Strategy for mixed periodicity: Fully periodic Velocity vs. Inflow/Outflow Passive Scalars

[tuanliu wenxin]

Dear Nek5000 Community,

I am currently setting up a simulation involving turbulent flow with passive scalar transport (pollutant dispersion) using a mesh generated by Gmsh and imported via gmsh2nek.

My Physical Configuration:

• Velocity Field: I require periodicity in both the streamwise ($x$) and spanwise ($y$) directions to simulate a fully developed infinite domain.

• Passive Scalar Fields: I require periodicity in the spanwise ($y$) direction, but Inflow/Outflow conditions in the streamwise ($x$) direction. (i.e., pollutants exiting the outlet should not re-enter the inlet).

The Problem:

I am facing a dilemma regarding the mesh topology setup in gmsh2nek:

1. If I define the streamwise periodic pairs in gmsh2nek (setting physical connectivity), the mesh topology becomes periodic. Consequently, the passive scalars inherit this periodicity, causing pollutants to recycle back to the inlet, which violates my physical model.

2. If I do not define the periodic pairs in gmsh2nek (leaving the mesh topologically disconnected), I can easily set Inflow/Outflow BCs for the scalars. However, I am unsure how to properly enforce high-order periodic boundary conditions for the velocity field in this configuration without the underlying topological connectivity.

My Question:

Is there a standard way in Nek5000 to decouple the periodicity of the velocity field from the scalar fields?

Specifically:

• Should I use a fully periodic mesh (defined in gmsh2nek) and apply a Fringe Region (Sponge Layer) with a sink term in userq to damp the scalars to zero before they recycle?

• Or is there a method in usrdat2 or userbc to enforce "pseudo-periodicity" for velocity on a topologically open mesh while keeping scalars as standard Dirichlet/Outflow?

Any advice or best practices for this "mixed topology" setup would be greatly appreciated.

Thank you for your time.

Best regards