Search

This thesis presents three studies focusing on boosted topologies that utilise machine learning techniques for boosted H → b¯b reconstruction using the ATLAS detector. The measurement of the t¯tH cross-section is a direct way of accessing the Higgs top Yukawa coupling (yt). Firstly, an all-hadronic feasibility study is shown, aimed at assessing boosted topologies in the all-hadronic t¯tH decay channel. It was found to have low statistical significance, with considerable efforts and data driven techniques required to reduce the QCD-multijet background. Secondly, the boosted contribution to the recent t¯tH, H → b¯b measurement using the full Run-2 ATLAS data set, 139f b−1 at √s = 13 TeV, is analysed. There is a considerable contribution from the boosted region to this result, particularly to the differential cross-section measurement of the Simplified Template Cross-Section (STXS) bins [300, 450) and [450, ∞) GeV. The result of the inclusive profile-likelihood fit is μ = 0.35+0.36−0.34, which corresponds to σ = 1.0(2.7) observed(expected) significance compared to the background-only hypothesis. Thirdly work on retraining the boosted H → b¯b reconstruction deep neural network (DNN) is shown for the Run-2 Legacy re-analysis. The bespoke DNN trained for the analysis showed some improvements over the previous round due to the updated analysis algorithms. It also outperformed the general purpose H → b¯b Xbb tagger. The final motivation for use of the bespoke DNN is that it allows the choice of boosted jet collection (RC-jets vs LR-jets). RC-jets re cluster "small" (∆R = 0.4) jets with ∆R = 1.0 while LR-jets directly cluster the calorimeter clusters with ∆R = 1.0, both using the anti-kt algorithm. The RC-jets jets are found to be advantageous. This is due to the ease of propagating systematics for combining with resolved regions and the good modelling observed using samples made with the Atlfast-2 detector simulation.

Solar activity sporadically erupts due to the release of magnetic energy that manifest as impulsive bursts of electromagnetic emission. Solar radio bursts provide a valuable diagnostic of the low corona where the energy release originates and of the environment through which the exciter of radio emission propagates. However, the corona is a turbulent environment such that the escaping radiation is significantly modulated, concealing the intrinsic exciter characteristics and spatial location. In this thesis, radio burst fine structures that can be embedded within or separate from broadband bursts are investigated. In particular, solar radio spikes are analysed using the LOw Frequency ARray (LOFAR), for the first time providing much needed time and frequency resolved imaging of individual spikes over sub-second scales at decametre frequencies. The characteristics of spikes are statistically compared across decades in frequency, demonstrating the prevalence of radio-wave scattering up to ∼ 1 GHz that governs the observed decay time, rather than collisional damping. Consequently, the findings suggest that the duration of energy release of which spikes may be a direct signature, could be shorter than implied from observations, particularly at decametre frequencies. Analysis of the characteristics of decametre spikes and striae from the same event presents similarity in morphology, spatial location, and polarisation, indicating that the spikes are likely generated via plasma emission. The escaping spike and striae radiation presents superluminal centroid motion directed non-radially which implies the presence of an extended coronal loop with strong anisotropic turbulence. Imaging observations suggest that the magnetic structure is perturbed via the passage of a CME shock front, with the loop structure slowly restoring towards the prior configuration over time, exciting frequent magnetic reconnection that manifests as radio spike emission. As a result, the site of electron beam acceleration, the characteristics of the beam, and the prevailing turbulent conditions together determine the emitting location. However, the scattering component obscures it, making interpretation of the source in radio images challenging. A quantitative analysis of the spatial and spectral evolution of solar radio burst fine structures as the radiation escapes through a dipolar magnetic structure with anisotropic turbulence is presented via radio-wave scattering simulations. The observed spatial location and motion of the spikes and striae is replicated, as well as the suppressed fine structure drift rate in dynamic spectra. The combination of high resolution observations with state-of-the-art simulations show that sub-second solar radio burst fine structures must be decoupled from anisotropic scattering effects in order to assess the intrinsic emitter. As such, the apparent source location, magnetic field structure, and turbulent conditions must be considered simultaneously.

In recent years, J/ψ photoproduction in the near-threshold region has seen a renewed theoretical interest due to the wealth of information it has to offer. Near-threshold J/ψ photoproduction proceeds through the exchange of gluons in the t-channel and is expected to provide unique insight about the nucleon gluonic gravitational form factors (GFFs). Previous studies at the Thomas Jefferson National Accelerator Facility (JLab) in Virginia, USA, have already measured the total and differential cross section of J/ψ near-threshold photoproduction on the free proton. This thesis presents the first measurements of J/ψ near-threshold photoproduction on the bound proton and bound neutron in a deuteron target. The CEBAF Large Acceptance Spectrometer at 12 GeV (CLAS12) based at JLab's Hall B uses a 11 GeV electron beam impinging on a fixed liquid deuteron target. J/ψ is then produced via the exchange of a quasi-real photon and decays to a lepton pair which is detected alongside the recoil nucleon by CLAS12. Analysis procedures and in particular machine learning based techniques were developed in order to correctly identify final state particles and select J/ψ photoproduction events. The total and differential cross section of J/ψ near-threshold photoproduction on the bound proton and bound neutron were then measured from the detected J/ψ photoproduction events. The measurements of the cross sections on the bound proton and bound neutron agree well within their statistical uncertainty. This is consistent with the assumed two-gluon exchange production mechanism which is isospin invariant. Overall a first measurement of the J/ψ near-threshold photoproduction cross sections on the bound proton and bound neutron was achieved. A better understanding of the mechanical properties of the nucleon, such as pressure and mass distributions, can be obtained by relating J/ψ production to the nucleon GFFs. This opens the way for exciting new insights into the internal structure of the nucleon, and in particular of the nucleon's gluonic content. An upcoming overhaul of the CLAS12 reconstruction will increase the reconstruction efficiency and the statistical precision of the preliminary measurements described in this thesis, with an expected gain in statistics of at least 50%. Future upgrades at JLab will allow to test some of the theoretical assumptions made in relating J/ψ near-threshold photoproduction to the nucleon GFFs.

A major challenge in gravitational-wave astrophysics is the interpretation of observations, which requires accurate inference of the astrophysical parameters and a rigorous statistical framework. The main focus of this thesis is the analysis of modelled gravitational-wave sources and its enhancement with machine learning and other statistical techniques. Bayesian statistics is at the base of gravitational-wave analysis and interpretation since each observation is unique and can often be assumed to be independent of all others. The unifying thread of this thesis is Bayes's theorem: how it is routinely leveraged for gravitational-wave analysis, allowing much of the work presented here, and how its use can be extended to develop new analysis techniques. The most notable application of Bayesian statistics in the field is the parameter estimation of compact binary coalescence. Chapter 2 reports the work done to reproduce the first Gravitational-Wave Transients Catalogue (GWTC-1) with the Bayesian Inference Library: bilby. The rigorous comparison between previous GWTC-1 results and the one presented here allowed bilby's specific tuning towards the gravitational-wave inference problem. Chapter 3, presents the author's work related to the discovery of the first neutronstar black-hole (NSBH) mergers GW200105 and GW200115, where bilby was used to estimate the parameter of the observed sources. This chapter also illustrates the role of gravitational-wave observations in our understanding of the astrophysical origins of binary sources. Chapter 4 describes a novel effective likelihood method to quantitively compare astrophysical distributions inferred from gravitational-wave observations and distributions obtained with theoretical simulations. This method, which is driven by a Bayesian philosophy, is applied to a set of globular cluster simulations and real data from the third Gravitational-Wave Transients Catalogue (GWTC-3). Chapter 5, presents a novel density estimation tool for parameter estimation products from gravitational-wave observations, based on Gaussian Processes which are a Bayesian machine learning technique. This density estimation method was found to be advantageous over other traditional methods for several gravitational wave applications since we need both the accurate treatment of individual event samples, e.g. standard siren analysis, but also robust propagation of systematics when combining multiple observations, e.g. measure of systematic errors. Finally, Chapter 6 presents a study that makes use of bilby to re-analyse the binary neutron star (BNS) event GW190425, in light of its potential electromagnetic counterpart FRB20190425A, and makes use of a Gaussian Process density estimator to calculate the Bayesian odds of the claimed association. This work is extended by performing a standard siren measurement for GW190425 and its potential host galaxy to determine the value of the Hubble constant.

Most massive stars are found in hierarchical triples or higher multiplicity systems, in which a close inner binary is orbited by one or more distant companions. Here, we present a new triple stellar evolution code, TSE, which simultaneously takes into account the physics of the stars and their gravitational interaction. TSE is used to simulate the evolution of massive stellar triples in the galactic field from the zero-age-main-sequence until they form compact objects. To this end, we implement initial conditions that incorporate the observed high correlation between the orbital parameters of early-type stars. We show that the interaction with a tertiary companion can significantly impact the evolution of the inner binary. For instance, high eccentricities can be induced by the third-body dynamical effects, leading to a Roche-lobe overflow or even to a stellar merger from initial inner binary separations 10³-10⁵ R⊙. Focusing on the evolution subsequent to a stellar merger, we find that binaries composed of the merger product star and the tertiary companion provide a new source to form binary black hole mergers in the Galactic field. By means of a population synthesis, we estimate their contribution to the total black hole merger rate to be R(z=0) = 0.3-25.2 Gpc⁻³yr⁻¹. Merging binary black holes that form from stellar post-merger binaries have exceptionally low mass ratios. We identify a critical mass ratio q≃0.5 below which they could dominate the total black hole merger rate in the field. Lastly, we study the dynamical evolution of the spin vector of a massive donor star that stably transfers mass to a binary companion. Assuming that the donor star loses its mass along the instantaneous interstellar axis, we find that the ejection of ≳30 per cent of the donor's initial mass causes its spin to nearly flip onto the orbital plane of the binary, independently of the initial spin-orbit alignment.

This thesis presents a collection of works, some methodological and others observational, with the common theme of strong gravitational lensing. We have leveraged a wide variety of techniques, from machine learning algorithms to uv-plane modelling, and used a multitude of simulated and real data products to achieve the results contained within this thesis. In Chapter 2, we explored the feasibility of combining the predictions of an approximate Bayesian neural network with the parametric profile fitting of PyAutoLens. By using the predicted 1-σ uncertainties on the lens model parameters from the CNN to inform our choice of prior within PyAutoLens, we were able to show that a technique that combines both of these tools is able to outperform either method alone. We applied our methodology to a series of increasingly complex and realistic simulated strong gravitational lenses, beginning with simple parametric lenses and sources, and graduating to HUDF sources and lenses extracted from the EAGLE simulation. In Chapter 3, we addressed the issue of the existence of unphysical source reconstructions by applying a convolutional neural network to detect these solutions, and developed a simple prescription to re-initialise the lens modelling process in a new region of parameter space to help ensure convergence upon the global solution. Such tools are necessary components to developing a truly automated lens modelling pipeline, as will become increasingly necessary in the era of LSST and Euclid. In Chapter 4, we presented and modelled ∼0.1 arcsec resolution ALMA imaging of seven strong gravitationally lensed galaxies detected by the Herschel Space Observatory. We inferred the mass profiles of the lensing galaxies and by determining the magnification factors, we investigated the intrinsic properties and morphologies of the lensed submillimetre sources. We found that these submillimetre sources all have ratios of star formation rate to dust mass that are consistent with, or in excess of, the mean ratio for high-redshift submillimetre galaxies and low redshift ultra-luminous infrared galaxies. Reconstructions of the background sources reveal that the majority of our sample display disturbed morphologies. The majority of our lens models have mass density slopes close to isothermal, but some systems show significant differences. In Chapter 5, we present a comparison between performing lens modelling in the uv-plane versus the image-plane. When dealing with interferometer observations, one must make a choice of whether to perform the analysis on the raw visibility data, or to employ an algorithm such as CLEAN to produce an image to work with. When producing an image from visibility data, there are several choices one must make that can impact the final image, such as the adopted lead to biases in the inferred lens model parameters, and whether some choices outperform others. We found that in general, direct modelling of the visibilities provided the most robust means for recovering the lens model parameters, but also that using the Briggs weighting scheme performed better than the natural weighting scheme.

