[FLASH-USERS] Refinement bug on boundaries

[Mark Richardson]

>
> Hello everyone,
>
>   I've been noticing an issue in blocks that refine on the boundaries, and
> so I've reduced it to a simple setup that seems to recreate the issue.
>
> In this simple set up I have a rectangular volume with nxb=32, nyb=16,
> nzb=8, so that all dimensions have equal resolution. In each block I have a
> density gradient in the x-direction from 2e-6 g/cc in the first cell
> (including guard cells), to 2e-4 g/cc in the last cell (including guard
> cells). I also have an x-velocity of +1e4 cm/s. I am using diode boundary
> conditions. I have an initial uniform pressure of 2e-2 dyne/cm^2.
>
> At init, I force everything to be a level lmin=3=lmax-1. After nref=2
> steps I force the entire volume to refine to level lmax=4. This results in
> a strip of high energy (or pressure) cells on the inner-x boundary that are
> approximately 3 orders of magnitude higher than the uniform value (see
> attached).
>
> [Aside: This also leads to an increasing in the integrated internal energy
> of the box by more than a factor of 20. While the density appears to remain
> the same (but smoother) it does look like it shifts slightly 'downhill', so
> when density is increasing from left to right, a refinement is activated
> such that a band of same colour (say 3e-5 g/cc) shifts to lower x. This may
> be an optical illusion though.]
>
>  These refinements are done at the end of step=2, and while the high
> pressure strip is present in the images I make of the checkpoint file at
> step=2, they are not present in the images I make of the plotfile at
> step=2, as the plotfile is generated before the refinements are done. This
> has convinced me it is the refinements causing the incorrect pressure.
>
> I've done a few tests so far:
>    If there is a constant density, and no velocity, then the problem
> doesn't appear.
>    If there is a density gradient, but no velocity, then the problem
> doesn't appear.
>    If there is a constant density, and a positive x velocity, then it does
> appear.
>    This is true if the gradients and velocity are in the y direction, but
> the pressure issue is on the inner y boundary.
>    If the density gradient is negative, and the x velocity positive, the
> problem doesn't seem to occur.
>    If the density gradient is negative, and the x velocity negative, the
> problem appears as before, but on the upper x boundary.
>
> Attached is a tarball of the setup files found in their respective
> directory in SimulationMain, including a flash.par and a setup_call that
> was used. I've also included my sites file so you can see my
> optimisations/options. Currently it's -O3, but I have tested this with
> -debug on, with the same results. I am currently using ifort/icc/icpc
> (IFORT/ICC/) 15.0.2 20150121.
>
> Any help in understanding why this is happening would be very useful! Is
> this the expected result for a prolongation on a simulation boundary with
> diode condition and motion away from the boundary? Or could there be a bug
> in gr_updateRefinement?
>
> -Mark
>
> --
>
> Mark Richardson
> Beecroft Postdoctoral Fellow, Department of Physics
> Denys Wilkinson Building - 555F
> University of Oxford
> Mark.Richardson at physics.ox.ac.uk
> +44 1865 280793
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://flash.rochester.edu/pipermail/flash-users/attachments/20170407/e3a9cee6/attachment.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Dens_Pres.pdf
Type: application/pdf
Size: 766624 bytes
Desc: not available
URL: <http://flash.rochester.edu/pipermail/flash-users/attachments/20170407/e3a9cee6/attachment.pdf>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: DebugRef.tgz
Type: application/x-gzip
Size: 1361624 bytes
Desc: not available
URL: <http://flash.rochester.edu/pipermail/flash-users/attachments/20170407/e3a9cee6/attachment.bin>