Lecture notes PH4508

PH4508 Introduction to general relativity

Part I: Special relativity and Minkowski spacetime
Dr Chen Yu (y.chen@ntu.edu.sg)
School of Physical and Mathematical Sciences
Nanyang Technological University

Abstract
In this part, we briefly review some basic results of special relativity. The emphasis is on the Minkowski
spacetime structure that special relativity gives rise to. In general relativity, this structure of spacetime is
generalised to be dynamical, representing the effects of gravitation.

Contents
1 Fundamentals of special relativity
1.1 Einstein’s two postulates
1.2 Lorentz transformations
1.3 Consequences of Lorentz transformations
2 Minkowski spacetime and four-vectors
2.1 Minkowski spacetime
2.2 Four-vectors
2.3 Lorentz group
3 Relativistic dynamics and electromagnetism
3.1 Relativistic dynamics
3.2 Tensors in Minkowski spacetime
3.3 Electromagnetism

Page 1 of 22
Lecture notes PH4508

1 FUNDAMENTALS OF SPECIAL RELATIVITY

1.1 EINSTEIN’S TWO POSTULATES

Einstein’s special relativity is based on the following two postulates:

1. Relativity principle: The laws of physics are the same in all inertial reference frames.
2. The speed of light in vacuum is the same in all inertial reference frames, regardless of the motion
of the light source.

Relativity principle

The relativity principle is the requirement that laws of physics are the same in all inertial reference frames.
If a statement of physics is only valid in one particular inertial reference frame, it cannot be used by
another inertial observer to make predictions about what will happen in his reference frame. Such a
statement of physics cannot be called a law.

The relativity principle has a history before Einstein, going all the way back to Galileo and Newton. In the
Galileo version, laws considered are those in mechanics, which were pretty much everything known in
those days. Suppose that there are two inertial reference frames moving with constant velocities relative
to each other, each with its own stationary observer. These observers can conduct various sorts of
mechanical experiments in their reference frames. The relativity principle says that no experiments in
mechanics allow these observers to tell “which reference frame is really moving”. There is no such concept
as “absolute reference frame” and “absolute velocity”. Each inertial observer deduces the same laws of
mechanics—Newton’s laws of motion and has every right to believe that he/she is at rest while the
observer in the other reference frame is moving. No inertial reference frame is better than another.

Note that we need to distinguish physics phenomena from physics laws. The aforementioned observers
deduce the same laws of mechanics even though they may observe different physics phenomena in
mechanics. For example, one inertial observer may find that a ball is in free fall in the vertical direction,
but the second observer may find the same ball is in projectile motion. Both observers agree that the
motion of the ball obeys Newton’s second law.

Galilean transformations

The relativity principle of Galileo can be alternatively stated that Newton’s laws of motion are valid in all
inertial reference frames. This form of the principle can be proved if one accepts the following structure
of space and time. We define the notion of events first. An event is something that happens at a certain
time and a certain place, one example being a firecracker explodes at a certain place and a certain instant
of time. Suppose that we have two inertial reference frames S and S’. The reference frame S’ is moving in
the reference frame S with a velocity 𝑢 along the 𝑥-axis. The axes of the two reference frames are aligned.
The clocks in the two frames are synchronized such that when the two origins cross each other, the clocks
in both reference frames at the origins read time equal to zero. Such two reference frames appear
throughout our discussion of special relativity and is called the standard configuration, see Fig. 1. Suppose
that we have an event happening with time and space coordinates (𝑡, 𝑥, 𝑦, 𝑧) recorded in the S-frame and
coordinates (𝑡′ , 𝑥′ , 𝑦′ , 𝑧′ ) recorded in the S’ -frame.

Page 2 of 22
Lecture notes PH4508

Figure 1: Standard configuration, in both Galilean transformations and Lorentz

transformations. The star shape denotes an event.

In Newtonian mechanics, there is one time for all observers: time is absolute and is the same as measured
by S and S’, so one has 𝑡’ = 𝑡. Distances measured by different observers are also believed to be the
same in Newtonian mechanics, so 𝑥’ = 𝑥 − 𝑢𝑡. The space and time coordinates observed by different
observers in the standard configuration of an event are related by the so-called Galilean transformations:
𝑡′ = 𝑡
𝑥 ′ = 𝑥 − 𝑢𝑡
{ ′ (1)
𝑦 =𝑦
𝑧′ = 𝑧
Galilean transformations can be used to find how velocities and accelerations of an object as measured
by the two observers S and S’ are related. Suppose that the object is moving with a velocity 𝑣⃗ ′ and
acceleration 𝑎⃗′ in the S’ -frame, which is to say

𝑑𝑟⃗ ′ 𝑑2 𝑟⃗ ′
= 𝑣⃗ ′ , = 𝑎⃗′ (2)
𝑑𝑡 ′ 𝑑𝑡 ′2
The velocity and acceleration of the object observed by S will be

𝑑𝑥 𝑑 𝑑2 𝑥 𝑑𝑣𝑥 𝑑
𝑣𝑥 = = (𝑥 ′ + 𝑢𝑡) = 𝑣𝑥′ + 𝑢 𝑎𝑥 = 2
= = ′ (𝑣𝑥′ + 𝑢) = 𝑎𝑥′
𝑑𝑡 𝑑𝑡 𝑑𝑡 𝑑𝑡 𝑑𝑡
, (3)
𝑣𝑦 = 𝑣𝑦′ 𝑎𝑦 = 𝑎𝑦′
{𝑣𝑧 = 𝑣𝑧′ {𝑎𝑧 = 𝑎𝑧′
where the Galilean transformations (1) have been used.

Now suppose that observer S’ finds that Newton’s second law holds in his/her reference frame

𝐹⃗ ′ = 𝑚𝑎⃗′ (4)
In Newtonian mechanics, forces and masses are assumed to be invariant, so observer S will measure the
same force 𝐹⃗ = 𝐹⃗ ′ and the same mass as S’ does. Combining these results with (3), we can see that
observer S will also find that Newton’s second law holds in his/her reference frame

Page 3 of 22
Lecture notes PH4508

𝐹⃗ = 𝑚𝑎⃗ (5)
It is also clear that Newton’s first and third laws will hold in the S-frame if they hold in the S’-frame. Despite
the fact that the motion of objects looks different to different inertial observers (e.g., the velocity of an
object appears to have different numerical values according to different observers), the laws of physics
are, however, the same.

Constancy of speed of light

Everything works well and consistently until the discovery of electromagnetic waves. As you have learned
from your EM courses, Maxwell’s theory predicts the presence of electromagnetic waves which travel in
vacuum at a value 𝑐 = 3 × 108 m/s. This is the speed of light, known already at the time. Light was then
recognized as one form of electromagnetic waves. Now the question was, to which observer is the speed
of light equal to 3 × 108 m/s? People proposed that there is some medium called ether through which
the light is travelling. The speed 𝑐 = 3 × 108 m/s is the speed of light with respect to ether. Then
experiments were designed to measure the velocity of Earth relative to ether and found that it is zero all
the time. Since we know that Earth is moving and rotating, it is impossible to find a zero velocity of Earth
relative to ether all the time.

Einstein came into the scene, and said that, the speed of light, as measured by any inertial observer, is
equal to 3 × 108 m/s. There is no absolute reference frame like ether that is more privileged in
electromagnetism. Maxwell equations are valid in all inertial reference frames, and any inertial observer
makes the same prediction about the speed of light. In this way, Einstein solved the puzzle and saved the
principle of relativity. According to him, the principle of relativity applies to all laws of physics, including
those of electromagnetism.

