Unified Hybridization Of Discontinuous Galerkin, Mixed, And Continuous Galerkin Methods For Second Order Elliptic Problems.

Similar documents
Hybridized DG methods

Hybridized Discontinuous Galerkin Methods

Local discontinuous Galerkin methods for elliptic problems

A note on discontinuous Galerkin divergence-free solutions of the Navier-Stokes equations

Analysis of Hybrid Discontinuous Galerkin Methods for Incompressible Flow Problems

A Mixed Nonconforming Finite Element for Linear Elasticity

ENERGY NORM A POSTERIORI ERROR ESTIMATES FOR MIXED FINITE ELEMENT METHODS

Discontinuous Galerkin Methods

Static condensation, hybridization, and the devising of the HDG methods

arxiv: v1 [math.na] 19 Dec 2018

Discontinuous Galerkin Methods: Theory, Computation and Applications

New class of finite element methods: weak Galerkin methods

WEAK GALERKIN FINITE ELEMENT METHODS ON POLYTOPAL MESHES

ENERGY NORM A POSTERIORI ERROR ESTIMATES FOR MIXED FINITE ELEMENT METHODS

ETNA Kent State University

To CG or to HDG: A Comparative Study

Fourier Type Error Analysis of the Direct Discontinuous Galerkin Method and Its Variations for Diffusion Equations

b i (x) u + c(x)u = f in Ω,

arxiv: v1 [math.na] 15 Nov 2017

On an Approximation Result for Piecewise Polynomial Functions. O. Karakashian

Multilevel Preconditioning of Graph-Laplacians: Polynomial Approximation of the Pivot Blocks Inverses

A PROJECTION-BASED ERROR ANALYSIS OF HDG METHODS BERNARDO COCKBURN, JAYADEEP GOPALAKRISHNAN, AND FRANCISCO-JAVIER SAYAS

Performance comparison between hybridizable DG and classical DG methods for elastic waves simulation in harmonic domain

Hybrid Discontinuous Galerkin methods for incompressible flow problems

Lecture Note III: Least-Squares Method

arxiv: v1 [math.na] 14 Apr 2011

AN EQUILIBRATED A POSTERIORI ERROR ESTIMATOR FOR THE INTERIOR PENALTY DISCONTINUOUS GALERKIN METHOD

Motivations. Outline. Finite element exterior calculus and the geometrical basis of numerical stability. References. Douglas N.

SECOND ORDER TIME DISCONTINUOUS GALERKIN METHOD FOR NONLINEAR CONVECTION-DIFFUSION PROBLEMS

An Implementation of Hybrid Discontinuous Galerkin Methods in DUNE

Weak Galerkin Finite Element Scheme and Its Applications

An Equal-order DG Method for the Incompressible Navier-Stokes Equations

MIXED FINITE ELEMENT METHODS FOR PROBLEMS WITH ROBIN BOUNDARY CONDITIONS

Mixed Finite Element Methods. Douglas N. Arnold, University of Minnesota The 41st Woudschoten Conference 5 October 2016

A note on accurate and efficient higher order Galerkin time stepping schemes for the nonstationary Stokes equations

Discontinuous Galerkin method for a class of elliptic multi-scale problems

Superconvergence and H(div) Projection for Discontinuous Galerkin Methods

MATH 676. Finite element methods in scientific computing

Finite element exterior calculus: A new approach to the stability of finite elements

ANALYSIS OF AN INTERFACE STABILISED FINITE ELEMENT METHOD: THE ADVECTION-DIFFUSION-REACTION EQUATION

Discontinuous Galerkin and Finite Difference Methods for the Acoustic Equations with Smooth Coefficients. Mario Bencomo TRIP Review Meeting 2013

The Mortar Boundary Element Method

A super convergent hybridisable discontinuous Galerkin method for linear elasticity

arxiv: v2 [math.na] 23 Apr 2016

THE MULTISCALE DISCONTINUOUS GALERKIN METHOD FOR SOLVING A CLASS OF SECOND ORDER ELLIPTIC PROBLEMS WITH ROUGH COEFFICIENTS

Discontinuous Galerkin Methods: An Overview and Some Applications. Daya Reddy UNIVERSITY OF CAPE TOWN

