A sharp version of Kahan's theorem on clustered eigenvalues (Q1923193)

From MaRDI portal
scientific article
Language Label Description Also known as
English
A sharp version of Kahan's theorem on clustered eigenvalues
scientific article

    Statements

    A sharp version of Kahan's theorem on clustered eigenvalues (English)
    0 references
    0 references
    0 references
    0 references
    7 October 1996
    0 references
    Let \(H\) be a \(k\times k\) Hermitian matrix with eigenvalues \(\mu_1\leq \mu_2\leq \cdots \leq\mu_k\), let \(A\) be an \(n\times n\) Hermitian matrix with \(n\geq k\), and let \(Q\) be an \(n\times k\) matrix with full column rank. \textit{W. Kahan} [Inclusion theorems for clusters of eigenvalues of Hermitian matrices. Techn. Rep., Computer Sci. Dept., Univ. of Toronto (1967)] proved that there exist \(k\) eigenvalues \(\lambda_1\leq \lambda_2\leq \cdots\leq \lambda_k\) of \(A\) such that \[ \max_{1\leq j \leq k} |\mu_j-\lambda_j |\leq {c \over \sigma_{\min} (Q)} |AQ-QH |_2 \] for some constant \(c\leq\sqrt 2\), where \(\sigma_{\min} (Q)\) is the smallest singular value of \(Q\) and \(|\cdots |_2\) denote the biggest singular value of a matrix. In some special cases Kahan was able to show that \(c=1\), and he conjectured that \(c=1\) always. In this paper the authors prove this conjecture.
    0 references
    0 references
    clustered eigenvalues
    0 references
    Hermitian matrix
    0 references
    singular value
    0 references
    0 references