The idea that the speed of light is the same in all inertial reference frames contradicts with many of our
cherished notions in Newtonian mechanics and has far-reaching and revolutionary consequences.
Consider a light pulse travelling in the S’-frame along the 𝑥′-axis. According to the Galilean velocity
transformation (3), the light pulse, when measured in the S-frame, should move to the right with a velocity
𝑢 + 𝑐. Now Einstein said that the velocity of the light pulse is still equal to 𝑐 when measured by S. The
transformations (3) and (1) must be wrong if one accepts the constancy of the speed of light. Galilean
transformations of space and time coordinates should be abandoned and replaced by new transformation
rules. You all know that the new rules are called Lorentz transformations. Newton’s laws of motion,
consistent with Galilean transformations, should also be modified to be consistent with the new
transformation rules. The derivation of Lorentz transformations based on the two postulates of Einstein
can be found in standard textbooks, which will be not repeated here. In the following section, we give a
quick and pedagogical derivation instead.

1.2 LORENTZ TRANSFORMATIONS

Notes: Unless otherwise specified, standard configuration is assumed in the remaining of this part.

We are thus led to find transformation rules that relate the coordinates (𝑡, 𝑥, 𝑦, 𝑧) of an event in the frame
S to the coordinates (𝑡′, 𝑥′, 𝑦′, 𝑧′ ) of the same event in the other frame 𝑆′ in the standard configuration.
The transformations should produce the same speed of light according to both observers. For this to be
possible, we must give up our good old notion of the absolute time coordinate. So, we must leave open

Page 4 of 22
Lecture notes PH4508

the possibility that 𝑡 ′ ≠ 𝑡 . We must also abandon the notion that a meter stick is a meter stick, and a
clock is a clock, and they should measure the same length and time regardless of their own motion.
Observer 𝑆′ may disagree on the measurement of length and time done by observer S, and vice versa.
Whatever length that S measures, 𝑆′ may think that it is not accurate and want to correct it by multiplying
it by a constant, and vice versa.

Suppose that, in the S-frame, an event is observed to have coordinates (𝑡, 𝑥, 𝑦, 𝑧). Then to S, the distance
of this event to the origin of the 𝑆′-frame along the 𝑥-direction is 𝑥 − 𝑢𝑡. But the observer 𝑆′ may disagree
on the distance and believe that he/she should multiply it by a factor denoted as 𝛾 to find the location of
the event:

𝑥 ′ = 𝛾(𝑥 − 𝑢𝑡) (6)

Similarly, if the same event is observed to have coordinates (𝑡′, 𝑥′, 𝑦′, 𝑧′) in the 𝑆′-frame, S may believe
that the distance of this event to his/her origin along the 𝑥-direction is

𝑥 = 𝛾(𝑥 ′ + 𝑢𝑡 ′ ) (7)

The factor 𝛾 appearing in the above two equations are the same, since neither S nor 𝑆′ can claim that
he/she is a more privileged observer.

The equations (6) and (7) should apply to any events that occur in space and time. Suppose that a light
pulse is traveling in the S-frame to the right with a speed 𝑐. According to the second postulate of Einstein,
the light pulse should also travel with the same speed 𝑐 in the 𝑆′-frame. The equations of trajectories of
the light pulse are then

𝑥 = 𝑐𝑡, 𝑥 ′ = 𝑐𝑡 ′ , (8)

respectively in the two reference frames. Substituting (8) in Eq. (6) and (7) yields

𝑐𝑡 ′ = 𝛾(𝑐 − 𝑢)𝑡, 𝑐𝑡 = 𝛾(𝑐 + 𝑢)𝑡 ′ . (9)

Taking the product of the above two equations and eliminating the factor 𝑡𝑡′, one can obtain

𝑐 2 = 𝛾 2 (𝑐 2 − 𝑢2 ), (10)
which can be solved to determine the factor 𝛾 as
1
𝛾= (11)
√1 − 𝑢2 /𝑐 2

Once 𝛾 is found, one can solve for 𝑡′ from Eq. (6) and (7) in terms of the unprimed coordinates
𝑢𝑥
𝑡 ′ = 𝛾 (𝑡 − ). (12)
𝑐2
Details of the derivation are left to the reader.

Since S and S’ are equally privileged, we should not expect any disagreement on the lengths they measure
along the y and z-directions. To summarise, we have found the transformation rules that relate the spatial
and time coordinates of the same event in two inertial reference frames S and S’ of the standard
configuration as follows:

Page 5 of 22
Lecture notes PH4508

𝑢𝑥
𝑡 ′ = 𝛾 (𝑡 − ),
𝑐2
𝑥 ′ = 𝛾(𝑥 − 𝑢𝑡), (13)
𝑦 ′ = 𝑦,
{𝑧 ′ = 𝑧,
The inverse transformation rules can be derived easily, and are given by

𝑢𝑥 ′
𝑡 = 𝛾 (𝑡 ′ + ),
𝑐2
𝑥 = 𝛾(𝑥 ′ + 𝑢𝑡 ′ ), (14)
𝑦 = 𝑦′,
{𝑧 = 𝑧 ′ .
The transformations (13) and (14) are known as Lorentz transformations. Almost everything that you
need to know and all the funny stuff that you have heard about special relativity can be derived from
these transformations. We now explore some of the most important consequences of Lorentz
transformations. You will see how special relativity revolutionises our understanding of the fundamental
concepts of nature such as space and time.

1.3 CONSEQUENCES OF LORENTZ TRANSFORMATIONS

Relativity of simultaneity

In Newtonian mechanics, it is believed that there is one time for every and all inertial observers. The clocks
carried by all inertial observes, once synchronized, will tick at the same rate and record the same time.
Two events occur simultaneously according to one inertial observer, will occur simultaneously according
to another inertial observer.

In special relativity, this is no longer true. Two events that occur simultaneously in one inertial reference
frame will, in general, not occur simultaneously in another inertial reference frame. One should note,
however, that two events that occur at the same place and time in one inertial reference frame appear to
occur at the same place and time in another inertial reference frame.

Exercise 1. Consider two events that occur simultaneously and are separated by a distance 𝑙 along the 𝑥-
direction according to S. Using Lorentz transformations, show that they will not occur simultaneously
according to 𝑆′ unless 𝑙 = 0.

Time dilation

In special relativity, the inertial observer S does not agree on the time unit used by the other observer 𝑆′
in the standard configuration. More explicitly, suppose that both S and 𝑆′ are carrying their own clocks, S
will find the clock carried by 𝑆′ run slower than his/her own clock. Neither of them can claim that they are
a reference frame more privileged than the other, so 𝑆′ will also find the clock carried by S runs slower
than his/her own. This interesting phenomenon is called time dilation in special relativity.

Time dilation is a nature of space and time in special relativity and affects anything that is in space and
time. It does not matter what kind of clocks the two observers are using: S can use a mechanical clock and
𝑆′ can use a biological clock, and they will not agree with each other’s clock once they have a relative

Page 6 of 22
Lecture notes PH4508

motion. If you are moving together with 𝑆′ with a speed u close to the speed of light, your clock will run
very slowly according to S. Your heartbeat will be very slow, your fingernails grow slow too, and everything
will be slow according to S. Time dilation is not science friction but real things that are happening all the
time now in high energy laboratories.

Exercise 2. If you are 𝑆′ and you take a nap in the afternoon, you measure the duration of your nap with
your clock and find it is equal to 𝜏 . Find the time S will measure for the duration of your nap.

Answer: The duration of your nap ∆𝑡 measured by S is

Δ𝑡 = 𝛾𝜏. (15)

Length contraction

A meter stick, once moving, is measured to be shorter in special relativity. This phenomenon is known as
length contraction. We measure the length of a rod, moving or not, by recording the spatial coordinates
of the two ends of the rod simultaneously and taking the difference of the two coordinates. The length of
an object, when measured in its own reference frame, is called the proper length of the object. The length
of a moving rod appears to be shorter than its proper length along the direction of motion.

