5.1. The lithium abundance
In main-sequence turnoff stars the lithium lines are always weak and difficult to measure. All the more so when the spectrum of a component is veiled by the underlying flux of a companion star, as it the case for CS 22873-139. In an effort to increase the signal to noise ratio, we co-added the three EMMI spectra of the star in the lithium region although the distance of the lines of the two components is different. First we co-added the spectra centered on the line of the A component (to measure the lithium line of the A component), and second we co-added the spectra centered on the line of the B component (to measure the lithium line of the B component). In neither of these cases could the lithium line be detected. However, if the models 6300/5750 are adopted the upper limit of the lithium line remains compatible with a lithium abundance log (Li) dex (Fig. 6). It happens that this lithium abundance is in agreement with the trend predicted by Ryan et al. (1996), but this agreement would have a real meaning only if our temperature determination was in the same scale as the one used for defining this trend (our temperature is derived from the line fitting). If the models 6760/6160 are adopted, the lithium profile would be even compatible with a lithium abundance up to log (Li) dex.
Let us note that Thorburn (1994), finds an equivalent width of 17 mA, from the measure of the line (on a spectrum with S/N = 75), amplified by enhancement through an estimated dilution factor. The two components are assumed to be near the turnoff, a temperature of 6423 K is estimated (exactly the same as the one provided by Schuster et al. 1996), resulting in a lithium abundance log(Li) = 2.28 dex. Ryan et al. (1996a) adopt a similar temperature, and derive a lithium abundance log (Li) = 2.15 dex.
5.2. Abundance anomalies
In Table 6, two abundance anomalies may be recognized. The most striking is the very low abundance of Sr (Fig. 7). In main-sequence turn-off stars, the strong resonance lines of Sr II at 4077 and 4215 Å usually persist even in very metal-poor stars (see Fig. 11 of Ryan et al. 1991). The second anomaly is the low abundance of the elements Mg and Ca, (but an overabundance of Ti), which is not the usual behavior in most metal-poor stars.
A few extremely metal-poor stars ([Fe/H]) are known with a very low abundance of strontium (Molaro & Bonifacio 1990, Ryan et al. 1991, Norris et al. 1993, Primas et al. 1994, McWilliam et al. 1995a, 1995b and Ryan et al. 1996b). Abundance ratios for a number of elements in these stars are given in Table 7, where we list the stars with . The stars CS 22885-096 and CS 22968-014 have been studied by three different authors. The agreement between the different determinations is rather good for CS 22885-096, but for CS 22968-014 the spread of the element-ratios from author to author is large, although the models used are very close. To improve the quality of the abundances we have averaged the three lists of published equivalent widths, and we re-computed the abundances, in a homogeneous manner, directly from the equivalent widths. We adopted , log and (Primas et al. 1994). This star is a cool giant (4900 K) and thus the equivalent widths of the magnesium lines are about 100m Å and the magnesium abundance is very sensitive to a small variation of the equivalent widths. On the other hand, the Ti II lines are all weak and a good signal to noise is important to measure weak lines. The result of these computations is given in Table 7.
Table 7. Relative abundances of the elements in CS 22873-139 compared to the seven other extremely metal-poor stars with very low strontium abundance . For CS22885-096 the last line gives the mean value of the abundances computed by the different authors. For CS 22968-014 where the results were rather discordant, the equivalent widths have been averaged and the abundances re-computed in an homogeneous way.
All the stars known to exhibit a very low strontium abundance are extremely metal-poor. Ryan et al. 1991 remark that their Sr-weak stars are among the most metal-poor stars of their sample and that, as a consequence, a possible explanation of their Sr deficiency is that they have been formed out of material issued from stars initially deprived of iron (and of course of Sr), i.e. from Pop III stars (first generation stars). The first generation of stars, when becoming SN II, would have synthesized iron-peak elements which could be transformed into heavy elements like Sr only in a second generation of (massive) stars. Our binary star, like the seven other Sr-weak stars in Table 7, could thus be a second generation (low-mass) object which did not initially possess any secondary elements, and for which even the amount of primary heavy elements is very small. Also, Cayrel (1996) has noted a global correlation between the abundance of Sr and the metallicity in very metal-poor stars.
If we suppose that CS 22873-139 is a second generation star, we could expect that the material from which it has been formed has been enriched only by SN IIe, and probably, only by a few SN IIe (or even by a single one) and it is important to compare the detailed abundances with those predicted by recent nucleosynthesis computations.
In Fig. 8 we compare the abundances (relative to iron) to the solar abundances, in the seven Sr-poor stars of Table 7 and in CS 22873-139. In the Sr-poor stars,the elements Mg, Ca, and Ti are generally overabundant relative to iron as it is normally observed in the Pop II stars. In CS 22873-139. [Mg/Fe] and [Ca/Fe] are close to zero, while Ti is enhanced. Titanium is not a typical -process element (contrary to Mg and Ca), and indeed may also be formed by the e-process (e.g. Wallerstein et al. 1997).
