5 Conclusions

We have presented spectra covering the wavelength range 2.28 to 2.36 $\mu$m for a sample of 24 cool evolved stars. The sample comprises 8 M supergiants, 5 M giants, 3 S stars, 6 carbon stars, and 2 RV Tauri variables. The wavelengths covered include the main parts of the ¹²C¹⁶O v = 2-0 and 3-1 overtone bands, as well as the v = 4-2 and ¹³CO v = 2-0 bandhead regions. The observed transitions arise from a wide range of energy levels extending from the ground state to E/k > 20000 K. The spectra have a resolution of $\Delta \lambda = 0.0007~\mu$m, or R = 3500. CO lines dominate the spectrum for all the stars observed, and at this resolution most of the observed features can be identified with individual CO R- or P-branch lines or blends. We looked for correlations between the intensities of various CO absorption line features and with other stellar properties, including IR colors and mass loss rates. Two of the most useful CO line features appear to be the 2-0 R14 line which is well-resolved from any other CO or atomic features; and the CO 2-0 bandhead, which, as the shortest wavelength of the first overtone bandheads, is not blended with any other CO lines nor any known atomic lines. The most significant conclusions are:

(1) The 2-0 R14 intensity shows a non-linear correlation with the 2-0 bandhead intensity, with an apparent leveling off in R14 for the deepest 2-0 bandheads (see Fig. 27).

(2) The ratio of the R14 line depth (below the continuum) to the 2-0 bandhead depth converges to a value of $\sim$0.5 for the deepest 2-0 bandheads ( I(2-0BH) <40%). Stars with shallower 2-0 bandheads ( I(2-0BH)>50%) show a large scatter in the ratio of line depths (Fig. 28).

(3) The published gas mass loss rates for both the AGB stars and the red supergiants in our sample correlate with the K-[12] color, consistent with other studies. The intensity of the 2-0 bandhead shows a trend with K-[12] color such that the reddest stars (K-[12] > 3 mag) exhibit a wide range in 2-0 bandhead depth, while the least reddened have the deepest 2-0 bandheads, with a small range of variation from star to star.

(4) Similarly, the reddest stars (K-[12] > 3 mag) exhibit a large range in the ratio of the depth of the 2-0 R14 line to the 2-0 bandhead depth. Stars with less reddening (K-[12] < 3 mag) show about one-third the range in this ratio compared to the redder sample. If we interpret the K-[12] color in terms of a mass loss rate, the data imply that stars with $\dot M_{\rm gas} < 5\times 10^{-7}~M_{\odot}$ y^-1 exhibit a much narrower range in the relative strengths of CO 2-0 band features than stars with higher mass loss rates.

(5) The wide range in CO feature intensities and intensity ratios for the most reddened stars suggests that dust continuum is not filling in the absorption, at least not in every case, but this effect needs careful modelling in any synthetic spectrum calculation.

(6) There is no evidence for a correlation of the CO band strengths with phase of the light cycle for the Miras, but intrinsic differences from star to star, or cycle-to-cycle variations, could mask any real effect. Our spectrum of AC Her, a post-AGB star, differs significantly from comparable spectra taken at other epochs. Well-sampled monitoring of selected stars is needed to determine the extent of spectral variability.

(7) Although the ¹²C¹⁶O overtone bands dominate the observed spectra, we note that other atomic and molecular species are present and need to be considered. First, the ¹³C¹⁶O 2-0 bandhead is closely blended with an atomic line of Ti I, and with the 3-1 R17 and 2-0 R0 lines of ¹²C¹⁶O. The observed ¹³C¹⁶O 2-0 bandhead depth will be affected by these features. Second, the carbon star spectra show significant modulation by other molecular lines in this wavelength region. The likely contributors include C₂, CN, CH, CS, HCN, C₃, and C₂H₂. We have not attempted to identify features of these molecules, and in fact almost all of the discrete features we observe in the carbon star spectra can be identified with the CO overtone bands. Precise modelling of this spectral region for carbon stars, however, will require that these other molecular species be included.

The range in spectral properties that we observe for this sample of cool giant and supergiant stars implies that there are significant differences in atmospheric structure. Spectra of the CO molecule should be good diagnostics for the structure of the extended atmospheres of these pulsating, mass-losing stars. Such data may be used to test hydrodynamic models which are being developed with more realistic treatment of stellar pulsations, shocks, dust formation, and molecular chemistry.

Acknowledgements

JHB thanks C. Engelbracht for invaluable help with data processing; and acknowledges support from the U.S. National Science Foundation through grants AST-9618523 and AST-9987408. We also thank the referee, Dr. M. Scholz, and Dr. J. M. Winters for helpful comments. This research has made use of the SIMBAD database, operated at CDS, Strasbourg, France.