3 Uncertainties on interpreting the kinematic properties of the extended gas in HzRG

The strongest lines in the optical (observed) spectra of HzRG are Ly$\alpha$, CIV$\lambda$1550, HeII$\lambda$1640 and CIII]$\lambda$1909, while [OII]$\lambda$3727, [OIII] $\lambda\lambda$5007, 4959, H$\alpha$dominate in the NIR. There are some physical effects that might make the interpretation of the line profiles and velocity shifts uncertain.

-

Presence of multiple kinematic components;

-

Broad scattered lines;

-

Resonant scattering of Ly$\alpha$ and (maybe) CIV photons.

3.1 Multiple kinematic components

We mentioned in Sect. 1 that the interaction between the radio jet and the ambient gas could be responsible for the extreme motions in HzRG. Other authors favour a gravitational origin for the kinematics of the gas. One interpretation or another has important consequences: if the velocity fields reflect the gravitational potential, the derived dynamical mass turns out to be correlated with redshift and/or radio power (Baum & McCarthy 2000). As the authors pointed out, this result is not valid if shocks are responsible for the kinematics.

Villar-Martín et al. (1999) carried out a detailed spectroscopic study of the intermediate redshift radio galaxy PKS 2250-51 (z=0.31) where a strong interaction between the radio and optical structures occurs. The authors resolved two main kinematic components, spatially extended and detected in all optical (rest frame) emission lines. One of the components is narrow ( $\sim 150-200$ km s^-1), the line ratios are consistent with photoionization and it extends beyond the radio structures. The other component is broad (FWHM as large as 900 km s^-1at some spatial positions), the line ratios are consistent with shock ionization and it is emitted inside the radio structures. The properties of the narrow component suggest that it is emitted by ambient photoionized gas that has not been perturbed, while the properties of the broad component are consistent with shocked gas.

Figure 2: Velocity curves of Ly$\alpha$, CIV and HeII emission lines in 4C40.36 (z=2.27) (top panels) and B3 0731+438 (z=2.43) (bottom panels). The spatial zero has been defined at the position of the centroid of continuum emission in both cases. These velocity curves suggest rotation of the ionized gas

A similar spectroscopic study could prove or disprove shocks as responsible for the kinematics in HzRG. The studies done so far have been based on the emission line kinematics along the radio axis and they have not provided a definitive answer. We propose that the joint study of the line kinematics (with spectral decomposition of the line profiles) and the line ratios may give the answer. If shocks perturb the kinematics in HzRG, then we should expect similar components and with similar flux ratios as those observed in PKS 2250-41. With 2D spectrographs it will be possible to extend the kinematic studies to regions far from the radio axis, where the interactions are expected to be non existent. The detection of broad lines far from the radio axis would confirm that jet/cloud interactions are not responsible for the extreme motions.

The spectral resolution we use is crucial to be able to isolate different kinematic components in the extended gas. A velocity resolution of $\sim 250$ km s^-1 ( $R=\lambda /\Delta\lambda \sim 1200$) would be ideal since it will allow an accurate decomposition of the emission line profiles. The instrumental profile (IP) is in this case well matched with the expected FWHM of the non perturbed gas (so that we avoid unnecessary instrumental broadening), such as the diffuse haloes extending beyond the radio structures. The study of such haloes can be very useful, since they show the gas properties before any perturbation. With the new $1024\times1024$ NIR arrays it should be possible to cover the interesting spectral range (HeII$\lambda$4686 to [OIII]$\lambda$5007) for a good number of objects.

3.1.1 Apparent rotation curves

The presence of several kinematic components can lead us to derive false rotation curves.

Figure 2 shows two examples of HzRG where the emission lines show a resemblance with rotation curves (see also Figs. 1 and 2 in Villar-Martín et al. 1999). This could be an exciting evidence for merger events (e.g. Hernquist 1993). However, the apparent rotation is an artifact, consequence of the presence of at least two kinematic components. To illustrate this, we have used the spectrum of the radio galaxy B3 0731+438 (Fig. 1). We have extracted 1-D spectra from those spatial pixels where the abrupt jump in the apparent rotation curve occurs. Figure 3 shows the Ly$\alpha$ spectral profile at each pixel.

Figure 3: Spatial variation of the spectral profile of Ly$\alpha$ in B3 0731+438. Each curve is the Ly$\alpha$spectral profile (redshifted) at a given spatial position (pixel). Two kinematic components are present whose relative contributions change from pixel to pixel

The profile changes dramatically due to the presence of at least two kinematic components that contribute with different relative intensity at different pixels. Figure 1 shows clearly the two main components. The fact that they are apparent also in HeII proves that it is not an effect of resonant scattering of Ly$\alpha$ and CIV. This makes the Ly$\alpha$ velocity centroid change in space, giving the appearance of a rotation curve. If we isolate the two components at every spatial position the rotation curve disappears (see Fig. 4).

