Scott A.

Hughes Introduction to relativity and spacetime physics

Massachusetts Institute of Technology
Department of Physics
8.033 Fall 2021

Lecture 11
A covariant formulation of electromagnetics (part I)

11.1 Electric and magnetic fields and forces: Background

Our pivot from Galileo’s relativity to Einstein’s relativity began by considering electrody-
namics. Let’s write out again the critical equations which govern electrodynamics — the
Maxwell equations which connect the the fields to their sources, and the Lorentz force law
which shows how these fields act on charges:

∇ · E = ρ/0 , ∇·B=0, (11.1)

∂B ∂E
∇×E=− , ∇ × B = µ0 J + µ0 0 ; (11.2)
∂t ∂t
F = q (E + v × B) . (11.3)
It should be emphasized very strongly that these equations are fully compatible with special
relativity. Indeed, all of the modifications to various physical concepts that Einstein’s rela-
tivity requires were introduced because it became clear that important aspects of Newtonian
mechanics were not compatible with electrodynamics. Electrodynamics is one of the most
successfully and accurate theories of nature we have developed. Once it has been updated
to account for the fact that our universe is quantum mechanical in nature (a topic for a
different course!), we end up with a version of electrodynamics that is perhaps humanity’s
most precisely-tested description of nature.
That said, Eqs. (11.1), (11.2), and (11.3) are not written in a way that makes it clear
they are compatible with Lorentz covariance. The fields and the force are written using
3-vectors, which depend upon us choosing a particular observer’s “space” coordinates; the
field equations are expressed using a particular observer’s time and space derivatives. These
equations are formulated for one particular reference frame, and it is not obvious how they
will transform to another reference frame. The goal of the next two lectures is to think how
to organize the structures expressed in Eqs. (11.1), (11.2), and (11.3) in a way that clearly
shows electrodynamics is a Lorentz covariant theory.

11.2 How to organize the fields

11.2.1 General considerations
So far, when we’ve translated a physical quantity into Lorentz covariant language, we have
found a way of taking quantities which are 3-vectors and mapping them into 4-vectors.
Examples so far are displacement (add ct as the “zeroth” component), the 4-velocity (change
d/dt to d/dτ so that we use a clock whose meaning is invariant to describe time derivatives;
add c dt/dτ = γc as the zeroth component), and the 4-momentum (add energy as the zeroth
component, dividing by c to make sure the dimensions are sensible). Can we do this with
the electric and magnetic fields?

81
 We have several problems here. First, we know that E and B fields must transform into
one another when we change frames: what is pure magnetic field in one frame is a mixture
of magnetic and electric fields in another; and vice versa. The classic example of this is a
charge moving in a magnetic field. Consider a charge moving parallel to a current-carrying
wire, as illustrated in Fig. 1:

Charge q moves with velocity v = vex

Separation r

Wire carries current I

Figure 1: A charge q moving parallel to a wire carrying a current I.

For concreteness, let’s define ex as pointing to the right, ey as pointing into the page, and ez
as pointing up. Then, in what we will call the “lab” frame L, we have a charge q that moves
to the right. The charge is a distance r from a wire that carries a current flowing to the left.
As we learned in 8.02/8.022, this wire generates a magnetic field that circulates around the
wire. At the location of the charge, this field takes the value
µ0 I
B= ey . (11.4)
2πr
The wire is neutral, so the charge q does not feel any electric force — it only feels a magnetic
force, whose value is
µ0 qIv
F = qv × B = ez . (11.5)
2πr
This force points “up” in the figure — the charge is repelled from the wire.
Let’s now change frames, and think about what must happen. First, we require q to
be the same in all reference frames. If changing frames changed the value of charge, the
elementary charge would vary for moving charges. Imagine the effect this would cause for a
system in which there are members whose charges are equal and opposite, but are moving at
different relative speeds. A system which is neutral when its members “sit still” might have
net charge when they are in motion! In addition to feeling absurd, the fact is that we have
no experimental evidence for anything like this whatsoever. Observations and measurements
indicate that a body’s charge is unchanged no matter how fast we observe it to move.
So, let’s jump into a reference frame that moves with v = vex — i.e., the frame C in
which the charge is at rest. In this frame there can be no magnetic force. The magnetic
force is proportional to the charge’s speed. If the speed is zero, the magnetic force is zero.
However, a repulsive force in one frame of reference is not consistent with no force in another.

