Counterexamples to uniqueness in the inverse fractional conductivity problem with partial data

TL;DR

This paper constructs explicit counterexamples showing non-uniqueness in the inverse fractional conductivity problem with partial data, highlighting limitations of current methods in recovering conductivities from exterior measurements.

Contribution

The authors provide the first explicit counterexamples for the partial data inverse fractional conductivity problem on general bounded domains, demonstrating non-uniqueness.

Findings

Counterexamples exist for all dimensions and general bounded domains.

Equal partial exterior Dirichlet-to-Neumann maps do not guarantee unique conductivities.

Counterexamples also constructed on specific domains under certain conditions.

Abstract

We construct counterexamples for the partial data inverse problem for the fractional conductivity equation in all dimensions on general bounded open sets. In particular, we show that for any bounded domain $\Omega \subset \mathbb{R}^n$ and any disjoint open sets $W_1,W_2 \Subset \mathbb{R}^n \setminus \bar{\Omega}$ there always exist two positive, bounded, smooth, conductivities $\gamma_1,\gamma_2$, $\gamma_1 \neq \gamma_2$, with equal partial exterior Dirichlet-to-Neumann maps $\Lambda_{\gamma_1}f|_{W_2} = \Lambda_{\gamma_2}f|_{W_2}$ for all $f \in C_c^\infty(W_1)$. The proof uses the characterization of equal exterior data from another work of the authors in combination with the maximum principle of fractional Laplacians. The main technical difficulty arises from the requirement that the conductivities should be strictly positive and have a special regularity property $\gamma_i^{1/2}-1 \in H^{2s,\frac{n}{2s}}(\mathbb{R}^n)$ for $i=1,2$. We also provide counterexamples on domains that are bounded in one direction when $n \geq 4$ or $s \in (0,n/4]$ when $n=2,3$ using a modification of the argument on bounded domains.