[FLASH-USERS] resolving small scales on a large domain

[Mario Manuel]

Hi Roman,

Thanks for the quick response. Using the lrefine_min_init=2 (I also did =3) worked fine and with lrefine_max set 8 gives me a few resolution elements for the amplitude of the perturbation. I have also updated the domain and nblockx/nblocky to reflect a square cell.

… now if I only had some more processors to run the simulation on. :)

Any thoughts on setting -maxblocks in the setup file? I originally had the default (1000 I think), and I’ve now upped it to 5000. My computer hasn’t exploded yet, but I’m not really familiar with what determines an optimum number for maxblocks.

Cheers,
Mario

______________________________________
                                                                                                                                
Dr. Mario Manuel
Einstein Postdoctoral Fellow, Smithsonian Astrophysical Observatory
mmanuel at umich.edu
617-899-8381

University of Michigan
Space Research Building
2455 Hayward Street
Ann Arbor, MI 48109
Office: 734-764-4585
______________________________________

On Sep 16, 2014, at 4:40 AM, Roman Yurchak <roman.yurchak at crans.org> wrote:

> Hi Mario,
> 
> I think you must be missing the the lrefine_min_init parameter in your
> flash.par that sets the minimum number of refinement levels at init time.
> 
> For instance with nxb==nxy==16, to resolve a 0.27um perturbation, I
> assume you would want at least
> lrefine_max = 8 # with max resolution  300./(16*(2**(8-1)) = 0.146 µm
> then you would have to set
> lrefine_min_init = 2  # i.e. 9.375 µm, as with a default value of 1
> (=18.75µm) you are on the limit of detecting the foil at init time.
> 
> Besides, it is advised to have square simulations cells, that can be
> achieved for a 300 um X 1200 um domain with
> nblockx = 1
> nblocky = 4
> 
> Best,
> -- 
> Roman Yurchak
> PhD student
> Laboratoire LULI
> Ecole Polytechnique, France
> 
> 
> On 15/09/14 23:49, Mario Manuel wrote:
>> Hi,
>> 
>> I am trying to setup a simulation that looks at RT growth of seeded
>> perturbations on a laser-accelerated foil. The foil parameters I'd like
>> to implement is a 21um thick foil with a 0.27um sinusoidal amplitude
>> perturbation on it. I have altered the LaserSlab simulation initBlock to
>> do this and I can successfully create a 21um thick foil with a 2um
>> perturbation, but as I try to make the perturbation smaller, the foil
>> entirely disappears from the domain. I have tried to change the
>> lrefine_max parameter to get the resolution I need, but have been thus
>> far unsuccessful at getting down to the sub micron level that i need to
>> run the simulation. I also ran the ./setup with nxb=32 and nyb=32 to try
>> to get at the problem, but still to no avail.
>> 
>> The domain is rather large, 300 um X 1100 um, due to the fact that I
>> need to have some computational space given to the laser-plasma
>> interaction and to allow the foil to accelerate and move during the
>> simulation.
>> 
>> Any thoughts on how I could better implement my problem?
>> 
>> Cheers,
>> Mario 
>> 
>> ______________________________________
>> 
>> 
>> Dr. Mario Manuel
>> Einstein Postdoctoral Fellow, Smithsonian Astrophysical Observatory
>> mmanuel at umich.edu <mailto:mmanuel at umich.edu>
>> 617-899-8381
>> 
>> University of Michigan
>> Space Research Building
>> 2455 Hayward Street
>> Ann Arbor, MI 48109
>> Office: 734-764-4585
>> ______________________________________
>> 
> 
> 
> -- 
> Roman

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://flash.rochester.edu/pipermail/flash-users/attachments/20140916/4c614130/attachment-0001.htm>