On Wed, Jul 18, 2012 at 5:41 PM, Eloisa Bentivegna bentivegna@cct.lsu.edu wrote:

On Jul 18, 2012, at 6:19 PM, Petr Tsatsin wrote:

...

Hello, I'm trying to use AHFinderDirect in the simulation of the NSs merger and was wondering if somebody can suggest how to set up the parameters for this thorn? More precisely I simulate two neutron stars tracked by moving grids. I can evolve this until merger and simulation fail once BH horizon has formed. I would like to use AHFinderDirect to look for a AH.

Hi Petr,

first off: is the simulation failing due to AHFinderDirect?

...

Particularly I have a several questions:

1. Do I need 3 centres of refinement in order for AHFinder to work? By 3 centres I meant 2 moving grids and one stationary at the origin. (I just looked at qc0 parameter file as an example)

If you're moving your grids based on the maximum-density tracker, then AHFinderDirect doesn't have anything to do with them. You can choose how many horizons to look for, where, and when in a completely independent way. You can look at AHFinderDirect's param.ccl to find out about all the different options you can use, but in most cases what's in qc0-mclachlan.par should be all you need:

AHFinderDirect::origin_x [1] = +1.168642873 AHFinderDirect::initial_guess__coord_sphere__x_center[1] = +1.168642873 AHFinderDirect::initial_guess__coord_sphere__radius [1] = 0.25 AHFinderDirect::which_surface_to_store_info [1] = 0 AHFinderDirect::reset_horizon_after_not_finding [1] = no AHFinderDirect::dont_find_after_individual_time [1] = 30.0

Naturally, depending on where you expect your initial horizon to be, you may also need to use origin_y and origin_z (and initial_guess__coord_sphere__y_center and initial_guess__coord_sphere__z_center). The same holds for the "find_after_individual_time" parameter. Notice the important parameter "which_surface_to_store_info": this tells AHFinderDirect which surface you have set up (through SphericalSurface) to store the horizon's properties. Make sure you request a surface exclusively for this purpose in your parameter file.

Petr

In addition to what Eloisa explained, I want to add that you need to guess not only the location and size, but also roughly the shape of the initial horizon; if it is very distorted, AHFinderDirect will not find it with a spherical guess.

I would save a checkpoint file, and then (as Eloisa suggested) run a job that examines only this time step, looking for ten or twenty initial guesses. Without matter, the surface where \alpha=0.3 is a reasonable guess.

If you get stuck, then I can explain (maybe better over Skype) how to use pretracking. In short, this is a method to track certain spacetime features (starting before the horizon forms) that may evolve into apparent horizons at some point. However, I expect that a bit of guessing will be simpler, and will work just fine.

...

2. What's going to happen with a 2 moving grids when apparent horizon was formed? Since I use maximum density tracking mechanism to track the positions of the NSs I want to switch the moving grids off after the formation of the apparent horizon and have only one centre of refinement at the origin.

This behavior is also not controlled by AHFinderDirect, but by CarpetRegrid2. The only way I know to trigger grid changes like you describe is to write your own thorn which inherits from CarpetRegrid2 and changes its variables num_levels[i] as desired. Perhaps someone else can give better advice here.

You do not need to switch off these additional grids, unless the greatly enlarge the fine grid you are evolving. Carpet combines the refined regions, and if you have three approximately equal regions, the combined region will work just fine. Is there a particular reason you want to switch them off?

-erik

On Thu, Jul 19, 2012 at 12:37 AM, Bruno Giacomazzo bruno.giacomazzo@jila.colorado.edu wrote:

2. What's going to happen with a 2 moving grids when apparent horizon was

formed? Since I use maximum density tracking mechanism to track the positions of the NSs I want to switch the moving grids off after the formation of the apparent horizon and have only one centre of refinement at the origin.

This behavior is also not controlled by AHFinderDirect, but by CarpetRegrid2. The only way I know to trigger grid changes like you describe is to write your own thorn which inherits from CarpetRegrid2 and changes its variables num_levels[i] as desired. Perhaps someone else can give better advice here.

You do not need to switch off these additional grids, unless the greatly enlarge the fine grid you are evolving. Carpet combines the refined regions, and if you have three approximately equal regions, the combined region will work just fine. Is there a particular reason you want to switch them off?

I think the problem in his case is that the grids follow the position of the maximum of the density. What we do with Whisky simulations of NS-NS mergers is to stop the movement of the grids after the merger (so you will end up having a set of fine grids fixed where the BH will form). If you don't do that, after the collapse to BH the positions of the grids would be quite "random". For example, if you are left with a disk outside the BH, then the centers of the grids will follow the position of the maximum density in the disk, which is not what you want usually.

Thorn CarpetRegrid2 can be "steered" at run time by modifying its grid scalars (NOT its parameters!). You will then need to write a thorn to e.g. set active[0] and active[1] to zero.

-erik