A SIMPLE TUTORIAL ON DISCONTINUOUS GALERKIN METHODS. Jennifer Proft CERMICS, ENPC. J. Proft CERMICS, ENPC

STABILIZED DISCONTINUOUS FINITE ELEMENT APPROXIMATIONS FOR STOKES EQUATIONS

An Iterative Substructuring Method for Mortar Nonconforming Discretization of a Fourth-Order Elliptic Problem in two dimensions

Bubble stabilized discontinuous Galerkin methods on conforming and non-conforming meshes

UNIFIED A POSTERIORI ERROR ESTIMATOR FOR FINITE ELEMENT METHODS FOR THE STOKES EQUATIONS

Energy norm a-posteriori error estimation for divergence-free discontinuous Galerkin approximations of the Navier-Stokes equations

AN ANALYSIS OF THE EMBEDDED DISCONTINUOUS GALERKIN METHOD FOR SECOND ORDER ELLIPTIC PROBLEMS

Error analysis of piecewise constant approximations of Darcy s law

STABILIZATION IN RELATION TO WAVENUMBER IN HDG METHODS

A primer on Numerical methods for elasticity

Unified A Posteriori Error Control for all Nonstandard Finite Elements 1

Discontinuous Galerkin Methods: Theory, Computation and Applications

Discontinuous Galerkin methods for nonlinear elasticity

arxiv: v1 [math.na] 29 Feb 2016

THE COMPACT DISCONTINUOUS GALERKIN (CDG) METHOD FOR ELLIPTIC PROBLEMS

A LAGRANGE MULTIPLIER METHOD FOR ELLIPTIC INTERFACE PROBLEMS USING NON-MATCHING MESHES

ASM-BDDC Preconditioners with variable polynomial degree for CG- and DG-SEM

IMA Preprint Series # 2146

An a posteriori error estimate and a Comparison Theorem for the nonconforming P 1 element

AMS subject classifications. Primary, 65N15, 65N30, 76D07; Secondary, 35B45, 35J50

Two-scale Dirichlet-Neumann preconditioners for boundary refinements

Yongdeok Kim and Seki Kim

1. Introduction. The Stokes problem seeks unknown functions u and p satisfying

FINITE ELEMENT APPROXIMATION OF STOKES-LIKE SYSTEMS WITH IMPLICIT CONSTITUTIVE RELATION

A SECOND ELASTICITY ELEMENT USING THE MATRIX BUBBLE WITH TIGHTENED STRESS SYMMETRY

ICES REPORT Analysis of the DPG Method for the Poisson Equation

A Hybrid Discontinuous Galerkin Method 2 for Darcy-Stokes Problems 3 UNCORRECTED PROOF

Mixed Hybrid Finite Element Method: an introduction

Mixed Discontinuous Galerkin Methods for Darcy Flow

PARTITION OF UNITY FOR THE STOKES PROBLEM ON NONMATCHING GRIDS

Outline. 1 Motivating examples. 2 Hilbert complexes and their discretization. 3 Finite element differential forms. 4 The elasticity complex 1 / 25

The CG1-DG2 method for conservation laws

BUBBLE STABILIZED DISCONTINUOUS GALERKIN METHOD FOR PARABOLIC AND ELLIPTIC PROBLEMS

General linear methods for time-dependent PDEs

Non-Conforming Finite Element Methods for Nonmatching Grids in Three Dimensions

The Discontinuous Galerkin Finite Element Method

Hybrid (DG) Methods for the Helmholtz Equation

ADAPTIVE HYBRIDIZED INTERIOR PENALTY DISCONTINUOUS GALERKIN METHODS FOR H(CURL)-ELLIPTIC PROBLEMS

A gradient recovery method based on an oblique projection and boundary modification

PRECONDITIONING OF DISCONTINUOUS GALERKIN METHODS FOR SECOND ORDER ELLIPTIC PROBLEMS. A Dissertation VESELIN ASENOV DOBREV

Chapter 5 A priori error estimates for nonconforming finite element approximations 5.1 Strang s first lemma

Multigrid Methods for Maxwell s Equations

Stochastic multiscale modeling of subsurface and surface flows. Part III: Multiscale mortar finite elements for coupled Stokes-Darcy flows

Interior superconvergence in mortar and non-mortar mixed finite element methods on non-matching grids

