Star-forming galaxies are considered to be the leading candidate sources that dominate the cosmic reionization at z > 7, and the search for analogs at moderate redshift showing Lyman continuum (LyC) leakage is currently a active line of research. We have observed a star-forming galaxy at z = 3.2 with Hubble/WFC3 in the F336W filter, corresponding to the 730-890Å rest-frame, and detect LyC emission. This galaxy is very compact and also has large Oxygen ratio [OIII]λ5007/[OII]λ3727 ( 10). No nuclear activity is revealed from optical/nearinfrared spectroscopy and deep multi-band photometry (including the 6Ms X-ray Chandra). The measured escape fraction of ionizing radiation spans the range 50-100%, depending on the IGM attenuation. The LyC emission is detected with m F336W = 27.57 ± 0.11 (S/N=10) and it is spatially unresolved, with effective radius R e < 200pc. Predictions from photoionization and radiative transfer models are in line with the properties reported here, indicating that stellar winds and supernova explosions in a nucleated star-forming region can blow cavities generating density-bounded conditions compatible with optically thin media. Irrespective to the nature of the ionizing radiation, spectral signatures of these sources over the entire electromagnetic spectrum are of central importance for their identification during the epoch of reionization, when the LyC is unobservable. Intriguingly, the Spitzer/IRAC photometric signature of intense rest-frame optical emissions ([O III]λλ4959, 5007 + Hβ) observed recently at z 7.5 − 8.5 is similar to what is observed in this galaxy. Only the James Webb Space Telescope will measure optical line ratios at z > 7 allowing a direct comparison with lower redshift LyC emitters, as reported here.

We have obtained deep ultraviolet imaging of the lensing cluster A1689 with the WFC3/UVIS camera onboard the Hubble Space Telescope in the F275W (30 orbits) and F336W (4 orbits) filters. These images are used to identify z ∼ 2 star-forming galaxies via their Lyman break, in the same manner that galaxies are typically selected at z 3. Because of the unprecedented depth of the images and the large magnification provided by the lensing cluster, we detect galaxies 100× fainter than previous surveys at this redshift. After removing all multiple images, we have 58 galaxies in our sample in the range −19.5 < M 1500 < −13 AB mag. Because the mass distribution of A1689 is well constrained, we are able to calculate the intrinsic sensitivity of the observations as a function of source plane position, allowing for accurate determinations of effective volume as a function of luminosity. We fit the faint-end slope of the luminosity function to be α = −1.74 ± 0.08, which is consistent with the values obtained for 2.5 < z < 6. Notably, there is no turnover in the luminosity function down to M 1500 = −13 AB mag. We fit the UV spectral slopes with photometry from existing Hubble optical imaging. The observed trend of increasingly redder slopes with luminosity at higher redshifts is observed in our sample, but with redder slopes at all luminosities and average reddening of E(B − V ) = 0.15 mag. We assume the stars in these galaxies are metal poor (0.2 Z ) compared to their brighter counterparts (Z ), resulting in bluer assumed intrinsic UV slopes and larger derived values for dust extinction. The total UV luminosity density at z ∼ 2 is 4.31 +0.68 −0.60 × 10 26 erg s −1 Hz −1 Mpc −3 , more than 70% of which is emitted by galaxies in the luminosity range of our sample. Finally, we determine the global star formation rate density from UV-selected galaxies at z ∼ 2 (assuming a constant dust extinction correction of 4.2 over all luminosities and a Kroupa initial mass function) of 0.148 +0.023 −0.020 M yr −1 Mpc −3 , significantly higher than previous determinations because of the additional population of fainter galaxies and the larger dust correction factors.

Narrowband imaging of the rest-frame Lyman continuum (LyC) of galaxies at z 3.1 has produced a large number of candidate LyC-emitting galaxies. These samples are contaminated by galaxies at lower redshift. To better understand LyC escape, we need an uncontaminated sample of galaxies that emit strongly in the LyC. Here we present deep Hubble Space Telescope imaging of five bright galaxies at z 3.1 that had previously been identified as candidate LyC emitters with ground-based images. The WFC3 F336W images probe the LyC of galaxies at > z 3.06 and provide an order-of-magnitude increase in spatial resolution over ground-based imaging. The non-ionizing UV images often show multiple galaxies (or components) within ~ 1 of the candidate LyC emission seen from the ground. In each case, only one of the components is emitting light in the F336W filter, which would indicate LyC escape if that component is at > z 3.06. We use Keck/NIRSPEC near-IR spectroscopy to measure redshifts of these components to distinguish LyC emitters from foreground contamination. We find that two candidates are low-redshift contaminants, one candidate had a previously misidentified redshift, and the other two cannot be confirmed as LyC emitters. The level of contamination is consistent with previous estimates. For the galaxies with > z 3.06, we derive strong s 1 limits on the relative escape fraction between 0.07 and 0.09. We still do not have a sample of definitive LyC emitters, and a much larger study of low-luminosity galaxies is required. The combination of high-resolution imaging and deep spectroscopy is critical for distinguishing LyC emitters from foreground contaminants.

