Improving the IMEX method with a residual balanced decomposition

Abstract

In numerical time-integration with implicit-explicit (IMEX) methods, a within-step adaptable decomposition called residual balanced decomposition is introduced. With this decomposition, the requirement of a small enough residual in the iterative solver can be removed, consequently, this allows to exchange stability for efficiency. This decomposition transfers any residual occurring in the implicit equation of the implicit-step into the explicit part of the decomposition. By balancing the residual, the accuracy of the local truncation error of the time-stepping method becomes independent from the accuracy by which the implicit equation is solved. In order to balance the residual, the original IMEX decomposition is adjusted after the iterative solver has been stopped. For this to work, the traditional IMEX time-stepping algorithm needs to be changed. We call this new method the shortcut-IMEX (SIMEX). SIMEX can gain computational efficiency by exploring the trade-off between the computational effort placed in solving the implicit equation and the size of the numerically stable time-step. Typically, increasing the number of solver iterations increases the largest stable step-size. Both multi-step and Runge-Kutta (RK) methods are suitable for use with SIMEX. Here, we show the efficiency of a SIMEX-RK method in overcoming parabolic stiffness by applying it to a nonlinear reaction-advection-diffusion equation. In order to define a stability region for SIMEX, a region in the complex plane is depicted by applying SIMEX to a suitable PDE model containing diffusion and dispersion. A myriad of stability regions can be reached by changing the RK tableau and the solver.