Summary.
Considered are Hankel, Vandermonde, and Krylov basis matrices. It is proved that for any real positive definite Hankel matrix of order \(n\), its spectral condition number is bounded from below by\(3 \cdot 2^{n-6}\) . Also proved is that the spectral condition number of a Krylov basis matrix is bounded from below by\(3^{\frac{1}{2}} \cdot 2^{\frac{n}{2}-3}\) . For \(V = V(x_1, \ldots, x_n)\), a Vandermonde matrix with arbitrary but pairwise distinct nodes \(x_1, \ldots, x_n\), we show that \(\mbox{\rm cond}_2\,V \geq 2^{n-2}/n^{\frac{1}{2}}\); if either \(|x_j| \leq 1\) or\(|x_j| \geq 1\) for all \(j\), then \(\mbox{\rm cond}_2\,V \geq 2^{n-2}\).
Similar content being viewed by others
Author information
Authors and Affiliations
Additional information
Received January 24, 1993/Revised version received July 19, 1993
Rights and permissions
About this article
Cite this article
Tyrtyshnikov, E. How bad are Hankel matrices?. Numer. Math. 67, 261–269 (1994). https://doi.org/10.1007/s002110050027
Issue Date:
DOI: https://doi.org/10.1007/s002110050027