Laboratoire de Mathématiques de Besançon - UMR 6623 CNRS
UFC
CNRS


Accueil > Activités > Séminaires > Analyse Numérique et Calcul Scientifique

Séminaire d’Analyse Numérique et Calcul Scientifique

par alozinski - publié le , mis à jour le

Le séminaire a lieu le jeudi, à 11h, en salle 316 du bâtiment de
Métrologie (plan d’accès).

Vous trouverez ci-dessous le planning du séminaire
d’Analyse Numérique et Calcul Scientifique pour l’année universitaire en cours.

Pour contacter le responsable (Alexei Lozinski) :
alexei.lozinski@univ-fcomte.fr.

Exposés à venir :

  • Jeudi 9 novembre 2017, à 14h en salle 324-B1  : Christian Klein (Institut de Mathématiques de Bourgogne)
    Multi-domain spectral methods for Green’s functions for the Maxwell equations in matter

    We present a multi-domain spectral approach for the Maxwell equations in matter with a Sommerfeld radiation condition at infinity. The situation to be studied is a conductor in a matter distribution. We concentrate here on an axisymmetric situation where the boundaries of conductor and matter are formed by spherical shells. For the time dependence, a mode analysis is used, i.e., after a Fourier transformation in time the frequency is treated as a parameter in the equations. The Maxwell equations for this situation are formulated in spherical coordinates, the matter is characterized by a diecletric function.
    In the numerical approach, the angular dependence is treated via a Chebychev collocation method in physical space. The axis is singular in this setting, thus no boundary conditions are needed there. For the radial dependence we introduce three domains, the first inside the conductor, the second between conductor and the boundary of the matter, the third in vacuum. Each of these domains in the radial coordinate is mapped to the interval [-1,1], in the last one we use $1/\rho$ as a coordinate thus compactifying the outer domain. This compactification allows the treatment of infinity as a point on the grid. In each domain we use a Chebychev collocation method in coefficient space. The Sommerfeld condition is imposed by writing the fields as an outgoing wave times a function and solving only for the latter. This allows for an exact implementation of the Sommerfeld condition at infinity which is a singular point of the equation. At the domain boundaries, the usual boundary conditions for the Maxwell equations are imposed via a tau method.

  • Jeudi 19 octobre 2017, à 11h : Samuel Dubuis (EPFL)
    Reporté sine die
    An adaptive algorithm for the time dependent transport equation with anisotropic finite elements and the Crank-Nicolson scheme

    The time dependent transport equation is solved with stabilized continuous, piecewise linear finite elements and the Crank-Nicolson scheme [1]. Finite elements with large aspect ratio are advocated in order to account for boundary layers. The error due to space discretization has already been studied in [2]. Here, the error due to the use of the Crank-Nicolson scheme is taken into account. Anisotropic a priori and a posteriori error estimates are proved. The a posteriori upper bound is obtained using a quadratic reconstruction in time as in [3].
    The quality of the error estimator is first validated on non adapted meshes and constant time steps. An adaptive algorithm in space and time is then proposed, with goal to build a sequence of anisotropic meshes and time steps, so that the final error is close to a preset tolerance. Numerical results on adapted, anisotropic meshes and
    time steps show the efficiency of the method.

    [1] E. Burman, Consistent SUPG-method for transient transport problems : Stability
    and convergences, Comput. Methods Appl. Mech. Engrg., 199 (2010).
    [2] Y. Bourgault and M. Picasso, Anisotropic error estimates and space adaptivity for a semidiscrete finite element approximation of the transient transport equation SIAM J. Sci. Comput., 35 (2013).
    [3] A. Lozinski, M.Picasso and V. Prachittham, An anisotropic error estimator for the Crank-Nicolson method : application to a parabolic problem, SIAM J. Sci. Comput., 31 (2009).

Exposés passés :

Agenda

iCal