I have developed a new general-purpose deterministic 2D radiative transfer code for astrophysical dusty environments named LELUYA (www.leluya.org). It can provide the solution to an arbitrary axially symmetric multi-grain dust distribution around an arbitrary heating source. By employing a new numerical method, the implemented algorithm automatically traces the dust density and optical depth gradients, creating the optimal unstructured triangular grid. The radiative transfer equation includes dust scattering, absorption and emission. Unique to LELUYA is also its ability to self-consistently reshape the sublimation/condensation dust cavity around the source to accommodate for the anisotropic diffuse radiation. LELUYAs capabilities are demonstrated in the study of the asymptotic giant branch (AGB) star IRC+10011. The stellar winds emanating from AGB stars are mostly spherically symmetric, but they evolve into largely asymmetric planetary nebulae during later evolutionary phases. The initiation of this symmetry breaking process is still unexplained. IRC+10011 represents a rare example of a clearly visible asymmetry in high-resolution near-infrared images of the circumstellar dusty AGB wind. LELUYA shows that this asymmetry is produced by two bipolar cones with 1/r0.5 density profile, imbedded in the standard 1/r2 dusty wind profile. The cones are still breaking though the 1/r2 wind, suggesting they are driven by bipolar jets. They are about 200 years old, thus a very recent episode in the final phase of AGB evolution before turning into a proto-planetary nebula, where the jets finally break out from the confining spherical wind. IRC+10011 provides the earliest example of this symmetry breaking thus far.

In the framework of active galactic nuclei (AGNs), a galaxy’s supermassive black hole is surrounded by a dusty torus whose clumpy configuration allows for either direct or obscured views toward the central engine. Viewing AGNs from different angles gives rise to a variety of AGN classifications; for example, the generic Type 1 AGN class requires the detection of optically broad emission lines, which arise from quickly moving material within the torus, whereas Type 2 AGNs lack these observations. While these viewing angles are not directly observable, synthetic torus models generated with CLUMPY provide a means to determine them along with other parameters that describe the nature and characteristics of the torus in general. Employing CLUMPY models with mid-infrared spectroscopic observations of a large sample of both Type 1 and Type 2 AGNs allows us to acquire a further understanding of the clumpy torus structure and its viewing angles.

Radiative transfer plays an important role in turbulent combustion and should be incorporatedin numerical simulations. However, as combustion and radiation are characterized bydifferent time scales and different spatial and chemical treatments, and the complexity of theturbulent combustion flow, radiation effect is often neglected or roughly modelled. Couplinga large eddy simulation combustion solver and a radiation solver through a dedicated languageCORBA is investigated. Four formulations of Monte Carlo method (Forward Method,Emission Reciprocity Method, Absorption Reciprocity Method and Optimized ReciprocityMethod) employed to resolve RTE have been compared in a one-dimensional flame testcase using three-dimensional calculation grids with absorbing and emitting medium in orderto validate the Monte Carlo radiative solver and to choose the most efficient model forcoupling. In order to improve the performance of Monte Carlo solver, two techniques havebeen developed. After that, a new code dedicated to adapt the coupling work has beenproposed. Then results obtained using two different RTE solvers (Reciprocity Monte Carlomethod and Discrete Ordinate Method) applied to a three-dimensional turbulent reactingflow stabilized downstream of a triangular flame holder with a correlated-k distributionmodel describing the real gas medium spectral radiative properties are compared not onlyin terms of physical behavior of the flame but also in computational performance (storagerequirement, CPU time and parallelization efficiency). Finally, the impact of boundary conditionstaking into account the actual wall emissivity and temperature has been discussed.

