We propose a finite volume scheme to discretize the one-dimensional Vlasov-Poisson system. We prove that, if the initial data are positive, bounded, continuous, and have their first moment bounded, then the numerical approximation converges to the weak solution of the system for the weak topology of L ∞ . Moreover, if the initial data belong to BV , the convergence is strong in C 0 (0, T ; L 1 loc ). To prove the convergence of the discrete electric field, we obtain an estimation in W 1,∞ (Ω T ). Then we havewhere (E, f ) is the unique weak solution of the Vlasov-Poisson system.
Introduction. The Vlasov-Poisson system is a model for the motion of a collisionless plasma of electrons in a uniform background of ions and describes the evolution of the distribution function of electrons (solution of the Vlasov equation)under the effects of the transport and self-consistent electric field (solution of the Poisson equation). The coupling between both equations gives a nonlinear problem.The numerical resolution of the Vlasov equation is most often performed using particle methods (PIC), which consist of approximating the plasma by a finite number of particles. The trajectories of these particles are computed from the characteristic curves given by the Vlasov equation. The interactions with self-consistent and external fields are computed by a numerical method using a mesh of the physical space (see, e.g., Birdsall and Langdon [2] or Cottet and Raviart [5]). This method enables us to get satisfying results with few particles.Methods relying on a discretization of the phase space have also been proposed (see, e.g., Shoucri and Knorr [15] and Klimas and Farrell [11]) and seem to be more efficient in some cases, for example, when particles in the tail of the distribution play an important physical role, or when the numerical noise due to the finite number of particles becomes too important. Among them, the semi-Lagrangian method [16] consists of directly computing the distribution function on a grid of the phase space. This computation is done by following the characteristic curves at each time step and interpolating the value at the origin of the characteristics by a cubic spline method.This interpolation method works well for simple geometries of the physical space but does not seem to be well suited to more complex geometries.To remedy this problem a possible approach is to use the finite volume method which is known to be a robust and computationally cheap method for the discretization of conservation laws (see, e.g., Eymard, Gallouet, and Herbin [9] and the references