We consider a model of N particles interacting through a Kac-style collision process, with m particles among them interacting, in addition, with a thermostat. When m = N, we show exponential approach to the equilibrium canonical distribution in terms of the L2 norm, in relative entropy, and in the Gabetta-Toscani-Wennberg (GTW) metric, at a rate independent of N. When m < N , the exponential rate of approach to equilibrium in L2 is shown to behave as m/N for N large, while the relative entropy and the GTW distance from equilibrium exhibit (at least) an "eventually exponential” decay, with a rate scaling as m/N^2 for large N. As an allied project, we obtain a rigorous microscopic description of the thermostat used, based on a model of a tagged particle colliding with an infinite gas in equilibrium at the thermostat temperature. These results are based on joint work with Federico Bonetto, Michael Loss and Hagop Tossounian.

La courbure de Ricci grossière d’un processus markovien sur un espace polonais est définie comme un taux de contraction local de la distance de Wasserstein W1 entre les lois du processus partant de deux points distincts. La première partie de cette thèse traite de résultats valables dans le cas d’espaces polonais quelconques. On montre que l’infimum de la courbure de Ricci grossière est un taux de contraction global du semigroupe du processus pour la distance W1. Quoiqu’intuitif, ce résultat est difficile à démontrer en temps continu. La preuve de ce résultat, ses conséquences sur le trou spectral du générateur font l’objet du chapitre 1. Un autre résultat intéressant, faisant intervenir les valeurs de la courbure de Ricci grossière en différents points, et pas seulement son infimum, est un résultat de concentration des mesures d’équilibre, valable uniquement en temps discret. Il sera traité dans le chapitre 2. La seconde partie de cette thèse traite du cas particulier des diffusions sur les variétés riemanniennes. Une formule est donnée permettant d’obtenir la courbure de Ricci grossière à partir du générateur. Dans le cas où la métrique est adaptée à la diffusion, nous montrons l’existence d’un couplage entre les trajectoires tel que la courbure de Ricci grossière est exactement le taux de décroissance de la distance entre ces trajectoires. Le trou spectral du générateur de la diffusion est alors plus grand que la moyenne harmonique de la courbure de Ricci. Ce résultat peut être généralisé lorsque la métrique n’est pas celle induite par le générateur, mais il nécessite une hypothèse contraignante, et la courbure que l'on doit considérer est plus faible. / The coarse Ricci curvature of a Markov process on a Polish space is defined as a local contraction rate of the W1 Wasserstein distance between the laws of the process starting at two different points. The first part of this thesis deals with results holding in the case of general Polish spaces. The simplest of them is that the infimum of the coarse Ricci curvature is a global contraction rate of the semigroup of the process for the W1 distance between probability measures. Though intuitive, this result is diffucult to prove in continuous time. The proof of this result, and the following consequences for the spectral gap of the generator are the subject of Chapter 1. Another interesting result, using the values of the coarse Ricci curvature at different points, and not only its infimum, is a concentration result for the equilibrium measures, only holding in a discrete time framework. That will be the topic of Chapter 2. The second part of this thesis deals with the particular case of diffusions on Riemannian manifolds. A formula is given, allowing to get the coarse Ricci curvature from the generator of the diffusion. In the case when the metric is adapted to the diffusion, we show the existence of a coupling between the paths starting at two different points, such that the coarse Ricci curvature is exactly the decreasing rate of the distance between these paths. We can then show that the spectral gap of the generator is at least the harmonic mean of the Ricci curvature. This result can be generalized when the metric is not the one induced by the generator, but it needs a very restricting hypothesis, and the curvature we have to choose is smaller.

Over the last twenty years, as biological, technological, and social net- works have risen in prominence and importance, the study of complex networks has attracted researchers from a wide range of fields. As a result, there is a large and diverse body of literature concerning the properties and development of models for complex networks. However, many of the models that have been previously developed, although quite successful at capturing many observed properties of complex networks, have failed to capture the fundamental semantics of the networks. In this thesis, we propose a robust and general model for complex networks that incorporates at a fundamental level semantic information. We show that for a large range of average degrees and with a suitable choice of parameters, this model exhibits the three hallmark properties of complex networks: small diameter, clustering, and skewed degree distribution. Additionally, we provide a structural interpretation of assortativity and apply this strucutral assortativity to the random dot product graph model. We also extend the results of Chung, Lu, and Vu on the spectral gap of the expected degree sequence model to a general class of random graph models with independent edges. We apply this result to the recently developed Stochastic Kronecker graph model of Leskovec, Chakrabarti, Kleinberg, and Faloutsos.

