4 Discussion

4.1 The blue continuum in Seyfert 2 galaxies

Koski (1978) and Kay (1994) found that all Seyfert 2 galaxies show an ultraviolet excess and weak absorption lines when compared with galaxies with no emission lines, indicating the presence of a blue featureless continuum. Boisson & Durret (1986) and Vaceli et al. (1997) suggested that this continuum is a non-thermal power-law continuum. Kinney et al. (1991) argued that most of the Seyfert 2s in which a blue continuum has been observed are of type Sb or earlier, suggesting that it is truly associated with the Seyfert nucleus. Shuder (1981) showed that its strength and the $\rm H\alpha$ luminosity are strongly correlated suggesting that a direct physical connection exists between the two; studying a sample of 28 Seyfert 2s, Yee (1980) found that the $\rm H\beta$ and continuum fluxes (rather than luminosities) are proportional over two orders of magnitude, with, however, a relatively large dispersion; but a number of those objects are now known to be Seyfert 1 galaxies.

Martin et al. (1983) discovered that a small fraction of all Seyfert 2 galaxies have a highly polarized continuum. Subsequently, Antonucci & Miller (1985); Miller & Goodrich (1990) and Tran et al. (1992) showed that these objects harbour a hidden Seyfert 1 nucleus, the observed polarized continuum arising from scattering of the nuclear continuum by dust or warm electrons. But most Seyfert 2s have very little polarization (Martin et al. 1983), much less than expected in the reflection model (Miller & Goodrich 1990).

On the other hand, Terlevich et al. (1990) showed that in Seyfert 2 galaxies, the IR Ca II triplet is equal or, in some cases, higher than in normal elliptical galaxies, which is most naturally explained by the presence of young stars contributing heavily to the nuclear light at near-IR wavelengths.

Heckman et al. (1995) used International Ultraviolet Explorer (IUE) spectra of 20 of the brightest type 2 Seyfert nuclei to build an ultraviolet template for this class; while the continuum was well detected in the template, there was no detectable broad line region (BLR), implying that no more than 20% of the template continuum could be light from a hidden Seyfert 1 nucleus scattered by dust; they suggested that either most of the nuclei in their sample were "pure'' type 2 Seyfert galaxies for which we have a direct view of the central engine and which simply lack of BLR, or that most of the observed ultraviolet continuum is produced by starbursts. From the absence of polarization of the continuum of most Seyfert 2 galaxies and of broad Balmer lines, Cid Fernandez & Terlevich (1995) concluded that, most probably, this continuum was due to a population of young stars in the vicinity of the nucleus. Colina et al. (1997) obtained ultraviolet HST images of four nearby Seyfert 2 galaxies known to have circumstellar star-forming rings, providing direct empirical evidence that the UV flux emitted by these galaxies is dominated by radiation coming from clusters of young hot stars distributed along the star-forming ring. If similar rings are a common characteristic of Seyfert 2 galaxies, the large IUE aperture would include both the Seyfert 2 nucleus and the rings for distances larger than 25 Mpc. Gonzalez Delgado et al. (1998) presented HST images and ultraviolet spectra of three Seyfert 2 nuclei (IC 3639, NGC 5135 and IC 5135); the data show the existence of nuclear starbursts (with absorption features formed in the photosphere of late O and early B stars) dominating the ultraviolet light. It is remarkable that, of the three observed galaxies, two (NGC 5135 and IC 5135) have a "composite'' nuclear emission spectrum, while the third (IC 3639), which has the largest UV nuclear flux (associated with the Seyfert nucleus) relative to the total UV flux, has a pure Seyfert 2 spectrum due to the relative weakness of the starburst emission component.

We conclude that there is ample evidence for the presence of young, hot stars in the nuclear region of many Seyfert 2 galaxies. When the continuum is relatively bright, the associated H II region could be strong enough to displace the object into the "transition'' zone in the diagnostic diagrams.

AGNs are more frequent in early type galaxies while starbursts are more often found in late-type galaxies (Véron & Véron-Cetty 1986; Ho et al. 1997b; Vaceli et al. 1997). It is therefore rather surprising to find almost systematically a population of young stars in Seyfert 2 galaxies; perhaps the nuclear activity triggers the star formation?

