Relkvtyuuant

Lecture Notes in Relativistic Quantum
Mechanics
by
Lars Bergstrom and Hans Hansson
Department of Physics, Stockholm University
1999
2
Contents
1 Special Relativity 1
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Review of basic relativity . . . . . . . . . . . . . . . . . . . . . 2
1.2.1 Electromagnetic vectors and tensors . . . . . . . . . . . 6
1.3 Relativistic kinematics . . . . . . . . . . . . . . . . . . . . . . 8
2 Relativistic Lagrangians 11
2.1 Classical mechanics . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 Classical elds . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3 Relativistic Quantum Fields 19
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.2 Scalar elds . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.2.1 The Klein-Gordon eld . . . . . . . . . . . . . . . . . . 20
3.2.2 Charged scalar eld . . . . . . . . . . . . . . . . . . . . 22
4 The Quantum Theory of Light 25
4.1 Quantum theory of monochromatic light . . . . . . . . . . . . 25
4.1.1 General theory . . . . . . . . . . . . . . . . . . . . . . 25
4.1.2 General polarization . . . . . . . . . . . . . . . . . . . 31
4.1.3 Interaction with matter . . . . . . . . . . . . . . . . . . 32
4.1.4 Some examples . . . . . . . . . . . . . . . . . . . . . . 34
4.1.5 The classical limit . . . . . . . . . . . . . . . . . . . . . 36
4.2 The full quantum eld theory . . . . . . . . . . . . . . . . . . 39
4.2.1 Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.2.2 The vacuum energy and the Casimir eect . . . . . . . 42
4.2.3 More about the classical limit . . . . . . . . . . . . . . 45
4.2.4 Momentum and angular momentum . . . . . . . . . . 45
4.3 Coherent states of the harmonic oscillator . . . . . . . . . . . 48
4.3.1 What is a coherent state ? . . . . . . . . . . . . . . . . 48
3
4
4.3.2 Coherent states in the n-representation . . . . . . . . . 50
4.3.3 Orthogonality and completeness relations . . . . . . . . 51
4.3.4 Coherent states in the x-representation . . . . . . . . . 52
4.3.5 Time evolution of coherent states . . . . . . . . . . . . 53
4.4 States of the quantized electromagnetic eld . . . . . . . . . . 55
4.4.1 Squeezed States . . . . . . . . . . . . . . . . . . . . . . 57
5 The Dirac Equation 59
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
5.1.1 Short biography of Dirac . . . . . . . . . . . . . . . . . 60
5.2 Constructing the Dirac equation . . . . . . . . . . . . . . . . . 61
5.2.1 Plane-wave solutions . . . . . . . . . . . . . . . . . . . 65
5.2.2 Coupling to Electromagnetism . . . . . . . . . . . . . . 69
5.2.3 Another way to arrive at the Dirac equation . . . . . . 69
5.3 Lorentz invariance . . . . . . . . . . . . . . . . . . . . . . . . 72
5.3.1 Bilinear forms . . . . . . . . . . . . . . . . . . . . . . . 74
5.3.2 Spin and energy projection operators . . . . . . . . . . 76
5.4 Non-relativistic limit . . . . . . . . . . . . . . . . . . . . . . . 78
5.5 Problems with the Dirac equation . . . . . . . . . . . . . . . . 81
5.5.1 The Dirac sea . . . . . . . . . . . . . . . . . . . . . . . 81
5.5.2 The Klein paradox . . . . . . . . . . . . . . . . . . . . 83
5.6 Some applications of the Dirac equation . . . . . . . . . . . . 85
5.6.1 The hydrogen atom . . . . . . . . . . . . . . . . . . . . 85
5.6.2 Coulomb scattering . . . . . . . . . . . . . . . . . . . . 87
5.6.3 Trace formulas . . . . . . . . . . . . . . . . . . . . . . 89
5.7 Quantization of the Dirac eld . . . . . . . . . . . . . . . . . . 91
5.7.1 Majorana particles . . . . . . . . . . . . . . . . . . . . 92
5.8 Lagrangian formulation . . . . . . . . . . . . . . . . . . . . . . 94
Chapter 1
Special Relativity
1.1 Introduction
Although quantum mechanics as it was developed up to around 1925 was an
extremely successful theory, one basic problem remained: it was not in agree-
ment with Einsteins theory of relativity. Since his special theory of relativity
had been experimentally tested and had won acceptance as the correct de-
scription of phenomena at high velocities (v/c 1, where c is the velocity
of light, c = 3 10
8
m/s), it was clear that quantum mechanics somehow had
to be modied to get the correct relativistic limit. One particular problem
was how to treat Einsteins famous relation E = mc
2
. It indicates that mat-
ter can be transformed to energy and the other way around, i.e. the number
of particles can change during collision or other interaction processes. The
non-relativistic quantum mechanics as you have been taught until now does
not have this feature built in the conservation of the probability current has
implied conservation of the number of particles.
The programme of formulating and solving a fully relativistic quantum
theory of electrons and photons, quantum electrodynamics (QED), took almost
20 years to conclude. It is, however, one of the most remarkable achievements
of the 20th century (involving people like Feynman, Schwinger, Tomonaga,
Dirac, Klein, Jordan, Dyson, Wick and many others). The agreement between
measurements and QED predictions is in many cases better than one part in
a million!
In this course, we will not expose you to the full mathematical formalism
of QED. We will, however, try to give you a avour of the most important
physical ingredients and an ability to calculate quite a number of dierent
important processes.
1
2 Special relativity
The notation used in these notes is essentially the same as in the two text-
books by J. J. Sakurai, [1, 2] Modern Quantum Mechanics, 2
nd
ed., Addison-
Wesley, 1994, and Advanced Quantum Mechanics, Addison-Wesley, 1967. In
some places we will refer to formulas in the rst of these books as e.g. (MQM
2.2.23). We will, however, mostly use c = 1, for the speed of light, and h = 1
for the Planck constant divided by 2. We furthermore use real components
for the space-time four-vector, and the metric diag(1, 1, 1, 1).
A fuller account of the the quantum theory of the e.m. eld, can be found
in almost any book on QFT, as e.g. F. Mandl and G. Shaw, Quantum Field
Theory, John Wiley & Sons, 1993. A very good popular book is written by
(guess who?): R. P. Feynman, QED - the Strange Theory of Light and Matter,
Princeton Univ. Press., 1985. An introduction to quantum optics and lasers
can be found in: R. Loudon, The Quantum Theory of Light, Oxford University
Press, 1973.
1.2 Review of basic relativity
Before entering into the problem of formulating relativistic quantum theories,
it may be useful to review some of the most important basic concepts of special
relativity
1
- for a more thorough treatment, see standard textbooks [6, 7].
We introduce four coordinates that parametrize space-time in a frame with-
out inertial forces (an inertial frame), x
0
= ct, x
1
= x, x
2
= y, x
3
= z, or
x
= (ct, r) = (ct, x
i
) (1.1)
where the greek index runs over the values 0, 1, 2, 3 and the latin index i
takes the values 1, 2, 3.
The basic invariant in special relativity is
ds
2
= c
2
dt
2
[r[
2
= (dx
0
)
2
(dx
1
)
2
(dx
2
)
2
(dx
3
)
2
(1.2)
which has the same value in all inertial frames.
Suppose that we rotate the spatial part of the coordinate system
r
= Rr, (1.3)
1
The special relativity was not Einsteins nal contribution to physics. It was superseded
by his general theory of relativity which is valid also for systems where gravity is important.
Since we will deal mostly with processes involving elementary particles, the eects of gravity
are completely negligible. For example, the gravitational force between the electron and
proton in a hydrogen atom is 10
40
times smaller than the electrostatic force. In fact, we
are lucky that this is the case, because still today no fully satisfactory theory for quantum
gravity has been found.
Review 3
with R a 3 3 matrix, which keeps the axes orthogonal and normalized. As
is known from basic linear algebra, this means that
R
1
= R
T
, (1.4)
and we know that just rotating the coordinate system does not change dis-
tances between points.
As an example, the orthogonal matrix that rotates an angle around the
x
3
-axis can be written
R(z; ) =
_
_
_
cos sin 0
sin cos 0
0 0 1
_
_
_. (1.5)
The four-dimensional (4 4) matrix giving space rotation is
rot
=
_
1 0
0 R
_
, (1.6)
where R is a 3 3 rotation matrix.
Suppose that instead we keep the directions of the axes xed, but let a new
inertial system move with constant velocity v in the direction of the x
1
-axis.
This system is said to be Lorentz-boosted along the x
1
-axis with respect to the
rst. Then the four-dimensional rotation (or Lorentz boost) is given by
(x
1
; v/c) =
_
_
_
_
_
0 0
0 0
0 0 1 0
0 0 0 1
_
_
_
_
_
, (1.7)
where
=
v
c
, (1.8)
=
1
1
2
. (1.9)
We can see that this transformation leaves the form ds
2
= (dx
0
)
2
(dx
1
)
2
(dx
2
)
2
(dx
3
)
2
invariant; and specically it will also then leave the velocity
of a light ray invariant as required by special relativity. The four-dimensional
metric for an inertial frame is thus
=
_
_
_
_
_
1 0 0 0
0 1 0 0
0 0 1 0
0 0 0 1
_
_
_
_
_
, (1.10)
where and run from 0 to 3. This metric is called the Minkowski metric,
and space-time with this metric is called Minkowski space. The reader should
be warned that there is unfortunately no universal denition of the Minkowski
metric. Quite often the denition ds
2
= (dx
0
)
2
+ (dx
1
)
2
+ (dx
2
)
2
+ (dx
3
)
2
is
used, and then all the components of (1.10) change sign. Of course, no physical
result will depend on this convention, however some intermediate results may
look dierent.
If we dene x
(we see that this just amounts to changing the

sign of the three space-components), then we can conveniently write ds
2
=
dx
dx
= dx
dx
.
The condition for any Lorentz transformation (rotation or boost) to pre-
serve the distance ds
2
is:
T
= , (1.11)
or
, (1.12)
Taking the determinant of both sides of this equation, one sees that det() =
1. Often, we will only consider so-called pure Lorentz transformations which
have det() = +1. An example of a transformation which has det() = 1 is
the reection or parity operator
P
, which takes x
= (x
0
, r) to x
= (x
0
, r).
Unlike the pure Lorentz transformations, which depend on a continuous set of
parameters (the velocities v and the rotation angles), the reection operator is
an example of a discrete operator. Acting twice with that operator gives back
the original coordinates, i.e.
2
P
= I, with I the unit operator (i.e. the unit
matrix). From a mathematical point of view, Lorentz transformations form a
group, and the reection transformation together with the unit operator cor-
responds to a discrete subgroup of the full Lorentz group. (The term discrete
means that the subgroup contains a nite set of elements - in this case only
2.)
Since the Lorentz transformations are linear transformations of the coor-
dinates
x
=
x
, (1.13)
we see that the matrix
can be expressed as
=
x
, (1.14)
and the inverse transformation
=
x
. (1.15)
Review 5
When studying properties of the continuous part of the Lorentz group it is
usually enough to look at innitesimal transformations. Writing
(1.16)
and demanding a
2
= a
2
for an arbitrary four-vector a
, one nds by inserting

(1.16)
a
2
= a
2
= a
2
+
+O(
2
). (1.17)
Thus,
. (1.18)
There are thus 6 independent parameters (e.g. 3 rotation angles and 3 boost
parameters) characterizing the part of the Lorentz group which is continuously
connected to the unit transformation. If we had considered also translations
(inhomogeneous transformations) of the coordinates, x
+ b
with
b
a constant four-vector, we would have 4 additional parameters. The full

inhomogeneous Lorentz group (or the Poincare group) is thus described by 10
parameters.
The space-time that we have given a metric with the help of
has a
so-called causal structure. The light cone with respect to an event P
1
, say, at
the origin of an inertial reference frame at time t = 0, is dened by s
2
= 0,
where s is the space-time distance between P
1
and another event P
2
.
Points inside the light cone have s
2
> 0, and are said to be causally
connected to the observer. This is because ct > r so that the event P
2
can
be reached from P
1
by travelling slower than the velocity of light. If s
2
< 0
(on the outside of the lightcone) P
1
and P
2
cannot be connected by any means
- the events are causally disconnected.
A set of four quantities that transform in the same way as the dierential
dx
under Lorentz transformations is called a contravariant four-vector. A set

that transforms like dx
is called a covariant four-vector. A quantity that,

like ds
2
, is invariant under Lorentz transformations is called a four-scalar or
Lorentz scalar. An important example of a covariant four-vector is provided
by the four-gradient of a four-scalar:
,
. (1.19)
An important operator is the so-called dAlembertian,
2
=
1
c
2
2
t
2

2
. (1.20)
Since it transforms as a Lorentz scalar, it can be used to construct rela-
tivistically invariant wave equations.
In general, we can for any contravariant four-vector A
use the metric

tensor
to lower the index of A
to obtain the covariant four-vector A
. The usefulness of four-vectors and scalars in special relativity is due to

the fact that if we can write the laws of nature in terms of such objects, they
will have the same form in all inertial systems (they will be so-called manifestly
covariant), in agreement with one of Einsteins postulates of special relativity.
It is easy to generalize to higher tensors, one example is the direct product
of two four-vectors, which is called a second-rank tensor, T
= A
. There
are second-rank tensors that cannot be written as simply a direct product of
two four-vectors, but in each index the transformation property is as if it were:
C
(x
) =
(x). (1.21)
Out of two four-vectors one can also form a four-scalar, the scalar product of
the two vectors, S = A.B
= A
= A
. This can also be

seen as the contraction of indices of the tensor T
= A
, T
= A
.
As a special example, the scalar obtained by contracting a four-vector with
itself, A
2
A.A = A
is an invariant under Lorentz transformations, i.e.,

its numerical value is unchanged when we change inertial system. In special
relativity, it may seem a bit luxurious to introduce both covariant and con-
travariant versions of the same four-vector, since they only dier in the sign
of the space components, A
= (A
0
, A
1
, A
2
, A
3
). Therefore, it
was some time ago popular to introduce imaginary time, i.e., to let ct ict
(this is, for instance, done in [2]). Then
is proportional to the unit matrix

and there is no dierence between covariant and contravariant components.
However, there is no way generalize this to general relativity, so this practice
is avoided nowadays.
1.2.1 Electromagnetic vectors and tensors
In electromagnetism, an important four-vector is the four-potential
A
= (, A), (1.22)
where is the electrostatic potential and Athe 3-vector potential. The electric
eld vector E and magnetic eld vector B are, one the other hand, parts of
Relativistic kinematics 7
an antisymmetric second-rank tensor F
,
F
=
_
_
_
_
_
0 E
x
E
y
E
z
E
x
0 B
z
B
y
E
y
B
z
0 B
x
E
z
B
y
B
x
0
_
_
_
_
_
. (1.23)
F
can be expressed in terms of A
through
F
. (1.24)
The current four-vector is given by
j
= (, j) (1.25)
where is the electrostatic charge density and j is the 3-current density.
Maxwells equations can be summarized by
= j
(1.26)
= 0. (1.27)
We see that the antisymmetry of F
gives from (1.26)
= 0. Thus,
consistency requires that we only couple the electromagnetic eld to conserved
currents.
A special tensor is of course
, which has the same constant value in all

inertial frames. Another important tensor is the 4-dimensional generalization
of the Levi-Civita tensor
ijk
, namely
, which is antisymmetric in all

indices and has
0123
= 1. With its help we can, for instance, dene the
so-called dual tensor

F to F:
=
1
2
. (1.28)
Then the second Maxwell equation (1.27) can simply be written as
= 0. (1.29)
In terms of A
the rst Maxwell equations become

2A
) = j
. (1.30)
Since (1.24) is unchanged if we make the gauge transformation A
f(r, t), we can use this freedom to impose a gauge condition on A
, e.g.,
= 0 (Lorentz gauge), A = 0 (Coulomb gauge) or A

0
= = 0
(axial gauge). In the absence of charges, both the axial and Coulomb gauge
conditions can be chosen simultaneously (radiation gauge).
1.3 Relativistic kinematics
For a point particle with mass m, an important four-vector is the four-momentum
p
,
p
0
= E/c, p
1
= p
x
, p
2
= p
y
, p
3
= p
z
, (1.31)
with E the total energy of the particle
E = mc
2
(1.32)
and p the relativistic linear momentum
p = mv. (1.33)
This means that
v =
c
2
p
E
. (1.34)
The four-distance between two events on the trajectory of a massive particle
moving with constant velocity v is time-like and is most easily calculated in
the rest frame of the particle s
2
= (x
0
)
2
= (c)
2
, where is the time
measured by a clock following the particle (called the proper time). Thus,
since for a clock that moves with the particle dx
i
= 0, we nd in such a frame
trivially that
ds
2
= c
2
d
2
=
dx
dx
, (1.35)
and since this is written in an explicitly Lorentz-invariant (or covariant) form,
it has to be valid in all inertial frames.
The four-velocity v can be dened as v = dx/d, which is the obvious four-
dimensional generalization of the non-relativistic three-velocity v = dr/dt.
The four-momentum is simply p = mv. Using p
2
= m
2
c
2
, it is easy to see that
E =
_
p
2
c
2
+ m
2
c
4
. (1.36)
In particular, for a massless particle (such as a photon), E = c[p[.
Sometimes it is convenient not to have to keep track of all the factors of
c, the velocity of light, in relativistic formulas. One way to achieve this is to
choose units such that c = 1. (Alternatively, this means that one measures all
velocities in fractions of the light velocity.) In the particle physics literature,
this is very common. In addition, h, Plancks constant divided by 2, is con-
veniently put equal to 1 in quantum mechanical problems. Since all physical
units can be expressed using combinations of length, time and mass, and we
have made two units dimensionless it means that we can choose just one to
x all physical dimensions. Usually, mass is used for this purpose.
Relativistic kinematics 9
EXAMPLE 1.3.1 What is the dimension of length and time in a system of
units where h = c = 1?
Answer: In SI units, [ h]=kgm
2
s
1
and [c] = ms
1
. Since we have chosen c
to be dimensionless, length and time must have the same dimensions. Since
h is also chosen to be dimensionless, and thus also the ratio h/c, with [ h/c]
=kgm, we see that mass and length must have inverse dimensions. Thus
[time]=[length]=[mass]
1
.
Chapter 2
Relativistic Lagrangians
2.1 Classical mechanics
In classical mechanics, we are usually interested in how some coordinates
change with time. This can be the positions of a set of point particles, but
also for instance the angle that a pendulum makes with respect to the vertical
direction. Thus, we use some set of generalized coordinates to describe the mo-
tion as a function of time. For example, if we consider N particles moving in
three space dimension, we describe their positions by generalized coordinates
q
i
(i = 1, 2, . . . N). E.g., we can use
q
i
= r
i
(2.1)
or
q
i
= r
i
,
i
,
i
. (2.2)
For each of the generalized coordinates, we dene the generalized velocities by
q
i

dq
i
dt
. (2.3)
We can thus formulate the fundamental dynamical problem in classical me-
chanics:
Given q
i
, q
i
at a time t = t
0
, what is the time development that follows?
It is shown in textbooks in classical mechanics (e.g. [8]) that the solution
is given by Hamiltons principle:
There is a quantity L (called the Lagrangian) such that the integral (the
action)
S =
_
t
1
t
0
L(q
i
, q
i
, t)dt (2.4)
11
12 Relativistic Lagrangians
q
(t)
i
cl
0
(t ) q
i
1
(t ) q
i
q
(t) +
i
cl
(t)
q
i
Figure 2.1: Variation of the path followed by a system, around the classical
path. Note that the endpoints are required to be xed.
takes an extremum when the system moves from q
i
(t
0
) to q
i
(t
1
). The path
followed by the system along this extremal solution is called the classical path
q
cl
i
(t).
Let us see how Hamiltons principle will give us the equations of motion
of the system. Let us perturb the classical path in the action integral by an
amount q
i
(t) (see Fig. 2.1):
q
i
(t) = q
cl
i
(t) + q
i
(t). (2.5)
According to Hamiltons principle, the classical path is at an extremum,
and therefore S = 0,
S =
_
t
1
t
0
_
L
q
i
q
i
+
L
q
i
q
i
_
dt = 0. (2.6)
Noting that q
i
= d/dtq
i
, integrating by parts, and using that q
i
= 0 at the
endpoints this becomes
S =
_
t
1
t
0
_
L
q
i
d
dt
L
q
i
_
q
i
dt = 0. (2.7)
Classical mechanics 13
Since this has to be true for an arbitrary variation q
i
(t) the only possibility
is that the integrand in (2.7) vanishes, i.e.
L
q
i
d
dt
L
q
i
= 0 (2.8)
These are the equations of motion of the system, the Euler Lagrange equa-
tions. Since they are a system of second order dierential equations the solu-
tions are specied by giving both q
i
and q
i
at a given time t
0
.
The fundamental problem of classical mechanics is now reduced to the
problem of nding the Lagrangian and solving the Euler-Lagrange equations.
As a simple example we look at a singe particle of mass m moving in a potential
V (q). In that case the Lagrangian is given by [8]
L =
1
2
m q
2
V (q) (2.9)
(according to the general rule L = T V where T is the kinetic and V the
potential energy). The Euler-Lagrange equation (2.8) then gives
m q =
V (q)
q
F, (2.10)
which is nothing but Newtons law (F is the force).
We have treated coordinates and velocities in this Lagrangian formulation
of classical mechanics. Sometimes we also want the Hamiltonian formulation,
which involves coordinates and momenta. We thereby start by dening canon-
ically conjugate momenta to q
i
by
p
i

