[FLASH-BUGS] possible bugs in Poisson routines

[Paul Ricker]

Artur,

The dust collapse problem requires isolated boundary conditions, which
are not supported for multigrid in 2D Cartesian coordinates.  What
boundary conditions did you use for the 2D/3D multigrid case?  If the
boundary conditions are periodic or Dirichlet, instead of isolated,
you will get a different potential, and the collapse time will be
different.

For multipole, the problem is spherically symmetric, so you shouldn't
need to use such high mpole_lmax values.  I suspect the problem in
this case is that, because of the high power of r for high multipole
order, errors in the computation of moments near the center are
leading to force errors there.

Paul

-- 
---------------------------------------------------------------------
Paul M. Ricker                                Department of Astronomy
Assistant Professor   National Center for Supercomputing Applications
pmricker at uiuc.edu          University of Illinois at Urbana-Champaign
http://www.astro.uiuc.edu/~pmricker              Urbana IL 61801-3074
---------------------------------------------------------------------

On Thu, 2003-08-07 at 15:57, Artur Gawryszczak wrote:
> Hi,
> 
> I've attached two setups derived from dust_coll, which shows quite strange 
> results.
> 
> In both cases I'm using "natural" units, so G*density = 1, and box walls are 
> set at +/-1
> 
> dust_coll.2d_cart_mgrd - sets up a collapse on 2-D cartesian grid using 
> multigrid submodule. Everything works fine, except that free-fall time is 
> longer than expected for given initial density. It takes 0.5 unit of time to 
> collapse, while in my opinion it should take 1/sqrt(2*pi) (this is a collapse 
> of an infinite cylinder). I've tried to run this problem in 3D, changed 
> init_block.F90 to fill a cylinder instead of sphere and put periodic BC in 
> z-direction. The free-fall time did not change.
> 
> dust_coll.3d_cart_mpol - sets up a collapse on 3-D cartesian grid using 
> multipole submodule. Strange bubble forms at the center of the sphere when 
> "mpole_lmax" parameter is set high enough, i.e. 10 or more.