Astron. Astrophys. 323, 461-468 (1997)

5. Discussion

In Table 3, we have gathered the C, N, O and iron abundance from all the available high resolution analyses in the Magellanic Clouds, for F supergiants (LL92, RB89, Spite et al. (1989b) and SBS89, Spite et al. 1993, Hill et al. 1995), B stars (Rolleston et al. 1993), together with the mean abundance for the Supernovae Remnants (Russell and Dopita 1990) and for the H II regions (Dufour 1984; Garnett et al. 1995). It should be noted that the more recent abundance determination of the Clouds' H II regions performed by Russell and Dopita (1990) gave results within 0.1 dex relative to the Dufour (1984) review. Moreover, from HST observations, Garnett et al. (1995) also found values of C and O compatible with Dufour. The values for Canopus (Luck & Lambert 1985) and B stars in Orion (Cunha & Lambert 1994) are also reported as Galactic references. We have corrected the abundances (relative to the Sun) in this table, to the presently accepted solar value from Grevesse et al. (1996).

Table 3. CNO abundances derived in supergiants from high resolution work available in the literature, and for reference objects (H II regions, supernovae remnants, Sun, Canopus and Orion)

5.1. Oxygen abundances

Before discussing the oxygen abundances in the Magellanic Clouds, it is worth recalling that oxygen is an element mostly produced in high mass Type II supernovae (SN II), while iron is thought to be most efficiently produced in Type Ia supernovae (SN I), exploding from longer-lived lower-mass binary progenitors. A mean underabundance of -0.18 0.10 dex relative to iron is found for our sample stars. In Fig. 6, we plot the oxygen-to-iron ratio against the iron abundance for the SMC and the LMC supergiants (Hill et al. 1995). The location of Galactic supergiants (hatched area) and their mean value, with the abundance of the B stars in the Orion Nebulae are also indicated.

Fig. 6. [O/Fe] versus [Fe/H]. For the Galaxy: Sun and Canopus (asterisks) and a mean for Orion B stars from Cunha & Lambert (1994) (open star), where the vertical line shows the r.m.s. For the SMC: K supergiants (filled triangles; this work), F supergiants (filled squares; SBS89). For the LMC: K supergiants (open triangles; Spite et al. 1993), F supergiants (open squares; Hill et al. 1995). The mean for SMC (filled circle) and LMC (open circles) SNR (Russell & Dopita 1990) and LMC type I planetary nebulae (Freitas Pacheco et al. 1993) (open diamond) are also displayed. The chemical evolution models by RD92 are shown for the Galaxy (dashed line), the LMC and the SMC (solid lines).

All the supergiants in our Galaxy and in the Magellanic Clouds are massive objects which have a very short life-time, and thus their atmospheres a representative of the young material of the galaxies, as the H II regions.

The most striking feature of Fig. 6 is the uniformly low [O/Fe] ratio in the young objects of the three galaxies: an oxygen-to-iron overdeficiency of -0.3 to -0.2 dex is found in Galactic supergiants and ISM of the solar neighbourhood. Therefore, the present day -0.18 dex (SMC) and -0.15 dex (LMC) oxygen overdeficiencies are similar to what is observed in similar objects in the Galaxy. In this picture, there is no need for an IMF different in the Clouds and the Galaxy: the Magellanic Clouds could have had a continuous star formation, but with a lower rate (per unit gas mass),than in our Galaxy. This picture would also be compatible with a formation rate occurring in bursts: to discriminate between star formation occurring continuously or in bursts, the only probes lie in the past. At the time when the burst(s) occur, the [O/Fe] must have changed very rapidly upon very small change of [Fe/H] (Tsujimoto et al. 1995): the observed [O/Fe] should therefore be very dispersed for a given [Fe/H] value (corresponding to the time of burst).