L(q, q)
q
i
. (2.11)
We assume that this relation is invertible so that the q
i
can be expressed
in terms of the q
i
and p
i
. We now form the Hamiltonian by
H
i
p
i
q
i
L(q, q(p, q)). (2.12)
We now compute how H is changed if we make a small change of the
coordinates and momenta:
dH =
_
q
i
+
q
j
p
i
_
p
j
L
q
j
__
dp
i
+
_
L
q
i
q
j
q
i
_
L
q
j
p
j
__
dq
i
. (2.13)
Both the expressions in round brackets vanish because of (2.11), and by com-
paring (2.13) with the general expression
dH =
H
p
i
dp
i
+
H
q
i
dq
i
, (2.14)
we nd
_
q
i
=
H
p
i
;
L
q
i
=
H
q
i
(2.15)
But according to (2.8) and (2.11), we can replace L/q
i
by p
i
to arrive at the
Euler-Lagrange equations in the Hamiltonian formulation:
_
q
i
=
H
p
i
; p
i
=
H
q
i
(2.16)
Let us return to our simple example. We had
L =
1
2
m q
2
V (q) (2.17)
which gives
p =
L
q
= m q, (2.18)
i.e., q = p/m. Then
H = p q L =
p
2
2m
+ V (q) = T + V. (2.19)
The Euler-Lagrange equations
_
q =
H
p
=
p
m
; p =
H
q
=
V
q
= F (2.20)
then give m q = F as before.
2.2 Classical elds
Let us consider an interesting application of our formalism. We look at a
system of N particles, moving in one dimension, connected to each other by
almost massless springs with force constant k (see Fig. 2.2). The particles
are separated by a distance a from each other in equilibrium. Let
i
be the
Classical elds 15
displacement from the equilibrium position of particle i. Then the Lagrangian
is
L =
i
(T
i
V
i
) =
1
2
i
_
m

2
i
k(
i+1
i
)
2
_
=
a
i
1
2
_
_
m
a
2
i
ka
_
i+1
i
a
_
2
_
_
=
i
aL
i
, (2.21)
where L
i
is the Lagrangian density (Lagrangian per unit length) contributed
by particle i. We can now take the continuum limit, i x,

i

_
dx,
i
(x), (
i+1

i
)/a /x, m/a (mass density) and ka
(Youngs modulus). The continuum Lagrangian then becomes
L =
_
dx
1
2
_
_
x
_
2
_
_
. (2.22)
We can thus write the action
S =
_
dtL =
_
dt
_
dxL
_
,

,

x
_
, (2.23)
with the Lagrangian density
L
_
,

,

x
_
=
1
2
_
_
x
_
2
_
_
. (2.24)
a
m
Figure 2.2: A system of identical classical particles of mass m, connected by
springs with force constant k. The distance between the particles in equilib-
rium is a.
In this example, (x, t) is the displacement eld. A eld is a function of
space and time, i.e., it contains an innite number of degrees of freedom. (To
completely specify a eld, we have to give its value in every point in space,
at every time.) The construction we made in this one-dimensional example is
easy to generalize to three space dimensions. Then
S =
_
dtL =
_
dt
_
d
3
xL
_
,

,
_
, (2.25)
which gives rise to the Euler-Lagrange equations
3
k=1
x
k
L
(/x
k
)
+

t
L
(/t)

L
= 0. (2.26)
In fact, we notice that (2.26) can be written in the relativistically invariant
form
_
L
(/x
)
_
= 0. (2.27)
This is Lorentz invariant if L is a scalar density (i.e. if L
(x
) = L(x
)).
As the rst realistic example we consider a real scalar eld (x) (this can
for instance be the so-called Higgs eld of particle physics). The simplest
non-trivial Lagrangian density is given by
L =
1
2
_
(
) (
)
2
2
_
. (2.28)
The Euler-Lagrange equation (2.27) gives the equation of motion
+
2
= 0, (2.29)
or
(2+
2
) = 0, (2.30)
which as we remarked before is a relativistic wave equation, the Klein-Gordon
equation
1
Just as ordinary particle dynamics can be described either by a Lagrangian
or an Hamiltonian, this is true also for a classical eld theory. Corresponding
to the coordinate there is a canonical momentum,, that is also labeled
by a continous variable x and dened in complete analogy with the discrete
case. For the real scalar eld we nd,
(r, t) =
L
(r, t)
= (r, t). (2.31)
1
The Swedish physicist Oskar Klein was an assistant to Niels Bohr and returned to
Stockholmas professor of theoretical physics at Stockholm University from 1931-1962. Every
year, the Klein memorial lecture is given at the University.
Quantum elds 17
The Hamiltonian density can now also be easily calculated,
H = L =
1
2
[
2
+ ()
2
+
2
2
]. (2.32)
By integrating this over space, we get the full Hamiltonian, H =
_
d
3
xH,
which gives the total energy. As usual, the time evolution of functions on the
phase space follows from Hamiltons equation,
d
dt
O((x, t), (x, t)) = O, H (2.33)
where the canonical Poisson bracket is given by
(r, t), (r
, t) =
3
(r r
). (2.34)
In the following, the hamiltonian formulation will be of special importance,
since we will use canonical quantization to construct a quantum theory of elds.
Chapter 3
Relativistic Quantum Fields
3.1 Introduction
The quantum nature of electromagnetic radiation, and especially light, is
stressed in all introductory treatments of quantum mechanics. It was the
ultraviolet catastrophe of black-body radiation that originally led Planck to
introduce the concept of quantum of energy, and it was the quantum nature of
light, as revealed in the photoelectric eect, that led Einstein to the concept
of photon. Since electromagnetic radiation was already very well understood
in terms of electric and magnetic elds satisfying the Maxwell equations, the
pioneers of quantum physics were faced with a great problem: How can light
be both particles and waves? A satisfactory answer to this question was not
given before the introduction of quantum eld theory about 1930, with central
contributions given by Dirac, Heisenberg, Jordan, Klein, Wigner, and others.
Also the solution to the problem came in a roundabout way, rst the particle-
wave duality was understood for electrons, and only afterwards for photons.
Electrons were originally of course thought of as (classical) particles, while
the matter-wave concept, introduced 1924 by de Broglie, was central in the
development of the quantum theory. For light, the classical description was
that of a wave, while the quantum theory introduced the particle aspect.
As we shall se later, it is completely natural that the classical limit of a
quantum theory of massive particles, like electrons, is a particle theory, while
the classical limit of a quantum theory of massless (or at least very light)
particles is a eld theory, like that of Maxwell. Very roughly speaking, in
a massless theory the number of particles is not xed in most experimental
situations, and the classical electromagnetic elds obeying Maxwells equations
are quantum states with a large uncertainty in the number of photons. The
19
quantum mechanics you study in introductory courses deals with systems with
a xed number of particles, and it was also this quantum theory that was
originally developed by Bohr, Schrodinger, Heisenberg, Dirac and others.
To describe the quantum theory of systems where the number of particles
can change, one needs techniques (of the previous chapter) that are usually
referred to as quantum eld theory (although sometimes one still comes across
the older, and somewhat misleading, notion of second quantization).
The main goal of these notes is to give a selfcontained presentation of the
basic ideas and results of quantum electrodynamics, which is the quantum the-
ory of relativistic electrons and lightparticles or photons. There are, however,
several technical diculties in dealing with both electrons and photons. First
they both carry an intrinsic spin, and in the case of photons there are dicul-
ties related to the gauge invariance of Maxwell theory. Electrons are fermions,
and it turns out that the usual canonical quantization procedure has to be
modied. All this will be explained later, but in order to present the basic
idea of eld quantization with a minimum of technical obstacles we shall start
by constructing a scalar quantum eld theory, i.e. a theory where the par-
ticles do not carry any intrinsic spin. In later courses you will see that this
theory is needed to describe the so called Higgs particle, which is an important
ingredient in the so called Standard model for elementary particles
1
.
3.2 Scalar elds
3.2.1 The Klein-Gordon eld
When we quantize a point particle in elementary quantum mechanics, we treat
the coordinates q (usually chosen to be the three cartesian coordinates x
i
) and
momenta p
i
as operators. They fulll the quantum mechanical commutation
relations
[x
i
, p
j
] = i h
ij
, (3.1)
and the wave function is a representation of the state vector, on which
these operators act. In the eld theory developped in the previous section,
the coordinate x
i
plays a very dierent role - it is simply a way to label the
(innitely) many coordinates (x, t). and momenta (x, t). To quantize this
1
Another important application of scalar eld theories is in condensed matter physics.
It should be clear from the derivation of the classical scalar eld theory in Section 2.2 that
it describes vibrations in a crystal, i.e. sound waves. If we replace the speed of light with
the speed of sound, we can use the scalar quantum eld theory to be developed below to
describe the quantized vibrations that are are called phonons.
Quantum elds 21
system, we simply turn the canonical Poisson brackets (2.34) into commutators
which are the natural generalizations of (3.1)
[(r, t), (r
, t)] = i h
3
(r r
). (3.2)
The Klein-Gordon equation, being a wave equation, should have plane-
wave solutions. We can imagine that our system is enclosed in a large box of
volume V , and we impose periodic boundary conditions (we will eventually
take the limit V , of course). We then expand the real scalar eld at
a given time t in terms of plane waves:
(r, t) =
k
1
2V
[a
k
e
i(tkr)
+ a
k
e
i(tkr
)], (3.3)
where we have introduced 1/
2V as a convenient normalization, and where

insertion into the Klein-Gordon equation shows that = (k) =
2
+k
2
,
which describes the energy of a relativistic particle with mass and momentum
k. We can thus write the factors in the exponentials as kx k
with k
0
= .
Since the K-G eld is written as a sum of independent components which
all fulll a wave equation, i.e., have harmonic motion, it is natural to quantize
each mode in the same way as we usually quantize the harmonic oscillator. By
inserting (3.3) in (2.32) and performing the integration over the whole volume
V , one gets
H =
k
h(k)a
k
a
k
(3.4)
where we have temporarily reinserted a factor h to show the similarity with
the expression of the energy of the quantum harmonical oscillator. Indeed, if
we interpret a
k
and a
k
a
k
as lowering and raising operators, fullling the
commutation relations
_
a
k
, a
_
=
k,k
, (3.5)
then and will fulll the commutations relations (3.1). Recomputing the
Hamiltonian using (3.5) one nds
H =
k
h(k)(a
k
a
k
+
1
2
). (3.6)
Usually, we will just ignore the constant contribution coming from summing
the zero-modes (i.e. the terms h/2). There are cases, however, when they
should be kept as we will see later.
Now we can use all the machinery that we have learned when studying
the harmonic oscillator in non-relativistic quantum mechanics. We dene the
ground state as the one annihilated by all a
k
, i.e.
a
k
[0) = 0 (3.7)
for all k. A normalized state with n
k
excitations in the mode k is given by
[n
k
) =
[a
k
]
n
k
n
k
!
[0). (3.8)
Since H is a sum of non-interacting harmonic oscillators, the eigenstates are
direct products:
[ . . . n
k
i
, . . . n
k
j
, . . .) =
k
i
[n
k
i
). (3.9)
The (enormous) Hilbert space spanned by all these basis states is called a Fock
space.
To give a physical interpretation of the states (3.9), we note
H[ . . . n
k
i
, . . . n
k
j
, . . .) =
k
n
k
(k)[ . . . n
k
i
, . . . n
k
j
, . . .) (3.10)
where (k) = h(k). Thus we can interpret (3.9) as a state with many par-
ticles, each of mass m, where n
k
1
have momentum k
1
, n
k
2
have momentum
k
2
, etc. This is also why the raising and lowering operators, a
k
and a
k
are
usually referred to as cration and annihilation operators: a
k
acting on the
vacuum state creates one particle with wave vector k. With this formalism we
will be able to treat processes where particles of given momenta are created
or destroyed, e.g. in collision processes.
3.2.2 Charged scalar eld
It turns out that one cannot describe electrically charged particles with a real
scalar eld. To do so, we must use a complex scalar eld.
2
The calculations
are very similar, however. A suitable classical Lagrangian density is given by
L =
_
(
) (
)
2
[[
2
_
. (3.11)
2
The reason is basically the same as in non-relativistic quantum mechanics. There one
must have a complex wave function to get a non zero current to couple to the vector
potential. In quantum eld theory on needs a complex quantum eld in order to form a non
zero current operator.
Charged eld 23
Treating and
as independent elds, the Euler-Lagrange equations become

_
2 +
2
_
(x) = 0, (3.12)
(x now stands for x
) and
_
2+
2
_
(x) = 0. (3.13)
The Fourier expansion now must describe a non-hermitian eld, and takes the
form
(x) =
k
1
2V
[a
k
e
ikx
+ b
k
e
ikx
], (3.14)
and from the canonical commutation relations of the classical elds and
one deduces
[a
k
, a
k
] = [b
k
, b
k
] =
kk
. (3.15)
It thus appears that we have two types of particles, of the same mass ,
created by a
and b
, respectively. When we couple electromagnetism to this

complex scalar eld, we will see that the a-particles and b-particles have op-
posite signs for the electric charge. This formalism thus can include (and in
fact even predicts) the existence of antiparticles!
Historical Note Early on it appeared that there was a way to make
relativistic quantum mechanics for the real scalar eld, by making the usual
substitutions E i h/t and p i h and inserting into the relativistic
equation relating energy and momentum (the square of (1.36))
E
2
p
2
c
2
= m
2
c
4
(3.16)
gives indeed (2.30) with = mc/ h. Maybe we can now interpret (2.30) as the
relativistic Schrodinger equation? It turns out that this is impossible. One
obvious problem is that we have squared the energy operator. This means that
whenever we have a solution with positive E there also has to be a solution
with negative E. We know, however, that the energy of a free particle must
be strictly positive (otherwise we could gain arbitrary amounts of energy by
creating more and more particles an unstable situation indeed!). It is not
possible to just forbid the negative-energy solutions, because without them
the solutions to (2.30) do not form a complete set. Another way to put it: If
we start with a solution of positive energy, that may evolve with time into a
negative-energy one.
Also Klein showed (through the so-called Klein paradox) that if one scatters
a plane wave solution to the Klein-Gordon equation against a steep potential
step, the reected wave can be larger than the incoming one, violating prob-
ability conservation. However, this is a problem only because we treat as a
Schrodinger-like wave function. From the way we have presented the deriva-
tion of the scalar eld theory it should be clear that this is not the correct
interpretation. Rather, as already stressed, (x
) is a coordinate labeled by
the space point x.
Chapter 4
The Quantum Theory of Light
4.1 Quantum theory of monochromatic light
4.1.1 General theory
We now want to develop the quantum theory of light step by step using only
simple tools, like classical electrodynamics and basic quantum mechanics. As
the story unfolds, you will recognize how we again encounter the concepts
that were introduced when discussing the quantization of the scalar eld. The
strategy is rst to develop the quantum theory of monochromatic light. This
is technically much simpler than the full theory, but it still contains all the
important concepts, like creation and annihilation operators for photons, the
nature of the classical limit etc.
From experiments we know that excited atoms emit radiation with very
sharply dened frequencies. For example, a hydrogen atom in its rst excited
states emit photons with energy h = 10.2 eV, and a lifetime 10
9
s, corresponding to a natural line width of about 4 10
7
eV. A laser is a
device to amplify such monochromatic light, and later we shall discuss the
characteristics of laser light. From this it should be clear that even though
monochromatic light is a very special case of electromagnetic radiation, it is an
important one and well worth studying for its own sake. It also turns out that
by understanding this special case we will be very well prepared to construct
the full quantum theory of light, because as we have seen in the previous
chapter, by making a Fourier expansion we can quantized a more general eld.
In classical theory, linearly polarized monochromatic
1
light is described by
1
The notion of monochromatic is slightly misleading, since we will consider elds with
a xed wave vector

k, i.e. both energy and direction of propagation is xed. We shall
assume that the eld is enclosed in a large periodic box, so there will be several possible
25
26 Quantum Theory of Light
perpendicular electric and magnetic elds,
2
E(x, t) =
1
2
E e
i(t
k x)
+ c.c. (4.1)
B(x, t) =
1
2
B

k e
i(t
k x)
+ c.c. (4.2)
where = [
k[ = k, [B[ = [E[ and the polarization vector , satises

k = 0.
Later we shall use a basis of polarization vectors,
, = 1, 2, chosen so that
the triad (
1
,
2
,
k) form a right-handed ON-basis. Light with circular, or in

general elliptic, polarization is obtained by (complex) linear combinations of
the two polarization vectors.
We shall also need the description in terms of potentials. In classical theory
this is merely a matter of convenience, but in quantum theory it is more than
so. As you already know from the Aharonov-Bohm eect, the vector potential
is necessary in order to formulate a local quantum mechanical description
of charged particles interacting with magnetic elds. Here we will also see
that the vector potential is needed in order to get a simple description of the
electromagnetic eld itself. In order to specify the potentials we must choose
a gauge condition (see Section 1.2.1). For the pure electromagnetic eld the
one that is simplest to handle is the so called Coulomb gauge condition,
i
A
i
=

