A Godunov scheme is derived for two-dimensional scalar conservation laws without or with source terms following ideas originally proposed by Boukadida and LeRoux [Math. Comput., 63 (1994), pp. 541-553] in the context of a staggered Lax-Friedrichs scheme. In both situations, the numerical fluxes are obtained at each interface of a uniform Cartesian computational grid just by means of the "external waves" involved in the entropy solution of the elementary two-dimensional (2D) Riemann problems; in particular, all the wave-interaction phenomena are overlooked. This restriction of the wave pattern suffices for deriving the exact numerical fluxes of the staggered Lax-Friedrichs scheme, but it furnishes only an approximation for the Godunov scheme: we show that under convenient assumptions, these flux functions are smooth and the resulting discretization process is stable under nearly optimal CFL restriction. A well-balanced extension is presented, relying on the Curl-free component of the Helmholtz decomposition of the source term. Several numerical tests against exact 2D solutions are performed for convex, nonconvex, and inhomogeneous equations and the time-evolution of the L1 truncation error is displayed.

When numerically simulating a kinetic model of an n+nn+ semiconductor device, obtaining a constant macroscopic current at steady state is still a challenging task. Part of the difficulty comes from the multiscale, discontinuous nature of both p|n junctions, which create spikes in the electric field and enclose a channel where corresponding depletion layers glue together. The kinetic formalism furnishes a model holding inside the whole domain, but at the price of strongly varying parameters. By concentrating both the electric acceleration and the linear collision terms at each interface of a Cartesian computational grid, we can treat them by means of a Godunov scheme involving two types of scattering matrices. Combining both these mechanisms into a global Smatrix can be achieved thanks to "Redheffer's star-product." Assuming that the resulting S-matrix is stochastic permits us to prove maximum principles under a mild CFL restriction. Numerical illustrations of collisional Landau damping and various n+nn+ devices are provided on coarse grids.

The numerical approximation of one-dimensional relativistic Dirac wave equations is considered within the recent framework consisting in deriving local scattering matrices at each interface of the uniform Cartesian computational grid. For a Courant number equal to unity, it is rigorously shown that such a discretization preserves exactly the (Formula presented.) norm despite being explicit in time. This construction is well-suited for particles for which the reference velocity is of the order of (Formula presented.), the speed of light. Moreover, when (Formula presented.) diverges, that is to say, for slow particles (the characteristic scale of the motion is non-relativistic), Dirac equations are naturally written so as to let a "diffusive limit" emerge numerically, like for discrete 2-velocity kinetic models. It is shown that an asymptotic-preserving scheme can be deduced from the aforementioned well-balanced one, with the following properties: it yields unconditionally a classical Schrödinger equation for free particles, but it handles the more intricate case with an external potential only conditionally (the grid should be such that (Formula presented.)). Such a stringent restriction on the computational grid can be circumvented easily in order to derive a seemingly original Schrödinger scheme still containing tiny relativistic features. Numerical tests (on both linear and nonlinear equations) are displayed. © 2014 Springer Science+Business Media Dordrecht.

Numerical approximation of one-dimensional kinetic models for directed motion of bacterial populations in response to a chemical gradient, usually called {\it chemotaxis}, is considered in the framework of well-balanced (WB) schemes. The validity of one-dimensional models have been shown to be relevant for the simulation of more general situations with symmetry in all but one direction along which appears the chemical attractant gradient. Two main categories are considered depending on whether or not the kinetic equation with specular boundary conditions admits non-constant macroscopic densities for large times. The WB schemes are endowed with the property of having zero artificial viscosity at steady-state; in particular they furnish numerical solutions for which the macroscopic flux vanishes, a feature that more conventional discretizations can miss. A class of equations which admit constant asymptotic states can be treated by a slight variation of the method of Case's elementary solutions originally developed for radiative transfer problems. More involved models which can display concentrations are handled through a different, but closely related, treatment of the tumbling term at the computational grid's interfaces. Both types of WB algorithms can be implemented efficiently relying on the Sherman-Morrison formula for computing interface values. Transient and stationary numerical results are displayed for several test-cases.