NONCONFORMING MIXED ELEMENTS FOR ELASTICITY

Numerical Simulation of Flows in Highly Heterogeneous Porous Media

Electronic Transactions on Numerical Analysis Volume 32, 2008

EQUADIFF 6. Jean-Claude Nédélec Mixed finite element in 3D in H(div) and H(curl) Terms of use:

C e n t r u m v o o r W i s k u n d e e n I n f o r m a t i c a

Fourier analysis for discontinuous Galerkin and related methods. Abstract

Adaptive Finite Elements with Thin Obstacles

Numerical Analysis of Higher Order Discontinuous Galerkin Finite Element Methods

A Least-Squares Finite Element Approximation for the Compressible Stokes Equations

FIRST-ORDER SYSTEM LEAST SQUARES (FOSLS) FOR PLANAR LINEAR ELASTICITY: PURE TRACTION PROBLEM

Transcription:

Unified Hybridization Of Discontinuous Galerkin, Mixed, And Continuous Galerkin Methods For Second Order Elliptic Problems. Stefan Girke WWU Münster Institut für Numerische und Angewandte Mathematik 10th of January, 2011 Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 1 / 43

Inhalt 1 introduction (an example) Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 2 / 43

Inhalt 1 introduction (an example) 2 the framework Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 2 / 43

Inhalt 1 introduction (an example) 2 the framework 3 examples for hybridizable methods Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 2 / 43

Inhalt 1 introduction (an example) 2 the framework 3 examples for hybridizable methods 4 Other novel methods hybridizable methods well suited for adaptivity The RT-method on meshes with hanging nodes Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 2 / 43

Inhalt literature D.N. Arnold, F. Brezzi, B. Cockburn, and L.D. Marini. Unified analysis of discontinuous Galerkin methods for elliptic problems. SIAM journal on numerical analysis, 39(5):1749 1779, 2002. B. Cockburn and J. Gopalakrishnan. A characterization of hybridized mixed methods for second order elliptic problems. SIAM Journal on Numerical Analysis, 42(1):283 301, 2005. B. Cockburn, J. Gopalakrishnan, and R. Lazarov. Unified hybridization of discontinuous Galerkin, mixed and continuous Galerkin methods for second order elliptic problems. SIAM J. Numer. Anal, 47(2):1319 1365, 2009. R.M. Kirby, S.J. Sherwin, and B. Cockburn. To CG or to HDG: A Comparative Study. M. Vohraĺık, J. Maryska, and O. Severýn. Mixed and nonconforming finite element methods on a system of polygons. Applied Numerical Mathematics, 57(2):176 193, 2007. OC Zienkiewicz. Displacement and equilibrium models in the finite element method by B. Fraeijs de Veubeke, Chapter 9, Pages 145-197 of Stress Analysis, Edited by OC Zienkiewicz and GS Holister, Published by John Wiley & Sons, 1965. International Journal for Numerical Methods in Engineering, 52(3):287 342, 2001. Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 3 / 43

introduction (an example) 1 introduction (an example) 2 the framework 3 examples for hybridizable methods 4 Other novel methods hybridizable methods well suited for adaptivity The RT-method on meshes with hanging nodes Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 4 / 43

introduction (an example) Second order elliptic boundary value problem Search solution u for (a u)+du = f in Ω Ê n, u = g on Ω, where a(x) is a bounded symmetric positive definite matrix-valued function, Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 5 / 43

introduction (an example) Second order elliptic boundary value problem Search solution u for (a u)+du = f in Ω Ê n, u = g on Ω, where a(x) is a bounded symmetric positive definite matrix-valued function, d(x) is a bounded nonnegative function. Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 5 / 43

introduction (an example) Second order elliptic boundary value problem Search solution u for (a u)+du = f in Ω Ê n, u = g on Ω, where a(x) is a bounded symmetric positive definite matrix-valued function, d(x) is a bounded nonnegative function. Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 5 / 43

introduction (an example) Second order elliptic boundary value problem Search solution u for where (a u)+du = f in Ω Ê n, u = g on Ω, a(x) is a bounded symmetric positive definite matrix-valued function, d(x) is a bounded nonnegative function. This can be rewritten mixed form Search solution (q, u) for q+a u = 0 in Ω, q+du = f in Ω, u = g on Ω. Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 5 / 43

introduction (an example) We re choosing a triangulation of Ω, T h T h Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 6 / 43

