2 Observations

Figure 1: The first half of the observed spectrum of R Dor is shown as the upper spectrum. The modeled spectrum, which is shifted down for clarity, is shown below. The modeled spectrum is convolved with a Gaussian whose width matches the appropriate resolution of the bands observed. The abscissa scale shows the wavelengths in vacuum.

Figure 2: The second half of the observed wavelength range of R Dor. The observations are shown above and the model spectrum is shifted down for clarity. The first half is shown in Fig. 1. The observed spectrum is a composite of several sub-spectra which sometimes overlap. However, at $3.51~\mbox{$\mu$ m}$ there is a gap in the observations. The various observed bands have different spectral resolutions.

The spectrum of R Dor was observed with the SWS (de Graauw et al. 1996) on board ISO (Kessler et al. 1996). The spectrometer was used in the grating scan mode (SWS06), which provides a resolution of $R\sim 2000$-2500, depending on the wavelength in the observed region.

The 44 min of observations were performed on the 23rd June 1997 during ISO revolution number 585. The reductions were made using the pipeline, basic reduction package OLP (version 9.5) and the ISO Spectral Analysis Package (ISAP version 2.0). The pipeline processing of the data, such as the flux calibration, is described in the ISO-SWS Handbook. The accuracy of the calibration of the absolute flux is better than 7% ($1\sigma$).

We also observed a low-resolution spectrum from $2.4~\mbox{$\mu$ m}$ to $45~\mbox{$\mu$ m}$ of R Dor with the ISO-SWS (in the SWS01 mode) on 27th June 1997 (orbit number 589). Also here, the reductions were made using OLP version 9.5 and ISAP version 2.0. Our SWS06 and SWS01 spectra of R Dor were thus observed within 4 days of each other. The variation in the spectra due to the periodic variations in the star can therefore safely be ignored, R Dor having a period of 338 days (Kholopov 1988). The SWS01 observation, in the wavelength region of our SWS06 observation, is composed of two sub-spectra; bands 1B and 1D. These two sub-spectra could be merged into one spectrum directly.

Due to the way the spectrometer works in the SWS06 mode, our region was observed in five spectral bands: 2.60- $3.02~\mbox{$\mu$ m}$, 3.02- $3.08~\mbox{$\mu$ m}$, 3.08- $3.19~\mbox{$\mu$ m}$, 3.19- $3.50~\mbox{$\mu$ m}$, and 3.52- $3.66~\mbox{$\mu$ m}$ (for details, see the ISO Handbook de Graauw et al. 2000). There are mismatches in the fluxes between these bands due to uncertainties in the detector gains in every band. Furthermore, the overlap between the bands is small. Therefore, in order to align the five bands, we used the low-resolution SWS01 spectrum, convolved to $\Delta\lambda=0.1~\mbox{$\mu$ m}$ or a resolution of $R \sim 30$, to outline the spectral shape of the region we observed . In this way we are able to scale the different bands of the SWS06 observation with correct factors in order to merge the bands into one spectrum. The flux levels of the bands observed, lie within a factor of 1.04 of each other. As a check, after having merged the SWS06 sub-spectra, we also convolved the merged spectrum to $R \sim 30$, which resulted in the same shape as the SWS01 observation convolved to $R \sim 30$. This indicates that we were successful in merging the five sub-spectra into one spectrum over the entire region and that we can rely on the overall shape of the observed spectrum.

In Figs. 1 and 2 we show the observed ISO spectra. The resolution of the five bands are R = 2300, 2500, 2000, 2500, and 2000, respectively (de Graauw et al. 2000).