In this thesis we investigate the redshift distribution and derived cosmological
properties of Extremely Red Galaxies (ERGs), through the study of a sample
selected with Ks ≤ 22 and (I₇₇₅ - Ks) > 3.92 (Vega) in 50.4 arcmin² of the
GOODS/CDFS field. We also study the properties of the total parent sample
of Ks-selected galaxies in the same field, with the aim of understanding their
evolution and clarifying the role of ERGs within this population. We determine
that the bright end of the Ks-band luminosity function, which is progressively
well reproduced by the ERGs with increasing redshift, shows no sign of decline
up to at least redshift z ~ 2.5. We also explore the evolution of massive systems
present in our sample: ~ 20% - 25% of the population of local galaxies with
assembled stellar mass M > 1 x 10¹¹M⨀ were formed before redshift z ~ 4, and
contain ~ 45% to 70% of the stellar mass density of the Universe at that redshift.
Within our sample, the comoving number density of these massive systems is
then essentially constant down to redshift z ~ 1.5. The remaining massive
systems observed in the local Universe are assembled later, at redshifts z < 1.5.
Thus, a two-fold assembly history for massive galaxies is suggested, in which
galaxy/star formation proceeds very efficiently in high mass haloes at very high
redshift. It is the massive ERGs at redshift z > 1 which contain the imprints of
the most efficient period of galaxy formation. Finally, from the morphological
study of our galaxies with estimated mass M > 10¹¹M⨀ and redshifts z < 2, we
explore the bridge between massive ERGs at z > 1 and local massive galaxies.
~ 50% - 60% of all the massive galaxies at 0.5 ≲ z ≲ 2.0 have surface brightness
profiles close to a de Vaucouleurs law. All our results are consistent with a
scenario in which the most massive ERGs are the progenitors of local cluster