Since, magnesium, like oxygen, is belived to be produced in massive SN II, the [Mg/Fe] ratio is expected to be about the same in the young objects of the SMC, the LMC and the Galaxy, as is indeed observed. Let us remark that in this picture, there is no need for a different IMF in the Clouds and the Galaxy.

The [O/Fe] ratios measured in the SMC and LMC (see also Barbuy et al. 1994) are also in agreement with more complex scenarios of chemical evolution in the Clouds (Russell & Dopita 1992 hereafter RD92; Tsujimoto et al. 1995). This is not very surprising since these models were fitted to the (O) for the H II regions of each Cloud: it only reflects the good agreement between the stars and H II regions.

To achieve the low metallicity of the young material in the Clouds, the RD92 models assume that the formation of stars in the Clouds began later than in our own Galaxy (about 8 Gyrs ago instead of 15 Gyrs in our Galaxy). The young material in the Clouds should thus be similar to material as it was 7 Gyrs ago in our Galaxy. Since, it is well known that [O/Fe] decreases with time (Edvardsson et al. 1993), we should expect a higher [O/Fe] ratio in MC's young objects than in their Galactic counterparts. From Fig. 6, no difference is observed. To achieve the low observed [O/Fe] ratio, the RD92 models used an IMF steeper in the Clouds than in the Galaxy (exponents of the power laws of respectively 1.8, 2.2, and 2.35 in the Galaxy, LMC and SMC).

However, other -elements such as Mg, Si and Ca (which are also produced efficiently in massive SN II) do not show the same over-deficiency with respect to iron (Paper I; Hill et al. 1995), and these steep-IMF models would have problems to explain.

5.2. Convective mixing effects on C and N abundances

Convective mixing brings CNO-processed material to the outer atmospheric layers during stellar evolution along the red giant branch. Such process can be detected through carbon deficiencies accompanied by nitrogen enhancements (the effect on oxygen is negligible or less pronounced, since the ON-cycle occurs in deeper layers relative to the CN-process). We find a mean carbon to iron deficiency of =-0.30 0.07 dex compatible with mixing effects expected in such stars. The nitrogen enhancement, however is mild ( =0.22 0.12 dex). Low values of ¹² C/¹³ C=10-20, determined for three sample stars confirm that convective mixing has occurred in these stars.

Our mean (C) = 7.54 is lower than the mean value from high resolution work for 7 F supergiants and 3 B supergiants (cf. Table 3) of (C) = 7.7 and the values found from low resolution spectra of (C) = 7.93 for 3 stars by Thévenin & Jasniewicz (1992), and of (C) = 7.85 for 40 K supergiants by Meliani et al. (1995). Therefore we find a value closer to the one by Dufour (1984) of (C) = 7.16, although not quite as low. A comparison with the recent HST data by Garnett et al. (1995) shows that log(C/O) for our stars is systematically higher by around 0.3 dex (see Table 2 and 3). The question of the carbon abundance in the Small Cloud thus appears to be still open.

A more secure way to consider the carbon and nitrogen is through the C+N abundance, which is only marginally dependent on the C₂ feature and mostly determined by the fit of the reliable CN feature: an overestimation of the carbon abundance by 0.2 dex leads (by fitting the CN feature) to an underestimation of the nitrogen abundance by 0.25 dex, but the (C+N) abundance is then only overestimated by 0.04 dex; under such circumstances, the C/N ratio would be overestimated by a factor of 3.

Our results constitute the first high resolution CNO derivation of field (cool) K supergiants in the SMC. LL92 give both C and N abundances for only two supergiants (non-Cepheid) (AzV121 and AzV369) showing a mean value of C/N = 0.65. The mean value for our sample is C/N = 1.27 0.28, clearly below the solar value, but not as mixed as the two LL92 stars.

