A Black-Box Iterative Solver Based on a Two-Level Schwarz Method

We propose a black-box parallel iterative method suitable for solving both elliptic and certain non-elliptic problems discretized on unstructured meshes. The method is analyzed in the case of the second order elliptic problems discretized on quasiuniform P1 and Q1 finite element meshes. The numerical experiments confirm the validity of the proved convegence estimate and show that the method can successfully be used for more difficult problems (e.g. plates, shells and Helmholtz equation in high-frequency domain.) Received: July 28, 1997; revised June 20, 1999     

Iterative Methods for Linear Complementarity Problems with Interval Data

A ] and an interval vector [b]. If all A∈[A] are H-matrices with positive diagonal elements, these methods are all convergent to the same interval vector [x ^*]. This interval vector includes the solution x of the linear complementarity problem defined by any fixed A∈[A] and any fixed b∈[b]. If all A∈[A] are M-matrices, then we will show, that [x ^*] is optimal in a precisely defined sense. We also consider modifications of those methods, which under certain assumptions on the starting vector deliver nested sequences converging to [x ^*]. We close our paper with some examples which illustrate our theoretical results. Received October 7, 2002; revised April 15, 2003 Published online: June 23, 2003 RID="*" ID="*" Dedicated to U. Kulisch on the occasion of his 70th birthday. We are grateful to the referee who has given a series of valuable comments.     

On the Convergence Rate of a Preconditioned Subspace Eigensolver

In this paper we present a proof of convergence for a preconditioned subspace method which shows the dependency of the convergence rate on the preconditioner used. This convergence rate depends only on the condition of the pre-conditioned system and the relative separation of the first two eigenvalues . This means that, for example, multigrid preconditioners can be used to find eigenvalues of elliptic PDE's at a grid-independent rate. Received: March 9, 1999, revised June 23, 1999     

Variable Additive Preconditioning Procedures

Iterative methods with variable preconditioners of additive type are proposed. The scaling factors of each summand in the additive preconditioners are optimized within each iteration step. It is proved that the presented methods converge at least as fast as the Richardson's iterative method with the corresponding additive preconditioner with optimal scaling factors. In the presented numerical experiments the suggested methods need nearly the same number of iterations as the usual preconditioned conjugate gradient method with the corresponding additive preconditioner with numerically determined fixed optimal scaling factors. Received: June 10, 1998; revised October 16, 1998     

Non-Nested Multi-Level Solvers for Finite Element Discretisations of Mixed Problems

We consider a general framework for analysing the convergence of multi-grid solvers applied to finite element discretisations of mixed problems, both of conforming and nonconforming type. As a basic new feature, our approach allows to use different finite element discretisations on each level of the multi-grid hierarchy. Thus, in our multi-level approach, accurate higher order finite element discretisations can be combined with fast multi-level solvers based on lower order (nonconforming) finite element discretisations. This leads to the design of efficient multi-level solvers for higher order finite element discretisations. Received May 17, 2001; revised February 2, 2002 Published online April 25, 2002     

A Class of Smoothers for Saddle Point Problems

In this paper smoothing properties are shown for a class of iterative methods for saddle point problems with smoothing rates of the order 1/m, where m is the number of smoothing steps. This generalizes recent results by Braess and Sarazin, who could prove this rates for methods where, in the context of the Stokes problem, the pressure correction equation is solved exactly, which is not needed here. Received December 4, 1998; revised April 14, 2000     

A Cascadic Multigrid Algorithm for Semilinear Indefinite Elliptic Problems

We propose a cascadic multigrid algorithm for a semilinear indefinite elliptic problem. We use a standard finite element discretization with piecewise linear finite elements. The arising nonlinear equations are solved by a cascadic organization of Newton's method with frozen derivative on a sequence of nested grids. This gives a simple version of a multigrid method without projections on coarser grids. The cascadic multigrid algorithm starts on a comparatively coarse grid where the number of unknowns is small enough to obtain an approximate solution within sufficiently high precision without substantial computational effort. On each finer grid we perform exactly one Newton step taking the approximate solution from the coarsest grid as initial guess. The linear Newton systems are solved iteratively by a Jacobi-type iteration with special parameters using the approximate solution from the previous grid as initial guess. We prove that for a sufficiently fine initial grid and for a sufficiently good start approximation the algorithm yields an approximate solution within the discretization error on the finest grid and that the method has multigrid complexity with logarithmic multiplier. Received February 1999, revised July 13, 1999     

