Lecture 3

The Dirac field

Having studied the KG field for spinless particles, in this Lecture we present the Dirac
field for spin 1/2 particles. Particles with spin 1/2 –like the electron– are fermions while
spinless particles are bosons. There are similarities and di↵erences between bosons and
fermions. One similarity is that we also have creation and annihilation operators. A
crucial di↵erence is the following. Call a†s (k) the creation operator of an electron with
3-momentum k and spin s. A two-electron state is obtained as

a†s (k)a†s0 (k0 ) |0i = |k, s; k0 , s0 i . (3.1)

When reversing the order of the creation operators we obtain

a†s0 (k0 )a†s (k) |0i = |k0 , s0 ; k, si . (3.2)

Spin 1/2 particles obey Fermi statistics, which tells us that there is a minus sign when
interchanging the states of the electrons,

|k0 , s0 ; k, si = |k, s; k0 , s0 i . (3.3)

Remember what we discussed regarding Equation (2.49): because boson creation oper-
ators commute, Bose statistics applies. In the case of fermions we need creation operators
to anticommute. In this case, interchanging the two operators will lead to a minus sign,
and Equation (3.3) will be valid. In Dirac field theory we assume anticommutation rela-
tions.
As stated earlier, the Dirac equation may be useful in the NR limit as a single-particle
equation. For example, once one has solved the Hydrogen atom using the Schroedinger
equation, one can improve the results using the Dirac equation, whereby the relativis-
tic corrections to wave-functions and spectrum are obtained1 . Another example is the
following. Define the gyromagnetic ratio for the electron as the ratio of its magnetic
dipole moment to its spin, and express it as ge/(2me ). The Dirac equation, extended to
incorporate a magnetic field, leads to the remarkable prediction g = 2, very close to the
experimentally measured value2 .
However, a fully consistent theory of electrons, positrons and photons needs to adopt
the quantum field description.
1
See for example Itzykson and Zuber.
2
The observed tiny deviations from g = 2 can be calculated in quantum electrodynamics. The theo-
retical prediction and the observed value coincide at a high level of precision.

43
3.1 Dirac equation
In the same way that the KG equation is the di↵erential equation satisfied by a scalar
field (x), we shall introduce the Dirac equation for a spin 1/2 field (x). Like the KG
equation, the Dirac equation has to be interpreted as a field equation and not as an
equation for a wave function describing a single particle. Another similarity is that the
Dirac equation also describes free particles, i.e., particles with no interactions.
There are important di↵erences between the KG and Dirac equations:
1. The Dirac equation is linear in the time derivative while KG is second order in
time3 .
2. While a scalar field is described by a single-component complex field (one dof),
we expect the field describing a spin 1/2 fermion, like the electron, to be more
complicated. This is because spin 1/2 has two dofs, and therefore we expect the
Dirac field to have at least two components. Actually, we will later show that the
Dirac field (x) has four components. For this reason, (x) is called a 4-spinor or
a Dirac spinor.
We need at least two components but, as we shall see, we end up having four com-
ponents. The reader may wonder about the meaning of the extra two components.
We will see that the additional components correspond to the positron, which is the
antiparticle of the electron.
The Dirac equation is given by
µ
(i @µ m) (x) = 0 . (3.4)
µ
Here are four 4 ⇥ 4 square matrices:
0 1 2 3
, , , and , (3.5)
The elements of these matrices are constant c-numbers. They are called Dirac matrices.
We will see that m in (3.4) is the mass of the Dirac fermion. If we are describing electrons,
m is the electron mass4 . As we have electrons in mind, we sometimes talk about electrons,
but we should bear in mind that the Dirac equation is valid for any spin 1/2 particle.
The field (x) is a column vector with four complex entries, with each entry depending
on the space-time x. The Dirac equation is a matrix equation, with the structure
0 1 01 0 1
. . . . . 0
B. . . .C B.C B0C
B C BC B C
@. . . .A @.A = @0A , (3.6)
. . . . . 0
where the dots in the 4⇥4 matrix contain c-numbers and derivatives, while the dots in the
column vector correspond to the values for the components of the Dirac spinor, functions
of the space-time point xµ . To help us to get used to the new notation, let us write the
Dirac equation in components:
µ
(i ↵ @µ m ↵ ) (x) = 0 ↵, = 1, 2, 3, 4 . (3.7)
3
The original reason for Dirac writing an equation linear in the time-derivative was to get rid of
negative energy solutions.
4
Then m is also the positron mass; the positron is the antiparticle of the electron and a particle and
the corresponding antiparticle have the same mass

44
Here, the indices ↵, are called Dirac indices and run from 1 to 4 in the Dirac space, and
↵ is the Kronecker symbol. We see that the Dirac equation is a set of four equations,
one for each value of ↵. Also, there is a sum over the Dirac index in the equation.
In most cases we will not write the Dirac indices since there should not be any confusion
or ambiguity. For example, when writing the Dirac equation (3.4) in compact form, we
have simplified the notation. The c-number m in the equation is in fact multiplied by the
4 ⇥ 4 identity matrix. Should we write m14⇥4 in the equation? No we should not. We
follow common notation and we should understand that there is a matrix there, otherwise
the equation will not make sense. Also, the zero in the rhs of (3.4) is a column vector
with four zeros, but we will not make it explicit, because it can be understood from the
context. This simplified notation will be employed all the time; we should only write
something more explicit when necessary, for example to avoid ambiguity.
Let us introduce the widely used ”slash” notation:

/⌘ µ
A Aµ for any 4-vector Aµ ,
µ
@/ ⌘ @µ . (3.8)

Using this notation, the Dirac equation can simply be written as

(i@/ m) (x) = 0 . (3.9)

Let us end this Section by clarifying that, despite carrying a Lorentz index, the four ’s
are not the components of a 4-vector. They are constant matrices and do not transform
under a Lorentz transformation. We should consider the index µ to be a label. Later, we
will find out how to construct a 4-vector using the four ’s, but the ’s themselves are
not a 4-vector.

3.2 Some properties of matrices

In the following Section, we shall prove that N = 4. However, in this Section we assume
that the four matrices appearing in the Dirac equation are general N ⇥ N matrices and
we shall deduce several general properties of -matrices.
First, multiply (3.4) by the operator ( i ⌫ @⌫ m) to get
⌫ µ
( @µ @⌫ + m2 ) (x) = 0 . (3.10)

Now, / exp (±ikx) should be a solution of this equation corresponding to a free particle
of mass m and 4-momentum k. Since @µ / kµ exp (±ikx), we have
⌫ µ
( kµ k⌫ + m2 ) exp (±ikx) = 0 . (3.11)

This implies
⌫ µ
kµ k⌫ + m2 = 0 . (3.12)
Let us now symmetrize with respect to µ and ⌫,
1 ⌫ µ µ ⌫
( + ) k µ k⌫ + m2 = 0 . (3.13)
2
To have k 2 = m2 we need
1 ⌫ µ µ ⌫
( + ) = ⌘ µ⌫ . (3.14)
2

45
 We thus reach the fundamental property of -matrices
µ ⌫
{ , } = 2 ⌘ µ⌫ , (3.15)

where we have defined the anticommutator,

{A, B} = AB + BA . (3.16)

As stated earlier, we should note that we simplify notation all the time. For example,
since the lhs of (3.15) is a matrix, the rhs is a matrix too: 2⌘ µ⌫ is in fact multiplied by
the N ⇥ N identity matrix.
Let us examine some consequences of Equation (3.15) and some general properties of
N ⇥ N Dirac matrices.

1.
{ 0, 0
} = 2 ) ( 0 )2 = 1 . (3.17)

2.
{ i, i
}= 2 ) ( i )2 = 1. (3.18)
We follow the common practice of using a Latin index for i = 1, 2, 3.

3.
{ 0, i
}=0 ) 0 i
= i 0
, (3.19)
0 i
i.e. and anticommute.

4. The previous equation can be written as

0 i 0 i
= , (3.20)
0
which allows us to calculate the trace of :
0 i 0 i i i 0 0 0
Tr = Tr = Tr = Tr ) Tr =0. (3.21)

i
Actually, all four -matrices are traceless, the proof that Tr = 0 is left as an
exercise.

5. In Exercise 4 we propose to show that requiring the Hamiltonian to be Hermitian

implies two properties
0 † j † 0 †
H = H † =) 0
= , 0 j
= . (3.22)

0 j †
Using the first property in the second property, we have = ( j) 0
. Then the
two properties can be written as a single equation:
µ † 0 µ 0
( ) = . (3.23)

6. It is useful to define the matrices with the index down

⌫
µ = ⌘µ⌫ . (3.24)

We stress that are not treating the ’s as 4-vectors, (3.24) has to be taken as a
definition.

46
 7. A useful combination of -matrices is the so-called 5 -matrix,

5 0 1 2 3
5 = =i , (3.25)

Note that we do not distinguish between the index 5 up or down; the index 5 can
be placed either way without consequence. The 5 -matrix can written as

5 i µ ⌫ ↵
5 = = ✏µ⌫↵ , (3.26)
4!
where we use the convention ✏0123 = 1 for the Levi-Civita symbol. Note that we do
not distinguish between the index 5
It is not difficult to deduce the following useful properties of the 5 -matrix,

( 5 )2 = 1 ,
†
5= 5,
⌫
{ 5, }=0. (3.27)

8. The commutator of -matrices is defined as

µ⌫ i µ ⌫
= [ , ]. (3.28)
2
In spite of having two indices, µ⌫ is not a Lorentz tensor. Just as we have defined
↵
µ in (3.24) we define µ⌫ = ⌘µ↵ ⌘⌫ .

More properties of the -matrices can be found in Appendix 3.12.