Our mean = -0.15 0.08 dex, on the other hand, is close to the solar ratio, whereas for AzV121 and AzV369 there seems to be an excess of C+N ([(C+N)/Fe] = 0.31), arising from the very strong nitrogen abundance found by LL92 for these stars. Such a large difference is not easily understandable, even if the mixing is larger in these two stars. Since the lines used in the analysis are different, there could be a systematic effect in the derivations by LL92 and/or by us, and particularly for nitrogen.

In Fig. 7 we show (C+N) versus [Fe/H], or versus [O/H] for the H II regions and B stars, for the data reported in Tables 2 and 3. The solid line indicating [(C+N)/Fe] = [Fe/H] represents approximately the behaviour of dwarf stars in our Galaxy. The C+N overdeficiency in the SMC is of the same order as that found in our Galaxy between the Sun and the solar neighbourhood young objects such as main sequence B stars in Orion and supergiants. The H II regions abundances in the Galaxy are consistent with that of these young objects, whereas the H II regions in both Clouds show a strong overdeficiency of C+N relative to supergiants. In fact, Garnett et al. (1995) have found that C/O in the H II regions of metal-poor dwarf irregular galaxies (including the SMC) are low; could this be an indication that the Clouds are indeed depleted in carbon, or maybe locked into grains ?

Fig. 7. [(C+N)/H] versus [Fe/H] (or versus [O/H] for the H II regions and the B stars). For the SMC: K supergiants (filled triangles; this work), F supergiants (filled squares; LL92) and B stars (filled stars). For the LMC: K supergiants (open triangles; Spite et al. 1993), F supergiants (open triangles; Hill et al. 1995, and open squares; RB89, LL92). The mean for SMC, LMC and the Galaxy H II regions (Dufour 1984) are also displayed (crosses). The solid line indicates [(C+N)/Fe] = [Fe/H].

5.3. Lithium

Lithium is detectable in all the program stars, and its abundance ranges from (Li)=0.0 to 0.6 dex. In fact, two of our stars (PMMR 27 and PMMR 145) display a strong lithium abundance ( (Li)= 0.6 dex), while the four others show milder abundances ( (Li) 0.0 dex). In other supergiant stars of the Magellanic Clouds, very few results are available; LL92 obtained only upper limits for lithium abundance owing to the faintness of the line in the hotter F supergiants. Previous determinations in Magellanic K supergiants only concern three stars: in the LMC NGC1948:WBT 542 and NGC1818:B12 and in the SMC NGC330 A7 (Spite et al 1993; Richtler et al. 1989; Spite et al. 1986). We have recomputed the lithium abundance in these stars using the present line list and the atmospheric parameters from the above papers, and we found values of respectively 0.0, 0.3 and 0.0. Smith and Lambert (1990) observed very strong lithium lines for Magellanic M stars in the AGB phase, and only in a limited range of luminosity Therefore, for stars such as our supergiants which are not in the AGB phase, low lithium is expected.

Of course, the lithium that we observe is the original lithium abundance of the star, strongly diluted by convective mixing with the deep layers of the star, where the lithium has been destroyed. In our Galaxy, the massive ( 9 ) K supergiants analysed by Luck (1977) show lithium abundances in the range -0.6 to +1.0 dex (mean value (Li)=0.12 dex) for all the stars with 4500 K. If the initial Li abundance of these stars was the standard abundance of the young Pop I, the dilution would be -2.3 to -3.9 dex (including non-LTE corrections). The theoretical calculations for dilution of Li in the convective zone of massive stars have not made much progress in the recent years, and we must therefore recall Iben (1966) for an estimation of it. As advocated by Spite et al. 1986, this dilution and the possible non-LTE effects in the Li line brings the lithium abundance of the star up by +1.8  0.3 dex, leaving us with values of ranging from 1.8 to 2.4 dex for the stars in our sample. However, such a value is largely uncertain. If the dilution is similar to the dilution found in Luck's supergiants, the initial lithium abundance would be 2.9 to 3.9 dex.

© European Southern Observatory (ESO) 1997

Online publication: June 5, 1998

helpdesk.link@springer.de