[FLASH-USERS] Problem at refinement region boundary
Klaus Weide
klaus at flash.uchicago.edu
Mon Sep 17 17:44:27 EDT 2012
On Sun, 16 Sep 2012, Markus Haider wrote:
> Hi everyone,
>
> I am seeing some strange artefacts in plots of the temperature (and eint and
> pres) at the edge of refinement levels, and I don't really have an idea where
> I should start to look for errors.
>
> Here is a plot of the temperature (z=8):
> http://ubuntuone.com/1SebX1vAk0clRvPiXfWFP4
>
> In contrast, the density plot looks fine:
> http://ubuntuone.com/2lxUZ42iWdnqKWClKW335K
>
> In the output file I get many warnings like this (numbers change a bit):
> WARNING after gc filling: min. unk(EINT_VAR)=91065420.08702280
> PE
> =18 block=18 type=1
Markus,
These warnings are produced when some variables (in his case, internal
energy) are found to be lower than a lower bound - in this case, given
by the runtime parameter 'smalle'.
Is your 'smalle' value appropriate for your simulation?
If appropriate, are the values shown in the warnings significantly
lower than smalle, or only a little bit?
(The latter could be acceptable if it is just a small bit of undershooting
introduced in interpolation at fine-coarse boundaries.)
Since the relationship between density, temperature, eint etc. is given by
the EOS - which Eos implementation are you suing?
> Could this be related?
>
> Here is some information about the simulation: I simulate a galaxy cluster in
> a 20 Mpc/h box. I use a static mesh refinement with 5 nested levels. The
> simulation was done using the PPM solver and periodic boundary conditions. I
> used the FLASH 3.3 version. I should note that I ported this simulation setup
> from a modified version of FLASH 3.1.1, and there this problem did not occur.
>
> I would be happy if someone has an idea about what could be responsible for my
> problem,
Looking at he two images, the solution doesn't look periodic! This shows
up particularly at around y=5Mpc, in both density and temperature.
This may have somethign to do with the temperature jumps at boundaries.
Klaus
More information about the flash-users
mailing list