Chapter 3

Two Body Central Forces

Consider two particles of masses m1 and m2 , with the only forces those of
their mutual interaction, which we assume is given by a potential which is a
function only of the distance between them, U (|~r1 − ~r2 |). In a mathematical
sense this is a very strong restriction, but it applies very nicely to many
physical situations. The classical case is the motion of a planet around the
Sun, ignoring the effects mentioned at the beginning of the book. But it
also applies to electrostatic forces and to many effective representations of
nonrelativistic interparticle forces.

3.1 Reduction to a one dimensional problem

Our original problem has six degrees of freedom, but because of the sym-
metries in the problem, many of these can be simply separated and solved
for, reducing the problem to a mathematically equivalent problem of a single
particle moving in one dimension. First we reduce it to a one-body problem,
and then we reduce the dimensionality.

3.1.1 Reduction to a one-body problem

As there are no external forces, we expect the center of mass coordinate to
be in uniform motion, and it behoves us to use

~ = m1~r1 + m2~r2
R
m1 + m2

65
66 CHAPTER 3. TWO BODY CENTRAL FORCES

as three of our generalized coordinates. For the other three, we first use the
cartesian components of the relative coordinate

~r := ~r2 − ~r1 ,

although we will soon change to spherical coordinates for this vector. In

~ and ~r, the particle positions are
terms of R

~ − m2 ~r,
~r1 = R ~ + m1 ~r,
~r2 = R where M = m1 + m2 .
M M
The kinetic energy is
1 1
T = m1 ṙ12 + m2 ṙ22
2 2
   
1 ˙~ m2 ˙ 2 1 ~˙ m1 ˙ 2
= m1 R − ~r + m2 R + ~r
2 M 2 M
1 2
= (m1 + m2 )R ~˙ + 1 m1 m2 ~r˙ 2
2 2 M
1 ~˙ 2 1 ˙ 2
= M R + µ~r ,
2 2
where
m1 m2
µ :=
m1 + m2
is called the reduced mass. Thus the kinetic energy is transformed to the
form for two effective particles of mass M and µ, which is neither simpler
nor more complicated than it was in the original variables.
For the potential energy, however, the new variables are to be preferred,
~ whose three components are
for U (~r1 − ~r2 ) = U (~r) is independent of R,
therefore ignorable coordinates, and their conjugate momenta
  ∂(T − U )
P~cm = = M Ṙi
i ∂ Ṙi
are conserved. This reduces half of the motion to triviality, leaving an effec-
tive one-body problem with T = 12 µṙ2 , and the given potential U (~r).
We have not yet made use of the fact that U only depends on the mag-
nitude of ~r. In fact, the above reduction applies to any two-body system
without external forces, as long as Newton’s Third Law holds.
3.1. REDUCTION TO A ONE DIMENSIONAL PROBLEM 67

3.1.2 Reduction to one dimension

In the problem under discussion, however, there is the additional restriction
that the potential depends only on the magnitude of ~r, that is, on the distance
between the two particles, and not on the direction of ~r. Thus we now convert
from cartesian to spherical coordinates (r, θ, φ) for ~r. In terms of the cartesian
coordinates (x, y, z)
1
r= (x2 + y 2 + z 2 ) 2 x= r sin θ cos φ
θ= cos−1 (z/r) y= r sin θ sin φ
φ= tan−1 (y/x) z= r cos θ
Plugging into the kinetic energy is messy but eventually reduces to a rather
simple form
1 h 2 i
T = µ ẋ1 + ẋ22 + ẋ23
2
1 h
= µ (ṙ sin θ cos φ + θ̇r cos θ cos φ − φ̇r sin θ sin φ)2
2
+(ṙ sin θ sin φ + θ̇r cos θ sin φ + φ̇r sin θ cos φ)2
i
+(ṙ cos θ − θ̇r sin θ)2
1 h 2 i
= µ ṙ + r2 θ̇2 + r2 sin2 θφ̇2 (3.1)
2
Notice that in spherical coordinates T is a funtion of r and θ as well as ṙ, θ̇,
and φ̇, but it is not a function of φ, which is therefore an ignorable coordinate,
and
∂L
Pφ = = µr2 sin2 θφ̇ = constant.
∂ φ̇
Note that r sin θ is the distance of the particle from the z-axis, so Pφ is just
the z-component of the angular momentum, Lz . Of course all of L ~ = ~r × p~
is conserved, because in our effective one body problem there is no torque
about the origin. Thus L ~ is a constant1 , and the motion must remain in a
plane perpendicular to L ~ and passing through the origin, as a consequence
1 ~ = 0, p~ and ~r are in the same direction, to which the motion is then confined.
If L
In this case it is more appropriate to use Cartesian coordinates with this direction as x,
reducing the problem to a one-dimensional problem with potential U (x) = U (r = |x|). In
the rest of this chapter we assume L~ 6= 0.
68 CHAPTER 3. TWO BODY CENTRAL FORCES

