[Archive] Convergence of Newton algorithm

Message by engwall:

When trying a simulation with a sphere in a plasma with 1000 cc, Ti=Te=1 eV, spacecraft radius=2m, Dimension of computational box=20 m, I get the following warning:

WARNING: Newton did not reached convergence'

and the spacecraft potential gets unrealistic values. I use the automatic integration timesteps. What should I change to let the Newton algorithm reach convergence?
Grateful for any help

Message by roussel:

I am a little surprised by this problem, because Newton implicit algorithm for solving Poisson non linear equation is stable in this range of parameters (tested and used quite a lot). We may yet have already met unstabilities similars to yours, which were indeed due to the extremely fast change of absolute potential (due to a small Csat, cf your “Spacecraft capacitance” post), which result in very large SC potential, and in turn generate convergence issues in the Poisson solver.
So, please:

  • try to check what is the evolution of the potential of your SC (if it get very large before the Poisson solver issue, it is problem source)
  • you may also simply try to leave a fixed potential and see if you Poisson solver issue persists

thanks for your testings!

Message by engwall :

OK, when increasing the spacecraft capacitance the problems disappear.
Thanks for the help, Erik