Chapter 5

The Relativistic Point Particle

To formulate the dynamics of a system we can write either the equations of

motion, or alternatively, an action. In the case of the relativistic point par-
ticle, it is rather easy to write the equations of motion. But the action is
so physical and geometrical that it is worth pursuing in its own right. More
importantly, while it is dicult to guess the equations of motion for the rela-
tivistic string, the action is a natural generalization of the relativistic particle
action that we will study in this chapter. We conclude with a discussion of
the charged relativistic particle.

5.1 Action for a relativistic point particle

How can we nd the action S that governs the dynamics of a free relativis-
tic particle? To get started we rst think about units. The action is the
Lagrangian integrated over time, so the units of action are just the units of
the Lagrangian multiplied by the units of time. The Lagrangian has units of
energy, so the units of action are

L2 M L2
[S] = M T = . (5.1.1)
T2 T
Recall that the action Snr for a free non-relativistic particle is given by the
time integral of the kinetic energy:


1 2 d
x
Snr = mv (t) dt , v 2
v
v ,
v = . (5.1.2)
2 dt

105
106 CHAPTER 5. THE RELATIVISTIC POINT PARTICLE

The equation of motion following by Hamiltons principle is

d
v
= 0. (5.1.3)
dt
The free particle moves with constant velocity and that is the end of the
story. Since even a free relativistic particle must move with constant veloc-
ity, how do we know that the action Snr is not correct in relativity? Perhaps
the simplest answer is that this action allows the particle to move with any
constant velocity, even one that exceeds that of light. The velocity of light
does not even appear in this action. Snr cannot be the action for a relativistic
point particle.

We now construct a relativistic action S for the free point particle. We

will do this by making an educated guess and then showing that it works
properly. But rst, how should we describe the motion of the particle? Since
we are interested in relativistic physics, it is convenient to represent the
motion in spacetime. The path traced out in spacetime by the motion of a
particle is called its world-line. In spacetime even a static particle traces a
line, since time always ows.
A key physical requirement is that the action must yield Lorentz invariant
equations of motion. Lets elaborate. Suppose a particular Lorentz observer
tells you that the particle is moving according to its equations of motion,
that is, that the particle is performing physical motion. Then, you should
expect that any other Lorentz observer will tell you that the particle is doing
physical motion. It would be inconsistent for one observer to state that a
certain motion is allowed and for another to state that the same motion is
forbidden. If the equations of motion hold in a xed Lorentz frame, they
must hold in all Lorentz frames. This is what is meant by Lorentz invariance
of the equations of motion.
We are going to write an action, and we are going to take our time to
nd the equations of motion. Is there any way to impose a constraint on the
action that will result in the Lorentz invariance of the equations of motion?
Yes, there is. We require the action to be a Lorentz scalar: for any particle
world-line all Lorentz observers must compute the same value for the action.
Since the action has no spacetime indices, it is indeed reasonable to demand
that it be a Lorentz scalar. If the action is a Lorentz scalar, the equations of
motion will be Lorentz invariant. The reason is simple and neat. Suppose one
Lorentz observer states that, for a given world-line, the action is stationary
5.1. ACTION FOR A RELATIVISTIC POINT PARTICLE 107

against all variations. Since all Lorentz observers agree on the values of
the action for all world-lines, they will all agree that the action is stationary
about the world-line in question. By Hamiltons principle, the world-line that
makes the action stationary satises the equations of motion, and therefore
all Lorentz observers will agree that the equations of motion are satised for
the world-line in question.
Asking the action to be a Lorentz scalar is a very strong constraint, and
there are actually valid grounds for suspecting that it may be too strong.
The action in (5.1.2), for example, is not invariant under a Galilean boost

v
v +
v0 with constant
v0 . Such a boost is a symmetry of the theory,
since the equation of motion (5.1.3) is invariant. We will nd, however,
that this complication does not arise in relativity. We can indeed nd fully
Lorentz-invariant actions.
Quick Calculation 5.1. Show that the variation of the action Snr under a
boost is a total time derivative.

