Relativistic Waves and Quantum Fields

Andreas Brandhuber

Department of Physics
Queen Mary, University of London
Mile End Road, London, E1 4NS
United Kingdom
email: [email protected]

Abstract

Lecture notes for the course Relativistic Waves and Quantum Fields MSci 4242/QMUL
PHY-415. (These are preliminary notes and may contain typos, so use with care!)
Contents

0 Why Quantum Field Theory? 3

1 Quantum Mechanics and Special Relativity 5

1.1 Quantum Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

1.1.1 Principles of Non-Relativistic Quantum Mechanics . . . . . . . . . . 5

1.1.2 Symmetries and Angular Momentum . . . . . . . . . . . . . . . . . 7

1.2 Special Relativity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

1.2.1 Lorentz Transformations . . . . . . . . . . . . . . . . . . . . . . . . 14

1.2.2 4-Vector Notation and Tensors . . . . . . . . . . . . . . . . . . . . . 15

1.2.3 Lorentz transformations, again . . . . . . . . . . . . . . . . . . . . . 17

1.2.4 Lorentz Group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

1.2.5 Poincaré Group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

1.2.6 The Maxwell Equation — First Glimpse at a Relativistic Wave

Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

2 Relativistic Quantum Mechanics 25

2.1 The Klein-Gordon Equation . . . . . . . . . . . . . . . . . . . . . . . . . . 25

2.2 The Dirac Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

1
2 CONTENTS

2.3 Representation of the Dirac Matrices . . . . . . . . . . . . . . . . . . . . . 28

2.4 Probability Density for the Dirac Equation . . . . . . . . . . . . . . . . . . 29

2.5 Extreme Non-Relativistic Limit of the Dirac Equation . . . . . . . . . . . . 29

2.6 Spin of the Dirac Particles . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

2.7 The Covariant Form of the Dirac Equation . . . . . . . . . . . . . . . . . . 31

2.8 Properties of the γ-Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . 32

2.9 The Dirac Equation and Lorentz Transformations . . . . . . . . . . . . . . 33

2.10 Plane Wave Solutions of the Dirac Equation . . . . . . . . . . . . . . . . . 36

2.11 Properties of Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

2.12 Anti-Particles — Hole Theory . . . . . . . . . . . . . . . . . . . . . . . . . 40

2.13 Vacuum Polarization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

2.14 Charge Conjugation Symmetry C . . . . . . . . . . . . . . . . . . . . . . . 41

2.15 Space Inversion P . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

2.16 Time Reversal T . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

2.17 Dirac Covariants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

2.18 Neutrinos . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

2.19 Feynman’s Interpretation of the Klein-Gordon Equation . . . . . . . . . . . 47

2.20 Dirac Equation in an Electromagnetic Field . . . . . . . . . . . . . . . . . 48

2.21 The Magnetic Moment of the Electron . . . . . . . . . . . . . . . . . . . . 50

2.22 Hydrogen Atom Spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

Chapter 0

Why Quantum Field Theory?

The fundamental challenge of every physical theory is to describe Nature and its phenom-
ena in as much detail as possible. Furthermore, this should be possible for the microcosm
and the macrocosm i.e. we are looking for theories that are valid at short distance scales
and long distance scales.

Classical physics is able to give a satisfactory description of many physical phenomena,

and its validity is not limited to the macroscopic regime alone. Certain aspects of the
microcosm can be treated with classical physics, e.g. certain aspects of kinetic gas theory.
But in general classical physics gives only an approximation that breaks down when we
reach a certain energy or distance scale.

The deficiencies of the classical viewpoint are most pronounced in the physics of the
microcosm and in the subnuclear regime. Many phenomena of molecular physics, atomic
physics, nuclear physics and elementary particle physics simply cannot be explained using
ideas from classical physics. Also the inclusion of Quantum Mechanics (QM) alone is often
insufficient, to explain physical problems of the microcosm. As was realised in the last
century, only the inclusion of Einstein’s Theory of Special Relativity (SR) leads to the
desired success.

Today Quantum Field Theory (QFT) provides the unified framework to describe all
particles and its interactions (forces) that we observe in Nature, including electromag-
netism, the weak and the strong nuclear force, and probably gravity. QFT was born out
of the attempt to combine the rules of Quantum Mechanics with the principles of Special
Relativity and field theory. The prime example for the success of QFT is Quantum Elec-
trodynamics (QED), which with its extremely accurate predictions of physical quantities,
like the anomalous magnetic moment of the electron, is also called ”the jewel of physics”.

3
4 CHAPTER 0. WHY QUANTUM FIELD THEORY?
Chapter 1

Quantum Mechanics and Special

Relativity

In this chapter we want to review and introduce basic concepts of Quantum Mechanics and
Special Relativity. At this stage we treat these two topics separately before we attempt
to unify them in Chapter 2 trying to preserve most features of usual non-relativistic QM.
This will lead only to partial success since a couple of conceptual problems will arise.
As was understood in the last century, the main reason for these problems is that the
inclusion of SR implies the existence of anti-particles and leads to a multi-particle theory
in which probability is not preserved, because particles can be destroyed and created
(death and birth of particles). These issues can be addressed properly if we adopt the
formalism of Quantum Field Theory, which will be developed from Chapter 3 on.

1.1 Quantum Mechanics

1.1.1 Principles of Non-Relativistic Quantum Mechanics

In this section I will remind you of the most important axioms of Quantum Mechanics
(QM). I expect that you are familiar with most of these facts, but it is good to repeat
them to see which ones we will be forced to give up once we include SR.

Axioms of QM:

• The state of the system is represented by a wave function |Ψi (also called a
vector in Hilbert space). Note that Ψ = Ψ(qi , si , t) is a function of the coordinates

5
6 CHAPTER 1. QUANTUM MECHANICS AND SPECIAL RELATIVITY

of individual particles qi , some internal degrees of freedom like spin si and, if we are
in the Schödinger picture also on time t (if you are not familiar with this concept
either wait until I will review it later in the course or check your favourite book
on QM). Most importantly Ψ itself has no direct physical meaning whatsover; only
expression like the probability density ρ = |Ψ|2 ≥ 0.

• Physical observables are in one-to-one correspondence with linear, Hermitian

operators. Hermiticity A
b=A b† implies that all eigenvalues of the operators are
∂
real. Example: momentum pbx = −i~ ∂x .

• The system can be in an eigenstate ψn which obeys Aψ b n = λn ψn with real eigen-

value λn . A general state Ψ can be written as a linear combination of a complete
set of eigenfunctions of a complete set of commuting P operators [Abi , A
bj ] = 0
(thanks to linearity of the QM operators). So Ψ = n an ψn where the an are com-
plex coefficients. If the eigenfunctions are properly orthonormalised (= orthogonal
and normalised to one) i.e. d3 xψn∗ ψm = δnm then |an |2 is the probability to find
R

the system in the state ψn .

• The result of a QM measurement of Ab is one of its eigenvalues λn (with probability

|an |2 ). The QM expectation value — the average over many measurements — is

Z X
hAi = hΨ|A|Ψi
b = d3 xΨ∗ AΨ
b = |an |2 λn . (1.1.1)
n

• The time evolution of the system is desribed by the Schrödinger equation i~ ∂Ψ

∂t
=
HΨ. The QM Hamiltionian H is obtained from its classical counterpart H by
b b
the replacements p~ → −i~∇ ~ and E → +i~ ∂ . For example: H = p~2 + V (~x) →
∂t 2m
~2 ~ 2
H = − 2m ∇ + V (~x). Since H
b is linear we can apply the superposition principle
to wavefunctions. Furthermore, Hermiticiy H b = Hb † implies the conservation of
probability. Proof:

∂Ψ∗
Z Z
d d ∗ ∂Ψ
hΨ|Ψi = Ψ Ψ= ( Ψ + Ψ∗ Ψ)
dt dt ∂t ∂t
Z
i b ∗ i b
= ((− HΨ) Ψ + Ψ∗ (− HΨ))
~ ~
Z
i b ∗ Ψ − Ψ∗ HΨ) i b
= ((HΨ) b = (hHΨ|Ψi − hΨ|HΨi)
b = 0 . (1.1.2)
~ ~
1.1. QUANTUM MECHANICS 7

1.1.2 Symmetries and Angular Momentum

Symmetries and Conservation Laws

In Quantum Mechanics (QM), for an observable A without explicit time dependence1 we

have
dhAi
i~ = hΨ|[A,
b H]|Ψi
b , (1.1.3)
dt

for any state |Ψi and where hAi = hΨ|A|Ψi

b , is the QM expectation value of A.
Proof:
Z
dhAi d
i~ = i~ Ψ∗ AΨ
b
dt dt
∂Ψ∗ b
Z
= i~ AΨ + Ψ∗ Ab ∂Ψ
∂t ∂t
Z
∂Ψ ∗ b b ∂Ψ )
= −(i~ ) AΨ + Ψ∗ A(i~
∂t ∂t
Z
= b ∗ AΨ
−(HΨ) b + Ψ∗ AbHΨ
b
Z
= −Ψ∗ H b + Ψ∗ A
b AΨ bHΨb
h i
= hΨ| A,bH b |Ψi . (1.1.4)

Therefore, if
[A,
b H]
b = 0, (1.1.5)

then
dhAi
= 0, (1.1.6)
dt

and we say that A is a conserved quantitity or constant of motion.

A symmetry is a transformation on the coordinates of a system which leaves the

Hamiltonian H invariant. We shall see that conservation laws are the consequence of
symmetries of a system. Symmetries are very powerful since they can be used to derive
results for a system even when we do not know the details of the dynamics involved.

1
i.e. the QM operator A
b corresponding to the observable A obeys ∂ A/∂t
b = 0.
8 CHAPTER 1. QUANTUM MECHANICS AND SPECIAL RELATIVITY

Translational Invariance

First consider a single particle. If ~x is the position vector, then a translation is the
operation
~x → ~x0 = ~x + ~a . (1.1.7)
If H
b is invariant then
b x0 ) = H(~
H(~ b x + ~a) = H(~
b x) . (1.1.8)
For an infinitesimal displacement we can make a Taylor expansion2
b x + ~a) ∼
H(~ = H(~ ~ H(~
b x) + ~a · ∇ b x) , (1.1.9)

ignoring higher powers of ~a. Thus, if H

b is invariant,

b x + ~a) − H(~
0 = H(~ ~ H(~
b x) = ~a · ∇ b x) . (1.1.10)

In general for the momentum operator P~ and any other operator O(~
b x) we have
b

[P~ , O]Ψ ~ O]Ψ

= [−i~∇,
b b b
~ OΨ)
= −i~∇( b − O(−i~Ψ)
b
~ O)Ψ
= −i~(∇ b , (1.1.11)

where we have suppressed the explicit ~x and t dependence. Since eqn. (1.1.11) is true for
arbitrary wavefunctions Ψ
[P~ , O(~
x)] = −i~∇~ O(~
b x) . (1.1.12)
b b

In particular for O
b=H
b
[P~ , H] ~H
= −i~∇ b. (1.1.13)
b b

Now,
~H
0 = −i~~a · ∇ ~ , H]
b = ~a · [P
b b (1.1.14)
and since this is true for an arbitrary displacement vector ~a we find

[P~ , H] = 0. (1.1.15)
b b

~i
dhP
We conclude that momentum is a conserved quantity dt
= 0 if the H
b is translationally
invariant.

