models for all three AGB present in all of the observed sources. This is an aston- stars of our sample are displayed in Fig. 2. Derived values ishing result, since O-rich and C-rich AGB stars were up of the varied model parameters are given in Tab. 2. to now believed to evolve separately from that moment Using our standard model, no good fit for the spectra on the AGB, when a thermal pulse makes the C/O ratio of the three post-AGB stars of our sample was found. of the stellar envelope exceed unity. Our model calcula- Even under the assumption of very low dust tempera- tions, requiring the inner CSE layers of O-rich post-AGB tures, the extreme redness of the spectra could not be stars to consist of C-rich dust, indicate that, even if the explained. To reproduce this redness, another type of O-rich chemistry survives all thermal pulses at the end of model had to be invented. Thus, we applied model cal- AGB evolution, a significant amount of carbonaceous dust culations for an inner C-rich shell with standard-model grains forms in the inner CSE. The origin of this C-rich parameters, consisting purely of amorphous carbon dust, dust is so far unknown. In the following, we describe two and used the resulting model spectrum as an input to an scenarios we consider to explain this C-rich inner shell. outer shell of mixed chemistry modelled in slab geome- The first scenario (see Fig.4) is a cessation of the hot bottom try. In this configuration the model mimics a carbon star burning (HBB) at the end of AGB evolution. HBB prevents seen through a screen consisting of a mixed dust com- massive AGB stars, which are all O-rich, from being con- position. AKARI/IRC spectra as well as obtained final verted to carbon stars. Their convective envelope pene- DUSTY models for all three post-AGB stars of our sample trates the hydrogen-burning shell, so that C-rich material, are displayed in Fig. 3. Derived values of the varied model which was dredged up into the envelope during a thermal parameters are given in Tab. 3. The temperature for the pulse, is being converted into nitrogen via the CNO cycle. outer shell is much lower than for the inner one, since its Thus, the abundance of carbon in the stellar envelope is distance to the central radiation source is much larger and prevented from outnumbering the abundance of oxygen. its density has decreased due to expansion of the shell. It However, as the star evolves along the AGB, it loses huge is conspicuous that the outer shell still seems to contain parts of its stellar envelope, what leads to a decrease in more C-rich than O-rich material. temperature at the bottom of the envelope, where HBB takes place. This might cause the HBB to stop at some 5. Implications on the final stage of AGB evolution point, when the is no longer able to pen- etrate the hydrogen-burning shell. Subsequent final ther- Model calculations presented in the previous section sug- mal pulses would now be able to change the chemistry gest a certain amount of carbonaceous dust grains to be of the remaining stellar envelope and C-rich dust could 4 Felix Bunzel, Dieter Engels

Fig. 4 – A cessation of the hot bottom burning (HBB) at the end of the AGB evolution is one possible scenario for the hidden transition to the post-AGB stage (see text)

Fig. 5 – Another scenario for the end of AGB evolution:The CSE of a massive AGB star initially remains O-rich after the whole stellar envelope was released (see text) be produced after the next mass ejection. If this happens Calderon´ for providing us with observational data ob- just before the convective envelope is totally lost and AGB tained by Spitzer, a NASA’s Great Observatories Program. evolution ends, the inner part of the remnant CSE would FB wants to thank the ICYA2009 team for organizing a be C-rich, while the outer parts are still O-rich. great conference. The second scenario (see Fig.5) is that the CSE of a mas- sive AGB star initially remains O-rich after the whole stel- References lar envelope was released. A subsequent ejection of the remnant hydrogen- and helium-burning shells as well as Engels, D. & Lewis, B. M., March 1996, A&AS, 116, 117–155. the intershell zone, might form the inner part of the CSE, Habing, H. J. & Olofsson, H., editors, 2003, Asymptotic giant branch which would then be made up of C-rich material. For stars. a short period of time the star will resemble our model Imai, H., Obara, K., Diamond, P. J., Omodaka, T., & Sasao, T., June made up of an inner C-rich shell and an outer O-rich (or 2002, Nature, 417, 829–831. mixed) shell. Imai, H., Morris, M., Sahai, R., Hachisuka, K., & Azzollini F., J. R., June 2004, A&A, 420, 265–271. These two scenarios would cause the formation of a geo- Ivezic, Z., Nenkova, M., & Elitzur, M., October 1999, ArXiv As- metrically thin C-rich layer at the inner edge of the out- trophysics e-prints, arXiv:astro-ph/9910475. flowing remnant CSE. Although the inner C-rich shell Likkel, L., Morris, M., & Maddalena, R. J., March 1992, A&A, 256, might be thin compared to the outer O-rich shell, it dra- 581–594. matically affects the spectral energy distribution of the Murakami, H., Baba, H., Barthel, P., Clements, D. L., Cohen, M., star, since its density is much higher than that of the dis- Doi, Y., & others, , October 2007, PASJ, 59, 369–+. sipated outer shell. Ohyama, Y., Onaka, T., Matsuhara, H., Wada, T., Kim, W., Fu- jishiro, N., & others, , October 2007, PASJ, 59, 411–+. Onaka, T., Matsuhara, H., Wada, T., Fujishiro, N., Fujiwara, H., ACKNOWLEDGMENTS Ishigaki, M., & others, , October 2007, PASJ, 59, 401–+. Salpeter, E. E., November 1974, ApJ, 193, 579–584. This research is based on observations with AKARI, a van der Veen, W. E. C. J. & Habing, H. J., April 1988, A&A, 194, JAXA project with participation of ESA. We want to thank 125–134. D.A. Garc´ıa-Hernandez,´ P. Garc´ıa-Lario and J.V. Perea-