Exercise 3. Suppose that the observer 𝑆′ is carrying a rod with length 𝑙0 aligned along the 𝑥′-axis. Find the
length of the rod ∆𝑙 that S will measure.

Answer: The length of the rod ∆𝑙 measured by S is

𝑙0
Δ𝑙 = . (16)
𝛾

Lorentz velocity transformations

The Newtonian velocity addition formula (3) has to be modified in special relativity to at least reconcile
the constancy of the speed of light in all inertial reference frames. We can in fact derive the relativistic
generalisation of (3) fairly easily using the Lorentz transformations (13) and (14). Suppose that a particle
is measured to have velocity 𝑣⃗ ′ = (𝑣𝑥′ , 𝑣𝑦′ , 𝑣𝑧′ ) in the 𝑆′-frame, namely,

𝑑𝑥 ′ 𝑑𝑦 ′ 𝑑𝑧 ′
= 𝑣𝑥′ , = 𝑣𝑦′ , = 𝑣𝑧′ (17)
𝑑𝑡 ′ 𝑑𝑡 ′ 𝑑𝑡 ′
The 𝑥-component of the velocity of the particle as measured by S can be worked out as follows:

𝑑𝑥 𝑑(𝛾(𝑥 ′ + 𝑢𝑡 ′ )) 𝑑(𝛾(𝑥 ′ + 𝑢𝑡 ′ ))/𝑑𝑡 ′ 𝑣𝑥′ + 𝑢

𝑣𝑥 = = = = . (18)
𝑑𝑡 𝑑(𝛾(𝑡 ′ + 𝑢𝑥 ′ /𝑐 2 )) 𝑑(𝛾(𝑡 ′ + 𝑢𝑥 ′ /𝑐 2 ))/𝑑𝑡 ′ 𝑢𝑣𝑥′
1+ 2
𝑐
One can similarly work out the 𝑦 and 𝑧 -components of the velocity to find the Lorentz velocity
transformations and the inverse as follows:

Page 7 of 22
Lecture notes PH4508

𝑣𝑥′ + 𝑢
𝑣𝑥 = 𝑣𝑥 − 𝑢
𝑢𝑣 ′ 𝑣𝑥′ = 𝑢𝑣
1 + 2𝑥 1 − 2𝑥
𝑐 𝑐
𝑣𝑦′ 𝑣𝑦
𝑣𝑦 = ′
𝑣𝑦 =
𝑢𝑣 ′ and 𝑢𝑣
γ (1 − 2𝑥 )
(19)
γ (1 + 2𝑥 ) 𝑐
𝑐
𝑣𝑧
𝑣𝑧′ ′
𝑣𝑧 =
𝑣𝑧 = 𝑢𝑣
𝑢𝑣𝑥′ { γ (1 − 2𝑥 )
γ (1 + ) 𝑐
{ 𝑐2
The funny factors appearing in the numerators of these transformations ensure that no signal, including
light, can travel faster than 𝑐.

2 MINKOWSKI SPACETIME AND FOUR-VECTORS

Lorentz transformations mix time and spatial coordinates of an event: The transformation of time
coordinate involves the spatial coordinates, and the transformation of spatial coordinates involves time.
Space by itself and time by itself have no observer-independent meaning in special relativity. Quoted from
Minkowski, “Henceforth space by itself, and time by itself, are doomed to fade away into mere shadows,
and only a kind of union of the two will preserve an independent reality.”

2.1 MINKOWSKI SPACETIME

Position four-vectors

Spacetime coordinates (𝑡, 𝑥, 𝑦, 𝑧) of an event in standard configuration transform in the particular way
according to the Lorentz transformations (13). Their transformations can be made more symmetric as
follows
′ 𝑢
𝑥 0 = 𝛾 (𝑥 0 − 𝑥 1 ) ,
𝑐
1′ 1
𝑢 0
𝑥 = γ (𝑥 − 𝑥 ) , (20)
𝑐
2′ 2
𝑥 =𝑥 ,
{𝑥 3 ′ = 𝑥 3 .
′ ′ ′ ′
if one defines ( 𝑥 0 = 𝑐𝑡, 𝑥 1 = 𝑥, 𝑥 2 = 𝑦, 𝑥 3 = 𝑧 ) and ( 𝑥 0 = 𝑐𝑡 ′ , 𝑥 1 = 𝑥 ′ , 𝑥 2 = 𝑦 ′ , 𝑥 3 = 𝑧 ′ ). The
quantity (𝑥 0 , 𝑥 1 , 𝑥 2 , 𝑥 3 ) defines what is known as the position four-vector in spacetime.

A quantity with four components that transform in the same way as the components (𝑥 0 , 𝑥 1 , 𝑥 2 , 𝑥 3 ) of
the position four-vector do under (all) Lorentz transformations is called a four-vector. We use 𝑉 μ to
denote a four-vector with the index μ understood to take values 0, 1, 2 and 3, where 0 is associated to
the time coordinate and 1, 2, and 3 are associated to spatial coordinates.

Minkowski spacetime

Page 8 of 22
Lecture notes PH4508

𝜇 𝜇 𝜇 𝜇
If 𝑥𝐴 and 𝑥𝐵 are two four-vectors, their difference is also a four-vector. Now suppose that 𝑥𝐴 and 𝑥𝐵 are
𝜇 𝜇
the position four-vectors of two neigbouring events in spacetime, the displacement Δ 𝑥 𝜇 = 𝑥𝐵 − 𝑥𝐴 is
also a four-vector, which transforms as
′ 𝑢
Δ𝑥 0 = 𝛾 (Δ𝑥 0 − Δ𝑥 1 ) ,
𝑐
1′ 1
𝑢 0
Δ𝑥 = 𝛾 (Δ𝑥 − Δ𝑥 ) , (21)
𝑐
′
Δ𝑥 2 = Δ𝑥 2 ,
{Δ𝑥 3 ′ = Δ𝑥 3 .
if one switches to 𝑆′-frame. It turns out that, although space itself or time itself loses its independent
significance in special relativity, a combination of the two does have an observer-independent meaning.
The displacement four-vector has an invariant

Δ𝑠 2 = −(𝑐Δ𝑡)2 + Δ𝑥 2 + Δ𝑦 2 + Δ𝑧 2 = −(Δ𝑥 0 )2 + (Δ𝑥 1 )2 + (Δ𝑥 2 )2 + (Δ𝑥 3 )2 , (22)

under Lorentz transformations. It is straightforward to show that the two inertial observers S and S’ in the
standard configuration will agree on their measurement of the quantity Δ𝑠 2 :
′ 2 ′ 2 ′ 2 ′ 2
−(Δx 0 )2 + (Δx1 )2 + (Δx 2 )2 + (Δx 3 )2 = −(Δx 0 ) + (Δx1 ) + (Δx 2 ) + (Δx 3 ) . (23)

Exercise 4. Prove that the quantity 𝛥𝑠 2 defined in (22) is invariant under Lorentz transformations.
Namely, prove (23).
μ μ
Two nearby events A and B, whose coordinates are 𝑥𝐴 and 𝑥𝐵 respectively, are called time-like separated
(space-like separated or light-like separated) if their corresponding invariant (22) is negative (positive or
zero). The invariant has the following intuitive meaning. If A and B are time-like separated, we can find an
inertial observer 𝑆̃ that moves with constant velocity and passes through these points along its worldline.
To this observer, events A and B occur at the same location but at different times. The time difference Δ𝑡̃
between the two events to this observer is called the proper time and is denoted by Δτ:

Δs 2 = −(cΔτ)2 + 02 + 02 + 02 ⇒ cΔτ = √−Δ𝑠 2 . (24)

