[FLASH-BUGS] Bug in quadratic_cartesian interpolation scheme?

[Mike Zingale]

Robi, if it is not too much trouble to rerun, try it with the prolongation
monotonicity constraint turned off,

monotone = .false.

I would suspect that this is where the problem enters, but I am not
certain.  I'll look at the pictures some more and thing about this.

Mike

------------------------------------------------------------------------------
Michael Zingale
UCO/Lick Observatory
UCSC
Santa Cruz, CA 95064

phone:  (831) 459-5246 
fax:    (831) 459-5265
e-mail: zingale at ucolick.org
web:    http://www.ucolick.org/~zingale

"Once something has been approved by the Government, It's no longer immoral."
   -- Reverend Lovejoy







On Thu, 6 Nov 2003, Robi Banerjee wrote:

> Hi Mike,
> 
>   I have set up a web page showing the outcome of the quadratic_cartesian
> interpolation scheme compared with the old second order one. You can
> find the page at: 
> http://physwww.physics.mcmaster.ca/~banerjee/problem.html
> 
>   I used FLASH 2.3 on our Alpha with Tru64 Unix. I initialized the
> the data with a simulation done with Zeus, therefore it is not easy
> for me to rewrite the code for a 2D simulation. Here are some more
> information of the modules I used:
> 
> REQUIRES driver/time_dep
> REQUIRES hydro
> REQUIRES gravity/poisson/multigrid
> REQUIRES materials/eos/gamma
> REQUIRES source_terms/cool/radloss
> REQUIRES mesh/amr/paramesh2.0
> 
> 
>   The 'stripes' appear in all variables but are most prominent
> in the velocities. Maybe the 'stripes' show up only in regions
> with a large gradient of the variables?
> 
>   Robi
> 
> 
> On Wed, 5 Nov 2003, Mike Zingale wrote:
> 
> > Hi Robi, the only problem that I am aware of with those prolongation
> > routines is that the monotonicity constraint does not operate on the mass
> > fractions as a whole, so it does not guarantee that they sum to one at the
> > end -- this is something that I've been meaning to fix, but have not had
> > the time.  
> > 
> > I have not seen a problem like you describe.  More information would be
> > useful -- especially if you could reproduce the problem in 2-d, as it
> > would be easier to debug then.
> > 
> > Mike
> > 
> > ------------------------------------------------------------------------------
> > Michael Zingale
> > UCO/Lick Observatory
> > UCSC
> > Santa Cruz, CA 95064
> > 
> > phone:  (831) 459-5246 
> > fax:    (831) 459-5265
> > e-mail: zingale at ucolick.org
> > web:    http://www.ucolick.org/~zingale
> > 
> > "Once something has been approved by the Government, It's no longer immoral."
> >    -- Reverend Lovejoy
> > 
> > 
> > 
> > 
> > 
> > 
> > 
> > On Wed, 5 Nov 2003, Robi Banerjee wrote:
> > 
> > > Hi,
> > > 
> > >   I ran into a problem (non convergence in the Riemann solver)
> > > using the new 'quadratic_cartesian' interpolation scheme in a
> > > 3D simulation including selfgravity. 
> > >   Looking into the details after a refinement step I figured 
> > > that the data in the newly refined blocks develop stripe-like
> > > structures. This leads, for instance, to overshooting in 
> > > velocities, resp. mach numbers. 
> > >   Using the old interpolation scheme (second_order_old) 
> > > everything is fine and the variables are smoothly prolongated 
> > > onto the newly refined blocks.
> > > 
> > >   Is this a known problem, or is there a problem with the
> > > new (quadratic_cartesian) interpolation scheme?
> > > 
> > >   If you need more information, please don't hesitate to
> > > ask for (e.g. 2D slices of the simulation data with the
> > > two different interpolations schemes)
> > > 
> > >   Thank you,
> > > Robi
> > > 
> > > ====================================
> > > Robi Banerjee
> > > Department of Physics and Astronomy
> > > ABB-320, McMaster University
> > > Hamilton, ON L8S 4M1
> > > CANADA
> > > e-mail: banerjee at physics.mcmaster.ca
> > > phone : (905) 525-9140 x 23189
> > > fax   : (905) 546-1252
> > >  
> > > 
> > 
> 
>