Stability of Calderón inverse conductivity problem in the plane (Q2465762): Difference between revisions

Let \(\Omega\) be a bounded domain in \(\mathbb C\) with connected complement. Given \(\gamma \in L^\infty(\partial \Omega)\), consider the Dirichlet to Neumann map \(\Lambda_\gamma\), defined as \(\gamma \frac{\partial u}{\partial \nu}\), in a proper generalized sense, with \(u\) being a solution of the Dirichlet problem \[ \begin{cases}\nabla \cdot (\gamma \nabla u) = 0, \\ u_{|\partial \Omega} = f, \end{cases} \tag{1} \] where \(\frac{\partial}{\partial \nu}\) is outer normal derivative. \(\Lambda_\gamma\) is thought as a linear bounded operator from \(H^{1/2}(\partial \Omega)\) to \(H^{-1/2}(\partial \Omega)\). It is known by results of K. Astala and L. Päivärinta that, in the two dimensional case, the map \(\gamma \to \Lambda_\gamma\) is injective, so that the inverse problem of reconstructing \(\gamma\) from \(\Lambda_\gamma\) has, at most, one solution. The problem of stability is of interest in applications, that is, the problem of establishing whether \(\Lambda_\gamma \to \gamma\) is continuous or not. The main result of the paper states that, if \(\gamma_1, \gamma_2\) are a priori assumed to be elements of \(C^\alpha(\overline{\Omega})\), with \(\alpha > 0\), \(\|\frac{1-\gamma_j}{1 + \gamma_j}\|_{L^\infty} < \kappa < 1\), and \(\|\gamma_j\|_{C^\alpha} < \Gamma_0\), then \[ \|\gamma_1 - \gamma_2\|_{L^\infty(\Omega)} \leq V(\|\Lambda_{\gamma_1} - \Lambda_{\gamma_2}\|_{H^{1/2}(\partial \Omega) \rightarrow H^{-1/2}(\partial \Omega))}), \] with \(V(\rho) = C \log(\rho)^{-a}\), where \(C\) and \(a\) depend on \(\kappa\), \(\alpha\) and \(\Gamma_0\). This improves the previous results of Alessandrini and Liu.