For a better understanding of the results shown in this Paper it is worth to study first, making use of the data available in the literature (Gezari et al. 1993), the characteristics of the near infrared emission observed in the different type of objects which are expected to be present in our sample. As refered above, among them, we can find not only PNe and other post-main sequence stars evolutionary connected with PNe, such as late-AGB and post-AGB stars, but also, although in a small proportion, a wide variety of young stellar objects, like T-Tauri stars, Herbig Ae/Be stars or compact H II regions, and a few active galaxies.
For an adequate interpretation of the data we need to consider the different sources of near infrared flux which can contribute to the observed emission:
a) Thermal emission from plasma: basically recombination continuum and free-free emission from nebular hydrogen and helium, only expected when ionization is present.
b) Thermal emission from dust: the emission due to the cold dust
present in circumstellar envelopes at characteristic temperatures
lower than 300 K is not expected to appreciably contribute to the
emission observed in the near infrared. However, the presence of hot dust
) can easily be detected in the near infrared, specially
in the reddest bands, and may be interpreted as a consequence of
recent mass loss or, in the case of young stellar objects, as an indicator
of the presence of a circumstellar disk.
c) Stellar continuum: it can dominate only in the case of stars not heavily obscured by the circumstellar dust. The hot central stars of PNe emit basically in the ultraviolet, and thus, must be extremely bright to detect its Rayleigh-Jeans tail emission in the near infrared.
d) Emission lines: basically recombination lines from hydrogen and
helium. They can be the main source of near infrared emission in the range
between 1 and 
for ionized envelopes. The neutral helium triplet at
can sometimes be bright enough to completely dominate the
emission in the J band, even although this emission feature is located close
to the edge of the photometric band. In addition, the Paschen recombination
lines of hydrogen can also contribute to the J emission, while Brackett and
Pfund lines and, in some cases, molecular hydrogen emission would
affect the emission observed at the H and K bands.
Figure 1: J-H vs. H-K two-colour diagram where we show the position
of very well identified objects present in our sample for which data are
available in the literature. The solid line in Region I indicates
the position corresponding to main sequence and giant stars, for comparison
(see text for more details)
In Fig. 1 (click here) we show a near infrared two-colour diagram (J-H vs. H-K) where we have plotted together the position of the different types of objects present in our sample for which data are available in the literature. We have divided this diagram into several regions (from I to V), which will be used in our subsequent analysis. The colours are not extinction corrected, since the value of the extinction is, in most cases, not known or bad determined. As an indication, a vector representing the effect of reddening is included in this diagram, together with the positions corresponding to main-sequence and giant stars, black-body emission at different temperatures and that corresponding to a plasma at a temperature of 104 K (Whitelock 1985).
| Class | Region I | Region II | Region III | Region IV | Region V | Total |
| Planetary Nebulae | 14 | 7 | 1 | 22 | 69 | 113 |
| Late-AGB/Post-AGB stars | 9 | 12 | 3 | 10 | 0 | 34 |
| Young Stellar Objects | 4 | 13 | 21 | 14 | 0 | 52 |
| Galaxies | 8 | 1 | 0 | 9 | 0 | 18 |
The distribution of the various types of objects found in the near infrared two-colour diagram is presented in Table 2 (click here). In the following we will try to characterize each class of object according to their near infrared properties.
PNe have only recently been systematically observed in the near infrared. Before 1985 only a few PNe, the brightest ones, had been observed in this wavelength range (Willner et al. 1972; Allen 1973; Persson & Frogel 1973; Allen & Glass 1974). More recently, several surveys have been carried out providing near infrared photometry for more than 200 PNe in the J, H and K bands and, in some cases, also in L' and M (Whitelock 1985; Kwok et al. 1986; Peña & Torres-Peimbert 1987; Persi et al. 1987). Most of these PNe satisfy our selection criteria and are, therefore, included in our sample. In Fig. 1 (click here) we show their position in the near infrared two-colour diagram J-H vs. H-K.
