Global inverse spectral problems for rational matrix functions (Q1821845): Difference between revisions

Let Y be a Hilbert space, \(\dim Y=m<\infty\), and let \(L^ 2_ Y\) be the space of norm square integrable functions f(\(\lambda)\) on the unit circle \(\{\lambda|\;| \lambda | =1\}\) such that \(f(\lambda)\in Y\) for any fixed \(\lambda\). \(H^ 2_ Y\) is the Hardy subspace of \(L^ 2_ Y\), i.e. \(H^ 2_ Y\) is the subspace of functions which are radial limits of functions analytic on the unit disk \({\mathcal D}=\{\lambda | | \lambda | <1\}\), and \(H_ Y^{2\bot}\) is the orthogonal complement of \(H^ 2_ Y\) in \(L^ 2_ Y\). The authors consider rational \(m\times m\) matrix valued functions \(W(\lambda)\) analytic and invertible on the unit circle and at \(\infty\). Using minimal realizations \(W(\lambda)=I+C(\lambda I-A_ p)^{-1}\tilde B\) and \(W(\lambda)^{- 1}=I-\tilde C(\lambda I-A_ z)^{-1}B\) the authors introduce the notion of canonical sets of spectral data (c.s.s.d.) on \({\mathcal D}\) and on \({\mathcal D}_ e=\{\lambda | | \lambda | >1\}\) for W. The following results are stated. (i) Given a c.s.s.d. \(\{(C_+,A_{p+}),(A_{z+},B_+),\hat T\}\) on \({\mathcal D}\) and a c.s.s.d. \(\{(C_-,A_{p-}),(A_{z-},B_-),\hat T^ x\}\) on \({\mathcal D}_ e\) for W, then the shift-invariant subspaces \(WH^ 2_ Y\) and \(WH_ Y^{2\bot}\) can be described explicitly in terms of the c.s.s.d. (ii) A sufficient and necessary condition is given for a collection of operators \(\{(C_+,A_{p+}),(A_{z+},B_+),\hat T\}\) to be a c.s.s.d. for some W on \({\mathcal D}\). (iii) A sufficient and necessary condition is given for two collections of operators, \(\{(C_+,A_{p+}),(A_{z+},B_+),\hat T\}\) and \(\{(C_-,A_{p- }),(A_{z-},B_-),\hat T^ x\}\) to be c.s.s.d. for the same W on \({\mathcal D}\) and on \({\mathcal D}_ e\), respectively. The applications include canonical Wiener-Hopf factorization, J-inner-outer factorization and coprime factorization of a known rational matrix function \(W(\lambda)\).