To date more than 1500 exoplanets have been discovered. A large number of these are hot Jupiters, Jupiter-sized planets orbiting < 0.1 au from their parent stars, due to limitations in observational techniques making them easier to detect than smaller planets in wider orbits. This is also, for the same reasons, the class of exoplanets with the most observational constraints. Due to the very large interaction between these planets and their parent stars they are believed to be tidally locked, causing a large temperature contrast between the permanently hot day side and colder night side. There are still many open questions about these planets. Many are observed to have inflated radii, i.e. the observed radius is larger for a given mass than evolutionary models predict. A mechanism that can transport some of the stellar heating into the interior of the planet may be able to explain this. The presence of hazes or clouds has been inferred on some planets, but their composition and distribution remain unknown. According to chemical equilibrium models TiO and VO should be present on the day side of the hottest of these planets, but these molecules have not yet been detected. Cold traps, where these molecules condense out on the night side, have been suggested to explain this. The efficiency of the heat redistribution from the day side to the night side has been found to vary significantly between different planets; the mechanism behind this is still unknown. To begin to answer many of these questions we need models capturing the three-dimensional nature of the atmospheres of these planets. General circulation models (GCMs) do this by solving the equations of fluid dynamics for the atmosphere coupled to a radiative transfer scheme. GCMs have previously been applied to several exoplanets, but many solve simplified fluid equations (shallow water or primitive equations) or highly parametrised radiation schemes (temperature-forcing, gray or band-averaged opacities). We here present an adaptation of the Met Office Unified Model (UM), a GCM used for weather predictions and climate studies for the Earth, to hot Jupiters. The UM solves the full 3D Euler equations for the fluid, and the radiation scheme uses the two-stream approximation and correlated-k method, which are state of the art for both Earth and exoplanet GCMs. This makes it ideally suited for the study of hot Jupiters. An important part of this work is devoted to the adaptation of the radiation scheme of the UM to hot Jupiters. This includes calculation of opacities for the main absorbers in these atmospheres from state-of-the-art high temperature line lists, the calculation of k-coefficients from these opacities, and making sure all aspects of the scheme perform satisfactorily at high temperatures and pressures. We have tested approximations made in previous works such as the two-stream approximation, use of band-averaged opacities and different treatments of gaseous overlap. Uncertainties in current models, such as the lack of high temperature line broadening parameters for these atmospheres, are discussed. We couple the adapted radiation scheme to the UM dynamical core, which has been tested independently. Our first application is devoted to one of the most well-observed hot Jupiters, HD 209458b. Differences between previous modelling works and our model are discussed, and we compare results from the full coupled model with results obtained using a temperature-forcing scheme. We have also developed a tool to calculate synthetic phase curves, and emission and transmission spectra from the output of our 3D model. This enables us to directly compare our model results to observations and test the effect of various parameters and model choices on observable quantities.

Over a large range of equilibrium temperatures, clouds shape the transmission spectrum of hot Jupiter atmospheres, yet their composition remains unknown. Recent observations show that the Kepler light. curves of some hot Jupiters are asymmetric: for the hottest planets, the light. curve peaks before secondary eclipse, whereas for planets cooler than similar to 1900 K, it peaks after secondary eclipse. We use the thermal structure from 3D global circulation models to determine the expected cloud distribution and Kepler light. curves of hot Jupiters. We demonstrate that the change from an optical light. curve dominated by thermal emission to one dominated by scattering (reflection) naturally explains the observed trend from negative to positive offset. For the cool planets the presence of an asymmetry in the Kepler light curve is a telltale sign of the cloud composition, because each cloud species can produce an offset only over a narrow range of effective temperatures. By comparing our models and the observations, we show that the cloud composition of hot Jupiters likely varies with equilibrium temperature. We suggest that a transition occurs between silicate and manganese sulfide clouds at a temperature near 1600 K, analogous to the L/T transition on brown dwarfs. The cold trapping of cloud species below the photosphere naturally produces such a transition and predicts similar transitions for other condensates, including TiO. We predict that most hot Jupiters should have cloudy nightsides, that partial cloudiness should be common at the limb, and that the dayside hot spot should often be cloud-free.

