Title: Carles%20Bona
1Checking AwA tests with Z4 (comenzando la
revolucion rapida)
- Carles Bona
- Tomas Ledvinka
- Carlos Palenzuela
- Miroslav Zacek
- Mexico, December 2003
2The Z4 system Physical Review D67, 104005 (2003)
-
- 10 Field equations
- R?? ??Z? ??Z? 8? (T?? T/2 g?? )
- 14 dynamical fields g?? , Z?
Covariant formulation with Z quantities to
monitorize (and maybe enforce in the future) the
constraint violations
3Z4 evolution equations
- (?t - L?) Kij - ?idj ? ? (3)Rij ?iZj
?jZi - - 2 K2ij (trK - 2?) Kij - Sij ½ (trS -
?) ?ij -
- (?t - L?) Zi ? ?k (Kki - trK ?ki) - 2 Kik Zk
- ?i ? - ? ?i/? - Si
- (?t - L?) ? ?/2 (3)R (trK - 2?) trK -
tr(K2) - 2 ?kZk 2 Zk ?k/? - 2?
-
? ? n?Z? ? Z0
4Generalized harmonic slicings
- 31 covariance
- t f(t) x g(x,t)
- (31)-covariant generalization
- (?t - L?) ln ? - ? f (trK - m ? )
Strongly hyperbolic iff fgt0 (harmonic, 1log,...)
5First order version of Z4gr-qc/0307067
- 1rst order variables
- (? , ?ij , Kij , ? , Zk , Ak , Dkij)
- Ak ? ?k(ln?) Dkij ? ½ ?k ?ij
- more constraints!
-
- supplementary evolution equations
- ?t Dkij ?k ? Kij 0
- ?t Ak ?k ? f (trK - m ?) 0
6Robust stability test
- Full 3D code with random small initial data
(almost linear regime --gt theorem) and periodic
boundaries - Finite differencing Method of lines
- Standard 3rd order Runge-Kutta in time
- 1st order systems standard centered 2nd order in
space - 2nd order systems there is an ambiguity (3 point
stencil or 5 point stencil?)
7Strong vs Weak Hyperbolicity
(dt0.03dx) slope of weak hyperbolic systems
grows with the resolution
8ICN results
(dt0.03dx) Numerical dissipation mask the
linear growth change the time integrator to RK3!!
9At the very end everything blows up
T 5 A(-1/3) for ADM T 4 A(-1/2) for
weakly Z4 T A(-3/2) for strongly
Z4---cosmological collapse?
10Suggestions to clarify Robust
- Changing the time integrator to RK3 and/or using
smaller courant factor - Using appropiate initial data (distribute
energy) for clear convergence tests - 2nd order systems using the 5 points scheme in
order to recover the theorem results or at least
comparing with the known results with 3 points
scheme - Plotting trK is enough to see if it works or not
11Gauge waves
- Go to http//stat.uib.es
- We can check the linear and nonlinear regime, the
numerical method, study the numerical
instability - Change A0.1 to A0.5
- Study with one fixed formulation the different
numerical methods (second or fourth order in
space, 3 and 5 points scheme for second order
systems, dissipation,.)
12Collapsing Gowdy waves
- Cosmological solution (vacuum) with periodic
boundaries - ds2 t-1/2 eQ/2 (-dt2 dz2) t (eP dx2 e-P
dy2) - P(t,z), Q(t,z) periodic in z (pp wave)
- Harmonic slicing
- t t0 exp(-t/t0)
- Testing the source terms
13Lapse collapse (Harmonic slicing)
14Oscillation Collapse
Things starts to be different at 2000 crossing
times..then evolve up to 10.000
15Z3 parameter space n
Studying the sources of the formulation (adding
energy, redefining variables,...)
16Conclusions
- Plot trK with robust and gauge waves should be
enough - Use RK3 for the tests to avoid dissipation effect
that can mask the formulations - Remove/replace the linear waves they do not give
any new information - Be careful with the stencil scheme (3-5) if you
use second order systems!! (do you want to test
the formulation or the numerical method?) - Change the gauge waves amplitude (A0.1 to A0.5
to study a strong non linear regime) - Evolve the Gowdy up to 10.000 crossing times
17Boundary test suggestions
- Robust stability with boundaries define exactly
the domain, face-edge-corners,.. - 2D radial gauge wave (or gauge wave packet) with
boundaries exact solution not known, but a lot
of things to see (constraint violation,
reflections,...)! - Static solution (without excision or too large
gradients) with boundaries (ideas, suggestions?) - Wave moving in the static previous solution with
boundaries
18General suggestions
- We need more agressive (but isolated) tests
with/without boundaries (it does not matter if we
do not know the exact solution! Convergence tests
are there) - We have to study in more detail some of the tests
like gauge waves to see what we can expect - Hurry, hurry, hurry! It is not difficult make all
the tests, we can not wait more than few months
(2-3) to see the results, compare and take some
results.
Check with hyperbolic system
Suggest new test
If it is not useful
If it is useful
Everybody make the test and compare