A = 0. (4.3)
If no matter is present, it follows from Maxwells equations that the scalar
potential is a constant, usually taken to be zero. The gauge condition (4.3)
together with the axial gauge condition = 0, is the radiation gauge, and the
elds are given by,
3
E
i
=

A
i
(4.4)
B
i
=
ijk
j
A
k
.
In the presence of matter is no longer constant, and there is a longitudinal
component to

E (i.e. a part for which
E ,= 0) so that Gauss law is satised.

wave vectors corresponding to a given energy. However, instead of all the time saying,
linearly polarized light with a xed wave vector, we will simply say monochromatic.
2
We will consistently use complex notation; c.c. denotes complex conjugation, and for
operators h.c. denotes hermitian conjugation.
3
We will often use component rather than vector notation, in the latter case these equa-
tion would read

E =
t

A and

B =
A. If you are still not familiar with the conventions

of repeated index summation, and the manipulation of epsilon symbols, this is a good time
to learn it.
Monochromatic light 27
Still, however, is completely determined by the distribution of charges and
is thus no independent degree of freedom. The radiation eld is still entirely
described by the transverse vector potential

A.
The vector potential in radiation gauge corresponding to the elds (4.14.2)
is given by
A(x, t) =
1
2
A e
i(t
k x)
+ c.c. (4.5)
=
1
2
A(t) e
i
k x
+ c.c. ,
where E = iA, and where we introduced the time-dependent amplitude
A(t) = e
it
A.
To construct any physical theory, we must rst nd the relevant variables
to describe our system, and then the correct Hamiltonian. The rst part, i.e.
the kinematics, is in many cases rather obvious. For instance, for a particle
moving in a potential it is clear that the position and its conjugate variable,
the momentum, describes the state of the particle. For more complicated sys-
tems it can be harder to gure out what the relevant variables are. The basic
rule is that they are always related to what is called physical or dynamical
degrees of freedom, i.e. variables that correspond to real physical changes in
the system. Here it is important that the variables are chosen to be indepen-
dent of each other, so it would e.g. be wrong to use both the position, the
momentum and the angular momentum, to describe the motion of a particle.
In quantum theory we have the additional diculty of nding the correct
commutation relations between the operators that describe the physical ob-
servables. Here the guiding principle is the correspondence with the classical
theory. The most elegant way to proceed in the case of electromagnetism is to
rst formulate Maxwells theory in the language of analytical mechanics and
then use the correspondence between Poisson brackets and commutators. We
saw in Section 3.2.1 that decomposing the generalized coordinates of the
Klein-Gordon eld in its Fourier components, we could quantize by promoting
the Fourier coecients to annihilation and creation operators for each partic-
ular monochromatic mode. We will try to proceed in a similar way with the
vector potential A.
In classical theory, the intensity of an e.m. wave depends on the amplitude,
while in the quantum description it is quantized in units of h, and propor-
tional to the number of photons. It is thus natural to expect that somehow the
amplitude of the eld will become a quantum variable. For a monochromatic
wave it is sucient to specify the amplitude at one point, since that determines
the complete eld. We shall use the amplitudes at x =
0 as variables,
4
A(
0, t) =

A(
0, t) =
1
2
[A(t) + A
(t)] (4.6)
E(
0, t) =

E(
0, t) =
i
2
[A(t) A
(t)] . (4.7)
Often the complex amplitudes A(t) and A
(t) are more convenient to use than

the real amplitudes A(
0, t)and E(
0, t). Recall that in the radiation gauge,
A(
0, t) = E(
0, t) , (4.8)
which suggests that A(
0, t) and E(
0, t) can be considered as generalized po-

sition and momentum coordinates.
5
Maxwells equations also give,
E(
0, t) = k
2
A(
0, t) , (4.9)
which combined with (4.8), implies harmonic time dependence. It is thus
very natural to compare with a classical harmonic oscillator described by the
Hamiltonian
H
ho
=
p
2
2m
+
1
2
m
2
x
2
, (4.10)
which gives the following Hamiltons equations
p(t) = m
2
x ; x =
p
m
. (4.11)
We also compare (4.10) with the Hamiltonian for the electromagnetic eld
in vacuum. The total eld energy given by,
c =
_
d
3
x
1
2
_
E(x, t)
2
+

B(x, t)
2
_
=
_
d
3
x
1
2
_
E(x, t)
2
+ k
2

A(x, t)
2
_
,(4.12)
so for a monochromatic eld the Hamiltonian takes the very simple form,
H =
V
2
_
E(
0, t)
2
+
2
A(
0, t)
2
_
=
V
2

2
A
(t)A(t) , (4.13)
where V is some large box with periodic boundary conditions that encloses
our system.
6
With the following identication,
x(t)
V A(
0, t)
p(t)
V E(
0, t) (4.14)
m 1 ,
4
Of course we could have picked any other point.
5
This can, of course, be derived from the Lagrangian density using the canonical quan-
tization formalism.
6
From A(t) = Ae
it
, it is clear that the Hamiltonian is time independent.
H
ho
(x(t), p(t)) H(A(
0, t), E(
0, t)), and the equations (4.8) and (4.9) de-

scribing the e.m. eld become identical to the harmonic oscillator equations
(4.11). Or in words: The canonical variables for a monochromatic electromag-
netic wave are the amplitudes of the vector potential and the electric elds. It
can never be stressed suciently, that this concept of coordinate and momen-
tum has nothing to do with the position coordinate x in the electromagnetic
eld

E(x, t).
In the quantum description of the harmonic oscillator, the canonical dy-
namical variables x(t) and p(t) is turned into Heisenberg operators x(t) and
p(t), satisfying the equal time canonical commutation relation
[ x(t), p(t)] = i h . (4.15)
In analogy with the particle case, we replace the classical vector potential and
electric eld into (Heisenberg) operators and postulate the following canonical
equal time commutation relation,
[

A(
0, t),
E(
0, t)] =
1
V
i h . (4.16)
Note that the amplitudes A(x, t) and E(x, t) are dened to be real, so the
operators

A(
0, t) and

E(
0, t) are hermitian.
We now have a candidate quantum theory for plane polarized monochro-
matic light that is constructed purely in analogy with the quantum mechanics
for a single particle.
7
Before continuing the formal development, we must ask
if this theory is consistent with what we know about light.
Since we can now freely borrow results from the quantum theory of the
harmonic oscillator, we immediately know that the Hilbert space of our quan-
tum theory is spanned by n-states [n) with energies E
n
= h(n +
1
2
). This
agrees with the observed fact that at very low intensities light is quantized in
units of h. Thus, we interpret [n) as a state of n free photons. It might come
as a surprise that the lowest energy state is not zero, but
1
2
h, and we will
come back to this later. Can we understand the concept of a classical e.m.
wave in our quantum theory? To do this, the crucial observation is that elec-
tromagnetic elds and potentials are quantum operators. Thus classical elds
must correspond to expectation values of these operators. This is precisely
in analogy with x and p, in our prototype quantum mechanics example, and
7
For those worried by the seemingly loose arguments based on analogies, we again stress
that in more advanced texts it is shown more rigorously how to give a Hamiltonian or
Lagrangian formulation of classical electrodynamics. The analogies given here is just an
intuitive short cut to those results.
there we know that the quantum states appropriate for the classical limit are
the coherent states. The same will be true for the e.m. eld. In the rest of
this section we shall dress these words in equations.
To proceed we need explicit expressions for the raising and lowering oper-
ators that connect the dierent n-states. As expected from the form of the
Hamiltonian (4.13), these (nonhermitian) operators are simply related to the
operators

A(t) and

A
(t), corresponding to the (complex) amplitudes A(t) and

A
(t),
a(t)
V
2 h
A(t) =
V
2 h
_

A(
0, t)
i
E(
0, t)
_
(4.17)
a
(t)
V
2 h
(t) =
V
2 h
_

A(
0, t) +
i
E(
0, t)
_
, (4.18)
where the scaling factor between a(t) and

A(t) is chosen as to get the canonical
commutation relation,
[a, a
] = 1 , (4.19)
and the standard form,
H = h
_
a
a +
1
2
_
, (4.20)
of the Hamiltonian (4.13). For simplicity we shall write a instead of a. Since
we use the Heisenberg picture, the photon state vectors are time independent,
while the time dependence of the operators is obtained from Heisenbergs
equation,
a(t) =
1
i h
[a(t), H] = ia(t), (4.21)
with the solution
a(t) = ae
it
, (4.22)
where we used the notation a = a(0). We can now express the vector potential
as,
A
op
(x, t) =
1
2
A(t) e
i
k x
+ h.c. =
h
2V
a e
i(t
k x)
+ h.c. . (4.23)
Since the operators a and a
changes the number of photons, they will be

referred to as annihilation and creation operators.
8
Starting from the vacuum
state [0), we dene n-photon states by repeated action of the creation operator,
[n) =
1
n!
(a
)
n
[0) (4.24)
with energies E
n
= h(n +
1
2
).
For most applications, the n-representation is the appropriate one, but
one can also use the

A representation, where the state is represented by time
dependent (Schrodinger picture) wave functions,
(A, t) = A, t[) (4.25)
and the time-independent Schrodinger operators

A =

A
op
(
0, 0) and

E =
E
op
(
0, 0) are represented by multiplication by A and the derivative operator,
E =
i
h
V
d
dA
. (4.26)
In complete analogy with the harmonic oscillator, the wave function for the
vacuum, or no photon, state will be
vac
(A, t) =
_
V
h
_
1
4
e
V
2 h
A
2
e
i
2
ht
. (4.27)
4.1.2 General polarization
It is straightforward to slightly extend the above theory to include monochro-
matic light in a general state of polarization,
A(x, t) =
1
2
(t)
e
i
k x
+ c.c. , (4.28)
and similarly for the electric and magnetic elds. The Hamiltonian becomes
a sum over two independent harmonic oscillators, one for each polarization,
H =
V
2
_
E
2
0, t) +
2
A
2
0, t)
_
, (4.29)
and the equal time commutation relation reads,
[

A
0, t),

E
0, t)] =
1
V
i h
. (4.30)
8
This language is quite often used also for the harmonic oscillator, where one thinks of
the action of a and a
as creating, or annihilating quanta of energy.

It should be clear how to extend all the previous formulas for the plane po-
larized case to a general polarization. For instance, a state with n
1
photons
polarized along
1
and n
2
along
2
, is (in obvious notation) given by,
[n
1
, n
2
) =
1
n
1
!
n
2
!
(a
1
)
n
(a
2
)
n
[0) . (4.31)
We can remark already now that it will be straightforward also to treat
a general non-monochromatic electromagnetic eld. All we have to do is to
Fourier-decompose the eld into a sum of components of given wave vectors,
and quantize each Fourier mode independently.
4.1.3 Interaction with matter
To demonstrate the power of the theory developed so far, we shall use it to
show how the quantized e.m. eld interacts with matter, typically an electron
bound to an atom. In this case the full Hamiltonian is given by the sum,
H = H
rad
+ H
mat
+ H
int
, (4.32)
where H
rad
is given by (4.20) and H
mat
and H
int
by (cf. MQM, 2.6.21 and
5.7.1),
H
mat
=
p
2
op
2m
e
+ e(x) (4.33)
H
int
=
e
2m
e
_
A
op
(x, t) p
op
+ p
op

A
op
(x, t)
_
+
e
2
2m
2
e
A
2
(x, t). (4.34)
We will use the radiation gauge, which means that we can replace p
op
A
op
(x, t)
by

A
op
(x, t) p
op
, so that the interaction Hamiltonian in radiation gauge can
be written simply as
H
int
=
e
m
e
A
op
(x, t) p
op
+
e
2
2m
2
e
A
2
(x, t). (4.35)
This diers from the treatment you have seen previously, in that the vector
potential is an operator. The total state vector for the system is now a direct
product of the photon state and the matter (here the electron) state,
[
a
; n) = [
a
) [n) (4.36)
Since the dipole force is weak, the interaction term can be treated as a pertur-
bation (you will learn more about this later), and the relevant term will be the
matrix element of H
int
between an initial state, characterized by the electron
in the state [
i
) and with n
i
photons, and a nal state described by [
f
), and
n
f
,
H
fi
(t) =
f
; n
f
[H
int
[
i
; n
i
) . (4.37)
Substituting (4.34) and (4.23) in (4.37) and using a[n) =

n[n) and a
[n) =
n + 1[n), we obtain, neglecting the term of order e

2
H
fi
(t) = H
emit
(t) + H
abs
(t) (4.38)
with
H
emit
(t) =
e
m
e
h
2V
f
[e
i
k x
p
op
[
i
) e
it
n
i
+ 1
n
f
,n
i+1
(4.39)
and
H
abs
(t) =
e
m
e
h
2V
f
[e
i
k x
p
op
[
i
) e
it
n
i
n
f
,n
i1
. (4.40)
From these expressions we already learn several things: First, to lowest or-
der in perturbation theory (i.e., lowest order in the electromagnetic coupling
constant e), atomic transitions take place via absorption or emission of one
photon. Note that this prediction of the theory, which is conrmed by exper-
iments, is not put in by hand, but follows from the form of the interaction
Hamiltonian that was constructed on the basis of gauge invariance, and the
expression (4.23) for the radiation eld operator that was derived without any
reference to matter.
9
Second, the term in (4.38) proportional to
n
f
,n
i+1
is
nonzero even if there are no photons in the initial state. This corresponds
to spontaneous emission and cannot be explained by coupling a classical e.m.
eld to quantized matter. For n
i
,= 0 the rst term in (4.38) corresponds to
absorption, and the second to stimulated emission.
The polarization vector has two independent degrees of freedom, corre-
sponding to the two possible linear polarizations of the electric (and magnetic)
eld vectors. (The third polarization has been removed by the gauge condi-
tion, which is seen to imply that the polarization vector is perpendicular to
the direction of propagation.)
9
To higher order in perturbation theory, there are processes where two or more photons
are emitted. These are however very rare, since the coupling involves an extra power of the
ne structure constant. In our formalism, the term proportional to e
2
in H
int
will contribute
to two-photon processes, together with the term linear in e computed to second order in
perturbation theory.
4.1.4 Some examples
In (4.38) we have a very useful formula applicable to a large number of pro-
cesses where a photon is either absorbed or emitted. Let us rst consider
emission processes. We recall that according to time-dependent perturbation
theory, the transition probability per unit time induced by a harmonic pertur-
bation of the form of H
int
is given by Fermis golden rule:
w
fi
= 2[f[H
int
[i)[
2
(E
f
E
i
)
, (4.41)
where we have taken out the explicit time dependence of the interaction Hamil-
tonian:
H
int
(t) = H
int
e
i
, (4.42)
where the plus sign hold for emission and the minus sign for absorption. (It
is the integral over time of those factors that gives an energy-conserving delta
function with the argument E
f
E
i
.)
For a photon of given polarization emitted into a small energy interval
between and + d, and solid angle element d the density of nal states
is given by
=
V
2
(2)
3
d. (4.43)
We can now compute the novel process of spontaneous emission, i.e., the
situation when an atom in the state A makes a transition to the state B under
the emission of a photon. We can use our eective interaction H
emit
with n
i
put equal to zero (we assume that there is no external electromagnetic eld
present during the process). The transition probability for a photon emitted
in the solid angle interval d is, according to the golden rule (we denote the
photon by the symbol )
w
AB+
=
e
2
m
2
e
V
[
j
B[e
i
k x
j
p
j
[A)[
2
V
2
(2)
3
d, (4.44)
where the sum is over all atomic electrons involved in the transitions, and
where the energy conservation condition xes to have the value E
A
E
B
. If
we assume for simplicity that we have a hydrogen atom we can omit the sum
over j. Also, like you have seen in previous courses of quantum mechanics,
the dipole approximation can usually be made. The reason for this is that
the wave-length of photon emitted in the transition is much larger than the
dimension of the atom, so that the exponential does not vary very much over
the region of emission. Thus,
e
i
k x
1 i
k x
1
2
_
k x
_
2
+ . . . (4.45)
can be replaced by the rst term (unity). The only matrix element to compute
is thus B[ p[A), which can be transformed using
_
p
2
, x
_
= 2i p (4.46)
to
B[ p[A) = im
e
B[ [H
0
, x] [A) = im
e
(E
B
E
A
) B[x[A) = im
e
x
BA
. (4.47)
This is what one would obtain if the interaction term would be of the dipole
form
e
m
e
A p ex

E. (4.48)
After some calculations (choosing a coordinate system, evaluating x
BA
,
squaring, summing over the two orthogonal polarization unit vectors and in-
tegrating over solid angle) one nds
w
AB+
=
e
2
4
4
3
3
[x
BA
[
2
=
4
3
3
[x
BA
[
2
, (4.49)
where we have introduced the dimensionless ne-structure constant

e
2
4
=
e
2
4 hc

1
137
. (4.50)
We note that since x
BA
is a vector operator (corresponding to spin one),
the selection rules for these transitions is that parity changes and [J
B
J
A
[ = 0
or 1, with the exception that 0 0 transitions are forbidden.
In fact, helicity conservation forbids 0 0 transition to all orders in pertur-
bation theory. Parity-conserving transitions (so-called quadrupole transitions)
can be induced by the terms we omitted in (4.45), however.
Let us apply our formalism to the processes A B + which can take
place in thermal equilibrium. According to the general principles of statistical
mechanics, equilibrium means that
N
B
w
abs
= N
A
w
emit
(4.51)
and the levels A and B are populated according to
N
B
N
A
=
e
E
B
/kT
e
E
A
/kT
= e
/kT
(4.52)
where T is the temperature and k is Boltzmanns constant. Using our eective
Hamiltonians (4.39) and (4.40), we can easily compute the ratio of transition
rates:
w
abs
w
emit
=
n
i
n
i
+ 1
, (4.53)
where hermiticity of the Hamiltonian was used to deduce that the square of
the two atomic matrix elements, that for emission and that for absorption, are
equal. We can now use
N
A
N
B
=
w
abs
w
emit
=
n
i
n
i
+ 1
(4.54)
to solve for n
i
,
n
i
=
1
e
/kT
1
. (4.55)
This is the famous Bose-Einstein distribution. If we consider a black body,
which has the capability of emitting and absorbing at all energies , we can
nd the total thermal emission at a given temperature. Since each photon
carries an energy , the total radiated energy per unit volume is
U()d = 2
1
L
3
e
/kT
1
_
L
2
_
3
4
2
d, (4.56)
where the factor 2 comes from the two possible polarizations. We thus have
arrived at Plancks famous radiation law which was the rst sign of the need
for quantum mechanics:
U() =
1
3
e
/kT
1
. (4.57)
4.1.5 The classical limit
In this section we shall elaborate on the earlier statement that the classical
e.m. eld is a coherent state of the quantum theory. We start by dening
quantum observables that are closely related to the classical amplitude and
phase by,
a =
aa
e
i
N + 1 e
i
(4.58)
a
= e
i
aa
= e
i
N + 1 , (4.59)
where we consider the operators at some xed time and introduced the number
operator N = a
a. From the commutation relation (4.19), we can now derive

