Astron. Astrophys. 343, 943-952 (1999)

4. Discussion

4.1. Comparison with shock models

Explicit predictions for a few FIR lines have been included in the models for relatively slow (150 km s^-1) shock interacting with low density material by Raymond (1979) and Shull & McKee (1979). The first lists [NeII], [SiII] and [FeII] while the latter include [SiII], [SIII] and [SIV]. The predicted line ratios from the above mentioned models are not in good agreement with our results. In particular, the observed [SIV]/[SIII] ratio is a factor 5 larger than the computed values. Also, most models predict a factor of 3 too strong [SiII] (relative to [NeII], [FeII], [SIII]), but this could be attributed to uncertainties in the atomic parameters of SiII which have been updated several times since the publication of the shock model results. The most important discrepancy, however, is that the predicted surface brightnesses are always a factor 10 lower than the observed values, which simply reflects the fact that the shock of RCW103 is much faster than the values used in the above models (cf. the Introduction and below).

Models of slow shocks interacting with very dense gas ( cm^-3) were developed by Hollenbach & McKee (1989) who also include explicit predictions for all the FIR lines of singly ionized and neutral species. The main problem with these models is that they span preshock densities much larger than the 300 cm^-3 required to account for the measured electron densities in the post-shock region. Consequently, lines with low critical densities, e.g. [SiII], are predicted too faint. Moreover, the models underestimate the flux of [FeII] lines by a factor of 5.

The most recent models of DS96, which cover shock velocities of 200-500 km s^-1 and are more representative of the conditions of RCW103, do not however give explicit predictions for the FIR lines. Nevertheless, reasonably accurate line ratios can be computed from the values of ionic column densities and mean temperatures listed in the above paper. To a first approximation, the ratio of two lines from the post-shock region is

where N and T are the column densities and temperature of the emitting ions and j is the line emission coefficient which, for FIR lines, is very little dependent on the gas temperature. The contribution from the photoionized precursor can be also computed from the published tables of N and T in the pre-shock region, while the compression factor (i.e. the ratio between the electron density in the post and pre shock regions) can be estimated imposing that the regions have similar surface brightnesses (cf. Sect. 3.2 of DS96), i.e.

Out of the many FIR line ratios we have identified those which are most sensitive to the shock speed and to the presence of the photoionized precursor. The behaviour of the selected line ratios is plotted in Fig. 10 where the most remarkable result is that models including the emission from the precursor largely overpredict the strength of [SIV] and [NeV]. The velocity dependence of the [NeV]/[NeIII] ratio in the post-shock region may appear at first sight surprising, but can be easily understood as follows. The post-shock [NeV] emission always occurs in collisionally ionized gas at K and whose column density, which primarily depends on the shape of the gas cooling curve, does not strongly vary with the shock speed. The [NeIII] line, on the contrary, could be strongly enhanced by emission from photoionized gas in the post-shock region, but this only occurs at km s^-1 while slower shocks do not produce enough ionizing photons to support a large NeIII zone. In short, the sharp decrease of [NeV]/[NeIII] between 200 and 300 km s^-1 is because the [NeIII] line emission rapidly increases in this velocity interval.

Fig. 10. Behaviour of velocity and precursor sensitive line ratios. The theoretical values are computed from the shock models of DS96 as described in Sect. 4.1. B= is the magnetic parameter (cf. DS96) and the dashed regions show the observed values.

The main conclusion of this analysis is that all ionic lines can be reasonably well reproduced by post-shock emission. This conclusion also agrees with the broad line profiles observed by SWS (Sect. 2.1) and imaging-spectroscopy observations of [OIII]5007 which show complex dynamical structures, similar to those seen in [FeII] (Fig. 5) and incompatible with emission from the precursor (Moorwood et al. 1987). The only ionic line which cannot be accounted for by post-shock emission is [SIV] which should be a factor of 10 fainter, but can be reproduced adding a quite `incomplete precursor', i.e. 5% the SIV column density of the precursor predicted by the DS96 models (cf. Fig. 10).

