Iterative reduction of the Dirichlet problem to the Neumann problem (Q4940479)

Let \(Q\) be a domain in \(\mathbb{R}^{m}\) and let \(\Gamma \subset \partial Q\) be an open part of its boundary. Consider the boundary value problem \[ (1) \quad Au=f\quad \text{in } Q; \quad \quad (2) \quad u|_{\Gamma }=0, \] where \(A\) is a second-order differential operator involving necessary initial and (or) boundary conditions on \(\partial Q\setminus \Gamma \). An iteration procedure for reduction of problem (1), (2) to a Neumann problem is suggested. The main idea is as follows. Let \(D\) be a differential operator of first order on \(\Gamma \). Instead of (2) consider the boundary condition \[ (1-\varepsilon)Du|_{\Gamma }+\varepsilon \alpha u|_{\Gamma }=(1-\varepsilon)\psi \tag{3} \] with \(\varepsilon \in[0,1]\), \(\alpha \neq 0\), \(\psi :\Gamma \to {\mathbb{R}}\). One can represent a solution to problem (1), (3) in the form \(u=\sum_{k=0}^{\infty }\varepsilon ^{k}v_{k}\). The function \(u_{N}=\sum_{k=0}^{N}v_{k}\), which satisfies \[ Au_{N}=f, Du_{N}|_{\Gamma }=Du_{N-1}|_{\Gamma }-\alpha u_{N-1}|_{\Gamma }, \quad N\geq 1, \] is treated as an approximate solution to (1), (2), and can be obtained by successive solving of \(N+1\) Neumann problems. Estimates of the convergence rate for the suggested iteration procedure are established in two cases: 1) elliptic Dirichlet boundary value problem; 2) boundary value problem for a parabolic equation.