Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
I. INTRODUCTION
The motion of two point masses interacting gravitationally
according to Newtons laws is simple: the bodies revolve in
ellipses about their common center of mass. The motion of
three such bodies is surprisingly complex. Even when two of
the bodies are constrained to orbit each other in circles at
constant speed and a third is restricted to the plane of their
orbits, the dynamics can be arbitrarily complicated. The famous analysis of this restricted three-body problem by
Poincare1 in 188990 led to the discovery of dynamical
chaos generations before the advent of digital computers and
underscored the futility of seeking a general solution in terms
of a finite number of known functions. Still, by 1912, Sundman had discovered an infinite series solution.2 Unfortunately, it converged too slowly to be practical and was too
unwieldy to be insightful. By 1967, Szebehely and others
adopted a statistical perspective, and concluded that the generic solution of the unrestricted three-body problem is the
escape of one of the three bodies.3 For almost all initial
conditions, as kinetic energy shuffles back and forth among
the bodies, one body eventually gains sufficient energy to
enter an unbounded hyperbolic orbit, while the other two
lose energy and enter the bounded elliptical orbits of a binary
system.
In the 1980s, a controversy arose concerning a seemingly
elementary special case of the three-body problem: What is
the minimum speed needed to escape from the solar system
envisioned as a three-body problem involving the earth, sun,
and a projectile? Hendel4 pointed out that many of the standard textbook solutions to this problem were incorrect. His
own approximate solution was later criticized by some5 and
defended by others.6 Hendel and Longo7 subsequently presented a computer simulation that bypassed the previous
mathematical approximations and vindicated Hendels earlier
work. Scudder significantly elaborated these results.8
We revisit solar escape in the context of the restricted
three-body problem and elucidate some of its features using
both theoretical and computational techniques. We consider a
solar system of two bodies of arbitrary mass and size, which
we still call earth and sun. This model places the special case of our actual earth and sun in a broader context. We
neglect the rotation, but not the revolution, of the earth.
Section II introduces the restricted three-body problem in
stages, derives the equations of motion, integrates them to
find a famous constant of the motion, and uses the constant
871
Mm
r
r G
Mm
r3
r ,
GM
r3
GM
r3
x,
2a
y,
2b
http://ojps.aip.org/ajp/
871
ma F G
M Em
d E2
d E G
M Sm
d 2S
d S .
GM E
d E3
GM E
d E3
GM S
xr E cos t
yr E sin t
xr S cos t , 7a
d 3S
GM S
d 3S
yr S sin t . 7b
In rotating coordinates (x R ,y R ), the earth and sun are stationary and the equations of motion do not explicitly depend
on time. In fact, by substituting the rotation transformation of
Eq. 5 into Eq. 7a and equating the coefficients of cos t
and sin t, we obtain
Fig. 1. Geometry of the three-body problem. The large masses M E and M S
move in concentric circles at constant speed. They attract, but do not react
to, the tiny mass m.
2
EC
M EM S
F E G
,
rE
r2
5a
yx R sin ty R cos t,
5b
y R 2 x R 2 y R
GM E
d E3
GM E
d E3
x R r E
y R
GM S
d 3S
GM S
d 3S
x R r S ,
8a
8b
yR .
E. Constant of motion
As in the derivation of Eq. 3, we multiply Eq. 8a by
x R , Eq. 8b by y R , add the two equations, and integrate
with respect to time to find
R2 x R2 y R2 2 x R2 y R2 2G
ME MS
C,
dE
dS
GM E GM S
3 1
1
C x R2 y R2
2
2 2
dE
dS
872
x R 2 y R 2 x R
12 2 x R2 y R2 23
10
872
projectile is trapped near the earth and the sun. The latter
numerically computed10 trapping energy is plotted in Fig. 2
bottom versus f. Each such energy corresponds to a speed
below which escape is impossible. We refer to these lower
limit speeds as trapping speeds.
III. INDEPENDENT ESCAPES
A. Mixed formula
We can approximate the minimum escape speed for the
projectile as a function of launch angle using the technique
of Hendel.4 We will assume that the radius of the earth is
much less than the distance between the earth and the sun,
R E r. We can then effectively divide the escape into two
parts: an escape from the earth alone, followed by an escape
from the sun alone. With respect to the earth, energy conservation implies that
GM E m 1
1
2
m PE
,
m 2PE
2
RE
2
11
2
r
12
PC PE EC ,
13
Fig. 3. Rectangular top and polar bottom plots of the minimum escape
speeds in the independent escapes approximation, with respect to the earth
dashed lines and with respect to the solar system center of mass solid
lines. The minimum escape speeds in the earth frame are largest in the
backward direction and smallest in the forward direction. To magnify the
angular variations, the center of the polar plot corresponds to a normalized
speed of /V E 0.9.
2
PE V E2 V 2S EC
2V S EC cos C ,
14
PC EC cos C EC cos C V S 2 V E2 ,
15
873
C. Qualifications
The independent escapes approximation should be used
with a couple of caveats. In the unrestricted three-body problem, even with this approximation, as the projectile leaves
the earth, it slows down and the earth speeds up due to their
mutual gravitational attraction.4 However, in the restricted
three-body problem, the earth is assumed to move at constant
speed. Furthermore, the approximation assumes energy is
conserved, but in the restricted problem only the pseudoenergy of Eq. 10 is conserved. Nevertheless, as we shall see
below, the approximation is reasonably good at reproducing
the coarse features of the escape speeds variations with
angle.
IV. SIMULATION
A. The
ESCAPE
program
As Poincare demonstrated, the restricted three-body problem exhibits extreme sensitivity to initial conditions, a phenomenon that is now known as chaos. Exact solutions are
rare and numerical integration is an invaluable tool. Consequently, we created a computer simulation called ESCAPE11 to
further investigate the escape problem.