82
The details of how the force behaves in this frame might differ1 (perhaps its magnitude will
be different), but there still must be an overall repulsive force. If there is no magnetic force,
then there must instead be an electric force.
This means there must be an electric field in the charge’s rest frame, even though there
was no such field in the lab frame. Something that we measured to be pure magnetic field
transforms to a mixture of electric and magnetic field. Whatever “entity” we will use to
describe electric and magnetic fields in special relativity must be able to transform magnetic
fields into electric fields (and vice versa).

11.2.2 A covariant representation of the force and fields

Our root issue is essentially one of simple counting. We have had success fitting important
physical quantities into 4-vectors so far, it just isn’t going to work for the electric and
magnetic field. They have 6 components, and we just cannot fit these 6 pieces of information
into the 4 components of a 4-vector. We need something bigger.
Let’s consider a “larger” geometric object. A 2nd-rank tensor has 16 components. That’s
too many; but, we can reduce the number of free components by imposing symmetry. If
we use a symmetric tensor, then it has 10 free components — still too many. But an
antisymmetric 2nd-rank tensor has 6 free components — exactly what we need.
So let’s think how we can fit the 6 components (E x , E y , E z ), (B x , B y , B z ) into an
antisymmetric 2nd-rank tensor which we will call F αβ . To guide us, let’s deduce how the
Lorentz force law, F = q(E + v × B), can be written in a fully covariant manner.
First, we “upgrade” the force. We start with F = dp/dt. Clearly, we will want to take
the 3-momentum p over to the 4-momentum, whose components are pα . We also need to
upgrade the time derivative with one that uses a notion of time that all frames can reference.
Just as we did in defining the 4-velocity, let’s replace d/dt with d/dτ , where τ is the proper
time measured by the body which is experiencing the force.
What about the right-hand side, q(E + v × B)? This a quantity that is linear in q,
linear in the fields, and — if we think about this carefully — linear in the components of the
velocity. “Wait a minute,” I imagine you protesting, “the B term is linear in components of
velocity, but what about the E term?” Well, note that E and B have different dimensions:
the dimensions of E are force over charge, but the dimensions of B are force over speed times
charge. When we assemble these quantities into a single tensor, we’ll need to account for
the difference in dimensions. We often do this by throwing in factors of the speed of light.
This suggests that we think about the Lorentz force law as
   
E
F=q c +v×B . (11.6)
c

Bearing in mind that the components of uα are given approximately by (c, v x , v y , v z ) for a
body that is not moving very fast relative to us, this suggests that in the Lorentz force law,
the electric field is being multiplied by the timelike component of the 4-momentum.
Putting all this together, we want the covariant formulation of the Lorentz force to be
dpα
= qF αβ uβ . (11.7)
dτ
1
In a few lectures we will look carefully at forces and accelerations in special relativity; we briefly discuss
a handful of important issues a little later in this lecture.

83
Let’s now figure out how to “fill up” the tensor F αβ so that this is consistent with the Lorentz
force law that we learned about in 8.02/8.022 by going through the spatial components,
α = 1, 2, 3, one by one. (We’ll come back to the α = 0 component later.) First look at
α = 1, or α = x:
dpx
= q F 10 u0 + F 12 u2 + F 13 u3 .


(11.8)
dτ
There is no F 11 term because of this tensor’s antisymmetry – all diagonal elements are zero.
Let’s further use the fact that u0 = −u0 = −γc, u2 = γ(dy/dt), and u3 = γ(dz/dt):
dpx
 
10 12 dy 13 dz
= γq −cF + F +F . (11.9)
dτ dt dt
Next, using the fact that an interval of time dt measured by clocks in this frame is γdτ ,
dpx
 
10 12 dy 13 dz
= q −cF + F +F . (11.10)
dt dt dt
Compare this to the x component of the Lorentz force law:
dpx
 
x z dy y dz
=q E +B −B . (11.11)
dt dt dt
This allows us to read off

F 10 = −E x /c , F 12 = B z , F 13 = −B y . (11.12)

Repeating this exercise for the y and z force components and noting that the tensor is
antisymmetric allows us to fill it in entirely:
 
0 E x /c E y /c E z /c
. −E x /c 0 B z −B y 
F αβ =  y z
 . (11.13)
−E /c −B 0 Bx 
−E z /c B y −B x 0
This tensor is often called the Faraday tensor. It replaces the 3-vectors which describe electric
and magnetic fields according to some particular observer’s reference frame with a geometric
object whose components can be readily translated to any reference frame; and, it connects
to 4-vectors whose components can likewise be readily translated to any reference frame.

