Astron. Astrophys. 342, 87-100 (1999)

5. Discussion

5.1. Element abundances

Deriving abundances of AGN clouds from photoionization models has generally been considered to be unreliable because the shape of the AGN ionizing continuum and the density distribution of the emitting regions (clouds) are both basically unknown. The more typical approach therefore has been to assume a metallicity and use photoionization models to constrain the AGN spectrum and/or the gas density distribution, or just to demonstrate that the gas is photoionized. Although explicit statements concerning the nitrogen abundance are often found in the literature (e.g. Storchi-Bergmann & Pastoriza 1989, Simpson & Ward 1996) these are based on a very limited choice of photoionization model parameters and in most cases find oxygen abundances close to solar, in disagreement with what is derived here. The only other piece of work which covers a model parameter range comparable to that presented here is that by Komossa & Schulz (1997) who analyzes a much more limited numbers of lines, e.g. do not include [ArIV,V], in a large sample of Seyferts. They find that, on average, oxygen is underabundant by a factor of 2 and that the N/O ratio is only a factor of 1.5-2.0 above the solar value. We believe that the analysis presented here leads to more reliable metallicity estimates.

The results presented here indicate that, in spite of the above uncertainties, reliable metallicities can indeed be derived from spectra including a large enough number of lines, such as that of knot C. Our method has been described in Sect. 4.2 and, in short, is based on the computation of a large number of photoionization models with a minimal personal bias and preconceived ideas on the AGN ionizing spectrum and the gas density distribution. In particular, we also considered mixed models with combinations of density and radiation bounded clouds (B96) as well as models with multiple density components. We spanned a very wide range of model parameters which were varied randomly, and selected the relatively few (about 200) good models which came closest to reproducing the observed line ratios.

The main results are summarized in Figs. 10, 11 which show the distribution of element abundances required to reproduce the observed line ratios. The most remarkable feature is that the metal abundances are quite well constrained in spite of the very different assumptions made for the gas density distribution and shape of the AGN continua. In other words, models with very different abundances fail to match the observed lines ratios in knot C regardless of the AGN spectral shape and/or gas density distribution assumed.

Fig. 10. Element abundances (in log of solar units) as derived from a comparison between the observed line strengths in knot C, and the predictions of a large grid of randomly generated photoionization models. The three columns refer to clouds with different gas density distributions, and the "good models" are those coming closer to reproducing the observed line ratios. See text, Sects. 5.1 and 4.2 for details.

Fig. 11. Same as Fig. 10 but using photoionization models where the rate coefficient for N-H charge exchange reactions were arbitrarily set to zero. Note that this provides a much better match for the [NII]/[NI] line ratio while has a small effect on the Nitrogen abundance derived from [NII] lines. See text, Sects. 5.1 and 4.2 for details.

Another encouraging result is that lines from different ionization stages yield similar abundances which simply reflects the fact that the models reproduce the observed line ratios reasonably well. There are however remarkable exceptions, such as the [NII]/[NI] and [FeII]/[FeVII] ratios which are both predicted too high. Possible explanations for these differences have been discussed above (Sects. 4.5, 4.6). We stress here, however, that the uncertainties on [NI] have little effects on the derived nitrogen abundance because N⁺ is the most abundant ion within the partially ionized region. Therefore, the best-fit N abundance decreases by only a factor 1.3 once the N⁺/N⁰ is increased to match the observed [NII]/[NI] ratio (cf. Fig. 11). Note also that the He/H abundance is only poorly constrained by the models, and although models with He/H0.1 are somewhat favoured, no firm conclusion about He overabundance can be drawn from the data.

The derived abundances are summarized in Table 6 where the most striking result is the large overabundance of nitrogen relative to oxygen, +0.7 dex above the solar value, whose implications are discussed below.

Table 6. Metal abundances in knot C  All values are 0.2 dex. Iron is not included because its relative abundance is very uncertain (cf. Sect. 4.5)  Absolute abundance by number  Log abundance by number relative to H=12.0, N=7.97, O=8.87, Ne=8.07, S=7.21, Ar=6.60, Fe=7.51, the adopted set of solar abundances  [X/O]=log(X/O)-log(X/O)  Predicted for a yr old starburst (Fig. 4 of Matteucci & Padovani 1993), see Sect. 5.3 for details  From Fig. 11  Element not included in the chemical evolution model  Derived from CO stellar absorption features (cf. Sect. 5.2)  0.3 dex

5.2. Comparison with other abundance estimates

An independent estimate of metallicity can be derived from the measured equivalent widths of CO stellar absorption features, using the new metallicity scale proposed and successfully applied to young LMC/SMC clusters by Oliva & Origlia (1998). In short, the method is based on the strength of the CO(6,3) band-head at 1.62 µm whose behaviour with metallicity is modelled using synthetic spectra of red supergiants. The equivalent width of the stellar CO lines from the central 100100 pc² of Circinus are reported in Table 2 of Oliva et al. (1995) and yield an average metallicity of , a value remarkably close to the oxygen abundance derived above (cf. Table 6).

5.3. Nitrogen overabundance and starburst activity

The nitrogen overabundance is of particular interest in view of its possible relationship with the (circum)nuclear starburst and N-enrichment from material processed through the CNO cycle. According to chemical evolutionary models of starburst events, the N/O relative abundance reaches a maximum value of [N/O] (i.e. 4 times the solar value) at about yr and remains roughly constant for several years (cf. Fig. 4 of Matteucci & Padovani 1993). The nitrogen overabundance mostly reflects the effect of the winds from He burning red supergiants whose surface composition is strongly N-enriched by gas dredged-up from the shell where hydrogen was burned through the CNO cycle. The amount and temporal evolution of the N/O abundance depends on model details, e.g. the shape of the IMF and the duration of the starburst, as well as on poorly known parameters such as the efficiency of the dredge-up and the contribution of primary N production by massive stars (e.g. Matteucci 1986). It is however encouraging to find that the observed N/O abundance (Table 6) is very close to that predicted at a time which is compatible with the age of the starburst in Circinus (cf. Fig. 9 of Maiolino et al. 1998). It should also be noticed that the observed absolute abundances are about an order of magnitude lower than the model predicted values, but this can be readily explained if the starburst transformed only of the available gas into stars, in which case the chemical enrichment was diluted by a similar factor. This hypothesis is in good agreement with the fact that Circinus is a very gas rich galaxy (e.g. Freeman et al. 1977).

In short, the observed N/O overabundance is fully compatible with what expected for a quite old (several yr) starburst. The [NII]/H and other line ratios measured in the cone of Circinus are similar to those observed in many others Sy2's and several observational results indicate that relatively old starburst events are common in type 2 Seyferts (e.g. Maiolino et al. 1995). This may therefore indicate that nitrogen is typically overabundant in Sy2's due to enrichment by the starburst associated with the AGNs. This tantalizing conclusion should be however verified by detailed spectroscopic studies and analysis of a sufficiently large number of objects.

© European Southern Observatory (ESO) 1999

Online publication: December 22, 1998
helpdesk.link@springer.de