Take e.g.
2
b =−~ ∇
H ~ 2 + V (~x) (1.1.16)
2m
2
The symbol ∼
= indicates that we only expand to first order in ~a and suppress all higher order terms.
1.1. QUANTUM MECHANICS 9

For the translation, ~x0 = ~x + ~a, in particular x0 = x + a

∂ ∂x0 ∂ ∂
⇒ = 0
= (1.1.17)
∂x ∂x ∂x ∂x0
~ and ∇
and similarly for y and z. Thus ∇ ~ 2 are invariant under translations and
2
b x + ~a) = − ~ ∇
b x0 ) = H(~
H(~ ~ 02 + V (~x0 )
2m
~2 ~ 2
= − ∇ + V (~x + ~a) . (1.1.18)
2m
Hence, for a translationally invariant Hamiltonian we must require

V (~x + ~a) = V (~x) , (1.1.19)

which is only true for a (trivial) constant potential, i.e. for a free particle. Thus, the
momentum of a free particle is conserved in QM in the sense

dhP~ i
= 0. (1.1.20)
dt

Consider now a two particle system (easily generalised to N particles). If the two
particles have position vectors ~x1 and ~x2 , the invariance condition for the translation of
the system through ~a reads

H(~
b x1 , ~x2 ) = H(~
b x1 + ~a, ~x2 + ~a) . (1.1.21)

Then for an infinitesimal translation

b x1 + ~a, ~x2 + ~a) ∼
H(~ = H(~ ~ 1 H(~
b x1 , ~x2 ) + ~a · ∇ ~ 2 H(~
b x1 , ~x2 ) + ~a · ∇ b x1 , ~x2 ) , (1.1.22)

we find that translational invariance implies

~1+∇
0 = ~a · (∇ ~ 2 )H(~
b x1 , ~x2 ) , (1.1.23)

and the total momentum operator is

P~ = P~ 1 + P~ 2 , (1.1.24)
b b b

where
P~ 1 = −i~∇
~ 1 and P
~ 2 = −i~∇
~2. (1.1.25)
b b

Identical to the one particle case, for any operator O(~

b x1 , ~x2 ),

[P~ 1 , O(~ ~ 1 O(~

x1 , ~x2 )] = −i~∇ ~ 2 , O(~
b x1 , ~x2 ) and [P ~ 2 O(~
b x1 , ~x2 )] = −i~∇ b x1 , ~x2 ) , (1.1.26)
b b b
10 CHAPTER 1. QUANTUM MECHANICS AND SPECIAL RELATIVITY

and so
[P~ , O(~ ~1+∇
x1 , ~x2 )] = −i~(∇ ~ 2 )O(~
b x1 , ~x2 ) . (1.1.27)
b b

This is true in particular for Ob = H. b Thus,

~1+∇
0 = −i~~a · (∇ ~ 2 )H ~ , H]
b = ~a · [P
b b . (1.1.28)
Since this must be true for arbitrary translation vector ~a, we have

[P~ , H] = 0, (1.1.29)
b b
~i
dhP
and total momentum is conserved in the sense of QM i.e. dt
= 0.

Rotational Invariance

Just as translational invariande is associated with conservation of momentum, it turns out

that rotational invariance is associated with conservation of angular momentum (AM).

Take spherical polar coordinates and take the axis of rotation to be the z-axis. Spec-
ify the position vector ~x in spherical polar coordinates (r, θ, φ) of the point. Then the
symmetry operation ~x → ~x0 corresponding to a rotation by an angle α about the z−axis
is
(r, θ, φ) → (r0 , θ0 , φ0 ) = (r, θ, φ + α) . (1.1.30)
For the Hamiltonian to be invariant under rotations about the z-axis
b x0 ) = H(~
H(~ b x)
b 0 , θ0 , φ0 ) = H(r,
⇐⇒ H(r b θ, φ)
⇐⇒ H(r,
b θ, φ + α) = H(r, b θ, φ) . (1.1.31)
For an infinitesimal rotation
b θ, φ + α) ∼
H(r, b θ, φ) + α ∂ H(r,
= H(r, b θ, φ) , (1.1.32)
∂φ
and, hence, for invariance of H
b

0 = b θ, φ + α) − H(r,
H(r, b θ, φ) = α ∂ H(r,
b θ, φ)
∂φ
∂ b
−→ H(r, θ, φ) = 0 . (1.1.33)
∂φ
The z−component of the orbital AM operator written in spherical coordinates is3
bz = −i~ ∂
L (1.1.34)
∂φ
3
In cartesian coordinates ~x = (x, y, z): L
b z = (~x ~ )z = −i~(x ∂ − y ∂ ) . The other components, L
b×P
b bx
∂y ∂x
and Lb y , can be obtained by cyclic permutation of (x, y, z).
1.1. QUANTUM MECHANICS 11

In general, for any other operator O,

b

b = −i~[ ∂ , O]
b = −i~( ∂ (OΨ) b ∂ Ψ) = −i~ ∂ O Ψ .
b
[L
bz , O]Ψ b −O (1.1.35)
∂φ ∂φ ∂φ ∂φ

This is true for arbitrary wavefuntions Ψ, thus

b = −i~ ∂ O ,
b
[L
bz , O] (1.1.36)
∂φ

and in particular for O

b = H,
b

b = −i~ ∂H
b
[L
bz , H] . (1.1.37)
∂φ
If the Hamiltonian is invariant under rotation about the z-axis, we now conclude that

[L
bz , H]
b = 0. (1.1.38)

We can define angles φx and φy analogous to φz ≡ φ for rotations about the x and y-axis.
If H
b is also invariant under rotations about the x and y-axis we will conclude that

[L
bx , H]
b = [L
by , H]
b = [L
bz , H] ~b H]
b = 0 i.e. [L, b = 0. (1.1.39)

Thus AM is a constant of motion

~
dhLi
= 0. (1.1.40)
dt
Any rotation can be built out of successive rotations about the x, y and z-axis.

Whenever H b is invariant under arbitrary rotations the AM L~ is a conserved quantity.

We must construct such an H out of scalars i.e. invariants under rotations. The simplest
b
examples of scalars are the magnitude (length) of a vector or the scalar product of two
vectors. For example, consider the Hamiltonian for a particle moving in a central potential
such as the Coulomb potential

~2 ~ 2
H=−
b ∇ + V (|~x|) . (1.1.41)
2m
~ 2 = ∇·
Because ∇ ~ ∇~ and |~x| are scalars, so is the Hamiltonian and orbital AM is conserved.

This discussion generalizes immediately to two (or more) particles:

For 2 particles, the Hamiltonian is a function of two sets of spherical coordinates

(r1 , θ1 , φ1 ) and (r2 , θ2 , φ2 ) so that

H
b = H(r
b 1 , θ1 , φ1 ; r2 , θ2 , φ2 ) . (1.1.42)
12 CHAPTER 1. QUANTUM MECHANICS AND SPECIAL RELATIVITY

The invariance condition for rotation of the system by an angle α about the z-axis is

H(r
b 1 , θ1 , φ1 ; r2 , θ2 , φ2 ) = H(r
b 1 , θ1 , φ1 + α; r2 , θ2 , φ2 + α) , (1.1.43)

and for an infinitesimal rotation

b 1 , θ1 , φ1 + α; r2 , θ2 , φ2 + α) ∼
H(r = H(r
b 1 , θ1 , φ1 ; r2 , θ2 , φ2 ) + α∂φ1 H
b + α∂φ2 H
b, (1.1.44)
∂
where we defined ∂φ ≡ ∂φ
. Invariance of H
b gives

0 = b 1 , θ1 , φ1 + α; r2 , θ2 , φ2 + α) − H(r
H(r b 1 , θ1 , φ1 ; r2 , θ2 , φ2 )
= α(∂φ1 + ∂φ2 )H b
−→ (∂φ1 + ∂φ2 )H
b = 0. (1.1.45)

The z-components of the orbital AM operator for the two particles are
b1z = −i~∂φ1 , L
L b2z = −i~∂φ2 . (1.1.46)

The z-component of the total orbital AM is:

L
bz = L
b1z + L
b2z . (1.1.47)

Much as before, for any operator O,

[L b = −i~(∂φ1 + ∂φ2 )O
bz , O] b, (1.1.48)

and in particular
[L b = −i~(∂φ1 + ∂φ2 )H
bz , H] b. (1.1.49)
If the Hamiltonian is invariant under rotations about the z-axis, we now conclude that

[L
bz , H]
b = 0. (1.1.50)

By also considering rotations about the x and y-axis we conclude that if H

b is invariant
under arbitrary rotations then
~b H]
[L, b = 0, (1.1.51)
so total AM is a constant of motion
~
dhLi
= 0. (1.1.52)
dt

As usual the orbital AM operators obey the (SO(3) or SU (2)) algebra

[L
bx , L
by ] = i~L
b z , [L
bx , L
by ] = i~L
b z , [L
bx , L
by ] = i~L
bz . (1.1.53)
1.2. SPECIAL RELATIVITY 13

Consequently, it is only possible to know simultaneously the values of one component L ~

~ 2 , for example, Lz and L
and L ~ 2 . Therefore, we may take the conserved quantities to be
~
Lz and L . 2

An atom with atomic number Z with Coulomb forces between the nucleus and the
electrons and between the electrons is an example of a system with the necessary rotational
invariance for conservation of AM. In this case:
Z Z Z
~2 X ~ 2 X Ze2 X e2
H=−
b ∇i − + , (1.1.54)
2m i=1 i=1
4π0 |~xi | i,j=1,i<j 4π0 |~xi − ~xj |

where ∇~ i acts on the coordinates of the i-th electron, and ~xi is the position vector of
the i-th electron relative to the nucleus. As before, ∇ ~ 2i and |~xi | are scalars (rotationally
invariant) and so is |~xi − ~xj |, so that H
b is rotationally invariant.

Remark: Generally for a system with spin it is the total AM

J~ = L
b ~b b ~
+S (1.1.55)
that commutes with the Hamiltonian
b~ b
[J, H] = 0 (1.1.56)
b is rotationally invariant, and then hJz i and hJ~2 i are constants of motion.
if H

For the atomic (non-relativistic) Hamiltonian above, hLz i and hL ~ 2 i are also constants
of motion, because the spin does not appear explicitly in the Hamiltion. This is a mani-
festation of the fact that spin is an effect of Special Relativity as we will see later in the
course. However, if we include the Spin-Orbit interaction due to Relativistic effects
1 1 dV ~b b ~,
H
b Spin−Orbit = L·S (1.1.57)
2m2e c2 r dr

~b does not commute with the

for an electron moving in a central potential V (r), then L
complete Hamiltonian. Then, only hJz i and hJ~2 i are conserved quantities.

1.2 Special Relativity

Topics we are covering in this section include: Lorentz transformations, 4-vectors, covari-
ance, contravariance, forminvariance, Lorentz and Poincare group, Maxwell equations as
an example of a relativistic wave equations.

Einstein’s Theory of Special Relativity (1905) rests on two postulates:

14 CHAPTER 1. QUANTUM MECHANICS AND SPECIAL RELATIVITY

1. The speed of light c in vacuum is absolut. It is the same in all inertial frames (these
are reference frames moving at constant velocity with respect to each other).
2. Principle of Relativity. The laws of physics are the same in all inertial frames. Or
in other words there is no preferred reference frame and, in particular, there is no
absolute time.

Two observers in two different inertial frames K and K 0 using coordinate systems
(t, x, y, z) and (t0 , x0 , y 0 , z 0 ) should agree on the predicted results of all experiments. In
other words the form of all dynamical equations should be invariant (Forminvariance).

An important example are the Maxwell equations written for E ~ and B.

~ The fields
do transform under Lorentz transformations, but the Maxwell equations take the same
form when written in terms of the new (t0 , x0 , y 0 , z 0 ) coordinates and the Lorentz trans-
~ 0 and B
formed fields E ~ 0 . Actually, if Maxwell had looked more carefully he should have
discovered special relativity, since Maxwell equations are not forminvariant under Galilei
transformations but only under Lorentz transformations!

1.2.1 Lorentz Transformations

You are all familiar with Lorentz transformations (LT), say a boost in the x direction:
ct − vx/c
ct0 = p = γ(ct − βx)
1 − v 2 /c2
x − vt
x0 = p = γ(−βct + x)
1 − v 2 /c2
y0 = y
z0 = z (1.2.1)
p
with β = v/c and γ = 1/ 1 − v 2 /c2 . This LT can be neatly rewritten in matrix form as
 0    
ct γ −γβ 0 0 ct
 x0   −γβ γ 0 0   x  .
 
 0 = (1.2.2)
 y   0 0 1 0  y 
z0 0 0 0 1 z
Note that general LTs are linear transformations of the coordinates. If we were to consider
general relativity we would have to relax this condition and consider general non-linear
transformations.

An important consequence of Einstein’s postulates is that the quantity

(ct)2 − x2 − y 2 − z 2 (1.2.3)
1.2. SPECIAL RELATIVITY 15

is Lorentz invariant. Try this with the LT given above

(ct0 )2 − (x0 )2 − (y 0 )2 − (z 0 )2 = γ 2 (c2 t2 − 2βctx + β 2 x2 ) + γ 2 (β 2 c2 t2 + x2 − 2cβtx) − y 2 − z 2
= γ 2 (c2 (1 − β 2 )t2 − (1 − β 2 )x2 ) − y 2 − z 2
= c2 t2 − x2 − y 2 − z 2 , (1.2.4)
where we have used γ 2 (1 − β 2 ) = 1.