~ It simplifies things if we choose our coordinates so

of the fact that ~r ⊥ L.
that L is in the z-direction. Then θ = π/2, θ̇ = 0, L = µr2 φ̇. The r equation
~
of motion is then
L2
µr̈ − µrφ̇2 + dU/dr = 0 = µr̈ − + dU/dr.
µr3
This is the one-dimensional motion of body in an effective potential
L2
Ueff (r) = U (r) + .
2µr2
Thus we have reduced a two-body three-dimensional problem to one with
a single degree of freedom, without any additional complication except the
addition of a centrifugal barrier term L2 /2µr2 to the potential.
Before we proceed, a comment may be useful in retrospect about the re-
duction in variables in going from the three dimensional one-body problem
to a one dimensional problem. Here we reduced the phase space from six
variables to two, in a problem which had four conserved quantities, L ~ and
H. But we have not yet used the conservation of H in this reduction, we
have only used the three conserved quantities L. ~ Where have these dimen-
~
sions gone? From L conservation, by choosing our axes with L ~ k z, the
two constraints Lx = 0 and Ly = 0 ( with Lz 6= 0) do imply z = pz = 0,
thereby eliminating two of the coordinates of phase space. The conservation
of Lz , however, is a consequence of an ignorable coordinate φ, with conserved
conjugate momentum Pφ = Lz . In this case, not only is the corresponding
momentum restricted to a constant value, eliminating one dimension of vari-
ation in phase space, but the corresponding coordinate, φ, while not fixed,
drops out of consideration because it does not appear in the remaining one
dimensional problem. This is generally true for an ignorable coordinate —
the corresponding momentum becomes a time-constant parameter, and the
coordinate disappears from the remaining problem.

3.2 Integrating the motion

We can simplify the problem even more by using the one conservation law
left, that of energy. Because the energy of the effective motion is a constant,
1
E = µṙ2 + Ueff = constant
2
3.2. INTEGRATING THE MOTION 69

we can immediately solve for

( )1/2
dr 2
=± (E − Ueff (r)) .
dt µ

This can be inverted and integrated over r, to give

Z
dr
t = t0 ± q , (3.2)
2 (E − Ueff (r)) /µ

which is the inverse function of the solution to the radial motion problem
r(t). We can also find the orbit because

dφ φ̇ L dt
= = 2
dr dr/dt µr dr
so
Z r dr
φ = φ0 ± L q . (3.3)
r0 r2 2µ (E − Ueff (r))

The sign ambiguity from the square root is only because r may be increasing
or decreasing, but time, and usually φ/L, are always increasing.
Qualitative features of the motion are largely determined by the range
over which the argument of the square root is positive, as for other values of
r we would have imaginary velocities. Thus the motion is restricted to this
allowed region. Unless L = 0 or the potential U (r) is very strongly attractive
for small r, the centrifugal barrier will dominate there, so Ueff −→ +∞, and
r→0
there must be a smallest radius rp > 0 for which E ≥ Ueff . Generically the
force will not vanish there, so E −Ueff ≈ c(r −rp ) for r ≈ rp , and the integrals
in (3.2) and (3.3) are convergent. Thus an incoming orbit reaches r = rp at a
finite time and finite angle, and the motion then continues with r increasing
and the ± signs reversed. The radius rp is called a turning point of the
motion. If there is also a maximum value of r for which the velocity is real,
it is also a turning point, and an outgoing orbit will reach this maximum and
then r will start to decrease, confining the orbit to the allowed values of r.
If there are both minimum and maximum values, this interpretation of
Eq. (3.3) gives φ as a multiple valued function of r, with an “inverse” r(φ)
which is a periodic function of φ. But there is no particular reason for this
70 CHAPTER 3. TWO BODY CENTRAL FORCES

period to be the geometrically natural periodicity 2π of φ, so that different

values of r may be expected in successive passes through the same angle in
the plane of the motion. There would need to be something very special
about the attractive potential for the period to turn out to be just 2π, but
indeed that is the case for Newtonian gravity.
We have reduced the problem of the motion to doing integrals. In general
that is all we can do explicitly, but in some cases we can do the integral
analytically, and two of these special cases are very important physically.

