On Mon, Apr 19, 2021 at 10:39 AM Konrad Topolski k.topolski2@student.uw.edu.pl wrote:
I have been trying a variety of initial guesses, but to no avail.
The setup is, in short: par_b = 3.0 target_m_plus = 0.25 target_m_minus = 0.75
And the smaller black hole's apparent horizon gets detected easily, which cannot be said of the bigger one. My estimates for the AHFinderDirect were at -1.05 to -0.95 for the center of the ellipsoid/sphere, with initial guesses for the x_radius ranging from 0.25 to 0.4, with y_radius, z _radius about 10-20% smaller.
Am I still not getting enough resolution? The refinement levels in my finest setup around x=-1 are:
Carpet::max_refinement_levels =8 CarpetRegrid2::movement_threshold_1 = 0.16 CarpetRegrid2::num_levels_2 = 8 CarpetRegrid2::position_x_2 = -1.0 CarpetRegrid2::radius_2[ 1] = 64.0 CarpetRegrid2::radius_2[ 2] = 16.0 CarpetRegrid2::radius_2[ 3] = 8.0 CarpetRegrid2::radius_2[ 4] = 4.0 CarpetRegrid2::radius_2[ 5] = 3.0 CarpetRegrid2::radius_2[ 6] = 2.5 CarpetRegrid2::radius_2[ 7] = 2.0 CarpetRegrid2::movement_threshold_2 = 0.16
I do not know a-priori if this apparent horizon is best approximated by an ellipsoid (can one see this before performing the numerical simulation?) or not, but I don't know if I'm on the right track. It seems weird to me that I cannot find the x=-1 AH, even though the corresponding black hole it is ~3 times more massive.
In another attempt, I tried to copy the philosophy of the setup in the GW150914.rpar file, setting the center_offset for the two punctures and other parameters, but that didn't work either. I took target_mp=0.75, target_mm=0.25, separation D=6, so xp=D*mm = 1.5, xm=D*mp=-4.5, par_b=D/2 = 3.0, center_offset=xp-half_D=-1.5 (all here: https://gist.github.com/konrad-topolski/1c208415d3edac2f428ded637153f639) I'm starting to doubt if any of these setups is correct.
By the way, can I somehow redirect the outputs of all MPI processes to one file at real time, better still, override the creation of CCTK_Proc files altogether?
No, this is not possible. This would be very expensive because it would be necessary to lock this file for every output. However, since this is just a test, you can run on a single MPI process.
-erik