Background
Gravitational softening is a technique used in many N-body simulations
to 1) make the system more collisionless (that is, less responsive to
close encounters between particles) and/or 2) to keep the integration
time well-behaved. Softening makes forces between particles have a
maximum value when they are close together. In this way, two
particles do not follow Keplerian orbits during close encounters and
timesteps have lower bounds. An argument for taking this approach is
that each particle in an N-body simulation actually represents an
unresolved distribution of mass. When these distributions get close,
their interaction will not be the same as for point masses. Former
UW-L student Jacob Gloe wanted to know if this is actually the case.
He helped implement the graphics-processing-unit-enhanced version of
the NBODY6 code (which does not involve softened forces), create
initial conditions, and run simulations of clusters of particles that
represent the unresolved point masses of softened N-body simulations.
I have used four different softening prescriptions. Prescriptions one
through three are "Plummer-like" in that they modify the
inverse-square law of gravity with a scalelength -- mass separations
smaller than the scalelength do not lead to increased gravitational
accelerations. Prescription four is a "spline" softening favored by a
commonly used N-body simulation package.
Simulations
The table below lists the various simulations we have run so far.
Clusters with N=2α (where 10 <= α <= 14)
have been created and placed so that they will travel past one
another, as would point masses in a softened N-body simulation. The
impact parameters (b) and relative speeds of
the interactions are also varied. The relative speed of two clusters
is determined by the parameter Q; larger
Q means larger speed. We
have also varied the initial separation
Δxinit
along the initial velocity direction.
The relative sparseness of the 'COMPLETED' tags results from the
insensitivity of results to the N value and the nearly-independent
behavior of results from different Q values.
Any links from 'COMPLETED' tags will take a reader to animations of
the corresponding simulation. You will see representative cluster
particles (small green and white dots) and the calculated locations of
the cluster centers-of-mass ('x' symbols). The softened point mass
locations are marked by diamond symbols. The specifics of the type of
softening prescription used and the adopted softening length are
shown. Each movie contains three different softening values; the
optimal softening, one-tenth of the lower uncertainty value, and twice
the upper uncertainty value. In this way, one can see how the two
kinds of motion diverge.
Δxinit = 4.0
|
b = 1.0
|
N
|
Q = 0.2
|
Q = 0.5
|
Q = 0.8
|
Q = 1.5
|
Q = 5.0
|
1024
|
-
|
COMPLETED
|
-
|
-
|
COMPLETED
|
2048
|
-
|
COMPLETED
|
-
|
-
|
-
|
4096
|
-
|
COMPLETED
|
-
|
-
|
-
|
8192
|
COMPLETED
|
COMPLETED
|
