This thesis presents multi-frequency data, galaxy identifications, estimated redshifts, and derived physical properties for the sub-millimetre source sample produced by the SCUBA HAlf Degree Extragalactic Survey (SHADES). SHADES is the largest, complete, sub-millimetre survey undertaken to date, and the aim of this work is to exploit this survey to study the evolution of sub-mm selected galaxies at high redshift, explore their possible connectionwith localmassive ellipticals, and to test current models of galaxy formation. The SHADES sample was selected from 850 micron images made with the submillimetre camera SCUBA at the James Clerk Maxwell Telescope. These submillimetre maps cover a total area of 720 arcmin2 split between two well-studied extra-galactic survey fields, the Lockman Hole East and the Subaru/XMMNewton Deep Field (SXDF). The resulting sample of 120 sub-millimetre sources is the focus of this thesis. Here the wealth of information provided by deep radio, optical, near-infrared and mid-infrared imaging of the two SHADES fields is exploited to complete the identification of the SHADES sample, and then to derive a robust redshift estimate for every sub-millimetre source. Where possible this is achieved from the optical+ infrared photometry using a new two-component redshift estimation code developed specifically to deal with starbursting galaxies with potentially highly stochastic star-formation histories. The effectiveness of this code is demonstrated via comparison with the small subset of SHADES source which possess robust spectroscopic redshifts. For those galaxies which are too faint for effective redshift constraints to be provided by the existing optical+infrared photometry, the information on the long-wavelength spectral energy distribution provided by the radio+submm photometry is utilised to provide cruder constraints or limits on redshift. The result is the first complete and unbiased estimate of the redshift distribution of the bright extragalactic sub-millimetre galaxy population. It is found that the brightest sub-mm sources are confined to the redshift range 2 < z < 4, while more moderate luminosity sources span the full range of redshift out to z ∼ 5. The fits to themulti-frequency photometry provided by the redshift estimation technique are also used to derive estimates of the stellar mass, and star-formation history of each SHADES galaxy. The average derived stellar mass is ∼ 3 × 1011 M⊙ and it is found that the violent starburst powering the sub-millimetre emission typically contributes less than 10% of the stellar mass of the galaxy which has been assembled prior to the “current” starburst event. The distributions of redshift, stellar mass, and star-burst ages are compared with the predictions of a range of galaxy models, including the suite of models originally used to motivate the SHADES survey in van Kampen et al. (2005), and themost recent incarnation of the Durhamsemi-analytic galaxy formationmodels described by Swinbank et al. (2008). It is found that the redshift distribution and sub-mmflux versus redshift for bright sub-mmgalaxies can be reproduced best by one of the van Kampen models, which is based on semi-analytic modelling with a Chabrier IMF. We can rule out the non-semi-analytic prediction models and the Durham semi-analytic model with a top-heavy IMF. However the stellar masses are systematically underpredicted by all of the models. Either the stellar masses derived from the SHADES data have been systematically over-estimated, or the models need to be modified (perhaps by the inclusion of AGN feedback) to allow larger galaxy masses to assembled prior to z ∼ 2. Finally, it is demonstrated that themass in place prior to the observed starburst cannot have been produced by an analogous super-burst at higher redshift, but rather requires to have been assembledmore gradually over a timescale of ∼ 1−2 Gyr. It is thus concluded thatmassive galaxies undergo theirmost violent phase of star formation at redshifts 2 < z < 4, but that the enormous starbursts which lead to detection in current sub-millimetre surveys can only take place in the potential well provided by an already massive galaxy. This supports a scenario in which bright sub-millimetre galaxies are indeed the progenitors of the massive elliptical galaxies found in the local Universe.

