Multigrid Methods in Lattice Field Computations

Abstract

The multigrid methodology is reviewed. By integrating numerical processes at all scales of a problem, it seeks to perform various computational tasks at a cost that rises as slowly as possible as a function of $n$, the number of degrees of freedom in the problem. Current and potential benefits for lattice field computations are outlined. They include: $O(n)$ solution of Dirac equations; just $O(1)$ operations in updating the solution (upon any local change of data, including the gauge field); similar efficiency in gauge fixing and updating; $O(1)$ operations in updating the inverse matrix and in calculating the change in the logarithm of its determinant; $O(n)$ operations per producing each independent configuration in statistical simulations (eliminating CSD), and, more important, effectively just $O(1)$ operations per each independent measurement (eliminating the volume factor as well). These potential capabilities have been demonstrated on simple model problems. Extensions to real life are explored.

44

Cited by in corpus (7)

• Stability of disk galaxies in the modified dynamics
• The Microscopic Representation of Complex Macroscopic Phenomena (Critical Slowing Down - A Blessing in Disguise)
• Kinematics of Multigrid Monte Carlo
• Linear Algebra with Disordered Sparse Matrices that have Spatial Structure: Theory and Computation
• Galerkin Formulation of Path Integrals in Lattice Field Theory
• Error analysis of coarse-grained kinetic Monte Carlo method
• Multilevel Monte Carlo method for simulations of fluids