3.2.1 The Kepler problem

Consider first the force of Newtonian gravity, or equivalently the Coulomb
attraction of unlike charged particles. The force F (r) = −K/r2 has a poten-
tial
K
U (r) = − .
r
Then the φ integral is
Z ( )−1/2
L 2E 2K L2
φ = φ0 ± dr + − 2 2
µr2 µ r µr
Z
du
= φ0 ± √ (3.4)
γ + αu − u2

where we have made the variable substitution u = 1/r which simplifies the
form, and have introduced abbreviations γ = 2µE/L2 , α = 2Kµ2 /L2 .
As dφ/dr must be real the motion will clearly be confined to regions for
which the argument of the square root is nonnegative, and the motion in
r will reverse at the turning points where the argument vanishes. The ar-
gument is clearly negative as u → ∞, which is r = 0. We have assumed
L 6= 0, so the angular momentum barrier dominates over the Coulomb at-
traction, and always prevents the particle from reaching the origin. Thus
there is always at least one turning point, umax , corresponding to the min-
imum distance rp . Then the argument of the square root must factor into
[−(u − umax )(u − umin )], although if umin is negative it is not really the min-
imum u, which can never get past zero. The integral (3.4) can be done2 with
2
Of course it can also be done by looking in a good table of integrals. For example, see
2.261(c) of Gradshtein and Ryzhik[7].
3.2. INTEGRATING THE MOTION 71

the substitution sin2 β = (umax − u)/(umax − umin ). This shows φ = φ0 ± 2β,

where φ0 is the angle at r = rmin , u = umax . Then
1
u≡ = A cos(φ − φ0 ) + B
r
where A and B are constants which could be followed from our sequence of
substitutions, but are better evaluated in terms of the conserved quantities
E and L directly. φ = φ0 corresponds to the minimum r, r = rp , the point
of closest approach, or perigee3 , so rp−1 = A + B, and A > 0. Let θ = φ − φ0
be the angle from this minimum, with the x axis along θ = 0. Then
 
1 1 e 1 1 + e cos θ
= A cos θ + B = 1− (1 − cos θ) =
r rp 1+e rp 1 + e

where e = A/B.
What is this orbit? Clearly rp just sets the scale of the whole orbit. From
rp (1 + e) = r + er cos θ = r + ex, if we subtract ex and square, we get
rp2 + 2rp e(rp − x) + e2 (rp − x)2 = r2 = x2 + y 2 , which is clearly quadratic in
x and y. It is therefore a conic section,

y 2 + (1 − e2 )x2 + 2e(1 + e)xrp − (1 + e)2 rp2 = 0.

The nature of the curve depends on the coefficient of x2 . For

• |e| < 1, the coefficient is > 0, and we have an ellipse.

• e = ±1, the coefficient vanishes and y 2 = ax + b is a parabola.

• |e| > 1, the coefficient is < 0, and we have a hyperbola.

All of these are possible motions. The bound orbits are ellipses, which
describe planetary motion and also the motion of comets. But objects which
have enough energy to escape from the sun, such as Voyager 2, are in hyper-
bolic orbit, or in the dividing case where the total energy is exactly zero, a
parabolic orbit. Then as time goes to ∞, φ goes to a finite value, φ → π for
a parabola, or some constant less than π for a hyperbolic orbit.
3
Perigee is the correct word if the heavier of the two is the Earth, perihelion if it is
the sun, periastron for some other star. Pericenter is also used, but not as generally as it
ought to be.
72 CHAPTER 3. TWO BODY CENTRAL FORCES

Let us return to the elliptic case. The closest approach, or perigee,

is r = rp , while the furthest apart the objects get is at θ = π, r = ra =
rp (1 + e)/(1 − e), which is called the apogee or aphelion. e is the eccentric-
ity of the ellipse. An ellipse is a circle stretched uniformly in one direction;
the diameter in that direction becomes the major axis of the ellipse, while
the perpendicular diameter becomes the minor axis.
One half the length of the major
axis is the semi-major axis and
is denoted by a.
 
1 1+e rp
a = rp + rp = ,
2 1−e 1−e
so

rp = (1 − e)a, ra = (1 + e)a.
Properties of an ellipse. The large
Notice that the center of the el- dots are the foci. The eccentricity
lipse is ea away from the Sun. is e and a is the semi-major axis.