What drives the transition of galaxies from the disc dominated, star forming blue cloud to the elliptical dominated, quiescent red sequence? What role does the morphology, central supermassive black hole and galaxy environment play in this transition? I have attempted to answer these questions by using Bayesian statistics to infer a simple star formation history (SFH) describing the time, t_q, and exponential rate, &tau;, that quenching occurs in a galaxy. I use both the optical and NUV photometry of a galaxy in order to infer the posterior distribution of its SFH across the two dimensional [t_q, &tau;] parameter space. I then utilise the Galaxy Zoo 2 morphological classifications to obtain a morphology weighted, combined population distribution across each quenching parameter for a sample of galaxies. I apply this method across the blue cloud, green valley and red sequence of a sample of 126,316 galaxies and find a clear difference between the quenching timescales preferred by smooth and disc weighted populations, with three major routes through the green valley dominated by smooth (rapid rates, attributed to major mergers), intermediately classified (intermediate rates, attributed to galaxy interactions) and disc morphologies (slow rates, attributed to secular evolution). I hypothesise that morphological changes occur in systems which have undergone quenching with an exponential rate, &tau; &LT; 1.5 Gyr, in order for the evolution of galaxies in the green valley to match the ratio of smooth to disc galaxies observed in the red sequence. I repeat this SFH analysis for a sample of 1,244 Type 2 AGN host galaxies and find statistical evidence for recent, rapid quenching, suggesting that this may be caused by AGN feedback. However I find that rapid quenching rates cannot account for all the quenching across the AGN host population; slow quenching rates, attributed to secular evolution, are also significant in the evolution of AGN host galaxies. I investigate this possible secular co-evolution of galaxies and black holes further by measuring the black hole masses of a sample of 101 bulgeless AGN host galaxies and compare them to typical black hole-galaxy scaling relations. I find that the measured black holes of the bulgeless galaxies are ~1-2 dex more massive than they should be, given their lack of bulges. This suggests that black hole-galaxy scaling relations may arise due to mutual correlations to the overall gravitational potential of the dark matter halo of the galaxy. I also considered the effect of the group environment on the time and rate that quenching occurs, with respect to the group-centric radius, for 4,629 satellite galaxies. I find that although mergers, mass quenching and morphological quenching are all occurring in groups, environmentally driven quenching mechanisms are also prevalent. However, I find that these environmentally driven quenching processes are not correlated with the velocity of a satellite within a group, ruling out ram pressure stripping as a possible mechanism. I discuss how all of these quenching mechanisms are likely to affect a galaxy across its lifetime, acting in concert to reduce the SFR, which in turn produces the wide distribution of quenching timescales seen across the colour-magnitude diagram. I discuss ideas for future work using the method employed in this work, including applying it to forthcoming data from large integral field unit surveys.

Galaxies, and clusters of galaxies, can act as gravitational lenses and magnify the light of objects behind them. The effect enables observations of very distant supernovae, that otherwise would be too faint to be detected by existing telescopes, and allows studies of the frequency and properties of these rare phenomena when the universe was young. Under the right circumstances, multiple images of the lensed supernovae can be observed, and due to the variable nature of the objects, the difference between the arrival times of the images can be measured. Since the images have taken different paths through space before reaching us, the time-differences are sensitive to the expansion rate of the universe. One class of supernovae, Type Ia, are of particular interest to detect. Their well known brightness can be used to determine the magnification, which can be used to understand the lensing systems. In this thesis, galaxy clusters are used as gravitational telescopes to search for lensed supernovae at high redshift. Ground-based, near-infrared and optical search campaigns are described of the massive clusters Abell 1689 and 370, which are among the most powerful gravitational telescopes known. The search resulted in the discovery of five photometrically classified, core-collapse supernovae at redshifts of 0.671&lt;z&lt;1.703 with significant magnification from the cluster. Owing to the power of the lensing cluster, the volumetric core-collapse supernova rates for 0.4 ≤ z &lt; 2.9 were calculated, and found to be in good agreement with previous estimates and predictions from cosmic star formation history. During the survey, two Type Ia supernovae in A1689 cluster members were also discovered, which allowed the Type Ia explosion rate in galaxy clusters to be estimated. Furthermore, the expectations of finding lensed supernovae at high redshift in simulated search campaigns that can be conducted with upcoming ground- and space-based telescopes, are discussed. Magnification from a galaxy lens also allows for detailed studies of the supernova properties at high redshift that otherwise would not be possible. Spectroscopic observations of lensed high-redshift supernovae Type Ia are of special interest since they can be used to test for evolution of the standard candle nature of these objects. If systematic redshift-dependent properties are found, their utility for future surveys could be challenged. In the thesis it is shown that the strongly lensed and very distant supernova Type Ia PS1-10afx at z=1.4, does not deviate from the well-studied nearby and intermediate populations of normal supernovae Type Ia. In a different study, the discovery of the first resolved multiply-imaged gravitationally lensed supernova Type Ia is also reported. / <p>At the time of the doctoral defense, the following papers were unpublished and had a status as follows: Paper 3: Manuscript. Paper 4: Manuscript.</p>

