Due to the higher data quality the sources surface density at K-band near
the maser positions in the 1994 observations
is a factor 3-20 (depending on the field)
higher than in the 1991 observations (see Paper I). As a consequence,
for all the
maser components we can find a NIR source within 
.
Hence we do not have strong statistical support for the uniqueness of
the association between the maser components and the NIR sources
based only on positional coincidence.
In fact, due to the high densities of sources
we expect some contamination from background/foreground sources.
The fields in which the mean separation between the K-band sources is
much smaller than have been marked in Table 2 (click here)
with a "C'' in the last column (note that in all those cases the
distance between the maser components and the NIR sources is
).
In order to find additional criteria to establish a physical association we have examined the NIR colours of the close coincident sources and those of background sources.
Since many of the sources coincident with the maser components are not detected at J and H, we will use the (H-K, J-K) colour-colour diagram to investigate the nature of the infrared sources, as in Paper I.
Since there is no systematic difference in the colour characteristics of the NIR sources associated to the H2O and OH masers (in fact in many cases both masers are associated with the same NIR source), we will discuss the nature of the sources associated with both masers together.
In Fig. 4 (click here) the (H-K, J-K) colour-colour diagram and the (H-K, K) colour magnitude diagram for the sources coincident with the maser components are presented. The colours of main sequence (MS) stars are from Koornneef (1983), the magnitudes from Schmidt-Kaler (1981) and the reddening vectors have been plotted assuming a Rieke & Lebofski (1985) extinction law. The K magnitudes of all the sources have been normalized to a distance of 5 kpc from the Sun, assuming the distances quoted in Table 1 (click here). In Fig. 5 (click here) similar plots are given for the sources detected within 1 arcmin from the maser reference of FC89, presumably field stars.
From a comparison of Figs. 4 (click here) and 5 (click here) it can be clearly seen that the field stars are distributed close with the locus of main sequence and reddened main sequence stars, while almost all the sources coincident with the maser components show a conspicuous infrared excess.
Figure 4: Colour-colour (H-K, J-K) and colour-magnitude
(H-K, K) diagrams for the sources associated with the maser components.
In the colour-colour diagram (left) the continuous line represents the
colours of Main Sequence stars (from Koornneef 1983), the
reddening vector for 20 mag of extinction in the V-band has been plotted
assuming a Rieke & Lebofski (1985) exctinction
curve. In the colour-magnitude diagram (right)
the K magnitudes have been scaled to a distance of 5 kpc without
correction for interstellar reddening. The continuous line marks the locus
of main sequence stars (Koornneef 1983 and
Schmidt-Kaler 1981)
Figure 5: Colour-colour (H-K, J-K) and colour-magnitude
diagrams (H-K, K) for all the sources detected within 1 arcmin from the
H2O maser reference in the 1994 data. The distance of the field stars is not
known, of course, but, in order to compare with Fig. 4 (click here), we have scaled
the K magnitude to 5 kpc using the maser distance
Consequently, independent support to a physical association between maser components and NIR sources is given by the fact that the NIR sources close coincident with the maser components show NIR excesses, while the field stars are located along the reddening line.
This provides proof of a distinct property of the maser associated
sources.
As discussed in Paper I and in Testi et al. (1997) the
NIR excess emission is probably the result of the superposition of (at
least) three main contributions: i) the heavily extincted stellar
photosphere of a massive young stellar object (the ultimate energy source of
the entire system), ii) a contribution from free-free and free-bound continum
emission from ionized gas and iii) the emission of the hot dust surrounding
the YSO.
In particular, in the earliest evolutionary stages, when dust is still
present very close to the exciting star,
a hot dust shell dominates the
emission at wavelengths longward of 
m and the
object is effectively detectable at K band even if still embedded in the
parental molecular clump.
Strong support for the association procedure is also provided by the three sources of this sample that have been studied in detail by Persi et al. (1996, 1997) and Testi et al. (1997). The arcsecond resolution comparison of the NIR images with mid infrared and/or thermal molecular line emission and mm-wave continuum observations show that the NIR sources associated with the maser emission using the "nearest neighbour'' criterion are also associated with high density molecular clumps and dense cool dust clouds, which are other indicators of the presence of a YSO. Due to the absence of a UCHII region (in spite of the high luminosities inferred from the IRAS far infrared observations), the three NIR sources discussed above are believed to be in a very early phase, when the size of the ionized region around the young star is so small that the radio continuum emission is strongly self-absorbed. Consequently, the UC HII region is undetectable at radio wavelengths (see also the discussion in Testi et al. 1997), while hot dust emission can be detected even through strong absorption.
In a forthcoming paper (Testi et al. 1997 in preparation) we will discuss the NIR properties of the sources associated with the masers in the framework of a model which takes into account the various emission components.