Several authors (e.g. Netzer & Elitzur 1993; Young 1995) have
shown that AGB stars with higher mass loss rates have higher
outflow velocities. The corresponding relationship for S stars
is shown in Fig. 19 (click here) and compared with the results for oxygen-rich
Miras (Young 1995) and carbon stars
(Olofsson et al. 1993). The
relationship between 
and 
found for oxygen stars by Young (1995) is also shown.
Figure 19:
Wind outflow speed versus mass loss rate
for: open circles: oxygen-rich Miras (Young 1995); filled circles:
S stars (current work); crosses: carbon stars (Olofsson et al. 1993).
The relationship for oxygen Miras found by Young (1995) is also shown
It is not clear at present whether the relationship between
and demonstrated by these data
is real, in the sense that it has a physical origin, or is
due to selection effects. For the present purpose, we will
simply treat it as a convenient way to display the data and compare
samples. The mass loss rates in Fig. 19 (click here) have been calculated
assuming a CO abundance of 
for oxygen stars (Young 1995), 
for S stars and 
for carbon stars (Olofsson et al. 1993). The S and oxygen stars
show essentially the same correspondence between and (though there is a lot more scatter in the
S star data), giving some confidence in the assumed value
of f and the resulting mass loss rates. The apparently higher
values of for a given mass loss rate for carbon stars
(see also Fig. 17 (click here)) will be discussed elsewhere.
The color excess can also be used to give the mass loss rate if the
gas-to-dust ratio and the outflow speed are known. This is
illustrated in Fig. 20 (click here), which shows the envelope density
versus the 
/
color. At low mass loss rates (
y-1) the broad-band colors are dominated
by the colors of the photosphere, as expected, while the stars with
higher mass loss rates show a strong correspondence between the
mass loss rate and the 12 
color, showing that the gas
to dust ratio in these envelopes is roughly constant. The
relationship in Fig. 20 (click here) is in agreement with that
for oxygen stars shown by Habing (1996).
Figure 20:
Envelope density (measured
in 
and 
) versus the ratio of 12 to 2.2 flux densities. The vertical dotted line shows
the photospheric color for a star of temperature 3000 K. Open symbols:
stars in Regions A, B and C of the IR color-color diagram
(Fig. 3 (click here));
filled squares: stars in Region D; filled circles: stars in Region E.
The inverted triangles show upper limits
The two stars TT Cen and RZ Sgr in Region E markedly depart from this relationship, however. These are probably in a transitory phase of evolution; TT Cen is a rare CS star where ZrO bands seem to have disappeared while C2 bands appeared (see Sect. 2.1 (click here) and Stephenson 1973), while RZ Sgr is surrounded by an optical (Whitelock 1994) and IR (YPK) nebula.
Stars with roughly photospheric 12 /2.2
colors show a wide range in envelope densities. This large
scatter may be due to the imperfect coupling between dust and
gas at these low densities, which sets a lower limit to the
mass loss rate for a radiation-pressure driven wind (cf.
Netzer & Elitzur 1993; SL95). Empirically, this limit is a few
y-1 (Fig. 20 (click here) and
Table 6 (click here)).
Figure 21 (click here) presents the mass loss rates of S stars
as a function of their location in the (K - [12], [25] - [60]) diagram.
None of the stars in Region A (stars with
photospheric colors) has detectable circumstellar CO emission, with limits
on the mass loss rates of 
and envelope densities well below those of the detected stars
(Fig. 20 (click here)).
This confirms the lower limit at which a star can lose mass by a
radiation-pressure driven wind estimated by Netzer & Elitzur (1993). The low
mass-loss rates inferred for extrinsic S stars in Region A
also confirm an earlier suggestion (Paper I) that these stars are
less evolved than the intrinsic S stars populating the other Regions of
the IR color-color diagram.
Figure 21:
Mass loss rates of S stars (as measured from CO) in the (K - [12], [25] - [60])
diagram. The diameter of
the circle is proportional to the mass loss rate, as labeled.
Inverted triangles correspond to upper limits
Stars in Regions B and D have moderate mass loss rates, in the range
to 
y-1. One star from Region B
(RS Cnc) has been found to exhibit a double wind (Table 6 (click here)).
