This thesis studies two Markov processes describing the evolution of a system of many interacting random components. These processes undergo an absorbing-state phase transition, i.e., as one variates the parameter values, the process exhibits a transition from a convergence regime to one of the absorbing-states to an active regime. In Chapter 2 we study Activated Random Walk, which is an interacting particle system where the particles can be of two types and their number is conserved. Firstly, we provide a new lower bound for the critical density on Z as a function of the jump distribution and of the sleeping rate and we prove that the critical density is not a constant function of the jump distribution. Secondly, we prove that on Zd in the case of biased jump distribution the critical density is strictly less than one, provided that the sleeping rate is small enough. This answers a question that has been asked by Dickman, Rolla, Sidoravicius [9, 28] in the case of biased jump distribution. Our results have been presented in [33]. In Chapter 3 we study a class of probabilistic cellular automata which are related by a natural coupling to a special type of oriented percolation model. Firstly, we consider the process on a finite torus of size n, which is ergodic for any parameter value. By employing dynamic-renormalization techniques, we prove that the average absorption time grows exponentially (resp. logarithmically) with n when the model on Z is in the active (resp. absorbing) regime. This answers a question that has been asked by Toom [37]. Secondly, we study how the neighbourhood of the model affects the critical probability for the process on Z. We provide a lower bound for the critical probability as a function of the neighbourhood and we show that our estimates are sharp by comparing them with our numerical estimates. Our results have been presented in [34, 35].

Nosso trabalho considera o processo de exclusão simples do vizinho mais próximo evoluindo com taxas de salto aleatórias . Demonstramos o limite hidrodinâmico deste processo. Este resultado è obtido através do limite hidrodinâmico do processo de exclusão onde as taxas de salto iniciais são substituidas pelas taxas cx,N que tem a mesma distribuição para cada N maior ou igal a 1. Fazemos algumas suposições no meio c_N e consideramos que as partículas estão inicialmente distribuidas de acordo à medida produto de Bernoulli associada a um perfil inicial. / Consider a Poisson process with rate equal to 1 in IR. Consider the nearest neighbor simple exclusion process with random jump rates § where §x =\\lambda, \\lambda > 0 if there is a Poisson mark between (x, x + 1) and §x = 1 otherwise. We prove the hydrodynamic limit of this process. This result follows from the hydrodynamic limit in the case that the jump rates {§x : x inteiro} are replaced by an array {cx,N : x inteiro} having the same distribution for each N >=1.

Nosso trabalho considera o processo de exclusão simples do vizinho mais próximo evoluindo com taxas de salto aleatórias . Demonstramos o limite hidrodinâmico deste processo. Este resultado è obtido através do limite hidrodinâmico do processo de exclusão onde as taxas de salto iniciais são substituidas pelas taxas cx,N que tem a mesma distribuição para cada N maior ou igal a 1. Fazemos algumas suposições no meio c_N e consideramos que as partículas estão inicialmente distribuidas de acordo à medida produto de Bernoulli associada a um perfil inicial. / Consider a Poisson process with rate equal to 1 in IR. Consider the nearest neighbor simple exclusion process with random jump rates § where §x =\\lambda, \\lambda > 0 if there is a Poisson mark between (x, x + 1) and §x = 1 otherwise. We prove the hydrodynamic limit of this process. This result follows from the hydrodynamic limit in the case that the jump rates {§x : x inteiro} are replaced by an array {cx,N : x inteiro} having the same distribution for each N >=1.

This thesis studies two Markov processes describing the evolution of a system of many interacting random components. These processes undergo an absorbing-state phase transition, i.e., as one variates the parameter values, the process exhibits a transition from a convergence regime to one of the absorbing-states to an active regime. In Chapter 2 we study Activated Random Walk, which is an interacting particle system where the particles can be of two types and their number is conserved. Firstly, we provide a new lower bound for the critical density on Z as a function of the jump distribution and of the sleeping rate and we prove that the critical density is not a constant function of the jump distribution. Secondly, we prove that on Zd in the case of biased jump distribution the critical density is strictly less than one, provided that the sleeping rate is small enough. This answers a question that has been asked by Dickman, Rolla, Sidoravicius [9, 28] in the case of biased jump distribution. Our results have been presented in [33]. In Chapter 3 we study a class of probabilistic cellular automata which are related by a natural coupling to a special type of oriented percolation model. Firstly, we consider the process on a finite torus of size n, which is ergodic for any parameter value. By employing dynamic-renormalization techniques, we prove that the average absorption time grows exponentially (resp. logarithmically) with n when the model on Z is in the active (resp. absorbing) regime. This answers a question that has been asked by Toom [37]. Secondly, we study how the neighbourhood of the model affects the critical probability for the process on Z. We provide a lower bound for the critical probability as a function of the neighbourhood and we show that our estimates are sharp by comparing them with our numerical estimates. Our results have been presented in [34, 35].

