Observations of the CO(2-1) line at 230 GHz and of the CO(3-2)
line at 345 GHz were made in January 1996 and January 1997 with the 10.4 m
telescope of the Caltech Submillimeter Observatory (CSO) on Mauna Kea,
Hawaii. The CSO is equipped with SIS
junction receivers cooled to liquid helium temperatures. The effective
single-sideband system temperatures for these observations, including the
effects of atmospheric emission and absorption, were about 500 K and 800 K
at 230 and 345 GHz, the telescope half-power beamwidths were respectively
and
and the main-beam efficiencies 76% and 65%.
The spectra were observed using two 1024 channel acousto-optic spectrographs
(AOS) simultaneously. The first has a total bandwidth of 500 MHz
(
at 230 GHz) and a velocity resolution of
. The second has a bandwidth of
50 MHz
and a velocity resolution of
. The observations were
made by chopping between the star position and an adjacent sky position,
offset
in azimuth, at a rate of 1 Hz, and consisted
of pairs of chopped observations with the source placed alternately in
each beam. The spectral baselines resulting from this procedure are linear
to within the rms noise. The telescope pointing errors were measured by
mapping the spectral line emission from a nearby CO bright star before
each observation was made, and the pointing accuracy is better than
for all of the observations.
The temperature scale and atmospheric opacity were measured by chopping against
a hot (room temperature) load. The line temperature was corrected for the
main-beam efficiency, and the resulting scale is the Rayleigh-Jeans
equivalent main beam brightness temperature , i.e. that
measured by a perfect 10.4 m antenna above the atmosphere. The spectrometer
frequency was calibrated using an internally generated frequency comb,
and the velocity scale is corrected to the Local Standard of Rest (LSR).
Figure 14:
CO(2-1) line profiles of four S stars observed at CSO.
The abscissa is velocity with respect to the LSR and the ordinate main
beam brightness temperature. The profiles for GCGSS 212 and 589 are observed
with a velocity resolution of , while GCGSS 283
and GCGSS 704 are observed with 1
resolution. The horizontal
bars show the velocity range observed in optical spectra (Table 4). There
is no optical velocity available for GCGSS 704
Twelve S stars were observed at the CSO (Table 4 (click here)), nine in the CO(2-1) line and three in the CO(3-2) line, and emission was detected from four (Table 5 (click here)). The line profiles for the detected stars are shown in Fig. 14 (click here).
Table 4 (click here) lists the stars that were observed:
the GCGSS number
and the variable star name are in Cols. 1 and 2,
respectively.
Next is the observed position: we used positions accurate to
from the HST Guide Star Catalog and other sources (see
Chen et al. 1995). Columns 5-9 list the galactic longitude and latitude,
the spectral type from the GCGSS, the variable type and the period
from the GCVS. Column 10 gives the stellar radial velocity with
respect to the LSR measured from optical spectra; since the radial velocities
of red giants vary as they pulsate, Table 4 (click here) lists the range of
reported radial velocities. Finally, Col. 11 gives the channel-to-channel
rms noise in the 500 MHz AOS.
Table 5 (click here) gives the line parameters of the four detected
stars: the CO(2-1)
line flux in , the peak line temperature, the central
velocity
and the half-width of the line at zero power,
,
which gives the terminal wind outflow speed. These quantities are determined
by fitting a parabolic line model to the data. The agreement between the
optical and CO radial velocities is good.
Seven of the stars in Table 4 (click here), GCGSS 89, 117 (GP Ori), 422 (NQ Pup) 626 (FM Hya), 704 (Z Ant), 796 (HR 4755) and 803 (S UMa) have not previously been observed. We detect one of these, Z Ant. GCGSS 816 (UY Cen) is weakly detected by Sahai & Liechti (1995, SL95) with good agreement between the CO and optical radial velocities. This star was not detected in the present observations, but our sensitivity is lower. The detection of circumstellar CO for this star is of particular interest, given its rare SC spectral type (see Sect. 6.3 (click here)). GCGSS 149 (NO Aur) and 796 (HR 4755) were previously observed by SL95 and by Bieging & Latter (1994, BL94); like the present observations, these did not detect CO emission.
RS Cnc has been observed by many authors (see Loup et al. 1993, for example). Margulis et al. (1990) point out that the CO line profile for this star, as for several others, more closely resembles a triangle or gaussian in shape than the parabolic profile typical of circumstellar winds. The high velocity resolution observations in Fig. 14 (click here) show that the line profile actually consists of two parabolic components of different widths centered at the same velocity. The parameters for these profiles, estimated by eye, are given in Table 5 (click here). This line shape may indicate the presence of two molecular winds. Such line profiles have been seen for several other stars (e.g. Margulis et al. 1990; SL95; Kahane & Jura 1996; Knapp et al. 1997a) and may be quite common.
GCGSS 283, R Lyn, has previously been detected by BL94, and the data
in Table 5 (click here)
are in good agreement. FU Mon (GCGSS 212) was detected by SL95, who observe
two narrow features at -44 and +14 which they attribute to
the blue- and red-shifted components of an expanding circumstellar shell.