The 21-cm line has a potential to be a powerful probe of the intergalactic medium (IGM) during the Cosmic Dawn and the Epoch of Reionization. Various observatories such as LOFAR and SKA aim to observe this line. To aid these observational efforts, I use cosmological simulations drawn from the Sherwood-Relics programme in which a novel hybrid approach of coupling hydrodynamical simulations with radiative transfer is implemented to model the 21-cm forest during the Epoch of Reionization and sky-averaged 21-cm signal during Cosmic Dawn. Recent Lya line observations can be explained by reionization which is completed at z ≃ 5.3 (i.e. later than previously thought). In this scenario, large islands of neutral hydrogen are expected to persist in the diffuse IGM until z ≃ 6. In this context, I predict the incidence of strong 21-cm forest absorbers (τ₂₁≳10⁻²) and find that if the IGM is not pre-heated above ≃ 10²K, the 21-cm forest should be detectable at redshift as low as z ≃ 6. I consider the effect of the pressure smoothing arising from the patchiness of reionization, redshift space distortions, Lyα coupling and soft X-ray background pre-heating of the IGM on the observability of the 21-cm forest signal. While the pressure smoothing affects the 21-cm forest signal only modestly, inclusion of redshift space distortion increases the largest τ₂₁ by up to a factor of ~10. In addition, the soft X-ray background can completely suppress the signal. However, a null detection of strong 21-cm forest absorbers at z ≃ 6 in the spectra of ~10 sufficiently radio-bright background sources with SKA1-low and possibly LOFAR can provide informative, model-dependent lower limits on the soft X-ray background at high redshift. Furthermore, I model the effect of quasar radiation on the ionization and thermal state of the IGM surrounding it. The gas in the proximity of a quasar is ionized enough such that it becomes transparent for Lya photons. The size of this Lyα transmission window, the Lya near-zone, has been used to inform us about the quasar optical/UV bright lifetime. The majority of measurements are consistent with lifetimes of ~10⁵-10⁷yr, however, the smallest Lyα near-zones suggest lifetimes shorter than ~10⁴yr. Such short lifetimes pose a challenge for the growth of black holes with masses of ~10⁹M_⊙ at z≳6. Models of black holes accreting in an obscured phase or quasars with time-varying spectrum (flickering) have been proposed to alleviate this problem. However, it is challenging to discern between young quasars and old quasars driven by episodic accretion with the Lyα forest alone. Motivated by this, I model the 21-cm forest absorption in the vicinity of z≳6 radio-loud quasars. The distance between a quasar and the closest 21-cm forest absorber with τ₂₁ ≳ 10⁻² is sensitive to the heating of the IGM by quasar X-rays. Hence, the extent of the proximate 21-cm forest is sensitive to the integrated quasar lifetime because of a very long gas cooling time. I find that a detection of a strong 21-cm forest absorber within ~3pMpc of a quasar would suggest quasar lifetime of ≲ 10⁵yr. On the other hand, a measurement of large distance between a quasar and the closest strong 21-cm forest absorber in combination with a small Lya near-zone would be consistent with a quasar that is old and flickering. Finally, I study the effect of IGM density fluctuations on the amplitude of the cosmic dawn absorption feature in the sky-averaged 21-cm spectrum. The density distribution of the IGM is extracted from Sherwood-Relics simulations. I find that the density fluctuations suppress the signal by 6.4-11.3% depending on the mass resolution of the simulation. Higher mass resolution of the simulation results in larger suppression of the signal. This effect makes the discrepancy between standard theoretical models and the EDGES measurement even larger.

Recent developments in the field of gravitational physics, including the emergence of gravitational wave astronomy, black hole images, and more accurate telescopes, have allowed us to probe the strong-field character of gravity in a novel and revolutionary manner. This accessibility related to strong gravity brings into the foreground discussions about potential modifications to General Relativity (GR) that are particularly relevant in high curvature regimes. The most straightforward way to generalize GR is to consider an additional degree of freedom, in the form of a scalar field. In this thesis, we study generalized scalar tensor theories that predict interesting strong-gravity phenomenology. First, we review scalar no-hair theorems and the conditions under which they can be evaded. Next, we study solutions of black holes with scalar hair and the way in which higher derivative terms alter their properties. We then move our discussion to the spontaneously scalarized solutions, which only deviate from GR in the strong-field regime. We propose a model consistent with compact object scalarization, that allows for a GR attractor at late times, without fine-tuning (EsRGB model). Then, we proceed to study properties of black holes and neutron stars in this theory, revealing the interesting phenomenology of the solutions. We also study the radial stability of black holes in EsRGB and perform a preliminary analysis of the hyperbolicity of the problem. Finally, we take a look at the shadows of black holes and wormholes in theories with scalar fields, in light of recent observations of black hole shadows.

This thesis is prepared in two parts. In the first half (Chapter 2 and Chapter 3) we discuss the evolution of galaxy mergers at 0.5 < z < 3.0 in all the CANDELS fields based on a supervised deep learning model trained on the IllustrisTNG cosmological simulations. The second half is dedicated to the rest-frame optical morphological evolution of galaxies from z = 1.5 to 8 as observed by JWST in the SMACS 0723 field, and in the early observations of the CEERS program. In Chapter 2 we describe a supervised deep learning framework designed for the classification of high redshift galaxy mergers based on data from the IllustrisTNG cosmological simulations. We generate a large dataset of before mergers/post- mergers/non-mergers galaxy mocks labeled with information from IllustrisTNG 300-1 merger trees. These imaging data are prepared to be CANDELS-like and are then used to train deep learning models capable of achieving 90% of accuracy within the simulations. Using these them we describe the evolution of the galaxy merger fractions and rates in the CANDELS fields and we discuss how these deep learning classifications are related to visual classifications. We report the first agreement of galaxy merger rates between galaxy pair statistics methods and morphologically selected mergers, with R(z) = 0.02 ± 0.004 × (1 + z)2.76±0.21, showing that the highest merger rates are found at the highest redshifts. We tackle the challenging problem of separating recently coalesced galaxy mergers from non-interacting highly star forming galaxies in Chapter 3. These two populations present ambiguous morphologies due to asymmetric features. We refine our methods reported in Chapter 2 for this particular question, generating a dataset of TNG100-1 post-mergers and star forming galaxies at 0.5 < z < 3.0, including a full radiative transfer treatment with the SKIRT code, producing ∼ 160,000 images with realistic morphologies. We explore the relative populations of post-mergers and non interacting star forming galaxies in this redshift range. We show that the population of high redshift asymmetric galaxies are more likely to be of post-merger origin than their low redshift counterparts. The interpretabil- ity of our models is discussed by exploring the feature space extracted from the mock imaging and the real CANDELS galaxies. We show that for this particu- lar problem, deep learning models provide an 30% improvement over quantitative morphology methods. We focus on the early release JWST observations of the SMACS 0723 cluster in Chapter 4. We report the first ever morphological study of rest-frame optical structure in 1.5 < z < 6 with NIRCam, within the wavelength range λ = 0.9μm - 4.4μm. We conduct visual classifications and quantitative morphology measure- ments on a sample of ∼ 200 galaxies previously detected with HST. We report a surprising mismatch between the number of disk galaxies detected with HST and JWST. Around ten times more disks are found. We briefly discuss the implica- tions of this result and how it fits in the galaxy formation and evolution evolution Over Chapter 5 we expand the framework of Chapter 4 to the early CEERS JWST observations that have overlap with the EGS observations from the CAN- DELS fields with HST. We release to the community the biggest sample of visually classified galaxies observed with JWST to-date, with 4265 galaxies that are both observed by HST and JWST. With this dataset, we carefully discuss the evolu- tion of the Hubble sequence up to z ∼ 8, finding that it is already present at the earliest of times for low to intermediate mass galaxies, while evolution driven by mergers is observed for massive galaxies. We detail the quantitative morphology characteristics of this sample, and how it correlates with visual optical morphology. We finish with a brief discussion on the results presented in this thesis, how the merger evolution at 0.5 < z < 3.0 and the general morphological evolution at z > 3.0 are linked, and what are the next steps to explore this connection further.

This thesis contains an analysis of the problem of time in quantum cosmology and its application to a cosmological minisuperspace model. In the first part, we introduce the problem of time and the theoretical foundations of minisuperspace models. In the second part, we focus on a specific minisuperspace universe, analyse it classically, and quantise it using the canonical quantisation method. The chosen model is a flat FLRW universe with a free massless scalar field and a perfect fluid. We explain how different types of perfect fluid can be accommodated in our model. We extract the Wheeler-DeWitt equation, and calculate its solutions. There are three dynamical variables that may be used as clock parameters, namely a coordinate t conjugated to the perfect fluid mass, the massless scalar field φ, and v, a positive power of the scale factor. We define three quantum theories, each one based on assuming one of the previous dynamical quantities as the clock. This quantisation method is then compared with the Dirac quantisation. We find that, in each quantisation procedure, covariance is broken, leading to inequivalent quantum theories. In the third part, the properties of each theory are analysed. Unitarity of each theory is implemented by adding a boundary condition on the allowed states. The solutions to the boundary conditions are calculated and their properties are listed. Requiring unitarity is what breaks general covariance in the quantum theory. In the fourth part, we study the numerical properties of the wave functions in the three theories, paying special attention to singularity resolution and other divergences from the classical theory. The t-clock theory is able to resolve the singularity, the φ-clock theory presents some non trivial dynamics that can be associated with a resolution of spatial infinity, and the v-clock theory does not show significant deviations from the classical theory. In the last part, we expand our analysis in order to include another quantisation method: path integralquantisation, and finally, we conclude.

In this thesis, I report on my work on galaxy evolution in proto-clusters. In Chapter 1, I describe previous knowledge about clusters and proto-clusters of galaxies. The next three Chapters correspond to three different projects I have completed during my PhD that are connected to the topic of galaxy evolution in proto-clusters. In Chapter 2, I quantify the relative importance of environmental quenching versus pre-processing in z~1 clusters by analysing the infalling galaxy population in the outskirts of 15 galaxy clusters at 0.8 < z < 1.4 drawn from the GOGREEN and GCLASS surveys. I find significant differences between the infalling galaxies and a control sample; in particular, an excess of massive quiescent galaxies in the infalling region. These massive infalling galaxies likely reside in larger dark matter haloes than similar-mass control galaxies because they have twice as many satellite galaxies. Based on these findings, I conclude that it may not be appropriate to use 'field' galaxies as a substitute for infalling pre-cluster galaxies when calculating the efficiency and mass dependency of environmental quenching in high redshift clusters. By comparing the quiescent fraction of infalling galaxies at 1 < R/Rv < 3 to the cluster sample (R/Rv < 1) I find that almost all quiescent galaxies with masses >10¹¹ M☉ were quenched prior to infall, whilst up to half of lower mass galaxies were environmentally quenched after passing the virial radius. This means most of the massive quiescent galaxies in z~1 clusters were self-quenched or pre-processed prior to infall. In Chapter 3, I report on my work on intracluster light in proto-clusters at z~2. In contrast to theoretical expectations, I report on the detection of intracluster light within two proto-clusters at z=2 using deep HST images. I use the colour of the intracluster light to estimate its mass-to-light ratio in annuli around the brightest cluster galaxies (BCG), up to a radius of 100 kpc. I find that 54±5% and 71±3% of the stellar mass in these regions is located more than 10 kpc away from the BCGs in the two proto-clusters. This low concentration is similar to BCGs in lower redshift clusters, and distinct from other massive proto-cluster galaxies. This suggests that the proto-cluster BCGs have already experienced a special merger history similar to their lower redshift counterparts. We compare these observations to the Hydrangea hydrodynamical galaxy cluster simulations. In contrast to semi-analytic models, they predict that intracluster stars are a generic feature of massive halos since at least z=2. This implies that intracluster light is a natural consequence of hierarchical structure formation. Over Chapter 4, I report on my work on the luminosity function and Dn4000 measurements of proto-cluster galaxies at 1.3 < z < 3.0. Using HST grism data, I found that the luminosity function of proto-clusters differ from the field, such that proto-clusters have an excess of luminous galaxies. I also found that proto-cluster galaxies have higher values of Dn4000 compared to field galaxies. We interpret that proto-cluster galaxies have older stellar populations than field galaxies. In Chapter 5, I summarise my main findings on this work and plans for future work.

Over the past several decades, researchers have explored a family of mathematical analogies that relate inaccessible astrophysical phenomena to systems that can be manipulated in laboratory settings. Now, a variety of these phenomena have been successfully scrutinised in a laboratory setting, allowing for a deeper investigation of these analogies. This thesis presents results from a series of studies of instabilities with astrophysical implications, undertaken in the context of analogue gravity. From the perspective of analogue gravity, vortices in quantum fluids provide space-times with discrete and topological features. It has long been known that a multiply charged quantum vortex will decay into a cluster of singly charged vortices. Recently, it was pointed out that this instability is similar to that of the rotational superradiance of black holes. This relationship is interpreted further, and through numerical observations, a new phenomenon is encountered in the late stages of the decay. Finally, an upper bound on the orbital frequency of a vortex pair is found, and related to the sound wave responsible for the decay of a doubly charged vortex. On the other hand, the relaxation of compact clusters of quantum vortices involves a complex interplay between vortices and waves, with energy released as sound radiation. By applying techniques from gravitational physics, the study uncovers the emergence of circular sound trajectories, a large-scale feature that enables a straightforward prediction of radiated sound. This phenomenon, called sound-rings, is closely related to the ringdown process of black holes. Furthermore, the linear scaling of the sound-rings with the net charge of the cluster allows them to be located well outside the vortex core. In a completely different system, we investigate cosmological preheating via the parametric instability appearing from applying a vertical oscillation to two-fluid interfaces. Using methods adapted from field theories, a non-linear model is presented and compared with experimental results and numerical simulations. The appearance of secondary instabilities created by the nonlinear contributions to the primary instability is observed and is well predicted by the model. Experimental results suggest that the analogy with cosmological preheating persists in the nonlinear regime. Then, in preparation for the next generation of analogue gravity experiments, a new technique of digital holography for the measurement of deformations of fluid interfaces is introduced. Due to partial reflections of the optical beam, coherent light impinging on a fluid interface returns as a multitude of rays. By applying, or exploiting, a tilt of the bed on which the fluid rests, multiple off-axis holograms are formed and can independently be used to measure the interface. The method is realisable with only a few basic optical components and provides a versatile scheme for high-precision measurement of fluid interfaces.