Dans cette thèse nous nous intéressons principalement aux comportements diffusifs et à la vitesse de convergence vers l'équilibre au sens de la variance de différents modèles de systèmes de particules interagissantes ainsi qu'à un problème de percolation. Nous commençons par introduire informellement le premier sujet. Dans l'étude des systèmes dynamiques, un processus de Markov apériodique et irréductible admettant une mesure invariante converge vers celle-ci en temps long. Dans ce travail, nous nous intéressons ici à la quantification de la vitesse de cette convergence en étudiant la variance du semigroupe associé à la dynamique appliqué à certains ensembles de fonctions. Deux vitesses de convergence sont envisagées ici : la vitesse de de convergence exponentielle impliquée par un trou spectral dans le générateur du processus; une vitesse de convergence polynomiale dite diffusive lorsque le trou spectral est nul.Dans le deuxième chapitre, nous nous étudions le modèle de marche aléatoire en milieu aléatoire et nous prouvons dans ce cadre une vitesse de décroissance de type diffusive.Dans le troisième chapitre, nous étudions le modèle d'exclusion simple à taux dégénérés en dimension 1 appelé ka1f. Nous prouvons des bornes sur le trou spectral en volume fini et une vitesse de décroissance sous-diffusive en volume infini.Dans le quatrième chapitre, nous étudions un modèle à spins non bornés. Nous prouvons une correspondance entre la covariance de l'évolution de deux masses et une marche aléatoire en milieu aléatoire dynamique. Dans le dernier chapitre, nous nous intéressons à un modèle de percolation et à l'étude d'une conjecture étudiant la distance de graphe au sens de la percolation. / In this thesis, we are interested mainly by the diffusive behaviours and the speed of convergence towards equilibrium in the sense of the variance of different models of interacting particles systems and a problem of percolation.We start by introducing unformally the first subject of interest. In the study of dynamic systems, a markov process aperiodic and irreducible having an invariant measure converges towards it in a long time. In this work, we are interested to quantify the speed of this convergence by studying the variance of the semigroup associated to the dynamic applied to some set of functions. Two speeds of convergence are considered: the exponential speed of convergence implied by a spectral gap in the generator of the process; a polynomial tome of convergence called diffusive when the spectral gap is null.In the second chapter, we study the model of random walk in random environment and we prove in this context a diffusive behavior of the speed of convergence.in the third chapter, we study the simple exclusion process with degenerate rates in dimension 1 called ka1F. We prove bounds on the spectral gap in finite volume and a sub-diffusive behavior in infinite volume. In the fourth chapter, we study an unbounded spin model. We prove a relation betweden the covariance of the evolution of two masses and a random walk in a dynamic random environment.In the last chapter, we are interested in the model of percolation and the study of a conjecture studying the distance of graph in the sense of the percolation.

