[Archive] How to increase a simulation speed

Message by Hodong Lee:
In addition to the previous question of CEX,
I also have a problem with too slow simulation speed.
I am running SPIS 6.0.0 with 24 GB RAM, but what bothers me is too small simulation dt while running the simulation.
The number of tetrahedrons in the mesh I made is about 65,000, with a not bad mesh quality.
I want to simulate a relatively short duration of 1 ms for a hall thruster, but simulation dt is in the range of 10^(-10) ~ 10^(-9).
I tried to decrease the number of avPartNbPerCell to 2.0, and change electronDistrib and ionDistrib as UnlimitedGlobalMaxwellBoltzmannVolDistrib, with other particles related to thruster (Xe+, CEX Xe+, etc) as PIC.
Are there other parameters I should change in order to get a reasonable simulation speed?
Or, is it originally (inevitably) so slow?
I am desperately waiting for a reply.

The following information is the parameters I used in my simulation:
About maximum intergration time step
electron dt: 1.0E-7
electron duration: 2.5E-7
ion dt : 1.0E-6
ion duration : 5.0E-6
plasma dt: 5.0E-7
plasma duration: 1.0E-6
simulation dt: 1.0E-5
simulation dt init: 1.0E-6
duration: 1.0E-3
VolInteract Parameters
volInteract#: 0.5
crossSectionVolInteract1 (for Xe+): 8.0E-19
crossSectionVolInteract2 (for Xe++): 2.0E-19
About Poisson Equation:
iterGradient: 1500
iterGradientNI: 2000
iterLinearSys: 10000
iterNewton: 200
regular Poisson computation
tolGradientNI: 1.0E-5
TolLinearSys: 1.0E-12
tolNewton: 0.005
variableTeConstant: 6.07E-4

Thank you.
Hodong Lee.

Message by S├ębastien Hess:
EP simulations are very slow in the first steps of the simulation because the spacecraft potential evolves very quickly due to the huge currents delivered by the thruster and cathodes. Once stabilized, the simulation steps should be much larger.
There are three phases with small dt: - the thruster/cathode ignition

  • the coupling with the environment when the plume reach the simulation boundaries
  • the coupling with the backflow when CEX reach the solar panels
Regards, Sebastien