Figure 5.1: A spacetime diagram with a series of world-lines connecting the origin
to the spacetime point (ctf , xf ).

We know that the action is a functional it takes as input a set of

functions describing the world-line and outputs a number S. Now let us
imagine a particle that is moving relativistically in spacetime, starting at
the initial position (0, 0) and ending at (ctf , xf ). There are many possible
world-lines between the starting and ending points, as shown in Figure 5.1
(our use of just one spatial dimension is only for ease of representation).
108 CHAPTER 5. THE RELATIVISTIC POINT PARTICLE

We would like our action to assign a number to each of these world-lines, a

number which all Lorentz observers agree on. Let P denote one world-line.
What quantity related to P do all Lorentz observers agree on? The answer is
simple: the elapsed proper time! All Lorentz observers agree on the amount
of time that elapsed on a clock carried by the moving particle. So lets take
the action of the world-line P to be the proper time associated to it.
To formulate this idea quantitatively, we recall that

ds2 = c2 dt2 + (dx1 )2 + (dx2 )2 + (dx3 )2 , (5.1.4)

where the innitesimal proper time is equal to ds/c. Of course, if we simply

take the integral of the proper time, then we will not have the correct units
for an action. Suppose instead that we integrate ds, which has the units of
length. To get the units of action we need a multiplicative factor with units
of mass times velocity. This factor should be Lorentz invariant, to preserve
the Lorentz invariance of our partial guess ds. For the mass we can use m,
the rest mass of the particle, and for velocity we can use c, the fundamental
velocity in relativity. If we had instead used the velocity of the particle, the
Lorentz invariance would have been spoiled. Therefore, our guess for the
action is the integral of (mc ds). Of course, there is still the possibility that
a dimensionless coecient is missing. It turns out that there should be a
minus sign, but the unit coecient is correct. We therefore claim that


S = mc ds , (5.1.5)
P

is the correct action. This action is so simple that it may be baing! It

probably looks nothing like the actions you have seen before. We can make
its content more familiar by choosing a particular Lorentz observer and ex-
pressing the action as an integral over time. With the help of (5.1.4) we can
relate ds to dt as 
v2
ds = c dt 1 2 . (5.1.6)
c
This allows us to write the action in (5.1.5) as an integral over time:


tf
v2
S = mc 2
dt 1 , (5.1.7)
ti c2
5.1. ACTION FOR A RELATIVISTIC POINT PARTICLE 109

where ti and tf are the values of time at the initial and nal points of the
world-line P, respectively. From this version of the action, we see that the
relativistic Lagrangian for the point particle is given by

v2
L = mc 1 2 .
2
(5.1.8)
c
The Lagrangian makes no sense when |
v | > c since it ceases to be real. The
constraint of maximal velocity is therefore implemented. This could have
been anticipated: proper time is only dened for motion where the velocity
does not exceed the velocity of light. The paths shown in Figure 5.1 all rep-
resent motion where the velocity of the particle never exceeds the velocity of
light. Only for such paths the action is dened. At any point in any of those
paths, the tangent vector to the path is a timelike vector.

To show that this Lagrangian gives the familiar physics in the limit of
small velocities, we expand the square root assuming |
v | << c. Keeping just
the rst term in the expansion gives
 1 v2  1
L  mc2 1 2 = mc2 + mv 2 . (5.1.9)
2c 2
Constant terms in a Lagrangian do not aect the equations of motion, so
when velocities are small the relativistic Lagrangian gives the same physics
as the the non-relativistic Lagrangian in (5.1.2). This also conrms that we
normalized our relativistic Lagrangian correctly.
The canonical momentum is the derivative of the Lagrangian with respect
to the velocity. Using (5.1.8) we nd

