Normal matrices with a dominant eigenvalue and an eigenvector with no zero entries (Q1855405)

A square complex matrix is said to be dominant if it has an algebraically simple eigenvalue whose module is strictly greater than the module of any other eigenvalues; this eigenvalue and any associated eigenvector are called dominant. The classical Perron theorem asserts that a square matrix with positive entries is dominant, its dominant eigenvalue is a positive scalar equal to the spectral radius and has an associated eigenvector with positive entries. Inequalities that are sufficient to ensure that a normal or Hermitian matrix is dominant and has a dominant eigenvector with no zero entries are provided. In particular, it is proved that if \(A\) is normal with eigenvalues \(\lambda_1,\ldots, \lambda _n\), \(|\lambda_1|\geq \ldots\geq |\lambda _n|\), and there exists a vector \(v\) such that \(|v^*Av|> \frac{(\text{tr} (A^*A))^{1/2} \|v\|^2}{\sqrt{2}}\) then \(|\lambda_1|> \frac{(\text{tr} (A^*A))^{1/2} }{\sqrt{2}} > |\lambda _2|\), so \(A\) is dominant. For real symmetric matrices these inequalities force the entries of dominant real eigenvector to have a prescribed sign pattern. The cases of equality in explored inequalities are considered. It is proved that in general case \(A\) is dominant and the dominant eigenvector has at most one zero entry. The author shows that exceptional extremal matrices and extremal matrices possessing a dominant eigenvector with a zero entry have certain special forms.