Inverse wave scattering in the time domain for point scatterers

TL;DR

This paper demonstrates how to determine the locations of point scatterers in three-dimensional space using time-domain wave scattering data, accounting for multiple scattering effects and employing a MUSIC-based algorithm.

Contribution

It introduces a method to identify point scatterer locations from finite-dimensional scattering data, considering multiple scattering effects and using a factorization approach.

Findings

Locations of point scatterers can be determined from scattering data.

The method accounts for multiple scattering effects.

The approach employs a MUSIC algorithm variation.

Abstract

Let $\Delta_{\alpha,Y}$ be the bounded from above self-adjoint realization in $L^{2}({\mathbb R}^{3})$ of the Laplacian with $n$ point scatterers placed at $Y=\{y_{1},\dots,y_{n}\}\subset{\mathbb R}^{3}$, the parameters $(\alpha_{1},\dots\alpha_{n})\equiv\alpha\in {\mathbb R}^{n}$ being related to the scattering properties of the obstacles. Let $u^{\alpha,Y}_{f_{\epsilon}}$ and $u^{\varnothing}_{f_{\epsilon}}$ denote the solutions of the wave equations corresponding to $\Delta_{\alpha,Y}$ and to the free Laplacian $\Delta$ respectively, with a source term given by the pulse $f_{\epsilon}(x)=\sum_{k=1}^{N}f_{k}\,\varphi_{\epsilon}(x-x_{k}) $ supported in $\epsilon$-neighborhoods of the points in $X_{N}=\{x_{1},\dots, x_{N}\}$, $X_{N}\cap Y=\varnothing$. We show that, for any fixed $\lambda>\sup\sigma(\Delta_{\alpha,Y})$, there exits $N_{\circ}\ge 1$ such that the locations of the points in $Y$ can be determined by the knowledge of the finite-dimensional scattering data operator $F^{N}_{\lambda}:{\mathbb R}^{N}\to{\mathbb R}^{N}$, $N\ge N_{\circ}$, $$ (F^{N}_{\lambda}f)_{k}:=\lim_{\epsilon\searrow 0}\int_{0}^{\infty}e^{-\sqrt\lambda\,t}\big(u^{\alpha,Y}_{f_{\epsilon}}(t,x_{k})-u^{\varnothing}_{f_{\epsilon}}(t,x_{k})\big)\,dt\,. $$ We exploit the factorized form of the resolvent difference $(-\Delta_{\alpha,Y}+\lambda)^{-1}-(-\Delta+\lambda)^{-1}$ and a variation on the finite-dimensional factorization in the MUSIC algorithm; multiple scattering effects are not neglected.