L 
v  1 m
v
p
= = mc 2 
2
= . (5.1.10)

v c 1 v /c
2 2 1 v 2 /c2

This is just the relativistic momentum of the point particle. What about the
Hamiltonian? It is given by

mv 2 
H = p

v L =  + mc2 1 v 2 /c2
1 v 2 /c2
mc2
=  . (5.1.11)
1 v 2 /c2
110 CHAPTER 5. THE RELATIVISTIC POINT PARTICLE

This coincides with the relativistic energy (2.4.2) of the point particle.
We have therefore recovered the familiar physics of a relativistic particle
from the rather remarkable action (5.1.5). This action is very elegant: it is
briey written in terms of the geometrical quantity ds, it has a clear physical
interpretation as total proper time, and it manifestly guarantees the Lorentz
invariance of the physics it describes.

5.2 Reparametrization invariance

In this section we will explore an important property of the point particle
action (5.1.5). This property is called reparametrization invariance. To eval-
uate the integral in the action, an observer will nd it useful to parametrize
the particle world-line. Reparametrization invariance of the action means
that the value of the action is independent of parametrization chosen to cal-
culate it. It should be so, since the action is dened independently of any
parametrization. We now investigate this point in detail and learn how to
rewrite the action in such a way that this property is manifest.

Figure 5.2: A world-line fully parametrized by . All spacetime coordinates x

are functions of .

To integrate ds, we parameterize the world-line P using a parameter ,

as shown in Figure 5.2. This parameter must be strictly increasing as the
world-line goes from the initial point xi to the nal point xf , but is otherwise
arbitrary. As ranges in the interval [i , f ] it describes the motion of the
5.2. REPARAMETRIZATION INVARIANCE 111

particle. To have a parametrization of the world-line means that we have

expressions for the coordinates x as functions of :
x = x ( ) . (5.2.1)
We also require
xi = x (i ) , xf = x (f ) . (5.2.2)
0
Note that even the time coordinate x is parameterized. Normally, we use
time as a parameter and describe position as a function of time. This is what
we did in section 5.1. But if we want to treat space and time coordinates
on the same footing, we must parameterize both in terms of an additional
parameter .
We now reexpress the integrand ds using the parametrized world-line. To
this end, we use ds2 = dx dx to write
dx dx
ds2 = (d )2 . (5.2.3)
d d
For any motion where the velocity does not exceed the velocity of light ds2 =
(ds)2 , and therefore the action (5.1.5) takes the form

f 
dx dx
S = mc d . (5.2.4)
i d d
This is the explicit form of the action when the path has been specied by a
parametrization x ( ) with parameter .
We have already seen that the value of the action is the same for all
Lorentz observers. We have now xed an observer, who has calculated the
action using some parameter . Does the value of the action depend on
the choice of parameter? It does not. The observer can reparametrize the
world-line, and the value of the action will be the same. Thus S is reparam-
eterization invariant.
To see this, suppose we change the parameter from to
. Then, by the
chain rule,
dx dx d

= . (5.2.5)
d d
d
Substituting back into (5.2.4), we get

f 
f

dx dx d
dx dx

S = mc
d = mc d ,
i d d
d i
d
d

(5.2.6)
112 CHAPTER 5. THE RELATIVISTIC POINT PARTICLE

which has the same form as (5.2.4), thus establishing the reparametrization
invariance. Because the verication of this property is quite simple, we say
that the action is manifestly reparameterization invariant.

5.3 Equation of motion

We now move on to the equations of motion. For this we must calculate the
variation S of the action (5.1.5), when the world-line of the particle is varied
by a small amount x . The variation is simply given by


S = mc (ds) . (5.3.1)

The variation of ds can be found from the simpler variation of ds2 . Since
(ds)2 = ds2 = dx dx , we nd

(ds)2 = 2 ds (ds) = (dx dx ) = 2 (dx )dx . (5.3.2)

The factor of two on the right-hand side arises because, by symmetry, the
variations of dx and dx give the same result. Simplifying a little,
dx
(ds) = (dx ) . (5.3.3)
ds
We now wish to understand why

