Relativistic Quantum Mechanics

J.Pearson
November 18, 2009
Abstract
These are a set of notes I have made, based on lectures given by G.Shaw at the Uni-
versity of Manchester Jan-May ’09. Please e-mail me with any comments/corrections:
jon@jpoffline.com. These notes may be found at www.jpoffline.com.
ii
CONTENTS iii
Contents
1 Preliminaries 1
1.1 Relativistic Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 The Dirac δ-function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Homogeneous Lorentz Transformations . . . . . . . . . . . . . . . . . . . . . 4
1.4 Equation of Continuity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.4.1 Electromagnetic Fields . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.5 Quantum Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.5.1 Interpretation of the Wavefunction . . . . . . . . . . . . . . . . . . . 8
1.5.2 Minimal Electromagnetic Interactions . . . . . . . . . . . . . . . . . . 9
1.6 Relativistic Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.7 Natural Units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2 The Klein-Gordon Equation 13
2.1 Negative Energies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2 Conserved Current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3 EM Fields and the “Hydrogen Atom” . . . . . . . . . . . . . . . . . . . . . . 15
2.3.1 The Klein Paradox . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3 The Dirac Equation 21
3.0.2 Properties of α
i
, β & Dirac Representation . . . . . . . . . . . . . . . 22
3.1 Conserved Current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.2 Angular Momentum & Spin . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.3 Plane Wave States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.4 Dirac Hole Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.5 Relativistic Covariance of The Dirac Equation . . . . . . . . . . . . . . . . . 32
3.5.1 Covariant Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.5.2 Proof of Covariance . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.5.3 Lorentz Boost . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
iv CONTENTS
3.5.4 Parity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.6 Interaction With Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.6.1 Magnetic Moment of Electrons . . . . . . . . . . . . . . . . . . . . . 38
3.7 Particles in a Spherical Potential . . . . . . . . . . . . . . . . . . . . . . . . 41
3.7.1 The MIT Bag Model . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.7.2 Hydrogen-like Atoms . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4 Quantum Fields 47
4.1 Quantum Mechanics of a String . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.1.1 Phonons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.1.2 Field Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.2 The Klein-Gordon Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.3 The Dirac Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.3.1 Fermions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.3.2 The Complex Dirac Field ψ(x) . . . . . . . . . . . . . . . . . . . . . 58
4.4 The Feynman Propagator G
F
(x) . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.5 Interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.5.1 The Interaction Picture . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.5.2 The S-matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4.5.3 The S-matrix Expansion . . . . . . . . . . . . . . . . . . . . . . . . . 69
5 Decays 73
5.1 Simple 2-body Decay: φ
3
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
5.2 Decay Rates & Lifetimes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
5.3 Example: Decay of a Scalar to two Scalars . . . . . . . . . . . . . . . . . . . 82
6 Scattering Processes 89
6.1 An Outline of Particle Exchange . . . . . . . . . . . . . . . . . . . . . . . . . 89
6.2 Other Scattering Processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
6.3 Scattering Cross-sections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
CONTENTS v
6.3.1 Centre of Momentum Frame . . . . . . . . . . . . . . . . . . . . . . . 96
6.4 Scattering From a Classical Field . . . . . . . . . . . . . . . . . . . . . . . . 97
7 Feynman Rules 101
vi CONTENTS
1
Non-relativistic quantum mechanics is concerned with systems for whom the kinetic energy
is a lot less than the rest mass, so that
T ≈ E −mc
2
=
1
2
mv
2
=
p
2
2m
.
Under this, no particles are created, and everything moves slowly. That is, if a system starts
out with one particle, it will always have one particle.
Now, relativistic quantum mechanics is concerned with systems for whom the energy is
more than the rest mass,
T ,= mv
2
.
Following this, one can make two distinctions:
T < mc
2
, T > mc
2
.
The former case does not have enough energy to create particles, the latter is able to produce
particles. The former case requires relativity, but not particle creation. The latter needs the
full formalism of quantum field theory.
In discussing a relativistic quantum theory, we are able to understand electron spin, the
electrons magnetic moment, predict the existence of anti-particles, the spin-statistics theorem
etc.
1 Preliminaries
Let us introduce the notation we shall be adopting.
1.1 Relativistic Notation
Let us introduce the contravariant 4-vector
x
µ
= (ct, x, y, z) = (ct, x).
Let us also introduce the metric tensor, such that
g
µν
= diag(1, −1, −1, −1);
where all off-diagonal components are zero. Using this, we can define a covariant 4-vector,
x
µ
= g
µν
x
ν
,
so that
x
µ
= (ct, −x).
2 1 PRELIMINARIES
We shall use the Einstein convention of implied summation over repeated indices.
We further define the inverse metric, as
g
λµ
g
µν
= δ
λ
ν
.
Hence, using these results, we can see
g
σµ
x
µ
= g
σµ
g
µν
x
ν
= δ
σ
ν
x
ν
= x
σ
,
which is
x
σ
= g
σµ
x
µ
.
Therefore, we see that we can use the metric to lower raised indices, and the inverse metric
to raise lower indices.
By the form of our metric and inverse, we see that they have the same components,
g
µν
= g
µν
.
We will commonly write expressions such as
px ≡ p
µ
x
µ
= Et −p x,
and φ(x) ≡ φ(t, x). Thus, by writing x, we mean the 4-vector (i.e. all four components), and
by writing x, only the 3-spatial components.
1.2 The Dirac δ-function
As we shall use the δ-function regularly, it is worth our noting a few of its properties. One
of its definitions is that
_

−∞
dx f(x)δ(x −x

) = f(x

). (1.1)
That is, one can think about the δ-function as only returning a value when its argument is
zero. In the context of the integral above, the integral “sweeps” over the real line, making the
variable x take all possible values. Only when x = x

does the δ-function return a non-zero
value. Hence, the only contribution from the integral is at x = x

, so that f(x

) is returned.
Thinking about the δ-function as a filtering mechanism can be very useful. For the δ-function
to be used directly in this way, its argument must be of the form (x − x

), where x is the
integration variable and x

some constant. If instead we have a functional argument, such as
δ (g(x)), then we must employ a certain tactic to get it into a useable form.
Suppose we have, as a simple case, δ(a(x −x

)), where a is a constant. Then, let us work
out how to evaluate
_

−∞
dx f(x)δ(a(x −x

)). (1.2)
1.2 The Dirac δ-function 3
Let us first change variables, to
y = ax, y

= ax

,
then, upon substitution into (1.2), we see that
_
∞×a
−∞×a
dy
a
f(
y
a
)δ(y −y

).
Because we always want this form of the integral (i.e. of the limits), we must take the
modulus of a; hence, we have
1
[a[
_

−∞
dyf(
y
a
)δ(y −y

) =
1
[a[
f(
y

a
),
where we have just used the defining property of the δ-function (1.1). An immediate con-
sequence of this is that the δ-function is even, δ(−x) = δ(x). Let us then consider a full
functional argument, δ(g(x)). Now, the δ-function only returns a non-zero value when its
argument is zero (before this was just at x = x

). Hence, we must find the zeros of the
function g(x). Supposing that we now have the zeros, at, say, x
i
, so that g(x
i
) = 0, let us
make a Taylor expansion of g(x) about the zero,
g(x
i
+ ) = g(x
i
) + (x −x
i
)g

(x
i
) + . . . .
The first term is zero, by definition of the zero. Hence,
g(x) ≈ (x −x

)g

(x
i
).
Therefore,
δ(g(x)) = δ (g

(x
i
)(x −x
i
)) .
Now, as g

(x
i
) is “just a number”, we can treat it exactly as our a before, so that
δ(g(x)) = δ (g

(x
i
)(x −x
i
)) =
1
[g

(x
i
)[
δ(x −x
i
).
If there are multiple zeros of the function g(x), we must sum over all zeros. Hence,
δ(g(x)) =

x
0
1
[g

(x
i
)[
δ(x −x
i
), g(x
i
) = 0. (1.3)
Therefore, we have a (very useful) relation which allows us to recast a δ-function with func-
tional argument, into a useable form.
We may have a multi-dimensional δ-function δ
(n)
(x −x

), which we define
δ
(n)
(x −x

) =
n

i=1
δ(x
i
−x

i
).
4 1 PRELIMINARIES
So, for example,
δ
(4)
(x −x

) = δ(x
0
−x
0
)δ(x
1
−x
1
)δ(x
2
−x
2
)δ(x
3
−x
3
)
= δ(t −t


(3)
(x −x

).
We can also define the δ-function to be the Fourier transform of unity:
δ
(4)
(x −x

) =
_
d
4
k
(2π)
4
e
−ik(x−x

)
.
Of course, the inverse Fourier transform is then
δ
(4)
(k −k

) =
_
d
4
xe
ix(k−k

)
.
At this stage we have merely given a few properties of an abstract function; these properties
will prove invaluable when discussing quantum fields.
1.3 Homogeneous Lorentz Transformations
Such transformations (between frames that have coinciding origins) are of the general form
x
µ
→x
µ
= Λ
µ
ν
x
ν
,
where Λ
µ
ν
∈ R and depend on the relative velocity and relative orientation of the axes. Using
previous relations, we see that
x
µ
→x

µ
= Λ
ν
µ
x
ν
,
where
Λ
ν
µ
= g
µσ
g
ντ
Λ
σ
τ
.
Now, Lorentz transformations (LTs) leave the interval s
2
= c
2
t
2
−x
2
invariant. That is,
s
2
= x
µ
x
µ
= x
µ
x

µ
.
Using these relations, it is simple to see that
Λ
µ
ν
Λ
σ
µ
x
ν
x
σ
= x
µ
x
µ
,
from which we read off that we require
Λ
µ
ν
Λ
σ
µ
= δ
σ
ν
.
The interval s
2
is an example of a Lorentz scalar – that is, something unchanged by a LT.
A four component object A
µ
which transforms like x
µ
under an LT, is, by definition, a
contravariant 4-vector. That is, if it transforms like
A
µ
→A
µ
= Λ
µ
ν
A
ν
.
1.3 Homogeneous Lorentz Transformations 5
Similarly, one can define a covariant 4-vector if it transforms as
A
µ
→A

µ
= Λ
ν
µ
A
ν
.
The scalar product of any two 4-vectors can be written in a variety of ways:
AB = A
µ
B
µ
= A
0
B
0
−A B,
and also
AB = A
µ
B
µ
= g
µν
A
µ
B
ν
= g
µν
A
µ
A
ν
= A
µ
B
µ
.
The scalar product is a Lorentz invariant, since one can easily show that
A
µ
B

µ
= A
µ
B
µ
.
For example, the energy-momentum in the 4-momentum
p
µ
= (E/c, p)
is a 4-vector. The scalar product with itself is just
p
2
= p
µ
p
µ
=
E
2
c
2
−p
2
= m
2
c
2
,
where m
2
c
2
is the thing we will call the Lorentz invariant. Hence,
p
2
= p
µ
p

µ
= m
2
c
2
is a Lorentz invariant.
All equations written consistently in terms of 4-vectors and scalars are automatically
Lorentz invariant. That is, those equations for whom indices balance automatically satisfies
the principles of special relativity.
So, what about derivatives? Consider a Lorentz scalar function, whereby its value at a
point in one coordinate system is the same at the same point in another coordinate system,
φ

(x

) = φ(x).
Then,
δφ =
∂φ
∂x
µ
δx
µ
.
So, the term on the LHS is a Lorentz scalar (as we defined it to be so). Also, the term on
the far RHS is a contravariant 4-vector. So, in order that the LHS is a scalar, we must have
that
∂φ
∂x
µ
is a covariant 4-vector. Hence,

µ
φ ≡
∂φ
∂x
µ
=
_
1
c
∂φ
∂t
, ∇φ
_
6 1 PRELIMINARIES
is a covariant vector. It thus transforms as

µ
−→∂

µ
= Λ
ν
µ

ν
.
Similarly, we define the contravariant differential 4-vector

µ
φ ≡
∂φ
∂x
µ
=
_
1
c
∂φ
∂t
, −∇φ
_
.
We further define the D’Alembertian,
≡ ∂
µ

µ
=
_
1
c

2
∂t
2
, −∇
2
_
,
which is a Lorentz scalar operator. We can obtain the Lorentz covariant form of the wave
equation, if φ is a Lorentz scalar;
φ = ∂
µ

µ
φ = 0 ⇒
1
c
2

2
φ
∂t
2
= ∇
2
φ.
1.4 Equation of Continuity
Suppose we have some charge density ρ(x, t), such that
Q

=
_

dV ρ(x, t) (1.4)
is unchanged under a Lorentz transformation (i.e. is a Lorentz scalar); where Ω is some
angular region of space, bounded by the surface S. This could be, for example, mass or
electric charge. Then, the charge Q is conserved, provided

µ
j
µ
(x) = 0 (1.5)
is satisfied; where
j
µ
(x) = (cρ, j).
To see this continuity equation, rewrite (1.5) using the definition of ∂
µ
,
∂ρ
∂t
+∇ j = 0. (1.6)
Then, by (1.4),
dQ
dt
=

∂t
_

dV ρ(x, t)
= −
_

dV ∇ j
= −
_
S
j dS.
The first step follows by (1.6) and the second by the divergence theorem. Now, note that as
(1.5) is a Lorentz invariant equation, then j
µ
(x) is a contravariant 4-vector.
1.4 Equation of Continuity 7
1.4.1 Electromagnetic Fields
We shall take space to be “free”, so that the relative permittivities and permeabilities are
ε
r
= µ
r
= 1. Now, Maxwell’s equations, in SI units, read
∇ E =
1
ε
0
ρ, ∇ B = 0,
∇E = −
∂B
∂t
, ∇B = µ
0
j + µ
0
ε
0
∂E
∂t
.
We note the standard relation
µ
0
ε
0
=
1
c
2
, µ
0
= 4π 10
−7
NA
−2
.
Now, the equations in SI units, are not used practically. Instead, we use the rationalised
Gaussian units,
ε
0
= 1 ⇒ µ
0
=
1
c
2
,
and we redefine the magnetic field,
B ≡
B
SI
c
.
Hence, under these rationalised SI units, the four Maxwell equations are
∇ E = ρ, ∇ B = 0,
∇E = −
1
c
∂B
∂t
, ∇B =
1
c
j +
1
c
∂E
∂t
.
We introduce the electromagnetic potentials φ and A, so that the fields are related to the
potentials via
E = −∇φ −
1
c
∂A
∂t
,
B = ∇A.
Notice that upon subsitution of these expressions into each of the Maxwell equations, they
are satisfied (for example, the curl of grad is zero). Note that upon substitution of these
equations, into Gauss’ and Amperes’ law (i.e the first and fourth of Maxwell’s equations),
one finds

2
φ +
1
c

∂t
∇ A = −ρ,

2
A−
1
c

2
A
∂t
2
−∇
_
∇ A+
1
c
∂φ
∂t
_
= −
1
c
j,
which may be written in the Lorentz covariant (and indeed, more succinct) form
A
µ
(x) −∂
µ

ν
A
ν
(x) =
1
c
j
µ
,
where the 4-vector potential A
µ
is
A
µ
(x) = (φ, A).
8 1 PRELIMINARIES
1.5 Quantum Mechanics
1.5.1 Interpretation of the Wavefunction
We tend to define that
ρ(x, t)d
3
x = [ψ(x, t)[
2
d
3
x
is the probability of finding a particle in volume d
3
x, at position x, and at time t. Now, this
requires that probabilities are positive;
ρ(x, t) = [ψ(x, t)[
2
≥ 0,
and that the wavefunction is normalised;
_
ρ(x, t)d
3
x =
_
[ψ(x, t)[
2
d
3
x = 1,
where integrals are taken over all space. Notice that this last statement means that proba-
bilities are conserved, as the derivative of unity is zero;

∂t
_
ρ(x, t)d
3
x =

∂t
_
[ψ(x, t)[
2
d
3
x = 0.
Now, we can get to the equation of continuity, with a little work.
Consider the Schrodinger equation,
i
∂ψ
∂t
= −

2
2m

2
ψ + V (x)ψ, (1.7)
and its Hermitian conjugate,
−i
∂ψ

∂t
= −

2
2m

2
ψ

+ V (x)ψ

. (1.8)
Now, take the Schrodinger equation (1.7), multiply it by ψ

, and subtract from it the conju-
gated equation (1.8) multiplied by ψ, to give
i
_
ψ

∂ψ
∂t
+ ψ
∂ψ

∂t
_
= −

2
2m
_
ψ


2
ψ −ψ∇
2
ψ

_
. (1.9)
The LHS of this is just
i
_
ψ

∂ψ
∂t
+ ψ
∂ψ

∂t
_
= i

∂t
[ψ[
2
= i

∂t
ψ

ψ.
And the RHS of (1.9) can be trivially written

2
2m
∇ (ψ

∇ψ −ψ∇ψ

) .
1.5 Quantum Mechanics 9
Hence, bringing this together, we have that

∂t
[ψ[
2
=
i
2m
∇ (ψ

∇ψ −ψ∇ψ

) .
Hence, this is of the form of a continuity equation,
∂ρ
∂t
+∇ j = 0, (1.10)
where
ρ(x, t) = [ψ(x, t)[
2
, j =
i
2m
(ψ∇ψ

−ψ

∇ψ) . (1.11)
This is the Born interpretation of quantum mechanics; that the modulus squared of the
wavefunction is a probability density. Hence, we see that the continuity equation is satisfied
in the Schrodinger equation, but we shall have to recheck this when we come to discuss
relativistic quantum mechanics.
1.5.2 Minimal Electromagnetic Interactions
The classical Hamiltonian, for a point particle of mass m, charge q, in an EM field, is given
by
H =
1
2m
_
p −
q
c
A
_
2
+ qφ. (1.12)
Let us compute Hamilton’s equations from this. That is, let us consider
˙ x
i
=
∂H
∂p
i
, ˙ p
i
= −
∂H
∂x
i
. (1.13)
So, the first gives
˙ x
i
=
1
m
_
p
i

q
c
A
i
_
,
which easily rearranges into
p
i
= m˙ x
i
+
q
c
A
i
,
or,
p = m˙ x +
q
c
A. (1.14)
The second of (1.13) provides us with
m¨ x
i
= m˙ x
i
+ qE
i
,
or generally with
m¨ x = qE +
q
c
v B.
10 1 PRELIMINARIES
Notice that if we insert (1.14) into (1.12), then we have
H(x, ˙ x) = E(x, ˙ x) =
1
2
m˙ x
2
+ qφ,
which is just the total energy of the system. Notice that the total energy is independent of the
magnetic vector potential A; this is a consequence of the magnetic field being perpendicular
to motion.
We call (1.14) the conjugate momentum; notice that it is the “usual” expression of mo-
mentum, with an additive term due to the magnetic vector potential.
Let us take the Schrodinger equation,
Hψ = i
∂ψ
∂t
,
and use the “correspondence principle” (also known as first quantisation, or canonical quan-
tisation) to turn the momentum vector p in (1.12) into the momentum operator ˆ p = −i∇,
so that the Schrodinger equation reads (using the Hamiltonian),
1
2m
_
−i∇−
q
c
A
_
2
ψ + qφψ = i
∂ψ
∂t
,
which rearranges fairly simply into

2
2m
_
∇−
iq
c
A
_
2
ψ = i
_

∂t
+
iq

φ
_
ψ.
Now, this is related back to the “free equation” (i.e. the Schrodinger equation with no EM
fields) by the substitution of space and time derivatives,
∇ −→∇−
iq
c
A,

∂t
−→

∂t
+
iq

φ.
Indeed, this may be written as

µ
−→∂
µ
+
iq

A
µ
, (1.15)
where
A
µ
= (φ, −A), ∂
µ
=
_
1
c

∂t
, ∇
_
.
This substitution, (1.15) is called minimal substitution.
1.6 Relativistic Case 11
1.6 Relativistic Case
By a similar argument as above, the relativistic Hamiltonian is
H =
_
(cp −qA)
2
+ m
2
c
4
+ qφ,
or, by the minimal substitution,
H =
¸

2
c
2
_
∇−
iq
c
A
_
2
+ m
2
c
4
+ qφ.
Now, the problem is: how do we interpret this square root? For example, the free particle
Schrodinger equation reads


2
c
2

2
+ m
2
c
4
ψ = i
∂ψ
∂t
. (1.16)
How should we interpret this? This is the main thrust of all we look at.
1.7 Natural Units
From now on, we use natural units, whereby
= 1, c = 1.
To get back to practical units, we use dimensions to restore ’s and c’s; and then use
= 6.582 10
−22
MeV s, c = 1.973 10
−13
MeV m.
12 1 PRELIMINARIES
13
2 The Klein-Gordon Equation
Let us work in free space, so that (φ, A) = 0; and in the natural units previously discussed.
So, the relativistic Schrodinger equation (1.16) reads
Hφ(x, t) = i
∂φ(x, t)
∂t
, (2.1)
where the Hamiltonian is
H =

−∇
2
+ m
2
.
Now, we can avoid this square-root, by noting that the Hamiltonian is independent of time
t. So, if we act H upon (2.1), we get
H
2
φ = Hi
∂φ
∂t
= i

∂t

= i

∂t
i
∂φ
∂t
= −

2
φ
∂t
2
.
Therefore,
H
2
φ =
_
−∇
2
+ m
2
_
φ = −

2
φ
∂t
2
.
Hence,
_

2
∂t
2
−∇
2
+ m
2
_
φ(x, t) = 0. (2.2)
This is the Klein-Gordon (KG) equation. Using the D’Alembertian notation, this is just
_
+ m
2
_
φ(x, t) = 0, ≡ ∂
µ

µ
.
This KG equation is mathematically correct, and is indeed Lorentz invariant (for a scalar φ).
But, what does it mean? What is the interpretation of φ?
2.1 Negative Energies
Substituting a plane wave solution
φ(x, t) = e
i(p·x−Et)
into the KG equation, we find the relation
E
2
= p
2
+ m
2
,
14 2 THE KLEIN-GORDON EQUATION
so that the allowed energies are
E = +
_
p
2
+ m
2
> m, E = −
_
p
2
+ m
2
< −m.
Hence, the spectrum (i.e. allowed energies of the system) is discontinuous. There is an energy
gap between mc
2
< E < mc
2
(i.e. for energies below the rest mass, and above the negative of
the rest mass). In classical physics, such a gap is a bit of a problem. In quantum mechanics,
interactions allow quantum jumps over this gap, releasing quanta E
γ
> mc
2
, as radiation or
other particles; such transitions carry on to give an infinite energy release.
2.2 Conserved Current
Let us compute the conserved current and charge j, ρ associated with the KG equation. Let
us write the KG equation (2.2) as

2
φ
∂t
2
=
_

2
−m
2
_
φ, (2.3)
also taking its complex conjugate,

2
φ

∂t
2
=
_

2
−m
2
_
φ

. (2.4)
Taking the first equation (2.3), multiplying it by φ

, and subtracting from it (2.4) multiplied
by φ, gives

∂t
_
φ

∂φ
∂t
−φ
∂φ

∂t
_
= ∇ (φ

∇φ −φ∇φ

) .
Multiplying through by i, and bringing the LHS over to the RHS,
i

∂t
_
φ

∂φ
∂t
−φ
∂φ

∂t
_
−i∇ (φ

∇φ −φ∇φ

) = 0.
Hence, we can compare this with the continuity equation (1.10) to identify the charge and
current;
ρ = i
_
φ

∂φ
∂t
−φ
∂φ

∂t
_
, (2.5)
j = −i [φ

∇φ −φ∇φ

] . (2.6)
Now, if we compare these with those found in the non-relativistic case, (1.11), we see that we
cannot interpret ρ as a probability density (as we did do in the non-relativistic case). In our
case, (2.5) can be negative as well as positive (thus negating its ability to be a probability).
Hence, we see that φ cannot (generally) be interpreted as a probability amplitude (as we did
for the non-relativistic ψ).
So, to summarise, we have found that the KG equation is a covariant wave equation; and
with it is a conserved charge j
µ
= (ρ, j) such that ∂
µ
j
µ
= 0 is satisfied. But, we have problems
with negative energies, and interpreting the “wavefunction” φ. For the moment, let us carry
on, regardless of these issues.
2.3 EM Fields and the “Hydrogen Atom” 15
2.3 EM Fields and the “Hydrogen Atom”
If the 4-vector potential is non-zero, A
µ
,= 0, then we make the minimal substitution

µ
−→∂
µ
+ iqA
µ
,
so that the KG equation becomes
_
(∂
µ
+ iqA
µ
) (∂
µ
+ iqA
µ
) + m
2
¸
φ = 0. (2.7)
Now, consider electrostatic fields, so that A
µ
= (Φ, 0) only (i.e. only a static term, and no
magnetic term); such that
A = 0, V (x) = qA
0
(x) = qΦ(x),
and V is not a function of time t. Then, substituting this into the KG equation (2.7), one
easily obtains
_
_

∂t
+ iV
_
2
−∇
2
+ m
2
_
φ(x, t) = 0. (2.8)
Now, if we substitute a solution of the form
φ(x, t) = ψ(x)e
−iEt
into equation (2.8), we easily find that
_
−(E −V )
2
+−∇
2
+ m
2
¸
ψ = 0. (2.9)
Now, let us solve for the energy levels, where the potential is the central Coulomb
V = −

r
, α =
e
2

=
1
137
. (2.10)
We can solve by comparing with the non-relativistic Schrodinger equation
_
−∇
2

2mZα
r
_
ψ = 2mψ, (2.11)
where = E −mc
2
. The way to solve this, is to make the separable ansatz
ψ =
U(r)
r
Y
m
(θ, φ), (2.12)
and use that the Laplacian in spherical polars can be written

2
=
1
r

2
∂r
2
r −
ˆ
L
2
r
2
. (2.13)
16 2 THE KLEIN-GORDON EQUATION
So, substituting these into (2.11) fairly easily results in
_

d
2
dr
2
+
( + 1)
r
2

2mZα
r
_
U(r) = 2mU(r). (2.14)
Making the same ansatz (2.12), after putting in the central Coulomb potential (2.10), we see
that we can write the KG equation (2.9) as
_
−∇
2

(Zα)
2
r
2

2EZα
r
_
U(r)
r
Y
m
(θ, φ) = (E
2
−m
2
)
U(r)
r
Y
m
(θ, φ),
using the form of the Laplacian (2.13) gives
_

d
2
dr
2
+
( + 1) −(Zα)
2
r
2

2EZα
r
_
U(r) = (E
2
−m
2
)U(r). (2.15)
Hence, upon comparison of this with the non-relativistic version (2.14), we see that we have
the relations
2m −→ 2E,
2m −→ E
2
−m
2
,
( + 1) −→

(

+ 1) = ( + 1) −(Zα)
2
. (2.16)
Notice that (2.16) may be solved to give

= −
1
2
+
¸
_
+
1
2
_
2
−(Zα)
2
. (2.17)
Now, the energy levels of the non-relativistic Schrodinger equation (2.14) are given by
2m = −
2m(Zα)
2
2n
2
m,
where the principle quantum number is given in terms of the number of radial modes and
angular quantum number, as
n = n
r
+ + 1, n
r
= 0, 1, 2 . . . .
Hence, using our relations between the solutions of the relativistic and non-relativistic equa-
tions, we see that the equivalent expression for the relativistic KG equation is
E
2
−m
2
= −
2E(Zα)
2
2n
2
E, (2.18)
where
n

= n
r
+

+ 1, n
r
= 0, 1, 2, . . . .
We can solve (2.18) for E,
E = m
_
1 +
(Zα)
2
n
2
_
−1/2
. (2.19)
We can consider two cases: the strong and weak potential, in order that we may expand this
energy.
2.3 EM Fields and the “Hydrogen Atom” 17
Weak potential If we take
Zα ¸1,
then (2.19) can be expanded in Zα
E = m
_
1 −
(Zα)
2
2n
2
+
3
8
(Zα)
4
n
4
+ . . .
_
.
Now, by (2.17),

= −
1
2
+
_
+
1
2
_

(Zα)
2
2
_
+
1
2
_ + . . .
So, to first order in Zα,

= , which means that to first order, n

= n. To second order,

= −
(Zα)
2
2
_
+
1
2
_,
so that
n

= n
r
+

+ 1
= n
r
+ −
(Zα)
2
2
_
+
1
2
_ + 1
= n −
(Zα)
2
2
_
+
1
2
_,
and hence that
n
4
= n
4
_
1 −
4(Zα)
2
2n
_
+
1
2
_ + . . .
_
.
One can then show that
E = m
_
1 −
(Zα)
2
2n
2

(Zα)
4
2n
4
_
n
+
1
2

3
4
_
+ . . .
_
.
The first term is the rest mass of the electron; the second the non-relativistic result. The
third term gives a splitting for modes with different , and is thus the fine-splitting term.
This fine-splitting term does not agree with experimental results for Hydrogen, and thus
cannot be used as a description of the Hydrogen atom. However, the derivation for the term
is correct (infact, the result is correct for pionic atoms, spin-zero particles). Hence, this scalar
KG equation described spin-0 particles.
Strong potential Let us take
Zα > +
1
2
,
so that for = 0, this corresponds to Z ≥ 69. Notice that in this regime, (2.17) is complex,
which will give complex energies! To understand this, we shall take a look at a simpler
problem.
18 2 THE KLEIN-GORDON EQUATION
2.3.1 The Klein Paradox
z = 0
V
T = E −m
Figure 2.1: A beam with kinetic energy T travels towards a barrier which has height V , centred at
z = 0, along the ˆ z-axis,
Consider a beam as in Figure (2.1). Then, for z < 0, the KG equation (2.9) is just
_
−E
2

d
2
dz
2
+ m
2
_
ψ(z) = 0,
which is easily solved to find
ψ(z) = Ae
ipz
+ Be
−ipz
, p
2
= E
2
−m
2
.
Notice that A is the amplitude for the incident wave, and B for some reflected wave. For
z > 0, the KG equation is just
_
−(E −V )
2

d
2
dz
2
+ m
2
_
ψ(z) = 0,
which is solved to give
ψ(z) = Ce
ip

z
, p
2
= (E −V )
2
−m
2
;
where we have imposed that there should be no imaginary wave from the right.
Now, if p
2
> 0, then
(E −V )
2
> m
2
⇒ (E −m)
2
> V
2
.
Similarly, if p
2
< 0, then
(E −m)
2
< V
2
,
where we can thus write p