introduction (an example) We re choosing a triangulation of Ω, T h and using Raviart-Thomas elements P k (K) n +xp k (K) T h RT 0 RT 1 RT 2 Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 6 / 43

introduction (an example) Notation E h interior edges, E h boundary faces of Ω, Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 7 / 43

introduction (an example) Notation Eh interior edges, E h boundary faces of Ω, E h = Eh E h, Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 7 / 43

introduction (an example) Notation Eh interior edges, E h boundary faces of Ω, E h = Eh E h, (u,v) D := D uvdx for u,v L2 (D), D Ê n, Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 7 / 43

introduction (an example) Notation Eh interior edges, E h boundary faces of Ω, E h = Eh E h, (u,v) D := D uvdx for u,v L2 (D), D Ê n, u,v D := D uvdx for u,v L2 (D), D Ê n 1, Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 7 / 43

introduction (an example) Notation Eh interior edges, E h boundary faces of Ω, E h = Eh E h, (u,v) D := D uvdx for u,v L2 (D), D Ê n, u,v D := D uvdx for u,v L2 (D), D Ê n 1, (u,v) Th = K T h (v,w) K for u,v on Ω, Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 7 / 43

introduction (an example) Notation Eh interior edges, E h boundary faces of Ω, E h = Eh E h, (u,v) D := D uvdx for u,v L2 (D), D Ê n, u,v D := D uvdx for u,v L2 (D), D Ê n 1, (u,v) Th = K T h (v,w) K for u,v on Ω, µ,λ E = e E µ,λ e for µ,λ on E E h. Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 7 / 43

introduction (an example) Notation Eh interior edges, E h boundary faces of Ω, E h = Eh E h, (u,v) D := D uvdx for u,v L2 (D), D Ê n, u,v D := D uvdx for u,v L2 (D), D Ê n 1, (u,v) Th = K T h (v,w) K for u,v on Ω, µ,λ E = e E µ,λ e for µ,λ on E E h. H(div,Ω) := {v L 2 (Ω) n v L 2 (Ω)} Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 7 / 43

introduction (an example) RT finite element approximation The approximation (q h,u h ) is sought in the finite element space V RT h Wh RT given by V RT h = {v H(div,Ω) v K P k (K) n +xp k (K) K T h }, W RT h = {w L 2 (Ω) w K P k (K) K T h }. and is defined by requiring that, Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 8 / 43

introduction (an example) RT finite element approximation The approximation (q h,u h ) is sought in the finite element space V RT h Wh RT given by V RT h = {v H(div,Ω) v K P k (K) n +xp k (K) K T h }, W RT h = {w L 2 (Ω) w K P k (K) K T h }. and is defined by requiring that, (cq h,v) Ω (u h, v) Ω = g,v n Ω (w, q h ) Ω (du h,w) Ω = (f,w) Ω v V RT h, w W RT h, with c = a 1 for each element K T h Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 8 / 43

introduction (an example) This yields to a matrix equation of the form: ( t )( ) ( É = Í À ), where É and Í are vectors of coefficients of q h and u h with respect to their corresponding finite element basis. Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 9 / 43

introduction (an example) problem System is not positive definite. Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 10 / 43

introduction (an example) problem System is not positive definite. solutions using expensive solver, Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 10 / 43

introduction (an example) problem System is not positive definite. solutions using expensive solver, using a positive definite system by elimination of É: solve equation ( 1 t + )Í = + 1 for Í. Required inversion 1 is difficult to compute and a full matrix. Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 10 / 43

introduction (an example) problem System is not positive definite. solutions using expensive solver, using a positive definite system by elimination of É: solve equation ( 1 t + )Í = + 1 for Í. Required inversion 1 is difficult to compute and a full matrix. hybridization: Introduce new unkowns λ h (Lagrangian multipliers) and relax the continuity constraints between element interfaces for q h. Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 10 / 43

introduction (an example) Definition (hybridized RT finite element approximation) The approximation (q h,u h,λ h ) is sought in the finite element space V h RT Ŵ h RT M h RT given by V h RT = {v L 2 (Ω) n v K P k (K) n +xp k (K) K T h }, Ŵ h RT = {w L 2 (Ω) w K P k (K) K T h }, M hrt = {µ L 2 (E h ) µ e P k (e) e E h } Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 11 / 43