𝜇 𝜇
If 𝑥𝐴 and 𝑥𝐵 are space-like separated, we can find an inertial observer that moves with constant velocity
and to which the two events occur simultaneously. This observer will measure the proper distance
between the two events equal to Δ𝑙:

Δs 2 = −02 + Δx 2 + Δy 2 + Δz 2 ⇒ Δl = √Δ𝑠 2 . (25)

Two light-like separated events can be connected by light signals and the above invariant Δ𝑠 2 is equal to
zero.
μ 𝜇
When the displacement four-vector between the two events 𝑥𝐴 and 𝑥𝐵 becomes infinitesimal 𝑑𝑥 μ , the
invariant (22) becomes

𝑑𝑠 2 = −(𝑑𝑥 0 )2 + (𝑑𝑥 1 )2 + (𝑑𝑥 2 )2 + (𝑑𝑥 3 )2 (26)

Eq. (26) is called the spacetime interval, or more often, the line element of the spacetime of special
relativity. The line element can also be represented in terms of what is known as the metric ημν as follows

Page 9 of 22
Lecture notes PH4508

𝑑𝑠 2 = ∑ 𝜂𝜇𝜈 𝑑𝑥 𝜇 𝑑𝑥 𝜈 = 𝜂𝜇𝜈 𝑑𝑥 𝜇 𝑑𝑥 𝜈 . (27)

𝜇,𝜈

It is worth emphasising that, in the above formula, 𝑑𝑥 𝜇 is the μ-th component of the infinitesimal
displacement four-vector between two nearby events. The metric 𝜂𝜇𝜈 carries two spacetime indices
μ, ν = 0 … 3, and can be represented as the following matrix:
−1 0 0 0
0 1 0 0 (28)
(𝜂𝜇𝜈 ) = ( )
0 0 1 0
0 0 0 1
In this representation, 𝜇 is a row index, while ν is a column index. The metric (tensor) 𝜂𝜇𝜈 is symmetric,
with diagonal entries equal to ±1. The entries +1 are associated with spatial coordinates, while the −1
entry is associated with the time coordinate.

In the last equality of the equation (27), we have used Einstein summation convention: TWO repeated
indices, EXACTLY one upper and one lower, are implicitly summed over. So 𝜂𝜇𝜈 𝑑𝑥 𝜇 𝑑𝑥 𝜈 means
∑μ,ν ημν 𝑑𝑥 𝜇 𝑑𝑥 𝜈 . Unless otherwise specified, Einstein summation convention is assumed throughout this
course. We say that the two repeated indices μ's are contracted (and so are the two ν's). The index ρ of a
ρσ
quantity say 𝑇 σ, is not contracted and labels different components of the quantity as its value varies
from ρ = 0 to ρ = 3. Such an index is called a free index. Two indices, once contracted, become dummy.
𝜌𝜎 𝜌𝜎
The indices σ's of 𝑇 𝜎 are dummy indices. So 𝑇 𝜎 is an object with one free upper index, and thus
contravariant vector. The quantity given by (27) has no free index so it is just one number and hence a
scalar (𝑑𝑠 2 is a scalar!).

Let me emphasise that the spacetime interval 𝑑𝑠 2 is an invariant in special relativity: all inertial reference
frames measure the same value of 𝑑𝑠 2 between two spacetime points. The line element (26) or the
metric (28) completely characterises the spacetime structure of special relativity. In particular, Lorentz
transformations and thus all the consequences that follow (such as time dilation and length contraction)
can be derived from it. The time and three-dimensional space are now unified into a four-dimensional
spacetime endowed with the metric structure (28) —the celebrated Minkowski spacetime. Minkowski
spacetime is often denoted as 𝑅 3,1 or 𝑅1,3 in the literature.

Euclidean space has a metric

You have actually encountered spaces with a metric structure. A simple example of such spaces is two-
dimensional Euclidean plane. Recall that two points in the Euclidean plane has a distance given by Δ𝑙 2 =
Δ𝑥 2 + Δ𝑦 2 . Another example is the usual three-dimensional Euclidean space, with distance between two
points given by Δ𝑙 2 = Δ𝑥 2 + Δ𝑦 2 + Δ𝑧 2. In such spaces, we have notions such as lengths, angles, straight
lines, parallel lines etc, all determined by the metric. A two-dimensional surface, such as an egg shell or
𝑥2 𝑦2
one given by the equation 𝑎2 + 𝑏2 + 2𝑧 = 1, in the three-dimensional Euclidean space has an induced
metric—the first fundamental form in the language of differential geometry.

Page 10 of 22
Lecture notes PH4508

2.2 FOUR-VECTORS
The Lorentz transformations (20) of spacetime coordinates in the two reference frames in the standard
configuration can be written as
′ ′
𝑥 𝜇 = Λ𝜇 𝜇 𝑥 μ , (29)
′
where Λμ μ are the components of the following matrix:
𝑢
γ −𝛾 0 0
𝑐
′ 𝑢
𝛬 = (𝛬𝜇 𝜇 ) = −𝛾 𝛾 0 0 (30)
𝑐
0 0 1 0
( 0 0 0 1)
In the expression above, μ′ and μ are understood as the row and column indices respectively. Since the
Lorentz transformations (29) are linear, the Jacobian matrix is just the Λ-matrix:
′
𝜇′
𝜕𝑥 𝜇
𝛬 𝜇 = (31)
𝜕𝑥 𝜇
A four-vector 𝑉𝜇 , by definition, also transforms as
′ ′
𝑉 𝜇 = 𝛬 𝜇 𝜇 𝑉𝜇 , (32)

under the Lorentz transformations represented by Λ.

Since a four-vector 𝑉 μ has the same transformation rule as the position four-vector does under Lorentz
transformations, it also has an invariant called the norm ‖𝑉𝜇 ‖ given by1

‖𝑉𝜇 ‖2 ≡ 𝜂𝜇𝜈 𝑉𝜇 𝑉 𝜈 = −(V 0 )2 + (V1 )2 + (V 2 )2 + (V 3 )2 . (33)

The norm of the four-vector ‖𝑉 μ ‖ defined above is the analogue of length in Euclidean space, so we will
often simply refer to ‖𝑉𝜇 ‖ as the length of 𝑉 μ in spacetime. The components of a four-vector change
under Lorentz transformations, but the norm is an invariant. Notice that, however, the norm of a four-
vector in spacetime can be imaginary. The metric does more than define the norm of a four-vector; it also
defines the dot product or the inner product of two four-vectors 𝑉𝜇 and 𝑈𝜇 in spacetime as
𝑉 ⋅ 𝑈 = 𝜂𝜇𝜈 𝑉𝜇 𝑈 𝜈 . (34)

All geometrical notions in Minkowski spacetime, such as straight lines, distance, volume, angle and parallel
transport, are derived from this dot product.

1
In these sets of notes, an equation 𝐴 ≡ 𝐵 means 𝐴 is defined as 𝐵.

Page 11 of 22
Lecture notes PH4508

2.3 LORENTZ GROUP

In Minkowski spacetime, different inertial observers measure different spacetime coordinates of the same
event (or spacetime point) but they agree on the spacetime interval. The invariance of the spacetime
interval
′ ′
𝑑𝑠 2 = 𝜂𝜇𝜈 𝑑𝑥 𝜇 𝑑𝑥 𝜈 = 𝜂𝜇′ 𝜈′ 𝑑𝑥 𝜇 𝑑𝑥 𝜈 (35)

under the Lorentz transformations (29) is equivalent to the requirement that

′ ′
𝜂𝜇𝜈 = Λ𝜇 𝜇 𝜂𝜇′ 𝜈′ Λ𝜈 𝜈 , (36)
or in the matrix form

𝜂 = Λ𝑇 𝜂Λ. (37)

Notice that, in both inertial reference frames, the metric has the same components, so (𝜂𝜇𝜈 ) = (𝜂𝜇′ 𝜈′ )
as two matrices.

