The classic Smoothed Particle Hydrodynamics (SPH) is a particle-based method commonly used to solve fluid dynamics equations. SPH is amenable to parallelization and it has been applied successfully to fluids with low viscosity. However, it can perform poorly when applied to low Reynolds number flows in complex geometry.
In this talk we present an accurate implicit implementation of SPH that can be used with unsteady low Reynolds number flows and that features good parallel scalability. The fluid dynamics equations are discretized in a Lagrangian fashion using an incremental pressure projection scheme and second-order accurate differential operators. The implementation uses the molecular dynamics library LAMMPS exploiting the common computational structure of particle methods. In our formulation of implicit SPH the solutions to Poisson and Helmholtz linear systems, required at each timestep, are performed using linear solvers provided in the Trilinos library. Using the same framework we solve the nonlinear Poisson-Bolztman (PB) equations which, in fact, reduce to a Poisson system after the application of the Newton method. We demonstrate the efficiency, accuracy and scalability of the implementation on large-scale three-dimensional simulations. Finally we show results for realistic electrokinetic applications in the micro-/nano-scale.