search

SmileiPIC / Smilei

Particle-in-cell code for plasma simulation
https://smileipic.github.io/Smilei
341 stars 120 forks source link

Constant Solenoid Background Fields #159

Closed Buddhabrot2 closed 4 years ago

Buddhabrot2 commented 5 years ago

Hello everyone!

I observe some strange behavior when trying to apply a constant magnetic background field.

I am trying to use Smilei to simulate a simple magnetic mirror. To do that i want to apply a static magnetic background field as it would be created by a solenoid.

The solenoid is approximated using the the loopfield module (see https://pypi.org/project/loopfield/), wrapped in a function and given to the ExternalFields block.

For a solenoid directly outside the simulation region, the field is initialized correctly... field_0

... but then shows strange dynamic behavior. (25 steps in) field_10

If the "solenoid" is placed inside the simulation region, the field seems to undergo some kind of reconnection:

Start: field_0 30 Steps in: field_6

From Issue "External field and moving window #59 " I learned, that the external field is only applied on startup, since it was originally intended to study reconnection, but may be made constant by adapting the boundary conditions somehow. How is this done? Would this also work for the second case, where the "solenoid" is inside the simulation region? If not, could you recommend any better way to apply the field? I thought about using the antenna, but this would limit me to solenoids within the simulation region.

Namelist: magnetic_mirror.py.txt

Thanks for any help!

mccoys commented 5 years ago

I don't understand your setup very well but you must ensure that your field is divergence-free.

Also expect the field generated by the particles to appear in your results. If the particles in your simulation make strong currents able to significantly modify the field, and induce currents in the solenoid, then you might not be using the right tool. I might be misunderstanding your point though.

Buddhabrot2 commented 5 years ago

Thanks for your very quick reply!

About the setup: I would like to model a cloud of thin plasma moving into a solenoid (the plasma is following the central axis of the solenoid). The plasma should be compressed and then reflected or transmitted. I expect single particle trajectories to play a significant role, so i chose a PIC code. Imagine a reverse magnetic nozzle for a thin plasma. I should mention that at this stage i am just playing around with the concept ;) https://en.wikipedia.org/wiki/Magnetic_nozzle

Induction of current in the solenoid can be neglected, since the solenoid current would be kept stationary by external circuitry and the plasma would (probably) be too thin anyway.

I hope to eventually see the field generated by particles in the results, but for now i set the particle density to zero, so i can check the correct application of the external field and so I can get a baseline to compare the eventual result against. The field should be divergence free (exept maybe for numerical errors) since it matches the field of a real world solenoid calculated via the Biot Savart Law.

So what i am looking for is a constant field just superimposed on anything the actual simulation does. Is there any way I can make this happen?

mccoys commented 5 years ago

ExternalFields should do what you expect. They are applied as initial fields over the whole box, then continuously applied at the boundaries.

Did you try without particles first ?

MickaelGrech commented 5 years ago

Ciao! So I think SMILEI is behaving correctly and that ExternalField (which actually are initial fields, defined at t=0, then evolving according to Maxwell's equations) cannot do what you would like it to do. Indeed, the magnetic field given by the solenoid requires a current to be maintained. The way you are running the code, the magnetic field is here at time t=0, but let to evolve without any current maintaining it. It's like you just turned off the current in the solenoid. Hence, the magnetic field dissipates. We have recently implemented the possibility to add a time-dependent external field. In that case, this external field is prescribed by the user (in the namelist) and does not go through the Maxwell's solver. We will release (soon) a new version of the code with this time-dependent external field in beta-version. This should solve your problem (you just will have to enter a constant-in-time field, but it will be applied to the particles at all times). Note however that @mccoys is right in stating that you have to ensure a divergence-free B-field (which it seems you do within the numerical error) and that should still be the case for the time-dependent B-field. I would also like to stress that one has also to be cautious with the electromagnetic boundary conditions.

mccoys commented 5 years ago

Isn't the field supposed to be maintained by the boundary conditions?

Buddhabrot2 commented 5 years ago

@mccoys I checked the field with the number density set to zero, and again with particles_per_cell = 0 for all species, with identical results. I think boundary conditions can only uphold a magnetic field which is not caused by currents within the simulation, since the maxwell solver would just "connect" the open field lines that are created at the boundary. If i place the "solenoid" far outside the simulation region, the field behaves much more sensible, too.

@MickaelGrech This explains the behaviour, of course. From the documentation, i did not realize that the external fields are initial fields. I am looking forward to the new version of the code! For now i will just experiment a bit with the antenna block to maintain a current and combine this with the external field block. Regarding the boundary conditions, i probably still have some reading to do. Could you point me to any good resources on this?

Thanks for your time and have a nice weekend!

talmiller2 commented 5 years ago

What about using Antenna to drive the solenoid current?

mccoys commented 5 years ago

@talmiller2 the antenna option was suggested earlier on, but they would require a solenoid to fit in the box

talmiller2 commented 4 years ago

@mccoys I missed that when going over the comments previously. I can see why fitting the solenoid in the simulation would be wasteful, and so would like to join on the request for a time dependent magnetic field boundary condition, or at least an option for a constant background field (to keep enforcing the initial field). I'm also interested in simulating a solenoid background field. Thanks a lot.

mccoys commented 4 years ago

The option for additional, time-varying fields is currently being tested. I hope it will be available soon

mccoys commented 4 years ago

@MickaelGrech isn't this already available? I thought it was done already

MickaelGrech commented 4 years ago

Yes. This will be available pretty soon, just not right now (my fault I'm lagging behind).

mccoys commented 4 years ago

@MickaelGrech This is done, right ?

MickaelGrech commented 4 years ago

Yes! Since v4.4 the user can define some prescribed fields. Here is the corresponding doc. Note that these fields are not considered by the Maxwell solver, but only to push particles (they appear in the Fields or Probe diagnostics though).

mccoys commented 4 years ago

An important fix was made to this feature in v4.5. Please upgrade.