5. The influence of the ionizing mechanism on the UV line ratios
5.1. AGN photoionization models
We have used the multi-purpose photoionization code MAPPINGS I to build these models. The version described in Binette et al. (1993a, b) is well suited for the conditions studied here, since it includes a detailed treatment, not only of the transfer of optically thin lines (HeII 1640 and CIII] 1909) but also of the resonant lines, Ly and CIV 1550, which have strongly-geometry dependent emission (VMBF96).
The first step to predict the spectrum emitted by a gaseous region is to define the physical conditions of the gas and the shape of the ionizing continuum. At very high redshift, there are large uncertainties in this respect. The optical lines, which have provided a great deal of information about the emitting gas and the ionization processes involved in low redshift radio galaxies, are redshifted into the IR. As mentioned before, the optical (rest-frame) spectra obtained for very high redshift so far are rare. Therefore, it is not possible to use optical line diagnostics to constrain the physical conditions of the emitting gas in our objects.
As a starting point, it is reasonable to assume that these regions are similar to the EELR in low redshift radio galaxies. The discrepancies, if they exist, between the data and the standard models which are successful for the low-z objects will allow us to build more coherent models and therefore, reveal information about the actual conditions of the emitting gas and the nature of the ionizing source. Therefore, the question we will answer in this section is: can the models which reproduce the optical line ratios of low redshift radio galaxies also explain the UV line ratios of very high redshift radio galaxies?
Analysis of the optical emission lines has shown that typical densities in EELR apparently photoionized by the central AGN at low redshift are lower than a few hundred. Indeed, if the line emitting gas in the EELR is in pressure equilibrium with the hot phase of the ISM, the EELR densities implied from X-ray observations are of the order or a few particles per cm-3 (Clark 1996). McCarthy (1993) reached a similar conclusion using a different approach based on the Ly luminosity. Shocks, however, can raise the density up to a few hundred (Clark et al. 1996). This increase in the density is not large enough that we need to worry about the suppression of the emission lines due to collisional deexcitation effects. In our "pure AGN" photoionization model, such an increase in the density will have equivalent effects to a decrease in the flux of the AGN ionizing photons by the same factor. This means that those objects where shocks have compressed the emitting gas - but not ionized it -can be interpreted in terms of a lower ionization parameter (see Eq. 1).
We have assumed a density of 10 cm-3 at the illuminated face. The behaviour of the gas is isobaric. Therefore, the density at every position in the cloud is adjusted with the temperature to keep the pressure equilibrium. The abundances are assumed to be solar unless another value is specified. The clouds are considered to be radiation bounded (except in Sect. 6.2) and with plane parallel geometry (see VMBF96 for a more detailed description).
5.1.1. Power Law models, =-1.5
We have first investigated whether -1.5 power-law (PL) photoionization models, which are so successful for low-z radio galaxies, are also able to explain the UV line ratios of our sample of high redshift objects. The results are presented in the diagnostic diagrams in Fig. 1. The -1.5 PL sequence is represented on the left diagrams by solid circles connected with a solid line.
A quick look to the left diagrams shows that the -1.5 sequence lies far away from the data points. However, comparing the shape of the model sequence and the trend defined by the data, the similarity suggests that the data sequence can be explained in terms of the variation in the ionization parameter, as is observed at low redshift.
The influence of geometrical effects
VMBF96 showed that -1.5 PL models can explain the sequence defined by the objects in the Ly /CIV vs. CIV/CIII] diagnostic diagram when geometrical effects are taken into account. The resonant character of the CIV 1550 and Ly lines makes the escape of the line photons very asymmetric, so that their emission is strongly dependent on the distribution of material inside (and outside, for Ly ) the ionized cones. Provided that the axis of extended emission line nebulosity is not in the plane of the sky, the spectrum we observe is emitted by a mixture of clouds observed from different viewing angles: the clouds further from the observer are seen preferentially from the illuminated face, while the clouds closer to the observer are seen from the rear. In this situation resonant lines are emitted very differently on the two sides of the nucleus; they appear stronger with respect to the other lines on the side which lies further from the observer, and fainter on the other side.
The previous sequences (Fig. 1, left) did not consider the geometrical effects described above: they represent a situation in which geometrical effects due to different orientation angles are cancelled.
In Fig.1 (right) we show the same diagnostic diagrams as before, but geometry is considered. The single PL sequence in the left diagrams has been replaced by two new ones: the long dashed line represents models for clouds observed directly from the illuminated face - this describes the case in which the spectrum of the gas on the far side of the source is dominant - while the dotted line corresponds to the opposite case in which we observe the clouds from the rear (i.e. the clouds on the near side dominate). Any intermediate case is described by a sequence of models intermediate between these.
The back perspective does not help at all to improve the fitting to the data because the resonant CIV line intensity is decreased relative to the other lines. For the "front" sequence, the line ratios involving CIV increase slightly, but not enough to explain the relative faintness of HeII. Moreover, CIII]/HeII is still a problem: geometrical effects do not affect this line ratio and cannot explain its high observed values.
The conclusion is that photoionization by a power law of index -1.5, which reasonably reproduces the optical line ratios of low redshift radio galaxies cannot explain the observed UV line ratios of HZRG.
