Assignment of fields from particles to mesh

TL;DR

This paper evaluates various field assignment methods from particles to mesh in CFD, emphasizing preservation of information and stability, and finds that symmetric mass assignment methods outperform others in accuracy and robustness.

Contribution

It introduces the concept of information preservation in particle-to-mesh assignment and compares multiple procedures, highlighting the superiority of symmetric mass assignment methods.

Findings

Symmetric mass assignment methods outperform others.

Mass assignments ensure invariance of field values after interpolation and assignment.

The methods are tested on 1D, 2D, and CFD scenarios with positive results.

Abstract

In Computational Fluid Dynamics there have been many attempts to combine the power of a fixed mesh on which to carry out spatial calculations with that of a set of particles that moves following the velocity field. These ideas indeed go back to Particle-in-Cell methods, proposed about 60 years ago. Of course, some procedure is needed to transfer field information between particles and mesh. There are many possible choices for this "assignment", or "projection". Several requirements may guide this choice. Two well-known ones are conservativity and stability, which apply to volume integrals of the fields. An additional one is here considered: preservation of information. This means that mesh interpolation, followed by mesh assignment, should leave the field values invariant. The resulting methods are termed "mass" assignments due to their strong similarities with the Finite Element Method. We test several procedures, including the well-known FLIP, on three scenarios: simple 1D convection, 2D convection of Zalesak's disk, and a CFD simulation of the Taylor-Green periodic vortex sheet. The most symmetric mass assignment is seen to be clearly superior to other methods.