Multigrid reduction-in-time convergence for advection problems: A Fourier analysis perspective

H. De Sterck,S. Friedhoff,O. A. Krzysik,Scott P. MacLachlan

2023-10-07

Abstract:A long-standing issue in the parallel-in-time community is the poor convergence of standard iterative parallel-in-time methods for hyperbolic partial differential equations (PDEs), and for advection-dominated PDEs more broadly. Here, a local Fourier analysis (LFA) convergence theory is derived for the two-level variant of the iterative parallel-in-time method of multigrid reduction-in-time (MGRIT). This closed-form theory allows for new insights into the poor convergence of MGRIT for advection-dominated PDEs when using the standard approach of rediscretizing the fine-grid problem on the coarse grid. Specifically, we show that this poor convergence arises, at least in part, from inadequate coarse-grid correction of certain smooth Fourier modes known as characteristic components, which was previously identified as causing poor convergence of classical spatial multigrid on steady-state advection-dominated PDEs. We apply this convergence theory to show that, for certain semi-Lagrangian discretizations of advection problems, MGRIT convergence using rediscretized coarse-grid operators cannot be robust with respect to CFL number or coarsening factor. A consequence of this analysis is that techniques developed for improving convergence in the spatial multigrid context can be re-purposed in the MGRIT context to develop more robust parallel-in-time solvers. This strategy has been used in recent work to great effect; here, we provide further theoretical evidence supporting the effectiveness of this approach.

Numerical Analysis

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is the poor convergence of standard iterative parallel - in - time methods (such as MGRIT and Parareal) for advection - dominated PDEs in parallel - in - time methods. Specifically, these problems include:

1. **Poor convergence in advection problems**: Traditional parallel - in - time methods show poor convergence when dealing with advection - dominated PDEs, which limits the efficiency improvement in practical applications. For diffusion - dominated PDEs, these methods can usually converge quickly and robustly and significantly accelerate the solution process; but for advection - dominated PDEs, the situation is the opposite.

2. **Insufficient coarse - grid correction**: Through local Fourier analysis (LFA), the author finds that the standard MGRIT method has convergence problems when dealing with advection - dominated PDEs because some smooth Fourier modes (called characteristic components) are insufficiently corrected on the coarse grid. This phenomenon is similar to the performance of the classical spatial multigrid method in steady - state advection - dominated PDEs.

3. **The influence of CFL number and coarsening factor**: For some advection problems with semi - Lagrangian discretization, the MGRIT method using the re - discretized coarse - grid operator cannot maintain robustness to the CFL number or the coarsening factor. This means that in these cases, the convergence of MGRIT may be seriously affected by the time and space discretization parameters.

To address these problems, the paper proposes a new theoretical framework, that is, to study the convergence of the two - level MGRIT algorithm through local Fourier analysis and reveal the root causes of these problems. In addition, the author also explores how to draw on the improvement strategies in the classical spatial multigrid method to develop more robust parallel - in - time solvers.

### Key contributions
- **Theoretical analysis**: A closed - form convergence theory is provided through local Fourier analysis, which explains the convergence problems of MGRIT in advection - dominated PDEs.
- **New insights**: It is found that insufficient coarse - grid correction is one of the main reasons for poor convergence, especially when dealing with advection - dominated PDEs.
- **Improvement directions**: It is proposed that the improvement strategies in the classical spatial multigrid method can be drawn on to improve the convergence and robustness of MGRIT in advection - dominated PDEs.

Through these analyses and insights, the paper provides a theoretical basis and support for the development of more efficient parallel - in - time solvers.