Efficient Spectral and Spectral Element Methods for Eigenvalue Problems of Schr\"{o}dinger Equations with an Inverse Square Potential

TL;DR

This paper introduces novel spectral and spectral element methods for efficiently approximating eigenvalues of Schrödinger operators with inverse square potentials, achieving exponential convergence on complex domains with singularities.

Contribution

The paper develops new spectral and spectral element techniques that incorporate singularity-aware basis functions, extending to polygonal domains with reentrant corners.

Findings

Achieves exponential convergence rates in all three stages.

Outperforms standard spectral and hp-adaptive methods in numerical experiments.

Effectively handles singularities in eigenfunctions on complex domains.

Abstract

In this article, we study numerical approximation of eigenvalue problems of the Schr\"{o}dinger operator $\displaystyle -\Delta u + \frac{c^2}{|x|^2}u$. There are three stages in our investigation: We start from a ball of any dimension, in which case the exact solution in the radial direction can be expressed by Bessel functions of fractional degrees. This knowledge helps us to design two novel spectral methods by modifying the polynomial basis to fit the singularities of the eigenfunctions. At the second stage, we move to circular sectors in the two dimensional setting. Again the radial direction can be expressed by Bessel functions of fractional degrees. Only in the tangential direction some modifications are needed from stage one. At the final stage, we extend the idea to arbitrary polygonal domains. We propose a mortar spectral element approach: a polygonal domain is decomposed into several sub-domains with each singular corner including the origin covered by a circular sector, in which origin and corner singularities are handled similarly as in the former stages, and the remaining domains are either a standard quadrilateral/triangle or a quadrilateral/triangle with a circular edge, in which the traditional polynomial based spectral method is applied. All sub-domains are linked by mortar elements (note that we may have hanging nodes). In all three stages, exponential convergence rates are achieved. Numerical experiments indicate that our new methods are superior to standard polynomial based spectral (or spectral element) methods and $hp$-adaptive methods. Our study offers a new and effective way to handle eigenvalue problems of the Schr\"{o}dinger operator including the Laplacian operator on polygonal domains with reentrant corners.