11.3 A brief aside on forces and accelerations

In this lecture, we’ve been talking about a specific force without yet having discussed forces in
special relativity in broader terms. We will discuss forces, accelerations, and the properties
of accelerated observers in more detail in an upcoming lecture. Certain aspects of this
discussion are needed now, so we pause in our discussion of electric and magnetic fields for
a brief digression to talk about forces and accelerations.
As we have discussed, a body of mass m moving with 4-velocity ~u has a 4-momentum
p~ = m~u. As you have seen in our discussion above, this momentum changes if the body is
acted on by a force or, more properly, a 4-force:
d~p
F~ = . (11.14)
dτ
84
If the body’s mass cannot change, then this leads to the body having a 4-acceleration:
1 ~ d~u
~a = F = . (11.15)
m dτ
When we discuss 3-velocities u and 3-accelerations a, these quantities can have largely any
value that we want them to have: the value of u is essentially an initial condition to our
analysis, and the value of a is only constrained by the mechanism providing the force F.
Not so for the 4-velocity and the 4-acceleration: there is a very interesting and important
constraint which these two quantities must always satisfy. To see where this comes from,
begin with the invariant that we can construct from ~u:
~u · ~u = −c2 . (11.16)
Take d/dτ of both sides of this equation:
~a · ~u + ~u · ~a = 0 , (11.17)
or
~a · ~u = 0 . (11.18)
The 4-velocity and the 4-acceleration are always “orthogonal” in spacetime. This important
constraint has important implications for the nature of any 4-force that you may compute —
if at the end of your analysis, you find that F~ · ~u 6= 0, you’ve made a mistake or overlooked
something important.

11.4 Some details of the electromagnetic 4-force

With the above discussion in mind, let’s examine the electromagnetic 4-force that we have
worked out. Is it the case that F~ · ~u = 0? The answer is yes, and we can show this using a
little bit of “index gymnastics”:
F~ · ~u = qF αβ uβ uα (11.19)
= −qF βα uβ uα (11.20)
βα
= −qF uα uβ (11.21)
= −qF αβ uβ uα . (11.22)
Let’s step through these lines of analysis carefully. On the first line, we have have contracted
the definition of the electromagnetic 4-force, Eq. (11.7), with the 4-velocity in order to make
the inner product. On the second line, we have used the fact that the Faraday tensor is
antisymmetric to swap the order of the indices on the tensor, introducing a minus sign. On
the third line, we have used the fact that uα uβ is symmetric to swap the order of their
indices. On the final line, we have used the fact that α and β are “dummy” indices — they
are being summed over, so it doesn’t matter how we label them. We can in fact change α
for β and β for α, as long as we do this consistently throughout the expression.
Now compare the first line with the fourth line. Their right-hand sides are identical ...
except for a minus sign. This is thus an expression of the form x = −x, whose only solution
is x = 0. We conclude that
F~ · ~u = 0 , (11.23)
So our 4-force indeed is spacetime orthogonal with the 4-velocity — as it should be.
Two remarks on this calculation:

85
 • This is our first encounter with a trick that gets used a lot: whenever you contract all
free indices of a totally antisymmetric mathematical object, like F αβ , against a totally
symmetric mathematical object, like uα uβ , the result is zero.
If this makes you nervous and you want to be totally confident in the result, you can
always go through an exercise like the one that I did above. But, the key point is
that by combining symmetric with antisymmetric, we add up terms that are equal and
opposite. If you expand out the Einstein summation that I did above, you find that
you can combine terms in pairs: F 10 u1 u0 + F 01 u0 u1 , F 23 u2 u3 + F 32 u3 u2 , etc. The
members of the pair are always going to be equal in magnitude and opposite in sign.

• When a force law is set up properly, it generally works out “automatically” that we find
F~ ·~u = 0, in much the way that it did for this electromagnetic 4-force. Finding F~ ·~u = 0
does not guarantee that your force law is correct, but not finding this guarantees that
your force law is wrong.

