[FLASH-USERS] Problem at refinement region boundary

[Markus Haider]

Hi,

I did some more testing regarding the artifacts at the refinement 
boundary. The problem is also present in FLASH 4.0. I used exactly the 
same simulation setup and the same parameter file and made 4 runs using 
different versions of FLASH and different solvers for comparison.

The initial conditions are at z=50 and I took the output after 20 time 
steps in the PPM and after 40 time steps in the USM, which corresponds 
to ~ z=20.8. The initial conditions are a uniform temperature and a 
rather smooth density field, box size is 20Mpc/h.

Here is how the density distribution looks like (there are hardly any 
noticeable differences between the runs):
http://ubuntuone.com/29SQg5m3kCnm9zkOV6KJzX

Here is how a temperature slice through the center of the domain looks 
in the different runs:

Flash 4 using the PPM solver:
http://ubuntuone.com/0NC31EMXM83dXyrDsyZPDR

Flash 4 using the unsplit staggerd mesh solver:
http://ubuntuone.com/0MTUuVcnpATZN7CdU3C12E

Flash3.1 with some patches using the PPM solver:
http://ubuntuone.com/2cuKPf0wMrahLeqSiVZjzV

Flash3.1 using the then private 3D version of the unsplit staggered mesh 
solver by Dongwook Lee:
http://ubuntuone.com/05gixjehhhTOKeYRTcj5sB

I should note that the artifacts increase with time. They are hardly 
noticeable in the first few time steps and become more and more 
prominent with time.

As you see, this raises some questions.

- All the solvers show artifacts at the refinement boundaries, except 
Flash3.1 using the 3D USM solver.
- There is a big discrepancy in temperature between between the USM and 
the PPM solver in Flash 4.
- The Flash3.1 USM mean temperature at z=20 is more than 5 times lower 
(13111K) compared to the USM solver in Flash 4 (74392K).

Let me know if I can do more tests which could explain these differences.

Thank you for your help,
Cheers,
Markus