FEM-DtN-SIM Method for Computing Resonances of Schrödinger Operators

Bo Gong 1 , Takumi Sato 2 , Jiguang Sun 3 , Xinming Wu 2 · 2025

Read FEM-DtN-SIM Method for Computing Resonances of Schrödinger Operators by Bo Gong 1 , Takumi Sato 2 , Jiguang Sun 3 , Xinming Wu 2 (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-11680.txt.

Introduction

Bo Gong and co-authors offer a new way to compute resonances of the Schrödinger operator, in this 2025 paper. Resonances mark states that linger for a while before fading, a key idea in quantum physics. Theory has mapped where they sit and how they spread out. Yet computing them past one dimension stays rare, since the problem is nonlinear and the space runs on without bound.

Their method joins three tools: finite elements, a boundary map, and a spectral indicator scheme. The boundary map is set on a cut-off edge of the working region to mimic open space. Finite elements handle tricky potential shapes with ease. Together the parts pick out the complex values that mark each resonance.

Catalog record

CREATORBo Gong 1 , Takumi Sato 2 , Jiguang Sun 3 , Xinming Wu 2
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE20 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-11680

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

FEM-DtN-SIM Method for Computing Resonances of Schrödinger Operators Bo Gong 1 , Takumi Sato 2 , Jiguang Sun 3 , Xinming Wu 2 Abstract The study of resonances of the Schrödinger operator has a long-standing tradition in mathematical physics. Extensive theoretical investigations have explored the proximity of resonances to the real axis, their distribution, and bounds on the counting functions. However, computational results beyond one dimension remain scarce due to the nonlinearity of the problem and the unbounded nature of the domain. We propose a novel approach that integrates finite elements, Dirichlet-to-Neumann (DtN) mapping, and the spectral indicator method. The DtN mapping, imposed on the boundary of a truncated computational domain, enforces the outgoing condition. Finite elements allow for the efficient handling of complicated potential functions. The spectral indicator method effectively computes (complex) eigenvalues of the resulting nonlinear algebraic system without introducing spectral pollution. The viability of this approach is demonstrated through a range of numerical examples. I 1. Introduction Resonances of Schrödinger operators have been a classical topic in mathematics and physics, generalizing bound states for systems where energy can scatter to infinity. They are characterized as poles of the meromorphic continuation of the resolvent operator, with their real parts representing oscillation frequencies and their imaginary parts corresponding to decay rates DyatlovZworski2019 . Extensive studies have investigated the proximity of resonances to the real axis, resonance-free regions, bounds, and counting functions, etc. BrietCombesDuclos1987JMAA ; DyatlovZworski2019 ; LiYaoZhao2024SIMA . The study of scattering resonances has important applications in various fields, including gravitational wave signatures, photonics, and plasmonic scattering Maakitalo2014 ; JMS2022PRL ; BBCGB2024PRB . In contrast to the extensive literature on the analysis of resonances, relatively few studies address the computation of resonances of Schrödinger operators beyond the one-dimensional case BindelZwordki2006 ; BMR2023 . The problem becomes significantly more challenging in higher dimensions. When restricting the computation to a bounded domain, it is crucial to impose the outgoing condition correctly. Additionally, the problem