On sinc discretization and banded preconditioning for linear third-order ordinary differential equations. (Q2889397)

From MaRDI portal





scientific article; zbMATH DE number 6043450
Language Label Description Also known as
default for all languages
No label defined
    English
    On sinc discretization and banded preconditioning for linear third-order ordinary differential equations.
    scientific article; zbMATH DE number 6043450

      Statements

      0 references
      0 references
      0 references
      7 June 2012
      0 references
      linear ordinary differential equation
      0 references
      sinc discretization
      0 references
      Toeplitz matrix
      0 references
      preconditioning
      0 references
      Krylov subspace method
      0 references
      boundary value problem
      0 references
      sinc-collocation
      0 references
      sinc-Galerkin
      0 references
      GMRES
      0 references
      BiCGSTAB
      0 references
      convergence
      0 references
      numerical experiments
      0 references
      0 references
      0 references
      0 references
      On sinc discretization and banded preconditioning for linear third-order ordinary differential equations. (English)
      0 references
      The paper considers solving a linear third-order ODE boundary value problem. First, the ODE is discretized using a combination of the sinc-collocation and sinc-Galerkin methods. It is shown, that the discrete solution converges to the exact solution of the problem. The resulting system of linear algebraic equations, \(Aw=p\), is generally nonsymmetric and ill-conditioned. The matrix \(A\) is a linear combination of diagonal and Toeplitz matrices. Thus the algebraic system is solved by Krylov subspace methods (GMRES and BiCGSTAB) taking advantage of the fast matrix-vector multiplication. To accelerate the convergence, a class of banded preconditioning matrices for the considered problem is derived. Under some special assumptions on the coefficients of the ODE, it is proved that the eigenvalues of the preconditioned matrix lie in a rectangle on the complex plane. This rectangle is independent of the size of the algebraic system. Convergence is illustrated on two numerical experiments.
      0 references

      Identifiers

      0 references
      0 references
      0 references
      0 references
      0 references
      0 references
      0 references
      0 references
      0 references