A coupled finite-volume solver for numerical simulation of electrically-driven flows

TL;DR

This paper introduces a coupled finite-volume solver for electrically-driven flows that improves accuracy and convergence in transient and steady-state simulations, demonstrating significant speedups over traditional segregated methods.

Contribution

The work develops a generic coupled solver framework within OpenFOAM for electrically-driven flows, enhancing stability and efficiency compared to segregated solvers.

Findings

Coupled solvers improve transient simulation accuracy.

Coupled solvers enable larger time-steps without divergence.

Overall speedup factor up to ~100 with semi-coupled solvers.

Abstract

The accuracy and stability of implicit CFD codes are frequently impaired by the decoupling between variables, which can ultimately lead to numerical divergence. Coupled solvers, which solve all the governing equations simultaneously, have the potential to fix this problem. In this work, we report the implementation of coupled solvers for transient and steady-state electrically-driven flow simulations in the finite-volumes framework. The numerical method, developed in OpenFOAM, is generic for Newtonian and viscoelastic fluids and is formulated for the Poisson-Nernst-Planck and Poisson-Boltzmann models. The resulting coupled systems of equations are solved efficiently with PETSc library. The performance of the coupled solvers is assessed in two test cases: induced-charge electroosmosis of a Newtonian fluid around a cylinder; electroosmotic flow of a PTT viscoelastic fluid in a contraction/expansion microchannel. The coupled solvers are more accurate in transient simulations and allow the use of larger time-steps without numerical divergence. For steady-state simulations, the coupled solvers converge in fewer iterations than segregated solvers. Although coupled solvers are much slower in a per time-step basis, the overall speedup factor obtained in this study reached a maximum value of ~100, where the highest factors have been obtained with semi-coupled solvers, which drop some coupling terms between equations. While further research is needed to improve the efficiency of the matrix solving stage, coupled solvers are already superior to segregated solvers in a number of cases.