(dx ) = d(x ) . (5.3.4)

You may already be familiar with this. A simpler version of this result states
that the variation of a velocity is just the derivative with respect to time of
the variation of the coordinate. Equation (5.3.4) becomes quite clear when
we spell out the meaning of d. For any -dependent quantity A( ) we have
dA = A( + d ) A( ). Using this on both sides of equation (5.3.4) we nd

[x ( + d ) x ( )] = x ( + d ) x ( ) . (5.3.5)

Since is linear, the two sides are indeed equal. We can understand the
equality (5.3.4) a little more geometrically by referring to Figure 5.3. The
original world-line of the particle is the lower curve, and the world-line after
variation is the upper curve. Shown in the lower curve are the points x ( )
5.3. EQUATION OF MOTION 113

and x ( + d ). By denition, they are separated by the innitesimal vector

dx . The variations x ( ) and x ( + d ) are also shown as vectors in the
gure, with their ends dening the varied points. In the upper curve, the sep-
aration between the varied points is the new dx: the old dx plus the variation
(dx). The little quadrilateral on the left-side of the gure is shown to the
right with labels a, a
, b, b
for the vectors that represent the corresponding
sides. We see that (dx ) corresponds to a
a. On the other hand d(x )
corresponds to b
b. The equality of these two quantities follows from the
vector identity a + b
= b + a
.

Figure 5.3: Relating (dx ) and d(x ) in the variation of a world-line.

Using (5.3.4) back in (5.3.3), we write the nal expression for (ds):

d(x ) dx
(ds) = ds . (5.3.6)
ds ds
We can now go ahead and vary the action using (5.3.1):

sf
d(x ) dx
S = mc ds . (5.3.7)
si ds ds

We introduced here explicit limits to the integration: si and sf denote the

values of the proper time parameter at the initial and nal points of the
world-line, respectively. To get an equation of motion we need to have x
multiplying an object under the integral the equation of motion is then
114 CHAPTER 5. THE RELATIVISTIC POINT PARTICLE

simply the vanishing of that object. Since there are still derivatives acting
on x , we must rewrite the integrand as a total derivative plus additional
terms where x appears multiplicatively :

sf
sf
d  dx
  d2 x 
S = mc ds x
ds x mc . (5.3.8)
si ds ds si ds2

The rst integral just gives some expression evaluated at the boundaries of
the world-line. But we x the coordinates on the boundaries, so the rst
term vanishes. The variation of the action then reduces to

sf  d2 x 
S = ds x (s) mc 2 . (5.3.9)
si ds

Since x (s) is arbitrary, everything multiplying it must vanish in order for

the variation of the action to be zero. We thus nd
d2 x
mc = 0. (5.3.10)
ds2
The constants are nonvanishing and can be removed. Even the Minkowski
metric can be removed: an equation of the form 0 = b , implies that
b = 0. Indeed, multiplying of the equation by you nd 0 = b =
b = b . Therefore, in its simplest form, the equation of motion is

d2 x
= 0. (5.3.11)
ds2
This is the equation of motion for a free particle. We obtained this equation
by varying the relativistic action for the point particle using fully-relativistic
notation. Equation (5.3.11) is formulated using the proper time parameter s.
It tells you that the four-velocity u is constant along the world-line:
d  dx  du dx
=0 = 0 , u = . (5.3.12)
ds ds ds ds
This means that if the path is marked by equal intervals of proper time, the
change in x between any successive pair of marks is the same. It also follows
from (5.3.12) that the relativistic momentum p = mu is constant along the
world-line:
dp
= 0. (5.3.13)
ds
5.4. RELATIVISTIC PARTICLE WITH ELECTRIC CHARGE 115

