Diese Arbeit ist ein Beitrag zur theoretischen Chemie sowie zur Oberflächenchemie. Durch Kombination von computergestützten und experimentellen Untersuchungen wird die atomare Struktur von dünnen SiO2-Filmen auf Ru(0001)-Unterlagen, von eisendotierten SiO2-Filmen auf diesen Unterlagen und von H2O-Filmen auf MgO(001)-Oberflächen bestimmt. Die atomaren Strukturmodelle wurden entweder mit dem neu entworfenen und im Paket DoDo implementierten genetischen Algorithmus oder mittels auf Sachkenntnis gestützter Vermutungen erhalten. Die simulierten Eigenschaften der so erhaltenen Strukturen stimmen sehr gut mit den experimentellen Daten (Raster-Tunnel-Mikroskopie, Infrarot-Spektroskopie) überein. Die erfolgreiche Strukturbestimmung mithilfe des DoDo-Programms zeigt, dass genetische Algorithmen zur systematischen und extensiven Erkundung der Energielandschaften 2D-periodischer Systeme geeignet sind. / This work is a contribution in the field of theoretical chemistry and surface science. The joint computational and experimental studies investigated the atomic structure of ultrathin silica and iron-doped silica films formed on the Ru(0001) surface and water films formed on the MgO(001) surface. The atomic structure models were obtained using either the educated guess approach or the genetic algorithm that was designed and implemented within the DoDo package. The properties simulated for the resulting models are in a very good agreement with the experimental data (scanning tunnelling microscopy, infrared spectroscopy). The successful structure determination using the DoDo program shows that the genetic algorithm technique is capable of systematic and extensive exploration of the energy landscapes for 2D-periodic systems.

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / Um código computacional para escoamentos bifásicos incorporando metodologia híbrida entre oMétodo dos Elementos Finitos e a descrição Lagrangeana-Euleriana Arbitrária do movimento é usado para simular a dinâmica de um jato transversal de gotas na zona primária de quebra. Os corpos dispersos são descritos por meio de um método do tipo front-tracking que produz interfaces de espessura zero através de malhas formadas pela união de elementos adjacentes em ambas as fases e de técnicas de refinamento adaptativo. Condições de contorno periódicas são implementadas de modo variacionalmente consistente para todos os campos envolvidos nas simulações apresentadas e uma versão modificada do campo de pressão é adicionada à formulação do tipo um-fluido usada na equação da quantidade de movimento linear. Simulações numéricas diretas em três dimensões são executadas para diferentes configurações de líquidos imiscí veis compatíveis com resultados experimentais encontrados na literatura. Análises da hidrodinâmica do jato transversal de gotas nessas configurações considerando trajetórias, variação de formato de gota, espectro de pequenas perturbações, além de aspectos complementares relativos à qualidade de malha são apresentados e discutidos. / A two-phase flow computational code taking a hybrid Arbitrary Lagrangian-Eulerian desciption of movement along with the Finite Element Method is used to simulate the dynamics of an incompressible drop jet in crossflow in the primary breakup zone. Dispersed entities are described by means of a front-tracking method which produces zero-thickness interfaces through contiguous element meshing and adaptive refinement techniques. Periodic boundary conditions are implemented in a variationally consistent way for all the scalar fields involved in the presented simulations and amodified version of the pressure field is added to the one-fluid formulation employed in the momentum equation. Three-dimensional direct numerical simulations for different flow configurations of immiscible liquids pertinent to experimental results found in literature. Analyses of the hydrodynamics of the drop jet in crossflow in these configurations considering trajectories, drop shape variations, spectrum of small disturbances, besides additional aspects relating to mesh quality are presented and discussed.

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / Um código computacional para escoamentos bifásicos incorporando metodologia híbrida entre oMétodo dos Elementos Finitos e a descrição Lagrangeana-Euleriana Arbitrária do movimento é usado para simular a dinâmica de um jato transversal de gotas na zona primária de quebra. Os corpos dispersos são descritos por meio de um método do tipo front-tracking que produz interfaces de espessura zero através de malhas formadas pela união de elementos adjacentes em ambas as fases e de técnicas de refinamento adaptativo. Condições de contorno periódicas são implementadas de modo variacionalmente consistente para todos os campos envolvidos nas simulações apresentadas e uma versão modificada do campo de pressão é adicionada à formulação do tipo um-fluido usada na equação da quantidade de movimento linear. Simulações numéricas diretas em três dimensões são executadas para diferentes configurações de líquidos imiscí veis compatíveis com resultados experimentais encontrados na literatura. Análises da hidrodinâmica do jato transversal de gotas nessas configurações considerando trajetórias, variação de formato de gota, espectro de pequenas perturbações, além de aspectos complementares relativos à qualidade de malha são apresentados e discutidos. / A two-phase flow computational code taking a hybrid Arbitrary Lagrangian-Eulerian desciption of movement along with the Finite Element Method is used to simulate the dynamics of an incompressible drop jet in crossflow in the primary breakup zone. Dispersed entities are described by means of a front-tracking method which produces zero-thickness interfaces through contiguous element meshing and adaptive refinement techniques. Periodic boundary conditions are implemented in a variationally consistent way for all the scalar fields involved in the presented simulations and amodified version of the pressure field is added to the one-fluid formulation employed in the momentum equation. Three-dimensional direct numerical simulations for different flow configurations of immiscible liquids pertinent to experimental results found in literature. Analyses of the hydrodynamics of the drop jet in crossflow in these configurations considering trajectories, drop shape variations, spectrum of small disturbances, besides additional aspects relating to mesh quality are presented and discussed.