Fig. 2: The transformation matrix (38) represents a rotation in the 𝑥-𝑦 plane in
spacetime by an angle θ. The Lorentz boost (39) can be understood as a
rotation in the 𝑡-𝑥 plane in spacetime by a boost parameter ϕ.

The transformations (29) and (30) are called a Lorentz boost along the 𝑥-direction and can be shown to
satisfy (37). There are two more Lorentz boost transformations along the y and z-directions respectively,
corresponding to inertial frames moving along these two directions relative to the S-frame. All
transformations Λ that satisfy (37) form the so-called Lorentz group2. Apart from Lorentz boosts, rotations
are obviously elements of the Lorentz group. Suppose that the S'-frame stays at rest but is rotated by an
angle θ in the 𝑥-𝑦 plane (or about the 𝑧-axis) of the S-frame (See Fig. 2), the transformation matrix that
relates the coordinate systems of the two observers is

2
A group is a set with an associative operation that have the following properties: it has an identity element; every
element has an inverse. The mathematical structure of groups has found many applications in physics.

Page 12 of 22
Lecture notes PH4508

1 0 0 0
𝜇′ 0 cos 𝜃 sin 𝜃 0 (38)
(𝛬 𝜇 ) = ( )
0 − sin 𝜃 cos 𝜃 0
0 0 0 1
This matrix can be shown to satisfy (37). There are two more rotation transformations corresponding to
the S'-frame rotated in the 𝑥-𝑧 plane and 𝑦-𝑧 plane of the S-frame respectively. While rotations rotate
axes in space only, Lorentz boosts can be thought of as to rotate space direction to time direction or vice
versa. So the Lorentz boost (30) along 𝑥-direction can be understood as a rotation in the 𝑡-𝑥 plane in
𝑢/𝑐
spacetime. In fact, one can define a boost parameter called ϕ given by = sinh ϕ so that the
√1−𝑢2 /𝑐 2
Lorentz boost (30) can be written as
cosh 𝜙 − sinh 𝜙 0 0
𝜇′ − sinh 𝜙 cosh 𝜙 0 0 (39)
(𝛬 𝜇 ) = ( )
0 0 1 0
0 0 0 1
whose form is analogous to that of a rotation (38).

Exercise 5. Verify that the rotation matrix (38) and the Lorentz boost matrix (39) satisfy (37). Identify
the range of the boost parameter ϕ in (39).

All rotations form a group itself, usually called the rotation group and denoted as 𝑆𝑂(3). All boosts and
rotations together with their combinations form what is known as the proper orthochronous group or
restricted Lorentz group 𝑆𝑂+ (3,1). The Lorentz group defined previously as matrices satisfying (37) is
often denoted as 𝑂(3,1) and contains the restricted Lorentz group as a subgroup. To fully generate the
Lorentz group, one should include the time-reversal and spatial-inversion transformations. These two
transformations are discrete transformations in the sense that they are not connected to the identity
transformation in a continuous way. They are not of particular interest in our course, but they do, however,
play an important role in areas of physics such as quantum field theory.

INFO Poincare group

A transformation that preserves the form of the metric of a given spacetime is called an isometry of the
spacetime. All isometries of Minkowski spacetime form what is known as the Poincare group. More
′
specifically, Poincare group constitutes all those transformations 𝑥 μ = 𝑥 μ (𝑥 μ ) such that ( 35) holds. All
translations of Minkowski spacetime form a subgroup called the translation group. The translation group
and the Lorentz group together with their combinations form the Poincare group.

3 RELATIVISTIC DYNAMICS AND ELECTROMAGNETISM

Newton's laws of motion which were shown to be invariant under Galilean transformations cannot be the
correct laws of motion in special relativity. They have to be modified to be invariant 3 under Lorentz

3
Invariance here should be interpreted as covariance in special relativity. This is because in special relativity, unlike
in Newtonian mechanics, neither the generalisation of force nor that of acceleration is invariant under Lorentz
transformations. Both acceleration and force transform in a specified way in special relativity—the meaning of
covariance.

Page 13 of 22
Lecture notes PH4508

transformations to comply with the relativity principle. Laws of physics are formulated as tensor equations
in special relativity. Tensors in special relativity have specified transformation rules under Lorentz
transformations, so if a tensor equation is true in one inertial reference, it is true in any other inertial
frames. Four-vectors and metric tensors are special types of tensors.

3.1 RELATIVISTIC DYNAMICS

Four-velocity

Now consider the motion of a particle in special relativity. The curve (or path) traced out by the particle
in spacetime is called the worldline of the particle. A small displacement 𝑑𝑥 μ along the worldline of the
particle is a four-vector. What is then the generalization of the concept of velocity in special relativity?
Notice that the time difference 𝑑𝑥 0 of the displacement four-vector is no longer an invariant as in
𝑑𝑥 μ
Newtonian mechanics, so the naive guess 𝑑𝑡
is not a four-vector. Recall that the proper time 𝑑τ defined
along the worldline of the particle is an invariant. We are thus lead to define the velocity four-vector of
the particle as
𝑑𝑥 𝜇
𝑈𝜇 = . (40)
𝑑𝜏
𝑑𝑟⃗ 𝑑𝑥 𝑑𝑦 𝑑𝑧
Suppose that the particle has a 3-velocity 𝑣⃗ ≡ 𝑑𝑡
≡ ( 𝑑𝑡 , 𝑑𝑡 , 𝑑𝑡 ). It is easy to show that the proper time
𝑑τ as measured by the particle itself is related to the coordinate time dt by

𝑣2
𝑑𝜏 = 𝑑𝑡√1 − , (41)
𝑐2

and the four-velocity defined above can be expressed in terms of the 3-velocity 𝑣⃗ as
1
𝑈𝜇 = (𝑐, 𝑣⃗). (42)
√1 − 𝑣 2 /𝑐 2

The norm of the four-velocity just defined satisfies

‖𝑈𝜇 ‖2 = −𝑐 2 . (43)
Exercise 6. Derive (41). What is the four-velocity of a stationary particle?

Energy-momentum four-vector

It is natural to define the momentum four-vector of a particle as

𝑃𝜇 = 𝑚0 𝑈𝜇 , (44)

where 𝑚0 is the (rest) mass of the particle. The momentum four-vector can be expressed in terms of the
3-velocity of the particle as

𝑚0 𝑐 𝑚0 𝑣⃗ 𝐸
𝑃𝜇 = ( , ) ≡ ( , 𝑝⃗) . (45)
√1 − 𝑣 2 /𝑐 2 √1 − 𝑣 2 /𝑐 2 𝑐

Page 14 of 22
Lecture notes PH4508

The last equality of the previous equation defines the relativistic energy 𝐸 and relativistic momentum 𝑝⃗
of the particle, which satisfy the famous energy-momentum relation

𝐸 2 = 𝑚02 𝑐 4 + 𝑝2 𝑐 2 . (46)

We can expand 𝐸 in terms of 𝑣 as

1 3𝑚0 𝑣 4
𝐸 = 𝑚0 𝑐 2 + 𝑚0 𝑣 2 + +⋯, (47)
2 8𝑐 2
from which we can see that the relativistic energy of a particle contains its rest energy, the Newtonian
kinetic energy and higher-order corrections.

Exercise 7. Derive (46) and (47).

The momentum four-vector is also called energy-momentum four-vector, since its first component is
essentially energy, while the other components represent momentum. The conservation law of energy
and momentum for a closed system in Newtonian mechanics can now be generalised and translated to
the conservation of the energy-momentum four-vector in special relativity. Because of this, the energy-
momentum four-vector is an extremely important and useful quantity in special relativity.

