On the basis of the previous results we conclude that the data can only be reproduced if we vary the abundances of nitrogen and sulfur alone, from 0.5 to 5 times the solar value; it is not possible to reproduce them by varying the abundance of the heavy elements all together. As these results depend on the models used, we have also checked the grid of photoionization models of Stasinska (1984b) and have not found any significant difference for the models with similar parameters to those used in this work.
The diagrams [N II] / H x [S II] / H show some correlation between these two ratios. A linear regression to the data points gives a correlation coefficient of 0.64. Should this mean that the nitrogen and sulfur abundances are correlated? The results of the previous section indicate that the answer is yes, because, as was shown above, a variation of U, gas density, and abundance of all the elements in the models cannot reproduce the data in this diagram. We also point out that it is not possible to cover the data by varying the abundance of only N or S; It is necessary to vary the abundance of both elements together. A better coverage of the data points can be obtained if we consider the coupled effect of a varying density, ionization parameter U, and N and S abundances, as shown in Figure 5. In this figure we have considered two typical values for U: 10-3 and 10-3.5, densities in the range 103-105, cm-3, and N and S abundances 0.5 times solar, solar, and two and three times solar. This figure shows that we can successfully reproduce the data, although the effect of a varying U intermixes somewhat with that of varying N and S abundances. It can also be seen that the broadness of the relation can be explained by the presence of clouds of different densities in the gas.
Figure 5. Sample galaxies (dots) and photoionization models (lines) showing the coupled effect of a varying gas density (103-105 cm-3) and ionization parameter (10-3 < U < 10-3.5) for nitrogen and sulfur abundances 0.5, 1, 2, and 3 times solar. The abundances of the other elements are solar.
In a previous work (Storchi-Bergmann 1990) it has been found a dependence of the [N II] / H ratio observed for this same sample of Seyfert 2's and LINERs with the effective aperture (l) used in the observations (square ratio of the area at the galaxy corresponding to the observing slit), it was found that for l < 1 kpc most galaxies present N overabundance in the nucleus. On the basis of the correlation between [N II] / H and [S II] / H, we have searched for the same effect for sulfur. Figure 6 shows the [S II] / H ratio plotted against effective aperture, where it can be seen that high [S II] / H values are preferentially found for l < 1 kpc. The histogram quantifies this result showing that 60% of the subsample with l < 0.5 kpc present [S II] / H > 0.6 (solar value). This result shows that indeed the same effect found previously for N is occurring with S: There is a sulfur overabundance in most Seyfert 2 and LINER nuclei, which is very localized, being restricted to a region with a typical diameter of 500 pc around the nucleus.
Figure 6. The dependence ofthe [S II] / H ratio on the effective aperture l (square root of the area at the galaxy corresponding to the slit used in the observations).
The above could indicate that high N and S abundances are somehow related to the nuclear activity in these galaxies. A preliminary calculation (Cid Fernandes and Dottori 1990) shows that a model nebula photolonized by a cluster of very hot stars (WARMERS) (Terlevich and Melnick 1985) should soon be contaminated by a great amount of nitrogen due to mass loss of the stars in the Wolf-Rayet N phase. This is one possibility for explaining the high [N II] / H ratios. In another possibility in which the gas is enriched by winds from normal nuclear stars, our results introduce important constraints in the stellar evolution at the nuclei.