Discussion of our results

Next we turn to the results of our observations of the three sources. In PKS 1353-341 the deepest absorption is in a broad feature (with possibly a narrow core) centered at -125 km s^-1 with respect to the optical emission line redshift. A shallower feature is centered at around +75 km s^-1. The overall absorption extends from -330 km s^-1 to +170 km s^-1.

PKS 1353-341 is a radio source with uncertain structural information. According to the VLBI data of Preston et al. (1985) at 2.29 GHz, most of the source is resolved at 3mas (20 pc) resolution (correlated flux <0.13 Jy, total flux 1.5$\pm$0.4 Jy). However their total flux is substantially above the value of 0.64 Jy at 2.7 GHz in the Parkes catalogue. If this were to indicate variability, the source would have to be rather small; if not, the upper limit on the visibility is increased to 0.2 which would still leave open what structure there may be at 3 mas resolution. In any case the flat spectrum suggests a compact source.

The absorbing column density N$_{\rm H I}$(100)=21 10²⁰ cm^-2 if the source were fully covered but could be higher if not. The large velocity width suggests that the gas is located at modest distances from the nucleus (<100 pc?) and could very well form part of the absorbing torus of the unified model of AGN.

In PKS 1814-637 most of the absorption is in a deep component with a FWHM of only 45 km s^-1 (in the rest frame) and N$_{\rm H I}$(100)=7.4 10²⁰ cm^-2. Its redshift is the same as our adopted emission line redshift but, as discussed before, the latter is very uncertain. An underlying broad asymmetric feature with an optical depth of about 0.009 extends from about -190 km s^-1 to +180 km s^-1 with respect to the centroid of the narrow line.

PKS 1814-637 at 2.3 GHz is a moderately compact double with a separation of about 300 mas (550 pc) (Tzioumis et al. 1996a), but with about half of the total flux in more diffuse structures. Without spectral information it is not clear which source in the VLBI map, if any, is the nucleus; if it is the source in between the two extreme components and if at 1.4 GHz the distribution is not too different, the deep absorption cannot be only in front of the nucleus since dilution effects would then give a lower optical depth; the broad component could be. Because of the large separation of the components, it is rather clear that the narrow absorption is likely to occur at a sizeable distance from the nucleus (> a few hundred pc).

In PKS 1934-638 only a narrow absorption line was detected with FWHM of some 18 km s^-1 and very low optical depth centered at +260 km s^-1 with respect to the optical emission line redshift of Morganti & Dickson (1999), but at -50 km s^-1 with respect to the earlier result of Fosbury et al. (1987).

PKS 1934-638 is a compact double with a separation of 42 mas (168 pc) and components about 2.5 mas (10 pc) across, but with some 40% of the total flux in a more diffuse uncertain structure between the components (Tzioumis et al. 1989,1996b). Even at 8.4 GHz, there is no evidence for a radio nucleus. In view of its narrow profile it is probable that the H I absorption covers only part of the source, but which part is unknown.

The precise location and distribution of the absorbing gas is difficult to determine. In the unified model it is generally assumed that the gas is in some torus around the nucleus which is geometrically thick. In Barthel's (1989) unification model the vertical thickness of the torus must be about equal to its radius to obtain the observed percentage of radio galaxies and quasars. If the torus is thick because of random motions of gas clouds, these must be comparable in magnitude to the rotational motions and if these are isotropic, the line widths of the absorption lines should give an indication of the latter.

Since the Broad Line Region is undoubtely of sub-pc scale, it has frequently been assumed that the radius of the torus is of the order of a pc. While there are some indications that this may be the case in typical Seyfert galaxies, there is accumulating evidence that the radii in strong radio galaxies are larger. A detailed discussion of the torus in Cyg A led to the conclusion (Conway & Blanco 1995; Maloney 1996) that it should be at least 50 pc from the nuclear X-ray source to avoid a large optical depth in free-free absorption at 1.4 GHz which would make the radio nucleus invisible. Of course, the "torus" may be a misnomer and it is likely that the gas is located in a more extended disk with a surface density and thickness as a function of radial distance to the nucleus still largely unknown, but to be determined by observation. In such a disk the larger velocities are likely to be found closer to the nucleus. This is confirmed by the VLBI data of Taylor et al. (1999) on PKS 2322-123 in which the nucleus is absorbed by H I with a FWHM velocity spread of 735 km s^-1, while one of the jet components 34 pc away has absorption with a FWHM of only 133 km s^-1. Interestingly Taylor et al. also find such a relatively narrow component in the nuclear absorption but with a five times lower N$_{\rm H}$ than in the jet component. This may be indicative of an absorbing disk in which the thickness increases at larger radii.

