4.

Central Forces
In this section we will study the three-dimensional motion of a particle in a central
force potential. Such a system obeys the equation of motion

mẍ = rV (r) (4.1)

where the potential depends only on r = |x|. Since both gravitational and electrostatic
forces are of this form, solutions to this equation contain some of the most important
results in classical physics.

Our first line of attack in solving (4.1) is to use angular momentum. Recall that this
is defined as

L = mx ⇥ ẋ

We already saw in Section 2.2.2 that angular momentum is conserved in a central

potential. The proof is straightforward:
dL
= mx ⇥ ẍ = x ⇥ rV = 0
dt
where the final equality follows because rV is parallel to x.

The conservation of angular momentum has an important consequence: all motion

takes place in a plane. This follows because L is a fixed, unchanging vector which, by
construction, obeys

L·x=0

So the position of the particle always lies in a plane perpendicular to L. By the same
argument, L · ẋ = 0 so the velocity of the particle also lies in the same plane. In this
way the three-dimensional dynamics is reduced to dynamics on a plane.

4.1 Polar Coordinates in the Plane

We’ve learned that the motion lies in a plane. It will turn out to be much easier if we
work with polar coordinates on the plane rather than Cartesian coordinates. For this
reason, we take a brief detour to explain some relevant aspects of polar coordinates.

To start, we rotate our coordinate system so that the angular momentum points in
the z-direction and all motion takes place in the (x, y) plane. We then define the usual
polar coordinates

x = r cos ✓ , y = r sin ✓

– 48 –
Our goal is to express both the velocity and acceleration y ^ ^
θ r
in polar coordinates. We introduce two unit vectors, r̂
ˆ in the direction of increasing r and ✓ respectively
and ✓
as shown in the diagram. Written in Cartesian form, θ x
these vectors are
! !
cos ✓ ˆ= sin ✓
r̂ = , ✓
sin ✓ cos ✓
Figure 11:
These vectors form an orthornormal basis at every point
on the plane. But the basis itself depends on which angle ✓ we sit at. Moving in the
radial direction doesn’t change the basis, but moving in the angular direction we have
! !
dr̂ sin ✓ d ˆ
✓ cos ✓
= =✓ˆ , = = r̂
d✓ cos ✓ d✓ sin ✓

This means that if the particle moves in a way such that ✓ changes with time, then the
basis vectors themselves will also change with time. Let’s see what this means for the
velocity expressed in these polar coordinates. The position of a particle is written as
the simple, if somewhat ugly, equation

x = rr̂

From this we can compute the velocity, remembering that both r and the basis vector
r̂ can change with time. We get
dr̂ ˙
ẋ = ṙr̂ + r ✓
d✓
= ṙr̂ + r✓˙✓
ˆ (4.2)

The second term in the above expression arises because the basis vectors change with
˙ (Strictly speaking, this is the angular
time and is proportional to the angular velocity, ✓.
speed. In the next section, we will introduce a vector quantity which is the angular
velocity).

Di↵erentiating once more gives us the expression for acceleration in polar coordinates,

dr̂ ˙ ˆ
ẍ = r̈r̂ + ṙ ✓ + ṙ✓˙✓ ˆ + r✓˙ d✓ ✓˙
ˆ + r✓¨✓
d✓ d✓
= (r̈ ˙ 2 ¨ ˙
r✓ )r̂ + (r✓ + 2ṙ✓)✓ ˆ (4.3)

The two expressions (4.2) and (4.3) will be important in what follows.

– 49 –
An Example: Circular Motion
Let’s look at an example that we’re already all familiar with. A particle moving in a
circle has ṙ = 0. If the particle travels with constant angular velocity ✓˙ = ! then the
velocity in the plane is
ˆ
ẋ = r! ✓

so the speed in the plane in v = |ẋ| = r!. Similarly, the acceleration in the plane is

ẍ = r! 2 r̂

The magnitude of the acceleration is a = |ẍ| = r! 2 = v 2 /r. From Newton’s second

law, if we want a particle to travel in a circle, we need to supply a force F = mv 2 /r
towards the origin. This is known as a centripetal force.