Kepler tells us not only that the orbit is an ellipse, but also that the
sun is at one focus. To verify that, note the other focus of an ellipse is
symmetrically located, at (−2ea, 0), and work out the sum of the distances
of any point on the ellipse from the two foci. This will verify that d + r = 2a
is a constant, showing that the orbit is indeed an ellipse with the sun at one
focus.
How are a and e related to the total energy E and the angular momentum
L? At apogee and perigee, dr/dφ vanishes, and so does ṙ, so E = U (r) +
L2 /2µr2 = −K/r + L2 /2µr2 , which holds at r = rp = a(1 − e) and at
r = ra = a(1 + e). Thus Ea2 (1 ± e)2 + Ka(1 ± e) − L2 /2µ = 0. These two
equations are easily solved for a and e in terms of the constants of the motion
E and L
K 2 2EL2
a=− , e =1+ .
2E µK 2
As expected for a bound orbit, we have found r as a periodic function
of φ, but it is surprising that the period is the natural period 2π. In other
words, as the planet makes its revolutions around the sun, its perihelion is
always in the same direction. That didn’t have to be the case — one could
3.2. INTEGRATING THE MOTION 73

imagine that each time around, the minimum distance occurred at a slightly
different (or very different) angle. Such an effect is called the precession
of the perihelion. We will discuss this for nearly circular orbits in other
potentials in section (3.2.2).
What about Kepler’s Third Law? The area of a triange with ~r as one
edge and the displacement during a small time interval δ~r = ~v δt is A =
1
2
|~r × ~v |δt = |~r × p~|δt/2µ, so the area swept out per unit time is

dA L
= .
dt 2µ
which is constant. The area of an ellipse made by stretching a circle is
stretched by the same amount, so A is π times the semimajor axis times the
semiminor axis.√ The endpoint of the semiminor axis is a away from each
focus, so it is a 1 − e2 from the center, and
v !
u
2
√ u
2t 2EL2
A = πa 1 − e = πa 1 −
2 1+
µK 2
s
L −2E
= πa2 .
K µ
Recall that for bound orbits E < 0, so A is real. The period is just the area
swept out in one revolution divided by the rate it is swept out, or
s
L −2E 2µ
2
T = πa
K µ L
2q
2πa π
= −2µE = K(2µ)1/2 (−E)−3/2 (3.5)
K 2
2πa2 q
= µK/a = 2πa3/2 (K)−1/2 µ1/2 , (3.6)
K
independent of L. The fact that T and a depend only on E and not on
L is another fascinating manifestation of the very subtle symmetries of the
Kepler/Coulomb problem.

3.2.2 Nearly Circular Orbits

For a general central potential we cannot find an analytic form for the motion,
which involves solving the effective one-dimensional problem with Ueff (r) =
74 CHAPTER 3. TWO BODY CENTRAL FORCES

U (r)+L2 /2µr2 . If Ueff (r) has a minimum at r = a, one solution is certainly a

circular orbit of radius a. The minimum requires dUeff (r)/dr = 0 = −F (r) −
L2 /µr3 , so

L2
F (a) = − .
µa3

We may also ask about trajectories which differ only slightly from this orbit,
for which |r − a| is small. Expanding Ueff (r) in a Taylor series about a,

1
Ueff (r) = Ueff (a) + (r − a)2 k,
2
where

d2 Ueff 
k = 
dr2 a
!
dF 3L2 dF 3F
= − + 4 =− + .
dr µa dr a

For r = a to be a minimum and the nearly circular orbits to be stable, the

second derivative and k must be positive, and therefore F 0 + 3F/a < 0. As
4
always when we treat a problem as small deviations from a stable
q equilibrium
we have harmonic oscillator motion, with a period Tosc = 2π µ/k.
As a simple class of examples, consider the case where the force law
depends on r with a simple power, F = −crn . Then k = (n + 3)can−1 , which
is positive and the orbit stable only if n > −3. For gravity, n = −2, c =
K, k = K/a3 , and
s
µa3
Tosc = 2π
K
agreeing with what we derived for the more general motion, not restricted to
small deviations from circularity. But for more general n, we find
v
u
u µa1−n
Tosc = 2π t .
c(n + 3)
4
This statement has an exception if the second derivative vanishes, k = 0.
3.2. INTEGRATING THE MOTION 75

