SIAM-SEAS 2005 Minisymposium MSDG
Discontinuous Galerkin Methods

Friday, 2:15-4:15, Saturday 8:45-10:15 and 2:00-3:30

Organized by Susanne Brenner brenner@math.sc.edu and Fengyan Li fli@math.sc.edu, University of South Carolina

The discontinuous Galerkin method is a class of finite element methods which use completely discontinuous functions as the numerical solutions and test functions. Stable and convergent discontinuous Galerkin methods have been designed in recent years for various types of partial differential equations, including hyperbolic, parabolic, elliptic equations and equations containing higher order spatial derivatives such as the KdV equations.

The flexibility in the choice of the local spaces and the minimum requirement on the triangulation of the domain make the method especially suitable for adaptivity. The method also has excellent parallel efficiency. Discontinuous Galerkin methods have found rapid applications recently, in such diverse areas as aeroacoustics, electromagnetism, gas dynamics, granular flows, meteorology, oceanography, semiconductor device simulation, turbulent flow, weather forecasting, among many others.

Speakers are named first.

    Friday afternoon 2:15-4:15 (session MSDG1)

  1. 2:15-2:40 Slimane Adjerid, Virginia Tech.
    Supeconvergence and error estimation for the discontinuous finite element method

    The discontinuous Galerkin method (DGM) is an appealing approach to address problems having discontinuities, such as those that arise in hyperbolic conservation laws. The DGM uses a discontinuous finite element basis which simplifies hp adaptivity and leads to a simple communication pattern across faces that makes it useful for parallel computation. In order for the DGM to be useful in an adaptive setting, techniques for estimating the discretization errors should be available both to guide adaptive enrichment and to provide a stopping criteria for the solution process. We will show that the p-degree DG finite element solution for hyperbolic problems exhibits superconvergence at the roots of Radau polynomials of degree p + 1 with the fixed endpoints selected at the downwind boundary of each quadrilateral element. We also show that the DG solution has strong superconvergence on average at the outflow boundary. We discuss an extension of these results to locally discontinuous finite element solutions of convection-diffusion problems. We use our superconvergence results to construct asymptotically exact a posteriori error estimates for first-order hyperbolic and convection-diffusion problems. Finally, we present numerical results for several computational examples with both continuous and discontinuous solutions that validate our theory.

  2. 2:45-3:10 Shanqin Chen, Weinan E and Chi-Wang Shu, Brown University
    The Heterogeneous Multi-Scale Method Based on the Discontinuous Galerkin Method for Hyperbolic and Parabolic Problems

    In this paper we develop a discontinuous Galerkin (DG) method, within the framework of the heterogeneous multi-scale method (HMM), for solving hyperbolic and parabolic multi-scale problems. Hyperbolic scalar equations and systems, and parabolic scalar problems are considered. Error estimates are given for the linear equations and numerical results are provided for the linear and nonlinear problems to demonstrate the capability of the method.

  3. 3:15-3:40 Katarina Jegdic, Department of Mathematics, University of Houston
    Convergence of a spacetime discontinuous Galerkin method for Temple class systems

    We consider a spacetime discontinuous Galerkin method for one-dimensional systems of hyperbolic conservation laws. We assume that the partitions of the spacetime domain satisfy a causality constraint and that the Galerkin basis consists of piecewise constant functions. In the case of genuinely nonlinear Temple class systems (such as equations of multicomponent chromatography), we show that given a causal spacetime mesh, a unique approximation satisfying local Riemann invariant bounds exist. This enables us to prove convergence of the method to a weak solution.

