Nonhomogeneous boundary conditions for the WarpX Ohm solver

[pimanov]

Hi,

I am looking for advice on simulating plasma acceleration devices such as magnetoplasmadynamic thrusters using WarpX. Plasma acceleration is driven by an electric current injected through coaxial electrodes and entering the domain through its boundaries.

A hybrid kinetic model seems appropriate, but imposing electrode-driven current may require nonhomogeneous magnetic field boundary conditions, which I could not find in the current WarpX implementation.

Is this supported for the hybrid solver? If not, what is the recommended way to model current injection in this setup? I would be happy to contribute an implementation if guidance is available.

Do you see any alternative approaches, or know of related work using WarpX for plasma acceleration devices?

[roelof-groenewald]

Hi @pimanov. That sounds like a cool application. My initial recommendation would be to use the "external current" functionality included for the hybrid solver to model the driven current. See hybrid_pic_model.J[x/y/z]_external_grid_function(x, y, z, t) in the docs (https://warpx.readthedocs.io/en/latest/usage/parameters.html#maxwell-solver-kinetic-fluid-hybrid).

With that input you can specify a spatial and temporally dependent current that would be included in Ohm's law but is not coming from actual particles in the system.

Would that serve your needs? Do you think you would still require additional magnetic field boundary support?

[pimanov]

Thank you @roelof-groenewald! I see that hybrid_pic_model.J[x/y/z]_external_grid_function with external_vector_potential and PMC boundaries can address this. However, the pec_insulator boundary option seems more straightforward for my case. I initially missed it because it's only available in the parameters list interface, not in PICMI Python.

[JustinRayAngus]

Can you achieve what you need using the insulator_pec boundary ?

See PR #4943 and https://warpx.readthedocs.io/en/latest/usage/parameters.html#domain-boundary-conditions

[pimanov]

Thank you @JustinRayAngus, the pec_insulator option solves the problem.

In my case, it was also important to maintain a non-zero hybrid_pic_model.plasma_resistivity(rho,J). Without resistivity, the boundary conditions had no effect on the solution because there was no mechanism for information to propagate from the wall into the domain (zero characteristic wave speeds in the wall-normal direction).

[pimanov]

@JustinRayAngus Is pec_insulator currently supported in the PICMI interface, or is support planned/in progress?