[FLASH-USERS] Problems with strong shocks

[Jeremy S Ritter]

Hello Dongwook,

That's correct, I did not have this same problem using FLASH3. I have not
run this same exact setup under FLASH3, but I have run many very similar
setups (same mass/energy blastwaves in similar environment). Unfortunately
it would take some effort to back-port this particular setup to FLASH3. The
primary difference in our setup between FLASH3 and FLASH4 is the move from
the split to the unsplit solver, but I don't know if that is causing the
issue here. The basic setup is a spherical blast wave with 10^51 erg
kinetic energy started in the free-expansion phase inside an HII region,
where we have zoomed in from a cosmological simulation. There are 128-256
cells across the diameter of the blast wave.

I've tried different Riemann solvers (Roe, HLLC,  and hybrid), with and
without shock detection and lowering the CFL. I've tried different
masses/velocities for the blastwaves. They all eventually run into this
problem at some point, usually after a few hundred years. I wonder if I
don't have some incorrect choices in flash.par? Below are the relevant
values from the Hydro section of my flash.par.

Cheers,
-Jeremy


# Hydro ---------------------------------------------------
useHydro = .true.
updateHydroFluxes = .true.
ppm_modifystates = .false.
use_auxEintEqn = .false.

eintSwitch = 1.0e-4
gamma = 1.66666666667
cfl = 0.8
cvisc = 0.1
nriem = 25

small = 1.0E-40
smlrho = 1.0E-40
smallp = 1.0E-40
smalle = 1.E8
smallt = 1.
smallu = 1.0E-40
smallx = 1.0E-40

##  SWITCHES SPECIFIC TO THE UNSPLIT HYDRO SOLVER               ##
#interpol_order = 2
order = 3 # 3rd PPM
slopeLimiter = "mc" # Slope limiters (minmod, mc, vanLeer, hybrid, limited)
LimitedSlopeBeta = 1. # Slope parameter for the "limited" slope by Toro
charLimiting = .true. # Characteristic limiting vs.Primitive limiting

use_avisc = .true. # use artificial viscosity (originally for PPM)
use_flattening = .false. # use flattening (dissipative) (originally for PPM)
use_steepening = .false. # use contact steepening (originally for PPM)
use_upwindTVD = .false. # use upwind biased TVD slope for PPM (need
nguard=6)

flux_correct = .true.
shockDetect = .true.
shockLowerCFL = .true.

#       II. RIEMANN SOLVERS:
RiemannSolver = "hybrid"       # Roe, HLL, HLLC, LLF, Marquina
entropy = .false.

use_gravHalfUpdate = .true.

#convertToConsvdInMeshInterp = .false.


# apply gr_smalle and don't print anything
#enableMaskedGCFill = .true.
gr_sanitizeDataMode = 3
gr_sanitizeVerbosity = 0





On Tue, Feb 20, 2018 at 2:40 PM, Dongwook Lee <dongwook at flash.uchicago.edu>
wrote:

> Dear Jeremy,
>
> Do you see this instability issue only with FLASH4, but NOT with FLASH3?
> More specifically, I wonder what happens if you run the exact same problem
> with FLASH3. Have you tried this type of FLASH3 and FLASH4 comparison tests?
>
> Cheers,
> Dongwook
>
> On Sun, Feb 18, 2018 at 10:41 AM, Jeremy S Ritter <jritter at mail.utexas.edu
> > wrote:
>
>> Hello Flash Users,
>>
>> I have been using FLASH since version 3 to run simulations of supernova
>> blastwaves in cosmological simulations. Since updating to version 4 I have
>> had a lot of stability problems with the unsplit solver and strong shocks.
>> The setup is similar to a Sedov example, with a spherical blastwave
>> inserted into a realistic cosmological environment. We zoom in to high
>> resolution to insert the blastwaves with radii of ~1 pc and 16-32 grid
>> blocks across the diameter. The background density/temperature/pressure in
>> the vicinity of the blastwaves is effectively constant when zoomed in to
>> this scale. We refine a large volume around the blastwave so that the
>> entirety of it is kept at the same refinement level while expanding.
>>
>> The blastwaves will evolve as they should for many years, doubling or
>> more in radius, and then suddenly within 1 step the solution will break
>> down. This happens in the free expansion phase, before the reverse
>> shock. The timestep dt will drop from ~1 year to 10^-7 seconds, tracer
>> particles in the blastwave will be flung outside of the simulation box, gas
>> density in some cells will drop by 20 orders of magnitude, etc.
>>
>> I have tried both the HLLC and the hybrid HLLC solver with shock
>> detection and shock lower CFL. By lowering the CFL even further, to 0.1, I
>> was able to push the simulation further but eventually it breaks down too.
>> I never had any similar issues in FLASH3. I was wondering if anybody has
>> dealt with anything similar in FLASH4?
>>
>> Cheers,
>> -Jeremy
>>
>>
>
>
> --
>
> =========================================
> Dongwook Lee, Ph.D., Assistant Professor
> Applied Mathematics and Statistics
> University of California, Santa Cruz
> Baskin Engineering, Room 353C
> 1156 High Street, Santa Cruz, CA 95064
> <https://maps.google.com/?q=1156+High+Street,+Santa+Cruz,+CA+95064&entry=gmail&source=g>
> https://users.soe.ucsc.edu/~dongwook/
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://flash.rochester.edu/pipermail/flash-users/attachments/20180221/c807abf1/attachment.htm>