Discussion

As it is shown in Table1, the metallicity for IRAS04296 is significantly decreased relative to the solar value: the average abundance for the elements of the iron-group with respect to the Sun is ${\rm [(Ti,V,Cr,Fe)/H]_{\odot}\,=\,-0.9}$ with the standard deviation ${\rm \sigma = 0.2}$ .

Recently Decin et al. (1998), using high resolution spectra and model atmospheres method, calculated abundances of 14 chemical elements in the IRAS04296 atmosphere. Their results are in qualitative agreement with these ones presented here, but there are some significant differences. Decin et al. (1998) calculated chemical composition of this object assuming =7000K, logg=1.0, $\xi_t$ =4km/s, essentially different from the model atmospheres parameters found in this work. It should be noted that we estimated the effective temperature by two independent methods, and it is worth stressing that we have obtained consistent values of the effective temperature: =6300K from numerous FeI, FeII spectral lines and =6500K from modelling of the spectral energy distribution of this source. The difference in effective temperature between Decin et al. (1998) and our estimation ( $\Delta$ =700K) is able to explain different metallicities estimated by Decin et al. (1998) and by us ( ${\rm \Delta\,log\,\epsilon(Fe)\approx\,0.2}$ ). The same is true for the case of the rare-earth element abundances: large differences, about 1dex, in the values could be explained by differences in model atmosphere parameters.

Let us consider now in more detail the peculiarities in the chemical composition of the object. For this purpose, in Table3 we present the logarithmic differences

$\begin{displaymath} {{\rm [X/Fe]_{\odot}\,=\,[log\,\epsilon(X) - log\,\epsilon (Fe)]_{\star} - [log\,\epsilon(X) - log \,\epsilon (Fe)]_{\odot}}}\end{displaymath}$

**Table 3:** Relative abundances of chemical elements for IRAS04296+3429 in comparison to related PPNe. The data by Grevesse and Noels (1993) are adopted for solar abundances.

	IRAS04296+3429	IRAS07134+1005^a	ROA24^b
		-1.00	-1.77
Element	3c
LiI
CI	+0.84	+1.08	+0.67
NI	+0.83	+1.03	+1.02
OI	+0.19	+0.63	+1.01
NaI	+0.42	+0.54	+0.71
MgI		+0.97	+0.31
MgII	+1.34		+0.09
AlI	+1.03	+1.48
SiI	+0.58	+0.95	+0.80
SiII	+0.26		+1.03
SI	+0.43	+0.63
CaI	+0.19	+0.45	+0.60
ScII	+0.18	-0.07	-0.13
TiII	-0.27		+0.33
VII	+0.10	-0.03	+0.15
CrII	+0.11		+0.65
CuI	+0.24	+1.03	-0.01
ZnI	+0.08
YII	+1.20	+1.70	+0.37
ZrII	+0.62
BaII	+2.49	+0.99	+0.96
LaII	+1.17	+1.59	+0.54
CeII	+0.82		+1.60
PrII	+0.74
NdII	+1.07	+1.30	+0.67
EuII	+0.34	+1.06	+0.25
4la - Klochkova (1995),
4lb - Gonzalez and Wallerstein (1992).

The carbon overabundance (revealed from intensities of 21 absorption lines with the standard deviation ) and the enhancement of nitrogen (from 4 lines, ) suggest that IRAS04296 underwent the third dredge-up episode.

The oxygen content based on intensity of 3 weak lines near 6155Å is determined with a small internal error.

From the Fe-deficiency and CNO abundances () we can conclude that IRAS04296 is a low mass object in advanced stage of evolution. For an unevolved metal-deficient object (with ) the average value of is only about -0.2 (Tomkin et al. 1995), the average value of is (Wheeler et al. 1989, Timmes et al. 1995) and the average value of is (Wheeler et al. 1989, Timmes et al. 1995, Klochkova & Panchuk 1996b). The atmospheres of the post-AGB stars IRAS07134 and ROA24 are also overabundant in both carbon and nitrogen. Note however, that for most of the PPNe candidates studied, strong relative changes between elements of the CNO-group are observed (Luck et al. 1983; Lambert et al. 1988; Klochkova 1995; Zacs et al. 1995, 1996; van Winckel et al. 1996a, 1996b; van Winckel 1997).

The abundances of some light metals (Na, Al, Mg, Si, Ca) are enhanced for all three stars. The average value for these elements is for IRAS04296; +0.9 for IRAS07134 and +0.6 for ROA24, with the standard deviations: , 0.4 and 0.36, respectively.

We did not still include the KI abundance into our results, since we suspect that the equivalent width of its line near ${\rm \lambda}$ 7699Å could be significantly distorted due to circumstellar and interstellar components.

The iron-group element zinc is the most important for determination of real (initial) value of the metallicity of a star since, firstly, its abundance follows that of iron in a wide [Fe/H] interval (Sneden & Crocker 1988; Wheeler et al. 1989, Sneden et al 1991) and, secondly, zinc having a low condensation temperature is not depleted by selective separation processes onto dust grains (Bond 1992). A close to solar abundance of Zn relative to iron () permits us to conclude about the inefficiency of the selective separation processes in the IRAS04296 envelope. This conclusion is based also on an absence of overdeficiency of light depleted elements (Ca, Sc). Besides, the relative abundance ( with the standard deviation ) of S, a chemical element which is not depleted by dust-gas separation, for IRAS04296 is close to the value for unevolved metal-deficient dwarfs (François 1987, Timmes et al 1995). This futher confirms the lack of selective separation in the envelope of the object studied.

