Multigrid Methods for Constrained Minimization Problems and Application to Saddle Point Problems
For researchers in numerical analysis and scientific computing, this work provides a theoretically grounded multigrid solver for saddle point problems with uniform convergence guarantees, though it is an incremental extension of existing subspace optimization methods.
This paper develops a multigrid method for constrained minimization problems, reformulated as saddle point problems, using a multiplicative Schwarz smoother. It achieves uniform convergence for mixed finite element methods applied to Poisson and Stokes equations without requiring full regularity of the underlying PDEs.
The first order condition of the constrained minimization problem leads to a saddle point problem. A multigrid method using a multiplicative Schwarz smoother for saddle point problems can thus be interpreted as a successive subspace optimization method based on a multilevel decomposition of the constraint space. Convergence theory is developed for successive subspace optimization methods based on two assumptions on the space decomposition: stable decomposition and strengthened Cauchy-Schwarz inequality, and successfully applied to the saddle point systems arising from mixed finite element methods for Poisson and Stokes equations. Uniform convergence is obtained without the full regularity assumption of the underlying partial differential equations. As a byproduct, a V-cycle multigrid method for non-conforming finite elements is developed and proved to be uniform convergent with even one smoothing step.