Les travaux exposés dans cette thèse sont consacrés à l’étude de méthodesprécises pour approcher des équations différentielles stochastiques ou deséquations aux dérivées partielles (EDP) déterministes. La première parties’inscrit dans le cadre du développement de méthodes visant à corriger le biaisdans les processus de diffusion paramétrique. Trois modèles sont étudiés enparticulier : Ornstein-Uhlenbeck, Auto-régressif et Moyenne mobile. A l’issuede ce travail, plusieurs approximations de biais ont été proposées suivant deuxapproches : la première consiste en un développement de Taylor del’estimateur obtenu alors que la seconde s'appuie sur une expansionstochastique de celui-ci.La deuxième partie de cette thèse porte sur l’approximation de l’équation de lachaleur obtenue après changement de variables à partir du modèle de Black etScholes. En général, on préfère utiliser des méthodes implicites pour résoudredes EDP paraboliques mais depuis quelques années, les méthodes dites deRunge-Kutta explicites stabilisées, sont de plus en plus utilisées. Nousmontrons que l’utilisation de ce type de méthodes explicites et notamment lesschémas ROCK donnent de très bons résultats même si les conditions initialessont peu régulières, ce qui est le cas dans les modèles financiers / The work presented in this thesis is devoted to the study of precise methods forapproximating stochastic differential equations (SDE) or deterministic partialdifferential equations (PDE). The first part is devoted to the development ofbias correction methods in parametric diffusion processes. Three models arestudied in particular : Ornstein-Uhlenbeck, auto-regressive and Movingaverage. At the end of this work, several approximations of bias have beenproposed following two approaches : the first consists in a Taylor developmentof the obtained estimator while the second one relies on a stochastic expansionof the latter.The second part of this thesis deals with the approximation of the heatequation obtained after changing variables from the Black-Scholes model. Likethe vast majority of PDE, this equation does not have an exact solution, sosolutions must be approached using explicit or implicit time schemes. Itis often customary to prefer the use of implicit methods to solve parabolic PDEsuch as the heat equation, but in the past few years, the stabilized explicitRunge-Kutta methods which have the largest possible domains of stabilityalong the negative real axis, are increasingly used. We show that the useof this type of explicit methods and in particular the ROCK (Runge-Orthogonal-Chebyshev-Kutta) schemes give very good results even if the initial conditionsare not very regular, which is the case in the financial models

In dieser Dissertation werden algorithmische Verbesserungen und Varianten für Simulationen der zwei-Flavor Gitter QCD mit dynamischen Fermionen studiert. Der O(a)-verbesserte Dirac-Wilson-Operator wird im Schrödinger Funktional mit einem Update des Hybrid Monte Carlo (HMC)-Typs verwendet. Sowohl der Hermitische als auch der nicht-Hermitische Operator werden betrachtet. Für den Hermitischen Dirac-Wilson-Operator untersuchen wir die Vorteile des symmetrischen gegenüber dem asymmetrischen Gerade-Ungerade-Präkonditionierens, wie man von einem mehr Zeitskalen-Integrator profitieren kann, sowie die Auswirkungen der kleinsten Eigenwerte auf die Stabilität des HMC Algorithmus. Im Fall des nicht-Hermitischen Operators leiten wir eine (semi)-analytische Schranke für das Spektrum her und zeigen eine Methode, um Informationen über den spektralen Rand zu gewinnen, indem wir komplexe Eigenwerte mit dem Lanczos-Algorithmus abschätzen. Diese spektralen Ränder erlauben es, Vorzüge des symmetrischen Gerade-Ungerade-Präkonditionierens oder den Effekt des Sheikholeslami-Wohlert-Terms für das Spektrum des nicht-Hermitischen Operators zu zeigen. Unter Verwendung der Informationen des spektralen Randes konstruieren wir angepasste, komplexe, skalierte und verschobene Tschebyschow Polynome zur Approximation des inversen Dirac-Wilson-Operators. Basierend auf diesen Polynomen entwickeln wir eine neue HMC-Variante, genannt nicht-Hermitischer polynomialer Hybrid Monte Carlo (NPHMC). Sie erlaubt, vom Importance Sampling unter Kompensation mit einem Gewichtungsfaktor abzuweichen. Zudem wird eine Erweiterung durch Anwendung des Hasenbusch-Tricks abgeleitet. Erste Größen der Leistungsfähigkeit, die die Abhängingkeit von den Eingabeparametern als auch einen Vergleich mit unserem Standard-HMC zeigen, werden präsentiert. Im Vergleich der beiden ein-Pseudofermion-Varianten ist der neue NPHMC etwas besser; eine eindeutige Aussage im Fall der zwei-Pseudofermion-Variante ist noch nicht möglich. / In this thesis algorithmic improvements and variants for two-flavor lattice QCD simulations with dynamical fermions are studied using the O(a)-improved Dirac-Wilson operator in the Schrödinger functional setup and employing a hybrid Monte Carlo-type (HMC) update. Both, the Hermitian and the Non-Hermitian operator are considered. For the Hermitian Dirac-Wilson operator we investigate the advantages of symmetric over asymmetric even-odd preconditioning, how to gain from multiple time scale integration as well as how the smallest eigenvalues affect the stability of the HMC algorithm. In case of the non-Hermitian operator we first derive (semi-)analytical bounds on the spectrum before demonstrating a method to obtain information on the spectral boundary by estimating complex eigenvalues with the Lanzcos algorithm. These spectral boundaries allow to visualize the advantage of symmetric even-odd preconditioning or the effect of the Sheikholeslami-Wohlert term on the spectrum of the non-Hermitian Dirac-Wilson operator. Taking advantage of the information of the spectral boundary we design best-suited, complex, scaled and translated Chebyshev polynomials to approximate the inverse Dirac-Wilson operator. Based on these polynomials we derive a new HMC variant, named non-Hermitian polynomial Hybrid Monte Carlo (NPHMC), which allows to deviate from importance sampling by compensation with a reweighting factor. Furthermore an extension employing the Hasenbusch-trick is derived. First performance figures showing the dependence on the input parameters as well as a comparison to our standard HMC are given. Comparing both algorithms with one pseudo-fermion, we find the new NPHMC to be slightly superior, whereas a clear statement for the two pseudo-fermion variants is yet not possible.

