4.4.8 Abundance ratios in BCGs

Oxygen is normally considered as representative of the metallicity of BCGs. However H II region abundance analysis can also provide abundances of other elements. Especially nitrogen, helium and carbon have been investigated. In addition elements such as argon, neon and sulphur may be studied. Lately, iron has been added to the list. The study of helium in BCGs offers a route towards determining the primordial He abundances, and will be discussed separately in Sect. 8.1. In Fig. 4 we show C/O and N/O vs. O/H for BCGs.

Figure 4. Element ratios in BCG and dIs: Top: The relation between oxygen abundance and C/O. Filled symbols are from Izotov and Thuan 1999, which includes a reanalysis of previously published data. The open circles are from Garnett et al. (1995). The cross is ESO 338-IG04 (Bergvall 1985, Masegosa et al. 1994). The open triangles show the location of the NW and SE regions in IZw18 from Garnett et al. (1997), while the filled triangles shows the same regions as derived by Izotov and Thuan (1999). Bottom: The relation between oxygen abundance and N/O for BCGs, data taken from Izotov and Thuan (1999).

The investigation of carbon abundances in the H II gas poses some difficulties since there are no strong emission lines in the optical regions. The investigation of carbon abundances in BCGs began with the International Ultraviolet Explorer (IUE) satellite and has continued with the HST. Garnett et al. (1995) presents C/O ratios for seven galaxies, including some BCGs, and three others are presented by Kobulnicky and Skillman (1998). Thus carbon abundances are still not very well explored in BCGs. Garnett et al. (1995) found that C/O increases with increasing oxygen abundance. The average value of this ratio is rather low, as compared to solar, except possibly for IZw18 which has C/O about a factor of two larger than predicted from stellar nucleosynthesis (Garnett et al. 1997).

The relative abundance of nitrogen to oxygen increases with O/H (Pagel and Edmunds 1981, Serrano and Peimbert 1983, Torres-Peimbert et al. 1989), implying a secondary origin of N in the CNO cycle. Such a behaviour was not seen at very low O/H (Lequeux et al. 1979; Kunth and Sargent 1983; Campbell et al. 1986) indicating that nitrogen is mainly a primary element in very metal-poor gas. The current interpretation of this behaviour from stellar nucleosynthesis models is that intermediate stars produce primary nitrogen by hot-bottom burning. In such a phase, the third dredge-up brings carbon-rich material from the core onto the hydrogen burning shell (Renzini and Voli 1981; van der Hoek and Groenewegen 1997). The scatter of the N/O versus O/H diagram has been considered as larger than the observational uncertainties (although they were nearly comparable two decades ago). Time delays between the production of oxygen due to massive stars and that of nitrogen is likely part of the explanation although this point of view has been challenged by recent data from Izotov and Thuan (1999). Indeed their high signal to noise observations not only suggest a small intrinsic dispersion of log N/O (± 0.02 dex) at low metallicities but a similar behaviour is found for C/O and other ratios, see Fig. 4. The disagreement with Garnett et al. (1995, 1997) comes mainly from the reassessment of C/O in IZw18, the abundances of which are thoroughly discussed in Sect. 5.1.1. Izotov and Thuan (1999) find positive correlations between C/O and N/O with O/H but for 12 + log(O/H) 7.6, C/O and N/O remain constant and independent of O/H. They conclude that galaxies with such low abundances are genuinely young (less than 40 Myr old), now making their first generation of stars. Moreover they suggest that all galaxies with 7.6 12 + log(O/H) 8.2 have ages from 100 to 500 Myr. Thus, the question raised by Searle and Sargent almost 30 years ago would after all have a positive answer. However, there are independent data suggesting that these galaxies do in fact contain old stars (see Sect. 5). Moreover, there are definitely many BCGs with 12 + log(O/H) < 8.2 which have been convincingly shown to be much older than 500 Myr, e.g. ESO 338-IG04 from its globular clusters (Östlin et al. 1998). Moreover, as we shall discuss below, there are alternative interpretations of the abundance patterns which do not require the galaxies to be young.

H II regions in the outskirts of spiral galaxies have C/O values as low as those of the most metal-poor galaxies, and H II regions in spiral galaxies follow the same C/O vs. O/H relation as dwarf galaxies (Garnett et al. 1999). This suggests that they evolve chemically in the same manner. Now, the discs of spiral galaxies are several Gyr old, still the C/O is as low as in the most metal-poor BCGs, clearly indicating that C/O is not simply a function of age. The observed trend of C/O vs. O/H could equally well be explained by a metallicity dependent yield (Maeder 1992). Gustafsson et al. (1999) studied the carbon abundances of disc stars in our Galaxy and concluded that the observed relation could be explained if carbon production occurs mainly in massive WR(WC) stars. In this scenario, C/O would be mainly a function of metallicity and not age.

A similar pattern is seen for N/O observations of H II regions in spirals. Outlying H II regions appear to have N/O similar to the most metal-poor galaxies (van Zee et al. 1998b). Moreover, the low surface brightness galaxies studied by Rönnback & Bergvall (1995) have N/O comparable to those of the most metal-poor BCGs, which are still fairly old systems (Bergvall et al. 1999). Pilyugin (1999) finds that if significant N production occurs in intermediate mass stars, and the heavy element abundances have not been polluted by the present star formation event (i.e. the time scale for cooling of fresh metals is longer than the typical lifetime of a giant H II region) the constant N/O found at low metallicity is consistent with the presence of previous starbursts, i.e. high ages. It is also worth commenting that if the time scale for recycling is longer than the duration of a typical burst of star formation, this can explain the lack of abundance gradients in dwarfs (Sect. 3.5). As for carbon, the net yield of nitrogen may be metallicity dependent due to an increased contribution from WR(WN) stars with increasing metallicity.

The elements Ne, Si, S and Ar all shows a constant abundance relative to oxygen, independent of O/H as expected from stellar nucleosynthesis, since they are products of -processes (Izotov & Thuan 1999). Finally the Fe/O abundance ratio in BCGs is on average 2.5 times smaller than in the Sun with a mean [O/Fe] = 0.40 ± 0.14 with no dependence on oxygen abundance (Izotov and Thuan 1999). The scatter is surprisingly small considering the short time scale for the production of oxygen as compared to iron production because different stellar masses are involved. If real, it would imply that Fe could have been produced by explosive nucleosynthesis of SNe type II for both O and Fe at the early stage of chemically unevolved galaxies.