Appendix A: Comments on individual stars

CL Car (Fig. B3)

In the GCVS, CLCar is classified as an LPV of variability SRc and spectral type M5Ia. Its spectrum indeed fits very well into the supergiant sequence, but it has relatively deep H₂O bands in addition to CN absorption, probably a consequence of pulsation (period = 513 days, amplitude $\delta V=2.5$ mag). These properties already make CLCar interesting as a particularly large amplitude red supergiant or alternatively as a particularly massive Mira.

A remarkable feature of this luminous late-type star is the strong absorption line at 1.083$\mu$m, the wavelength of the transition between the triplet levels 2³S and 2³P of HeI. The feature has an equivalent width of a few Å and is not an artifact since it is seen in all individual raw and reduced spectra of the object and in no other spectrum of the same programme. Searches for HeI absorption in this line in various types of cool stars have been reported by Zirin (1982) and O'Brien & Lambert (1986). Zirin concluded that absorption was rare in giants and supergiants later than type K3 except for symbiotic stars like Z And and R Aqr; CLCar might thus be a symbiotic star. With a better signal to noise ratio, O'Brien & Lambert found weak absorption in all K supergiants but no absorption in stars later than M1I. Note that only a few luminous M stars were observed in each of the samples.

No significant absorption is seen in the 2.058$\mu$m 2¹S - 2¹P transition of HeI in CLCar (although this line is located in a telluric CO₂ band which degrades the signal-to-noise ratio). The lower level of the 1.083$\mu$m transition of helium lies 626Å above ground level, but is metastable as opposed to the lower level of the other near-IR transition at 2.058$\mu$m (2¹S). Interpretations for the population of the 2³S level of HeI in non-coronal stars range from X-ray photoionization through collisional excitation by thermal electrons to stochastic shock waves (Cuntz & Luttermoser 1990). In the latter mechanism, outwards running shock fronts initiated with various velocities merge and build up, possibly leading to temperatures above 20000K. Similar merging processes of successive shock waves are predicted to occur in Miras, but no 1.083$\mu$m absorption has been observed in any of those stars.

U Equ (Fig. B24)

The observation of UEqu (= IRAS20547+0247) was suggested by A.Omont. This heavily obscured object has recently revealed strong peculiarities (Barnbaum et al. 1996): the spectral type of the obscured star could range from mid-G to early-K, the optical spectrum indicates the luminosity class of a giant but shows anomalously deep and narrow molecular absorption lines of TiO, AlO and VO as well as bright molecular emission lines of the same compounds. The $12 - 25\,\mu$m energy distribution indicates optically thin dust, while silicate absorption and a strong 60 $\mu$m excess are consistent with a thick dust envelope. OH and H₂O maser emission are present and highly variable (see also Lewis 1997). Circumstellar CO(2-1) emission has also been detected (Josselin et al. 1998). A non-spherical geometry seems necessary to account for all properties; Barnbaum et al. (1996) suggested the presence of a cold disk viewed nearly edge-on.

The 1.6 - 2.45 $\mu$m spectrum obtained in May 1996 (Fig. B24) has a red slope consistent with the photometry collected in Barnbaum et al. (1996). H₂O absorption is weak and narrow as compared to normal pulsating OH/IR stars. It is remarkable that no 2nd overtone CO absorption is detected longward of 2.29 $\mu$m: the bands seem to be filled up with CO emission. A similar filling of molecular bands usually seen in absorption has been observed in the mid-IR with the Infrared Space Observatory (A.Omont, private communication).

R Lep (Fig. B31)

RLep is a transition object between optically visible and dust-enshrouded C-rich Mira variables. In addition to its red colours, two particularities of this object are worth notice: the presence of an absorption feature at 1.53 $\mu$, not seen in other (warmer) carbon stars, and the significant variations of the spectrum with phase.

The four spectra obtained in 1995 and 1996 (GCVS phases 0.7 - 0.9, but $\sim 0.4-0.7$ according to AAVSO data, Mattei 1999) drew our attention to the relatively broad absorption feature centered at 1.53 $\mu$m. In order to check the phase dependance of the band, an additional spectrum was obtained in November 1997. The presence of the feature was confirmed (Fig. A1), and its strength found to have changed very little.

