4 Peculiar evolution of bright massive LPVs

The end of the evolution of massive LPVs on AGB induces complex phenomena and deserves a more detailed study, presented in this section.

4.1 Luminosity boundary of C stars

The luminosity boundary of carbon stars is an important constraint for the models. We find the brightest C stars around K=-9.2, independent of metallicity (Fig. 1), in agreement with the theoretical boundary: $M_{\rm bol}=-6.4$ (Boothroyd et al. 1993).

However, in our individual estimates of absolute magnitude some O-rich LPVs are brighter in K than the brightest C stars (see Table 3). This could be due to Hot Bottom Burning (HBB) that prevents carbon star formation. One of them, R Cen, is even brighter than the $3\sigma$ upper limit of the AGB population from our calibration (K=-9.4 for the disk 1 population, Paper I). The properties of this exceptional star will be a guide to investigate the massive bright LPVs.

   
4.2 R Cen and Hot Bottom Burning

R Cen has a very long period (more than 500 days) and its light curve presents a double maximum. It may have already changed from being a first overtone pulsator to a fundamental one, but this assumption does not agree with pulsation models (Ya'ari & Tuchmann 1996) for a high luminosity star with a period around 500 days. Moreover, the hypothesis that R cen is a first overtone pulsator accounts for the peculiar shape of its light curve by a resonance phenomenon with the ratio P1/P3 close to 2 (Barthes 1998). This author suggests that this peculiarly massive star (more than $3~{\cal M}_{\odot}$ and maybe $5~{\cal M}_{\odot}$) is a candidate star in the Hot Bottom Burning phase. Indeed this phenomenon stops the carbon-enrichment of the surface. This agrees with the fact that this star is nearly the brightest in K, 12 and 25 luminosities.

Two other candidates are proposed by Barthes (1998): T Cas and X Oph. In Paper I, we assign both these stars to the disk population (see Table 3), in agreement with young and initially massive LPVs. Our estimated luminosities confirm T Cas as an HBB candidate. X Oph is more intriguing because its mass is probably at the lowest limit of carbon star formation (Sect. 2.2.4).

Another peculiarity of R Cen is the appearance of the silicate band of its ISO-SWS spectrum (Justtanont et al. 1998). Let us examine the O-rich LPVs assigned to the disk 1 population with an estimated high luminosity. V774 Sgr and SV Cas have IRAS-LRS spectra (IRAS Science Team 1986) with a similar appearance as that of R Cen; they also have a relatively early spectral type (M5 and M6.5 respectively), and so they may also be HBB candidates. IM Cas and V539 Cas have a high luminosity, an early spectral type (M2) and their IRAS-LRS spectra are close to that of R Cen one in the same way as T Cas and X Oph. Another possible candidate could be Y Vel, which has a late spectral type LPV (M8-M9.5).

At least YZ Per has an IRAS-LRS spectrum close to that of R Cen, an early spectral type (M1-M3), but it is classified as a supergiant of class Iab and so it may be at the most advanced end of the AGB or may be a post-AGB. IRAS luminosities and 25-12 indices of V4028 Sgr and V613 Mon indicate that they probably are in the first stages on the AGB. This is confirmed by the IRAS-LRS spectrum of V4028 Sgr in which no SiO feature is present. Unfortunately no IRAS-LRS spectrum of V613 Mon is available.

   
4.3 R Cen and He-shell flash

Another interesting property of R Cen was studied by Hawkins et al. (2001) in a recent paper. They present evidence of a steadily decreasing period of R Cen from 550 to 505 days during the last 50 years. They suggest that it is caused by a He-shell flash, in a similar way to R Hya, R Aql, W Dra and T UMi (Wood & Zarro 1981). W Dra and T UMi have no available HIPPARCOS data and so we could not estimate their luminosities. R Aql is an OH emitter and a remark similar to X Oph (Sect. 4.2) applies. Wood & Zarro (1981) deduce a value between -5.3 and -5.5 mag. for the bolometric luminosity of R Hya, in agreement with our estimation (K=-8.58). They also find that the time scale for the period change of R Hya corresponds to a longer time after the maximum luminosity of the He-shell flash than that of R Aql. This could explain the Tc enrichment of R Hya and its M6-M9S spectral type. The period change for R Cen is steeper than for R Hya and thus Hawkins et al. (2001) give two possible explanations: either R Cen is in a stage right after the beginning of the flash, with a total mass less than 2-3  ${\cal M}_{\odot}$ or it is in a stage right after where the luminosity of the flash reaches the stellar surface with a much larger range of allowed stellar mass. We find that R Cen is the most K luminous LPV but its 25-12 index is this of an S star, at the limit between O and C-rich LPVs, and is assigned to BD, the group of the most massive stars. Therefore, our results strongly favor the second possibility: R Cen is in a stage right after the luminosity of the flash reaches the stellar surface. Furthermore the He-shell flash enhances the efficiency of the third dredge-up (Herwig 2000), that can explain the very peculiar location, compared to the one of that HBB candidates, of R Cen in the diagram (K, 25-12). This so luminous O-rich LPV might become a carbon star exceptionally brighter than the usual luminosity limit accepted for these stars. A few such luminous carbon stars exist in the Magellanic Clouds as observed by Van Loon et al. (1999b) and modeled by Frost et al. (1998).