4.2 Back to Central Forces

We’ve already seen that the three-dimensional motion in a central force potential ac-
tually takes place in a plane. Let’s write the equation of motion (4.1) using the plane
polar coordinates that we’ve just introduced. Since V = V (r), the force itself can be
written using
dV
rV = r̂
dr
and, from (4.3) the equation of motion becomes
dV
m(r̈ r✓˙2 )r̂ + m(r✓¨ + 2ṙ✓)
˙ ✓ˆ= r̂ (4.4)
dr
ˆ component of this is particularly simple. It is
The ✓
1 d ⇣ 2 ˙⌘
r✓¨ + 2ṙ✓˙ = 0 ) r ✓ =0
r dt
It looks as if we’ve found a new conserved quantity since we’ve learnt that

l = r2 ✓˙ (4.5)

does not change with time. However, we shouldn’t get too excited. This is something
that we already know. To see this, let’s look again at the angular momentum L. We
already used the fact that the direction of L is conserved when restricting motion to
the plane. But what about the magnitude of L? Using (4.2), we write
⇣ ⌘ ⇣ ⌘
L = mx ⇥ ẋ = mrr̂ ⇥ ṙr̂ + r✓˙✓ˆ = mr2 ✓˙ r̂ ⇥ ✓
ˆ

– 50 –
 ˆ are orthogonal, unit vectors, r̂ ⇥ ✓
Since r̂ and ✓ ˆ is also a unit vector. The magnitude
of the angular momentum vector is therefore

|L| = ml

and l, given in (4.5), is identified as the angular momentum per unit mass, although
we will often be lazy and refer to l simply as the angular momentum.

Let’s now look at the r̂ component of the equation of motion (4.4). It is

dV
m(r̈ r✓˙2 ) =
dr
Using the fact that l = r2 ✓˙ is conserved, we can write this as

dV ml2
mr̈ = + 3 (4.6)
dr r
It’s worth pausing to reflect on what’s happened here. We started in (4.1) with a com-
plicated, three dimensional problem. We used the direction of the angular momentum
to reduce it to a two dimensional problem, and the magnitude of the angular momen-
tum to reduce it to a one dimensional problem. This was all possible because angular
momentum is conserved.

This should give you some idea of how important conserved quantities are when it
comes to solving anything. Roughly speaking, this is also why it’s not usually possible
to solve the N -body problem with N 3. In Section 5.1.5, we’ll see that for the N = 2
mutually interacting particles, we can use the symmetry of translational invariance to
solve the problem. But for N 3, we don’t have any more conserved quantities to
come to our rescue.

Returning to our main storyline, we can write (4.6) in the suggestive form

dVe↵
mr̈ = (4.7)
dr
where Ve↵ (r) is called the e↵ective potential and is given by

ml2
Ve↵ (r) = V (r) + (4.8)
2r2
The extra term, ml2 /2r2 is called the angular momentum barrier (also known as the
centrifugal barrier). It stops the particle getting too close to the origin, since it must
pay a heavy price in “e↵ective energy”.

– 51 –
 Veff

r* r

E min

Figure 12: The e↵ective potential arising from the inverse square force law.

4.2.1 The E↵ective Potential: Getting a Feel for Orbits

Let’s just check that the e↵ective potential can indeed be thought of as part of the
energy of the full system. Using (4.2), we can write the energy of the full three dimen-
sional problem as
1
E = mẋ · ẋ + V (r)
2
1 1
= mṙ2 + mr2 ✓˙2 + V (r)
2 2
1 2 ml2
= mṙ + 2 + V (r)
2 2r
1 2
= mṙ + Ve↵ (r)
2
This tells us that the energy E of the three dimensional system does indeed coincide
with the energy of the e↵ective one dimensional system that we’ve reduced to. The
e↵ective potential energy is the real potential energy, together with a contribution from
the angular kinetic energy.

