New Gkeyll users for plasma acceleration?

[MaxThevenet]

Dear Gkeyll developers,

I am a researcher in plasma wakefield acceleration, one of the developers of the particle-in-cell (PIC) code WarpX https://ecp-warpx.github.io, and we are looking for an open-source plasma simulation code for several problems of interest to us, and have a few questions to see if Gkeyll could be the way to go. While PIC works like a charm to simulate plasma acceleration below a picosecond, we are currently interested in longer-term plasma evolution (up to a microsecond), where other models (in particular two-fluid) could provide a good description.

The image below shows what we are interested in: a high-energy particle beam perturbs an initially uniform neutral plasma, and drives an electron plasma wave (the "wake") in which gigantic accelerating fields can be used to accelerate another particle beam located in the first bubble behind the driver (the rightmost electron bucket in blue). We are currently interested in how this plasma wave relaxes long after the driver's passage. The important features for this process would be:

• The problem is axisymmetric, so an RZ description would be ideal.
• Collisions are likely to play a role long after the driver's passage, in particular electron-ion collisions.
• The background plasma is pre-ionized, with an ionization fraction of 1-10%, so collisional ionization is probably important.

Additional comments:

• The figure is in normalised units. In SI units, the beam would be ~10 microns wide, ~50 microns long, and the plasma density would be on the order 1.e15-1.e17 cm-3.
• Just modelling Hydrogen would be extremely helpful.
• The actual setup has some transverse boundaries, but simulating an infinitely-wide plasma would be a great first step.

Do you think Gkeyll could help for our problem?

Thanks in advance!

A high-energy, high-current electron beam (red) propagates (from left to right) at the speed of light through a neutral plasma and perturbs the electron density (blue), driving a non-linear wake. Ions are immobile on this short time scale, and start moving later on.

This discussion could also be interesting to @GregoryBoyle @SeverinDiederichs @MathisMewes.

[JunoRavin]

Thank you for your interest in Gkeyll! Based on your description of the problem there would be a few challenges in simulating this problem with the two-fluid solver in Gkeyll.

1. Axisymmetry and inter-species collisions are currently being implemented and will likely be in the main-line code in the coming weeks.
2. There is currently no fluid model for ionization implemented. The kinetic models for ionization could be ported to the fluid solver but this is a little bit of work infrastructure-wise to make all the pieces fit together.
3. The two-fluid solver is not currently relativistic. Including relativity is something we would like to do, but it's a non-trivial amount of work because of the additional complications that arise in implementing relativistic fluid solvers (for example, the conservative to primitive variable transformation is no longer trivial in relativity due to the presence of the relativistic enthalpy and Lorentz boost factors in the conserved variables).

If you would like, I can follow up on this issue in the coming weeks as axisymmetry and inter-species collisions are added to the main branch if you think you would like to try something anyway; however, I would understand if not having relativity is a bit of show-stopper for this application.

[MaxThevenet]

Thank you for your reply @JunoRavin! I am glad to hear that axisymmetric geometry and inter-species collisions are already planned for. Sure, I would appreciate if you could follow up on this in the coming weeks, thanks. As for the other points:
2. We anticipate that ionization plays a significant role in most cases, so this is a significant point IMO. Is there some interest to push this forward?
3. The driver is highly relativistic, and plasma electrons are slightly relativistic right behind the driver, but the further back we go, the less relativistic electrons are. One workflow could be to run a PIC simulation, extract fluid quantities from it several plasma periods after the driver passage (when electrons are no longer relativistic) and use these as input for a Gkeyll 2-fluid simulation. Do you think that could be possible/reasonable? I would have to check whether this is doable for our problem.

Also, since both Vlasov and 2-fluid models are implemented, is there a way to run these together in a Hybrid run (fluid for electrons, Vlasov for the ions)? I couldn't find it in the doc, I am just asking.