Fundação de Amparo à Pesquisa do Estado do Amazonas / The LBA (Large Scale Biosphere-Atmosphere in Amazônia) project has been using the eddy covariance technique since 1998 to continuously monitor the surface fluxes of energy, water vapor and carbon over Amazônia. The results obtained up to now indicate high level of uncertainties, especially regarding the role of the Amazonian ecosystem to the global carbon budget. Besides the problems related with the Eddy measure system (systematic error and nighttime stable conditions), there is an important factor associated with the averaging time scale or time window used by the scientific community to determine the surface fluxes. This work focuses on the determination of the nocturnal turbulence time scale for long term surface fluxes (carbon, energy and water) over the Amazon rainforest. We used the multi-resolution decomposition technique to project the signal into several time scales and determine when the spectral and co-spectral gap occurred. This technique permitted evaluating and separating the real contribution from turbulent and mesoscale fluxes to the total nocturnal surface fluxes. Our results indicate that the nighttime turbulence time scale is near 100 seconds. It suggests that the time averaging commonly used to calculate nocturnal surface fluxes (30 minutes), needs to be revised. Besides, our results show that, when the mesoscale flux contributions were included, the total nocturnal surface flux was generally underestimated. / Desde 1998 o projeto LBA (Large Scale Biosphere-Atmosphere in Amazônia) vem monitorando continuamente os fluxos de energia, água e carbono na Amazônia utilizando o sistema de Covariância dos Vórtices (CV). Os resultados obtidos até agora apresentam um alto grau de incerteza, especialmente no que diz respeito ao papel do ecossistema amazônico no ciclo global do carbono. Além desses problemas relacionados com as medidas do sistema CV (erros sistemáticos e durante condições estáveis), um outro fator extremamente importante está relacionado com a escala de tempo janela de tempo usada pela comunidade científica para determinar os fluxos de superfície. Este trabalho consiste no esforço inicial em determinar essas escalas para as transferências noturnas (carbono, energia e água) para séries de longo prazo sobre a floresta Amazônica. Nós utilizamos a técnica de multiresolução para projetar o sinal em várias escalas de tempo e determinar quando ocorre a falha coespectral. Esta técnica permite avaliar a real contribuição da turbulência, bem como da mesoescala para o fluxo noturno total. Nossos resultados indicam que a escala de turbulência noturna (comprimento da falha) foi em media 100 s. isto sugere que a escala comumente empregada, que é de 30 min, deve ser revisitada, sobretudo para o sítio estudado. Além disso, os resultados mostram que quando adicionamos a contribuição de mesoescala ao fluxo total, em geral, isto provoca subestimativa das transferências noturnas.

L'objet de cette thèse est l'introduction, l'étude et l'utilisation des cônes complexes multidimensionnels. Dans un premier temps, nous étudions la grassmannienne des espaces de Banach. Nous définissons une notion de bonne décomposition pour les espaces de dimension p et nous démontronsl'équivalence entre la distance de Hausdorff sur la grassmannienne et la distance fournie par une norme sur l'algèbre extérieure.Dans un deuxième temps, nous définissons les cônes complexes p-dimensionnels ainsi qu'une jauge sur les sous-espaces de dimension p de ces cônes. Nous montrons alors un principe de contraction pour cette jauge. Cela nous permet de prouver, pour un opérateur contractant un tel cône, l'existence d'un trou spectral séparant les p valeurs propres dominantes du reste du spectre. Nous utilisons cette théorie pourdémontrer un théorème de régularité analytique pour les exposants de Lyapunov d'un produit aléatoire d'opérateurs contractant un même cône.Nous donnons également une comparaison entre la distance de Hausdorff entre espaces vectoriels et notre jauge.Enfin, nous introduisons une notion de cône dual pour les cônes p-dimensionnels. Dans ce cadre, nous prouvons que les propriétéstopologiques d'un cône se traduisent en propriétés topologiques sur son dual, et réciproquement. Nous complétons le théorème de régularitéprécédent en démontrant l'existence et la régularité d'une décomposition de l'espace en "espace lent" et "espace rapide". / The subject of this thesis is the introduction, the study and the applications of multidimensional complex cones. First, we study the grassmannian of Banach space. We define a notion of right decomposition for p-dimensional spaces and we prove the equivalence between theHausdorff distance on the grassmannian and the distance given by a norm on the exterior algebra.Then, we define p-dimensional complex cones and a gauge on the subspaces of dimension p of these cones. We show a contraction principle for thisgauge. This allows us to prove, for an operator contracting such a cone, the existence of a spectral gap which isolate the p leading eigenvaluesfrom the rest of the spectrum. We use this theory to prove a theorem of analytic regularity for Lyapunov exponents of a random product ofoperators contracting a cone. We also give a comparison between the Hausdorff distance for vector spaces and our gauge.Finally, we introduce a notion of dual cone for p-dimensional cones. In this setting, we prove that the topological properties of a cone translateinto topological properties for its dual and conversely. We complete the previous regularity theorem by proving the existence and the regularity ofa dominated splitting of the space into a "fast space" and a "slow space".

