2010MNRAS.404.1672V

We investigate the possibility of contamination by lower redshift interlopers in the measure of the ionizing radiation escaping from high-redshift galaxies. Taking advantage of the new ultradeep Very Large Telescope/Visible Multiobject Spectrograph U-band number counts in the Great Observatories Origins Deep Survey (GOODS)-South field, we calculate the expected probability of contamination by low-redshift interlopers as a function of the U magnitude and the image spatial resolution (point spread function). Assuming that ground-based imaging or spectroscopy cannot resolve objects lying within a 0.5-arcsec radius of each other, then each z ≳ 3 galaxy has a 2.1 and 3.2 per cent chance of foreground contamination, adopting surface density U-band number counts down to 27.5 and 28.5, respectively. Those probabilities increase to 8.5 and 12.6 per cent, assuming 1.0-arcsec radius. If applied to the estimates reported in the literature at redshift ∼3 for which a Lyman continuum has been observed directly, the probability that at least one-third of them are affected by foreground contamination is larger than 50 per cent. From a Monte Carlo simulation, we estimate the median integrated contribution of foreground sources to the Lyman continuum flux (f900). Current estimations from stacked data are >2σ of the median integrated pollution by foreground sources. If the correction to the observed f900 flux is applied, the relative escape fraction decreases by a factor of ∼1.3 and 2, depending on the cases reported in literature. The spatial cross-correlation between the U-band ultradeep catalogue and a sample of galaxies at z ≳ 3.4 in the GOODS-South field produces a number of U-band detected systems fully consistent with the expected superposition statistics. Indeed, each of them shows the presence of at least one offset contaminant in the Advanced Camera for Surveys images. An exemplary case of a foreground contamination in the Hubble Ultra Deep Field at redshift 3.797 by a foreground blue compact source (U = 28.63±0.2) is reported; if observed with a low-resolution image (seeing larger than 0.5arcsec) the polluting source would mimic an observed (f1500/f900)_OBS∼ 38, erroneously ascribed to the source at higher redshift.