4. Results of the chemical analyses
In the Tables 2 - 5 we give the results of the abundance analysis for each programme star. For every ion we list the number of lines used, the mean equivalent width, the absolute abundance, the abundance ratio relative to the solar value and the internal scatter, if more than one line is used. A complete line list with the detailed atomic data can be obtained upon request. For the solar iron abundance we used the meteoric iron abundance of 7.51.
Table 2. Chemical analysis of HR 1865. In the last column we listed the difference with the analysis of Venn (1995b).
4.1. HR 1865 and HR 1017
Although our primary goal was to use the spectra of the narrow-lined supergiants HR 1865 and HR 1017 for line-identification purposes and to recognize possible blending in the spectra of the IRAS objects, we also performed a complete LTE abundance analysis of these two bright objects.
The non-variable HR 1865 is a very bright Galactic F-type supergiant which is often used as a standard star for Galactic and extragalactic research. The most recent extensive chemical study is by Venn (1995a, b). She deduced model-parameters of = 7400 K and log g = 1.1 and = 4.0 km s-1 which are similar to our findings. The difference between the two studies is given in Table 2. Generally the agreement is rather good with the noticeable exception of Mg I, S I and Sc II where the difference amounts to 0.26, 0.22 and 0.23 dex respectively. There are no Sc II and Mg I lines common in our analyses. The common S I lines (3 in total) have very similar equivalent widths, so the difference in results is due to slightly different atomic parameters and microturbulent velocity.
The most recent analysis of HR 1017 we found in the literature is by Luck & Lambert (1985). The atmospheric parameters they found are somewhat cooler than ours: = 6250 K, log g = 0.90 and = 3.0 km s-1. Their iron abundance is somewhat higher, but generally the agreement between the abundances is good.
4.1.1. Intermediate and heavy elements (Na-Zr)
Except for Na, the other elements do not seem to be enhanced (see Fig. 3). This corresponds to the abundance analysis of Venn and Luck & Lambert.
Both HR 1865 and HR 1017 display a strong sodium enrichment ([Na/H]=+0.39 and +0.68). Sodium enrichment in F-K supergiants was first reported by Luck (1977, 1978). Since then several authors have detected an enhancement of Na (see Venn 1995b and references therein). Boyarchuk & Lyubimkov (1983) proposed two possibilities for the Na-enrichment: 1) non-LTE effects in the tenuous and cool atmosphere, of which the corrections in the abundances are -0.1 dex for weak lines 2) during a NeNa proton capture reaction (22 Ne(p, )23 Na) Na could have been synthesized and after deep mixing with the interior (first dredge-up), Na-enrichment may occur in the stellar atmosphere (Venn 1995b). The Na-synthesis can only take place when the temperature is high enough. Therefore Na-enrichment is mass dependent: the more massive the star, the stronger Na will be enhanced. Takeda & Takada-Hidai (1994) calculated that HR 1865 is a less massive star than HR 1017 ( =17 and =14.5 ); the Na enrichment displays the same trend.
Interestingly in an LTE analysis of several high-luminosity LMC/SMC Cepheids, Hill et al. (1995) did not found a general Na enhancement. Since the same temperatures and densities apply for the LMC/SMC Cepheid models, there is a hint that the Na enhancement of the Galactic Cepheids is real and not due to non-LTE effects. Detailed evolutionary models coupled with accurate nucleosynthetic networks are, however, needed in order to quantify the dependance of the nucleosynthetic yields and mixing ratios on the overall chemical composition, or other as yet unexplained difference in the LMC/SMC and Galactic supergiants.
4.2. IRAS 22223+4327 and IRAS 04296+3429
Both IRAS 22223+4327 and IRAS 04296+3429 are iron deficient, [Fe/H]=-0.4 and -0.7 respectively. Together with the value of their galactic latitude (b=- for IRAS 22223+4327 and b=- for IRAS 04296+3429) this indicates that the IRAS-sources are low-mass objects of a relatively old population. The radial velocity deduced from the spectra are and for respectively IRAS 22223+4327 and IRAS 04296+3429 which are similar to values found in the literature (Omont et al. 1993). So far there is thus no evidence for radial velocity variations and binarity of the programme stars.