We present a robust measurement of the rest-frame UV luminosity function (LF) and its evolution during the peak epoch of cosmic star formation at 1 < z < 3. We use our deep near ultraviolet imaging from WFC3/UVIS on the Hubble Space Telescope (HST) and existing ACS/WFC and WFC3/IR imaging of three lensing galaxy clusters, Abell 2744 and MACSJ0717 from the Hubble Frontier Field survey and Abell 1689. Combining deep UV imaging and high magnification from strong gravitational lensing, we use photometric redshifts to identify 780 ultra-faint galaxies with M UV < −12.5 AB mag at 1 < z < 3. From these samples, we identified 5 new, faint, multiply imaged systems in A1689. We run a Monte Carlo simulation to estimate the completeness correction and effective volume for each cluster using the latest published lensing models. We compute the rest-frame UV LF and find the best-fit faint-end slopes of α = −1.56 ± 0.04, α = −1.72 ± 0.04 and α = −1.94 ± 0.06 at 1.0 < z < 1.6, 1.6 < z < 2.2 and 2.2 < z < 3.0, respectively. Our results demonstrate that the UV LF becomes steeper from z ∼ 1.3 to z ∼ 2.6 with no sign of a turnover down to M UV = −14 AB mag. We further derive the UV LFs using the Lyman break "dropout" selection and confirm the robustness of our conclusions against different selection methodologies. Because the sample sizes are so large, and extend to such faint luminosities, the statistical uncertainties are quite small, and systematic uncertainties (due to the assumed size distribution, for example), likely dominate. If we restrict our analysis to galaxies and volumes above > 50% completeness in order to minimize these systematics, we still find that the faint-end slope is steep and getting steeper with redshift, though with slightly shallower (less negative) values (α = −1.55 ± 0.06, −1.69 ± 0.07 and −1.79 ± 0.08 for z ∼ 1.3, 1.9 and 2.6, respectively). Finally, we conclude that the faint star-forming galaxies with UV magnitudes of −18.5 < M UV < −12.5 covered in this study, produce the majority (55%-60%) of the unobscured UV luminosity density at 1 < z < 3.

We present a strong-lensing analysis of MACSJ0717.5+3745 (hereafter MACS J0717), based on the full depth of the Hubble Frontier Field (HFF) observations, which brings the number of multiply imaged systems to 61, ten of which have been spectroscopically confirmed. The total number of images comprised in these systems rises to 165, compared to 48 images in 16 systems before the HFF observations. Our analysis uses a parametric mass reconstruction technique, as implemented in the L software, and the subset of the 132 most secure multiple images to constrain a mass distribution composed of four large-scale mass components (spatially aligned with the four main light concentrations) and a multitude of galaxy-scale perturbers. We find a superposition of cored isothermal mass components to provide a good fit to the observational constraints, resulting in a very shallow mass distribution for the smooth (large-scale) component. Given the implications of such a flat mass profile, we investigate whether a model composed of "peaky" non-cored mass components can also reproduce the observational constraints. We find that such a non-cored mass model reproduces the observational constraints equally well, in the sense that both models give comparable total rms. Although the total (smooth dark matter component plus galaxy-scale perturbers) mass distributions of both models are consistent, as are the integrated two-dimensional mass profiles, we find that the smooth and the galaxy-scale components are very different. We conclude that, even in the HFF era, the generic degeneracy between smooth and galaxy-scale components is not broken, in particular in such a complex galaxy cluster. Consequently, insights into the mass distribution of MACS J0717 remain limited, emphasizing the need for additional probes beyond strong lensing. Our findings also have implications for estimates of the lensing magnification. We show that the amplification difference between the two models is larger than the error associated with either model, and that this additional systematic uncertainty is approximately the difference in magnification obtained by the different groups of modelers using pre-HFF data. This uncertainty decreases the area of the image plane where we can reliably study the high-redshift Universe by 50 to 70%.