= i[p

[, i.e. that p

is complex. Thus, if p

is complex, then the
argument of the exponential e
ip

z
is real (and negative), and thus we have exponential decay
“inside the barrier” (this is called “evanescence”); a somewhat expected result.
If we impose continuity of the wavefunction at z = 0, where ψ and dψ/dz are continuous,
then one finds the relation
B =
p −p

p + p

A, C =
2p
p + p

A.
2.3 EM Fields and the “Hydrogen Atom” 19
Now, in 1D, the conserved current is
j = −i
_
ψ


dz
−ψ


dz
_
.
So, the current for the incident wave Ae
ipz
is
j
I
= 2p[A[
2
,
for the reflected Be
−ipz
is
j
R
= −2p[B[
2
,
and for the transmitted wave Ce
ip

z
j
T
= 2p

[C[
2
.
Notice that
j
I
= j
R
+ j
T
,
somewhat as expected. Hence, we have propagation through the barrier if (E − V )
2
> m
2
and evanescence if (E − V )
2
< m
2
(these cases correspond to p

being real and imaginary,
respectively).
Let us then consider two cases, as before.
Weak potential Consider V < E. Then, propagation if E − V > m, or equivalently,
E − m > V . Then, this requires the kinetic energy to be above the barrier, which is an
expected result.
Strong potential Consider V > E. Then, propagation can occur if E − V < −m. That
is, V > E+m, which just gives (E−V )
2
< m
2
. This result says that propagation is allowed,
for large potential, if the kinetic energy is below the barrier. This is the Klein paradox.
Consider waves on the right (i.e. for z > 0). For V > E + m, then
p
2
= (V −E)
2
−m
2
,
which we can differentiate with respect to p

, to give
2p

= 2(V −E)
_

dE
dp

_
.
Hence, the group velocity,
v
g
=
dE
dp

= −
p

V −E
.
20 2 THE KLEIN-GORDON EQUATION
Thus, we see that the group velocity (which is the speed of the movement of particles) is
“left”, whereas the wave is traveling “right”. If we substitute p

into the expression for the
conserved current, then one finds
z < 0 ⇒ [j
R
[ > [j
I
[,
and
z > 0 ⇒ j
T
< 0.
That is, current is flowing to the left, even though waves are traveling to the right.
waves current
Figure 2.2: The relative amplitudes of the waves and current at the barrier.
The suggested interpretation of this, is that particle/anti-particle pairs are created at the
barrier, by an interaction with the barrier. Particles move to the left, in the reflected wave,
and anti-particles move to the right in the transmitted wave (with opposite charge). This
suggests a similar effect in atoms, if Z is large.
So, for a full description for strong potentials, we must abandon a single particle description,
and we will need a field theory – which we will come back to later.
21
3 The Dirac Equation
Let us reconsider the Schrodinger equation,
Hψ = i
∂ψ
∂t
, (3.1)
where the relativistic Hamiltonian is
H =

−∇
2
+ m
2
,
which was derived using canonical quantisation. Recall that we got the Klein-Gordon equa-
tion by multiplying the Schrodinger equation through by H,
HHψ = (−∇
2
+ m
2
)ψ = −

2
ψ
∂t
2
. (3.2)
This made the KG equation second order (in time), which led to negative conserved quanti-
ties. To get around this, Dirac looked for an equation that would be first order in time, of
the form (3.1). So, he proposed a Hamiltonian
H = −iα ∇+ βm, (3.3)
so that (3.1) becomes
(−iα ∇+ βm) ψ = i
∂ψ
∂t
.
This is known as the Dirac equation. The four constant coefficients α = (α
1
, α
2
, α
3
) & β
are chosen such that the Dirac equation is consistent with (3.2), so that E
2
= p
2
+ m
2
is
recovered. They are also chosen such that the Hamiltonian H is Hermitian, resulting in real
energy eigenvalues.
To be consistent with the KG equation, we thus require
HHψ =
_
−i
3

j=1
α
j

∂x
j
+ βm
__
−i
3

k=1
α
k

∂x
k
+ βm
_
ψ = −

2
ψ
∂t
2
.
So, expanding out,
HHψ = −

j,k
α
j
α
k

2
ψ
∂x
j
∂x
k
−im

j
α
j
β
∂ψ
∂x
j
−im

k
βα
k
∂ψ
∂x
k
+ β
2
m
2
ψ
= −

j,k
α
j
α
k

2
ψ
∂x
j
∂x
k
−im

j

j
β + βα
j
)
∂ψ
∂x
j
+ β
2
m
2
ψ
= −

j
α
2
j

2
ψ
∂x
2
j

j=k

j
α
k
+ α
k
α
j
)

2
ψ
∂x
j
∂x
k
−im

j

j
β + βα
j
)
∂ψ
∂x
j
+ β
2
m
2
ψ.
22 3 THE DIRAC EQUATION
Hence,
HHψ = −

j
α
2
j

2
ψ
∂x
2
j

j=k

j
α
k
+ α
k
α
j
)

2
ψ
∂x
j
∂x
k
−im

j

j
β + βα
j
)
∂ψ
∂x
j
+ β
2
m
2
ψ = −

2
ψ
∂t
2
.
Now, upon comparison of this with (3.2), we see that we require
α
2
j
= 1, α
j
α
k
+ α
k
α
j
= 0 (j ,= k), α
j
β + βα
j
= 0, β
2
= 1.
Thus, we have the anti-commutation relations
¦α
i
, α
j
¦ = 2δ
ij
, (3.4)
¦β, α
i
¦ = 0, (3.5)
β
2
= 1. (3.6)
Thus, we see that α
i
, β cannot be numbers, but they can be matrices.
3.0.2 Properties of α
i
, β & Dirac Representation
Now, one can easily see that α
i
and β must be Hermitian, in order that the Hamiltonian is
Hermitian.
Consider the eigenvalue equation
α
i
ψ = λψ,
and further
α
2
i
ψ = λ
2
ψ,
but, from (3.4), we see that we have α
2
i
= 1. Hence, λ
2
= 1, which means that the eigenvalues
of α
i
are λ = ±1. This result also holds for β, by (3.6).
We shall state, then prove, that
Tr β = 0, Tr α
i
= 0. (3.7)
That is, the sum over diagonal components is zero. Now, consider that
Tr (ABC) = Tr (CAB) = Tr (BCA),
which is easily provable by
Tr (ABC) =

i,j,k
A
ij
B
jk
C
ki
=

i,j,k
C
ki
A
ij
B
jk
= Tr (CAB). (3.8)
23
Then, consider that (3.5) is
βα
i
= −α
i
β,
multiplying from the left by β gives
β
2
α
i
= α
i
= −βα
i
β,
where the first equality comes from (3.6). Hence,
Tr α
i
= −Tr (βα
i
β) = −Tr (β
2
α
i
) = −Tr α
i
,
where the second equality follows from (3.8) and the third by (3.6). Hence,
2Tr (α
i
) = 0 ⇒ Tr α
i
= 0,
which is the second of (3.7). The proof for Tr β = 0 is analogous.
The final property of α
i
and β is that they must be of even dimension. To see this, consider
that the trace of a matrix is the sum of its eigenvalues. But, we found that the eigenvalues
of both α
i
and β are λ = ±1. We also showed that the trace of α
i
and β is zero. Hence, the
only way to add such eigenvalues to give zero, is to have an even number of them. Hence, α
i
and β are of even dimension.
Let us consider how to represent the four matrices α
i
and β.
In dimension N = 2, there are only 3 linearly independent matrices of the form
_
a
1
a
2
+ ia
3
a
2
−ia
3
−a
1
_
, a
i
∈ R.
These are
σ
1
=
_
0 1
1 0
_
, σ
2
=
_
0 −i
i 0
_
, σ
3
=
_
1 0
0 −1
_
; (3.9)
notice that they satisfy
¦σ
i
, σ
j
¦ = 2δ
ij
, (3.10)
as required. They also satisfy the commutation relation

j
, σ
k
] = 2i
jkl
σ
l
(3.11)
If N = 4, there are a few ways in which we can represent the matrices. The way we shall
is called the Dirac representation, where
β =
_
1 0
0 −1
_
, α
i
=
_
0 σ
i
σ
i
0
_
. (3.12)
24 3 THE DIRAC EQUATION
Just to make this block-notation clear, we shall write some of these out, long hand.
β =
_
_
_
_
1 0 0 0
0 1 0 0
0 0 −1 0
0 0 0 −1
_
_
_
_
, α
1
=
_
_
_
_
0 0 0 1
0 0 1 0
0 1 0 0
1 0 0 0
_
_
_
_
, α
2
=
_
_
_
_
0 0 0 −i
0 0 i 0
0 −i 0 0
i 0 0 0
_
_
_
_
.
As we previously said, there are other representations. Infact, a specific representation is
unneeded, and one can just work with the commutation relations. That we use the Dirac
representation is a mere convenience for the non-relativistic limits, and observable predictions
are not affected by the choice of representation.
Hence, we have that the Dirac equation
i
∂ψ
∂t
= −iα ∇ψ + βmψ (3.13)
is a matrix equation with four components. That is, the wavefunction has four components
ψ(x) =
_
_
_
_
ψ
1
(x)
ψ
2
(x)
ψ
3
(x)
ψ
4
(x)
_
_
_
_
.
Since α
i
, β are Hermitian, the adjoint equation of (3.13) is
−i
∂ψ

∂t
= i∇
_
ψ

α
_
+ mψ

β. (3.14)
Notice that one can fairly easily see the anti-commutation relations,
¦β, α
i
¦ = 0, ¦α
i
, α
j
¦ = 2δ
ij
;
the second of which can be seen just by (3.10).
3.1 Conserved Current
We can compute the conserved quantities by the usual argument. We multiply (3.13) by ψ

from the left (we must be careful about this now), to give


∂ψ
∂t
= −iψ

α ∇ψ + mψ

βψ. (3.15)
In a similar way, we multiply ψ from the right onto the adjoint equation (3.14), to give
−i
∂ψ

∂t
ψ = i∇
_
ψ

α
_
ψ + mψ

βψ. (3.16)
3.2 Angular Momentum & Spin 25
Then, subtracting (3.16) from (3.15), we see that
2i

∂t
_
ψ

ψ
_
= −2i∇
_
ψ

αψ
_
,
or, upon comparison with the continuity equation
∂ρ
∂t
+∇ j = 0,
then we see that
ρ = ψ

ψ, j = ψ

αψ. (3.17)
In particular, notice that
ρ = ψ

ψ = (ψ

1
ψ

2
ψ

3
ψ

4
)
_
_
_
_
ψ
1
ψ
2
ψ
3
ψ
4
_
_
_
_
=
4

i=1

i
(x)[
2
≥ 0.
Hence, we see that we can treat ψ as a probability density function for position, in the usual
way (which we could not do for the KG equation).
3.2 Angular Momentum & Spin
Consider the orbital angular momentum operator,
L = x p,
or, in index notation,
L
i
=
ijk
x
j
p
k
.
Let us ask: is orbital angular momentum conserved? If it is, then we require that
[H, L] = 0.
So, using the Dirac Hamiltonian (3.3),
H = −iα ∇+ βm
= α p + βm,
and using the usual commutation relation
[x
j
, p
k
] = iδ
jk
,
26 3 THE DIRAC EQUATION
we consider the commutator
[H, L
k
] = [α
j
p
j
,
knm
x
n
p
m
]
=
knm
α
j
[p
j
, x
n
]p
m
= −i
knm
α
j
δ
jn
p
m
= −i
kjm
α
j
p
m
,
where the last equality allows us to see that
[H, L] = −iαp, (3.18)
which is non-zero. If this were zero, then we could have said that the Hamiltonian commuted
with the orbital angular momentum operator. However, it does not, in which case we must
say that the Hamiltonian does not commute with orbital angular momentum, and hence
orbital angular momentum is not conserved.
Now, we only considered orbital angular momentum. Instead, let us suppose that the total
angular momentum is conserved (i.e. orbital plus something-else).
Let us consider some matrix operators
Σ
j
=
_
σ
j
0
0 σ
j
_
. (3.19)
Then,
α
j
Σ
k
−Σ
k
α
j
=
_
0 σ
j
σ
k
σ
j
σ
k
0
_

_
0 σ
k
σ
j
σ
k
σ
j
0
_
=
_
0 [σ
j
, σ
k
]

j
, σ
k
] 0
_
= 2i
jkn
_
0 σ
n
σ
n
0
_
= 2i
jkn
α
n
.
Hence, we have used (3.11) to show that

j
, Σ
k
] = 2i
jkn
α
n
. (3.20)
In a much simpler, but analogous manner, one can also show that
[β, Σ
k
] = 0. (3.21)
Hence, the commutator of the Hamiltonian with this Σ
i
operator,
[H, Σ
k
] = [α
j
p
j
+ βm, Σ
k
]
= [α
j
, Σ
k
]p
j
= 2i
jkn
α
n
p
j
= 2i
knj
α
n
p
j
,
3.3 Plane Wave States 27
thus,
[H, Σ] = 2iαp. (3.22)
Then, consider the combination
J = L +
1
2
Σ, (3.23)
so that
[H, J] = −iαp +
1
2
2iαp = 0
from (3.22) and (3.18). Therefore, we see that the operator J, defined by (3.23) is conserved,
as it commutes with the Hamiltonian. We can define some spin operator,
S =
1
2
Σ =
1
2
_
σ 0
0 σ
_
,
in Dirac representation. Notice further, since σ
2
i
= 1, then Σ
2
i
= 1, and hence the eigenvalues
of S
i
are ±
1
2
. Thus, we see that the particle has spin−
1
2
.
3.3 Plane Wave States
Let us consider solutions to the Dirac equation, where the solutions are of the form
ψ(x) = Ne
−ipx
u(p), (3.24)
where px = p
µ
x
µ
= Et −p x, N is some normalisation and u(p) some 4-component spinor,
whose form is to be determined. We also require that p
2
= E
2
− m
2
, to satisfy the KG
equation.
We use Dirac representation, whereby
β =
_
1 0
0 −1
_
, α =
_
0 σ
σ 0
_
, (3.25)
and we write
u =
_
ξ
η
_
, (3.26)
where both ξ and η are two component spinors (thus making u a 4-component spinor). We
will substitute into the Dirac equation,
i
∂ψ
∂t
= −iα ∇ψ + βmψ, (3.27)
28 3 THE DIRAC EQUATION
and use the useful, easily derivable

µ
e
−ipx
= −i (E, −p) e
−ipx
= −ip
µ
e
−ipx
.
Notice that the Dirac equation (3.27) can be written as
βmψ = i
_

∂t
+α ∇
_
ψ
= i (∂
0
+ α
j

j
) Ne
−ipx
u(p)
= i (−iE + iα
j
p
j
) Ne
−ipx
u(p),
indeed as,
βmu(p) = (E −α p) u(p).
Hence, using the matrices (3.25) and (3.26)
m
_
1 0
0 −1
__
ξ
η
_
= E
_
ξ
η
_
+
_
0 σ p
σ p 0
__
ξ
η
_
,
rearranging
E
_
ξ
η
_
=
_
m1 σ p
σ p −m1
__
ξ
η
_
.
So, expanding out this matrix, we get two coupled simultaneous equations,
Eξ = mξ +σ pη, (3.28)
Eη = σ pξ −mη. (3.29)
So, solving (3.29) for η, gives
η =
σ p
E + m
ξ, (3.30)
so that upon substitution into (3.28), one finds
Eξ = mξ +σ p
σ p
E + m
ξ,
which rearranges easily into
(E −m)(E + m)ξ = (σ p)
2
ξ. (3.31)
Now, notice that
(σ p)
2
= (σ
1
p
1
+ σ
2
p
2
+ σ
3
p
3
) (σ
1
p
1
+ σ
2
p
2
+ σ
3
p
3
)
= σ
2
1
p
2
1
+ σ
2
2
p
2
2
+ σ
2
3
p
2
3
+(σ
1
σ
2
+ σ
2
σ
1
)p
1
p
2
+(σ
1
σ
3
+ σ
3
σ
1
)p
1
p
3
+(σ
2
σ
3
+ σ
3
σ
2
)p
2
p
3
.
3.3 Plane Wave States 29
The final three terms are all zero, by the anti-commutation relation (3.10), and also σ
2
i
= 1.
Hence, we see that
(σ p)
2
= p
2
1
+ p
2
2
+ p
2
3
= p
2
. (3.32)
Thus, substituting this into (3.31) – also expanding out the brackets on the LHS – gives
(E
2
−m
2
)ξ = p
2
ξ. (3.33)
Now, recall the positive energy solution, whereby
E(p) = +
_
m
2
+p
2
,
we then see that there are two independent solutions to (3.33), which we label
ξ
s
=

E + mχ
s
, (3.34)
where the χ
s
are such that
χ
s=
1
2
=
_
1
0
_
, χ
s=−
1
2
=
_
0
1
_
.
In writing (3.34), we have somewhat inserted the factor

E + m by forethought. Later on
we shall be writing the normalisation of the spinors, where the inserted factor will make
expressions a lot simpler. Hence, using this solution in (3.30), we see that (3.26) is
u
s
(p) =
_
ξ
η
_
=
_
ξ
σ·p
E+m
ξ
_
=
_ √
E + mχ
s
σ·p
E+m

E + mχ
s
_
=

E + m
_
χ
s
σ·p
E+m
χ
s
_
.
Therefore, we have derived that the positive energy solution is
u
s
(p) =

E + m
_
χ
s
σ·p
E+m
χ
s
_
. (3.35)
Notice that
u

s
(p) =

E + m
_
χ
s
(σ p)

E + m
χ
s
_
,
moreover that
u

s
(p)u
s
(p) = (E + m)
_
χ
2
s
+
(σ p)
2
(E + m)
2
χ
2
s
_
,
30 3 THE DIRAC EQUATION
which becomes, using (3.32),
u

s
(p)u
s
(p) = (E + m)
_
1 +
p
2
(E + m)
2
_
=
(E + m)
2
+p
2
E + m
=
(E + m)
2
+ E
2
−m
2
E + m
= 2E. (3.36)
If we recall the solution we were seeking, (3.24)
ψ(x) = Ne
−ipx
u(p),
and if we require its normalisation,
_
d
3
x ψ

(x)ψ(x) = [N[
2
_
d
3
x u

s
(p)u
s
(p)
= [N[
2
2EV
= 1,
where we normalise in some box of volume V , then we see that
N =
1

2EV
.
Therefore, the full positive energy solution is
ψ
(+)
p,s
(x) =
u
s
(p)e
−ipx

2EV
, (3.37)
where u
s
(p) is given by (3.35). This solution describes a particle with momentum p, energy
E > m and spin S
z
= s = ±1/2.
The negative energy solution is written as
ψ
(−)
p,s
(x) = Ne
ipx
v
s
(p).
One finds that the equivalent of (3.28) and (3.29) are
−Eξ = mξ +σ pη,
−Eη = σ pξ −mη.
Then, one can easily rearrange to find that
ξ = −
σ p
E + m
η,
3.4 Dirac Hole Theory 31
which results in the equivalent of (3.33) being
p
2
η = (E
2
−m
2
)η.
Hence, with a negative energy, one still has the relation E
2
= p
2
+ m
2
. This equation has
similar solutions, and so one arrives at
v
s
(p) =

E + m
_
σ·p
E+m
χ
−s
χ
−s
_
, (3.38)
and thus the wavefunction
ψ
(−)
p,s
(x) =
v
s
(p)e
ipx

2EV
. (3.39)
This corresponds to a particle with momentum (−p), energy E = −
_
p
2
+ m
2
and spin
(−s), in volume V .
Notice that in the non-relativistic limit p → 0 (and the energy E is of the order the rest
mass m), (3.35) and (3.38) become
u
s
−→

2m
_
χ
s
0
_
,
v
s
−→

2m
_
0
χ
−s
_
.
3.4 Dirac Hole Theory
As spin−
1
2
particles are fermions, they obey the exclusion principle, which states that no two
fermions can be in the same state. Now, imagine the “vacuum” as being a completely filled
“sea” of negative energy states. That the sea is completely filled is important, as no positive
energy particle can drop in there, as there is no free state for it to occupy.
If the negative energy states are full, then no positive energy particle can transition into
the negative states. But, a negative energy particle can transition into the positive energy
states, if it is given enough energy E > 2m. Then, there will be a hole in the negative
energy states, which looks like an anti-particle. That is, the absence of a e

particle with
−E, −p, −s, which looks like a e
+
particle with E, p, s.
See Figure (3.1) for a depiction of the theory.
32 3 THE DIRAC EQUATION
full in vacuum
empty in vacuum
E
p
m
−m
Figure 3.1: A schematic of the Dirac hole picture. In vacuum, negative energy states – given
by E = −
_
p
2
+ m
2
are full, so that no positive energy particles can occupy the negative states.
However, a negative energy particle can be promoted to the positive states, and appears as an
anti-particle
3.5 Relativistic Covariance of The Dirac Equation
3.5.1 Covariant Notation
The Dirac equation written in the form
i
∂ψ
∂t
= (−iα ∇+ βm) ψ
is not obviously covariant. Now, if we multiply from the left by β, and recall that β
2
= 1,
then we see that

∂ψ
∂t
= (−iβα ∇+ m) ψ,
writing βα ∇ as
βα ∇ = βα
i

∂x
i
,
then we see that the Dirac equation looks like

∂ψ
∂t
=
_
−iβα
i

∂x
i
+ m
_
ψ. (3.40)
Now, if we make the definitions,
γ
0
≡ β, γ
i
≡ βα
i
, (3.41)
3.5 Relativistic Covariance of The Dirac Equation 33
then we see that (3.40) can be written

0
∂ψ
∂t
=
_
−iγ
i

∂x
i
+ m
_
ψ. (3.42)
Then, if we write
γ
µ
= (γ
0
, γ
i
) = (β, βα
i
),
we see that (3.42) can be written as
(iγ
µ

µ
−m) ψ = 0. (3.43)
The properties of the γ
µ
matrices can be deduced from the properties of β, α
i
(i.e. the
commutation relations). It is easy to see that ¦α
i
, β¦ = 0, so that
¦γ
0
, γ
i
¦ = ¦β, βα
i
¦
= β
2
α
i
+ βα
i
β
= α
i
+ β(−βα
i
)
= 0.
Similarly, one can see that ¦γ
0
, γ
0
¦ = 2. Also,
¦γ
i
, γ
j
¦ = ¦βα
i
, βα
j
¦
= βα
i
βα
j
+ βα
j
βα
i
= −β
2
α
i
α
j
−β
2
α
j
α
i
= −¦α
i
, α
j
¦.
Hence, one can deduce that
¦γ
µ
, γ
ν
¦ = γ
µ
γ
ν
+ γ
ν
γ
µ
= 2g
µν
. (3.44)
Also, one can see that
γ
µ†
= γ
0
γ
µ
γ
0
, (3.45)
which is equivalent to saying γ
0†
= γ
0
, γ
i†
= −γ
i
. Also, as β
2
= 1, then (γ
0
)
2
= 1.
It is also useful to introduce the Dirac adjoint,
¯
ψ(x) ≡ ψ

(x)γ
0
. (3.46)
If we recall the components of j
µ
, from (3.17),
j
µ
= (ψ

ψ, ψ

αψ),
then we see that using the Dirac adjoint, we can write j
µ
as
j
µ
(x) =
¯
ψ(x)γ
µ
ψ(x), (3.47)
and so the equation of continuity is ∂
µ
j
µ
= 0.
Now, this is not obvious, but the γ
µ
do not form the components of a 4-vector. Hence, it is
still not obvious as to whether (3.43) is covariant or not. If (3.43) is a valid physical equation
(which it is), the principle of special relativity says that it must be covariant. Hence, we
must look for transformation properties of ψ so that (3.43) is covariant.
34 3 THE DIRAC EQUATION
3.5.2 Proof of Covariance
Consider a Lorentz transformation from one frame to another, O →O

, such that
x −→x

= ax,
so that
x
µ
= a
µ
ν
x
ν
,
where a
µ
ν
is some transformation matrix that satisfies
a
µ
ν
a
σ
µ
= δ
σ
ν
. (3.48)
So, let us define some operator S(a) by
ψ(x) −→ψ

(x

) = S(a)ψ(x), (3.49)
and, correspondingly,
ψ(x) = S
−1
(a)ψ

(x

) = S(a
−1
)ψ(x

).
We shall state, and then prove, that the Dirac equation (3.43) and (3.47) are covariant,
provided,
S
−1
(a)γ
µ
S(a) = a
µ
ν
γ
ν
, (3.50)
γ
0
S

(a)γ
0
= S
−1
(a). (3.51)
Those familiar with the terminology of group theory will notice that (3.50) is just the state-
ment of equivalence. That is, we require a
µ
ν
γ
ν
and γ
µ
to be equivalent, with “equivalence
element” S(a). Physically, we are requiring a γ-matrix to still be a γ-matrix, even after
transformation (which is not too much to ask of a matrix). One will also notice that (3.51)
is just the statement that S
−1
(a) is equivalent to S

(a), with equivalence element γ
0
.
So, prove the covariance, we want to show
(iγ
µ

µ
−m) ψ(x) = 0

_

µ

µ
−m
_
ψ

(x

) = 0, (3.52)
where

µ
= a
ν
µ

ν
. (3.53)
Let us substitute (3.49) and (3.53) into (3.52), to give
_

µ
a
ν
µ

ν
−m
_
S(a)ψ(x) = 0,
multiplying from the left by S
−1
(a),
_
iS
−1
(a)γ
µ
S(a)a
ν
µ

ν
−m
_
ψ(x) = 0. (3.54)
3.5 Relativistic Covariance of The Dirac Equation 35
Now, if we multiply (3.50) by a
σ
µ
from the right, then we see that
S
−1
(a)γ
µ
S(a)a
σ
µ
= a
µ
ν
γ
ν
a
σ
µ
= γ
ν
δ
σ
ν
= γ
σ
,
after using (3.48) in the second equality. So, using this result in (3.54), gives
(iγ
µ

µ
−m) ψ(x) = 0,
as required. Thus, the Dirac equation is covariant.
We still need to show that (3.47) is covariant. That is, show that
j
µ
=
¯
ψ

γ
µ
ψ

= a
µ
ν
¯
ψγ
ν
ψ (3.55)
holds.
Let us consider the transformation law of
¯
ψ(x),
ψ

(x

) = S(a)ψ(x) ⇒ ψ

(x

) = ψ

(x)S

(a),
and multiplying by γ
0
from the right,
ψ

γ
0
=
¯
ψ

= ψ

S

γ
0
,
where the first equality follows by (3.46). Now, if we insert (γ
0
)
2
= 1, between ψ

and S

,
then
¯
ψ

= ψ

γ
0
γ
0
S

γ
0
=
¯
ψγ
0
S

γ
0
,
where the second equality again follows by the definition of the Dirac adjoint. If we now use
(3.51) for γ
0
S

γ
0
, then we see that
¯
ψ

=
¯
ψS
−1
(a). (3.56)
So, inserting this for the
¯
ψ

in the middle term of (3.55), gives
¯
ψ

γ
µ
ψ

=
¯
ψS
−1
γ
µ
ψ

,
using (3.49) for ψ

on the far RHS,
¯
ψS
−1
γ
µ
ψ

=
¯
ψS
−1
γ
µ
Sψ,
using (3.50) for S
−1
γ
µ
S, gives
¯
ψS
−1
γ
µ
Sψ = a
µ
ν
¯
ψγ
ν
ψ.
And therefore,
j
µ
= a
µ
ν
¯
ψγ
ν
ψ, (3.57)
as required.
Notice that if we multiply (3.56) by ψ

from the right,
¯
ψ

ψ

=
¯
ψS
−1
ψ

=
¯
ψS
−1
Sψ =
¯
ψψ.
That is,
¯
ψψ is a Lorentz scalar: its value is the same in all coordinate systems.
36 3 THE DIRAC EQUATION
3.5.3 Lorentz Boost
Consider a Lorentz boost along the x-axis. Then, we have a coordinate transformation
x

= ax,
where
a =
_
_
_
_
cosh ω −sinh ω 0 0
−sinh ω cosh ω 0 0
0 0 1 0
0 0 0 1
_
_
_
_
,
so that for a frame O

moving at speed v, relative to O, we have
sinh ω =
v

1 −v
2
= γv, cosh ω =
1

1 −v
2
= γ.
The form of the S(a) is
S(a) = e
ωα
1
/2
.
Now, consider that
sinh x = x +
x
3
3!
+
x
5
5!
+ . . . , cosh x = 1 +
x
2
2!
+
x
4
4!
+ . . . ,
and that
e
x
= 1 + x +
x
2
2!
+
x
3
3!
+ . . . = cosh x + sinh x.
Hence,
sinh
_
ωα
1
2
_
+ cosh
_
ωα
1
2
_
= 1 +
ωα
1
2
+
_
ωα
1
2
_
2
1
2!
+
_
ωα
1
2
_
3
1
3!
+
_
ωα
1
2
_
4
1
4!
+ . . .
= 1 +
ω
2
α
1
+
_
ω
2
_
2
1
2!
1 +
_
ω
2
_
3
α
1
1
3!
+
_
ω
2
_
4
1
4!
1 + . . .
= 1cosh
_
ω
2
_
+ α
1
sinh
_
ω
2
_
,
where we have made use of α
2n
1
= 1 and α
2n+1
1
= α
1
. Hence,
S(a) = e
ωα
1
/2
= 1cosh
_
ω
2
_
+ α
1
sinh
_
ω
2
_
, (3.58)
for a Lorentz boost.
3.5.4 Parity
Consider the transformation
t −→t

= t, x
i
−→x

i
= −x
i
,
3.5 Relativistic Covariance of The Dirac Equation 37
so that x
µ
= a
µ
ν
x
ν
, where
a =
_
_
_
_
1 0 0 0
0 −1 0 0
0 0 −1 0
0 0 0 −1
_
_
_
_
.
Notice that this matrix satisfies, as indeed it must,
a
µ
ν
a
σ
µ
= δ
σ
ν
.
Now, if we have that S(a) = P, then
ψ