5.1.2. Hot black body models
A very hot black body (T 130,000 K) provides an even better fit to the optical line ratios of low redshift radio galaxies than the -1.5 power law usually assumed (Robinson et al. 1987). We have studied the predictions of these models in the UV spectral range. The results are presented in Fig. 1 (left pannels). Contrary to the conclusion obtained in the optical range, such a continuum cannot explain the position of the objects in the UV diagnostic diagrams. The fit is worse than the one produced by the -1.5 PL models - the T 130,000 K black body model falling even further from the data points. The dispersion produced by geometrical effects is similar to the -1.5 PL models. We do not present these models here for simplicity.
Therefore, the models which are successful at explaining the low redshift optical spectra do not work in the UV spectral range for very high redshift objects. An alternative possibility is that the ionizing continuum emitted by the central AGN has a different shape at low and high redshifts. We investigate now the effects that a harder continuum have on the UV line ratios.
5.1.3. A hard ionizing continuum: Power law with index =-1.0
The new models, with a power law of index =-1.0, are shown in the diagnostic diagrams in Fig. 2. As in Fig. 1, the diagrams on the left do not take into account geometrical effects, which, on the contrary, are considered on the right plots. The agreement with the data is excellent. Most of the data points are rather well described by the =-1.0 U-sequence.
The influence of geometrical effects
When considering geometrical effects with the =-1.0 models (Fig. 2, right), most objects lie between the two sequences which represent the back and the front perspective, as expected. For a fixed U we find that, while CIII] /HeII is the same for back and front, the dispersion predicted by the models for different viewing angles is large when line ratios involving CIV are considered and the ionization level of the gas is high. This trend is also suggested by the data: the models seem to envelope a very similar area to the one occupied by the data. Note that the -1.0 PL models can explain the data points in the CIV/Ly vs. CIV/CIII] diagnostic diagram as well as the 1.5PL models.
In summary, a power law of index =-1.0, where the sequence is defined by the variation of the ionization level of the gas, can explain the observed UV line ratios of HZRG.
If this result is confirmed, it indicates an evolution of the spectral shape of the ionizing continuum emitted by the central AGN with redshift or luminosity (we are biased to very powerful objects).
There are already some indications of a correlation between the hardness of the AGN continuum and redshift. O'Brien et al. (1988) found that quasars of higher redshift show harder UV continua. Francis (1993) concluded that high redshift (z=2) AGNs have intrinsically harder mean continuum slopes ( 0.8) than low redshift AGNs. He also concludes that this is a correlation with redshift and not with absolute magnitude. It is important to mention that these authors study a spectral range longward (in wavelength) of the Lyman limit and, therefore, they do not include the UV ionizing continuum of interest to us in this paper. However, extrapolating their results to shorter wavelengths, the ionizing continuum should also get harder with increasing redshifts.
5.2. The shock models
Another strong possibility to explain the discrepancies between the -1.5 PL (and hot black body) models and the observed UV line ratios is shocks. A high velocity radiative shock can influence the emission line processes through two different mechanisms: a) the generation of a strong local UV photon field in the hot post-shock zone, which can ionize the surrounding medium both upstream and downstream; b) line emission during the radiative cooling of gas behind the hot post-shock zone. We have used the new shock models by Dopita and Sutherland (1995) which take into account both effects. The two main parameters which influence the predicted spectrum are the velocity of the shock and the magnetic parameter defined as G.cm-2, where B is the pre-shock magnetic transverse field and n is the pre-shock density. Note that we have assumed an ionized helium recombination ratio of HeII 1640/He 7.7 in order to calculate the HeII 1640 strength from the HeII 4686 strengths published in Dopita & Sutherland (1995). Our experience with photoionization models shows that the value of the HeII 1640/He recombination ratio is insensitive to the physical conditions and the ionization state of the gas.
Shock model sequences are presented in the diagnostic diagrams in Fig. 3. The diagrams on the left show the UV line ratios of the cooling region (post-shock material). The velocity varies between 150 and 500 km s-1 and the magnetic parameter between 0 and 4 µG cm . The density adopted is n(H)=1cm-3. The diagrams on the right combine the post-shock and the precursor gas emission, that is, material which has not entered the shock but is ionized by its UV photon flux. The velocity is varied (200-500 km s-1) with a fixed magnetic parameter (1 µG cm ).
The agreement between the shock and precursor models with the data is very poor. For the range of models which reproduce the observed range of CIV/HeII values, CIII] is too faint, both with respect CIV and HeII, so that the sequences cannot cover the area occupied by the objects. This can be said both for the post-shock and post-shock + precursor models.
Another problem evident from the diagrams is the difficulty of the models at reproducing the sequences defined by the data. In contrast, the trends on the diagnostic diagrams are naturally explained in the AGN photoionization context by a sequence in the ionization parameter U.
The conclusion is that the published shock models cannot explain the observed UV line ratios. However, shock models are complex and we cannot rule out the possibility that a combination of parameters will be found that reproduces the observed spectra.
© European Southern Observatory (ESO) 1997
Online publication: June 5, 1998