A discrete Liouville identity for numerical reconstruction of Schr\"odinger potentials

TL;DR

This paper introduces a novel discrete Liouville identity approach for reconstructing Schr"odinger potentials from boundary measurements, enabling direct coarse reconstructions and improved optimization-based solutions with better convergence.

Contribution

The paper develops a discrete Liouville identity that links resistor networks to Schr"odinger potentials, enhancing inverse problem solutions with a new, effective computational method.

Findings

Discrete Schr"odinger potential provides accurate coarse reconstructions.

Reformulating the inverse problem improves convergence and stability.

Numerical results demonstrate high-quality potential reconstructions.

Abstract

We propose a discrete approach for solving an inverse problem for Schr\"odinger's equation in two dimensions, where the unknown potential is to be determined from boundary measurements of the Dirichlet to Neumann map. For absorptive potentials, and in the continuum, it is known that by using the Liouville identity we obtain an inverse conductivity problem. Its discrete analogue is to find a resistor network that matches the measurements, and is well understood. Here we show how to use a discrete Liouville identity to transform its solution to that of Schr\"odinger's problem. The discrete Schr\"odinger potential given by the discrete Liouville identity can be used to reconstruct the potential in the continuum in two ways. First, we can obtain a direct but coarse reconstruction by interpreting the values of the discrete Schr\"odinger potential as averages of the continuum Schr\"odinger potential on a special sensitivity grid. Second, the discrete Schr\"odinger potential may be used to reformulate the conventional nonlinear output least squares optimization formulation of the inverse Schr\"odinger problem. Instead of minimizing the boundary measurement misfit, we minimize the misfit between the discrete Schr\"odinger potentials. This results in a better behaved optimization problem that converges in a single Gauss-Newton iteration, and gives good quality reconstructions of the potential, as illustrated by the numerical results.