Le but de cette thèse est d'étudier les trois problèmes suivantes : 1) On s'intéresse à la stabilité des cônes normaux et des sous-différentiels via deux types de convergence d'ensembles et de fonctions : La convergence au sens de Mosco et celle d'Attouch-Wets. Les résultats obtenus peuvent être vus comme une extension du théorème d'Attouch aux fonctions non nécessairement convexes sur des espaces de Banach localement uniformément convexes. 2) Pour une bornologie β donnée sur un espace de Banach X, on étudie la validité de la formule suivante (…). Ici Tβ(C; x) et Tc(C; x) désignent le β -cône tangent et le cône tangent de Clarke à C en x. On montre que si, X x X est ∂β-« trusted » alors cette formule est valable pour tout ensemble fermé non vide C ⊂ X et x ∈ C. Cette classe d'espaces contient les espaces ayant une norme équivalent β-différentiable, etplus généralement les espaces possédant une fonction "bosse" lipschitzienne et β-différentiable). Comme conséquence, on obtient que pour la bornologie de Fréchet, cette formule caractérise les espaces d'Asplund. 3) On examine la convexité des ensembles de Chebyshev. Il est bien connu que, dans un espace normé réflexif ayant la propriété Kadec-Klee, tout ensemble de Chebyshev faiblement fermé est convexe. On démontre que la condition de faible fermeture peut être remplacée par la fermeture faible locale, c'est-à-dire pour tout x ∈ C il existe ∈ > 0 tel que C ∩ B(x, ε) est faiblement fermé. On montre aussi que la propriété Kadec-Klee n'est plus exigée lorsque l'ensemble de Chebyshev est représenté comme une union d'ensembles convexes fermés. / The aim of this thesis is to study the following three problems: 1) We are concerned with the behavior of normal cones and subdifferentials with respect to two types of convergence of sets and functions: Mosco and Attouch-Wets convergences. Our analysis is devoted to proximal, Fréchet, and Mordukhovich limiting normal cones and subdifferentials. The results obtained can be seen as extensions of Attouch theorem to the context of non-convex functions on locally uniformly convex Banach space. 2) For a given bornology β on a Banach space X we are interested in the validity of the following "lim inf" formula (…).Here Tβ(C; x) and Tc(C; x) denote the β-tangent cone and the Clarke tangent cone to C at x. We proved that it holds true for every closed set C ⊂ X and any x ∈ C, provided that the space X x X is ∂β-trusted. The trustworthiness includes spaces with an equivalent β-differentiable norm or more generally with a Lipschitz β-differentiable bump function. As a consequence, we show that for the Fréchet bornology, this "lim inf" formula characterizes in fact the Asplund property of X. 3) We investigate the convexity of Chebyshev sets. It is well known that in a smooth reflexive Banach space with the Kadec-Klee property every weakly closed Chebyshev subset is convex. We prove that the condition of the weak closedness can be replaced by the local weak closedness, that is, for any x ∈ C there is ∈ > 0 such that C ∩ B(x, ε) is weakly closed. We also prove that the Kadec-Klee property is not required when the Chebyshev set is represented by a finite union of closed convex sets.

