A note on \(J\)-positive block operator matrices (Q487211)

Block operators of the form \( {\mathcal A} = \left(\begin{smallmatrix} A & C^*\\ C & B \end{smallmatrix}\right) \) are considered in a Hilbert space \({\mathfrak H} = {\mathcal H} \times {\mathcal H}\) based on another Hilbert space \({\mathcal H}\). General conditions are assumed such that \({\mathcal A}\) is self-adjoint (i.e., the closure of an essentially self-adjoint operator, which is more precisely defined). These conditions include that \(A\) and \(B\) are self-adjoint and the form \(\langle {\mathcal A} f , f \rangle\) has at most finitely many negative squares. The fundamental symmetry \( {\mathcal J} = \left(\begin{smallmatrix} 0 & i I\\ -i I & 0 \end{smallmatrix}\right) \) turns \({\mathfrak H}\) into a Krein space and \({\mathcal L} = {\mathcal J} {\mathcal A}\) is \({\mathcal J}\)-self-adjoint. The spectra of \({\mathcal A}\) and of \({\mathcal L}\) are studied. The definitizability of \({\mathcal L}\) (i.e., here \(\rho ({\mathcal L}) \neq \emptyset\)) is characterized in terms of \(T(z) := B - (C + iz) A^{-1}(C^* - iz)\) and examples of non-definitizable operators \({\mathcal L}\) are presented. Under the assumption \(\sigma ({\mathcal L}) \subset {\mathbb R}\), the main focus is on the similarity of \({\mathcal L}\) to a self-adjoint operator, or, in other words, on the regularity of the critical points of \({\mathcal L}\), in particular, of \(\infty\). A well-known necessary condition is the so-called linear resolvent growth condition which is here reformulated in terms of \(T(z)\). This condition gives rise to a number of examples where \(\infty\) is a singular critical point of \({\mathcal L}\). The main example is the operator \({\mathcal L}\) with \(A = - d^2/dx^2 + m^2 + V(x)\), \(C = \nu \, d/dx\), \(B = I\) and \({\mathcal H} = L^2({\mathbb R})\) induced by the nonlinear relativistic Ginzburg-Landau equation. Explicit conditions on the coefficients are obtained such that \(\infty\) is a singular critical point. This result indicates the limits of earlier research on eigenfunction expansions [\textit{A. Komech} and \textit{E. Kopylova}, J. Stat. Phys. 154, No. 1--2, 503--521 (2014; Zbl 1300.34195)].