Overview
This work introduces an accelerated approach for the parareal algorithm, centered on formulating and optimizing the coarse propagator as a two-step method. The framework focuses on achieving favorable convergence properties by minimizing the convergence factor. Theoretical analysis providing a rigorous error estimate guides the design of these propagators, which is then validated through numerical experiments.
Research Context
The parareal algorithm is a method for parallelizing time integration. Its effectiveness hinges on the performance of its coarse propagator, which typically governs the algorithm's convergence rate. Existing practical coarse propagators often yield convergence factors that may limit the overall acceleration. The presented research addresses this by proposing a modified coarse propagator structure intended to improve this critical metric.
Approach
The core of the approach involves re-conceptualizing the coarse propagator within the parareal algorithm as a two-step method. This two-step formulation is then subjected to an optimization process aimed at minimizing its convergence factor. A rigorous error estimate was derived for this proposed two-step parareal algorithm. This estimate provides an explicit bound on the linear convergence factor, serving as a quantitative guideline for selecting and designing appropriate coarse propagators. To illustrate and validate this, the linear parabolic equation was considered as an example. An optimized two-step coarse propagator (O2CP) was constructed based on the insights from the error estimate. This O2CP was then integrated into numerical experiments on both linear and nonlinear parabolic equations.
Findings
- The proposed framework formulates the coarse propagator as a two-step method, optimized with respect to the convergence factor.
- A rigorous error estimate was derived for the two-step parareal algorithm, providing an explicit bound on the linear convergence factor. This estimate aids in the selection and design of coarse propagators.
- For the illustrative case of the linear parabolic equation, an Optimized Two-Step Coarse Propagator (O2CP) was constructed.
- The O2CP achieved an optimized convergence factor of approximately $0.0064$.
- This convergence factor is substantially smaller than that of commonly used practical coarse propagators within the classical parareal setting.
- The computational cost associated with the O2CP remains moderate.
- Numerical experiments conducted on linear and nonlinear parabolic equations supported the theoretical analysis.
- These experiments demonstrated rapid convergence using the two-step parareal algorithm equipped with the O2CP.