1 Introduction

The element abundances in the most metal-poor stars provide the fossil record of the earliest nucleosynthesis events in the Galaxy, and hence allow the study of Galactic chemical evolution in its earliest phases. Moreover, such data can have cosmological implications: radioactive age determinations (Cowan et al. 1999; Cayrel et al. 2001) provide an independent lower limit to the age of the Universe, and the lithium abundance in metal-poor dwarfs is an important constraint on the baryonic content of the Universe (Spite & Spite 1982). The dispersion of the observed abundance ratios for a number of elements in extremely metal-poor (XMP) stars is very large, reflecting the yields of progenitors of different masses, and indicating that each star has been formed from material processed by a small number of supernova events - perhaps only one. Thus, the detailed chemical composition of XMP stars provides constraints on the properties of the very first supernovae (and hypernovae?) events in our Galaxy. Accordingly, the past decade has seen a rapidly growing interest in the detailed chemical composition of XMP stars.

BPS CS 22947-037, discovered in the HK survey of Beers et al. (1992, 1999), is one of the lowest metallicity halo giants known ([Fe/H] $\approx-4.0$). Several studies (McWilliam et al. 1995; Norris et al. 2001) have shown that this star exhibits a highly unusual chemical composition, characterised by exceptionally large enhancements of the lightest $\alpha$ elements: Norris et al. (2001) found $\rm [Mg/Fe] = +1.2$, $\rm [Si/Fe] = +1.0$, $\rm [Ca/Fe] = +0.45$, and even more dramatic values for carbon and especially nitrogen: $\rm [C/Fe] = +1.1$ and $\rm [N/Fe] = +2.7$. In contrast, the abundance of the neutron-capture elements was found to be rather low: $\rm [Ba/Fe] = -0.77$.

These abundance ratios cannot be explained by classical mixing processes between the surface of the star and deep CNO-processed layers, or by the enrichment of primordial material by the ejecta of standard supernovae. Canonical supernova and Galactic Chemical Evolution (GCE) models (Timmes et al. 1995) predict that [C/Fe] should be solar, and [N/Fe] subsolar (since N is produced as a secondary element). Norris et al. (2001) discuss a variety of scenarios to explain this abundance pattern, and finally suggest that the material from which the star was formed was enriched by the ejecta from a massive zero-heavy-element hypernova ( ${>}200~ M_{\odot}$), which might be able to produce large amounts of primary nitrogen via proton capture on dredged-up carbon.

However, no previous analysis has yielded a measurement of the abundance of oxygen, an extremely important element, both for the precise determination of the nitrogen abundance and as a key diagnostic of alternative progenitor scenarios. We have therefore observed CS 22949-037 in the framework of a large, systematic programme on XMP stars with the ESO VLT and its UVES spectrograph. The extended wavelength coverage (including the red region), superior resolution, and S/N of our spectra enable us to determine abundances for a large sample of elements with unprecedented accuracy, accounting for the abundance anomalies of the star in a self-consistent manner, especially in the opacity computation.

This paper presents these new results and discusses their astrophysical implications. In Sect. 2 the observations are sumarized. Section 3 presents a description of the model atmosphere calculations, and the methods used in the elemental abundance analyses. In Sect. 4 the observed abundance pattern for CS 22949-037 is compared with predictions of recent supernova and hypernova models.

 

 

Table 1: Log of the UVES observations. S/N refers to the signal-to-noise ratio pixel at 420, 630, and 720 nm in the mean spectrum. There are between 6.7 and 8.3 pixels per resolution element.

			Exp.	$V\rm _r$	420	630	720
date	UT	Setting	time	km s-1	nm	nm	nm
2000-08-08	04:54	396-573	1 h	-125.68 $\pm$ 0.2
2000-08-08	05:57	396-850	1 h
2000-08-09	05:04	396-573	1 h	-125.64 $\pm$ 0.2
2000-08-09	04:01	396-850	1 h
2000-08-11	04:27	396-573	1 h	-125.60 $\pm$ 0.2
2000-08-11	05:36	396-850	1 h
Global S/N per pixel (2000/08)					110	170	110
Global S/N per resolution element					285	490	320
2001-09-06	05:06	396-573	1 h	-125.62 $\pm$ 0.2
S/N per pixel (2001/09)					40	90	-