The centrepiece of this thesis is a deep, new, high-resolution 1.4-GHz image covering the United Kingdom Infrared (IR) Telescope IR Deep Sky Survey (UKIDSS) Ultra Deep Survey (UDS) legacy field. Deep pseudo-continuum observations were made using the Very Large Array, prior to its recent upgrade, in its A, B and DnC configurations. The resulting mosaic has a full-width-at-half-maximum synthesised beam width of ≈ 1.7 arcsec and a point-source sensitivity of ≈ 60μJy (6σ ) across the central 0.6 deg2, while conserving flux from sources of extended emission. The full image covers 1.3 deg2. I also present a catalogue containing over 1,000 radio emitters, having chosen the 6-σ threshold by maximising the number of radio sources with secure optical/near-IR counterparts. Most of the sources in the catalogue (≈ 90 per cent) lie in the sub-mJy flux density regime. Deep, complementary data covering a wide range of wavelengths was used to explore this faint radio population, whose nature remains controversial. It was found that 53 per cent of the sample comprise active galactic nuclei (AGN). AGN dominate at & 0.2mJy and remain a significant population down to 0.1mJy; at lower fluxes – the so-called μJy radio population – star-forming galaxies become dominant. The radio sample presented here was also matched to Hubble Space Telescope imaging of the UDS field (which is part of the Cosmic Assembly Near-IR Deep Extragalactic Legacy Survey – CANDELS) to classify the faint radio population morphologically. These classifications were done using the Gini–M20 method. It was found that a low fraction of AGN and SFGs are undergoing interactions and mergers, 33 ± 9 and 13 ± 7 per cent, respectively. The merger fraction does not appear to have evolved significantly since z ∼ 3. This suggests that mergers have played a relatively minor role in the assembly of galaxies and super-massive black holes – certainly less significant than previously thought. Finally, I present a study of cosmic star-formation activity as a function of stellar mass and redshift, exploiting panchromatic stacking. Mid-IR–through–radio images, including new data from Herschel, are stacked at the positions of a K-selected (i.e. an approximately mass-selected) sample in the UDS field. Specific star-formation rates (SSFR, i.e. star-formation rate per stellar mass, or the rate at which a galaxy is converting its gas into stars) were derived from UDS radio luminosities measured here and stellar masses from the literature. The SSFR was found to be poorly correlated with stellar mass; it decreases with decreasing redshift; at a given mass, SSFR rises with redshift. These results indicate that at early epochs, galaxies were forming stars more efficiently and at a higher rate.

The Great Observatories Origins Deep Survey North (GOODS-N) field, first surveyed by the HST, has been observed across numerous wavebands revealing populations of both Star Forming Galaxies (SFG) and Active Galactic Nuclei (AGN) over wide ranges of luminosities. It has been surmised that the evolution in the star forming population appears to diverge from that in the AGN population leading to a domination of SFGs at low flux densities. The number of starbursts can only be disentangled from the entire population if each source can be classified individually, which usually requires high angular resolution imaging. This is the motivation behind the e-MERLIN Galaxy Evolution survey, e-MERGE, which expands the depth of high resolution radio imaging in the GOODS-N field to increase the number of potentially classifiable sources. By use of wide-field imaging techniques, including a new high-speed mapping tool, together with a new semi-empirical primary beam-shape model for the e-MERLIN array, a deep wide-field high-resolution map is derived. This is the widest and deepest contiguous imaging yet obtained from e-MERLIN and JVLA observations, and yet contains less than 25% of the e-MERLIN data so far observed. The majority of the objects are shown to exhibit extended structure, and the angular size distribution place the median size around 1.2 arcsec, peaking between 0.5 and 0.7 arcsec. Automated algorithms are utilised to facilitate a new probabilistic classification tool based on multi-parameter correlations. 248 sources could be classified using the tool, each deriving a probability of AGN or SFG rather than forcing a binary category. Linear sizes of star-formation dominated sources are determined to lie in a range of 4 - 11 kpc, within the optical extent of galaxies. Differential source counting based on probabilistic classifications reveals that an increase in the luminosity evolution of SFGs is likely, although an apparent upturn in AGN may also exist to some lesser degree at low flux densities. The thesis establishes a clear roadmap for the remainder of the e-MERGE survey and a path to determine the star formation rate history of the Universe.

