[Users] question about ML_CCZ4 parameters

[Federico Guercilena]

Dear Bruno,

contrary to Erik, I was not around when ML_CCZ4 was implemented, but I used
it (or tried to use it) rather extensively.

Regarding the parameter named k3 in the papers, I'm pretty sure that's the
one named GammaShift in the code. It is not hard coded, you can change it
in the parfile, but its default value is 0.5 . As noted in the two original
CCZ4 papers, this value breaks the covariance of the system (it should be
set to 1. for it to be restored), but it allows to handle BH singularities.

The values of 0.05 and 0 for k1 and k2 (which do correspond to the k1 and
k2 of the papers) are "standard" as far as I can tell. Another value I've
seen a lot in parfiles for k1 is 0.036, while as far as I know k2 has
always been set to 0. I cannot recall any real difference in BNS runs with
k1=0.05 or k1=0.036, but I never tried to systematically tune these
parameters to "optimal" values.

Finally, I want to note that I had a lot of subtle problems in using
ML_CCZ4 that made me switch to ML_BSSN. These would not be outright
crashes, but larger than expected oscillations of stars that in the long
term made some simulations unusable to get realistic data, despite the GR
constraints violations being much lower than ML_BSSN. When I first
encountered this problem, I somehow got the impression that they might be
related to some conflict with boundary conditions, but now I'm starting to
think that they are related to the values of k1, k2 and k3 not being tuned
properly, or maybe even by the system not being strongly hyperbolic (note
that as far as I know, there is no actual proof of strong hyperbolicity for
the original version of CCZ4 as implemented in McLachaln in the literature.
Other versions of the system have been proved hyperbolic, but not this one).

Actually, if you and your student took the time to do some tuning and
managed to improve the performance of this code, that would be pretty
interesting in its own right, I think.

I hope this was of some help.

Best,
Federico
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.einsteintoolkit.org/pipermail/users/attachments/20220602/4a241fdc/attachment.html