(x

) = Pψ(x).
Now, for covariance, we require (3.50) and (3.51) to hold,
P
−1
γ
µ
P = a
µ
ν
γ
ν
,
P
−1
= γ
0
P

γ
0
.
Now, by the form of the matrix a, we see that P
−1
= P, so that P
2
= 1. All of these
conditions are satisfied by the choice
P = ηγ
0
, η = ±1.
Recall that in Dirac representation,
γ
0
=
_
1 0
0 −1
_
.
Particles at Rest Consider a particle at rest. Then, the positive and negative energy
solutions (3.37) and (3.39) become
ψ
(+)
=
_
χ
s
0
_
e
−imt
, ψ
(−)
=
_
0
χ
−s
_
e
imt
.
So, acting P upon both states, gives

(+)
= ηψ
(+)
, Pψ
(−)
= −ηψ
(−)
.
Therefore, we see that both the positive and negative energy solutions are eigenstates of
parity P, but particles and anti-particles have opposite “intrinsic” parity. This has been
experimentally verified, and is crucial for understanding positronium e
+
e

and mesons q¯ q.
We cannot measure η, and we conventionally set it η = 1, so that P = γ
0
.
38 3 THE DIRAC EQUATION
Particles in Motion We now consider the full wavefunctions,
ψ
(+)
p,s
(x) = N
_
χ
s
σ·p
E+m
χ
s
_
e
−i(Et−p·x)
, ψ
(−)
p,s
(x) = N
_
σ·p
E+m
χ
−s
χ
−s
_
e
i(Et−p·x)
.
Hence, using the parity transformation, P : x −→−x,

(+)
p,s
(x) = N
_
χ
s
σ·p
E+m
χ
s
_
e
−i(Et

+p·x

)
,
= ψ
(+)
−p,s
(x

)

(−)
p,s
(x) = −N
_
σ·p
E+m
χ
−s
χ
−s
_
e
i(Et

+p·x

)
= −ψ
(−)
−p,s
(x

).
Thus again, we see the intrinsic parity difference between particles and anti-particles.
3.6 Interaction With Fields
Consider interactions with an electromagnetic field, A
µ
(x) = (A
0
, A), which can be intro-
duced via the minimal substitution

µ
−→∂
µ
+ iqA
µ
,
where q is some charge. We could also introduce interactions with a scalar field, via
m −→m + S(x).
Hence, the Dirac equation is rewritten into
[iγ
µ
(∂
µ
+ iqA
µ
) −m−S] ψ(x) = 0,
or, indeed,
i
∂ψ
∂t
=
_
α (−i∇−qA) + β(m + S) + qA
0
¸
ψ. (3.59)
3.6.1 Magnetic Moment of Electrons
Let us consider interactions with EM fields only, so that S = 0, and look for energy eigen-
functions of the form
ψ(x) =
_
φ(x)
η(x)
_
e
−iEt
. (3.60)
3.6 Interaction With Fields 39
If we substitute (3.60) into (3.59), we see that
σ (−i∇−qA) η + (qA
0
+ m)φ = Eφ, (3.61)
σ (−i∇−qA) φ + (qA
0
−m)η = Eη; (3.62)
where the −m in the second expression comes from the −1 in the β matrix.
We shall consider the non-relativistic weak-field limit
= E −m ¸m, [qA
µ
[ ¸m.
We neglect the qA
0
term in (3.62), and take mη to the RHS, to give
σ (−i∇−qA) φ = (E + m)η,
but, E + m = + 2m, however, ¸m, so that E + m ≈ 2m. Therefore,
η =
1
2m
σ (−i∇−qA) φ. (3.63)
Now, consider (3.61); we can easily rearrange it into
σ (−i∇−qA) η + qA
0
φ = (E −m)φ,
and we use that = E − m (notice that we do not have that qA
0
¸ , so that we cannot
neglect the qA
0
term). So, inserting (3.63) for η,
1
2m
[σ (−i∇−qA)]
2
φ + qA
0
φ = φ. (3.64)
Our task now is to tidy this expression up, so that we can read off terms we can recognise.
We first use the easily provable
(σ a)(σ b) = a b + iσ (a b),
so that
[σ (−i∇−qA)]
2
φ = (−i∇−qA) (−i∇−qA) φ
+iσ (−i∇−qA) (−i∇−qA) φ. (3.65)
Consider the second term,
iσ (−i∇−qA) (−i∇−qA) φ = −qσ [∇(Aφ) +A(∇φ)] ,
after noting that curl grad is zero, and that a a = 0. Let us then use the vector identity
∇(Aφ) = φ∇A−A∇φ,
40 3 THE DIRAC EQUATION
so that the second term of (3.65) is just
−qσ (∇A) φ = −qσ Bφ,
where we have used that B = ∇A. We can also use that S =
1
2
σ, to see that this term is
just
−2qS Bφ. (3.66)
Let us now consider the first term of (3.65),
(−i∇−qA) (−i∇−qA) φ = −∇
2
φ + iq∇ (Aφ) + iqA ∇φ + q
2
A
2
φ. (3.67)
Now, notice that we have
∇ (Aφ) = ∇ Aφ +A ∇φ,
and if we use the Coulomb gauge,
∇ A = 0,
then
∇ (Aφ) = A ∇φ.
Hence, (3.67) becomes
_
−∇
2
+ 2iqA ∇+ q
2
A
2
_
φ. (3.68)
Thus far, our discussion has been very general, but, we shall now specify ourselves to a
homogeneous magnetic field; which one can show is given by
A =
1
2
Bx.
Notice then that
2iqA ∇ = iq (Bx) ∇ = iq (x ∇) B,
but,
L = x p = −ix ∇,
and hence that
2iqA∇ = −qL B.
Thus, using this in (3.68), we see that the first term of (3.65) is
_
−∇
2
−qL B+ q
2
A
2
_
φ.
Hence, using this and (3.66) in the first and second terms of (3.65), we see that (3.65) becomes
[σ (−i∇−qA)]
2
φ =
_
−∇
2
−2qS B−qL B+ q
2
A
2
_
φ.
3.7 Particles in a Spherical Potential 41
Hence, we can rewrite (3.64) with things that we recognise;
_

1
2m

2
+ qA
0
−µ B+
q
2
2m
A
2
_
φ = φ,
where
µ =
q
2m
(L + 2S) .
The factor of 2 infront of S is the so-called g-factor (gyromagnetic ratio) of the electron. This
prediction has been verified by experiment.
3.7 Particles in a Spherical Potential
Consider the time independent Dirac equation,
[−iα ∇+ β(m + S(r)) + V (r)] ψ = Eψ. (3.69)
Notice that S(r), V (r) depend on r = [x[ only. We have that
V (r) = qA
0
(r), A = 0,
so that we have an interaction with static fields only.
Let us consider solutions with j = 1/2, P = 1. Now, wavefunctions in Dirac representation
have the form
ψ(r) =
_
f(r)χ
s
g(r)iσ ˆrχ
s
_
, (3.70)
so that f(r) is large for positive energy solutions, and g(r) is large for negative energy
solutions, and both f, g are scalars. Let us then substitute (3.70) into (3.69),
_
m + S + V −iσ ∇
−iσ ∇ −m−S + V
__
f(r)χ
s
g(r)iσ ˆrχ
s
_
= E
_
f(r)χ
s
g(r)iσ ˆrχ
s
_
. (3.71)
We can use that
(σ ∇)(σ ˆr)g(r) =
2g(r)
r
+
dg
dr
.
and expand (3.71) to see that we get two coupled differential equations,
_
d
dr
+
2
r
_
g + (m + S + V −E) f = 0, (3.72)
df
dr
+ (m + S −V + E) g = 0. (3.73)
These coupled differential equations can then be solved for f, g. Notice that particles have f
large, and anti-particles have g being large.
42 3 THE DIRAC EQUATION
3.7.1 The MIT Bag Model
This is a very crude model of hadrons. For example, the uud quarks are confined to make
a proton p, with mass m
p
= 940MeV/c
2
and radius R
p
= 1fm. If we consider typical quark
momenta, taking λ = 2R and p = h/λ = 600MeV, we see that quarks inside a protons are
highly relativistic. We also know that quarks strongly interact at distances r ∼ 1fm, and
interact weakly at r ¸ 1fm (their so-called asymptotic freedom). So, the simple model we
will discuss – the MIT bag model – binds quarks in a scalar potential
S(r) =
_
0 r < R,
S
0
r > R;
where R ∼ 1fm. So, we have some scalar potential, which will allow quarks and anti-quarks
to attract (note, we want anti-quarks to attract as well, as we would like to be able to allow
anti-protons to exist).
We shall set m
q
= 0 (which is a good approximation for u, d-quarks). Notice that we are
not setting the constituent mass to zero, which is the mass of the bare quark and the virtual
gluons.
So, the radial equations (3.72), (3.73) become for the various limits:
Inside r < R
_
d
dr
+
2
r
_
g −Ef = 0,
df
dr
+ Eg = 0,
we can substitute the second,
g = −
1
E
df
dr
, (3.74)
into the first,
d
2
f
dr
2
+
2
r
df
dr
+ E
2
f = 0. (3.75)
We can solve this by setting
f(r) =
u(r)
r
,
so that upon substitution into (3.75), one can easily find
d
2
u
dr
2
= −E
2
u,
3.7 Particles in a Spherical Potential 43
which has solution
u(r) = A
1
sin Er + A
2
cos Er.
Now, under the requirement that f(r) is finite at r = 0, we must therefore set A
2
= 0, and
hence
f(r) =
Asin Er
r
. (3.76)
Then, by (3.74), we see that
g(r) = A
_
sin Er
Er
2

cos Er
r
_
. (3.77)
Outside r > R
_
d
dr
+
2
r
_
g + (S
0
−E) f = 0,
df
dr
+ (S
0
+ E) g = 0,
again, substituting the second into the first to give
d
2
f
dr
2
+
2
r
df
dr
+
_
E
2
−S
2
0
_
f = 0.
To solve this, we make a similar ansatz, and we find
f(r) =
Be
−κr
r
, κ ≡
_
S
2
0
−E
2
,
where we used the boundary condition that f should not diverge as r →∞, and
g(r) =
B
S
0
+ E
_
κe
−κr
r

e
−κr
r
2
_
.
If we assume that S
0
¸E (i.e. strong potential), then κ ≈ S
0
and
g(r) =
Be
−κr
r
≈ f(r).
We can now apply continuity at the boundary r = R, so that (3.76) and (3.77) are equal
from below,
f(R) = g(R) ⇒
sin ER
R
=
sin ER
ER
2

cos ER
R
,
or,
tan ER =
ER
1 −ER
.
44 3 THE DIRAC EQUATION
This can be graphically solved to give
ER = 2.04.
Therefore,
E
s
1
2
=
2.04
R
,
or, if we have 3 quarks,
3E =
6.12
R
.
Hence, the mass of the proton, in this crude model, is m
p
= 3E = 1200MeV/c
2
. This isnt too
far off, given that we have completely ignored things like the quarks interact with nothing
except the binding bag – so, ignoring the gluons around them.
3.7.2 Hydrogen-like Atoms
Now consider a system with
S(r) = 0, V (r) = −

r
Then, in a similar way to above, one can solve the radial equations, and find
E
n
= m
_
_
1 +
Z
2
α
2
n −
_
j +
1
2
_
+
_
_
j +
1
2
_
2
−Z
2
α
2
_
_
−1/2
. (3.78)
Notice that does not appear in this expression, only j (and n, obviously), and this is
because the Dirac equation “knows about” spin. Also notice that for Z
2
α
2
> j +
1
2
, the
energy becomes imaginary (for j = 1/2, this corresponds to atoms with Z > 137) – this is
another instance of the Klein paradox for strong potentials.
Consider a hydrogen atom, with Z = 1, then, expanding in powers of α, one finds

n
= E
n
−m = −

2
2n
2
_
1 +
α
2
n
_
1
j +
1
2

3
4n
_
+O(α
4
)
_
.
The first term gives the Rydberg energy, and the α
2
-term is the first relativistic correction
and is O(10
−5
).
With reference to Figure (3.2), we see the spectrum of the n = 2 level, and the splitting.
In the non-relativistic case, the spin-orbit interaction is added in manually, as is g
s
= 2.
Notice that the 2s
1/2
level splits into 2p
1/2
and 2p
3/2
, and those two levels are separated by
an energy gap ∆.
Notice that in the relativistic spectrum, the splitting gives a degeneracy between 2s
1/2
and 2p
1/2
; this is because in (3.78), there is no mention of , and all energy-information is
3.7 Particles in a Spherical Potential 45
non-relativistic

2p
3/2
2p
1/2
2s
1/2
relativistic

2p
3/2
2p
1/2
2s
1/2
Figure 3.2: The spectrum of splittings in the H-atom. In the non-relativistic case, the spin-orbit
coupling is put in “by hand”, but in the relativistic case, the Dirac equation “knows about” spin
inherently.
contained in j. This degeneracy is exact to all orders of expansion. The states have the
same splitting ∆ as in the non-relativistic case. In this relativistic case, the Dirac equation
already “knew about” spin, and orbital angular momentum, and one can predict the magnetic
moment g
s
= 2, so that one does not need to add them in “by hand”, as one did for the
non-relativistic case.
simple perturbative correction
e

e

e

e

Figure 3.3: A difference in the simple Dirac model and a more complicated – but needed – pertur-
bative correction.
There was fantastic agreement in the relativistic prediction of these splittings, until 1947,
when it was observed that the 2s
1/2
-level was shifted down by O(α
3
). This shift is called the
Lamb shift, and was promptly awarded a Nobel prize. The reason behind the failure is that
one needs to go further than the rather simple Dirac equation, and quantum fluctuations
come into play; for example, as in Figure (3.3). Also, one must think that problems are
present when one gets complex energies for strong potentials!
Also, the magnetic moment is not exactly g
s
= 2 in the more complicated model;
g = 2
_
1 +
α

+ . . .
_
,
which is again confirmed by experiment.
This highlights the need for a more complicated description of relativistic quantum sys-
tems.
46 3 THE DIRAC EQUATION
47
4 Quantum Fields
We now want to move on to a many-body theory, in which particle numbers can change.
An idea of how to do this, lies in that photons are quanta of the electromagnetic field, and
that the KG and Dirac equations are wave equations. But, how do we quantise waves? The
general idea is to expand a field in terms of modes of harmonic oscillators, where the modes
are operators. It is the technicalities of doing this, and some of the results of doing this,
which we shall consider in this section. The idea of assigning an operator to a field, or to a
quantity in general, is called secondary quantisation.
We use fields, where a field is some continuous dynamical variable which has a value
at each point in space. For example, the displacement field of a string, or density of a
compressible fluid, or the electromagnetic field. Fields can support waves, and can transmit
energy, momentum and forces; and have infinitely many degrees of freedom.
4.1 Quantum Mechanics of a String
Consider a string, length L, with tension T having transverse displacement φ(t, z). The string
will satisfy the wave equation
1
c
2

2
φ
∂t
2
=

2
φ
∂z
2
,
where the speed of sound is
c =
¸
T
µ
,
where µ is the mass per unit length of the string. The energy of the string is given by
E =
_
L
0
dz
_
1
2
µ
_
∂φ
∂t
_
2
+
1
2
T
_
∂φ
∂z
_
2
_
. (4.1)
For simplicity, we shall set T = µ = 1, so that c = 1 and so the wave equation and energy
become

2
φ
∂t
2
=

2
φ
∂z
2
, E =
_
L
0
dz
_
1
2
_
∂φ
∂t
_
2
+
1
2
_
∂φ
∂z
_
2
_
.
The first terms is a kinetic energy, and the second a potential energy density. Notice that
the wave equation is like a KG equation, with m = 0: massless.
Now, we want to impose travelling waves. We let our theory do so by imposing periodic
boundaries,
φ(t, z) = φ(t, z + L),
where we could let L → ∞, which we will do at the end of the calculation. Now, we can
expand the solution in terms of normal modes,
φ
n
= e
±i(ωnt−knz)
,
48 4 QUANTUM FIELDS
where
k
n
=
2πn
L
,
and n is an integer. Then, with this choice of k
n
, the periodic boundary condition is satisfied.
Upon substitution of the normal mode into the wave equation, we find the requirement,
ω
n
= [k
n
[,
a dispersion relation. Notice that at each position z along the string, the string performs
simple harmonic motion. We can expand an arbitrary φ in terms of the normal modes φ
n
,
φ(t, z) =

n=−∞
1


n
L
_
a
n
e
−i(ωnt−knz)
+ a

n
e
i(ωnt−knz)
¸
. (4.2)
The prefactor 1/


n
L is there purely for later convenience – if we did not put it in, we
would have found that equations become more complicated than if we did. The first term
in the brackets bears resemblance with the positive energy solutions, and the second term to
negative energy solutions.
We now want to work out the Hamiltonian – i.e. the energy – in terms of the coefficients
a
n
.
So, the kinetic energy, where we evaluate at t = 0,
1
2
_
L
0
dz
_
∂φ
∂t
_
2
=
1
2

n,m
_
L
0
dz
1
2L

ω
n
ω
m
__
−iω
n
a
n
e
iknz
+ iω
n
a

n
e
−iknz
¸

_
−iω
m
a
m
e
ikmz
+ iω
m
a

m
e
−ikmz
¸_
=

n,m
_
L
0
dz
1
4L

ω
n
ω
m
ω
n
ω
m
_
−a
n
a
m
e
i(kn+km)z
+ a
n
a

m
e
i(kn−km)z
+a

n
a
m
e
i(km−kn)z
−a

n
a

m
e
−i(kn+km)z
_
= −

n,m

ω
n
ω
m
4L
L¦a
n
a
m
δ
n,−m
−a
n
a

m
δ
nm
−a

n
a
m
δ
mn
+ a

n
a

m
δ
n,−m
¦ ,
where in the last equality, we made use of the orthogonality relation
_
L
0
e
i(kn−km)z
dz = δ
nm
L.
Hence, performing the sum over m, we find an expression for the kinetic energy,
1
2
_
L
0
dz
_
∂φ
∂t
_
2
= −

n
ω
n
4
_
a
n
a
−n
−2a

n
a
n
+ a

n
a

−n
_
.
4.1 Quantum Mechanics of a String 49
In a very similar fashion, we compute the potential energy,
1
2
_
L
0
dz
_
∂φ
∂z
_
2
=

n
ω
n
4
_
a
n
a
−n
+ 2a

n
a
n
+ a

n
a

−n
_
,
where we have used [k
−n
[ = k
n
and ω
−n
= ω
n
. Notice that we have not been careful with
allowing the coefficients a
n
to commute – this is because they are purely classical in nature,
and therefore commute. We then add the potential to the kinetic energy to find the total
energy,
H =

n
ω
n
a

n
a
n
. (4.3)
To make this look more familiar, we could introduce the real variables,
q
n

a
n
+ a

n


n
, p
n
≡ −i
_
ω
2
(a
n
−a

n
) ,
so that upon rearrangement and insertion into the total energy (4.3), we find
H =

n
1
2
_
p
2
n
+ ω
2
n
q
2
n
_
.
We can then compare this with the usual expression for the energy of a simple harmonic
oscillator,
H =
p
2
2m
+
1
2
ω
2
x
2
,
to see that (4.3) is actually equivalent to an infinite discrete sum over modes of the harmonic
oscillator.
So far, our analysis has been completely classical in nature. To quantise the theory, we
promote q
n
, p
n
to Hermitian operators, and a
n
, a

n
to the a
n
, a

n
operators. We also impose
the canonical commutation relations
[p
n
, q
m
] = −iδ
nm
, [p
n
, p
m
] = [q
n
, q
m
] = 0; (4.4)
and hence,
_
a
n
, a

m
¸
= δ
nm
, [a
n
, a
m
] = [a

n
, a

m
] = 0. (4.5)
To find the Hamiltonian of this theory (i.e. replacement of quantities with non-commuting
operators), we repeat the procedure that resulted in (4.3), but being careful with ordering;
and we find
H =

n
ω
n
2
_
a

n
a
n
+ a
n
a

n
_
, (4.6)
50 4 QUANTUM FIELDS
or, using the commutator [a
n
, a

n
] = 1,
H =

n
ω
n
_
a

n
a
n
+
1
2
_
. (4.7)
The identity
[AB, C] = A[B, C] + [A, C]B (4.8)
is very useful in computing
[H, a

n
] = ω
n
a

n
, [H, a
n
] = −ω
n
a
n
. (4.9)
4.1.1 Phonons
Let us consider the state [ψ), and suppose that the Hamiltonian acting upon the state gives
the energy,
H [ψ) = E [ψ) .
Now, consider
H
_
a

n
[ψ)
_
= Ha

n
[ψ)
=
_
ω
n
a

n
+ a

n
H
_
[ψ)
= (E + ω
n
) a

n
[ψ) .
The second equality follows after using the first of the commutation relations (4.9). Similarly,
one can compute that
Ha
n
[ψ) = (E −ω
n
) a
n
[ψ) .
Therefore, we see that a

n
is an eigenstate of the Hamiltonian, with eigenvalue E + ω
n
, and
that a
n
is an eigenstate with eigenvalue E − ω
n
. Therefore, we say that a

n
increases the
energy of a system by ω
n
, and a
n
decreases the energy by ω
n
.
We define the ground state of a system to be the state with lowest possible energy, denoted
by [0). Hence, as it is the state with lowest energy, any attempt to lower the energy with a
n
results in zero;
a
n
[0) = 0, ∀n. (4.10)
We can then construct excited states from the ground state, by applying a

n
to the ground
state as many times as we wish,
[¦N
n
¦) =

n
_
a

n
_
Nn

N
n
!
[0) , (4.11)
4.1 Quantum Mechanics of a String 51
so that we will have N
n
quanta in the mode n. The energy eigenvalues are then given by
E (¦N
n
¦) =

n=−∞
_
N
n
+
1
2
_
ω
n
.
Notice that in this expression for the energy, we have an infinity; the zero-point energy,
H [0) = E
ZP
=
1
2

n=−∞
ω
n
,
this sum is over an infinite range, and hence gives an infinite value. This infinity can be
“taken care of”, by redefining the energy,
E −→E −E
ZP
,
or, equivalently, redefining the Hamiltonian,
H −→H =

n
ω
n
a

n
a
n
.
In practice, this infinite energy is not a problem: we measure energies relative to some value
– we measure energy differences, as opposed to absolute values. Hence, in this definition, the
energy is just the sum over the occupation number of each mode,
E =

n
N
n
ω
n
.
This has the interpretation of looking like a state with N
n
particles (or phonons) with energy
ω
n
. We say that a

n
creates a particle, and a
n
destroys a particle. Just as we say that photons
are the quanta of electromagnetic waves, we say that phonons are the quanta of elastic waves.
So, what about momentum, as waves carry momentum. We will make a guess – and confirm
later – that the momentum operator in the z-direction is
P
z
=

n
k
n
a

n
a
n
. (4.12)
Recall that k
n
= 2π/λ
n
, so that we recover the de Broglie relation P = k, with = 1. By
analogy with (4.9), we can easily see that
[P
z
, a

n
] = k
n
a

n
, [P
z
, a
n
] = −k
n
a
n
. (4.13)
Thus, we see that a

n
increases, and a
n
decreases, momenta by units of k
n
, and energy by
units of ω
n
. That is, the action of a

n
upon a system, is to create a massless particle with
ω
n
= k
n
. The particle is associated with the wavefunction e
−i(ωnt−knz)
, as usual, but, this is
not the probability amplitude. The states [ψ) are the probability amplitudes.
52 4 QUANTUM FIELDS
Finally, the states are symmetric under interchange of particles; since, for example,
[k
n
, k
m
) = a

n
a

m
[0) = a

m
a

n
[0) = [k
m
, k
n
) ,
where the second equality follows from the commutator [a

n
, a

m
] = 0, (4.5). We can also place
any number of particles in a given state, so that N
n
= 0, 1, 2, . . . is allowed, with identical
properties k
n
, ω
n
.
Hence, the particles described by this formalism (infact, it all stems from the usage of
commutators, rather than anti-commutators) are bosons.
4.1.2 Field Operators
Recall (4.2),
φ(t, z) =

n=−∞
1


n
L
_
a
n
e
−i(ωnt−knz)
+ a

n
e
i(ωnt−knz)
¸
,
where we have now promoted the coefficients to operators. Hence, the field φ is a field
operator. Infact, the operator is time dependent, but the states [¦N
n
¦) are independent of
time. Therefore, we are in the Heisenberg picture of quantum mechanics – operators evolve,
but states do not. Following from the commutators (4.9), we thus have
[H, φ] =

n
_
ω
n
2L
_
−a
n
e
−i(ωnt−knz)
+ a

n
e
i(ωnt−knz)
¸
,
and therefore, by inspection,
[H, φ] = −i
∂φ
∂t
, (4.14)
which is the Heisenberg equation of motion. Hence, we have H in terms of φ. So, we have
the (verifiable)
P
z
=

n
k
n
a

n
a
n
= −
_
L
0
dz
∂φ
∂t
∂φ
∂z
,
and, using the commutators (4.13), we can easily verify that
[P
z
, φ] = i
∂φ
∂z
. (4.15)
Now, consider some wave equation,

2
φ
∂t
2
=

2
φ
∂z
2
,
where φ is now a Lorentz scalar, and using a unit speed of light. That is, this is just the
invariant Klein-Gordon equation, for a massless particle,
φ = ∂
µ

µ
φ = 0,
4.2 The Klein-Gordon Field 53
in (1 + 1)-dimensions; where x
µ
= (t, z) and x
µ
= (t, −z). So, let us say that H = E
must then be a component of the energy-momentum vector P
µ
= (H, P
z
), and the invariant
generalisation of the Heisenberg equation of motion is
i
∂φ
∂x
µ
= [φ, P
µ
]; (4.16)
that is, it has two components,
i
∂φ
∂t
= [φ, H], i
∂φ
∂z
= −[φ, P
z
],
which is in accordance with (4.14) and (4.15), after noting that [φ, H] = −[H, φ].
4.2 The Klein-Gordon Field
Let us consider quantising fields that satisfy the Klein-Gordon equation. We shall work in a
box, with volume V = L
3
, under periodic boundaries. Hence, the normal modes are discrete
(as a finite sized box), and are of the form
e
±ipx
, px = p
µ
x
µ
= Et −p x.
The momentum and energy are given by
p =

L
(n
x
, n
y
, n
z
) , E
p
≡ E(p) =
¸
¸
¸
_
p
2
+ m
2
¸
¸
¸ .
We shall consider complex scalar fields, so that the field operator is
φ(x) =

p
1
_
2V E
p
_
a(p)e
−ipx
+ c

(p)e
ipx
_
(4.17)
We have that a(p) ,= c(p), as we do not want φ = φ

(in general, at least). We have that the
coefficients a, a

, c, c

are operators, and we assume the usual commutation relations
_
a(p), a

(p

)
¸
= δ
pp
=
_
c(p), c

(p

)
¸
. (4.18)
This holds on a component-by-component basis, so that δ
pp
= δ
p
1
p

1
δ
p
2
p

2
δ
p
3
p

3
. All other
commutators vanish.
So, what about the Hamiltonian? The essential thing is that we require the Heisenberg
equation of motion,
i
∂φ
∂t
= [φ, H] ,
to hold. Consider the Hamiltonian for a 1D classical massless real string, (4.1),
H =
_
L
0
dz
_
1
2
_
∂φ
∂t
_
2
+
1
2
_
∂φ
∂z
_
2
_
.
54 4 QUANTUM FIELDS
Notice that we have appropriately chosen units to make the propagation speed unity. The
generalisation of this, to a massive complex 3D field, is simple,
H =
_
d
3
x
_
∂φ

∂t
∂φ
∂t
+ (∇φ)

(∇φ) + m
2
φ

φ
_
. (4.19)
Notice where the “mass term” comes in (i.e. the prefactor of φ

φ); also notice that there is
no factor of one-half (this is because we treat φ and φ

independently). We then expand this
Hamiltonian by inserting our field operator (4.17).
We will make use of the orthogonality relations
_
d
3
xe
−ipx
e
ip

x
= V δ
pp
,
_
d
3
xe
ipx
e
ip

x
= V δ
p,−p
.
So, we have
∂φ

∂t
∂φ
∂t
=

p,p

E
p
E
p

2V
_
E
p
E
p

_
a

(p)a(p

)e
ipx
e
−ip

x
−a

(p)c

(p

)e
ipx
e
ip

x
−c(p)a(p

)e
−ipx
e
−ip

x
+c(p)c

(p

)e
−ipx
e
ip

x
_
,
_
∇φ

_
(∇φ) =

p,p

p
i
p

i
2V
_
E
p
E
p

_
a

(p)a(p

)e
ipx
e
−ip

x
−a

(p)c

(p

)e
ipx
e
ip

x
−c(p)a(p

)e
−ipx
e
−ip

x
+c(p)c

(p

)e
−ipx
e
ip

x
_
;
and hence, using the orthogonality condition,
_
d
3
x
∂φ

∂t
∂φ
∂t
=

p
_
E
p
2
_
a

(p)a(p) + c(p)c

(p)
¸

E
p
E
−p
2
_
E
p
E
−p
_
a

(p)c

(−p) + c(p)a(−p)
¸
_
,
(4.20)
_
d
3
x
_
∇φ

_
(∇φ) =
1
2

p
_
p
2
E
p
_
a

(p)a(p) + c(p)c

(p)
¸
+
p
2
_
E
p
E
−p
_
a

(p)c

(−p) + c(p)a(−p)
¸
_
.
(4.21)
4.2 The Klein-Gordon Field 55
We can also compute that
_
d
3
x m
2
φ

φ =
1
2

p
m
2
_
1
E
p
_
a

(p)a(p) + c(p)c

(p)
¸
+
1
_
E
p
E
−p
_
a

(p)c

(−p) + c(p)a(−p)
¸
_
. (4.22)
Now, notice that
E
p
=
¸
¸
¸
_
p
2
+ m
2
¸
¸
¸ ⇒ E
p
= E
−p
,
which can be used to see that upon adding (4.21) and (4.22), we have a factor of E
p
which
cancels from the second term of (4.20). And hence, using these last two results, we can write
that the Hamiltonian is
H =
_
d
3
x
_
∂φ

