Exploring the infrared/radio correlation at high redshift

We have analysed the 24-mu m properties of a radio-selected sample in the Subaru-XMM-Newton Deep Field in order to explore the behaviour of the far-infrared/radio (FIR/radio) relation at high redshifts. Statistically, the correlation is described by q(24), the ratio between the observed flux densities at 24 mu m and 1.4 GHz, respectively. Using 24-mu m data results in considerably more scatter in the correlation than previous work using data at 60-70 mu m. Nevertheless, we do observe a steady correlation as a function of redshift, up to z approximate to 3.5, suggesting its validity back to primeval times. We find q(24) = 0.30 +/- 0.56 for the observed and q(24) = 0.71 +/- 0.47 for the k-corrected radio sample, based on sources with 300 mu Jy < S(1.4 GHz) < 3.2 mJy and 24-mu m detections. A suitable k-correction given by a M82-like mid-infrared (mid-IR) template suggests no extreme silicate absorption in the bulk of our radio sample. Using thresholds in q(24) to identify radio-excess sources, we have been able to characterize the transition from radio-loud active galactic nuclei (AGN) to star-forming galaxies and radio-quiet AGN at faint (less than or similar to 1 mJy) radio-flux densities. Our results are in broad agreement with previous studies which show a dominant radio-loud AGN population at > 1 mJy. The rest-frame U- B colours of the expected radio-excess population have a redder distribution than those that follow the correlation. This is therefore a promising way to select obscured type 2 AGN, with a radio-loud nature, missed by deep X-ray observations. Spectroscopic follow-up of these sources is required to fully test this method.