An analytical method for linear elliptic PDEs and its numerical implementation (Q596169)

The authors construct a method which uses the specific analytical information about the solution along with numerical techniques in order to solve elliptic boundary value problems with constant coefficients on polygonal domains. The key of the method is a ``global condition'' which couples specified and unknown values of the solution or its derivatives on the boundary. With respect to this condition the aim of the paper is twofold. First, the authors show that it is a necessary and sufficient condition for the existence of the solution and then they consider, in details, a discrete technique in order to solve this condition numerically. To illustrate the effectiveness of the method, it is applied to the Laplace equation on some convex polygonal domains with different boundary conditions. In spite of the fact that a rigorous proof of the convergence remains open, the numerical results are fairly close to the analytical solution.