L'illumination par la face arrière amincie des imageurs CMOS est une des voies étudiées pour accroître le rapport signal à bruit et ainsi la sensibilité de ce capteur. Or cette configuration est adaptée à la détection des électrons dans la gamme d'énergie [[1 ; 12 keV]. L'électron incident crée, par multiplication, plusieurs centaines d'électrons secondaires, proche de la surface. Une couche de passivation par surdopage P++ de la face arrière est nécessaire afin de réduire le nombre de recombinaisons de surface des électrons. Par effet de champ électrique, la couche de passivation augmente le nombre de charges collectées, et ainsi le gain de collection du capteur. L'objectif de cette thèse est de développer des moyens de caractérisation pour déterminer in situ les performances sur le gain de collection de six procédés de passivation. Préalablement, le profil de dépôt d'énergie de l'électron incident est étudié au moyen d'une simulation Monte-Carlo puis d'un modèle analytique. Un modèle associé du gain de collection indique qu'à forte énergie, l'effet miroir de la passivation est déterminant tandis qu'à faible énergie, l'épaisseur de la passivation est un facteur clef. Une première expérience d'irradiation de diodes étendues P++=N permet de dégager l'influence du procédé de passivation sur les recombinaisons de surface. Grâce à une seconde caractérisation de type < événement unique >, directement sur capteur CMOS aminci, les passivations sont discriminées quant à leur effet miroir et l'étalement de la charge qu'elles induisent. Le recuit laser d'activation des dopants peut s'avérer une source d'inhomogénéités du gain sur la surface de la matrice

The modeling of Micro- and NanoElectroMechanical Systems (MEMS and NEMS) requires new computational techniques that can deal efficiently with geometric complexity and scale dependent effects that may arise. Reduced feature sizes increase the coupling of physical phenomena and noncontinuum behavior, often requiring models based on molecular descriptions and/or first principles. Furthermore, noncontinuum effects are often localized to small regions of (relatively) large systemsprecluding the global application of microscale models due to computational expense. Multiscale modeling couples efficient continuum solvers with detailed microscale models to providing accurate and efficient models of complete systems. This thesis presents the development of multiscale modeling techniques for nonequilibrium microscale gas phase phenomena, especially thermally driven microflows. Much of this focuses on improving the ability of the Information Preserving DSMC (IP-DSMC) to model thermally driven flows. The IP-DSMC is a recent technique that seeks to accelerate the solution of direct simulation Monte Carlo (DSMC) simulations by preserving and transporting certain macroscopic quantities within each simulation molecules. The primary contribution of this work is the development of the Octant Splitting IP-DSMC (OSIP-DSMC) which recovers previously unavailable information from the preserved quantities and the microscopic velocities. The OSIP-DSMC can efficiently simulate flow fields induced by nonequilibrium systems, including phenomena such as thermal transpiration. The OSIP-DSMC provides an efficient method to explore rarefied gas transport phenomena which may lead to a greater understanding of these phenomena and new concepts for how these may be utilized in practical engineering systems. Multiscale modeling is demonstrated utilizing the OSIP-DSMC and a 2D BEM solver for the continuum (heat transfer) model coupled with a modified Alternating Schwarz coupling scheme. An interesting application for this modeling technique is Thermal Sensing Atomic Force Microscopy (TSAFM). TSAFM relies on gas phase heat transfer between heated cantilever probes and the scanned surface to determine the scan height, and thus the surface topography. Accurate models of the heat transfer phenomena are required to correctly interpret scan data. This thesis presents results demonstrating the effect of subcontinuum heat transfer on TSAFM operation and explores the mechanical effects of the Knudsen Force on the heated cantilevers.

Thesis (Ph. D.)--University of California, Riverside, 2009. / Includes abstract. Available via ProQuest Digital Dissertations. Title from first page of PDF file (viewed March 12, 2010). Includes bibliographical references (p. 100-106). Also issued in print.