The field of gravitational wave astronomy has grown remarkably since the first direct detection of gravitational waves on 14th September 2015. The signal, originating from the merger of two black holes, was detected by the two US-based Advanced LIGO interferometers in Hanford (Washington State) and Livingston (Louisiana). The second observing run of the Advanced LIGO and Virgo detectors marked the first detection of a binary neutron star merger, along with its electromagnetic counterparts. The optical follow-up of the merger led to the first confirmed observations of a kilonova, an electromagnetic counterpart to binary neutron star and neutron star-black hole mergers whose existence was first predicted in 1970s. Following the multimessenger observations of the binary neutron star merger GW170817, constraints were put on the rate of expansion of the Universe using both gravitational wave and electromagnetic data. These measurements could help us understand the current tension between early-Universe and late-Universe measurements of the Hubble constant H0. The use of gravitational wave signals for measuring the rate of expansion of the Universe was proposed by Schutz in 1986. Compact binary coalescences can be used as distance markers, a gravitational wave analogue to standard candles: "Standard Sirens". Measurements of the Hubble constant from standard sirens are independent from previous methods of constraining H0. Bright sirens are gravitational wave signals that are detected coincidentally with electromagnetic signatures. These "bright" gravitational wave sirens are powerful cosmological probes, allowing us to extract information on both the distance and the redshift of the source. It is therefore important to maximise these coincident detections, and to carefully calibrate the data extracted from any standard siren. The work presented in this thesis can be divided into three main topics, all under the umbrella of maximising scientific returns from observations of compact binary coalescences. These three topics are: kilonova parameter estimation, cosmology with gravitational waves, and calibration of advanced gravitational wave detectors. We present work on inferring parameters from kilonova light curves. Ejecta parameters and information about the merging time of the progenitor is extracted from simulated kilonova light curves. We explore the consequence of neglecting some aspects of microphysics on the resulting parameter estimation. We also present new results on the inference of the Hubble constant through the application of a robust test of galaxy catalogue completeness to the current gravitational wave cosmology pipeline. We explore the impact of adopting a robust estimate of the apparent magnitude threshold mthr for the galaxy catalogues used in gravitational wave cosmology on the final inference of the Hubble constant H0 from standard sirens, and compare the results to those obtained when adopting a conservative estimate for mthr. Finally, we present the first results from the prototype of a Newtonian Calibrator at the LIGO Hanford detector. Calibrating the LIGO detectors is crucial to the extraction of the gravitational wave source parameters that are used in cosmology with standard sirens.

The eruption of a classical nova (CN) is an extremely energetic transient event that produces a rapid optical brightening of 10-15 magnitudes, followed by a slower decline in luminosity. A CN is a binary system consisting of a white dwarf (WD) primary that accretes stellar material from the less-evolved donor star. In the majority of systems, mass transfer onto the WD takes place via an accretion disk. A thermonuclear runaway is triggered when sufficient mass has accumulated on the WD, and the energy thus injected into the WD envelope causes the high velocity expulsion of the envelope in the nova eruption. Due to the rapid ejection of this shell of material, the WD photosphere expands and then contracts, which is observable as the brightening and subsequent fading of the nova light curve. A dwarf nova (DN) outburst is less luminous than a CN eruption, and occurs when material in the accretion disk is suddenly deposited onto the WD due to thermal or tidal instabilities within the disk. The corresponding release of gravitational potential energy powers the increase in luminosity. V392 Persei is a known DN that underwent a CN eruption on April 29 2018, with γ-ray emission detected from the system the following day. V392 Per provided the first opportunity to study the γ-ray emission processes in a previously studied nova system. Here we report ground-based optical, Swift UV and X-ray, and Fermi -LAT γ-ray observations following the eruption for almost three years. The optical light curve reveals that V392 Per is one of the fastest evolving novae yet observed, with a t₂ decline time of 2 days. Early spectra present evidence for multiple and interacting mass ejections, with the associated shocks driving both the γ-ray and early optical luminosity. V392 Per entered Sun constraint within days of eruption. Upon exit, the nova had evolved to the nebular phase, and we saw the tail of the super-soft X-ray phase. Subsequent optical emission captured the fading ejecta alongside a persistent narrow line emission spectrum from the accretion disk. Ongoing hard X-ray emission is characteristic of a standing accretion shock in an intermediate polar. Analysis of the optical data reveals an orbital period of 3.230 ± 0.003 days, but we see no evidence for a WD spin period. The optical and X-ray data suggest a high mass WD, the pre-nova spectral energy distribution (SED) indicates an evolved donor, and the post-nova SED points to a high mass accretion rate. Following eruption, the system has remained in a nova-like high mass transfer state, rather than returning to the pre-nova DN low mass transfer configuration. We suggest that this high state is driven by irradiation of the donor by the nova eruption. In many ways, V392 Per shows similarity to the well-studied nova and DN GK Persei. A preliminary photoionization analysis of the early nebular spectra was performed in an attempt to constrain the ionization conditions within the nova shell. Three key emission line flux ratios were measured from the spectra. The plasma simulation and spectral synthesis code cloudy was used to produce an array of models that varied the effective temperature of the WD (the ionizing source), and the electron density and metallicity of the nova shell. The measured line ratios were compared with the predicted ratios for the models. Although the results were inconclusive, they indicated some constraints on the ionization conditions that were consistent with what we might expect for a nova shell. Finally, some suggested developments of the work discussed in this thesis are presented. The first extension considered is a more complete analysis of the photoionization conditions within the shell of V392 Per, accompanied by morpho-kinematic modelling to constrain the geometry of the nova shell. Another avenue to progress this work is to conduct a further monitoring campaign on V392 Per and ascertain the ongoing mass transfer state of the system. Polarimetric observations may reveal signals of the WD magnetic field, or of a degree of dust production within the expanding shell. Perhaps the most exciting possibility would be to apply the same analytical techniques to observations of a system similar to V392 Per, but which does not experience Sun constraint at such an early stage of its evolution.

This thesis covers a number of numerical experiments exploring the first generation of Population III star formation. After briefly reviewing the field to date, we present a resolution test for primordial gas. We find that the gas fragments to a higher degree as the resolution of the simulations is increased, leading to lower mass stars than predicted by previous, lower resolution studies. A number of stars are ejected with masses capable of surviving until the present day, providing targets for observations. We continue to present a study exploring the effects of primordial magnetic fields on Population III star formation, finding that the small-scale, tangled fields are incapable of providing support against fragmentation or significantly altering the collapse. Lastly, we present the results of cosmological zoom-in simulations around halos exposed to Lyman-Werner (LW) radiation. While the LW fields result in higher halo masses, it has little effect on the Pop III IMF. The primordial IMF therefore likely remains constant during the transition from pure Pop III.1 stars to the Pop III.2 stars that form in the presence of their radiation.

The data from ground-based gravitational-wave interferometers has been used for direct astrophysical observations of transient signals for almost a decade. With this data, over 90 compact binary systems have been observed and studied via their gravitational-wave emission. These observations have, and continue to, provide new solutions to astrophysical questions. Gravitational waves can provide information about astrophysical systems that has previously been inaccessible through electromagnetic observations. The aim of this thesis is to outline some of the dangers of making incorrect assumptions about observed signals in gravitational-wave interferometer data. In Chapter 1 I provide a summary of gravitational-wave interferometer data; from the basic design of the interferometers, through the form of the data, and some methods can be used to manipulate the data for analysis. I then describe the types of signals that can be observed in this data, how these signals are studied, and what further analysis can be performed on the results of these studies. In Chapter 2 I calculate the probability that transient gravitational-wave signals will overlap with each other in the data. In the chapters that follow I outline the potential problems that time-overlapping transients will cause for the signal detection and analysis methods currently implemented by the LIGO-Virgo-KAGRA collaboration. Chapter 3 shows how the presence of a second, time-overlapping, binary black hole signal can cause inaccuracies in the estimation of the parameters of the other binary black hole system. I show how this problem manifests in the estimated parameter distributions for different relative parameters between two compact binary signals. In Chapters 4 and 5 I consider whether it is possible to detect these cases of time-overlapping transients with current methods, determining what cases will be missed without modifications to current algorithms. I also discuss possible modifications to these algorithms, and a separate bespoke method, designed to identify which detected signals contain time-overlapping transient signals. Chapter 6 presents an analysis of the data of a gravitational-wave interferometer for the purposes of a search for scalar dark matter signals. In this work, we produced extremely precise estimations of the noise floor of the interferometer. We used this to identify and reject possible candidate signals from scalar dark matter. Finally, in Chapter 7, I provide a summary of the key findings in each of the chapters of this thesis. This chapter includes recommendations of extensions and adaptations to the described investigations to expand and improve upon this work.

Sulphur is the tenth most abundant element in the universe and is known to play many important roles in biochemical, geochemical, and atmospheric processes. However, the astrochemistry of sulphur presents many as yet unresolved problems. In the dense interstellar medium, for instance, there exists a notable paucity of sulphur in both the gas phase as well as within interstellar icy grain mantles, compared to the expected cosmic abundance of the element. Moreover, no conclusive detection of solid interstellar H2S has yet been made, despite the known efficiency of hydrogenation reactions on the surface of interstellar dust grains. Within the Solar System, solid SO2 is known to be a component of many icy bodies, including the Galilean moons of Jupiter and several comets, but the mechanism leading to its formation still largely eludes contemporary astronomers, chemists, and spectroscopists. This thesis probes a number of questions relating to sulphur ice astrochemistry in both interstellar and Solar System environments. It begins with a review of the current state of knowledge of extra-terrestrial sulphur chemistry, before delving into a broader discussion of the chemical characteristics of the cosmos and a brief history of astrochemistry as an independent field of research. A background to the relevant scientific concepts and principles routinely used in laboratory astrochemistry then follows, which includes detailed discussions on molecular structure and symmetry, spectroscopy (with an emphasis on measurements made in the mid-infrared spectral range), and radiation chemistry. A thorough description of a new experimental facility for ion and electron irradiation studies of astrophysical ice analogues based at the Institute for Nuclear Research (Atomki) is provided, along with various experiments performed to validate the set-up. The results of systematic studies on the mid-infrared absorption spectra of H2S and SO2 astrophysical ice analogues, both pure and mixed with H2O, are then presented and discussed in the context of their applicability to the detection of these sulphur-bearing species in different astrophysical environments. The results of the first ever systematic comparisons of the radiation chemistry and physics of the amorphous and crystalline phases of a number of pure astrophysical ice analogues, including some that bear sulphur atoms, are described in the subsequent chapter. The observed greater radiolytic decay rates of amorphous ices and the more rapid formation of molecular products as a result of their irradiation is discussed both in the context of the differences in the strength and extent of the intermolecular interactions between the amorphous and crystalline phases, as well as in light of the recent discoveries of complex organic molecules in astrophysical environments in which space radiationinduced amorphisation is thought to out-compete thermal crystallisation of ices. The results of experiments investigating the implantation of high-energy sulphur ions into oxygen-bearing ices as a potential mechanism towards the formation of icy SO2 on the surfaces of the Galilean moons of Jupiter are also presented, with the analysis of the data demonstrating that this is not likely to be a major contributor towards solid SO2 on these moons. This contrasts with the results obtained from experiments investigating the irradiation of oxygen-bearing ices deposited on top of elemental sulphur layers, which was shown to result in the formation of a number of volatile sulphur-bearing molecules such as SO2, CS2, and OCS. The thesis concludes with a discussion of potential directions for future work, as well as a number of suggestions on improving the present experimental set-up.

