The jump at zero of the spectral shift function (Q1883223): Difference between revisions

The author considers an \(n\)-dimensional Riemannian manifold which is cone-like at infinity, i.e., there is a neighbourhood of infinity of the form \(\Omega_R = (R,\infty) \times \Sigma\) with metric \(dr^2+r^2h\) where \((\Sigma, h)\) is a compact Riemannian manifold and a Schrödinger operator \(L = \Delta^g + V\) where \(V\) is homogeneous of degree \(-2\) at infinity (i.e., \(V(r,\theta)=q(\theta)/r^2\) on \(\Omega_R\)). He compares \(L\) with the operator \(L_{\Omega_R}\) where the latter operator is given by \(L\) on \(\Omega_R\) with Dirichlet boundary condition at \(\partial \Omega_R\). The author considers the spectral shift functions for the pair \((L, L_{\Omega_R})\). Note that the difference of the corresponding heat kernel operators is of trace class [cf. \textit{G. Carron}, Am. J. Math. 124, No. 2, 307--352 (2002; Zbl 1014.34015)]. In addition, it is assumed that \(\Delta^h + q\) (the Laplacian on \(\Sigma\)) is bounded from below by \(-(n-2)^2/4\). This implies that \(L_{\Omega_R}\) is positive and that \(L\) has only a finite number of negative eigenvalues. The main theorem expresses the difference of the left and right limit of the spectral shift function at the spectral value \(0\) as the dimension of solutions of \(L \Phi=0\) with a certain mild growth condition (including \(L_2\)-solutions). If the spectrum of \(\Delta^h+q\) is included in \((-(n-2)^2/4+1,\infty)\) then the difference is exactly the dimension of the \(L_2\)-solution space of \(L\Phi=0\). A solution therefore contributes to the jump by \(\nu\) if \(r^{\nu - 1 -\varepsilon} \Phi\) is in \(L_2\) for all \(\varepsilon\), but \(r^{\nu-1}\Phi\) is not. In particular, solutions with \(\nu=0\) do not contribute to the spectral shift function and is invisible for the trace formula. Finally, the author shows that these solutions contribute to the singularity of the scattering operator in zero. Therefore, solutions of \(L\Phi=0\) are almost \(L_2\) (i.e., \(r^{-\varepsilon} \Phi \in L_2\) for all \(\varepsilon>0\) but \(\Phi \notin L_2\)) are detectable in the scattering operator; such solutions are called zero resonance states. The techniques used is an asymptotic analysis of the Dirichlet-to-Neumann operator associated to the problem.