We already saw in Section 2.1.1 how we can understand qualitative aspects of one
dimensional motion simply by plotting the potential energy. Let’s play the same game
here. We start with the most useful example of a central potential: V (r) = k/r,
corresponding to an attractive inverse square law for k > 0. The e↵ective potential is
k ml2
Ve↵ = + 2
r 2r
and is drawn in the figure.

– 52 –
 The minimum of the e↵ective potential occurs at r? = ml2 /k and takes the value
Ve↵ (r? ) = k 2 /2ml2 . The possible forms of the motion can be characterised by their
energy E.

• E = Emin = k 2 /2ml2 : Here the particle sits at the bottom of the well r? and
stays there for all time. However, remember that the particle also has angular
velocity, given by ✓˙ = l/r?2 . So although the particle has fixed radial position, it
is moving in the angular direction. In other words, the trajectory of the particle
is a circular orbit about the origin.

Notice that the radial position of the minimum depends on the angular mo-
mentum l. The higher the angular momentum, the further away the minimum.
If there is no angular momentum, and l = 0, then Ve↵ = V and the potential has
no minimum. This is telling us the obvious fact that there is no way that r can
be constant unless the particle is moving in the ✓ direction. In a similar vein,
notice that there is a relationship between the angular velocity ✓˙ and the size of
the orbit, r? , which we get by eliminating l: it is ✓˙2 = k/mr?3 . We’ll come back
to this relationship in Section 4.3.2 when we discuss Kepler’s laws of planetary
motion.

• Emin < E < 0: Here the 1d system sits in the dip, oscillating backwards and
forwards between two points. Of course, since l 6= 0, the particle also has angular
velocity in the plane. This describes an orbit in which the radial distance r
depends on time. Although it is not yet obvious, we will soon show that for
V = k/r, this orbit is an ellipse.

The smallest value of r that the particle reaches is called the periapsis. The
furthest distance is called the apoapsis. Together, these two points are referred
to as the apsides. In the case of motion around the Sun, the periapsis is called
the perihelion and the apoapsis the aphelion.

• E > 0. Now the particle can sit above the horizontal axis. It comes in from
infinity, reaches some minimum distance r, then rolls back out to infinity. We
will see later that, for the V = k/r potential, this trajectory is hyperbola.

4.2.2 The Stability of Circular Orbits

Consider a general potential V (r). We can ask: when do circular orbits exist? And
when are they stable?

– 53 –
 The first question is quite easy. Circular orbits exist whenever there exists a solution
with l 6= 0 and ṙ = 0 for all time. The latter condition means that r̈ = 0 which, in
turn, requires
0
Ve↵ (r? ) = 0
In other words, circular orbits correspond to critical points, r? , of Ve↵ . The orbit is
stable if small perturbations return us back to the critical point. This is the same kind
of analysis that we did in Section 2.1.2: stability requires that we sit at the minimum
of the e↵ective potential. This usually translates to the requirement that
00
Ve↵ (r? ) > 0
If this condition holds, small radial deviations from the circular orbit will oscillate
about r? with simple harmonic motion.

Although the criterion for circular orbits is most elegantly expressed in terms of the
e↵ective potential, sometimes it’s necessary to go back to our original potential V (r).
In this language, circular orbits exist at points r? obeying
ml2
V 0 (r? ) =
r?3
These orbits are stable if
3ml2 3
V 00 (r? ) + 4
= V 00 (r? ) + V 0 (r? ) > 0 (4.9)
r? r?
We can even go right back to basics and express this in terms of the force (remember
that?!), F (r) = V 0 (r). A circular orbit is stable if
3
F 0 (r? ) + F (r? ) < 0
r?
An Example
Consider a central potential which takes the form
k
V (r) = n 1
rn
For what powers of n are the circular orbits stable? By our criterion (4.9), stability
requires
3 ⇣ ⌘ k
V 00 + V 0 = n(n + 1) 3n n+2 > 0
r r
which holds only for n < 2. We can easily see this pictorially in the figures where we’ve
plotted the e↵ective potential for n = 1 and n = 3.