Supernovae are the explosive death of stars. Massive stars (M > 8M⊙) end their lives as core-collapse supernovae. A white dwarf that gains mass and reaches the Chandrasekhar mass (∼1.4M⊙) explodes as a thermonuclear supernova. Supernovae can be broadly split into two classes based on whether hydrogen is present in the spectrum. Type I supernovae lack hydrogen and type II supernovae exhibit hydrogen. Type IIn supernovae are a mysterious subclass, they are highly inhomogeneous and are characterised by complex Balmer line profiles with a narrow component which is interpreted as interaction between the SN ejecta and a dense, pre-existing hydrogen-rich circumstellar medium. The progenitor paths for type IIn supernovae are unclear, at least one transient, SN2005gl has pre-explosion imaging revealing the progenitor to be a luminous blue variable. These luminous blue variables are massive evolved stars that undergo episodic mass loss which may form the required circumstellar material for the type IIn phenomenon. However recent studies on the environments of type IIn supernovae reveal that these transients are not strongly associated with regions of ongoing star formation. These environmental studies are a powerful indirect method to constrain progenitor paths. In the first part of this thesis, I set out a classification scheme for type IIn supernovae and apply it to archival spectral data. Then using this sample with robust classifications, I observe the hosts of type IIn supernovae with the Liverpool Telescope, the Isaac Newton Telescope and the Las Cumbres Observatory Global Telescope Network 2m. Using these data, I create continuum subtracted Hα images and apply the novel pixel statistics technique, normalised cumulative ranking. This method is used to gauge the association of a SN position with the Hα emission in the host. There is a mass ladder in terms of this association, the more massive the progenitor, the better the supernovae follow the emission. I present the results of these pixel statistics as well as the radial distributions of type IIn supernovae in order to investigate possible progenitor routes. In the second part of this thesis, I investigate the environments of classical novae in the Andromeda galaxy, M31. Classical novae are a subset of cataclysmic variable where a white dwarf accretes hydrogen-rich material from a companion via Roche-lobe overflow, Once sufficient material has been accreted, a thermonuclear runaway occurs on the surface of the white dwarf and a portion of the accreted material is ejected. Generally, classical novae can be split into two spectral classes, Fe II and He/N, based on their characteristic non-Balmer lines. Previous work has suggested the existence of different populations of classical novae in terms of their radial distribution, or association to the bulge or disc of M31. I investigate the possibility that the spectral classes of classical novae can be separated based on their radial distributions. The progenitor systems of He/N classical novae may have a higher mass white dwarf and may be expected to be associate with the younger populations in the disc of M31. Firstly I present the largest spectroscopically confirmed M31 classical nova sample. Then as well as a radial analysis, for the first time, I will implement the normalised cumulative ranking method on classical novae. In this case with GALEX NUV and FUV. I compare the spectral classes to each other in terms of both their radial distributions and association to the UV emission.

The era of gravitational wave (GW) observations began with the ground-breaking detection at the Laser interferometer GravitationalWave detector (LIGO), we are now exploring more of the GW power spectrum. Upcoming space-based detectors such as the Laser Interferometer Space Antenna (LISA) will probe, for the first time, the millihertz window of the GW spectrum with the hope of detecting astrophysical and cosmological sources. The source of interest in this thesis is a cosmological first order phase transition at the electroweak scale. Detecting these GWs from the early universe would provide the opportunity to delve further back in the history of the universe than ever before. In theories beyond the Standard Model a cosmological first order phase transition occurs when, below a critical temperature, bubbles of stable phase spontaneously nucleate in the surrounding metastable phase. These bubbles expand, collide, and merge until only the stable phase remains. This phenomenon produces a stochastic gravitational wave background (SGWB) that, once scaled due to the expansion of the universe, peaks in the frequency band corresponding to LISA. In the Standard Model this transition is a crossover and no GWs are produced. Thus, a detection of such a SGWB would be a discovery of new physics. In the following thesis, using the latest analytical model of a SGWB from a first order phase transition (the Sound Shell Model) and information from cutting edge simulations I explore LISA's ability to perform parameter estimation on key phase transition parameters. I focus on two different parameterisations of the phase transition SGWB, the thermodynamic and spectral. The thermodynamic parameters are derived from the physics of the phase transition and consequently are related to the beyond standard model theory they inhabit. The spectral parameterisation is computationally cheaper than the thermodynamic parameterisation, which is advantageous when performing MCMC simulations. However, the connection from the spectral to thermodynamic parameters is an unanswered question. Here, we present a method for reconstructing the thermodynamic parameters from the spectral parameters. I use statistical methods including Fisher analysis and Markov chain Monte Carlo (MCMC) simulations to estimate parameter uncertainties on the two parameterisations. The impact of astrophysical foregrounds on resolving a SGWB for a first order phase transition are also investigated.

Solar prominences have been observed for years, with observations dating back to the 12th century. Only in recent history however, have we had the spectroscopic tools required to probe and understand their structure and dynamics. Even then, with ground bases observations we have been unable to observe the hottest parts of solar prominences due to the atmosphere which cruelly hides this ultraviolet radiation from us. Several observational satellites have been launched over the years so that we may observe these high temperatures. With the launch of the Interface Region and Imaging Telescope (IRIS), we were able to observe the Mg II h&k lines. Which gives us a unique view into the atmosphere surrounding solar prominences. In Chap 1, we give a brief overview of the solar atmosphere and what we know about the life and evolution of solar prominences. We cover their morphology, typical thermodynamic parameters, velocity distributions, and surrounding magnetic field strength. This is followed by a brief aside into radiative transfer where the basic equations are outlined with the LTE solution for radiative transfer through some emitting material. The principle ion in this work, Mg II h&k, is also presented with mention of its key features. The instrumentation used in the study are then introduced; IRIS, the X-Ray Telescope (XRT) onboard Hinode, Atmospheric Imaging Assembly (AIA) and the Helioseismic Magnetic Imager (HMI) onboard the Solar Dynamics Observatory (SDO), and the Sun-Earth Connection Coronal and Heliospheric Imager (SECCHI) onboard the Solar Terrestrial Relations Observatory Ahead (STEREO-A). In Chap 2, we cover the coordinated observation of a solar prominence on the 19 April 2018 by IRIS, AIA, and XRT.We show the distribution of the following statistical measures of the prominence, the integrated line intensity, Doppler velocity, line widths, and asymmetries using the quantile method. We also present the distribution of line profile types, whether they be singe, double, or complexly peaked. A method for filtering out coronal pixels using the line widths, peak intensities, and pixel connectivity is presented and shown to work to effectively isolate the prominence. We draw conclusions surrounding its observed dynamics and the relationships between its statistical measures. Chap 3 sees the introduction of a new method used to ascertain the properties of solar prominences observed in Mg II h&k. This is achieved by the point-for-point comparison of 1007 line profiles generated by the 1D non-local thermodynamic equilibrium (NLTE) radiative transfer code, PROM, with that of the observations. PROM treats its resonance lines in PRD and allows us to generate the Mg II h&k line profiles, with the three Mg II triplet lines and the principle Hydrogen lines, which are not used in this method. This matching method is named rRMS. Not only do we find the best fitting profiles in our bank, but we also find some measure of the 'goodness' of these fits. Therefore we can set a cut-off for what is considered a good or a bad fit. This is the first time a prominence has been inverted in this way. An updated version of rRMS called xRMS is then presented. It demonstrates a significant computational improvement over rRMS, allowing us to use a larger grid of 23940 line profiles. We present both of these results and compare what is found for the smaller and bigger grids. While both achieve similar levels of successful fits, the parameters which they recover are not exactly the same. Chap 4 introduces the radiative transfer code RTCY. It is a 2D cylindrical radiative transfer code. With which we are able to simulate an array of geometric configurations and velocity fields. As such, we explore the effect that these velocities have on line formation through the use of plots similar to that of Figs 4 through 7 of (Carlsson & Stein 1997). The most interesting result is that of the expanding velocity field where the prominence moves radially away from the axis of rotational symmetry of the cylinder. Following this, we explore multithread simulations where we stack these cylinders behind each other. We see how this affects the observed line profiles and how adding random line-of-sight velocities produces interesting asymmetries. In addition to the latter, the spectral and spatial point spread function (PSF) of the IRIS spectrograph is convolved with our final profiles demonstrating what IRIS would see if presented with these profiles. Finally to close the chapter, we perform a manual multithread forward fit. In Chap 5, we explore the formation of a coronal bright point (CBP) due to flux emergence and the associated filament channel of this emergence. Using Fourier Local Correlation tracking on HMI intensity images, we are able to recover the overall global velocity vectors of magnetic patches predict their global movement. The flows of these patches are seen to influence the stability of a minifilaments which form in the filament channel. These minifilaments are both seen to erupt when certain magnetic phenomena, such as flux cancellation occur. Chap 6 offers our conclusions where we outline the main results found and present our plans for future work.

Fluctuations in the light we receive from the Sun happen constantly. Much of this variation can be thought of as random, however within solar timeseries there are many true periodic signals to be found, if one knows where to look. Identifying the source of these oscillations can give us a wealth of information about the underlying physical environment of the solar atmosphere. One area which has been particularly exploited is the search for oscillations due to solar flares which can be short-lived, bursty, and variable. Periodic signals detected during solar flares can help us understand the physics of flares, as well as how they interact with the lower layers of the atmosphere. The lower layers of the atmosphere have been known to present periodic fluctuations on timescales of a few minutes in quiet Sun conditions, and there have been some recent reports of energetic events such as flares affecting the quiescent oscillations, even exciting them. This thesis aims to investigate the effect of flares on the lower atmosphere by identifying and categorising oscillatory signals in a variety of solar data. In Chapter 1 an overview of necessary background physics is given. The layers of the Sun's atmosphere and the standard model of solar flares are described. An introduction to different periodic signals which are present in solar data is given, including quiet Sun periodicities and those observed during transient phenomena. The methods used throughout the thesis are presented in Chapter 2, starting with descriptions of the instrumentation which obtained the analysed datasets. A discussion is presented on best practices when searching for oscillations in data, and the spectral fitting method is outlined. Techniques used to obtain physical solar parameters from observations are also introduced. In Chapter 3, a study of chromospheric intensity oscillations in a flaring active region is presented. Maps were produced showing the oscillatory signals which were present before and after the occurrence of the flare. When comparing the results after the flare to those before, oscillations were found to have changed their locations and typical periods in the vicinity of a sunspot located near the flare ribbons. These changes were interpreted as the result of a changing magnetic environment connected to the flare. The interpretations of these results were investigated using magnetohydrodynamic simulations in Chapter 4. Three different sets of simulations were carried out, to identify the most likely effect to have been responsible for the results of Chapter 3. The sets of simulations were concerned with the magnetic field inclination, the chromospheric temperature profile, and the length of the chromospheric cavity. The inclination angle was found to be the most likely of these three effects to change the periodic signals. Chapter 5 presents analysis of a second dataset, from an active region which featured much less powerful flare activity than the event from Chapter 3. A similar analysis was performed, and extended to also feature Doppler velocity data. Velocity oscillations were identified in pixels which exhibited some of the strongest flare heating. The possible cause of the induced velocity oscillations was explored using measurements of the magnetic field from spectropolarimetric observations, and inversions using the STiC code. In Chapter 6 concluding remarks are made, including a summary of the main results of the thesis and descriptions of possible future avenues for investigation.

This thesis is based on research into primordial non-Gaussianity produced during inflation. It focuses on predictions of the bispectrum from D3-brane inflation, and constraints on the squeezed-trispectrum derived from observational data.

Gravitational lensing is a cornerstone prediction of Einstein's General Theory of Relativity, which provided us with the first piece of direct evidence that led to his theory becoming the accepted interpretation of gravity. Now, over one hundred years later, the study of gravitational lensing has evolved into a diverse field that is continually utilised to unveil the wondrous subtleties of our universe. Specifically, the gravitational lensing of explosive transient sources is a field which is set to make major contributions to some of the biggest open debates in astrophysics -- including the Hubble Tension, the physics and nature of transient progenitors, the redshift evolution of explosive transient rates, and new tests of Einstein's theory. The next few years are pivotal for the subject, when observatories such as Rubin will transform studies of these events by increasing the number of discovered lensed transients by two orders of magnitude. In this thesis, I give a description of my original contributions to the field, which are focused around optimising the detection of gravitationally-lensed explosive transient sources such as supernovae, gamma-ray bursts, gravitational waves and kilonovae using a watchlist-based approach. The outcome of my research has led to a new understanding of how a lensed transient watchlist should be constructed, and the development of an algorithm that is capable of locating the galaxy clusters relevant to the lensing of such transients within existing all-sky survey data, which can then be used to populate a lensed transient watchlist. In addition, I describe the searches for lensed transient sources I have been active in alongside my research.

Since the initial detection of gravitational waves in 2015, a slow and steady stream of events has been recorded. The next step is to detect more and weaker sources, probing deeper into the universe. In order to do so, our interferometers must be pushed to even greater extremes, to suppress the noise that would otherwise obscure such faint signals. Quantum noise is perhaps the most fundamental source of noise in a laser interferometer. Arising from the inherent uncertainty in the amplitude and phase of any light source, it limits the sensitivity of all current and proposed future interferometers, especially at higher signal frequencies. The second generation detectors currently in use, such as Advanced LIGO and Virgo, are operating around the standard quantum limit---the maximum sensitivity that can be reached by employing classical techniques. For future upgrades and new detectors, advanced quantum techniques must be employed to further increase sensitivity. Advanced LIGO has already demonstrated its ability to surpass the standard quantum limit over a small range of frequencies by making use of squeezed states of light, and future improvements will increase the magnitude and frequency range of this sensitivity enhancement. In this thesis, I will detail my work on these quantum techniques, aimed particularly at the upcoming third generation of gravitational wave detectors. This covers a wide variety of topics: different arrangements of filter cavities for improved squeezing, an alternate detection scheme that reduces both quantum and more technical noises, a mechanism that operates via complex optomechanical interactions, and more. In each of these subjects, my role has largely been to perform validation of new ideas and techniques, and to investigate the effects of varying detector parameters on the proposed scheme. To this end, I have made extensive use of a numerical frequency domain interferometer simulation tool known as finesse. Widely used throughout the interferometer community, finesse allows the fast and accurate modelling of optical layouts. Certain advanced techniques, however, were beyond finesse's capabilities when I started my work. This thesis therefore also describes my contributions towards a full rewrite of this tool, called finesse 3. I performed a complete reimplementation of the quantum noise calculations in finesse 3, and added features necessary to model new and exciting ideas. I will explain the mathematics that finesse 3 uses to simulate interferometers, before finally giving a broader overview of its design, usage, and the great potential it has to allow investigation of ever more complex techniques for the foreseeable future.