This explains why the speed of light is constant in two frames K and K 0 . Assume that
a spherical light wave starts at t = 0 expanding from the origin x = y = z = 0. Then for
t > 0 the equation c2 t2 − x2 − y 2 − z 2 describes a spherical wave front with radius r = ct
i.e. it is expanding with the speed of light. Obviously, using the identity derived above
this holds true also in the frame K 0 and actually for any other frame, hence the speed of
light is consant.

It is very useful to think of Lorentz Transformations (LTs) as generalised rotations.

In order to see this remember γ 2 − γ 2 β 2 = 1 = γ 2 − (βγ)2 and compare this with the
identity cosh2 ω − sinh2 ω = 1. Hence, we can set
γ ≡ cosh ω , βγ ≡ sinh ω , (1.2.5)
and the boost in the x direction can be written as
 
cosh ω − sinh ω 0 0
 − sinh ω cosh ω 0 0 
  (1.2.6)
 0 0 1 0 
0 0 0 1
where similar expressions exist for boosts in the y and z directions. Compare this now
with a spatial rotation around, say, the z axis by an angle θ
 
1 0 0 0
 0 cos θ − sin θ 0 

 0 sin θ cos θ 0 
 (1.2.7)
0 0 0 1
where cos2 θ + sin2 θ = 1. Note that under a spatial rotation also x2 + y 2 + z 2 is invariant.
Also there are similar transformation matrices for rotations around the x and y axis.
This implies that the Lorentz group is six dimensional: there are 3 proper LTs (boost)
and 3 spatial rotations. Why is this a group? Because any combination of two LTs which
corresponds to matrix multiplication of the transformation matrices produces another LT.

1.2.2 4-Vector Notation and Tensors

Recall that c2 t2 − x2 − y 2 − z 2 is rather similar to the dot-product of 3-vectors ~x · ~x =

x2 + y 2 + z 2 (which is invariant under spatial rotations) except for the funny minus signs.
16 CHAPTER 1. QUANTUM MECHANICS AND SPECIAL RELATIVITY

In order to deal with this one introduces a metric tensor η, which is a 4-by-4 matrix, and
two types of 4-vectors, covariant ones and contravariant ones.

We can write the relativistic inner product c2 t2 − x2 − y 2 − x2 as

  
1 0 0 0 ct
 0 −1 0 0  x 
 
(ct, x, y, z) 
 0 0 −1 0   y  (1.2.8)
0 0 0 −1 z
| {z }
≡η

Another way to deal with this is to introduce two types of vectors

xµ ≡ (x0 , x1 , x2 , x3 ) = (ct, x, y, z) = (ct, ~x) contravariant
xµ ≡ (x0 , x1 , x2 , x3 ) = (ct, −x, −y, −z) = (ct, −~x) covariant (1.2.9)
so that
3
X
xµ xµ = c2 t2 − x2 − y 2 − x2 ≡ x · x . (1.2.10)
µ=0

Also if we introduce the Minkowski metric η µν then

3 X
X 3
η µν xν xµ = c2 t2 − x2 − y 2 − x2 ≡ x · x , (1.2.11)
µ=0 ν=0

because
    
1 0 0 0 ct ct
3
 0 −1 0 0    −x  =  x  = xµ .
X    
η µν xν = 
 0 0 −1 0   −y   y  (1.2.12)
ν=0
0 0 0 −1 −z z
   
1 0 0 0 1 0 0 0
 0 −1 0 0   0 −1 0 0
 µν
 
The metric tensor ηµν =  0 0 −1 0  and the inverse metric η =  0 0 −1 0


0 0 0 −1 0 0 0 −1
are used to raise and lower indices:
X 3
η µν xν = xµ ,
ν=0
X 3
ηµν xν = xµ . (1.2.13)
ν=0

Note also that the metric is symmetric ηµν = ηνµ and

 
1 0 0 0
3
X  0 1 0 0  ν
ηµα η αν = 
 0 0 1 0  ≡ δµ .
 (1.2.14)
α=0
0 0 0 1
1.2. SPECIAL RELATIVITY 17

Finally, one introduces the Einstein summation convention in order to avoid writing the
boring sums over and over again: repeated upper and lower index pairs are automatically
summed over (unless stated otherwise). Note that this only makes sense if one index
is upper and the other is lower, otherwise one gets expressions that do not transform
properly under LTs.

Examples
c2 t2 − x2 − y 2 − z 2 = xµ xµ = xµ xµ = η µν xµ xν (1.2.15)
µ µν
x = η xν (1.2.16)
ν
xµ = ηµν x (1.2.17)

1.2.3 Lorentz transformations, again

Now LTs can be written very compactly as

x0µ = Λµ ν xν , (1.2.18)

where Λ can be any LT matrix.

This was for a contravariant vector, what about a covariant one?

x0α = ηαµ x0 µ = ηαµ Λµ ρ xρ = ηαµ Λµ ρ η ρβ xβ ≡ Λα β xβ (1.2.19)

Therefore,
x0α = Λα β xβ with Λα β = ηαµ Λµ ν η νβ . (1.2.20)

Now the Lorentz transformations are those Λ’s that leave xµ xµ invariant. Note that a
priori Λ has 16 components which should be reduced to 6. Let’s see how this works. We
require that the norm of a 4-vector xµ is Lorentz invariant. So we have

ηµν x0µ x0ν = ηµν Λµ α Λν β xα xβ = ηαβ xα xβ , (1.2.21)

which implies
ηµν Λµ α Λν β = ηαβ . (1.2.22)
This is a matrix equation for Λ with 16 components. But only 10 components are inde-
pendent because both sides of the equation are symmetric under exchange of α and β.
Hence the number of independent components of Λ is reduced to 16 − 10 = 6. This is
just the right number to incorporate 3 boosts and 3 spatial rotations!

General tensors with n contravariant indices and m covariant indices transform as

follows

T 0µ1 ...µn ν1 ...νm = Λµ1 α1 . . . Λµn αn Λν1 β1 . . . Λνm βm T α1 ...αn β1 ...βm . (1.2.23)
18 CHAPTER 1. QUANTUM MECHANICS AND SPECIAL RELATIVITY

This is the general transformation law for tensors that transform covariantly, however we
usually do not account tensors with more than two indices (unless we consider General
Relativity).

It is also important to know how differentials and partial derivatives transform under
LTs.

From
x0µ = Λµ ν xν (1.2.24)
we learn that
dx0µ = Λµ ν dxν , (1.2.25)
so dx0µ transforms like a contra-variant vector. It is also known that for general coordinate
transformations we have
∂x0µ ν
dx0µ = dx , (1.2.26)
∂xν
0µ
where for the special case of LTs ∂x∂xν
= Λµ ν .

Furthermore, application of the chain rule yields

 0µ −1
∂ ∂xν ∂ ∂x ∂
0µ
= 0µ ν
= ν
. (1.2.27)
∂x ∂x ∂x ∂x ∂xν
Hence, partial derivatives ∂x∂ µ transform inversely to dxµ i.e. they transform like a co-
variant vector (index downstairs). It is common to introduce the co-variant 4-vector
 
∂ ∂ ~
∂µ = = ,∇ (1.2.28)
∂xµ ∂t
and the contra-variant vector
 
µ ∂ ∂ ~ .
∂ = = , −∇ (1.2.29)
∂xµ ∂t

Important Lorentz invariant contractions:

• The d’Alembertian operator or ”box”

1 ∂2 ~2=
∂µ ∂ µ ≡ ∂ · ∂ = −∇ (1.2.30)
c2 ∂t2
is a relativistic wave operator.
• The energy momentum 4-vector is pµ = (E/c, p~) from which we can form the in-
2
variant pµ pµ = p · p = p2 = Ec2 − p~2 = m2 c2 . Another important invariant is
p · x = pµ xµ = Et − p~ · ~x. Note that Ψ = exp(±ip · x) is a plane wave solution to
the wave equation Ψ = 0 if p2 = 0.
• Note that any dot product of two 4-vectors is a Lorentz invariant a · b = aµ bµ =
η µν aµ bν .
1.2. SPECIAL RELATIVITY 19

1.2.4 Lorentz Group

We will set ~ = c = 1 from now on.

Let us think a moment about infinitesimal LTs Λµ ν = δ µ ν + ω µ ν , where ω µ ν is small

and obeys ω µν = −ω νµ .

For a boost in the x-direction

   
cosh ω − sinh ω 0 0 0 −i 0 0
 − sinh ω cosh ω 0 0   −i 0 0 0 
  ∼ I4 − iωKx , with Kx =   , (1.2.31)
 0 0 1 0   0 0 0 0 
0 0 0 1 0 0 0 0

where we call Kx the generator of the boost in the x-direction. (Check for yourself
that exp(−iωKx ) gives back the finite Lorentz transformation.) Similarly we find the
generators for boosts in the y and z direction.
   
0 0 −i 0 0 0 0 −i
 0 0 0 0 
 , Kz =  0 0 0 0  .
 
Ky =  −i 0 0 0   0 0 0 0  (1.2.32)
0 0 0 0 −i 0 0 0

For a rotation around the x-axis

   
1 0 0 0 0 0 0 0
 0 1 0 0   0 0 0 0 
 0 0 cos φ − sin φ  ∼ I4 − iφJx , with Jx =  0 0
   , (1.2.33)
0 −i 
0 0 sin φ cos φ 0 0 i 0

where Jx is the generator for rotations around the x axis. Similarly we find the generators
for rotations around the y and z axis.
   
0 0 0 0 0 0 0 0
 , Jz =  0 0 −i 0  .
 0 0 0 i   
Jy = 
 0 0 0 0   0 i 0 0  (1.2.34)
0 −i 0 0 0 0 0 0

The boost and rotation generators obey the following set of commutation relations.

[Jx , Jy ] = iJz , +cyclic permutations

[Kx , Ky ] = −iJz , +cyclic permutations
[Jx , Kx ] = [Jy , Ky ] = [Jz , Kz ] = 0
[Jx , Ky ] = iKz , +cyclic permutations (1.2.35)
20 CHAPTER 1. QUANTUM MECHANICS AND SPECIAL RELATIVITY

Note that the first line is just the standard SU (2) algebra for angular momentum opera-
tors.

This commutator algebra reveals a more interesting structure if we introduce the linear
combinations

~ = 1 J~ + iK
 
N ~
2
~ = 1  
M J~ − iK
~ . (1.2.36)
2
Note that these are two 3-vectors where each component of the 3-vectors is a 4-by-4
matrix. These two sets of matrix generators obey the following commutator algebra

[Nx , Mx ] = [Nx , My ] = [Nx , Mz ] = . . . = 0

[Nx , Ny ] = iNz , +cyclic permutations
[Mx , My ] = iMz , +cyclic permutations , (1.2.37)

which are two commuting sets of SU (2) algebras.

Until now we have considered a particular representation of the operators Kx,y,z and
Jx,y,z in terms of 4-by-4 matrices which are the generators of boosts and rotations acting
on 4-vectors (and tensors by multiple action of Λs). But this algebra could be represented
by differential operators (as we are used to from QM) or different sets of 4-by-4 matrices
(as we will find later for the spinors which are spin 1/2 particles). All these have one
thing in common: they obey the same algebra. So let us now consider the operators
abstractly and assume that the only thing we know about them is that they obey the
Lorentz algebra (1.2.37).

Then we know on general grounds that N ~ 2 and M

~ 2 have eigenvalues jN (jN + 1) and
jM (jM + 1) respectively, where jN , jM can be integer or half-integer. This statement is
true for any set of matrices or operators that obey the same algebra as above and is not
restricted to our particular choice of generators for boost and rotations. Representations
(particles) are labelled by (jN , jM ), where the spin of the particle is jN + jM .

Important examples include:

• (0, 0): spin zero, scalar particle

• ( 21 , 0): spin 21 , left handed fermion

• (0, 21 ): spin 21 , right handed fermion

• ( 21 , 12 ): spin 1, transforms like a 4-vector, e.g. photon, gluon, W and Z bosons

1.2. SPECIAL RELATIVITY 21

1.2.5 Poincaré Group

This is a bonus section and will NOT be part of the exam. A nice exposition of the
material in this and the previous section can be found in the book by Ryder [4].

