The essential uniqueness property for operators on Kreĭn spaces (Q1310824): Difference between revisions

For \(T\) a bounded linear operator between Krein spaces \(\mathcal H\) and \(\mathcal K\) one can always construct a unitary operator \(U\) of the form \(U=\bigl[{T\atop\widetilde D^*}{D\atop -L^*}\bigr]\) from the Krein space \({\mathcal H}\oplus{\mathcal D}\) into the Krein space \({\mathcal K}\otimes\widetilde{\mathcal D}\) (where \(\mathcal D\) and \(\widetilde{\mathcal D}\) are auxiliary Krein spaces) so that \(\text{ker }D=\{0\}\) and \(\text{ker }\widetilde D=\{0\}\). In this setup \(\widetilde D\) is called a defect operator for \(T\), \(D\) is called a defect operator for \(T^*\) and \(U\) is called a Julia operator (sometimes also elementary rotation or unitary colligation) for \(T\). Clearly, for any Krein space unitary operators \(W\) and \(V\), \(U'=\bigl[{I\atop 0}{0\atop W^*}\bigr] U\bigl[{I\atop 0}{0\atop V}\bigr]\) is also a Julia operator for \(T\) whenever \(U\) is. If this is the extent of nonuniqueness of the Julia operator for \(T\) we say that the Julia operator is essentially unique. A result of Dritschel and Rovnyak is that essential uniqueness holds if either \(\widetilde{\mathcal D}\) or \(\widetilde{\mathcal D}\) is a Pontryagin space. One of the main results of the paper under review is that essential uniqueness holds if and only if, given any defect operator \(\widetilde D\) of \(T\), the range of \(\widetilde D^*\) contains either a maximal uniformly negative or a maximal uniformly positive subspace. This is done through a detailed analysis of the ranges of operators with dense range in a Krein space.