[N, e
i
] = e
i
(4.60)
which implies,
e
i
[n) = [n 1) , (4.61)
where [ 1) = 0.
We see that quantum mechanically the amplitude and the phase of an
e.m. wave do not commute, and hence will satisfy an uncertainty relation.
The operators e
i
are not hermitian, so we will rather consider the hermitian

combinations,
cos

=
1
2
(e
i
+ e
i
) (4.62)
sin

=
1
2i
(e
i
e
i
) ,
with the commutation relations,
[N, cos

] = i sin

(4.63)
[N, sin

] = i cos

,
that follow from (4.60). From the fundamental uncertainty relation,

A

B
1
2
[[

A,

B])[ (4.64)
where

A = [(

A
A))
2
)]
1
2
(cf. MQM 1.4.53), we derive the following un-
certainty relation between the number and phase operators,
N cos

1
2
[sin

)[ (4.65)
N sin

1
2
[cos

)[ .
In an [n) state we can calculate the phase uncertainty from the easily
proven relations,
n[ cos

[n) = n[ sin

[n) = 0 (4.66)
n[ cos
2

[n) = n[ sin
2

[n) =
1
2
n ,= 0 . (4.67)
The result is (for n ,= 0),
cos

= sin

=
1
2
, (4.68)
corresponding to a completely undetermined phase. Similarly one can form
eigenstates of the phase-operators (4.62) that have a big uncertainty in the
number of photons (the expectation value of N is in fact divergent).
As in the case of the harmonic oscillator (see Section 4.3), the minimum
uncertainty states are the coherent states dened by,
a[) = [) . (4.69)
The classical electromagnetic eld corresponding to this state is given by,
A
cl
(x, t) = [
A(x, t)
op
[) =
h
2V
[[ e
i(t
k x)
+ c.c. (4.70)
where = [[e
i
, i.e. the amplitude and phase angle of the classical e.m. wave
is nothing but the amplitude and phase of the complex number that labels
the coherent photon state. Similar expressions can be given for the electric
and magnetic eld strengths. Let us now consider the quantum uctuations.
It is easy to establish (do it!) that [N[) = [[
2
and [N
2
[) = [[
4
+[[
2
,
so N = [[. The main observation is now that the relative uncertainty in
the particle number,
N
[N[)
=
1
[[
, (4.71)
vanish in the limit [[ 1, corresponding to large amplitudes, or a large
number of photons. After we have discussed the full quantum eld theory in
Section 4.2, we shall estimate how large n has to be for the classical description
to make sense.
Full quantum theory 39
To show that the coherent photon states indeed have minimum uncertainty,
we need the following relations which hold in the limit of large [[,
10
[ cos

[) cos
_
1
1
8[[
2
_
(4.72)
[ cos
2

[) cos
2
_
1
1
2[[
2
_
+
1
4[[
2
. (4.73)
The corresponding phase uncertainty is cos

=
1
2||
sin , and by combining
this with N = [[, one can easily demonstrate that the uncertainty relations
(4.65) are saturated.
4.2 The full quantum eld theory
4.2.1 Basics
We now treat the case of a general electromagnetic eld. Classically this can
always be resolved in eigenmodes by the following fourier decomposition,
A(x, t) =
1
2
k,
A
k
(t)
k
n
)e
i
k x
+ c.c. (4.74)
where

k = (n
x
, n
y
, n
z
)
2
L
, if we use periodic boundary conditions in a box with
volume V = L
3
. From (4.74) we easily obtain the corresponding expressions
for the electric and magnetic elds. The classical Hamiltonian is obtained by
generalizing (4.13),
H =
_
d
3
x
1
2
_
A(x, t)
2
+
2
A(x, t)
2
_
(4.75)
=
V
2
k,
2
k
A
k
(t)A
k
(t) ,
and we make the transition to the quantum theory by turning all the fourier
components of (4.75) into operators using (4.23),
A
op
(x, t) =
1
k,
h
2
k

k
(t)e
i
k x
+ h.c. . (4.76)
10
For the proofs of these relations, see e.g. the book by Loudon and references given
there.
The annihilation and creation operators satisfy,
[a
k
i
(t), a
k
j
(t)] =
k
i
k
j
. (4.77)
In terms of these operators, the quantum Hamiltonian reads,
H =
k,
h
k
(a
k
a
k
+
1
2
) , (4.78)
and from the Heisenberg equation of motion we deduce the time dependence,
a
k
(t) = e
i
k
t
a
k
, (4.79)
exactly as in the monochromatic case.
The Hilbert space, H
F
, obtained by acting repeatedly by the a
k
:s, is very
large, but constructed in complete analogy with the previous, monochromatic,
case. Clearly it will be the Fock space discussed in Section 3.2.1, the direct
sum of an innite tower of Hilbert spaces describing states with xed number
of photons,
H
F
= H
0
H
1
H
2
.... (4.80)
where H
0
only contains the vacuum state [0), where H
1
is spanned by a
k
[0)
and so on.
From (4.76,4.77) we can work out the following equal time commutation
relations between the operator elds

A
op
and

E
op
,
[

A
i
(x, t),

E
j
( y, t)] = i h(
ij
2
)
3
(x y) . (4.81)
Note the structure (
ij
2
) in the right hand side that is necessary for the
commutator to be consistent with the gauge condition

A = 0 (to see this,
apply
i
on both sides of (4.81)).
11
A quantum theory where the canonical variables are operator elds rather
than usual operators is called a quantum eld theory (QFT). It is clear from
(4.77) and (4.78) that the full radiation eld has innitely many degrees of
freedom (d.o.f.) corresponding to all the fourier modes. Alternatively (4.81)
shows that we can think of the

A
i
(x, t) and

E
j
(x, t) as dening a pair of
canonical variables at each point in space, i.e. one should regard the space
11
If you are worried about the denition of the operator 1/
2
, think of it in fourier space
where it is simply k
2
.
coordinate x as a label on the dynamical variables, in the same way as the
polarization index . From this way of counting d.o.f. it is clear that not only
does the whole system posses an innite number of d.o.f., but in a quantum
eld theory, there is an innite number of degrees of freedom in any nite
spatial volume. Of course, for a small box all the modes with wave lengths
larger than the box size will be absent, but there are still innitely many
high frequency modes. That an innite number of d.o.f. can cause trouble is
easy to understand in perturbation theory where one has to do sums over all
intermediate states. In QFT these sums often diverge, and for a long time this
made many physicists doubt its usefulness. Through work by Dyson, Feynman,
Tomonaga, Schwinger and others around 1950, and by Kadano, Wilson and
others in the early 1970-ies we now have a very good understanding of how
to tame these divergences. The fascinating theory of renormalization and
the renormalization group are central topics in any advanced course in QFT.
The Casimir eect, to be discussed later, provides a simple example of how
to handle a divergence.
From the way we constructed the photon theory, it is natural to consider the
non-hermitian annihilation and creation operators a
k
and a
k
, as the primary
variables, and the hermitian eld operators

A
op
(x, t) and

E
op
(x, t), merely as
their fourier transforms. However, in many cases, especially in the theory of
elementary particles, it is most convenient to start from the eld formulation.
The main thing to remember is nevertheless that QFT is the natural language
for describing quantum systems where the number of particles can change. This
is the reason why QFT is a very important tool also to describe condensed
matter systems. There the particles that can be created and destroyed are e.g.
phonons, magnons (quantized spin-waves), and Cooper pairs. Quite often it is
also convenient to use QFT to describe electrons in the context of condensed
matter theory. In this case particles cannot really be created or destroyed,
but in a large system with xed chemical potential the particle number is not
xed, and QFT turns out to be an appropriate language.
The QFT of photons developed above is very simple, and ideally suited to
understand many aspects of the quantum theory of light and matter, such as
radiation from classical sources and the function of lasers and masers. There
are, however, drawbacks with our formulation that become apparent if one
tries to combine it with relativistic electron theory (the so called Dirac the-
ory) to the full theory of quantum electrodynamics. The problem is that
the gauge condition (4.3) breaks Lorentz invariance, and this causes many
technical diculties. There are alternative formulations where the Lorentz
invariance is explicitly maintained (e.g. by using the so called Landau gauge
condition
= 0), but the proper place to study this is in a special course

in relativistic QFT.
4.2.2 The vacuum energy and the Casimir eect
From (4.78) it is clear that the Hamiltonian is diagonal in the states with a
xed number of photons of each frequency - each additional photon simply
adds h
k
to the total energy. Now let us nally also consider the zero point
energy
1
2
h
k
. Taken literally it implies the following expression for the energy
of the vacuum,
c
vac
=
1
2
k,
h
k
, (4.82)
which clearly diverges. One way out of this dilemma is to realize that only
energy dierences are physically meaningful, and thus simply subtract the
divergent c
vac
from all energies. That this procedure is too simple minded was
pointed out 1948 by H. Casimir, who studied the quantized radiation eld in
the presence of two innite conducting planes separated by a distance a. The
wave vectors allowed by the boundary conditions on the electric and magnetic
elds at the surface of a perfect conductor are
k = (k
x
, k
y
,
n
a
) , (4.83)
if the planes are at z = 0 and z = a. In order to calculate the vacuum energy,
we regulate the sum (4.82), and dene c
vac
by a limiting procedure:
c
vac
=
h
2
lim
0
d
d
k,
e

h
2
lim
0
d
d
() , (4.84)
where () is the thermodynamical partition function if we identify = =
1/kT. In the present case of two large plates, each with surface area A, and
separated by a distance a, () is given by
12
() =
A
4
2
_
dk
x
dk
y
_
2
n=0
e
[k
2
x
+k
2
y
+(n/a)
2
]
1
2
e
[k
2
x
+k
2
y
]
1
2
_
(4.85)
where A is assumed to be large enough for boundary eects to be ignored, and
at the end we shall take the limit A corresponding to innite planes.
12
If you have diculties deriving this expression, consult any textbook on classical elec-
trodynamics for the boundary conditions on a perfect conductor, and pay special attention
to the polarizations.
With a bit of eort one can evaluate (4.85) and by substitution in (4.84) we
obtain,
c
vac
=
3 haA
4

2
720
hA
a
3
. (4.86)
Note that the rst term, which diverges as 0, is aA, i.e. to the volume
enclosed by the two planes. The second term is nite as 0, and there are
no divergences 1/ or 1/
2
.
a
A
L
Figure 4.1: Schematic picture of a demonstration of the existence of the
Casimir eect. Two electrically neutral, conducting plates of area A are placed
at a distance a from each other, and the whole system is enclosed in a box of
side-length L with periodic boundary conditions. The presence of the plates
changes the total zero-point energy of the electromagnetic eld in such a way
that an attracive force appears between the plates.
Let us now assume that the plates are enclosed in a very large periodic box
as shown in the gure, and ask what is the dierence between the box with
the plates, and the box without the plates. We now redo the above calculation
in the regions in and out to get the result
c
in
vac
+c
out
vac
=
3 hV
4

2
720
hA
a
3

2
720
hA
(L a)
3
, (4.87)
where V = L
3
is the total volume of the large box. The Casimir energy,
c
Cas
is dened as the dierence between the ground state energy for the box
with the plates, and without them, in the limit of a very large outer box. The
divergent term V cancels, and in the limit L , we get
c
Cas
=

2
720
hA
a
3
. (4.88)
Since only energy dierences are experimentally observable, this is clearly the
important quantity. Another way to express the same thing is to note that
the force (per unit area) between the plates, i.e. the derivative of the energy
w.r.t. a, only depends on the Casimir energy, and is given by,
F/A =
1
A
dc
Cas
da
=

2
240
1
a
4
. (4.89)
The sign of this Casimir force correspond to an attraction between the plates
and has been measured in the laboratory (see, e.g. M. J. Sparnay, Nature,
180, 334 (1957)).
13
Clearly, it would be wrong to throw out the zero point
energy from the start, since then there would be no Casimir force, and the
lesson we learn is that the seemingly trivial empty space, or vacuum, is in a
sense not empty at all, but lled with uctuating electric and magnetic elds.
14
It actually turns out that in more complicated eld theories one of the main
(and often unsolved!) problems is to nd the vacuum state.
You might also worry about the assumption that the walls are perfect con-
ductors. All real materials are characterized by a complex dielectric function
(
k, ). The conductivity, which is the imaginary part of (
k, ), vanishes
at high frequencies, so the walls are transparent to the high energy modes.
We imposed the boundary conditions for all the modes up to innity, does
this not make our calculation invalid? It turns out not - the very fact that
c
Cas
does not depend on the cuto means that it is insensitive to the high
energy modes, and thus it does not matter whether they are aected by the
boundaries or not. The important lesson to be learned is that in spite of all
manipulations of divergent sums and integrals, the origin of the the Casimir
eect is the modication of the long wave length modes due to the boundaries,
and is thus a low energy phenomenon.
13
As a curiosity we mention that the corresponding force on a thin spherical conducting
shell is repulsive.
14
Those who feel uncomfortable with hand waving arguments can forget the words and
just remember the calculation. They might also be comforted by knowing that there is
claimed to be another way to explain the Casimir eect as an induced dipole - dipole
interaction between the charge carriers in the plates.
4.2.3 More about the classical limit
As promised earlier we shall now explain in more detail what is meant with a
large amplitude or large number of photons, when discussing the classical
limit. For this we rst notice that the divergence in the vacuum energy (4.82) is
also reected in that 0[
E
op
(x, t)
2
[0) and 0[
B
op
(x, t)
2
[0) diverge, even though
0[
E
op
(x, t)[0) = 0[
B
op
(x, t)[0) = 0. This does not look good, since there
seems to be no natural way to get rid of this divergence by a subtraction.
The way out is to realise that the eld in a point is not an observable, since
any experiment always measures the average value of the eld in some volume
V (l)
3
. For the average value,
E
av
(x, t) =
1
V
_
yV
E
op
(x +y, t) , (4.90)
one can derive the following relation for the uctuations,
0[
E
2
av
(x, t)[0)
h
(l)
4
. (4.91)
This should be compared with the (space and time) average of the square of
the classical electric eld corresponding to (4.70),
(
E
2
cl
(x, t))
av
=
[N[)
V
h = n
h
. (4.92)
where n is the average number of photons in the volume V . Taking l = ,
and requiring that the vacuum uctuations (4.91) are negligible compared with
the average of the classical eld (4.92), we arrive at the condition,
3
n 1 , (4.93)
i.e. for the classical description to be valid, the average number of photons in
a volume
3
must be much larger that one.
4.2.4 Momentum and angular momentum
We shall end the formal development by deriving expressions for the momen-
tum and angular momentum of the photon states. In particular we shall show
that the photon carries one unit of h of internal angular momentum, or spin.
In classical e.m. theory the eld momentum is given by the Poynting
vector,
P =
_
d
3
x
_
E(x)

B(x)
_
. (4.94)
We turn this into an operator by substituting for

E and

B, the corresponding
operator elds. A bit of algebra gives,
P
op
=
k,
h
k(a
k
a
k
+
1
2
) =
k,
h
kN
k
, (4.95)
where we introduced the photon number operator N
k
= a
k
a
k
. This ex-
pression is again in perfect agreement with the photon interpretation of light:
The momentum of the eld is associated with a collection of photons, each
one carrying momentum h
k. Also note that in this case, there is no problem

with the
1
2
h
k coming from the zero point uctuations, since these occur with
equal probability in all directions, which is just to say that the vacuum is
rotationally invariant.
A classical e.m. eld also carries angular momentum given by,
J =
_
d
3
xx [
E(x)

B(x)] , (4.96)
and this can also be turned into a quantum operator by direct substitution,
J
op
=
i
h
_
d
3
x[

E
i
(x p)

A
i
+

E
i
o
ij

A
j
] =

J
orb
+

J
sp
, (4.97)
where
o
a
ij
= i h
aij
. (4.98)
In (4.97) we decomposed the total angular momentum in an orbital and a spin
part as suggested by the operator

L = x p in the rst term and the explicit
h in the second. We shall now take a closer look at

J
sp
by expressing it in
creation and annihilation operators,
J
sp
= i h
k
[a
k1
a
k2
a
k2
a
k1
]
k . (4.99)
Introducing creation and annihilation operators for circularly polarized light
by
a
k
=
1
2
(a
k1
ia
k2
) (4.100)
and the corresponding a
k
and number operators N
k
, we can express the spin
angular momentum as,
J
sp
= h
k
(N
k+
N
k
)
k . (4.101)
Again the interpretation is clear, each circularly polarized photon carries h or
h of angular momentum around the direction of propagation.
We end with a comment on operator ordering. The expressions (4.94)
and (4.96) are ambiguous on the quantum level since the operators