So far we have only considered rotations and generalised rotations i.e. boosts, but
you might have wondered what has happened to translations in space and time that
we discussed in the context of non-relativistic QM. We will remedy this situation now
and introduce the Poincaré Group which is the natural extension of the Lorentz group
including translations in space-time:

x0µ = Λµ ν xν + aµ (1.2.38)

The first part of this transformation are just the LTs, which are generated by Jx,y,z
and Kx,y,z and the second, inhomogenous, part are the translations generated by the
4-momentum pµ .

In QM we are used to expressions for Jbx,y,z or b

p~ in terms of differential operators. For
example p~ → b ~ and E → i∂/∂t which can be combined into the 4-vector
p~ = −i∇

pµ → pbµ = i∇µ . (1.2.39)

 
∂ ∂ ∂ ∂


The QM angular momentum operators are Jbx = −i y ∂z − z ∂y , Jby = −i z ∂x − x ∂z ,
 
∂ ∂
Jbz = −i x ∂y − y ∂x . Similar operators exist for the boosts (however they are not her-
 
∂ ∂ ∂ ∂ bz = i t ∂ + z ∂ .



mitian): Kx = i t ∂x + x ∂t , Ky = i t ∂y + y ∂t , K
b b
∂z ∂t

Notably, this set of operators obeys the same algebra (1.2.35) that we found in the
last section for the generators Jx,y,z and Kx,y,z in matrix form.

These differential operators are defined through their action on functions, while the
matrices act directly on the coordinates which appear as arguments of spacetime depen-
dent functions. So these two viewpoints are naturally related, e.g. consider an infinitesi-
mal boost in the x-direction. Recalling

Λµ ν = δ µ ν + iωx (Kx )µ ν
x0µ = xµ + iωx (Kx )µ ν xν , (1.2.40)

it is easy to see that (infinitesimally)

b x f (x) = f (x0 ) − f (x) ,

iωx K (1.2.41)

where the argument x of f (x) stands for a point in spacetime i.e. a 4-vector.
22 CHAPTER 1. QUANTUM MECHANICS AND SPECIAL RELATIVITY

Another common form of the algebra of the K 0 s and J 0 s is written in terms of the
anti-symmetric tensor M
cµν which is defined as follows:

cij = −M
M cji = ijk Jbk
ci0 = −M
M c0i = −K bi (1.2.42)

where i, j, k = 1 . . . 3. Now the Lorentz algebra becomes

h i  
Mµν , Mρσ = i ηνρ Mµσ − ηµρ Mνσ + ηµσ Mνρ − ηνσ Mµρ .
c c c c c c (1.2.43)

Including translations generated by pbµ = i∂µ one finds the additional commutation rela-
tions
[b
pµ , pbν ] = 0 , (1.2.44)
h i
cρσ = i (ηµρ pbσ − ηµσ pbρ ) .
pbµ , M (1.2.45)
This is the Poincaré algebra (the Poincaré group is obtained by exponentiating the gen-
erators). A mathematical problem with the Lorentz group is that it does not have finite
dimensional unitary representations; the representations we listed in the last section are
finite dimensional but non-unitary! The problem comes from the boosts, whose generators
are not Hermitian and hence their exponentiation gives non-unitary matrices or opera-
tors. But in QM we seek unitary finite dimensional representations and here the Poincaré
group comes to the rescue, since it does have finite dimensional unitary representations
corresponding to particles with various spins.

This result allows us to answer with mathematical rigour the following fundamental
question: What is a particle?
The mathematicians answer: An irreducible representation of the Poincaré group.

In order to classify representations we have to look for operators that commute with
b2
all operators of the Poincaré group. (This is the analogue to the operator J~ for the
angular momentum operators.) It turns out that there exist exactly two such operators,
which are also called Casimir operators of the Poincaré algebra:

C1 = pbµ pbµ , (1.2.46)

which for a momentum eigenstate becomes p2 = E 2 − p~2 = m2 , so this is just the mass
b2
squared. Now interestingly J~ is not a Casimir since it does not commute with boosts K.
~
b
To constructe the second Casimir one introduces the Pauli-Lubanski vector Wcµ

cµ = − 1 µνρσ M
W cνρ Pbσ
2
cµ Pbµ = 0 ,
W (1.2.47)

so that
C2 = W cµ .
cµ W (1.2.48)
1.2. SPECIAL RELATIVITY 23

As an example consider a massive, spin s particle in its rest frame where pµ = (m, ~0).
Now in the rest frame the orbital angular momentum L ~ = 0 so that the total angular
momentum J~ = L ~ +S~ = S.
~ For Wµ we find

1
W0 = − 0ijk J ij P k = 0
2
1 m m
Wi = − iµνρ J µν P ρ = − iµν0 J µν = − ijk J jk = −mJi (1.2.49)
2 2 2
and, hence,
~ 2 = −m2 J~2 = −m2 S
C2 = Wµ W µ = (W0 )2 − W ~ 2 = −m2 s(s + 1) . (1.2.50)

So we find that a relativistic particle is characterised by its mass m and its spin s.

1.2.6 The Maxwell Equation — First Glimpse at a Relativistic

Wave Equation

Remember Maxwell’s equations (where again ~ = c = 1)

~ ·E
∇ ~ = ρ
~ ·B
∇ ~ = 0
~
~ = − ∂B
~ ×E
∇
∂t
~
∇ ~ = ∂ E + ~j
~ ×B (1.2.51)
∂t
This does not look manisfestly Lorentz covariant since it is written in terms of 3-vectors
~ B
E, ~ and ~j.

~ in terms of which B
Let us introduce the 4-vector potential Aµ = (φ, A) ~ =∇
~ ×A
~ and
~
~ = − ∂ A − ∇φ.
~
E ∂t
This implies the 2nd and 3rd Maxwell equation since
~ ·B
∇ ~ =∇
~ · (∇
~ × A)
~ = 0,
~ ~ ~
~ × (− ∂ A − ∇φ)
∇ ~ = −∇ ~ × ∂A = − ∂B . (1.2.52)
∂t ∂t ∂t
Now we define the anti-symmetric electro-magnetic fieldstrength tensor F µν (remember
∂
∂ µ = ( ∂t ~
, −∇):

F µν = −F νµ = ∂ µ Aν − ∂ ν Aµ
~
∂A
F 0i = ∂ 0 Ai − ∂ i A0 = ( ~ i = −Ei
+ ∇φ)
∂t
F ij = ∂ i Aj − ∂ j Ai = −ijk Bk . (1.2.53)
24 CHAPTER 1. QUANTUM MECHANICS AND SPECIAL RELATIVITY

So  
0 −E1 −E2 −E3
 E1 0 −B3 B2 
F µν =  (1.2.54)
 E2 B3 0 −B1 
E3 −B2 B1 0
Under Lorentz transformations
F µν → F 0µν = Λµ α Λν β F αβ (1.2.55)
which can also be written as a matrix equation F → F 0 = Λ · F · ΛT . The second and the
third Maxwell equation are encoded in the so-called Bianchi identity
∂ µ F νρ + ∂ ν F ρµ + ∂ ρ F µν = 0, (1.2.56)
while the first and the fourth Maxwell equation are encoded in the equation
∂µ F µν = j ν (1.2.57)

with j ν = (ρ, ~j). This is an example of an equation that is forminvariant under Lorentz
transformations, i.e. it transforms covariantly under LTs.

An important feature of the electromagnetic fieldstrength tensor F µν is that it is

invariant under gauge transformations. Gauge transformations are redefinition of Aµ by
a total derivative
Aµ → Aµ + ∂ µ χ(x) . (1.2.58)
Proof: under such a gauge transformation
F µν → ∂ µ (Aν + ∂ ν χ(x)) − ∂ ν (Aµ + ∂ µ χ(x)) =
F µν + (∂ µ ∂ ν − ∂ ν ∂ µ )χ(x) = F µν . (1.2.59)
This gauge invariance implies that out of the four degrees of freedom (components) of
Aµ one is redundant and we can impose a convenient gauge condition to eliminate this
redundancy. A popular choice is the so-called Lorentz gauge ∂ · A = ∂µ Aµ = 0 because it
is a Lorentz invariant condition.

In this gauge
∂µ F µν = ∂µ (∂ µ Aν − ∂ ν Aµ ) = Aν − ∂ ν (∂µ Aµ ) = j ν (1.2.60)
becomes
Aν = j ν . (1.2.61)
Furthermore, for j ν = 0 (in vacuo) we obtain
Aν , (1.2.62)
which is a relativistic wave equation for massless spin 1 particles called photons. The
equation implies in particular that the waves/photons are travelling at the speed of light
as any massless particle does (independent of its spin).
Chapter 2

Relativistic Quantum Mechanics

2.1 The Klein-Gordon Equation

~2
p
In non-relativistic QM, the free Hamiltonian H = E = 2m
is quantised by the substitution
∂ ~
H → i~ , p~ → −i~∇ (2.1.1)
∂t
to give the Schrödinger equation
∂Ψ ~2 ~ 2
i~ =− ∇ Ψ. (2.1.2)
∂t 2m

A relativistic free particle has Hamiltonian

p
H = E = p~2 c2 + m2 c4 (2.1.3)
and hence the same, naive substitution gives
q
∂Ψ ~ 2Ψ .
i~ = m2 c4 − ~2 c2 ∇ (2.1.4)
∂t
But what to do about the square root of the operator? One interpretation is to make a
series expansion, but then we get a Hamiltonian with derivatives of arbitrarily high order.

A more sensible route is to start from H 2 = p~2 c2 + m2 c4 to get

∂2  
~ 2 + m2 c4 Ψ
−~2 2 Ψ = −~2 c2 ∇
∂t
1 ∂2
   mc 2
⇒ − ~
∇ 2
Ψ + Ψ=0
c2 ∂t2 ~
  mc 2 
⇒ + Ψ=0 (2.1.5)
~

25
26 CHAPTER 2. RELATIVISTIC QUANTUM MECHANICS

By analogy with the Schrödinger equation it is possible to derive a continuity equation

for the Klein-Gordon (KG) equation
∂ρ ~ ~
+∇·J =0 (2.1.6)
∂t
where ρ is the probability density and J~ is the probability current.1

Derivation: Subtracting the following two equations

  mc 2 
∗
Ψ + Ψ=0
~
  mc 2 
Ψ + Ψ∗ = 0 (2.1.7)
~
gives
Ψ∗ ∂µ ∂ µ Ψ − Ψ∂µ ∂ µ Ψ∗ = 0 . (2.1.8)
Furthermore, this implies
∂µ (Ψ∗ ∂ µ Ψ − Ψ∂ µ Ψ∗ ) = 0
 
∂ 1 ∗ ∗ ~

∗~ ~ ∗

⇒ (Ψ ∂t Ψ − Ψ∂ t Ψ ) − ∇ · Ψ ∇Ψ − Ψ∇Ψ = 0. (2.1.9)
∂t c2
After multiplying through by ic2 , in order to make ρ real, we can write this as desired as
a continuity equation with
 
ρ = i (Ψ∗ ∂t Ψ − Ψ∂t Ψ∗ ) and J~ = −ic2 Ψ∗ ∇Ψ
~ − Ψ∇Ψ ~ ∗ . (2.1.10)

Now because of the negative sign between the two terms in ρ, the probability density can
both take positive and negative values (in contrast to non-relativistic QM where ρ = |Ψ|2
is positive definite)! This is an absurd and nonsensical result for a probability density!!

Junk the KG equation for the moment and try harder. Schrödinger was the first to
write down this relativistic wave equation, but discarded it for a different reason; the
spectrum is not bounded from below.

2.2 The Dirac Equation

Let us go back to our starting point

b = i~ ∂Ψ
HΨ (2.2.1)
∂t
Integrating the probability density over a volume V bounded by the surface S we find V ∂ρ
1 3
R
∂t d x =
d 3 ~ ~ 3 ~ ~
R R R
dt V ρd x = − V ∇ · jd x = − S j · dS. This implies that probability cannot be created or destroyed;
it can only flow from one point to another.
2.2. THE DIRAC EQUATION 27

and try to give a meaning to the square root in

q q
b = m2 c4 + b
H ~2.
p~c2 = m2 c4 − ~2 c2 ∇ (2.2.2)

∂
Since the equation is linear in ∂t
, Lorentz covariance suggests it should be linear also
in the ∂x∂ i , i = 1, 2, 3.

