Limits on the LyC signal from z~3 sources with secure redshift and HST coverage in the E-CDFS field

Aim: We aim to measure the LyC signal from a sample of sources in the Chandra deep field south. We collect star-forming galaxies (SFGs) and active galactic nuclei (AGN) with accurate spectroscopic redshifts, for which Hubble Space Telescope (HST) coverage and multi-wavelength photometry are available. Method: We selected a sample of about 200 sources at z~3. Taking advantage of HST resolution, we applied a careful cleaning procedure and rejected sources showing nearby clumps with different colours, which could be lower-z interlopers. Our clean sample consisted of 86 SFGs (including 19 narrow-band selected Lya emitters) and 8 AGN (including 6 detected in X-rays). We measured the LyC flux from aperture photometry in four narrow-band filters covering wavelengths below a 912 A rest frame (3.11<z<3.53). We estimated the ratio between ionizing (LyC flux) and 1400 A non-ionizing emissions for AGN and galaxies. Results: By running population synthesis models, we assume an average intrinsic L(1400 A)/L(900 A) ratio of 5 as the representative value for our sample. With this value and an average treatment of the lines of sight of the inter-galactic medium, we estimate the LyC escape fraction relative to the intrinsic value (fesc_rel(LyC)). We do not directly detect ionizing radiation from any individual SFG, but we are able to set a 1(2)sigma upper limit of fesc_rel(LyC)<12(24)%. This result is consistent with other non-detections published in the literature. No meaningful limits can be calculated for the sub-sample of Lya emitters. We obtain one significant direct detection for an AGN at z=3.46, with fesc_rel(LyC) = (72+/-18)%. Conclusions: Our upper limit on fescrel(LyC) implies that the SFGs studied here do not present either the physical properties or the geometric conditions suitable for efficient LyC-photon escape.