Most of the observed stars, and most of the detections, lie in Region C,
which contains stars with moderately optically thick circumstellar envelopes,
likely containing silicate dust. CO emission is detected from 20 of these 23
stars, and the mass loss rates are typically larger than several 
. The undetected stars have
upper limits greater than this value, so that the data are consistent with the
conclusion that all stars in this region lose mass at a rate larger than several
10-7 y-1.
The stars in Region D, which show roughly photospheric 12 /2.2 
m colors despite large [25] - [60] indices,
generally have low mass-loss rates (like Region B).
Stars in Region E have mixed properties. Some, like TT Cen and RZ Sgr,
lose mass at a very large rate (several 10-6 y-1),
while others like FU Mon lose mass at a more moderate rate (a few
10-7 y-1).
Figure 22 (click here) presents the variation of the wind velocity
across the (K - [12], [25] - [60]) diagram, and shows that stars in Region
E also have a wide range of wind velocities, from very low (FU Mon:
2.8 km s
) to very large (TT Cen: 24.7 km s).
Figure 22:
The outflow velocity as measured from CO for S stars in the (K -
[12], [25] - [60]) diagram.
The diameter of
the circle is proportional to the outflow velocity.
Compare to Fig. 10 of Olofsson et al. (1993) for C stars
FU Mon also has a resolved IR envelope (Sect. 2.4 (click here)),
and this fact taken together with a low mass loss rate and a small
outflow velocity suggest that mass loss has just resumed in that
star. Olofsson et al. (1990; 1993) detected several carbon stars (S
Sct, U Ant, TT Cyg) with
a detached shell and a double wind that would also fall in our Region
E. The older, detached shell was produced by a massive, fast wind,
whereas the recent shell is caused by a slow (5 km s), low mass loss rate
wind. Since FU Mon is
an SC star with a C/O ratio close to unity (Dominy et al. 1986),
the mass loss may have come to a halt when the
the C/O ratio approached unity, as already suggested by
Willems & de Jong (1988) and Chan & Kwok (1988).
This scenario may well hold for all SC
stars, even though some, like UY Cen, are in fact located in Region C.
As argued in Sect. 4.2 (click here), that star may be at the end of
the loop in the IR color-color diagram. The horn-shaped CO line
profile of UY Cen observed by SL95, indicative of a detached shell and
fossil mass loss, supports this idea. Finally, other stars lying in
Region E like TT Cen, RZ Sgr and DK Vul have peculiar CO line profiles
(narrow central feature superimposed on a broader less intense
feature) suggesting that they have multiple winds.
Several intrinsic S stars in the sample considered
in this paper are binaries with main sequence
companions, as revealed by the composite nature of their spectrum at
minimum light (WY Cas, W Aql, T Sgr; Herbig 1965;
Culver & Ianna 1975). A composite spectrum is also suspected for S Lyr from its
shallow lightcurve (Merrill 1956). The star 
Gru has a close
G0V visual companion (Feast 1953).
W Aql, WY Cas and S Lyr are surely among the S stars with the largest
mass loss rates, while T Sgr and Gru have mass loss rates
close to the average for stars in Region C. Clearly, a definite
conclusion as to whether binarity can indeed reinforce the mass loss rate,
e.g. by the companion-reinforced attrition process (Eggleton 1986),
would require the knowledge of the orbital separation of
these systems.
It has sometimes been argued (e.g. Eggleton 1986; Tout & Eggleton 1988; Kenyon 1994; Han et al. 1995) that the mass loss rates of giant stars in binary systems must be larger than those of single red giants. More precisely, Han et al. (1995) suggest that the mass loss rate of a giant approaching its Roche lobe in a binary system exceeds by more than a factor 103 the rate predicted by the Reimers formula (Reimers 1975). As far as the binary, extrinsic S stars are concerned, this effect, if present, is clearly not large enough to bring their mass loss rates to the level of the intrinsic S stars. The difference in the evolutionary stages of extrinsic and intrinsic S stars, believed to be RGB (or Early-AGB, according to the terminology of Iben & Renzini 1983) and TP-AGB stars, respectively, thus appears to be of greater importance for the mass loss rate than their binary or non-binary character.