exhibits a nonlinear dependence on the eigen-parameter. In one dimension, the outgoing condition results in a quadratic eigenvalue problem, which can be linearized and efficiently solved using standard numerical linear algebra techniques BindelZwordki2006 . This linearization is no longer feasible in higher dimensions. Moreover, formulations that employ absorbing boundary conditions, such as the perfectly matched layer (PML) or impedance boundary conditions, are not equivalent to the original problem. As a result, they can introduce significant spectral pollution or even yield entirely incorrect resonance computations. In this paper, we study the computation of resonances of the Schrödinger operator with a bounded (complex) potential of compact support. By reformulating the problem on a truncated domain and incorporating the Dirichlet-to-Neumann (DtN) mapping to enforce the outgoing condition, we obtain an eigenvalue problem for a holomorphic operator-valued function. We then propose a finite element method to solve this nonlinear system, where the computed eigenvalues correspond to the resonances Hsiao2011 ; SunZhou2017 . The discrete eigenvalue problem is computed using a parallel multistep spectral indicator method based on contour integrals XiSun2023 . We provide numerical examples for various potentials to validate the proposed method. While our focus is on two-dimensional cases, the scheme is also applicable to three-dimensional problems. The computational framework developed here serves as a quantitative tool for studying Schrödinger resonances. Notably, it has also been successfully applied to compute scattering resonances for acoustic obstacles and metallic grating structures XiGongSun2024 ; XiLinSun2024CMA . The convergence of the proposed method can be established using abstract approximation theory for eigenvalue problems of holomorphic operator functions, following the same approach as in the analysis of scattering resonances for acoustic obstacles karma1996a ; XiGongSun2024 . Specifically, the computed resonances converge to the true resonances at a certain rate as the mesh size h\to 0 , which is further demonstrated through numerical experiments. The remainder of the paper is organized as follows. In Section 2, we introduce the concept of resonances of the Schrödinger operator. In Section 3, we truncate the infinite domain using a circle and impose the Dirichlet-to-Neumann (DtN) mapping on the boundary. Section 4 contains the finite element formulation and the definition of a holomorphic operator-valued function whose eigenvalues correspond to the resonances. Finally, in Section 5, we provide numerical examples to validate the proposed approach. II 2. Schrödinger Resonances Let V(x)\in L^{\infty}(D) be a potential function with compact support D\subset\mathbb{R}^{2} . The Schrödinger operator is defined as H_{V}:=-\triangle+V(x). (1) The resonance problem for the Schrödinger operator is to find k\in\mathbb{C} and a nontrivial outgoing solution u to the Schrödinger equation H_{V}u(x)-k^{2}u(x)=0,\quad x\in\mathbb{R}^{2}. (2) For any k\in\mathbb{C} , an outgoing solution u to the Schrödinger equation H_{V}u(x)-k^{2}u(x)=f satisfies Zworski1999NAMS u(x)=\sum_{n=-\infty}^{\infty}a_{n}H_{n}^{(1)}(kr)e^{in\theta},\ \ \ |x|>r_{0}, (3) where \theta=\arg(x) , r_{0}>0 is a constant large enough, and H_{n}^{(1)}(\cdot) is the first kind Hankel function of order n . The resonances are the poles of the resolvent operator R(k):=(-\triangle+V-k^{2})^{-1} . Note that R(k) is analytic for {\rm Im}k\gg 0 and can be continued as a meromorphic function on \mathbb{C} . Here we are interested in the numerical computation of the resonances k ’s in the lower half complex plane. When D is a disk with radius r_{0} and V(x)=V_{0}\geq 0 , the resonances can be derived analytically. The outgoing solution u can be written in the form u(x)=\left\{\begin{array}[]{ll}\sum_{n=-\infty}^{\infty}a_{n}H_{n}^{(1)}(kr)e^{in\theta},&r=|x|>r_{0},\\ \sum_{n=-\infty}^{\infty}b_{n}J_{n}(\sqrt{k^{2}-V_{0}}r)e^{in\theta},&r=|x|\leq r_{0}.