This thesis is about the field of exoplanets, with a particular focus on the search for circumbinary exoplanets with Radial Velocity (RV) observations. Chapter 1 introduces the field of exoplanet science, provides an overview of relevant techniques for exoplanet detection, discusses what we have learnt about exoplanets through their demographics, and provides background information on circumbinary exoplanets. Chapter 2 covers the data analysis techniques that were utilised in this work. It introduces Bayesian statistics and nested sampling, explains the Kima package used to fit Keplerian signals to RV data in this work, best practises in its use, and how planetary parameters are obtained through this analysis. Chapter 3 describes my contribution to the analysis of the RV data for the star HD-16417 (λ²) Fornacis) and includes the full paper. In this work we constrain the parameters of a known planet host star, and in turn update the planet's parameters. Chapter 4 provides a detailed description of the Binaries Escorted By Orbiting Planets (BEBOP) survey, the main topic of work in this thesis. Previous attempts to discover circumbinary planets with RV observations are discussed, before describing how the BEBOP survey is carried out. In Section 4.3, I present my work on the calculation of detection limits for the BEBOP survey, along with the calculation of circumbinary planet occurrence rates, and present some preliminary candidate circumbinary signals. I find the BEBOP survey is sensitive to planets with masses down to that of Saturn and Neptune, and that our circumbinary planet occurrence rates agree with those from other works, including those of gas giants around single stars. Chapter 5 details the detection of the first circumbinary planet with ground-based RV observations, along with my contributions to this work. We are able to independently detect the circumbinary planet Kepler-16b, confirm its orbital parameters, and place constraints on the presence of additional planets in the system. In Chapter 6, I describe the first discovery of a circumbinary planet with RV observations alone, BEBOP-1c. This second planet in the system has a mass of 0.2 M_J and an orbital period of 215 days. We are also able to place an upper limit on the smaller inner transiting planet's mass at 23.6 M_⊕ with 99% confidence. In this chapter I also describe an attempt to view the signal of the secondary star in the binary star system. Finally, in Chapter 7, I describe additional contributions I have made to other bodies of work during my PhD, and conclude the thesis while discussing future avenues of work to increase our sensitivity to circumbinary planets with RV observations.

The study of exoplanets planets as a field is only three decades old, however, it is now one of the biggest areas of research in astrophysics. This is because fundamentally all of the exoplanet research attempts to answer one of the biggest questions in humanity, are we alone in the Universe? One of the only ways for scientists to fully answer that question lies in the understanding of the gaseous envelope that surrounds these planets, known as the atmosphere. These atmospheres hold key information about their formation and migration histories through their primordial disks. The key to unlocking that information is through spectroscopic observations which can be used to determine the chemistry and physical processes within these atmospheres. I present two separate analyses of high resolution observations of two nontransiting hot Jupiters, HD179949 b (chapter 3) and τ Bo¨otis b (chapter 4), in the near-infrared. For HD179949 b, I present the detection of water vapour in the Lband using the CRIRES instrument. For τ Bo¨otis b, I also present the detection of water vapour in the atmosphere using the CARMENES instrument. The latter detection is in disagreement with recent observations taken with the SPIRou instrument which is discussed in chapter 4.8. In chapter 5, I present a simulated analysis of high resolution spectroscopic observations of an Earth-like planet from the ELT 39m telescope around M-dwarf stars. This study aimed to simulate the typical data analysis techniques on a time series data-set currently used on high resolution spectroscopic data typically used for faster orbiting planets. This analysis used principal component analysis (PCA) on these simulations to remove time varying flux variations. It was found that PCA removed most, if not all, of the planets signal due to the fact that a habitable planet, even around later type M-dwarfs move too slowly for this methodology to be feasible.

The intrinsic nature of noisy and complex data sets is often concealed in low-dimensional structures embedded in a higher dimensional space. Number of methodologies have been developed to extract and represent such structures in the form of manifolds (i.e. geometric structures that locally resemble continuously deformable intervals of Rj 1). Usually a-priori knowledge of the manifold's intrinsic dimensionality is required. Additionally, their performance can often be hampered by the presence of a significant high-dimensional noise aligned along the low-dimensional core manifold. In real-world applications, the data can contain several low-dimensional structures of different dimensionalities. This work describes a framework for dimensionality estimation and reconstruction of multiple noisy manifolds embedded in a noisy environment. The workings of the framework are demonstrated on different synthetic data sets, presenting challenging features for state-of-the-art techniques in Multi-Manifold learning. Through these worked examples it is shown how the proposed methodology is able to model abstract and topologically challenging manifolds such as Möbius strip and toroids (also higher dimensional). The comparison with existing techniques is organized along the two separate aspects of the methodology, namely manifold approximation and probabilistic modelling. The framework is then applied to astronomical complex data set containing simulated gas volume particles from a particle simulation of a dwarf galaxy interacting with its host galaxy cluster and a Dark Matter simulation of the Large Scale Structure (LSS) of the Universe (Cosmic Web). Detailed analysis of the recovered 1D and 2D manifolds can help in understanding the nature of Star Formation in such complex systems and to link the kinematic properties of Dark Matter filaments with their galaxies. An additional case study is presented for the modelling of cavities in a Jellyfish galaxy. Tracking their evolution through time may help in understanding their origin.

In the thirty years since the discovery of the first exoplanet, over 5000 verified exoplanets have been discovered, unveiling a rich array of different exoplanetary architectures. However, there are still many unanswered questions regarding the formation and evolution pathways which have led to the observed population. To understand these processes it is imperative to trace exoplanets across time, both over galactic and human timescales. This thesis presents work in both of these areas, using data from the Transiting Exoplanet Survey Satellite (TESS). The bulk of this thesis focuses on the challenge of discovering new young exoplanets (age < 1 Gyr) and understanding the variability of their potential host stars. This begins with building an extended population of young stars around which to search for exoplanets, illustrating the kinematic power of the recently launched Gaia satellite and resulting in a target list of over three million young stars. A dedicated new young star detrending pipeline is then presented, which is in turn used to search for new young exoplanets in stellar associations within TESS sectors 1-5. Although no new exoplanets were found, the pipeline's effectiveness is demonstrated by recovering the previously known young exoplanets DS Tuc Ab and AU Mic b, alongside all other 2 min Targets of Interest (TOIs) from the 30 min cadence data alone. The completed young exoplanet search highlighted the challenging diversity of young stellar variability. To understand this variability, Kohonen Self-Organising Maps are used for the first time on a dedicated sample of young stars observed in the first year of TESS's primary mission, in order to sort light-curves by topology and look for distinct variability classes. This analysis forms the first step in the YOUNGSTER programme, aiming to use knowledge of young star variability to inform more targeted detrending in future young exoplanet searches. Finally, this thesis presents work on tracing exoplanets through time on human timescales, updating and improving the ephemerides for all previously known Kepler planets (22) and candidates (4) which were observed with sufficient signal to noise in TESS. It also explores any transit-timing variations seen for these objects, including intriguing results for HAT-P-7b, Kepler-411d, K00075.01 and K00076.01.

In this thesis I study stellar obliquities across a range of companion masses, and in new regimes, that aims at constraining theories of planet formation and evolution. I begin in Chapter 1 with an introduction to the state of the field of extrasolar planets, key discoveries that have motivated previous studies on the misalignments between planetary orbits and stellar spins, and highlight the gaps in our knowledge where this thesis aims to make an impact. In Chapter 2, I outline the models and tools that underpin the analysis of transit light curves and high-resolution spectra in subsequent chapters. In Chapter 3, I apply these tools to the discovery of binary systems of various mass ratios. Two such systems are rare brown dwarfs whose discoveries help calibrate models of sub-stellar evolution, and the connection to giant planet formation and evolution. In Chapters 4-6, I present new measurements of stellar obliquities across a range of companion masses. In Chapter 4 I consider two systems hosting small planets. I demonstrate a misalignment the stellar spin and the orbit of a planet twice the size of Earth. This discovery is consistent with some disc-free migration scenarios, and provides the first observational evidence of its kind that super-Earths may form far from their star. In Chapter 5, I consider a sample of 13 giant planets orbiting cool stars in weak-tide regimes. I show that their host stars display a variety of obliquities, contrary to similar planets orbiting closer to their star. Such an effect is consistent with the expectation from tidal evolution, but has not yet been tested on this scale. In Chapter 6, I study the spin angular momentum of the primary component of a binary star hosting a circumbinary planet. I demonstrate that the star is aligned with the binary and planet orbit, providing an important constraint on the formation of binary stars and circumbinary planets. Finally, in Chapter 7, I conclude and offer some thoughts on future prospects.

There are two main goals of this thesis. The first is developing our so-called 'downsampling' procedure: an approximation technique for reproducing the likelihood function (in simulation) in relatively short time, of the sort of signals we expect to observe with the LISA Gravitational Wave (GW) detector, such as the inspiral part of binary black-hole mergers. The procedure is tested on a variety of signals and shown to provide highly accurate reproductions of the likelihood function up to a few thousand times faster than our estimates of convergence time considering the fastest type of analysis one could perform (frequency domain, stationary noise) using all the data points. The second goal is to understand the features and effectiveness of using multi-band GW data in data analysis, in particular on expansion of the signal model to include anticipated physical effects that modify the low-frequency part of the waveform, such as acceleration and relativistic time-delays. In addition, we extend the model to cover two well-known modified gravity theories, to test the assertion that multi-band GW data analysis will provide strong theory constraints. Since the degree to which the various time delays (which have their own parameter space) modify the waveform ranges from being negligible to significant and essential to model, we devise a formalism for splitting the time-delay parameter space into distinct regions that either fully model, 'dimensionally reduce', or neglect the time-delays from the model. The advantages of doing this are that one acquires posteriors that are considerably more informative, since the waveform is not being 'over-modelled'. The posteriors we obtain confirm what one would expect to see (particularly in terms of parameter degeneracies) after considering and comparing the effects on the GW phase that arise from varying different combinations of parameters. In the fully modelled region, it is possible to recover well constrained time-delay parameters (i.e., the Keplerian orbital parameters, including supermassive black hole mass), but it is very difficult to derive general results about the expected behaviours of posteriors since the time-delay functions are complicated functions of time. The behaviour of posteriors of the 'parameter reduced' models are far easier to predict and understand, but the new parameters are relegated to nuisance parameters. The modified gravity theories we analyse are not well-constrained simply by multi-band observations alone; very high SNR appears to be the more important factor, but even then, different aspects of the nature of the waveform modifications (depending on the theory) can lead to significant bias, or render the effects too weak for GW astronomy to provide any useful constraints. By inspecting one-parameter families of posterior distributions (treating the posteriors as functions of some model parameter or signal property) and observing their structural evolution as that parameter is varied, we uncover and discuss some interesting features and behaviours of the distributions. Topics for future study are highlighted and include extensions and refinements of the waveform, population, and detector models, further studies of the Kepler parameter space division scheme, and posterior sampling issues to be addressed.

The evolution of galaxies is heavily influenced by the environment in which they are found. Dense regions of the Universe can quench a galaxy's star formation by removing its gas, and transform a galaxy's morphology from being disk-dominated to bulge-dominated. Galaxy clusters represent the most extreme example of this: compared to the cosmic field, these highly dense environments contain a much greater fraction of red, elliptical galaxies. However, a subtlety of this is that a galaxy's evolution is not just driven by the environment in which it is currently found, but is also influenced by the environments through which it has previously passed. For example, a galaxy in the outskirts of a cluster may also have been quenched during pre-processing, meaning that it was accreted onto the cluster through a cosmic filament or as a member of a group, both of which can quench star formation. Alternatively, it could be a backsplash galaxy: one that has passed through the dense cluster environment in the past, but subsequently left and now resides in the cluster outskirts. Pre-processing, backsplash galaxies, and the direct impact of a cluster environment are difficult to disentangle from each other, because observations cannot provide us with the full histories of galaxies. This makes observational studies of galaxy environment difficult, particularly nearby to clusters. For instance, it is not clear how common are backsplash galaxies around clusters, how their frequency varies between clusters, or whether we can identify which galaxies are indeed backsplash. It is also not fully understood what happens to galaxy groups nearby to clusters, and how a cluster can influence the impact of a group on its constituent galaxies. These questions are challenging to answer with observational data, but can be approached using cosmological simulations that complement the available observations. In this thesis, we use data from The Three Hundred project, a suite of hydrodynamical simulations of large galaxy clusters, to study the environmental histories of galaxies in and around clusters. We begin by establishing that these simulations are fit-for-purpose, by comparing them to equivalent dark matter-only simulations. We find that, compared to the hydrodynamical runs, our dark matter-only simulations underestimate the number density of galaxies in the central regions of both groups and clusters, for which we discuss several potential causes. This indicates that hydrodynamical simulations are necessary in studying cluster substructure, as the evolution of galaxy groups will be different in dark matter-only simulations. Having established this, we then use these hydrodynamical simulations to examine how galaxy groups evolve as they approach, enter, and pass through a cluster. These galaxy groups become gravitationally unbound very quickly, losing most of their member galaxies less than 1 Gyr after entering a cluster. In fact, the overwhelming majority of groups do not survive a full passage through a cluster, meaning that any groups nearby to a cluster are almost certainly on their first infall towards the cluster centre. We then investigate backsplash galaxies and find that, on average, over half of all galaxies between R200 and 2R200 from their host at z=0 are backsplash galaxies. However, this fraction depends on the dynamical state of a cluster; dynamically relaxed clusters, which are isolated and accreting new material slowly, have a far greater fraction of backsplash galaxies in their outskirts. This backsplash population is mostly developed in the last few Gyr, and is dependent on the recent dynamical history of a cluster. This work uses simulations to shed light on the different processes that galaxies can experience during their accretion onto a galaxy cluster. More importantly though, the findings from these simulations can be applied to real, observational studies. The dynamical state of a cluster is a measurable property, and so can be used to infer how 'contaminated' the population of infalling galaxies in an observed cluster's outskirts will be by backsplash galaxies. Furthermore, galaxy groups observed nearby to a cluster are on their first infall, and so will contain very few backsplash galaxies. Any galaxies that are members of a group inside of a cluster will have experienced the central region of a group, but have likely only joined the cluster very recently. Simulations will be a valuable tool to complement upcoming surveys like the WEAVE Wide-Field Cluster Survey and Euclid -- due to begin in late 2022 and 2023 respectively -- and will allow us to more deeply interpret this observational data, and infer the environmental histories of galaxies.