The abundance ratios relative to iron in CS 22873-139 are not very different from CS 22968-014, which presents only a small overabundance of Mg and Ca relative to iron, but the ratio [Ti/Fe] is (probably) significantly different.
Some kind of similarity in the abundance pattern seems to appear in Fig. 8, blurred by variations of Fe abundance, and it was hoped that, by cancelling these variations of Fe, the similarity would show up: in Fig. 9, the ratios were normalized to Mg (rather than to Fe), but the spread of abundances for Ca is not reduced, and the Ti spread is worse. Adding or not adding some Fe, does not solve the lack of similarity.
An explanation of a low ratio -elements/Fe could be the addition of Fe by a SN Ia but the very low iron abundance in the stars considered here is against this hypothesis, as well as the fact that the spread in the abundances Mg/Fe, Ca/Fe and Ti/Fe does not show the signature of a simple iron addition, as noted before.
It is also interesting to compare the abundances of CS 22873-139 to the abundances of CS 22876-32 another extremely metal-poor binary star recently studied by Norris et al. (2000) with [Fe/H]. In this star the [Sr/Fe] is and thus it does not appear in Table 7 (stars with [Fe/H] and [Sr/Fe]), but the strontium abundance is certainly rather low in this star. However CS 22876-32 unlike CS 22873-139, is magnesium rich: [Mg/Fe]=+0.50. It is only moderately Ca and Ti rich: [Ca/Fe]=+0.13 and [Ti/Fe]=+0.11.
Two common proper motion stars (HD 134439 - HD 134440) show abundance anomalies similar to those of our binary, they have retrograde orbits (which are sometimes interpreted as a sign of accretion and particular nucleosynthesis history). The deficiency of HD 134439 is moderate: only [Fe/H] (King 1997, Ryan et al. 1991) but [Mg/Fe] , [Ca/Fe] and [Ti/Fe] . Strontium is also rather weak in HD 134439, [Sr/Fe]. The abundance pattern of CS 22873-139 is therefore strikingly similar to the pattern found in HD 134439.
It has to be noted that all the very Sr-poor stars have to be formed before the addition of the ejecta of massive AGB (which form and eject s-process elements) and therefore relatively rapidly after the explosion of the SN IIe, owing to the relatively short time scale of the evolution of the AGB, and of the mixing of the AGB products by their wind.
Since our star does not show the usual enhancement of -elements (found in the observations of Pop II stars and in the predictions of SN II yields averaged over the IMF), we can try to compare the abundancce pattern to the predictions of nucleosynthesis in individual SN II. For example, recent computations by Umeda et al. (2000) predict no overabundance of Mg and Ca relative to Fe in the ejecta of a 20 SN II of zero metallicity. The models of Umeda et al. predict for such SN IIe, large deficiencies of Na and Al (odd-even effect) and these deficiencies are found in our star (and in most of the other Sr-poor stars, but not all). The predicted odd-even effect is however about twice as large as observed in our star, although a NLTE analysis would provide a lower Na abundance, following Baumüller et al. (1998). Moreover, Ti is predicted to be deficient (and observed abundant). Umeda et al. consider (cf. their Sect. 8.3) that the yield of hypernovae could possibly explain the relatively high abundance of Ti.
Perhaps the ratios of the abundances to Fe are not essential, since the mass-cut parameter is rather uncertain, the relative abundances of the elements Mg, Ca and Ti (and the deficiencies of Na and Al) are maybe more informative.
The SN II models predict, for other masses, an enhancement of the -elements: for example the 25 model of Umeda et al. (2000) predicts Ca more enhanced than Mg (Al and Na remaining deficient): this could fit for example, the star CS 22885-96.
Also, models of Nomoto et al. (1997) for SN IIe more massive than 18 , predict an enhancement of Mg/Fe and even, for masses of 40 , an enhancement of both Mg/Fe and Ca/Fe, but strong deficiencies of Na and Al are not predicted: these models could fit stars such as CS 22891-200. On the contrary, most of the Sr-poor stars do not fit the abundances shown by the global product (averaged over the IMF) of the SN IIe of various masses (Fig. 8 of Nomoto et al. 1997).
Let us recall that one explanation proposed in the literature for stars with a low -elements/Fe ratio, is a late star formation, from primordial matter polluted by already evolved matter, i. e. matter which has already reached solar abundance ratios. But this explanation does not fit well the Sr-poor stars which are presumably second generation stars.
Another explanation is that the stars are from a region of low star formation rate (or of infrequent bursts) so that solar ratios (due to the addition of the ejecta of SN Ia) can be reached at low metallicity. Again this explanation does not fit well the very iron-poor and Sr-poor stars (as noted above, the stars have to be formed quickly after the explosion of the SN II, before a large formation of s-elements by the AGB).
Finally, it appears that the the most likely explanation could be that the very metal deficient and Sr-poor stars seem to reflect the products of few (or even of a single) zero-metal SN II, or hypernovae, and do not follow the mean abundances obtained by the integration of the products of the SN IIe, averaged over the standard IMF.
© European Southern Observatory (ESO) 2000
Online publication: August 23, 2000