Figure 4: Comparison between the results obtained by fitting 1 and 2 Gaussians to the spectral profile of the lines. The individual kinematic components are shown as open triangles and the result of fitting a single Gaussian is shown as open circles. The gradient in the velocity curve is due to the change on the contribution of one component relative to the other. The "apparent" rotation curve is an artifact consequence of the blend of both components in some pixels

3.2 Broad scattered lines

Many HzRG are highly polarized in the optical due to the scattering of nuclear emission (both continuum and emission lines from the broad line region) by (probably) dust in the extended gas (e.g. Cimatti et al. 1997, 1998; Fosbury et al. 1999). The radiation from the extended gas is therefore a mixture of direct and scattered light. Can the scattered broad lines affect our conclusions on the kinematic studies if neglected?

As an example we present the spectrum of TXS0211-122 (z=2.34), one of the most highly polarized ( $P\sim 20\%$ longward Ly$\alpha$) radio galaxies at high redshift. We have analysed the profile of CIV, which is efficiently emitted in the broad line region and therefore is subject to scattering. Figure 5 shows the spectrum in the region of CIV. There is an underlying broad component with $FWHM \sim 4000$ km s^-1. Similar FWHM are often observed in high redshift quasars suggesting that it is scattered light.

The fit shows that the CIV line profile is not seriously affected by the underlying broad component due to the prominence of the "narrow" emission. This is usually the case (also for Ly$\alpha$) and the studies that have found $FWHM\geq1000$ km s^-1in the extended gas of HzRG are not affected by the effects of scattered light (McCarthy et al. 1996; Villar-Martín et al. 1999). However, we cannot neglect this contribution when studying the line profiles in more detail (looking for different kinematic components, for instance); we would interpret the broad scattered components as due to extreme motions in the extended gas.

Figure 5: TXS0211-122 is one the most highly polarized HzRG. Although broad scattered CIV seems to be present, it does not affect the narrow component, which dominates the emission. Right panel: fit and data. Left panel: individual components of the fit and data

The best way to avoid any possible uncertainty on this issue is to compare with lines that are not emitted efficiently in the BLR (and, therefore, they are not scattered), such as HeII$\lambda$1640 (Foltz et al. 1998; Heckman et al. 1991) or optical forbidden lines such as [OII] and [OIII]$\lambda$5007, 4959 (the [OIII] lines might have a minor contribution from scattered light, di Serego Alighieri et al. 1997).

3.3 Resonant scattering of Ly$\alpha$ and CIV

Ly$\alpha$ and CIV$\lambda$1550 are resonant lines and intervening HI and CIV respectively can absorb the emission. Ly$\alpha$ is particularly sensitive to this effect. The discovery of absorption troughs in the profile of the line across the whole extension of the Ly$\alpha$ emitting gas in many HzRGs has proved that Ly$\alpha$ absorption is a common phenomenon (van Ojik et al. 1997). CIV absorption troughs have aslo been observed (e.g. Binette et al. 2000). Van Ojik et al. found a correlation between the radio size and the occurrence of absorption, suggesting that in radio galaxies with large radio sources the effect is less worrying.

Absorption of Ly$\alpha$ photons can modify dramatically the profile of the line (van Ojik et al. 1997). Röttgering et al. (1997) concluded that the effects of associated HI absorption may be responsible for the shift of the Ly$\alpha$ line with respect to the high ionization lines in some objects. By using non resonant lines (HeII$\lambda$1640, forbidden lines) we will avoid this problem. The CIV doublet, although resonant, is less sensitive and more reliable than Ly$\alpha$. The use of high spectral resolution ($\sim3$ Å) can also help, since we will be able to resolve the absorption troughs (van Ojik et al. 1997).

3.4 The line doublets

Figure 6: The inclined solid lines indicate the variation with redshift of the observed wavelength for several important emission lines. The hatched regions show the spectral coverage of the atmospheric bands where the absorption is more than 50% of the incoming radiation. The redshifts for which one of the emission lines lies in these dark regions can be easily deduced fromthe plot

The comparison between the FWHM of the emission lines provides information about the processes responsible for the line emission and the kinematics. The detection of different line velocity widths for different lines will lead us to the conclusion that different mechanisms are at work and/or there are regions of different physical conditions. As an example, some radio galaxies with radio/optical interactions show that low ionization lines are broader than high ionization lines (Clark et al. 1998). This has been interpreted as a consequence of shocks: the shocked gas has lower ionization level and more perturbed kinematics than the non perturbed gas.

In this sense, care must be taken with the doublets (CIV$\lambda$1550, CIII]$\lambda$1909 and [OII]$\lambda$3727). At high redshifts the separation between the two components is large (7 Å at z=2.5 and 10 Å at z=4) and the doublet as a whole can show a profile apparently broader than single lines. This will be the case when the lines are similar in (observed) width or narrower than the doublet separation. If the lines are much broader the two components will be severely blended, whatever the resolution we use and the broadening will not be important. This should often be the case since gas motions can produce lines of intrinsic $FWHM \geq20$ Å at z=2.5.