Efficient recovery of smooth functions which are s-sparse with respect to the basis of so-called prolate spheroidal wave functions from a small number of random sampling points is considered. The main ingredient in the design of both the algorithms we propose here consists in establishing a uniform L? bound on the measurement ensembles which constitute the columns of the sensingmatrix. Such a bound provides us with the restricted isometry property for this rectangular random matrix, which leads to either the exact recovery property or the "best s-term approximation" of the original signal by means of the 1 minimization program. The first algorithm considers only a restricted number of columns for which the L? holds as a consequence of the fact that eigenvalues of the Bergman's restriction operator are close to 1 whereas the second one allows for a wider system of PSWF by taking advantage of a preconditioning technique. Numerical examples are spread throughout the text to illustrate the results.

The ability of Well-Balanced (WB) schemes to capture very accurately steady-state regimes of non-resonant hyperbolic systems of balance laws has been thoroughly illustrated since its introduction by Greenberg and LeRoux (1996) [15] (see also the anterior WB Glimm scheme in E, 1992 [8]). This paper aims at showing, by means of rigorous C0 t (L1x ) estimates, that these schemes deliver an increased accuracy in transient regimes too. Namely, after explaining that for the vast majority of non-resonant scalar balance laws, the C0 t (L1x ) error of conventional fractional-step (Tang and Teng, 1995 [45]) numerical approximations grows exponentially in time like exp(max(g )t) ? x (as a consequence of the use of Gronwall's lemma), it is shown that WB schemes involving an exact Riemann solver suffer from a much smaller error amplification: thanks to strict hyperbolicity, their error grows at most only linearly in time (see also Layton, 1984 [30]). Numerical results on several testcases of increasing difficulty (including the classical LeVeque-Yee's benchmark problem (LeVeque and Yee, 1990 [34]) in the non-stiff case) confirm the analysis.

We derive a sufficient condition by means of which one can recover a scale-limited signal from the knowledge of a truncated version of it in a stable manner following the canvas introduced by Donoho and Stark (1989) [4]. The proof follows from simple computations involving the Zak transform, well-known in solid-state physics. Geometric harmonics (in the terminology of Coifman and Lafon (2006) [22]) for scale-limited subspaces of L2(R) are also displayed for several test-cases. Finally, some algorithms are studied for the treatment of zero-angle problems.

An original well-balanced (WB) Godunov scheme relying on an exact Riemann solver involving a non-conservative (NC) product is developed. It is meant to solve accurately the time-dependent one-dimensional radiative transfer equation in the discrete ordinates approximation with an arbitrary even number of velocities. The collision term is thus concentrated onto a discrete lattice by means of Dirac masses; this induces steady contact discontinuities which are integral curves of the stationary problem. One solves it by taking advantage of the method of elementary solutions mainly developed by Case, Zweifel and Cercignani. This approach produces a rather simple scheme that compares advantageously to standard existing upwind schemes, especially for the decay in time toward a Maxwellian distribution. It is possible to reformulate this scheme in order to handle properly the parabolic scaling in order to generate a so-called asymptotic-preserving (AP) discretization. Consistency with the diffusive approximation holds independently of the computational grid. Several numerical results are displayed to show the realizability and the efficiency of the method.

We consider the problem of short-time extrapolation of blue chips' stocks indexes in the context of wavelet subspaces following the theory proposed by X.-G. Xia and co-workers in a series of papers \cite{XLK,XKZ,LK,LXK}. The idea is first to approximate the oscillations of the corresponding stock index at some scale by means of the scaling function which is part of a given multi-resolution analysis of $L^2(\Re)$. Then, since oscillations at a finer scale are discarded, it becomes possible to extend such a signal up to a certain time in the future; the finer the approximation, the shorter this extrapolation interval. At the numerical level, a so--called Generalized Gerchberg-Papoulis (GGP) algorithm is set up which is shown to converge toward the minimum $L^2$ norm solution of the extrapolation problem. When it comes to implementation, an acceleration by means of a Conjugate Gradient (CG) routine is necessary in order to obtain quickly a satisfying accuracy. Several examples are investigated with different international stock market indexes.

