A Generalized Rayleigh Quotient Iteration for Computing Simple Eigenvalues of Nonnormal Matrices (Q4944712)

From MaRDI portal

!

This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:

scientific article; zbMATH DE number 1420437
Language Label Description Also known as
default for all languages
No label defined
    English
    A Generalized Rayleigh Quotient Iteration for Computing Simple Eigenvalues of Nonnormal Matrices
    scientific article; zbMATH DE number 1420437

      Statements

      0 references
      0 references
      21 November 2000
      0 references
      nonnormal eigenvalue problem
      0 references
      Newton iteration
      0 references
      simple bifurcation point
      0 references
      Rayleigh quotient iteration
      0 references
      inverse iteration
      0 references
      convergence
      0 references
      numerical examples
      0 references
      algorithm
      0 references
      eigenvectors
      0 references
      A Generalized Rayleigh Quotient Iteration for Computing Simple Eigenvalues of Nonnormal Matrices (English)
      0 references
      The paper presents a new algorithm for iterative computation of eigenvalues and the corresponding left and right eigenvectors of nonnormal matrices. The new algorithm is based on Newton's method for solving a parameter dependent nonlinear system of the form \(F(x,\lambda)= Ax-x\lambda=0\). If \(\lambda\) is an eigenvalue of \(A\) then \((0,\lambda)\) is a bifurcation point on the solution set of the system. The well-known bifurcation point algorithm of \textit{A. Griewank} and \textit{G. W. Reddien} [SIAM J. Numer. Anal. 21, 176-185 (1984; Zbl 0536.65031)], later modified by \textit{E. L. Allgower} and \textit{H. Schwetlick} [Z. Angew. Math. Mech. 77, No. 2, 83-97 (1997; Zbl 0915.65042)], is adopted by the authors to the problem considered. The derived algorithm is studied theoretically w.r.t. convergence and approximation errors. Further extension leads to another method with updated bordering vectors, which can be viewed as a modification of the Rayleigh quotient iterations. The last algorithm is able to compute ill-conditioned eigenvalues and the corresponding left and right eigenvectors. It is shown that the eigenvalue approximations converge \(Q\)-quadratically, whereas the eigenvector iterations converge \(R\)-quadratically to the corresponding exact solutions. The advantages of the new algorithm are demonstrated on various numerical problems and comparisons with related methods are given.
      0 references

      Identifiers