∂t
∂φ
∂t
+ (∇φ)

(∇φ) + m
2
φ

φ
_
=

p
E
p
_
a

a + c

c + 1
_
,
where in the last step we used the commutator (4.18). Hence, the Hamiltonian is
H =

p
E(p)
_
a

(p)a(p) + c

(p)c(p) + 1
_
, (4.23)
Thus, this is like two infinite sets of simple harmonic oscillators – this corresponds to the two
degrees of freedom in a complex field. We shall ignore the zero-point energy in what follows.
Thus, we can then easily compute the following commutation relations, using (4.18),
_
H, a

(p)
¸
= E(p)a

(p), [H, a(p)] = −E(p)a(p),
_
H, c

(p)
¸
= E(p)c

(p), [H, c(p)] = −E(p)c(p); (4.24)
which are similar to (4.9). Hence, we can use these to show that
i
∂φ
∂t
= [φ, H]
holds. That is, our formalism of a quantised complex scalar field, with the assumed commu-
tators, is consistent with the KG equation of motion. By covariance of the theory, we have
that the 4-momentum is P
µ
= (H, P), where
P =

p
p
_
a

a + c

c
_
;
and we have a set of commutators exactly like (4.24), except H is replaced by P and E
p
by
p, for example
_
P, a

(p)
¸
= pa

(p).
56 4 QUANTUM FIELDS
So, what we see, is that a

(p) and c

(p) both create spin-0 particles, with 4-momentum
(E(p), p) and mass m; and correspondingly, a(p) and c(p) both destroy spin-0 particles,
with 4-momentum (E(p), p) and mass m. Then, we may ask, what is the difference between
a and c. To answer this, consider the conserved 4-current for the KG equation,
j
µ
(x) = iq
_
φ

(∂
µ
φ) −
_

µ
φ

_
φ
_
,
where we have merely inserted a constant q. Now, corresponding to this, is a conserved
charge,
Q =
_
d
3
x j
0
(x) (4.25)
= iq
_
d
3
x
_
φ

∂φ
∂t

∂φ

∂t
φ
_
. (4.26)
We can then insert the fields, and expand just as we did for the Hamiltonian (and we still
ignore the vacuum energy part); and we find
Q = q

p
_
a

(p)a(p) −c

(p)c(p)
_
. (4.27)
Hence, we see that the first term is the number operator for particles with charge q, and
the second term is the number operator for particles with charge −q. We can now use this
conserved charge to construct a set of commutators (which we derive either directly or by
analogy – they are effectively trivial to see, given (4.24));
_
Q, a

(p)
¸
= qa

(p), [Q, a(p)] = −qa(p),
_
Q, c

(p)
¸
= −qc

(p), [Q, c(p)] = qc(p). (4.28)
Therefore, we can see that a

and c increase the charge by q, and a, c

decrease the charge
by q.
Hence, putting this all together, we say that a

creates particles with charge Q = +q, c

creates anti-particles with charge Q = −q; and a destroys particles with charge Q = +q, c
destroys anti-particles with charge Q = −q. It is interesting to notice that if the field φ were
Hermitian, so that we only had a(p)-type coefficients (i.e no c(p)-type), then the charge Q
is zero. Hence, to have charge, we require non-Hermitian fields; with Hermitian fields being
uncharged.
4.3 The Dirac Field 57
In deriving the charge operator Q, one computes the quantities
φ

∂φ
∂t
= −i

p,p

E
p

2V
_
E
p
E
p

_
a

(p)a(p

)e
ipx
e
−ip

x
−a

(p)c

(p

)e
ipx
e
ip

x
+c(p)a(p

)e
−ipx
e
−ip

x
−c(p)c

(p

)e
−ipx
e
ip

x
_
, (4.29)
∂φ

∂t
φ = i

p,p

E
p

2V
_
E
p
E
p

_
a

(p)a(p

)e
ipx
e
−ip

x
+a

(p)c

(p

)e
ipx
e
ip

x
−c(p)a(p

)e
−ipx
e
−ip

x
−c(p)c

(p

)e
−ipx
e
ip

x
_
. (4.30)
Then, integrating the first and fourth terms of both expressions (where the second expression
is subtracted from the first), one finds

i
2

p
_
a

(p)a(p) + a

(p)a(p) −c(p)c

(p) −c(p)c

(p)
_
,
which one can simplify using commutation relations, to
−i

p
_
a

(p)a(p) −c

(p)c(p) −1
_
.
Then, the final terms which need integrating, are the second and third terms of both (4.29)
and (4.30); these infact cancel due to the symmetry of the summation. Hence, the charge Q
is found by multiplying the above by iq.
In a very similar fashion to the above, one can compute the equal-time commutator,
_
∂φ
∂t
(t, x), φ(t, x

)
_
= −iδ
(3)
(x −x

),
which is essentially the statement that one cannot know the position and velocity of a field
at the same time, unless they are at the same place.
It is worth noting, to be inkeeping with the literature, that the format of an operator, when
we get the modes to have destruction on the right and creation on the left, is called “time
ordered”.
4.3 The Dirac Field
In our previous discussion on the Klein-Gordon field, we looked at fields which permit particles
(and anti-particles) which are spinless. Naturally, we also want fields that allow particles
58 4 QUANTUM FIELDS
which have spin. Hence, the obvious choice is some Dirac field (given that the Dirac equation
permitted particles with spin).
4.3.1 Fermions
Suppose we have some operators b

n
and b
n
which create and destroy spin-
1
2
fermions, in the
state n = p, s (i.e. a given value of momentum and spin). That is,
b
n
[0) = 0, ∀n,
b

m
[0) = [n) ,
b

n
b

n

[0) = [n, n

) ; (4.31)
where [n) is a single-particle state, in n = p, s and [n, n

) is a two-particle state.
Now, by the Pauli exclusion principle (which states that no two fermions can be in the
same quantum state; and that fermions are anti-symmetric), we require
[n, n

) = −[n

, n) = −b

n

b

n
[0) .
Now, this and (4.31) mean that the b

n
operators are inconsistent with the usual commutation
relation [b

n
, b

n

] = 0, but are consistent with the anticommutation relation,
_
b

n
, b

n

_
= b

n
b

n

+ b

n

b

n
= 0. (4.32)
We also require that
[n, n) = b

n
b

n
[0) = 0,
so that two fermions cannot be created in the same state.
Therefore, this suggests that we should quantise fermionic fields using anticommutation
relations,
_
b
n
, b

m
_
= δ
nm
, ¦b
n
, b
m
¦ =
_
b

n
, b

m
_
= 0, (4.33)
for the creation and destruction operators. We use these anticommutation relations instead
of the commutation relations we used for bosonic fields. So, we must ask: does this work?
To answer this, we shall need to quantise a theory, using anticommutation relations, and see
if the answers are consistent.
4.3.2 The Complex Dirac Field ψ(x)
Recall that the Dirac equation reads
i
∂ψ
∂t
= (−iα ∇+ βm) ψ,
4.3 The Dirac Field 59
and has plane wave solutions,
u
s
(p)e
−ipx
_
2E
p
V
,
v
s
(p)e
ipx
_
2E
p
V
,
with the spinors being orthogonal;
u

s
(p)u
s
(p) = 2E
p
δ
ss
= v

s
(p)v
s
(p),
which follows from (3.36). Hence, the general solution of the Dirac equation can be written
as a sum over modes;
ψ(x) =

s,p
1
_
2E
p
V
_
b
s
(p)u
s
(p)e
−ipx
+ d

s
(p)v
s
(p)e
ipx
_
. (4.34)
The first term (will have) the interpretation of being a positive energy – or, alternatively,
frequency – solution, with destruction operator b
s
, and the second term being a negative
frequency solution, with creation operator d

s
. To quantise this field operator properly, we
must specify a set of commutation relations. Specifically, in this case, as we want to describe
fermions, we specify the anticommutation relations (4.33), so that
_
b
s
(p), b

s

(p

)
_
= δ
ss
δ
pp
=
_
d
s
(p), d

s

(p

)
_
, (4.35)
with all other anticommutators vanishing.
Now we have a field operator, and a set of anticommutation relations, we want to check
if the Hamiltonian (after being expanded in terms of the field operators modes) satisfies the
Heisenberg equation of motion,
i
∂ψ
∂t
= [ψ, H]. (4.36)
The Hamiltonian for the Dirac equation was
H = −iα ∇+ βm.
Hence, a good “guess”, is to integrate this over all space. Therefore, we suppose that the
Hamiltonian in the quantum Dirac field theory is
H =
_
d
3
x ψ

(−iα ∇+ βm) ψ. (4.37)
Notice that this Hamiltonian is just the expectation value of the thing we “used to” call
the Hamiltonian (i.e. is the energy). So, upon substitution of the field (4.34) into this
Hamiltonian, one finds that
H =

s,p
E(p)
_
b

s
(p)b
s
(p) + d

s
(p)d
s
(p)
_
, (4.38)
60 4 QUANTUM FIELDS
after ignoring the vacuum energy. The easiest way to show that this is the Hamiltonian, is
to recall that the Dirac equation is
i
∂ψ
∂t
= (−iα ∇+ βm) ψ,
so that upon contraction with ψ

from the left,


∂ψ
∂t
= ψ

(−iα ∇+ βm) ψ,
and integrating,
i
_
d
3


∂ψ
∂t
=
_
d
3


(−iα ∇+ βm) ψ,
we have on the RHS the Hamiltonian (4.37). Thus, in order that the RHS is true, the LHS
must also be true, and it is easier to compute the LHS.
We can use the easily provable identity
[AB, C] = A¦B, C¦ −¦A, C¦B
to find the commutators of the Hamiltonian with the creation and destruction operators (to
infact verify their interpretation). Then, one finds
_
H, b

s
(p)
¸
= E(p)b

s
(p), [H, b
s
(p)] = −E(p)b
s
(p),
_
H, d

s
(p)
¸
= E(p)d

s
(p), [H, d
s
(p)] = −E(p)d
s
(p). (4.39)
Hence, these can be used to show that the Hamiltonian (4.37) with the field quantised as
(4.34), satisfies the Heisenberg equation of motion (4.36). Therefore, we now see that we
have the interpretation of b

s
(p), d

s
(p) creating particles of energy E and momentum p; and
b
s
(p), d
s
(p) destroying particles of energy E and momentum p.
Now, recall that the conserved charge for the Dirac equation was
ρ = ψ

(x)ψ(x).
Hence, the conserved charge operator for the quantised Dirac field is
Q = q
_
d
3
x ψ

(x)ψ(x),
where we have merely inserted a multiplicative charge q. Then, if we insert our field expansion
(4.34), we find that
Q = q

s,p
_
b

s
(p)b
s
(p) −d

s
(p)d
s
(p)
_
. (4.40)
And hence, we can compute the commutator of this charge operator with the creation oper-
ators;
_
Q, b

s
(p)
¸
= qb

s
(p),
_
Q, d

s
(p)
¸
= −qd

s
(p);
4.4 The Feynman Propagator G
F
(x) 61
and obvious equivalents for the destruction operators. This allows us to see what the differ-
ence is in the particles that b

s
and d

s
create; which we summarise below.
Therefore, we have the interpretation that b

s
(p) creates a spin-
1
2
particle with charge Q =
+q, energy E and momentum p; and b
s
(p) destroys that particle. We also have that d

s
(p)
creates a spin-
1
2
antiparticle with charge Q = −q, energy E and momentum p; and d
s
(p)
destroys that antiparticle.
It is worth noting again that the vacuum state (which is the state with lowest energy)
satisfies
b
s
(p) [0) = d
s
(p) [0) = 0, ∀s, p.
And also that single particle states
b

s
(p) [0) , d

s
(p) [0) ,
have energies E(p) = +
_
p
2
+ m
2
; notice that neither the particle or anti-particle have
negative energies any more, relative to the vacuum.
4.4 The Feynman Propagator G
F
(x)
So far we have operators which create and destroy particles, but, we still need a way of
propagating from one place to another. The crucial quantity which does this is the Feynman
propagator. We define it for the complex KG field,
G
F
(x) = −i ¸0[ T
_
φ(x)φ

(0)
_
[0) , (4.41)
where the Dyson (also called the time-ordered) product is
T
_
φ(x)φ

(0)
_
= θ(t)φ(x)φ

(0) + θ(−t)φ

(0)φ(x), (4.42)
and the θ-function is defined
θ(t) =
_
1 t > 0,
0 t < 0.
(4.43)
Basically, what the Dyson product does, is to ensure that particles are created before they
are annihilated. φ(x) destroys particles, and φ

(x) creates particles. Hence, if t > 0, then G
F
represents the creation of a particle at t = 0 and subsequent destruction at time t. Similarly,
if t < 0, then an antiparticle is created at t and destroyed at time t = 0. That is, it ensures
causality is not violated.
Now,
φ(x)φ

(0) =

p,p

1
2V
_
E
p
E
p

_
a(p)a

(p

)e
−ipx
+ a(p)c(p

)e
−ipx
+c

(p)a

(p

)e
ipx
+ c

(p)c(p

)e
ipx
_
,
62 4 QUANTUM FIELDS
so that the only term to contribute upon contracting with the vacuum state is the first,
¸0[ φ(x)φ

(0) [0) =

p,p

1
2V
_
E
p
E
p

¸0[ a(p)a

(p

) [0) e
−ipx
=

p,p

1
2V
_
E
p
E
p

δ
pp
e
−ipx
=

p
e
−ipx
2V E
p
.
Similarly,
¸0[ φ

(0)φ(x) [0) =

p
e
ipx
2V E
p
.
Hence, we can use these to see that the Feynman propagator (4.41) in its two limits is
G
F
= −i

p
e
−ipx
2V E
p
, t > 0, (4.44)
G
F
= −i

p
e
ipx
2V E
p
, t < 0; (4.45)
where
px = p
µ
x
µ
= Et −p x.
We can write this using the θ-function as
G
F
(x) = −i

p
_
θ(t)
e
−ipx
2E
p
V
+ θ(−t)
e
ipx
2E
p
V
_
. (4.46)
These can be represented by diagrams, see f4.1.
t > 0
t
x
t
x
O O
x x
t < 0
Figure 4.1: Representations of the Feynman propagator. The direction of the arrow only denotes
whether the moving object is a particle or antiparticle. The arrow is sometimes emitted for bosons.
4.4 The Feynman Propagator G
F
(x) 63
Another way of evaluating the propagator is as a single term. Let us differentiate the
propagator (4.41) with respect to t, using
dθ(t)
dt
= δ(t),
dθ(−t)
dt
= −δ(t).
So,
∂G
F
(x)
∂t
= −iδ(t) ¸0[ φ(x)φ

(0) −φ

(0)φ(x) [0)
−i ¸0[ θ(t)
∂φ(x)
∂t
φ

(0) + θ(−t)φ

(0)
∂φ(x)
∂t
[0) .
Now, the first term is zero as
_
φ(t, x), φ

(t, x

)
¸
= 0. (4.47)
Hence,
∂G
F
(x)
∂t
= −i ¸0[ θ(t)
∂φ(x)
∂t
φ

(0) + θ(−t)φ

(0)
∂φ(x)
∂t
[0) .
Let us differentiate again,

2
G
F
∂t
2
= −i ¸0[ δ(t)
∂φ(x)
∂t
φ

(0) [0) −i ¸0[ θ(t)

2
φ(x)
∂t
2
φ

(0) [0)
+i ¸0[ δ(t)φ

(0)
∂φ(x)
∂t
[0) −i ¸0[ θ(−t)φ

(0)

2
φ(x)
∂t
2
[0)
= −iδ(t) ¸0[
_
∂φ(x)
∂t
, φ

(0)
_
[0) −i ¸0[ θ(t)

2
φ(x)
∂t
2
φ

(0) [0)
−i ¸0[ θ(−t)φ

(0)

2
φ(x)
∂t
2
[0) ,
where we have noticed the appearance of the commutator. Now, recall the KG equation,

2
φ
∂t
2
= (∇
2
−m
2
)φ,
which we can use in the last two terms. So, explicitly,
−i ¸0[ θ(t)

2
φ(x)
∂t
2
φ

(0) [0) = −i ¸0[ θ(t)(∇
2
−m
2
)φ(x)φ
+
(0) [0)
= −i(∇
2
−m
2
) ¸0[ θ(t)φ(x)φ
+
(0) [0) ,
−i ¸0[ θ(−t)φ

(0)

2
φ(x)
∂t
2
[0) = −i(∇
2
−m
2
) ¸0[ θ(−t)φ
+
(0)φ(x) [0) ;
so that their sum is just the Feynman propagator (4.41). Hence,

2
G
F
∂t
2
= −iδ(t) ¸0[
_
∂φ(x)
∂t
, φ

(0)
_
[0) + (∇
2
−m
2
)G
F
.
64 4 QUANTUM FIELDS
Therefore, taking the far right term over to the LHS, we see that we have
_

2
∂t
2
−∇
2
+ m
2
_
G
F
= −iδ(t) ¸0[
_
∂φ(x)
∂t
, φ

(0)
_
[0) ,
which is just
_

µ

µ
+ m
2
_
G
F
= −iδ(t) ¸0[
_
∂φ(x)
∂t
, φ

(0)
_
[0) . (4.48)
We can show, by simply expanding in terms of modes, that
_
∂φ(t, x)
∂t
, φ

(t, x

)
_
= −iδ
(3)
(x −x

). (4.49)
Hence, upon substitution of this into (4.48), we see that
_

µ

µ
+ m
2
_
G
F
(x) = −δ
(4)
(x). (4.50)
This is a point-source Klein-Gordon equation, and its solution is the Feynman propagator.
This can also be derived by differentiating (4.46). One should recall that such solutions are
Green functions, and that Green functions can be integrated over a charge distribution to
give the solution for the inhomogeneous Klein-Gordon equation. That is, we can use the
Green function G
F
(which is the solution to the above point-source equation) to solve
_

µ

µ
+ m
2
_
φ(x) = −ρ(x),
where ρ(x) is an arbitrary source.
To find the Green function G
F
(x), we solve (4.50) by Fourier transforms. So, we have
G
F
(x) =
_
d
4
k
(2π)
4
e
−ikx
˜
G
F
(k) (4.51)
δ
(4)
(x) =
_
d
4
k
(2π)
4
e
−ikx
,
noting that a δ-function can be written as the Fourier tranform of unity. Now,

µ

µ
e
−ikx
= −k
2
e
−ikx
,
so that upon substitution of these transforms into (4.50), we have
_
d
4
k
(2π)
4
_
−k
2
+ m
2
_
e
−ikx
˜
G
F
(k) = −
_
d
4
k
(2π)
4
e
−ikx
.
Thus, as the integrals are identical, we equate the integrands;
˜
G
F
(k)
_
−k
2
+ m
2
_
= −1,
4.4 The Feynman Propagator G
F
(x) 65
and hence,
˜
G
F
(k) =
1
k
2
−m
2
, k
2
≡ k
µ
k
µ
.
We then substitute this back into (4.51),
G
F
(x) =
_
d
4
k
(2π)
4
e
−ikx
k
2
−m
2
. (4.52)
This then gives us an expression for the Green function; once we have performed the inte-
gral, we have an expression which allows us to solve any inhomogeneous KG equation, by
integrating over it. However, this expression has a singularity at k
2
= m
2
;
singularity : k
0
= ±ω
k
≡ ±

k
2
+ m
2
.
This pole is on the real axis; to get around this problem, we use the Feynman prescription,
which is to displace the pole from the real axis, do the integral, then put everything back as
it was. That is, we modify the Green function (4.52) into
G
F
(x) =
_
d
4
k
(2π)
4
e
−ikx
k
2
−m
2
+ i
, > 0, (4.53)
perform the integral, and at the end of any calculations, set = 0. We evaluate the time-
component,
I =
_

−∞
dk
0
e
−ik
0
t
(k
0
)
2
−ω
2
k
+ i
, ω
k
=

k
2
+ m
2
, (4.54)
by contour integration, with poles at
k
0
= ±(ω
k
−i) .
With reference to Figure (4.2), we see the complex k
0
plane, with poles, and the contour
we shall integrate over. We separate the cases t < 0 and t > 0.
t > 0: Lower-half-plane In this case, we close the contour in the lower half plane, and
let the radius of the semi-circle go to infinity. There is no contribution from the semi-circle,
as k
0
< 0. We do the contour integral by noting that Cauchy’s theorem states that the
integral of any analytic function, over an (anti-clockwise) closed path, is just 2πi multiplied
by the sum of the residues enclosed by the path. Thus, for our contour which is closed in the
lower-half plane, we only have a pole at k
0
= ω
k
− i. Therefore, the value of the integral
(4.54) is
I = −2πi
e
−iω
k
t

k
.
Notice that the minus-sign is due to our contour being clockwise, and that we have set = 0.
66 4 QUANTUM FIELDS
−ω
k
+ i
ω
k
−i
Im(k
0
)
Re(k
0
)
Figure 4.2: The complex plane, in k
0
-space; with the poles marked on. We evaluate (4.54) over the
contour shown. This is the t > 0 case.
t < 0: Upper-half-plane In this case, we close the contour in the upper-half-plane, so
that the only contributing pole is at k
0
= −ω
k
+ i.
Hence, using these results, we can write down (4.53), in the two regimes,
G
F
(x) = −
i
(2π)
3
_
d
3
p
e
−i(Et−p·x)
2E
p
, t > 0, (4.55)
G
F
(x) = −
i
(2π)
3
_
d
3
p
e
i(Et−p·x)
2E
p
, t < 0. (4.56)
Notice that these are the same as (4.44), (4.45); only now we have moved to continuous
momentum. Thus, this is the same as the causal result.
We shall leave propagators for the moment, and discuss interactions. We shall come back
to them when we need a term which allows the creation and destruction of a particle within
a process (i.e. when we discuss exchange bosons).
4.5 Interactions
4.5.1 The Interaction Picture
So far, we have two ways of describing the time evolution of a quantum system: the
Schrodinger and Heisenberg pictures. In the Schrodinger picture, only the states evolve
and not the operators, where the states obey
i

∂t
[ψ(t))
S
= H[ψ(t))
S
.
4.5 Interactions 67
The formal solution to this is just
[ψ(t))
S
= e
−iHt
[ψ(0))
S
.
Hence, we can use this to find the expectation value of an operator,
S
¸ψ(t)[ A
S
[ψ(t))
S
=
S
¸ψ(0)[ e
iHt
A
S
e
−iHt
[ψ(0))
S
.
We can take this, and redefine a few terms,
[ψ)
H
≡ [ψ(0))
S
, A
H
(t) ≡ e
iHt
A
S
e
−iHt
, (4.57)
so that the expected value becomes just
S
¸ψ(t)[ A
S
[ψ(t))
S
=
H
¸ψ[ A
H
(t)[ψ)
H
.
Hence, we see that by writing (4.57), we have preserved the form of the expectation value of
an operator. What we have done, is to change the object that is evolving in time: from the
state to the operator. In the Heisenberg picture, the state is constant, and all time evolution
is carried out by the operators. Simply by differentiating the second of (4.57), we can derive
the Heisenberg equation of motion,
−i
dA
H
dt
= [H, A
H
(t)] . (4.58)
For perturbation theory, we use a third picture, the interaction picture. To define the picture,
we write the Hamiltonian as a sum of a free Hamiltonian and an interaction Hamiltonian,
H = H
0
+ H
I
. (4.59)
We also define the interaction picture states and operators as
[ψ(t))
I
= e
iH
0
t
[ψ(t))
S
, A
I
(t) = e
iH
0
t
A
S
e
−iH
0
t
. (4.60)
Thus, differentiating, we get
i
d
dt
[ψ(t))
I
= H
I
(t)[ψ(t))
I
, −i
dA
I
dt
= [H
0
, A
I
(t)] , (4.61)
where
H
I
(t) = e
−iH
0
t
H
I,S
e
iH
0
t
. (4.62)
If we have that H = H
0
, then the Heisenberg and interaction pictures are identical. If
H = H
0
+H
I
, then we have an intermediate picture in which operators (which includes field
operators) evolve according to the free Hamiltonian H
0
. Hence, our free-field results in the
Heisenberg picture hold for interacting fields, in the interaction picture.
68 4 QUANTUM FIELDS
4.5.2 The S-matrix
For time evolution, it is useful to write
[ψ(t))
I
= U(t, t
0
) [ψ(t
0
)) ,
where
i
d
dt
U(t, t
0
) = H
I
(t)U(t, t
0
), (4.63)
under the boundary condition U(t
0
, t
0
) = 1. We define the S-matrix via
S = lim
τ→∞
U(
τ
2
, −
τ
2
) = U(∞, −∞). (4.64)
Hence, we start long before particles get close enough to interact, and finish longer after
particles have finished interacting. The initial state (i.e. at t = −∞) is just
[ψ(−∞))
I
= [ψ
i
)
I
,
which is a free-particle state (i.e. an eigenfunction of the free Hamiltonian H
0
). At finite t,
the state looks like
[ψ(t))
I
= U(t, −∞)[ψ
i
)
I
,
and is a complicated state of interacting particles. The final state (i.e. at t = ∞), is just
[ψ(∞))
I
= S[ψ
t=−∞
)
I
,
after using the definition of the S-matrix, (4.64). Thus, we can write the final state as a
linear sum of free particle states,
[ψ(∞))
I
= [ψ
f
)
I
=

i
o


i
)
I
.
We can write the final state like this, as we have stated that the S-matrix ends a long time
after the particles have finished interacting. Hence, we can easily see that the amplitude to
end up in the final state f is
o

= ¸ψ
f
[ S [ψ
i
) , (4.65)
which occurs with probability
P
fi
= [o

[
2
. (4.66)
These matrix elements are the only observables of the system, and summarise the physics of
a set of interactions.
One can think about the S-matrix as a “mixing matrix”, which describes the transitions
from the initial state that took place to give the final state. Those familiar with statistical
physics can think about this matrix as being like the Markov chain transition matrix. The
matrix elements o

describe the amplitude for a particular initial and final state.
4.5 Interactions 69
The Unitarity of the S-matrix Recall that
[ψ(t))
I
= e
iH
0
t
[ψ(t))
S
from the first of (4.60). Hence, we can write
[ψ(t))
I
= e
iH
0
t
[ψ(t))
S
= e
iH
0
t
e
−iHt
[ψ(t
0
))
S
= e
−H
0
t
e
−iHt
e
iHt
0
[ψ(t
0
))
I
= e
iH
0
t
e
−iH(t−t
0
)
[ψ(t
0
))
I
= U(t, t
0
)[ψ(t
0
))
I
.
Now, as H and H
0
are both Hermitian, we thus have that U is unitary;
U
−1
(t, t
0
) = U(t
0
, t) = U

.
Hence, taking the limit t →∞ and t
0
→−∞, we arrive at the statement that the S-matrix
is unitary,
S

S = 1. (4.67)
The consequence of this, is that probabilities are conserved. For example,

f
P
fi
=

f
[¸ψ
f
[ S [ψ
i
)[
2
=

f
¸ψ
i
[ S


f
) ¸ψ
f
[ S [ψ
i
)
= ¸ψ
i
[ S

S [ψ
i
)
= ¸ψ
i

i
).
The third equality follows from the second due to the completeness of eigenfunctions. Hence,
we see that if a state is normalised, then the normalisation is preserved.
4.5.3 The S-matrix Expansion
Now, from (4.63), we have the formal solution
U(t, t
0
) = 1 −i
_
t
t
0
dt
1
H
I
(t
1
)U(t
1
, t
0
), (4.68)
under the boundary condition U(t
0
, t
0
) = 1. The reason we say “formal”, is that the thing
we are trying to solve for, U(t, t
0
), appears within the integral itself. Now, we can form an
iterative solution, by inserting the formal solution into itself,
U(t, t
0
) = 1 −i
_
t
t
0
dt
1
H
I
(t
1
) + (−i)
2
_
t
t
0
dt
1
H
I
(t
1
)
_
t
1
t
0
dt
2
H
I
(t
2
)U(t
2
, t
0
).
70 4 QUANTUM FIELDS
We can imagine doing this iterative insertion many times,
U(t, t
0
) = 1 −i
_
t
t
0
dt
1
H
I
(t
1
) + (−i)
2
_
t
t
0
dt
1
H
I
(t
1
)
_
t
1
t
0
dt
2
H
I
(t
2
) + . . . .
So, if we take t → ∞, and t
0
→ −∞, this becomes the expansion of the S-matrix,
U(∞, −∞) = S, by definition. Hence, we write this as a sum of terms of incrementing
orders,
S = S
(0)
+ S
(1)
+ S
(2)
+ . . . ,
where S
(0)
= 1 denotes no interaction, and for example,
S
(1)
= −i
_

−∞
dt
1
H
I
(t
1
), (4.69)
S
(2)
= (−i)
2
_

−∞
dt
1
H
I
(t
1
)
_
t
1
−∞
dt
2
H
I
(t
2
). (4.70)
Now, as the expression stands, it is not covariant: we have singled out the time coordinate.
Let us rewrite S
(2)
in a more symmetric form by interchanging t
1
and t
2
,
S
(2)
=
(−i)
2
2
_

−∞
dt
1
H
I
(t
1
)
_
t
1
−∞
dt
2
H
I
(t
2
) +
(−i)
2
2
_

−∞
dt
2
H
I
(t
2
)
_
t
2
−∞
dt
1
H
I
(t
1
).
We can notice that in the first term, t
2
< t
1
and in the second term, t
1
< t
2
. We can then
notice that in the both terms, the “earlier” Hamiltonian appears to the right of the “later”
Hamiltonian;
t
1
> t
2
⇒ H
I
(t
1
)H
I
(t
2
),
t
2
> t
1
⇒ H
I
(t
2
)H
I
(t
1
).
Hence, we have an expression which is of the time-ordered form (or Dyson product),
S
(2)
=
(−i)
2
2
_

−∞
dt
1
_

−∞
dt
2
T ¦H
I
(t
1
)H
I
(t
2
)¦ . (4.71)
We have set the upper-limit on the second integral to ∞ as we notice that the whole t
1
−t
2
-
plane will be covered by such a Dyson product. More generally, for the n
th
-order term,
S
(n)
=
(−i)
n
n!
_

