4. The nature of the IRAS counterparts
In this section, we classify the IRAS counterparts as mass-losing AGB stars that are oxygen rich (O), carbon rich (C), or that are not distinguishable between oxygen stars and carbon stars (OC). We argue that some of the stars may be post-AGB or thermal pulse stars (post/TP). The arguments used in classifying the stars are explained below, and the results are tabulated in Table 4.
4.1. Chemical classification from colours and magnitudes
4.1.1. Galactic comparison sample
In paper II we show how in a K- versus H-K diagram carbon stars are distinguished from oxygen stars. A similar diagram can be made using J-K. In Fig. 5 we present the K- versus J-K diagnostic diagram, for the carbon stars (Fig. 5a) and the oxygen stars (Fig. 5b) from the Galactic sample described in paper II. It is clear that in the Milky Way the K- versus J-K diagram can also be used to separate carbon stars from oxygen stars. The Galactic carbon star sequence (dotted straight line) is described by the empirical relation
The Galactic oxygen star sequence (solid curved line) approximately satisfies an infinite series of the form
There is a hint a secondary sequence of oxygen stars, indicated by the dashed line in Fig. 5b. Later we discuss the stars at this side of the principal sequence in the K- versus J-K diagram in more detail. From now, we assume that there be no substantial differences between the K- versus J-K diagrams for mass-losing AGB stars in the Milky Way and in the LMC.
4.1.2. LMC sample
The K- versus J-K diagram for the stars of the present sample in the direction of the LMC is given in Fig. 6, together with a diagram of the K-band magnitudes versus J-K. The sequences we have drawn in the latter diagram are estimated from the positions of AGB stars and red supergiants in the LMC as compiled by Loup & Groenewegen (1994). LMC red supergiants and foreground stars are predominantly occupying a linear sequence, approximated in Fig. 6a by:
Mass-losing AGB stars in the LMC follow the curved sequence towards red J-K colours, approximated in Fig. 6a:
Using the K- versus J-K diagram, we can classify several stars as being mass-losing AGB stars either on the oxygen star sequence (solid), or on the carbon star sequence (dotted), with the carbon stars to be found exclusively amongst the optically thickest sources. But there are several stars that have too large a 12 µm excess for their J-K colour, or alternatively are too blue in J-K for their 12 µm excess to be on the AGB. Of these, LI-LMC1821 is far too bright in the K-band to be an AGB star in the LMC, and it is probably a foreground star. Another peculiar source is LI-LMC0530, which lies on the Loup & Groenewegen (1994) sequence at a blue J-K colour. We will come back to this source later. The remaining six outliers are all faint in the K-band, and five of them are redder than J-K . Up to three of the stars with lower limits to their J-K colours might be similar to the group of six outliers as well. The field stars are distributed over a larger range of K-band magnitudes, and like the galaxies they are all bluer than J-K . This suggests that the outliers and the field stars are of different nature.
Another way of separating AGB carbon stars from AGB oxygen stars is the J-K versus - colour-colour diagram. For this purpose, we have estimated the carbon star and oxygen star sequences from Le Bertre (1993) and Le Sidaner & Le Bertre (1994). The average carbon star sequence may be approximated by:
The average oxygen star sequence may be approximated by:
The J-K versus - diagram for the stars of our sample is given in Fig. 7, together with diagrams of the K- colours and the 12 µm magnitudes versus -. For the errors on the IRAS flux densities we have adopted formal values of 1 Jy. The J-K versus - diagram does not work well for our sample mainly because the lower limits to the J-K colours of several stars allow these stars to lie either on the oxygen star sequence (solid), or on the carbon star sequence (dotted). The outliers of the K- versus J-K diagram have relatively blue J-K colours. Alternatively they may be characterised by relatively cool dust envelopes because of their relatively red - colours. From the  versus - diagram it appears that the IRAS sources for which no counterpart was found (triangles) may have cooler dust envelopes and/or may experience less severe mass loss, but these indications are very marginal (see also paper II). We also note that if the counterpart would not have been particularly red or bright as compared to the other stars in the field, we may have failed to recognise the counterpart as such. This may be the case for PNe, which can have blue J-K colours between 0 and 0.5 mag due to strong He I line emission at 1.083 µm (Whitelock 1985a).
4.2. Post-AGB star candidates
It is interesting to compare the identified NIR counterparts of the IRAS sources as found in the present study, with the sample from van der Veen et al. (1989: VHG-89). The latter sample consists of a compilation of Galactic objects thought to be in the transition from the AGB to the planetary nebula phase. Their distances and absolute magnitudes are not known, so that we can use this sample in the colour-colour diagrams but not in the colour-magnitude diagrams.
The J-K, K-, and - colours are plotted versus each other in Fig. 8 (circles for the present sample, and dots for the VHG-89 sample). The VHG-89 stars have cool dust envelopes, and have larger 12 µm excesses for their J-K colours than would be reconcilable with mass-losing AGB stars. The six outliers of the present sample roughly overlap with the VHG-89 sample. We classify the six outliers therefore tentatively as post-AGB stars. They would be related to the VHG-89 post-AGB stars with relatively warm dust envelopes, and either have relatively large 12 µm excesses or be optically thick in the K-band. We cannot exclude that one or more of the LMC post-AGB star candidates are actually AGB stars that have recently experienced a thermal pulse (see section Discussion).
4.3. The stellar counterpart of LI-LMC1821
The IRAS point source LI-LMC1821 is identified with a bright NIR star. Our BVRi photometry is presented together with the NIR and IRAS photometry in Table 3. The B-V and V-i colours indicate a non-reddened, early-M type spectrum (Iyengar & Parthasarathy 1997). We plot the spectral energy distribution in Fig. 9 (squares), together with an arbitrarily scaled 3000 K blackbody. We also plotted mean fluxes of the M0 (open circles) and M5 (solid circles) giant from Fluks et al. (1994), scaled to a distance of 1.5 kpc. LI-LMC1821 may be a late-K or early-M giant at a similar distance, and a member of the Galactic halo or thick disk population (Robin et al. 1996). Alternatively, it may be a very luminous red supergiant in the LMC, with a bolometric luminosity of mag. The small IR excess indicates a thin CSE as a result of modest mass loss. The IRAS colours are consistent with a detached CSE (van der Veen & Habing 1988), so that it may be a Galactic post-AGB star, but as the source is not detected at 60 µm this is not conclusive. It is surprising that it has not been listed in any stellar catalogue or survey, despite that it is bright at optical wavelengths. It would be interesting to obtain a radial velocity measurement discriminating between membership of the LMC or the Milky Way.
Table 3. BVRiJK-band magnitudes, and IRAS 12 and 25 µm flux densities of the star LI-LMC1821.
Table 4. Bolometric magnitudes and object classes of the positive identifications in the LMC: oxygen (O), carbon (C), or inconclusive (CO) mass-losing AGB stars, and post-AGB or Thermal Pulse (post/TP) stars.
© European Southern Observatory (ESO) 1997
Online publication: April 28, 1998