Astron. Astrophys. 333, L75-L78 (1998)

4. Discussion

At a level of just a percent of the strongest low excitation lines, a variety of possible excitation mechanisms for the [O iv] line must be considered.

4.1. Weak AGNs

Because of the great strength of [O iv] emission in Seyfert galaxies, quite faint and perhaps obscured AGNs embedded in a more luminous starburst would contribute sufficient [O iv], with the additional constraint that their narrow line width would have to be small. In fact, hard X-ray observations may indicate AGNs deeply hidden in some of our sources. The case is convincing for NGC 4945 (Iwasawa et al. 1993 ), but less so for M 82 (Tsuru et al. 1997 ). For M 82 , definite proof against an AGN origin of [O iv] is provided, however, by the fact that the emitting region is spatially resolved and similar in size to the starburst region. If it were illuminated by a central source, sufficient [O iv] could not be produced without exceeding the observed relatively low [Ne iii]/[Ne ii] ratio.

This constraint is illustrated in Fig. 4, which shows predicted line ratios for a simple AGN photoionization model with varying ionization parameter computed using CLOUDY (Ferland 1996 ). When [O iv] reaches 1% of the low-excitation neon lines, [Ne iii]/[Ne ii] is already much too high to be consistent with starbursts like M 82 ([Ne iii]/[Ne ii] 0.17, Förster-Schreiber et al., in preparation). This is a fairly general problem in any photoionisation scenario, which persists if one adds a small hard component to a soft starburst spectrum (as discussed below in the context of hot stars). The way to circumvent this problem - postulate small region(s) with very strong [O iv] but small contribution to the total [Ne ii] - is not viable here since this would imply a central small NLR which is inconsistent with the observations of M 82 . For individual galaxies with [O iv] detections but lacking spatial information, a weak central AGN remains possible. However, we emphasize that the fairly uniform level of the [O iv] detections (Fig. 2) requires an unlikely finetuning of AGN and starburst activity to fit our sample as a whole.

Fig. 4. Photoionization models for [O iv]/([Ne ii]+0.44 [Ne iii]) (continuous) and [Ne iii]/[Ne ii] (dashed) as function of ionization parameter in an AGN narrow line region

4.2. Super-hot stars

The ionization edge for creation of [O iv] is just beyond the He ii edge; at higher energies the spectral energy distributions of most stars drop precipitously. However, a small component of hotter (e.g. Wolf-Rayet) stars might provide the necessary high energy photons. We have run a photoionization model for an H ii region excited by a 39000 K main sequence star (represented by a Kurucz model atmosphere), plus an additional 80000 K blackbody to represent a harder component. The blackbody is an ad-hoc choice selected for ease of implementation; however, other strong sources of photons beyond 54 eV would give similar results. As Fig. 5 shows, this attempt fails to explain [O iv] in low-excitation starbursts since the predicted [Ne iii]/[Ne ii] ratio ( 0.5 to 1) exceeds the observations ( 0.1) when [O iv] reaches 1% of the low-excitation lines. For the high-excitation starbursting dwarfs, such a discrepancy does not arise, and hot stars remain an option.

Fig. 5. Photoionization models for [O iv]/([Ne ii]+0.44 [Ne iii]) (continuous) and [Ne iii]/[Ne ii] (dashed) in an H ii region excited by a 39000 K star, with addition of an 80000 K blackbody contributing different fractions of the total Lyman continuum luminosity

Again, the inconsistency could be alleviated if small H ii regions with relatively stronger [O iv] emission were dispersed in a lower excitation background. In fact, such a scenario is qualitatively consistent with the observations, as are others with distributed local sources of [O iv]. The major reason to consider it unlikely is that we have failed up to now to detect [O iv] emission even at a similar level in local star forming regions, while we would have to postulate regions with stronger emission. The Galactic center, which is closest to starburst galaxies in many aspects, still shows [O iv], though even fainter than in the starbursts (Lutz et al. 1996b ). In the massive star forming regions W51 IRS2 and 30 Doradus , for which [Ne iii]/[Ne ii] indicates high excitation, we were unable to detect [O iv] at a level of 0.01 and 0.005 of [Ne ii]+0.44[Ne iii], respectively (Thornley et al., in preparation).

4.3. Planetary nebulae

High excitation planetary nebulae are a known source of [O iv] emission. A young starburst will, of course, not contain planetary nebulae and it is easy to show that their integrated contribution from the old stellar population is too faint. Evolutionary calculations (e.g. Charlot & Bruzual 1991 ) show that the contribution of post-AGB stages to the bolometric luminosity is less than 1% even in old populations. Making the extreme assumptions that 10% of the bolometric luminosity is due to an old population and that all PAGB objects are like NGC 7027 , one of the highest excitation planetary nebulae, we estimate a robust upper limit of 10^-20 W cm^-2 for the [O iv] emission from planetary nebulae in M 82 , based on [O iv] flux, luminosity and distance of NGC 7027 as given by Shure et al. (1983 ) and Beintema et al. (1996 ).

4.4. Ionizing shocks

There is ample evidence for ionizing shocks in starburst galaxies. Spatially extended, 'Liner'-type optical emission lines can be attributed to shocks, and kinematic mapping sometimes provides direct evidence for outflowing 'superwinds' (Heckman et al. 1990 ). [O iv] column densities approaching 10 cm^-2 are expected for modest velocity shocks (100-200 km/s, e.g. Shull & McKee 1979 , Dopita & Sutherland 1996 ). Assuming postshock values of n=1000 cm^-3 and T=50000 K, we estimate a 25.90 µm intensity of erg s^-1 cm^-2 sr^-1, equivalent to W cm^-2 for the SWS beam. For the assumed conditions, the covering factor of such shocks in the starburst region of M 82 would have to be of the order unity. At higher shock velocities, the [O iv] column would be increasingly dominated by material 'at rest' in the photoionized precursor in the material ahead of the shock front (Dopita & Sutherland 1996 ). The shock models predict that intensities similar to those estimated for [O iv] are emitted in optical shock tracers like [S ii] 6716/31Å. This is fully consistent with optical spectroscopy of M 82 (e.g. Götz et al. 1990 ). We note that the faint shock emission predicted for the [Ne ii] and [Ne iii] lines will be completely dominated by the emission from H ii regions.

It is instructive to compare the [O iv] results for M 82 with the SWS observations for RCW 103 , a bright supernova remnant interacting with a dense molecular cloud (Oliva et al., in preparation). The [O iv] intensities are very similar. The RCW 103 ionic lines are just resolved at the SWS spectral resolving power, again similar to M 82 . Ionizing shocks hence are a plausible origin for the M 82 [O iv] emission if their total covering factor approaches unity in the central starburst region of M 82 .

© European Southern Observatory (ESO) 1998

Online publication: April 28, 1998

helpdesk@link.springer.de