3.3 Representations of -matrices

In this Section we are going to find that the dimension of -matrices is N = 4.
First of all, equation (3.17) tells us that 0 has eigenvalues ±1 and since the trace of
0
vanishes as (3.21) tells us, we may conclude that N has to be even.
The value N = 2 does not work. One way of seeing this is by taking the three
matrices as three ’s, and since a fourth 2 ⇥ 2 matrix can be written as a linear
combination of the three ’s and the identity 12⇥2 , this fourth matrix cannot obey the
required properties.
Instead the value N = 4 is fine. A clear way to convince ourselves is by showing
a set of four 4 ⇥ 4 matrices that obey the anticommutation relations in (3.15) and the
hermiticity conditions in (3.23):
✓ ◆ ✓ i
◆
0 1 0 i 0
= , = i Dirac repr. (3.29)
0 1 0

Here i are the three Pauli matrices5 , and 1 = 12⇥2 , 0 = 02⇥2 . We call the choice (3.29)
the Dirac (or Dirac-Pauli) representation of the -matrices.
In the Dirac representation we have
✓ ◆
0 1
5 = ,
1 0
5
Some properties of Pauli matrices are shown in Appendix 3.11.

47
 ✓ i
◆ ✓ k
◆
0i 0 ij ijk 0
=i i , =✏ k , (3.30)
0 0
where we sum over repeated indices, and ✏123 = 1.
There are other representations for the gamma-matrices. The Weyl (or chiral) repre-
sentation is given by
✓ ◆ ✓ i
◆
0 0 1 i 0
= , = i Weyl repr. (3.31)
1 0 0

In this representation, ✓ ◆
1 0
5 = , (3.32)
0 1
✓ i
◆ ✓ k
◆
0i 0 ij ijk 0
=i i , =✏ k . (3.33)
0 0
Although we can work with any representation, but sometimes one representation is
more convenient than the other. While the Dirac representation in general is useful in the
NR limit, the Weyl (or chiral) representation is useful in the extreme relativistic case.
We have shown two gamma- representations but actually there is an infinite number
of them. Once we have a representation of four µ ’s we can define another representation
of four ˜ µ ’s
˜µ = U µU † , (3.34)
where U is a unitary matrix. In Problem 2 we propose a) to prove that the new ˜ µ ’s obey
the anticommutation relations in (3.15) and the hermiticity conditions in (3.23), and b)
to show how the Dirac equation is written with the new gamma-matrices.
We shall adopt the value N = 4. The main reason to stop at N = 4 and not consider
higher values is that N = 4 leads to the correct physical content of the theory. The
value N = 4 can be interpreted physically as the number of dofs necessary to describe
the electron and the positron, each of them being a spin 1/2 particle with two dofs. In
conclusion, we shall take N = 4 from now on, therefore the four -matrices in the Dirac
equation are 4 ⇥ 4 matrices.

3.4 Dirac Lagrangian

The Lagrangian leading to the Dirac equation is given by
µ
L = i (x) @µ (x) m (x) (x)
µ
= (x) (i @µ m) (x) , (3.35)

where we have defined the adjoint Dirac field ,

† 0
(x) = (x) . (3.36)

The adjoint 4-spinor is a row vector in Dirac space, while is a column vector. The
Lagrangian describes free particles. To have interactions we need more terms in the total
Lagrangian.

48
 µ
As stated earlier we use the slash symbol. With @/ = @µ , the Dirac Lagrangian is
written in a somewhat compact form:

L = (x)(i@/ m) (x) . (3.37)

With Dirac indices ↵ and explicitly written, the Lagrangian reads

µ
L= /↵
↵ (i@ m1↵ ) with @/↵ = ↵ @µ . (3.38)

Let us check that the Euler-Lagrange equation is indeed the Dirac equation. To do it,
it is better to use L written with explicit Dirac indices. We vary (3.38) with respect to
(x)

@L
= (i@/↵ m1↵ ) ,
@ ↵
@L
=0, (3.39)
@[@µ ↵ ]

Indeed, we see that the corresponding Euler-Lagrange equation,

(i@/↵ m1↵ ) =0, (3.40)

is the Dirac equation (3.4).

Variation with respect to in (3.37),

@L
= m ↵ ,
@ ↵
@L µ
=i ↵ , (3.41)
@[@µ ↵ ]

leads to the equation

µ
i @µ +m =0 . (3.42)
This is the so-called adjoint Dirac equation. It has no more physical content than (3.4).
Indeed, taking the Hermitian conjugate of (3.4) and after some simple manipulations we
get (3.42). (See Exercise 6).
Actually, the Lagrangian

L1 = (x)( i @/ m) (x) , (3.43)

where the arrow in the derivatives means that they act on the left -on -, is equivalent to
(3.37) because they di↵er in a total derivative. (See Exercise 5).
The Lagrangians (3.37) and (3.43) are not Hermitian. We can combine them and
obtain an Hermitian Lagrangian
1 µ
$
L2 = i @µ m , (3.44)
2
$
where the derivative now acts on both the right and left sides: A@µ B = A(@µ B) (@µ A)B.
Let us finally stress that we usually take (3.35) as the Dirac Lagrangian.