From the analysis of Fig. 1 (click here) it is clear that there is a strong
concentration of sources in Region V, specially around the
so-called nebulae box, as defined by
Whitelock (1985), which has also
been represented in this figure with a solid line. Around two thirds of
the PNe observed in the near infrared fall inside or in the surroundings
of this box. This confined region of the two-colour diagram is well
separated from that where main-sequence and giant stars are located and shows
no overlap with any other type of stellar object. PNe in Region
V show a characteristic J band excess with respect to the emission expected
from a plasma at an electronic temperature of 
. This effect cannot be
attributed to a different 
in the plasma, since this would only produce
a small displacement up and left (if increases) or down and right
(if decreases) in the diagram, but it is probably due to the presence
of the strong He I triplet at .
This hypothesis was confirmed by Whitelock (1985) and
Peña &
Torres-Peimbert (1987), who found a clear correlation between the
J-H extinction corrected infrared colour and the 
abundance.
In fact, the emission excess in the J band is more pronounced in
intermediate excitation class PNe. In very high excitation PNe the He I
triplet is very weak, although the He II emission line at 1.162 
m
can still contribute to a small J excess. On the other hand, in very low
excitation PNe, most of the helium is in neutral state, producing a much
weaker He I triplet and, thus, a smaller J excess.
A considerable number of PNe lie close to the borderline between Regions IV and V above and to the right of the nebulae box, and a few of them can also be found in Region III, as we show in Fig. 1 (click here). Although this anomalous location could simply be the result of interstellar reddening, it can also be explained as the effect of hot dust present in the circumstellar envelope. This would produce a characteristic excess in the longer wavelengths and, thus, a larger value of the H-K colour index. The presence of hot dust at temperatures around 1000 K in most of these PNe is confirmed by the fact that they also show large values of K-L' which cannot be explained by the effect of interstellar reddening. Remarkably, most of these dust-type PNe are considered to belong to the youngest group of PNe, such as M2-9, Vy2-2, IC 418, Hu2-1, Sw St1, K3-62, IC 5117 or Tc 1. We interpret the presence of hot dust in their envelopes as the result of mass loss processes suffered in the very recent past or even still taking place, as the P-Cygni profiles found in some cases in their spectra indicate.
Finally, a small number of PNe show near infrared colours similar to those
of main sequence stars and giants. Probably the central stars of these
stellar-type PNe are binary systems and we are only detecting the
emission coming from the companion star in the near infrared. Another
possibility is the presence of a foreground source in the aperture which
is contaminating the near infrared photometry. This kind of near infrared
emission is observed, for instance, in 
, IC 3568, 
,
NGC 5315,
and Pb 8, all them located in Region I of our diagram. Clearly,
additional observations are needed for all them.
The peculiar location of a few PNe in Region II deserves a special attention.
Most of them, such as 
or 
, are close to Region I, and the
near infrared emission observed can be explained as the combination of a
moderate infrared excess with a stellar-type emission. The most extremely
reddened colours observed are those of 
, whose optical spectrum
corresponds to a very low excitation PN
with its central star showing an effective temperature of only 25000 K
(Kwok
1985). Its location in the HR diagram
(West & Kohoutek 1985) and the
optical and radio compact appearance indicates that is a very young
PN, for which the circumstellar extinction is still very high.
Very similar characteristics are observed in the very well known proto-PN
CRL 618 (Latter et al. 1992).
While still on the AGB, stars are strongly variable, due to stellar pulsation. They are then called "variable OH/IR stars", since most of them show a double-peaked OH maser emission at 1612 MHz and are very bright in the infrared, heavily obscured by thick circumstellar envelopes formed as a consequence of the strong mass loss. Shortly after the end of the AGB, the mass loss suddenly stops and for no longer the star is variable, while the effective temperature of the central star increases. The star is now in the post-AGB phase and recognized as a "non-variable OH/IR star" while the OH maser emission is still detectable. After the partial dilution of the circumstellar envelope in the interstellar medium the OH maser emission dissapears and the central post-AGB star becomes observable again in the optical, in its way to become a new PN.
