No description available.

Cosmological hydrodynamic simulations have shown that galaxies are fed by dense, cold gas streams at high redshift. However, the presence of such gas has never been observationally confirmed. Using the Horizon- MareNostrum simulation, I examined whether cold flows are detectable with low-ionisation metal absorption lines, such as C II 1334. It is concluded that due to their low metallicity and density, it is extremely difficult to prove/disprove the presence of cold flows using the metal absorption lines. Revisiting the acquisition of angular momentum in disc galaxies using high resolution simulations, I found that at the time of accretion, gas and dark matter do carry a similar amount of specific angular momentum which is systematically and significantly higher (at minimum by a factor of 2) than that of the dark matter halo as a whole. Whereas cold streams directly deposit this large amount of angular momentum within a sphere of radius r~0.1 Rvir, dark matter particles easily pass through the central region, depositing their angular momentum over a much more spatially extended region. As a result, in our simulations neither the total specific angular momentum of the baryons nor its radial profile ever follows that of the virialised dark matter halo, contrary to what is typically assumed in the standard theory of disc galaxy formation. In order to better understand the formation of disc galaxies and the missing baryon problem in a LCDM universe, continuous, collective galactic winds are implemented. It is demonstrated that stellar feedback processes are able to suppress star formation by ~30% at z=3, compared to that from the run without feedback sources, but it still produces an unrealistic central peak in the rotation curve. Although inclusion of hypernovae further suppresses star formation, it is unable to quench the formation of low-angular momentum stars enough to remove the peaked rotation curves at high redshift. Finally, feedback from active galactic nuclei turns out to be effective at suppressing star formation in massive galaxies at 1<z<2, reproducing their observed number densities in the redshift range. However, further suppression of residual star formation is required to form quiescent galaxies at z=2.

This thesis explores the evolution of galaxies from the onset of star formation shortly after the Big Bang until the present day. Particular emphasis lies on the redshift range z = 1.4 2.5, the so-called “redshift desert”, as it coincides with the peak epoch of cosmic star formation activity and mass assembly. Most of the information about galaxies and their evolution arrives in the form of their integrated light, i.e. the conglomeration of light emitted by stars of various ages and metallicities. In order to interpret the observed spectra and magnitudes, and to extract the physical parameters we therefore require models. This holds true in particular for galaxies too faint to target them spectroscopically, and for which redshifts and physical parameters derived from only their photometry is the only feasible way to study them in more detail. This thesis is concerned with such models, and describes how GALEV evolutionary synthesis models describe the spectral and chemical evolution of galaxies, accounting for gaseous emission and the increasing initial abundances of successive stellar generations, how they compare to observations and what we can learn from their application. Based on a large model grid, covering all observed galaxy evolution stages, I find that sub-solar metallicities have significant impact on the spectra of galaxies, and can lead to systematic errors and biases if not accounted for. A comparison of models with different metallicities furthermore reveals that photometric redshifts are also systematically biased if sub-solar metallicities are not properly accounted for. I also note that even a small mass-fractions of young stars can dominate the overall spectrum, leading to a large underestimation of the mass and age of the stellar population. The models explain not only the colour evolution of galaxies observed at a range of redshifts, but also their physical parameters. I show that with magnitudes in only a few bands we can successfully explain not only the masses of galaxies, but also their star formation rates and, where available from observations, their metallicities. If additional data are available, the grid of models can be used to refine colour selection criteria and to break degeneracies, e.g. between dust-reddened actively star-forming galaxies and intrinsically old, passively evolving galaxies. Using GAZELLE, a photometric redshift code that is purpose-tailored to harmonise with these models, I can extract accurate redshifts and a wealth of physical parameters from the largest ever sample of observed multi-wavelength photometry of galaxies. I then compare our findings with semi-analytical models that trace the evolution of individual galaxies based on cosmological simulations. In my sample I find a significant population of high-mass galaxies that is not accounted for by this class of models. Furthermore a small percentage of massive, yet starforming galaxies challenges our idea on how these galaxies form and evolve. In an appendix to this thesis I present a complementary approach to reconstruct the evolution of galaxies, using star clusters as tracers. I introduce a new technique to break the age-metallicity degeneracy and obtain individual ages and metallicities for a sample of globular clusters, revealing a merger of two Sb/Sc-type spirals 2 Gyrs ago in NGC 4570, a lenticular galaxy in the Virgo cluster. Also in the appendix I show that, at least in the studied galaxy Arp 78, the initial mass function conforms with our assumptions and does not change in low-density environments as recently predicted. Although studies of galaxy evolution are a major field in astronomy, there is still a lot more to be done to reveal the inner workings of these island universes, and this thesis also addresses how to continue and improve the work presented herein.

This thesis presents spectroscopic observations of z ≥ 7 galaxy candidates in the Hubble Ultra Deep Field, which were selected with HST WFC3 imaging, using the Lyman-Break technique. Four z-band (z ≈ 7) dropout galaxies were targeted with Gemini/GNIRS, one z-band dropout galaxy and three Y -band (z ≈ 8 − 9) dropout galaxies with VLT/XSHOOTER, and 22 z-band dropouts with VLT/FORS2, where 15 of the latter are strong candidates. No evidence of Lyman-α emission is found, and the upper limits on the Lyman-α flux and the broad-band magnitudes are used to constrain the rest-frame equivalent widths for this line emission. Amongst the targeted objects, observations were made of HUDF.YD3, a relatively bright Y -band dropout galaxy likely to be at z ≈ 8 − 9 on the basis of its colours in the HST ACS and WFC3 images. Lehnert et al. (2010) observed this galaxy using the VLT/SINFONI integral field spectrograph and claim that it exhibits Lyman-α emission at z = 8.55. In observations of this object described in this thesis, which were made with VLT/XSHOOTER and Subaru/MOIRCS, this line was not reproduced despite the expected signal in the combined MOIRCS & XSHOOTER data being 5σ. Hence it appears unlikely that the reported Lyman-α line emission at z > 8 is real. Accounting for incomplete spectral coverage, in total (across all spectro- graphs) 9.63 z-band dropouts and 1.15 Y -band dropouts are surveyed to a Lyman-α rest-frame Equivalent Width better than 75 ̊A. A model where the fraction of high rest-frame equivalent width emitters follows the trend seen at z = 3−6.5 is inconsistent with these non-detections at z = 7−9 at a confidence level of ∼ 91%, which may indicate that a significant neutral HI fraction (χHI) in the intergalactic medium suppresses the Lyman-α line at z > 7. In particular, the lack of detection of Lyman-α emission in this spectroscopy is compared with results at lower redshift by Stark et al. (2010), who derive a mapping between Lyman-α fractions and χHI based on radiative transfer simulations by McQuinn et al. (2007). These results suggest a lower limit of χHI ~ 0.5.