In the case of 4C 31.04 a compact double with separation of some 120 pc, Conway (1996a,b) found H I absorption which fully covers one component, but only part of the other. The FWHM was 133 km s^-1. Conway interpreted this as due to a disk 100 pc thick and with inner radius about 100 pc.

It seems possible that PKS 1814-637 is similar to 4C 31.04. With a FWHM of only 45 km s^-1, the absorption might be inferred to occur rather far out in the disk which would be expected in any case because of the 550 pc separation of the components. With such a low velocity dispersion the disk would be too thin to cover both components, and it seems likely that the one component covered would have twice the average optical depth. The inferred value of N$_{\rm H I}$ is comparable to that in 4C 31.04. It would be interesting to confirm by VLBI measurements that only one of the components is absorbed. Some uncertainty is introduced in this discussion by the fact that not all the flux comes from the two components. As we noted before the shallow broad absorption could be from material in front of the faint nucleus.

Some quantitative estimates may be made about the kind of torus or disk responsible for the absorption in PKS 1353-341. The nature of such a torus depends on the column density of matter, the local pressure and the intensity of the X-ray irradiation. Depending on these, such a torus could be atomic or mainly molecular (Neufeld et al. 1994; Maloney et al. 1996). The simplest case is the atomic torus. Here the temperature is typically 8000K or somewhat less and the degree of ionization of the order of 0.03. Two important effects need be considered: the radiative excitation of the spin levels by the non-thermal continuum of the radio source may change the spin temperature of the H I and the ionized component of the torus may become optically thick at 1.4 GHz.

With regards to the former, according to the results of Field (1958) and Bahcall & Ekers (1969) the spin temperature (in the absence of Ly $\alpha$ excitation which in the torus should be negligible on account of the high dust opacity) may be written as $\rm T_{\rm S}=\frac{\rm T_{\rm R}+\rm y_{\rm C} \rm T_{\rm K}}{1+\rm y_{\rm C}}$with T$_{\rm R}$ the radiation temperature, T$_{\rm K}$ the kinetic temperature and y$_{\rm C}$=2100 n$_{\rm H I}$ T$_{\rm K}^{-0.8}$ where electron de-excitation has been included (accounting for 1/4 of the value of y$_{\rm C}$) assuming n$_{\rm e}$=0.03 n(HI)$_{\rm H I}$ and T$_{\rm K}$ of the order of 8000 K, but its slightly different temperature dependence neglected. The factor y$_{\rm C}$ is proportional to the collisional de-excitation rate and the formula given assumes only collisions with H atoms and electrons. If the H₂ abundance is significant, the expression for y$_{\rm C}$ may well have to be changed, but to our knowledge the appropriate cross section has never been calculated. If n$_{\rm H}\gg$1cm^-3 and T$_{\rm K}<10^{4}$K, we have y$_{\rm C} \gg$1. Replacing then 1 + y$_{\rm C}$by y$_{\rm C}$ and inserting the appropriate expression for T$_{\rm R}$ we may write T$_{\rm S}$=F T$_{\rm K}$ with

$$\rm F=0.57 \times 10^{-9} (1+z) (\rm D/r)^{2} \rm S_{1.4/(1+z)} n_{\rm HI}^{-1} T_{\rm K}^{-0.2} + 1.$$

(D is the luminosity distance and r the distance of the absorber from the radio emitter). Writing 10⁶n$_{\rm H I}$=n₆, taking T$_{\rm K}$=8000K and inserting parameters appropriate for PKS 1353-341, we obtain F=126 n₆^-1 r$_{\rm pc}^{-2}+1$.

From Spitzer (1978), the free free optical depth may be written as $\tau _{\rm ff}$=0.106 n$_{\rm e}^{2}$ T^-3/2 $\nu ^{-2}$ d, where we have evaluated the Gaunt factor for T=10⁴ and $\nu$=1GHz and neglected its logarithmic dependence on T and $\nu$ for slightly different values, and where d is the path length. Taking n_e=0.03 n$_{\rm H I}$, T$_{\rm K}$=8000 K, $\nu$=1.41GHz and writing the column density as N₂₄=10^-24 n d, we obtain $\tau _{\rm ff}$=67 n₆ N₂₄.

