[FLASH-USERS] overwriting default density floor on FLASH

[Jung Lin Doris Lee]

Hi Christoph,

Thanks for the reply. We have switched to the unsplit hydro option and used
the HLL solver as you suggested and was able to override the default
smlrho, smalle, smallp settings. This allowed us to successfully adjust for
a smaller floor in our simulation.

Regards,
Doris and Aaron

On Wed, Mar 23, 2016 at 5:57 PM, Christoph Federrath <
christoph.federrath at gmail.com> wrote:

>
> Hi Doris, Hi Aaron,
>
> from my experience with this, it might not help much anyway to reset
> smlrho. The actual problem is that the PPM solver crashes, because it is
> not very robust (I’ve had a lot of these non-convergence problems in the
> past when I was still using PPM). You can try to set the CFL number smaller
> and/or nriem higher to push it further, but eventually, you’d need a more
> diffusive solver (e.g., HLLE) or (actually better) a stable solver that
> keeps density and pressure positive by construction (
> http://adsabs.harvard.edu/abs/2011JCoPh.230.3331W). The latter is not in
> the public version, but if you’d like to collaborate, I can send the solver
> to you. It is an MHD solver, but of course, you can use it with B=0 and I
> also have a stripped pure-hydro version of it.
>
> Also, since this is a collapse, eventually you will need to create a sink
> particle in the centre to cut off the collapse in a controlled way in order
> to follow what’s happening on larger scales (see sink particle technique:
> http://adsabs.harvard.edu/abs/2010ApJ...713..269F). The sink particle
> method is already in the public version, so feel free to switch it on and
> see if you can get further. The way you’d use the sink particles is
> explained in the user guide. Depending on what your science goal with this
> collapse simulation is, this is certainly something to consider in any case.
>
> I hope this helps.
>
> Kind regards,
> Christoph
>
> On 24 Mar 2016, at 11:26, Jung Lin Doris Lee <dorislee at berkeley.edu>
> wrote:
>
> Flash community,
>
> We have a (hopefully simple) question about how to best change the
> numerical floors for state quantities like density and pressure. For
> context, we are trying to reproduce the cloud collapse problem of Foster &
> Chevalier (ADS <http://adsabs.harvard.edu/abs/1993ApJ...416..303F>),
> which require densities of 10^-19~10^-27g/cm^3. The setup has a spherical
> core in the central ~1/2 of the domain with a radial profile in density.
> Outside the sphere is a tenuous and hotter medium (a reduced density but
> increased temperature so that it remains in pressure balance with the edge
> of the sphere).
>
> For example, when we initialize the data, we receive the following warning:
>
> WARNING after gc filling: min. unk(DENS_VAR)=0.7259099999999999388370E-20
> PE=0 block=10 type=1
>
> And the initial outputs show that the density has become entirely uniform
> after which the program crashes because of non-convergence in the Riemann
> subroutine. Grep-ing around, we see that there are quantities like “smallp”
> or “smlrho”, whose default values are 10^-10g/cm^3, indeed larger than the
> values we are initializing. We tried setting a value for smlrho in
> flash.par, but it did not appear to do anything.
>
> Can these values be reset in flash.par versus going into the source code?
> We are glancing through the user guide and example problems and can not
> find an example of how to do this.
>
> Best,
>
> Doris and Aaron
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://flash.rochester.edu/pipermail/flash-users/attachments/20160325/4a326c47/attachment.htm>