E
i
(x) and
B
i
(x) do not commute. In the case of momentum and angular momentum, this
is no source of concern, since the correct form is dictated by the symmetry
properties of the ground state. In the case of the energy the situation is
more subtle. The form (4.78) corresponds to turning 2AA
into the operator

aa
+a
a, corresponding to the Hamiltonian E

2
+
2
A
2
, which does not suer
from ordering ambiguities. The ultimate argument for the correct ordering is,
as in the case of the harmonic oscillator, agreement with experiments.
4.3 Coherent states of the harmonic oscillator
Since we now have quantized the scalar eld and the electromagnetic eld with
the help of harmonic oscillator-like basis states, it is worthwhile to look back
at some solutions of the non-relativistic quantum harmonic oscillator system,
before discussing the dierent types of states of the quantized electromagnetic
eld (which we will do in Section 4.4).
In this Section we will assume knowledge about the operator method for the
harmonic oscillator (for example, Sect 2.3 in Modern Quantum Mechanics
by J.J. Sakurai). At a couple of places we refer to this book, and we also use
the same notation, notably x and p are operators, while the corresponding
eigenkets are [x
) etc.
4.3.1 What is a coherent state ?
Remember that the ground state [0), being a gaussian, is a minimum uncer-
tainty wavepacket:
Proof:
x
2
=
h
2m
(a + a
)
2
p
2
=
m h
2
(a a
)
2
Since
0[(a + a
)(a + a
)[0) = 0[aa
[0) = 1 (4.102)
0[(a a
)(a a
)[0) = 0[aa
[0) = 1 (4.103)
it follows that
x
2
)
0
p
2
)
0
=
h
2
4
1(1) =
h
2
4
and nally since x)
0
= p)
0
= 0, it follows that
(x)
2
)
0
(p)
2
)
0
=
h
2
4
(4.104)
We can now ask whether [n) is also a minimum uncertainty wave packet.
Corresponding to (4.102) and (4.103) we have
n[(a + a
)(a + a
)[n) = n[aa
+ a
a[n) = n[2a
a + [a, a
][n) = 2n + 1
Coherent states of harmonic oscillator 49
and similarly
n[(a a
)(a a
)[n) = (2n + 1)
which implies
(x)
2
)
n
(p)
2
)
n
=
h
2
4
(2n + 1)
2
(4.105)
so [n) is not minimal !
Clearly a crucial part in [0) being a minimal wave packet was
a[0) = 0 0[a
a[0) = 0
this corresponds to a sharp eigenvalue for the non-Hermitian operator mx+ip
even though, as we saw, there was (minimal) dispersion in both x and p. It
is natural to expect other minimal wave-packets with non zero expectation
values for x and p but still eigenfunctions of a, i.e.
a[) = [) (4.106)
which implies [a
a[) = [[
2
. It is trivial to check that this indeed denes a
minimal wave-packet
[(a + a
)[) = ( +
)
[(a a
)[) = (
)
[(a + a
)(a + a
)[) = ( +
)
2
+ 1
[(a a
)(a a
)[) = (
)
2
1
from which follows
(x)
2
)
= x
2
)
x)
2
=
h
2m
(p)
2
)
= p
2
)
p)
2
=
hm
2
and accordingly
(x)
2
)
(p)
2
)
=
h
4
(4.107)
So the states [), dened by (4.106), satisfy the minimum uncertainty rela-
tion. They are called coherent states and we shall now proceed to study them
in detail.
4.3.2 Coherent states in the n-representation
In the [n) base the coherent state look like:
[) =
n
c
n
[n) =
n
[n)n[) (4.108)
Since
[n) =
(a
)
n
n!
[0) (4.109)
we have
n[) =

n
n!
0[) (4.110)
and thus
[) = 0[)
n=0
n!
[n) (4.111)
The constant 0[) is determined by normalization as follows:
1 =
n
[n)n[) = [0[)[
2
m=0
[[
2m
m!
= [0[)[
2
e
||
2
solving for 0[) we get:
0[) = e
1
2
||
2
(4.112)
up to a phase factor. Substituting into (10) we obtain the nal form:
[) = e
1
2
||
2
n=0
n!
[n) (4.113)
Obviously [) are not stationary states of the harmonic oscillator, but we
shall see that they are the appropriate states for taking the classical limit.
A very convenient expression can be derived by using the explicit expression
(4.109) for [n):
n=0
n!
[n) =
n=0
n
n!
(a
)
n
[0) = e
a
[0)
which implies
[) = e
1
2
||
2
+a
[0) (4.114)
It is a good exercise to show that by using
e
A
e
B
= e
A+

B+
1
2
[

A,

B]
(4.115)
which holds if both

A and

B commute with [

A,

B], that this can also be written
as
[) = e
a
a
[0) = D()[0) (4.116)
where the exponential displacement operator D() acts as a creation op-
erator for the coherent state [).
4.3.3 Orthogonality and completeness relations
We proceed to calculate the overlap between the coherent states using (4.113).
[) =
n
[n)[) = e
1
2
||
2
1
2
||
2

n
(
)
n
n!
= exp (
1
2
[[
2
1
2
[[
2
+
) (4.117)
and similarly
[) = exp (
1
2
[[
2
1
2
[[
2
+
) (4.118)
so
[[)[
2
= [)[) = exp ([[
2
[[
2
+
)
or
[[)[
2
= e
||
2
. (4.119)
Since [) ,= 0 for ,= , we say that the set [) is overcomplete. There is
still, however, a closure relation:
_
d
2
[)[ =
_
d
2
e
||
2

m,n
(
)
n
n! m!
[m)n[ (4.120)
where the measure d
2
means summing over all complex values of , i.e.
integrating over the whole complex plane. Now, writing in polar form:
= re
i
d
2
= ddr r (4.121)
we get
_
d
2
e
||
2
(
)
n
m
=
_

0
dr re
r
2
r
m+n
_
2
0
de
i(mn)
= 2
m,n
1
2
_

0
dr
2
(r
2
)
m
e
r
2
= m!
m,n
Using this we nally get:
_
d
2
[)[ =
n
[n)n[ =
or equivalently
_
d
2
[)[ = 1 (4.122)
4.3.4 Coherent states in the x-representation
Remember that x
[0) is a minimal gaussian wave packet with x) = p) = 0.

Since x
[) is also a minimal wave packet and

'() = [
a + a
2
[) =
_
m
2 h
[x[) (4.123)
() = [
a a
2i
[) =
1
2m h
[p[) (4.124)
or
_
_
_
x)
=
_
h
m
2'()
p)
m h
2()
(4.125)
it is natural to expect that x
[) is a displaced gaussian moving with velocity

v = p/m. We now proceed to show this. Using the previously derived form
(4.114) of the coherent states expressed in terms of the ground state of the
harmonic oscillator we get
x
[) = x
[e
1
2
||
2
+a
[0) = e
1
2
||
2
x
[e
m
2 h
(x
ip
m
)
[0)
Acting from the left with x
[ gives us
15
:
x
[) = e
1
2
||
2
e
m
2 h
(x
i
m
(ih
d
dx
))
x
[0)
= Ne
1
2
||
2
e
x
0
2
(x
x
2
0
d
dx
)
e
1
2
(
x
x
0
)
2
(4.126)
15
c.f. MQM p.93
where we have used the explicit form of x
[0) and introduced the constants:

_
_
x
0
=
_
h
m
N =
4
_
m
h
(4.127)
For notational simplicity we also put y
= x
/x
0
. With these substitutions
equation (4.126) will look like:
x
[) = Ne
1
2
||
2
e
2
(y
d
dy
)
e
1
2
y
2
(4.128)
Using the commutator relation
e
A+

B
= e
1
2
[

A,

B]
e
A
e
B
(4.129)
which is equivalent to (4.115) and thus valid if both

A and

B commute with
[

A,

B], we get
e
2
(y
d
dy
)
e
1
2
y
2
= e
||
2
4
+

2
y
2
d
dy
1
2
y
2
(4.130)
and thus, noting that e
a
d
dy
, is a translation operator
x
[) = Ne
1
2
||
2
1
2
y
2
1
2
2
+
2y
= N exp (
1
2
(y
2'())
2
+ i
2()y
i()'()) (4.131)
Using (4.125) the resulting expression for the wavefunction of the coherent
state is
x
[) = Ne
m
2 h
(x
x)
2
+
i
h
px
i
2 h
px
(4.132)
and since the last term is a constant phase it can be ignored and we nally get
(x
) = (
m
h
)
1/4
e
i
h
px
m
2 h
(x
x)
2
, (4.133)
which is the promised result.
4.3.5 Time evolution of coherent states
The time evolution of a state is given by the time evolution operator
16
|(t).
Using what we know about this operator and what we have learned so far
about the coherent states we can write:
[, t) = |(t, 0)[(0)) = e
i
h
Ht
[(0)) = e
i
h
Ht
e
1
2
|(0)|
2

n
((0))
n
n!
[n)(4.134)
16
c.f. MQM chapter 2.1
But the [n):s are eigenstates of the Hamiltonian so:
[, t) = e
1
2
|(0)|
2

n
((0))
n
n!
e
i
h
h(n+
1
2
)t
(a
)
n
n!
[0) (4.135)
which is the same as
[, t) = e
1
2
|(0)|
2
e
i
2
t
n
((0)e
it
a
)
n
n!
[0) =
exp (
1
2
[(0)[
2
i
2
t + (0)e
it
a
)[0) (4.136)
Comparing this expression with (4.114), it is obvious that the rst and the
third term in the exponent, operating on the ground state, will give us a
coherent state with the time dependent eigenvalue e
it
(0) while the second
term only will contribute with a phase factor. Thus we have:
[, t) = e
i
2
t
[e
it
(0)) = [(t)) (4.137)
So the coherent state remains coherent under time evolution. Furthermore,
(t) = e
it
(0)
d
dt
(t) = i(t) (4.138)
or in components
_
d
dt
'() = ()
d
dt
() = '()
(4.139)
Dening the expectation values,
_
x(t) = (t)[x[, t)
p(t) = (t)[p[, t)
(4.140)
we get
_
_
_
d
dt
x(t) =
_
h
2m
2
d
dt
'() =
_
h
2m
2() =
p(t)
m
d
dt
p(t) = i
_
mh
2
(2i)
d
dt
() =
_
mh
2
2'() = m
2
x(t)
(4.141)
or in a more familiar form
_
p(t) = m
d
dt
x(t) = mv(t)
d
dt
p(t) = m
2
x(t)
(4.142)
i.e. x(t) and p(t) satisfy the classical equations of motion, as expected from
Ehrenfests theorem.
In summary, we have seen that the coherent states are minimal uncertainty
wavepackets which remain minimal under time evolution. Furthermore, the
time dependent expectation values of x and p satisfy the classical equations
of motion. From this point of view, the coherent states are very natural for
studying the classical limit of quantum mechanics.
States of quantized e.m. eld 55
4.4 States of the quantized electromagnetic eld
When we quantized the electromagnetic eld, it was convenient to use the
states [n), the eigenstates of the number operator N = a
a. However, we also
introduced the phase operator

dened by
a =
aa
e
i
(4.143)
One may ask oneself whether there exist states which are eigenstates of

.
This turns out not to be possible. However, one can construct states [) which
are nearly states of well-dened phase , namely
[) = lim
n
n
m=0
e
im
[m) (4.144)
It is not dicult to show that in the limit when n
[ cos

[) = cos (4.145)
[ sin

[) = sin (4.146)
and, furthermore,
[ cos
2

[) = cos
2
(4.147)
[ sin
2

[) = sin
2
(4.148)
This means that for the states [) the uncertainties of both sin and cos are
zero,
sin = cos = 0 (4.149)
However, since the state [) is formed by a superposition of innitely many
[n) states of dierent [n), it is easy to understand that n is completely unde-
ned. For the electric eld operator of a state with given wave-vector

k and
polarization r, Equation (4.76) gives
E(x, t) = i
_

2V
_
ae
it+i
k x
a
e
+iti
k x
_
(4.150)
which means that (verify this!) the phase is well-dened (remember that we
write = [[ exp(i)):
[
E[) sin
_
k x t +
_
(4.151)
Such a state is quite dierent from a classical state, since the amplitude
(which, at least for large n, where the dierence between n and n + 1 is
negligible is proportional to
n) is completely undened, but the phase of the

eld (e.g. the location of nodes) is well-dened.
The situation is the reverse for a state of given n. As we saw in Eq. 4.68,
there the phase is completely undened whereas
n[
E[n) = 0 (4.152)
n[
E
2
[n) =

V
_
n +
1
2
_
(4.153)
Thus, for a state with a xed n, the electric eld has an amplitude which can be
taken to be the square root of the quantity in (4.153). It is clear that both the
states [n) and [) are very dierent from the classical electromagnetic waves
you have encountered in classical electrodynamics. This was of course the
motivation behind the introduction of the coherent states of the last Section.
Since we have quantized the electromagnetic eld using the harmonic os-
cillator states [n), we can immediately take over the formalism of the non-
relativistic harmonic oscillator. The coherent states [) for the non-relativistic
oscillator have the property of minimizing the uncertainty relation between x
and p. In addition, it can be seen by making x and p dimensionless through
x
= x
m, p
= p/
m that the uncertainty is equally shared,

x
=
1
2
(4.154)
with
x
= p
=
1
2
(4.155)
For the electromagnetic eld, the coherent states also minimize the uncertainty
relation bewteen n and sin or cos . In addition, the uncertainty is equally
distributed in n and . For large values of [[, the fractional uncertainty in
both the photon number n and the phase decreases as 1/[[, i.e., the eld
looks more and more like a classical eld with growing .
It is interesting to note that according to (4.113) the probability to nd
the photon number n in a single-mode
17
coherent state [) is given by
[n[)[
2
= e
||
2 [[
2n
n!
(4.156)
17
Single-mode: we only consider the case when just one mode

k has an occupation number
dierent from zero.
Introduction 57
which is a Poisson distribution around the average [[
2
. In a laser, the state
of the electromagnetic eld is usually well described by such a single-mode
coherent state with Poissonian statistics corresponding to a high intensity, and
therefore high . For a steady intensity, the distribution of photon counts
in a detector which registers photons during a given time interval can also be
shown to be distributed according to a Poisson distribution around the average.
This is an intrinsic feature of the quantized electromagnetic eld which does
not only appear for coherent light and which cannot be eliminated in general.
If one tries to detect very faint light, the number of photons uctuates, creating
a stochastic noise (sometimes called shot noise) in the intensity signal. For
astronomical observations using CCD devices this is often a much noticeable
eect.
4.4.1 Squeezed States
The coherent states have the attractive property that they minimize the quan-
tum mechanical uncertainty, i.e., in the case of the non-relativistic harmonic
oscillator we had for the dimesionless variables x
and p
that
x
=
1
2
(4.157)
with
x
= p
=
1
2
. (4.158)
In, e.g., optical communications problems it may be that one wants to
minimize the uncertainty in one of the two conjugate variables x
and p
(it
may be that information is carried mainly by one of them). Is it possible to
nd states which have minimum uncertainty, but instead of (4.158) a smaller
uncertainty in, say x
at the expense of a larger uncertainty in p
? The answer
is yes, and we state without proof that such squeezed states are generated by
acting on the vacuum with the squeeze operator, which contains besides the
linear terms in a and a
in the exponential of the displacement operator in

(4.116) also quadratic terms:
[, ) = e
a
a
e
a
2
a
2
[0) (4.159)
Such states can be achieved through various procedures which most often in-
volve strong laser elds (so that the non-linear terms in the interaction Hamil-
tonian become excited).
58 The Dirac Equation
Chapter 5
The Dirac Equation
5.1 Introduction
The English physicist P.A.M. Dirac (1902-1984) was one of the founders of the
modern quantum eld theory. Many of his papers belong to the category of
classical papers and are well worth reading still today. His most important
contribution to physics is the relativistic equation for spin-1/2 particles (like
the electron, the quarks, the neutrinos and so on) which bears his name. It is
an ironical fact that it introduced partly for wrong reasons, but the equation
itself has stood the test of time. We will follow partly the historical path in
which the equation was for quite some time regarded as a Schrodinger-like
single-particle equation, before being seen as the equations of motion for a
relativistic quantum eld, similar to the scalar eld (x
) that we treated in
Chapter 3. It is an amazing fact that for spin-1/2 particles (fermions) which
obey the Pauli principle, one can in fact obtain a relativistic single-particle
equation which makes sense, by postulating that all the negative-energy states,
which caused the problems for the scalar eld, are already occupied (forming
the so-called Dirac sea). The Pauli principle will then forbid positive-energy
particles from making transitions to the negative-energy states in the Dirac
sea. For many applications, there is therefore no need to use the full machin-
ery of quantum eld theory. We will thus treat Dirac particles similarly to
how we treated the classical radiation eld in elementary quantum mechanics
(i.e. before turning the classical vector potential eld into a eld operator).
Towards the end of the chapter we will, however carry out the quantization of
the Dirac eld. Then you will have all the basic tools needed for a full course
in quantum eld theory.
59
Figure 5.1: P.A.M. Dirac (1902-1984).
5.1.1 Short biography of Dirac
Paul Adrien Maurice Dirac
1
was born on 8th August, 1902, at Bristol, Eng-
land, his father being Swiss and his mother English. He was educated at
the Merchant Venturers Secondary School, Bristol, then went on to Bristol
University. Here, he studied electrical engineering, obtaining the B.Sc. (Engi-
neering) degree in 1921. He then studied mathematics for two years at Bristol
University, later going on to St.Johns College, Cambridge, as a research stu-
dent in mathematics. He received his Ph.D. degree in 1926. The following
year he became a Fellow of St.Johns College and, in 1932, Lucasian Professor
of Mathematics at Cambridge.
Diracs work was concerned with the mathematical and theoretical aspects
of quantum mechanics. He began work on the new quantum mechanics as
soon as it was introduced by Heisenberg in 1928 - independently producing
a mathematical equivalent which consisted essentially of a noncommutative
algebra for calculating atomic properties - and wrote a series of papers on the
subject, published mainly in the Proceedings of the Royal Society, leading to
his relativistic theory of the electron (1928) and the theory of holes (1930).
This latter theory required the existence of a positive particle having the same
mass and charge as the known (negative) electron. This, the positron was
discovered experimentally at a later date (1932) by C. D. Anderson, while
its existence was likewise proven by Blackett and Occhialini (1933) in the
1
This section is taken from the web page http://www.nobel.se/laureates/physics-1933-
2-bio.html, c The Nobel Foundation.
Constructing the Dirac equation 61
phenomena of pair production and annihilation.
The importance of Diracs work lies essentially in his famous wave equation,
which introduced special relativity into Schrodingers equation. Taking into
account the fact that, mathematically speaking, relativity theory and quantum
theory are not only distinct from each other, but also oppose each other,
Diracs work could be considered a fruitful reconciliation between the two
theories.
Diracs publications include the books Quantum Theory of the Electron
(1928) and The Principles of Quantum Mechanics (1930; 3rd ed. 1947).
He was elected a Fellow of the Royal Society in 1930, being awarded the
Societys Royal Medal and the Copley Medal. He was awarded the Nobel
prize in physics in 1933. He was elected a member of the Pontical Academy
of Sciences in 1961.
Dirac travelled extensively and studied at various foreign universities, in-
cluding Copenhagen, Gttingen, Leyden, Wisconsin, Michigan, and Princeton
(in 1934, as Visiting Professor). In 1929, after having spent ve months in
America, he went round the world, visiting Japan together with Heisenberg,
and then returned across Siberia.
In 1937 he married Margit Wigner, of Budapest.
5.2 Constructing the Dirac equation
The Schrodinger equation for a free particle (V (r) = 0) is based on the non-
relativistic expression of kinetic energy
E
kin
=
p
2
2m
, (5.1)
which inserted into the energy equation H = E
kin
gives, with the prescription
H i h/t, p i h,
h
2
2
2m
(r, t) = i h
(r, t)
t
. (5.2)
An important property of this equation, and the reason why we can interpret
it as a single-particle equation, is that it implies a conserved current (the
probability current). The continuity equation for this current can be derived
by multiplying (5.2) from the left by
, its complex conjugate from the right

by and subtracting:
t
+ j = 0, (5.3)
where the probability current is
j(r, t) =
i h
2m
(
) , (5.4)
and the probability density
(r, t) = [(r, t)[
2
. (5.5)
Thus, if we start with a particle described by a wave function normalized
to unity, it will remain normalized during the time evolution.
We now check what will happen if we instead use the relativistic Klein-
Gordon equation (2.30)
_
h
2
2
+ m
2
c
2
_
= h
2