introduction (an example) Definition (hybridized RT finite element approximation) The approximation (q h,u h,λ h ) is sought in the finite element space V h RT Ŵ h RT M h RT given by RT V h = {v L 2 (Ω) n v K P k (K) n +xp k (K) K T h }, RT Ŵ h = {w L 2 (Ω) w K P k (K) K T h }, M hrt = {µ L 2 (Eh ) µ e P k (e) e Eh } and is defined by requiring that, (cq h,v) Ω (u h, v) Ω + λ h, v E h = g,v n Ω v V RT h, (w, q h ) Ω (du h,w) Ω = (f,w) Ω w ŴhRT, µ, q h E h = 0 µ M hrt. Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 11 / 43

introduction (an example) This yields to a matrix of the form A B t C t Q G B D 0 U = F, C 0 0 Λ 0 where Λ is the vector of dofs associated to the multiplier λ h. Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 12 / 43

introduction (an example) This yields to a matrix of the form A B t C t Q G B D 0 U = F, C 0 0 Λ 0 where Λ is the vector of dofs associated to the multiplier λ h. New vectors of dofs Q and U define the same approximation (q h,u h ) as the original method. Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 12 / 43

introduction (an example) This yields to a matrix of the form A B t C t Q G B D 0 U = F, C 0 0 Λ 0 where Λ is the vector of dofs associated to the multiplier λ h. New vectors of dofs Q and U define the same approximation (q h,u h ) as the original method. Both Q and U can be easily eliminated Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 12 / 43

introduction (an example) This yields to a matrix of the form A B t C t Q G B D 0 U = F, C 0 0 Λ 0 where Λ is the vector of dofs associated to the multiplier λ h. New vectors of dofs Q and U define the same approximation (q h,u h ) as the original method. Both Q and U can be easily eliminated Λ = À, =CA 1 (A B t (BA 1 B t +D) 1 B)A 1 C t, À = CA 1 (A B t (BA 1 B t +D) 1 B)A 1 G CA 1 B t (BA 1 B t +D) 1 F. Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 12 / 43

introduction (an example) advantages of hybridization A is now a block diagonal matrix, Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 13 / 43

introduction (an example) advantages of hybridization A is now a block diagonal matrix, is symmetric positive definite, Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 13 / 43

introduction (an example) advantages of hybridization A is now a block diagonal matrix, is symmetric positive definite, number of dofs is remarkably smaller, Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 13 / 43

introduction (an example) advantages of hybridization A is now a block diagonal matrix, is symmetric positive definite, number of dofs is remarkably smaller, once Λ has been obtained, both Q and U can be computed efficiently element by element, Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 13 / 43

introduction (an example) advantages of hybridization A is now a block diagonal matrix, is symmetric positive definite, number of dofs is remarkably smaller, once Λ has been obtained, both Q and U can be computed efficiently element by element, the multiplier λ h can be used to improve the approximation to u by means of local post-processing. Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 13 / 43

introduction (an example) history of hybridization year event 1965 first hybridization of finite elements for solving equations of linear elasticity (often: static condensation)/ implementation trick [6] 1985 proof: hybrid variable contains extra information about the exact solution ( local post-processing)[5] 2004 hybridized RT and BDM methods of arbitrary order[2] 2005 extended to finite element methods for stationary Stokes equation (DG, mixed) 2009 unifying framework[3] 2010 comparison CG-H vs. LDG-H[4] Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 14 / 43

the framework 1 introduction (an example) 2 the framework 3 examples for hybridizable methods 4 Other novel methods hybridizable methods well suited for adaptivity The RT-method on meshes with hanging nodes Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 15 / 43

the framework Unified framework provides approximations for 1 (q,u) in the interior of the elements K T h, (q h,u h ), 2 u on the interior border of the elements, λ h. Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 16 / 43

the framework Definition (local solvers) For any single valued function m L 2 ( K), the functions (Qm,Um) are the solutions of cqm+ Um = 0 on K, Qm+dUm = 0 on K, Um = m on K Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 17 / 43

the framework Definition (local solvers) For any single valued function m L 2 ( K), the functions (Qm,Um) are the solutions of cqm+ Um = 0 on K, Qm+dUm = 0 on K, Um = m on K and for any single valued function f L 2 (K), the functions (Qf,Uf) are the solutions of cqf + Uf = 0 on K, Qf +duf = f on K, Uf = 0 on K. Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 17 / 43

