Coupled Multiphysics Simulations of Charged Particle Electrophoresis for Massively Parallel Supercomputers
share
The article deals with the multiphysics simulation of electrokinetic flows. When charged particles are immersed in a fluid and are additionally subjected to electric fields, this results in a complex coupling of several physical phenomena. In a direct numerical simulation, the dynamics of moving and geometrically resolved particles, the hydrodynamics of the fluid, and the electric field must be suitably resolved and their coupling must be realized algorithmically. Here the two-relaxation-time variant of the lattice Boltzmann method is employed together with a momentum-exchange coupling to the particulate phase. For the electric field that varies in time according to the particle trajectories, a quasistatic continuum model and its discretization with finite volumes is chosen. This field is coupled to the particulate phase in the form of an acceleration due to electrostatic forces and conversely via the respective charges as boundary conditions for the electric potential equation. The electric field is also coupled to the fluid phase by modeling the effect of the ion transport on fluid motion. With the multiphysics algorithm presented in this article, the resulting multiply coupled, interacting system can be simulated efficiently on massively parallel supercomputers. This algorithm is implemented in the waLBerla framework, whose modular software structure naturally supports multiphysics simulations by allowing to flexibly combine different models. The largest simulation of the complete system reported here performs more than 70000 time steps on more than five billion (5 × 10^9) mesh cells for both the hydrodynamics, as represented by a D3Q19 lattice Boltzmann automaton, and the scalar electric field. The computations are executed in a fully scalable fashion on up to 8192 processor cores of a current supercomputer.
READ FULL TEXT