Creating internal boundary where particles that hit it are removed from simulation

[Poto45]

Hello, I'm running dielectric wakefield simulations where the dielectric is an enclosed rectangular tube. I have the epsilon function for the dielectric and a "conductor" layer surrounding the dielectric using the sigma function. One issue I'm running into is that the electrons' y locations extend "into" the dielectric. I've been searching through for a scraping function, but am only finding functions that are for the external boundary of the simulation (like the outside surface of the dielectric). Is there a function I can use where I can manually enter a boundary surface at the inside surface of the dielectric that deletes any particles that hit it?

Also, I've been having issues with the epsilon and sigma functions such that it cannot be described in z, which is the direction of the moving window I'm using. For example, if I want the dielectric to go from 10cm to 160 cm instead of 0cm to 150cm, warpx just doesn't read the string correctly and there ends up being no dielectric anywhere. If people use a moving window and a function in all three coordinates, would you be able to share it with me? I'd really like to use a moving window with epsilon and sigma functions being dependent on x, y, and z. Thanks!

[ax3l]

I've been having issues with the epsilon and sigma functions such that it cannot be described in z,

Thanks for the reproducer, let us debug this in #4952

[RemiLehe]

Regarding the issue with the dielectric, I think one solution would be for you to implement a quick hack of WarpX:

• In your input script, place the "conductor" boundary at the inside surface of the dielectric. (This will make sure that particles are removed at this exact location and cannot propagate inside the dielectric.)
• In WarpX, remove the line below and recompile the code. This will make sure that the field solver ignores this "conductor" boundary and uses the dielectric instead:
  https://github.com/ECP-WarpX/WarpX/blob/development/Source/FieldSolver/FiniteDifferenceSolver/MacroscopicEvolveE.cpp#L177-L180
  https://github.com/ECP-WarpX/WarpX/blob/development/Source/FieldSolver/FiniteDifferenceSolver/MacroscopicEvolveE.cpp#L197-L203
  https://github.com/ECP-WarpX/WarpX/blob/development/Source/FieldSolver/FiniteDifferenceSolver/MacroscopicEvolveE.cpp#L221-L224

[ax3l]

Hi @Poto45, @RemiLehe and I discussed the two options we spoke about at IPAC for the particle absorption at the dielectric. We think using the EBs with the small modification to not let them act on fields (but just on particles) in the source code is the quickest way to success. After #4865 is merged, we could even make this an input parameters

Alternatively, you can use a regular simulation and no EBs, by adding the particle removal logic from Python. For this, you just need to add a callback once per step that removes particles that go out of bounds. In the callback, loop over the particles tiles
https://warpx.readthedocs.io/en/latest/usage/workflows/python_extend.html#particles
and create a short array mask that stores the particles to mark invalid, e.g.

x = soa.real[0]
y = soa.real[1]
mask = np.abs(x) > 1e-6 and np.abs(y) > 1e-6  # for 2mu box geometry, adjust as needed

Then in the same inner loop, set soa.idcpu[mask] = 16777216 which is this value (sorry, will provide a Python constant or function for that) for the particles that are out of bounds according to their positions. This will mark them for removal in the step.

Complete example:

from pywarpx import picmi
from pywarpx.callbacks import callfromafterstep

# Preparation: set up the simulation
#   sim = picmi.Simulation(...)
#   ...

@callfromafterstep
def my_after_step_callback():
    warpx = sim.extension.warpx
    Config = sim.extension.Config
    if Config.have_gpu:
        import cupy as xp
    else:
        import numpy as xp

    # data access
    multi_pc = warpx.multi_particle_container()
    pc = multi_pc.get_particle_container_from_name("electrons")

    # iterate over mesh-refinement levels
    for lvl in range(pc.finest_level + 1):
        # get every local chunk of particles
        for pti in pc.iterator(pc, level=lvl):
            soa = pti.soa().to_xp()

            x = soa.real[0]
            y = soa.real[1]

            # mark particles that reach the dielectric
            mask = xp.abs(x) > 1e-6 and xp.abs(y) > 1e-6  # for 2mu box geometry, adjust as needed

            # remove particles that are out-of-bounds
            soa.idcpu[mask] = 16777216


sim.step(nsteps=100)

[Poto45]

Great! Thank you for that! I'll look into running it within python and try your suggestion. It was great meeting you at IPAC!