Cette thèse est consacrée à l'étude de systèmes subissant des coagulations et fragmentations successives. Dans le cas déterministe, on travaille avec des solutions mesures de l'équation de coagulation - multifragmentation. On étudie aussi la contrepartie stochastique de ces systèmes : les processus de coalescence - multifragmentation qui sont des processus de Markov à sauts. Dans un premier temps, on étudie le phénomène de coagulation seul. D'un côté, l'équation de Smoluchowski est une équation intégro-différentielle déterministe. D'un autre côté, on considère le processus stochastique connu sous le nom de Marcus-Lushnikov qui peut être regardé comme une approximation de la solution de l'équation de Smoluchowski. Nous étudions la vitesse de convergence par rapport à la distance de type Wassertein $d_{lambda}$ entre les mesures lorsque le nombre de particules tend vers l'infini. Notre étude est basée sur l'homogénéité du noyau de coagulation $K$.On complémente les calculs pour obtenir un résultat qui peut être interprété comme une généralisation de la Loi des Grands Nombres. Des conditions générales et suffisantes sur des mesures discrètes et continues $mu_0$ sont données pour qu'une suite de mesures $mu_0^n$ à support compact existe. On a donc trouvé un taux de convergence satisfaisant du processus Marcus-Lushnikov vers la solution de l'équation de Smoluchowski par rapport à la distance de type Wassertein $d_{lambda}$ égale à $1/sqrt{n}$.Dans un deuxième temps on présente les résultats des simulations ayant pour objectif de vérifier numériquement le taux de convergence déduit précédemment pour les noyaux de coagulation qui y sont étudiés. Finalement, on considère un modèle prenant en compte aussi un phénomène de fragmentation où un nombre infini de fragments à chaque dislocation est permis. Dans la première partie on considère le cas déterministe, dans la deuxième partie on étudie un processus stochastique qui peut être interprété comme la version macroscopique de ce modèle. D'abord, on considère l'équation intégro-partielle différentielle de coagulation - multifragmentation qui décrit l'évolution en temps de la concentration $mu_t(x)$ de particules de masse $x>0$. Le noyau de coagulation $K$ est supposé satisfaire une propriété de $lambda$-homogénéité pour $lambdain(0,1]$, le noyau de fragmentation $F$ est supposé borné et la mesure $beta$ sur l'ensemble de ratios est conservative. Lorsque le moment d'ordre $lambda$ de la condition initial $mu_0$ est fini, on est capable de montrer existence et unicité d'une solution mesure de l'équation de coagulation - multifragmentation. Ensuite, on considère la version stochastique de cette équation, le processus de coalescence - fragmentation est un processus de Markov càdlàg avec espace d'états l'ensemble de suites ordonnées et est défini par un générateur infinitésimal donné. On a utilisé une représentation Poissonienne de ce processus et la distance $delta_{lambda}$ entre deux processus. Grâce à cette méthode on est capable de construire une version finie de ce processus et de coupler deux processus démarrant d'états initiaux différents. Lorsque l'état initial possède un moment d'ordre $lambda$ fini, on prouve existence et unicité de ces processus comme la limite de suites de processus finis. Tout comme dans le cas déterministe, le noyau de coagulation $K$ est supposé satisfaire une propriété d'homogénéité. Les hypothèses concernant la mesure $beta$ sont exactement les mêmes. D'un autre côté, le noyau de fragmentation $F$ est supposé borné sur tout compact dans $(0,infty)$. Ce résultat est meilleur que celui du cas déterministe, cette amélioration est due à la propriété intrinsèque de masse totale non-explosive que possède un système avec un moment fini d'ordre $lambda$ / This thesis is devoted to the study of systems of particles undergoing successive coagulations and fragmentations. In the deterministic case, we deal with measure-valued solutions of the coagulation - multifragmentation equation. We also study, on the other hand, its stochastic counterpart: coalescence - multifragmentation Markov processes. A first chapter is devoted to the presentation of the mathematical tools used in this thesis and to the discussion on some topics treated in the following chapters. n Chapter 1 we only take into account coagulation phenomena. We consider the Smoluchowski equation (which is deterministic) and the Marcus-Lushnikov process (the stochastic version) which can be seen as an approximation of the Smoluchowski equation. We derive a satisfying rate of convergence of the Marcus-Lushnikov process toward the solution to Smoluchowski's coagulation equation. The result applies to a class of homogeneous-like coagulation kernels with homogeneity degree ranging in $(-infty,1]$. It relies on the use of the Wasserstein-type distance $d_{lambda}$, which has shown to be particularly well-adapted to coalescence phenomena. It was introduced and used in preceding works. In Chapter 2 we perform some simulations in order to confirm numerically the rate of convergence deduced in Chapter ref{Chapter1} for the kernels studied in this chapter.medskip Finally, in Chapter 3 we add a fragmentation phenomena and consider a coagulation multiple-fragmentation equation, which describes the concentration $c_t(x)$ of particles of mass $x in (0,infty)$ at the instant $t geq 0$. We study the existence and uniqueness of measured-valued solutions to this equation for homogeneous-like kernels of homogeneity parameter $lambda in (0,1]$ and bounded fragmentation kernels, although a possibly infinite number of fragments is considered. We also study a stochastic counterpart of this equation where a similar result is shown. We prove existence of such a process for a larger set of fragmentation kernels, namely we relax the boundedness hypothesis. In both cases, the initial state has a finite $lambda$-moment