the framework With u = λ + g and the linearity of the problem we have that (q,u) = (Qλ+Qg +Qf,Uλ+Ug +Uf). Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 18 / 43

the framework With u = λ + g and the linearity of the problem we have that (q,u) = (Qλ+Qg +Qf,Uλ+Ug +Uf). The above property only holds iff Definition (transmission condition) This completely characterizes λ. Qλ+Qg +Qf = 0 Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 18 / 43

the framework Definition (discrete local solvers) (Qm,Um) and (Qf,Uf) are the discrete versions of the local solvers. Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 19 / 43

the framework Definition (discrete local solvers) (Qm,Um) and (Qf,Uf) are the discrete versions of the local solvers. Then the discrete solution can be written as (q h,u h ) = (Qλ h +Qg h +Qf,Uλ h +Ug h +Uf), where λ h M h and g h M h are approximations to the values of u. u h λ h M h g h M h Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 19 / 43

the framework Definition (discrete version of transmission condition (weak)) a h (λ h,µ) = b h (µ) µ M h is the discrete version of the transmission condition. Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 20 / 43

the framework Notation V(K) polynomial space in which q is approximated, Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 21 / 43

the framework Notation V(K) polynomial space in which q is approximated, W(K) polynomial space in which u is approximated, Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 21 / 43

the framework Notation V(K) polynomial space in which q is approximated, W(K) polynomial space in which u is approximated, V h = {v v K V(K)}, Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 21 / 43

the framework Notation V(K) polynomial space in which q is approximated, W(K) polynomial space in which u is approximated, V h = {v v K V(K)}, W h = {w w K W(K)}. Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 21 / 43

the framework Notation V(K) polynomial space in which q is approximated, W(K) polynomial space in which u is approximated, V h = {v v K V(K)}, W h = {w w K W(K)}. In general: A function h on K is double-valued, each branch is denoted by h K + or h K. Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 21 / 43

the framework To define a hybridizable method, we had to define the discrete local solvers (Qm,Um) and (Qf,Uf), Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 22 / 43

the framework To define a hybridizable method, we had to define the discrete local solvers (Qm,Um) and (Qf,Uf), the transmission condition determining λ h, a h and b h, Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 22 / 43

the framework To define a hybridizable method, we had to define the discrete local solvers (Qm,Um) and (Qf,Uf), the transmission condition determining λ h, a h and b h, the finite element spaces Mh and M h, Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 22 / 43

the framework To define a hybridizable method, we had to define the discrete local solvers (Qm,Um) and (Qf,Uf), the transmission condition determining λ h, a h and b h, the finite element spaces Mh and M h, the finite element spaces V(K) and W(K), Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 22 / 43

the framework To define a hybridizable method, we had to define the discrete local solvers (Qm,Um) and (Qf,Uf), the transmission condition determining λ h, a h and b h, the finite element spaces Mh and M h, the finite element spaces V(K) and W(K), the trace spaces ˆQm and ˆQf (used in discrete local solvers). Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 22 / 43

the framework To define a hybridizable method, we had to define the discrete local solvers (Qm,Um) and (Qf,Uf), the transmission condition determining λ h, a h and b h, the finite element spaces Mh and M h, the finite element spaces V(K) and W(K), the trace spaces ˆQm and ˆQf (used in discrete local solvers). Finally, think about existence and uniqueness of λ h. Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 22 / 43

the framework Definition (trace space M h ) M h := {µ M h µ = 0 on Ω}. Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 23 / 43

the framework Definition (discrete local solver) 1 maps m M h to the function (Qm,Um): (cqm,v) K (Um, v) K = m,v n K ( w,qm) K + w, ˆQm n K +(dum,w) K = 0 v V(K), w W(K). Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 24 / 43

the framework Definition (discrete local solver) 1 maps m M h to the function (Qm,Um): (cqm,v) K (Um, v) K = m,v n K ( w,qm) K + w, ˆQm n K +(dum,w) K = 0 v V(K), w W(K). 2 maps f L 2 (Ω) to the pair (Qf,Uf): (cqf,v) K (Uf, v) K = 0 ( w,qf) K + w, ˆQf n K +(duf,w) K = (f,w) K v V(K), w W(K). Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 24 / 43