In this dissertation we focus on the application of thermodynamic formalism to non-autonomous and random dynamical systems. Specifically we use the thermodynamic formalism to investigate the dimension of various fractal constructions via the, now standard, technique of Bowen which he developed in his 1979 paper on quasi-Fuchsian groups. Bowen showed, roughly speaking, that the dimension of a fractal is equal to the zero of the relevant topological pressure function. We generalize the results of Rempe-Gillen and Urbanski on non-autonomous iterated function systems to the setting of non-autonomous graph directed Markov systems and then show that the Hausdorff dimension of the fractal limit set is equal to the zero of the associated pressure function provided the size of the alphabets at each time step do not grow too quickly. In trying to remove these growth restrictions, we present several other systems for which Bowen's formula holds, most notably ascending systems. We then use these various constructions to investigate the Hausdorff dimension of various subsets of the Julia set for different large classes of transcendental meromorphic functions of finite order which have been perturbed non-autonomously. In particular we find lower and upper bounds for the dimension of the subset of the Julia set whose points escape to infinity, and in many cases we find the exact dimension. While the upper bound was known previously in the autonomous case, the lower bound was not known in this setting, and all of these results are new in the non-autonomous setting. We also use transfer operator techniques to prove an almost sure invariance principle for random dynamical systems for which the thermodynamical formalism has been well established. In particular, we see that if a system exhibits a fiberwise spectral gap property and the base dynamical system is sufficiently well behaved, i.e. it exhibits an exponential decay of correlations, then the almost sure invariance principle holds. We then apply these results to uniformly expanding random systems like those studied by Mayer, Skorulski, and Urbanski and Denker and Gordin.

Neste trabalho consideramos um processo de Markov a tempo contínuo com espaço de estados finito em um meio aleatório, a saber, a dinâmica de Metropolis para o Modelo de Energia Aleatória Generalizado (GREM) com um número de níveis finito e discutimos o comportamento do seu tempo de chegada ao equilíbrio, o qual é dado pelo inverso da lacuna espectral de sua matriz de probabilidades de transição. No principal resultado desta tese provamos que o quociente entre o volume do sistema e o logaritmo do inverso da lacuna é quase sempre limitado, por cima, por uma função da temperatura, que também é a que descreve a energia livre do GREM sob o regime de temperaturas baixas. Como um estudo adicional, também é discutido um correspondente limitante inferior em um caso particular do GREM com 2 níveis. / In this work we consider a finite state continuous-time Markov process in a random environment, namely, the Metropolis dynamics for the Generalized Random Energy Model (GREM) with a finite number of levels, and we discuss the behavior of its reaching time to equilibrium which is given by inverse of the spectral gap of its transition probability matrix. On the main result of this thesis, we prove the division between the system volume and the logarithm of the inverse of the gap is almost surely upper bounded by a function of the temperature that it is also the function that describe the free energy of the GREM at low temperature. As an additional study, it is also discuss the corresponding limiting lower in a particular case of the 2-level GREM.

Neste trabalho consideramos um processo de Markov a tempo contínuo com espaço de estados finito em um meio aleatório, a saber, a dinâmica de Metropolis para o Modelo de Energia Aleatória Generalizado (GREM) com um número de níveis finito e discutimos o comportamento do seu tempo de chegada ao equilíbrio, o qual é dado pelo inverso da lacuna espectral de sua matriz de probabilidades de transição. No principal resultado desta tese provamos que o quociente entre o volume do sistema e o logaritmo do inverso da lacuna é quase sempre limitado, por cima, por uma função da temperatura, que também é a que descreve a energia livre do GREM sob o regime de temperaturas baixas. Como um estudo adicional, também é discutido um correspondente limitante inferior em um caso particular do GREM com 2 níveis. / In this work we consider a finite state continuous-time Markov process in a random environment, namely, the Metropolis dynamics for the Generalized Random Energy Model (GREM) with a finite number of levels, and we discuss the behavior of its reaching time to equilibrium which is given by inverse of the spectral gap of its transition probability matrix. On the main result of this thesis, we prove the division between the system volume and the logarithm of the inverse of the gap is almost surely upper bounded by a function of the temperature that it is also the function that describe the free energy of the GREM at low temperature. As an additional study, it is also discuss the corresponding limiting lower in a particular case of the 2-level GREM.

