We study multigrid preconditioning of matrix-free Newton-Krylov methods as
a means of developing more efficient nonlinear iterative methods for large
scale simulation. Newton-Krylov methods have proven dependable in solving n
onlinear systems while not requiring the explicit formation or storage of t
he complete Jacobian. However, the standard algorithmic scaling of Krylov m
ethods is nonoptimal, with increasing linear system dimension. This motivat
es our use of multigrid-based preconditioning. It is demonstrated that a si
mple multigrid-based preconditioner can effectively limit the growth of Kry
lov iterations as the dimension of the linear system is increased. Differen
t performance aspects of the proposed algorithm are investigated on three n
onlinear, nonsymmetric, boundary value problems. Our goal is to develop a h
ybrid methodology which has Newton-Krylov nonlinear convergence properties
and multigrid-like linear convergence scaling for large scale simulation.