So we write

H
b = c~ p~ + βmc2
α·b
= −i~c~α·∇~ + βmc2 , (2.2.3)

where αi and β are coefficients to be determined. More explicitly this equation can be
written  
1 ∂ 2 ∂ 3 ∂
b = −i~c α
H 1
+α 2
+α 3
+ βmc2 . (2.2.4)
∂x ∂x ∂x
Now we determine the coefficients αi and β by requiring that this linear operator ”squares”
to the KG operator
H ~ 2 + m2 c4 .
b 2 = −~2 c2 ∇ (2.2.5)
We find

∂2 2 2
 
b2 2 2 1 2 2 2 ∂ 3 2 ∂
H = −~ c (α ) + (α ) + (α ) + β 2 m 2 c4
∂(x1 )2 ∂(x2 )2 ∂(x3 )2
 
∂
−i~mc3 (α1 β + βα1 ) 1 + . . .
∂x
∂2
 
2 2 1 2 2 1
−~ c (α α + α α ) 1 2 + . . . . (2.2.6)
∂x ∂x

Thus we need to solve

(αi )2 = I , i = 1, 2, 3
αi αj + αj αi = 0 , i 6= j
β2 = I
αi β + βαi = 0 , i = 1, 2, 3 , (2.2.7)

where I denotes a unit matrix (if a subscript is added it denotes the dimensionality, e.g.
I2 denotes a 2 × 2 unit matrix).

It is obviously NOT possible to solve those equations if the coefficients are simply
complex numbers. So let us assume that they are N ×N matrices. With some (guess)work
it can be shown that the smallest value of N for which eq. (2.2.7) can be solved is N = 4.
28 CHAPTER 2. RELATIVISTIC QUANTUM MECHANICS

This implies that the Dirac wave function is a 4-component column vector
 
Ψ1 (x)
 Ψ2 (x) 
Ψ(x) =  Ψ3 (x) 
 (2.2.8)
Ψ4 (x)

where x ≡ (x0 , ~x) and the Dirac equation becomes a matrix equation

∂Ψ ~ + βmc2 )Ψ .
i~ = HΨ
b = (c~ p~ + βmc2 )Ψ = (−i~c~
α·b α·∇ (2.2.9)
∂t
This is a set of 4 first order linear differential equations to determine Ψ1 , . . . , Ψ4 .

2.3 Representation of the Dirac Matrices

A particular set of solutions of (2.2.7) for the 4 × 4 matrices αi , β can be written with
the help of the 2 × 2 Pauli matrices σ i , i = 1, 2, 3,
     
1 0 1 2 0 −i 3 1 0
σ = , σ = , σ = (2.3.1)
1 0 i 0 0 −1

which obey the following identities

2
σ i = I2 , i = 1, 2, 3
σ i σ j + σ j σ i = 0 , i 6= j . (2.3.2)

We may satisfy the first two lines in eq. (2.2.7) by taking αi to be the 4 × 4 matrices
 
i 0 σi
α = , i = 1, 2, 3 . (2.3.3)
σi 0

Now we may satisfy the remaining two lines in eq. (2.2.7) by taking
 
I2 0
β= . (2.3.4)
0 −I2

Because the σ i are Hermitian, so are the αi and β, i.e.

(αi )† = αi , β † = β . (2.3.5)

(† ↔ complex conjugate transposed)

2.4. PROBABILITY DENSITY FOR THE DIRAC EQUATION 29

2.4 Probability Density for the Dirac Equation

The Dirac equation is given by

∂Ψ ~ + mc2 βΨ
i~ = −i~~
α · ∇Ψ (2.4.1)
∂t
and its Hermitian conjugate is

∂Ψ† ~ †·α
−i~ = i~∇Ψ ~ + mc2 Ψ† β . (2.4.2)
∂t

(Recall that αi and β are hermitian and (AB)† = B † A† .)

Now take Ψ† ×(Dirac eqn.) and (Hermitian conjugate eqn.)×Ψ and subtract the two
to obtain:
∂(Ψ† )
 
† ∂Ψ
 
i~ Ψ + Ψ = −i~c Ψ† α ~ + ∇(Ψ
~ · ∇Ψ ~ †) · α
~Ψ . (2.4.3)
∂t ∂t
Dividing this equation by i~ we obtain a Continuity Equation

∂ρ ~ ~
+∇·J =0 (2.4.4)
∂t
with the positive definite probability density given by

4
X 4
X
†
ρ=ΨΨ= Ψ∗k Ψk = |Ψk |2 > 0 , (2.4.5)
k=1 k=1

and the probability current

J~ = cΨ† α
~Ψ . (2.4.6)

2.5 Extreme Non-Relativistic Limit of the Dirac Equa-

tion

For a particle at rest (~p = 0) the Dirac equation becomes

∂Ψ
i~ = βmc2 Ψ
∂t
∂Ψ mc2
⇒ = −i βΨ . (2.5.1)
∂t ~
30 CHAPTER 2. RELATIVISTIC QUANTUM MECHANICS
 
I2 0
Taking β = the four equations for the components of Ψ turn into
0 −I2
∂Ψ1 mc2
= −i Ψ1
∂t ~
∂Ψ2 mc2
= −i Ψ2
∂t ~
∂Ψ3 mc2
= +i Ψ3
∂t ~
∂Ψ4 mc2
= +i Ψ4 (2.5.2)
∂t ~
mc2
⇒ Ψ1 = c1 e−i ~
t
e.t.c. (2.5.3)
where c1 is an arbitrary constant.

Thus the general solution takes the form

mc2
 
c1 e−i ~ t
2
 c2 e−i mc~ t
 

Ψ= 
mc2
 (2.5.4)
 c3 ei ~ t


mc2
c4 ei ~
t

which can be rewritten as

   
c1 0
mc2  c2  2 
0 
i mc
Ψ = e−i ~ t 
 0 +e
 ~
t  (2.5.5)
 c3 
0 c4

By acting with the Hamiltonian operator H b = i~ ∂ we find that the first term in the
∂t
2
solution (2.5.5) carries positive energy (+mc ) whereas the second term carries negative
energy (−mc2 ).

Although we found a positive probability density (contrary to the KG equation) we

find that also the Dirac equation has both positive and negative energy solutions. We
shall later interpret the negative energy part as due to Anti-particles.

2.6 Spin of the Dirac Particles

The (free) Dirac equation is

i~∂t Ψ = HΨ
b with H
b = c~ p~ + βmc2
α·b (2.6.1)
2.7. THE COVARIANT FORM OF THE DIRAC EQUATION 31

~
p~ = −i~∇.
with b

Consider the total angular momentum operator

 
~~ ~σ 0
J~ = L
~ +S
~ = ~x × b
p~ + Σ ~ =
where Σ (2.6.2)
2 0 ~σ

Using the Uncertainty Principle

[xi , pbj ] = i~δij , (2.6.3)

it can be shown with some effort that

[H, ~ = i~c(~
b S] α×b
p~) (2.6.4)

and
[H, ~ = −i~c(~
b L] α×b
p~) , (2.6.5)
b~
thus, [H,
b J]=0. So J~ is a conserved quantity which we interpret as the total angular
momentum.

The 3-component of Spin, Sz , is

~ 3
S3 = Σ (2.6.6)
2
 
1 0 0 0
 0 −1 0 0 
 0 0 1 0  which has eigenvalues 2 , − 2 ,
is the matrix ~2  and − ~2 .
 ~ ~ ~
2

0 0 0 −1

1
⇒ we are describing spin 2
particles (and anti-particles).

2.7 The Covariant Form of the Dirac Equation

Multiply the Dirac equation

3
∂Ψ X ∂Ψ
i~ = −i~c αi i + βmc2 Ψ (2.7.1)
∂t i=1
∂x
β
with c
to obtain !
3
∂Ψ X ∂Ψ
i~ β + βαi i = mcΨ . (2.7.2)
∂(ct) i=1 ∂x
32 CHAPTER 2. RELATIVISTIC QUANTUM MECHANICS

Next define the matrices

γ 0 = β , γ i = βαi , i = 1, 2, 3 (2.7.3)
which are also called Gamma-Matrices and allow us to rewrite the Dirac equation as
3
!
∂Ψ X ∂Ψ
i~ γ 0 0 + γ i i Ψ = mcΨ (2.7.4)
∂x i=1
∂x

→ i~γ · ∂Ψ = mcΨ (2.7.5)

 
where ∂ µ ≡ ∂ ~
, −∇ as defined in section 2.1 and
∂x0

γ µ ≡ γ 0 , ~γ = γ 0 , γ 1 , γ 2 , γ 3 ,



(2.7.6)

which makes the Dirac equation now look Lorentz covariant.

Dirac or Feynman Slash Notation

For any 4-vector Aµ , we define A ~ Then, the Dirac equation is

/ ≡ γ · A = γ 0 A0 − ~γ · A.

i~/
∂ Ψ = mcΨ (2.7.7)

or
p/bΨ = mcΨ (2.7.8)
⇒ (/
pb − mc) Ψ = 0 (2.7.9)
where pbµ = i~∇µ .

2.8 Properties of the γ-Matrices

Using the properties of the αi and β we may show that the gamma-matrices obey the
following anti-commutation identity

{γ µ , γ ν } ≡ γ µ γ ν + γ ν γ µ = 2g µν I4 , (2.8.1)
 
1 0 0 0
 0 −1 0 0 
with g µν =  .
 0 0 −1 0 
0 0 0 −1

In particular, (γ 0 )2 = I4 , (γ i )2 = −I4 , i = 1, 2, 3.

Furthermore (γ 0 ) = β † = β i.e. γ 0 is Hermitian, whereas (γ i )† = (βαi )† = (αi )† β † =

α β = −βαi = −γ i i.e. γ i is anti-Hermitian, i = 1, 2, 3.
i
2.9. THE DIRAC EQUATION AND LORENTZ TRANSFORMATIONS 33

Using the explicit matrices αi and β of Section 2.4 we find

   
0 I2 0 i 0 σi
γ = , γ = . (2.8.2)
0 −I2 −σ i 0

2.9 The Dirac Equation and Lorentz Transformations

In the previous sections we have introduced the covariant form of the Dirac equation

(iγ µ ∂µ − m) Ψ(x) = 0 , (2.9.1)

without actually justifying that name. Let us clarify this point now.

We start with two comments:

1. A Lorentz transformatin takes

Ψ(x) → Ψ0 (x0 ) = Ψ0 (Λ · x) . (2.9.2)

The Dirac wave function has four components but it does not transform like a
4-vector. Nevertheless we expect the transformation to be linear.

2. In the primed system Ψ0 (x0 ) should obey an equation that has the same form as the
Dirac equation in the un-primed system.
 
µ ∂
ie
γ − m Ψ0 (x0 ) = 0 , (2.9.3)
∂x0µ

where the γ eµ = U † γ µ U with

eµ are related to γ µ by a unitary transformation γ
U † = U −1 and obey {eγµ, γ
eν } = 2I4 η µν .

From the first comment we see

Ψ0 (x0 ) = Ψ0 (Λ · x) = S(Λ)Ψ(x) = S(Λ)Ψ(Λ−1 · x0 ) (2.9.4)

or
Ψ(x) = S −1 (Λ)Ψ0 (x0 ) = S −1 (Λ)Ψ0 (Λ · x) . (2.9.5)

Combining eqns. (2.9.1) and (2.9.5) we can find an equation for Ψ0 (x0 )

(iγ µ ∂µ − m) S −1 (Λ)Ψ0 (x0 ) = 0 . (2.9.6)

34 CHAPTER 2. RELATIVISTIC QUANTUM MECHANICS

Multiplying this equation from the left with S we find

iSγ µ S −1 ∂µ − m Ψ0 (x0 ) = 0 .


(2.9.7)
∂x0ν ∂
Now ∂µ = ∂xµ ∂x0ν
= Λν µ ∂x∂0ν = Λν µ ∂ν0 and, hence, we get

iSγ µ S −1 Λν µ ∂ν0 − m Ψ0 (x0 ) = 0 .

