Fix #688 2D Restart Fields

[ax3l]

Fix #688 2D Restart Fields: in 2D simulations, the last component (z) of FieldE and FieldB was not initialized again (default: zero).

• verify with 2D KHI simulation and check EnergyFields.dat before/after restart (especially B_z)

unrelated:

• wrong docs string and index names in hdf5/writer/Field.hpp -> extra pull HDF5 WriteField: Fix Wrong Doc String and Index #690

the index names in hdf5/writer/Field.hpp are not ideal (even if the implementation is correct): once d is used as an index [0; simDim-1] and once as an index [0; nComponents-1] and same mixed with i.

[PrometheusPi]

I will test this with KHI 2D restart.

[ax3l]

ok cool.
I suggest running until step 1000 with dumping energy_fields every 10 steps.
then two restart(s): one with old version and one with fixed.

The plot of Energy B_z would be interesting for both restarts.

[PrometheusPi]

I ran the simulation (KHI, 2D, electron/positron, total of 2000 time steps, restart after 1000th time step)

The numerical heating shows that the fix worsens the numerical heating 😜.

Since we loose energy with the restart, we counteract the numerical heating.

If we have a look at the energy stored in the driving magnetic field B_z:

we see a clear reaction to the restart error. The feedback loop misses one compenent and restarts iteslf. We see strong oscilation untill the error is compensated. B_z is now higher but we lost more energy in the electrons and the other field components.

The total field energy looks similar:

[ax3l]

thank you for running the tests on taurus, holy moly - what a bug.

[PrometheusPi]

No problem. I used hypnos/k20d not taurus/gpu.

[ax3l]

oh! good catch! I want that, too.