In Fig. 1 (click here) we can see that IRAS sources recently identified as late-AGB
or post-AGB stars in the literature show a wide distribution in the near
infrared two-colour diagram and can be located everywhere, with the only
exception of Region V. Fortunately, objects in different regions
of the diagram show peculiar characteristics which can help us in our
identification purposes. For instance, most of the strongly reddened objects
found in Region II are identified as OH/IR stars since they show OH maser
emission and no optical counterparts. Variable OH/IR stars in this region
of the diagram show near infrared properties which are an extension
towards more extreme values of those observed in optically bright Mira
variables, which are also stars in the AGB. Mira variables are not plotted in
our diagram because they do not fulfill our selection criteria, but many of
them are also known to be located in Region II, inmediately above but
close to Region I (Feast & Whitelock 1987). Objects in Regions III and IV,
on the other hand, are also affected by a strong circumstellar reddening but
their slightly different position can simply be due to an excess of emission
in the K band produced by the presence of hot dust in the envelope. Some
of them are not variable in the near infrared and may be identified as early
post-AGB stars. While objects in Region III usually show OH maser emission
and no optical counterparts, those in Region IV are not so frequently
detected in the OH maser line. Moreover, they are not so strongly obscured
and sometimes show a faint optical counterpart.
Finally, the objects found in Region I, showing a stellar-like
emission with little or no reddening, are identified as evolved post-AGB stars,
now observable again in the optical after the dilution of the
circumstellar envelope in the interstellar medium. They show a small
irregular variability and no OH maser emission. Some
near infrared excess is observed in a few of them located to the right of
the main-sequence. This is probably indicative of recent post-AGB mass-loss,
something which, for some of these objects, has been
confirmed through the detection of H
emission in the optical spectrum.
Under this category we can find both heavily obscured young stellar objects showing the most extremely reddened colours, together with optically bright stars with little or just a moderate near infrared excess.
Among the first group it is possible to identify deeply obscured compact HII regions and Herbig-Haro objects still embedded in the molecular clouds in which they have been originated. They are predominantly located in Region III of the near infrared two-colour diagram, although a few are also found in Region II, always close to the position expected for black-bodies emitting at temperatures between 800 and 1500 K.
The second group is basically formed by T-Tauri and Herbig Ae/Be stars, which are not so heavily obscured. Most of them are located in Region IV, although we also detect a few in Regions I and II. The circumstellar disks usually associated to these objects are probably the responsible for the presence of the near infrared excess observed (Strom et al. 1989; Hillenbrand et al. 1992). Finally, it is important to remark that, again, none of these objects is found in Region V of the diagram.
The small number of galaxies found satisfying our selection criteria are known to show active nuclei and most of them are classified in the literature as bright Seyfert galaxies. Although we expect to find a very small number of them among the unidentified objects in our sample it is worth to investigate whether they show peculiar near infrared colours which could be used for their identification.
As we can see in Fig. 1 (click here), they are all well confined in a relatively small region of the diagram in the intersection of Regions I and IV. Unfortunately, this is the same location in which we can also find, as we have already mentioned, evolved stars, young stellar objects and, sometimes, even PNe.
Seyfert galaxies detected by IRAS have previously been studied in this wavelength range by different authors (Sanders et al. 1988; Carico et al. 1990). They have shown that those with the higher luminosities usually show a near infrared excess originated in the circumnuclear regions, probably associated to a strong star-forming activity. According to this, we suggest that the group of galaxies in our sample with a moderate near infrared excess, located in Region IV, must possess starburst nuclei. On the other hand, the galaxies found in Region I of our diagram, showing stellar-like colours, may correspond to those with just a moderate nuclear activity, in which the dominant emission observed is originated in the outer disk and is basically due to the stellar content of the galaxy.
From the results above shown it is clear that, in most cases, it is not possible to determine, based on near infrared data alone, the nature of previously unidentified IRAS sources in our sample. Thus, additional criteria are needed to be used in combination with the near infrared photometry. Unfortunately, the only information available for many of the sources observed comes from IRAS data, since no observations in other spectral ranges are yet available.