No description available.

In non-rotating fluids, boundary-layer separation occurs when the nearly inviscid flow just outside a viscous boundary-layer experiences an appreciable deceleration due to a region of adverse pressure gradient. The fluid ceases to flow along the boundary due to a flow recirculation region close to the boundary. The flow is then said to be "detached." In recent decades, attention has shifted to the study of boundary-layer separation in a rotating reference frame due to its significance in Geophysical Fluid Dynamics (GFD). Since the Earth is a rotating sphere, the so-called β-plane approximation f = f0 + βy is often used to account for the inherent meridional variation of the Coriolis parameter, f, while still solving the governing equations on a plane. Numerical simulations of currents on the β-plane have been useful in understanding ocean currents such as the Gulf Stream, the Brazil Current, and the Antarctic Circumpolar Current to name a few. In this thesis, we first consider the problem of prograde flow past a cylindrical obstacle on the β-plane. The problem is governed by the barotropic vorticity equation and is solved using a numerical method that is a combination of a finite difference method and a spectral method. A modified form of the β-plane approximation is proposed to avoid computational difficulties. Results are given and discussed for flow past a circular cylinder at selected Reynolds numbers (Re) and non-dimensional β-parameters (β^). Results are then given and discussed for flow past an elliptic cylinder of a fixed aspect ratio (r = 0.2) and at two angles of inclination (90°, 15°) at selected Re and β^. In general, it is found that the β-effect acts to suppress boundary-layer separation and to allow Rossby waves to form in the exterior flow field. In the asymmetrical case of an inclined elliptic cylinder, the β-effect was found to constrain the region of vortex shedding to a small region near the trailing edge of the cylinder. The shed vortices were found to propagate around the trailing edge instead of in the expected downstream direction, as observed in the non-rotating case. The second problem considered in this thesis is the separation of western boundary currents from a curved coastline. This problem is also governed by the barotropic vorticity equation, and it is solved on an idealized model domain suitable for investigating the effects that boundary curvature has on the tendency of a boundary current to separate. The numerical method employed is a two-dimensional Chebyshev spectral collocation method and yields high order accuracy that helps to better resolve the boundary-layer dynamics in comparison to low-order methods. Results are given for a selection of boundary curvatures, non-dimensional β-parameters (β^), Reynolds numbers (Re), and Munk Numbers (Mu). In general, it is found than an increase in β^ will act to suppress boundary-layer separation. However, a sufficiently sharp obstacle can overcome the β-effect and force the boundary current to separate regardless of the value of β^. It is also found that in the inertial limit (small Mu, large Re) the flow region to the east of the primary boundary current is dominated by strong wave interactions and large eddies which form as a result of shear instabilities. In an interesting case of the inertial limit, strong waves were found to interact with the separation region, causing it to expand and propagate to the east as a large eddy. This idealized the mechanism by which western boundary currents such as the Gulf Stream generate eddies in the world's oceans.