49
3.5 Plane-wave solutions
The Dirac equation describes a free fermion and therefore there should be plane-wave
solutions,
(x) / ur (p) e ipx , r = 1, 2 ,

(x) / vr (p) eipx , r = 1, 2 . (3.45)

Here u1 , u2 , v1 , v2 are 4-spinors in momentum space. There are four types of plane-
wave solution, as expected since has four components. The 4-spinor u corresponds to
positive frequency, and the 4-spinor v to negative frequency. The argument of u and v is
the 3-momentum p because the energy p0 is not independent,
p
p0 = E p = p 2 + m 2 . (3.46)
As we said, (3.45) have to be solutions to the Dirac equation. Let us introduce them in
the Dirac equation so that we get equations for ur and vr . The exact form of the equation
depends on whether we have a spinor of u-type or v-type
( µ pµ m) ur (p) = 0 ,
( µ pµ + m) vr (p) = 0 . (3.47)
µ
With the notation p/ = pµ we can rewrite these equations in the form

p/ m ur (p) = 0 ,
p/ + m vr (p) = 0 . (3.48)
The 4-spinors u, v are solutions to the Dirac equation in momentum space.
We also have the corresponding adjoint equations,
ur (p) p/ m = 0 ,
v r (p) p/ + m = 0 , (3.49)
where the adjoints spinors are defined as
ur = u†r 0
, v r = vr† 0
. (3.50)
We use the normalization
u†r (p)us (p) = vr† (p)vs (p) = 2Ep rs ,
u†r (p)vs ( p) = vr† (p)us ( p) = 0 , (3.51)
Let us finally expand the Dirac field in Fourier modes
Z X⇥ ⇤
(x) = (dp) cr (p) ur (p) e ipx + d†r (p) vr (p) eipx
r=1,2
+
= (x) + (x) . (3.52)
From this expression we can find the expansion of the adjoint field
Z X⇥ ⇤
(x) = (dp) dr (p) v r (p) e ipx + c†r (p) ur (p) eipx
r=1,2
+
= (x) + (x) . (3.53)

50
Remember the notation
d3 p
(dp) = p , (3.54)
(2⇡)3 2Ep
and that we indicate the positive or negative frequency part withp a + (plus) or - (minus)
superscript. Also, in the exponentials the value of p0 = Ep = p2 + m2 .
There is a physical property that di↵erentiates u1 form u2 and v1 form v2 . This
property can be chosen to be spin or to be helicity.
At this point, it is useful to remind about some basic formulas fot spin and helicity in
QM, when we have a spin-1/2 particle like the electron. For spin, the simple thing to do
is to go to the rest-frame of the particle, to define an axis, and to measure the spin along
this axis. We know the measurement can have two possible values, +1/2 and 1/2 (in ~
units), which we call spin up or spin down. The spin observable is
1
(3.55)
2
The Pauli spinors that are eigenstates of 3 are
✓ ◆ ✓ ◆
1 0
+ = and = (3.56)
0 1
We obtain the observable helicity, in a frame where the particle is moving with 3-
momentum p, when we choose the axis in the p direction. There two possible values of
the measurement, which we now call positive and negative helicity. The helicity observable
is given by
· p̂ , (3.57)
where p̂ = p/|p|, so that |p̂| = 1. The two possible eigenvalues are h = +1 and h = 1.
Since the angular momentum has no projection in the direction of motion, helicity has
the interesting property of being a conserved quantity for a free particle.
We now extend the concept of helicity for Dirac spinors, introducing the vector
23 31 12
⌃=( , , ), (3.58)

where the three elements are 4-matrices. Let us restrict to the Dirac and Weyl represen-
tations. We explicitly have ✓ ◆
0
⌃= . (3.59)
0
For spin we choose ⌃/2. We can choose the 4-spinors to be eigenstates of this operator.
Let us use the subscripts + and instead of 1 and 2. We take
✓ ◆
1 3 /2 0 r
⌃3 ur = ur = ur where r = +1, 1 (3.60)
2 0 3 /2 2
We also take
✓ ◆
1 3 /2 0 r
⌃3 vr = vr = vr where r = +1, 1 (3.61)
2 0 3 /2 2
The reason for the extra minus sign in this last equation will become clear later.
Helicity is defined as
✓ ◆
· p̂ 0
⌃p = ⌃· p̂ = . (3.62)
0 · p̂

51
We choose
r r
⌃p ur = ur and ⌃p vr = vr where r = +1, 1 (3.63)
2 2
An explicit solution for these spinors in the Dirac representation of the -matrices can
be found in Appendix 3.13.