Before moving on to other aspects of the covariant formulation of electric and magnetic
fields, let’s clean up one last detail. We saw in our calculation above that the α = 1, 2, and 3
components of the 4-force correspond perfectly to the x, y, and z components of the Lorentz
force. What is the α = 0 component? Let’s write this out:

dp0
= qF 0β uβ = q F 01 u1 + F 02 u2 + F 03 u3


dτ
γq x x
= (E (u) + E y (u)y + E z (u)z )
c
γq
= E·u. (11.24)
c
Using the fact that p0 = E/c, where E with no indices and no boldface means the energy of
the charged body, and using dt = γdτ , this becomes
dE
= qE · u . (11.25)
dt
This expression tells us about the rate at which work is done on the charge by the electric
field. If you need a reminder of where this comes from, remember that the differential of
work done in moving a displacement dr in an E field is

dW = F · dr = qE · dr . (11.26)

If the charge moves through this interval in a time dt, then

dW dr
= qE · , (11.27)
dt dt
in agreement with Eq. (11.25).

11.5 Transforming electric and magnetic fields

By fitting the electric and magnetic fields into a rank-2 tensor, it becomes simple to deduce
how these fields transform when we change frames. Let observer O measure fields described

86
by the tensor F αβ ; let O0 in a different inertial frame measure fields described by the tensor
0 0
F µ ν . These are related by converting using Lorentz transformation matrices:
0 0 0 0
F µ ν = F αβ Λµ α Λν β . (11.28)

Let’s work through this using the Lorentz transformation matrix

 
γ −γβ 0 0
0 −γβ γ 0 0
Λµ α = 
 0
 . (11.29)
0 1 0
0 0 0 1

In other words, we take O0 to be moving with v = vex relative to O. Let’s use this calculate
how the components of the Faraday tensor translate between frames. Start by working
through the transformation for the (00 10 ) component:
0 0 0 0 0 0
F 0 1 = Λ0 0 Λ1 1 F 01 + Λ0 1 Λ1 0 F 10
= γ 2 − γ 2 β 2 F 01

= F 01 . (11.30)

On the first line, we expanded the transformation rule to write out all the non-zero terms
0 0
that contribute to F 0 1 . This amounts to all the lambda matrix elements that have 00 on
the first index, and all the matrix elements that have 10 on the first index. A total of 4 such
0 0 0 0
elements exist: Λ0 0 = γ, Λ0 1 = −γβ, Λ1 0 = −γβ, and Λ1 1 = γ; all the others ones with 00
or 10 inpthe first position are zero. We then used antisymmetry, and then used the fact that
γ = 1/ 1 − β 2 to clean this expression up. Translating back into electric and magnetic field
components, this tells us
0
Ex = Ex . (11.31)
Move on to the (00 20 ) component:
0 0 0 0 0 0
F 0 2 = Λ0 0 Λ2 2 F 02 + Λ0 1 Λ2 2 F 12
= γF 02 − γβF 12 . (11.32)

We cannot simplify this any further, so we now translate back into electric and magnetic
field components:
0
E y = γ(E y − vB z ) . (11.33)
Next the (00 30 ) component:
0 0 0 0 0 0
F 0 3 = Λ0 0 Λ3 3 F 03 + Λ0 1 Λ3 3 F 13
= γF 03 − γβF 13 . (11.34)

This becomes
0
E z = γ(E z + vB y ) . (11.35)
Doing a similar exercise for the components of the Faraday tensor which map to the
magnetic fields, we find
0 0 0
Bx = Bx , B y = γ(B y + vE z /c2 ) , B z = γ(B z − vE y /c2 ) . (11.36)

87
 By repeating this analysis for frames moving with v = vey and v = vez , it’s not too
difficult to work out the general rule for transforming between frames. For completely general
v, we have
E0k = Ek , E0⊥ = γ (E⊥ + v × B⊥ ) ; (11.37)
B0k = Bk , B0⊥ = γ B⊥ − v × E⊥ /c2 .


(11.38)
Here, Ek denotes the component of E that is parallel v. Let ev ≡ v/v denote the unit vector
along the velocity vector; then, Ek = (E·ev )ev . The other component, E⊥ = E−Ek , denotes
the part of E that is orthogonal to v. The magnetic field vectors Bk and B⊥ are defined
likewise.

88