When a lander approaches a dusty surface, the plume from the descent engine impinges on the ground and entrains loose regolith into a high velocity spray. This problem exhibits a wide variety of complex phenomena such as highly under-expanded plume impingement, transition from continuum to free molecular flow, erosion, coupled gas-dust motions, and granular collisions for a polydisperse distribution of aerosolized particles. The focus of this work is to identify and model the important physical phenomena and to characterize the dust motion that would result during typical lunar landings. A hybrid continuum-kinetic solver is used, but most of the complex physics are simulated using the direct simulation Monte Carlo method. A descent engine of comparable size and thrust to the Lunar Module Descent Engine is simulated because it allows for direct comparison to Apollo observations. Steady axisymmetric impingement was first studied for different thrust engines and different hovering altitudes. The erosion profiles are obtained from empirically derived scaling relationships and calibrated to closely match the net erosion observed during the Apollo missions. Once entrained, the dust motion is strongly influenced by particle-particle collisions and the collision elasticity. The effects of two-way coupling between the dust and gas motions are also studied. Small particles less than 1 µm in diameter are accelerated to speeds that exceed 1000 m/s. The larger particles have more inertia and are accelerated to slower speeds, approximately 350 m/s for 11 µm grains, but all particle sizes tend obtain their maximum speed within approximately 40 m from the lander. The maximum particle speeds and erosion rates tend to increase as the lander approaches the lunar surface. The erosion rates scale linearly with engine thrust and the maximum particle speed increases for higher thrust engines. Dust particles are able to travel very far from the lander because there is no background atmosphere on the moon to inhibit their motion. The far field deposition is obtained by using a staged calculation, where the first stages are in the near field where the flow is quasi-steady and the outer stages are unsteady. A realistic landing trajectory is approximated by a set of discrete hovering altitudes which range from 20 m to 3 m. Larger particles are accelerated to slower speeds and are deposited closer to the lander than smaller particles. Many of the gas molecules exceed lunar escape speed, but some gas molecules become trapped within the dust cloud and remain on the moon. The high velocity particulate sprays can be damaging to nearby structures, such as a lunar outpost. One way of mitigating this damage is to use a berm or fence to shield nearby structures from the dust spray. This work attempts to predict the effectiveness of such a fence. The effects of fence height, placement, and angle as well as the model sensitivity to the fence restitution coefficient are discussed. The expected forces exerted on fences placed at various locations are computed. The pressure forces were found to be relatively small at fences placed at practical distances from the landing site. The trajectories of particles that narrowly avoid the fence were not significantly altered by the fence, suggesting that the dust motion is weakly coupled to the gas in the near vicinity of the fence. Future landers may use multi-engine configurations that can form 3-dimensional gas and dust flows. There are multiple plume-plume and plume-surface interactions that affect the erosion rates and directionality of the dust sprays. A 4-engine configuration is simulated in this work for different hovering altitudes. The focusing of dust along certain trajectories depends on the lander hovering altitude, where at lower altitudes the dust particles focus along symmetry planes while at higher altitudes the sprays are more uniform. The surface erosion and trenching behavior for a 4-engine lander are also discussed. / text

L'hadronthérapie est une technique de traitement de cancer basée sur l'utilisation de particules lourdes chargées. Les caractéristiques physiques de ces particules permettent un ciblage plus précis des tumeurs et une efficacité biologique supérieure à celle des photons et des électrons. Ce travail de thèse traite la problématique de la modélisation des effets biologiques induits par ce type de particules. Une partie de ce travail est consacrée à l'analyse de l'outil de simulation Monte‐Carlo " Geant4 ", utilisé pour simuler la phase physique de l'interaction des particules avec le milieu biologique. Nous avons évalué la capacité de " Geant4 " à simuler la distribution microscopique des dépôts d'énergie des particules chargées et confronté ces résultats à ceux d'un autre code de simulation dédié aux applications de radiobiologie. L'autre partie du travail est dédiée à l'étude de deux modèles radiobiologiques basés sur deux approches différentes qui sont le modèle LEM (Local Effect Model) basé sur une approche de trace amorphe et le modèle MKM (Microdosimetric Kinetic Model) basé sur une approche microdosimétrique. Une analyse théorique des deux modèles est effectuée ainsi qu'une comparaison de leurs concepts. Ensuite, nous nous sommes focalisé sur le modèle microdosimétrique " MKM " que nous avons analysé de manière plus approfondie. Enfin, nous avons appliqué le modèle MKM pour reproduire les résultats expérimentaux d'irradiation cellulaire obtenus au GANIL avec des ions carbone et argon sur des cellules tumorales (lignées SCC61 et SQ20B) de radiosensibilité différente.