The chemical fingerprints of old, metal-poor stars can be used to unravel the events of the newborn Universe and help us understand the properties of the first stars and star clusters. The study of nearby stars to infer properties in the distant past is often referred to as Galactic archaeology. However, the chemical composition of stars cannot be observed directly, but must be inferred by means of spectroscopic modelling. Traditionally, this modelling utilises one-dimensional (1D) stellar atmospheres in hydrostatic and local thermodynamic equilibrium (LTE). Today, we know that departures from LTE (known as NLTE), and differences between 1D model atmospheres and their hydrodynamical three-dimensional (3D) counterparts, become increasingly severe at lower metallicity. The development of NLTE modelling of spectral line formation in 3D atmospheres is still in its infancy, but constitutes a remarkable step forward that has been made possible by parallelised codes and supercomputers. The central theme of this thesis is the application of NLTE analyses to metal-poor stars, to help usher the field of Galactic archaeology forward with important consequences for the nature of the first stellar generations. I present a theoretical NLTE study of aluminium, where I validate the analysis using a set of bright standard stars and provide calculated NLTE effects for a large parameter space. I perform 3D NLTE calculations for the solar spectrum to better constrain the zero-point of the cosmic abundance scale, and find excellent agreement with the meteoritic aluminium abundance. I also present NLTE analyses of metal-poor stars in the globular clusters NGC 6397 and M4. While globular cluster stars were long expected to form from a chemically homogeneous medium, star-to-star abundance variations of light elements indicate multiple epochs of star formation. Massive first-generation stars polluted the interstellar medium from which later generations formed, and I use the observed abundance variations to deduce the properties of the polluting stars. Among the heavier elements, I uncover evolutionary abundance variations that match predictions of stellar evolution models with atomic diffusion. The results indicate that the chemical abundance ratios of unevolved metal-poor stars are affected by gravitational settling, with a bias of the order 25-50 %, increasing towards lower metallicity. This atmospheric depletion mechanism is a probable explanation to why the stellar abundances of lithium fall short of the predictions from standard Big Bang nucleosynthesis. Finally, I apply a 3D NLTE abundance analysis to the red giant SMSS 0313-6708, which is the most iron-deficient star known. The chemical abundance pattern of this star indicates that it formed from gas affected only by Big Bang nucleosynthesis and a single faint supernova. Comparison of the inferred abundance pattern to theoretical predictions leads to constraints on the explosion mechanism and the mass of the metal-free progenitor star.

We present an outline of basic assumptions and governing structural equations describing atmospheres of sub-stellar mass objects, in particular the extrasolar giant planets and brown dwarfs. Although most of the presentation of the physical and numerical background is generic, details of the implementation pertain mostly to the code COOLTLUSTY. We also present a review of numerical approaches and computer codes devised to solve the structural equations, and make a critical evaluation of their efficiency and accuracy.

Thermal radiation is ubiquitous to all matter at finite temperature and controlling the radiative nature of that matter has been a key enabling factor in the development of several recent technologies, such as thermal diodes, thermal antennae, thermophotovoltaics, heat-assisted magnetic recording, and contactless cooling in microelectromechanical systems. At the micro/nano-scale, thermal radiation does not reliably behave in the way Planck's blackbody law predicts, due to near-field effects such as the diffraction, interference, and tunneling of light. In fact, the so-called blackbody limit can routinely be broken by several orders of magnitude when objects of dimensions or separation distances much smaller than the peak thermal wavelength (approximately 10 \si{\micro\meter} at room temperature) exchange thermal radiation. A deeper theory is required to understand near-field thermal radiation: Maxwell's equations. Maxwell's equations allow for a direct connection between the thermally induced current fluctuations and radiative transfer. In this dissertation, I investigate radiative transfer among spherical bodies aligned in a linear chain. The chain may be composed of any number of spheres, and the spheres themselves may be composed of any linear isotropic material, may be of any size and separation distance, and may each have any number of spherically symmetric layers. Using a dyadic Green's function formalism, I derive numerically exact formulas for heat transfer between pairs of spheres in the chain and between any sphere in the chain and its environment. My work clearly demonstrates that adding coatings to spherical objects can drastically impact the spectrum of radiative transfer, enhancing or diminishing it in various cases. This degree of tailoring makes coated spheres a flexible, yet unexplored, platform for future experiments in near-field radiative heat transfer. My work also demonstrates that, in an experiment measuring the distance dependent heat transfer between two spheres, heat transfer from the spheres to their environment can also have a strong distance dependence, which must be considered carefully when designing an experiment and analyzing its results. This demonstrates a cautious but optimistic outlook for the near-field radiative heat transfer community moving beyond traditional plane-plane and sphere-plane experimental configurations.