The period of revolution Trev can be calculated for the circular orbit, as

2π q
L = µa2 φ̇ = µa2 = µa3 |F (a)|,
Trev
so
s
µa
Trev = 2π
|F (a)|

which for the power law case is

s
µa1−n
Trev = 2π .
c

Thus the two periods Tosc and Trev are not equal unless n = −2, as in the
gravitational case. Let us define the apsidal angle ψ as the angle between
√
an apogee and the next perigee. It is therefore ψ = πTosc /Trev = π/ 3 + n.
For the gravitational case ψ = π, the apogee and perigee are on opposite sides
of the orbit. For a two- or three-dimensional harmonic oscillator F (r) = −kr
we have n = 1, ψ = 12 π, and now an orbit contains two apogees and two
perigees, and is again an ellipse, but now with the center-of-force at the
center of the ellipse rather than at one focus.
Note that if ψ/π is not rational, the orbit never closes, while if ψ/π = p/q,
the orbit will close after p revolutions, having reached q apogees and perigees.
The orbit will then be closed, but unless p = 1 it will be self-intersecting.
This exact closure is also only true in the small deviation approximation;
more generally, Bertrand’s Theorem states that only for the n = −2 and
n = 1 cases are the generic orbits closed.
In the treatment of planetary motion, the precession of the perihelion is
the angle though which the perihelion slowly moves, so it is 2ψ −2π per orbit.
We have seen that it is zero for the pure inverse force law. There is actually
some precession of the planets, due mostly to perturbative effects of the other
planets, but also in part due to corrections to Newtonian mechanics found
from Einstein’s theory of general relativity. In the late nineteenth century
discrepancies in the precession of Mercury’s orbit remained unexplained, and
the resolution by Einstein was one of the important initial successes of general
relativity.
76 CHAPTER 3. TWO BODY CENTRAL FORCES

3.3 The Laplace-Runge-Lenz Vector

The remarkable simplicity of the motion for the Kepler and harmonic oscilla-
tor central force problems is in each case connected with a hidden symmetry.
We now explore this for the Kepler problem.
For any central force problem F~ = p~˙ = f (r)êr we have a conserved
angular momentum L ~˙ = µ~r˙ × ~r˙ + (f (r)/r)~r × ~r = 0. The
~ = µ(~r × ~r˙), for L
motion is therefore confined to a plane perpendicular to L, ~ and the vector
~
p~ × L is always in the plane of motion, as are ~r and p~. Consider the evolution
~ with time5
of p~ × L
d  
~ = p~˙ × L
~ = F~ × L ~ = µf (r)êr × (~r × ~r˙)
p~ × L
dt  
= µf (r) ~rêr · ~r˙ − ~r˙êr · ~r = µf (r)(ṙ~r − r~r˙)

On the other hand, the time variation of the unit vector êr = ~r/r is

d d ~r ~r˙ ṙ~r ṙ~r − r~r˙

êr = = − 2 =− .
dt dt r r r r2
For the Kepler case, where f (r) = −K/r2 , these are proportional to each
other with a constant ratio, so we can combine them to form a conserved
quantity A ~ − µK êr , called6 the Laplace-Runge-Lenz vector,
~ = p~ × L
~
dA/dt = 0.
While we have just found three conserved quantities in addition to the
conserved energy and the three conserved components of L, ~ these cannot all
~
be independent. Indeed we have already noted that A lies in the plane of
~ so A
motion and is perpendicular to L, ~ ·L
~ = 0. If we dot A
~ into the position
vector,
~ · ~r = ~r · (~p × (~r × p~)) − µKr = (~r × p~)2 − µKr = L2 − µKr,
A
~ and ~r, we have Ar cos θ + µKr = L2 , or
so if θ is the angle between A
!
1 µK A
= 2 1+ cos θ ,
r L µK
5 ~ B× ~ C)~ = B(
~ A·
~ C)−
~ C( ~ A· ~ and êr ·~r˙ = (1/r)~r ·~r˙ = (1/2r)d(r2 )/dt =
~ B),
Some hints: A×(
ṙ. The first equation, known as the bac-cab equation, is shown in Appendix A.1.
6
by Goldstein, at least. While others often use only the last two names, Laplace clearly
has priority.
3.4. THE VIRIAL THEOREM 77

which is an elegant way of deriving the formula we found previously by

integration, with A = µKe. Note θ = 0 is the perigee, so A ~ is a constant
vector pointing towards the perigee.
We also see that the magnitude of A~ is given in terms of e, which we have
previously related to L and E, so A2 = µ2 K 2 + 2µEL2 is a further relation
among the seven conserved quantities, showing that only five are indepen-
dent. There could not be more than five independent conserved functions
depending analytically on the six variables of phase space (for the relative
motion only), for otherwise the point representing the system in phase space
would be unable to move. In fact, the five independent conserved quantities
on the six dimensional dimensional phase space confine a generic invariant
set of states, or orbit, to a one dimensional subspace. For power laws other
than n = −2 and n = 1, as the orbits do not close, they are dense in a two
dimensional region of phase space, indicating that there cannot be more than
four independent conserved analytic functions on phase space. So we see the
connection between the existence of the conserved A ~ in the Kepler case and
the fact that the orbits are closed.

