The faint radio sky: VLBA observations of the COSMOS field^⋆

¹ Astronomisches Institut, Ruhr-Universität Bochum, Universitätstrasse 150, 44801 Bochum, Germany
e-mail: herrera@astro.rub.de
² The Netherlands Institute for Radio Astronomy (ASTRON), 7991 Dwingeloo, The Netherlands
³ Centre for Astrophysics and Supercomputing, Swinburne University of Technology, PO Box 218, Hawthorn, VIC 3122, Australia
⁴ CSIRO Australia Telescope National Facility, PO Box 76, Epping, NSW 1710, Australia
⁵ Western Sydney University, Locked Bag 1797, Penrith South, NSW 1797, Australia
⁶ SUPA, Institute for Astronomy, Royal Observatory Edinburgh, Blackford Hill, Edinburgh EH9 3HJ, UK
⁷ National Radio Astronomy Observatory, PO Box O, Socorro, NM 87801, USA
⁸ MPI for Astronomy, Königstuhl 17, 69117 Heidelberg, Germany
⁹ Department of Physics, Faculty of Science, University of Zagreb, Bijenička cesta 32, 10000 Zagreb, Croatia
¹⁰ California Institute of Technology, MC 249-17, 1200 East California Boulevard, Pasadena, CA 91125, USA

Received: 12 May 2017
Accepted: 13 July 2017

Abstract

Context. Quantifying the fraction of active galactic nuclei (AGN) in the faint radio population and understanding their relation with star-forming activity are fundamental to studies of galaxy evolution. Very long baseline interferometry (VLBI) observations are able to identify AGN above relatively low redshifts (z> 0.1) since they provide milli-arcsecond resolution.

Aims. We have created an AGN catalogue from 2865 known radio sources observed in the Cosmic Evolution Survey (COSMOS) field, which has exceptional multi-wavelength coverage. With this catalogue we intend to study the faint radio sky with statistically relevant numbers and to analyse the AGN – host galaxy co-evolution, making use of the large amount of ancillary data available in the field.

Methods. Wide-field VLBI observations were made of all known radio sources in the COSMOS field at 1.4 GHz to measure the AGN fraction, in particular in the faint radio population. We describe in detail the observations, data calibration, source detection and flux density measurements, parts of which we have developed for this survey. The combination of number of sources, sensitivity, and area covered with this project are unprecedented.

Results. We have detected 468 radio sources, expected to be AGN, with the Very Long Baseline Array (VLBA). This is, to date, the largest sample assembled of VLBI detected sources in the sub-mJy regime. The input sample was taken from previous observations with the Very Large Array (VLA). We present the catalogue with additional optical, infrared and X-ray information.

Conclusions. We find a detection fraction of 20 ± 1%, considering only those sources from the input catalogue which were in principle detectable with the VLBA (2361). As a function of the VLA flux density, the detection fraction is higher for higher flux densities, since at high flux densities a source could be detected even if the VLBI core accounts for a small percentage of the total flux density. As a function of redshift, we see no evolution of the detection fraction over the redshift range 0.5 <z< 3. In addition, we find that faint radio sources typically have a greater fraction of their radio luminosity in a compact core – ~70% of the sub-mJy sources detected with the VLBA have more than half of their total radio luminosity in a VLBI-scale component, whereas this is true for only ~30% of the sources that are brighter than 10 mJy. This suggests that fainter radio sources differ intrinsically from brighter ones. Across our entire sample, we find the predominant morphological classification of the host galaxies of the VLBA detected sources to be early type (57%), although this varies with redshift and at z> 1.5 we find that spiral galaxies become the most prevalent (48%). The number of detections is high enough to study the faint radio population with statistically significant numbers. We demonstrate that wide-field VLBI observations, together with new calibration methods such as multi-source self-calibration and mosaicing, result in information which is difficult or impossible to obtain otherwise.