“…For example, Morel et al [6] propose a cell-centered diffusion scheme (MDHW) for quadrilateral (logically orthogonal) grids which has two advantages over the K&P schemes: (1) If u is linear, the discretization of ∇ · ∇u does vanish and (2) the solution seems to converge with secondorder accuracy, regardless of the smoothness of the mesh. However, MDHW has the following disadvantages: (1) In addition to the cell-centered unknowns, edge-centered unknowns are introduced which triples the size of the linear system.…”