Four-force and Newton's second law

𝑑𝑝⃗
Newton's second law 𝐹⃗ = 𝑑𝑡 has to be modified to be Lorentz covariant (refer to footnote 3). The
relativistic version of Newton's second law is
𝑑𝑃𝜇
𝐾𝜇 = , (48)
𝑑𝜏
where 𝑃μ is the energy-momentum four-vector of the object under consideration and 𝐾 μ is called the
force four-vector. Both sides of the equation are four-vectors. The Lorentz covariance of the equation is
manifest.

3.2 TENSORS IN MINKOWSKI SPACETIME

As already mentioned, tensors in Minkowski spacetime are defined to formulate laws of physics in a
manifestly covariant form. A physics law written as a tensor equation is true in all inertial reference frames
if it is true in one inertial reference frame—The relativity principle will be obeyed. In this section, we give
a gentle introduction to tensors in Minkowski spacetime. More general tensors in arbitrary spacetimes
(not necessarily Minkowski) will be the subject in Part II of this course.

Tensors at a point

Tensors at a given point in Minkowski spacetime in special relativity are a set of quantities known as the
components that transform in a specified way under Lorentz transformations. Depending on the
transformation properties, tensors are divided into different types. The simplest tensor is a scalar. At a
given point in spacetime, a scalar φ is a quantity that does not change under Lorentz transformations.

Exercise 8. Is the relativistic energy E of a particle a scalar quantity in special relativity?

Page 15 of 22
Lecture notes PH4508

Four-vectors are the next simplest type of tensors in special relativity, the prototype of which is the
position four-vector. Recall that a four-vector 𝑉 μ is a set of four quantities (𝑉 0 , 𝑉 1 , 𝑉 2 , 𝑉 3 ) known as the
components of 𝑉 that transform in the same way as the coordinates of spacetime do under Lorentz
transformations. More specifically, the components 𝑉 μ transform as
′ ′
𝑉 𝜇 = Λ𝜇 𝜇 𝑉 𝜇 , (49)

under arbitrary Lorentz transformations

′ ′
𝑥 𝜇 = Λ𝜇 𝜇 𝑥 𝜇 . (50)

A four-vector defined in this way is known as a contravariant four-vector. Examples of contravariant

vectors include velocity and energy-momentum four-vectors.

There is another type of four-vectors in Minkowski spacetime known as covariant four-vectors, whose
components transform in the opposite way as those of contravariant vectors do. The meaning of
“opposite way” is as follows. We first find the transformations inverse to (50), namely the relation
′
expressing 𝑥 μ in terms of 𝑥 μ
′
𝑥 𝜇 = (Λ−1 )𝜇𝜇′ 𝑥 𝜇 . (51)
μ
Substituting (50) in (51), we can see that the transformation matrix ((Λ−1 )μ′ ) in (51) is indeed the matrix
μ′
inverse of the transfmation matrix (Λ μ ) in (50):
′ 𝜇
(Λ−1 )𝜇𝜇′ Λ𝜇 𝜈 = 𝛿𝜈 . (52)
μ
Here, δν is the Kronecker delta defined as
𝜇 1 if 𝜇 = 𝜈
𝛿𝜈 = { (53)
0 otherwise
As an example, for the Lorentz boost transformations ( 50 ) or the boost matrix ( 30 ), the inverse
transformations are given by (51) or the following inverse matrix
𝛾 𝛾𝑢/𝑐 0 0
𝛾𝑢/𝑐 𝛾 0 0
𝛬−1 = ((𝛬−1 )𝜇𝜇′ ) = ( ) (54)
0 0 1 0
0 0 0 1
A covariant four-vector ωμ is a set of four quantities (ω0 , ω1 , ω2 , ω3 ) known as the components of ω that
transform as

𝜔𝜇′ = (Λ−1 )𝜇𝜇′ 𝜔𝜇 , (55)

under the Lorentz transformations (50). Further discussions on covariant vectors will be deferred to Part
II of the course. From now on, a vector means a contravariant vector unless indicated otherwise.

We next generalise the concept of four-vectors 𝑉 μ to quantities with more indices—tensors of general
type. A tensor of type (2,0) or (2,0)-tensor is a set of 4 × 4 quantities 𝑇 μν known as the components of T
that transform as

Page 16 of 22
Lecture notes PH4508

′𝜈′ ′ ′
𝑇𝜇 = Λ𝜇 𝜇 Λ𝜈 𝜈 𝑇𝜇𝜈 , (56)

under Lorentz transformations (50). Similarly, a tensor of type (0,2) or (0,2)-tensor is a set of 4 × 4
quantities 𝑇μν known as the components of 𝑇 that transform as

𝑇𝜇′ 𝜈′ = (Λ−1 )𝜇𝜇′ (Λ−1 )𝜈 𝜈′ 𝑇𝜇𝜈 , (57)

under the Lorentz transformations (50). A tensor of type (1,1) or (1,1)-tensor is a set of 4 × 4 quantities
𝑇 μ ν known as the components of 𝑇 that transform as
′ ′
𝑇𝜇 𝜈′ = Λ𝜇 𝜇 (Λ−1 )𝜈 𝜈′ 𝑇𝜇 𝜈 , (58)

under the Lorentz transformations (50). In the transformation rules (56), (57) and (58) each upper index
is associated with one factor of Λ and each lower index is associated with one factor of Λ−1.

The components of a general tensor in Minkowski spacetime will transform and thus be different in
different inertial reference frames. There are some tensors whose components remain the same in all
inertial reference frames. These tensors are called Lorentz-invariant tensors in Minkowski spacetime.
Some examples of these tensors can be found in the exercises below.

Exercise 9. Show that the quantity 𝑇μν whose components are all equal to zero in every inertial coordinate
system is a tensor. Such a tensor and its obvious generalisations to arbitrary types are called zero tensors.

Exercise 10. Show that the quantity 𝑇μν whose components are given by the following matrix:

1 0 0 0
0 1 0 0 (59)
(𝑇𝜇𝜈 ) = ( ),
0 0 1 0
0 0 0 1
in every inertial coordinate system is NOT a (0,2)-tensor. Show that it is instead a (1,1)-tensor, i.e., the
μ
quantity 𝑇 μ ν ≡ δν in every inertial coordinate system is a tensor of type (1,1).

The Minkowski metric tensor

Recall that, from the invariance of spacetime interval, we derived the important property (36) of the
Minkowski metric. This property can be rewritten as

𝜂𝜇′ 𝜈′ = (𝛬−1 )𝜇𝜇′ 𝜂𝜇𝜈 (𝛬−1 )𝜈 𝜈′ , (60)

for an arbitrary Lorentz transformation Λ. Comparing (57) and (60), we conclude that the Minkowski
metric ημν given by (28) in every inertial coordinate system is a (0,2)-tensor.

The inverse metric, defined as the set of quantities arranged as the following matrix in every inertial
coordinate system 𝑥 μ
−1 0 0 0
(𝜂 𝜇𝜈 ) = ( 0 1 0 0
), (61)
0 0 1 0
0 0 0 1

Page 17 of 22
Lecture notes PH4508

is a (2,0)-tensor (Please verify this result yourself). The metric tensor arranged as a matrix (𝜂𝜇𝜈 ) and the
inversed metric tensor arranged as a matrix (𝜂 𝜇𝜈 )are the matrix inverse of each other, i.e.,
𝜇
𝜂 𝜇𝜈 𝜂𝜈𝜌 = 𝛿𝜌 . (62)

Notice that the metric tensor and inverse metric tensor are different objects, although they are identical
as two matrices (the difference will be apparent if spherical coordinates are used).

