Extremely metal-poor stars .4. The carbon-rich objects

Abundances are presented for some 20 elements in three extremely metal-poor, carbon-rich stars. All have [Fe/H] < -2.5. Based on their high proper motions and spectroscopic gravities two of them have relatively low luminosity: LP 706-7 is on or near the main sequence, while LP 625-44 is a subgiant. The third is the giant CS 22892-052, discovered to be r-processes enriched by Sneden et al. (1994). All three stars have large C and N overabundances, and large enhancements of the heavy neutron-capture elements - similar to 1-2 dex. In contrast to the r-process signature observed in CS 22892-052, however, both LP 625-44 and LP 706-7 are clearly s-process enriched, suggesting that they may be progenitors of the well-known halo CH giants, the peculiarities of which are believed to result from mass transfer across a binary system containing an asymptotic giant branch star. In both LP 625-44 and LP 706-7 the distribution of s-process elements is heavily weighted toward higher atomic number. [hs/ls] similar to 1.5, considerably larger than the values less than or similar to 0.5 found in the s-process enhanced near-main-sequence stars of the disk populations. This implies a much higher neutron exposure per seed nucleus in the Population II objects and identifies C-13(alpha, n)O-16 as the neutron source. For LP 625-44 radial velocity and Li abundance data are consistent with the binary hypothesis. LP 706-7, however, remains something of an enigma in these respects: it shows no clear evidence for velocity variations, and its Li abundance lies precisely on the Spite Plateau. We estimate that the probability of this occurring in the Ba/CH class of objects is, roughly, less than or similar to 1%. In metal-poor stars the incidence of carbon enrichment appears to increase toward the lowest metallicities, and below [Fe/H] = -2.5 supersolar values of [C/Fe] are not uncommon. Comparison of the available observational material with the Galactic chemical enrichment model of Timmes et al. (1995) shows that the model produces too little carbon. While the difference may result in the sensitivity of carbon production to modeling uncertainties such as the treatment of convection, we also discuss the possible role of a class of carbon producing zero heavy element supernovae and of massive ''hypernovae'' discussed by Woosley & Weaver (1982, 1995) in explaining this result. The carbon problem is also implicit in the suggestion that the r-process signature seen in CS 22892-052 results from normal supernovae enrichment-where then does its large carbon overabundance originate? The models one might invoke to produce carbon overabundances leave black hole remnants in which the layers containing the seed nuclei for the r-process are not available for ejection.