However, the optical radial velocity of the star is about -43
(Table 4 (click here)),
and further, the narrow component at +14
is likely to be interstellar, as shown by our observations
with the 500 MHz AOS in Fig. 15 (click here). We conclude that the emission
at
(best seen on the high resolution profile
in Fig. 14 (click here))
is from the circumstellar envelope. The outflow speed,
3
, is very low, but similar low values are found for
some other Mira and semi-regular variables (Wallerstein & Dominy 1988; Young 1995;
Kerschbaum et al. 1996).
Figure 15:
Broad-band CO(2-1) spectrum in the direction of GCGSS 212 (FU Mon).
The horizontal bar shows the range of the observed optical velocities
Table 6 (click here) summarizes (under the header "Observations'') CO millimeter wavelength observations of S stars published since 1990, including those in the present paper. Data observed prior to 1990 can be found in Loup et al. (1993). The stars in Table 6 (click here) are grouped according to their location in the (K - [12], [25] - [60]) color-color diagram (see Sect. 2.3 (click here)) and arranged in order of right ascension within these groups. Several observations are not listed in Table 6 (click here) because they were made at positions which are too discrepant from the optical position; the stars are GCGSS 133, GZ Peg, and T Cam.
Table 6 (click here) gives the IRAS name (an asterisk before the IRAS
name refers to a note at the end of the table),
the star name and the results of CO observations of the star:
the line observed; the telescope half power beamwidth in arcseconds;
the channel-to-channel rms noise in K; the integrated line brightness
in ; the peak brightness temperature in K;
the central velocity
with respect to the LSR; the wind outflow
speed
; and the reference. All temperatures are expressed in
main-beam brightness temperature. Dashes for any of these quantities
mean either that no emission was detected from the star or that the quantity
in question was not quoted in the paper. Table 6 (click here)
contains observations of 56 stars, with 35 detections.
To first order, the peak brightness temperatures and the integrated CO line intensities should scale inversely with the square of the telescope beamwidth, since the CO lines are usually fairly optically thick and the envelopes in general smaller than the beam. As Table 6 (click here) shows, this is roughly the case, and the agreement among the observations is in general good, with no serious discrepancies.
Figure 16 (click here) shows the histogram of the stellar systemic
radial velocities (with respect to the LSR).
The values adopted for stars with multiple observations are straight
averages of the individual values. The mean velocity for
S stars detected in CO
is and the
radial-velocity dispersion is
. This value refers to a sample of intrinsic S stars,
since no extrinsic S stars have been
detected in CO. This dispersion is typical of a young-disk population
(Mihalas & Binney 1981). Such a population
should have a scale-height above the galactic plane of about 200 pc,
in agreement with the results of Van Eck et al. (1997).
Figure 16:
Distribution of radial velocities (with respect to the LSR)
from CO observations of intrinsic S stars
Figure 17 (click here)
shows the histogram of the wind terminal velocity
compared with the distributions for three other sets of molecular
line observations; those for nearby oxygen-rich Mira variables
(Young 1995), for semi-regular (SRa and SRb) variables
(Kerschbaum et al. 1996) and for carbon stars (Olofsson et al. 1993).
The outflow speeds
for the S stars are taken from
Table 6 (click here). We used average values for stars with
multiple observations except for RZ Sgr (20120-4433), for
which CO(1-0) and (2-1) observations give discrepant values
(14 and 8.8
; SL95). We use the velocity derived
from the CO(2-1) observation since this has a much higher
signal-to-noise ratio, but note that the CO(1-0) line may really be
broader; the larger telescope beam at this wavelength could be
detecting gas at larger distances from the extended envelope of this
star, which could have a larger outflow velocity.
Figure 17 (click here) shows that oxygen-rich Miras have the smallest outflow
velocities (median 6.5 , largest value 12.7
), while those of the SRVs cover a similar range (median
8.0
, largest value 15.6
). Carbon
stars have the largest outflow velocities (median 12.0
,
largest value 33.2
) while as expected S stars are
intermediate (median 8.5
, largest value 24.7
). The largest
in our sample of S stars
is observed for the CS star TT Cen, a border case between S and C stars
(see Sect. 2.1 (click here)).
Figure 18 (click here) shows that, among Mira S stars, the outflow velocity
correlates well with the period of the photometric variations, a
result already discussed by Heske (1990) and by
Olofsson et al. (1993).
Jura (1988) finds an almost identical dependence on period for the ratio of
flux densities at 25 and 2.2
m.
Figure 17:
Normalized histogram of terminal wind outflow speeds
of a) oxygen-rich Miras (Young 1995) and semiregular variables
(Kerschbaum et al. 1996) and b) S stars (present work) and carbon stars
(Olofsson et al. 1993)
Figure 18:
Terminal wind outflow speed of S stars measured from CO millimeter
wavelength emission
lines versus period of photometric variations (from the GCVS).
