Abstract

In order to derive the AGN contribution to the cosmological ionizing emissivity we have selected faint AGN candidates at $z>4$ in the CANDELS GOODS-South field which is one of the deepest fields with extensive multiwavelength coverage from Chandra, HST, Spitzer and various groundbased telescopes. We have adopted a relatively novel criterion. As a first step high redshift galaxies are selected in the NIR $H$ band down to very faint levels ($H\leq27$) using reliable photometric redshifts. This corresponds at $z>4$ to a selection criterion based on the galaxy rest-frame UV flux. AGN candidates are then picked up from this parent sample if they show X-ray fluxes above a threshold of $F_X\sim 1.5\times 10^{-17}$ cgs (0.5-2 keV). We have found 22 AGN candidates at $z>4$ and we have derived the first estimate of the UV luminosity function in the redshift interval $4<z<6.5$ and absolute magnitude interval $-22.5\lesssim M_{1450} \lesssim -18.5$ typical of local Seyfert galaxies. The faint end of the derived luminosity function is about two/four magnitudes fainter at $z\sim 4-6$ than that derived from previous UV surveys. We have then estimated ionizing emissivities and hydrogen photoionization rates in the same redshift interval under reasonable assumptions and after discussion of possible caveats, the most important being the large uncertainties involved in the estimate of photometric redshift for sources with featureless, almost power-law SEDs and/or low average escape fraction of ionizing photons from the AGN host galaxies. We argue that, under reasonable evaluations of possible biases, the probed AGN population can produce at $z=4-6.5$ photoionization rates consistent with that required to keep highly ionized the intergalactic medium observed in the Lyman-$\alpha$ forest of high redshift QSO spectra, providing an important contribution to the cosmic reionization.