Spectral factorization of rectangular rational matrix functions with application to discrete Wiener-Hopf equations (Q1207136)

Generalizing the common notion for regular matrix functions the (possibly nonregular and nonsquare) matrix function \(A(z)\), which is continuous on the contour \(\Gamma\) and satisfies the condition: \(\text{rank } A(z)=k\) is constant on \(\Gamma\), is said to have a (right) Wiener-Hopf factorization with respect to \(\Gamma\) if \(A(z)=A_ -(z) D(z) A_ +(z)\) on \(\Gamma\), with \(D(z)= \text{diag}(({{z-z_ +} \over {z-z_ - }})^{\kappa_ i})_{i=1}^ k\), \(z_ -(z_ +)\) outside (inside) \(\Gamma\), \(\kappa_ 1\geq\dots \geq\kappa_ k\) integers and \(A_ -(z)\) \((A_ +(z))\) continuous outside (inside) and on \(\Gamma\), analytic outside (inside) \(\Gamma\), with left (right) inverses \(\widetilde{A}_ - (z)\) \((\widetilde {A}_ +(z))\) having the same properties. The indices \(\kappa_ 1\geq \dots\geq \kappa_ k\) are uniquely determined; if \(\kappa_ 1= \dots= \kappa_ k=0\) (i.e., if \(D(z)\equiv I_ k\)) then the factorization is called canonical. A rational matrix function \(W(z)\) has such a factorization with respect to \(\Gamma\) if it has no poles and no ``zeros'' (points \(z_ 0\) with \(\text{rank } A(z_ 0)< \max A(z)\)) on \(\Gamma\). The paper presents necessary and sufficient conditions (in terms of suitable ``matching'' subspaces of the state space in a minimal realization) in order that \(W\) has a canonical right Wiener-Hopf factorization \(W= W_ 1W_ 2\), and formulae for the factors \(W_ 1\), \(W_ 2\) are provided. The results are applied to the Wiener-Hopf equation \[ \sum_{k=0}^ \infty A_{j-k} x_ k= c_ j, \qquad j\in\mathbb{N}, \] where \(A_ i\), \(i\in\mathbb{Z}\), are -- possibly nonsquare -- matrices with \(\sum \| A_ i\|<\infty\), \((c_ j)\in \ell_ p^ m\).