the framework Definition (characterization of λ h ) a h (η,µ) = (cqη,qµ) Th +(duη,uµ) Th + 1, (Uµ µ)(ˆqη Qη) Eh, b h (µ) = g h, ˆQµ Eh +(f,uµ) Th 1, (Uµ µ)(ˆqf Qf) Eh + 1, Uf(ˆQµ Qµ) Eh 1, (Uµ µ)(ˆqg h Qg h ) Eh + 1, (Ug h g)(ˆqµ Qµ) Eh. Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 25 / 43

the framework Theorem (existence and uniqueness of λ h ) If the three assumptions 1 2 3 are fulfilled, then there is a unique solution λ h of the weak formulation. Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 26 / 43

the framework Theorem (existence and uniqueness of λ h ) If the three assumptions 1 existence and uniqueness of the local solvers, 2 3 are fulfilled, then there is a unique solution λ h of the weak formulation. Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 26 / 43

the framework Theorem (existence and uniqueness of λ h ) If the three assumptions 1 existence and uniqueness of the local solvers, 2 positive semidefiniteness of the local solvers and 3 are fulfilled, then there is a unique solution λ h of the weak formulation. Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 26 / 43

the framework Theorem (existence and uniqueness of λ h ) If the three assumptions 1 existence and uniqueness of the local solvers, 2 positive semidefiniteness of the local solvers and 3 the gluing condition (If µ Mh, then for every interior face there exists a branch K with P K µ = µ.) are fulfilled, then there is a unique solution λ h of the weak formulation. Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 26 / 43

examples for hybridizable methods 1 introduction (an example) 2 the framework 3 examples for hybridizable methods 4 Other novel methods hybridizable methods well suited for adaptivity The RT-method on meshes with hanging nodes Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 27 / 43

examples for hybridizable methods general assumptions same local solvers in every element K, Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 28 / 43

examples for hybridizable methods general assumptions same local solvers in every element K, conforming simplicial triangulation. Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 28 / 43

examples for hybridizable methods general assumptions same local solvers in every element K, conforming simplicial triangulation. M c h,k := {µ C(E h) µ e P k (e) for all faces e E h } M h,k := {µ L 2 (E h ) µ e P k (e) for all faces e E h } Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 28 / 43

examples for hybridizable methods RT-H (Raviart Thomas): V(K) W(K) P k (K) n +xp k (K) P k (K) M h M h,k ˆQm Qm ˆQf Qf a h (η,µ) (cqη,qµ) Th +(duη,uµ) Th b h (µ) g h,qµ n Ω +(f,uµ) Th Conservativity strong Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 29 / 43

examples for hybridizable methods BDM-H (Brezzi Douglas Marini): V(K) W(K) P k (K) n P k 1 (K) M h M h,k ˆQm Qm ˆQf Qf a h (η,µ) (cqη,qµ) Th +(duη,uµ) Th b h (µ) g h,qµ n Ω +(f,uµ) Th Conservativity strong Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 30 / 43

examples for hybridizable methods LDG-H (Local Discontinuous Galerkin): V(K) W(K) 1.) P k (K) n P k 1 (K) 2.) P k (K) n P k (K) 3.) P k 1 (K) n P k (K) M h M h,k ˆQm Qm+τ K (Um m)n ˆQf Qf +τ K (Uf)n a h (η,µ) (cqη,qµ) Th +(duη,uµ) Th + 1, (Uµ µ)(τ K (Uη η)n) Eh b h (µ) g h,qµ n+τ K Uµ Ω +(f,uµ) Th Conservativity strong The stabilization parameter τ K is nonnegative constant on each face in E h, double-valued on Eh. Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 31 / 43

examples for hybridizable methods CG-H (Continuous Galerkin): V(K) W(K) P k 1 (K) n P k (K) M h M c h,k ˆQm a new unkown variable ˆQf a new unkown variable a h (η,µ) (a Uη, Uµ) Th +(duη,uµ) Th b h (µ) g h, ˆQµ Eh +(f,uµ) Th Conservativity weak Assume that a(x) is a constant on each element. CG-H is an LDG-H method with τ. Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 32 / 43

examples for hybridizable methods Are all methods hybridizable? Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 33 / 43

examples for hybridizable methods Are all methods hybridizable? No. Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 33 / 43

examples for hybridizable methods Are all methods hybridizable? No. Example (DG-methods) Remember the lecture Scientific Computing or see [1]. A hybrizable method has to be single valued for û h (= λ h ). Method û h Bassi-Rebay {u h } Brezzi et al. {u h } LDG {u h } β u h IP {u h } Bassi et al. {u h } Baumann-Oden {u h }+n K u h NIPG {u h }+n K u h Babuska-Zlamal (u h K ) K Brezzi et al. (u h K ) K Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 33 / 43

