Surface Diffusion Parameters 

#include <process/psSurfaceDiffusion.hpp>

SurfaceDiffusionParameters configure the graph-based surface diffusion solver used by process models that expose surface diffusion coefficients.

Surface diffusion is not a standalone pre-built ProcessModel. It is a process-level capability used by compatible surface models, and it can also be used directly through SurfaceDiffusionStencil and SurfaceDiffusionSolver.

Process Parameters

The parameters configure graph construction for process-integrated surface diffusion. They are set through the standard Process::setParameters(...) interface.

Field Type Default Description
stabilityFactor double 1.0 Factor used by process-integrated diffusion when choosing stable explicit substeps.
kNeighbors int 16 Number of nearest neighbors considered when radius <= 0.
radius double 0.0 Radius-based neighbor search cutoff. 0.0 uses kNeighbors instead.
normalCutoff double 0.25 Minimum normal dot product allowed between neighboring points.
sigmaNormal double 0.35 Controls how strongly normal differences reduce diffusion weights.
normalizeByLocalScale bool true Scale weights by the local point spacing.
symmetrizeWeights bool true Build symmetric graph weights from the directed neighbor graph.

Example Usage

C++

 
using namespace viennaps;

SurfaceDiffusionParameters diffusion;
diffusion.kNeighbors = 16;
diffusion.radius = 0.0;
diffusion.normalCutoff = 0.25;

Process<double, 3> process(domain, model, processTime);
process.setParameters(diffusion);
process.apply();

Python

 
import viennaps as vps

diffusion = vps.SurfaceDiffusionParameters()
diffusion.kNeighbors = 16
diffusion.radius = 0.0
diffusion.normalCutoff = 0.25

process = vps.Process(domain, model, process_time)
process.setParameters(diffusion)
process.apply()

Standalone Solver

ViennaPS also exposes the underlying graph solver directly. It builds a neighborhood graph from surface point positions and normals, then applies a graph Laplacian to diffuse scalar data along the surface while reducing coupling across sharp changes in surface normal.

The checked-in example computes a particle flux on a trench geometry, builds a PointCloud from the flux mesh nodes and normals, and repeatedly applies SurfaceDiffusionSolver::stepExplicit(...) to smooth the particleFlux field.

PointCloud<double> cloud;
cloud.positions = flux->getNodes();
cloud.normals = *flux->getNormals();

SurfaceDiffusionParameters params;
params.kNeighbors = 16;

SurfaceDiffusionSolver<double> solver(
    SurfaceDiffusionStencil<double>(cloud, params));

auto currentFlux = *flux->getCellData().getScalarData("particleFlux");
currentFlux = solver.stepExplicit(currentFlux, 1e-3, 1.0);

Practical Notes

  • Use radius when a physical cutoff is known; otherwise use kNeighbors.
  • Increase kNeighbors or radius if the graph is too disconnected.
  • Raise normalCutoff to reduce diffusion across corners and steep surface changes.
  • SurfaceDiffusionParameters only affect process-integrated diffusion when the selected process model provides surface diffusion coefficients.