Modify

Opened 10 months ago

Last modified 3 months ago

#2119 assigned defect

The binary neutron star gallery example gives different results with the release candiate.

Reported by: Peter Diener Owned by: Peter Diener
Priority: blocker Milestone: ET_2018_02
Component: Other Version: development version
Keywords: Cc: frank.loeffler@…

Description (last modified by Peter Diener)

The late part of the waveform plot looks very different in the release
candidate than on the gallery page. For l=2, m=2 mode of psi_4 extracted
at R=300 (that seems to be the correct data file as otherwise the scale on the y-axis don't match) the merger seem to happen slightly later and the waveform after merger is completely different.

The gallery page plot:

Gallery plot of psi_4

The new plot: New plot of psi_4

Attachments (5)

mp_Psi4_l2_m2_r300.00.png (11.2 KB) - added by Peter Diener 10 months ago.
Gallery plot of psi_4
mp_Psi4_l2_m2_r300.00_new.png (1.8 KB) - added by Peter Diener 10 months ago.
New plot of psi_4
mp_Psi4_l2_m2_r300.00_newer.png (2.1 KB) - added by Peter Diener 10 months ago.
With GRHydro::sources_spatial_order = 4
dens_compare.png (10.6 KB) - added by Peter Diener 3 months ago.
Plot of maximum of density as function of time
psi4_compare.png (14.9 KB) - added by Peter Diener 3 months ago.
Plot of l=2, m=2 of psi_4 as function of time.

Download all attachments as: .zip

Change History (16)

Changed 10 months ago by Peter Diener

Attachment: mp_Psi4_l2_m2_r300.00.png added

Gallery plot of psi_4

Changed 10 months ago by Peter Diener

New plot of psi_4

comment:1 Changed 10 months ago by Peter Diener

Description: modified (diff)

comment:2 Changed 10 months ago by Peter Diener

I will try to add "GRHydro::sources_spatial_order = 4" and see if this fixes this problem as well.

comment:3 Changed 10 months ago by Roland Haas

This may well be the same issue as with the single NS #2118, namely having to set GRHydro::sources_spatial_order = 4 since http://einsteintoolkit.org/gallery/bns/nsnstohmns.par does not set it and the default is 2. Admittedly normally a too low order should be more dissipative and should lead to an earlier merger but still, having and order of 2 makes little sense. One would have to go back and check whether at the time the example was created (by me) GRHydro used ADMBase::spatial_order or GRHydro::sources_spatial_order.

Changed 10 months ago by Peter Diener

With GRHydro::sources_spatial_order = 4

comment:4 Changed 10 months ago by Peter Diener

With GRHydro::sources_spatial_order = 4 the plot of psi_4 look more similar to the gallery page than before. However,there are still differences after merger as can be seen from this plot:
With GRHydro::sources_spatial_order = 4

comment:5 Changed 10 months ago by roberto.depetri@…

Quite likely the new run use the c++ code instead of the fortran one (that was the default at that time) and small difference can be present. A better comparison (for the post-merger phase) should be to consider (also) the module of psi_4 instead of just the real part. In the post merger phase there are more than one modes with diffrent constructive-distructive interference) and a small phase differnce on the two-three excited mode may show up. Running the Gallery example on Marconi.

comment:6 Changed 9 months ago by Steven R. Brandt

Owner: set to Peter Diener
Status: newassigned

comment:7 Changed 8 months ago by Steven R. Brandt

Peter, what is the status of this ticket?

comment:8 Changed 6 months ago by Frank Löffler

Cc: frank.loeffler@… added

Putting in my 2 cents, but no solution. I am currently also investigating a difference in our NSNS production runs when we move from the 2015_05 release to the current one. Details are of course different, but the most notable one is also a difference in the post BH-formation GW form, so CCing me here.

Last edited 6 months ago by Frank Löffler (previous) (diff)

comment:9 Changed 6 months ago by Steven R. Brandt

Peter, any progress?

Changed 3 months ago by Peter Diener

Attachment: dens_compare.png added

Plot of maximum of density as function of time

Changed 3 months ago by Peter Diener

Attachment: psi4_compare.png added

Plot of l=2, m=2 of psi_4 as function of time.

comment:10 Changed 3 months ago by Peter Diener

A part of the difference comes from a change to Meudon_Bin_NS to use EOS_Omni in commit: 1b9d6a3eff34224a357a61e3b74d5468f8efa82b. After this change some variables have slightly different profiles on the initial data slice. This affects: hydrobase::press, hydrobase::eps, grhydro::tau and grhydro::scon. The following plot shows the maximum density as function of time for 3 different runs. The red curve is for the original code at the time of the creation of the gallery example. The green curve is for the current development version of the toolkit. The blue curve is for the current development version with Meudon_Bin_NS reverted to the commit before the EOS_Omni commit.

Plot of maximum of density as function of time

As can be seen the blue and red curves agrees much better up to the merger, but still some differences appear after that.

The affect on the Psi_4 waveform can be seen here where I have zoomed in to the merger and post-merger phase:

Plot of l=2, m=2 of psi_4 as function of time.

Also in this case, the agreement is better between the red and blue curves, but significant differences still appear late in the waveform.

I don't know if Meudon_Bin_NS is more correct before or after the commit mentioned above. And I don't know what then causes the additional differences.

comment:11 Changed 3 months ago by Roland Haas

The ticket discussing the change Peter identified is #2039 .

Changes in Meudon_Bin seems quite likely to be capable of changing the initial density, these changes having an effect that only showsup very late (at t=1600) is certainly unexpected and somewhat disturbing.

Modify Ticket

Change Properties
Set your email in Preferences
Action
as assigned The owner will remain Peter Diener.
Next status will be 'review'.
as The resolution will be set.
to The owner will be changed from Peter Diener to the specified user.
The owner will be changed from Peter Diener to anonymous.

Add Comment


E-mail address and name can be saved in the Preferences.

 
Note: See TracTickets for help on using tickets.