IRAS 23128-5919

4 Discussion and concluding remarks

 

A wealth of observational data available has shown that ULIRGs have high concentrations of gas and dust in their nuclei, sufficient to account for most of their observed infrared luminosity (see Sanders & Mirabel 1996 for a review). Whether the energy source of ULIRGs is a dust enshrouded AGN or a starburst still remains an open issue. However, recent indirect evidence is beginning to favour the existence of bright extremely red point-like sources in the nuclear regions of ULIRGs. More specifically near-infrared observations of luminous infrared galaxies have shown that their flux at 2.2 $\mu$m is more concentrated towards the center than at 1.3 $\mu$m (Carico et al. 1990b; Scoville et al. 2000). Furthermore, recent high resolution MIR observations using Keck of a sample of ULIRGs reveal compact sub-arcsecond sources (with linear scales of $\sim$100-300 pc) which contain 30% to 100% of the observed MIR energy of these galaxies (Soifer et al. 2000). This contrasts with the LIRGs ( $10^{11}~L_{\odot}\leq L_{\rm IR} \leq 10^{12}$ $L_{\odot}$), in which the infrared energy seems to be generated over somewhat larger scales ($\sim$100 pc-1 kpc, Soifer et al. 2001) and sometimes can be found in extra-nuclear regions associated with the physical interaction of merging pairs of galaxies. Furthermore, there are galaxies such as VV 114 where it has even been found that a substantial fraction of the MIR flux originates from an extended component of hot dust emission spread over several kpc scales (Soifer et al. 2001; Le Floc'h et al. 2002). ULIRGs are thus not simply a scaled-up version of LIRGs and require further dynamical compression of the molecular gas responsible for the IR luminosity within very compact regions. A plausible mechanism would be one where the shocks and tidal forces of the interaction first lead to star formation over galactic scales, leading to IR luminosities up to a few 10¹¹ $L_{\odot}$. Subsequently, gravitational instabilities and the formation of a bar, strip the gas of its angular momentum, funneling large quantities towards the nuclear regions of galaxies, which can feed circumnuclear starbursts or AGNs and trigger the ultraluminous phase in the infrared (Combes 2001).

Even though the above scenario is appealing, given the high extinction in the nuclei of ULIRGs, the limited atmospheric transmission in the MIR windows, and the limited sensitivity of ground-based instruments, questions related to the direct probing of the nuclear activity such as "does all MIR emission from those systems originate from the nuclei?'' and if not "what are the spectral properties of any extended component?'' still remain unanswered.

This is where the superb sensitivity of space instruments, such as ISO, is essential. We have found that in the ULIRGs studied here more than $\sim$95% of the MIR emission seen by IRAS is confined within a few arcsecs of their central region. Obviously the relatively large pixel size of the ISOCAM detector places limitations in interpreting these findings. However, deconvolution tests of the central point source in each galaxy suggest that the corresponding nuclei are resolved and the physical diameter of the emitting region is contained within 1 to 2 kpc. Moreover, with the exception of the Superantennae where the MIR spectrum is dominated by the emission arising from the AGN of the southern galaxy, the bulk of the IR luminosity of IRAS 23128-5919 and IRAS 14348-1447 is powered by massive star formation. The fact that starbursts can dominate the MIR emission in galaxies with IR luminosities as high $\sim$10¹² $L_{\odot}$ had already been demonstrated in other ISOCAM-CVF (Tran et al. 2001) and ISO-SWS (Genzel et al. 1998) observations of ULIRGs, and is supported by our results. Given that an active nucleus appears to be always present in the most energetic objects of the local Universe (Lutz et al. 1998), our MIR data favor a luminosity threshold for the transition between starburst- and AGN-dominated galaxies which is higher than the IR luminosity of the galaxies in our sample. This is in agreement with the results of Tran et al. (2001) who proposed that this transition takes place at $L_{\rm IR}\sim10^{12.5}$  $L_{\odot}$ and also found individual starbursts up to 10^12.65  $L_{\odot}$. Our data also indicate is that such starbursts can be confined to the very central nuclear regions which may have important consequences in the probing how the instabilities fuel the inner regions of galaxies (e.g. Combes 2001), as well as determining the nature of high redshift dusty sources (e.g. Ivison et al. 2000).

Another striking feature revealed in our observations is that in all three cases one galaxy seems to dominate the MIR energy output of the system by more than 75%. Could this be a record of the initial distribution of the amount of molecular gas available in each merging progenitor or could this suggest that in the later stages of interaction, the gas finally merges towards one component? If the latter were true one would expect that a sufficiently large quantity of gas could trigger and fuel both circumnuclear star forming activity and AGN-type activity at the core of a single object. This is evident in the southern galaxy of IRAS 19254-7245 which harbors an active nucleus as well as numerous massive star forming regions. As we mentioned in the introduction though the presence of a Seyfert nucleus is correlated with a MIR flux increase relative to the FIR luminosity of the entire galaxy, which is what one can actually derive from our observations when we compare the Superantennae with IRAS 14348-1447. IRAS 14348-1447 has indeed a much higher total IR luminosity despite its MIR flux being lower than that of the southern source of IRAS 19254-7245. Furthermore, using the $f_{15~\mu\rm m}/\rm H\alpha$ and $f_{15~\mu\rm m}/K$ ratios as probes of dust absorption and hot dust emission normalized to the mass of the galaxy, we find that in each interacting system it is always the most active galaxy of the system that exhibits the higher ratios. In each system, the most luminous galaxy contains a larger amount of molecular gas leading to the triggering/feeding of the starburst activity and/or an active nucleus.

Finally, we wish to stress once more that because of the limited spatial resolution in studying such distant sources, the diagnostics we have used in this paper address only the integrated MIR emission of each galaxy. Our difficulties to identify whether an active nucleus is solely responsible for the increase in the MIR luminosity relative to the FIR emission will not be resolved unless we can either clearly map the extent of the emitting region or obtain MIR spectra using very narrow slits. The upcoming launch of SIRTF which, despite the fact it has comparable spatial resolution to ISO, is equipped with a new generation of detectors of smaller pixel size, and in particular the use of its infrared spectrograph will help us improve upon our current results and provide conclusive answers to the issues which still remain unresolved to date.

Acknowledgements

We wish to thank P.-A Duc for providing his K band images of IRAS 19254-7245 and IRAS 23128-5919 as well as D. Rigopoulou for providing the ISOPHOT data. We greatly appreciated the comments of the referee which helped us improve the manuscript. VC would like to acknowledge the partial support of JPL contract 960803.

IRAS 23128-5919