Counting individual galaxies from deep 24 micron Spitzer surveys: beyond the confusion limit

We address the question of how to deal with confusion limited surveys in the mid-infrared domain by using informations from higher frequency observations over the same sky regions. Such informations, once applied to apparently extended mid-infrared sources, which are indeed ``blends'' of two or more different sources, allow us to disentangle the single counterparts and to split the measured flux density into different components. We present the application of this method to the 24 micron Spitzer archival data in the GOODS-EN1 test field, where apparently extended, ``blended'' sources constitute about 20% of a reliable sample of 983 sources detected above the 5-sigma threshold down to 23 microJy. As higher frequency data-set we have considered the public IRAC images and catalogues on the same field. We show that the 24 micron sample is almost unbiased down to ~40 microJy and the careful application of the deblending procedure does not require any statistical completeness correction (at least at the flux level considered). This is probed by direct comparison of our results with those of Chary et al. (2004), who analysed the same data-set through extensive Monte Carlo simulations. The deblending procedure reduces of about 30% the confusion limit of the MIPS 24 micron survey, allowing one to obtain reliable source counts down to ~40 microJy. The extrapolation of the source counts down to fainter fluxes suggests that our 24 micron sample is able to resolve ~62% of the cosmic background down to a flux level of 38 microJy.
Comment: 10 pages, submitted to MNRAS on July 2005