Lower bounds for the spread of a matrix (Q1066236): Difference between revisions

This paper refers to some lower bounds for the spread \(s(A)=\max_{i,j}| \lambda_ i-\lambda_ j|\) of an \(n\times n\) matrix \(A=(a_{ij})\), \(\lambda_ i\) being its eigenvalues. They proceed from the following results due to \textit{L. Mirsky} [Duke Math. J. 24, 591- 599 (1957; Zbl 0081.251)]: s(A)\(\geq \sqrt{3} \sup (u,Av)\) for A normal and \(s(A)=2 \sup (u,Av)\) for A Hermitian; sup is taken with respect to all orthonormal vectors u,v. By appropriate choices of u and v one derives four results. For instance s(A)\(\geq | \sum_{i\neq j}a_{ij} | /(n-1)\) and, A being symmetric, s(A)\(\geq 2\nu\), where \(\nu\) denotes the standard deviation of the row sums of A. Comparisons of the bounds derived, with several known bounds, are made in the final part of the paper.