N-BODY PROBLEMS Math118, O. Knill THE 1. KEPLER LAW.

An ellipse with focal points S 0 =

(−a, 0), S = (a, 0) is the set of points (x, y) whose distances r 0
ABSTRACT. The Newtonian n-body problem influenced the development of mathematics at several occasions. and r to S 0 and S satisfy r0 + r = 2a. The number  is called the
For example, it was the catalisator for the development of calculus or topology. In this section, we look at eccentricity. From (2a−r)2 = r02 = r2 sin2 (θ)+(2a+r cos(θ))2 ,
general facts about the n-body problem like existence of solutions and the nature of singularities like Zeipel we obtain r = a(1 − 2 )/(1 +  cos(θ)), the polar form of
theorem which distinguishes collision and noncollision singularities by the convergence of the moment of inertia. the ellipse. Differentiation of this with respect to time, using (x,y)

θ̇ = L/(mr2 ) leads to ṙ = a(1−2 ) sin(θ)(1+ cos(θ)−2 L/(mr2 ) =

L sin(θ)/(ma(1 − 2 )) and r̈ = L2  cos(θ)/(m2 a(1 − 2 )r2 ).
NEWTON EQUATIONS. r’ r
With n = x/r, one has ẍ = (ẍ · n)n and x = nr gives ẋ =
HAMILTON EQUATIONS FOR THE N-BODY
ṅr + nṙ, ẍ = n̈r + 2ṅṙ + nr̈ so that ẍ · n = n̈ · nr + 2ṅ· nṙ + r̈. Using θ
Celestial mechanics is the study of the Newto- PROBLEM.
n · n = 1 ⇒ n · ṅ = 0, n̈ · n + ṅ · ṅ = 0 and ṅ · ṅ = θ̇2 n⊥ · n⊥ =
nian n-body problem, the study of the differential (-aε,0)= S’ (0,0) S =(aε,0)
θ̇2 = L2 /(mr2 )2 , we have
equations A point (x, y) with x = (x1 , . . . , xn ), y = (y1 , ..., yn )
P m m (x −x ) in the phase space encodes both the positions xi and
mj ẍj = −G i6=j j|xii−xjj |3 i ẍ · n = r̈ − L2 /(m2 r4 )
the momenta yi = mi ẋi of the bodies. The function
The vectors xj are the positions of the bodies with Xn X mi mj With 1/r4 = (1/r)(1/r 3 ) = (1 + )/(r 3 a(1 − 2 )) and the formula
yj2
mass mj and G is the gravitational constant. If H(x, y) = − U (x), U (x) = G for r̈ we get ẍ · n = L2  cos(θ)/(m2 a(1 − 2 )r2 ) − (L2 /(m2 ))(1 +
the initial positions and velocities of the bodies are j=1
2m j i<j
|xi − xj | )/(r3 a(1 − 2 )) = −L2 /(a(1 − 2 )r2 so that ẍ = (ẍ · n)n =
known, then the equations determine the position of −L2 /(a(1 − 2 ))x/r3 = −Gx/r3 .
the bodies at later times as long as solutions exist. on the phase space is the energy of the particle
The phase space of the system is the 6n-dimensional system. One calls it the Hamiltonian. The Newton
THE 3. KEPLER LAW. If T is the period of the orbit, the the third Kepler law states that T 2 /a3 is constant.
space M × R3n , where M = R3n \ ∆ with the colli- equations can be rewritten as Hamilton equations ˙
Indeed, if f (t) is √
the area swept by the radial vector from
√ time 0 to time t, then f (t) = L implies that the area
sion set of the ellipse πa2 1 − 2 is equal to LT . From T = πa2 1 − 2 /L, we get
[ [ ẋj = ∇yj H(x, y), ẏj = −∇xj H(x, y) .
∆= ∆ij = {x ∈ R3n | xi = xj } . T 2 /a3 = π 2 a(1 − 2 )/L2 = π 2 /G
i6=j i6=j
The third Kepler law allows to determine the gravitational constant G from the period and the geometry of the
ellipse.
10 CLASSICAL INTEGRALS. An integral of motion of a Hamiltonian system is a quantity which is conserved
along the orbits.
Proofs.
a) Every term in the sum Ẏ appears twice but with
The n-body problem in three dimensions has the 10
opposite sign.
classical integrals of motion:
b) From Ẏ = C follows that if Y = 0 then C is con- EXAMPLE. A Mars year is 1.88 earth years. How much longer
P stant. P is the length of the major semiaxes of the Mars orbit than the
a) The total momentum Y = ni=1 y . P
Pn i n
c) L̇ = i=1 ẋi × yi + i xi × ẏi . The second sum is semiaxes of the earth orbit?
b) If Y = 0, the position C = i=1 m x
i i of the 2 3 2 3
zero because xi × (xi − xj ) = −xi × xj and because Answer: we know Tmars /rmars = Tearth /rearth so that rmars =
center of mass. P each term in the remaining sum appears twice with rearth (Tmars /Tearth )2/3 = rearth 1.882/3 = 1.523.... Mars is about
c) The total angular momentum L = i xi × yi .
opposite sign.
Pn Pn one and a half times further away from the sun then the earth.
d) The total energy H.
d) Ḣ = i=1 Hxi ẋi + Hyi ẏi = i=1 Hxi Hyi −
Hyi Hxi = 0.