4.2 Excitations and abundances in ionHII nuclei and AGNs

The $\lambda 5007/{\rm H}\beta$ and $\lambda 6583/{\rm H}\alpha$ ratios are strongly correlated in H II regions. Theoretical studies show that the heavy-metal abundances change continuously along this sequence, a low $\lambda 5007/{\rm H}\beta$ ratio indicating a high metal abundance and a high $\lambda 5007/{\rm H}\beta$ ratio, a low metal abundance, with the heavy metal abundances changing from $1.5 \,Z_{\hbox{$\odot$}}$ at the lower right of Fig. 4a to $0.25 \,Z_{\hbox{$\odot$}}$ at the upper left (see for instance Dopita & Evans 1986; Ho et al. 1997b). However, Stasinska & Leitherer (1996) have shown that most startbusts and H II galaxies can be described as being produced by an evolving starburst with an universal initial mass function embedded in a gas cloud of the same metallicity. The emission line ratios depend mainly on two independent parameters: the age of the starburst and the metallicity. In this scenario, the $\lambda 5007/{\rm H}\beta$ ratio effectively changes with these two parameters and therefore is not a direct measurement of metallicity. The metallicity is strongly correlated with luminosity, luminous galaxies having higher metallicities; this correlation is also valid for elliptical galaxies, for which the metallicity is determined from absorption lines with [O/H] $\sim$ 1 at M_B = -21 (Salzer et al. 1989; Zaritsky et al. 1994).

AGNs are known to occur preferentially in high luminosity (Ho et al. 1997b), early-type (Véron & Véron-Cetty 1986; Vacali et al. 1997) galaxies; they are therefore expected to have high metallicities. Indeed, the NLRs of active galactic nuclei have enhanced nitrogen abundances (Storchi-Bergmann & Pastoriza 1989, 1990; Storchi-Bergmann et al. 1992; Schmitt et al. 1994). In these NLRs, [N/O] correlates with [O/H] in a manner identical to H II regions in normal galaxies, with nuclear [O/H] and [N/O] values ranging from $1 \,Z_{\hbox{$\odot$}}$ to $2\, Z_{\hbox{$\odot$}}$ (Storchi-Bergmann et al. 1996b). Storchi-Bergmann et al. (1996b,c) have determined the chemical composition of the H II regions in the ring surrounding the nucleus of several AGNs, as well as in the nuclei; high metallicities were found ([O/H] $\sim 2 \,Z_{\hbox{$\odot$}}$ and [N/O] $\sim 3\, Z_{\hbox{$\odot$}}$) both in the H II regions and in the AGNs, these abundances being similar to those found in the nuclei of non-active galaxies with the same morphological type and absolute magnitude. Further work by Storchi-Bergmann et al. (1998) has shown that, in fact, oxygen abundances derived for Seyfert 2 nebulosities and neighbouring H II regions (assuming that the emission lines in the active nucleus are due to photoionization by a typical active galactic nucleus continuum) are well correlated, while this is not the case for Liners. This suggests that the gas in AGNs and in the neighbouring H II regions has the same origin and that the scatter observed in the Seyfert 2 region in the diagnostic diagrams, involving the $\lambda 6583/{\rm H}\alpha$ ratio, is due to variations in the nitrogen abundance. In NGC 6300, in which $\lambda 6583/{\rm H}\alpha$ = 3.4, the nitrogen abundance is estimated to be $\sim 5\, Z_{\hbox{$\odot$}}$.

We have seen that nuclear H II regions and Seyfert 2 nebulosities, when appearing in the same galaxy, have the same high metallicity; as a result of their metallicity, the H II regions have a low excitation, while the Seyfert 2 nebulosities have a high excitation. This explains why it is relatively easy to separate the two components in "transition'' spectra.

4.3 Objects with weak [ionNII] lines

  Figure 4 shows a small number of objects which have very weak [N II] lines for Seyfert 2 galaxies; their [O I] lines are however normal for this class of objects.

The first photoionization models invoked to explain the narrow emission lines in AGNs assumed a single density cloud. However, new observations quickly suggested the presence of several emitting clouds, ruling out single component models. Most of the multicloud models first studied were such that the emitting gas, as a whole, was ionization-bounded and thus the He II$\lambda$4686 line intensity relative to $\rm H\beta$ was determined by the hardness of the ionizing spectrum. In these models, the extreme values reached by the $\lambda 4686/{\rm H}\beta$ ratio are not well reproduced. A number of objects have $\lambda 4686/{\rm H}\beta$ of the order of 0.2 or more; such high values cannot be accounted for unless the line emitting clouds are matter-bounded (Stasinska 1984). On the basis of a weak trend for the low excitation lines to become weaker as $\lambda 4686/{\rm H}\beta$ gets larger, Viegas-Aldrovandi (1988) and Viegas & Prieto (1992) argued in favor of a model in which matter-bounded clouds are present; indeed, if the gas is not optically thick to all the ionizing continuum (i.e., is matter bounded), the H⁺ emitting volume is smaller, but the He⁺⁺ volume is not, leading to a higher $\lambda 4686/{\rm H}\beta$ line ratio. Moreover, Viegas-Aldrovandi & Gruenwald (1988) and Rodriguez-Ardila et al. (1998) showed that, for most AGNs, the observed low-excitation lines are better explained by matter-bounded models with about 50% of the $\rm H\beta$ luminosity produced in ionization-bounded clouds.

