The cascadic multigrid method for elliptic problems (Q1358121)

The numerical method discussed here may be viewed as multigrid without correction cycles. Consequently, more iterates are needed on the coarse grids than would be used in traditional multigrid methods, and the stopping criterion on each grid is one of the issues faced in this paper. The authors obtain estimates of the rate of convergence and of the computational complexity for their method, with conjugate gradient and with symmetric Gauss-Seidel smoothers. They also consider an adaptive version, with the gridding dependent on a dynamic estimate of the smoothness of the solution. Example computations illustrate the effectiveness of the adaptive method for the Laplace equation on a domain with a re-entrant corner, specifically, a square with a slit.