(2.9.8)

Therefore, if we are able to find an S(Λ) such that

Λν µ Sγ µ S −1 = γ ν (2.9.9)

then we would have proven form invariance (or commonly called convariance) of the Dirac
equation! We can write this condition also as

S −1 γ ν S = Λν µ γ µ . (2.9.10)

So let us consider infinitesimal Lorentz Transformations

Λν µ = δ ν µ − iδω ν µ , (2.9.11)

and write the transformation matrix acting on the Dirac spinor also in infinitesimal form

S = I4 − iδσ . (2.9.12)

Its inverse is
S −1 = I4 + iδσ . (2.9.13)
Plugging these expressions into (2.9.10) we find

(I4 + iδσ) γ ν (I4 − iδσ) = (δ ν µ − iδω ν µ ) γ µ

⇒ [γ ν , σ] = ω ν µ γ µ (2.9.14)

Solutions for this equation are easy to find and well known. For ω ν µ = Kx,y,z ; Jx,y,z
we find σ(Kx ), σ(Ky ), . . . , σ(Jz ) with
 
i  0 i i 0 σi
σ(Ki ) = γ ,γ = , (2.9.15)
4 2 σi 0
 
i  j k  1 σi 0 1
σ(Ji ) = ijk γ , γ = = Σi . (2.9.16)
8 2 0 σi 2

Note that the σ(Ki ) and σ(Ji ) obey the same commutation relations as the Ki and
Ji . They correspond to a particular representation of the Lorentz group called the spinor
2.9. THE DIRAC EQUATION AND LORENTZ TRANSFORMATIONS 35

representation. The σ(Ki ) and σ(Ji ) correspond to the vector representation which acts
by generalised rotations on Lorentz 4-vectors.

Finite LTs for Dirac spinors are again obtained by exponentiating the infinitesimal
LTs:
S = exp(−iδσ) , (2.9.17)

so that under a LT
Ψ → SΨ . (2.9.18)

Rotation of Dirac fermion:

S = exp(−i(φx σ(Jx ) + φy σ(Jy ) + φz σ(Jz )))

!
i~
~ · Σ/2)
~ exp(− 2
φ · ~
σ ) 0
= exp(−iφ = ~ · ~σ ) (2.9.19)
0 exp(− 2i φ

For a rotation around the z-axis φx = φy = 0 and φz = φ this becomes

 
exp(−iφ/2) 0 0 0
 0 exp(+iφ/2) 0 0 
.
 (2.9.20)
 0 0 exp(−iφ/2) 0 
0 0 0 exp(+iφ/2)

In particular if φ = 2π, S = −I4 , i.e. under a 2π-rotation Ψ → −Ψ and a Dirac fermion

is only invariant under a 4π rotation. This is in contrast to a 4-vector that comes back
to itself after a 2π rotation. One consequence of this is that physical observables always
contain an even number of fermions which is invariant under 2π rotations.

Boost of Dirac fermion: Consider a boost in the z direction

S = exp(−iωσ(Kz ))
0 ω2 σz
 
= exp ω
σ
2 z
0
 
cosh(ω/2) 0 sinh(ω/2) 0
 0 cosh(ω/2) 0 − sinh(ω/2) 
= 
 sinh(ω/2)
 (2.9.21)
0 cosh(ω/2) 0 
0 − sinh(ω/2) 0 cosh(ω/2)

Acting with this S on a plane wave solution with p~ = (0, 0, 0) we should obtain a plane
wave solutionqwith p~ = (0, 0, pz ). In order to show this it is helpful to use the identities
and sinh(ω/2) = √ pz
p
cosh(ω/2) = E+m 2m
with E = m2 + p2z . (Homework)
2m(E+m)
36 CHAPTER 2. RELATIVISTIC QUANTUM MECHANICS

2.10 Plane Wave Solutions of the Dirac Equation

From now on we will work in natural units ~ = c = 1. We look for plane wave solutions
of the Dirac equation of the form
 
∓ip·x φ
Ψ=e (2.10.1)
χ

where φ are the upper two components and χ the lower two components of Ψ and p · x =
Et − p~ · ~x, with E > 0.

The factor e−ip·x gives solutions with positive energy E and momentum p~, and the
factor e+ip·x gives solutions with negative energy −E and momentum −~p. Substituting
back into the Dirac equation

~
 ∂Ψ
−i~
α · ∇ + βm Ψ = i , (2.10.2)
∂t
and using ∂t Ψ = ∓iEΨ and ∂x Ψ = ±ipx Ψ e.t.c., we obtain

α · p~ + βm)Ψ = i(∓iE)Ψ
(−i(±i)~
→ (±~α · p~ + βm)Ψ = ±EΨ . (2.10.3)

If we use the standard representation for the αi and β from Section 2.4 we obtain
    
mI ±~σ · p~ φ φ
= ±E (2.10.4)
±~σ · p~ −mI χ χ

which gives the coupled set of equations

(m ∓ E)φ ± ~σ · p~χ = 0
±~σ · p~φ − (m ± E)χ = 0 . (2.10.5)

We will construct the solutions in such a way that they have a straighforward p~ = 0 limit.

Positive Energy Solutions

(m − E)φ + ~σ · p~χ = 0
~σ · p~φ − (m + E)χ = 0 . (2.10.6)

For p~ = 0, E = m and χ = 0 (in agreement with Section 2.6).

2.10. PLANE WAVE SOLUTIONS OF THE DIRAC EQUATION 37

For p~ 6= 0 it is convenient to solve for χ in terms of φ using the second equation in

(2.10.6), i.e.
~σ · p~
χ= φ (2.10.7)
(E + m)
then,
 
−ip·x φ
Ψ=e σ ·~
~ p . (2.10.8)
(E+m)
φ

Note that the first equation in (2.10.6) only gives the on-(mass)shell condition E 2 =
p~2 + m2 .
   
1 0
We can write φ in terms of φ1 = and φ2 = .
0 1

There are thus two independent positive energy solutions

Ψ = e−ip·x U (p, s) , s = 1, 2 (2.10.9)

where
√
 
φs
U (p, s) = E+m σ ·~
~ p (2.10.10)
(E+m) s
φ
is a positive energy Dirac spinor. In the last expression a convenient normalization factor
has been introduced.

It can be checked that the first equation in (2.10.6) is automatically satisfied by using
the identity (~σ · p~)2 = p~2 I2 .

Negative Energy Solutions

(m + E)φ − ~σ · p~χ = 0
−~σ · p~φ − (m − E)χ = 0 . (2.10.11)

For p~ = 0, E = m and φ = 0 (in agreement with Section 2.6).

For p~ 6= 0 it is convenient to solve for φ in terms of χ using the first equation in

(2.10.11), i.e.
~σ · p~
φ= χ (2.10.12)
(E + m)
then,
 σ ·~
~ p 
+ip·x (E+m)
χ
Ψ=e . (2.10.13)
χ
38 CHAPTER 2. RELATIVISTIC QUANTUM MECHANICS

Note that the second equation in (2.10.11) only gives the on-(mass)shell condition E 2 =
p~2 + m2 .
   
0 1
We can write χ in terms of χ1 = and χ2 = .
1 0

There are thus two independent negative energy solutions

Ψ = e+ip·x V (p, s) , s = 1, 2 (2.10.14)

where
√ σ ·~
~ p
 
(E+m) s
χ
V (p, s) = E+m (2.10.15)
χs

is a negative energy Dirac spinor. In the last expression a convenient normalization factor
has been introduced.

It can be checked that the second equation in (2.10.11) is automatically satisfied by

using the identity (~σ · p~)2 = p~2 I2 .

Interpretation

To find the physical interpretation for the four independent solutions we consider the
rest frame p~ = 0. Then:
 
1
√ √
 
φ1  0 
U (p, 1) = 2m = 2m  
0  0 
0
 
0
√ √
 
φ2  1 
U (p, 2) = 2m = 2m  
0  0 
0
 
0
√ √
 
0  0 
V (p, 1) = 2m = 2m  
χ1  0 
1
 
0
√ √
 
0  0 
V (p, 2) = 2m = 2m   (2.10.16)
χ2  1 
0

~ = ~x × p~ = 0 so that the total angular momentum

Furthermore, for p~ = 0 we have L
2.11. PROPERTIES OF SOLUTIONS 39

operator becomes
1~
J~ = L
~ +S~=S ~= Σ
 1 2 
σ
2 3
0
⇒ Sz = 1 (2.10.17)
0 σ
2 3

Thus, U (p, 1) and U (p, 2) are positive energy solutions with Sz eigenvalues sz = +1/2
and sz = −1/2 respectively, whereas V (p, 1) and V (p, 2) are negative energy solutions
with Sz eigenvalues sz = −1/2 and sz = +1/2 respectively.

In general, U and V are the Lorentz boosts of these solutions to a frame where p~ 6= 0.
Interpret the negative energy solutions later.

2.11 Properties of Solutions

Since e−ip·x U (p, s) is a solution of the Dirac equation

(iγ · ∂ − m)(e−p·x U (p, s)) = 0
 
0 ∂ ~
→ iγ + i~γ · ∇ − m e−i(Et−~p·~x) U (p, s) = 0
∂t
→ (Eγ 0 − p~ · ~γ − m)U = 0
→ (p · γ − m)U = 0
⇒ (p/ − m)U (p, s) = 0 (2.11.1)
i.e. U (p, s) obeys the Dirac equations with pbµ simply replaced by pµ . Similarly, we find
(p/ + m)V (p, s) = 0 . (2.11.2)

We will often make use of the adjoint spinors which are defined as
U (p, s) ≡ U † (p, s)γ 0 , V (p, s) ≡ V † (p, s)γ 0 . (2.11.3)
By taking the Hermitian adjoint of the equation obeyed by U and V we find
U (p, s)(p/ − m) = 0 , V (p, s)(p
/ + m) = 0 . (2.11.4)

One may also check directly (using again (~σ · p~)2 = p~2 I2 ) that
U † (p, s)U (p, s) = V † (p, s)V (p, s) = 2E , s = 1, 2 , (2.11.5)
and
U (p, s)U (p, s) = 2m , V (p, s)V (p, s) = −2m , s = 1, 2 . (2.11.6)
40 CHAPTER 2. RELATIVISTIC QUANTUM MECHANICS

2.12 Anti-Particles — Hole Theory

Since the Dirac equation has negative energy solutions, why do positive energy electrons
not radiate energy and fall into a negative energy state? Dirac: Negative energy states
are completely filled and the Pauli exclusion principle (which applies to fermions) forbids
the transition. Consequently, the Vacuum is a state with all positive energy states empty
but all negative energy states filled.

(picture goes here)

If a photon excites a negative energy e− of energy −|E2 | into a positive energy e−

of energy |E1 |, we observe the production of an e− of mass m, charge −|e| and energy
|E1 |, and a Hole in the negative energy sea (Pair production). Note that there is a gap
of 2mc2 between the negative and positive energy states and, hence, the photon energy
hν = |E1 | + |E2 | must be larger than 2mc2 for this to happen. The hole appears as a
particle of mass m, charge +|e| and energy +|E2 |.
⇒ The existence of the Positron (and Anti-particles in general) is predicted!

(picture goes here)

The absence of a spin-up electron of energy −|E| and momentum −~p is equivalent to
the presence of a spin-down positron of energy +|E| and momentum +~p. (Think about
time running backwards or the arrow in a Feynman diagram reversed)

(picture goes here)

Thus, the electron wavefunction eip·x V (p, s) corresponding to energy −E and momen-
tum −~p describes a positron of energy +E and momentum +~p. Also, V (p, 1) and V (p, 2)
which describe spin down and spin up negative energy electrons must describe spin up
and spin down positrons.

2.13 Vacuum Polarization

In general the infinite negative charge of the vacuum produces no effect because the
distribution of charge is homogeneous.

However, consider the effect of a positive energy electron with charge −|e| on the
vacuum. It repels the negative energy electrons and electrically polarises the vacuum.
Thus the physical charge −|e| seen by a test charge at a large distance from the electron
is numerically smaller than the bare charge −|e0 |, i.e. |e| < |e0 |.
2.14. CHARGE CONJUGATION SYMMETRY C 41

(picture goes here)

However, if the test charge comes very close it will see the bare charge −|e0 |. For
S-wave electrons (l = 0) in an atom, the proton sees a charge numerically greater than
the ordinary electric charge |e|. Note that for l > 0 the wavefunction vanishes at the
origin and the proton feels a numerically smaller charge. This effect leads to measurable
shifts of the energy levels of atoms.

2.14 Charge Conjugation Symmetry C

We construct an operator acting on the Dirac wave function

C : Ψ → ΨC (2.14.1)

which turns a positive energy electron wavefunction (e− ) into a negative energy wave
function (e+ ) with the same momentum and spin state. If Ψ = e−ip·x U (p, s) then ΨC =
eip·x V (p, s). The required operation turns out to be

C : Ψ → ΨC = Cγ 0 Ψ∗ , (2.14.2)

where C = iγ 2 γ 0 . Useful properties: C † = −C, C 2 = −I, C −1 = −C and Cγ µ C = (γ µ )T .

