6 Summary and conclusions

We have presented the cross-correlation between mid-IR and X-ray observations in the Lockman Hole- and HDF-N-centred regions. ISOCAM and XMM-Newton observed a common region of more than 200 square arcminutes in the Lockman Hole. A total of 24 galaxies out of 76 XMM-Newton sources in this field show mid-IR emission. In particular, the percentage of hard X-ray sources with 15 $\mu$m emission is around 60%. On the other hand, only around 10% of the mid-IR sources show X-ray emission in the different XMM-Newton bands. Deep Chandra observations (Brandt et al. 2001a) completely cover the ISOCAM observations of the HDF-N and flanking fields. In a region of 24 square arcminutes, 25% of the mid-IR sources have been detected in the X-ray for a total of 24 sources. A comparison of the Lockman Hole, HDF-N and Elais-S1 surveys (Alexander et al. 2001) shows that these surveys are compatible in terms of source density taking into account their respective detection limits. While the HDF-N survey is so sensitive to the detection of even normal galaxies and the Elais-S1 survey detects only very powerful and rare type-1 AGNs, the Lockman Hole survey is able to detect a population of galaxies whose emission is mostly dominated by AGNs of types 1 and 2. In particular, thanks to the increased sensitivity of XMM-Newton with respect to ROSAT, nearly half of the sources with mid-IR emission are new XMM-Newton sources. Most of the sources which are optically studied are type-1 AGNs. Relying on optical colours and X-ray hardness ratio diagrams, we conclude that about 70% of the detected sources are type-2 AGNs. Nevertheless, XMM-Newton observations are not deep enough to detect all the obscured AGNs in the sample as are the Chandra observations in the HDF-N. Only forthcoming observations of the Lockman Hole with XMM-Newton will be able to detect the population of faint obscured X-ray sources visible with Chandra in the HDF-N. We have studied how the mid-IR to hard X-ray index ( $\alpha _{\rm IX}$) varies with redshift, comparing it with the expected behaviour from local templates. Most of the Lockman sources detected lie in a region of the diagram occupied by type-2 AGN local templates and several are optically classified as type-1 AGN. Since in general these sources do not appear highly extincted using X-ray hardness diagnostic diagrams, a possible explanation is that emission from the host galaxies contributes a fraction of their IR-optical emission greater than that of local templates.

Finally, we have evaluated how much the integrated emission of AGN contributes to the total extragalactic mid-IR background light using two independent methods. A direct estimation gives a percentage of $(15\pm5)\%$ for the Lockman survey ( $0.5<F_{\rm 15\,\mu m}<3$ mJy) and a value of $(18\pm7)\%$ for the HDF-N survey ( $0.1<F_{15\,{\rm\mu m}}<0.5$ mJy), hence a contribution of $(17\pm6)\%$ in the interval of fluxes $0.1<F_{\rm 15\,\mu m}<3$ mJy. Considering median mid-IR to X-ray spectral indices for two hard-X flux ranges, we estimate that the population of AGNs making up $\sim$85% of the 2-10 keV X-ray background contribute $(17\pm 2)\%$ of the mid-IR extragalactic background.

This fraction could be higher if there exists a population of AGNs that is highly obscured at X-ray wavelengths. In particular, since the X-ray background peaks at 30-40 keV while we can now observe only up to 5-10 keV, we expect that deeper X-ray observations (over a wider spectral range) will unveil more highly extincted AGNs. Considering that IR spectra of typical AGNs peak around 20 $\mu$m while starburst spectra peak at significantly longer infrared wavelengths, this figure may be considered as an upper limit to the AGN contribution to the cosmic IR background (CIRB) energy density.

We conclude that the bulk of the mid-IR extragalactic emission comes from star formation and that the luminous galaxies seen by ISOCAM in the deep surveys are essentially starbursts obscured by dust. Nevertheless, this result does not exclude the possibility that the majority of the galaxies in the Universe have both AGN and starforming contributions. The results obtained by Elbaz et al. (2002) about the origin of the infrared background light based on a set of observed correlations and by Flores et al. (1999) evaluating the mid-IR part of the star formation are not significantly affected by the AGN contribution to the mid-IR extragalactic light.

Acknowledgements

F. D. dedicates this work to the memory of his professor Giuliano Giuricin, recently deceased, who introduced him to the study of AGNs. F. D. acknowledges support from the network ISO SURVEY set up by the European Commission under contract ERBFNRXCT960068 or its TMR program. H. F. was supported by a grant of the "Académie de la science''. We thank the referee D. M. Alexander for his careful reading of the manuscript, interesting comments and suggestions which greatly improved the paper. We are grateful to H. Aussel for providing us with his flux list before publication. We also thank M. Arnaud, I. Perez-Fournon and F. La Franca for fruitful discussions and interesting suggestions.