THE 2 BODY PROBLEM. After a change of coordinates, one can assume that the center of mass
C = m1 x1 + m2 x2 is at the origin. If q = x1 − x2 , then q̈ = ẍ1 − ẍ2 = m2 G(x2 − x1 )/|x2 − x1 |3 − m1 G(x1 − REMARKS
x2 )/|x2 − x1 |3 = −(m1 + m2 )Gq/|q|3 . This is a 1-body problem for a particle with position q and mass
m = m1 + m2 moving in a central field. The angular momentum L = mx × ẋ and the energy 2E/m = ẋ2 + G/|x| • To derive the first Kepler law starting with the ellipse is easier than taking off from the differential
are conserved quantities. equations. The later approach is possible but the steps are harder to motivate.
• All Kepler laws crucially depend on the conservation of L.
P(t)
P(t+T)
4

THE 2. KEPLER LAW. Because ẍ is parallel to x, we get L̇ = 0. 3

From the conservation of L follows that the vector x stays in a CULOMB CASE. The case  > 1 corresponds to a negative G, 2

plane, where we can us polar coordinates x = (r cos(θ), r sin(θ)). where particles repel each other. The third Kepler law does then 1

The constant quantity L = mr 2 θ̇ can be interpreted as df /dt, no more apply and the curve ”ellipse” will be a ”hyperbola” in -2 -1 1 2 3 4

where f is the area swept over by the vector x. We have derived (-aε,0)= S’ S =(aε,0) the first law. The second law is unchanged. In this Coulomb -1

the ”area law”, Keplers second law: ”the radius vector x passes case of the n-body problem, the total energy is always positive. -2

-3

the same area in the same time.” -4

OTHER POTENTIALS. SOME HISTORY.

