| Black
Hole Merger Simulations
Physicists at the Max Planck Institute for Gravitational Physics
are performing simulations of the spiraling coalescence of two
black holes, a problem of particular importance for interpreting
the gravitational wave signatures that will soon be seen by
new laser interferometric detectors around the world. Detection
of the first gravitational waves (or failure to do so) will
strongly test Einstein’s Theory of General Relativity,
the results of which will have ramifications that extend throughout
the world of physics. The Cactus simulation code is being used
to perform the calculations. This is the first time ever that
a spiraling merger of this type has been accurately simulated
(Figure 5). The results so far indicate that the Meudon model
for coalescence seems to match the simulation data more accurately
than the competing Cook-Baumgarte model.
 |
|
 |
|
| Figure
5 Visualization of binary black hole
inspiral. |
|
Collisions between black holes should theoretically create
propagating gravitational waves, similar to the electromagnetic
waves given off by distant stars. These ripples in space-time
should be seen as subtle variations in the length of objects
as they move through space. Recently built laser interferometric
detectors such as LIGO and VIRGO are capable of measuring
these subtle ripples in space. However, the gravitational
wave signal that can be detected by these interferometers
is so faint that it is very close to the level of noise in
these devices. So simulations of the kinds of events that
might produce gravitational waves can provide important insights
into the gravitational wave signature produced by these events,
potentially making the instruments more productive.
The Cactus code performs a direct evolution of Einstein’s
equations, which are a system of coupled nonlinear elliptic
hyperbolic equations that contain thousands of terms if fully
expanded. Consequently, the simulation resource requirements
are enormous just to do the most basic of simulations. The
simulations have been limited by both the memory and CPU performance
of supercomputers as they attempt to move from calibrating
against analytic black hole solutions to non-analytic astrophysically
relevant cases in full 3D. The spiraling merger is just such
a non-analytic case.
These simulations must use more than one-third of the NERSC
IBM SP’s available aggregate memory of 4.3 TB in order
to achieve the resolution required to accurately simulate
these phenomena. This simulation uses 1.5 TB of memory and
more than 2 TB of disk space for each run. These runs typically
consume 64 of the large-memory nodes of the SP (a total of
1,024 processors) for 48 wall-clock hours at a stretch. The
simulation can use all 184 nodes, but this would only allow
simulations that are fractionally larger than using the large-memory
nodes due to memory/load-balancing issues.
NERSC provided access to a special queue to improve turnaround,
opened ports to allow remote steering and Grid access, and
provided consulting support for 64-bit integration and code
debugging. In the space of two months, this simulation consumed
700,000 CPU hours, simulating three-fourths of a full orbit
before coalescence. In the near future, this project could
use 10 TB of disk for each run, 5 TB of uniform, user-available
memory, and 15 million CPU hours.
INVESTIGATORS
E. Seidel, M. Alcubierre, G. Allen, B. Brügmann, P. Diener,
D. Pollney, T. Radke, and R. Takahashi, Max Planck Institute
for Gravitational Physics; J. Shalf, Lawrence Berkeley National
Laboratory.
PUBLICATIONS
M. Alcubierre, B. Brügmann, P. Diener, M. Koppitz, D.
Pollney, E. Seidel, and R. Takahashi, “Gauge conditions
for long-term numerical black hole evolutions without excision,”
Phys. Rev. D (in press).
M. Alcubierre B. Brügmann, P. Diener, F. Guzman, S. Hawley,
M. Koppitz, D. Pollney, and E. Seidel, “Dynamical evolution
of binary black hole data,” in preparation (2002).
URL
http://www.aei-potsdam.mpg.de/
|