This thesis outlines the work I undertook at Cardiff University as part of my degree for doctor of philosophy. The document is split into two distinct parts. The first section covers the development of new photodetector electronics that successfully reduced the dark noise of the LIGO experiment by a factor of 6.5 at 2 kHz. This reduction in the dark noise is mainly achieved by filtering the majority of the DC photocurrent from the audio readout path and increasing the transimpedance resistor by a factor of 100. The new readouts are projected to improve the overall DARM sensitivity of LIGO by 10% at 2 kHz in the next observing run. The second section covers the implementation of a laser auto-alignment system using differential wavefront sensing for the ALPS II dark matter search. The autoalignment system is currently undergoing further commissioning but has already shown an improvement in low frequency cavity noise of roughly an order of magnitude which will help the ALPS II detector reach the target design sensitivity in its search for axion-like particles in the upcoming science run at the end of 2022. Both projects involved the development of advanced techniques for ultra-precise interferometry experiments but the goals and techniques implemented were vastly different. This document outlines the approach in tackling these projects and will hopefully prove a useful guide for others in the future.

The solar atmosphere has a peculiar property where its outer layer, the corona, is hotter than its base, the photosphere; this issue is termed the coronal heating problem. One possible solution is that small solar flares-microflares and nanoflares-occur in such a frequency that they produce more net energy than their larger counterparts, that are insufficient in themselves, to heat the corona. To investigate this premise, microflare extreme ultraviolet (EUV) and X-ray emission must be studied to determine their production mechanisms and energetics. In this thesis, we focus on the observations of, and analytical tools for, the study of high energy microflare emission with particular emphasis on X-ray spectroscopy. In Chapter 1, we introduce the necessary context for microflares with respect to the solar atmosphere as well as a description of their high energy EUV and X-ray emission and how this can be modelled mathematically. Chapter 2 provides an overview of the hard X-ray, soft X-ray, and EUV instruments used throughout this thesis with specific attention given to the Nuclear Spectroscopic Telescope ARray (NuSTAR). NuSTAR is an astrophysical X-ray telescope capable of observing the Sun with direct imaging spectroscopy providing a unique sensitivity >2.5 keV. In Chapter 3, we investigate ten NuSTAR microflares that originated from active region AR12721 and occurred between 2018 September 9-10. The ten microflares are all weak sub-A GOES class and still reach temperatures up to ∼10MK. One microflare shows direct evidence for non-thermal emission and eight of the ten show indications of photospheric magnetic flux cancellation in proximity to their footpoints. Chapter 4 then investigates the weakest microflare from Chapter 3 further, finding it to be the weakest active region X-ray microflare. Chapter 5 provides an overview of X-ray spectral fitting approaches where the nuances of X-ray spectral data analysis, utilising different fitting and statistical methods, are discussed. We then describe two widely used X-ray spectral fitting programs, OSPEX and XSPEC, detailing their advantages and limitations before introducing a new Python X-ray spectral fitting tool called Sunxspex. Sunxspex is optimised for solar data products and aims to combine the capabilities of OSPEX and XSPEC into one program. Using examples fromNuSTAR, the Reuven Ramaty High-Energy Solar Spectroscopic Imager (RHESSI), and Solar Orbiter's Spectrometer/Telescope for Imaging X-rays (STIX), we showcase the abilities of Sunxspex in Chapter 6. We re-analyse data from existing NuSTAR and RHESSI studies, finding good agreement between fitted model parameter values, while executing new analysis with the same data and STIX data that was not possible with software like OSPEX and XSPEC. Chapter 7 describes the nested sampling algorithm which is made available in Sunxspex and allows for quantitative comparisons between different models being fitted to the same data. Finally, Chapter 8 presents the analysis of five microflares and a jet observed by NuSTAR. The first two microflares are observed during the 2020 January solar observation campaign while the last three microflares and the jet are from the 2021 November campaign. The microflares show evidence of temperatures>10MKbeing present with several of the microflares showing potential direct evidence of non-thermal emission. The jet reaches quiescent, non-flaring active region temperatures of ∼4MK while being located far from the closest active region.

The field of hadron spectroscopy is informed by experiments which study excited states of the nucleon by detecting the decay products following excitation with a hadronic, leptonic or electromagnetic probe. Experiments using a polarised beam of photons provide an important contribution to the world data. This thesis describes the analysis of data from one such experiment, the CEBAF Large Acceptance Spectrometer at Jefferson Laboratory, Virginia, USA. In the experiment, an incident beam of electrons passes through a diamond radiator producing a linearly polarised photon beam of energy from 1.1 to 2.1 GeV on a liquid hydrogen target. The analysis method and results for two reaction channels are presented. In the first analysis, the polarisation observables {Σ,P,T,Ox,Oz} for the reaction −→γ p → K 0 S Σ + are extracted simultaneously using likelihood sampling with Markov-Chain Monte-Carlo. The results are presented as angular distributions (cosθK 0 S ) for bins in photon energy Eγ for a total of 21 bins. The values extracted for T, Ox and Oz are a first measurement of these quantities for this reaction and the beam asymmetry (Σ) measurement supplements the one previous measurement for this quantity and extends the energy range over which it has been extracted. The measurement of the recoil polarisation, P, is consistent with previous measurements. In a second analysis, the spin density matrix elements (SDMEs) for γ p → p φ are extracted using similar analysis techniques. The results are presented as angular distributions (cosθφ ) for bins in photon energy Eγ for a total of 24 bins. This work is the first to extract all nine SDMEs for the full angular range of the φ meson production. A comparison is made to the predictions of the Vector Meson Dominance (VMD) model and the results provide evidence for contributing processes other than VMD, particularly at more backward angles of the φ meson production. The data also indicate non-helicity conserving process in both the s-channel and the t-channel. The results from the first analysis have been passed to the Juelich-Bonn theory group and the effect of including the new data within their dynamical coupled channel model is described. Several nucleon resonances are affected and the impact on the pole positions, widths and photon decay amplitudes of the affected resonances is described.

The first detection of a gravitational wave by LIGO and Virgo is a milestone for the study of compact objects in the Universe. Since it took place in 2015, a few tens of detections have been confirmed by the LIGO Virgo Collaboration (LVC). Afterwards, other collaborations have confirmed and extended such catalogues with independent analysis on the same datasets. Through exquisite experimental devices, sophisticated data analysis algorithms, and an accurate interpretational effort, these observations constitutes now an invaluable body of knowledge, and are a key piece of observational evidence for our understanding of the astrophysical population of binary systems. The science case is set to continuously expand as detections increase, and space-based gravitational-wave observatories are going to complement current ones with an otherwise inaccessible window to the "gravitational-wave sky". In this thesis I report on my contributions on a few scientific investigations, including: (i) the development of statistical tools for the simultaneous inference on multiple sources and multiple populations of compact binaries; (ii) the development of a framework for parameter estimation on space-based detector observations, focusing on stellar mass binary black-holes and binary white dwarfs systems; (iii) the predictions of yet unob- served phenomena (e.g. gravitational lensing of gravitational waves) or specific signals (e.g. the stochastic foreground of gravitational waves) and their mutual connections; while developing the tools above I have had the opportunity to provide some insight on: (i) the astrophysical population of binary black hole masses and spins, and their distribution across redshift, therefore providing observational evidence in support of different formation channels; (ii) the future detectability of binary white dwarfs in satellites galaxies of the Milky Way with space-based detectors, with implications on the assembly history of satellite galaxies.

Studying exoplanets enables us to better understand our place in the Universe. Transiting exoplanets, where the planet passes in front of its host star, provide us with a great wealth of information on the properties of the planet. In this thesis, I use high precision photometry from the Next Generation Transit Survey (NGTS) and the Transiting Exoplanet Survey Satellite (TESS) to discover and characterise transiting exoplanets. I also include spectroscopic observations to monitor the radial velocity of the host stars in order to measure the planet mass. I present the discovery of NGTS-12 b, a 1.05 RJ transiting exoplanet orbiting a V = 12.38 mag star with a period of 7.53 days. Using spectroscopic observations from HARPS, I measured the mass of the planet as being 0.21 MJ . I present a novel observing method that involves the use of multiple NGTS telescopes to simultaneously observe the same star. I tested this method on the bright (V = 9.35 mag) exoplanet hosting star WASP-166 achieving a photometric precision of 152 ppm per half hour. This is some of the highest photometric precision data for ground-based observations of bright stars, opening up multiple avenues for new NGTS science. Using this technique I was able to observe a transit of the 542 day period exoplanet HIP-41378 f. These observations revealed the presence of transit timing variations for this planet. Planet formation is generally well understood, however there are a number of planetary systems which challenge the current theories. One such system is the WASP-47 system, which contains a hot Jupiter with a close orbiting inner super-Earth companion. I present new spectroscopic observations obtained with ESPRESSO and analysis to improve our understanding of the properties of the WASP-47 planets. This analysis included the use of a Gaussian Process to model a stellar activity signal present in the radial velocity data. The Gaussian Process was motivated by measurements of the rotation period of the host star. The results of my analysis have begun to reveal the presence of a potential new population of exoplanets, one which if better studied might shed more light on how this remarkable system formed. Finally, I present a study of the occurrence rates of transiting gas giant exoplanets with low-mass stellar hosts. These occurrence rates are currently not well constrained and so the results of this project will be of great importance for understanding the formation mechanisms of giant planets. This search has yielded fourteen giant planet candidates from a sample of 91,306 low-mass stars, and radial velocity monitoring with ESPRESSO is underway to confirm the candidates. These candidates will form the basis of a calculation of the occurrence rates for these systems, which will be performed over the coming months.

This thesis focuses on how the combination of far infrared (FIR) and radio emission can be used to better understand galaxy evolution. Radio and FIR emission is linked to star formation, which plays a major role in the evolution of galaxies. I have combined FIR observations from Herschel and radio observation from the LOFAR radio telescope in order to take advantage of this. Chapter 2 details the construction of a joint FIR-radio catalogue, using the likelihood ratio method, to link the radio catalogue to a multiwavelength optical/NIR catalogue. Subsequently the FIR fluxes were measured by deblending the FIR maps using the radio galaxies as positional priors. This method creates a perfect sample for measuring the far infrared radio correlation (FIRC). Chapter 3 focuses on the measurement of the FIRC at 150MHz with a mass complete sample. I find that the FIRC is primarily dependent on the stellar mass of the host galaxy with a minor dependence on the redshift. Finally, chapter 4 describes how I used XID+ to measure the flux posterior for galaxies at 150MHz that were undetected in the LOFAR radio catalogues. I found that XID+ is able to accurately measure the flux of point like radio galaxies. These results were used to remeasure the FIRC and evaluate the best approach for extending XID+ to work on radio maps.