2
c
2
t
2
, (5.6)
thus
_
2+
2
_
= 0, (5.7)
with = mc/ h. We now set h = c = 1, so that = m. Multiplying (5.7)
from the left by
, the complex conjugate of (5.7) from the right by and

subtracting we nd
t
+ j = 0, (5.8)
with
j(r, t) =
i
2m
(
) (5.9)
and
(r, t) = i
_
t
_
. (5.10)
Although (5.8) looks like a continuity equation, it is not possible to inter-
pret in (5.10) as a probability density since it is not positive denite. For
instance, a plane wave with time dependence e
iEt
gives a negative value of ,
whereas the time dependence e
iEt
does not have this problem. The solution
is, as we saw in Chapter 3 to interpret as a quantum eld whose excita-
tions can be an arbitrary number of particles. Since the number of particles
can change, there is no reason to have a single-particle equation. Also, the
Hamiltonian of the quantum eld (3.6) is in fact positive denite. This was
not known to Dirac, however, and he tried another way out. The problem
of the negative energies comes from the fact that the time derivative in (5.7)
is of second order. In the Schrodinger equation, we have a rst order time
derivative but a second order space derivative. Could we perhaps nd a rst-
order equation? For a long time, this was thought to be impossible, because
relativity theory demands that we treat space and time in a similar way, and a
linear equation in space derivatives is not, for instance, invariant under space
rotations like the Schrodinger equation is.
The crucial step forward taken by Dirac was to introduce several wave-
function components into the theory and to nd a system of rst order dier-
ential equations in both space and time. Thus, he made the ansatz
H = i
t
= i
_
x
1
+
2
x
2
+
3
x
3
_
+ m, (5.11)
where and are some constant N N matrices to be determined, and
is a column vector with N components (as usual we do not write the unit
matrix explicitly). The idea is that each of the components
should satisfy
the relativistic Klein-Gordon equation (2+m
2
)
= 0. Multiplying (5.11) by
i/t, and using the equation itself to replace the time derivative on on the
right hand side, we nd
t
2
= M
ij
x
i
x
j
iN
j
m

x
j
+ m
2
2
(5.12)
(remember that we use the summation convention), where
M
ij
=

i
j
+
j
i
2
(5.13)
and
N
i
=
i
+
i
. (5.14)
We see that (5.12) becomes the diagonal Klein-Gordon equation only if
M
ij
=
ij
(5.15)
and
N
i
= 0, (5.16)
thus, introducing the anticommutator , :
A, B AB + BA, (5.17)
i
,
j
= 2
ij
, (5.18)
i
, = 0. (5.19)
In addition, we must have
2
= 1. (5.20)
From (5.18) it also follows that
2
i
= 1 (5.21)
for all i.
Of course, we still have to show that there exist matrices fullling these
relations. Also, we have to show that (5.12) is Lorentz invariant, i.e., that
it has the same form in all inertial frames. Some properties of the matrices
are easy to derive. Since the square of any of them is the unit matrix, the
eigenvalues have to be 1. From (5.19) we see that
i
=
i
. Taking the
trace of this equation and using Tr(AB) = Tr(BA) it is seen that the trace of
i
has to be zero. In the same way, Tr() = 0 is proven. Since the trace is
the sum of the eigenvalues, which are +1 or 1, we see that we must have an
even number of dimensions. The Pauli matrices
i
would almost do the job.
Since
i
,
j
= 2
ij
(5.22)
they could serve the role of
i
. However, the Pauli matrices span the space of
2 2 matrices together with the unit matrix
0
= 1. The unit matrix can not
be used as , however, since it commutes with all matrices in contradiction to
the requirement (5.19). We must thus consider 4 4 matrices. By trial-and-
error, Dirac found a solution which uses the Pauli matrices as 2 2 blocks in
the 4 4 matrices:
i
=
_
0
i
i
0
_
(5.23)
and
=
_
1 0
0 1
_
. (5.24)
The Dirac wave function thus can be considered as a column vector
consisting of four components
, = 1, 2, 3, 4. We dene as usual the

hermitian conjugate wave function
to be a row vector with components

(
1
,
2
,
3
,
4
). Let us check that we can now form a positive-denite prob-
ability from
. We multiply (5.11) from the left by
and the hermitian

conjugate of the equation from the right by and subtract the two equations
thus obtained. This gives (using the fact that both the
i
and matrices are
hermitian)
t
+ j = 0, (5.25)
where indeed is the positive denite quantity
, and
j =
, (5.26)
where we have assembled the component matrices
i
to a vector of matri-
ces .
To show Lorentz invariance, it is convenient to multiply (5.11) from the
left by and rearranging the terms to obtain
i
_
0

x
0
+
1

x
1
+
2

x
2
+
3

x
3
_
m = 0. (5.27)
Here we have introduced the very convenient notation
0
= and
i
=
i
.
This allows us to write the anticommutation relations in the suggestive form
= 2g
. (5.28)
In terms of block matrices,
i
=
_
0
i
i
0
_
(5.29)
and
0
=
_
1 0
0 1
_
. (5.30)
The Dirac equation (5.27) can now be written in a form that looks manifestly
Lorentz invariant:
(i , m) = 0, (5.31)
where the slash symbol will be used for an arbitrary contraction of a four-
vector with the set of gamma matrices, ,A
.
5.2.1 Plane-wave solutions
Let us consider solutions to the free Dirac equation (5.31) that can describe
plane waves of positive energy and momentum p, thus
p
(r, t) =
1
2EV
u(p)e
i(Etpr)
=
1
2EV
u(p)e
ipx
, (5.32)
where u(p) is a 4-column vector, and where we in the last expression use the
convenient abbreviation px = p
. Inserting this into (5.31) we nd the Dirac

equation in momentum space for u(p):
(,p m) u(p) = 0. (5.33)
In particular, we can ask for solutions at rest, p = 0. Then only the
0
/t
term contributes, and we nd with the representation (5.30) for the
0
matrix
(remember that E = m when p = 0)
2m
_
_
_
_
_
0 0 0 0
0 0 0 0
0 0 1 0
0 0 0 1
_
_
_
_
_
u(0) = 0, (5.34)
which shows that we can take as basis states only
u
1
(0) =
_
_
_
_
_
1
0
0
0
_
_
_
_
_
(5.35)
and
u
2
(0) =
_
_
_
_
_
0
1
0
0
_
_
_
_
_
. (5.36)
The two lowest components have to be zero to satisfy (5.34), but this leaves us
with the problem that we only have a basis for the two-dimensional subspace
spanned by the upper two components. Again, it is the negative energy states
that have to be involved. If we try instead the plane-wave ansatz corresponding
to negative energy (i.e., we let the four-momentum p p):
p
(x) =
1
2EV
v(p)e
+ipx
, (5.37)
then the Dirac equation at rest becomes
2m
_
_
_
_
_
1 0 0 0
0 1 0 0
0 0 0 0
0 0 0 0
_
_
_
_
_
v(0) = 0, (5.38)
which has solutions
v
1
(0) =
_
_
_
_
_
0
0
1
0
_
_
_
_
_
(5.39)
and
v
2
(0) =
_
_
_
_
_
0
0
0
1
_
_
_
_
_
. (5.40)
(We will see later that the negative-energy solutions are related to antipar-
ticles.) We can thus expand an arbitrary 4-component Dirac-spinor at rest
as
(t) = e
imt
_

0
_
+ e
+imt
_
0
_
, (5.41)
with and being 2-component spinors.
Let us now check what happens when p ,= 0. Then we can still write
u(p) =
_

_
(5.42)
where and now will depend on p. The Dirac equation was constructed to
give E = p
0
where we dene p
0
to be the positive quantity p
0
= +
m
2
+p
2
.
If we consider positive-energy solutions, inserting (5.42) into (5.33) implies
=
p
p
0
+ m
, (5.43)
and therefore
u(p) = N
_

p
p
0
+m
,
_
(5.44)
where N is a normalization constant. Similarly, the negative energy solutions
v(p) are given by
v(p) = N
_
p
p
0
+m
_
. (5.45)
The Dirac equation becomes, for these spinors
(,p m) u(p) = 0 (5.46)
and
(,p + m) v(p) = 0 (5.47)
respectively.
Let us dene the conjugate Dirac spinor

by
(x)
(x)
0
(5.48)
and thus
u(p) u(p)
0
(5.49)
and
v(p) v(p)
0
. (5.50)
Then using the identity
=
0
(
0
=
, (5.51)
which can be easily veried, we obtain for the conjugate spinors u and v the
equations
u(p) (,p m) = 0 (5.52)
and
v(p) (,p + m) = 0. (5.53)
By multiplying (5.46) from the left by u
, (5.52) from the right by
u,
summing these equations and using the anticommutation relation (easily de-
rived from (5.28))
, ,p = 2p
, (5.54)
we nd
m u(p)
u(p) = p
u(p)u(p) (5.55)
and
m v(p)
v(p) = p
v(p)v(p). (5.56)
Normalizing the 2-spinors and to unity (
= 1), we choose to
normalize the 4-spinors not to unity but rather by the condition
u(p)u(p) = 2m, (5.57)
v(p)v(p) = 2m (5.58)
which are relativistically covariant conditions. This means, for instance, that
u
u = 2E. Then, according to (5.55) and (5.56)

u(p)
u(p) = 2p
(5.59)
and
v(p)
v(p) = 2p
. (5.60)
This xes the normalization constant N to be
p
0
+ m, so that
u
r
(p) =
_
p
0
+ m
_

r
p
p
0
+m
r
,
_
, (5.61)
where we choose the two independent 2-spinor basis states to be
1
=
_
1
0
_
(5.62)
and
2
=
_
0
1
_
. (5.63)
Similiarly,
v
s
(p) =
_
p
0
+ m
_
p
p
0
+m
s
_
. (5.64)
5.2.2 Coupling to Electromagnetism
The easiest way to introduce electromagetism is as usual to make the minimal
coupling substitution
p
eA
, (5.65)
where A
is the electromagnetic four-potential. Going back to the Dirac equa-

tion in Hamiltonian form (5.11), this gives
i
t
= [ (p A) + m+ e] . (5.66)
5.2.3 Another way to arrive at the Dirac equation
Before going on to investigate the transformation properties of the Dirac wave
function under Lorentz transformations, it may be useful to see how one can
arrive at the equation using a quite dierent route. This starts with the
well-known fact from nonrelativistic quantum mechanics that to describe the
two spin degrees of freedom of, e.g., an electron, a spinor is introduced.
That is, we have a wave function which factorizes for a free particle into a
spin part describes by a two-component Pauli spinor and a space part that
is described by a space-dependent function (e.g., a plane-wave). The kinetic
energy operator for the system is given by
H
kin
=
p
2
2m
. (5.67)
However, since H
kin
is really a 2 2 matrix (in this case proportional to the
unit matrix, since the kinetic energy is the same for both the spin states), we
could also write this as
H
kin
=
( p)( p)
2m
. (5.68)
This is because the formula for the Pauli matrices
( A) ( B) = A B+ i (AB) , (5.69)
which is valid also if A and B are operators. In fact, (5.68) seems preferable to
use because doing the minimal coupling to the electromagnetic vector potential
A through
p p eA (5.70)
in (5.68) one gets automatically
2
the correct Pauli term
H
spin
=
e
2m
B (5.71)
where B is the magnetic eld.
Doing the same trick in the relativistic equation
E
2
p
2
= m
2
(5.72)
gives
(E p) (E + p) = m
2
, (5.73)
or substituting as usual E i/dt, p i,
_
i

x
0
+ i
__
i

x
0
i
_
= m (5.74)
where is a two-component wave function. We now have the problem that
it is a second-order equation, so we rewrite it as a system of two rst-order
equations. Introducing
=
1
m
_
i

x
0
i
_
(5.75)
the equation (5.74) becomes
_
i

x
0
i
_
= m (5.76)
_
i

x
0
+ i
_
= m. (5.77)
2
It is a good exercise to derive this, by using (5.69) and the operator formula p A =
i(A) Ap.
Introducing the new two-component spinors
A,B
= , it is easily shown
that this is nothing but the Dirac equation with the same representation of
the gamma matrices as we had before, i.e.,
(i
m) = 0, (5.78)
with
=
_

A
B
_
, (5.79)
so that
A
and
B
are the same as and of (5.42).
5.3 Lorentz invariance
Although we have written the Dirac equation (5.31) in a way that looks Lorentz
invariant:
(i
m) (x) = 0, (5.80)
we have to demand that the Dirac matrices and the Dirac wave function trans-
form in the required way. According to the relativity postulate, we have to
nd in a primed system
x
(5.81)
(e.g. a Lorentz-boosted inertial frame) an equation of the same form:
_
i
m
_
(x
) = 0, (5.82)
(that m
= m follows from the fact that the rest mass is itself an invariant
quantity). Let us try if we can use the same constant Dirac matrices (i.e.
- that this is possible this can in fact be proven) but make a linear
transformation of the Dirac spinor:
(x
) = S(x), (5.83)
with inversion
(x) = S
1
(x
). (5.84)
Then insertion into (5.31) gives
_
i
m
_
S
1
(x
) = 0, (5.85)
where we used (1.14). Multiplying from the left by S we nd
_
iS
S
1
m
_
(x
) = 0. (5.86)
This is of the required form (5.82) if we can nd an S such that
S
S
1
, (5.87)
or
= S
1
S. (5.88)
We now have to nd S such that (5.88) is fullled. We rst consider a
Lorentz boost along the x
1
direction (cf. Eq. 1.7), but write for convenience
= tanh . (5.89)
Lorentz invariance 73
Then
(x
1
; ) =
_
_
_
_
_
cosh sinh 0 0
sinh cosh 0 0
0 0 1 0
0 0 0 1
_
_
_
_
_
, (5.90)
and we see that for (5.88) to be fullled
S
1
0
S = (cosh )
0
(sinh )
1
S
1
1
S = (sinh )
0
+ (cosh )
1
S
1
2
S =
2
(5.91)
S
1
3
S =
3
.
Using the anticommutation relations of the gamma matrices, this can be sum-
marized as (exercise: prove this!)
S
1
S =
_
cosh

2
+
0
1
sinh

2
_
_
cosh

2

0
1
sinh

2
_
, (5.92)
which shows that
S = cosh

2

0
1
sinh

2
. (5.93)
The transformation matrices S for other Lorentz transformations can be found
in a similar way, they can be summarized by
S = cosh

2

0
i
sinh

2
(5.94)
for a boost with boost parameter along the x
i
axis and
S = cos

2

j
k
sin

2
(5.95)
for a rotation of an angle around the x
i
-axis (i, j, k = 1, 2, 3 cyclic).
In a similar way, it can be shown that the conjugate Dirac equation
i

(x)
(x) = 0 (5.96)
will be of the same form in a primed system provided that
(x
) =

(x)S
1
. (5.97)
5.3.1 Bilinear forms
To summarize, we have found the transformation matrix S that species how
the four components of a Dirac spinor mix under a Lorentz transformation.
Since S is not equal to , the Dirac spinors are not four-vectors (in fact, the
four spinor components live in an internal space which is not space-time).
However, we see and interesting thing from (5.88). Using the fact that
and

transform as (5.83) and (5.97) respectively, it is seen that
V

(x)

(x)
(x) (5.98)
transforms as a four-vector:
V
(x
) =
(x). (5.99)
Simliarly, we see that with
s(x)

(x)(x) (5.100)
s
(x
) = s(x), (5.101)
i.e. s(x) is a scalar quantity.
There exists an important 4 4 matrix, which is lineary independent from
the set
of four matrices, but can be constructed from their product:
5
i
0
3
. (5.102)
In our standard representation it becomes
5
=
_
0 1
1 0
_
. (5.103)
It satises
_
5
_
2
= 1 (5.104)
and
5
,
= 0. (5.105)
The
5
matrix plays an important role when discussing how Dirac spinors
transform under parity. Let us see if we can nd an operator P acting on a
Dirac spinor that can represent the parity transformation of the space-time co-
ordinates t t, r r, i.e. x
= (t, r). The Lorentz transformation

matrix which achieves this is
(
P
)
=
_
_
_
_
_
1 0 0 0
0 1 0 0
0 0 1 0
0 0 0 1
_
_
_
_
_
, (5.106)
which has det() = 1 and thus not a proper transformation (the latter is de-
ned as one which can be continuously reached from the unit transformation).
The requirement that the Dirac equation has the same form in the primed
system, Eq. 5.88, is seen to be solved by
P =
P
0
, (5.107)
where
P
is a phase factor
3
.
Since the
5
matrix anticommutes with
0
,
P
1
5
P =
5
= det(
P
)
5
. (5.108)
For a general Lorentz transformation the
5
matrix thus transforms as
S
1
5
S = det()
5
. (5.109)
Thus, it changes sign if we make a parity transformation x x, (but t t)
and remains unchanged for proper Lorentz tranformations. It has the same
properties as the box product a(b c) for 3-vectors. The bilinear form
s
5
(x)

(x)
5
(x) (5.110)
consequently behaves as a pseudoscalar quantity. The bilinear
A

(x)
5
(x) (5.111)
transforms under proper Lorentz tranformations like a four-vector, but is a
pseudovector under parity transformations. Thus,
A
(x
) = det ()
(x). (5.112)
Let us compute how many 44 matrices we have now introduced. Each of the
sets
and
5
contains 4 matrices. Together with
5
and the unit matrix we
thus have 10 matrices. There should thus exist 6 more linearly independent
matrice to span the 16-dimensional space of 4 4 matrices. A convenient
choice of the remaining matrices is
i
2
[
] . (5.113)
3
For a spin-1/2 particle, it can be shown to have the possible values 1 and i. This
means that in general it takes four reections to bring the wave function back to its original
value, similar for the case of rotations where for fermions it takes a rotation through 4
radians.
Since this set is antisymmetric in and it indeed contains 6 independent
matrices. The bilinear constructed from
,
T
(x)

