Corrected explicit-implicit domain decomposition algorithms for two-dimensional semilinear parabolic equations

Corrected explicit-implicit domain decomposition (CEIDD) algorithms are studied for parallel approximation of semilinear parabolic problems on distributed memory processors. It is natural to divide the spatial domain into some smaller parallel strips and cells using the simplest straight-line interface (SI). By using the Leray-Schauder fixed-point theorem and the discrete energy method, it is shown that the resulting CEIDD-SI algorithm is uniquely solvable, unconditionally stable and convergent. The CEIDD-SI method always suffers from the globalization of data communication when interior boundaries cross into each other inside the domain. To overcome this disadvantage, a composite interface (CI) that consists of straight segments and zigzag fractions is suggested. The corresponding CEIDD-CI algorithm is proven to be solvable, stable and convergent. Numerical experiments are presented to support the theoretical results.