Circles: Mira variables (open circles are for stars for which there
is only one measurement of ); crosses: semi-regular variables. The
points corresponding to the two outflow speeds for RS Cnc are connected
by a vertical line. The data for DY Gem, 06331+1414 (SRa; P = 1145 d),
are not included in the figure
The CO data from Table 6 (click here)
were used to calculate mass loss rates for the
detected S stars and upper limits for the non-detected stars.
The CO lines were modeled
using a code based on that of Morris (1980) which assumes
spherically-symmetric mass loss at a constant rate and constant
outflow
speed
with excitation by collisions and by infrared photons at
4.6
. The envelope
outer radius was taken to be that at which the CO is photodestroyed by
the diffuse interstellar radiation field using the calculations of
Mamon et al. (1988). The details are given by
Knapp et al. (1997b).
The relative abundance of CO to
was assumed to
be
for all stars (Lambert et al. 1986;
Smith & Lambert 1990).
The infrared radiation field was approximated as that of a black body
of temperature 2500 K and radius cm. Models show that
the CO line strength is only weakly dependent on the radiation
field, so this simplifying assumption is unlikely to produce an
uncertainty of more than 20% in the derived mass loss rates.
The distances are derived by adopting absolute magnitudes which
depend on location in the IR color-color diagram.
For stars in Region A (photospheric colors) we assume , i.e. one magnitude below the RGB tip for solar-metallicity
stars with M = 1
(Schaller et al. 1992). The bolometric
correction
is derived from
the apparent bolometric
flux for BD Cam (HR 1105)
obtained by integrating the flux densities corresponding to
the UBVRIJKL magnitudes from Lee (1970) plus the IRAS flux densities.
The corresponding absolute K magnitude is -4.6. Other authors, e.g.
Jura
(1988) have used
for all S stars; this absolute
magnitude is derived from carbon stars in the solar neighborhood
and AGB stars in the Magellanic Clouds.
For stars in the other regions, which are supposedly more evolved than
are those in Region A, we adopt .
That value of
yields distance moduli that are consistent
with direct determinations when available.
For the composite system
Gru (S5,7e + G0V),
Ake & Johnson (1992) derive a distance modulus of
6.0 from a fit to the UV spectrum, corresponding to a distance of 160 pc,
identical to the value derived from
the K magnitude (Table 1 (click here)). For T Sgr, a distance of 1000 pc (as
compared to 810 pc from the K distance modulus) is derived by Culver & Ianna
(1975) from the spectral type F3IV assigned to its companion.
For
Cyg, a distance of 136 pc is obtained from the distance
modulus in the K band, consistent with that (106 pc)
derived from the Hipparcos parallax (
mas;
van Leeuwen
et al. 1997).
The evolutionary status of stars in Regions D and E is unclear, as is
their relationship to stars in other regions of the color-color diagram,
and so their absolute magnitudes are uncertain. Since these are generally
stars of spectral type SC, the choice appears
to be a reasonable one.
We modeled the wind from RS Cnc as two separate components, fit only to the CO(2-1) observations described in the previous section.
The upper limits to the mass loss rate of S stars not detected in CO
were calculated using the median outflow speed of
8.5 found for the detected sample and assuming
that we can detect a line of brightness temperature three times the
rms noise (examination of the data in Table 6 (click here)
suggests that this is reasonable).
The best fit mass loss rates were found by calculating the model peak antenna temperature and integrated line intensities for a given input mass loss rate and comparing it with the observations. The mass loss rates were adjusted until reasonable agreement with all of the observations was found, and were corrected for the mass of helium (see Knapp & Morris 1985).
The results are listed in the rightmost columns of
Table 6 (click here) (under the header "Model''), which list
the distance, the mean outflow speed (the same for all lines; the
value of 8.5 used for calculating the upper limits
is given in parentheses), the CO photodissociation radius in cm and in arcseconds,
and the predicted CO peak line temperature and integrated CO
line intensity for each of the observed lines.
Comparison between the calculations and observations
shows that uncertainties in the observations introduce about a
factor of two uncertainty into the mass loss rate.
Table 4:
S stars observed in CO at the Caltech Submillimeter Observatory
Notes to Table 4
1. The spectral type in column 8 is from the GCGSS
2. Column 10 gives the observed range of optical velocities, taken from the
following references: Brown et al. 1990 (GCGSS 149 and 589),
Udry et al. (1998) and Jorissen et al. (1998) (GCGSS 117 and 212),
Wilson 1953 (GCGSS 283, 803 and 816),
Hoffleit (1982) (GCGSS 796)
3. All stars observed in the CO(2-1) line except GP Ori, NQ Pup and UY Cen,
which were observed in the CO(3-2) line.
Table 5:
CO(2-1) data for the four S stars detected by
the CSO observations
Table 6:
A compilation of CO data for S stars (Cols. 3-10; for data prior to 1990,
see Loup et al. 1993) and mass loss rates
fitting these data (Cols. 11-19), assuming = 2500 K,
cm, and
Table 6: continued
Table 6: continued
Table 6: continued
Table 6: continued
Table 6: continued