Help Understanding pgkyl data format

[BrendanMicheal]

I have been trying to understand how Gkeyll formats data for evolved species with the moments app. I've been running the rt_5m_gem.lua from the moments luareg for testing and am confused on what is being plotted when analyzing with postgykl. For example, the field plots return the E and B fields along each axis, but I can't tell what the last two graphs are depicting. They are initialized as 0.0 without a variable name in the lua file itself. When exporting the field Gkeyll file as a .txt using write -f the resulting file has 10 rows of data. If you could let me know what is being recorded and plotted, I would greatly appreciate it.

Snippet where the initial conditions are fed to Moments.Field

field = Moments.Field.new {
    ...

    -- Initial conditions function.
    init = function (t, xn)
      ...
      return Ex, Ey, Ez, Bx, By, Bz, 0.0, 0.0
    end,

    evolve = true, 
  }

c6 and c7 are the plots I am struggling to understand

[JunoRavin]

The extra two fields are because we solve the perfectly hyperbolic variation of Maxwell's equation to fold in the divergence constraints directly to the evolution: https://gkeyll.readthedocs.io/en/latest/dev/maxwell-eigensystem.html

The simulation you ran does not utilize div(E) cleaning, but does use div(B) cleaning (that's why the last component there takes on value as we damp away div(B) errors that are generated by the numerical scheme).

[BrendanMicheal]

Thank you for your quick reply! That makes a lot of sense. To make sure I am understanding correctly, the mgnErrorSpeedFactor and elcErrorSpeedFactor variables are used to make corrections to the fields to meet divergence criteria at each time step, which are then displayed separately in c6 and c7? What would these variables correspond to in the methodology?

[JunoRavin]

Glad to help! Yes those two inputs are chi and gamma respectively in the note I sent and they are best thought of as factors which scale the speed with respect divergence errors are propagated away from their source (see the highlight piece of the eigensystem that notes what the eigenvalues are of the flux Jacobian, which tell us the characteristic velocities of the system).

Although we allow any number as inputs (since these are just numerical factors) all simulations we have ever run give these factors one of two values: 0 (no divergence cleaning) and 1 (divergence errors are propagated away from their source at the speed of light).