Understanding the time evolution of the star formation in the Universe is one of the main aims of observational astronomy. Since a significant portion of the UV starlight is being absorbed by dust and re-emitted in the IR, we need to understand both of those regimes to properly describe the cosmic star formation history. In UV, the depth and the resolution of the data permits calculations of the star formation rate densities out to very high redshifts (z ∼ 8 − 9). In IR however, the large beam sizes and the relatively shallow data limits these calculations to z ∼ 2. In this thesis, I explore the SMA and PdBI high-resolution follow-up of 30 bright sources originally selected by AzTEC and LABOCA instruments at 1.1 mm and 870 μm respectively in conjunction with the SCUBA-2 Cosmology Legacy Survey (S2CLS) deep COSMOS and wide UDS maps, where 106 and 283 sources were detected, with the signal-to-noise ratio of > 5 and > 3.5 at 850 μm respectively. I find that the (sub)mm-selected galaxies reside and the mean redshifts of ¯z ≃ 2.5±0.05 with the exception of the brightest sources which seem to lie at higher redshifts (¯z ≃ 3.5 ± 0.2), most likely due to the apparent correlation of the (sub)mm flux with redshift, where brighter sources tend to lie at higher redshifts. Stellar masses, M⋆, and star formation rates, SFRs, were found (M⋆ & 1010M⊙ and SFR & 100M⊙ yr−1) and used to calculate the specific SFRs. I determine that the (sub)mm-selected sources mostly lie on the high-mass end of the star formation ‘main-sequence’ which makes them a high-mass extension of normal star forming galaxies. I also find that the specific SFR slightly evolves at redshifts 2−4, suggesting that the efficiency of the star formation seems to be increasing at these redshifts. Using the S2CLS data, the bolometric IR luminosity functions (IR LFs) were found for a range of redshifts z = 1.2 − 4.2 and the contribution of the SMGs to the total star formation rate density (SFRD) was calculated. The IR LFs were found to evolve out to redshift ∼ 2.5. The star formation activity in the Universe was found to peak at z ≃ 2 followed by a slight decline. Assuming the IR to total SFRD correction found in the literature the SFRD found in this work closely follows the best-fitting function of Madau & Dickinson (2014).

Until recently, obtaining rest-frame optical spectra of galaxies at z > 1 was a time consuming and challenging observation due to the difficult nature of near-infrared (near-IR) spectroscopy. However, with the advent of second generation ground-based near-IR spectrographs (e.g. KMOS, MOSFIRE), and the new low resolution near-IR grisms on the Hubble Space Telescope (HST), we have entered a new era in the study of high redshift galaxies. This thesis explores the physical properties of star-forming galaxies in the redshift range 1 < z < 3 by utilising a custom reduction of the 3D-HST near-IR grism spectroscopic survey. One of the most important observational constraints on the evolution of galaxies is the mass-metallicity relation (MZR), which is sensitive to both the star-formation history and various inflow/outflow processes. I use the 3D-HST spectra to provide a new constraint on the MZR at 2:0 < z < 2:3, and moreover measure the O/H abundance directly from the oxygen and hydrogen emission lines ([OII], [OIII] and Hβ) as opposed to the more common method at high redshift of inferring O/H from the N/H ratio (via [NII] and Hα). I show that the traditional form of the MZR is recovered from the 3D-HST data, with metallicity increasing with the stellar mass of a galaxy. However, the absolute metallicity values I derive are inconsistent with previous N/H-based measurements of metallicity at these redshifts. Moreover, I show that the 3D-HST data is inconsistent with the `fundamental metallicity relation' (FMR), and that, contrary to previous claims, this local Universe relation may not hold out to z & 2. To investigate this metallicity discrepancy further, I measure the evolution of the [OIII]/Hβ nebular emission line ratio in the 3D-HST spectra over the redshift range 1:3 < z < 2:3. I compare this observed line ratio evolution with state-of-the- art theoretical models which take into account the independent evolution of the ionization parameter, electron density and metallicity of star-forming regions with redshift. The homogeneous 3D-HST dataset allows me to perform a consistent analysis of this evolution which takes into account line luminosity selection effects. I show that, according to models, the observed [OIII]/Hβ evolution cannot be accounted for by pure metallicity evolution. Instead I am able to infer that the line ratio evolution is more consistent with, at the very least, an evolution to stronger ionizing conditions at high redshift, and perhaps even denser star-forming regions. I explore how this result can also explain the observed discrepancy between high redshift metallicity measurements. In light of this finding, I revisit the MZR at z >~ 2 and employ a purely theoretical approach to inferring metallicities from nebular lines, which is able to account for an evolution in ionization conditions. I then use a selection of galaxies from the local Universe, which mimic the properties of high redshift galaxies, to derive a more robust ionization sensitive, conversion, between N/H and O/H. With this new conversion which I am able to bring the previous inconsistent metallicity measurements at z >~ 2 back into agreement. Finally, I am able to show that, in this new formalism, the metallicity evolution between z = 2 and z = 3 is perhaps not as large as previously reported. To conclude I discuss ongoing work as part of the KMOS Deep Survey (KDS) being undertaken with the near near-IR Multi-Object Spectrograph KMOS on the VLT. I describe the observations and data reduction that has been completed to date and describe how this instrument will allow me to extend the work presented in this thesis to z > 3. I also introduce FIGS, a new HST near-IR grism survey seeking to spectroscopically identify galaxies at 5:5 < z < 8:5 and work I have begun in exploring this dataset.

