Abstract
It is well known that the projection of a matrix \(A\) onto a Krylov subspace span \(\left\{ \mathbf {h}, A\mathbf {h}, A^2\mathbf {h}, \ldots , A^{k-1}\mathbf {h}\right\} \), with \(A \in \mathbb {C}^{n \times n}\) and \(\mathbf {h} \in \mathbb {C}^n\), results in a Hessenberg matrix. We show that the projection of the matrix \(A\) onto an extended Krylov subspace, which is of the form span \(\left\{ A^{-k_r}\mathbf {h}, \ldots , A^{-2}\mathbf {h},A^{-1}\mathbf {h}, \mathbf {h}, A\mathbf {h}, A^2 \mathbf {h}, \ldots , A^{k_\ell } \mathbf {h} \right\} \), is a matrix of so-called extended Hessenberg form which can be characterized uniquely by its \(QR\)-factorization. This \(QR\)-factorization will be presented by means of a pattern of \(2 \times 2\) unitary rotations. We will show how this rotation pattern leads to new insights and allows to elegantly predict the structure of the matrix. In case \(A\) is Hermitian or unitary, this extended Hessenberg matrix is banded and structured, allowing the design of short recurrence relations. For the unitary case, coupled two term recurrence relations are derived of which the coefficients capture all information necessary for a sparse factorization of the corresponding extended Hessenberg matrix.
Similar content being viewed by others
Notes
These matrices were termed compressed matrices in [14], as they admit an equally expensive \(QR\)-factorization as a Hessenberg matrix.
Given a rotation \(Q\) and an upper triangular matrix \(R\), the product \(QR\) can be rewritten as \({\tilde{R}}{\tilde{Q}}\), where \({\tilde{R}}\) is upper triangular. Hence the rotation has been passed from the left to the right of the upper triangular matrix (of course rotations can also be passed from the right to the left). This shows that one can pass an entire twisted shape of rotations through an upper triangular matrix without changing the shape.
References
Antoulas, A.C.: Approximation of Large-Scale Dynamical Systems. SIAM, Philadelphia (2005)
Barroso, R.C., Delvaux, S.: Orthogonal Laurent polynomials on the unit circle and snake-shaped matrix factorizations. J. Approx. Theory 161(1), 65–87 (2009)
Cantero, M.J., Moral, L., Velazquez, L.: Five-diagonal matrices and zeros of orthogonal polynomials on the unit circle. Linear Algebra Appl. 362, 29–56 (2003)
Druskin, V., Knizhnerman, L.: Extended Krylov subspaces: approximations of the matrix square root and related functions. SIAM J. Matrix Anal. Appl. 19(3), 755–771 (1998)
Gragg, W.B.: The QR-algorithm for unitary Hessenberg matrices. J. Comput. Appl. Math. 16, 1–8 (1986)
Gragg, W.B.: Positive definite Toeplitz matrices, the Arnoldi process for isometric operators, and Gaussian quadrature on the unit circle. J. Comput. Appl. Math. 46(1–2), 183–198 (1993)
Jagels, C., Reichel, L.: The extended Krylov subspace method and orthogonal Laurent polynomials. Linear Algebra Appl. 431(3), 441–458 (2009)
Knizhnerman, L., Simoncini, V.: A new investigation of the extended Krylov subspace method for matrix function evaluations. Numer. Linear Algebra Appl. 17(4), 615–638 (2010)
Knizhnerman, L., Simoncini, V.: Convergence analysis of the extended Krylov subspace method for the Lyapunov equation. Numer. Math. 118(3), 567–586 (2011)
Simon, B.: CMV matrices: five years after. J. Comput. Appl. Math. 208(1), 120–154 (2007)
Simoncini, V.: A new iterative method for solving large scale Lyapunov matrix equations. SIAM J. Sci. Comput. 29(3), 1268–1288 (2007)
Vandebril, R.: Chasing bulges or rotations? A metamorphosis of the QR-algorithm. SIAM J. Matrix Anal. Appl. 32, 217–247 (2011)
Vandebril, R., Van Barel, M., Mastronardi, N.: Matrix Computations and Semiseparable Matrices, Volume I: Linear Systems. Johns Hopkins University Press, Baltimore (2008)
Vandebril, R., Watkins, D.S.: A generalization of the multishift QR algorithm. SIAM J. Matrix Anal. Appl. 33(3), 759–779 (2012)
Watkins, D.S.: Some perspectives on the eigenvalue problem. SIAM Rev. 35(3), 430–471 (1993)
Watkins, D.S.: Unitary orthogonalization processes. J. Comput. Appl. Math. 86(1), 335–345 (1997)
Author information
Authors and Affiliations
Corresponding author
Additional information
The research was partially supported by the Research Council KU Leuven, projects OT/11/055 (Spectral Properties of Perturbed Normal Matrices and their Applications), CREA/13/012 (Can Unconventional Eigenvalue Algorithms Supersede the State-Of-The-Art), PFV/10/002 (Optimization in Engineering, OPTEC) , by the Fund for Scientific Research–Flanders (Belgium) project G034212N (Reestablishing Smoothness for Matrix Manifold Optimization via Resolution of Singularities), and by the Interuniversity Attraction Poles Programme, initiated by the Belgian State, Science Policy Office, Belgian Network DYSCO (Dynamical Systems, Control, and Optimization).
Rights and permissions
About this article
Cite this article
Mertens, C., Vandebril, R. Short recurrences for computing extended Krylov bases for Hermitian and unitary matrices. Numer. Math. 131, 303–328 (2015). https://doi.org/10.1007/s00211-014-0692-3
Received:
Revised:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00211-014-0692-3