In interacting galaxies, strong tidal forces disturb the global morphology of the progenitors and give birth to the long stellar, gaseous and dusty tails often observed. In addition to this destructive effect, tidal forces can morph into a transient, protective setting called compressive mode. Such modes then shelter the matter in their midst by increasing its gravitational binding energy. This thesis focuses on the study of this poorly known regime by quantifying its properties thanks to numerical and analytical tools applied to a spectacular merging system of two galaxies, commonly known as the Antennae galaxies. N-body simulations of this pair yield compressive modes in the regions where observations reveal a burst of star formation. Furthermore, characteristic time- and energy scales of these modes match well those of self-gravitating substructures such as star clusters and tidal dwarf galaxies. These results suggest that the compressive modes of tidal fields plays an important role in the formation and evolution of young clusters, at least in a statistical sense, over a lapse of ~10 million years. Preliminary results from simulations of stellar associations highlight the importance of embedding the clusters in the evolving background galaxies to account precisely for their morphology and internal evolution.

We present a candidate for the most distant galaxy known to date with a photometric redshift z = 10.7 +0.6 / -0.4 (95% confidence limits; with z < 9.5 galaxies of known types ruled out at 7.2-sigma). This J-dropout Lyman Break Galaxy, named MACS0647-JD, was discovered as part of the Cluster Lensing and Supernova survey with Hubble (CLASH). We observe three magnified images of this galaxy due to strong gravitational lensing by the galaxy cluster MACSJ0647.7+7015 at z = 0.591. The images are magnified by factors of ~8, 7, and 2, with the brighter two observed at ~26th magnitude AB (~0.15 uJy) in the WFC3/IR F160W filter (~1.4 - 1.7 um) where they are detected at >~ 12-sigma. All three images are also confidently detected at >~ 6-sigma in F140W (~1.2 - 1.6 um), dropping out of detection from 15 lower wavelength HST filters (~0.2 - 1.4 um), and lacking bright detections in Spitzer/IRAC 3.6um and 4.5um imaging (~3.2 - 5.0 um). We rule out a broad range of possible lower redshift interlopers, including some previously published as high redshift candidates. Our high redshift conclusion is more conservative than if we had neglected a Bayesian photometric redshift prior. Given CLASH observations of 17 high mass clusters to date, our discoveries of MACS0647-JD at z ~ 10.8 and MACS1149-JD1 at z ~ 9.6 are consistent with a lensed luminosity function extrapolated from lower redshifts. This would suggest that low luminosity galaxies could have reionized the universe. However given the significant uncertainties based on only two galaxies, we cannot yet rule out the sharp drop off in number counts at z >~ 10 suggested by field searches.