−∞
dt
1
. . .
_

−∞
dt
n
T ¦H
I
(t
1
) . . . H
I
(t
n
)¦ . (4.72)
This now does not single out a specific time value, as the case was before we wrote the sym-
metric sum, but does single out time from space; hence, the expression still is not explicitly
covariant. We now note that in field theory the Hamiltonian H is the spatial integral of the
Hamiltonian density H,
H
I
(t) =
_
d
3
x H
I
,
4.5 Interactions 71
where we may have multiple fields which are functions of the 4-position vector x,
H
I
(x) = H
i

i
, ∂
µ
φ
i
) .
Finally, using the Hamiltonian density, rather than the Hamiltonian, the expressions (4.69)
and (4.70) can be written in a covariant form,
S
(1)
= (−i)
_
d
4
x H
I
(x), (4.73)
S
(2)
=
(−i)
2
2
_
d
4
x
1
_
d
4
x
2
T ¦H
I
(x
1
)H
I
(x
2
)¦ , (4.74)
S
(n)
=
(−i)
n
n!
_
d
4
x
1
. . .
_
d
4
x
n
T ¦H
I
(x
1
) . . . H
I
(x
n
)¦ . (4.75)
Hence, we have an explicitly covariant form of the S-matrix expansion.
72 4 QUANTUM FIELDS
73
5 Decays
We shall illustrate calculations by considering examples.
5.1 Simple 2-body Decay: φ
3
Let us consider an example of a Higgs boson decaying to two tauons,
H
0
−→τ
+
+ τ

. (5.1)
The scalar uncharged Higgs boson H
0
has zero spin; the tauons τ
±
are spin-
1
2
leptons (i.e.
fermions), with mass m
τ
≈ 1.8GeV/c
2
. In terms of momentum and spin, (p, s), the decay
can be written
(p, 0) −→(k

, s

) + (k, s).
H
0
τ
+
τ

p
k

, s

k, s
Figure 5.1: Schematic of the 2-body decay (5.1).
Now, in the standard model, the interaction is of the form
H
I
(x) = gN
_
¯
ψ(x)ψ(x)φ(x)
_
, (5.2)
where N denotes normal (or time) ordering. The choice of interaction term is effectively “the
problem”; different interaction terms produce different physics. Hence, a given interaction
term is “the theory” we are investigating. Thus, above, we are considering a φ
3
-theory.
Notice that
¯
ψ(x)ψ(x) is a Lorentz scalar density. As H
0
is an uncharged boson, we have that
φ(x) is a Hermitian scalar field; as τ
±
are charged particles with spin, we have that ψ(x) is
a Dirac spinor field. The dimensionless coupling parameter g is
g = m
τ
_
(

2
_
1/2
, ( ≡ 1.16 10
−5
GeV
−2
; (5.3)
74 5 DECAYS
where ( is Fermi’s coupling constant. To first order, we need to evaluate the matrix element
o
(1)

= −i
_
d
4
x
¸
f = τ
+
τ

¸
¸
H
I
(x)
¸
¸
i = H
0
_
,
which denotes the amplitude for a decay of H
0
→τ
+


, with given momenta. Notice that
other matrix elements are possible, such as ¸f = τ
+
τ

H
0
[ H
I
(x) [i = 0), which would denote
particle creation. We are interested therefore in
2 body decay : ¸k, s; k

, s

[ H
I
[p) .
The neutral scalar field has its expansion being
φ(x) = φ
(+)
(x) + φ
(−)
(x)
=

p
1
(2V ω
p
)
1/2
_
a(p)e
−ipx
+ a

(p)e
ipx
_
,
so that φ
(+)
destroys and φ
(−)
creates netural scalar bosons; where
ω
p
= ω(p) =
_
p
2
+ m
2
.
The charged fermionic field expansion is
ψ(x) = ψ
(+)
(x) + ψ
(−)
(x)
=

p,s
1
(2V E
p
)
1/2
_
b
s
(p)u
s
(p)e
−ipx
+ d

s
(p)v
s
(p)e
ipx
_
,
so that ψ
(+)
destroys τ

and ψ
(−)
creates τ
+
. As the field is a Dirac field, we also note that
¯
ψ
(−)
creates τ

, and
¯
ψ
(+)
destroys τ
+
, where an over-bar denotes the Dirac adjoint, so that
¯
ψ(x) =
¯
ψ
(−)
(x) +
¯
ψ
(+)
(x).
So, our initial state is a single Higgs boson, with momentum p, which we can create from
the vacuum,
[i) =
¸
¸
H
0
, p
_
= a

(p) [0) .
The final state similarly is
[f) =
¸
¸
τ

, k, s; τ
+
, k

s

_
= b

s
(k)d

s

(k

) [0) .
So, to first order in H
I
, the matrix element is
o

= −ig
_
d
4
x ¸f[ N
_
¯
ψ(x)ψ(x)φ(x)
_
[i) .
Now, if we were to plug in the expressions for the field expansions, and expand out, we would
have 8 terms. In fact, each of these terms correspond to a different process, but will only
5.1 Simple 2-body Decay: φ
3
75
give non-zero contribution if they correspond to the process we are currently dealing with
(due to the orthonormality of states, and our specific choice of initial and final states). One
can see that the only contributing term is
o

= −ig
_
d
4
x ¸f[
¯
ψ
(−)
(x)ψ
(−)
(x)φ
(+)
(x) [i) , (5.4)
where the first term creates a τ

, the second creates a τ
+
and the third destroys a H
0
.
So, let us treat each field in turn. The term on the far RHS is
φ
(+)
(x) [i) = φ
(+)
(x) [p)
= φ
(+)
(x)a

(p) [0) ,
but,
φ
(+)
(x) =

p
1
(2V ω
p
)
1/2
a(p)e
−ipx
,
and hence
φ
(+)
(x) [i) =

p

1
(2V ω
p
)
1/2
a(p

)e
−ip

x
a

(p) [0) .
Now, we can write a(p

)a

(p) = a

(p)a(p

)+δ
pp
, by the usual commutation relation. Hence,
doing so,
φ
(+)
(x) [i) =

p

1
(2V ω
p
)
1/2
e
−ip

x
_
a

(p)a(p

) + δ
pp

_
[0) .
The first bracketed term gives zero upon action with the vacuum state, as a(p) [0) = 0. The
Kronecker-delta term simply filters out p

= p in the summation to leave
φ
(+)
(x) [i) =
1
(2V ω
p
)
1/2
e
−ipx
[0) . (5.5)
In an entirely analogous manner, one can easily see that
ψ
(+)
(x) [k, s) =
1
(2V E
k
)
1/2
u
s
(k)e
−ikx
[0) .
Now, it is the adjoint of this term which appears in (5.4); hence, taking the adjoint,
¸k, s[
¯
ψ
(−)
(x) =
1
(2V E
k
)
1/2
¯ u
s
(k)e
ikx
¸0[ . (5.6)
And the final term can be seen by analogy again,
¸k

, s

[ ψ
(−)
(x) =
1
(2V E

k

)
1/2
v
s
(k

)e
ik

x
¸0[ . (5.7)
76 5 DECAYS
Hence, putting together (5.5), (5.6) and (5.7), we see that
¸f[
¯
ψ
(−)
ψ
(−)
φ
(+)
[i) =
1
(2V E
k
)
1/2
1
(2V E

k

)
1/2
1
(2V ω
p
)
1/2
e
ix(k+k

−p)
¯ u
s
(k)v
s
(k

), (5.8)
where we have been careful to maintain the order of the spinors, and noting that ¸0[0) = 1.
Now, in (5.4), we have the integral over space of this quantity: this will only have the effect
on the exponential,
_
d
4
x e
ix(k+k

−p)
= (2π)
4
δ
(4)
(k + k

−p). (5.9)
If we are not in the limit of an infinite volume/time box, we will have that
_
d
4
x e
ix(k+k

−p)
= V Tδ
k+k

,p
.
So, putting together (5.8) and (5.9) in (5.4), we see that
o

=
1
(2V E
k
)
1/2
1
(2V E

k

)
1/2
1
(2V ω
p
)
1/2
(2π)
4
δ
(4)
(k + k

−p)/

, (5.10)
where we have defined the Feynman amplitude
/

= −ig¯ u
s
(k)v
s
(k

). (5.11)
Let us now consider the three types of term in (5.10); its structure is
o

= normalisation energy-momentum conservation dynamics.
The initial three factors involving energy are the normalisation, and will be common to all
processes. The second term with the delta-function is infact the statement of conservation
of energy-momentum. The final term /

is the only term that depends on the specific
interaction, and is where all dynamical information is kept. We can represent /

by a
Feynman graph, as in Figure (5.1). See Figure (5.2) for the assignment of mathematical
factors to specific features of a Feynman graph. See Figure (5.3) to see the whole amplitude
for a given graph.
5.2 Decay Rates & Lifetimes
We now want to calculate the measured decay rate for a given process. The probability of
decay per unit time is
Γ =

f
Γ
fi
=

f
[o

[
2
T
.
5.2 Decay Rates & Lifetimes 77
outgoing fermion
outgoing antifermion
¯ u
s
(k)
v
s
(k)
vertex factor
−ig
Figure 5.2: The assignment of factors to the elements of a Feynman graph.
M
fi
= −ig¯ u
s
(k)v
s
(k

)
k, s
k

, s

Figure 5.3: The Feynman amplitude for the given graph.
Hence, inserting (5.10), we see that
Γ =

k,s,k

,s

1
T
1
8V
3
ω
p
E
k
E

k

T
2
V
2
δ
k+k

,p
[/

[
2
,
where we are in the finite volume/time limit, to allow us to have a meaning in squaring the
delta-function. Now we have squared the delta-function, we can take the limit V, T →∞, so
that the energy-momentum conservation term becomes just
V Tδ
k+k

,p
−→(2π)
4
δ
(4)
(k + k

−p),
and the two phase-space terms
1
V

k
−→
_
d
3
k
(2π)
3
.
78 5 DECAYS
Thus, all factors that related to the volume and time of the box disappear, leaving us with
Γ =
1

p

s,s

_
d
3
k
(2π)
3
2E
k
_
d
3
k

(2π)
3
2E

k

(2π)
4
δ
(4)
(k + k

−p) [/

[
2
.
(5.12)
To continue, we need to evaluate the spin sum and do the phase space integrals. Let us do
the spin sum first.
We must evaluate

s,s

[/

[
2
= g
2

s,s

[¯ u
s
(k)v
s
(k

)[
2
.
So, to begin,

s,s

[/

[
2
=

/

/


= g
2

¯ u(k)v(k

) (¯ u(k)v(k

))

,
where we drop the s, s

, as they are implied by the argument of the spinor. Now, notice that
the bracketed term is a scalar (indeed, as is the term directly before it). Therefore, we can
treat complex conjugation exactly as Hermitian conjugation,
(¯ u(k)v(k

))

= (¯ u(k)v(k

))

.
If we recall that the Dirac adjoint was defined to be
¯
A = A

γ
0
, we thus have
(¯ u(k)v(k

))

=
_
u

(k)γ
0
v(k

)
_

.
Hence, taking the Hermitian conjugate (which has the effect of conjugating everything, and
swapping the order of all terms),
_
u

(k)γ
0
v(k

)
_

= v

(k


0†
u(k),
but, γ
0†
= γ
0
, so that we see the Dirac adjoint appear again,
_
u

(k)γ
0
v(k

)
_

= ¯ v(k

)u(k).
Therefore,

s,s

[/

[
2
= g
2

¯ u(k)v(k

)¯ v(k

)u(k).
Now, a term such as ¯ u(k)v(k

) is a scalar; hence, this is just two scalars multiplied together.
Let us then write this with indices, so that all terms involved are numbers, and can hence
be commuted at will,

s,s

[/

[
2
= g
2

¯ u
a
(k)v
a
(k

)¯ v
b
(k

)u
b
(k)
= g
2

u
b
(k)¯ u
a
(k)v
a
(k

)¯ v
b
(k

).
5.2 Decay Rates & Lifetimes 79
We now use two identities of the spinors,

u
b
(k)¯ u
a
(k) = (k/ + m
τ
)
ba
,

v
a
(k

)¯ v
b
(k

) = (k/

−m
τ
)
ab
,
where k/ ≡ γ
µ
k
µ
; and hence

s,s

[/

[
2
= g
2
(k/ + m
τ
)
ba
(k/

−m
τ
)
ab
.
Now, an expression with indices arranged so is just a trace, so that

s,s

[/

[
2
= g
2
Tr [(k/ + m
τ
) (k/

−m
τ
)]
= g
2
_
Tr (k/k/

) −m
τ
Tr (k/) + m
τ
Tr (k/

) −m
2
τ
Tr (1
4
)
_
.
(5.13)
Now the traces of the middle two terms are zero, by noting
Tr (k/) = Tr (γ
µ
k
µ
) = k
µ
Tr (γ
µ
) = 0,
where we first note that k
µ
is not a matrix, and secondly that the γ-matrices are traceless.
The last term in (5.13) is fairly obviously just Tr (1
4
) = 4. Hence, using these two results,
we see that (5.13) reduces down to

s,s

[/

[
2
= g
2
_
Tr (k/k/

) −4m
2
τ
_
. (5.14)
To evaluate the first term, we write
k/k/

= γ
µ
γ
ν
k
µ
k
ν
= γ
ν
γ
µ
k
ν
k
µ
.
We now use ¦γ
µ
, γ
ν
¦ = 2g
µν
1
4
(where we must remember that the LHS of this is a matrix
equation, and hence so must the RHS – therefore the introduction of the unit matrix). Hence,
we write
Tr (γ
µ
γ
ν
) + Tr (γ
ν
γ
µ
) = 2g
µν
Tr (1
4
).
Now, due to the cyclicality of the trace, the LHS is just 2Tr (γ
µ
γ
ν
). Therefore,
Tr (γ
µ
γ
ν
) = g
µν
Tr (1
4
) = 4g
µν
.
Therefore,
Tr (k/k/

) = 4g
µν
k
µ
k

ν
.
Hence, using this in (5.14), we see that

s,s

[/

[
2
= 4g
2
_
g
µν
k
µ
k

ν
−m
2
τ
_
,
80 5 DECAYS
and therefore,
g
2

s,s

[¯ u
s
(k)v
s
(k

)[
2
= 4g
2
_
kk

−m
2
τ
_
, (5.15)
where kk

= k
µ
k

µ
.
We now specialise to the rest-frame of the Higgs boson, P
µ
= (m
H
, 0). Therefore, conser-
vation of 4-momentum implies that
k = −k

, E(k) = E(k

) =
m
H
2
. (5.16)
Hence,
E
2
(k) = k
2
+ m
2
τ
⇒ k
2
= E
2
(k) −m
2
τ
=
m
2
H
4
−m
2
τ
. (5.17)
Therefore, noting that the kk

term in (5.15) can be written
kk

= k
µ
k

µ
= E
2
(k) +k
2
= 2k
2
+ m
2
τ
,
where we have taken careful note of the change in sign. Hence, inserting (5.17) into this
expression, we see that
kk

=
m
2
H
2
−m
2
τ
,
which we can insert into (5.15) to see that

s,s

[/

[
2
= g
2

s,s

[¯ u
s
(k)v
s
(k

)[
2
= 8g
2
_
m
2
H
4
−m
2
τ
_
. (5.18)
Let us now do the phase-space integral (5.12), in the rest-frame of the Higgs. So, we must
compute
I =
_
d
3
k
2E
k
_
d
3
k

2E

k

δ (E(k) + E(k

) −m
H
) δ
3
(k +k

),
where we have ignored the factors of 2π (to be put back in later), and we have also split
the delta-function into its space and time parts. The δ
(3)
(k +k

) term, whilst evaluating the
d
3
k

-integral, filters out k = −k

as the only non-zero contribution, so that
I =
_
d
3
k
4E
k
E
−k
δ (E(k) + E(−k) −m
H
) .
We now use the second of (5.16) to write
I =
_
d
3
k
4E
2
k
δ (2E(k) −m
H
) ,
5.2 Decay Rates & Lifetimes 81
and then E(k) =
_
k
2
+ m
2
τ
, to see that
I =
_
d
3
k
4E
2
k
δ
_
2
_
k
2
+ m
2
τ
−m
H
_
. (5.19)
To evaluate the delta-function, we use the usual relation for a functional argument,
δ (f(x)) =

x
0
δ(x −x
0
)
[f

(x
0
)[
, f(x
0
) = 0.
Now, the argument of the delta-function in (5.19) is zero when
k = k
0
=
_
m
2
H
4
−m
2
τ
,
and the value of the derivative of the argument, at k
0
is
f

(k
0
) =
2k
E(k)
.
Therefore, using these results, we have that (5.19) reads
I =
_
d
3
k
4E
2
k
E([k[)
2[k[
δ
_
[k[ −
_
m
2
H
4
−m
2
τ
_
,
or, if we assume no angular dependence,
I =
_
4π[k[
2
4E
2
k
d[k[
E([k[)
2[k[
δ
_
[k[ −
_
m
2
H
4
−m
2
τ
_
=
4π[k
0
[
8E(k
0
)
, k
0

_
m
2
H
4
−m
2
τ
.
Then, using the second of (5.16), this cancels down to
I =
π[k
0
[
m
H
, k
0

_
m
2
H
4
−m
2
τ
.
Hence, using this result and (5.18), we see that the decay rate (5.12) becomes
Γ =
1
2m
H
1
(2π)
2
8g
2
_
m
2
H
4
−m
2
τ
_
π[k
0
[
m
H
=
g
2
_
m
2
H
4
−m
2
τ
_
3/2
πm
2
H
, (5.20)
82 5 DECAYS
after using ω
p
= m
H
in the rest-frame of the Higgs. Now, in the limit m
H
¸m
τ
(which is a
very good approximation),
Γ
_
H
0
→τ
+
τ

_
=
g
2
m
H

.
We can use g from (5.3) to see that
Γ
_
H
0
→τ
+
τ

_
= /m
H
m
2
τ
, / ≡
(
8


. (5.21)
Therefore, we have found that the decay rate for the process H
0
→τ
+
τ

increases with the
mass of the Higgs particle (infact, it will increase for the larger product mass as well, as there
is a larger phase space to “decay into”). We can somewhat generalise this decay rate to any
lepton = e

, µ

, τ

, to give the decay rate
Γ
_
H
0

+


_
= /m
H
m
2

. (5.22)
For quarks, we have to multiply the answer by 3, as there are 3 colour states that could be
“decayed into”, hence increasing the phase space further;
Γ
_
H
0
→q¯ q
_
= 3/m
H
m
2
q
.
As the top t quark has the highest mass, the dominant decay channel is H
0
→t
¯
t.
5.3 Example: Decay of a Scalar to two Scalars
Let us consider a second example, of a neutral scalar particle decaying into two charged
scalars. We will compute the matrix element /

, the general decay rate Γ and the decay
rate of a stationary particle.
So, consider
σ −→φ

+ φ
+
p −→k +k

.
The interaction Hamiltonian we are considering is
H
I
= gφ

(x)φ(x)σ(x),
where the field expansions are
σ(x) =

p
1
(2V E
p
)
1/2
_
a(p)e
−ipx
+ a

(p)e
ipx
_
,
φ(x) =

k
1
(2V ω
k
)
1/2
_
b(k)e
−ikx
+ d

(k)e
ikx
_
;
and
px = p
µ
x
µ
= Et −p x,
E
p
=
_
p
2
+ m
2
σ
, ω
k
=
_
k
2
+ m
2
φ
.
5.3 Example: Decay of a Scalar to two Scalars 83
The important features of these expansions are: the σ-field is uncharged, which is modeled
by being Hermitian (i.e. we use a and a

). In contrast, the φ-field is charged, modeled by
not being Hermitian, so we use b and d

.
We need to first compute o

,
o

= −i
_
d
4
x ¸f[ H
I
(x) [i) ,
where the initial and final states are just
[i) = [σ, p) , [f) =
¸
¸
φ

, k; φ
+
, k

_
.
These states can be created from the vacuum via application of the creation operators;
[i) = a

(p) [0) , [f) = b

(k)d

(k

) [0) .
Then, using our interaction Hamiltonian, we want to compute
g ¸f[ φ

φσ [i) = g ¸0[ d(k

)b(k)φ

φσa

(p) [0) .
The only non-zero terms in the field expansion can then be deduced by requiring those which
“null out” those operators which are already present. The term from σ will be a(p)e
−ipx
, as
that will destroy the particle the already present a

(p)-term has created. Similarly, we also
need d

(k

)- and b

(k)-terms to create the particles that are destroyed by the already present
d(k

)b(k)-terms. We can get d

(k

) simply from the field expansion as it stands (also picking
up a factor of e
ik

x
). We get the b

(k)-term by conjugating the field expansion, in which case
we pick up a factor of e
ikx
as well. Hence, we have (also inserting the factors of energy and
volume we have neglected to mention thus far),
g ¸f[ φ

φσ [i) = g ¸0[ d(k

)b(k)b

(k)d

(k

)a(p)a

(p) [0)

1
(2V E
p
)
1/2
1
(2V ω
k
)
1/2
1
(2V ω
k
)
1/2
e
ix(k+k

−p)
.
The entire “state” part contracts to unity (by construction), to leave just
g ¸f[ φ

φσ [i) = g
1
(2V E
p
)
1/2
1
(2V ω
k
)
1/2
1
(2V ω
k
)
1/2
e
ix(k+k

−p)
.
Thus, the integral of this is just o

, so that
o

= −i
_
d
4
x g ¸f[ φ

φσ [i)
=
_
d
4
x
1
(2V E
p
)
1/2
1
(2V ω
k
)
1/2
1
(2V ω
k
)
1/2
e
ix(k+k

−p)
/

,
84 5 DECAYS
where we have defined the Feynman amplitude to be
/

≡ −ig.
This rather simple amplitude is just because every particle involved is scalar, and hence we
only have a contribution from the vertex itself. We now use the property that
_
d
4
x e
ix(k+k

−p)
= (2π)
4
δ
(4)
(k + k

−p),
so that we can do the integral in o

;
o

=
1
(2V E
p
)
1/2
1
(2V ω
k
)
1/2
1
(2V ω
k
)
1/2
(2π)
4
δ
(4)
(k + k

−p)/

.
Hence, we have computed an expression for the matrix element for neutral scalar decay into
charged scalar particles.
Let us now compute the decay rate. To do so, we perform a summation over all possible
final states;
Γ =

f
[o

[
2
T
.
To be able to compute [o

[
2
, (infact, more precisely, to compute the square of the delta-
function), we go into the finite-box limit, so that
(2π)
4
δ
(4)
(k + k

−p) = V Tδ
k+k

,p
.
Hence, squaring this:
_
(2π)
4
δ
(4)
(k + k

−p)
_
2
= V
2
T
2

k+k

,p
)
2
,
but

k+k

,p
)
2
= δ
k+k

,p
.
So,
_
(2π)
4
δ
(4)
(k + k

−p)
_
2
= V
2
T
2
δ
k+k

,p
.
Hence, upon computing the square of the matrix element, we easily see that
[o

[
2
=
1
(2V E
p
)
1
(2V ω
k
)
1
(2V ω
k
)
V
2
T
2
δ
k+k

,p
[/

[
2
.
Let us cancel off one of the factors of V , and divide by T, so that
[o

[
2
T
=
1
2E
p
1
(2V ω
k
)
1
(2V ω
k
)
V Tδ
k+k

,p
[/

[
2
,
5.3 Example: Decay of a Scalar to two Scalars 85
and we now send V Tδ
k+k

,p
→(2π)
4
δ
(4)
(k + k

−p), so that
[o

[
2
T
=
1
2E
p
1
(2V ω
k
)
1
(2V ω
k
)
(2π)
4
δ
(4)
(k + k

−p)[/

[
2
.
To compute the decay rate, we must sum this expression over all possible final states, so that
Γ =

f
[o

[
2
T
=

k,k

1
2E
p
1
(2V ω
k
)
1
(2V ω
k
)
(2π)
4
δ
(4)
(k + k

−p)[/

[
2
.
To perform the summation, we take the infinite-box limit, so that

k
1
V
−→
_
d
3
k
(2π)
3
.
Hence,
Γ =
1
2E
p
_
d
3
k
(2π)
3

k
_
d
3
k

(2π)
3

k

(2π)
4
δ
(4)
(k + k

−p)[/

[
2
.
This is a very general expression for the decay rate. We now specialise to the rest frame of
the decaying particle; that is, we will compute the decay rate of a σ-particle at rest. This
will allow us to compute the delta-function.
In this frame, we have p = 0, so that
E
p
=
_
p
2
+ m
2
σ
= m
σ
.
Now, we can always split up the 4-dimensional delta-function, thus
δ
(4)
(k + k

−p) = δ(ω
k
+ ω
k
−E
p

(3)
(k +k

−p) .
In the rest frame of the decaying particle, the second 3-D delta-function just becomes δ
(3)
(k+
k

). Hence, using this, the decay rate becomes
Γ =
1
2E
p
_
d
3
k
(2π)
3

k
_
d
3
k

(2π)
3

k

_
(2π)
4
δ(ω
k
+ ω
k
−E
p

(3)
(k +k

)[/

[
2
_
.
The first d
3
k

-integral will only return a non-zero value if k

= −k, due to the 3-D delta-
function. Hence, after doing this integral, we have
Γ =
1
2E
p
_
d
3
k
(2π)
3

k
1
(2π)
3

−k
(2π)
4
δ(ω
k
+ ω
−k
−E
p
)[/

[
2
.
86 5 DECAYS
By inspection of ω
k
=
_
k
2
+ m
2
φ
, we see that ω
−k
= ω
k
. Hence, the remaining delta-function
becomes just
δ(2ω
k
−E
p
) = δ(2ω
k
−m
σ
).
Inserting ω
k
just gives
δ(2ω
k
−m
σ
) = δ
_
2
_
k
2
+ m
2
φ
−m
σ
_
.
To make this delta-function useable, we must get it into the form δ(k − k
0
), so that upon
integration, only k = k
0
will give a non-zero contribution. To do this, we use the property
δ(f(x)) =

x
0
δ(x −x
0
)
[f

(x
0
)[
, f(x
0
) = 0.
Hence, given our delta-function, we see that its argument is zero when
[k[ = [k
0
[ =
_
m
2
σ
4
−m
2
φ
.
We also see that
f

([k[) =
2[k[
ω
k
.
Hence, using this,
δ
_
2
_
k
2
+ m
2
φ
−m
σ
_
=
ω
k
0
2[k
0
[
δ ([k[ −[k
0
[) .
Therefore, we can insert this expression for the delta-function in the decay rate, to see that
Γ =
1
2E
p
_
d
3
k
(2π)
3

2
k
1
(2π)
3
(2π)
4
ω
k
0
2[k
0
[
δ ([k[ −[k
0
[) [/

[
2
.
Let us rearrange slightly,
Γ =
1
2E
p
ω
k
0
2[k
0
[
_
d
3
k
(2π)
2

2
k
δ ([k[ −[k
0
[) [/

[
2
.
We now make the integral spherically symmetric, so that
_
d
3
k −→
_
d[k[ [k[
2
4π.
Hence,
Γ =
1
2E
p
ω
k
0
2[k
0
[
_
d[k[ [k[
2

(2π)
2

2
k
δ ([k[ −[k
0
[) [/

[
2
,
where integration will now just filter out [k[ = [k
0
[, so that
Γ =
1
2E
p
ω
k
0
2[k
0
[
[k
0
[
2
π
(2π)
2
ω
2
k
0
[/

[
2
=
1
2E
p
[k
0
[
8πω
k
0
[/

[
2
.
5.3 Example: Decay of a Scalar to two Scalars 87
We can tidy this expression up slightly, by recalling that we are in the frame where the
decaying particle is at rest; so that E
p
= m
σ
. Now, we have that
ω
2
k
0
= k
2
0
+ m
2
φ
,
where we defined
k
2
0
=
m
2
σ
4
−m
2
φ
.
Hence,
ω
k
0
=
m
σ
2
.
Thus,
[k
0
[
8πω
k
0
=
1

2
m
σ
_
m
2
σ
4
−m
2
φ
.
Therefore, we can use this in the decay rate, to see that
Γ =
1
4πm
2
σ
1
2
_
m
2
σ
4
−m
2
φ
[/

[
2
.
Due to the simple form of our /

= −ig, we easily have
[/

[
2
= g
2
.
Hence, we have computed that the complete expression for the decay rate of a scalar σ-particle
to a pair of scalar charged φ-particles is
Γ
_
σ →φ
+
φ

_
=
g
2
8πm
2
σ
_
m
2
σ
4
−m
2
φ
. (5.23)
Let us recall the equivalent expression we derived for a scalar decaying to two fermions,
(5.20),
Γ(σ →τ
+
τ

) =
g
2
πm
2
σ
_
m
2
σ
4
−m
2
τ
_
3/2
.
We can now make a comparison of the two expressions. First, we note that the overall
structure of the decay rate is the same. One of the differences is that the bracketed factor is
raised to a different power. This is a general feature of fermionic decays. Secondly, the scalar
decay rate has a factor 1/8 suppression relative to the fermionic. This is because there are
more fermionic states in the phase space to decay to, than in the scalar phase space (that is,
each fermion has two degrees of freedom).
88 5 DECAYS
89
6 Scattering Processes
6.1 An Outline of Particle Exchange
Let us extend our previous interaction Hamiltonian, to
H
I
= g
τ
¯
ψ
τ
(x)ψ
τ
(x)φ(x) + g
e
¯
ψ
e
(x)ψ
e
(x)φ(x). (6.1)
Hence, we have two types of lepton, τ

and e

, which can scatter by boson exchange in the
process
e

(p
1
, s
1
) + τ

(p
2
, s
2
) −→e

(p
3
, s
3
) + τ

(p
4
, s
4
).
So, the initial state is just
[i) = b

e,s
1
(p
1
)b

τ,s
2
(p
2
) [0) ,
and the final state
[f) = b

e,s
3
(p
3
)b

τ,s
4
(p
4
) [0) .
So that we can describe the process, we obviously need two electron and two tauon field
operators. The lowest order term in the S-matrix expansion that can do this is the second
order,
S
(2)
=
(−i)
2
2
_
d
4
x
_
d
4
x