This thesis deals with the direct simulation Monte Carlo (DSMC) method of analysing gas flows. The DSMC method was initially proposed as a method for predicting rarefied flows where the Navier-Stokes equations are inaccurate. It has now been extended to near continuum flows. The method models gas flows using simulation molecules which represent a large number of real molecules in a probabilistic simulation to solve the Boltzmann equation. Molecules are moved through a simulation of physical space in a realistic manner that is directly coupled to physical time such that unsteady flow characteristics are modelled. Intermolecular collisions and moleculesurface collisions are calculated using probabilistic, phenomenological models. The fundamental assumption of the DSMC method is that the molecular movement and collision phases can be decoupled over time periods that are smaller than the mean collision time. Two obstacles to the wide spread use of the DSMC method as an engineering tool are in the areas of simulation configuration, which is the configuration of the simulation parameters to provide a valid solution, and the time required to obtain a solution. For complex problems, the simulation will need to be run multiple times, with the simulation configuration being modified between runs to provide an accurate solution for the previous run�s results, until the solution converges. This task is time consuming and requires the user to have a good understanding of the DSMC method. Furthermore, the computational resources required by a DSMC simulation increase rapidly as the simulation approaches the continuum regime. Similarly, the computational requirements of three-dimensional problems are generally two orders of magnitude more than two-dimensional problems. These large computational requirements significantly limit the range of problems that can be practically solved on an engineering workstation or desktop computer. The first major contribution of this thesis is in the development of a DSMC implementation that automatically adapts the simulation. Rather than modifying the simulation configuration between solution runs, this thesis presents the formulation of algorithms that allow the simulation configuration to be automatically adapted during a single run. These adaption algorithms adjust the three main parameters that effect the accuracy of a DSMC simulation, namely the solution grid, the time step and the simulation molecule number density. The second major contribution extends the parallelisation of the DSMC method. The implementation developed in this thesis combines the capability to use a cluster of computers to increase the maximum size of problem that can be solved while simultaneously allowing excess computational resources to decrease the total solution time. Results are presented to verify the accuracy of the underlying DSMC implementation, the utility of the solution adaption algorithms and the efficiency of the parallelisation implementation.

Les alliages fer-chrome présentent des comportements thermodynamiques et cinétiques particuliers liés à leurs propriétés magnétiques. La décomposition de l’alliage par vieillissement thermique a été étudiée dans cette thèse. Pour cela, un modèle cinétique de la décomposition de l’alliage à l’échelle atomique a été développé en modélisant de façon détaillée les propriétés thermodynamiques et de diffusion des espèces chimiques dans l’alliage. L’évolution des propriétés de diffusion des éléments avec la transition ferro-paramagnétique de l’alliage a en particulier été modélisée. Les décompositions simulées par ce modèle ont été comparées à des décompositions expérimentales pour une large gamme de concentrations et de températures. Cette comparaison a montré un bon accord entre les cinétiques simulées et les expériences et a permis de mettre en évidence le rôle central de la transition ferro-paramagnétique dans la cinétique de décomposition des alliages fortement concentrés en chrome. Cette étude a également montré que la diffusion des éléments à l’interface des phases est responsable de la cinétique de décomposition de l’alliage aux temps longs. Une étude de la ségrégation induite par l’irradiation au voisinage des puits de défauts a également été commencée. Pour cela, un modèle de l’évolution de l’alliage sous irradiation contenant un puits de défauts a été développé. Il a été montré, en accord avec la littérature, que pour les cas faiblement concentrés en chrome, l’alliage a tendance à s’enrichir en chrome au voisinage des puits de défauts à basse température et à s’appauvrir en chrome à haute température. / Iron-Chromium alloys have a peculiar thermodynamic and diffusion behavior which is due to their magnetic properties. The alloy decomposition under thermal ageing has been studied in this thesis. An atomistic kinetic model has been performed in this aim in which we have modeled in details the chemical species thermodynamic and diffusion properties. In particular, the evolution of elements diffusion properties with the ferro-paramagnetic transition has been introduced in the model. Simulated decompositions have been compared with experiments for a large range of concentrations and temperatures. A good agreement between simulations and experiments was observed and these comparisons have highlighted the ferro to paramagnetic transition key role in the concentrated alloys kinetic decomposition. This study has also evidenced that the elements diffusion at phases interfaces is responsible for the alloy decomposition kinetic in long lasting. We have also started a study on the alloy radiation induced segregation. For that purpose, atomistic kinetic model has been performed modeling defects migration through a perfect planar sink. It have been shown, I agreement with former studies, that chromium tends to segregate in the vicinity of sinks a low temperatures and deplete at high temperature.