Many uncertainties remain as to how the most massive stars are formed and how they interact with their environment via radiative and mechanical processes. This feedback may affect future generations of star formation -- triggering it by compressing gas, or hindering it by dispersing reservoirs. These scenarios can be simulated by solving the equations of hydrodynamics and radiative transfer. However, the latter is usually simplified due to its computational expense, despite its importance in determining the dynamics. In this thesis, I describe how I increased the efficiency of the radiation hydrodynamics code, TORUS, which uses a Monte Carlo approach to solving the radiative transfer. Tens of millions of energy packets are propagated through a domain split over hundreds of processors running in parallel with Message Passing Interface (MPI). By re-examining and improving communication algorithms, I lowered the radiation run time by about a factor of ten, making it tractable to run three-dimensional simulations of massive star feedback in clusters. This includes both the stellar and diffuse radiation fields, with multiple atomic species and silicate dust grains. The full ionization states and temperatures can then be fed in to produce self-consistent synthetic observations. I applied this to clouds of 1000 and 10,000 solar masses with surface density 0.01 g/cm^2, containing a 34 solar mass star, with photoionization and radiation pressure feedback. Photoionization is efficient at shaping and dispersing clouds. The expanding ionization front forms dense, spherical knots with pillars pointing away from the emitting star. These resemble the Pillars of Creation in the Eagle Nebula, and the proplyds observed in the Orion Nebula. In the lower-mass model, almost all material is removed from the (15.5 pc)^3 grid within 1.6 Myr; the higher mass cloud is somewhat more resistant, with 25 per cent remaining inside (32.3 pc)^3 after 4.3 Myr. Radiation pressure has a negligible effect, but will be more important for denser clouds or higher luminosities.

In this thesis, I apply techniques from the field of computer graphics to ray tracing in astrophysical simulations, and introduce the grace software library. This is combined with an extant radiative transfer solver to produce a new package, taranis. It allows for fully-parallel particle updates via per-particle accumulation of rates, followed by a forward Euler integration step, and is manifestly photon-conserving. To my knowledge, taranis is the first ray-traced radiative transfer code to run on graphics processing units and target cosmological-scale smooth particle hydrodynamics (SPH) datasets. A significant optimization effort is undertaken in developing grace. Contrary to typical results in computer graphics, it is found that the bounding volume hierarchies (BVHs) used to accelerate the ray tracing procedure need not be of high quality; as a result, extremely fast BVH construction times are possible (< 0.02 microseconds per particle in an SPH dataset). I show that this exceeds the performance researchers might expect from CPU codes by at least an order of magnitude, and compares favourably to a state-of-the-art ray tracing solution. Similar results are found for the ray-tracing itself, where again techniques from computer graphics are examined for effectiveness with SPH datasets, and new optimizations proposed. For high per-source ray counts (≳ 104), grace can reduce ray tracing run times by up to two orders of magnitude compared to extant CPU solutions developed within the astrophysics community, and by a factor of a few compared to a state-of-the-art solution. taranis is shown to produce expected results in a suite of de facto cosmological radiative transfer tests cases. For some cases, it currently out-performs a serial, CPU-based alternative by a factor of a few. Unfortunately, for the most realistic test its performance is extremely poor, making the current taranis code unsuitable for cosmological radiative transfer. The primary reason for this failing is found to be a small minority of particles which always dominate the timestep criteria. Several plausible routes to mitigate this problem, while retaining parallelism, are put forward.