ESCAPE numerically integrates the differential equations of
motion using a fourth-order RungeKutta algorithm with a
variable time step dt. Guided by Keplers third law, we let
dt
d
k
T
r
3/2
16
Fig. 4. Each of our solar system models can be characterized by the earths
mass and radius, which corresponds to a point in the parameter space
(M E ,R E ). In this space, solar systems with the same earth-alone escape
speed V E lie along lines of slope 1, while solar systems with the same earth
surface gravity g lie along lines of slope 1/2.
PC / 0 1 f cos C
1 f cos C 2 1 f 2 V E / 0 2 ,
17
where the mass fraction f M E /M implies 1 f M S /M .
Note that if the escape speed from the earth alone is high, so
that V E 0 , then Eq. 17 implies PC V E .
C. Parameter space
We first have to choose the mass and size of the earth. We
did this systematically in several ways as illustrated in Fig. 4.
One technique was to adjust the radius of the earth so that
the escape speed from the earth alone was a constant, say
V E 20 0 600 km/s. Because R E /r2(M E /M )( 0 /V E ) 2
f 1 , varying the mass of the earth at constant V E traces out
a straight line of slope one in the loglog plot of Fig. 4.
Another technique was to adjust the radius of the earth so
that the acceleration due to gravity at the earths surface was
a constant, say g1650g 0 9.8 m/s2 . Because R E /r
(M E /M )(g 0 /g) f 1/2, varying the mass of the earth at
constant g traces out a straight line of slope 1/2 in the log
log plot.
We obtained similar results for both strategies. Here we
present our data for constant V E 20 0 . Our actual solar
system has V E 0.37 0 , but this relation cannot be extrapolated to high mass fractions without the earth and the sun
overlapping.
D. Direct escapes
We first used the simulation to investigate direct escapes.
Direct means leaving the earth and exiting the solar sysHarmon, Leidel, and Lindner
874
tem without falling back. We took escape to mean exceeding a distance that is large compared to the size of the solar
system. In practice, we chose the escape distance to be
1000r. We ran copies of the program concurrently on clusters of small computers, each copy investigating a different
range of angles for a given mass fraction. The program used
a binary search algorithm to bracket the true minimum escape speed to any desired accuracy.
Figure 5 illustrates the minimum direct escape speeds as a
function of angle for three mass fractions. Launching at an
angle of zero corresponds to launching in the direction of the
earths motion tangent to its orbit. We noticed two interest-
Fig. 7. Rectangular top and polar bottom plots of trapping speed, numerically computed escape speeds, and independent escapes approximation
speeds, for a mass fraction of f 0.3. As in Fig. 3, the polar plot magnifies
the angular variations.
875
Fig. 9. Two diverging orbits, with nearby initial conditions from the chaotic
region of the Escape Set, in an inertial reference frame.
Fig. 8. Rectangular top and polar bottom plots of the Escape Set.
Gradations correspond to the distance of the projectile from the center of
mass of the solar system after 5 years, for various initial speeds and angles.
At low speeds the projectile is trapped in the solar system. At large speeds,
the projectile easily escapes. At intermediate speeds above the trapping
speed and below the minimum direct escape speed, the projectile typically
executes intricate chaotic orbits with many returns and close encounters, but
also arbitrarily large excursions.
V. CONCLUSIONS
We have investigated the optimal escape of a projectile
launched from the solar system within the context of the
restricted three-body problem. Noteworthy features include
the minimum escape speeds at small nonzero angles, the observation that larger angle escapes utilizing gravity assists
are more effective at large mass fractions, and the visualization of the chaotic region between the direct escape and
trapped orbits.
We have assumed that the projectile is confined to the
plane of the orbits of the earth and the sun, which move at
constant speed in concentric circles. Future work may relax
some of these restrictions. Because the generic solution of
the unrestricted three-body problem is the escape of one of
the three bodies from the other two, relaxing our constraints
to allow nonplanar motion or primaries with noncircular orbits or a projectile with noninfinitesimal mass should increase the number of possible escape scenarios. Initial conditions that trap the projectile under the restricted three-body
problem might free the projectile for the general three-body
problem. The motion of the projectile could cause the primaries to move together, decreasing their mutual gravitational
potential energy, and thereby freeing more kinetic energy for
the projectile. Interesting effects might also be obtained by
including the earths spin, by launching the projectile out of
the plane of the earth and the sun, or by modeling the earth
and the sun as solid spheres with which the projectile might
collide.
Escape problems in the Newtonian N-body problem are
currently of practical concern in other contexts, such as estimating the rate at which Earth-crossing asteroids escape
from the asteroid belt between Mars and Jupiter.15 From a
theoretical point of view, Xia has recently rigorously demonstrated a noncollision singularity in the Newtonian five-body
problem.16 His scenario involves one point mass shuttling
back and forth between two binary systems and being ejected
to infinity in a finite time. Although Painleve proved that the
three-body problem is free of such singularities,17 Xias sceHarmon, Leidel, and Lindner
876
Reflection of Particles. From the 1881 catalogue of James W. Queen of Philadelphia: Apparatus to Illustrate the Reflexion of Motion. Consisting of a
semicircular table of cherry, a marble slab for reflector, a spring pistol, ivory ball, and pocket to catch the ball after the rebound. Edge of table graduated to
show the relation between the angles of incidence and reflexion. ...$20.00. Two similar pieces of apparatus exist: John Tyndalls demonstration to show the
specular reflection of visible light, and his apparatus for showing the specular reflection of sound. It seems that when introducing students to specular
reflection, a particle model describes the situation well. Newton thought so, too. Photograph and notes by Thomas B. Greenslade, Jr., Kenyon College
877
877