– 54 –
 r r

Figure 13: Ve↵ for V = 1/r Figure 14: Ve↵ for V = 1/r3

Curiously, in a Universe with d spatial dimensions, the law of gravity would be

F ⇠ 1/rd 1 corresponding to a potential energy V ⇠ 1/rd 2 . We see that circular
planetary orbits are only stable in d < 4 spatial dimensions. Fortunately, this includes
our Universe. We should all be feeling very lucky right now.

4.3 The Orbit Equation

Let’s return to the case of general Ve↵ . If we want to understand how the radial position
r(t) changes with time, then the problem is essentially solved. Since the energy E is
conserved, we have
1
E = mṙ2 + Ve↵ (r)
2
which we can view as a first order di↵erential equation for dr/dt. Integrating then gives
r Z
m dr
t=± p
2 E Ve↵ (r)
However, except for a few very special choices of Ve↵ (r), the integral is kind of a
pain. What’s more, often trying to figure out r(t) is not necessarily the information
that we’re looking for. It’s better to take a more global approach, and try to learn
something about the whole trajectory of the particle, rather than its position at any
given time. Mathematically, this means that we’ll try to understand something about
the shape of the orbit by computing r(✓).

In fact, to proceed, we’ll also need a little trick. It’s trivial, but it turns out to make
the resulting equations much simpler. We introduce the new coordinate
1
u=
r
I wish I had a reason to motivate this trick. Unfortunately, I don’t. You’ll just have to
trust me and we’ll see that it helps.

– 55 –
 Let’s put these things together. Firstly, we can rewrite the radial velocity as

dr dr dr l du
= ✓˙ = = l
dt d✓ d✓ r2 d✓
Meanwhile, the acceleration is
✓ ◆
d2 r d du d2 u ˙ d2 u 1 d2 u
= l = l ✓= l2 = l 2 u2 (4.10)
dt2 dt d✓ d✓2 d✓2 r2 d✓2

The equation of motion for the radial position, which we first derived back in (4.6), is

ml2
mr̈ = F (r)
r3
where, we’ve reverted to expressing the right-hand side in terms of the force F (r) =
dV /dr. Using (4.10), and doing a little bit of algebra (basically dividing by ml2 u2 ),
we get the second order di↵erential equation

d2 u 1
+u= F (1/u) (4.11)
d✓2 ml2 u2
This is the orbit equation. Our goal is to solve this for u(✓). If we want to subsequently
figure out the time dependence, we can always extract it from the equation ✓˙ = lu2 .

4.3.1 The Kepler Problem

The Kepler problem is the name given to understanding planetary orbits about a star.
It is named after the astronomer Johannes Kepler – we’ll see his contribution to the
subject in the next section.

We saw in Section 2.3 that the inverse-square force law of gravitation is described by
the central potential

km
V (r) = (4.12)
r
where k = GM . However, the results that we will now derive will equally well apply to
motion of a charged particle in a Coulomb potential if we instead use k = qQ/4⇡✏0 m.

For the potential (4.12), the orbit equation (4.11) becomes very easy to solve. It is
just

d2 u k
2
+u= 2
d✓ l

– 56 –
But this is just the equation for a harmonic oscillator, albeit with its centre displaced
by k/l2 . We can write the most general solution as
k
u = A cos (✓ ✓0 ) + (4.13)
l2
with A and ✓0 integration constants. (You might be tempted instead to write u =
A cos ✓ + B sin ✓ + k/l2 with A and B as integration constants. This is equivalent to
our result above but, as we will now see, it’s much more useful to use ✓0 as the second
integration constant).

