Lightning Talk

Concurrent Coupling of Kinetic Monte Carlo and Fluctuating Hydrodynamics (SPPARKS-MUI-FHDeX)


Changho Kim
University of California, Merced
  • Wednesday, 11 Aug 2021
  • 14:09 - 14:12 EDT
  • Prerecorded Video

We develop a computational methodology to simulate multi-physics phenomena occurring at gas-solid interfaces, focusing on heterogeneous catalysis applications. In particular, we aim to construct numerical methods that can efficiently simulate computationally challenging cases where complex interactions between gas flows and chemical reactions occur at time and length scales that are much smaller than those of conventional heterogeneous catalysis systems. To this end, we propose to concurrently couple the kinetic Monte Carlo (KMC) method (for an atomistic-resolution description of chemical reactions on the catalytic surface) with the fluctuating hydrodynamics (FHD) method (for hydrodynamic transport at the mesoscopic scale). To implement the coupling method and deliver scalable computer software that can efficiently utilize emerging high-performance computer architectures, we base our development on FHDeX and SPPARKS that provide massively parallel implementations of FHD and KMC, respectively, and integrate these heterogeneous solvers via the multiscale universal interface (MUI) library. FHDeX is based on the massively parallel AMReX framework and SPPARKS includes the synchronous sublattice algorithm for parallel KMC simulations. Favorable scalability test results of MUI have been reported up to a minimum of about 10^4 MPI tasks. One distinguishing feature of our implementation is that heterogeneous solvers are compiled as separate executables and run concurrently via MPI’s multiple program-multiple data (MPMD) support. This feature will facilitate the code development effort while computational efficiency is maintained. This work is in collaboration with Dr. John Bell’s research group (developer of FHDeX, LBNL) and Dr. Yu-Hang Tang (main developer of MUI, LBNL).