\end{array}\right. The continuity of u and its derivative at the boundary of D leads to \displaystyle a_{n}H_{n}^{(1)}(kr)\Big|_{r=r_{0}} \displaystyle= \displaystyle b_{n}J_{n}(\sqrt{k^{2}-V_{0}}r)\Big|_{r=r_{0}}, \displaystyle a_{n}\frac{\partial H_{n}^{(1)}(kr)}{\partial r}\Big|_{r=r_{0}} \displaystyle= \displaystyle b_{n}\frac{\partial J_{n}(\sqrt{k^{2}-V_{0}}r)}{\partial r}\Big|_{r=r_{0}}. The above equations can be written as a linear system \begin{pmatrix}H_{n}^{(1)}(kr_{0})&-J_{n}(\sqrt{k^{2}-V_{0}}r_{0})\\ kH_{n+1}^{(1)}(kr_{0})&-\sqrt{k^{2}-V_{0}}J_{n+1}(\sqrt{k^{2}-V_{0}}r_{0})\end{pmatrix}\begin{pmatrix}a_{n}\\ b_{n}\end{pmatrix}=\begin{pmatrix}0\\ 0\end{pmatrix}. The scattering resonances are k ’s such that \displaystyle\sqrt{k^{2}-V_{0}}J_{n+1}\left(\sqrt{k^{2}-V_{0}}r_{0}\right)H_{n}^{(1)}(kr_{0}) \displaystyle\qquad\qquad\qquad-kH_{n+1}^{(1)}(kr_{0})J_{n}\left(\sqrt{k^{2}-V_{0}}r_{0}\right)=0. If r_{0}=1 , one has that \displaystyle d_{n}(k) \displaystyle:= \displaystyle\sqrt{k^{2}-V_{0}}J_{n+1}\left(\sqrt{k^{2}-V_{0}}\right)H_{n}^{(1)}(k) \displaystyle\qquad-kH_{n+1}^{(1)}(k)J_{n}\left(\sqrt{k^{2}-V_{0}}\right). The contours of the absolute values of the above determinants d_{n}(k),n=0,\ldots,10 are shown in Fig. 1 . The minima (zeros) indicate the locations of the exact resonances. Figure 1: Contour plots of |d_{n}(k)| , n=0,\ldots,10 ( r_{0}=1,V(x)=2 ). III 3. Dirichlet-to-Neumann Mapping Since the problem is posed on the infinite domain \mathbb{R}^{2} , we first reformulate it as an equivalent problem on a bounded domain to facilitate finite element computation. Let \Gamma_{R} be a circle centered at the origin with radius R and D inside. Denote by \Omega the disk such that \partial\Omega=\Gamma_{R} . We need to impose a suitable boundary condition on \Gamma_{R} . To this end, we employ the DtN mapping on \Gamma_{R} . Let H^{1}(\Omega) be the Sobolev space of integrable functions defined on \Omega with integrable gradients. Let H^{1/2}(\Gamma_{R}) be the trace space of H^{1}(\Omega) and H^{-1/2}(\Gamma_{R}) be its dual space. Following Hsiao2011 , the DtN operator T(k):H^{1/2}(\Gamma_{R})\rightarrow H^{-1/2}(\Gamma_{R}) is defined as T(k)\varphi=\sum_{n=0}^{+\infty}{\!}{}^{{}^{\,\scriptstyle{\prime}}}\frac{k}{\pi}\frac{{H^{(1)}_{n}}^{\prime}(kR)}{H^{(1)}_{n}(kR)}\int_{0}^{2\pi}\varphi(\phi)\cos(n(\theta-\phi))\text{d}\phi, where the summation with a prime ′ means that the zero-th term is factored by 1/2 . The operator T(k) is bounded. For real number s\geq 1/2 , T(k):H^{s}(\Gamma_{R})\rightarrow H^{s-1}(\Gamma_{R}) , satisfies \|T(k)\varphi\|_{H^{s-1}(\Gamma_{R})}\leq C\|\varphi\|_{H^{s}(\Gamma_{R})}, (4) where C>0 is a constant depending on kR but not \varphi for almost all k\in\mathbb{C} . Using T(k) , we obtain a problem defined on the bounded domain \Omega , which is equivalent to ( 2 ): Find k\in\mathbb{C} and nontrivial u\in H^{1}(\Omega) such that \left\{\begin{array}[]{rlll}-\Delta u+V(x)u-k^{2}u&=&0&\rm{in}\ \Omega,\\ \frac{\partial u}{\partial\nu}&=&T(k)u&\rm{on}\ \Gamma_{R}.\end{array}\right. (5) The variational formulation for ( 5 ) is to find k\in\mathbb{C} and u\in H^{1}(\Omega) such that (\nabla u,\nabla v)+(V(x)u,v)-k^{2}(u,v)-\langle

Read the complete text