What if you used some arbitrary parameter instead of s? Equation (5.3.11)

would not hold with s replaced by . This is reasonable: changing arbitrarily
the parameter is like changing arbitrarily the marks, and as a consequence,
the change in x between any successive pair of new marks will not be the
same. It is actually possible to write a slightly more complicated version of
the equation of motion (5.3.11). That version uses an arbitrary parameter,
and is manifestly reparametrization invariant (see Problem 5.2).
Quick Calculation 5.2. Show that equation (5.3.13) implies that

dp
= 0, (5.3.14)
d
d
holds for an arbitrary parameter (s). Is there a constraint on ds
?
Our goal in this section has been achieved: we have shown how to derive
the physically expected equation of motion (5.3.11) starting from the Lorentz
invariant action (5.1.5). As we explained earlier, the resulting equation of
motion is guaranteed to be Lorentz invariant. Let us check this explicitly.
Under a Lorentz transformation, the coordinates x transform as indi-
cated in equation (2.2.32): x
= L x , where the constants L can be
viewed as the entries of an invertible matrix L. Since ds is the same in all
Lorentz frames, the equation of motion in primed coordinates is (5.3.11),
with x replaced by x
:

d2 x
d2 2
d x
0= = (L x ) = L . (5.3.15)
ds2 ds2 ds2
Since the matrix L is invertible, the above equation implies equation (5.3.11).
Namely, if the equation of motion holds in the primed coordinates, it holds
in the unprimed coordinates as well. This is the Lorentz invariance of the
equations of motion.

5.4 Relativistic particle with electric charge

The point particle we studied so far in this chapter is free and it moves with
constant four-velocity or four-momentum. If a point particle is electrically
charged, and there are nontrivial electromagnetic elds, the particle will ex-
perience forces and the four-momentum will not be constant. You know, in
fact, how the momentum of such particle varies in time. Its time derivative
116 CHAPTER 5. THE RELATIVISTIC POINT PARTICLE

is governed by the Lorentz force equation (3.1.5). The relativistic version of

the Lorentz force equation was given in Problem 3.1:
dp q
= 2 F u . (5.4.1)
ds c
This is an intricate equation. In the spirit of our previous analysis we try
to write an action that gives this equation upon variation. The action turns
out to be remarkably simple.

Since the Maxwell eld couples to the point particle along its world-
line, we should add to the action (5.1.5) an integral over P representing the
interaction of the particle with the electromagnetic eld. The integral must
be Lorentz invariant, and must involve the four-velocity of the particle. Since
the four-velocity has one spacetime index, to obtain a Lorentz scalar we must
multiply it against another object with one index. The natural candidate is
the gauge potential A . I claim that the interaction term in the action is


q dx
ds A (x(s)) (s) . (5.4.2)
c P ds
Here q is the electric charge, and the integral is over the world-line P,
parametrized with proper time. At each s, the four-velocity (dx /ds) is
multiplied by the gauge potential A evaluated at the position x(s) of the
particle. The integrand can be written more briey as A dx , by cancelling
the factors of ds. In this form, the interaction term is manifestly independent
of parametrization. The world-line of the particle is a one-dimensional space,
and the natural eld that can couple to a particle in a Lorentz invariant way
is a eld with one index. This will have an interesting generalization when
we consider the motion of strings. Since strings are one-dimensional, they
trace out two-dimensional world-sheets in spacetime. We will see that they
couple naturally to elds with two Lorentz indices!
The full action for the electrically charged point particle is obtained by
adding the term in (5.4.2) to (5.1.5):



q
S = mc ds + A (x)dx . (5.4.3)
P c P

This Lorentz invariant action is simple and elegant. It is the correct action,
and the equation of motion (5.4.1) arises by setting to zero the variation of
5.4. RELATIVISTIC PARTICLE WITH ELECTRIC CHARGE 117

S under a change x of the particle world-line. I do not want to take away

from you the satisfaction of deriving this important result. Therefore, I have
left to Problem 5.5 the task of varying the action (5.4.3).
118 CHAPTER 5. THE RELATIVISTIC POINT PARTICLE