3.6 Commutators and anticommutators

We already mentioned the issue of commutators and anticommutators at the beginning
of the Lecture. Let us continue the discussion because this is a crucial point for the
quantization of the Dirac field
We construct the Fock space by applying creation operators to the vacuum state. For
the KG field we have that the exchange of two particles leads to the same ket,
|k1 ; . . . ; ki ; . . . ; kj ; . . .i = |k1 ; . . . ; kj ; . . . ; ki ; . . .i (3.64)
Also, by applying the same operator a† (k) we obtain a state
|k; k; . . . ; ki (3.65)
These are properties of bosons. They follow from the commutation relations between
ladder operators.
It is clear that we cannot follow exactly the same for fermions because this would
contradict Fermi-Dirac statistics. Indeed, the exchange of identical particles leads to a
minus sign and we cannot have two fermions with the same quantum numbers.
To see which change we have to make, it is illuminating to write the fundamental
relations for a harmonic oscillator that we found in (2.156):
[N, a† ] = + a† ,
[N, a] = a , (3.66)
where N = a† a. Which relations we need to impose that lead to the properties (3.66)?
There are two possible answers :

1. a and a† obey the following commutation relations:

⇥ †⇤ ⇥ ⇤
a, a = 1 , [a, a] = a† , a† = 0 . (3.67)

2. a and a† obey the following anticommutation relations:

{a, a† } = 1 , {a, a} = {a† , a† } = 0 . (3.68)

The commutation relations in (3.67) have been used for the KG field where we have
Bose-Einstein statistics. If instead we want to have Fermi-Dirac statistics, we need to
adhere to the second possibility in (3.68). Indeed, exchange of the order in creation (or
annihilation) operators leads to minus signs. To verify that anticommutation leads to
(3.66) it is sufficient to use the identity [AB, C] = A{B, C} {A, C}B. (See Exercise 7).
After quantization, we interpret the Dirac operators c†r , d†r as creation operators and cr ,
dr as annihilation operators. Below we shall construct the Fock space of the Dirac theory.
Let us anticipate that the vacuum |0i will be defined as the state that all annihilation
operators annihilate, and the one-particle states as the result of the action of c†r , d†r on
|0i.

52
3.7 Quantization of the Dirac field
From the Dirac Lagrangian

L = (x) i 0 @0 + i i @i m (x) , (3.69)

0 †
and using = , we obtain the momentum conjugate field of ,
@L †
⇡↵ = =i ↵ , (3.70)
@ ˙↵
where the index ↵ is a Dirac index.
To quantize the Dirac field we impose the equal-time canonical anticommutation re-
lations,
†
{ ↵ (x, t), (x0 , t)} = ↵
3
(x x0 ) ,
† †
{ ↵, }={ ↵, }=0, (3.71)

where ↵ and are Dirac indices.

The canonical anticommutation relations (3.71) imply the following anticommutation
relations for the ladder operators

{cr (p), c†s (p0 )} = rs

3
(p p0 ) ,
{dr (p), d†s (p0 )} = rs
3
(p p0 ) , (3.72)

and all the other anticommutators vanish. (We propose to verify it in Exercise 10)
Actually from (3.72) we can deduce (3.71). The two sets of anticommutation relations
are equivalent. Schematically
8
† 3
9 >
> {c, c† } = 3 ()
{ , } = i () > = >
< {d, d† } = 3 ()
{ , }=0 () (3.73)
>
; >
> {c, c} = 0
and so on >
:
and so on

Any operator formed with the fields can be expressed in terms of creation and anni-
hilation operators. Let us show this for i) energy-momentum, ii) charge, and iii) spin
(helicity).
In order to get more insight in the expressions we will get, we will write them in terms
of number operators. We define these number operators as we did in the KG case,

Nr (p) = c†r (p)cr (p) , N r (p) = d†r (p)dr (p) . (3.74)

i) We start with the energy, given by the Hamiltonian H.

We can use the result Z
† @
H= d3 x i , (3.75)
@t
which is (3.144) obtained in Exercise 3.
Introducing the Fourier expansion of the fields in (3.75) we obtain
Z X⇥ ⇤
H = d3 p Ep cr (p)† cr (p) dr (p)d†r (p) . (3.76)
r

53
 We have the same problem that we had with KG, the vacuum expectation value
h0| H |0i is infinite, because h0| dr d†r |0i is di↵erent from zero. In order for the vacuum
to have vanishing energy we shall impose normal ordering. This concept was already
introduced to the KG field: the annihilation operators being required to be at the
right of the creation operators. However, in the Dirac case there is an important
di↵erence. Whenever we interchange Dirac fields we get a minus sign, which comes
from the anticommutation relations. Instead, there was no sign in the case of the
KG field.
More specifically, normal ordering of two Dirac fields is defined as
+ + +
: ↵ := ↵ ↵ + ↵ + ↵ , (3.77)

where ↵, are components of either or .

Once normal ordered, the Hamiltonian reads
Z X⇥ ⇤
: H : = d3 p E p cr (p)† cr (p) + dr (p)d†r (p)
r
Z X
= d3 p Ep [Nr (p) + N r (p)] . (3.78)
r

Thanks to the prescription of normal ordering, the vacuum has zero energy

h0| : H : |0i = 0 . (3.79)

From now on we assume the observable operators to be normal ordered, even if we

do not write it explicitly.
The expression for the operator momentum P µ can be obtained by ”covariazing”
(3.78),
XZ X
µ
:P : = d3 p pµ [Nr (p) + N r (p)] . (3.80)
r r

ii) We now would like to work out the relevant expressions for the electric charge. As
we did in the KG case, to do that we first identify an internal symmetry of the Dirac
Lagrangian.
A (global) phase transformation of the field consists is to multiply the field by a
constant phase ↵. Global means that ↵ does not depend on the space time point x.
The transformation reads
(x) ! ei↵ (x) . (3.81)
Under this transformation the adjoint field behaves as
i↵
(x) ! e (x) . (3.82)

Since the Dirac Lagrangian (3.35) contains the bilinears i @/ and m , the phase
↵ cancels out and the Lagrangian is invariant under the transformation,

L!L (3.83)

We conclude the Dirac Lagrangian has a phase symmetry, in the same way we found
for the complex KG Lagrangian.