3.4 The virial theorem

P
Consider a system of particles and the quantity G = i p~i · ~ri . Then the rate
at which this changes is
dG X ~
= Fi · ~ri + 2T.
dt
If the system returns to a region in phase space where it had been, after some
time, G returns to what it was, and the average value of dG/dt vanishes,
* +
dG DX E
= F~i · ~ri + 2 hT i = 0.
dt
This average will also be zero if the region stays in some bounded part of
phase space for which G can only take bounded values, and the averaging time
is taken to infinity. This is appropriate for a system in thermal equilibrium,
for example.
Consider a gas of particles which interact only with the fixed walls of the
container, so that the force acts only on the surface, and the sum becomes
an integral over dF~ = −pdA, ~ where p is the uniform pressure and dA ~ is
78 CHAPTER 3. TWO BODY CENTRAL FORCES

an outward pointing vector representing a small piece of the surface of the

volume. Then
DX E Z Z
F~i · ~ri = − ~ = −p
p~r · dA ∇ · ~rdV = −3pV
δV V

so h2T i = 3pV . In thermodynamics we have the equipartition theorem

which states that hT i = 32 N kB τ , where N is the number of particles, kB is
Boltzmann’s constant and τ the temperature, so pV = N kB τ .
A very different application occurs for a power law central force between
pairs of particles, say for a potential U (~ri , ~rj ) = c|~ri − ~rj |n+1 . Then this
action and reaction contribute F~ij · ~rj + F~ji · ~ri = F~ji · (~ri − ~rj ) =
−(n + 1)c|~ri − ~rj |n+1 = −(n + 1)U (~ri , ~rj ). So summing over all the particles
P
and using h2T i = −h F~ · ~ri, we have
n+1
hT i = hU i.
2
For Kepler, n = −2, so hT i = − 12 hU i = −hT +U i = −E must hold for closed
orbits or for large systems of particles which remain bound and uncollapsed.
It is not true, of course, for unbound systems which have E > 0.
The fact that the average value of the kinetic energy in a bound system
gives a measure of the potential energy is the basis of the measurements
of the missing mass, or dark matter, in galaxies and in clusters of galaxies.
This remains a useful tool despite the fact that a multiparticle gravitationally
bound system can generally throw off some particles by bringing others closer
together, so that, strictly speaking, G does not return to its original value or
remain bounded.

3.5 Rutherford Scattering

We have discussed the 1/r potential in terms of Newtonian gravity, but of
course it is equally applicable to Coulomb’s law of electrostatic forces. The
force between nonrelativistic charges Q and q is given7 by
1 Qq
F~ = ~r,
4π0 r3
7
Here we use S. I. or rationalized MKS units. For Gaussian units drop the 4π0 , or for
Heaviside-Lorentz units drop only the 0 .
3.5. RUTHERFORD SCATTERING 79

and the potential energy is U (r) = −K/r with K = −Qq/4π0 .

Unlike gravity, the force is not al-

ways attractive (K > 0), and for θ
like sign charges we have K < 0,
and therefore U and the total en-
ergy are always positive, and there
are no bound motions. Whatever α
φ
the relative signs, we are going to
consider scattering here, and there- α
fore positive energy solutions with
the initial state of finite speed v0
and r → ∞. Thus the relative mo-
tion is a hyperbola, with
1+e
r = rp
1 + e cos φ
s
2EL2
e = ± 1+ .
µK 2
This starts and ends with r → ∞, Rutherford scattering. An α par-
at φ → ±α = ± cos−1 (−1/e), and ticle approaches a heavy nucleus
the angle θ through which the ve- with an impact parameter b, scat-
locity changes is called the scat- tering through an angle θ. The
tering angle. For simplicity we cross sectional area dσ of the inci-
will consider the repulsive case, dent beam is scattered through an-
with e < 0 so that α < π/2. gles ∈ [θ, θ + dθ].
We see that θ = π − 2α, so
s
θ cos α |e|−1 1 µK 2
tan = cot α = √ = q =√ 2 = .
2 1 − cos α
2
1 − |e|−2 e −1 2EL2

We have K = Qq/4π0 . We need to evaluate E and L. At r = ∞, U → 0,

E = 12 µv02 , L = µbv0 , where b is the impact parameter, the distance by
which the asymptotic line of the initial motion misses the scattering center.
Thus
s
θ µ |K|
tan = |K| 2
= . (3.7)
2 µv0 (µbv0 )2 µbv02
80 CHAPTER 3. TWO BODY CENTRAL FORCES