¸f[ T (H
I
(x)H
I
(x

)) [i) . (6.2)
From this, we need the operators
¯
ψ
(−)
e
(x),
¯
ψ
(−)
τ
(x

), ψ
(+)
e
(x) and ψ
(+)
τ
(x

), which create elec-
trons/tauons and destroy electrons/tauons; notice that we have electron destruction/creation
at a different place to tauon destruction/creation. Hence, the term we want is
2 ¸f[
¯
ψ
(−)
e
(x)
¯
ψ
(−)
τ
(x


(+)
e
(x)ψ
(+)
τ
(x

) [i) ,
where we have multiplied by 2 as the choice of which is x and which x

is arbitrary. Therefore,
the matrix element to be computed is
S
(2)
= (−i)
2
_
d
4
x
_
d
4
x

g
e
g
τ
¸f[
¯
ψ
(−)
e
(x)
¯
ψ
(−)
τ
(x


(+)
e
(x)ψ
(+)
τ
(x

)T (φ(x)φ(x

)) [i) .
(6.3)
The fields act upon the initial and final states in an identical manner to our discussion on
decays, so that for example,
ψ
(+)
e
(x)
¸
¸
e

, p
1
, s
1
_
=
u
e,s
1
(p
1
)e
−ip
1
x
_
2V E(p
1
)
[0) .
90 6 SCATTERING PROCESSES
Then, plugging all of the relevant terms into (6.3), we end up with
o

= (−ig
e
)¯ u
e,s
3
(p
3
)u
e,s
1
(p
1
)(−ig
τ
)¯ u
τ,s
4
(p
4
)u
τ,s
2
(p
2
)

4

i=1
1
(2V E
i
)
1/2
_
d
4
x
_
d
4
x

e
ip
3
x
e
−ip
1
x
e
ip
4
x

e
−ip
2
x

¸0[ T (φ(x)φ(x

)) [0) . (6.4)
p
4
, s
4
p
3
, s
3
p
1
, s
1
p
2
, s
2
τ

τ

e

e

Figure 6.1: Feynman graph for e

τ

→e

τ

scattering.
These terms perhaps become a little more understandable if we consider Figure (6.1). The
spinor factors ¯ u
e,s
3
(p
3
)u
e,s
1
(p
1
) can be thought about as first destroying an electron (i.e.
a particle), with an un-barred spinor on the far right, then an electron created after the
destruction, with a barred spinor. The exponential factors e
ip
3
x
e
−ip
1
x
can be thought about
equivalently. Notice that the exponential factors (and these are the only spatial terms) for
the tauons have primes on them, denoting their interaction happens at a different place to
the electrons interaction. The final term is the Feynman propagator, which as we saw in
Section (4.4), can be written in a variety of ways:
¸0[ T (φ(x)φ(x

)) [0) = iG
F
(x −x

) =
_
d
4
q
(2π)
4
e
−iq(x−x

)
i
q
2
−m
2
H
+ i
.
We now recall that we interpreted ¸0[ T (φ(x)φ(x

)) [0) as creating a particle at x and destroy-
ing at x

. Therefore, this Feynman propagator term describes the exchange of a “virtual”
H
0
. Substituting the propagator into (6.4), we can begin to do integrals. First, integrating
over x only concerns the terms
_
d
4
xe
ip
3
x
e
−ip
1
x
e
−iqx
= (2π)
4
δ
(4)
(p
3
−p
1
−q) ,
6.1 An Outline of Particle Exchange 91
and integrating over x

only
_
d
4
x

e
i(p
4
−p
2
+q)x

= (2π)
4
δ
(4)
(p
4
−p
2
+ q) .
Finally, integrating over q,
_
d
4
q
(2π)
4
_
(2π)
4
δ
(4)
(p
3
−p
1
−q) (2π)
4
δ
(4)
(p
4
−p
2
+ q)
i
q
2
−m
2
H
+ i
_
,
which simply gives
(2π)
4
δ
(4)
(p
3
+ p
4
−p
1
−p
2
)
i
q
2
−m
2
H
+ i
, q = p
3
−p
1
= p
4
−p
2
.
Hence, putting all of this together, we have that (6.4) becomes
o
(2)

=
_
4

i=1
1
(2V E
i
)
1/2
_
(2π)
4
δ
(4)
(p
3
+ p
4
−p
1
−p
2
) /

, (6.5)
where we have defined the Feynman amplitude to be
/

= (−ig
e
)¯ u
e,s
3
(p
3
)u
e,s
1
(p
1
)
i
q
2
−m
2
H
+ i
(−ig
τ
)¯ u
τ,s
4
(p
4
)u
τ,s
2
(p
2
).
(6.6)
Notice that (6.5) has an identical structure to the matrix element we derived for decays; nor-
malisation times energy-momentum conservation times the dynamic parts. We can represent
the amplitude (6.6) by a Feynman diagram, as in Figure (6.2).
We can note that the factors in Figure (6.2) are very similar to those we found for the
decay, except we now have a factor for the virtual exchange boson. We should also note that
at each vertex, charge and lepton numbers are conserved (where lepton numbers stay within
their own species); Figure (6.3) has some examples.
This conservation stems from the
¯
ψ
e
(x)ψ
e
(x) structured-terms in the Hamiltonian; there
are no terms like
¯
ψ
τ
(x)ψ
e
(x) (which would create a τ

and destroy a e

).
We should also note that 4-momentum is conserved at each vertex (there are two vertices
in Figure (6.2)), as q = p
3
−p
1
= p
2
−p
4
, something which came about due to delta-functions
in the integral. Finally, we should note that the Feynman amplitude represents two time-
orderings; which comes from the Feynman propagator allowing for x or x

to come first (i.e.
there is no reason which to choose to be first). Figure (6.4) has the two equivalent diagrams
(the first two), which we understand to be present when we conventionally draw the third.
This highlights the fact that is dosent matter how the diagrams are drawn; and only the
connections (i.e. all initial states on the left, and final on the right), and the number of
connections. It is interesting to note that if we had derived the Feynman rules without the
time ordering, we would get a single diagram out, which would not be Lorentz invariant (the
time ordered double-diagram is Lorentz invariant).
92 6 SCATTERING PROCESSES
i
q
2
−m
2
H
+i
¯ u
e,s
3
(p
3
)
(−ig
e
)
(−ig
τ
)
u
e,s
1
(p
1
)
¯ u
τ,s
4
(p
4
) u
τ,s
2
(p
2
)
q
e

, p
3
τ

, p
4
e

, p
1
τ

, p
2
Figure 6.2: Feynman graph for e

τ

→e

τ

scattering, with Feynman rules.
e

e

τ

τ

(a) Valid diagrams.
e

e
+
e

τ

(b) Invalid diagrams.
Figure 6.3: Examples of charge and lepton conserving diagrams, as well as invalid diagrams.
6.2 Other Scattering Processes
We have discussed non-identical particle scatterings. We could discuss identical particle
scatterings as well, such as
τ

(p
1
) + τ

(p
2
) −→τ

(p
3
) + τ

(p
4
).
There are now two possibilities. We can now have that when the particles are created, a τ

can be created with either momentum p
3
or p
4
.
6.2 Other Scattering Processes 93
≡ ≡
x

x x
x

x

x
Figure 6.4: The equivalent diagrams represented by the time ordering of the Feynman propagator.
1
2
3
4
2
1
3
4
Figure 6.5: The equivalent diagrams represented by the time ordering of the Feynman propagator
for scattering of identical particles. The diagram on the left is the direct diagram, on the right the
exchange term.
If we do the calculation and keep track of fermion field order, we would find a relative
minus sign from the anti-commutation relations (which reflects the Pauli principle). The full
invariant amplitude is
/

= /

direct
+/

exchange
,
where
/

direct
(p
1
p
2
p
3
p
4
) = (−ig)¯ u(p
3
)u(p
1
)
i
q
2
−m
2
H
+ i
(−ig)¯ u(p
4
)u(p
2
),
as before, and
/

exchange
(p
1
p
2
p
3
p
4
) = −/

direct
(p
1
p
2
p
4
p
3
).
The minus sign is simply because we are exchanging identical fermions. If we dealt with
bosons, one would have the exact same thing, but without the minus sign (as bosonic field
commute).
We could also have particle anti-particle scattering,
τ

+ τ
+
−→τ

+ τ
+
,
where, as in Figure (6.6), everything is conserved.
94 6 SCATTERING PROCESSES
u ¯ u
¯ v v
Figure 6.6: The two diagrams corresponding to particle anti-particle scattering.
6.3 Scattering Cross-sections
We now outline how to compute quantities we can actually observe. Let us consider elastic
scattering, where the same particles come out as go in, but they may have different momenta
and spin; see Figure (6.7).
p

1
p

2
p
2
p
1
Figure 6.7: Schematic of elastic scattering.
The cross-section σ is defined as the transition rate into a given set of final states per unit
flux of initial particles. Hence, in terms of dimensions,
[σ] =
T
−1
T
−1
L
−2
= L
2
.
Hence, the cross-section has the units of an area.
For decays, we have that the transition rate is defined as
transition rate = w
fi
=
[o

[
2
T
, (6.7)
for some unique final state f.
Now, the flux for co-linear collisions (i.e. where the colliding particles hit each other head
on, with 180

between them) is
flux ≡ f =
1
V
[v
1
−v
2
[,
6.3 Scattering Cross-sections 95
where V is some volume, and v
i
are the velocities of the colliding particles. This is simply
the statement that flux is the density times relative speed.
Let us consider an element of the final phase space,
V
(2π)
3
d
3
p

1
V
(2π)
3
d
3
p

2
,
so that the corresponding elemental cross-section is
dσ =
1
flux
transition rate element size (6.8)
=
1
f
[o

[
2
T
V
(2π)
3
d
3
p

1
V
(2π)
3
d
3
p

2
.
We know o

, by adapting (6.5) to be in the finite volume/time limit (which we do such that
we can use δ
2
ij
= δ
ij
, rather than having to compute the square of the delta-function),
o

=
2

i=1
_
1
(2V E
i
)
1/2
1
(2V E

i
)
1/2
_
TV δ
p
1
+p
2
,p

1
+p

2
/

.
Hence, we write the elemental cross-section
dσ =
V
[v
1
−v
2
[
1
T
2

i=1
_
1
2V E
i
1
2V E

i
_
T
2
V
2
δ
p
1
+p
2
,p

1
+p

2
[/

[
2
V
(2π)
3
d
3
p

1
V
(2π)
3
d
3
p

2
=
V
[v
1
−v
2
[
1
T
1
2V E
1
1
2V E

1
1
2V E
2
1
2V E

2
T
2
V
2
δ
p
1
+p
2
,p

1
+p

2
[/

[
2
V
(2π)
3
d
3
p

1
V
(2π)
3
d
3
p

2
, (6.9)
which we clean up using
dQ ≡ TV δ
p
1
+p
2
,p

1
+p

2
d
3
p

1
(2π)
3
2E

1
d
3
p

2
(2π)
3
2E

2
= (2π)
4
δ
(4)
(p
1
+ p
2
−p

1
−p

2
)
d
3
p

1
(2π)
3
2E

1
d
3
p

2
(2π)
3
2E

2
, (6.10)
where we have now taken the infinite box-size limit. Hence, using this definition, (6.9)
becomes
dσ =
1
4E
1
E
2
[v
1
−v
2
[
[/

[
2
dQ. (6.11)
96 6 SCATTERING PROCESSES
We call (6.10) the invariant phase space factor. Now, upon comparison of this with our
prototype (6.8), we see that we can write a “modified” flux factor,
F ≡ 4E
1
E
2
[v
1
−v
2
[. (6.12)
This can also be written as
F = 4
_
(p
µ
1
p

)
2
−m
2
1
m
2
2
.
The flux factor F and invariant phase space dQ are both Lorentz invariant, as is /

. Hence,
dσ (6.11) is Lorentz invariant under co-linear transformations. Notice that dQ (6.10) has only
two independent scalar variables. The six from having two d
3
p’s (each has three variables)
are constrained by the delta-function, eliminating four of them. Thus, the remaining two are
usually chosen to be the polar angle θ, φ relative to the beam direction of particle 1 (say).
One must note that the angles are frame dependent.
6.3.1 Centre of Momentum Frame
Let us now go into the centre of momentum frame, where p
1
+p
2
= 0 and thus p

1
+p

2
= 0.
See Figure (6.8).
θ, φ
p

1
p
2
= −p
1
p

2
= −p

1
p
1
Figure 6.8: Colinear scattering in the centre of momentum frame.
Energy conservation requires that [p
1
[ = [p

1
[, as we have the same particles come out that
go in. The flux factor (6.12) can be written as
F = 4E
1
E
2
_
[p
1
[
E
1
+
[p
2
[
E
2
_
= 4(E
1
+ E
2
)[p
1
[
We can rewrite the phase space factor (6.10) by integrating over redundant variables. Notice
that we take take out the zeroth-component of the delta-function,
δ(E

1
+ E

2
−E
1
−E
2
) = δ
_
_
m
2
1
+[p

1
[
2
+
_
m
2
2
+[p

2
[
2
−E
1
−E
2
_
= δ(f([p[)).
6.4 Scattering From a Classical Field 97
Then, we rewrite this delta-function using the usual method for the argument of the function,
δ(E

1
+ E

2
−E
1
−E
2
) =
E
1
E
2
(E
1
+ E
2
)[p
1
[
δ([p

1
[ −[p
1
[).
Hence, doing this, we find
dQ =
1
(2π)
2
[p
1
[
4(E
1
+ E
2
)
dΩ
cm
,
so that the differential cross-section is

dΩ
cm
=
[/

[
2
64π
2
(E
1
+ E
2
)
2
. (6.13)
To go further, we have to evaluate [/

[
2
, which will depend upon spins and momentum. If
we “dont care about” spins (i.e. polarisation), then we average over initial spins and sum
over final spins – this is done if a detector does not pick up different polarisations. Finally,
if we want the total cross-section we integrate over the solid angles θ, φ.
6.4 Scattering From a Classical Field
Here we shall consider scattering from a static classical field – classical means that the field
is so “big” that one can ignore quantum fluctuations. This could be the case, for example, if
electrons are scattered from a large magnetic field, or the Coulomb field of a heavy nucleus.
Let us recall the previous interaction Hamiltonian, for the fermionic fields scattering from
a bosonic field,
H
I
= gN
_
¯
ψ(x)ψ(x)φ(x)
_
.
Suppose we take our scalar field φ(x) to now be static, and let us denote it by V (x). Then,
the first order S-matrix element for scattering from [i) to [f) is
o

= −ig
_
d
4
x ¸f[ N
_
¯
ψ(x)ψ(x)
_
[i) V (x). (6.14)
Now, suppose we have the initial and final states being
[i) = [p, s) , [f) = [p

, s

) ,
then, the only terms in the field expansion of (6.14) to contribute are
o

= −ig
_
d
4
x ¸f[
¯
ψ
(−)
(x)ψ
(+)
(x) [i) V (x).
Now, we know that
ψ
(+)
(x) [p, s) =
e
−ipx
u
s
(p)

2EV
[0) ,
¸p

, s

[
¯
ψ
(−)
(x) = ¸0[
e
ip

x
¯ u
s
(p

)

2EV
.
98 6 SCATTERING PROCESSES
Hence, we use these in the matrix element to easily obtain
o

= −ig
_
d
4
x V (x)¯ u
s
(p

)u
s
(p)
e
ix(p

−p)
2V

EE

. (6.15)
The dx
0
integral is easy to compute, as the only term to contribute is the exponential,
_
dx
0
e
x
0
(E

−E)
= Tδ
E,E
= (2π)δ(E −E

).
The only terms in the dx
i
integral are the potential and exponential, which form the integral
V (q) ≡
_
d
3
x V (x)e
−ix·(p

−p)
, (6.16)
where we have introduced the Fourier transform of the scattering potential, V (q), where
q ≡ p

−p. So, using these, (6.15) becomes
o

= −igV (q)¯ u
s
(p

)u
s
(p)Tδ
E,E

1
2V

EE

.
Now, the transition rate is just
w
fi
=
[o

[
2
T
=
g
2
4V
2
EE

T
2
δ
2
E,E

T
[¯ u
s
(p

)u
s
(p)[
2
[V (q)[
2
. (6.17)
We now take advantage of

2
E,E
= Tδ
E,E
= (2π)δ(E −E

),
so that (6.17) becomes
w
fi
= [/[
2

δ(E −E

)
V
2
,
where
[/[
2

g
2
4EE

[¯ u
s
(p

)u
s
(p)[
2
[V (q)[
2
.
To calculate the cross-section, we need both the flux and density of states. The flux is just
f = ρv =
1
V
v,
where v is the difference in velocities. The density of (final) states is simply
V d
3
p

(2π)
3
.
6.4 Scattering From a Classical Field 99
Hence, following the same structure as (6.8), the elemental cross-section is
dσ =
V
v
[/[
2

δ(E −E

)
V
2
V d
3
p

(2π)
3
=
1
v
[/[
2
2πδ(E −E

)
d
3
p

(2π)
3
. (6.18)
Now, let us use a trick to convert d
3
p

into an integral over energy. We can easily recall that
E
2
= [p[
2
+ m
2
,
and hence
dE
dp
=
p
E
,
where p ≡ [p[ (i.e. not the 4-vector). So,
d
3
p

= p
2
dp

dΩ

= p

E

dE

dΩ

.
Hence, inserting this into the elemental cross-section (6.18), we see that
dσ =
1
v
[/[
2
2πδ(E −E

)
1
(2π)
3
p

E

dE

dΩ

.
If we integrate over E

now, we simply filter out E = E

and p = p

, so that the differential
cross-section is

dΩ
=
Ep
(2π)
2
v
[/[
2
, (6.19)
where
[/[
2

g
2
4EE

[¯ u
s
(p

)u
s
(p)[
2
[V (q)[
2
. (6.20)
Basically, one of the points of this, is that the differential cross-section is proportional to the
modulus-squared of the Fourier transform of the scattering potential – this is very similar to
the Born approximation in non-relativistic quantum mechanics.
If we consider the low energy limit, then the spinors reduce
u
s
(p →0) =

2m
_
χ
s
0
_
,
so that
¯ u
s
(p

)u
s
(p) = 2mδ
ss
.
Thus, in the low energy limit, the spins of the outgoing particles is the same as that of
the incoming particles. That is, spin is conserved; in which case the result reduces to the
non-relativistic Born approximation obtained from the Schrodinger equation.
We can also get the same result from our previous completely quantised calculation of e

τ

scattering by one-particle exchange; where we make V (q) proportional to the propagator.
That is, the scattering potential is proportional to the Fourier transform of the propagator
of the exchanged particle.
100 6 SCATTERING PROCESSES
101
7 Feynman Rules
The Feynman rules can be used to construct the invariant Feynman amplitude /

for a
given process. These only apply up to second order perturbation theory, and do not include
any QED, weak or QCD interactions. When drawing Feynman graphs, we use the convention
that the arrow only denotes whether the “particle” described is a particle or anti-particle.
We also use the convention that a solid line denotes a fermion, and a dashed line a boson.
All factors are pretty self-explanatory; but it is worth noting that m denotes the mass of
the exchange boson, and q the 4-momentum carried by the exchange boson. One write a
vertex factor for every vertex in the diagram.
102 7 FEYNMAN RULES
outgoing ¯ u
s
(k)
vertex factor
−ig
fermion
fermion
incoming u
s
(k)
anti-fermion
incoming ¯ v
s
(k)
outgoing v
s
(k)
anti-fermion
exchange boson
i
q
2
−m
2
+i
Figure 7.1: Summary of the Feynman rules we have derived.

ii

CONTENTS

iii

Contents
1 Preliminaries 1.1 1.2 1.3 1.4 1.5 Relativistic Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Dirac δ-function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Homogeneous Lorentz Transformations . . . . . . . . . . . . . . . . . . . . . Equation of Continuity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.4.1 1.5.1 1.5.2 1.6 1.7 Electromagnetic Fields . . . . . . . . . . . . . . . . . . . . . . . . . . Interpretation of the Wavefunction . . . . . . . . . . . . . . . . . . . Minimal Electromagnetic Interactions . . . . . . . . . . . . . . . . . . Quantum Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1 2 4 6 7 8 8 9 11 11 13 13 14 15 18 21 22 24 25 27 31 32 32 34 36

Relativistic Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Natural Units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2 The Klein-Gordon Equation 2.1 2.2 2.3 Negative Energies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conserved Current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . EM Fields and the “Hydrogen Atom” . . . . . . . . . . . . . . . . . . . . . . 2.3.1 The Klein Paradox . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3 The Dirac Equation 3.0.2 3.1 3.2 3.3 3.4 3.5 Properties of αi , β & Dirac Representation . . . . . . . . . . . . . . . Conserved Current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Angular Momentum & Spin . . . . . . . . . . . . . . . . . . . . . . . . . . . Plane Wave States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dirac Hole Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Relativistic Covariance of The Dirac Equation . . . . . . . . . . . . . . . . . 3.5.1 3.5.2 3.5.3 Covariant Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . Proof of Covariance . . . . . . . . . . . . . . . . . . . . . . . . . . . . Lorentz Boost . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . Hydrogen-like Atoms . . . . . .5 The Feynman Propagator GF (x) . . . 36 38 38 41 42 44 47 47 50 52 53 57 58 58 61 66 66 68 69 73 73 76 82 89 89 92 94 Interaction With Fields . . . . . . . . . . . . . . . . . . . . . . . . The Klein-Gordon Field . . . . . . Interactions . .1. . . . . . . . .1 4. . . . . . .2 CONTENTS Parity . . . . . . . . . . . . . .5. . . . . . 4. . . . . . . . . . . . . . . . . . . . . . . . . Scattering Cross-sections . . . . . . . . . . . .1 5. . . .3. . . . . . . . . . . . . . . . . . . . . .7 3. . .iv 3. . . . . . . . . . . . . . . . . . . . . . . . The S-matrix Expansion . . The S-matrix . . . . . . . . . . . . Particles in a Spherical Potential . . . Field Operators .1 3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4 4. . . . . . Example: Decay of a Scalar to two Scalars .1 4. . . . . . . . . .2 4. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1. . . . . . . The MIT Bag Model . 4. . . . . . . . . . . . . . . . . . . . . . . . . .7. . . . . Decay Rates & Lifetimes . . . . . . . . . . . . . . . . . . .6 3. . . . . . . . . . . . . . . . . . .5. . . . . . . 5 Decays 5. . . . . . . . .2 5. . . . . . . . . . . . . . . .5. . . . . . . . . . .5. . . . . . . . . . . . . . .2 4. . . . . .2 Fermions . . . . . . . . . . . . . . . . . . Other Scattering Processes .3. .1 6. . . . . . . .3 An Outline of Particle Exchange . . . . . . . . . . .7. . . . .4 3. . . . . . . 4. . . . . . . . . . . . . . . . . . . . .1 3. . . . . . . . . . . . . . Magnetic Moment of Electrons . . . . . . . . . . . . .2 4. . . . . .6. 6 Scattering Processes 6. . . . . . . . . . . . . . . . . .1 Quantum Mechanics of a String . .3 Phonons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 6. . . . . . . . . . . . . . . . . . . . . . . . . . .1 4. . . . . . . . . . . . . . . . 4 Quantum Fields 4. . . . . . . . . 4. . . . . . . . . . . .3 Simple 2-body Decay: φ3 . . . . . . . . . . . . . . . . . . . . . . . The Complex Dirac Field ψ(x) . . . . . . . . . . .3 The Interaction Picture . . . . . . . . The Dirac Field . . . . . . . . . . . . . . . .

v 96 97 101 Scattering From a Classical Field . . . . . . . .4 Centre of Momentum Frame . . . . . . . . . . . . . . . .1 6. . . . . .3. . . . . . . . . . . . 7 Feynman Rules . .CONTENTS 6. . . . . . .

vi CONTENTS .

the spin-statistics theorem etc. x). The latter needs the full formalism of quantum field theory. T > mc2 .1 Non-relativistic quantum mechanics is concerned with systems for whom the kinetic energy is a lot less than the rest mass.1 Relativistic Notation Let us introduce the contravariant 4-vector xµ = (ct. T = mv 2 . −x). The former case does not have enough energy to create particles. if a system starts out with one particle. but not particle creation. we are able to understand electron spin. no particles are created. such that gµν = diag(1. In discussing a relativistic quantum theory. That is. Using this. and everything moves slowly. predict the existence of anti-particles. −1. the latter is able to produce particles. 2 2m Under this. −1. where all off-diagonal components are zero. we can define a covariant 4-vector. so that 1 p2 T ≈ E − mc2 = mv 2 = . z) = (ct. 1. . −1). relativistic quantum mechanics is concerned with systems for whom the energy is more than the rest mass. Let us also introduce the metric tensor. The former case requires relativity. Now. Following this. 1 Preliminaries Let us introduce the notation we shall be adopting. so that xµ = (ct. the electrons magnetic moment. x. it will always have one particle. y. xµ = gµν xν . one can make two distinctions: T < mc2 .

−∞ (1. only the 3-spatial components. Only when x = x does the δ-function return a non-zero value. For the δ-function to be used directly in this way. If instead we have a functional argument.2 The Dirac δ-function As we shall use the δ-function regularly.2 1 PRELIMINARIES We shall use the Einstein convention of implied summation over repeated indices. Hence. we see that we can use the metric to lower raised indices. Therefore. −∞ (1. and φ(x) ≡ φ(t.1) That is.2) . Thus. We will commonly write expressions such as px ≡ pµ xµ = Et − p · x. as a simple case. we see that they have the same components. Thinking about the δ-function as a filtering mechanism can be very useful. Suppose we have. by writing x. one can think about the δ-function as only returning a value when its argument is zero. By the form of our metric and inverse. x). and by writing x. all four components). such as δ (g(x)). let us work out how to evaluate ∞ dx f (x)δ(a(x − x )). In the context of the integral above.e. we can see σ g σµ xµ = g σµ gµν xν = δν xν = xσ . making the variable x take all possible values. and the inverse metric to raise lower indices. its argument must be of the form (x − x ). so that f (x ) is returned. δ(a(x − x )). the integral “sweeps” over the real line. then we must employ a certain tactic to get it into a useable form. One of its definitions is that ∞ dx f (x)δ(x − x ) = f (x ). where x is the integration variable and x some constant. the only contribution from the integral is at x = x . which is xσ = g σµ xµ . as λ g λµ gµν = δν . where a is a constant. it is worth our noting a few of its properties. g µν = gµν . 1. We further define the inverse metric. using these results. Hence. we mean the 4-vector (i. Then.

let us make a Taylor expansion of g(x) about the zero. xi . as g (xi ) is “just a number”. we have a (very useful) relation which allows us to recast a δ-function with functional argument. Hence. We may have a multi-dimensional δ-function δ (n) (x − x ). upon substitution into (1. g(xi + ) = g(xi ) + (x − xi )g (xi ) + . say. of the limits). which we define n δ (n) (x − x ) = i=1 δ(xi − xi ). δ(−x) = δ(x). . then. we must find the zeros of the function g(x). . Supposing that we now have the zeros.1). δ(g(x)) = δ (g (xi )(x − xi )) . g(x) ≈ (x − x )g (xi ). An immediate consequence of this is that the δ-function is even. . we see that ∞×a −∞×a dy y f ( )δ(y − y ). Therefore. a a Because we always want this form of the integral (i. Hence.2 The Dirac δ-function Let us first change variables. |g (xi )| g(xi ) = 0. to 3 y = ax. Now. we can treat it exactly as our a before. we must sum over all zeros. Hence. |a| a where we have just used the defining property of the δ-function (1. by definition of the zero. so that g(xi ) = 0. δ(g(x)). . so that δ(g(x)) = δ (g (xi )(x − xi )) = If there are multiple zeros of the function g(x).e.2). hence. we have 1 |a| ∞ −∞ y dyf ( a )δ(y − y ) = 1 y f ( ). (1.3) Therefore. Let us then consider a full functional argument. y = ax . at. the δ-function only returns a non-zero value when its argument is zero (before this was just at x = x ). δ(g(x)) = x0 1 δ(x − xi ). The first term is zero. we must take the modulus of a.1. |g (xi )| Now. 1 δ(x − xi ). into a useable form.

it is simple to see that Λ µ ν Λ µ σ xν xσ = xµ xµ . s 2 = xµ xµ = x µ xµ .4 So. by definition. That is. (2π)4 Of course. something unchanged by a LT. where Λµ ν ∈ R and depend on the relative velocity and relative orientation of the axes. these properties will prove invaluable when discussing quantum fields. . from which we read off that we require σ Λ µ ν Λ µ σ = δν . is. for example. The interval s2 is an example of a Lorentz scalar – that is. We can also define the δ-function to be the Fourier transform of unity: δ (4) (x − x ) = d4 k −ik(x−x ) e . 1 PRELIMINARIES δ (4) (x − x ) = δ(x0 − x 0 )δ(x1 − x 1 )δ(x2 − x 2 )δ(x3 − x 3 ) = δ(t − t )δ (3) (x − x ). That is.3 Homogeneous Lorentz Transformations Such transformations (between frames that have coinciding origins) are of the general form xµ → x µ = Λµ ν xν . the inverse Fourier transform is then δ (4) (k − k ) = d4 xeix(k−k ) . where Λµ ν = gµσ g ντ Λσ τ . if it transforms like Aµ → A µ = Λµ ν Aν . At this stage we have merely given a few properties of an abstract function. Lorentz transformations (LTs) leave the interval s2 = c2 t2 − x2 invariant. a contravariant 4-vector. Using these relations. A four component object Aµ which transforms like xµ under an LT. we see that xµ → xµ = Λµ ν xν . 1. Using previous relations. Now.