For PKS 1353-341 we have found N$_{\rm H I}$(100)=21 10²⁰ cm^-2 or in an atomic torus with T$_{\rm K}$=8000 K, N₂₄= 0.17 F cm^-2. Inserting the expression for F we may write this as n₆ N₂₄=21 r$_{\rm pc}^{-2}$+0.17 n₆. The condition that the source have a free free optical depth less than about unity to be observable then becomes 1420 r$_{\rm pc}^{-2}$+ 11.4 n₆<1, from which it follows that in any case r>38 pc and n₆<0.09 cm^-3. Hence such a torus must have a radius much larger than that assumed for Seyfert galaxies and a pressure n T < 7 10⁸ K cm^-3, compatible with the pressures found in the Narrow Line Region. The mass of the atomic gas then becomes at least a few times 10⁷ M$_{\odot}$.Similar conclusions were previously reached for the torus around Cyg A (Conway & Blanco 1995; Maloney 1996).

If the torus were molecular, it should still have an atomic layer with at 10 GHz a free free optical depth given by approximately $\tau$=0.5 F$_{\rm X5}^{1.1}$ p₁₁^-0.1 with F$_{\rm X5}$ the hard X-ray flux in units of 10⁵ erg cm^-2 s^-1 and P₁₁ the pressure (defined as n T) in units of 10¹¹ K cm^-3 (Neufeld et al. 1994). The value of P₁₁ is not very critical and we take P₁₁=0.01. For the optical depth at 1.4 GHz we may then write $\tau _{\rm ff}$=200 L₄₄^1.1 r$_{\rm pc}^{-2.2}$ with L₄₄ the 2-100 keV X-ray luminosity in units of 10⁴⁴ erg s^-1.

PKS 1353-341 has been identified with a weak ROSAT source (RX J13560-3420, Brinkman et al. 1994) with a 0.1-2 keV flux of 1.5 10^-12 erg cm^-2 s^-1, corresponding to a soft X-ray luminosity of 3.2 10⁴⁴ erg s^-1. It is unclear if this is associated with the radio source or with hot gas in the cD galaxy and associated cluster. If we estimate the hard X-ray flux from the absorption corrected [O III] flux, we would obtain L₄₄ $\approx$3, evidently a very uncertain value. Inserting this into the expression for $\tau _{\rm ff}$, the condition $\tau _{\rm ff}<1$ would become 670 r$_{\rm pc}^{-2.2}<1$ or r>19 pc. Thus the conclusion is confirmed: the torus or the disk in which the H I absorption takes place in PKS 1353-341 should have a radius of tens of pc. With similar conclusions for Cyg A (Conway & Blanco 1995) and probably also for S5 1946+708 (Peck et al. 1999), it seems that this may well be a general result for strong radio galaxies.

It would be of particular interest to determine if the H I in strong radio galaxies is falling in towards the nucleus to feed the central engine or if it flows out as a result of energetic processes there. Both processes may occur in the same object as seen, for example, in the Seyfert 2 galaxy IC 5063 (Morganti et al. 1998).

However the question is not easy to answer because the systemic velocities of radio galaxies tend to be uncertain. It was noted by Mirabel & Wilson (1984) that in a sample of 20 nearby Seyfert galaxies the difference between the optical emission line velocities and the systemic velocities (determined from extended H I emission disks) ranged from more than -200 km s^-1 to +50 km s^-1 with an average of -74 km s^-1 (blueshift). In more remote strong radio galaxies such effects may well be still larger, and the only way to obtain reliable systemic redshifts is to measure stellar absorption lines. We have attempted to do this for the three galaxies we have studied, but because of poor weather conditions no sufficiently reliable results were obtained.

From the notes to table 3, it appears that the only strong radio galaxy for which infall has been demonstrated is Cyg A. This is an illustrative case, because Conway & Blanco (1995) had found a blueshifted component at -149 km s^-1, which became redshifted at +110 km s^-1 when a reliable systemic redshift was obtained. The value of the H I absorption data would be much increased if more reliable systemic redshifts were obtained.