Les études détaillées d'amas de galaxies confirmés à grands redshifts sont peu nombreuses. L’objectif de cette Thèse est d’établir le premier catalogue d'amas confirmés spectroscopiquement à grand redshift et, pour la première fois à ces redshifts, d’étudier de manière statistique les propriétés des galaxies membres des amas. Dans cette Thèse, nous caractérisons et étudions 20 candidats amas à redshifts 1.4 < z < 2.8 parmi les candidats les plus prometteurs de l’échantillon CARLA. Nous réduisons et analysons des données spectroscopiques grism sans fente et imagerie proche-infrarouge des amas candidats, obtenues avec le télescope spatial Hubble. Nous mesurons plus de 700 redshifts au sein des champs observés, et confirmons spectroscopiquement 16 amas CARLA dans l’intervalle 1.4 < z < 2.8; ces amas sont associés à des noyaux galactiques actifs à fortes émissions radios (RLAGN) en leur centre, par sélection. Cet effort fait plus que doubler le nombre d’amas confirmés à ces redshifts. Nous étudions également le taux de formation stellaire des galaxies membres des amas en fonction de leur masses stellaires, et de la distance aux RLAGN. Nous trouvons que les galaxies membres massives sont situées sous la séquence principale jusqu’à z=2, ce qui suggère déjà à ces redshifts une évolution accélérée des galaxies massives au sein des amas. Nous trouvons également une concentration plus importante de membres actifs à plus petits rayons des RLAGN, jusqu’à z=2. Ceci est en accord avec un renversement de la relation densité vs. taux de formation stellaire pour nos amas CARLA à 1.4 < z < 2.0, ce qui suggère que les amas CARLA représentent une phase de transition de l’évolution des galaxies au sein des amas. Nous étudions également les populations stellaires de deux de nos amas confirmés à redshift z=2.0. Nous analysons les relations couleurs-couleurs et couleurs-magnitudes de ces deux amas et montrons que l’une des structures à z=2 possède une séquence rouge de galaxies passives. Globalement, nos résultats démontrent que les amas CARLA représentent des structures riches comprenant des populations mixtes de galaxies évoluées et massives sans formation stellaire, et des galaxies actives formant des étoiles. Cet échantillon sans précédent de 16 amas confirmés spectroscopiquement dans l’intervalle de redshift 1.4 < z < 2.8 constitue un échantillon idéal pour étudier statistiquement la phase de transition des amas de galaxies, ainsi que les mécanismes de suppression de la formation stellaire. (Abrégé) / Detailed studies of high-redshift confirmed galaxy clusters are based on a few individual objects. In this Thesis, we therefore aim at building the first sample of spectroscopically confirmed clusters at high-redshifts and, for the first time at these redshifts, statistically infer cluster member galaxy properties. In this Thesis, we study and characterize 20 cluster candidates at redshifts 1.4 < z < 2.8, which represent the most promising cluster candidates from the CARLA sample. We reduce and analyze slitless grism spectroscopic and near-infrared imaging data of the fields, obtained with the Hubble Space Telescope. We measure redshifts for over 700 star-forming sources in the 20 fields, and we spectroscopically confirm 16 CARLA clusters in the range 1.4 < z < 2.8; by selection, these clusters are associated with powerful radio-loud active galactic nuclei (RLAGN) at their center. This effort alone more than doubles the number of confirmed clusters at these redshifts. We study cluster member star-formation rates (SFRs) as a function of their stellar masses and distances from the RLAGN. We find that massive members are located below their star-forming main-sequence up to z=2. This implies that the massive star-forming end of the cluster population already followed an accelerated evolution at these high redshifts. We also find an increasing concentration of star-forming members with smaller radii relative to the RLAGN, at all redshifts up to z=2. Our 1.4 < z < 2.0 cluster members are therefore consistent with a reversal of the SFR-density relation. This is a first evidence showing that CARLA clusters represent a transition phase for cluster galaxy evolution. We also study stellar populations of two of our confirmed CARLA clusters at z=2.0. We study their color-color and color-magnitude relations and show that one of the two structures is comprised of a z=2 red sequence of passive candidate members. Together, these results provide clear evidence that our confirmed CARLA clusters represent rich structures comprised of mixed populations, including both evolved, passive, massive galaxies, and galaxies with ongoing star formation. Together, this unprecedented sample of 16 confirmed clusters at 1.4 < z < 2.8 constitutes an ideal sample for further statistical investigation of the cluster transition phase, including study of quenching mechanisms. (Abridged)