Given the importance that the photoionized precursor may have in modelling the spectra of active galaxy nuclei (Sect. 4.3), it is of interest to investigate why little or no ionic line emission is observed from the precursor of RCW103. We envisage the following possibilities.

• The precursor in RCW103 is very thin to UV ionizing photons, but this is very difficult to reconcile with the fact that H₂ emission is observed from pre-shock molecular gas lying outside of the shock front (cf. Fig. 7). For H₂ to exist, the molecules must be shielded from the strong field of UV ionizing radiation from the shock front or, equivalently, the preshock region must be optically thick.

• The shock front in RCW103 is significantly slower than so far assumed and below 150 km s^-1, the minimum speed required to produce a prominent phototoionized precursor. This is in strict constrast with the observed line widths and filament dynamics (cf. Sect. 2 and Fig. 5). Moreover, slow shocks cannot explain the very large surface brightness of the lines which require a large mechanical power of the shock, i.e. a large product, n being the preshock density and the shock speed. More specifically, the average surface brightness of Br within the ISO beam corresponds to = erg cm^-2 s^{- 1} sr^-1 which, coupled to the predicted values from shock models (Eq. 3.4 of DS96), yields

  

  or, equivalently, a shock speed of about km s^-1 for a pre-shock density of 300 cm^-3. Larger pre-shock densities are effectively excluded by the measured electron densities (Table 2) in the post-shock region, i.e. after the gas has been compressed by the shock front. The factor takes into account projection effects such as those modelled in details by Hester (1987) who interpreted the bright filaments in IC443 and Cygnus-Loop in therms of relatively slow shocks seen quasi edge-on and found that small filaments amplified by a factor 10-100 should be quite common. However, this model cannot hold for RCW103 for the following reasons. This remnant is much brighter (factor of 10) than IC443 and Cygnus-Loop. The average surface brightness within the relatively large ISO-SWS beam (i.e. the value used in Eq. 3) is already a factor of 4 lower than that observed on arcsec scales in optical/IR line images of RCW103. The most largely amplified edge-on filaments should have small radial velocities (FWHM40 km, cf. Fig. 2 of Hester 1987) amd this is not compatible with the observed line widths and dynamics.

• The shock models largely overpredict the contribution of the photoionized precursor. Indeed, DS96state that the column density of ionized gas in the precursor might be overestimated due to a possibly incorrect treatment of the transfer of the UV ionizing photons (cf. end of Sect. 4.2 of DS96). Moreover, the ionization structure of the precursor could be much different than computed in DS96if the shock evolves on time scales shorter than 100 yr, i.e. the recombination time in the pre-shock gas.

4.2. Comparison with the Galactic center

The region on the line of sight of the GC has a rich spectrum of prominent IR lines which are believed to arise from gas with an unusually large Fe gas phase abundance and which is primarily photoionized by quite hot stars (Lutz et al. 1996). Table 4 is a comparison between the most significant line ratios measured in RCW103 and in the GC.

Table 4. Comparison between RCW103 (SNR), the Galactic center and the Circinus galaxy.
Notes:
(1) Line fluxes from this paper
(2) Data from Lutz et al. (1996)
(3) Data from Moorwood et al. (1996b)

The [FeIII]22.9/[FeII]26.0 ratio is much higher (a factor of 42) in the GC spectrum. This implies Fe/Fe⁺ 1 and a factor 10 larger than in RCW103, regardless of the assumed gas density in the GC. This simply reflects the fact that a region predominantly photoionized by stars, such as those near to the GC, contains only a relatively small fraction of partially ionized gas. The recombining region behind the SNR shock front, on the contrary, has a large zone of partially ionized gas, which is heated by photoionization from the shock front radiation, and where most of iron is forced into Fe⁺ by the very rapid charge exchange reactions with neutral hydrogen.