This thesis presents a starburst model for far-infrared/sub-millimeter/millimeter (FIR/sub-mm/mm) line emission of molecular and atomic gas in an evolving starburst region, which is treated as an ensemble of non-interacting hot bubbles which drive spherical shells of swept-up gas into a surrounding uniform gas medium. These bubbles and shells are driven by winds and supernovae within massive star clusters formed during an instantaneous starburst. The underlying stellar radiation from the evolving clusters affects the properties and structure of photodissociation regions (PDRs) in the shells, and hence the spectral energy distributions (SEDs) of the molecular and atomic line emission from these swept-up shells and the associated parent giant molecular clouds (GMCs) contains a signature of the stage evolution of the starburst. The physical and chemical properties of the shells and their structure are computed using a a simple well known similarity solution for the shell expansion, a stellar population synthesis code, and a time-dependent PDR chemistry model. The SEDs for several molecular and atomic lines ($^{12}$CO and its isotope $^{13}$CO, HCN, HCO$^+$, C, O, and C$^+$) are computed using a non-local thermodynamic equilibrium (non-LTE) line radiative transfer model. By comparing our models with the available observed data of nearby infrared bright galaxies, especially M 82, we constrain the models and in the case of M 82, provide estimates for the age of the recent starburst activity. We also derive the total H$_2$ gas mass in the measured regions of the central 1 kpc starburst disk of M 82. In addition, we apply the model to represent various stages of starburst evolution in a well known sample of nearby luminous infrared galaxies (LIRGs). In this way, we interpret the relationship between the degree of molecular excitation and ratio of FIR to CO luminosity to possibly reflect different stages of the evolution of star-forming activity within their nuclear regions. We conclude with an assessment of the strengths and weaknesses of this approach to dating starbursts, and suggest future work for improving the model.

This thesis presents a starburst model for far-infrared/sub-millimeter/millimeter (FIR/sub-mm/mm) line emission of molecular and atomic gas in an evolving starburst region, which is treated as an ensemble of non-interacting hot bubbles which drive spherical shells of swept-up gas into a surrounding uniform gas medium. These bubbles and shells are driven by winds and supernovae within massive star clusters formed during an instantaneous starburst. The underlying stellar radiation from the evolving clusters affects the properties and structure of photodissociation regions (PDRs) in the shells, and hence the spectral energy distributions (SEDs) of the molecular and atomic line emission from these swept-up shells and the associated parent giant molecular clouds (GMCs) contains a signature of the stage evolution of the starburst. The physical and chemical properties of the shells and their structure are computed using a a simple well known similarity solution for the shell expansion, a stellar population synthesis code, and a time-dependent PDR chemistry model. The SEDs for several molecular and atomic lines ($^{12}$CO and its isotope $^{13}$CO, HCN, HCO$^+$, C, O, and C$^+$) are computed using a non-local thermodynamic equilibrium (non-LTE) line radiative transfer model. By comparing our models with the available observed data of nearby infrared bright galaxies, especially M 82, we constrain the models and in the case of M 82, provide estimates for the age of the recent starburst activity. We also derive the total H$_2$ gas mass in the measured regions of the central 1 kpc starburst disk of M 82. In addition, we apply the model to represent various stages of starburst evolution in a well known sample of nearby luminous infrared galaxies (LIRGs). In this way, we interpret the relationship between the degree of molecular excitation and ratio of FIR to CO luminosity to possibly reflect different stages of the evolution of star-forming activity within their nuclear regions. We conclude with an assessment of the strengths and weaknesses of this approach to dating starbursts, and suggest future work for improving the model.

The rate at which stars have formed throughout the history of theuniverse is not constant, it started out slow, increased until around redshift ∼ 2 when it reversed and became slower again. The reason for this behaviour is still being investigated with various models and simulations usually based upon dark matter halos. The aim of this study is to instead investigate whether there is a correlation between the cosmic star formation rate and the evolution of cosmic voids. This is achieved by comparing the total mass flow from voids with the amount of matter forming stars. A simple model of void mass flow is created and compared with observational data of star formation. The model is shown to exhibit the same behaviour as the star formation rate indicating that there is indeed a correlation between void evolution and star formation. This suggests it to be fruitful to create a more involved, alternative model of star formation based upon void evolution as opposed to the common halo evolution / Hur snabbt stjärnor bildas har genom universums historia förändrats över tid, det började långsamt och ökade sedan fram till rödförskutning ∼ 2 då trenden vände och saktade ner igen. Förklaringen till detta beteende utforskas fortfarande genom diverse modeller och simularingar som vanligtvis bygger på mörk materia halos. Syftet med detta arbete är att istället undersöka ifall det finns en korrelation mellan tomrumsutveckling och den kosmiska stjärnbildningen. Detta åstadkoms genom att jämföra det totala massflödet från tomrum med den massa som bildar stjärnorna. En simpel model för tomrumsutveckling skapas och jämförs med observationell data för stjärnbildningshastighet. Denna modell visar samma beteende som stjärnbildningen och tyder på att det finns en korrelation mellan denna och tommrumsutveckling. Som slutsats pekar denna studie på att det kan vara fruktbart att utveckla en mer anancerad modell för den kosmiska stjärnbildningen som bygger på tomrumsutveckling istället för mörk materia halos.