2 Detection of water on ISO spectra

We used the spectra listed in Table 1 observed with the ISO SWS by its highest resolution grating mode, which gives a resolution of $R = \lambda/{\Delta \lambda} \approx 1600$ ( ${\it FWHM} \approx 188$kms^-1). The sample shown in Table 1 is probably all the red giants earlier than M4III observed with the ISO SWS by this high resolution grating mode, even though more spectra were observed by the lower resolutions. Also, some spectra of late M giants were observed by the high resolution (e.g. Tsuji et al. 1997), but we concentrate in this Letter to the case of red giant stars earlier than about M4 for which the presence of water is not clear yet. The spectra are reduced with the use of OSIA and the resulting spectra are shown in Fig. 1. For comparison, we show in Fig. 2 the spectra of H₂O in the form of log  $B_{\nu}(T_{*}) {\rm e}^{-\tau_{\nu}}$ with $\tau_{\nu} = \kappa_{\nu}(T) N_{\rm col}$, where $\kappa_{\nu}(T)$ is the absorption cross-section of H₂O and $N_{\rm col}$ is the column density of H₂O (assumed to be 10¹⁸cm^-2 throughout). The spectra for T = 1000, 1500, and 2000K are shown and some features to be used as signatures of H₂O absorption are indicated by a - e in Fig. 2.

Figure 1: Spectra observed with the ISO SWS. The first three stars may serve as references in which no signature of water can be seen. The other five objects all show the signatures of water a - e predicted from the spectroscopic data of H2O in Fig. 2.

The spectrum of $\alpha$ CMa (A1Vm; $T_{\rm eff} \approx 10\,000$K) shown at the top of Fig. 1 should show no stellar feature in this spectral region, and the features shown may simply be noise whose variations are within about 0.01dex. ($\pm$1.2%). The next two spectra of K giant stars $\alpha$ Boo (K1IIIb) and $\gamma$ Dra (K5III) may show some features which, however, do not agree with the signatures of water a - e noted in Fig. 2. The features may be due to stellar CO lines (see Fig. 3) and/or to noise. The spectrum No.4 of the K5 giant Aldebaran is quite different and shows, if very weak, most signatures a - e of water againt noise. Also, the overall pattern of the spectrum of $\alpha$ Tau is clearly different from that of the spectra Nos.1-3, which are rather similar to each other. Then, the spectrum No.5 of the M0 giant $\beta$ And shows the H₂O signatures a - e more clearly. The presence of water absorption in the spectra Nos.6, 7, and 8 of $\alpha$ Cet (M1.5IIIa), $\beta$ Peg (M2.5II-III), and $\gamma$ Cru (M3.5III), respectively, is definite and we thus find convincing evidence for water in the early M giant stars. The water features are the strongest in $\beta$ Peg rather than in $\gamma$ Cru, the latest M giant in the present sample. The identification of molecular absorption on stellar spectra is a simple problem of pattern recognition, and the presence of water in $\alpha$ Tau can be convinced if we compare the spectrum No.4 with the spectra Nos.5-8.

Figure 2: Spectra of water evaluated at high resolution ( $R \approx 10^{5}$) and convolved with the slit function of the ISO SWS (Gaussian with ${\it FWHM} \approx 188$kms-1) are shown for three temperatures. The linelists by Rothman (1997) and by Schwenke & Partridge (1997) are used and the results are almost the same.

The water spectra shown in Fig. 2 are well sensitive to temperature, since the features a and c are mainly contributed by the low excitation lines (typically L.E.P.< 2000cm^-1) while the features b, d, and e by the higher excitation lines (L.E.P.> 2000cm^-1). For this reason, the relative intensities of b + d +e against a +c are larger at higher temperatures. We notice that the observed spectra in Fig. 1 do not agree with the trend of the predicted spectrum based on T = 1000K in Fig. 2, and the excitation temperature of the water gas in the observed red giants cannot be as low as 1000K. Instead, the relative intensities of the observed features appear to be more consistent with $T \approx 1500$K or somewhat higher. For evaluating water spectra shown in Fig. 2, we used a calculated water linelist HITEMP (Rothman 1997), but its accuracy is unknown. Then, we also used a more extensive linelist by Schwenke & Partridge (1997), and confirmed that the resulting spectra show little difference with those based on HITEMP at the resolution of Fig. 2. This consistency of the available linelists is encouraging, although the accuracy of the linelists of hot water should be verified by laboratory data in future. Once temperatures can be known, the column densities can be estimated by comparisons of the observed and calculated water spectra. We found $N_{\rm col}$ between $2 \times 10^{17}$ ($\alpha$ Tau) and $2 \times 10^{18}$ ($\beta$ Peg) cm^-2.

Figure 3: Predicted spectra by model photospheres whose basic parameters ( $T_{\rm eff}$/mass in unit of $M_{\odot }$/radius in unit of $R_{\odot }$) are indicated. The synthetic spectra are evaluated with the spectral line database including about a million of lines of CO, OH, CN, SiO, H2O etc., and reduced to the resolution of the observed spectra as in Fig. 2.