Estudamos um modelo de passeios aleatórios simples em grafos, conhecido como modelo dos sapos. Esse modelo pode ser descrito de maneira geral da seguinte forma: existem partículas ativas e partículas desativadas num grafo G. Cada partícula ativa desempenha um passeio aleatório simples a tempo discreto e a cada momento ela pode morrer com probabilidade 1-p. Quando uma partícula ativa entra em contato com uma partícula desativada, esta é ativada e também passa a realizar, de maneira independente, um passeio aleatório pelo grafo. Apresentamos limites superior e inferior para o parâmetro crítico de sobrevivência do modelo dos sapos na árvore, e demonstramos que este parâmetro crítico não é uma função monótona do grafo em que está definido. / We study a system of simple random walks on graphs, known as frog model. This model can be described generally speaking as follows: there are active and sleeping particles living on some graph G. Each particle performs a simple random walk with discrete time and at each moment it may disappear with probability 1 - p. When an active particle hits a sleeping particle, the latter becomes active and starts to perform, independently, a simple random walk on the graph. We present lower and upper bounds for the surviving critical parameter on the tree, and we show that this parameter is not a monotonic function of the graph it is defined on.

This thesis examines the mixing times for one-dimensional interacting particle systems. We use the coupling method to study the mixing rates for particle systems on the circle which move according to specific permutations e.g., transpositions and 3-cycles. / Graduation date: 2008

Estudamos um modelo de passeios aleatórios simples em grafos, conhecido como modelo dos sapos. Esse modelo pode ser descrito de maneira geral da seguinte forma: existem partículas ativas e partículas desativadas num grafo G. Cada partícula ativa desempenha um passeio aleatório simples a tempo discreto e a cada momento ela pode morrer com probabilidade 1-p. Quando uma partícula ativa entra em contato com uma partícula desativada, esta é ativada e também passa a realizar, de maneira independente, um passeio aleatório pelo grafo. Apresentamos limites superior e inferior para o parâmetro crítico de sobrevivência do modelo dos sapos na árvore, e demonstramos que este parâmetro crítico não é uma função monótona do grafo em que está definido. / We study a system of simple random walks on graphs, known as frog model. This model can be described generally speaking as follows: there are active and sleeping particles living on some graph G. Each particle performs a simple random walk with discrete time and at each moment it may disappear with probability 1 - p. When an active particle hits a sleeping particle, the latter becomes active and starts to perform, independently, a simple random walk on the graph. We present lower and upper bounds for the surviving critical parameter on the tree, and we show that this parameter is not a monotonic function of the graph it is defined on.

O propósito desta dissertação é combinar tópicos de percolação e processo de contato para formular e obter resultados em um modelo de sistema de partículas que é inspirado no fenômeno de difusão de uma inovação em uma população estruturada. Mais precisamente, propomos uma cadeia de Markov a tempo contínuo definida na rede hipercúbica d-dimensional. Cada indivíduo da população deve estar em algum dos três estados pertencentes ao conjunto {0; 1; 2}. Nesse modelo, 0 representa ignorante, 1 consciente e 2 adotador. Serão estudados argumentos que permitam encontrar condições suficientes nas quais a inovação se espalha ou não com probabilidade positiva. Isto envolve o estudo de modelos de percolação e do processo de contato. / The purpose of this work is to combine percolation and contact process topics to formulate and achieve results in a particle system model that is inspired by the diffusion phenomenon of an innovation in a structured population. More precisely, we proposed a continuous time Markov chain defined in a population represented by the d-dimensional integer lattice. Each agent of population may be in any of the three states belonging to the set {0; 1; 2}. In this model, 0 stands for ignorant, 1 for aware and 2 for adopter. The arguments, that allow to obtain sufficient conditions under which the innovation either becomes extinct or survives with positive probability, will be studied. This involves the study of percolation models and contact process.

O propósito desta dissertação é combinar tópicos de percolação e processo de contato para formular e obter resultados em um modelo de sistema de partículas que é inspirado no fenômeno de difusão de uma inovação em uma população estruturada. Mais precisamente, propomos uma cadeia de Markov a tempo contínuo definida na rede hipercúbica d-dimensional. Cada indivíduo da população deve estar em algum dos três estados pertencentes ao conjunto {0; 1; 2}. Nesse modelo, 0 representa ignorante, 1 consciente e 2 adotador. Serão estudados argumentos que permitam encontrar condições suficientes nas quais a inovação se espalha ou não com probabilidade positiva. Isto envolve o estudo de modelos de percolação e do processo de contato. / The purpose of this work is to combine percolation and contact process topics to formulate and achieve results in a particle system model that is inspired by the diffusion phenomenon of an innovation in a structured population. More precisely, we proposed a continuous time Markov chain defined in a population represented by the d-dimensional integer lattice. Each agent of population may be in any of the three states belonging to the set {0; 1; 2}. In this model, 0 stands for ignorant, 1 for aware and 2 for adopter. The arguments, that allow to obtain sufficient conditions under which the innovation either becomes extinct or survives with positive probability, will be studied. This involves the study of percolation models and contact process.