Double Sobolev gradient preconditioning for nonlinear elliptic problems

O. Axelsson, J. Karátson

Research output: Contribution to journalArticle

Abstract

A mixed variable formulation of a second-order nonlinear diffusion problem leads to a finite element matrix in a product form. This form enables the efficient updating of the nonlinearity in a Picard type iteration method, in which the preconditioner involves twice a discrete Laplacian. The article gives a conditioning analysis of this method, based on analytic investigations in the corresponding Sobolev function space that reveal the behaviour of this preconditioning. The further generalization of the preconditioner can produce arbitrarily low condition numbers by proper subdivisions of Ω, while still no differentiability of the nonlinear diffusion coefficient is required.

Original languageEnglish
Pages (from-to)1018-1036
Number of pages19
JournalNumerical Methods for Partial Differential Equations
Volume23
Issue number5
DOIs
Publication statusPublished - Sep 2007

Fingerprint

Sobolev Gradient
Nonlinear Elliptic Problems
Nonlinear Diffusion
Preconditioning
Preconditioner
Discrete Laplacian
Product Form
Diffusion Problem
Iteration Method
Condition number
Differentiability
Subdivision
Conditioning
Function Space
Diffusion Coefficient
Sobolev Spaces
Updating
Nonlinear Problem
Nonlinearity
Finite Element

Keywords

  • Discrete laplacian
  • Nonlinear diffusion problem
  • Picard-type iteration
  • Sobolev gradient preconditioning

ASJC Scopus subject areas

  • Applied Mathematics
  • Analysis
  • Computational Mathematics

Cite this

Double Sobolev gradient preconditioning for nonlinear elliptic problems. / Axelsson, O.; Karátson, J.

In: Numerical Methods for Partial Differential Equations, Vol. 23, No. 5, 09.2007, p. 1018-1036.

Research output: Contribution to journalArticle

@article{ca68abeef66948038718de1f2284bf74,
title = "Double Sobolev gradient preconditioning for nonlinear elliptic problems",
abstract = "A mixed variable formulation of a second-order nonlinear diffusion problem leads to a finite element matrix in a product form. This form enables the efficient updating of the nonlinearity in a Picard type iteration method, in which the preconditioner involves twice a discrete Laplacian. The article gives a conditioning analysis of this method, based on analytic investigations in the corresponding Sobolev function space that reveal the behaviour of this preconditioning. The further generalization of the preconditioner can produce arbitrarily low condition numbers by proper subdivisions of Ω, while still no differentiability of the nonlinear diffusion coefficient is required.",
keywords = "Discrete laplacian, Nonlinear diffusion problem, Picard-type iteration, Sobolev gradient preconditioning",
author = "O. Axelsson and J. Kar{\'a}tson",
year = "2007",
month = "9",
doi = "10.1002/num.20207",
language = "English",
volume = "23",
pages = "1018--1036",
journal = "Numerical Methods for Partial Differential Equations",
issn = "0749-159X",
publisher = "John Wiley and Sons Inc.",
number = "5",

}

TY - JOUR

T1 - Double Sobolev gradient preconditioning for nonlinear elliptic problems

AU - Axelsson, O.

AU - Karátson, J.

PY - 2007/9

Y1 - 2007/9

N2 - A mixed variable formulation of a second-order nonlinear diffusion problem leads to a finite element matrix in a product form. This form enables the efficient updating of the nonlinearity in a Picard type iteration method, in which the preconditioner involves twice a discrete Laplacian. The article gives a conditioning analysis of this method, based on analytic investigations in the corresponding Sobolev function space that reveal the behaviour of this preconditioning. The further generalization of the preconditioner can produce arbitrarily low condition numbers by proper subdivisions of Ω, while still no differentiability of the nonlinear diffusion coefficient is required.

AB - A mixed variable formulation of a second-order nonlinear diffusion problem leads to a finite element matrix in a product form. This form enables the efficient updating of the nonlinearity in a Picard type iteration method, in which the preconditioner involves twice a discrete Laplacian. The article gives a conditioning analysis of this method, based on analytic investigations in the corresponding Sobolev function space that reveal the behaviour of this preconditioning. The further generalization of the preconditioner can produce arbitrarily low condition numbers by proper subdivisions of Ω, while still no differentiability of the nonlinear diffusion coefficient is required.

KW - Discrete laplacian

KW - Nonlinear diffusion problem

KW - Picard-type iteration

KW - Sobolev gradient preconditioning

UR - http://www.scopus.com/inward/record.url?scp=34548530288&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=34548530288&partnerID=8YFLogxK

U2 - 10.1002/num.20207

DO - 10.1002/num.20207

M3 - Article

VL - 23

SP - 1018

EP - 1036

JO - Numerical Methods for Partial Differential Equations

JF - Numerical Methods for Partial Differential Equations

SN - 0749-159X

IS - 5

ER -