Numerical simulation of high frequency waves in highly heterogeneous media is a challenging problem. Resolving the fine structure of the wave field typically requires extremely small time steps and spatial meshes. We show that capturing macroscopic quantities of the wave field, such as the wave energy density, is achievable with much coarser discretizations. We obtain such a result using a time splitting algorithm that solves separately and successively propagation and scattering in the simplified regime of the parabolic wave equation in a random medium. The mathematical theory of the convergence and statistical properties of the algorithm is based on the analysis of the Wigner transforms in random media. Our results provide a step toward understanding time and space discretizations that are needed in order for the numerical algorithm to capture the correct macroscopic statistics of the wave energy density in a random medium.

We analyze residual and hierarchicala posteriori error estimates for nonconforming finite elementapproximations of elliptic problems with variable coefficients.We consider a finite volume box scheme equivalent toa nonconforming mixed finite element method in a Petrov–Galerkinsetting. We prove thatall the estimators yield global upper and local lower bounds for the discretizationerror. Finally, we present results illustrating the efficiency of theestimators, for instance, in the simulation of Darcy flows throughheterogeneous porous media.

We present a hybrid finite-volume-particle numerical method for computing the transport of a passive pollutant by a flow. The flow is modeled by the one- and two-dimensional Saint-Venant system of shallow water equations and the pollutant propagation is described by a transport equation.This paper is an extension of our previous work [Chertock, Kurganov and Petrova, J. Sci. Comput. (to appear)], where the one-dimensional finite-volume-particle method has been proposed.The core idea behind the finite-volume-particle method is to use different schemes for the flow and pollution computations: the shallow water equations are numerically integrated using a finite-volume scheme, while the transport equation is solved by a particle method. This way the specific advantages of each scheme are utilizedat the right place. A special attention is given to the recovery of the point values of the numerical solution from its particle distribution. The reconstruction is obtained using a dual equation for the pollutant concentration.This results in a significantly enhanced resolution of the computed solution and also makes it much easier to extendthe finite-volume-particle method to the two-dimensional case.

In this paper we study a model problem describing the movement ofa glacier under Glen's flow law and investigated by Colinge andRappaz [Colinge and Rappaz, ESAIM: M2AN 33 (1999) 395–406]. We establish error estimates for finiteelement approximation using the results of Chow [Chow, SIAM J. Numer. Analysis 29 (1992) 769–780] andLiu and Barrett [Liu and Barrett, SIAM J. Numer. Analysis33 (1996) 98–106] and give an analysis of theconvergence of the successive approximations used in [Colinge and Rappaz, ESAIM: M2AN 33 (1999) 395–406].Supporting numerical convergence studies are carried out and wealso demonstrate the numerical performance of an aposteriori error estimator in adaptive mesh refinementcomputation of the problem.

The incompressible Navier-Stokes problem is discretized in time by the two-step backward differentiation formula. Error estimates are proved under feasible assumptions on the regularity of the exact solution avoiding hardly fulfillable compatibility conditions. Whereas the time-weighted velocity error is of optimal second order, the time-weighted error in the pressure is of first order. Suboptimal estimates are shown for a linearisation. The results cover both the two- and three-dimensional case.

A linearly convergent iterative algorithm that approximates therank-1 convex envelope  $f^{rc}$ of a given function $f:\mathbb{R}^{n\times m} \to \mathbb{R}$ , i.e. the largest function below f which is convex along all rank-1 lines, isestablished. The proposed algorithm is a modified version of an approximationscheme due to Dolzmann and Walkington.

In this paper we present two-level overlapping domain decomposition preconditioners for the finite-element discretisation of elliptic problems in two and three dimensions. The computational domain is partitioned into overlapping subdomains, and a coarse space correction is added. We present an algebraic way to define the coarse space, based on the concept of aggregation. This employs a (smoothed) aggregation technique and does not require the introduction of a coarse grid. We consider a set of assumptions on the coarse basis functions, to ensure bound for the resulting preconditioned system. These assumptions only involve geometrical quantities associated to the aggregates, namely their diameter and the overlap. A condition number which depends on the product of the relative overlap among the subdomains and the relative overlap among the aggregates is proved. Numerical experiments on a model problem are reported to illustrate the performance of the proposed preconditioners.

We analyze a new formulation of the Stokes equations inthree-dimensional axisymmetric geometries, relying on Fourier expansion with respect tothe angular variable: the problem for each Fourier coefficient is two-dimensional and hassix scalar unknowns, corresponding to the vector potential and the vorticity. Aspectral discretization is built on this formulation, which leads to an exactlydivergence-free discrete velocity. We prove optimal error estimates.

