Adaptive time-stepping schemes for the solution of the Poisson-Nernst-Planck equations (Q1995976)

The authors investigate the Poisson-Nernst-Planck equations with generalized Frumkin-Butler-Volmer reaction kinetics at the electrodes. A solver that dynamically chooses the time-step size so that the approximate local truncation error is within user-specified bounds is developed. Here, the spatial discretization can be nonuniform, so that the finer mesh needed near the electrodes to resolve the boundary layers, are permitted. The adaptive time-stepping permits the solver to choose small time-steps during the initial transient period during which the boundary layers may be forming quickly in response to the boundary conditions. This is also suitable, when one studies physical situations in which the imposed voltage or current has sudden fast changes. Two adaptive time-stepping schemes are considered in this work. The semi-implicit adaptive time-stepper is shown to have a stability restriction which causes its time-step sizes to stabilize to a limiting value, in the long-time limit of constant voltage or current, whereas the fully-implicit scheme is shown to have no restriction on the time step beyond a user-specified value. However, the fully-implicit time-stepper requires a computationally expensive nonlinear solve per step, resulting in longer computation time per time step. A profile is presented where one could identify regimes of the perturbation parameter where one method is favourable over the other.