Complex eigenvalue bounds for a Schrödinger operator on the half line (Q2180282)

For \(\nu\geq0\) and complex-valued, scalar potentials \(V\), the authors are locating complex eigenvalues of the perturbed Hardy operator \[ H_{0,\nu}+V:= -\frac{d^2}{dx^2} + \frac{\nu^2-1/4}{x^2} +V \quad \text{in}\ L^2(0,\infty)\,, \] subject to a Dirichlet boundary condition at the origin. The optimal shape of the location of complex eigenvalues of \(H_{0,1/2}+V\) was already found in [\textit{R. L. Frank} et al., Oper. Theory: Adv. Appl. 214, 39--44 (2011; Zbl 1253.35086)]. A fundamental difference between the operator \(H_{0,1/2}+V\) in \(L^2(0,\infty)\) (with Dirichlet boundary condition) and in \(L^2(\mathbb{R})\) is that in the latter case, the optimal shape for locations of complex eigenvalues is a disc centered at the origin, whereas the bounds in the former case generally also depend on the phase of the complex eigenvalue. Their main results, which are contained in Theorems 1 and 2, generalize those of [\textit{R. L. Frank} et al., Oper. Theory: Adv. Appl. 214, 39--44 (2011; Zbl 1253.35086)], when \(\nu\neq1/2\). By the Birman-Schwinger principle and precise bounds on the Green's function of \(H_{0,\nu}\), they obtain eigenvalue bounds for \(V\in L^1(0,\infty)\) whenever \(\nu\geq0\) (which are sharp for \(V=c\delta(x-b)\)), and, by an additional application of weighted Hölder's inequality, for \(V\in L^{\gamma+d/2}((0,\infty),r^{d-1}\,dr)\) for all \(\nu>0\) and \(d\geq2(1-\gamma)\) such that \(\gamma\in(0,d/2)\). More precisely, for an eigenvalue \(\lambda=|\lambda|\mathrm{e}^{i\theta}\), their results read \[ |\lambda|^{1/2} \leq \frac{\pi}{2}\sup_{0\leq y\leq x<\infty} \left(\sqrt x |H_\nu^{(1)}(x\mathrm{e}^{i\theta/2})| \sqrt y|J_\nu(y\mathrm{e}^{i\theta/2})|\right) \int_0^\infty |V(x)|\,dx\,, \] respectively \[ |\lambda|^\gamma \leq \left(\int_0^\infty \tilde G_\nu(r,\theta)^\beta r^{-\alpha\beta} \,dr\right)^{\frac{d/2+\gamma}{\beta}}\int_0^\infty r^{d-1}|V(r)|^{\frac{d}{2}+\gamma}dr, \quad \alpha = \frac{2(d-1)}{d+2\gamma} ,\, \beta = \frac{d+2\gamma}{d+2\gamma-2} \] where \[ \tilde G_\nu(x,\theta) = \frac{\pi}{2} \max \left(\max_{y\in(0,x)}\sqrt x |H_\nu^{(1)}(x\mathrm{e}^{\theta/2})| \sqrt y |J_\nu(y\mathrm{e}^{\theta/2})|, \, \sup_{y\in(x,\infty)}\sqrt y |H_\nu^{(1)}(y\mathrm{e}^{\theta/2})| \sqrt x |J_\nu(x\mathrm{e}^{\theta/2})|\right)\,. \] Qualitatively, this agrees with the Laptev-Safronov conjecture [\textit{A. Laptev} and \textit{O. Safronov}, Commun. Math. Phys. 292, No. 1, 29--54 (2009; Zbl 1185.35045)] which was also proven in Theorems 4.1 and 4.2 in [\textit{R. L. Frank} and \textit{B. Simon}, J. Spectr. Theory 7, No. 3, 633--658 (2017; Zbl 1386.35061)] for \(-\Delta+V\) in \(L^2(\mathbb{R}^d)\) when \(V\) is radial. For \(V\in L^{\gamma+d/2}((0,\infty),r^{d-1}\,dr)\) their numerical analysis suggests that the contours \(\mathcal{S}_{d,\nu}\) of the shapes where a complex eigenvalue may be found, are contained in each other for different values of \(\nu\). More precisely, they conjecture that, as it is the case for real-valued \(V\), one has \(\mathcal{S}_{d,\nu'}\subseteq\mathcal{S}_{d,\nu}\) whenever \(\nu'>\nu\). Theorem 2 readily generalizes to \(-\Delta+V\) in \(L^2(\mathbb{R}^d)\), whenever \(V\) is a radial complex-valued potential. Assuming that the inclusion property \(\mathcal{S}_{d,\nu'}\subseteq\mathcal{S}_{d,\nu}\) for \(\nu'>\nu\) holds, their result reads \(\sigma(-\Delta+V(|x|))\subseteq\mathcal{S}_{d,N(d,0)}\), where \(N(d,\ell)=\sqrt{c_{\ell,d}+\xi(\xi-1)+1/4}\), \(c_{\ell,d}=\ell(\ell+d-2)\), and \(\xi=(d-1)/2\), i.e., \(N(d,0)=(d-2)/2\).