(x)
(x) (5.114)
can be shown to transform like a rank-2 tensor (it is an antisymmetric tensor):
T
(x
) =
(x). (5.115)
5.3.2 Spin and energy projection operators
Since the Dirac equation describes spin-1/2 particles, we should be able to
nd 4-dimensional analogs to the 22 Pauli matrices
i
. In fact, they are not
dicult to nd. If we dene
i
by
i
=
1
2
ijk
jk
=
i
2
ijk
k
, (5.116)
they have the form (in our usual representation of the gamma matrices)
i
=
_

i
0
0
i
_
. (5.117)
The commutation relations for
i
are then the same as for
i
_
i
,
j
_
= 2i
ijk
k
, (5.118)
which means that
i
/2 satises the commutation relations of angular momen-
tum. We also nd trivially (since it is the same as for the matrices) that
i
_
i
2
_
2
=
3
4
=
1
2
_
1
2
+ 1
_
, (5.119)
so that we indeed are dealing with a spin-1/2 particle.
In calculations it is sometimes helpful to use the formula
i
=
5
0
, (5.120)
which is easy to derive using the denition of
5
. We have in (5.120) only de-
ned the space-like part of the spin operator. To describe the spin operator in a
covariant way we should nd a four-vector that becomes
= (0,
1
,
2
,
3
)
in the rest frame. The four-dimensional generalization of
ijk
is the four-
dimensional Levi-Civita tensor
. Since the four-momentum vector is

p
= (m, 0, 0, 0) in the rest frame we see that
=
1
2m
(5.121)
is the operator we are looking for. To get the spin operator for a given direction
n we can similarly introduce a four-vector n
which has the value n
= (0, n)
in the rest frame. Then the covariant operator n
will refer to this spin

direction in any inertial frame. In particular, we can nd the relativistic analog
of the two-component spin-up and spin-down projection operators in the z-
direction
P
=
1
z
2
(5.122)
by introducing n
z
= (0, 0, 0, 1) and writing
P
=
1
z
2
=
1
5
,n
z
0
2
. (5.123)
Here we can write the
0
matrix as ,p/m and use that ,p acting on a free spinor
gives m. If we dene the spinor u
1
(p) to have spin up in the z-direction and
v
1
(p) to have spin down, and vice versa for u
2
(p), v
2
(p) (we will see later
the motivation for this opposite assignment of spin), we can for both types of
spinors use
P
=
1
5
,n
z
2
(5.124)
as spin projection operators.
We can nd projectors also for positive and negative energies. The Dirac
operator for positive-energy spinors ,p m gives zero on any spinor which
has been multiplied by ,p + m (see Eq. 5.46; use p
2
= m
2
) and since the
operator ,p + m gives in a similar way gives negative-energy spinors (see
Eq. 5.47), we can expect them to be suitable projection operators, apart from
normalization. Let us call the normalized projection operators
and
+
respectively. Then we must have (as for all projection operators)
2
2
+
=
+
,
+
+
= 1 and
+
=
+
= 0. We see that
+
=
,p + m
2m
(5.125)
and
=
,p + m
2m
(5.126)
have these required properties.
A related set of equations, often convenient to use in calculations, is
r
u
r
(p) u
r
(p) = (,p + m)
(5.127)
and
s
v
s
(p) v
s
(p) = ( ,p + m)
. (5.128)
Another set of projection operators which are important, e.g., when de-
scribing the weak interactions of fermions are the so-called chirality operators
P
L

1
5
2
(5.129)
and
P
R

1 +
5
2
. (5.130)
They enable a Lorentz-invariant decomposition of an arbitrary Dirac spinor in
to left- and right-chirality components:
L,R
P
L,R
. (5.131)
It can be shown that in the limit of zero mass chirality is equal to helicity, i.e.
the projection of the spin onto the direction of motion. An eigenstate of P
L
, for
instance, has its spin oriented opposite to the direction of motion (and is said to
describe a left-handed fermion). In the modern theory of elementary particles
(quarks and leptons) it seems that the chiral states are the fundamental states
of the theory. (For instance, the left- and right-chirality states may interact
with dierent strength - for neutrinos it seems that the right-handed states do
not interact at all!)
5.4 Non-relativistic limit
We now investigate the non-relativistic limit of the Dirac equation coupled to
electromagnetism
i
t
= [ (p eA) + m+ e] . (5.132)
If we start with a positive-energy solution the time dependent Dirac spinor
wave function is of the form
= e
mt
_

_
(5.133)
Non-relativistic limit 79
where we have taken out the large t dependence coming from the rest energy
m, and where thus and should be slowly varying functions of t. Inserting
this into (5.132) one nds
i

t
_

_
= (p eA)
_

_
+ e
_

_
2m
_
0
_
. (5.134)
Here we can approximate the solution to the lower equation by
=
(p eA)
2m
(5.135)
which inserted into the lower equation gives
i
t
=
_
(p eA) (p eA)
2m
+ e
_
, (5.136)
which, as we noted in connection with (5.71), gives the correct Pauli term
for a spin-1/2 particle. This is seen even more clearly by checking what the
equation gives for the interaction with a weak magnetic eld B in which case
A =
1
2
Br, expanding (5.136) to lowest order in A:
i
t
=
_
p
2
2m

e
2m
[L + 2S] B
_
. (5.137)
Here we see that the total angular momentum J consists of two parts, the
orbital angular momentum L = r p and a spin part gS with S
i
=
i
/2 the
spin operator appropriate for a spin-1/2 particle, and the gyromagnetic ratio
g = 2.
Thus, we see that the Dirac equation has the attractive property of being a
relativistic equation with the correct non-relativistic limit to describe spin-1/2
particles like electrons.
If we want to compute the next order corrections to (5.136), it is conve-
nient to choose a slightly dierent basis for the Dirac matrix which makes
the Hamiltonian diagonal to that order. A systematic scheme for doing this
order by order was devised by Foldy and Wouthuysen in 1950. The idea is to
transform

= e
iS
, (5.138)
leading to
i
t
=
_
e
iS
_
H i

t
_
e
iS
_
= H
. (5.139)
By using
H
= H + i[S, H] +
(i)
2
2!
[S, [S, H]] + . . . (5.140)
and choosing S so that the nondiagonal terms vanish one nds after a straight-
forward though tedious calculation (see [9]) to second order:
H
2nd
=
_
m+
(p eA)
2
2m

p
4
8m
3
_
+ e
e
2m
B
ie
8m
2
(E)
e
4m
2
(E p)
e
8m
2
E. (5.141)
In this formula one observes relativistic corrections to the kinetic energy and
spin-orbit interactions. The last term (the so-called Darwin term) is specic
to the Dirac equation and can be interpreted as a smearing of the electrostatic
potential caused by the very fast vibrations (zitter bewegung) of the electron
at short distances.
Problems with Dirac equation 81
5.5 Problems with the Dirac equation
5.5.1 The Dirac sea
We have seen that we can nd a relativistic rst-order linear wave equation
which describes spin-1/2 particles such as electrons. The problem is that as
we have seen it also allows negative-energy solutions. How shall we interpret
them? Dirac found a nice intermediate solution before quantum eld the-
ory for electrons was constructed. He postulated the existence of a sea of
electrons occupying all the negative-energy states. Since electrons obey the
Pauli exclusion principle, there is no risk that positive-energy electrons will
make transitions into this sea by radiating photons (gamma rays) of huge en-
ergy > 2m, thus avoiding a catastrophic instability of the theory. However,
the inverse process should be possible in this picture: by absorbing a gamma
ray of energy greater than 2m, a negative-energy electron could be lifted to
a positive-energy unoccupied state, leaving a hole in the Dirac sea behind
(see Fig. 5.2). If the sea electron had negative-energy E, charge e, and
momentum p, the new sea state has energy +E, charge +e and momentum
p more than the original sea state. The hole will thus behave as a positive-
energy particle with positive charge, with momentum p. It thus behaves as a
positron! The process where one negative-electron is lifted by a gamma-ray to
a positive-energy state can thus be interpreted as the creation of an e
+
e
pair
from the vacuum (allowed by the energy provided by the gamma ray). Indeed,
if there is a hole already present, a positive-energy electron can fall into that
hole under emission of a gamma ray. It is then equivalent to electron-positron
annihilation.
How can we describe the presence of positrons (i.e. the absence of negative-
energy electrons in the Dirac sea) by the Dirac equation? The key point is
that the charge is opposite. Therefore, since the negatively charged electrons
are described by
(i , e ,A m) = 0 (5.142)
we would like to nd a charge conjugation transformation on such that
the transformed Dirac wave function
c
fulls instead
(i , + e ,A m)
c
= 0. (5.143)
If we take the complex conjugate of (5.142) we nd, using that A
is real,
__
i

x
+ eA
_
(
+ m
_
= 0. (5.144)
E
+m
-m
Figure 5.2: The energy spectrum of the free Dirac equation. Solutions with
E > m and E < m are possible (there is a mass gap of size 2m). However,
in the picture of the Dirac sea, the negative-energy states are all occupied.
By absorption of a photon of energy greater than 2m, one of the negative-
energy electrons can be lifted to an unoccupied positive-energy state, leaving
a hole behind. The physical interpretation of such a process is production of
an electron-positron pair.
This gives a hope to obtain (5.143) if we can nd a matrix S
C
with the property
S
C
(
S
1
C
=
. (5.145)
It is a good exercise to check that the matrix
S
C
= i
2
(5.146)
fulls this equation. Consequently, the charge conjugate Dirac eld
c
can be
represented by
c
= i
2
. (5.147)
If we introduce C by S
C
= C
0
, which is solved by C = i
2
0
, we see that we
can also write
c
= C

T
(5.148)
where we recall the denition (5.48) for

.
Let us check how the charge-conjugation operator acts on our basis states.
Problems with Dirac equation 83
Consider a free negative-energy electron at rest with spin down, i.e.

_
_
_
_
_
0
0
0
1
_
_
_
_
_
e
+imt
. (5.149)
The charge-conjugated state will have the wave function
c
= C

T
=
_
_
_
_
_
1
0
0
0
_
_
_
_
_
e
imt
. (5.150)
We see that
c
are the appropriate wave functions to use for the hole states, and
we get the reasonable result that the absence of a spin-down negative-energy
electron is equivalent to the presence of a spin-up positive-energy positron.
5.5.2 The Klein paradox
Although we can make sense of the Dirac equation by postulating a lled
Dirac sea for the unwanted negative-energy states, we have at the same time
sacriced the possibility of treating the Dirac spinor (x) as a single-particle
wave function like the Schrodinger equation. This leads us almost inevitably to
the view of a quantized electron eld, which we will treat in the next Section.
Before doing this, we will point out another strange feature of the Dirac
equation, rst found by Oskar Klein.
Suppose that we solve the Dirac equation for a potential which is a step
function in the z-direction:
V (z) = 0 z < 0
V (z) = V
0
z > 0, (5.151)
(5.152)
where V
0
> 0. This is analogous to the scattering problems you have seen in
elementary quantum mechnics courses. let us prepare the incident wave to be
a positive-energy plane-wave with spin up in the z-direction, i.e.
inc
= Ae
ik
i
z
_
_
_
_
_
1
0
k
i
E+m
0
_
_
_
_
_
. (5.153)
Then the reected wave will be
refl
= Be
ik
i
z
_
_
_
_
_
1
0
k
i
E+m
0
_
_
_
_
_
, (5.154)
and there may also be a transmitted wave for z > 0:
trans
= Ce
iktrz
_
_
_
_
_
_
1
0
ktr
EV
0
+m
0
_
_
_
_
_
_
, (5.155)
where we have chosen the phases such that A, B and C are real and positive
and where the Dirac equation requires that
k
2
tr
= (E V
0
)
2
m
2
. (5.156)
Since the Dirac equation is rst order in the derivatives, we can only demand
that the function (not its derivative) is continuous at the step. This gives
A + B = C (5.157)
and
AB =
k
tr
k
i
E + m
E V
0
+ m
C. (5.158)
From (5.156) and (5.158) we see a mysterious eect if V
0
> E +m, namely
that A B is real and negative, i.e. the reected wave is bigger than the
incident wave! This eect is the Klein paradox, and it was casting a shadow
on the Dirac equation for a long time before eld quantization was introduced.
The interpretation is that the extremely strong forces due to the steep potential
barrier cause spontaneous production of electron-positron pairs. The Dirac
equation as we have treated it until now thus has its limitations when dealing
with very strong forces and thus very large potential gradients. However, for
a wealth of other problems, e.g. in atomic physics and quantum chemistry,
it is often good enough to treat the Dirac spinor (x) as a single-particle
wave function and calculate relativistic corrections along the line discussed in
Section 5.4.
Some applications 85
5.6 Some applications of the Dirac equation
5.6.1 The hydrogen atom
When treating the hydrogen atom in non-relativistic quantum mechanics, it is
convenient to use the fact that for a central potential, the angular momentum
operator
L = r p (5.159)
commutes with the Hamiltonian and can therefore be simultaneously diago-
nalized with it. For a given l we can couple the spin of an electron to a given
j = l 1/2. Let us see if we can do the same in the Dirac theory. With
H = p + m+ V (r), (5.160)
where V (r) is a spherically symmetric potential, we nd that
[H, L] = i p, (5.161)
so L does in fact not commute with H. However, if we compute the commu-
tator of H with the spin operator (see Eq. 5.117) we nd
[H, ] = 2i p, (5.162)
so that the total angular momentum operator
J = L +
1
2
(5.163)
commutes with the Hamiltonian.
Thus, we can nd simultaneous eigenfunctions of H, J
2
and J
z
with corre-
sponding eigenvalues E, j(j + 1) and m, but not to L
2
and L
z
. We therefore
must allow each of the two-spinors and to be linear combinations of the
two possible values of l corresponding to j = l + 1/2,
(+)
j,m
=
_
_
_
l+1/2+m
2l+1
Y
m1/2
l
_
l+1/2m
2l+1
Y
m+1/2
l
_
_
(5.164)
and j = l 1/2 (only for l > 0)
()
j,m
=
_
_

_
l+1/2m
2l+1
Y
m1/2
l
_
l+1/2+m
2l+1
Y
m+1/2
l
_
_
. (5.165)
Here the Y functions are the usual spherical harmonics. Including also the
radial dependence of the wave functions, we can write a general solution as
j,m
=
_
_
g
+
j
(r)
r

(+)
j,m
+
g
j
(r)
r

()
j,m
if
+
j
(r)
r

(+)
j,m
+
if
j
(r)
r

()
j,m
_
_
. (5.166)
In fact,
(+)
j,m
and
()
j,m
have opposite parity and are related to each other by
(+)
j,m
=
r
r

()
j,m
. (5.167)
Using these results, and inserting this form of
j,m
into the Dirac equation
with electrostatic potential energy
V (r) =
Z
em
r
, (5.168)
where the ne-structure constant
em
1/137, one nds the coupled system
of equations:
_
E m+
Z
em
r
_
G
+
j
=
dF
+
j
dr
+
dF
+
j
r
(5.169)
_
E + m+
Z
em
r
_
F
+
j
=
dG
+
j
dr
+
dG
+
j
r
, (5.170)
(and corresponding ones for F
, G
), with
= 2(j l)(j + 1/2). (5.171)
These equations can be solved using standard methods of mathematical physics.
The eigenvalues are
E
n
=
m
1 +
_
Zem
n(j+1/2)+
(j+1/2)
2
(Zem)
2
_
2
. (5.172)
The quantum number n = 1, 2, . . . is a positive integer, n j + 1/2. An
expansion in powers of Z
em
gives
E
n
= m
_
1
1
2
(Z
em
)
2
n
2

1
2
(Z
em
)
4
n
3
_
1
j + 1/2

3
4n
_
. . .
_
. (5.173)
We see that this formula gives the correct non-relativistic result for the
energy levels. The relativistic corrections are small and have the property
(thanks to the 1/(j + 1/2) term in Eq. 5.173) that states with higher j are
at higher energy. This so-called ne-structure splitting is in agreement with
observations.
To get an even more correct spectrum one has to add a couple of small
eects. First, there is a hyper-ne shift (a triplet-singlet splitting of the ground
state) due to the interaction between proton and electron spins. Secondly there
is an interaction between the electron and the uctuating electromagnetic eld
(th Lamb shift). This last eect is an interesting application of the zero-
point energy of the innitely many quantum harmonic oscillators describing
the electromagnetic eld, and is one of the many triumphs of quantum eld
theory.
5.6.2 Coulomb scattering
We now compute the dierential cross section for the scattering of an electron
on a heavy nucleus. If the electron energy is low compared to the mass of the
nucleus, we can regard the latter to be stationary. The eect of the nucleus
can therefore be represented by a static Coulomb potential
V (r) =
Z
r
(5.174)
(we now drop the subscript on
em
).
To use the formalism of time-dependent perturbation theory, we should
write the Hamiltonian as H = H
0
+H
I
, where H
0
is the free Dirac Hamiltonian,
and (see Eq. 5.160) H
I
= V (r). The transition matrix element between initial
momentum and spin p
i
, r and nal momentum and spin p
f
, s caused by H
I
is
thus
M
fi
= p
f
; s[H
I
[p
i
; r) =
_
d
3
r
f
(r)V (r)
i
(r). (5.175)
Here we insert our unperturbed plane-wave Dirac spinors:
i
(r) =
1
2E
i
V
u
(r)
(p
i
)e
i(E
i
tp
i
r)
, (5.176)
f
(r) =
1
_
2E
f
V
u
(s)
(p
f
)e
i(E
f
tp
f
r)
, (5.177)
where we have quantized in a large volume V , and where the four-spinor u is
normalized such that u
u = 2E.
When inserting these expressions into (5.175) we nd the integration over
d
3
r to give essentially the Fourier transform of the potential V (r) (this is of
course the Born approximation):
_
d
3
r
r
e
iqr
=
4
q
2
, (5.178)
where q = p
f
p
i
. According to Fermis golden rule, we should now sum over
the possible nal states of the same energy. In the momentum interval d
3
p
f
there are V d
3
p
f
/(2)
3
such states. Using the energy delta function (E
f
E
i
)
which comes from the time independence of the Coulomb potential, using
p
f
dp
f
= E
f
dE
f
(which follows from the energy equation E
2
= p
2
+ m
2
), we
nd after inserting the various factors in the golden rule (exercise: ll in the
intermediate steps)
d
d
=
(Z)
2
q
4
u
(s)
(p
f
)u
(r)
(p
i
)
2
. (5.179)
For given initial and nal momenta and spins, we can compute the value
of [u
u[
2
to get the dierential cross section. Often we are, however, interested
in the unpolarized cross section which means that we take the average over
the two possible inital spins and sum over the two possible nal spins. Thus,
we want to compute
d
unpol
d
=
(Z)
2
q
4
1
2
r,s
u
(s)
(p
f
)u
(r)
(p
i
)
2
. (5.180)
Of course we can do this by inserting all the four combinations of spins r, s
and summing them separately. We will, however introduce a technique which
has proven to be very useful also in more complicated cases (and which is easy
to implement using computer algebra). First, we note that (since u = u
0
and (
0
)
2
= 1)
u
(s)
(p
f
)u
(r)
(p
i
) = u
(s)
(p
f
)
0
u
(r)
(p
i
). (5.181)
Also, for any 4 4 matrix , we can write
u
(s)
(p
f
)u
(r)
(p
i
)
2
=
_
u
(s)
(p
f
)u
(r)
(p
i
)
_
_
u
(s)
(p
f
)u
(r)
(p
i
)
_
. (5.182)
However,
_
u
(s)
(p
f
)u
(r)
(p
i
)
_
=
_
u
(r)
(p
i
)
_
u
(s)
(p
f
)
_
_
=
_
u
(r)
(p
i
)
u
(s)
(p
f
)
_
,
(5.183)
with
=
0
0
. (5.184)
We now use that
s
u
(s)
(p
f
) u
(s)
(p
f
) =,p
f
+ m, (5.185)
so that (writing out the Dirac indices and - remember that we use the
summation convention also for them)
r,s
u
(s)
(p
f
)u
(r)
(p
i
)
2
=
r
u
(r)
(p
i
)
_
0
[ ,p
f
+ m]
0
_
u
(r)
(p
i
). (5.186)
Here we can move u
(r)
(p
i
) to the far right, and using that
r
u
(r)
(p
i
) u
(r)
(p
i
) =,p
i
+ m, (5.187)
we see that the unpolarized spin sum can be written as
r,s
u
(s)
(p
f
)u
(r)
(p
i
)
2
= Tr
_
0
(,p
f
+ m)
0
(,p
i
+ m)
_
. (5.188)
5.6.3 Trace formulas
We see that according to (5.188) an unpolarized cross section can be written in
terms of a trace of a product of matrices acting in the four-dimensional Dirac
spinor space. Since this is a very general method to compute cross sections,
it is convenient to summarize some of the properties of such traces. We rst
note that
Tr(I) = 4, (5.189)
where I is the 44 unit matrix. Also, we saw before that due to the anticom-
mutation relations
Tr (
) = 0 (5.190)
for all the four gamma matrices. Also
Tr
_
5
_
= 0, (5.191)
as is obvious from the representation (5.103). We can use the fact that (
5
)
2
=
I and that
5
anticommutes with all
to prove that the trace of any product