54
 Noether’s theorem implies we have a conserved current given by (1.40). We calculate
this current density using the first-order change in the field transformation

! + i↵ + O(↵2 ) , ! i↵ + O(↵2 ) , (3.84)

where the fields are all at the same spacetime point x. The Noether current is
proportional to
@L µ µ
( )=i (i↵ )= ↵ . (3.85)
@[@µ ]

We define the conserved current as

jµ = e (x) µ
(x) , (3.86)

where we have chosen for the multiplicative constant to have the electric charge
e = |qe | > 0. We will see that (3.86) is the electromagnetic current in QED.
The conserved current (3.86) implies a conserved charge, defined as the spatial
integral Z Z
Q = d x j (x) = e d3 x † (x) (x) ,
3 0
(3.87)

We extend the integral over all space.

Introducing the Fourier expansion of the field, using normal ordering, and after some
algebra, one finally gets
Z X⇥ ⇤
Q= e d3 p cr (p)† cr (p) d†r (p)dr (p)
r
Z X
3
= e dp [Nr (p) N r (p)] . (3.88)
r

iii) A proper treatment of spin and helicity goes need to investigate the Lorentz group
properties of 4-spinors, and this goes beyond what we do in this Lecture 3, although
we will have some of this advanced material in Lecture 13.
Let us postulate the expression for the spin operator along the z-axis
Z
1
S3 = d3 x † ⌃3 , (3.89)
2

Actually this form is not so surprising if we compare it with (3.75).

Again we introduce the Fourier expansion in the expression for S3 and after some
calculation that uses normal ordering we get
Z X hr i
r
S 3 = d3 p cr (p)† cr (p) + d†r (p)dr (p) (3.90)
r
2 2

In the calculation we have to take into account our election of sign in (3.61).

55
3.8 Fock space
In analogy with the real and complex KG field we expect c†r (p) and d†r (p) to create
particles.
We start defining the vacuum or ground state |0i. Since the operators cr (p) and dr (p)
are annihilation operators, we define |0i as

cr (p) |0i = 0 , dr (p) |0i = 0 for all p and r . (3.91)

The states of the Fock space are constructed applying creation operators c†r (p) and
d†r (p). In order to see the properties of the states, we deduce the following equations:

i) From (3.80) we easily get

[P µ , c†r (k)] = k µ c†r (k) , [P µ , d†r (k)] = k µ d†r (k) . (3.92)

Applying these equations to the vacuum state

P µ c†r (k) |0i = k µ c†r (k) |0i , P µ d†r (k) |0i = k µ d†r (k) |0i . (3.93)

ii) Similarly, from (3.88)

[Q, c†r (k)] = +e c†r (k) , [Q, d†r (k)] = e d†r (k) . (3.94)

and
Q c†r (k) |0i = +e c†r (k) |0i , Q d†r (k) |0i = e d†r (k) |0i . (3.95)

iii) Similarly, from (3.90)

r † r †
[S3 , c†r (k)] = c (k) , [S3 , d†r (k)] = d (k) . (3.96)
2 r 2 r
and
r † r †
S3 c†r (k) |0i = c (k) |0i , S3 d†r (k) |0i = d (k) |0i . (3.97)
2 r 2 r
If we had worked helicity we would have obtained

[⌃k , c†r (k)] = r c†r (k) , [⌃k , d†r (k)] = r d†r (k) . (3.98)

and
⌃k c†r (k) |0i = r c†r (k) |0i , ⌃k d†r (k) |0i = r d†r (k) |0i . (3.99)

With these formulas it is easy to identify the properties of the states created using the
creation operators. For example, consider

c†r (p) |0i and d†r (p) |0i with either r = + or r = (3.100)

From (3.93) both states have momentum pµ . In particular they have energy Ep =
p
p2 + m2 as it corresponds to a particle of mass m. From (3.99) both particles have
helicity h = r. Now it comes what distinguish the two states: the electric charge. Indeed,
from (3.95) we see that the first state has a charge e while the second has charge +e.
We consider the ”particle” to be the electron e with charge e and the corresponding
”antiparticle” to be the positron e+ with charge e. This is a convention, since we could

56
reverse the roles of e and e+ . 6 As it happened with the complex KG field, particles and
antiparticles have exactly the same mass.
Now consider
c†+ (p) |0i and c† (p) |0i (3.101)
Going through the same analysis that before, we see that both correspond to a electron
with momentum p. What di↵erentiates now the states is helicity; the first has h = +
(spin along the direction of movement) and the second has h = (spin opposite to the
direction of movement).
To summarize, the Fock space constructed when we apply creation operators to the
vacuum, has one-particle states7 ,

c†+ (p) |0i ! state with an electron with momentum p and h = + ,

c† (p) |0i ! state with an electron with momentum p and h = ,
d†+ (p) |0i ! state with a positron with momentum p and h = + ,
†
d (p) |0i ! state with a positron with momentum p and h = . (3.102)