A Multigrid Method for Nonconforming FE-Discretisations with Application to Non-Matching Grids

Nonconforming finite element discretisations require special care in the construction of the prolongation and restriction in the multigrid process. In this paper, a general scheme is proposed, which guarantees the approximation property. As an example, the technique is applied to the discretisation by non-matching grids (mortar elements). Received: October 15, 1998     

On the Convergence of a Multigrid Method for Linear Reaction-Diffusion Problems

In this note we consider discrete linear reaction-diffusion problems. For the discretization a standard conforming finite element method is used. For the approximate solution of the resulting discrete problem a multigrid method with a damped Jacobi or symmetric Gauss-Seidel smoother is applied. We analyze the convergence of the multigrid V- and W-cycle in the framework of the approximation- and smoothing property. The multigrid method is shown to be robust in the sense that the contraction number can be bounded by a constant smaller than one which does not depend on the mesh size or on the diffusion-reaction ratio. Received June 15, 2000     

Stabilised hp-Finite Element Approximation of Partial Differential Equations with Nonnegative Characteristic Form

This paper is devoted to the a priori error analysis of the hp-version of a streamline-diffusion finite element method for partial differential equations with nonnegative characteristic form. This class of equations includes second-order elliptic and parabolic problems, first-order hyperbolic problems and second-order problems of mixed elliptic-parabolic-hyperbolic type. We derive error bounds which are simultaneously optimal in both the mesh size h and the spectral order p. Numerical examples are presented to confirm the theoretical results. Received October 28, 1999; revised May 26, 2000     

Condition Numbers of Approximate Schur Complements in Two- and Three-Dimensional Discretizations on Hierarchically Ordered Grids

We investigate multilevel incomplete factorizations of M-matrices arising from finite difference discretizations. The nonzero patterns are based on special orderings of the grid points. Hence, the Schur complements that result from block elimination of unknowns refer to a sequence of hierarchical grids. Having reached the coarsest grid, Gaussian elimination yields a complete decomposition of the last Schur complement. The main focus of this paper is a generalization of the recursive five-point/nine-point factorization method (which can be applied in two-dimensional problems) to matrices that stem from discretizations on three-dimensional cartesian grids. Moreover, we present a local analysis that considers fundamental grid cells. Our analysis allows to derive sharp bounds for the condition number associated with one factorization level (two-grid estimates). A comparison in case of the Laplace operator with Dirichlet boundary conditions shows: Estimating the relative condition number of the multilevel preconditioner by multiplying corresponding two-grid values gives the asymptotic bound O(h ^−0.347) for the two- respectively O(h ^−4/5) for the three-dimensional model problem. Received October 19, 1998; revised September 27, 1999     

Approximating Local Averages of Fluid Velocities: The Stokes Problem

As a first step to developing mathematical support for finite element approximation to the large eddies in fluid motion we consider herein the Stokes problem. We show that the local average of the usual approximate flow field u ^h over radius δ provides a very accurate approximation to the flow structures of O(δ) or greater. The extra accuracy appears for quadratic or higher velocity elements and degrades to the usual finite element accuracy as the averaging radius δ→h (the local meshwidth). We give both a priori and a posteriori error estimates incorporating this effect. Received December 3, 1999; revised October 16, 2000     

An Iterative Substructuring Method for div-stable Finite Element Approximations of the Oseen Problem

We apply an iterative substructuring algorithm with transmission conditions of Robin–Robin type to the discretized Oseen problem appearing as a linearized variant of the incompressible Navier–Stokes equations. Here we consider finite element approximations using velocity/pressure pairs which satisfy the Babuška–Brezzi stability condition. After proving well-posedness and strong convergence of the method, we derive an a-posteriori error estimate which controls convergence of the discrete subdomain solutions to the global discrete solution by measuring the jumps of the velocities at the interface. Additionally we obtain information how to design a parameter of the Robin interface condition which essentially influences the convergence speed. Numerical experiments confirm the theoretical results and the applicability of the method. Received February 18, 2000; revised February 21, 2001     