Individual abundances of the heavy s-process metals Y and Zr are determined with a relatively large error because of the small number of lines measured. However, the average value for Y and Zr is sufficiently reliable. In addition, the abundance of heavy s-process element Ba () derived from the equivalent width of strong lines could be altered by a systematic error due to the complexity of the outer regions of the stellar atmopshere as discussed above. Nevertheless, we conclude that there is a Ba excess.

The abundance of lanthanides (La, Ce, Pr, Nd) are strong enhanced relative to iron for the objects from Table3. For these heavy metals the average value is for IRAS04296, IRAS07134 and ROA24, respectively, with the standard deviations 0.2 and 0.6 for IRAS04296 and ROA24. Moreover, for all these objects we see the overabundance of Eu which is predominantly produced by the r-process.

Excess of s-process elements has been reliably found up to now in three objects investigated at the 6m telescope: IRAS04296+3429, IRAS07134+1005 and IRAS22272+5435. Besides, similar conclusions have appeared for another four PPN candidates (and for one object in common): HD158616 (van Winckel et al. 1995); IRAS19500-1709=HD187885 (van Winckel 1997); IRAS05341+0852 (Reddy et al. 1997); IRAS22223+4327 and IRAS04296+3429 (Decin et al. 1998). In atmospheres of most PPN candidates overdeficiency (with respect to their metallicity) of heavy nuclei is generally observed (Klochkova 1995; van Winckel et al. 1996a, 1996b; Klochkova and Panchuk 1996a; van Winckel 1997), whose existence in the atmospheres of post-AGB low-mass supergiants has not yet found a clear explanation.

In consequence, we can state that chemical abundances pattern for the source IRAS04296 is related to its old galactic population membership and dredge-up of matter enriched by the nucleosynthesis products. It may be part of the old disk population.

As has been concluded already by Decin et al. (1998) all the post-AGB candidates mentioned above (only these, up to now, show an s-process element enhancement!) belong to the small group of PPNe (Kwok et al. 1989; Kwok et al. 1995) which have in their IR spectrum an unidentified emission band at about ${\rm 21\,\mu}$ m. This feature is neither found in the spectra of their predecessors, AGB stars, nor in the spectra of PNe. Note, once more, that the search by means of the ISO for the new 21 ${\mu}$ m emitters among candidates selected by Henning et al. (1996) failed to find any new sources with the feature (Henning, private communication). As has been stated in the papers by Kwok et al. (1989, 1995), the objects whose spectra contain the ${\rm 21\,\mu}$ m band are carbon-rich stars. Our investigations based on the spectra from the 6m telescope, for IRAS07134 (Klochkova 1995), IRAS22272+5435 (Zacs et al. 1995) and IRAS04296 (Klochkova et al. 1997b), confirmed that for all of them. In this context, the conclusion that the carrier of the 21 ${\mu}$ m band is related to C is natural. For example, Buss et al. (1990) have supposed that this feature may be caused by polycyclic aromatic hydrocarbons. On the other hand, Goebel (1993) has identified the ${\rm 21\,\mu}$ m band with the vibrational band of the SiS₂ molecule, the presence of which is consistent with the temperature in the envelope.

Taking into account the available results on chemical composition for subclass of PPNe with the 21 ${\mu}$ m feature: IRAS07134+1005 (Parthasarathy et al. 1992, Klochkova 1995), IRAS22272+5435 (Zacs et al. 1995), IRAS19500-1709 (van Winckel et al. 1996a), IRAS05341+0852 (Reddy et al. 1997), IRAS22223+4327 (Decin et al. 1998), and IRAS04296 (Decin et al. 1998; Klochkova et al. 1997b; this paper) we see that the carbon-rich atmospheres of these objects are also enriched by s-process elements. It is evident that there is a strong correlation between presence of the 21 ${\mu}$ m feature, ${\rm C_2}$ , ${\rm C_3}$ molecular bands, and excess of the s-process elements. Decin et al. (1998) were the first who pointed out this relationship. What is even more important, an excess of s-process elements was not found for a number of IRAS sources with altered CNO-content but without the 21 ${\mu}$ m feature (some of which are oxygen-rich stars rather than carbon-rich stars): IRAS06338+5333 (Luck & Bond 1984; Bond & Luck 1987), IRAS07331+0021 (Luck & Bond 1989; Klochkova & Panchuk 1996a), IRAS09276+4454 (Klochkova & Mishenina 1998), IRAS12175-5338 (van Winckel 1997), IRAS 12538-2611 (Luck et al. 1983; Klochkova & Panchuk 1988b; Giridhar et al. 1997), IRAS15039-4806 (van Winckel et al. 1996b), IRAS17436+5003 (Klochkova & Panchuk 1988a; Luck et al. 1990; Klochkova 1998), IRAS18095+2704 (Klochkova 1995), and IRAS19114+0002 (Zacs et al. 1996, Klochkova 1998). Therefore, it seems that carrier of 21 ${\mu}$ m feature is strongly related to the whole chemical composition pattern typical for the third dredge-up (excess of s-process elements), and not only to the C-richness of the photosphere.

That 21 ${\mu}$ m feature is not observed around AGB-stars showing s-process elements could be explained by the physical conditions which are inappropriate for the excitation of this band, while its non-presence in planetary nebulae may be a result of carrier destruction by the highly energetic photons.