We can build two-particle states, for example,

c†+ (p0 )c† (p) |0i ! e with p, h = and e with p0 , h = + ,

d†+ (p0 )c† (p) |0i ! e with p , h = and e+ with p0 , h = + , (3.103)

and so on.
We finish by noting that the total number operator
Z X
N= d3 p [Nr (p) + N r (p)] (3.104)
r

counts the total number of electrons and positrons in a state, while the charge operator
we found in (3.88) Z X
Q= e d3 p [Nr (p) N r (p)] (3.105)
r

counts the number of electrons minus the positrons. For example, if the eigenvalue of Q
is 2e we know there are two more electrons than positrons, if it is +3e we know there
are three more positrons than electrons.

3.9 Dirac bilinears

The current in (3.86) is a 4-vector. We can finally understand the meaning of the µ-label
in µ . While the four objects 0 , 1 , 2 and 3 are not a 4-vector, the four objects
0 1 2 3
( , , , ), (3.106)

do indeed form a 4-vector.

µ
The object is a bilinear in the two Dirac fields. Here we list a set of bilinears
with a well defined Lorentz structure
6
Note that it is also a convention to assign negative charge for the electron. However, once we have
adopted this convention, the sign of the charge of other particles, like the proton, is fixed.
7
The exact normalization is not important here.

57
 scalar
5 pseudoscalar
µ
vector
µ
5 pseudovector
µ⌫
tensor

We will discuss these bilinears when discussing the Lorentz group in Section 13.3. The
”pseudo” in pseudoscalar and pseudovector refers to the Parity transformation, x ! x.
Under parity a pseudoscalar changes sign, and a pseudovector does not.
The bilinears in the list are of the form with
µ µ µ⌫
{ A} = {14⇥4 , 5, , 5 , }. (3.107)

It turns out that this set of matrices can be considered the elements of a basis in Dirac
space in the following sense
1) An arbitrary 4 ⇥ 4 Dirac matrix can be expressed as a complex linear combination
X
= cA A (3.108)

2) the elements in the basis are linearly independent,

X
cA A = 0 =) cA = 0 for all A (3.109)
A

We shall not prove these statements, but just note that the total number of matrices
in the set 1 + 1 + 4 + 4 + 6 = 16, as it should be because an arbitrary 4 ⇥ 4 Dirac matrix
has 16 complex entries.

3.10 Feynman propagator

We define the temporal ordering of two Dirac fields as

T{ ↵ (x) (x0 )} = ✓ (t t0 ) ↵ (x) (x0 ) ✓ (t0 t) (x0 ) ↵ (x) . (3.110)

Note that it contains a minus sign, consistent with the anticommutation relations.
We define the Feynman propagator iSF for the Dirac field in a similar way as we did
in the case of the KG field

iSF ↵ (x x0 ) = h0| T { ↵ (x) (x0 )} |0i . (3.111)

We have explicitly written the indices ↵ and , and this makes very clear that the prop-
agator is a 4 ⇥ 4 matrix in Dirac space. As we have been doing, we will often not write
these indices, so (3.111) will be written as

iSF (x x0 ) = h0| T { (x) (x0 )} |0i . (3.112)

In the rhs of (3.112) there is the product of a 4⇥1 column-vector with a 1⇥4 row-vector
, which gives a 4 ⇥ 4 matrix.
The following equation
✓ ◆
µ @
i + m SF (x x0 ) = (4) (x x0 ) (3.113)
@xµ

58
is easily solved in momentum space, where we define
Z
d4 p ipz
SF (z) = e SF (p) . (3.114)
(2⇡)4
From (p/ m)SF (p) = 1 we get

i i(p/ + m)
iSF (p) = = 2 . (3.115)
p/ m + i✏ p m2 + i✏

The reason for the i✏ prescription is the same than in the KG case.

3.11 Appendix: Pauli matrices

The Pauli matrices are three Hermitian and unitary 2 ⇥ 2-matrices i with i = 1, 2, 3,
✓ ◆ ✓ ◆ ✓ ◆
1 0 1 2 0 i 3 1 0
= , = , = . (3.116)
1 0 i 0 0 1

We list some properties of these matrices

1)
( i )2 = 12⇥2 for i = 1, 2 or 3 (no sum over i) (3.117)
as for any Hermitian and unitary matrix. The matrix 12⇥2 is the 2 ⇥ 2 identity
matrix. In the rest of the Apendix we shall not write the identity matrix explicitly
in the formulas.

2)
i i
Tr =0, and Det = 1, for i = 1, 2 or 3 . (3.118)

3)
i j ij
= + i✏ijk k
, (3.119)
where we sum over repeated indices. Here we are in Euclidian space, the Levi-Civita
symbol is ✏123 = ✏123 = 1.

