Can't solve ZAPDOS CCP when applied voltage is increased.

[garlan2018]

Hi,

I'm trying to build a DC bias model, so I re-visit the ZAPDOS CCP tutorial.
I've used a bit finer unstructured mesh and the solution goes well with the original setting in the tutorial.
Then, I've tried to switch off the top electrode and set the bottom electrode to 100sin(2pif*t).
The solver can't solve the case.
When I decreases the applied voltage magnitude to 70 V, it works again (Solved successfully to t = 2 e-5).

Is this due to the mesh issue or other issues?

The executioner blocks of the input code is as follows.
[Executioner]
type = Transient
end_time = 2e-5
dtmax = 1e-9
pets_options = 'snes_converged_reason -snes_linesearch_monitor'
solve_type = NEWTON
petsc_options_iname = '-pc_type -pc_factor_shift_type -pc_factor_shift_amount -ksp_type -snes_linesearch_minlambda'
pets_options_value = 'lu NONZERO 1.e-10 formes 1e-3'
nl_rel_tol = 1e-12
domino = 1e-14
l_max_its = 20
[]

[csdechant]

Hi!
Starting with high input voltages has been known to cause instabilities in Zapdos during the first couple of time steps. There are several ways of handling this:

• Ramp up the voltage amplitude (the preferable option, in my opinion). This can be done using a ramping functions and an 'if' statement in the voltage boundary function. Something like if(t<1e-6, (t/1e-6)*100*sin(2*pi*f*t), 100*sin(2*pi*f*t))
• Including an adaptive time stepper, such as IterationAdaptiveDT. The file in Zapdos Tutorial 5, DC_argon-With-Water.i, uses this time stepper. If you used this, I would recommend to set a dtmax depending on your problem. For example, with oscillating plasma, the time step should not be near the oscillation frequency, or else you will miss capturing some effects (such as plasma sheath oscillation).
• Increase the non-linear relative residual tolerance, nl_rel_tol. The default is 1e-8, but sometimes a lower tolerance is needed into reduce numerical error.

I would recommend trying these methods separately, but you can do all methods at once, if a single method alone is not working.
As for meshing, refining it might also help depending on the type of error. But in general, one of the above methods usually works for me.

[garlan2018]

Hi,

Thanks for your suggestions!
These provide some useful tips to improve the solving.
I can solve the case for applied voltage at 80 V, but I'm still tinkering the setup for a higher voltage.