Problems
Problem 5.1. Point particle equation of motion and reparametrizations.
If we parametrize the path of a point particle using proper time, the
equation of motion is (5.3.11). Consider now a new parameter = f (s).
What is the most general function f for which (5.3.11) implies
d2 x
= 0.
d 2
Problem 5.2. Particle equation of motion with arbitrary parametrization.
Vary the point particle action (5.2.4) to nd a manifestly reparametriza-
tion invariant form of the free particle equation of motion.
Problem 5.3. Current of a charged point particle.
Consider a point particle with charge q moving in a D = d+1 dimensional
spacetime as described by functions x ( ) = {x0 ( ),
x( )} where is param-
eter. Recall that the electromagnetic current j is dened as j = (c,
j)
where is the charge density (charge per unit volume) and
j is the current
density (current per unit area).
(a) Use delta functions to write expressions for j 0 (
x, t) and j i (
x, t) describ-
ing the electromagnetic current associated to the point particle.
(b) Show that the expressions you wrote in (a) arise from the following
integral representation


dx ( )
j (
x, t) = qc d d+1 (x x( ))

(1)
d
Here d+1 (x) (x0 )(x1 ) (xd ).
Problem 5.4. Hamiltonian for a non-relativistic charged particle.
The action for a non-relativistic particle of mass m and charge q coupled
to electromagnetic elds is obtained by replacing the rst term in (5.4.3) by
the non-relativistic action for a free point particle:



1 2 q dx
S= mv dt + A (x) dt . (1)
2 c dt
We have also chosen to use time to parametrize the second integral.
Problems for Chapter 5 119

and the or-

(a) Rewrite the above action in term of the potentials (, A)
dinary velocity
v . What is the Lagrangian?

(b) Calculate the canonical momentum p
conjugate to the position of the

particle and show that it is given by
q

p
= m
v + A . (2)
c

(c) Construct the Hamiltonian for the charged point particle and show it
is given by
1  q
2
H= p
A + q. (3)
2m c
Problem 5.5. Equations of motion for a charged point particle.
Consider the variation of the action (5.4.3) under a variation x (x) of the
particle trajectory. The variation of the rst term in the action was obtained
in section 5.3. Vary the second term (written more explicitly in (5.4.2)) and
show that the equation of motion for the point particle in the presence of an
electromagnetic eld is (5.4.1). Begin your calculation by explaining why
A
A (x(s)) = (x(s)) x (s) .
x
Problem 5.6. Electromagnetic eld dynamics with charged particle.
The action for the dynamics of both a charged point particle and the EM
eld is given by



 

q 1
S = mc ds + A (x)dx
dd+1 x F F .
P c P 16c

Here dd+1 x = dx0 dx1 dxd . Notice that the total action S
is a hybrid. The
last term is an integral over spacetime and the rst two terms are integrals
over the particle world-line. While included for completeness, the rst term
will play no role here. Vary the action S
under a uctuation A (x) and
obtain the equation of motion for the electromagnetic eld in the presence of
the charged point particle. The answer should be equation (3.3.23), where
the current is the one calculated in Problem 5.3. [ Hint: to vary A (x) in
the world-line action it is useful to rst turn this term into a full spacetime
integral with the help of delta functions].
120 CHAPTER 5. THE RELATIVISTIC POINT PARTICLE

Problem 5.7. Point particle action in curved space

In section 3.6 we considered the invariant interval ds2 = g (x)dx dx

in curved-space. The motion of a point particle of mass m on curved space
is studied using the action


S = mc ds .

Show that the equation of motion takes the form

d2 x
dx dx

+ = 0, (1)
ds2 ds ds
where
1  g g g 
= g + .
2 x x x
The Christoel coecients are symmetric in the lower indices = ,
and are built from the metric and its rst derivatives. When the metric
is constant the Christoel coecients vanish and we recover the familiar
equation of motion of the free point particle in Minkowski space. Equation (1)
is called the geodesic equation. It is a rather nontrivial dierential equation
that follows from the very simple action S.