of an odd number of gamma matrices vanishes (here we do not write which
index each gamma matrix has):
Tr (
1
2
. . .
2k+1
) = Tr
_
2
. . .
2k+1
_
=
Tr
_
2
. . .
2k+1
5
_
= (1)
2k+1
Tr (
1
2
. . .
2k+1
) = 0, (5.192)
where in second step we used the cyclic property of the trace, and in the nal
step we anticommuted one of the
5
matrices through all the 2k + 1 gamma
matrices, each step giving a factor of (1).
For the product of two gamma matrices, the result is easy to compute:
Tr (
) = Tr (
) =
1
2
Tr (
) = g
Tr(I) = 4g
. (5.193)
If a four-vector p
n
is contracted with each gamma matrix this formula can be
written
Tr (,p
1
,p
2
) = 4 (p
1
p
2
) . (5.194)
For an even number n > 2 we can use a recursion formula:
Tr (,p
1
,p
2
. . . ,p
n
) = p
1
p
2
Tr (,p
3
,p
4
. . . ,p
n
)
p
1
p
3
Tr (,p
2
,p
4
. . . ,p
n
) + . . . + p
1
p
n
Tr (,p
2
,p
3
. . . ,p
n1
) . (5.195)
This formula is proven by using
,p
1
,p
2
= ,p
2
,p
1
+ 2p
1
p
2
(5.196)
to move p
1
successively one step to the right at a time until it is last in the
product. When that is the case we use the cyclic property to move it back to
the rst position, and then the formula follows.
An important example is for n = 4, which gives
Tr (,p
1
,p
2
,p
3
,p
4
) = 4 (p
1
p
2
p
3
p
4
p
1
p
3
p
2
p
4
+ p
1
p
4
p
2
p
3
) . (5.197)
It is easy to show that (exercise: do this, using the denition of
5
)
Tr
_
5
,p
1
,p
2
_
= 0. (5.198)
The trace of a product of
5
with four gamma matrices is, however, non-zero:
Tr
_
5
,p
1
,p
2
,p
3
,p
4
_
= i
1
p
2
p
3
p
4
, (5.199)
a rule that can be veried for one particular order of gamma matrices (e.g.
3
), and using the antisymmetry property of the result.
With this detour, we are now ready to compute the result (5.180), (5.188).
It becomes
d
unpol
d
=
(Z)
2
2q
4
Tr
_
0
,p
f
0
,p
i
+ m
2
_
0
_
2
_
. (5.200)
Quantization 91
Using our rules for computing traces, this can now be evaluated:
d
unpol
d
=
(Z)
2
2q
4
_
8E
f
E
i
4p
f
p
i
+ 4m
2
_
. (5.201)
Introducing spherical coordinates with the polar axis along the incident direc-
tion, one nds, with = v/c = p/E
d
unpol
d
=
(Z)
2
4p
2
2
sin
4
_
2
_
_
1
2
sin
2
_
2
__
. (5.202)
This is called the Mott cross section, and reduces to the Rutherford formula
in the non-relativistic limit 0.
5.7 Quantization of the Dirac eld
So far, we have treated the dirac eld as a classical eld, without quantizing
it. The rst guess would be that to quantize the Dirac eld we can use the
canonical formalism and interpret the coecients of the plane-wave expansion
as creation and annihilation operators. Since we have a complex eld with
four components
, we would need to introduce a
and b
similarly to how
we treated the complex Klein-Gordon eld in (3.14). However, this does not
really work for the Dirac eld, which as we have seen describes spin-1/2 par-
ticles. This is due to the requirement of Fermi statistics which means that
the Pauli principle has to be obeyed. For instance, a two-particle state must
be antisymmetric in the exchange of the two particles. Also, we cannot have
more than one particle in any state of given spin and momentum.
We have seen that in our description of scalar eld quanta in terms of ele-
mentary excitations obtained by acting on the vacuum with creation operators
a
k
, we can get a state with n quanta of the same energy and momentum by
acting n times with a
k
. This is perfectly OK for bosons, but of course violates
the Pauli princple for fermions. Moreover, if we describe a state with two
particles of dierent momenta k and k
by acting on the vacuum state:

[0 . . . 1
k
. . . 1
k
. . .) = a
k
a
k
[0) , (5.203)
we see that it is symmetric when we exchange the particle labels:
a
k
a
k
[0) = a
k
a
k
[0) (5.204)
because of the commutation relation
_
a
k
, a
_
= 0. (5.205)
If we want to describe fermions with creation operators c
r,k
and c
s,k
, we
would instead expect antisymmetry:
c
r,k
c
s,k
[0) = c
s,k
c
r,k
[0) . (5.206)
Furthermore, we should have
_
c
r,k
_
2
= 0 (5.207)
for all r and k since we can only put one particle in each state. Both of these
requirements can be met if we postulate that the creation and annihilation
operators for fermions obey anticommutation relations instead of commutation
relations:
_
c
r,k
, c
s,k
_
=
_
c
r,k
, c
s,k
_
= 0. (5.208)
Instead of
_
a
r
, a
s,k
_
=
rs
k,k
(5.209)
it is then reasonable to postulate also
_
c
r,k
, c
s,k
_
=
rs
k,k
. (5.210)
Indeed, it turns out that this way of quantizing the Dirac eld is the correct
one. (The so-called spin-statistics theorem states that one has to quantize
elds of half-integer spin with anticommutators and elds of integer spins
with commutators to get a theory which is relativistically invariant and which
respects causality.) To describe the quantized complex Dirac we thus write
(r, t) =
r,p
1
2EV
_
c
r,p
u
(r)
(p) e
ipx
+ d
r,p
v
(r)
(p) e
ipx
_
, (5.211)
where
_
c
r,p
, c
s,p
_
=
_
c
r,p
c
s,p
_
=
_
d
r,p
, d
s,p
_
=
_
d
r,p
d
s,p
_
= 0 (5.212)
and
_
c
r,p
, c
s,p
_
=
_
d
r,p
, d
s,p
_
=
rs
p,p
. (5.213)
5.7.1 Majorana particles
We have assumed that the Dirac eld is complex, which is indeed necessary
to describe charged particles as we saw also for the scalar Klein-Gordon eld.
Lagrangian 93
However, there are other spin-1/2 particles in nature which do not carry elec-
tric charge. Examples are neutrons and neutrinos. Are they their own an-
tiparticles and can they in that case be described by a real Dirac eld? In the
case of the neutron, the answer is no. There is another type of charge, called
baryon number, which seems to be conserved to accuracy in nature, and which
distinguishes a neutron from an antineutron.
In the case of neutrinos, the question is still open. In the Standard Model of
particle physics neutrinos carry a lepton number which also makes a dierence
between particles and antiparticles. However, in extensions of the model,
neutrinos could be their own antiparticles and should then be decribed by a
real eld, called a Majorana eld. Actually, whether the eld is redepends
on the choice of the representation of the gamma matrices. There exists one
representation, the Majorana representation where the gamma matrices are in
fact real, and then the Dirac-Majoran eld in that representation is real. If we
stick to our standard representation of gamma matrices, the requirement (the
so-called Majorana condition) is that the eld is equal to its charge-conjugate,
i.e., (see Eq. 5.148)
=
c
= C

T
(5.214)
with C = i
0
2
. Since (check this using explicit expressions for u and v!)
u
(s)
(p) = C v
(s)T
(p) (5.215)
and
v
(s)
(p) = C u
(s)T
(p) (5.216)
it follows that the appropriate expansion of a Majorana eld
M
is
M
(r, t) =
r,p
1
2EV
_
c
r,p
u
(r)
(p) e
ipx
+ c
r,p
v
(r)
(p) e
ipx
_
. (5.217)
Whether the neutrino is a Majorana particle or not can have observable conse-
quences. Since a Majorana particle can have no conserved additive quantum
number (like electric charge or lepton number), exotic decays of nuclei are
possible which violate lepton number conservation. One example is neutrino-
less double beta decay, where a nucleus changes its charge by two units upon
emission of two beta particles (electrons). So far, none has been observed.
In so-called supersymmetric theories, Majorana fermions play a crucial role.
In particular, the supersymmetric partner of the photon, which can be stable
and perhaps make up the dark matter in the Universe, is a Majorana particle.
5.8 Lagrangian formulation
You now have all the basic tools for computing quantities in quantum eld the-
ory. What you will learn in courses on that subject is how to do perturbation
theory (and sometimes even estimate nonperturbative eects) in a Lorentz
invariant way. It has turned out to be very convenient to use the Lagrangian
formulation of eld theory, since that makes Lorentz invariance manifest (and
also various other symmetries that may exist in the theory).
It is not dicult to nd a Lagrangian density which has the Dirac equation
as its Euler-Lagrange equations of motion. If we consider the complex Dirac
eld, we should treat
and
as independent variables, just as we treated

and
as independent variables for the complex scalar eld. Since
and
are the same degrees of freedom (linearly related by the matrix

0
), and

is more useful when constructing Lorentz covariant bilinears, we prefer to use
and

as independent variables. You should check that the Lagrangian

density
L =

(x) [i , m] (x) (5.218)
gives the Dirac equation. From this we can also verify that the Hamiltonian
has the form that we expect from (5.11). The canonically conjugate momenta
are
=
L
= i
. (5.219)
and

=
L
0. (5.220)
The Hamiltonian is then
H =
_
d
3
r

(x)
_
i
j

x
j
+ m
_
(x) . (5.221)
Inserting the quantized Dirac eld (5.211) and performing the integration,
one nds with the use of the anticommutation relations (excluding the zero-
mode contribution)
H =
r,p
E(p)
_
c
r,p
c
r,p
+ d
r,p
d
r,p
_
. (5.222)
If we introduce the minimal coupling to the electromagnetic eld through
i
eA
we see that the interaction term becomes of the form eA
,
with the electric four-current density
j
(x) =

(x)
(x) (5.223)
Bibliography 95
and the total electric charge
Q = q
_
d
3
rj
0
= e
_
d
3
r

0
= e
r,p
_
c
r,p
c
r,p
d
r,p
d
r,p
_
. (5.224)
These relations show that although negative energy states appear in the so-
lutions to the Dirac equation, the total energy associated with the quantized
complex Dirac eld is positive denite. Also, although the energy contributed
by the d-particles is positive, the charge has the opposite sign to that of
the c-particles. Thus, the natural interpretation is that the theory contains
positive energy electrons and positive energy positrons.
96 Bibliography
Bibliography
[1] J. J. Sakurai, Modern Quantum Mechanics, 2
nd
ed., Addison-Wesley,
1994.
[2] J. J. Sakurai,Advanced Quantum Mechanics, Addison-Wesley, 1967.
[3] F. Mandl and G. Shaw, Quantum Field Theory, John Wiley & Sons.
[4] R. P. Feynman, QED - the Strange Theory of Light and Matter, Princeton
Univ. Press., 1985.
[5] R. Loudon, The Quantum Theory of Light, Oxford University Press, 1973.
[6] W. Rindler, Introduction to Special Relativity, Oxford University Press,
1982; E.F. Taylor and J.A. Wheeler, Spacetime Physics, Freeman, 1992.
[7] J.D. Jackson, Classical Electrodynamics, Wiley, 1975.
[8] H. Goldstein, Classical Mechanics, Addison-Wesley, 1980.
[9] J.D. Bjorken and S.D. Drell, Relativistic Quantum Mechanics,
McGraw-Hill 1964.
97
Index
absorption
of light, 32
action, 11
Aharonov-Bohm eect, 26
angular momentum
of electromagnetic eld, 42
anticommutation relations, 90
antiparticles, 23
bilinear forms
of Dirac spinors, 73
Bohr, N., 20
canonical quantization, 17
Casimir eect, 37, 38
Casimir energy, 40
Casimir, H., 39
charge conjugation
in Dirac theory, 80
chirality projectors
in Dirac theory, 77
classical limit, 41
of quantized electromagnetic eld,
32
coherent state
wavefunction of, 49
coherent states, 44
completeness and orthogonality
properties of, 47
in x-representation, 48
matrix elements of, 51
of quantized electromagnetic eld,
32
of the electromagnetic eld, 30
time evolution of, 50
conjugate Dirac spinor, 68
contravariant
see four-vector, 5
Coulomb gauge, 26
Coulomb scattering
in Dirac theory, 86
covariant
see four-vector, 5
dark matter, 92
de Broglie, L., 19
Dirac equation, 63
alternative derivation of, 69
coupling to electromagnetism, 69
Lorentz invariance of, 71
Lorentz transformations, 72
negative-energy solutions of, 67
nonrelativistic limit of, 78
plane-wave solutions of, 65
spin and energy projectors, 75
Dirac sea, 79
Dirac spinor, 67
normalization of, 69
Dirac, P.A.M., 19, 59, 60
biography of, 61
Ehrenfests theorem, 52
Einstein, Albert, 19
emission
of light, 32
Euler-Lagrange equations, 13, 14
98
Index 99
Fermis golden rule, 86
eld
denition of, 16
ne structure
in hydrogen atom, 85
ne-structure constant, 85
Fock space, 22
Foldy-Wouthuysen transformation,
79
four-scalar, 5
four-vector
contravariant, 5
covariant, 5
gamma matrices
denition of, 65
matrices constructed from, 73
generalized coordinates, 11
elds as, 28
ground state, 22
Hamiltons principle, 11
Hamiltonian, 27
denition of, 13
of electromagnetic eld in vac-
uum, 28
Heisenberg operators, 29
Heisenberg, W., 20
hole theory
in Dirac equation, 81
hydrogen atom
in Dirac theory, 83
Jordan, P., 19
Klein paradox, 23, 82
Klein, Oskar, 16, 19
Klein-Gordon equation, 16, 62
Klein-Gordon eld, 20, 27
Lagrangian, 11
Lorentz boost, 3
Lorentz covariance
of Dirac equation, 65
Lorentz scalar
see four-scalar, 5
Lorentz transformation
discrete, 4
group structure, 4
pure, 4
Majorana particle, 91
Maxwells equations, 19
Minkowski
metric, 4
monochromatic light, 26
negative-energy solutions, 23
number operator
parity operator
in Dirac theory, 74
Pauli matrices, 64
Pauli priciple, 90
Pauli term
in Dirac theory, 70
phase operator, 33
phonons, 20
photon
concept of, 19
energy and momentum of, 8
Poisson bracket, 17
polarization
of light, 26
of photons, 35
positrons
in Dirac theory, 80
positrons
in hole theory, 82
Poynting vector, 41
probability current, 62
100 Index
probability density, 62
QFT, 37
quantization
quantized Dirac eld, 91
quantum electrodynamics, 20
quantum eld theory, 19
quntum eld theory, 37
renormalization group, 37
rotation
3-dimensional matrix, 3
scalar eld
real, 16
Schrodinger, E., 20
second quantization, 20
Sparnay, M.J., 40
spin
of photon, 41
spontaneous emission, 32
stimulated emission
of radiation, 32
summation convention, 3
tensor
second rank, 6
trace formulas
for Dirac matrices, 88
ultraviolet catastrophe, 19
units
system of, 8
unpolarized cross section, 87
vacuum uctuations, 41
Wigner, E., 19
Young modulus, 15

Relkvtyuuant

Enviado por

Dados do documento

Título original

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

Relkvtyuuant

Enviado por

Direitos autorais:

Formatos disponíveis

Lecture Notes in Relativistic Quantum

(we see that this just amounts to changing the

, one nds by inserting

a constant four-vector, we would have 4 additional parameters. The full

under Lorentz transformations is called a contravariant four-vector. A set

is called a covariant four-vector. A quantity that,

use the metric

to lower the index of A

to obtain the covariant four-vector A

. The usefulness of four-vectors and scalars in special relativity is due to

. This can also be

is an invariant under Lorentz transformations, i.e.,

is proportional to the unit matrix

can be expressed in terms of A

gives from (1.26)

, which has the same constant value in all

, which is antisymmetric in all

the rst Maxwell equations become

f(r, t), we can use this freedom to impose a gauge condition on A

= 0 (Lorentz gauge), A = 0 (Coulomb gauge) or A

2V as a convenient normalization, and where

as independent elds, the Euler-Lagrange equations become

, respectively. When we couple electromagnetism to this

k[ = k, [B[ = [E[ and the polarization vector , satises

k) form a right-handed ON-basis. Light with circular, or in

E ,= 0) so that Gauss law is satised.

A. If you are still not familiar with the conventions

(t) are more convenient to use than

0, t). Recall that in the radiation gauge,

0, t) can be considered as generalized po-

0, t)), and the equations (4.8) and (4.9) de-

(t), corresponding to the (complex) amplitudes A(t) and

changes the number of photons, they will be

0, 0) are represented by multiplication by A and the derivative operator,

as creating, or annihilating quanta of energy.

n + 1[n), we obtain, neglecting the term of order e

a. From the commutation relation (4.19), we can now derive

are not hermitian, so we will rather consider the hermitian

= 0), but the proper place to study this is in a special course

k, ). The conductivity, which is the imaginary part of (

k. Also note that in this case, there is no problem

into the operator

a, corresponding to the Hamiltonian E

[0) is a minimal gaussian wave packet with x) = p) = 0.

[) is also a minimal wave packet and

[) is a displaced gaussian moving with velocity

[0) and introduced the constants:

n) is completely undened, but the phase of the

m that the uncertainty is equally shared,

at the expense of a larger uncertainty in p

in the exponential of the displacement operator in

, its complex conjugate from the right

, the complex conjugate of (5.7) from the right by and

, = 1, 2, 3, 4. We dene as usual the

to be a row vector with components

. We multiply (5.11) from the left by

and the hermitian

. Inserting this into (5.31) we nd the Dirac

, (5.52) from the right by

u = 2E. Then, according to (5.55) and (5.56)

is the electromagnetic four-potential. Going back to the Dirac equa-

of four matrices, but can be constructed from their product:

= (t, r). The Lorentz transformation