Located Operators

Operators from a Hilbert space \(H\) to itself have many applications, ranging from linear integral equations to quantum physics. Many of these applications use known properties of operators such as their spectral decomposition. Because of these applications, it is desirable to analyze when the corresponding mathematical existence theorems can be transformed into algorithms. In other words, it is desirable to design a constructive-mathematics version of operator theory. Such a version has been constructed for bounded operators. In the corresponding theory, a constructive operator \(A\) is defined as an algorithm that transforms a vector \(x\in H\) into \(Ax\in H\). In many applications, however, it is desirable to consider unbounded operators as well. For example, in quantum mechanics, momentum is described by an (unbounded) differentiation operator. An unbounded operator is only partially defined on \(H\); it is therefore difficult even to define a good constructive analog of an unbounded operator -- to say nothing about designing algorithms for thus defined objects. The author provides a new natural constructive definition of an unbounded operator. Namely, with each operator \(A\), we can associate its graph \(G=\{\langle x,Ax\rangle\,| \,x\in H\}\subseteq H^2\). It turns out that for practically useful operators, graphs are \textit{located}, i.e., there is an algorithm that maps every vector \(x\in H^2\) into a distance from \(x\) to \(G\). The authors prove that locatedness is a necessary and sufficient condition for a bounded operator to have an adjoint, and it is also a necessary and sufficient condition for an unbounded operator to have a constructive spectral decomposition. Thus, a located operator looks like an appropriate constructive analog of the operator from \(H\) to \(H\).

Recommendations

Cited in

(7)

This page was built for publication: Located Operators

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4787861)