Stellar halos of galaxies retain crucial clues to their mass assembly history. It is in these galactic components that the remains of cannibalised galactic building blocks are deposited. For the case of the Milky Way, the opportunity to analyse the stellar halo's structure on a star-by-star basis in a multi-faceted approach provides a basis from which to infer its past and assembly history in unrivalled detail. Moreover, the insights that can be gained about the formation of the Galaxy not only help constrain the evolution of our Milky Way, but may also help place constraints on the formation of other disc galaxies in the Universe. This thesis aims to make progress toward answering the most fundamental question in the field of Galactic archaeology: "How did the Milky Way form?" Through the effort to answer this question, this thesis presents new insights into aspects of the history of assembly and evolution of our Galaxy and measurements of the structure of various of its Galactic components. Providing further insight into the accretion history and mass assembly of the Milky Way, I present a detailed analysis of the properties of Milky Way halo stars in the heart of the Galaxy contained in both the APOGEE and Gaia data sets. I present evidence for the discovery of a new halo substructure (whose progenitor we attribute the name of "Heracles") that, given its chemical composition and dynamical properties, is likely to be the debris from a major building block of the Milky Way. I also compare its properties with expectations from the EAGLE numerical simulations to ascertain its nature, and make a quantitative prediction of the stellar mass of the disrupted satellite galaxy to comprise approximately one third of the estimated total stellar halo mass. To ascertain the reality and nature of halo substructures, and to place further constraints on the mass assembly history of the Galaxy, I perform a detailed qualitative and quantitative analysis of the chemical compositions of halo substructures in the Milky Way with APOGEE and Gaia data. The findings from this study revealed that many halo substructures identified in recent years, conjectured to be the debris from individual satellite accretions, likely belong to the Gaia-Enceladus/Sausage accretion event. They also showed that the Heracles halo substructure is statistically different from in situ populations given its chemical compositions, further confirming its accreted nature. To understand how much mass dissolved and/or evaporated globular clusters (GC) contribute to the total stellar halo mass budget, I perform a density modelling analysis of stellar halo populations. By identifying GC escapees using a Gaussian mixture modelling and chemical tagging procedure, I model their density distribution accounting for the APOGEE selection function and assess their ratio to the halo field. The main finding of this work showed that in the inner ∼2-3 kpc from the Galactic centre, there is a much higher incidence of dissolved/evaporated GC stars that is on the order of five to six times larger than in the outer ∼10 kpc region. In order to decipher the origin of the Galactic GC system, I undertook a study aimed at comparing the chemical composition of previously categorised GC subgroups classified based on their orbits. More specifically, by determining a homogeneous sample of GC star members in the APOGEE DR16 survey, and comparing the mean [α/Fe] and [Fe/H] abundances of GCs with field populations, I was able to place constraints on the origin of GCs in the Milky Way, that in turn help place constraints on the accretion history of the Galaxy. The above results place constraints on our current understanding of the accretion and mass assembly history of the Milky Way. In discovering new halo substructures, assessing the reality of known ones, modelling the density of stars contributed from dissolved/evaporated GCs, and deciphering the origin of the Galactic GC system, I have tackled the question of "How did the Milky Way form?" from numerous different angles. All the findings contained in this thesis help pave the way for future work towards the goal of fully reconstructing the assembly history of our Galaxy and using that understanding to formulate robust and general models for the formation of disc galaxies.

Over the past few decades numerical simulations with collisionless dynamics have reached a consensus about the form of the internal structure of dark matter (DM) haloes. However, the theoretical origin of these is still poorly understood. In this thesis I focus on studying the link between the initial primordial power density fluctuations and the internal structure of collapsed DM haloes today. This is the main focus of this thesis, and is split into three main parts; in the first chapter I study the dependence of the internal properties of DM haloes, primarily the mass density and pseudo phase space density (PPSD) profiles, on initial density fluctuation by systematically varying both the amplitude and slope of the linear power spectrum. It is observed that a number of previously assumed universal results break down when the initial power spectrum deviates from the CMB normalised case with the density profiled deviating strongly from an NFW form, with steeper slopes than −3, and the PPSD power law slope now exhibiting a clear cosmological dependence. In the second part of this thesis the simulations introduced in the first section are used to develop a model to predict the density profiles of DM haloes for a general mass, redshift and cosmology. To fully describe the density profiles observed in simulations two parameters are required: concentration, c, and an additional 'shape' parameter, α. I demonstrate that these two parameters can be expressed as a single, universal function of peak height using an appropriately chosen window function, allowing for a simple model to be developed. In the final section of this thesis I explore the joint effects of warm dark matter (WDM) and baryonic effects on the satellite populations of Milky Way mass systems. Here it is found that there is a strong degeneracy between the feedback (subgrid) parameters and the assumed WDM strength that should be taken into account when placing constraints on WDM and similar cosmological extensions.

Heart failure will contribute to the death of one in three people who read this thesis; and one in three of those who don't. Although in order to diagnose patients' heart condition cardiologists have access to electrocardiograms, chest X-rays, ultrasound imaging, MRI, Doppler techniques, angiography, and transesophageal echocardiography, these diagnostic techniques require a cardiologist's visit, are expensive, the examination time is long and so are the waiting lists. Furthermore abnormal events might be sporadic and thus constant monitoring would be needed to avoid fatalities. Therefore in this thesis we propose a cost effective device which can constantly monitor the heart condition based on the principles of phonocardiography, which is a cost-effective method which records heart sounds. Manual auscultation is not widely used to diagnose because it requires considerable training, it relies on the hearing abilities of the clinician and specificity and sensitivity for manual auscultation are low since results are qualitative and not reproducible. However we propose a cheap laser-based device which is contactless and can constantly monitor patients' heart sounds with a better SNR than the digital stethoscope. We also propose a Machine Learning (ML) aided software trained on data acquired with our device which can classify healthy from unhealthy heart sounds and can perform biometric authentication. This device might allow development of gadgets for remote monitoring of cardiovascular health in different settings.

The discovery of gravitational waves opened a new way to look at the Universe and offered new opportunities to shed light on the still unknown aspects of physical sciences. The work presented in this thesis wants to give a contribution to the development of this new type of research: the author chose to focus on the improvement of the instruments able to detect the gravitational waves. This field is important to make the detectors more sensitive, in order to see more gravitational-wave sources and help to complete the mosaic of the astrophysical science. In particular, the detectors currently in use are interferometers, which are especially blind in a range of frequency below 30 Hz: this affects the chance to detect sources emitting in this frequency band. This lack of sensitivity is mainly due to seismic motion, and the work presented in this thesis focussed on new techniques to lower the noise sources and allow the instruments to be sensitive below 30 Hz. During the studies, the development and test of devices capable of potentially reducing the seismic motion have been performed, such as optical levers for tilt motion reduction and laser stabilization for low frequency readout; a new concept of the seismic system on one of the interferometers (LIGO) has also been proposed. The optical levers can in principle reduce tilt motion below 1 Hz; the use of capacitive position sensors in a new software configuration for LIGO can help to suppress ground motion by a factor of 3 in order of magnitude below 0.1 Hz. A competitive frequency stabilization to 3.6 times 10³ Hz√Hz at 1 Hz for readout at low frequency is possible with a compact and easy to handle setup. These results are promising to provide suppression of the seismic motion in the bandwidth of interest and show that it is possible for a ground-based instrument to be seismically more stable and capable of detecting gravitational waves where it is now forbidden.

Cosmology is at a crossroads. Experiments are providing an unprecedented amount of data that, in theory, should lead to clear solutions to the many open questions in cosmology. However, with new data comes new questions and recently uncovered tensions between the predictions of the standard model of cosmology and observations are leading some to question the very foundations on which the standard model is built. To explore the vast cosmological landscape, numerical simulations are often employed, but given the broad parameter space that needs to be explored other faster (but more approximate) methods need to be adopted to maximise the coverage and the possible extensions surveyed. In this panorama one of the options is the halo model, a simple and elegant way to study the clustering of matter in the Universe. However, this method is not free from assumptions and associated uncertainties. In this thesis I explore the uncertainties associated with the halo model making use of cosmological numerical simulations. I use the BAHAMAS simulations to obtain data products such as the mass density profiles of the haloes and the number density of haloes over a wide range of masses and I use these quantities in the halo model formalism in order to make a self-consistent comparisons against the simulations results. Aside from this application, I calibrate a fitting function on the Einasto function, which has been shown to be a good representation of the matter distribution inside haloes, and I use a standard form for the halo mass function. Comparing against the simulation matter power spectrum at different redshift, I show the accuracy of the halo model predictions is strongly dependent on the mass definitions used with differences over 50%. In particular, the transition region between the 1-halo and the 2-halo terms and in the smallest scales sampled (k≈ 10 h/Mpc). This picture applies to both collisionless and hydrodynamical simulations, where galaxy formation processes are taken into account. In contrast to the poor ability in reproducing the matter clustering, the halo model can reproduce the relative impact of baryons on the matter clustering to a competitive accuracy (<5%) in line of next-generation observations predictions. In the second part of this work, I analyse the halo model applications of large-scale structure observables as gravitational weak lensing and thermal Sunyaev-Zel'Dovich effect. To explore these observables, I have built the halo model using the electron pressure inside haloes (relevant for the tSZ effect), and I have made several realisations of the matter power spectrum up to z=3 for the lensing observables, in both the collisionless and hydrodynamical cases. In this analysis, I have compared against observational data (e.g., KiDS-450 survey and Planck) and results obtained from light-cones from the BAHAMAS simulations. I examined the dependence of the results on the different mass definitions and the baryonic effects, in particular the baryonic suppression that can be inferred from this set of observables.

The study of the lower flaring atmosphere of the Sun is one facet of understanding the complex physics involved in solar flares and their effect on space weather and the Earth. Despite a rich history of investigation into the study of the lower flaring atmosphere, there are still many unanswered questions in regards to the mechanism of energy deposition and the response to such energy being injected into the atmosphere. This thesis aims to provide tools for future researchers to rigorously explore these problems. In particular, this thesis looks at how machine learning - with a particular focus on deep learning - can improve data storage and analysis pipelines as well as uncover new results from data that were not feasibly possible before. In Chap. 1, the standard model of a solar flare is introduced and its extension to three dimensions explained. This allows for the definition of a flare ribbon - the brightest points in the lower solar atmosphere resulting from direct heating from a flare - which is a key observational feature whose origin is explored in later chapters. A brief history of study on flare ribbons is then given with a particular focus on the asymmetries in spectral lines that show clear flare ribbons. These asymmetries link directly to the velocity field in the flaring atmosphere as a static atmosphere would yield symmetric profiles. This gives a direct diagnostic of the motion happening in the atmosphere as it is heated and the ribbons evolve. In Chap. 2, the field of deep learning is introduced from its inception to the current models used today. This chapter covers how to build and train deep neural networks and some best practices when implementing these tools. The telescopes and detectors used to obtain the data analysed in Chaps. 4 - 6 are described in Chap. 3. In this chapter, the inner workings of the Swedish 1-m Solar Telescope's CRisp Imaging SpectroPolarimeter (SST/CRISP), Hinode's Solar Optical Telescope (Hinode/SOT) and Solar Dynamics Observatory's Atmospheric Imaging iii Assembly (SDO/AIA) is described. Chap. 4 introduces a deep convolutional neural network (CNN) trained on Hα images from Hinode/SOT for solar image classification. This is trained to distinguish between five classes of solar features prominent in Hα: filaments, flare ribbons, prominences, sunspots and the absence of any of the other four features. The final model has a validation accuracy of 99.2% misclassifying only one image in the validation dataset. The trained CNN is then tested with adversarial examples from SDO/AIA UV continua and EUV spectral line images where the features look perceptually different but still identifiable to the human eye. This demonstrates that the network cannot identify these features in different wavelengths well and to extend this network to non-visible wavelengths, the training set must be expanded to include such wavelengths. The trained CNN in this chapter is used further in Chap. 5 for transfer learning - the process of using a trained deep learning model to influence the training of another, related deep learning model. In Chap. 5, a method based on deep learning for correcting the atmospheric effects in optical solar flare observations is presented. This takes the form of a fully convolutional autoencoder trained on data from SST/CRISP imbued with synthetic seeing described by the model developed in the first sections of the chapter. The trained model works well on the validation dataset showing accurate reconstruction of both spatial and spectral elements of the data. SST/CRISP data with real atmospheric seeing is then corrected by the trained model. The sources of error in this reconstruction are discussed with a coarse error estimate on the recovered intensity values used. Then in Chap. 6 a novel deep learning method for estimating the parameters of the flaring atmosphere from observations is presented - an Invertible Neural Network (INN). The INN is trained on synthetic flare data produced by the one dimensional radiation hydrodynamics code RADYN with near-perfect restoration of the atmospheric paramters during validation. This is then applied to a single pixel from a CRISP image to show the power of this method in disentangling the ambiguity in the velocity field responsible for observed asymmetry in the spectrum. This method is then applied to flare ribbons as a whole - which are selected through a combination of a Gaussian Mixture Model (GMM) and Density-Based Spatial Clustering of Applications with Noise (DBSCAN) - to determine the specific motions of the flaring velocity field responsible for the observed spectral line asymmetries.

