Astron. Astrophys. 339, 811-821 (1998)

2. Observational results

2.1. Source selection

The sample consists of stars in A General Catalog of S Stars, 2nd Edition (Stephenson 1984, hereafter S2), for which CO J=1-0 or J=2-1 emission was well-detected by Sahai & Liechti (1995) or BL, and with nominal distances 1 kpc. Table 1 lists the seven S stars included in the sample. The first 3 columns give the variable star name; the IRAS Point Source Catalog (PSC) designation; and the S2 catalog number. Column 4 gives the K-band (2.2 µm) magnitude from the Two Micron Sky Survey (TMSS-Neugebauer & Leighton 1969) or from Gezari et al. (1993) if not in the TMSS. The 12 µm and 60 µm flux densities from the IRAS PSC are listed in columns 5 and 6. The spectral type in column 7 is taken from S2. (Note that S2 uses more than one variant of the S classification system, as indicated by the different formats in Table 1.) The variable star type is listed in column 8.

Table 1. Source list and stellar properties

The estimated distance is given in column 9 of Table 1, and follows Jura (1988) and BL in assuming that all S stars have the same mean absolute K magnitude, = -8.1 mag. For Gru, the Hipparcos parallax yields a distance of 153 pc (van Eck et al. 1998), essentially identical to that in Table 1, which was derived from the observed K magnitude and the assumed of -8.1 mag. A typical variation in K-band luminosity of a factor of 2 between maximum and minimum light (cf. Jorissen & Knapp 1998) would cause a distance uncertainty of up to 40%. The assumed mean is another potential error source. Groenewegen & deJong (1998) have used Hipparcos data to derive a mean of -7.1 mag for semiregular and irregular variables among the intrinsic S stars. An error of 1 mag in our assumed would result in a 60% overestimate in the distance in Table 1. A comparison of the distances in Table 1 with those of Groenewegen & deJong (1998) shows that the Miras agree to within 10%, while the 3 SRs without a measured parallax differ by the expected factor 1.6. Gru, however, is an SRb variable, yet has = -8.0 (based on the Hipparcos distance), in good agreement with our adopted value of -8.1.

Finally, column 10 of Table 1 gives the systemic velocity (in the Local Standard of Rest) as determined from the CO 1-0 or 2-1 spectra obtained by BL or by Sahai & Liechti (1995). Column 11 lists our adopted envelope expansion velocity, , which we take as the average value derived from all published CO spectra, given by Jorissen & Knapp (1998).

All the stars in our sample show IR excesses in their IRAS fluxes, and all have relatively strong CO emission lines. Both characteristics imply that these stars possess dusty circumstellar envelopes produced by mass loss, and so are probably intrinsic S stars on the TP-AGB. The presence of technetium has been verified spectroscopically in R Gem (Little et al. 1987) and Gru (Jorissen et al. 1993), thereby confirming that these two objects are intrinsic S stars. The IR colors of the other stars in the sample provide strong evidence that they are also intrinsic S stars, even though spectroscopic data for Tc are lacking. Specifically, Jorissen & Knapp (1998) find that all the stars in our sample lie in regions of the (K-[12], [25]-[60]) IR color-color diagram which are populated by bona fide intrinsic S stars, for which Tc is identified spectroscopically. We conclude, therefore, that our sample consists exclusively of intrinsic S stars.

2.2. Observations and data reduction

The millimetre wavelength spectroscopic observations were made using the 15 metre Swedish-ESO Submillimetre Telescope (Booth et al. 1989) in October and December 1995. The SESIS closed-cycle cryogenic receiver was used in dual frequency mode, enabling simultaneous observations of the HCN(J=1-0) lines near 88.6 GHz (three hyperfine components) and the SiO(v=0, J=3-2) line at 130.3 GHz. The spectra were integrated in two wide band acousto-optic spectrometers (Schieder et al. 1989) having a spectral resolution of 1.4 MHz (4.7 km s^-1 at HCN, 3.2 km s^-1 at SiO). Typical single sideband system temperatures were 140 K at HCN and 180 K at SiO.

Observations were made using a symmetric dual beam switching technique (focal plane chopping at 6 Hz, telescope nodding once every 2 minutes of time, chopping/nodding throw arc minutes in azimuth) which provided excellent compensation for short term atmospheric fluctuations and resulted in consistently flat baselines. The observations were initially calibrated to the temperature scale (Kutner & Ulich 1981; Kutner et al. 1984) using the cooled chopper wheel technique. The absolute pointing accuracy of the telescope was 4 arc seconds rms, and was monitored by making occasional observations of and Cygni.

