On perturbations for oblique projection generalized inverses of closed linear operators in Banach spaces (Q2431162)

For Banach spaces \(X\) and \(Y\), let \(L(X,Y)\), \(C(X,Y)\) and \(B(X,Y)\) denote the space of all linear operators, the space of all closed linear operators with a dense domain, and the space of all bounded linear operators from \(X\) into \(Y\), respectively. Let \(T \in C(X,Y)\) with closed range \(R(T)\). Suppose that \(X\) is the topological direct sum of the subspaces \(N(T)\) and \(M\), and \(Y\) is the topological direct sum of the subspaces \(R(T)\) and \(L\), where \(N(T)\) denotes the null space of \(T\). Let \(P\) denote the projector of \(X\) onto \(N(T)\) along \(M\) and \(Q\) denote the projector of \(Y\) onto \(R(T)\) along \(L\). Then there is a unique operator \(S \in B(Y,X)\) satisfying the formulae \(TST=T\) on \(D(T)\); \(STS=S\) on \(Y\); \(ST=I-P\) on \(D(T)\); and \(TS=Q\) on \(Y\). In that case, \(S\) is denoted by \(T^{\dagger}\) and is called an oblique projection generalized inverse of \(T\) with respect to \(P\) and \(Q\) [cf.\ \textit{M.\,Z.\thinspace Nashed} (ed.), ``Generalized inverses and applications''.\ Proceedings of an advanced seminar sponsored by the Mathematics Research Center (The University of Wisconsin, Madison, October 8--10, 1973) (1976; Zbl 0324.00009)]. One of the main results of the paper under review is as follows. Let \(T\) be as above and let \(\delta T \in L(X,Y)\) satisfy the inequality \(\| \delta T u \| \leq \alpha \| u \| + \beta \| Tu \|\) for all \(u \in D(T)\), where \(\alpha\) and \(\beta\) are nonnegative constants with \(\beta < 1\). Let \(\overline T\) be the perturbation of \(T\) by \(\delta T\), namely, \(\overline T =T+\delta T\). Suppose that \(N(\overline T) \cap R(T^{\dagger}) = \{0\}\), \(R(\bar T) \cap N(T^{\dagger}) = \{0\}\) and \(Y=\overline T R(T^{\dagger}) + L\). Then \(R(\overline T)\) is a closed subspace and the operator \(B=T^{\dagger}(I + \delta TT^{\dagger})^{-1}\) is an oblique projection generalized inverse of \(\overline T\) with \(N(B)=N(T^{\dagger})\) and \(R(B)=R(T^{\dagger})\). A norm bound on the difference \(B-T^{\dagger}\) is also presented.