At the point where the orbit is closest to the origin (the periapsis), u is largest. From
our solution, we have umax = A + k/l2 . We will choose to orient our polar coordinates
so that the periapsis occurs at ✓ = 0. This choice means that set ✓0 = 0. In terms of
our original variable r = 1/u, we have the final expression for the orbit
r0
r= (4.14)
e cos ✓ + 1
where
l2 Al2
r0 = and e=
k k
Notice that r0 is fixed by the angular momentum, while the choice of e is now e↵ectively
the integration constant in the problem.

You have seen equation (4.14) before (in the Vectors and Matrices course): it de-
scribes a conic section. If you don’t remember this, don’t worry! We’ll derive all the
necessary properties of this equation below. The integration constant e is called the
eccentricity and it determines the shape of the orbit.

Ellipses: e < 1
For e < 1, the radial position of the particle is bounded in the interval
r0
2 [1 e, 1 + e]
r
We can convert (4.14) back to Cartesian coordinates x = r cos ✓ and y = r sin ✓, writing

r = r0 er cos ✓ ) x2 + y 2 = (r0 ex)2

Multiplying out the square, collecting terms, and rearranging allow us to write this
equation as
(x xc ) 2 y2
+ =1
a2 b2

– 57 –
 P
b

a xc O O

Figure 15: The elliptical orbit with the Figure 16: The distance from between
origin at a focus the two foci and a point on the orbit is
constant

with

er0 r02 r02

xc = and a2 = and b2 = < a2 (4.15)
1 e2 (1 e2 )2 1 e2

This is the formula for an ellipse, with its centre shifted to x = xc . The orbit is
drawn in the figure. The two semi-axes of the ellipse have lengths a and b. The centre
of attraction of the gravitational force (for example, the sun) sits at r = 0. This is
marked by the yellow disc in the figure. Notice that it is not the centre of the ellipse:
the two points di↵er by a distance
r0 e
|xc | = = ea
1 e2

The origin where the star sits has special geometric significance: it is called the focus of
the ellipse. In fact, it is one of two foci: the other, shown as O0 in Figure above, sits at
equal distance from the centre along the major axis. A rather nice geometric property
of the ellipse is that the distance OP O0 shown in the second figure is the same for all
points P on the orbit. (You can easily prove this with some messy algebra).

When e = 0, the focus sits at the centre of the ellipse and lengths of the two axes
coincide: a = b. This is a circular orbit.

In the Solar System, nearly all planets have e < 0.1. This means that the di↵erence
between the major and minor axes of their orbits is less than 1% and the orbits are
very nearly circular. The only exception is Mercury, the closest planet to the Sun,
which has e ⇡ 0.2. For very eccentric orbits, we need to look at comets. The most
famous, Halley’s comet, has e ⇡ 0.97, a fact which most scientists hold responsible for
the Chas and Dave lyric “Halley’s comet don’t come round every year, the next time
it comes into view will be the year 2062”. However, according to astronomers, it will
be the year 2061.

– 58 –
Hyperbolae: e > 1
For e > 1, there are two values of ✓ for which r ! 1. They are cos ✓ = 1/e.
Repeating the algebraic steps that lead to the ellipse equation, we instead find that the
orbit is described by
✓ ◆2
1 r0 e y2
x =1
a2 e2 1 b2

with a2 = r02 /(e2 1)2 and b2 = r02 /(e2 1). This

is the equation for a hyperbola. It is plotted in the
figure, where the dashed lines are the asymptotes.
They meet at the point x = r0 e/(e2 1). Again, the
centre of the gravitational attraction sits at the origin
denoted by the yellow disc. Notice that the orbit goes
o↵ to r ! 1 when cos ✓ = 1/e. Since the right-
hand side is negative, this must occur for some angle
Figure 17: A hyperbola
✓ > ⇡/2. This is one way to see why the orbit sits in
the left-hand quadrant as shown.