The metric tensor can be used to convert a contravariant vector to a covariant vector and vice versa. For
example, if 𝑉 μ a contravariant four-vector, the quantity 𝜂𝜇𝜈 𝑉 𝜈 is a covariant four-vector and is usually
denoted by the same symbol 𝑉, namely:
𝑉𝜇 ≡ 𝜂𝜇𝜈 𝑉 𝜈 . (63)

So the metric tensor is used to lower the index of a contravariant four-vector 𝑉 μ to give a covariant four-
vector 𝑉μ . Conversely, the inverse metric tensor can be used to raise the index of a covariant four-vector
ωμ to give the corresponding contravariant four-vector ωμ :

𝜔 𝜇 = 𝜂 𝜇𝜈 𝜔𝜈 . (64)

Exercise 11. What are the components of the covariant vectors 𝑥μ and 𝑃μ in terms of quantities 𝑡, 𝑥, 𝑦, 𝑧
and 𝐸, 𝑝𝑥 , 𝑝𝑦 , 𝑝𝑧 respectively?

The above procedure can be applied repeatedly to raise and lower more than one index of a tensor of
general type. In each step, the tensor is multiplied by the metric or the inverse metric, and one index of
the tensor is contracted with the index of the (inverse) metric tensor. For example, given a (2,0)-tensor
𝑇 μν, we can lower its indices to construct a (1,1)-tensor 𝑇 μ ν and (0,2)-tensor 𝑇μν as follows

𝑇𝜇 𝜈 = 𝜂𝜈𝜌 𝑇𝜇𝜌 ,
{ (65)
𝑇𝜇𝜈 = 𝜂𝜇𝜌 𝑇𝜌 𝜈 = 𝜂𝜇𝜌 𝜂𝜈𝜎 𝑇𝜌𝜎

Tensor fields

By assigning a tensor at each point in Minkowski spacetime we get a tensor field. In an inertial coordinate
system, a scalar field is a function φ of the coordinates 𝑥 μ . Under the Lorentz transformations (50), the
′
value of the scalar at a point with new coordinates 𝑥 μ is given by φ(𝑥 μ ) at the same point with 𝑥 μ =
′ ′
(Λ−1 )μ μ′ 𝑥 μ . So in the new coordinates 𝑥 μ , the scalar field is represented by the function
′ ′
𝜑(𝑥 𝜇 ) = 𝜑(𝑥 𝜇 ) where 𝑥 𝜇 = (𝛬−1 )𝜇𝜇′ 𝑥 𝜇 . (66)

Exercise 12. Is (electric) charge density ρ a scalar field in special relativity?

A vector field is the assignment of a vector at each point in spacetime, so it is a set of four functions 𝑉 μ
of the coordinates 𝑥 μ in an inertial coordinate system. Under the Lorentz transformations (50), the
′ ′ ′ ′
components 𝑉 μ of the vector at any point 𝑥 ν are given by Λμ μ 𝑉 μ at the same point 𝑥 ν = (Λ−1 )ν ν′ 𝑥 ν .
′
So in the new coordinates 𝑥 μ , the vector field is given by the following functions
′ ′ ′ ′
𝑉𝜇 (𝑥 𝜈 ) = 𝛬𝜇 𝜇 𝑉𝜇 (𝑥 𝜈 ) where 𝑥 𝜈 = (𝛬−1 )𝜈 𝜈′ 𝑥 𝜈 . (67)

Page 18 of 22
Lecture notes PH4508

Generalisation to tensors of general type is straightforward. For example, a (1,1)-tensor field given by the
functions 𝑇 μ ν (𝑥 ρ ) in inertial coordinates 𝑥 μ transforms accordings to
′ ′ ′ ′
𝑇𝜇 𝜈′ (𝑥 𝜌 ) = 𝛬𝜇 𝜇 (𝛬−1 )𝜈 𝜈′ 𝑇𝜇 𝜈 (𝑥 𝜌 ) where 𝑥 𝜌 = (𝛬−1 )𝜌𝜌′ 𝑥 𝜌 . (68)
under the Lorentz transformation (50).

We have a very important result on tensor fields in Minkowski spacetime: The partial derivative ∂ν 𝑇 … … of
a tensor field 𝑇 … … in Minkowski spacetime is a tensor field under Lorentz transformations. For example,
if 𝑉 μ is a vector field, then ∂ν 𝑉 μ is a tensor field of type (1,1). This example can be proved as follows. We
′ ′
know that in the new inertial coordinate system 𝑥 μ , the vector field has components 𝑉 μ given by (67).
Taking partial derivative, we get

′ ′ ′ ′ 𝜕𝑉𝜇 𝜕𝑥 𝜈 𝜇′
𝜕𝜈′ 𝑉𝜇 = 𝜕𝜈′ (𝛬𝜇 𝜇 𝑉𝜇 (𝑥 𝜈 )) = 𝛬𝜇 𝜇 𝜕𝜈′ 𝑉𝜇 (𝑥 𝜈 ) = 𝛬𝜇 𝜇 𝜈 𝜈 ′ = 𝛬 𝜇 (𝛬
−1 )𝜈 ′ 𝜇
𝜈 𝜕𝜈 𝑉 , (69)
𝜕𝑥 𝜕𝑥
where for the third equality we have used the chain rule of partial derivatives. This is the transformation
rule (6868) of a (1,1)-tensor field. Caution: In a curved spacetime, the result that partial derivative ∂ν 𝑇 … …
of a tensor field 𝑇 … … is a tensor field no longer holds. To generalise this result to curved spacetime, we
need to introduce the notion of covariant derivative, to be discussed in part II of the course.

INFO Realisation of tensor fields in physics

The underlying spacetime in standard quantum field theory (QFT) is Minkowski spacetime. So tensors (and
tensor fields) in QFT transform in the way specified here. The Klein-Gordon field ϕ is a scalar field. In
standard model of particle physics, Higgs bosons are the field quanta of the Higgs field, which is a scalar
field (the Higgs field also has internal degrees of freedom and is a representation of certain gauge groups).
Electromagnetic field can be described by a covariant vector field 𝐴μ or equivalently an antisymmetric (0,2)-
tensor field 𝐹μν as we are going to see in the following section. Metric tensor ημν is a (constant) symmetric
(0,2)-tensor field with no dynamics in standard QFT. You can consult any standard textbooks on quantum
field theory to learn more about them. General relativity gives dynamics to metric tensor which represents
effects of gravity.

3.3 ELECTROMAGNETISM
All laws of physics including laws of electromagnetism should obey Einstein's principle of relativity. In this
section, we recast the laws of electromagnetism in a form that is manifestly covariant under Lorentz
transformations in terms of tensors. In the new form, the fact that electromagnetism obeys the relativity
principle becomes transparent.