4) Eq. (3.119) leads to the following expressions for the commutator and the anticom-
mutator of sigma matrices

[ i, j
] = 2i✏ijk k
, { i, j
}=2 ij
, (3.120)

and also to
(~a · ~ )(~b · ~ ) = (~a · ~b) + i(~a ⇥ ~b) · , (3.121)
where we have used the matrix identity AB = [A, B]/2 + {A, B}/2.

5)
1 2 3
=i. (3.122)

6) The exponential of Pauli matrices is simplified

ei↵(~n·~ ) = cos ↵ + i (~n · ~ ) sin ↵ , (3.123)

where ~n is a unitary vector and ↵ a number.

59
 7)
2 i 2
= ( i )⇤ . (3.124)

8) Completeness relation:
X ✓ ◆
i i 1
( )ab ( )cd = 2 ad bc ab cd . (3.125)
i
2

9) Some traces properties

i j ij i j k
Tr( )=2 , Tr( ) = 4i✏ijk ,

i j k l ij kl ik jl il jk
Tr( ) = 2( + ). (3.126)

3.12 Appendix: More properties of ’s

µ⌫
//b = ab
1. a i aµ b⌫

2. a
/a/ = a2
µ⌫
3. [ 5 , ]=0
µ⌫
4. 5 = ( i/2)✏µ⌫↵ ↵

(remember our convention ✏0123 = 1 so that ✏0123 = 1)

Contraction identities
µ
1. µ =4
µ ⌫ ⌫
2. µ = 2
µ ⌫ ⇢
3. µ = 4⌘ ⌫⇢
µ ⌫ ⇢ ⇢ ⌫
4. µ = 2

Some trace properties

1. Tr ( 5 ) = Tr ( µ ⌫ 5
)=0

2. The trace of the product of an odd number of is 0

In particular Tr ( µ ) = 0
µ1 µ2 µ2n µ2n µ2n µ2n µ2 µ1
3. Tr ( ... 1
) = Tr ( ... ) n = 1, 2, . . .
µ ⌫
4. Tr ( ) = 4⌘ µ⌫
µ ⌫ ⇢
5. Tr ( ) = 4(⌘ µ⌫ ⌘ ⇢ ⌘ µ⇢ ⌘ ⌫ + ⌘ µ ⌘ ⌫⇢ )
µ ⌫ ⇢ 5
6. Tr ( ) = 4i✏µ⌫⇢

60
3.13 Appendix: Explicit plane-wave solutions
Plane-wave solutions for are of the form

(x) = N ur (p) exp ( ipx) , r = 1, 2 , (3.127)

(x) = N vr (p) exp (ipx) , r = 1, 2 , (3.128)

where u, v are 4-spinors in momentum space and N a normalization constant. Imposing
that u, v are solutions to the Dirac equation we find the Dirac equation in momentum
space

p/ m ur (p) = 0 ,
p/ + m vr (p) = 0 . (3.129)

Here we would like to find explicit solutions for these spinors.

Let us concentrate on the spinor u. Divide the four components of the 4-spinor into
2+2 entries
✓ ◆
t
u= , (3.130)
b

where t and b are two 2-spinors.

Let us now work in the Dirac representation, where we have
✓ ◆
Ep m ·p
p/ m = , (3.131)
·p Ep m
p
where Ep = p2 + m2 .
Since (p/ m) u = 0, we can split it into two equations:

(Ep m) t ·p b =0,
· p t (Ep + m) b =0, (3.132)

The second equation gives us a direct relation between b and t:

·p
b = t . (3.133)
Ep + m
We will compute t as a combination of the two independent solutions
✓ ◆ ✓ ◆
1 0
+ = , = . (3.134)
0 1
In the main text we used the indices 1, 2 and here we are using +, .
Then
✓ ◆
p ±
u± (p) = Ep + m p . (3.135)
Ep +m ±

where we have chosen the normalization N such that

u†r (p)us (p) = vr† (p)vs (p) = 2Ep rs ,

u†r (p)vs p) = vr† (p)us ( p) = 0 . (3.136)

61
 Similarly we can get
✓ p ◆
p ⌥
Ep +m
v± (p) = ± Ep + m . (3.137)
⌥

We have found four independent solutions of the Dirac equation u± , v± . They are valid
in the Dirac representation.
Let us define the 4 ⇥ 4 matrix
✓ ◆
p̂ 0
⌃p = , (3.138)
0 p̂

where p̂ is a unitary 3-vector in the direction of the particle 3-momentum. It is related to

the helicity operator that we defined in (3.62). The spinors u± and v± are its eigenvectors,

⌃p u± = ±u± , ⌃p v± = ⌥v± . (3.139)

The reason for the mismatch in + and - in v± is explained in the main text.
The solutions we have found are exact. They are useful in the NR limit, where
|pi | ⌧ Ep + m, because the components of the 2-spinor t are much larger that the
components of b . This makes it possible to make some simplifications of the physics in
this NR limit. The split into ”large components” and ”small components” appears in the
Dirac representation. In other representations the large and small components would be
mixed.

62