In the reliability theory, the availability ofa component, characterized by non constant failure and repair rates,is obtained, at a given time, thanks to the computation of the marginal distributions of asemi-Markov process. These measures are shown to satisfy classicaltransport equations, the approximation of which can be donethanks to a finite volume method.Within a uniqueness result for the continuous solution,the convergence of the numerical scheme isthen proven in the weak measure sense,and some numerical applications, which show the efficiency and theaccuracy of the method, are given.

In this paper, we derive and analyze a Reissner-Mindlin-like model for isotropic heterogeneous linearly elastic plates. The modeling procedure is based on a Hellinger-Reissner principle, which we modify to derive consistent models. Due to the material heterogeneity, the classical polynomial profiles for the plate shear stress are replaced by more sophisticated choices, that are asymptotically correct. In the homogeneous case we recover a Reissner-Mindlin modelwith 5/6 as shear correction factor. Asymptotic expansions are used to estimate the modeling error. We remark that our derivation is not based on asymptotic arguments only. Thus, the model obtained is more sophisticated (and accurate) than simply taking the asymptotic limit of the three dimensional problem. Moreover, we do not assume periodicity of the heterogeneities.

This paper is concerned with the numerical approximations of Cauchy problems for one-dimensional nonconservative hyperbolic systems.The first goal is to introduce a general concept of well-balancingfor numerical schemes solving this kind of systems. Once this concept stated, weinvestigate the well-balance properties of numerical schemes based on thegeneralized Roe linearizations introduced by [Toumi, J. Comp. Phys.102 (1992) 360–373]. Next, this general theoryis applied to obtain well-balanced schemes for solving coupled systems of conservation laws withsource terms. Finally, we focus on applications to shallow water systems: the numericalschemes obtained and their properties are compared, in the case of one layer flows, with those introduced by[Bermúdez and Vázquez-Cendón, Comput. Fluids23 (1994) 1049–1071]; in the case of two layer flows, they arecompared with the numerical scheme presented by [Castro, Macías and Parés, ESAIM: M2AN35 (2001) 107–127].

We consider a fully practical finite element approximation of the following degenerate system $$ {\frac{\partial }{\partial t}} \rho(u)- \nabla . ( \,\alpha(u) \,\nabla u ) \ni \sigma(u)\,|\nabla\phi|^2 ,\quad \nabla . (\, \sigma(u) \,\nabla \phi ) = 0$$ subject to an initial condition on the temperature, u,and boundary conditions on both u and the electric potential, ϕ. In the above p(u) is the enthalpy incorporating the latent heat of melting, α(u) > 0 is the temperature dependent heat conductivity, and σ(u) > 0is the electricalconductivity. The latter is zero in the frozen zone, u ≤ 0,which gives rise to the degeneracy in this Stefan system.In addition to showing stability bounds, we prove (subsequence) convergence of our finite element approximation intwo and three space dimensions. The latter is non-trivial due to the degeneracy in σ(u)and the quadratic nature of the Joule heating term forcing the Stefan problem.Finally, some numerical experiments are presented in two space dimensions.

This paper provides new results of consistence and convergence of the lumped parameters (ODE models) toward one-dimensional (hyperbolic or parabolic) models for blood flow. Indeed,lumped parameter models (exploiting the electric circuit analogy for the circulatory system) are shown to discretize continuous 1D models at first order in space. We derive the complete set of equations useful for the blood flow networks, new schemes for electric circuit analogy, the stability criteria that guarantee the convergence, and the energy estimates of the limit 1D equations.

In the present work, the symmetrized sequential-parallel decomposition method with the fourth order accuracy for the solution of Cauchy abstract problem with an operator under a split form is presented. The fourth order accuracy is reached by introducing a complex coefficient with the positive real part. For the considered scheme, the explicit a priori estimate is obtained.

A coupled finite/boundary element method to approximate the free vibration modes of an elastic structure containing an incompressible fluid is analyzed in this paper. The effect of the fluid is taken into account by means of one of the most usual procedures in engineering practice: an added mass formulation, which is posed in terms of boundary integral equations. Piecewise linear continuous elements are used to discretize the solid displacements and the fluid-solid interface variables. Spectral convergence is proved and error estimates are settled for the approximate eigenfunctions and their corresponding vibration frequencies. Implementation issues are also discussed and numerical experiments are reported.

This paper analyses the implementation of the generalizedfinite differences method for the HJB equation of stochastic control, introduced by two of the authors in [Bonnans and Zidani,SIAM J. Numer. Anal.41 (2003) 1008–1021]. The computation of coefficients needs tosolve at each point of the grid (and for each control)a linear programming problem.We show here that, for two dimensional problems, this linear programming problem can be solved in O(p max) operations, where p max is the size of the stencil. The method is based on a walk on the Stern-Brocot tree,and on the related filling of the set of positive semidefinite matrices of size two.