The scattering angle therefore depends on b, the perpendicular displace-

ment from the axis parallel to the beam through the nucleus. Particles
passing through a given area will be scattered through a given angle, with
a fixed angle θ corresponding to a circle centered on the axis, having radius
b(θ) given by 3.7. The area of the beam dσ in an annular ring of impact
parameters ∈ [b, b + db] is dσ = 2πb|db|. To relate db to dθ, we differentiate
the scattering equation for fixed v0 ,

1 θ −K
sec2 dθ = 2 2 db,
2 2 µv0 b

dσ µv02 b2 πµv02 b3
= 2πb =
dθ 2K cos2 (θ/2) K cos2 (θ/2)
!3 !3 !2
πµv02 K cos θ/2 K cos θ/2
= =π
K cos2 (θ/2) µv02 sin θ/2 µv02 sin3 θ/2
!2
π K sin θ
= .
2 µv02 sin4 θ/2

(The last expression is useful because sin θdθ is the “natural measure” for θ,
in the sense that integrating over volume in spherical coordinates is d3 V =
r2 dr sin θdθdφ.)
How do we measure dσ/dθ? There is a beam of N particles shot at
random impact parameters onto a foil with n scattering centers per unit
area, and we confine the beam to an area A. Each particle will be significantly
scattered only by the scattering center to which it comes closest, if the foil
is thin enough. The number of incident particles per unit area is N/A, and
the number of scatterers being bombarded is nA, so the number which get
scattered through an angle ∈ [θ, θ + dθ] is

N dσ dσ
× nA × dθ = N n dθ.
A dθ dθ
We have used the cylindrical symmetry of this problem to ignore the φ
dependance of the scattering. More generally, the scattering would not be
uniform in φ, so that the area of beam scattered into a given region of (θ,φ)
would be
dσ
dσ = sin θdθdφ,
dΩ
3.5. RUTHERFORD SCATTERING 81

where dσ/dΩ is called the differential cross section. For Rutherford scat-
tering we have
!2
dσ 1 K θ
= 2
csc4 .
dΩ 4 µv0 2

Scattering in other potentials

We see that the cross section depends on the angle through which the incident
particle is scattered for a given impact parameter. In Rutherford scattering
θ increases monotonically as b decreases, which is possible only because the
force is “hard”, and a particle aimed right at the center will turn around
rather than plowing through. This was a surprize to Rutherford, for the
concurrent model of the nucleus, Thompson’s plum pudding model, had the
nuclear charge spread out over some atomic-sized spherical region, and the
Coulomb force would have decreased once the alpha particle entered this
region. So sufficiently energetic alpha particles aimed at the center should
have passed through undeflected instead of scattered backwards. In fact, of
course, the nucleus does have a finite size, and this is still true, but at a much
smaller distance, and therefore a much larger energy.
If the scattering angle θ(b) does run smoothly from 0 at b = 0 to 0 at
b → ∞, as shown, then there is an extremal value for which dθ/db|b0 = 0,
and for θ < θ(b0 ), dσ/dθ can get contributions from several different b’s,

dσ X bi db 
=  .
dΩ 
i sin θ dθ i

It also means that the cross sec-

tion becomes infinite as θ → θ(b0 ),
and vanishes above that value of θ
θ. This effect is known as rain- θ
bow scattering, and is the cause
of rainbows, because the scattering
for a given color light off a water
droplet is very strongly peaked at
the maximum angle of scattering.
Another unusual effect occurs when θ(b) becomes 0 or π for some nonzero
value of b, with db/dθ finite. Then dσ/dΩ
R
blows up due to the sin θ in the
denominator, even though the integral (dσ/dΩ) sin θdθdφ is perfectly finite.
82 CHAPTER 3. TWO BODY CENTRAL FORCES

This effect is called glory scattering, and can be seen around the shadow
of a plane on the clouds below.

Exercises
3.1 A space ship is in circular orbit at radius R and speed v1 , with the period
of revolution τ1 . The crew wishes to go to planet X, which is in a circular orbit
of radius 2R, and to revolve around the Sun staying near planet X. They propose
to do this by firing two blasts, one putting them in an orbit with perigee R and
apogee 2R, and the second, when near X, to change their velocity so they will have
the same speed as X.
• (a) By how much must the first blast change their velocity? Express your
answer in terms of v1 .

