6 Conclusion

We have attempted to synthesis an observed spectral region, covering $1.06~\mbox{$\mu$ m}$ of the spectrum around $3~\mbox{$\mu$ m}$ of the red giant R Doradus at a spectral resolution of $R\sim 2000$-2500. The synthesised spectrum is based on a MARCS hydrostatic model-photosphere. We are successful in reproducing the 2.80- $3.66~\mbox{$\mu$ m}$ region, but encounter a discrepancy in the beginning of the spectral region of our observations; $\sim$2.6- $2.8~\mbox{$\mu$ m}$. We have discussed possible explanations of this discrepancy, but have not identified the cause.

The good agreement with the medium-resolution ISO-SWS06 observations, apart from the region with the strongest water bands, suggests the adequacy of using a hydrostatic model photosphere for this particular star, and the completeness and correctness of our input data in the form of the molecular opacity for the calculation of the spectrum. Given all the possible failures of the model to correctly describe the physical picture of the outer photospheres of red giants, the agreement is promising. It shows once again the accuracy and the strength of the new MARCS code also in the infrared wavelength region. Thus, the modeling of moderately varying red giants, such as semi-regular variables, in the near-infrared region with hydrostatic model photospheres may be a reasonable approximation. Note, however, that for the very variable Mira stars in general, hydrostatic model atmospheres certainly fail to reproduce observed spectra (see, for example, Aringer et al. 2001). Attention should also always be given to the extended atmospheres of these stars, which could affect the spectra and especially so in the infrared wavelength region. This extended atmosphere is a possible reason for the discrepancy in the 2.6- $2.8~\mbox{$\mu$ m}$ region.

The spectral signatures in the spectrum presented here are mostly due to photospheric water vapour but several photospheric CO and OH bands are also identified. The spectral region is found to be very temperature sensitive, leading us to an estimate of the effective temperature of the star of $(3000\pm 100)~\mbox{K}$. The best-fit synthetic spectrum was based on a model photosphere of a surface gravity of $\log g = 0 \pm 1$ (cgs). The temperature found here is slightly higher than the ones reported in the literature. A possible reason for this is a problem with the water vapour line list of Partridge & Schwenke (1997), as suggested by Jones et al. (2002).

Acknowledgements

We should like to thank Professors Bengt Gustafsson and David L. Lambert for inspiration and enlightening discussions and the referee for valuable comments and suggestions. This work was supported by the P. E. Lindahl Foundation Fund of the Royal Swedish Academy of Sciences and the Swedish Foundation for International Cooperation in Research and Higher Education.