  4. 3:45-4:10 Jennifer Ryan, Computer Science and Mathematics Division, Oak Ridge National Lab and Chi-Wang Shu, Division of Applied Mathematics, Brown University.
    Post-processing of the discontinuous Galerkin methods for nonuniform mesh

    A post-processing technique based on negative order norm estimates for the discontinuous Galerkin method was previously introduced Cockburn, Luskin, Shu, and Süli. The postprocessor allows improvement in accuracy of the discontinuous Galerkin method for time-dependent linear hyperbolic equations from order k+1 to order 2k+1 for a uniform mesh. This improvement in accuracy was extended to include superconvergence of the derivatives, two space dimensions, multi-domains with different mesh sizes, and variable and discontinuous coefficient hyperbolic equations. The uniform mesh assumption gives the convolution kernel a local form and thus allows for simple implementation using small matrix-vector multiplications. In this talk, we present an extension of this postprocessing technique to include smoothly varying and nonuniform meshes.

  5. Saturday morning 8:45-10:15 (session MSDG2)

  6. 8:45- 9:10 Guido Kanschat, Universitat Heidelberg
    Multilevel discontinuous Galerkin methods on locally refined meshes

    On hierarchical meshes generated by local refinement of a coarse mesh, a multilevel method must be devised carefully in order to maintain optimal complexity. Here, we show that the scheme involving local smoothing is particularly suited for discontinuous Galerkin discretizations of elliptic operators. While the efficiency of the method exhibits independence of the refinement pattern of the grid, the method can still be implemented in a general way. Results will be shown for the interior penalty method as well as for an extension to the LDG method.

  7. 9:15-9:40 Ohannes Karakashian, University of Tennessee
    A posteriori error estimates and convergent adaptive algorithms for a discontinuous Galerkin method

    We present some aposteriori error estimates for an interior penalty discontinuous Galerkin approximation of second order elliptic equations with general boundary conditions. We also present an adaptive refinement/coarsening procedure which is capable of achieving any finite error tolerance in a finite number of steps. While this number depends on the tolerance, we have shown that under certain conditions there is a guaranteed convergence rate.

  8. 9:45-10:10 Kening Wang, University of South Carolina
    Domain Decomposition Preconditioners for C0 Interior Penalty Methods

    We study both two-level additive Schwarz preconditioners and the Bramble-Pasciak-Schatz (BPS) preconditioner that can be used in the iterative solution of the discrete problems resulting from C0 interior penalty methods for fourth order elliptic boundary value problems.

  9. Saturday afternoon 2:00-3:30 (session MSDG3)

  10. 2:00-2:15 Bernardo Cockburn (University of Minnesota), Dominik Schoetzau, (British Columbia) and Guido Kanschat (Heidelberg)
    DG methods for the Navier-Stokes equations

    We uncover the strage fact that DG methods for the Navier-Stokes equations cannot be locally conservative and energy-stable at the same time. We then propose three new ways of dealing with such a problem. Two of them consist in working with exactly divergence-free velocities: In the first, the exactly divergence-free velocity is obtained by a simple postprocessing; in the second, we work directly with divergence-free velocity spaces but show how to avoid having to actually contruct them. The third approach consists in introducing a new pressure that renders the hyperbolic system associated with the Navier-Stokes equations symmetric.

  11. 2:30-2:55 Mark Sussman, Florida State University
    A discontinuous spectral element method for the level set equation

    Level set methodology is crucially pertinent to tracking moving singular surfaces or thin fronts with steep gradients in the numerical solutions of partial differential equations governing complex flow fields. This methodology must be consistent with the basic solution technique for the partial differential equations. To this end, a discontinuous spectral element approach is developed for level set advection and reinitialization as these methods are becoming increasingly popular for the solution of the fluid dynamic problems. Example computations are provided, which demonstrate the high order accuracy of the method.

  12. 3:00-3:25 Ramachandran Nair, NCAR
    A Discontinuous Galerkin Atmospheric Shallow Water Model

    Correction: wrong abstract posted previously. We consider the development of a shallow water model on the "cubed-sphere" based on discontinuous Galerkin method. The continuous flux-form shallow equations in non-orthogonal curvilinear coordinates are employed. The high-order spatial discretization employs a modal or nodal basis set consisting of Legendre polynomials. Time integration of the model is performed by an explicit Runge-Kutta scheme. The accuracy of the model is evaluated using the standard test problems suggested by Williamson et al. (JCP, 1992). Some of the test results are compared with a spectral element model and other global shallow water models.