The [OIV]25.9/[FeII]26.0 ratio is the same in the two objects, within the errors. Given the difficulties to produce both FeII and OIV with photoionization from normal stars, it seems not unreasonable to conclude that both species are primarily produced by shock excited gas in the line of sight of the GC.

The [NeIII]15.6/[NeII]12.8 ratio is a factor of 20 lower in the GC than in RCW103 which indicates that fast shocks are more effective than late O stars in producing NeIII. Moreover, the [NeIII]15.6/[FeII]26.0 ratio is only a factor of 2.6 higher in the GC than in RCW103 and this indicates that a non negligible fraction of the [NeIII] emission from the GC could come from shock excited gas.

4.3. Photoionized precursor and shocks in active galaxy nuclei

According to the shock models of DS96, the precursor could be an important source of lines from high ionization species (e.g. [OIII]5007), but its importance relative to the post-shock region may strongly depend on the column density of the pre-shock material. In a paper specifically dedicated to study the spectral signatures of shocks in active galaxies, Dopita & Sutherland (1995) consider the following limiting cases:

• Shock only, in which the precursor is very thin and its emission is effectively negligible relative to the post-shock region. This can fairly well reproduce the line ratios observed in low excitation AGNs (LINERS).

• Shock + precursor, where the pre-shock region is opaque to the ionizing photons from the shock front. Since the ionizing spectrum is quite hard and effectively similar to a typical AGN spectrum, the ionization structure of the precursor is similar to that of standard narrow line regions photoionized by the AGN. Consequently, the emerging line spectrum is similar to that of standard photoionization models and could explain, therefore, the high excitation lines from e.g. type 2 Seyferts.

In view of this proposed scenario, it is interesting to compare the spectra of RCW103 with that of the Circinus galaxy, an archetype Seyfert 2 galaxy whose observed line ratios are listed in Table 4. The most striking difference is that the high excitation (coronal) lines are much stronger in Circinus with, in particular, [NeV]/[NeIII]=1 and roughly 2 orders of magnitude larger than in RCW103. Such a strong [NeV] could be in principle compatible with emission from the precursor of a km s^-1 shock (cf. Fig. 10), while even higher velocities, i.e. 1000 km s^-1, could probably account for highest ionization coronal lines (e.g. [SiIX]). The main problem is that these shocks should also emit prominent low excitation lines from their fast moving post-shock gas, but this is incompatible with the observed line profiles which are remarkably narrow (FWHM150 km s^-1, Oliva et al. 1994) and similar for all ionization species. Therefore, a shock dominated model for the Circinus galaxy seems very unlikely and, more generally, the role played by the photoionized precursor in Seyferts could be questioned on the basis of the following arguments.

If dominated by photoionization, the low excitation lines from the post-shock region (e.g. [SII]) should be broader than those from the photoionized precursor (e.g. [OIII]), but this is in strict contrast with the observations which show that [OIII] and higher excitation lines are usually broader than those of [SII] and lower excitation species.

The ISM medium of Seyfert galaxies is well known to be quite "porous", especially within the ionization cones, and several arguments indicate that the line emitting clouds are probably density bounded (e.g. Binette et al. 1996). The host galaxies of LINERS, on the contrary, are often very rich in both gas and dust, a spectacular example being NGC4945 (e.g. Moorwood et al. 1996a). It seems therefore curious that the shocks in Seyferts should primarily impact onto the relatively few large clouds (i.e. those with large enough column density to absorb all the ionizing photons from the shock) while, in LINERS, the shocks should selectively avoid the largest clouds and only hit regions with low column densities (i.e. those which cannot produce a bright precursor).

The absence of significant emission from the pre-shock region in RCW103 indicates that shock models may overestimate the importance of the precursor region.

© European Southern Observatory (ESO) 1999

Online publication: March 1, 1999
helpdesk.link@springer.de