Circumstellar envelopes around evolved stars, in particular AGB stars and late-type supergiants, are a two-fluid system, composed of gas and dust. As gas and dust are expected to be coupled both dynamically and chemically, it is natural to look for correlations between their characteristics, and in particular their emission properties.
A consensus has grown (see e.g. Habing 1995, and references therein) concerning the essential role of dust driven out by radiation pressure in the generation of the largest mass-loss rates and the general complexity of mass-loss mechanisms. There is no simple universal estimator of mass-loss rate. Every estimator must be carefully tested with the help of both realistic physical models, and large statistical samples. However, it has been confirmed that among these estimators, the most useful for massive molecular and dusty envelopes are the intensity of millimeter CO lines and the far-infrared emission.
Because of its great stability, CO is thought to
concentrate almost all the carbon in AGB oxygen-rich
circumstellar envelopes,
with the limitation that the radial extension is limited by
photodissociation by interstellar UV radiation (see
Mamon et al. 1988, and references therein). The intensity of millimeter
lines was proposed by
Knapp & Morris (1985) as a precise estimator of
mass-loss rate.
In parallel, the study of the dust component has been greatly improved by the
use of IRAS colours
(see e.g. van der Veen & Habing 1988;
van der Veen 1991) and low resolution spectra
(LRS, IRAS Science Team, 1986). In
particular, a simple approximate relation using the
flux can be
used to infer the dust mass-loss rate of large samples of sources
(Jura 1986).
With the availability of large databases of CO observations
(Nyman et al. 1992;
Loup et al. 1993), it has been possible to
compare CO and far-infrared emissions for all types of
circumstellar envelopes. Both resulting determinations of the mass-loss
rates appeared roughly consistent, as shown by
Nyman et al. (1992),
with a narrow range of the ratio between the CO and
intensities
for most sources. However, it had been known for a long time that the
correlation between CO and
far-infrared emission does not hold for some "extreme'' objects. In
particular, Heske et al. (1990) have shown that in the coldest
OH/IR stars CO emission is generally weaker by at least one order of
magnitude than that expected from the mass-loss rate deduced from the
IRAS intensity. Heske et al. partly explained such deficient CO emission
by the very low kinetic temperature in those very cold envelopes or/and by
variations of their mass-loss rates (see also
Kastner 1992;
Groenewegen 1994;
Justtanont et al. 1994; Delfosse et al. 1996).
Our study focuses on a class of slightly warmer O-rich IRAS sources.
We attempted to avoid previous biases due to selection criteria, such as
the avoidance of the Galactic plane. Furthermore, our CO observations were
conducted with a better sensitivity.
In the first paper of this series,
(Josselin et al. 1996, hereafter Paper I),
we presented a statistical study of the CO observations of these
objects, with a discussion of the 
ratio.
The peak temperature of the CO line was preferred to the integrated
intensity as it is easier to derive an upper limit for
than for
for the numerous non-detections. Furthermore, it stresses the
distinction between AGB stars and supergiants, through the effect of the
expansion velocity 
.
We showed that a low CO emission with respect to the far-infrared emission
occurs in fact in an significant fraction of O-rich envelopes.
The objects in our sample have intermediate values of
mass-loss rates and IRAS colours. They are located in regions
IIIa1 and IIIa2 of the IRAS colour-colour diagram of Fig. 1 (click here)
(i.e. 0.69 < S25/S12 < 1.20, where S12and S25 are
the 12 and 
IRAS fluxes, respectively).
This is a region where the 
silicate feature is usually in emission
and where many OH/IR stars are
located (but not the coldest ones), as well as
dusty supergiants.
Their mass-loss rates are expected to be a few
.
In this second paper, we detail the analysis of the set of CO data now
available. In Sect. 2, we describe the observations of CO and OH.
In Sect. 3, we show correlations of
the 
ratio with stellar characteristics, such as galactic coordinates,
variability, and spectral type.
In Sect. 4, we discuss the possible causes of a high
value in an AGB star. In Sect. 5, we compare CO and OH
observations. Finally, we discuss some peculiar sources. Two forthcoming
papers will deal with visible spectroscopy (hereafter Paper III) and
infrared photometry (Paper IV), basis of the determination of luminosity,
opacities and mass-loss rates.