Figure 2: IRAS two-colour diagram where we show the position of
very well identified objects in our sample together with the regions
associated to a) Optically bright Mira variables; b) Variable OH/IR stars;
c) T-Tauri and Herbig Ae/Be stars; d)
Active galactic nuclei; and e)
Compact HII regions. The thick solid line indicates the limits used
for the selection of our sample
and the exponential curve represents the evolutionary track
followed by AGB stars with increasing mass loss (see text for details)
In order to investigate whether a more detailed study of the far infrared properties shown by the variety of objects found in our sample could be used to provide useful colour classification criteria, we have plotted in Fig. 2 (click here) an IRAS two-colour diagram [12]-[25] vs. [25]-[60], where
In this plot we show the region of the diagram satisfying our selection criteria which, as we know, basically corresponds to that where most of the well known PNe are located (Pottasch et al. 1988), together with those associated to: a) Optically bright Mira variables (Herman 1988) b) Variable OH/IR stars (Sivagnaman 1989; te Lintel Hekkert et al. 1991); c) T-Tauri and Herbig Ae/Be stars (Harris et al. 1988); d) Active galactic nuclei (de Grijp et al. 1987; Kailey & Lebofsky 1988); and e) Compact HII regions (Antonopoulos & Pottasch 1987). The exponential curve crossing the diagram corresponds to the sequence of colours expected from model predictions for stars losing mass at increasing rate at the end of the AGB (Bedijn 1987).
As we can see, although a strong overlap exists between different kind of objects in specific areas of the diagram, it is also possible to identify wide regions in it where this overlap is minimum or restricted to only two different groups. The consistency of this colour association have been checked by plotting in Fig. 2 (click here) the position of all the well identified sources in our sample. Galaxies and young stellar objects show characteristic far infrared colours in excellent agreement with the predictions, and appear concentrated in specific areas of the diagram, while PNe are widely distributed, as already known. On the other hand, most of the objects classified as late-AGB/post-AGB stars in the diagram are concentrated in the region associated with variable OH/IR stars, as expected. Most of them are known to show variable OH maser emission. The few objects in this class located well outside the limits of this Region are either heavily obscured non-variable OH/IR stars, already in the post-AGB stage, or optically bright post-AGB stars with a supergiant-like spectrum, as we will see in the following section.
Apart from the IRAS photometry, mid-infrared IRAS Low Resolution Spectra
(LRS) have also been used in our identification process, although they
are only available for the brightest sources in our sample
(Olnon & Raimond
1986). LRS spectra are classified according to the slope of the continuum
and the presence or absence of specific features in the spectrum (see
IRAS Explanatory Supplement 1985). The
LRS classes 3n and 7n, for instance, correspond to objects with
a very red continuum and a strong silicate absorption feature around
9.8 m, where n is a number from 1 to 9 increasing with
the strength of this absorption. Well identified heavily obscured OH/IR
stars in our sample are allways associated with one of these two LRS classes.
On the other hand, the LRS class 9n is characteristic of evolved PNe,
since it corresponds to emission line spectra, where n, in this case,
is a number from 1 to 6 increasing with the excitation class. Unfortunately,
very few young stellar objects and galaxies are bright
enough to have an available LRS. When this is the case, the LRS class
5n, which corresponds to featureless spectra with a red
continuum, is frequently observed.
Finally, we have also used the IRAS variability index as an additional source of information for the classification of the unidentified objects in our sample. This variability index is a number between 0 and 99 which indicates the likelihood of variability for a given IRAS source. It is based on the fact that the inclusion of an infrared source in the Point Source Catalogue required its detection in at least two different scans, which could be separated by hours, weeks or months. In this way it is possible to have some information about the variability of the source. Moreover, the way in which this index was computed favours the association of the highest values with long-period variables. Well known variable OH/IR stars showing smooth long-term variations are usually found associated to values well above 50%, with a strong concentration around 99%, while other variable objects in our sample, such as T-Tauri stars, with small amplitudes and irregular variations are associated to lower values of the variability index (below 50%). It is important to remark that a low variability index associated to an OH/IR star does not necessarily mean that the source is actually non-variable. Objects with very long periods may look like non-variable if they were detected in scans separated just a few weeks and there are ecliptic positions which were poorly scanned by IRAS where variable sources were missed (Whitelock et al. 1994). On the other hand, variable OH/IR stars may also be misidentified as non-variable when the observations were taken at different epochs corresponding to a similar phase in the light curve. In contrast, a high variability index can only be associated to true variable stars.
Figure 3: J-H vs. H-K two-colour diagram where we show the
position of the infrared sources observed