2007MNRAS.377..203G

We investigate claims that recent ultra-deep X-ray surveys are detecting starbursts at cosmologically interesting redshifts (z = 0-1). We combine X-ray data from the 2Ms Chandra Deep Field-North and multiwavelength observations obtained as part of the Great Observatories Origins Deep Survey (GOODS)-North to build the Spectral Energy Distributions (UV, optical, infrared) of X-ray sources in this field. These are fit with model templates providing an estimate of the total infrared luminosity (3-1000µm) of each source. We then exploit the tight correlation between infrared and X-ray luminosities for star-forming galaxies, established in the local Universe, to select sources that are dominated by star formation rather than supermassive black hole accretion. This approach is efficient in discriminating normal galaxies from active galactic nuclei (AGN) over a wide range of star formation rates, from quiescent systems to starbursts. The above methodology results in a sample of 45 X-ray selected star-forming systems at a median redshift z ~ 0.5, the majority of which (60 per cent) are either luminous or ultraluminous infrared galaxies. This sample is least affected by incompleteness and residual AGN contamination and is therefore well suited for cosmological studies. We quantify the X-ray evolution of these sources by constructing their differential X-ray counts, dN/dS, and comparing them with evolving luminosity function models. The results are consistent with luminosity evolution of the form (1 + z)^p with p ~ 2.4. This is similar to the evolution rate of star-forming galaxies selected at other wavelengths, suggesting that the deep X-ray surveys, like the Chandra Deep Field-North, are indeed finding the starburst galaxy population that drives the rapid evolution of the global star formation rate density in the range z ~ 0-1. Our analysis also reveals a separate population of infrared-faint X-ray sources at moderate z. These include old galaxies but also systems that are X-ray luminous for their stellar mass compared to local ellipticals. We argue that these may be post-starbursts that will, over time, become fainter at X-ray wavelengths and will eventually evolve into early-type systems (i.e. E/S0).