\(A\)-partial isometries and generalized inverses (Q2637115): Difference between revisions

Let \(A\) be a positive operator in \(B(H)\). One may consider the semi-inner product \(\langle x,y\rangle_A:=\langle Ax, y\rangle\). An operator \(T\in B(H)\) is called \(A\)-adjointable if it is adjointable with respect to \(\langle \cdot,\cdot\rangle_A\). Clearly, an \(A\)-adjoint \(T^\#\) of \(T\) exists if and only if \(AT^\#=T^*A\). Similarly, by an \(A\)-partial isometry we mean an operator \(T\in B(H)\) behaving as a partial isometry with respect to \(\langle \cdot,\cdot\rangle_A\). This concept was first introduced in [\textit{M. L. Arias} et al., Linear Algebra Appl. 428, No. 7, 1460--1475 (2008; Zbl 1140.46009)]. It is known that the classical partial isometries can be defined as those operators whose adjoint coincides with their Moore-Penrose generalized inverse. In the paper under review, the authors explore whether a similar equivalence holds for \(A\)-partial isometries. They also study similarity in the context of \(A\)-partial isometries.