Some results on numerical ranges and factorizations of matrix polynomials

From MaRDI portal
Publication:4342152





Let \(L(\lambda)= \lambda^mI_n+ \sum^{m-1}_{j=0} \lambda^jA_j\) be a matrix polynomial, where \(A_j\in\mathbb{C}^{n\times n}\). Denote by \((f,g)\) and \(|f|= \sqrt{(f,f)}\) the standard scalar product and the norm in \(\mathbb{C}^n\). Let \(S^{n-1}\subset\mathbb{C}^n\) be the unit sphere and \(W(L)= \{\lambda\in\mathbb{C}:(L(\lambda)f,f)= 0\) for some \(f\in S^{n-1}\}\) the numerical range of \(L\). The components of \(W(L)\) are the maximal connected subsets in \(W(L)\). Let \(\nu(L)\leq m\) be the number of components of \(W(L)\). It is known that the statement (1) ``\(\nu(L)= m\) implies (2) ``For every \(f\in S^{n-1}\) the polynomial \((L(\lambda)f,f)\) has \(m\) distinct roots. Also, if \(W(L)\) is real then (1) and (2) are equivalent. The authors show that (1) and (2) are equivalent if \(m=2\) and \(A_1\) is Hermitian. A number of other properties of quadratic matrix polynomials are considered. Conditions for the factorization \(L(\lambda)= (\lambda I_n- Y_1)\cdots(\lambda I_n-Y_m)\) are also given. Finally, the extension of these results is discussed for operator polynomials whose coefficients are operators in an infinite-dimensional Hilbert space.











This page was built for publication: Some results on numerical ranges and factorizations of matrix polynomials

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4342152)