The Stochastic wave equation: study of the law and approximations

Abstract

[eng] This dissertation is devoted to the study of some aspects of the theory of stochastic partial differential equations. More precisely, we mainly focus on the study of a stochastic wave equation perturbed by some random noise. The contents of the thesis may be split in two parts: firstly, we deal with a stochastic wave equation in spatial dimension three with a random perturbation given by a Gaussian noise. In this case, the main objective is to study the existence and smoothness properties of the density of the solution of the equation. Secondly, we handle a one-dimensional stochastic wave equation controlled by the so called space-time white noise. The main aim here corresponds to discretise the equation with respect to space and then study the convergence of the discretised process to the real solution. In the very first part of the dissertation, we introduce the subject of study, give the main mathematical motivations and summarise the goals that we have been able to attain. For this, as a preliminary part, we give the main definitions and state the main results concerning the theory of stochastic partial differential equations driven by Gaussian noises. We give also the main definitions and state the main criteria concerning the stochastic calculus of variations or Malliavin calculus. After a summary of their contents, the main results of the dissertation are included in several appendices. Indeed, the first work is devoted to the existence of density for the solution to a three-dimensional stochastic wave equation driven by a spatially homogeneous Gaussian noise. The main techniques used to prove this result are given by the Malliavin calculus' theory. Moreover, in order to give sense to the evolution equations satisfied by the Malliavin derivatives, we extend the theory of integration with respect to martingale measures to a Hilbert-valued setting. On the other hand, the main difficulty with respect to the studied cases, where the space dimension is one or two, is the fact that in the three-dimensional case the fundamental solution of the wave equation is no more a function but a distribution. The second work extends the results of the first one in the sense that we prove that the density of the solution at any fixed point not only exists but also is a smooth function. For this, again the techniques of the Malliavin calculus are applied, but with much more effort. In the framework of existence and smoothness of densities of solutions to stochastic partial differential equations, we have also devoted a small part of the thesis in extending some of the known results for the stochastic heat equation to general equations of parabolic type. We jump now to the third and last work that forms the body of the dissertation. Namely, we consider discretisation schemes of a stochastic Dirichlet problem given by a stochastic wave equation in spatial dimension one and driven by the space-time white noise. More precisely, the equation is discretised by means of a finite difference method in space and the random perturbation is formally discretised using an Euler scheme. Then, the main idea is to find out an evolution equation satisfied by the approximation process so as to be able to deal with mean and almost sure convergence to the real solution. Furthermore, we get suitable bounds for the rate of convergence that are tested numerically to be optimal. Eventually, the dissertation concludes with a summary of the contents in Catalan and the bibliography.