Title: RayleighTaylor Instability
1Rayleigh-Taylor Instability
Collaborators Vasily Zhakhovskii, M.
Horikoshi, K. Nishihara, Sergei Anisimov
2Motivation
There are several phenomena that influence the
RTI and RMI. They include viscosity, thermal
conduction, thermodynamic non-ideality, surface
tension, etc. Usual hydrodynamic approach does
not take into account these effects.
Quantitative comparison with experiment
requires a different approach. One such example
provides the method of molecular dynamics.
3Molecular Dynamics approach
- MD method is based on tracking of the atom
motions by means of - numerical solving Newtons equations
- MD method has great advantages over hydrodynamic
methods - Spatial mesh is not needed more
- EOS is not needed more
- The system under investigation can be far from
local thermodynamic equilibrium - Viscosity, heat conduction, surface tension are
taken into account automatically - Conservation laws are satisfied automatically
- MD method has the disadvantages (huge
computational time) - the total number of atoms
- so the system size 10TFLOPS)
4Molecular Dynamics Simulation Technique
z
Lennard-Jones pair potential
- Potential
- barrier as
- piston
L J atoms
Fij
-X
X
gravity
- Periodical boundary conditions are
- imposed on the system along z-axis
- The atoms interact via Lennard-Jones (L-J) pair
potential - with cutoff at rc
-
- The piston is simulated by an external potential
xi -X( fi )2
5Initial Condition
- Density Ratio 21 (Atwood Number A 1/3)
- Number of Atom 12 millions (8 millions of light
atoms) - Gravity 1011 cm/sec2
- Initial Amplitude 0.06? Single Mode
- Space Size 1274 s 858 s 13.4 s
- Pressure profile variable
- Computational Resource Pentium?-S 1.4GHz 80
(112GFLOPS) - Computational Time Almost 2 weeks
6Parallel Computing for load balancing.Dynamically
Re-distributed.
Actually atoms information was re-distributed.
Time Evolution
7Time Evolution of Interface.Density profile.
heavy atoms
light atoms
?
Each pixel represents a small domain, which is
occupied approximately 50 atoms
8Good Agreement with Theoretical Growthat linear
stage.
hExp(?t) where ?sqrt(AKg) L. Rayleigh (1900),
S.G. Taylor (1950)
9The thickness of mixed layer
10Conclusion
- We have developed the MD simulation code for RT
instability. - Parallelization for good load balancing by using
re-distribution - Simulation results agree with theoretical
prediction. - In the future
- More late time simulation (e.g., nonlinear
regime) - Making fully use of MD advantages (e.g.,
analysis around singularity) - New initial condition Hydro code does NOT
detail
11Interface evolution bydifferent hydro Schemes
Appendix
LL
CFLFh
JT
CLAW
WAFT
WENO
PPM
VH1
Ref http//www-troja.fjfi.cvut.cz/liska/CompareE
uler/compare8/