A potential function method for the Cauchy problem of elliptic operators (Q441963)

The authors consider the ill-posed problem to find the solution of a Helmholtz equation in \(\mathbb{R}^d\), \(d=2,3\), where Dirichlet and Neumann data are given on a part of the boundary. Assuming that the Dirichlet data, Neumann data and the solution are in corresponding \(H^1\), \(L^2\), \(H^{3/2}\) spaces, respectively, they prove that these three functions can be approximated to any given accuracy by single-layer functions. This leads to the necessity to solve two first-kind integral equations which is done using Tikhonov regularization with Morozov selection of the regularization parameter (assuming that a measure of the noise of the boundary data is given). Then they prove that this leads to a stable solution and estimate its accuracy. Finally, they report on a series of numerical experiments (where they don't define two important symbols, \(\Omega\) and \(\Sigma_1\)) which show the viability of the approach if the noise of the boundary data is rather small.