A symmetry of a wave equation is an operation on a wave function Ψ → Ψ0 and on the

space-time coordinates x → x0 such that Ψ0 obeys the same equation as Ψ, with x replaced
by x0 .

Ψ → ΨC , x → x0 can be shown to be a symmetry of the Dirac equation as follows:

Proof: We claim that the Dirac equation is charge conjugation invariant. The Dirac
equation may be written as
(iγ µ ∂µ − m)Ψ = 0, (2.14.3)
taking the complex conjugate gives

(−i(γ µ )∗ ∂µ − m)Ψ∗ = 0 . (2.14.4)

Now multiply from the left with Cγ 0

(−iCγ 0 (γ µ )∗ ∂µ − m)Ψ∗ = 0
→ (iγ µ (Cγ 0 ))∂µ − mcγ 0 )Ψ∗ = 0
→ (iγ µ ∂µ − m)ΨC = 0 , (2.14.5)

where we have used the identity Cγ 0 (γ µ )∗ = −γ µ (Cγ 0 ). This shows that ΨC obeys the
same equation as Ψ.
42 CHAPTER 2. RELATIVISTIC QUANTUM MECHANICS

2.15 Space Inversion P

The Dirac equation is also invariant under reflection of space coordinates in the origin

P : ~x → ~x0 = −~x , t → t0 = t . (2.15.1)

The corresponding operation2 on Dirac spinors is

P : Ψ → Ψ0 = P Ψ (2.15.2)

with P = γ 0 . It can be checked by direct calculation that P U (~p, s) = U (−~p, s) i.e.

p~ → −~p as expected for space inversion, but the spin state is unchanged. Also P V (~p, s) =
−V (−~p, s). The −1 factor indicates that anti-particles have opposite Parity to particles.

In this case, to check invariance of the Dirac equation, it is necessary to replace ∂x ,

∂y and ∂z by −∂x , −∂y and −∂z , as well as replacing Ψ by Ψ0 , i.e. Ψ0 obeys the same
equation as Ψ with ∂x , ∂y and ∂z replaced by −∂x , −∂y and −∂z .

2.16 Time Reversal T

The Dirac equation also has a symmetry under time reversal

T : t → t0 = −t , ~x → ~x0 = ~x , (2.16.1)

the appropriate transformation of Ψ is

T : Ψ → Ψ0 = T ψ ∗ (2.16.2)

with T = −γ 1 γ 3 . It can be checked directly that this is the correct transformation by

showing that
T U ∗ (~p, 1) = +U (−~p, 2) , T V ∗ (~p, 1) = −V (−~p, 2) (2.16.3)
Thus, the transformation changes a solution of the Dirac equation with momentum p~ and
spin up into a solution with momentum −~p and spin down.

~ = ~x × p~, and
p
This is as expected for time reversal, since p~ = m~v / 1 − ~v 2 /c2 and L
~ → −L
thus under time reversal p~ → −~p and L ~ and in particular Lz → −Lz . We assume
that this applies to any AM operator, so that in particular Sz → −Sz . In this case, Ψ0
obeys the same equation as Ψ with ∂t replaced by −∂t .
2
Since P also transforms the space time coordinates this operation should be written more properly
as Ψ(t, ~x) → Ψ0 (t0 , ~x0 ) = Ψ0 (t, −~x) = P Ψ(t, ~x). Hence Ψ0 (t, ~x) = P Ψ(t, −~x); a similar comment applies
to time reversal T .
2.17. DIRAC COVARIANTS 43

2.17 Dirac Covariants

It is important in the study of the Weak Interactions to know the properties of objects
like Ψγ µ Ψ, Ψγ µ γ5 Ψ, etc, where we introduced
 
0 1 2 3 0 I
γ5 ≡ iγ γ γ γ = . (2.17.1)
I 0

We list here the behaviour of some of the Dirac covariants under Lorentz transforma-
tions, P, C:

Covariant LT’s P C
ΨΨ scalar +ΨΨ −ΨΨ
Ψγ5 Ψ pseudoscalar −Ψγ5 Ψ −Ψγ5 Ψ
Ψγ µ Ψ 4-vector +Ψγ 0 Ψ +Ψγ µ Ψ
−Ψγ i Ψ
Ψγ µ γ5 Ψ (pseudo) 4-vector −Ψγ 0 γ5 Ψ −Ψγ µ γ5 Ψ
+Ψγ i γ5 Ψ

(γ5 has the properties {γ5 , γ µ } = 0, γ5† = γ5 )

For example under P the behaviour of the vector current is

0
Ψγ µ Ψ → Ψ γ µ Ψ0
= (Ψ0 )† γ 0 γ µ Ψ0
= Ψ† (γ 0 )† γ 0 γ µ γ 0 Ψ
= Ψ† γ 0 γ 0 γ µ γ 0 Ψ
= Ψγ 0 γ µ γ 0 Ψ (2.17.2)

Thus,
Ψγ 0 Ψ → Ψγ 0 γ 0 γ 0 Ψ = Ψγ 0 Ψ (2.17.3)

Ψγ i Ψ → Ψγ 0 γ i γ 0 Ψ
= −Ψ(γ 0 )2 γ i Ψ
= −Ψγ i Ψ (2.17.4)

In the Relativistic version of Time Dependent Perturbation Theory (Feynman Dia-

grams) the probability amplitudes for Electromagnetic Scattering of 2 particles via photon
44 CHAPTER 2. RELATIVISTIC QUANTUM MECHANICS

exchange contains a factor

U (p2 , s2 )γµ U (p1 , s1 )U (p4 , s4 )γ µ U (p3 , s3 )

X3
= U γ 0U U γ 0U − U γ iU U γ iU , (2.17.5)
i=1

which is invariant under both P and C because the two negative signs for the U γ i U
cancel for space inversion. Thus the Electromagnetic interactions are both space reflection
invariant and charge conjugation invariant.

(picture goes here)

The corresponding probability amplitude for the Weak Interaction has a factor

U (p2 , s2 )γµ (I − γ5 )U (p1 , s1 )U (p4 , s4 )γ µ (I − γ5 )U (p3 , s3 )

= U γµ U U γ µ U + U γµγ5 U U γ µ γ5 U
−U γµ γ5 U U γ µ U − U γµ U U γ µ γ5 U . (2.17.6)

The term

U γµ γ5 U U γ µ U
3
X
= U γ 0 γ5 U U γ 0 U − U γ i γ5 U U γ i U (2.17.7)
i=1

changes sign under both P and C transformations, because Ψγ µ Ψ and Ψγ µ γ5 Ψ transform

with opposite signs both for µ = 0 and µ = i. Thus, the weak interactions break both
space inversion and charge conjugation invariance. This manifests itself in the angular
dependence of scattering processes (e.g. cos θ changes sign under space inversion: 0 →
π − 0). Note however that the combined action of C and P, CP, is a symmetry of this
interaction.

Note, that in a general theory C, P and T are not preserved, but the combination of
the three transformations CPT is always a symmetry.

2.18 Neutrinos

Some modification of RQM is needed in the physically important case of massless spin-1/2
particles — Neutrinos (with todays experimental evidence of Neutrino oscillations this is
not quite true, nevertheless it is a very good approximation.)
2.18. NEUTRINOS 45

First, define the Helicity of a particle as the component of its AM J~ = L

~ + Σ/2
~ in its
direction of motion. For a Dirac particle,
p~ ~ p~
Σ
Helicity = J~ · = · (2.18.1)
|~p| 2 |~p|
~ · p~ = (~x × p~) · p~ = 0.
because L

Experimental observation shows that whereas an e− can have Helicity +1/2 or −1/2,
a Neutrino (which is massless) can only have Helicity −1/2 and an Anti-Neutrino can
only have Helicity +1/2.

Thus, whereas we need four degrees of freedom to describe the 2 spin states of an
electron or positron, we need only 2 degrees of freedom to describe the spin states of the
neutrino and anti-neutrino. We need to discard 2 spin states of the Dirac particle.

Now we return to the Dirac equation for a positive energy solution of energy E and
momentum p~. However, we choose a different representation of the Dirac matrices (and
hence a different representation of the gamma-matrices). This does not effect the physics
but makes the proof much easier. It may be checked that
   
0 I ~σ 0
β= , α
~= (2.18.2)
I 0 0 −~σ
also obey the Dirac Algebra (2.2.7).

−ip·xφ
For Ψ = e α · p~ + βm)Ψ = EΨ (see Section
the Dirac equation reduces to (~
χ
2.11), from which we get
    
~σ · p~ mI φ φ
=E , (2.18.3)
mI −~σ · p~ χ χ
which gives the coupled equations
~σ · p~φ + mχ = Eφ
mφ − ~σ · p~χ = Eχ . (2.18.4)
Now taking m = 0 for a massless neutrino decouples the two equations,
~σ · p~φ = Eφ
~σ · p~χ = −Eχ , (2.18.5)
and since E = |~p| for m = 0,
~σ · p~ 1
φ = φ
2|~p| 2
~σ · p~ 1
χ = − χ. (2.18.6)
2|~p| 2
46 CHAPTER 2. RELATIVISTIC QUANTUM MECHANICS

Thus the upper 2 components of Ψ describe Helicity 1/2, and the lower two describe
helicity −1/2 when Ψ has positive energy. To obtain an appropriate Ψ to describe a
Neutrino we perform a projection that removes the upper 2 components.

This may be achieved by using γ5 = iγ 0 γ 1 γ 2 γ 3 . With the above choice of α

~ and β,
   
0 I 0 −σ i
γ0 = β = , γ i = βαi = (2.18.7)
I 0 σi 0
and    
5 −σ 1 σ 2 σ 3 0 I 0
γ =i = (2.18.8)
0 σ1σ2σ3 0 −I

 
1 0 0
If we form 2
(I−γ5 )
= , then we may use it to project the upper 2 components
0 I
and leave the Helicity −1/2 components. Thus,
1
ΨL ≡ (I − γ5 )Ψ , (2.18.9)
2
with Ψ a positive energy spinor, may be used to describe the neutrino.

If instead we start from a negative energy solution Ψ, the from Section 2.11

α · p~ + βm)Ψ = −EΨ .
(−~ (2.18.10)
 
φ
For Ψ = eip·x we then have
χ
    
−~σ · p~ mI φ φ
= −E , (2.18.11)
mI ~σ · p~ χ χ
which gives the coupled equations

−~σ · p~φ + mχ = −Eφ

mφ + ~σ · p~χ = −Eχ . (2.18.12)

Now taking m = 0 for a massless anti-neutrino

~σ · p~φ = Eφ
~σ · p~χ = −Eχ , (2.18.13)

and since E = |~p| for m = 0,

~σ · p~ 1
φ = φ
2|~p| 2
~σ · p~ 1
χ = − χ. (2.18.14)
2|~p| 2
2.19. FEYNMAN’S INTERPRETATION OF THE KLEIN-GORDON EQUATION 47

This is the same as for the positive energy solution. Thus, the upper 2 components of Ψ
still describe Helicity +1/2 and the lower 2 components describe Helicity −1/2. To obtain
an appropriate Ψ to describe an Anti-Neutrino with Helicity +1/2 we need a negative
energy state with Helicity −1/2.

Thus, ΨL = 12 (I − γ5 )Ψ with Ψ a negative energy spinor, may be used to describe the

anti-neutrino with Helicity +1/2.

2.19 Feynman’s Interpretation of the Klein-Gordon

Equation

In Section 1.2 we abandoned the KG equation because the Probability density

ρ = i (φ∗ ∂t φ − φ∂t φ∗ ) (2.19.1)

could give negative values (we have renamed the wavefuntion Ψ by φ).

It can be checked by direct substitution that the KG equation

∂2 ~ 2 + m2 )φ = 0
( −∇ (2.19.2)
∂t2
has positive energy solutions

φ = N e−ip·x = N e−i(Et−~p·~x) (2.19.3)

and negative energy solutions

φ = N eip·x = N ei(Et−~p·~x) . (2.19.4)

The probability density of such solutions is

ρ = |N |2 (±2E) . (2.19.5)

Thus, negative probabilities come from negative energy solutions. These are (as usual)
the problem.

We need an interpretation for the negative energy solutions of the KG equation. Dirac
Hole theory will NOT work for the spin-0 Bosons described by the KG equation, because
they do not obey the Dirac exclusion principle to give a filled negative energy sea.

