4 Comparison of the abundance pattern of CS 22949-037 with various theoretical studies

This early generation star is not the first one to display an abundance pattern that is not easily accounted for by standard SN II nucleosynthesis computations. The HK survey has found several very iron-poor stars with large abundances of C, and N (e.g., Hill et al. 2000). Figure 8 compares the abundances in CS 22949-037 with those of 3 classical metal-poor stars, which are passably explained by current SN II nucleosynthesis (Tsujimoto et al. 1995; Woosley & Weaver 1995), although no nitrogen (which is by-passed in a pure helium core) is predicted in our star and in CD-38$^{\circ }$245, at contrast with the observations. Figure 9 gives the ratios of the abundances in CS 22949-037 to the mean of the 3 stars. Very clearly the major feature is a large relative overabundance with respect to iron of the light elements C, N, O, Na, and Mg, declining to almost insignificance at Si, and none for Z > 15, as already noted in Norris et al. (2001). Qualitalively, something very similar is occurring in the model Z35B of Woosley & Weaver (1995) which, because of insufficient explosion energy and partial "fallback", expels only C, O, Ne, Na, Mg, Al, a very small quantity of Si, and nothing heavier. Below we discuss several attempt to refine this idea.

Another path was followed by Norris et al. (2002) for explaining CS 22949-037: the pair-instability hypernova yields (see Fryer et al. 2001; and Heger & Woosley 2002). Here one important ingredient is the mixing of some of the carbon in the helium core with proton-rich material, producing a large amount of primary nitrogen. However, the other yields of pair-instability supernovae have some features which poorly fit the more complete pattern we have obtained here for CS 22949-037. In particular, they show a larger odd/even effect than the one seen in the star, and a small [Zn/Fe] , in contrast to the observed value of $\rm [Zn/Fe]=+0.7$. So, it seems that, if the idea of primary nitrogen production by mixing must be retained, the case for pair-instability hypernovae is not attractive.

A large body of other theoretical work is relevant to the nucleosynthesis in very low-metallicity stars, and we make no attemps to fully sumarize previous results in the present papaer. However, a few recent ideas are worth keeping in mind. For example, Umeda & Nomoto (2002) have tried to explain the $\rm [Zn/Fe] \approx 0.5$ found at very low metallicity. Their conclusion is that the solution is a combination of a proper mass cut, followed by mixing between the initial mass cut and the top of the incomplete Si-burning region, followed by a fallback of most of the Si-burning region. In order to produce the usual [O/Fe] value and $\rm [Zn/Fe] \approx 0.5$, it is necessary to have a progenitor mass of 25 or $30~ M_{\odot}$, and an energetic explosion of 10 to $30\times10^{51}$ ergs.

Chieffi & Limongi (2002) have explored the possibility of adjusting the free parameters in a single SN II event to fit the abundances of five individual very metal-poor stars (Norris et al. 2001, including CS 22949-037). Although in the end they discard CS 22949-037, they note that, except for the overabundance of C to Mg, the star is very similar to the other stars of the sample, and that the high [Co/Fe] value is apparently well explained in all C-rich stars by their computed yields.

Finally, we come back to the "fallback" explanation for the high, C,N,O, and Na abundances, which make this $\rm [Fe/H] =-4$ star a $Z= 0.01~Z_{\odot}$ star. An unpublished result (model Z35Z of Woosley & Heger, in preparation) was kindly communicated to us as a variant of the already cited model Z35C. This model has a slightly larger amount of fallback, and includes hydrodynamical mixing in the explosion. It shows a fairly good fit with our observations (crosses in Fig. 10), except for Al and Na, which have to be corrected for non-LTE effects, and for N, which is not expected to be formed in the Z35Z model. To improve this fit we corrected for the non-LTE effects on Na and Al (see Sect. 3.2.3), and we supposed (open circles in Fig. 10) that the observed abundance of nitrogen was the result of a transformation of carbon into nitrogen through the CN cycle (in the star itself or in its progenitor). After these corrections the agreement is much better. The discrepancy about the Zn abundance is probably curable (Umeda & Nomoto 2002) as explained here above.

At this point we must mention that rotation may be a source of mixing and CN processing (see Meynet & Maeder 2002), and that other non-standard mixing mechanisms have been investigated along the RGB, which may have altered the ¹²C/¹³C ratio and the C/N ratio in CS 22949-037 itself (Charbonnel 1995).

The computation of the supernova yields does not contain predictions for the neutron-capture elements. Generally speaking, these elements are not easily built in zero-metal supernovae (like Z35Z), nor in zero-metal very massive objects, owing to an inefficient neutron flux, a lack of neutron seeds or both. The main phenomenon observed in CS 22949-37 is the very rapid decline of the abundance of these elements with the atomic number. Such a decline is not observed in other very metal-poor stars (see Sect. 3.2.5), and it suggests an unusually "truncated" neutron exposure (very short relative to the neutron flux).

In summary, it appears that SNe II of mass near $30~ M_{\odot}$ , either primordial or of very low metallicity, offer good prospects for explaining stars like CS 22949-037. Enough ingredients are available. They have still to be assembled in the most economic way.

Figure 10: The logarithmic mass ratio of elements X to Mg (log  $M_{\rm X}/M_{\rm Mg}$) compared with those predicted by the zero heavy-element supernova model Z35Z (Woosley & Weaver 1995, as recently modified by Heger and Woosley). The measured abundances in CS 22949-037 have been corrected for NLTE effects (Na, Al), or for internal mixing in the star (C, N). The open circles represent the assumed initial abundances of the elements, while the crosses show the atmospheric abundances as derived in LTE. For K, only an upper limit is available.