Magnetic field in TrackParticle and ParticleBinning diagnostics about smilei HOT 21 CLOSED

Just a precision: the magnetic field is shifted by half a timestep compared to its value from Field diagnostics. Right? Is this what you meant?

from smilei.

Not exactly. It is shifted by half a timestep and half a spatial step. In particular, if particle moves along x direction inside the plane wave, which is propagating in -x direction, then the vector of the transverse magnetic field (B_y,B_z) in TrackParticles/ParticleBinning diagnotics is rotated by the angle phi=timestep*(1+v_x)/2

from smilei.

I don't understand how there could be a spatial shift. Is it due because of the wave displacement during that half timestep?

from smilei.

I think it is due to the particle displacement during the half timestep, since there is the term timestep*v_x/2

from smilei.

But what do you compare it with? The field diagnostic? If yes, i believe those are not properly aligned (by half a cell) in happi

from smilei.

No, actually I calculate the longitudinal component of the Lorentz force ([v x B]_x, for circular polarization it is also the ponderomotive force), and I know that (at least for rare plasma) the integral over time of it should be equal to zero. It is zero ideed after the rotation I described above.

from smilei.

@egelfer Could you try your test using the option keep_interpolated_fields? It probably changes the results as the interpolated fields are calculated at the correct time with this option.

from smilei.

Actually I used the option keep_interpolated_fields=["Ex","Ey","Ez","Bx","By","Bz"]. Do you want me to run the test without it?

from smilei.

yes that would be good.

Now your result is puzzling. Those fields are supposed to be exactly those calculated for pushing particles

from smilei.

Yes, I also assumed it and was very much surprized :)

from smilei.

Ah well in fact I forgot that the magnetic field is interpolated at integer timesteps while the velocity is known at half-integer timesteps

Thus the shift?

from smilei.

Yes, I suppose that might be the source of the shift

from smilei.

Did you manage to test without keep_interpolated_fields?

from smilei.

Did you test without keep_interpolated_fields?

from smilei.

I made some changes in the documentation in branch develop. Please reopen if necessary

from smilei.

Did you test without keep_interpolated_fields?

Sorry, I forgot to answer. In this case the simulation crushes.

I also wanted to ask, is it possible to output the magnetic field in a way similar to when the force acting on a field is calculated (https://smileipic.github.io/Smilei/Understand/algorithms.html#the-pic-loop), i.e. (B_p^{(n+1/2)}+B_p^{(n-1/2)})/2 ? Unfortunately my method to recover the correct field is valid only for a plane wave, and it is not always the case.

from smilei.

keep_interpolated_fields exactly retain the value of the fields that were obtained during interpolation. So you should have the force exactly.

from smilei.

That's what I assumed! :) But it is not, what I've got. Here is the plot of average longitudinal momentum (darkred, obtained from the ParticleBinning diag) in the collision of electron-positron bunch with a pulse plane wave. Magenta curve corresponds to the integral of the (averaged over the particles) force F_x= [v x B]_x+E_x with the E and B fields obtained from TrackParticles diag. Orange curve corresponds to the integral of F_x'= [v x B']_x+E_x, where B' is B shifted by a half step. However, as I wrote above, it works only for a plane wave field.

from smilei.

Are you using the velocity from tracked particles as well? As the fields and velocities differ by half a timestep that might be the problem

from smilei.

Yes, the velocities are taken from TrackParticles diag as well

from smilei.

Well then there you have it. It is not the fields that are out of phase. It's the velocities

from smilei.