4.2.2. 3rd dredge-up
Both stars display a strong enrichment of carbon ([C/Fe] = +0.5 and +0.9), an indication that the third dredge-up was effective. During this third dredge-up material out of the helium-burning shell is mixed into the stellar photosphere. Unfortunately, the O abundance is difficult to compute in this temperature-gravity domain. The only lines available are the O triplet at 6150 Å , but these are heavily blended with a Fe I and a Si I line. We only could determine the O abundance for the bright IRAS 22223+4327 and we obtain a C/O ratio of 1.3, but due to the large uncertainty of the O abundance this ratio is not very accurate. From the atomic photospheric lines alone the C/O ratio cannot be determined accurately enough to claim the objects to be real carbon stars (C/O 1). The sum of the CNO nuclei is therefore also rather uncertain and compared to the expected value for an unevolved star of the same metallicity we find only a CNO of +0.05.
The most convincing argument for mixing products of the helium-burning shell into the atmosphere are the large s-process elemental abundances. For IRAS 22223+4327 and IRAS 04296+3429 we could derive abundances of 9 respectively 7 s-process elements and all turned out to be significantly overabundant, even relative to the solar value. We will focus on the distribution of these elements in a separate section.
For metal deficient stars the abundances of s-process-elements scale with Fe for -1.5 [Fe/H] 0.0 (see Wheeler 1989). For IRAS 22223+4327 and IRAS 04296+3429 [s/Fe] is 1.0 and 1.4! There is thus no doubt that both IRAS stars are post 3rd dredge-up stars.
4.3. Intermediate-mass elements (Al-Ni)
For unevolved metal deficient stars a slight overabundance of the -elements reflects the chemical history of our Galaxy. The solar neighbourhood displays an increase of the [ /Fe] ratio from 0.0 to +0.4 in the metallicity region from about solar to [Fe/H]=-1.0 (Edvardsson et al. 1993). When we take into account the error bars in the -abundances, the overabundances of Si, S and Ni in IRAS 04296+3429 and Si and Ti IRAS 22223+4327 do not seem to be abnormal. We may conclude that the 3rd dredge-up did not enhance significantly the isotopes.
Also the odd elements are interesting tracers of internal nucleosynthesis and structure. Theoretical models predict an enrichment of Al and especially the Al/Mg ratio in stars where Hot Bottom Burning (HBB) has taken place (Lattanzio et al. 1996). In the deepest layers of the convective envelope the temperatures may reach about 82 million K, and substantial hydrogen burning via the CNO cycle will take place. Further, at these high temperatures 26 Mg suffers substantial proton captures and produces 27 Al during the interpulse phase. In theoretical evolutionary models, HBB is predicted only in intermediate mass stars, since the temperature of the bottom of the convective envelope only reaches in these models high enough values for the synthesis to take place. Although the high C abundance already indicates that HBB was not very effective in the IRAS objects, we carefully analysed the Al lines in our spectra of IRAS 22223+4327. The strongest optical lines at 8773 Å (multiplet number 9) are unfortunately heavily blended with the Phillips (2,0) band of the circumstellar C2. We therefore based our analysis of the lines of multiplet 10 (7835.3 and 7836.1 Å) and multiplet 5 (6696.0 and 6698.7 Å). The small overabundance of [Al/Fe] = 0.2 is not significant. Also the 7 Li resonance line at 6707 Å is not detected, so we conclude that there is no evidence for even moderate HBB in IRAS 22223+4327. Unfortunately, there are no useful Mg lines in our spectra.
4.4. s-proces element distribution
Several theoretical studies of large nuclear reaction networks coupled with accurate AGB evolutionary codes exist (Malaney 1987a, b; Busso et al. 1992, 1995) that enable us to characterize the s-process in the IRAS sources based on the photospheric s-process element distribution. The direct physical information on the efficiency of the internal nucleosythesis that can be deduced from the s-process distribution of an individual object, is unfortunately limited since the predicted photospheric distribution is not only determined by the nucleosythesis itself, but also by the theoretically less understood dredge-up process (Busso et al. 1995). The parameters governing the outcome of such chemical evolutionary models do therefore not only consist of nucleosynthetic quantities, but also of more ad-hoc adopted values governing the stellar evolution like mass-loss, dredged-up mass, frequency of dredge-ups etc. (Busso et al. 1992).