φ c ∂t Then. The scalar product of any two 4-vectors can be written in a variety of ways: AB = Aµ Bµ = A0 B0 − A · B. Hence. what about derivatives? Consider a Lorentz scalar function. the energy-momentum in the 4-momentum pµ = (E/c. we must have ∂φ that ∂xµ is a covariant 4-vector. Hence. and also AB = Aµ Bµ = g µν Aµ Bν = gµν Aµ Aν = Aµ B µ . those equations for whom indices balance automatically satisfies the principles of special relativity. ∂xµ So.3 Homogeneous Lorentz Transformations Similarly. in order that the LHS is a scalar. . For example.1. So. 2 c where m2 c2 is the thing we will call the Lorentz invariant. That is. p) is a 4-vector. the term on the LHS is a Lorentz scalar (as we defined it to be so). The scalar product with itself is just p 2 = pµ pµ = E2 − p2 = m2 c2 . 5 The scalar product is a Lorentz invariant. ∂φ µ δx . whereby its value at a point in one coordinate system is the same at the same point in another coordinate system. Also. one can define a covariant 4-vector if it transforms as Aµ → Aµ = Λµ ν Aν . φ (x ) = φ(x). δφ = ∂µ φ ≡ ∂φ = ∂xµ 1 ∂φ . the term on the far RHS is a contravariant 4-vector. All equations written consistently in terms of 4-vectors and scalars are automatically Lorentz invariant. since one can easily show that A µ Bµ = Aµ Bµ . p 2 = p µ p µ = m 2 c2 is a Lorentz invariant. So.

4). provided ∂µ j µ (x) = 0 is satisfied.6) dV ρ(x. bounded by the surface S.− c ∂t2 2 1 PRELIMINARIES ∂φ = ∂xµ 1 ∂φ . Similarly. dQ ∂ = dt ∂t = − Ω (1. . rewrite (1. The first step follows by (1. 1. mass or electric charge. the charge Q is conserved.6) and the second by the divergence theorem. Then. if φ is a Lorentz scalar. note that as (1. which is a Lorentz scalar operator. (1. j). by (1. It thus transforms as ∂µ −→ ∂µ = Λµ ν ∂ν . we define the contravariant differential 4-vector ∂ µφ ≡ We further define the D’Alembertian. where j µ (x) = (cρ. t) Ω dV ·j = − S j · dS. ≡ ∂µ ∂ µ = 1 ∂2 .5) is a Lorentz invariant equation.5) To see this continuity equation. t). then j µ (x) is a contravariant 4-vector.e. This could be. Now.6 is a covariant vector. · j = 0. φ = ∂µ ∂ φ = 0 µ ⇒ 1 ∂ 2φ = c2 ∂t2 2 φ. t) (1. c ∂t .4 Equation of Continuity Suppose we have some charge density ρ(x. is a Lorentz scalar).4) is unchanged under a Lorentz transformation (i. such that QΩ = Ω dV ρ(x. We can obtain the Lorentz covariant form of the wave equation.− φ . ∂ρ + ∂t Then. where Ω is some angular region of space.5) using the definition of ∂µ . for example.

1 1 ∂E 1 ∂B . Maxwell’s equations. ×E=− c ∂t c c ∂t We introduce the electromagnetic potentials φ and A. Notice that upon subsitution of these expressions into each of the Maxwell equations. 1 ε0 = 1 ⇒ µ 0 = 2 . c ∂t2 c ∂t c which may be written in the Lorentz covariant (and indeed. BSI B≡ . so that the fields are related to the potentials via 1 ∂A E = − φ− . c Now. under these rationalised SI units. ×B= j+ . the curl of grad is zero). read 1 · B = 0. c µ where the 4-vector potential A is Aµ (x) = (φ. so that the relative permittivities and permeabilities are εr = µr = 1. · E = ρ.4. c Hence. more succinct) form 1 Aµ (x) − ∂ µ ∂ν Aν (x) = j µ . Now. c ∂t B = × A. µ0 = 4π × 10−7 NA−2 . the four Maxwell equations are · E = ρ. ∂t ∂t We note the standard relation 1 µ0 ε0 = 2 . the equations in SI units. . one finds 1∂ 2 φ+ · A = −ρ. Instead. Note that upon substitution of these equations. A). c ∂t 1 ∂ 2A 1 ∂φ 1 2 A− − ·A+ = − j. c and we redefine the magnetic field. into Gauss’ and Amperes’ law (i. we use the rationalised Gaussian units. are not used practically. in SI units.1 Electromagnetic Fields 7 We shall take space to be “free”.e the first and fourth of Maxwell’s equations). · B = 0.4 Equation of Continuity 1. × B = µ0 j + µ0 ε0 . ε0 ∂B ∂E ×E=− .1. they are satisfied (for example.

Now. t) = |ψ(x. t)d3 x = |ψ(x. t)|2 d3 x = 0. multiply it by ψ ∗ .9) can be trivially written − 2 2m · (ψ ∗ ψ − ψ ψ ∗ ) . t)d3 x = |ψ(x. Consider the Schrodinger equation. ρ(x.8) Now.1 Quantum Mechanics Interpretation of the Wavefunction ρ(x.5. we can get to the equation of continuity. (1.9) And the RHS of (1. at position x. (1.8 1 PRELIMINARIES 1. and at time t. to give i The LHS of this is just i ψ∗ ∂ψ ∂ψ ∗ +ψ ∂t ∂t =i ∂ ∂ ∗ |ψ|2 = i ψ ψ. (1. with a little work. ∂ ∂t ρ(x. t)|2 d3 x = 1. and subtract from it the conjugated equation (1. i and its Hermitian conjugate. Now. . t)|2 ≥ 0.8) multiplied by ψ. t)d3 x = ∂ ∂t |ψ(x.5 1.7) ψ ∗ + V (x)ψ ∗ . take the Schrodinger equation (1. Notice that this last statement means that probabilities are conserved. −i 2 ∂ψ ∗ =− ∂t 2m 2 2 ∂ψ =− ∂t 2m 2 ψ + V (x)ψ. this requires that probabilities are positive. and that the wavefunction is normalised. ρ(x. as the derivative of unity is zero. where integrals are taken over all space. t)|2 d3 x We tend to define that is the probability of finding a particle in volume d3 x.7). ∂t ∂t ψ∗ ∂ψ ∂ψ ∗ +ψ ∂t ∂t =− 2 2m ψ∗ 2 ψ−ψ 2 ψ∗ .

we have that i ∂ |ψ|2 = ∂t 2m ∂ρ + ∂t where ρ(x. · j = 0.14) The second of (1. t)|2 . 1.5. charge q.1. bringing this together.5 Quantum Mechanics Hence. for a point particle of mass m. let us consider xi = ˙ So. (1. That is. t) = |ψ(x. we see that the continuity equation is satisfied in the Schrodinger equation. ∂xi (1. the first gives xi = ˙ which easily rearranges into or. pi = − ˙ ∂H . j= i (ψ ψ ∗ − ψ ∗ ψ) . is given by H= 1 q p− A 2m c ∂H . ∂pi 2 + qφ.13) 1 q pi − Ai . Hence. (1. but we shall have to recheck this when we come to discuss relativistic quantum mechanics. m c q pi = mxi + Ai .10) (1.11) This is the Born interpretation of quantum mechanics.13) provides us with m¨i = mxi + qEi . x ˙ or generally with q m¨ = qE + v × B.2 Minimal Electromagnetic Interactions The classical Hamiltonian. this is of the form of a continuity equation. 2m · (ψ ∗ ψ − ψ ψ ∗ ) . ˙ c q ˙ p = mx + A. x c . in an EM field. c (1. that the modulus squared of the wavefunction is a probability density. 9 Hence.12) Let us compute Hamilton’s equations from this.

12) into the momentum operator p = −i . (1. . this is related back to the “free equation” (i. (1. x) = mx2 + qφ.15) − iq A. notice that it is the “usual” expression of momentum. with an additive term due to the magnetic vector potential. c ∂t .15) is called minimal substitution. the Schrodinger equation with no EM fields) by the substitution of space and time derivatives. ∂t 2m iq − A c 2 ψ=i ∂ iq + φ ψ.14) into (1. We call (1. Notice that the total energy is independent of the magnetic vector potential A. This substitution. so that the Schrodinger equation reads (using the Hamiltonian). −→ Indeed.14) the conjugate momentum. Hψ = i ∂ψ .e. ∂t and use the “correspondence principle” (also known as first quantisation. c ∂ ∂ iq −→ + φ.10 Notice that if we insert (1. ∂t Now. Let us take the Schrodinger equation. this may be written as ∂µ −→ ∂µ + where Aµ = (φ. ∂µ = iq Aµ . x) = E(x. 1 −i 2m which rearranges fairly simply into − 2 q − A c 2 ψ + qφψ = i ∂ψ . −A). this is a consequence of the magnetic field being perpendicular to motion. ˙ ˙ ˙ 2 1 PRELIMINARIES which is just the total energy of the system. or canonical quanˆ tisation) to turn the momentum vector p in (1. ∂t ∂t 1∂ .12). then we have 1 H(x.

1. 1. Now.582 × 10−22 MeV s. c = 1. by the minimal substitution.973 × 10−13 MeV m. iq − A c 2 + m2 c4 + qφ. To get back to practical units. whereby = 1.6 Relativistic Case By a similar argument as above. we use natural units. . and then use = 6. c = 1. the free particle Schrodinger equation reads √ − 2 c2 2 + m 2 c4 ψ = i ∂ψ . H= − 2 c2 (cp − qA)2 + m2 c4 + qφ. ∂t (1. the relativistic Hamiltonian is H= or. we use dimensions to restore ’s and c’s.16) How should we interpret this? This is the main thrust of all we look at. the problem is: how do we interpret this square root? For example.6 Relativistic Case 11 1.7 Natural Units From now on.

12 1 PRELIMINARIES .

we find the relation E 2 = p2 + m 2 . we get H 2 φ = Hi = i ∂φ ∂t ∂ Hφ ∂t ∂ ∂φ = i i ∂t ∂t ∂ 2φ = − 2. the relativistic Schrodinger equation (1. if we act H upon (2. ≡ ∂µ ∂ µ . what does it mean? What is the interpretation of φ? 2. t) = 0. A) = 0. we can avoid this square-root.1). Now. and in the natural units previously discussed. t) = ei(p·x−Et) into the KG equation. t) . ∂t 2 Let us work in free space. So. ∂2 − ∂t2 2 2 + m2 φ = − ∂ 2φ .16) reads Hφ(x.2) + m2 φ(x. (2. by noting that the Hamiltonian is independent of time t. this is just + m2 φ(x. Using the D’Alembertian notation. t) = 0. But.1) where the Hamiltonian is H= √ − + m2 . H 2φ = − Hence. so that (φ. This KG equation is mathematically correct. ∂t Therefore. ∂t2 (2.13 2 The Klein-Gordon Equation So. This is the Klein-Gordon (KG) equation. and is indeed Lorentz invariant (for a scalar φ).1 Negative Energies Substituting a plane wave solution φ(x. t) = i ∂φ(x. .

∂ 2 φ∗ 2 = − m2 φ∗ . ρ = i φ ∂t ∂t j = −i [φ∗ φ − φ φ∗ ] . Let us write the KG equation (2. ρ associated with the KG equation. let us carry on. regardless of these issues. i ∂ ∂t φ∗ ∂φ ∂φ∗ −φ ∂t ∂t −i · (φ∗ φ − φ φ∗ ) = 0. and bringing the LHS over to the RHS. and interpreting the “wavefunction” φ. we have found that the KG equation is a covariant wave equation. multiplying it by φ∗ . and subtracting from it (2. j) such that ∂µ j µ = 0 is satisfied.4) multiplied by φ. In our case. In classical physics. 2. But.e. ∂t ∂t ∂t Multiplying through by i.e.5) can be negative as well as positive (thus negating its ability to be a probability). to summarise. (2. For the moment. In quantum mechanics. interactions allow quantum jumps over this gap. gives ∂φ∗ ∂ ∂φ −φ φ∗ = · (φ∗ φ − φ φ∗ ) . if we compare these with those found in the non-relativistic case.2 Conserved Current Let us compute the conserved current and charge j.5) (2. and above the negative of the rest mass). So. releasing quanta Eγ > mc2 . we can compare this with the continuity equation (1.2) as ∂ 2φ = ∂t2 also taking its complex conjugate. There is an energy gap between mc2 < E < mc2 (i. ∗ ∂φ (2. . allowed energies of the system) is discontinuous. (2. we have problems with negative energies.11).10) to identify the charge and current. the spectrum (i. and with it is a conserved charge j µ = (ρ.14 so that the allowed energies are E = + p2 + m2 > m. (2. we see that φ cannot (generally) be interpreted as a probability amplitude (as we did for the non-relativistic ψ).3). as radiation or other particles. such transitions carry on to give an infinite energy release. Hence. 2 − m2 φ.4) ∂t2 Taking the first equation (2.6) Now.3) Hence. 2 THE KLEIN-GORDON EQUATION E = − p2 + m2 < −m. Hence. we see that we cannot interpret ρ as a probability density (as we did do in the non-relativistic case). ∂φ∗ −φ . (1. for energies below the rest mass. such a gap is a bit of a problem.

φ).10) We can solve by comparing with the non-relativistic Schrodinger equation − 2 − 2mZα ψ = 2m ψ.12) and use that the Laplacian in spherical polars can be written 2 = (2. consider electrostatic fields. such that A = 0. Aµ = 0.9) Now. 0) only (i. t) = 0. if we substitute a solution of the form φ(x. one easily obtains ∂ + iV ∂t 2 − 2 + m2 φ(x. Then. let us solve for the energy levels.11) where = E − mc2 . is to make the separable ansatz ψ= U (r) Y m (θ. t) = ψ(x)e−iEt into equation (2. (2. The way to solve this.3 EM Fields and the “Hydrogen Atom” 15 2. r (2.8) Now. so that Aµ = (Φ. we easily find that −(E − V )2 + − 2 + m2 ψ = 0.8). 4π 137 (2.2. and no magnetic term). (2. only a static term.3 EM Fields and the “Hydrogen Atom” If the 4-vector potential is non-zero. so that the KG equation becomes (∂µ + iqAµ ) (∂ µ + iqAµ ) + m2 φ = 0.7) Now. r ∂r2 r (2. substituting this into the KG equation (2.7). and V is not a function of time t.e. (2. r α= e2 1 = . r ˆ 1 ∂2 L2 r− 2. V (x) = qA0 (x) = qΦ(x).13) . then we make the minimal substitution ∂µ −→ ∂µ + iqAµ . where the potential is the central Coulomb V =− Zα .

nr = 0.14).14) Making the same ansatz (2. . we see that we can write the KG equation (2. 1.18) for E. (2.10). .17) Now. as 2m = − n = nr + + 1. substituting these into (2. .11) fairly easily results in d2 ( + 1) 2mZα − 2+ U (r) = 2m U (r). φ) = (E 2 − m2 ) Y m (θ. 1. 2. . + − dr2 r2 r (2. we see that the equivalent expression for the relativistic KG equation is E 2 − m2 = − where We can solve (2.9) as − 2 − U (r) (Zα)2 2EZα U (r) Y m (θ.16) may be solved to give 1 =− + 2 1 + 2 2 (2. . in order that we may expand this energy. (2. φ). 2 . − dr r2 r (2.18) n = nr + + 1.13) gives − d2 ( + 1) − (Zα)2 2EZα U (r) = (E 2 − m2 )U (r). . 2m −→ E 2 − m2 . we see that we have the relations 2m −→ 2E. E =m 1+ (Zα)2 n2 −1/2 2E(Zα)2 E. − 2 r r r r using the form of the Laplacian (2.15) Hence.12).14) are given by 2m(Zα)2 m. ( + 1) −→ ( + 1) = ( + 1) − (Zα)2 .16 2 THE KLEIN-GORDON EQUATION So.16) − (Zα)2 . 2n 2 nr = 0. . . Notice that (2. upon comparison of this with the non-relativistic version (2. Hence. after putting in the central Coulomb potential (2. the energy levels of the non-relativistic Schrodinger equation (2. 2n2 where the principle quantum number is given in terms of the number of radial modes and angular quantum number. (2. .19) We can consider two cases: the strong and weak potential. using our relations between the solutions of the relativistic and non-relativistic equations.

Notice that in this regime. The first term is the rest mass of the electron. Hence. spin-zero particles)..3 EM Fields and the “Hydrogen Atom” Weak potential If we take then (2. this corresponds to Z ≥ 69. which will give complex energies! To understand this. the second the non-relativistic result.17). 2 so that for = 0. ..2. and is thus the fine-splitting term. n = n. and thus cannot be used as a description of the Hydrogen atom.. this scalar KG equation described spin-0 particles. 2n 2 8 n4 + 1 2 − (Zα)2 + . 1 =− + 2 (Zα)2 3 (Zα)4 + + . However. The third term gives a splitting for modes with different .. Strong potential Let us take 1 Zα > + .. = .. 1 2 +2 +1 so that n = nr + (Zα)2 +1 2 +1 2 (Zα)2 . 2 +1 2 17 Zα 1. the derivation for the term is correct (infact. (2.. .. by (2. This fine-splitting term does not agree with experimental results for Hydrogen. . which means that to first order. the result is correct for pionic atoms. = n− 2 +1 2 4(Zα)2 + .19) can be expanded in Zα E =m 1− Now. to first order in Zα. . So. To second order. = − (Zα)2 . we shall take a look at a simpler problem. 1 2n + 2 n + 3 4 = nr + − and hence that n 4 = n4 1 − One can then show that E =m 1− (Zα)2 (Zα)4 − 2n2 2n4 1 2 − + .17) is complex.

e. where we can thus write p = i|p |. if p is complex. then (E − V )2 > m2 Similarly. and B for some reflected wave. then ⇒ (E − m)2 > V 2 . then the argument of the exponential eip z is real (and negative). dz 2 where we have imposed that there should be no imaginary wave from the right. and thus we have exponential decay “inside the barrier” (this is called “evanescence”). then one finds the relation p−p 2p B= A. Thus. p+p p+p . (E − m)2 < V 2 . centred at ˆ z = 0. for z < 0.3. along the z-axis.1 The Klein Paradox T =E−m 2 THE KLEIN-GORDON EQUATION V z=0 Figure 2. Consider a beam as in Figure (2.18 2. C = A. if p 2 > 0.1: A beam with kinetic energy T travels towards a barrier which has height V . If we impose continuity of the wavefunction at z = 0. Then.1). a somewhat expected result. d2 + m2 ψ(z) = 0. if p 2 < 0. For z > 0. that p is complex. p 2 = (E − V )2 − m2 . the KG equation (2. the KG equation is just −(E − V )2 − which is solved to give ψ(z) = Ceip z . dz 2 Notice that A is the amplitude for the incident wave. d2 + m2 ψ(z) = 0. where ψ and dψ/dz are continuous. p2 = E 2 − m2 . i.9) is just −E 2 − which is easily solved to find ψ(z) = Aeipz + Be−ipz . Now.

dp V −E . Then. Then. Then. the conserved current is j = −i ψ ∗ dψ dψ ∗ −ψ dz dz . That is. E − m > V . Strong potential Consider V > E.3 EM Fields and the “Hydrogen Atom” Now. vg = dE dp . Weak potential Consider V < E. for z > 0). for the reflected Be−ipz is and for the transmitted wave Ceip z jT = 2p |C|2 . jR = −2p|B|2 . which just gives (E − V )2 < m2 . as before. which we can differentiate with respect to p . propagation can occur if E − V < −m. This result says that propagation is allowed. then p 2 = (V − E)2 − m2 . For V > E + m. 19 So. this requires the kinetic energy to be above the barrier. Hence. if the kinetic energy is below the barrier. propagation if E − V > m. the group velocity. Consider waves on the right (i. V > E + m.e. dE p =− . somewhat as expected. Notice that jI = jR + jT . we have propagation through the barrier if (E − V )2 > m2 and evanescence if (E − V )2 < m2 (these cases correspond to p being real and imaginary. for large potential. which is an expected result. This is the Klein paradox. in 1D. respectively).2. the current for the incident wave Aeipz is jI = 2p|A|2 . or equivalently. to give 2p = 2(V − E) − Hence. Let us then consider two cases.

This suggests a similar effect in atoms.20 2 THE KLEIN-GORDON EQUATION Thus. for a full description for strong potentials. So. we must abandon a single particle description. waves current Figure 2. jT < 0. Particles move to the left. current is flowing to the left. if Z is large. even though waves are traveling to the right. and we will need a field theory – which we will come back to later. by an interaction with the barrier. then one finds z<0 and z>0 ⇒ ⇒ |jR | > |jI |. and anti-particles move to the right in the transmitted wave (with opposite charge).2: The relative amplitudes of the waves and current at the barrier. is that particle/anti-particle pairs are created at the barrier. If we substitute p into the expression for the conserved current. . That is. we see that the group velocity (which is the speed of the movement of particles) is “left”. in the reflected wave. whereas the wave is traveling “right”. The suggested interpretation of this.

21

3

The Dirac Equation
∂ψ , ∂t
2

Let us reconsider the Schrodinger equation, Hψ = i where the relativistic Hamiltonian is H= √ − (3.1)

+ m2 ,

which was derived using canonical quantisation. Recall that we got the Klein-Gordon equation by multiplying the Schrodinger equation through by H, HHψ = (−
2

+ m2 )ψ = −

∂ 2ψ . ∂t2

(3.2)

This made the KG equation second order (in time), which led to negative conserved quantities. To get around this, Dirac looked for an equation that would be first order in time, of the form (3.1). So, he proposed a Hamiltonian H = −iα · so that (3.1) becomes (−iα · + βm, (3.3)

∂ψ . ∂t This is known as the Dirac equation. The four constant coefficients α = (α1 , α2 , α3 ) & β are chosen such that the Dirac equation is consistent with (3.2), so that E 2 = p2 + m2 is recovered. They are also chosen such that the Hamiltonian H is Hermitian, resulting in real energy eigenvalues. + βm) ψ = i To be consistent with the KG equation, we thus require
3

HHψ = So, expanding out, HHψ = −

−i
j=1

αj

∂ + βm ∂xj

3

−i
k=1

αk

∂ 2ψ ∂ + βm ψ = − 2 . ∂xk ∂t

αj αk
j,k

∂ 2ψ − im ∂xj ∂xk ∂ ψ − im ∂xj ∂xk
2

αj β
j

∂ψ − im ∂xj

βαk
k

∂ψ + β 2 m2 ψ ∂xk

= −
j,k

αj αk
2∂ αj j 2

(αj β + βαj )
j

∂ψ + β 2 m2 ψ ∂xj

= − −im

ψ − ∂x2 j

∂ 2ψ (αj αk + αk αj ) ∂xj ∂xk j=k ∂ψ + β 2 m2 ψ. ∂xj

(αj β + βαj )
j

22 Hence, HHψ = −
j 2 αj

3 THE DIRAC EQUATION

∂ 2ψ − ∂x2 j

(αj αk + αk αj )
j=k

∂ 2ψ ∂xj ∂xk

−im
j

∂ 2ψ ∂ψ 2 2 +β m ψ =− 2 . (αj β + βαj ) ∂xj ∂t

Now, upon comparison of this with (3.2), we see that we require
2 αj = 1,

αj αk + αk αj = 0 (j = k),

αj β + βαj = 0,

β 2 = 1.

Thus, we have the anti-commutation relations {αi , αj } = 2δij , {β, αi } = 0, β 2 = 1. Thus, we see that αi , β cannot be numbers, but they can be matrices. 3.0.2 Properties of αi , β & Dirac Representation (3.4) (3.5) (3.6)

Now, one can easily see that αi and β must be Hermitian, in order that the Hamiltonian is Hermitian. Consider the eigenvalue equation and further αi ψ = λψ,
2 αi ψ = λ2 ψ,

2 but, from (3.4), we see that we have αi = 1. Hence, λ2 = 1, which means that the eigenvalues of αi are λ = ±1. This result also holds for β, by (3.6).

We shall state, then prove, that Tr β = 0, Tr αi = 0. (3.7)

That is, the sum over diagonal components is zero. Now, consider that Tr (ABC) = Tr (CAB) = Tr (BCA), which is easily provable by Tr (ABC) =
i,j,k

Aij Bjk Cki =
i,j,k

Cki Aij Bjk = Tr (CAB).

(3.8)

23 Then, consider that (3.5) is multiplying from the left by β gives β 2 αi = αi = −βαi β, where the first equality comes from (3.6). Hence, Tr αi = −Tr (βαi β) = −Tr (β 2 αi ) = −Tr αi , where the second equality follows from (3.8) and the third by (3.6). Hence, 2Tr (αi ) = 0 ⇒ Tr αi = 0,

βαi = −αi β,

which is the second of (3.7). The proof for Tr β = 0 is analogous. The final property of αi and β is that they must be of even dimension. To see this, consider that the trace of a matrix is the sum of its eigenvalues. But, we found that the eigenvalues of both αi and β are λ = ±1. We also showed that the trace of αi and β is zero. Hence, the only way to add such eigenvalues to give zero, is to have an even number of them. Hence, αi and β are of even dimension. Let us consider how to represent the four matrices αi and β. In dimension N = 2, there are only 3 linearly independent matrices of the form a1 a2 + ia3 a2 − ia3 −a1 These are σ1 = notice that they satisfy {σi , σj } = 2δij , as required. They also satisfy the commutation relation [σj , σk ] = 2i
jkl σl

,

ai ∈ R.

0 1 1 0

,

σ2 =

0 −i i 0

,

σ3 =

1 0 0 −1

;

(3.9)

(3.10)

(3.11)

If N = 4, there are a few ways in which we can represent the matrices. The way we shall is called the Dirac representation, where β= 1 0 0 −1 , αi = 0 σi σi 0 . (3.12)

16) . That we use the Dirac representation is a mere convenience for the non-relativistic limits.10). a specific representation is unneeded. α1 =  0 1 0 0 0 0 0 −1 1 0 0 0 3 THE DIRAC EQUATION some of these out.24 Just to make this block-notation clear. the wavefunction has four components   ψ1 (x)  ψ (x)   ψ(x) =  2  ψ3 (x)  .15) In a similar way. 3. to give iψ † ∂ψ = −iψ † α · ∂t ψ + mψ † βψ.14) Notice that one can fairly easily see the anti-commutation relations. Infact. (3.1 Conserved Current We can compute the conserved quantities by the usual argument. αj } = 2δij . and observable predictions are not affected by the choice of representation. αi } = 0. the adjoint equation of (3. {αi . (3. That is.13) is −i ∂ψ † =i ∂t · ψ † α + mψ † β. α2 =  0 0 i 0  .14). to give −i ∂ψ † ψ=i ∂t · ψ † α ψ + mψ † βψ. We multiply (3. (3. and one can just work with the commutation relations. the second of which can be seen just by (3.13) by ψ † from the left (we must be careful about this now).13) is a matrix equation with four components. long hand. we have that the Dirac equation i ∂ψ = −iα · ∂t ψ + βmψ (3. ψ4 (x) Since αi . β are Hermitian. there are other representations.    0 0 0 −i     . {β. we multiply ψ from the right onto the adjoint equation (3.   0 −i 0 0  i 0 0 0 As we previously said. we shall write    1 0 0 0 0 0 0 1  0 1 0  0 0 1 0 0    β=  0 0 −1 0  . Hence.

subtracting (3. and using the usual commutation relation [xj . i=1 Hence. then we require that [H. Let us ask: is orbital angular momentum conserved? If it is. In particular.17) |ψi (x)|2 ≥ 0. 3. j = ψ † αψ. (3. upon comparison with the continuity equation ∂ρ + ∂t then we see that ρ = ψ † ψ. we see that 2i ∂ ψ † ψ = −2i ∂t · ψ † αψ . in the usual way (which we could not do for the KG equation).3. L = x × p.16) from (3. using the Dirac Hamiltonian (3.15).2 Angular Momentum & Spin Then. L] = 0.3).2 Angular Momentum & Spin Consider the orbital angular momentum operator. Li = ijk xj pk . notice that  ψ1  ψ  ∗ ∗ ∗ ∗ ρ = ψ † ψ = (ψ1 ψ2 ψ3 ψ4 )  2  =  ψ3  ψ4  4 · j = 0. pk ] = iδjk . H = −iα · + βm = α · p + βm. 25 or. . or. we see that we can treat ψ as a probability density function for position. So. in index notation.

However. Σk ] [αj . Σk ] = 2i jkn αn .26 we consider the commutator [H. Σk ] = = = = [αj pj + βm. Σk ]pj 2i jkn αn pj 2i knj αn pj . (3. orbital plus something-else).11) to show that [αj . it does not.20) In a much simpler. knm xn pm ] = knm αj [pj . αj Σk − Σk αj = = = 2i 0 σj σk σj σk 0 − 0 σk σj σk σj 0 σj 0 0 σj . Instead.21) . L] = −iα × p. σk ] 0 jkn 0 σn σn 0 = 2i jkn αn .e. Σk ] = 0. we have used (3. the commutator of the Hamiltonian with this Σi operator.18) which is non-zero. xn ]pm = −i knm αj δjn pm = −i kjm αj pm . Now. we only considered orbital angular momentum. [H. Let us consider some matrix operators Σj = Then. 3 THE DIRAC EQUATION (3. Hence. Hence. σk ] [σj . (3. and hence orbital angular momentum is not conserved.19) 0 [σj . one can also show that [β. where the last equality allows us to see that [H. but analogous manner. (3. Lk ] = [αj pj . in which case we must say that the Hamiltonian does not commute with orbital angular momentum. let us suppose that the total angular momentum is conserved (i. then we could have said that the Hamiltonian commuted with the orbital angular momentum operator. If this were zero.