The 1.53 $\mu$m feature has been previously noticed in a few low temperature carbon-rich Miras including V CrB, S Cep, U Cyg and SS Vir (Goebel et al. 1981) and V Cyg (Joyce 1998). Goebel et al. (1981) suggest HCN or C₂H₂ as the main carriers. Joyce (1998) also favours this interpretation, noting that the presence of the 1.53 $\mu$m band in a spectrum is associated with absorption around 2.45 $\mu$m, as expected from the laboratory spectrum of C₂H₂ (Goebel et al. 1981). However, HCN or C₂H₂ are also responsible for the 3.1 $\mu$m and no clear correlation with the 1.53 $\mu$m feature is found (Joyce 1998). Chemical equilibrium networks for C-rich atmospheres (Morris & Wyller 1967; Scalo 1973; Helling et al. 1996; Doty & Leung 1998) predict that under adequate circumstances a large variety of carbon bearing molecules may be abundant. In addition to the most obvious C₂, CN, HCN, CO, C₃ and C₂H₂, the species produced at low temperatures include C₂H, CH₃, C⁺ and more complex compounds. Goebel et al. (1983) suggest C₂H as the carrier of the 2.9 $\mu$m band seen in the R type carbon star HD 19557. They present a theoretical spectrum which, taking into account the uncertainties on the predicted wavelengths as estimated by the authors, could explain the 1.53 $\mu$m feature. The existence of C₂H in cool carbon stars is further supported by the high resolution spectrum of IRC +10216 around 2 $\mu$m, published by Keady & Hinkle (1998). Indirectly, the failure to reproduce the spectrum of V CrB with a network including HCN and C₂H₂ but neglecting C₂H enhances the probable role of the latter (e.g. Kipper 1998). Unfortunately, C₂H is not included in most current spectral synthesis codes for carbon stars, because of lacking accurate chemical data (Hron et al. 1998; Plez et al., private communications).

Among our C-rich sample, RLep is the only star displaying strong variations with phase. Circumstellar extinction may be significant for this object and it may possibly vary with phase; however, photospheric variations are required to explain the changes in the strengths of the CN bandheads at 1.1 and 1.4 $\mu$m. The C₂ Ballik-Ramsay bandhead (1.77 $\mu$m) and the CO band strengths vary little in comparison. They are similar to those of the S/C Mira BHCru (Fig. A1), indicating a relatively low C/O ratio for a carbon star. The efforts put into modelling the variations of cool C-stars in the ISO wavelength range need to be extended to the range of this data, in order to provide basic interpretation tools.

SV Lib (Fig. B5)  

This O-rich star displays a remarkably clean CO spectrum between 1.55 and 1.7 $\mu$m. OH lines and the Si line at 1.59 $\mu$m are particularly weak, similar to the S/C Mira BH Cru in this respect (Fig. A2). However, the star displays H₂O absorption bands, normal VO absorption bands for its energy distribution and normal/strong TiO bands (Alvarez et al. 2000).

All three IK spectra available in the sample display P$\beta$ in emission: they are obtained close to maximum light, though in different cycles. The optical spectrum obtained in March 1996 is cooler by some 200 K (Alvarez et al. 2000). Near-IR observations at shorter time intervals are necessary to find out whether the weak-lined H band spectrum is a usual property of the star (possibly a relatively large C/O ratio) or whether it is peculiar to phases around maximum light.

Unfortunately the present observations do not allow us to constrain the rather uncertain pulsation properties of SV Lib. The GCVS indicates P=402.7 and $\delta V \geq 1.3$; Shawl & Bord (1990) correct the period to 192 days (but mention aliases at 126 and 410 days in their data), and find an optical pulsation amplitude of about 5 mag. As shown in the following table, which lists the phases of the available spectra according to both references, the lack of data between June 95 and March 96 prevents a final conclusion.

Date	Phase
	GCVS	B&S90
June 95	0.80	0.80
March 96	0.46	0.18
May 96	0.67	0.63
July 96	0.77	0.84

The understanding of pulsating atmospheres is not good enough to allow us to distinguish between phase 0.46 and 0.18 from an optical spectrum alone. The low flux level of the March 96 spectrum as well as the drop in effective temperature for this spectrum appear rather large and tend to favour the longer, GCVS period.

Figure A1: RLep (November 1997; bottom) and BHCru (May 1996)

Figure A2: The spectrum of SV Lib (May 1996, but all available data are similar), compared to the Mira variable WW Sco (May 1996), the cool supergiant UY Sct (May 1996) and the S/C Mira BH Cru (Jan. 1996). The position of the 1.59 $\mu$m SiI line (blended with OH lines in cool O-rich objects) is indicated