We consider a rather simple algorithm to address the fascinating field of numerical extrapolation of (analytic) band-limited functions. It relies on two main elements: namely, the lower frequencies are treated by projecting the known part of the signal to be extended onto the space generated by ``Prolate Spheroidal Wave Functions" (PSWF, as originally proposed by Slepian), whereas the higher ones can be handled by the recent so--called ``Compressive Sampling" (CS, proposed by Cand\`es) algorithms which are independent of the largeness of the bandwidth. Slepian functions are recalled and their numerical computation is explained in full detail whereas $\ell^1$ regularization techniques are summarized together with a recent iterative algorithm which has been proved to work efficiently on so--called ``compressible signals" which appear to match rather well the class of smooth bandlimited functions. Numerical results are displayed for both numerical techniques and the accuracy of the process consisting in putting them altogether is studied for some test-signals showing a quite fast Fourier decay.

In this paper we present algorithms for approximating real band-limited signals by multiple Gaussian Chirps. These algorithms do not rely on matching pursuit ideas. They are hierarchial and, at each stage, the number of terms in a given approximation depends only on the number of positive-valued maxima and negative-valued minima of a signed amplitude function characterizing part of the signal. Like the algorithms used in \cite{gre2} and unlike previous methods, our chirplet approximations require neither a complete dictionary of chirps nor complicated multi-dimensional searches to obtain suitable choices of chirp parameters.

We consider a class of finite Markov moment problems with an arbitrary number of positive and negative branches. We show criteria for the existence and uniqueness of solutions, and we characterize in detail the nonunique solution families. Moreover, we present a constructive algorithm to solve the moment problems numerically and prove that the algorithm computes the right solution.

This paper investigates a simple one-dimensional model of incommensurate “harmonic crystal” in terms of the spectrum of the corresponding Schrödinger equation. Two angles of attack are studied: the first exploits techniques borrowed from the theory of quasi-periodic functions while the second relies on periodicity properties in a higher-dimensional space. It is shown that both approaches lead to essentially the same results; that is, the lower spectrum is split between “Cantor-like zones” and “impurity bands” to which correspond critical and extended eigenstates, respectively. These “new bands” seem to emerge inside the band gaps of the unperturbed problem when certain conditions are met and display a parabolic nature. Numerical tests are extensively performed on both steady and time-dependent problems.

We are concerned with the numerical study of a simple one-dimensional Schr\"odinger operator $-\frac 1 2 \Dxx + \alpha q(x)$ with $\alpha \in \Re$, $q(x)=\cos(x)+\eps \cos(kx)$, $\eps >0$ and $k$ being irrational. This governs the quantum wave function of an independent electron within a crystalline lattice perturbed by some impurities whose dissemination induces long-range order only, which is rendered by means of the quasi-periodic potential $q$. We study numerically what happens for various values of $k$ and $\eps$; it turns out that for $k > 1$ and $\eps\ll 1$, that is to say, in case more than one impurity shows up inside an elementary cell of the original lattice, ``impurity bands" appear and seem to be $k$-periodic. When $\eps$ grows bigger than one, the opposite case occurs.

We consider the approximation of a microelectronic device corresponding to a $n^+-n-n^+$ diode consisting in a channel flanked on both sides by two highly doped regions. This is modelled through a system of equations: ballistic for the channel and drift-diffusion elsewhere. The overall coupling stems from the Poisson equation for the self-consistent potential. We propose an original numerical method for its processing, being realizable, explicit in time and nonnegativity preserving on the density. In particular, the boundary conditions at the junctions express the continuity of the current and don't destabilize the general scheme. At last, efficiency is shown by presenting results on test-cases of some practical interest.

The aim of this text is to develop on the asymptotics of some 1-D nonlinear Schr\"odinger equations from both the theoretical and the numerical perspectives, when a caustic is formed. We review rigorous results in the field and give some heuristics in cases where justification is still needed. The scattering operator theory is recalled. Numerical experiments are carried out on the focus point singularity for which several results have been proven rigorously. Furthermore, the scattering operator is numerically studied. Finally, experiments on the cusp caustic are displayed, and similarities with the focus point are discussed. Several shortcomings of spectral time-splitting schemes are investigated.