[FLASH-USERS] 2D Spherical and Cartesian Geometry

[Michiel Bustraan]

Klaus,

I ran a few simulation the past few days that might help explain what was causing the issues and hopefully provide some closure.

So firstly, I did add +pm4dev to the setup command, and apart from an additional statement during the runtime, it's hard to say how much of an impact it's had.

Secondly, simulations without any intended motion (no temperature differences, no initial velocities) work perfectly fine.
I see small fluctuations in velocity, but only on the order of 10^-4.

I think the recent problems are due to my initial conditions.
The simulation I've been running, of spherical point explosions, start with a radial velocity that is proportional to radius.
I've attached a plot showing the initial velocity distribution. 
Because my simulation starts with a positive velocity in the inner radii (minus the innermost point), it leaves a hole behind.
Looking at the forced plotfiles from previous simulations, I noticed very low temperature and internal energies (~ 10^-19) at very small radii.
I think that the empty space left behind by the outwardly moving shock was leading to the low internal energies that created warnings and ultimately crashed the code.

However, it seems strange to me then that the crashing error message would be missing neighbors, which sounds like a grid issue.

In any case, a recent simulation I ran, with no AMR and a much higher radial resolution (2500 blocks), ran much better (no warnings, no crashes).
The density ends up being almost exactly identical to past 1D simulations, implying the code is working well.
However, the lower density/energy issues at very small radii still exist. In the density plot I've attached, you can see a few datapoints dropping in the density at the innermost radii.
I think it's because of the higher resolution that the velocity is distributed better (more of a linear increase instead of a step function), and this minimizes the issue.

Next week I'll run the same simulation with a lower base number of blocks and use AMR to reach the same resolution.
Then I should know whether the recent issues really were just due to having too few blocks and not because of the AMR.

I was wondering if this issue of having a positive velocity at very small radii could be fixed.
It always seems to leave a gap behind (I guess because it can't draw density from the boundaries) though the impact can be minimized with a larger number of blocks.

Kind Regards and sorry for the long email,
Michiel
________________________________________
From: Klaus Weide <klaus at flash.uchicago.edu>
Sent: Wednesday, September 28, 2016 5:38 PM
To: Michiel Bustraan
Cc: flash-users at flash.uchicago.edu
Subject: Re: [FLASH-USERS] 2D Spherical and Cartesian Geometry

Michiel,

Please try also adding '+pm4dev' to your setup command.

You are already using the Pemaesh4dev version of PARAMESH by default, but
requesting +pm4dev explicitly should enable an alternative way to
compute surr_blks (an important array inside the PARAMESH kernel of the
FLASH Grid).

Klaus
-------------- next part --------------
A non-text attachment was scrubbed...
Name: velx_1d_00.png
Type: image/png
Size: 19290 bytes
Desc: velx_1d_00.png
URL: <http://flash.rochester.edu/pipermail/flash-users/attachments/20160930/37ea06c3/attachment.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: density_1d_10.png
Type: image/png
Size: 27933 bytes
Desc: density_1d_10.png
URL: <http://flash.rochester.edu/pipermail/flash-users/attachments/20160930/37ea06c3/attachment-0001.png>