En application, on souhaite générer des nombres aléatoires avec une loi précise (méthode de Monte Carlo par chaines de Markov - MCMC (Markov Chaine Monte Carlo)). La méthode consiste à trouver une diffusion qui a la loi invariante souhaitée et à montrer la convergence de cette diffusion vers son équilibre avec une vitesse exponentielle. L’exposant de cette convergence est le trou spectral du générateur. Il a été montré par Chii-Ruey Hwang, Shu-Yin Hwang-Ma, et Shuenn-Jyi Sheu qu’on peut agrandir le trou spectral, en rajoutant un terme non-symétrique au générateur auto-adjoint (souvent utilisé en MCMC). Ceci correspond à passer d’une diffusion réversible (en detailed balance) à une diffusion non réversible. Un moyen de construire une diffusion non-réversible avec la même mesure invariante est de rajouter un flot incompressible à la dynamique de la diffusion réversible.Dans cette thèse, nous étudions le comportement de la diffusion lorsqu’on accélère le flot sous-jacent en multipliant le champ des vecteurs qui le décrit par une grande constante. P. Constantin, A.Kisekev, L.Ryzhik et A.Zlatoš (2008) ont montré que si le flot était faiblement mélangeant alors l’accélération du flot suffisait pour faire converger la diffusion vers son équilibre en un temps fini. Dans ce travail, on explicite la vitesse de ce phénomène sous une condition de corrélation du flot. L’article de B. Franke, C.-R.Hwang, H.-M. Pai et S.-J. Sheu (2010) donne l’expression asymptotique du trou spectral lorsque le flot sous-jacent est accéléré vers l’infini. Ici aussi, on s’intéresse à la vitesse avec laquelle le phénomène se manifeste. Dans un premier temps, nous étudions le cas particulier d’une diffusion du type Ornstein-Uhlenbeck qui est perturbée par un flot préservant la mesure gaussienne. Dans ce cas, grâce à un résultat de G. Metafune, D. Pallara et E. Priola (2002), nous pouvons réduire l’étude du spectre du générateur à des valeurs propres d’une famille de matrices. Nous étudions ce problème avec des méthodes de développement limité des valeurs propres. Ce problème est résolu explicitement dans cette thèse et nous donnons aussi une borne pour le rayon de convergence du développement. Nous généralisons ensuite cette méthode dans le cas d’une diffusion générale de façon formelle. Ces résultats peuvent être utiles pour avoir une première idée sur les vitesses de convergence du trou spectral décrites dans l’article de Franke et al. (2010). / In application, we would like to generate random numbers with a precise law MCMC (Markov Chaine Monte Carlo). The method consists in finding a diffusion which has the desired invariant law and in showing the convergence of this diffusion towards its equilibrium with an exponential rate. The exponent of this convergence is the spectral gap of the generator. It was shown by C.-R. Hwang, S.-Y. Hwang-Ma and S.-J. Sheu that the spectral gap can grow up by adding a non-symmetric term to the self-adjoint generator.This corresponds to passing from a reversible diffusion to a non-reversible diffusion. A means of constructing a non-reversible diffusion with the same invariant measure is to add an incompressible flow to the dynamics of the reversible diffusion.In this thesis, we study the behavior of diffusion when the flow is accelerated by multiplying the field of the vectors which describes it by a large constant. In 2008, P. Constantin, A. Kisekev, L. Ryzhik and A. Zlatoˇs have shown that if the flow was weakly mixing then the acceleration of the flow was sufficient to converge the diffusion towards its equilibrium after finite time. In this work, the speed of this phenomenon is explained under a condition of correlation of the flow. The article by B. Franke, C.-R.Hwang, H.-M. Pai and S.-J.Sheu (2010) gives the asymptotic expression of the spectral gap when the large constant goes to infinity. Here we are also interested in the speed with which the phenomenon manifests itself. First, we study the special case of an Ornstein-Uhlenbeck diffusion which is perturbed by a flow preserving the Gaussian measure. In this case, thanks to a result of G. Metafune, D. Pallara and E. Priola (2002), we can reduce the study of the generator spectrum to eigenvalues of a family of matrices. We study this problem with methods of limited development of eigenvalues. This problem is solved explicitly in this thesis and we also give a boundary for the convergence radius of the development. We then generalize this method in the case of a general diffusion in a formal way. These results may be useful to have a first idea on the speeds of convergence of the spectral gap described in the article by Franke et al. (2010).