Incomplete Block Triangular Decompositions of Order l

In this paper, we present a special technique for improving the robustness of block triangular decompositions, for example those described in [1], that have proved their efficiency as preconditioners for the conjugate gradient method in a wide class of problems. The thorough theoretical analysis of this technique is carried out for the model problem, and the practical efficiency in the case of more complex problems is illustrated by numerical examples. Received August 2, 1999; revised September 21, 1999     

Energy Optimization of Algebraic Multigrid Bases

We propose a fast iterative method to optimize coarse basis functions in algebraic multigrid by minimizing the sum of their energies, subject to the condition that linear combinations of the basis functions equal to given zero energy modes, and subject to restrictions on the supports of the coarse basis functions. For a particular selection of the supports, the first iteration gives exactly the same basis functions as our earlier method using smoothed aggregation. The convergence rate of the minimization algorithm is bounded independently of the mesh size under usual assumptions on finite elements. The construction is presented for scalar problems as well as for linear elasticity. Computational results on difficult industrial problems demonstrate that the use of energy minimal basis functions improves algebraic multigrid performance and yields a more robust multigrid algorithm than smoothed aggregation. Received: March 9, 1998; revised January 25, 1999     

A Dynamic Method for Weighted Linear Least Squares Problems

A new method for solving the weighted linear least squares problems with full rank is proposed. Based on the theory of Liapunov's stability, the method associates a dynamic system with a weighted linear least squares problem, whose solution we are interested in and integrates the former numerically by an A-stable numerical method. The numerical tests suggest that the new method is more than comparative with current conventional techniques based on the normal equations. Received August 4, 2000; revised August 29, 2001 Published online April 25, 2002     

A Robust Smoother for Convection-Diffusion Problems with Closed Characteristics

In this paper we analyze a model problem for the convection-diffusion equation where the reduced problem has closed characteristics. A full upwinding finite difference scheme is used to discretize the problem. Additionally to the strength of the convection, an arbitrary amount of crosswind-diffusion can be added on the discrete level. We present a smoother which is robust w.r.t. the strength of convection and the amount of crosswind-diffusion. It is of Gauss–Seidel type using a downwind ordering. To handle the cyclic dependencies a frequency-filtering algorithm is used. The algorithm is of nearly optimal complexity ?(n log n). It is proved that it fulfills a robust smoothing property.     

The Inf-Sup Condition for the Mapped Q k ?P k?1 disc Element in Arbitrary Space Dimensions

 One of the most popular pairs of finite elements for solving mixed formulations of the Stokes and Navier–Stokes problem is the Q _k −P _k−1 ^disc element. Two possible versions of the discontinuous pressure space can be considered: one can either use an unmapped version of the P _k−1 ^disc space consisting of piecewise polynomial functions of degree at most k−1 on each cell or define a mapped version where the pressure space is defined as the image of a polynomial space on a reference cell. Since the reference transformation is in general not affine but multilinear, the two variants are not equal on arbitrary meshes. It is well-known, that the inf-sup condition is satisfied for the first variant. In the present paper we show that the latter approach satisfies the inf-sup condition as well for k≥2 in any space dimension. Received January 31, 2001; revised May 2, 2002 Published online: July 26, 2002     

An Efficient Direct Solver for the Boundary Concentrated FEM in 2D

The boundary concentrated FEM, a variant of the hp-version of the finite element method, is proposed for the numerical treatment of elliptic boundary value problems. It is particularly suited for equations with smooth coefficients and non-smooth boundary conditions. In the two-dimensional case it is shown that the Cholesky factorization of the resulting stiffness matrix requires O(Nlog⁴ N) units of storage and can be computed with O(Nlog⁸ N) work, where N denotes the problem size. Numerical results confirm theoretical estimates. Received October 4, 2001; revised August 19, 2002 Published online: October 24, 2002     

Problem Dependent Generalized Prewavelets

In this paper, we present a new approach to construct robust multilevel algorithms for elliptic differential equations. The multilevel algorithms consist of multiplicative subspace corrections in spaces spanned by problem dependent generalized prewavelets. These generalized prewavelets are constructed by a local orthogonalization of hierarchical basis functions with respect to a so-called local coarse-grid space. Numerical results show that the local orthogonalization leads to a smaller constant in strengthened Cauchy-Schwarz inequality than the original hierarchical basis functions. This holds also for several equations with discontinuous coefficients. Thus, the corresponding multilevel algorithm is a fast and robust iterative solver. Received November 13, 2001; revised October 21, 2002 Published online: December 12, 2002