Far beyond stacking: Fully Bayesian constraints on sub-μJy radio source populations over the XMM-LSS-VIDEO field

Abstract

Measuring radio source counts is critical for characterizing new extragalactic populations, brings a wealth of science within reach and will inform forecasts for SKA and its pathfinders. Yet there is currently great debate (and few measurements) about the behaviour of the 1.4-GHz counts in the μJy regime. One way to push the counts to these levels is via 'stacking', the covariance of a map with a catalogue at higher resolution and (often) a different wavelength. For the first time, we cast stacking in a fully Bayesian framework, applying it to (i) the Square Kilometre Array Design Study (SKADS) simulation and (ii) Very Large Array (VLA) data stacked at the positions of sources from the VISTA Infra-red Deep Extragalactic Observations (VIDEO) survey. In the former case, the algorithm recovers the counts correctly when applied to the catalogue, but is biased high when confusion comes into play. This needs to be accounted for in the analysis of data from any relatively low-resolution Square Kilometre Array (SKA) pathfinders. For the latter case, the observed radio source counts remain flat below the 5-σ level of 85 μJy as far as 40 μJy, then fall offearlier than the flux hinted at by the SKADS simulations and a recent P(D) analysis (which is the only other measurement from the literature at these flux-density levels, itself extrapolated in frequency). Division into galaxy type via spectral-energy distribution reveals that normal spiral galaxies dominate the counts at these fluxes.