In this thesis, I use large scale state-of-the-art hydrodynamical cosmological simulations to investigate the Ly-α absorption properties of protocluster regions at redshift z∼2.4. Using the IllustrisTNG simulations, I investigate how the AGNproximity effect and hot, collisionally ionised gas arising from gravitational infall and black hole feedback changes the Ly-α absorption associated with the z≃2.4 progenitors of M≥10^14 M_☉ clusters. Within these protoclusters, I find that galaxy overdensities exhibit a weak anti-correlation with Ly-α transmission in tomographic maps of the intergalactic medium (IGM), but local HI ionisation enhancements due to hot T > 10⁶ K gas or nearby AGN can disrupt this relationship within individual protoclusters. On average, however, I find that strong reductions in the IGM neutral fraction are limited to within ≲5 h^−1 cMpc of the dark matter haloes. Local ionisation enhancements will therefore have a limited impact on the completeness of protocluster identification in tomographic surveys if smoothing Ly-α transmission maps over scales of ∼4 h^−1 cMpc, as is typically done in observations. However, if calibrating the relationship between the matter density and Ly-α transmission in tomographic maps using simple analytical models for the Ly-α forest opacity, the presence of hot gas around haloes may result in systematically lower estimates of the z=0 descendant mass of the most massive protoclusters. I then use simulations from the Sherwood, EAGLE and Illustris projects to examine the Ly-α absorption signatures of z∼2.4 protoclusters with a descendant mass of M≥10^14 M_☉. I find there is a weak correlation between the mass overdensity, δ_m, and the Ly-α effective optical depth relative to the mean, δ_τeff, averaged over 15 h^−1 cMpc scales; however, the scatter in the δ_m-δ_τeff plane means it is not possible to uniquely identify large scale overdensities with strong Ly-α absorption. Assuming perfect removal of contamination by Ly-α absorbers with damping wings, more than half of the remaining sightlines with δ_τeff>3.5 (CoSLAs) trace protoclusters. It is furthermore possible to identify a model dependent δ_τeff threshold that eliminates the contamination from filamentary structure and Lyman-limit systems that are preferentially aligned along the line of sight. However, such regions are rare: excluding absorption caused by damped systems, less than 0.1% of sightlines that pass through a protocluster have δ_τeff>3.5, meaning that any protocluster sample selected in this manner will also be highly incomplete. Finally, I use the TNG300-1 simulation from the IllustrisTNG project to study the extent to which simulation box size-and the resulting presence of the most massive protoclusters-is responsible for the differences between the CoSLA populations observed in large N-body simulations and those in comparatively small hydrodynamical simulations. Using the better statistics afforded by the larger simulation volume, I confirm that CoSLAs are rare objects, finding a number of CoSLAs per unit redshift dn/dz∼1×10^−2 at redshift z=2.444. Furthermore, I find the δτeff-δm relation in a 205^3 h^−3 Mpc^3 volume is consistent with that for a smaller 75^3 h^−3 Mpc^3 box size, with systems with δ_m>1 existing over a wide range of δ_τeff values. I find the CoSLAs in TNG300-1 to have a median δ_m=0.45±0.05, which remains somewhat lower than the δ_m=0.64±0.38 previously found in larger N-body simulations. These findings suggest that the differences between the CoSLA populations in hydrodynamical and N-body simulations are not solely the result of the differing simulation box sizes. Instead I suggest that these discrepancies are caused by the modelling of the hot, collisionally ionised, gas that surrounds massive haloes in protocluster regions.

The subject of this thesis is cosmological implications of string compactifications understood in a broad sense. In the first half of the thesis, we will begin by reviewing the four-dimensional description of the tree-level perturbative type IIB action. We will then introduce a number of open questions in cosmology and their relevance with regards to the remainder of the thesis. We will first explore some of these questions from the perspective of effective field theories motivated by supergravity. In particular, we provide a description of a naturally light dark energy field in terms of the clockwork mechanism and the Dvali-Kaloper-Sorbo four-form mixing. We study its possible UV completion and show a no-go for its embedding within perturbative type IIA supergravity. We also discuss the coincidence problem for dynamical models of dark energy consistent with a quintessence field slowly rolling down a potential slope, of the type one would expect from the asymptotics of moduli space. As it rolls, a tower of heavy states will generically descend, triggering a phase transition in the low energy cosmological dynamics after at most a few hundred Hubble times. As a result, dark energy domination cannot continue indefinitely and there is at least a percentage chance that we find ourselves in the first Hubble epoch. In the second half of the thesis, we introduce the effects of perturbative and nonperturbative corrections to the tree-level type IIB action. We then focus on obtaining a viable model of quintessence from the type IIB effective field theory. However, we are able to show that such a model must have a non-supersymmetric Minkowski vacuum at leading order. Furthermore, it must necessarily take the form of axion hilltop quintessence. When we consider the effects of quantum fluctuations during the early Universe, we see that such models must have extremely fine-tuned initial conditions to describe a slow-rolling scalar field at present times. We conclude that quintessence faces more challenges than a true cosmological constant, to the point that quintessence is very unattractive for model building modulo a ruling out of the cosmological constant by observations. Following this line of reasoning, we consider whether other perturbative corrections can generate de Sitter solutions in an appropriate setting. In particular, we consider the effects of higher curvature corrections in the Gauss-Bonnet term. Remarkably, we are able to show that, for the particular setting of a fluxed runaway potential motivated by heterotic supergravity, the curvature corrections reduce the space of solutions.

Primordial black holes (PBHs), black holes that are formed in the early Universe from the collapse of over-densities, are a candidate for non-baryonic cold dark matter. The fraction of dark matter made up of PBHs has been constrained using multiple mechanisms including evaporation, gravitational lensing, gravitational waves from mergers, large-scale structure, accretion and the disruption of wide binaries. In this thesis we review the work on wide binaries and improve the reliability of these constraints, primarily by introducing a more physically motivated initial semi-major axis distribution for our simulated binaries. We used Monte Carlo simulations to implement halo wide binary encounters with PBHs over a period of 10Gyr. The binaries are placed in a sea of perturbers and encounters are calculated using the impulse approximation. Broken binaries are kept in the simulations and used to calculate constraints, since their peculiar velocities are similar enough to be mistaken for binary stars. Our simulation results are consistent with previous work for a log flat initial semi-major axis distribution, and when broken binaries are removed from the simulations. We calculate constraints by using a modified χ² test to compare the simulated binary distribution with a catalog of observed binaries. A χ² test provides p-values whilst also quantifying the goodness of the best fit parameters. We use a modified version in order to correct for the small number of binaries used in the comparison. Our constraints are re-scaled to take into account the non-uniform dark matter density along the binary Galactic orbits. To find the value by which to re-scale the constraints, we calculate the time-averaged dark matter density along each binary orbit for which we have sufficient data. Our final constraints are much weaker than those calculated previously, primarily due to the inclusion of unbound binaries and a more physically motivated semi-major axis distribution. These constraints are subdominant to previously calculated constraints from other effects such as microlensing, gravitational waves and accretion in the PBH mass range 1M⊙-1000M⊙, allowing PBHs to make up at least 20% of dark matter. More detailed simulations are needed to find tighter constraints on the PBH mass fraction: simulations of the Galactic potential would take into account the Galactic tide and disk and the variable dark matter density. These constraints could also be made more reliable by using a larger catalog of binary stars, such as a subset of GAIA DR2.

Strong galaxy-scale gravitational lensing provides a powerful means of studying galaxy formation, constraining cosmology and understanding the evolution of large-scale structure. The presence of a foreground lensing galaxy deflects the light rays of a background source into multiple images, with their configuration dependent upon the lens mass distribution. Modelling of this distribution to reproduce the lensed images not only aids in measuring the dark matter content of the foreground galaxy, but allows for the reconstruction of the background source in order to study high-redshift galaxy populations. Modelling is typically performed by relatively slow parametric parameter-fitting techniques requiring manual inspection. However, upcoming large-scale surveys like the Legacy Survey of Space and Time (LSST) and Euclid will discover tens of thousands of strong lenses, with this vast quantity driving the growing use of automated machine learning for both identifying and modelling such systems extremely quickly. Convolutional neural networks (CNNs) can extract information from lensed images in order to predict parameters of the lens mass model, but require large simulated data sets for training and testing. In this thesis, I explore the effectiveness of deep learning CNNs for estimating strong galaxy-scale lens mass model parameters when applied to upcoming wide-field survey data, investigating the practicalities they face and comparing and combining them with conventional modelling methods. I construct a CNN and train it on my own simulated lensing images with the imaging characteristics of the Euclid VIS band, LSST r-band, and LSST gri multiband. The CNN is trained to predict parameters of a smooth singular isothermal ellipsoid (SIE) projected mass profile that would best fit the observed image. Multiple aspects of this method are investigated, beginning with a comparison of its accuracy and reliability when applied to the survey data sets. The impact of multiband imaging is analysed, as well as the impact of lens light subtraction commonly performed by conventional modelling techniques. I show that for images including lens galaxy light, the CNN recovers the lens model parameters with an acceptable accuracy, with precision improved on average by 34 ± 5 per cent when lens light is subtracted. Additionally, the inclusion of multiband data improves performance regardless of lens light subtraction. While similar accuracies and precision are obtained for single epoch Euclid VIS and LSST r-band data sets, adding g- and i-band images to the latter increases precision by 24 ± 2 per cent without lens light and by 20 ± 2 per cent with lens light. I also examine the gains in performance through stacking images, and the impact of lens mass-light alignment. For the latter, when orientation and ellipticity of the lens light profile are allowed to differ from those of its mass profile, just as with real galaxies, the network performs most consistently when trained with a moderate amount of scatter between the two profiles. I next seek ways of improving the method for application to real images, starting by implementing an existing technique to create a new Bayesian CNN that can predict both model parameters and their uncertainties. New data sets are then simulated for training and testing the network, allowing the network to either predict parameters of the SIE profile or predict those of the more general elliptical power law profile. To examine how the CNN performs at fitting these profiles to more complex mass models, these data sets feature a range of increasingly realistic lensing systems, from smooth parametric mass and light profiles to featuring real background sources, complex hydrodynamically-generated foreground mass distributions and line-of-sight structure. In order to assess the suitability of the Bayesian CNN as a whole, I compare its performance when tested on these data sets to that of a conventional fitting method: the semilinear inversion technique PyAutoLens. In addition, I present a method for combining the network with such inversion methods where the CNN provides initial priors on the latter's parameters. Across the test sets, I find that the CNN achieves errors 19 ± 22 per cent lower than when applying PyAutoLens blindly. Compared to PyAutoLens alone, the initial centring of its priors on CNN-predicted parameters instead achieves 27 ± 11 per cent lower errors. If the prior widths are additionally initialised according to CNN-predicted Bayesian uncertainties, errors are reduced further to 37 ± 11 per cent as the uncertainties help it to avoid local minima in parameter space, with errors also 17 ± 21 per cent lower than the CNN by itself. While the CNN is undoubtedly the fastest modelling method, the combination of the two increases the speed of conventional fitting alone by a factor of 1.73 and 1.19 with and without CNN-predicted uncertainties, respectively. This, combined with greatly improved precision, highlights the benefits obtainable through combining neural networks with conventional techniques in order to achieve an efficient automated modelling approach. I finish off this thesis by discussing the scientific applications requiring such an approach, and examining how tightly these methods can constrain cosmological parameters through the modelling of double source plane lenses to aid in our understanding of the Universe.

Although general relativity passes all precision tests to date, there are several reasons to go beyond the current model of gravitation and search for new fundamental physics. This means looking for new, so far undetected, fields. Scalars are the easiest fields to consider and they are ubiquitous in both extensions of the Standard Model and in alternative theories of gravity. That is why a lot of attention has been drawn towards the investigation of scalar-tensor theories, where gravity is described by both the metric tensor and a scalar field. A particularly interesting phenomenon that has recently gained increasing interest is spontaneous scalarization. In gravity theories that exhibit this mechanism, astrophysical objects are identical to their general relativistic counterpart until they reach a specific threshold, usually either in compactness, curvature or, as recently shown, in spin. Beyond this threshold, they acquire a nontrivial scalar configuration, which also affects their structure. In this thesis, we focus on the study of this mechanism in generalized scalar-tensor theories. We identify a minimal action that contains all of the terms that can potentially trigger spontaneous scalarization. We first focus on the onset of scalarization in this specific theory and determine the relevant thresholds in terms of the contributing coupling constants and the properties of the compact object. Finally, we study the effect of this model on the properties of both scalarized black holes and neutron stars, such as affecting their domain of existence or the amount of scalar charge they carry.

Analog models in physics utilize a conceptual metaphore and a mathematical similarity to describe one system in terms of another. In this thesis, we will present the work that has been done regarding a time-dependent analog examining classical fluid interface waves in order to test predictions from cosmology. We will first detail theoretical work regarding interface waves in a strong-gradient magnetic field, conceived to simulate cosmological inflation, and constituting the first proposal for analog cosmology using interface waves. Motivated by this proposal we will then shift our focus towards parametric resonance, a process by which interface waves are exponentially amplified when coupled to an oscillating gravitational field. We outline an experiment designed to study the effective field theory of the interface, subject to parametric amplification. In this, we demonstrate that it is possible to conduct hundreds of nearly identical experiments while carefully controlling and monitoring the mechanical, optical, chemical, and environmental conditions with previously unachievable levels of precision. Our measurements of the exponential growth and damping rates for the interface waves are believed to be the most precise ever reported. The precise repetitions in the experiment further allow us to comment on the distribution of initial state at sub-micrometer amplitudes, introduce a classical two-mode squeezing model to characterize the linear statistical evolution of the model, and we present preliminary results characterizing the degree of nonlinearity in the system. Our results show that it is possible to control and interact with two-fluid systems to the accuracy needed to mimic and investigate in depth cosmological processes in a controlled laboratory environment. At the heart of this thesis is the desire to gain a deeper understanding of effective or emergent field theories. Our vision is to establish a fluid interface metrology approach to drive theoretical developments in both, the effective field theories in fluids and cosmology alike.