Other novel methods 1 introduction (an example) 2 the framework 3 examples for hybridizable methods 4 Other novel methods hybridizable methods well suited for adaptivity The RT-method on meshes with hanging nodes Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 34 / 43

Other novel methods hybridizable methods well suited for adaptivity 1 introduction (an example) 2 the framework 3 examples for hybridizable methods 4 Other novel methods hybridizable methods well suited for adaptivity The RT-method on meshes with hanging nodes Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 35 / 43

Other novel methods hybridizable methods well suited for adaptivity hybridizable methods well suited for adaptivity: V(K) W(K) 1.) P k(k) (K) n +xp k(k) P k(k) (K) 2.) P k(k) (K) n P k(k) 1 (K) 3.) P k(k) (K) n P k(k) (K) 4.) P k(k) 1 (K) n P k(k) (K) Mh {µ µ e M h,k(e) e Eh } {µ µ K C({x K τ K (x) = })} ˆQm Qm+τ K (Um m)n ˆQf Qf +τ K (Uf)n a h (η,µ) depending on V(K),W(K) b h (µ) depending on V(K),W(K) with τ K 0 for 3.) (on at least one face) and 4.) (for alle faces). Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 36 / 43

Other novel methods hybridizable methods well suited for adaptivity For e = K + K, we set max{k(k + ),k(k )}, if τ + < and τ < k(e) := k(k ± ), if τ ± = and τ < min{k(k + ),k(k )}, if τ + = and τ =. Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 37 / 43

Other novel methods hybridizable methods well suited for adaptivity assumption K τ K (0, ) interior face E h [0, ] τ K face of T h corresponding to LDG-H depending on the selected method Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 38 / 43

Other novel methods hybridizable methods well suited for adaptivity Main features of this class of methods: 1 variable degree approximation spaces on conforming meshes (k(k)), Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 39 / 43

Other novel methods hybridizable methods well suited for adaptivity Main features of this class of methods: 1 variable degree approximation spaces on conforming meshes (k(k)), 2 automatical coupling of different methods on conforming meshes (τ K determines method), Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 39 / 43

Other novel methods hybridizable methods well suited for adaptivity Main features of this class of methods: 1 variable degree approximation spaces on conforming meshes (k(k)), 2 automatical coupling of different methods on conforming meshes (τ K determines method), 3 mortaring capabilities for nonconforming meshes (choice of τ for hanging nodes). Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 39 / 43

Other novel methods The RT-method on meshes with hanging nodes 1 introduction (an example) 2 the framework 3 examples for hybridizable methods 4 Other novel methods hybridizable methods well suited for adaptivity The RT-method on meshes with hanging nodes Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 40 / 43

Other novel methods The RT-method on meshes with hanging nodes Considere the case of variable degree RT-H method with τ 0 everywhere (assumption in last method is violated). Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 41 / 43

Other novel methods The RT-method on meshes with hanging nodes Considere the case of variable degree RT-H method with τ 0 everywhere (assumption in last method is violated). Mesh is locally refined into four congruent triangles. Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 41 / 43

Other novel methods The RT-method on meshes with hanging nodes Considere the case of variable degree RT-H method with τ 0 everywhere (assumption in last method is violated). Mesh is locally refined into four congruent triangles. Idea: Impose special conditions on the meshes and link the definition of k(k) to the structure of the mesh. Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 41 / 43

Other novel methods The RT-method on meshes with hanging nodes k(k 2 ) " " max{k(k 2 ),k(k 4 )} k(k 4 ) arbitrarily max{k(k 3 ),k(k 4 )} k(k 3 ) " " k(k 1 ) Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 42 / 43

Other novel methods The RT-method on meshes with hanging nodes Thanks for your attention! Stefan Girke (WWU Münster Institut für Numerische und Angewandte Hybridization Mathematik) 10th of January, 2011 43 / 43