RZ: `parse_density_function` incorrect on-axis density

[jjvdwOX]

Hi,

Using parse_density_function in the RZ geometry is resulting in a spike in density on-axis. Here is an example setting the density function to a constant. input.txt

[RemiLehe]

Thanks for reporting this. I think that what you are seeing is a known issue for RZ grid. You can read more about it in this paper: https://www.sciencedirect.com/science/article/pii/S0021999101969232 The algorithm described in this paper mitigates the issue, but does not suppress it entirely (at least for a finite number of macroparticles per cell) - and this is indeed the algorithm that we implemented in WarpX.

In order to confirm that this is the issue, you could try to run with more particles in the r direction, i.e.

electrons.num_particles_per_cell_each_dim = 4 12 2

In principle, using more macroparticles should reduce the amplitude of the effect that you are seeing here.

[jjvdwOX]

Thank you for your reply. I tried doubling the number of particles per cell in the r direction as advised, but it did not reduce the on-axis density error. input_1.txt

My main worry is that this on-axis density spike is not just a plotting error, but a real effect in the simulation which can affect on-axis physics.

[jjvdwOX]

As an example of how the density error affects the on-axis physics, here is a previous run of a gaussian laser pulse in a parabolic channel with high radial grid resolution input_RZ_channel.txt

The density error on-axis reacts to the laser pulse and causes wave-breaking, which makes it impossible to study injection

[ax3l]

@dpgrote @RemiLehe does one of you like to take the lead here or assign another developer?

[jlvay]

@jjvdwOX There is indeed an issue on axis with the RZ FDTD solver, which needs to be addressed. In the meantime, it is best that you use the RZ PSATD solver.

[jlvay]

@dpgrote @RemiLehe @ax3l The RZ FDTD solver should be disabled until the issue on axis is fixed.