Uniform convergence of V-cycle multigrid algorithms for two-dimensional fractional Feynman-Kac equation (Q1742680)

Convergence for the V-cycle multigrid algorithm for the fractional Feynman-Kac equation with traditional restriction and prolongation operators corresponding to the Galerkin approach is shown, based on works like the work of \textit{J. H. Bramble} and \textit{J. E. Pasciak} [Math. Comput. 49, 311--329 (1987; Zbl 0659.65098)]. For the smoothing, only damped Jacobi iteration are considered. From the two conditions for uniform convergence (one for the preconditioner in the smoothing iteration and one for the coarse grid operator) the second one is emplaced by a new condition (circumventing estimations of the coarse grid equations). Then, for sufficiently small \(h\) and separately for the one- and two-dimensional case (where properties of the involved symmetric positive definite Toeplitz block tridiagonal matrices are employed), convergence estimates are proved which depend only on constants in the two basic assumptions and on the number of smoothing steps. In their numerical experiments, along with the theoretically investigated Galerkin approach, the geometrical coarsening (by substituting \(2h\) for \(h\) to get the coarse grid equations) is also considered. The comparison shows tiny advantage for the doubling approach when looking at the required solution times.