The Direct Simulation Monte Carlo algorithm as coded in FENIX is used to model the transport of trace ions in the first vacuum stage of the inductively coupled plasma mass spectrometer. Haibin Ma of the Farnsworth group at Brigham Young University measured two radial trace density profiles: one 0.7 mm upstream of the sampling cone and the other 10 mm downstream. We compare simulation results from FENIX with the experimental results. We find that gas dynamic convection and diffusion are unable to account for the experimentally-measured profile changes from upstream to downstream. Including discharge quenching and ambipolar electric fields, however, makes it possible to account for the way the profiles change.

Les simulations de Monte-Carlo, bien que consommatrices en temps de calcul, restent un outil puissant qui permet d'évaluer les méthodes de correction des effets physiques en imagerie médicale. Nous avons optimisé et validé une méthode de reconstruction baptisée F3DMC (Fully3D Monte CARLO) dans laquelle les effets physiques perturbant le processus de formation de l'image en tomographie d'émission monophotonique sont modélisés par des méthodes de Monte-Carlo et intégrés dans la matrice-système. Le logiciel de simulation de Monte-Carlo utilidé est GATE. Nous avons validé GATE en SPECT en modélisant la gamma-caméra (Philips AXIS) utilisé en routine clinique. Des techniques de seuillage, filtrage par analyse en composantes principales et de reconstruction ciblée (régions fonctionnelles, région hybrides) ont été testées pour améliorer la précision de la matrice-système et réduire le nombre de photons ainsi que le temps de calcul nécessaires. Les infrastructures de la grille EGEE ont été utilisées pour déployer les simulations GATE afin de réduire leur temps de calcul. Les résultats obtenus avec F3DMC sont comparés avec les méthodes de reconstruction (FBP,ML-EM,ML-EMC) pour un fantôme simulé et avec la méthode OSEM-C pour un fantôme réel. Les résultats de cette étude montrent que la méthode F3DMC ainsi que ses variantes permettent d'améliorer la restauration des rapports d'activité et le rapport signal sur bruit. L'utilisation de la grille de calcul EGEE a permis d'obtenir un gain de l'ordre de 300 en temps de calcul. Dans la suite, ces résultats doivent être confirmés par des études sur des fantômes complexes et des patients et ouvrent la voie vers une méthode de reconstruction unifiée, pouvant être appliquée aussi bien en SPECT qu'en PET.

The processes by which anisotropic colloidal and nanoscale particles may come together to form ordered monodisperse structures are not well understood. The canonical example of such a system is provided by the assembly of virus capsids, in which tens to thousands of particles of one or a few types assemble efficiently into ordered, highly symmetrical shells. Other examples include a wide variety of protein oligomers, and there is interest in producing analogous systems of synthetic particles. In this thesis I study the self-assembly of simple model particles, consisting of spheres decorated with attractive patches. I consider in detail the assembly of clusters of particles corresponding to the Platonic solids. For the majority of these structures assembly is found to be efficient over a wide range of parameter space. The optimal conditions represent a compromise between thermodynamic stability and kinetic accessibility. We consider two versions of the model, with and without constraints on the torsion angle of bound particles. In both cases the structures with triangular faces are found to assemble most easily. In the absence of torsional constraints dodecahedra will not assemble under any set of parameters as a result of the preferential formation of disordered aggregates. With torsional constraints included all of the Platonic solids assemble successfully and the behaviour of the model is considerably changed. In particular disordered aggregates become far less favourable. I explore possible methods of assembling larger structures, either via “hierarchical” assembly where small clusters are first assembled and then used as building blocks in another stage of assembly, or by a templating method in which an inner cluster acts as a template for a larger outer cluster. These approaches are studied using the “Virtual Move Monte Carlo” cluster move algorithm, the behaviour of which we examine in some detail.