[FLASH-USERS] [EXTERNAL] USM Advection-only Test

[Ryan Farber]

Hi all,

Thanks for your replies! Interestingly, Eddie's idea of modifying EOS (I'm
using eos_idealGamma) worked pretty well although with reduced stability
(blew up for cfl >= 0.74, worked for cfl <= 0.7). Interestingly, I just
noticed there appears to develop asymmetry orthogonal to the motion with
ForceHydroLimit = .false. and asymmetry parallel to the motion with
ForceHydroLimit = .true. (I was initializing Bx=By=1e-10 so they'd be
dynamically unimportant and plots suggest they didn't change with time).

I was hoping to gain a more fundamental understanding of how FLASH works so
I really appreciate Adam's message! I modified hy_uhd_prim2flx to remove
the pressure terms. That ~worked in that the blob of gas energy density
maintained shape and size (inspected by eye), but the uniformity in density
was broken. Also that was at cfl = 0.1. At higher cfl (say 0.45) there were
ripple artefacts in all variables. The ripples don't blow up but they do
grow in spatial extent, leading to extraneous refinement.

Dean Townsley mentioned offline that I could just set the density profile
based on the gas energy density to initialize my blob in pressure
equilibrium. This worked quite nicely.

I expect the vortex method of balancing out the pressure gradient with
centrifugal force would work just as well.

Thanks again for the help!

Best,
--------
Ryan



On Wed, Jun 3, 2020 at 4:52 PM Adam Reyes <acreyes at uchicago.edu> wrote:

> You can remove the pressure terms inside of "hy_uhd_prim2flx.F90".
>
> Alternatively you can use a uniform pressure and similarly advect a
> density profile. This has an advantage of not needing to modify any code
> inside of Hydro and whatever unintended consequences might stem from that.
>
> You can also look at the IsentropicVortex test problem provided in
> Simulation/SimulationMain/IsentropicVortex. This is very similar to the
> problem you've described, except the pressure gradient is balanced by the
> centrifugal force of the vortex. All of which is embedded in a constant
> background flow so the initial conditions are simply advected through the
> domain.
>
> On Wed, Jun 3, 2020 at 3:30 PM Eddie Hansen <ehansen at pas.rochester.edu>
> wrote:
>
>> Maybe you could edit whatever EOS routine you're using to artificially
>> set the pressure(s) to some constant value. Not very physical, but it would
>> presumably get rid of motion due to pressure gradients.
>>
>>
>> On Wed, Jun 3, 2020, 3:15 PM Ryan Farber <rjfarber at umich.edu> wrote:
>>
>>> Hi flash-users,
>>>
>>> I'm working on a (2D) test problem in FLASH4.6.2 in which I initialize a
>>> gaussian of gas energy density on an otherwise uniform medium with
>>> vx=vy=1.0.
>>>
>>> I refine on gas energy density with lrefine_min = 6 and lrefine_max = 9
>>> and update refinement every two steps (nref=2).
>>>
>>> The purpose of the test is to confirm conservation of mass, momentum,
>>> and energy with AMR (and additionally to confirm AMR
>>> refinement/derefinement works correctly).
>>>
>>> I have successfully run the problem; however, the gaussian spreads out
>>> presumably due to the pressure gradient between the blob and the ambient
>>> medium.
>>>
>>> I've been perusing hy_uhd_: unsplit; unsplitUpdate; and now getFaceFlux
>>> but I haven't yet seen how I can turn off the inclusion of pressure
>>> gradients in the solution.
>>>
>>> If you have a hint, I'd welcome it :)
>>>
>>> Best,
>>> --------
>>> Ryan Farber
>>> PhD Candidate
>>> University of Michigan, Ann Arbor
>>>
>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://flash.rochester.edu/pipermail/flash-users/attachments/20200604/727c47cf/attachment.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ripples.png
Type: image/png
Size: 106335 bytes
Desc: not available
URL: <http://flash.rochester.edu/pipermail/flash-users/attachments/20200604/727c47cf/attachment.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: AdvectGaussian_hdf5_plt_cnt_0005_Slice_z_density.png
Type: image/png
Size: 67547 bytes
Desc: not available
URL: <http://flash.rochester.edu/pipermail/flash-users/attachments/20200604/727c47cf/attachment-0001.png>