2012MNRAS.420..901D

Following the discovery of the first significant samples of galaxies at z > 6.5 with Wide Field Camera 3/Infra-Red (WFC3/IR) on board Hubble Space Telescope (HST), it has been claimed that the faintest high-redshift galaxies display extremely blue ultraviolet (UV) continuum slopes, with a UV power-law index β =^ ≲ -3 (where f_λ∝λ^β). Such slopes are bluer than previously reported for any other galaxy population, and are most readily explained theoretically by extinction-free, young and very low metallicity stellar populations with a high ionizing photon escape fraction. Here we undertake a critical study of the evidence for such extreme values of β, combining three new WFC3/IR-selected samples of galaxies spanning nearly two decades in UV luminosity over the redshift range z =^ ≲ 4.5-8. We explore the impact of inclusion/exclusion of less robust high-redshift candidates and use the varying depths of the samples to explore the effects of noise and selection bias at a given UV luminosity. Simple data-consistency arguments suggest that artificially blue average values of β can result when the analysis is extended into the deepest =^ ≲ 0.5 mag bin of these WFC3/IR-selected galaxy samples, regardless of the actual luminosity or redshift range probed. By confining attention to robust high-redshift galaxy candidates, with at least one 8σ detection in the WFC3/IR imaging, we find that the average value of β is consistent with < β > =-2.05 ±0.10 over the redshift range z= 5-7 and the UV absolute magnitude range -22 < M_UV,AB< - 18, and that < β > shows no significant trend with either redshift or M_UV. We create and analyse a set of simple end-to-end simulations based on the WFC3/IR+ACS Hubble Ultra Deep Field (HUDF) and Early Release Science data sets which demonstrate that a bias towards artificially low/blue average values of β is indeed `expected' when the UV slope analysis is extended towards the source detection threshold, and conclude that there is as yet no clear evidence for UV slopes significantly bluer than β =^ ≲ -2, the typical value displayed by the bluest star-forming galaxies at more modest redshifts. A robust measurement of < β > for the faintest galaxies at z =^ ≲ 7 (and indeed z =^ ≲ 8) remains a key observational goal, as it provides a fundamental test for high escape fractions from a potentially abundant source of re-ionizing photons. This goal is achievable with HST, but requires still deeper WFC3/IR imaging in the HUDF.