Lorentz four-force and electromagnetic field tensor

The dynamics of an object can be determined only if the four-force acting on it in Eq. (48) is specified. We
know that a charged particle moving in an electromagnetic field with a velocity 𝑣⃗ experiences a Lorentz
force given by 𝑞(𝐸⃗⃗ + 𝑣⃗ × 𝐵
⃗⃗), where 𝐸⃗⃗ and 𝐵
⃗⃗ are the electric and magnetic fields respectively. What is the
relativistic version of the Lorentz force? We notice that the force is linear in terms of the velocity with
coefficients given by the electric field 𝐸⃗⃗ and magnetic field 𝐵 ⃗⃗. (Recall linear algebra: A linear map can be
represented by a matrix once a basis is chosen.) This leads us to propose the following form of the Lorentz
four-force

Page 19 of 22
Lecture notes PH4508

𝐾𝜇 = 𝑞𝐹𝜇 𝜈 𝑈 𝜈 , (70)

where the coefficients 𝐹 μ ν should be a certain package of the electric and magnetic fields and 𝑈 μ is the
velocity four-vector of the charge. Combing (48) and (70), the equation of motion of a charged particle in
an electromagnetic field then becomes
𝑑𝑃𝜇
= 𝑞𝐹𝜇 𝜈 𝑈 𝜈 . (71)
𝑑𝜏
In order to reproduce the familiar equation of motion (73) of a charged particle in an electromagnetic
field, the quantity 𝐹 μ ν can be shown to have the following form
𝐸𝑥 𝐸𝑦 𝐸𝑧
0
𝑐 𝑐 𝑐
𝐸𝑥
0 𝐵𝑧 −𝐵𝑦
𝐹𝜇 𝜈 = 𝐸𝑐 (72)
𝑦
−𝐵𝑧 0 𝐵𝑥
𝑐
𝐸𝑧
𝐵𝑦 −𝐵𝑥 0
(𝑐 )

The quantity 𝐹 μ ν, a (1,1)-tensor field, packages all information about the electric field 𝐸⃗⃗ and the magnetic
⃗⃗, and is known as the electromagnetic field tensor. By observing the motion of the charged particle,
field 𝐵
one can in principle determine all components of 𝐹 μ ν according to (71).

Exercise 13. Show that the equation of motion ( 71 ) of a charged particle in the presence of an
electromagnetic field is equivalent to the following two equations that you are familiar with

𝑑𝐸 𝑑𝑝⃗
= 𝑞𝐸⃗⃗ ⋅ 𝑣⃗, = 𝑞(𝐸⃗⃗ + 𝑣⃗ × 𝐵
⃗⃗). (73)
𝑑𝑡 𝑑𝑡
Notice that in these two equations 𝐸 is the relativistic energy of the charge, while 𝐸⃗⃗ is the electric field.

We can define the electromagnetic field tensor with two upper indices 𝐹 μν by raising the index ν in (72)
using the inverse metric tensor ημν as follows
𝐸𝑥 𝐸𝑦 𝐸𝑧
0
𝑐 𝑐 𝑐
𝐸𝑥
− 0 𝐵𝑧 −𝐵𝑦
F μν
≡F μ ρν 𝑐 (74)
ρη = 𝐸𝑦
− −𝐵𝑧 0 𝐵𝑥
𝑐
𝐸𝑧
(− 𝑐 𝐵𝑦 −𝐵𝑥 0
)
and that with two lower indices 𝐹μν by lowering the index μ in (72) using the metric tensor ημν as follows

Page 20 of 22
Lecture notes PH4508

𝐸𝑥 𝐸𝑦 𝐸𝑧
0 − − −
𝑐 𝑐 𝑐
𝐸𝑥
0 𝐵𝑧 −𝐵𝑦
𝐹𝜇𝜈 ≡ 𝜂𝜇𝜌 𝐹𝜌 𝜈 = 𝑐 (75)
𝐸𝑦
−𝐵𝑧 0 𝐵𝑥
𝑐
𝐸𝑧
𝐵𝑦 −𝐵𝑥 0
(𝑐 )
The quantities 𝐹 μν and 𝐹μν are tensors of type (2,0) and (0,2) respectively. So, for example, under the
Lorentz transformations (50), 𝐹 μν transforms as
′𝜈′ ′ ′
𝐹𝜇 = 𝛬𝜇 𝜇 𝛬𝜈 𝜈 𝐹𝜇𝜈 , (76)

where dependence on the arguments (which are spacetime coordinates) have been suppressed. This
transformation law determines how the electric field 𝐸⃗⃗ and magnetic field 𝐵 ⃗⃗ transform under Lorentz
transformations. When we say electromagnetic field tensor, we could mean any of the three quantities
𝐹μν , 𝐹 μν and 𝐹 μ ν , since these three quantities carry the same and full amount of information about
electromagnetic field.

Exercise 14. Is the electric field 𝐸⃗⃗ a contravariant vector in Minkowski spacetime? (Answer: No.) Is the
magnetic field 𝐵⃗⃗ a contravariant vector in Minkowski spacetime? (Answer: No.) What is the meaning of
the word “vector” when your lecturer always said “electric field vector” in your year one and year two EM
courses? (If you do not know the answer, go and ask your EM lecturer(s).)

Exercise 15. Show that under a Lorentz boost as in the standard configuration, electromagnetic fields
transform in the following way:

(𝐸𝑥′ , 𝐸𝑦′ , 𝐸𝑧′ ) = (𝐸𝑥 , 𝛾(𝐸𝑦 − 𝑢𝐵𝑧 ), 𝛾(𝐸𝑧 + 𝑢𝐵𝑦 )) , (77)

(𝐵𝑥′ , 𝐵𝑦′ , 𝐵𝑧′ ) = (𝐵𝑥 , 𝛾(𝐵𝑦 + 𝑢𝐸𝑧 /𝑐 2 ), 𝛾(𝐵𝑧 − 𝑢𝐸𝑦 /𝑐 2 )) . (78)

Exercise 16. Among the tensors 𝐹 μν, 𝐹μν and 𝐹 μ ν, which ones are antisymmetric?

Maxwell equations

You have learned from your EM courses that the dynamics of the electric and magnetic fields themselves
is governed by Maxwell equations as follows:
𝜌
𝛻 ⋅ 𝐸⃗⃗ = (79)
𝜖0
⃗⃗
𝜕𝐵
𝛻 × 𝐸⃗⃗ = − , (80)
𝜕𝑡
⃗⃗ = 0,
𝛻⋅𝐵 (81)

𝜕𝐸⃗⃗
⃗⃗ = 𝜇0 (𝑗⃗ + 𝜖0
𝛻×𝐵 ). (82)
𝜕𝑡

Page 21 of 22
Lecture notes PH4508

Maxwell equations in this form are not manifestly covariant. It is not immediately clear how they
transform under Lorentz transformations. Do Maxwell equations obey the relativity principle of Einstein?

It can be shown that, in terms of the electromagnetic field tensor, Maxwell equations can be written in
the following form:
𝜕𝛼 𝐹𝛽𝛾 + 𝜕𝛽 𝐹𝛾𝛼 + 𝜕𝛾 𝐹𝛼𝛽 = 0, (83)

𝜕𝛽 𝐹 𝛼𝛽 = 𝜇0 𝑗 𝛼 . (84)

Here, the quantity 𝑗 μ is called the current density four-vector and is given by

𝑗𝜇 = (𝑐𝜌, 𝑗𝑥 , 𝑗𝑦 , 𝑗𝑧 ). (85)

The time component of 𝑗 μ is essentially the charge density, while the spatial components form the current
density 𝑗⃗. Both sides of (83) and (84) are tensors. Maxwell equations in this form are manifestly covariant.
Electromagnetism, formulated as (71), (83) and (84), obeys Einstein's principle of relativity.

Exercise 17. How does electric charge density ρ transform under a Lorentz boost? Is it a scalar? Express
the charge conservation law:
𝜕𝜌
+ 𝛻 ⋅ 𝑗⃗ = 0, (86)
𝜕𝑡
in terms of the current density four-vector.

Exercise 18. Show that the covariant form of Maxwell equations (83) and (84) produces the familiar
equations (79)-(82).

Summary: The spacetime structure underlying special relativity is Minkowski spacetime. Minkowski
spacetime has a metric ημν = diag(−1,1,1,1) in inertial coordinate systems, from which geometrical
notions such as straight lines, length and angle can be defined. Basic quantities in Minkowski
spacetime are scalars, four-vectors and their generalisations—tensors. Laws of physics in special
relativity can be cast in a manifestly covariant form using tensors. A tensor equation is true in any
inertial coordinate system if it is true in one particular inertial coordinate system. Maxwell equations
and Lorentz force law can be written as tensor equations using the electromagnetic field tensor 𝐹 μν.

Page 22 of 22