[petsc-users] adaptive TS question

[Mark Adams]

I have a plasma problem of the form dU/dt+ F(U,t) = S(t). U is the density
of species in a plasma. S is, say, a step function scaling of a Maxwellian
for two of the species.  Using 2 or 3 species.

I'm happy with it for dU/dt+ F(U,t) = 0. My test problem starts with a time
step of about O(1) and goes to a quasi equilibrium with a time step about
O(100).

But when the source terms kick in the time step goes way down. This seems
strange because dU/dt = S(t) just adds S*time to U. (I have verified that
with U=0 one step does just add S * dt to U).

I use these parameters. FWIW, this TS does three stages, approximates the
error and adjusts accordingly. I notice that my S (RHS) function is called
after the first stage. So this algorithm solves  dU/dt+ F(U,t) = 0 in the
first stage. Apparently.

Any ideas?
Thanks,
Mark

-ts_type arkimex -ts_exact_final_time stepover -ts_arkimex_type 1bee
-ts_max_snes_failures -1 -ts_rtol 1e-6 -ts_dt 1e-1 -ts_adapt_clip .25,1.1
-ts_adapt_scale_solve_failed 0.75 -ts_adapt_time_step_increase_delay 5
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20200404/8a12e9f4/attachment.html>