Parabolae: e = 1
Finally, in the special case of e = 1, the algebra is particularly simple. The orbit is
described by the equation for a parabola,

y 2 = r02 2r0 x

The Energy of the Orbit Revisited

We can tally our solutions with the general picture of orbits that we built in Section
4.2.1 by looking at the e↵ective potential. The energy of a given orbit is

1 2 ml2 km
E= mṙ + 2
2 2r r
✓ ◆2
1 dr ml2 km
= m ✓˙2 + 2
2 d✓ 2r r
✓ ◆2 2
1 dr l ml2 km
= m +
2 d✓ r4 2r2 r

We can substitute in our solution (4.14) for the orbit to get

dr r0 e sin ✓
=
d✓ (1 + e cos ✓)2

– 59 –
After a couple of lines of algebra, we find that all the ✓ dependence vanishes in the
energy (as it must since the energy is a constant of the motion). We are left with the
pleasingly simple result
mk 2 2
E= (e 1) (4.16)
2l2
We can now compare this with the three cases we saw in Section 4.2.1:

• e < 1 ) E < 0: These are the trapped, or bounded, orbits that we now know
are ellipses.

• e > 1 ) E > 0: These are the unbounded orbits that we now know are
hyperbolae.

• e = 0 ) E = mk 2 /2l2 . This coincides with the minimum of the e↵ective

potential Ve↵ which we previously understood corresponds to a circular orbit.

A Repulsive Force
In the analysis above, we implicitly assumed that the
force is attractive, so k > 0. This, in turn, ensures
that r0 = l2 /k > 0. For a repulsive interaction, we
choose to write the solution (4.14) as
|r0 |
r= (4.17)
e cos ✓ 1
where |r0 | = l2 /|k| and e = Al2 /|k|. Note that with
Figure 18:
this choice of convention, e > 0. Since we must have
r > 0, we only find solutions in the case e > 1. This is nice: we wouldn’t expect to find
bound orbits between two particles which repel each other. For e > 1, the unbounded
hyperbolic orbits look like those shown in the figure. Notice that the orbits go o↵ to
r ! 1 when cos ✓ = 1/e which, since e > 0, must occur at an angle ✓ < ⇡/2. This is
the reason that the orbit sits in the right-hand quadrant.

4.3.2 Kepler’s Laws of Planetary Motion

In 1605, Kepler published three laws which are obeyed by the motion of all planets in
the Solar System. These laws were the culmination of decades of careful, painstaking
observations of the night sky, firstly by Tycho Brahe and later by Kepler himself. They
are:

• K1: Each planet moves in an ellipse, with the Sun at one focus.

– 60 –
 • K2: The line between the planet and the Sun sweeps out equal areas in equal
times.

• K3: The period of the orbit is proportional to the radius3/2 .

Now that we understand orbits, let’s see how Kepler’s laws can be derived from New-
ton’s inverse-square law of gravity.

We’ll start with Kepler’s second law. This is noth-

θ
ing more than the conservation of angular momentum.
From the figure, we see that in time t, the area swept
out is
1 dA 1 l Figure 19:
A = r2 ✓ ) = r2 ✓˙ =
2 dt 2 2
which we know is constant. This means that Kepler’s second law would hold for any
central force.

What about Kepler’s third law? This time, we do need the inverse-square law itself.
However, if we assume that the gravitational force takes the form F = GM m/r2 ,
then Kepler’s third law follows simply by dimensional analysis. The only parameter in
the game is GM which has dimensions

[GM ] = L3 T 2

So if we want to write down a formula relating the period of an orbit, T , with some
average radius of the orbit R (no matter how we define such a thing), the formula must
take the form
R3
T2 ⇠
GM
We already saw a version of this in Section 4.2.1 where we noted that, for circular
orbits, ✓˙2 ⇠ 1/r3 . For a general elliptical orbit, we can be more precise. The area of
an ellipse is
p ⇡r02
A = ⇡ab = ⇡a2 1 e2 =
(1 e2 )3/2

Since area is swept out at a constant rate, dA/dt = l/2, the time for a single period is
✓ ◆3/2
2A 2⇡r02 2⇡ r0
T = = = p
l l(1 e2 )3/2 GM 1 e2