In non-rotating fluids, boundary-layer separation occurs when the nearly inviscid flow just outside a viscous boundary-layer experiences an appreciable deceleration due to a region of adverse pressure gradient. The fluid ceases to flow along the boundary due to a flow recirculation region close to the boundary. The flow is then said to be "detached." In recent decades, attention has shifted to the study of boundary-layer separation in a rotating reference frame due to its significance in Geophysical Fluid Dynamics (GFD). Since the Earth is a rotating sphere, the so-called β-plane approximation f = f0 + βy is often used to account for the inherent meridional variation of the Coriolis parameter, f, while still solving the governing equations on a plane. Numerical simulations of currents on the β-plane have been useful in understanding ocean currents such as the Gulf Stream, the Brazil Current, and the Antarctic Circumpolar Current to name a few. In this thesis, we first consider the problem of prograde flow past a cylindrical obstacle on the β-plane. The problem is governed by the barotropic vorticity equation and is solved using a numerical method that is a combination of a finite difference method and a spectral method. A modified form of the β-plane approximation is proposed to avoid computational difficulties. Results are given and discussed for flow past a circular cylinder at selected Reynolds numbers (Re) and non-dimensional β-parameters (β^). Results are then given and discussed for flow past an elliptic cylinder of a fixed aspect ratio (r = 0.2) and at two angles of inclination (90°, 15°) at selected Re and β^. In general, it is found that the β-effect acts to suppress boundary-layer separation and to allow Rossby waves to form in the exterior flow field. In the asymmetrical case of an inclined elliptic cylinder, the β-effect was found to constrain the region of vortex shedding to a small region near the trailing edge of the cylinder. The shed vortices were found to propagate around the trailing edge instead of in the expected downstream direction, as observed in the non-rotating case. The second problem considered in this thesis is the separation of western boundary currents from a curved coastline. This problem is also governed by the barotropic vorticity equation, and it is solved on an idealized model domain suitable for investigating the effects that boundary curvature has on the tendency of a boundary current to separate. The numerical method employed is a two-dimensional Chebyshev spectral collocation method and yields high order accuracy that helps to better resolve the boundary-layer dynamics in comparison to low-order methods. Results are given for a selection of boundary curvatures, non-dimensional β-parameters (β^), Reynolds numbers (Re), and Munk Numbers (Mu). In general, it is found than an increase in β^ will act to suppress boundary-layer separation. However, a sufficiently sharp obstacle can overcome the β-effect and force the boundary current to separate regardless of the value of β^. It is also found that in the inertial limit (small Mu, large Re) the flow region to the east of the primary boundary current is dominated by strong wave interactions and large eddies which form as a result of shear instabilities. In an interesting case of the inertial limit, strong waves were found to interact with the separation region, causing it to expand and propagate to the east as a large eddy. This idealized the mechanism by which western boundary currents such as the Gulf Stream generate eddies in the world's oceans.

Numerical methods are often required to solve chemical problems, either to verify theoretical models or to access information that is not readily available experimentally. This thesis deals with both situations, though in differing levels of detail. A major component of this thesis is devoted to developing new methods to determine a full eigendecomposition of the matrices derived from "low temperature" unimolecular master equations. When transient behaviour is of interest achieving relative accuracy for more than just the eigenvector corresponding to the smallest eigenvalue is of central importance. Three new methods are presented. The first is based on a weighted implementation of subspace projection methods, in this case explored for the well-known Arnoldi method. This weighted inner product subspace projection methodology is demonstrated to

Numerical methods are often required to solve chemical problems, either to verify theoretical models or to access information that is not readily available experimentally. This thesis deals with both situations, though in differing levels of detail. A major component of this thesis is devoted to developing new methods to determine a full eigendecomposition of the matrices derived from "low temperature" unimolecular master equations. When transient behaviour is of interest achieving relative accuracy for more than just the eigenvector corresponding to the smallest eigenvalue is of central importance. Three new methods are presented. The first is based on a weighted implementation of subspace projection methods, in this case explored for the well-known Arnoldi method. This weighted inner product subspace projection methodology is demonstrated to

Numerical methods are often required to solve chemical problems, either to verify theoretical models or to access information that is not readily available experimentally. This thesis deals with both situations, though in differing levels of detail. A major component of this thesis is devoted to developing new methods to determine a full eigendecomposition of the matrices derived from "low temperature" unimolecular master equations. When transient behaviour is of interest achieving relative accuracy for more than just the eigenvector corresponding to the smallest eigenvalue is of central importance. Three new methods are presented. The first is based on a weighted implementation of subspace projection methods, in this case explored for the well-known Arnoldi method. This weighted inner product subspace projection methodology is demonstrated to

Numerical methods are often required to solve chemical problems, either to verify theoretical models or to access information that is not readily available experimentally. This thesis deals with both situations, though in differing levels of detail. A major component of this thesis is devoted to developing new methods to determine a full eigendecomposition of the matrices derived from "low temperature" unimolecular master equations. When transient behaviour is of interest achieving relative accuracy for more than just the eigenvector corresponding to the smallest eigenvalue is of central importance. Three new methods are presented. The first is based on a weighted implementation of subspace projection methods, in this case explored for the well-known Arnoldi method. This weighted inner product subspace projection methodology is demonstrated to