Analysis of a new dimension-wise splitting iteration with selective relaxation for saddle point problems (Q291892)

The authors study steady incompressible nonlinear Navier-Stokes equations in an open bounded domain in the plane or in the three-dimensional space. The nonlinearity is due to the transport term in the operator. A frequently used strategy is to perform a relaxation for the nonlinear term. After discretizing the resulting equations using a finite element or a finite difference method, one arrives at a system of linear equations whose coefficient matrix is a \(3 \times 3\) sparse block matrix. This is a typical saddle point problem that arises in many applications of numerical methods in physics and engineering. The authors present a new solution technique called dimension-wise splitting with selective relaxation (DSSR) method, which has the advantage that it could be used both as an iterative solver and as a preconditioner for a Krylov method. It is shown that the DSSR method converges unconditionally for any positive relaxation parameter over the plane. Fourier analysis is employed to optimize the relaxation parameter, whose optimal choice is shown to be inversely proportional to the viscosity coefficient. The results are extended to the case of three-dimensional Navier-Stokes and Oseen equations. Numerical comparisons between the DSSR method and the relaxed dimensional factorization preconditioner are presented.