Re: glass making

From: Volker Springel <volker_at_MPA-Garching.MPG.DE>
Date: Fri, 23 Jun 2006 17:42:50 +0200

Paul Taylor wrote:
> Hi,
> I'm having a bit of difficulty setting up a glass-like system. I
> eventually am just making a shock tube; I am assuming that I have to
> make the two halves of the tube individually, because they're
> different densities, and then just use the coordinates from each to
> make the composite tube in a periodic box.
> However, what are the necessary Makefile conditions for making
> glass? In addition to the recommended, I have been using -Dperiodic,
> -Dsynchronization, -DLong_x,_y,_z, and -Dmakeglass=(number of
> particles). I also turned on the PMgrid because the make was giving
> me troubles, but that still has not resolved. Also, I am not sure
> what the initial conditions file should look like- should I randomly
> assign particles? The users-guide says that the value of the
> makeglass option 'gives the particle load, which is initially
> generated as a poisson sample and then evolved', etc.


for making a glass, you should set

 PMGRID=(mesh resolution)
 MAKEGLASS=(particle number)

Note that you cannot use LONGX/Y/Z, because you can only make a
gravitational glass, and the gravity solver only supports cubical
volumes. It is possible to use a gravitational glass as an "SPH glass"
later on, because it provides a reasonable representation of a constant
density field without imposing the regularity of a cartesian grid.

To obtain a long box as a SPH initial conditions, one can produce a
cubical glass which one then replicates in one (or two) dimensions
several times. Likewise, to produce two regions of different densities,
one can produce two individual glasses with different particle
densities, replicate them if needed, and then put them adjacent to each
other in a bigger box.

The parameterfile for making a glass should be one appropriate for a
cosmological simulation with Omega=1.0, and fixed timesteps. The values
of boxsize and the softening (of species 'Halo') don't matter, provided
the softening is set reaonably, like ~1/25 of the mean particle spacing
or so. MaxSizeTimestep and MinSizeTimestep should be set to equal
values. Again, the detailed values don't matter at all, only the total
number of steps made to produce a glass matters. The more steps are
made, the better the glass becomes. This can be monitored by the
residual rms-displacements that occurs per step (as stated in the
"glass-making" line of the stdout), which will initially drop quickly
and then ever more slowly.

For example, if one were to start from the lcdm_gas.param example
parameter file, one would have to change:

 Omega0 1.0
 OmegaLambda 0.0
 MaxSizeTimestep 0.01
 MinSizeTimestep 0.01
 SofteningGas 0
 SofteningHalo 60.0
 SofteningGasMaxPhys 0
 SofteningHaloMaxPhys 60.0

which would be roughly appropriate for MAKEGLASS=32768
InitCondFile will be ignored, and OutputDir has to be set of course, as
well as the frequency of snapshots.

Normally, a good glass is established for 128-256 steps. One can then
use the last snapshot output and rescale the cube to any desired size,
together with a new setting of the particle mass. The type of the
particles can be changed to 0 if a use in SPH simulations is desired.


> Cheers,
> PT
> -----------------------------------------------------------
> If you wish to unsubscribe from this mailing, send mail to
> with a subject of: unsubscribe gadget-list
> A web-archive of this mailing list is available here:
Received on 2006-06-23 17:42:50

This archive was generated by hypermail 2.3.0 : 2023-01-10 10:01:30 CET