– 61 –
The quantity in brackets indeed has the dimension of a length. But what length is it?
In fact, it has a nice interpretation. Recall that the periapsis of the orbit occurs at
rmin = r0 /(1 + e) and the apoapsis at rmax = r0 /(1 e). It is then natural to define the
average radius of the orbit to be R = 12 (rmin + rmax ) = r0 /(1 e2 ). We have

2⇡
T =p R3/2
GM

The fact that the inverse-square law implies Kepler’s third law was likely known to
several of Newton’s contemporaries, including Hooke, Wren and Halley. However, the
proof that the inverse-square law also gives rise to Kepler’s first law – a proof which
we have spent much of this section deriving – was Newton’s alone. This is one of the
highlights of Newton’s famous Principia.

4.3.3 Orbital Precession

For extremely massive objects, Newton’s theory of gravity needs replacing. Its successor
is Einstein’s theory of general relativity which describes how gravity can be understood
as the bending of space and time. You will have to be patient if you want to learn
general relativity: it is o↵ered as a course in Part II.

However, for certain problems, the full structure of general relativity reduces to
something more familiar. It can be shown that for planets orbiting a star, much of
the e↵ect of the curvature of spacetime can be captured in a simple correction to the
Newtonian force law, with the force now arising from the potential3
✓ ◆
GM m 3GM
V (r) = 1+ 2
r cr

where c is the speed of light. For r GM/c2 , this extra term is negligible and we
return to the Newtonian result. Here we will see the e↵ect of keeping this extra term.

We again define k = GM . After a little bit of algebra, the orbit equation (4.11) can
be shown to be
✓ ◆
d2 u 6k 2 k
2
+ 1 2 2
u= 2
d✓ cl l
3
In the lecture notes on General Relativity we will actually derive a 1/r3 correction to Newton’s
law of gravity. But general relativity is subtle and there are di↵erent ways of parameterising the radial
distance r. A di↵erent choice leads to the 1/r2 correction described above. Both approaches result in
the same answer for the perihelion precession.

– 62 –
The solution to this equation is very similar to that of the Kepler problem (4.13). It is
r !
6k 2 k
u(✓) = A cos 1 2 2
✓ + 2
cl l 6k 2 /c2

where we have once again chosen our polar coordinates so that the integration constant
is ✓0 = 0.

This equation again describes an ellipse. But now the ellipse precesses, meaning that
the periapsis (the point of closest approach to the origin) does not sit at the same angle
on each orbit. This is simple to see. A periapsis occurs whenever the cos term is 1.
This first happens at ✓ = 0. But the next time round, it happens at
✓ ◆ 1/2 ✓ ◆
6k 2 3k 2
✓ = 2⇡ 1 ⇡ 2⇡ 1 + 2 2
c2 l 2 cl
We learn that the orbit does not close up. Instead the periapsis advances by an angle
of 6⇡G2 M 2 /c2 l2 each turn.

The general relativistic prediction of the perihelion advance of Mercury – the closest
planet to the sun – was one of the first successes of Einstein’s theory.

4.4 Scattering: Throwing Stu↵ at Other Stu↵

In the past century, physicists have developed a foolproof and powerful method to
understand everything and anything: you take the object that you’re interested in
and you throw something at it. Ideally, you throw something at it really hard. This
technique was pioneered by Rutherford who used it to understand the structure of the
atom. It was used by Franklin, Crick and Watson to understand the structure of DNA.
And, more recently, it was used at the LHC to demonstrate the existence of the Higgs
boson. In short, throwing stu↵ at other stu↵ is the single most important experimental
method available to science. Because of this, it is given a respectable sounding name:
it is called scattering.

Before we turn to any specific problem, there are a few aspects that apply equally
well to particles scattering o↵ any central potential V (r). We will only need to assume
V (r) ! 0 as r ! 1. We do our experiment and throw the particle from a large
distance which we will take to be r ! 1. We want to throw the particle towards the
origin, but our aim is not always spot on. If the interaction is repulsive, we expect
the particle to be deflected and its trajectory will be something like that shown in
the figure. (However, much of what we’re about to say will hold whether the force is
attractive or repulsive).

