Shift invariant manifolds and nonlinear analytic function theory (Q1112302): Difference between revisions

The purpose of this paper is, first of all, to extend the indefinite metric Beurling-Lax-Halmos theory on the representation of shift invariant (linear) subspaces of \(\ell^ 2_ N\) developed by the present authors [see for example \textit{J. W. Helton}, Operator theory, analytic functions, matrices and electrical engineering, Reg. Conf., Ser. Math. 68 (1987; Zbl 0639.47001)] to the case of nonlinear shift invariant manifolds. Let us present at least one result. Let M be a shift invariant (nonlinear) submanifold of \(\ell^ 2_ N\). A spray structure on M is a partition of M into a family \(\{S^{(n)}:\) \(n\geq 0\}\) of submanifolds (``sprays'') of M. Elements of \(S^{(n)}\) are indexed by elements \(a=(a_ 0,a_ 1,...,a_ n)\) of \(\ell^ 2_ N[0,n]\). Assume that: 1) \(S=\{S^{(n)}:\) \(n\geq 0\}\) is a spray structure on M such that a) the submanifolds \(S^{(n)}(a)\) depend on \(a\in \ell^ 2_ N[0,n]\) in a smooth way; b) \(S^{(0)}=\{S^{(0)}(0)\}=\{M\};\) c) if \(a,a'\in R^{(n+1)N}\) and \(a\neq a'\), then \(S^{(n)}(a)\cap S^{(n)}(a')=\emptyset;\) d) \(S^{(n)}(a)=\cup \{S^{(n+1)}(a,r):\) \(r\in R^ N\}\) for any \(a\in R^{(n+1)N};\) e) \(US^{(n)}(a)=S^{(n+1)}(0,a)\) for any \(a\in \ell^ 2_ N[0,n]\), where U is the bilateral shift on \(\ell^ 2_ N;\) 2) \(\rho\) is a shift invariant energy form on \(\ell^ 2_ N\) of class \(C^ 2.\) Then M is of the form \(\theta.\ell_ N^{2+}\) if and only if (S,\(\rho)\) satisfies the following three conditions: (CRIT) for each \(a\in \ell^ 2_ N[0,n]\), \(\rho | S^{(n)}(a)\) has a unique critical point, say \({\mathcal K}^{(n)}(a)\), and the function \(a\to {\mathcal K}^{(n)}(a)\) is of class \(C^ 2\), its derivative at each point is left invertible and \({\mathcal K}^{(0)}(0)=0;\) (HESS) for each \(a\in \ell^ 2_ N[0,n]\), the 2nd derivative \(D^ 2\rho ({\mathcal K}(a))\) is a nondegenerate bilinear form (on the square of \(T_{{\mathcal K}^{(n)}(a)}(S^{(n)}(a)));\) (DENS) the set \(\{\) \({\mathcal K}^{(n)}(a):\) \(a\in R^{(n+1)N}\), \(n\geq 0\}\) is dense in M. In the second part of the paper (over 50 pages) there are given applications to the phase-outer factorization, interpolation, \(H^{\infty}\)-approximation, commuting lifting, passivity of circuits and subcircuits and several electrical engineering problems.