New constraints on the average escape fraction of Lyman continuum radiation in

Context. Determining the average fraction of Lyman continuum (LyC) photons escaping high redshift galaxies is essential for understanding how reionization proceeded in the z> 6 Universe. Aims. We want to measure the LyC signal from a sample of sources in the Chandra Deep Field South (CDFS) and COSMOS fields for which ultra-deep VIMOS spectroscopy as well as multi-wavelength Hubble Space Telescope (HST) imaging are available. Methods. We select a sample of 46 galaxies at z ~ 4 from the VIMOS Ultra Deep Survey (VUDS) database, such that the VUDS spectra contain the LyC part, that is, the rest-frame range 880−910 Å. Taking advantage of the HST imaging, we apply a careful cleaning procedure and reject all the sources showing nearby clumps with different colours, that could potentially be lower-redshift interlopers. After this procedure, the sample is reduced to 33 galaxies. We measure the ratio between ionizing flux (LyC at 895 Å) and non-ionizing emission (at ~ 1500 Å) for all individual sources. We also produce a normalized stacked spectrum of all sources. Results. Assuming an intrinsic average Lν(1470) /Lν(895) of 3, we estimate the individual and average relative escape fraction. We do not detect ionizing radiation from any individual source, although we identify a possible LyC emitter with very high Lyα equivalent width (EW). From the stacked spectrum and assuming a mean transmissivity for the sample, we measure a relative escape fraction f escrel = 0.09 ± 0.04. We also look for correlations between the limits in the LyC flux and source properties and find a tentative correlation between LyC flux and the EW of the Lyα emission line. Conclusions. Our results imply that the LyC flux emitted by V = 25−26 star-forming galaxies at z ~ 4 is at most very modest, in agreement with previous upper limits from studies based on broad and narrow band imaging