The ratio of light (Sr, Y, Zr) to heavy (Ba, La, Nd and Sm) s-elements gives a measure of the neutron exposure rate which is defined as where is the neutron density, V the relative velocity of the neutrons and the seed nuclei. The pulsed neutron irradiation during the AGB evolution is parameterized by defining a mean neutron exposure defined a with the neutron exposure rate of a particular pulse and r the overlap factor which is the fraction of the inter-shell material that remains in the neutron-exposed region (Ulrich 1973). In most calculations, the neutron density and the overlap factor are adopted as constants and the efficiency of s-processing is indicated by one parameter .
For IRAS 22223+4327 we obtained the most complete s-process distribution. In the temperature-gravity domain of the moderately deficient programme stars, the Sr abundance is extremely difficult to measure due to the lack of weak lines. We therefore did not take this element into account to determine the [ls/Fe] ratio but following Busso et al. (1995) no correction factor is needed to account for unobserved elements from the light s-process trio. The mean abundance of the light elements Y and Zr is [ls/Fe] = +1.5 while for the heavy elements Ba, La, Nd and Sm [hs/Fe] = +0.9, hence the [hs/ls] = -0.6. Following Fig. 6 of Busso et al. (1995) the mean neutron exposure of the object can be estimated to be in between 0.2 and 0.25 .
IRAS 22223+4327 is located in this diagram in the locus of the Carbon stars and the high value of [ls/Fe] indicates a high mixing ratio between dredge-up material and residual mass of the envelope (Busso et al. 1995). This is not surprising since the post-AGB star is Carbon rich.
More surprising is, however, the low value of the mean neutron exposure of the object given its low metallicity. Indeed, several observational evidences exist that the neutron exposure increases with lower metallicity. In Fig. 1 of Busso et al. (1995), where they show the measured [hs/ls] values as a function of the iron abundance [Fe/H] of a sample of intrinsic and extrinsic S stars and Ba stars, a clear correlation is seen with an increase of 0.2 dex in [hs/ls] for a drop in metallicity from 0 to -0.5. While the mean trend of the Ba-stars indicate that an object with metallicity of -0.5, should have a [hs/ls] of 0.2, this ratio is only -0.6 for IRAS 22223+4327! Since the metallicity of Carbon stars is difficult to measure, no observational material is present for comparison.
In order to characterize more precisely the s-process efficiency, we made use of the abundance tables for s-process nucleosynthesis of Malaney (1987b). The tables list element abundance enhancements for exponential distribution of neutron exposures for different values using fixed neutron densities of and . To quantify the comparison between the observations and models we used the goodness-of-fit procedure as defined by Cowley & Downs (1980) and often used in the literature. For a description of the method we refer to Smith et al. (1997). In the comparison, we used the abundances of all the measured species: Y (Z=39), Zr(40), Ba(56), La(57), Ce(58), Pr(59), Nd(60), Sm(62) and Eu(63). For the uncertainty on the measured abundances we added quadratically a fixed value of 0.3 to the internal accuracy given in Table 5, this to account for uncertainties on the model atmosphere and uncertain values. Moreover, for species with less than 5 useful lines measured, we adopt an internal accuracy of 0.2 dex. We corrected the observed number densities for the initial abundance by adopting [s-process/Fe] = 0 and a metallicity of -0.5, but since the overabundance is large, this correction only marginally influences the result. For Eu, which is primarily an r-process element with a different chemical history than the s-process elements, we used the results of Woolf et al. (1995) to estimate the initial abundance. In this method, the quality of the fit is given by the quantity , with a lower value expressing a better fit. In Fig. 5 we plot our corrected abundance distribution together with Malaney's (1987b) model predictions for an exponential neutron exposure with of () with an value of 0.6. For higher neutron efficiencies, the parameter increases fast (e.g. for = 0.3, is already more than tripled) and for the = 0.1 model we have = 3.7. Also the model with = 0.05 gives a good fit but the absolute enhancement of the s-process elements is, in this model, predicted to be too low in comparison with the observed values. We can conclude that the distribution of s-process elements in IRAS 22223+4327 points to a low s-process efficiency characterized by a .
The s-process distribution of IRAS 04296+3429 also points to a low s-process efficiency with the best fit again obtained for the models with . In this object, only 7 elements were taken into account in the fitting.
© European Southern Observatory (ESO) 1998
Online publication: March 30, 1998