Cette thèse se place dans le contexte général de la modélisation de réseaux d’antennes de grande taille, avec pour objectif d’atteindre un niveau de précision suffisamment élevé pour permettre une optimisation complète des performances et en particulier une amélioration de l’efficacité énergétique. Partant du constat que l’optimisation électromagnétique de grands réseaux représente un verrou si les couplages doivent être modélisés efficacement, cette thèse propose la mise en œuvre d’une méthode permettant la modélisation fine de grands réseaux d'antennes tout en réduisant les temps de calcul et en conservant un haut degré de précision. L'objectif est de montrer qu'une approche dérivée des formalismes périodiques infinis connus de la littérature permet d'obtenir une matrice [S] complète d'un réseau depuis l'étude d'une cellule unitaire. Après avoir présenté un état de l'art sur les réseaux d'antennes et leurs méthodes d'analyse, l'approche de modélisation proposée est détaillée. Des véhicules de test numériques et expérimentaux, permettant de valider cette méthode de modélisation, sont ensuite réalisés. L'approche est enfin utilisée avec succès pour deux types d'application. / This thesis is organized in the general context of modeling a large antenna arrays with the aim of achieving a high level of precision. This modeling allows a complete optimization of the performances and an enhancement of the energy efficiency. Given that the electromagnetic optimization of large arrays still represents a challenge if the mutual coupling is not efficiently modeled. This work offers the implementation of a method allowing the precise modeling of large antenna arrays while reducing the computation time and maintaining a high degree of precision. The aim is to demonstrate that the derived approach from the infinite periodic formalisms makes it possible to obtain a full [S] matrix of an array based on the study of a unit cell. After presenting a state of the art of antenna arrays and their methods of analysis, the proposed modeling approach is explained. Numerical and Experimental demonstrators are then made for the validation. Finally, this method is successfully used for two types of applications.

The present thesis is dedicated to the efficient computation of electrostatic interactions in particle systems, which is of great importance in the field of molecular dynamics simulations. In order to compute the therefor required physical quantities with only O(N log N) arithmetic operations, so called particle-mesh methods make use of the well-known Ewald summation approach and the fast Fourier transform (FFT). Typically, such methods are able to handle systems of point charges subject to periodic boundary conditions in all spatial directions. However, periodicity is not always desired in all three dimensions and, moreover, also interactions to dipoles play an important role in many applications. Within the scope of the present work, we consider the particle-particle NFFT method (P²NFFT), a particle-mesh approach based on the fast Fourier transform for nonequispaced data (NFFT). An extension of this method for mixed periodic as well as open boundary conditions is presented. Furthermore, the method is appropriately modified in order to treat particle systems containing both charges and dipoles. Consequently, an efficient algorithm for mixed charge-dipole systems, that additionally allows a unified handling of various types of periodic boundary conditions, is presented for the first time. Appropriate error estimates as well as parameter tuning strategies are developed and verified by numerical examples. / Die vorliegende Arbeit widmet sich der Berechnung elektrostatischer Wechselwirkungen in Partikelsystemen, was beispielsweise im Bereich der molekulardynamischen Simulationen eine zentrale Rolle spielt. Um die dafür benötigten physikalischen Größen mit lediglich O(N log N) arithmetischen Operationen zu berechnen, nutzen sogenannte Teilchen-Gitter-Methoden die Ewald-Summation sowie die schnelle Fourier-Transformation (FFT). Typischerweise können derartige Verfahren Systeme von Punktladungen unter periodischen Randbedingungen in allen Raumrichtungen handhaben. Periodizität ist jedoch nicht immer bezüglich aller drei Dimensionen erwünscht. Des Weiteren spielen auch Wechselwirkungen zu Dipolen in vielen Anwendungen eine wichtige Rolle. Zentraler Gegenstand dieser Arbeit ist die Partikel-Partikel-NFFT Methode (P²NFFT), ein Teilchen-Gitter-Verfahren, welches auf der schnellen Fouriertransformation für nichtäquidistante Daten (NFFT) basiert. Eine Erweiterung dieses Verfahrens auf gemischt periodische sowie offene Randbedingungen wird vorgestellt. Außerdem wird die Methode für die Behandlung von Partikelsystemen, in denen sowohl Ladungen als auch Dipole vorliegen, angepasst. Somit wird erstmalig ein effizienter Algorithmus für gemischte Ladungs-Dipol-Systeme präsentiert, der zusätzlich die Behandlung sämtlicher Arten von Randbedingungen mit einem einheitlichen Zugang erlaubt. Entsprechende Fehlerabschätzungen sowie Strategien für die Parameterwahl werden entwickelt und anhand numerischer Beispiele verifiziert.