Feynman gave an alternative way of interpreting negative energy solutions which works
for both bosons and fermions!
48 CHAPTER 2. RELATIVISTIC QUANTUM MECHANICS

The emission/absorption of an anti-particle with 4-momentum pµ is equivalent to the

absorption/emission of a negative energy particle with 4-momentum −pµ .

In Feynman diagrams, which are the rules of calculating scattering and decay am-
plitudes in RQM, when Anti-Particles are involved we draw lines for negative energy
particles propagating backwards in time and use Feynman’s interpretation. E.g. for elec-
tromagnetic Electron-Positron scattering via photon exchange there are two diagrams
that contribute:

(picture goes here)

2.20 Dirac Equation in an Electromagnetic Field

In classical relativistic mechanics the interaction of a particle carrying charge q in an

external electromagnetic field can be obtained by substituting the momentum as

pµ → pµ + qAµ , (2.20.1)

where Aµ is the 4-vector potential

~ = (φ, A)
Aµ ≡ (A0 , A) ~ (2.20.2)

~ the vector potential. (Remember: E

with φ the scalar potential and A ~ = −∇φ
~ − ∂t A,
~
~ =∇
B ~ × A).
~

This works also for RQM

pbµ → pbµ + qAµ (2.20.3)
or equivalently
∂ µ → ∂ µ − iqAµ . (2.20.4)

The free particle Dirac equation is (iγ · ∂ − m)Ψ = 0. Making the above substitution

γ · ∂ → γ · ∂ − iqγ · A (2.20.5)

the Dirac equation in an electromagnetic field is

(iγ · ∂ − m)Ψ = −qγ · A (2.20.6)

or
(i∂/ − m)Ψ = −qA
/. (2.20.7)
2.20. DIRAC EQUATION IN AN ELECTROMAGNETIC FIELD 49

It is sometimes convenient to write the equation in terms of the Dirac matrices i.e. in
Hamiltonian form. Begin with the equation

∂ ~ − m)Ψ = −q(A0 γ 0 − A
~ · ~γ )Ψ
(iγ 0 + i~γ · ∇ (2.20.8)
∂t
and multiply from left with γ 0 = β. Since (γ 0 )2 = I and γ 0 γ i = ββαi = αi we obtain

∂Ψ  ~ ~

i = (−i∇ + q A) · α + βm Ψ − qA0 Ψ
∂t  
∂Ψ ~
→ i = α · Π + βm Ψ − qA0 Ψ , (2.20.9)
b
∂t

where
~b = −i∇
Π ~ + qA
~ =b ~.
p~ + q A (2.20.10)

Gauge Invariance

Remember that the EM field strength tensor Fµν = ∂µ Aν − ∂ν Aµ is invariant under

gauge transformations Aµ → Aµ + ∂µ χ ≡ õ . However the Dirac equation is not invariant
under gauge transformation

(i∂/ + qA/ − m)Ψ → (i∂/ + q Ã

/ − m)Ψ
= (i∂/ + qA / + q∂/χ − m)Ψ . (2.20.11)

This can be fixed if we transform the Dirac wavefunction by a spacetime dependent phase
factor
Ψ → eiα(x) Ψ = Ψ̃ . (2.20.12)
So we find

/ − m)Ψ = 0 → (i∂/ + q Ã/ − m)Ψ̃ = 0

(i∂/ + qA
→ eiα(x) (i∂/Ψ − (∂/α)Ψ + (qA
/ + q∂/χ − m)Ψ) = 0 .(2.20.13)

If we now require that the 2nd term cancels the 4th term in the last line of that equation,
which simply implies
α(x) = qχ(x) , (2.20.14)
then we get
eiqχ(x) (i∂/ + qA
/ − m) Ψ = 0 . (2.20.15)
So up to an irrelevant overall phase factor we have recovered the original Dirac equation!
Hence the Dirac equation is forminvariance under gauge transformations:

Aµ → õ = Aµ + ∂µ χ(x) ,
Ψ → Ψ̃ = eiqχ(x) Ψ . (2.20.16)
50 CHAPTER 2. RELATIVISTIC QUANTUM MECHANICS

Charge Conjugation

It is interesting to see what happens to the Dirac equation coupled to the EM field
under charge conjugation:
C : Ψ → ΨC = Cγ 0 Ψ∗ , (2.20.17)
with C = iγ 2 γ 0 .

It turns out that

(i∂/ + qA/ − m) Ψ = 0 → (i∂/ − qA

/ − m) ΨC = 0 (2.20.18)

i.e. under charge conjugation the charge of the particle flips sign. (Proof in homework)

2.21 The Magnetic Moment of the Electron

In the non-relativistic limit the rest mass mc2 is the largest energy in the problem (since
|~v |2 << c2 ) and we can write for a positive energy solution
 
−imt φ
Ψ=e (2.21.1)
χ

where φ and χ vary slowly with time and will be called large and small components for
reasons that will become clear in a moment.

Substituting in the Dirac equation (with the Dirac representation for β and αi , which
is more appropriat for studying non-relativistic limits) in an electromagnetic field
! 
~b
   
−imt φ −imt ∂t φ −imt (−qA0 + m)I ~σ · Π φ
me + ie =e ,
χ ∂t χ ~
~σ · Π
b
(−qA0 − m)I χ
(2.21.2)
+imt
multiplying with e and subtracting the first term on the left hand side from both sides
we obtain ! 
~
  0
i∂t φ −qA I ~σ · Π φ
b
= . (2.21.3)
i∂t χ ~ (−qA − 2m)I
~σ · Π
b 0 χ

The lower equtions is

~b − (qA0 + 2m)χ .
i∂t χ = ~σ · Πφ (2.21.4)
For χ varying slowly with time and under the assumption 2m >> qA0

~b
~σ · Π
χ∼ φ (2.21.5)
2m
2.21. THE MAGNETIC MOMENT OF THE ELECTRON 51

~b = b
where Π ~ Hence, for momenta and EM fields small compared to the rest mass
p~ + q A.

χ << φ . (2.21.6)

Now using the top equation

~b ,
i∂t φ = −qA0 φ + ~σ · Πχ (2.21.7)

we get, using eqn. (2.21.5)

~b 2
(~σ · Π)
0
⇒ i∂t φ = −qA φ + φ. (2.21.8)
2m

To simplify this further we may use the identity

(~σ · ~a)(~σ · ~b) = ~a · ~bI + i~σ · (~a × ~b) , (2.21.9)

which follows from

σ i σ j = δ ij I + iijk σ k (2.21.10)

1, i = j
where summation over k is understood. The Kronecker delta is defined as δ ij = ;
0, i 6= j
and 123 = 231 = 312 = +1, 132 = 213 = 321 = −1 and otherwise ijk = 0. In terms of
the -tensor  i
~
~a × b = ijk aj bk (2.21.11)
and, hence,

~b 2 φ = (~σ · Π)(~
(~σ · Π) ~b σ · Π)φ
~b = Π
~b · ΠIφ
~b + i~σ · (Π
~b × Π)φ
~b . (2.21.12)

Now (~ = 1),

~b Π)φ
(Π× ~b = (b ~
p~ +q A)×( ~ = (−i∇+q
p~ +q A)φ
b ~ ~
A)×(−i ~
∇+q ~ = −(∇+iq
A)φ ~ ~
A)×( ~
∇+iq ~ .
A)φ
(2.21.13)
The x component of this expression is

−(∂y + iqAy )(∂z + iqAz )φ + (∂z + iqAz )(∂y + iqAy )φ

= − [∂y (iqAz φ) − ∂z (iqAy φ) + iqAy ∂z φ − iqAz ∂y φ]
= −iq[∂y Az − ∂z Ay ]φ
= −iqBx φ , (2.21.14)

and hence
~b × Π)φ
(Π ~b = −iq Bφ
~ . (2.21.15)
52 CHAPTER 2. RELATIVISTIC QUANTUM MECHANICS

Now
~b 2 φ = Π
(~σ · Π) ~b · Πφ
~b + q~σ · Bφ
~ , (2.21.16)
hence, the non-relativistic limit of the Dirac equation in an EM field (also called Pauli
equation) may be written as
!
∂φ (p
b~ + q ~ 2 q~σ · B
A) ~
i = −qA0 + + φ (2.21.17)
∂t 2m 2m

~ = ~ ~σ for the spin of the electron (setting ~ = 1) we obtain

or writing S 2
!
∂φ (b ~ 2 qS
p~ + q A) ~ ·B
~
i = −qA0 + + φ (2.21.18)
∂t 2m m

where φ is a two-component wave function for the non-relativistic spin-1/2 particle.

Comparing with the usual form of the non-rel. Schrödinger equation

 
∂Ψ 1 ~2
i = − ∇ + V Ψ, (2.21.19)
∂t 2m

we can interpret the last term in eqn. (2.21.18) as a potential energy −~µspin · B ~ due to
the spin magnetic moment of the electron in an external magnetic field. Thus the spin
magnetic moment is
~
qS ~
qS
~µspin = − ≡ −g (2.21.20)
m 2m
where g is the so-called gyromagnetic ration. Hence, the Dirac equation predicts g = 2
whereas classically we would expect g = 1. This prediction was confirmed experimentally
and is one of the spectacular successes of the Dirac equation! Including radiative correction
from Quantum Electrodynamics (QED) yields a more precise value of g = 2(1.0011 . . .)
which agrees up to nine digits after the dot with experiment!

2.22 Hydrogen Atom Spectrum

In the presence of an electrostatic potential V (r) the Dirac equation becomes

∂Ψ
HΨ α·b
b = (~ p~ + βm + V (r))Ψ = i , (2.22.1)
∂t
for positive energy solutions with energy eigenvalue E > 0 we make an separation ansatz

Ψ = e−iEt Ψ0 (r, θ, φ) (2.22.2)

2.22. HYDROGEN ATOM SPECTRUM 53

so that
i∂t Ψ = EΨ (2.22.3)
which gives a time independent equation

α·b
(~ p~ + βm + V (r))Ψ = EΨ . (2.22.4)

For a Hydrogen-like atom we take

Zα e2
V (r) = − , α= . (2.22.5)
r 4π

The total AM operator J~ commutes with (~ p~ + βm) as in section 2.7. Also J~ commutes
α ·b
with V (r) because V (r) is independent of Spin and as in section 1.3, orbital AM operator
~ commutes with V (r). Thus [J,
L ~ H]
b = 0.

l
The problem can be solved using simultaneous eigenstates ψj,m of J~2 , Jz and the parity
operator P (which takes ~x → −~x). The corresponding quantum numbers are j(j + 1), m
and (−1)l , where l is orbital angular momentum.

For the spin-1/2 electron, the allowed values of j are j = l ± 1/2. The gory details of
the calculation can be found in section 2.3.2 of [6], with the result for the energy levels

1 Z 2 α2 1 Z 4 α4
   
1 3 6
En,j = me 1 − − − + O((Zα) ) (2.22.6)
2 n2 2 n3 j + 1/2 4n

This result predicts correctly the splitting of the energy levels with the same princi-
ple quantum number n but different j (Fine Splitting); It does not predict the observed
splitting of energy levels with the same n and j but different parity (−1)l (Lamb Shift).
This requires the quantization of the EM field Aµ and, hence, the use of Quantum Elec-
trodynamics (QED). Other important quantum corrections are discussed in [6].

(draw the energy levels with n = 1 and n = 2)

54 CHAPTER 2. RELATIVISTIC QUANTUM MECHANICS
Bibliography

[1] J. Bjorken and S. Drell, Relativistic Quantum Mechanics and Relativistic Quantum
Fields, McGraw-Hill.

[2] M.E. Peskin and D.V. Schröder, An Introduction to Quantum Field Theory, Addison-
Wesley.

[3] S. Weinberg, The Quantum Theory of Fields, Vol 1, Cambridge University Press.

[4] L.H. Ryder, Quantum Field Theory, Cambridge University Press.

[5] A. Zee, Quantum Field Theory in a Nutshell, Princeton University Press.

[6] C. Itzykson and J.-B. Zuber, Quantum Field Theory, McGraw-Hill.

[7] I.J. Aitchison and A.J. Hey, Gauge Theories in Particle Physics, Dover Publications.

[8] F. Mandl and G. Shaw, Quantum Field Theory, John Wiley & Sons.

[9] W. Greiner, Relativistic Quantum Mechanics: Wave Equations, Springer-Verlag.

55