On theoretical upper limits for valid timesteps of implicit ODE methods

Abstract

Implicit methods for the numerical solution of initial-value problems may admit multiple solutions at any given time step. Accordingly, their nonlinear solvers may converge to any of these solutions. Below a critical timestep, exactly one of the solutions (the consistent solution) occurs on a solution branch (the principal branch) that can be continuously and monotonically continued back to zero timestep. Standard step-size control can promote convergence to consistent solutions by adjusting the timestep to maintain an error estimate below a given tolerance. However, simulations for symplectic systems or large physical systems are often run with constant timesteps and are thus more susceptible to convergence to inconsistent solutions. Because simulations cannot be reliably continued from inconsistent solutions, the critical timestep is a theoretical upper bound for valid timesteps.