The data reduction followed standard procedures. Individual spectra were baselined (linear) and co-added with weights based upon the rms noise. The final spectra in were converted to main beam brightness temperatures using main beam efficiencies of 0.75 (HCN) and 0.69 (SiO). The FWHM beam size is for the HCN lines and for the SiO line.

2.3. Results

Table 2 summarizes the observational results for the program S stars, as well as for the carbon star IRC+10^o216, which was observed to check the telescope calibration by comparison with results from other telescopes. The integration time includes both source and reference beam measurements. Columns 3 - 7 give the following quantities for the SiO (v=0, J=3-2) transition : (a) peak main-beam brightness temperature; (b) line center velocity; (c) line full-width at half-maximum intensity; (d) rms noise level (mK-main beam brightness temperature) for off-line spectrometer channels; (e) intensity integrated over the line, in units of main-beam brightness temperature times velocity. Quantities (a), (b), and (c) were determined by a fit of the data to a gaussian line profile, using the least-squares fitting routine GAUSS in the CLASS software package of the Grenoble Astrophysics Group. Observations of M giants -e.g., Bujarrabal et al. 1989-show that SiO lines from these stars are typically gaussian-like, rather than parabolic or flat-topped profiles as are commonly found in CO emission from red giant envelopes. Within the noise, the gaussian fits give good representations of the observed SiO (v=0, J=3-2) line profiles.

Table 2. Summary of observations

Spectra of the SiO (v=0, J=3-2) detections for all the S stars and for IRC+10^o216 are shown in Fig. 1. For DK Vul and R Gem, the lines are weak relative to the noise level but show emission at velocities consistent with the well-detected CO emission lines, so we believe these stars are detected, though the derived line parameters are not well-determined (indicated by a colon in Table 2). The other 5 stars show clear detections at the velocities expected from the CO lines. The gaussian fits with parameters in Table 2 are shown as thin lines in Fig. 1.

Fig. 1. Spectra of SiO (v=0, J=3-2) lines for all detected S stars and for the carbon star IRC+10o216. For the S stars, the thin line shows the gaussian least-squares fit with parameters given in Table 2.

Parameters for the HCN (1-0) line are given in columns 8 -12 of Table  2. In this case, we have fitted parabolic profiles with the "SHELL" option of the CLASS software. Column 8 gives the peak intensity of the fit. Columns 9 and 10 show the center velocity and (FWZI)/2 of the fit. The rms noise of the off-line channels is in column 11, and the integrated intensity of the line (over the FWZI) is in the last column. We note that the HCN (1-0) line is significantly broadened by the presence of hyperfine structure, as is obvious in the spectrum of IRC+10^o216, though the line shape is basically parabolic. For W Aql and IRC+10^o216, hfs produces a (FWZI)/2 which is larger than the expansion velocity, V_e, while for RT Sco the width of the fit is affected by the low signal to noise ratio of the spectrum. For the detected S stars, the HCN spectra are too noisy to justify a deconvolution of the hfs components, so we have used only a single-component parabolic fit.

For non-detections of HCN, the upper limit to the integrated intensity, , is taken to be an assumed linewidth given by the FWZI of the CO line from Sahai & Liechti (1995) or BL, multiplied by a peak-to-peak value of the noise, of 4 RMS. These assumptions should give a conservative upper limit to .

Spectra of the HCN J=1-0 line for the 2 detected S stars as well as IRC+10^o216 are shown in Fig. 2. The thin curves show the parabolic fits to the spectra. The HCN line was detected previously in W Aql by BL, who reported an integrated line intensity, , of 2.75 K km s^-1, obtained with the NRAO 12-metre telescope. This value, when scaled by the ratio of the SEST and NRAO telescope areas, gives a predicted SEST intensity of 4.3 K km s^-1, provided the source is unresolved in both telescope beams. Our measured value (see Table 2) of 4.5 K km s^-1 agrees to better than 5% with the predicted value, which indicates that the calibrations of the 2 telescopes are consistent. (At the distance of W Aql-610 pc-the HCN emission is almost certainly unresolved in the SEST beam.) For comparison, the carbon star IRC+216 has an integrated HCN intensity from Table 2 of 199 K km s^-1, while the scaled intensity from BL is 176 K km s^-1. The discrepancy may be attributed to the large angular size of the HCN emission from IRC+216 (Dayal & Bieging 1995), which is significantly resolved by the SEST beam.

Fig. 2. Spectra of HCN J=1-0 detections. Thin line shows parabolic least-squares fit, with parameters given in Table 2.

© European Southern Observatory (ESO) 1998

Online publication: October 22, 1998
helpdesk.link@springer.de