– 63 –
 b
b

Figure 20:

Firstly, by energy conservation, the speed of the particle at the end of its trajectory
must be the same as the initial speed. (This is true since at r ! 1 at both the
beginning and end and there is no contribution from the potential energy). Let’s call
this initial/final speed v.

But, in a central potential, we also have conservation of angular momentum, L =

~
mr ⇥ ṙ. We can get an expression for l = |L|/m as follows: draw a straight line tangent
to the initial velocity. The closest this line gets to the origin is distance b, known as
the impact parameter. The modulus of the angular momentum is then

l = bv (4.18)

If this equation isn’t immediately obvious mathematically, the following words may
convince you. Suppose that there was no force acting on the particle at all. In this
case, the particle would indeed follow the straight line shown in the figure. When it’s
closest to the origin, its velocity ṙ is perpendicular to its position r and is its angular
momentum is obviously l = bv. But angular momentum is conserved for a free particle,
so this must also be its initial angular momentum. But, if this is the case, it is also the
angular momentum of the particle moving in the potential V (r) because there too the
angular momentum is conserved and can’t change from its initial value.

At the end of the trajectory, by the same kind of argument, the angular momentum
l is l = b0 v where b0 is the shortest distance from the origin to the exit asymptote as
shown in the figure. But since the angular momentum is conserved, we must have

b = b0

– 64 –
4.4.1 Rutherford Scattering
It was quite the most incredible event that ever happened to me in my life.
It was almost as incredible as if you fired a 15-inch shell at a piece of tissue
paper and it came back and hit you.

Ernest Rutherford

b
θ α
α
b φ

Figure 21:

Here we’ll look at the granddaddy of all scattering experiments. We take a particle
of charge q and mass m and throw it at a fixed particle of charge Q. We’ll ignore the
gravitational interaction and focus just on the repulsive Coulomb force. The potential
is
qQ
V =
4⇡✏0 r
This is mathematically identical to the gravitational force, so we can happily take all
the results from the last section and replace k = qQ/4⇡✏0 m in our previous equations.

Using our knowledge that b0 = b, we can draw another scatting event as shown. Here
✓ is the position of the particle. We will denote the total angle through which the
particle is deflected as . However, in the short term the angle ↵, shown in the figure,
will prove more useful. This is related to simply by

=⇡ 2↵ (4.19)

Our goal is to understand how the scattering angle depends on the impact parameter
b and the initial velocity v. Using the expression (4.17) for the orbit that we derived

– 65 –
earlier, we know that the particle asymptotes to r ! 1 when the angle is at ✓ = ↵.
This tells us that
1
cos ↵ =
e
As we mentioned previously, e > 1 which ensures that ↵ < ⇡/2 as shown in the figure.

There are a number of ways to proceed from here. Probably the easiest is if we use
the expression for energy. When the particle started its journey, it had E = 12 mv 2
(where v is the initial velocity). We can equate this with (4.16) to get

1 mk 2 mk 2
E = mv 2 = 2 (e2 1) = tan2 ↵
2 2l 2l2
Finally, we replace l = bv to get an the expression we wanted, relating the scattering
angle to the impact parameter b,
✓ ◆
1 |k|
= 2 tan (4.20)
bv 2

The result that we’ve derived here is for a potential with all the charge Q sitting at
the origin. We now know that this is a fairly good approximation to the nucleus of the
atom. But, in 1909, when Rutherford, Geiger and Marsden, first did this experiment,
firing alpha particles (Helium nuclei) at a thin film of gold, the standard lore was that
the charge of the nucleus was smeared throughout the atom in the so-called “plum
pudding model”. In that case, the deflection of the particle at high velocities would be
negligible. But, from (4.20), we see that, regardless of the initial velocity v, if you fire
a particle directly at the nucleus, so that b = 0, the particle will always be deflected by
a full = 180 . This was the result that so surprised Rutherford.

– 66 –