COMPLETED
|
COMPLETED
|
COMPLETED
|
16384
|
COMPLETED
|
COMPLETED
|
-
|
-
|
COMPLETED
|
b = 2.0
|
N
|
Q = 0.2
|
Q = 0.5
|
Q = 0.8
|
Q = 1.5
|
Q = 5.0
|
1024
|
-
|
COMPLETED
|
-
|
-
|
COMPLETED
|
2048
|
-
|
COMPLETED
|
-
|
-
|
-
|
4096
|
-
|
COMPLETED
|
-
|
-
|
-
|
8192
|
COMPLETED
|
COMPLETED
|
COMPLETED
|
COMPLETED
|
COMPLETED
|
16384
|
COMPLETED
|
COMPLETED
|
-
|
-
|
COMPLETED
|
b = 3.0
|
N
|
Q = 0.2
|
Q = 0.5
|
Q = 0.8
|
Q = 1.5
|
Q = 5.0
|
1024
|
-
|
COMPLETED
|
-
|
-
|
COMPLETED
|
2048
|
-
|
COMPLETED
|
-
|
-
|
-
|
4096
|
-
|
COMPLETED
|
-
|
-
|
-
|
8192
|
COMPLETED
|
COMPLETED
|
COMPLETED
|
COMPLETED
|
COMPLETED
|
16384
|
COMPLETED
|
COMPLETED
|
-
|
-
|
COMPLETED
|
b = 4.0
|
N
|
Q = 0.2
|
Q = 0.5
|
Q = 0.8
|
Q = 1.5
|
Q = 5.0
|
1024
|
-
|
COMPLETED
|
-
|
-
|
COMPLETED
|
2048
|
-
|
COMPLETED
|
-
|
-
|
-
|
4096
|
-
|
COMPLETED
|
-
|
-
|
-
|
8192
|
COMPLETED
|
COMPLETED
|
COMPLETED
|
COMPLETED
|
COMPLETED
|
16384
|
COMPLETED
|
COMPLETED
|
-
|
-
|
COMPLETED
|
b = 5.0
|
N
|
Q = 0.2
|
Q = 0.5
|
Q = 0.8
|
Q = 1.5
|
Q = 5.0
|
1024
|
-
|
COMPLETED
|
-
|
-
|
COMPLETED
|
2048
|
-
|
COMPLETED
|
-
|
-
|
-
|
4096
|
-
|
COMPLETED
|
-
|
-
|
-
|
8192
|
COMPLETED
|
COMPLETED
|
COMPLETED
|
COMPLETED
|
COMPLETED
|
16384
|
COMPLETED
|
COMPLETED
|
-
|
-
|
COMPLETED
|
Δxinit = 8.0
|
b = 1.0
|
N
|
Q = 0.2
|
Q = 0.5
|
Q = 0.8
|
Q = 1.5
|
Q = 5.0
|
1024
|
-
|
-
|
-
|
-
|
COMPLETED
|
8192
|
COMPLETED
|
COMPLETED
|
COMPLETED
|
COMPLETED
|
COMPLETED
|
b = 2.0
|
N
|
Q = 0.2
|
Q = 0.5
|
Q = 0.8
|
Q = 1.5
|
Q = 5.0
|
1024
|
-
|
-
|
-
|
-
|
COMPLETED
|
8192
|
COMPLETED
|
COMPLETED
|
COMPLETED
|
COMPLETED
|
COMPLETED
|
b = 3.0
|
N
|
Q = 0.2
|
Q = 0.5
|
Q = 0.8
|
Q = 1.5
|
Q = 5.0
|
1024
|
-
|
-
|
-
|
-
|
COMPLETED
|
8192
|
COMPLETED
|
COMPLETED
|
COMPLETED
|
COMPLETED
|
COMPLETED
|
b = 4.0
|
N
|
Q = 0.2
|
Q = 0.5
|
Q = 0.8
|
Q = 1.5
|
Q = 5.0
|
1024
|
-
|
-
|
-
|
-
|
COMPLETED
|
8192
|
COMPLETED
|
COMPLETED
|
COMPLETED
|
COMPLETED
|
COMPLETED
|
b = 5.0
|
N
|
Q = 0.2
|
Q = 0.5
|
Q = 0.8
|
Q = 1.5
|
Q = 5.0
|
1024
|
-
|
-
|
-
|
-
|
COMPLETED
|
8192
|
COMPLETED
|
COMPLETED
|
COMPLETED
|
COMPLETED
|
COMPLETED
|
b = 20.0
|
N
|
Q = 0.2
|
Q = 0.5
|
Q = 0.8
|
Q = 1.5
|
Q = 5.0
|
1024
|
-
|
-
|
-
|
-
|
COMPLETED
|
8192
|
-
|
COMPLETED
|
-
|
-
|
-
|
b = 24.0
|
N
|
Q = 0.2
|
Q = 0.5
|
Q = 0.8
|
Q = 1.5
|
Q = 5.0
|
1024
|
-
|
-
|
-
|
-
|
COMPLETED
|
8192
|
-
|
COMPLETED
|
-
|
-
|
-
|
b = 28.0
|
N
|
Q = 0.2
|
Q = 0.5
|
Q = 0.8
|
Q = 1.5
|
Q = 5.0
|
1024
|
-
|
-
|
-
|
-
|
COMPLETED
|
8192
|
-
|
COMPLETED
|
-
|
-
|
-
|
16384
|
-
|
-
|
-
|
-
|
COMPLETED
|
b = 32.0
|
N
|
Q = 0.2
|
Q = 0.5
|
Q = 0.8
|
Q = 1.5
|
Q = 5.0
|
1024
|
-
|
-
|
-
|
-
|
COMPLETED
|
8192
|
-
|
COMPLETED
|
-
|
-
|
-
|
b = 36.0
|
N
|
Q = 0.2
|
Q = 0.5
|
Q = 0.8
|
Q = 1.5
|
Q = 5.0
|
1024
|
-
|
-
|
-
|
-
|
COMPLETED
|
8192
|
-
|
COMPLETED
|
-
|
-
|
-
|
b = 40.0
|
N
|
Q = 0.2
|
Q = 0.5
|
Q = 0.8
|
Q = 1.5
|
Q = 5.0
|
1024
|
-
|
-
|
-
|
-
|
COMPLETED
|
8192
|
-
|
COMPLETED
|
-
|
-
|
-
|