Previous works by Elias et al. (1986), Wood et al. (1986,
1992), Reid et al. (1990), and Reid
(1991) have shown that some red supergiants and AGB stars have been
detected by IRAS in the LMC, at least at 
. As the distance
of the LMC is about 50 kpc, one may however be surprised
that IRAS could detect AGB stars so far away. A comparison
with stars of our Galaxy could allow us to answer this question
and to define the physical properties of such stars.
Reid et al. give part of the answer based on the most optically
thick OH/IR stars known in the Galaxy, and on the "prototype''
of optically thick carbon stars, IRC+10216. They conclude that AGB stars
with similar physical properties should "be detected with ease''
in the IRAS survey.
Lets consider as an example one
of the most extreme carbon star AFGL 3068 (Price & Walker
1976), and the well known O-rich star WX Psc. AFGL 3068 is
particularly optically thick as [K-L]=7 (le Bertre 1992). WX
Psc has a known optical counterpart but is not optically thin as
the 
silicate feature is slightly self-absorbed.
Assuming an intrinsic luminosity of 
,
their distances would be 0.95 and 0.54 kpc respectively
(see e.g. Loup et al. 1993). At 50 kpc,
such stars would have IRAS flux
densities of 0.25 and 0.13 Jy, and 
IRAS flux densities
of 0.28 and 0.11 Jy, respectively. The sensitivity limit of the IRAS-PSC
is 0.25 Jy at 12 and , and 0.15 and 0.22 Jy in
the catalogue of Schwering & Israel (1990).
Therefore a source like AFGL 3068
could be detected in the LMC, but not easily, and WX Psc would
not be detected (or by chance as Schwering & Israel report
a few 12 and detections at a level of 0.07 Jy).
The intrinsic luminosity of these 2 stars could be larger
than , but also smaller, and in addition
their luminosity varies by more than a factor 2.
In an attempt to get a more complete overview, we have
used the sample of Galactic sources whose CO emission in the
rotational transitions J=1-0 or/and J=2-1 has been detected
(Loup et al. 1993). Though this sample is strongly
observationally biased, it contains all the chemical types (O-rich,
C-rich, and S stars), and covers the whole range of mass loss rates
(10-7 to 
). This sample contains
about 400 AGB stars and a few M supergiants. Their bolometric
luminosities have been calculated from optical, JHKL(M), and IRAS
photometry, when enough data were available; ortherwise they were
estimated using the bolometric correction to IRAS data of van der
Veen & Rugers (1989). Distances have been estimated assuming an
intrinsic luminosity of for the AGB stars, and
for the supergiants, corresponding to bolometric
luminosities of -5.25 and -7.75, respectively.
Figure 1: Location of galactic AGB stars and M supergiants
detected in the CO (1-0) or CO (2-1) lines (Loup et al.
1993) in a [C21, S12] diagram as if they were located in
the LMC. Distances were calculated from bolometric fluxes and
assuming a luminosity of for AGB stars,
for M supergiants. S12 was then scaled to 50 kpc. Symbols are
defined in the figure. The two correlations seen for M and
C stars come from the bolometric correction of van der Veen
& Rugers and are not real (see also Sect. 2). The two dashed lines
correspond to the detection limit of the PSC (0.25 Jy at )
and of Schwering & Israel (0.15 Jy). Also plotted in
Fig. 1 (click here) are
the foreground stars listed in Table 6 (click here). One can see that, for most
of them, their location in the diagram would be sufficient
to determine that they do not belong to the LMC
Figure 1 (click here) shows their IRAS color 
(12S25/25S12)
as a function of their IRAS flux density scaled to 50 kpc.
The dashed lines indicate the IRAS-PSC and Schwering & Israel
sensitivity limits, 0.25 and 0.15 Jy. Note that the faintest
sources at in Schwering & Israel have S12=0.07 Jy.
The two correlations appearing in Fig. 1 (click here) for O-rich
and C-rich stars only reflect the bolometric correction
of van der Veen & Rugers (1989) and are not real.
It appears clearly that, whatever the value of C21, very few AGB
stars could be detected in the PSC if they are not more luminous
than . The situation is a little better with the
IRAS pointed observations, but we still expect that most AGB stars
with 
have not been detected by IRAS. The faintest
"obscured'' AGB star discovered until now actually has a bolometric
luminosity of -5.1 (
), and a flux density of
0.13 Jy (Reid 1991; Reid et al. 1990). The IRAS
color C21 can be considered as a rough estimator of the total dust
opacity in the circumstellar shell, and hence as a rough estimator
of the mass-loss rate (see e.g. Rowan-Robinson et al. 1986;
Bedijn 1987; Chan & Kwok 1990). As expected, one sees in Fig. 1 (click here)
that only optically thick AGB stars could be detected by IRAS
as the value of S12 decreases drastically when C21
decreases.
Comparing Fig. 1 (click here) with Fig. 9b in Loup et al. (1993),
we conclude that most AGB stars detected by IRAS in the LMC should
have a mass-loss rate larger than 
.
Sources with intermediate mass-loss rates are probably still almost
totally undiscovered in the MCs as they would already be too faint
for optical surveys, but not optically thick enough to have been seen by IRAS.
This is very well illustrated for carbon stars in the Fig. C1 of Groenewegen
& de Jong (1993) where they show the theoretical relation
between S12 and the I magnitude for various bolometric luminosities
and mass-loss rates.
The sample of sources detected in CO (Loup et al. 1993)
contains only a few supergiants, which does not allow a statistical
overview. We consider the example of 
Ori, VY CMa, and VX Sgr.
If their luminosity is , their mass-loss rates
estimated from CO observations are 
,
, and , respectively.
Their IRAS flux densities scaled to 50 kpc are 0.07, 1.7,
and 0.18 Jy. So even LMC red supergiants should have a relatively optically
thick dust envelope to be detected by IRAS.
From the previous considerations we expect the IRAS sample
of LMC AGB stars to be very incomplete, strongly biased towards luminous
(more than ) and optically thick sources
(without optical counterpart; mass-loss rate larger
than ). Most AGB stars with a bolometric
luminosity fainter than about -5.2, even if very optically thick,
have probably not been detected by IRAS. In particular, we therefore
expect to find far fewer C-rich stars than O-rich stars in the
IRAS sample, though carbon stars are more numerous than late M stars
in the MCs (Blanco et al. 1980). Groenewegen & de Jong
(1993) reach the same conclusion through a theoretical approach.