Storchi-Bergmann et al. (1996a) have obtained long-slit spectra of five active galaxies showing extended high excitation lines. At some positions, two of the objects (PKS 0349-27 and PKS 0634-20) show quite peculiar line ratios, with a strong He II$\lambda$4686 line ($\lambda 4686$/ $\rm H\beta \gt 0.3$) and weak [N II] lines (that is, $\lambda 6583/{\rm H}\alpha$ < 0.3). In fact, there seems to be a correlation between $\lambda 6583/{\rm H}\alpha$ and $\lambda 4686/{\rm H}\beta$, weak [N II] lines being associated with strong He II emission, suggesting that very small $\lambda 6583/{\rm H}\alpha$ ratios (as observed in the two above mentioned radiogalaxies) are not necessarily a signature of star-formation, but a natural consequence of having a region dominated by matter-bounded clouds (Binette et al. 1996, 1997). However, in the extranuclear regions of PKS 0349-278 in which strong He II$\lambda$4686 and weak [N II]$\lambda$6583 lines are observed, the [O I] $\lambda$6300 line is also reduced ($\lambda 6300/{\rm H}\alpha$ $\sim$ 0.05), which is a natural consequence of the model (Viegas-Aldrovandi 1988), while in our sample of weak [N II]$\lambda$6583 galaxies, we verify that the [O I]$\lambda$6300 line is not weakened in most of the objects.

In Table 8 we give the list of known AGNs with relatively weak [N II] lines ($\lambda 6583/{\rm H}\alpha$ < 0.45) with published values of the $\lambda 4686/{\rm H}\beta$ and $\lambda 6300/{\rm H}\alpha$ ratios. Three objects in this table (UM 85, MS 04124-0802 and Mark 699) have both weak [N II] lines ($\lambda 6583/{\rm H}\alpha$ < 0.20) and a strong He II line ($\lambda$4686/$\rm H\beta$ > 0.30). In the last two, the [O I] lines are also relatively weak ($\lambda 6300/{\rm H}\alpha$ $\le$ 0.05); these two objects could be dominated by matter-bounded clouds. Alternatively, in the other objects, the weakness of the [N II] lines could be due to a selective under-abundance of nitrogen. For a photoionized single cloud model with $U \sim$ 10^-2.5, Ferland & Netzer (1983) predicted $\lambda 6583/{\rm H}\alpha$ $\sim$ 1.0 for solar nitrogen abundances and $\sim$ 0.3 for nitrogen abundances $\sim$ 0.3 solar.

 

Table 8:  Known AGNs with weak [N II] lines

4.4 Seyfert 2s and Liners

It has been suggested by several authors (see for instance Ferland & Netzer 1983; Shields 1992; Ho et al. 1993a) that in Seyfert 2s, as well as in Liners, the ionized gas is excited by a non-thermal continuum, the only differences being the value of the ionizing parameter which would be $\sim\! 10^{-3.5}$ for Liners, and $\sim\!10^{-2.5}$ for Seyfert 2s. If this is the case, the discontinuity between Seyfert 2s and Liners is not easily understood. No reliable detection of the He II line in bona fide Liners has been reported suggesting that there could be a serious problem with the picture of simply reducing U in a standard power-law photoionization model predicting $\lambda 4686/{\rm H}\beta$ > 0.15 (Viegas-Aldrovandi & Gruenwald 1990), as the weakness of He II indicates that the continuum illuminating the NLR clouds must contain few photons more energetic than 54.4 eV, the ionization potential of He⁺ (Péquignot 1984). Binette et al. (1996) proposed that the emission spectrum of Liners is due to ionization-bounded clouds illuminated by a ionization spectrum filtered by matter-bounded clouds hidden from view by obscuring material. In this case, the He II emission is reduced ($\lambda 4686/{\rm H}\beta$ < 0.01). However, a nearly total obscuration of the matter-bounded component must then be invoked in order to keep the emission from He II at an acceptable low level, a scenario which seems to be rather unlikely to Barth et al. (1996).