Domain decomposition method for parabolic problems with Neumann conditions (Q858841)

Domain decomposition is a parallel algorithm to speed up the computations for solving partial differential equations. In this paper the authors consider a second order parabolic equation in two space dimensions on a Cartesian grid on the unit square. Both Dirichlet, Neumann and mixed boundary conditions are considered. The computational domain is divided into the same number of strips as the number of processes which will be used for the parallel computation. Within each strip a fully implicit scheme is used (BTCS, backward time central space). On the interfaces between the strips the \textit{modified implicit prediction} (MIP, a technique derived by the authors in a previous paper for Dirichlet boundary conditions) is used. In the test section of the paper the authors compare the spectral radii, the number of iterations and the speed up when the problem is solved with different numbers of processors.