α
If the interaction potential can be changed to ẍ = −Gx/r , where α is an integer. We have seen the case α = 3. Aristoteles (384-322 BC)
Hipparchus (190-120 BC)
For other α, the first Kepler law still applies. Formula θ̇ = L/(mr2 ) still applies. Also the derivation of the First model of solar system: had a moon theory built on
formula for ẍ · n = r̈ − L2 /(m2 r3 ) is still valid. The left hand side is −G/r α−1 which leads to the ordinary planets as well the sun move
epicycles. Still an earth cen-
differential equation around earth on perfect cir-
tered system.
cles.
r̈ = −G/rα−1 + L2 /(m2 r3 ) (*) Nicolas Copernicus (1473-
Claudius Ptolemaeus 1543) introduced a heliocen-
for r(t). Knowing r(t) gives then θ(t) from θ̇ = L/(mr2 ). The global behavior depends on the constants G, L. (78-150 AC) extended Hip- tric system as well as sec-
The case α = 4 corresponds to the natural Newton interaction in 4 dimensions. parchus’s system of epicycles ondary epicycles. This is
You show in the homework: to explain geocentric theory. a first step towards pertur-
Introduced 80 epicycles to bation theory (which later
In four dimensional space, planetary motion is unstable. explain the motions of sun, would be seen as the Fourier
moon and 5 planets. approximation of real mo-
α = 3 is the Kepler case with elliptic stable motion. tion).
The case α = 2 can physically be realized two massive parallel lines. (The general evolution of two rigid Galileo Galei (1564-1642) Tycho Brahe (1546-1601)
attracting lines in three dimensions is more complicated and form a special case of an interaction of two tops.) discovers Jupiter moons, suns revolutionized astronomy
pots etc. Famous for his fight with new instruments and
The case α = 1 can be realized by the motion of two massive parallel planes. Such planes attract each other for a Copernican theory with observations. For practical
with constant force independent of the distance. The equation of motion ẍ = −Gsign(x)x/|x|. The three body the inquisition. Mathemat- reasons, he used both he-
problem in this case is already interesting. In the case α = 0, each coordinate moves according to the harmonic ical work on moments and liocentric and earth centric
oscillator. center of gravity. coordinate systems.
Isaac Newton (1643-1727)
A theorem of Bertant states that only for α = 3 (the Kepler case) and α = 0 (the harmonic oscillator), all Johannes Kepler (1571- Put celestial mechanics on
bounded orbits are periodic. 1630) builds on the observa-
a solid mathematical foun-
tions of Tycho Brahe. He
dation and developed calcu-
finds the first and second Ke- lus simultaneously with Leib-
MORE REMARKS. pler law in 1609, the third in
niz. Derivation of Keplers
• To derive the first Kepler law starting with the ellipse is easier than taking off from the differential 1619.
laws from basic principles.
equations. The later approach is possible but the steps are harder to motivate. Joseph-Louis Lagrange
(1736-1813) Worked on the Leonard Euler (1707-1783)
• All Kepler laws crucially depend on the conservation of L. 3-body problem, the motion wrote a 775 page work on the
P mi mj of the moon, and pertur- motion of the moon. He won
• In d ≥ 2-dimensions, one would take the potential U (x) = i<j . In d = 2, the natural potential several prizes from the Paris
P |xi −xj |d−2
bations of comet orbits by
is U (x) = G i<j mi mj log |xi − xj |. Académie des Sciences in the
the planets as well as the
stability of the solar system. area of celestial mechanics.
• A natural regularisation of the singular potential is obtained by replacing the force by G · (|x| 2 + )−d/2 .
Pierre-Simon Laplace Siméon Denis Poisson
In that case, one does not have to exclude the collision set ∆.
(1749-1827) Investigated the (1781-1840) who had Laplace
• The phase space of the system is called with the fancy name cotangent bundle of M . Such terminology is inclination of planetary or- and Lagrange as teachers
not necessary when we deal with particles moving in the open region M of an Euclidean space. However, if bits, studied of planets were published in 1808 work on the
one would describe Newtonian particles on surfaces like the sphere or tori, where the interaction potential perturbed by their moons perturbations of the planets.
had to be modified, then the fancier notation is justified. We could for example look at the natural n and the stability of the solar He used series expansions to
body problem on a torus or the sphere. system. derive approximations.
Jean Le Rond d’Alembert
With Henry Poincaré
• One would need (6n − 1) integrals of motion to solve the n-body problem explicitly. The 10 classical (1717-1783) Improved New-
(1854-1912) at the end of
integrals are not enough to find explicit solutions if n > 2. The first mathematical proof of this fact was ton’s definition of force in the 19’th century, the n-
given by Poincaré in a special case of the three body problem using new qualitative methods. his Trait de dynamique pub-
body problem was studied
lished in 1743. This also con-
with new geometric and
tains d’Alembert’s principle
topological methods.
of mechanics.
Andrey Kolmogorov
George Birkhoff (1884-
(1903-1987) The beginning of
1944) Tools from probability
KAM-theory, which is named
theory statistical mechanics
THE THREE BODY PROBLEM. With 3 or more bodies, the after Kolmogorov, Arnold
lead to ergodic theory. An
problem becomes chaotic. On the right hand side, you see an and Moser. Kolmogorov also
example is Birkhoffs ergodic
orbit computed with the n-body solver ”xstar”. We will look at put probability theory on a
theorem. Poincare-Birkhoff
the restricted three body problem later in more detail and see in solid foundation and worked
fixed point theorem.
a special situation, the Sitnikov case, that chaos can occur. on a theory of turbulence.
Jürgen Moser (1928-1999) Vladimir Arnold (1937- )
The ”M” in KAM theory. Progress on stability ques-
Book with Siegel in Celestial tions with perturbative meth-
mechanics. Mosers contribu- ods (KAM). The concept of
tion to KAM is the twist map Arnold diffusion demon-
theorem. Worked also on in- strates a mechanism for insta-
tegrable n-body problems. bility.