A Study of the Numerical Stability of an ImEx Scheme with Application to the Poisson-Nernst-Planck Equations

Abstract

The Poisson-Nernst-Planck equations with generalized Frumkin-Butler-Volmer boundary conditions (PNP-FBV) describe ion transport with Faradaic reactions and have applications in a wide variety of fields. Using an adaptive time-stepper based on a second-order variable step-size, semi-implicit, backward differentiation formula (VSSBDF2), we observe that when the underlying dynamics is one that would have the solutions converge to a steady state solution, the adaptive time-stepper produces solutions that "nearly" converge to the steady state and that, simultaneously, the time-step sizes stabilize at a limiting size $d{t}_{\infty }$. Linearizing the SBDF2 scheme about the steady state solution, we demonstrate that the linearized scheme is conditionally stable and that this is the cause of the adaptive time-stepper's behaviour. Mesh-refinement, as well as a study of the eigenvectors corresponding to the critical eigenvalues, demonstrate that the conditional stability is not due to a time-step constraint caused by high-frequency contributions. We study the stability domain of the linearized scheme and find that it can have corners as well as jump discontinuities.

The earlier version, arXiv:1905.01368v1, also contained: a linear stability analysis of the SBDF2 scheme, a study of the effect of Richardson Extrapolation on numerical stability, and a study of the stability domain of the logistic equation. This is the 2nd of a pair of articles: "A Variable Step Size Implicit-Explicit Scheme for the Solution of the Poisson-Nernst-Planck Equations", is on arXiV