• (b) How long will it take until they reach the apogee? Express your answer
in terms of τ1

• (c) By how much must the second blast change their speed? Will they need
to slow down or speed up, relative to the sun.

3.2 Consider a spherical droplet of water in the sunlight. A ray of light with
impact parameter b is refracted, so by Snell’s Law n sin β = sin α. It is then
internally reflected once and refracted again on the way out.
(a) Express the scattering angle θ in terms of α and β.
(b) Find the scattering cross section
dσ/dΩ as a function of θ, α and β
(which is implicitly a function of θ
from (a) and Snell’s Law).
(c) The smallest value of θ is called
the rainbow scattering angle. Why? α
Find it numerically to first order in
δ if the index of refraction is n = β

1.333 + δ
(d) The visual spectrum runs from vi-
olet, where n = 1.343, to red, where
n = 1.331. Find the angular radius
of the rainbow’s circle, and the an- θ
gular width of the rainbow, and tell
whether the red or blue is on the out- One way light can scatter from a
side. spherical raindrop.
3.5. RUTHERFORD SCATTERING 83

3.3 Consider a particle constrained to move on the surface described in cylindrical

coordinates by z = αr3 , subject to a constant gravitational force F~ = −mgêz .
Find the Lagrangian, two conserved quantities, and reduce the problem to a one
dimensional problem. What is the condition for circular motion at constant r?

3.4 From the general expression for φ as an integral over r, applied to a three
dimensional symmetrical harmonic oscillator U (~r) = 12 kr2 , integrate the equation,
and show that the motion is an ellipse, with the center of force √ at the center of
the ellipse. Consider the three complex quantities Qi = pi − i kmri , and show
that each has a very simple equation of motion, as a consequence of which the
nine quantities Q∗i Qk are conserved. Identify as many as possible of these with
previously known conserved quantities.

3.5 Show that if a particle under the influence of a central force has an orbit
which is a circle passing through the point of attraction, then the force is a power
law with |F | ∝ r−5 . Assuming the potential is defined so that U (∞) = 0, show
that for this particular orbit E = 0. In terms of the diameter and the angular
momentum, find the period, and by expressing ẋ, ẏ and the speed as a function of
the angle measured from the center of the circle, and its derivative, show that ẋ, ẏ
and the speed all go to infinity as the particle passes through the center of force.

3.6 For the Kepler problem we have the relative position tracing out an ellipse.
What is the curve traced out by the momentum in momentum space? Show that
it is a circle centered at L ~ 2 , where L
~ × A/L ~ and A
~ are the angular momentum and
Runge-Lenz vectors respectively.

3.7 The Rutherford cross section implies all incident projectiles will be scattered
and emerge at some angle θ, but a real planet has a finite radius, and a projectile
that hits the surface is likely to be captured rather than scattered.
What is the capture cross section for an airless planet of radius R and mass M
for a projectile with a speed v0 ? How is the scattering differential cross section
modified from the Rutherford prediction?

3.8 In problem 2.12 we learned that the general-relativistic motion of a particle

in a gravitational field is given by Hamilton’s variational principle on the path
xµ (λ) with the action
v
Z uX
u dxµ dxν
S= dλL with L = mct gµν (xρ ) ,
µν dλ dλ
84 CHAPTER 3. TWO BODY CENTRAL FORCES

where we may freely choose the path parameter λ to be the proper time (after
√
doing the variation), so that the is c, the speed of light.
The gravitational field of a static point mass M is given by the
Schwartzschild metric
 
2GM 2GM
g00 = 1 − , grr = −1 1− , gθθ = −r2 , gφφ = −r2 sin2 θ,
rc2 rc2

where all other components of gµν are zero. Treating the four xµ (λ) as the coordi-
nates, with λ playing the role of time, find the four conjugate momenta pµ , show
that p0 and pφ = L are constants, and use the freedom to choose
v
Z uX
1 u dxµ dxν
λ=τ = t gµν (xρ )
c µν dλ dλ

P
to show m2 c2 = µν g µν pµ pν , where g µν is the inverse matrix to gαβ . Use this to
show that s  
dr 2GM L2 2GM L2
= κ− − + 2 2− 2 3 2 ,
dτ r m r m r c
where κ is a constant. For an almost circular orbit at the minimum r = a of
the effective potential this implies, show that the precession of the perihelion is
6πGM/ac2 .
Find the rate of precession for Mercury, with G = 6.67 × 10−11 Nm2 /kg2 , M =
1.99 × 1030 kg and a = 5.79 × 1010 m, per revolution, and also per century, using
the period of the orbit as 0.241 years.