2 in Dirac representation. α= 0 σ σ 0 . (3. [H. (3. N is some normalisation and u(p) some 4-component spinor. whose form is to be determined. consider the combination 1 J = L + Σ. i ∂ψ = −iα · ∂t ψ + βmψ. We also require that p2 = E 2 − m2 .18).22) (3.26) 1 0 0 −1 . where the solutions are of the form ψ(x) = N e−ipx u(p). Thus. (3. defined by (3.23) 1 [H.27) . as it commutes with the Hamiltonian. Then. then Σ2 = 1. and hence the eigenvalues i 1 1 of Si are ± 2 .22) and (3.25) where both ξ and η are two component spinors (thus making u a 4-component spinor).24) where px = pµ xµ = Et − p · x. since σi = 1. We use Dirac representation. we see that the particle has spin− 2 .23) is conserved. 1 1 S= Σ= 2 2 σ 0 0 σ . J] = −iα × p + 2iα × p = 0 2 from (3.3. Σ] = 2iα × p.3 Plane Wave States Let us consider solutions to the Dirac equation. Notice further. (3. We can define some spin operator. Therefore. to satisfy the KG equation. 3. we see that the operator J.3 Plane Wave States thus. We will substitute into the Dirac equation. whereby β= and we write u= ξ η . 2 so that 27 (3.

indeed as. Now.25) and (3. expanding out this matrix. −p) e−ipx = −ipµ e−ipx .28). βmu(p) = (E − α · p) u(p).30) (3. solving (3. notice that (σ · p)2 = = (σ1 p1 + σ2 p2 + σ3 p3 ) (σ1 p1 + σ2 p2 + σ3 p3 ) 2 2 2 σ1 p2 + σ2 p2 + σ3 p2 3 1 2 +(σ1 σ2 + σ2 σ1 )p1 p2 +(σ1 σ3 + σ3 σ1 )p1 p3 +(σ2 σ3 + σ3 σ2 )p2 p3 . Eη = σ · pξ − mη. ξ η Hence. we get two coupled simultaneous equations. one finds Eξ = mξ + σ · p which rearranges easily into (E − m)(E + m)ξ = (σ · p)2 ξ. Notice that the Dirac equation (3. E+m (3. 1 0 0 −1 E ξ η ξ η ξ η 0 σ·p σ·p 0 ξ η .29) for η. using the matrices (3. .27) can be written as βmψ = i ∂ +α· ψ ∂t = i (∂0 + αj ∂j ) N e−ipx u(p) = i (−iE + iαj pj ) N e−ipx u(p). So. = m1 σ · p σ · p −m1 So. Eξ = mξ + σ · pη.28) (3.31) (3.26) m rearranging =E + . easily derivable 3 THE DIRAC EQUATION ∂µ e−ipx = −i (E. gives η= σ·p ξ. E+m σ·p ξ.28 and use the useful.29) so that upon substitution into (3.

we have somewhat inserted the factor E + m by forethought.34) 0 . (3.33). (3. Hence. (E + m)2 s . we see that (3. we have derived that the positive energy solution is us (p) = Notice that u† (p) = s moreover that u† (p)us (p) = (E + m) χ2 + s s √ E+m χs σ·p χ E+m s . and also σi = 1. χs=− 1 = 2 us (p) = = = = √ ξ η σ·p ξ E+m ξ √ σ·p E+m E + mχs √ E + mχs χs σ·p χ E+m s .35) √ E + m χs (σ · p)∗ χs .30). using this solution in (3. Later on we shall be writing the normalisation of the spinors.31) – also expanding out the brackets on the LHS – gives (E 2 − m2 )ξ = p2 ξ. E+m Therefore.26) is χs= 1 = 2 1 0 . which we label √ ξs = E + mχs . Now. whereby E(p) = + m2 + p2 .33) we then see that there are two independent solutions to (3.3 Plane Wave States 29 2 The final three terms are all zero. by the anti-commutation relation (3.32) Thus. where the inserted factor will make expressions a lot simpler. 1 √ In writing (3.34). 3 2 1 (3. we see that (σ · p)2 = p2 + p2 + p2 = p2 . recall the positive energy solution. substituting this into (3. E+m (σ · p)2 2 χ . Hence.3. where the χs are such that (3.10).

d3 x ψ † (x)ψ(x) = |N |2 d3 x u† (p)us (p) s (3. where we normalise in some box of volume V . u† (p)us (p) = (E + m) 1 + s = 3 THE DIRAC EQUATION p2 (E + m)2 (E + m)2 + p2 E+m (E + m)2 + E 2 − m2 = E+m = 2E. Then.35).36) = |N |2 2EV = 1. using (3. 2EV (3. one can easily rearrange to find that ξ=− σ·p η. (3. One finds that the equivalent of (3.24) ψ(x) = N e−ipx u(p). then we see that N=√ Therefore. −Eη = σ · pξ − mη. This solution describes a particle with momentum p.29) are −Eξ = mξ + σ · pη.s (x) = 1 . If we recall the solution we were seeking. and if we require its normalisation. E+m . energy E > m and spin Sz = s = ±1/2.37) where us (p) is given by (3. 2EV us (p)e−ipx √ .28) and (3. The negative energy solution is written as (−) ψp.30 which becomes.s (x) = N eipx vs (p).32). the full positive energy solution is (+) ψp.

1) for a depiction of the theory. as no positive energy particle can drop in there.38) This corresponds to a particle with momentum (−p). as there is no free state for it to occupy. one still has the relation E 2 = p2 + m2 .35) and (3. But.4 Dirac Hole Theory As spin− 1 particles are fermions. energy E = − p2 + m2 and spin (−s). This equation has similar solutions. See Figure (3. 31 Hence. 3. which states that no two 2 fermions can be in the same state. they obey the exclusion principle. (3. if it is given enough energy E > 2m.39) √ E+m σ·p χ E+m −s χ−s .4 Dirac Hole Theory which results in the equivalent of (3. If the negative energy states are full. s. p. the absence of a e− particle with −E. a negative energy particle can transition into the positive energy states. That is. .3. Notice that in the non-relativistic limit p → 0 (and the energy E is of the order the rest mass m). which looks like an anti-particle. and so one arrives at vs (p) = and thus the wavefunction vs (p)eipx (−) . then no positive energy particle can transition into the negative states. ψp. . That the sea is completely filled is important. −s. (3. with a negative energy. which looks like a e+ particle with E. in volume V .38) become us −→ vs −→ √ 2m √ 2m χs 0 0 χ−s .s (x) = √ 2EV (3. Then. −p. imagine the “vacuum” as being a completely filled “sea” of negative energy states. Now.33) being p2 η = (E 2 − m2 )η. there will be a hole in the negative energy states.

and appears as an anti-particle 3.32 E 3 THE DIRAC EQUATION empty in vacuum m p −m full in vacuum Figure 3. and recall that β 2 = 1. negative energy states – given by E = − p2 + m2 are full. However. Now.1 Relativistic Covariance of The Dirac Equation Covariant Notation The Dirac equation written in the form i ∂ψ = (−iα · ∂t + βm) ψ is not obviously covariant.41) . ∂t writing βα · as βα · = βαi ∂ . In vacuum. if we multiply from the left by β. ∂xi then we see that the Dirac equation looks like iβ ∂ψ = ∂t −iβαi ∂ + m ψ. if we make the definitions. (3. a negative energy particle can be promoted to the positive states. γ i ≡ βαi .5. ∂xi (3. γ 0 ≡ β.40) Now. then we see that ∂ψ iβ = (−iβα · + m) ψ.5 3.1: A schematic of the Dirac hole picture. so that no positive energy particles can occupy the negative states.

βαj } βαi βαj + βαj βαi −β 2 αi αj − β 2 αj αi −{αi .43) is covariant or not. βαi } β 2 αi + βαi β αi + β(−βαi ) 0. αi (i. γ 0 } = 2. Also. if we write ∂ψ = ∂t −iγ i ∂ + m ψ. It is also useful to introduce the Dirac adjoint. the commutation relations). (3. βαi ). but the γ µ do not form the components of a 4-vector. one can deduce that {γ µ .3. (3. Similarly.5 Relativistic Covariance of The Dirac Equation then we see that (3. γ i } = = = = {γ i .40) can be written iγ 0 Then. ψ † αψ). then we see that using the Dirac adjoint. ∂xi 33 (3. this is not obvious. we can write j µ as ¯ j µ (x) = ψ(x)γ µ ψ(x). we must look for transformation properties of ψ so that (3.47) (3. and so the equation of continuity is ∂µ j µ = 0.43) we see that (3. αj }. Also. one can see that γ µ† = γ 0 γ µ γ 0 . so that {γ 0 .46) (3. j µ = (ψ † ψ. (iγ µ ∂µ − m) ψ = 0. If we recall the components of j µ . Hence. ¯ ψ(x) ≡ ψ † (x)γ 0 . γ i ) = (β. If (3. γ ν } = γ µ γ ν + γ ν γ µ = 2g µν .43) is a valid physical equation (which it is). Hence. γ i† = −γ i .44) {β. β} = 0. .e. {βαi . which is equivalent to saying γ 0† = γ 0 . as β 2 = 1. Now.45) (3. from (3. Also. one can see that {γ 0 . then (γ 0 )2 = 1. γ j } = = = = Hence. It is easy to see that {αi .42) γ µ = (γ 0 .17).43) is covariant.42) can be written as The properties of the γ µ matrices can be deduced from the properties of β. the principle of special relativity says that it must be covariant. it is still not obvious as to whether (3.

such that x −→ x = ax. One will also notice that (3. that the Dirac equation (3. provided. multiplying from the left by S −1 (a).49) and (3. (3. and. with “equivalence element” S(a). Physically. we want to show ⇒ where ∂µ = aµ ν ∂ν .53) into (3. O → O . even after transformation (which is not too much to ask of a matrix).51) Those familiar with the terminology of group theory will notice that (3. So.49) We shall state. Let us substitute (3. to give iγ µ aµ ν ∂ν − m S(a)ψ(x) = 0. and then prove. prove the covariance.43) and (3. (3.50) is just the statement of equivalence.52).52) .51) is just the statement that S −1 (a) is equivalent to S † (a). (3. we require aµ ν γ ν and γ µ to be equivalent.54) (3.2 Proof of Covariance 3 THE DIRAC EQUATION Consider a Lorentz transformation from one frame to another.47) are covariant.50) (3. we are requiring a γ-matrix to still be a γ-matrix. (3. γ 0 S † (a)γ 0 = S −1 (a).5.53) (iγ µ ∂µ − m) ψ(x) = 0 iγ µ ∂µ − m ψ (x ) = 0.34 3. (3. where aµ ν is some transformation matrix that satisfies σ aµ ν aµ σ = δν . S −1 (a)γ µ S(a) = aµ ν γ ν . let us define some operator S(a) by ψ(x) −→ ψ (x ) = S(a)ψ(x). with equivalence element γ 0 . iS −1 (a)γ µ S(a)aµ ν ∂ν − m ψ(x) = 0. so that x µ = aµ ν x ν . ψ(x) = S −1 (a)ψ (x ) = S(a−1 )ψ(x ). correspondingly. That is.48) So.

inserting this for the ψ in the middle term of (3. ψ (x ) = S(a)ψ(x) and multiplying by γ 0 from the right. if we insert (γ 0 )2 = 1. between ψ † and S † .5 Relativistic Covariance of The Dirac Equation Now.56) by ψ from the right. Notice that if we multiply (3. if we multiply (3. then we see that S −1 (a)γ µ S(a)aµ σ = aµ ν γ ν aµ σ σ = γ ν δν = γσ.51) for γ 0 S † γ 0 .50) for S −1 γ µ S.55) holds. using (3. then we see that ¯ ¯ ψ = ψS −1 (a). ¯ j µ = aµ ν ψγ ν ψ. as required.49) for ψ on the far RHS. Thus. where the second equality again follows by the definition of the Dirac adjoint. gives ¯ ¯ ψS −1 γ µ Sψ = aµ ν ψγ ν ψ. gives (iγ µ ∂µ − m) ψ(x) = 0. then ¯ ¯ ψ = ψ † γ 0 γ 0 S † γ 0 = ψγ 0 S † γ 0 .56) ¯ So. So. ¯ Let us consider the transformation law of ψ(x). ψψ is a Lorentz scalar: its value is the same in all coordinate systems. That is. where the first equality follows by (3. (3. using (3. after using (3.57) ⇒ ψ † (x ) = ψ † (x)S † (a). Now. as required. And therefore. the Dirac equation is covariant. show that ¯ ¯ j µ = ψ γ µ ψ = aµ ψγ ν ψ ν 35 (3. gives ¯ ¯ ψ γ µ ψ = ψS −1 γ µ ψ . using this result in (3.54). We still need to show that (3. ¯ ¯ ¯ ¯ ψ ψ = ψS −1 ψ = ψS −1 Sψ = ψψ.48) in the second equality.47) is covariant. ¯ That is. ¯ ψ †γ 0 = ψ = ψ†S †γ 0.50) by aµ σ from the right. ¯ ¯ ψS −1 γ µ ψ = ψS −1 γ µ Sψ. (3.55).3. . If we now use (3.46).

2 3! 2 4! ω 3 1 ω 4 1 α1 + 1 + . S(a) = eωα1 /2 = 1 cosh for a Lorentz boost.5. we have a coordinate transformation x = ax. 3! 5! cosh x = 1 + x2 x4 + + . relative to O. x2 x3 + + . 1 − v2 so that for a frame O moving at speed v.. . sinh ωα1 ωα1 + cosh 2 2 x3 x5 + + . consider that sinh x = x + and that ex = 1 + x + Hence.. ωα1 + 2 ω = 1 + α1 + 2 ω = 1 cosh 2 = 1+ 2n+1 2n where we have made use of α1 = 1 and α1 = α1 .5. 2 3! 2 4! .....58) Consider the transformation t −→ t = t. sinh ω = √ 1 − v2 The form of the S(a) is Now. .. Hence.. we have cosh ω = √ S(a) = eωα1 /2 . 2! 4! 0 0 1 0  0 0  ... . 3. xi −→ xi = −xi .4 Parity ω ω + α1 sinh . where cosh ω − sinh ω  − sinh ω cosh ω a=  0 0 0 0  v = γv.3 Lorentz Boost 3 THE DIRAC EQUATION Consider a Lorentz boost along the x-axis. = cosh x + sinh x. 2 2 (3. 0  1 1 = γ. 2! 3! ωα1 2 1 + 2 2! ω 2 1 1+ 2 2! ω + α1 sinh 2 ωα1 3 1 ωα1 4 1 + + . Then.36 3.

Particles at Rest Consider a particle at rest. we require (3.3. P −1 γ µ P = aµ ν γ ν . and is crucial for understanding positronium e+ e− and mesons q q . where  1 0 0 0  0 −1 0 0   a=  0 0 −1 0  . gives P ψ (+) = ηψ (+) . Therefore. P ψ (−) = −ηψ (−) . P −1 = γ 0 P † γ 0 . Then. and we conventionally set it η = 1. . 0 ψ (−) = 0 eimt .51) to hold. acting P upon both states. η = ±1. All of these conditions are satisfied by the choice P = ηγ 0 . 37 Now. by the form of the matrix a. so that P = γ . 0 0 0 −1  Notice that this matrix satisfies.5 Relativistic Covariance of The Dirac Equation so that x µ = aµ ν xν . χ−s So. but particles and anti-particles have opposite “intrinsic” parity. ¯ 0 We cannot measure η. Now.39) become ψ (+) = χs −imt e . Now.50) and (3. so that P 2 = 1.37) and (3. This has been experimentally verified. we see that P −1 = P . σ aµ ν aµ σ = δν . we see that both the positive and negative energy solutions are eigenstates of parity P . then ψ (x ) = P ψ(x). if we have that S(a) = P . for covariance. as indeed it must. the positive and negative energy solutions (3. Recall that in Dirac representation. γ0 = 1 0 0 −1 .

3. indeed. (3. where q is some charge. σ·p χs E+m (+) ψ−p. or. A).s (x) = N χs σ·p χ E+m s e−i(Et−p·x) . the Dirac equation is rewritten into [iγ µ (∂µ + iqAµ ) − m − S] ψ(x) = 0. η(x) (3.59) Magnetic Moment of Electrons Let us consider interactions with EM fields only.38 3 THE DIRAC EQUATION Particles in Motion We now consider the full wavefunctions. i 3. (−) ψp. using the parity transformation. so that S = 0.1 ∂ψ = α · (−i ∂t − qA) + β(m + S) + qA0 ψ. P : x −→ −x.6. (+) P ψp. Aµ (x) = (A0 . (−) −ψ−p.s (x = Thus again. we see the intrinsic parity difference between particles and anti-particles. which can be introduced via the minimal substitution ∂µ −→ ∂µ + iqAµ . We could also introduce interactions with a scalar field.6 Interaction With Fields Consider interactions with an electromagnetic field. and look for energy eigenfunctions of the form ψ(x) = φ(x) −iEt e .s (x ) σ·p χ E+m −s ei(Et +p·x ) −N χ−s ). via m −→ m + S(x).s (x) = χs e−i(Et +p·x ) .60) . (+) ψp.s (x) = N σ·p χ E+m −s χ−s ei(Et−p·x) .s (x) = N = (−) P ψp. Hence. Hence.

η= − qA) φ = (E + m)η. we can easily rearrange it into σ · (−i − qA) η + qA0 φ = (E − m)φ. − qA)]2 φ = (−i − qA) · (−i − qA) φ +iσ · (−i − qA) × (−i − qA) φ.63) for η.60) into (3. consider (3.59). |qAµ | m.65) Consider the second term.6 Interaction With Fields If we substitute (3. Let us then use the vector identity × (Aφ) = φ ×A−A× φ. E + m = + 2m. inserting (3. . so that we cannot and we use that = E − m (notice that we do not have that qA0 neglect the qA0 term). 39 (3.61). to give σ · (−i but. Therefore. 1 σ · (−i 2m − qA) φ.62). m. We shall consider the non-relativistic weak-field limit =E−m m. so that we can read off terms we can recognise. (3. We neglect the qA0 term in (3.61) (3. iσ · (−i − qA) × (−i − qA) φ = −qσ · [ × (Aφ) + A × ( φ)] . so that E + m ≈ 2m. 1 [σ · (−i 2m We first use the easily provable (σ · a)(σ · b) = a · b + iσ · (a × b). however. we see that σ · (−i σ · (−i − qA) η + (qA0 + m)φ = Eφ.62) where the −m in the second expression comes from the −1 in the β matrix.63) Now. so that [σ · (−i − qA)]2 φ + qA0 φ = φ. after noting that curl grad is zero.64) Our task now is to tidy this expression up. − qA) φ + (qA0 − m)η = Eη. (3. . (3. and take mη to the RHS. So. and that a × a = 0.3.

Hence. and if we use the Coulomb gauge. (−i − qA) · (−i − qA) φ = − · (Aφ) = 2 (3. Let us now consider the first term of (3. using this in (3. notice that we have · Aφ + A · · A = 0.67) becomes − 2 · (Aφ) = A · φ. and hence that 2iqA · = iq (B × x) · = iq (x × . + 2iqA · + q 2 A2 φ. using this and (3. but. we see that (3.65) is − 2 − qL · B + q 2 A2 φ.68) Thus far. (3. ) · B.65). our discussion has been very general. . to see that this term is − 2qS · Bφ. (3. 3 THE DIRAC EQUATION 1 × A.65) is just −qσ · ( where we have used that B = just × A) φ = −qσ · Bφ. (3. Thus.66) φ + iq · (Aφ) + iqA · φ. we shall now specify ourselves to a homogeneous magnetic field.65).40 so that the second term of (3.66) in the first and second terms of (3.68). L = x × p = −ix × 2iqA × = −qL · B. We can also use that S = 2 σ. which one can show is given by 1 A = B × x. we see that the first term of (3. then Hence.65) becomes [σ · (−i − qA)]2 φ = − 2 − 2qS · B − qL · B + q 2 A2 φ.67) Now. 2 Notice then that but. φ + q 2 A2 φ.

and both f.3.7 Particles in a Spherical Potential Hence.7 Particles in a Spherical Potential Consider the time independent Dirac equation.70) so that f (r) is large for positive energy solutions. Notice that particles have f large. g(r)iσ · ˆχs r (3. P = 1. Let us consider solutions with j = 1/2.71) to see that we get two coupled differential equations. and g(r) is large for negative energy solutions. r dr and expand (3.69).70) into (3. )(σ · ˆ)g(r) = r d 2 + dr r g + (m + S + V − E) f = 0. and anti-particles have g being large. m+S+V −iσ · We can use that −iσ · −m − S + V (σ · f (r)χs g(r)iσ · ˆχs r =E f (r)χs . g.72) (3. 2m q (L + 2S) . V (r) depend on r = |x| only. [−iα · + β(m + S(r)) + V (r)] ψ = Eψ. g are scalars. A = 0. .73) These coupled differential equations can then be solved for f. This prediction has been verified by experiment. − where 1 2m 2 41 + qA0 − µ · B + q2 2 A φ = φ. We have that V (r) = qA0 (r). we can rewrite (3. g(r)iσ · ˆχs r (3. so that we have an interaction with static fields only.69) Notice that S(r). Let us then substitute (3. µ= 3. wavefunctions in Dirac representation have the form ψ(r) = f (r)χs . df + (m + S − V + E) g = 0.71) 2g(r) dg + .64) with things that we recognise. Now. (3. dr (3. 2m The factor of 2 infront of S is the so-called g-factor (gyromagnetic ratio) of the electron.

we see that quarks inside a protons are highly relativistic. We shall set mq = 0 (which is a good approximation for u.75) 1 df . dr2 We can solve this by setting (3. d2 f 2 df + + E 2 f = 0.42 3.74) . the uud quarks are confined to make a proton p. we have some scalar potential. We also know that quarks strongly interact at distances r ∼ 1fm. as we would like to be able to allow anti-protons to exist). and interact weakly at r 1fm (their so-called asymptotic freedom). dr we can substitute the second. df + Eg = 0. d-quarks). Notice that we are not setting the constituent mass to zero. 2 dr r dr u(r) . So.1 The MIT Bag Model 3 THE DIRAC EQUATION This is a very crude model of hadrons.72).7. where R ∼ 1fm. taking λ = 2R and p = h/λ = 600MeV. E dr (3. which will allow quarks and anti-quarks to attract (note. the simple model we will discuss – the MIT bag model – binds quarks in a scalar potential S(r) = 0 S0 r < R. r so that upon substitution into (3. the radial equations (3. For example. (3. which is the mass of the bare quark and the virtual gluons. If we consider typical quark momenta. one can easily find f (r) = d2 u = −E 2 u.73) become for the various limits: Inside r < R 2 d + dr r g − Ef = 0. g=− into the first. So. r > R.75). with mass mp = 940MeV/c2 and radius Rp = 1fm. So. we want anti-quarks to attract as well.

1 − ER . under the requirement that f (r) is finite at r = 0. and g(r) = If we assume that S0 κe−κr e−κr − 2 r r .7 Particles in a Spherical Potential which has solution 43 u(r) = A1 sin Er + A2 cos Er. by (3. r We can now apply continuity at the boundary r = R.74). R ER2 R or. sin ER sin ER cos ER f (R) = g(R) ⇒ = − . and we find f (r) = Be−κr . r (3. so that (3. we must therefore set A2 = 0. substituting the second into the first to give d2 f 2 df 2 + + E 2 − S0 f = 0.3.e. then κ ≈ S0 and g(r) = Be−κr ≈ f (r). (3. r B S0 + E κ≡ 2 S0 − E 2 .77) are equal from below. we see that g(r) = A Outside r > R d 2 + dr r g + (S0 − E) f = 0. ER tan ER = .76) and (3. 2 dr r dr To solve this. df + (S0 + E) g = 0.76) again. where we used the boundary condition that f should not diverge as r → ∞. Now. we make a similar ansatz. and hence f (r) = Then. strong potential). dr sin Er cos Er − Er2 r .77) A sin Er . E (i.

44 This can be graphically solved to give ER = 2. In the non-relativistic case. Es 1 = 2 3 THE DIRAC EQUATION 2. Therefore. with Z = 1.04. in this crude model.78). R or. and this is because the Dirac equation “knows about” spin. V (r) = −  En = m 1 + n− j+ 1 2 Now consider a system with Z α 2 2 1 2 2 −1/2 − Z 2 α2  . the mass of the proton. the spin-orbit interaction is added in manually. and the α2 -term is the first relativistic correction and is O(10−5 ). This isnt too far off. R Hence. given that we have completely ignored things like the quarks interact with nothing except the binding bag – so. there is no mention of . 3E = 3. and the splitting.12 . one can solve the radial equations. in a similar way to above. and all energy-information is .2 Hydrogen-like Atoms Zα r Then. then. expanding in powers of α. ignoring the gluons around them. and those two levels are separated by an energy gap ∆. and find S(r) = 0.2).7. (3. the splitting gives a degeneracy between 2s1/2 and 2p1/2 . Notice that in the relativistic spectrum. Consider a hydrogen atom. this corresponds to atoms with Z > 137) – this is another instance of the Klein paradox for strong potentials. we see the spectrum of the n = 2 level. 6. Notice that the 2s1/2 level splits into 2p1/2 and 2p3/2 . is mp = 3E = 1200MeV/c2 . if we have 3 quarks. the 2 energy becomes imaginary (for j = 1/2. Also notice that for Z 2 α2 > j + 1 . obviously). as is gs = 2.04 .78) + j+ Notice that does not appear in this expression. one finds n = En − m = − mα2 α2 1+ 2n2 n 1 j+ 1 2 − 3 4n + O(α4 ) . The first term gives the Rydberg energy. this is because in (3. With reference to Figure (3. only j (and n.

7 Particles in a Spherical Potential 2p3/2 2p3/2 2s1/2 2p1/2 non-relativistic ∆ 2s1/2 ∆ 45 2p1/2 relativistic Figure 3. There was fantastic agreement in the relativistic prediction of these splittings. and one can predict the magnetic moment gs = 2. as one did for the non-relativistic case. for example.. α + . This highlights the need for a more complicated description of relativistic quantum systems. . In this relativistic case. but in the relativistic case. the Dirac equation “knows about” spin inherently. when it was observed that the 2s1/2 -level was shifted down by O(α3 ).. The states have the same splitting ∆ as in the non-relativistic case.2: The spectrum of splittings in the H-atom. and was promptly awarded a Nobel prize.3. the magnetic moment is not exactly gs = 2 in the more complicated model. e− simple e− e− e− perturbative correction Figure 3. In the non-relativistic case.3). This degeneracy is exact to all orders of expansion.3: A difference in the simple Dirac model and a more complicated – but needed – perturbative correction. as in Figure (3. g =2 1+ 2π which is again confirmed by experiment. so that one does not need to add them in “by hand”. contained in j. the Dirac equation already “knew about” spin. This shift is called the Lamb shift. and orbital angular momentum. The reason behind the failure is that one needs to go further than the rather simple Dirac equation. . the spin-orbit coupling is put in “by hand”. until 1947. Also. and quantum fluctuations come into play. one must think that problems are present when one gets complex energies for strong potentials! Also.

46 3 THE DIRAC EQUATION .

where we could let L → ∞. with tension T having transverse displacement φ(t. momentum and forces. c2 ∂t2 ∂z where the speed of sound is T c= . We let our theory do so by imposing periodic boundaries. or the electromagnetic field. where the modes are operators. The energy of the string is given by L E= 0 dz 1 µ 2 ∂φ ∂t 2 1 + T 2 ∂φ ∂z 2 . how do we quantise waves? The general idea is to expand a field in terms of modes of harmonic oscillators. lies in that photons are quanta of the electromagnetic field. and have infinitely many degrees of freedom. z) = φ(t. length L. which we shall consider in this section. For example. Now. The idea of assigning an operator to a field. we can expand the solution in terms of normal modes. and the second a potential energy density. and that the KG and Dirac equations are wave equations. z). We use fields.1 Quantum Mechanics of a String Consider a string. so that c = 1 and so the wave equation and energy become 2 2 L ∂ 2φ 1 ∂φ ∂ 2φ 1 ∂φ = 2. is called secondary quantisation. 4. Fields can support waves.1) For simplicity. and some of the results of doing this. The string will satisfy the wave equation ∂ 2φ 1 ∂ 2φ = 2. which we will do at the end of the calculation. φ(t. we want to impose travelling waves. we shall set T = µ = 1. or density of a compressible fluid. in which particle numbers can change. E = dz + . . φn = e±i(ωn t−kn z) . where a field is some continuous dynamical variable which has a value at each point in space. ∂t2 ∂z 2 ∂t 2 ∂z 0 The first terms is a kinetic energy. µ where µ is the mass per unit length of the string. Now. or to a quantity in general. Notice that the wave equation is like a KG equation.47 4 Quantum Fields We now want to move on to a many-body theory. An idea of how to do this. with m = 0: massless. z + L). (4. But. the displacement field of a string. It is the technicalities of doing this. and can transmit energy.

φ(t. Upon substitution of the normal mode into the wave equation. n 2ωn L n=−∞ ∞ (4. we made use of the orthogonality relation L ei(kn −km )z dz = δnm L. We now want to work out the Hamiltonian – i. 0 Hence. the periodic boundary condition is satisfied. 1 2 L dz 0 ∂φ ∂t 2 1 = 2 L L dz n. and the second term to negative energy solutions. m n n m 4L where in the last equality. where we evaluate at t = 0. with this choice of kn .m ωn ωm L {an am δn. Notice that at each position z along the string. the kinetic energy. So. we find an expression for the kinetic energy. L and n is an integer.−m − an a∗ δnm − a∗ am δmn + a∗ a∗ δn.2) √ The prefactor 1/ 2ωn L is there purely for later convenience – if we did not put it in. the string performs simple harmonic motion. performing the sum over m.e. we find the requirement.48 where 4 QUANTUM FIELDS 2πn .m 0 1 √ 2L ωn ωm −iωn an eikn z + iωn a∗ e−ikn z n × −iωm am eikm z + iωm a∗ e−ikm z m = n. n n −n 4 . The first term in the brackets bears resemblance with the positive energy solutions. we would have found that equations become more complicated than if we did. We can expand an arbitrary φ in terms of the normal modes φn . a dispersion relation. z) = 1 √ an e−i(ωn t−kn z) + a∗ ei(ωn t−kn z) . kn = ωn = |kn |.m 0 dz √ 1 ωn ωm −an am ei(kn +km )z + an a∗ ei(kn −km )z √ m 4L ωn ωm × +a∗ am ei(km −kn )z − a∗ a∗ e−i(kn +km )z n n m =− n. Then. the energy – in terms of the coefficients an .−m } . 1 2 L dz 0 ∂φ ∂t 2 =− n ωn an a−n − 2a∗ an + a∗ a∗ .

n 2 (4. am ] = [a† . m [an .4. we find H= n 1 2 2 2 p + ωn qn . n (4. We also impose n n the canonical commutation relations [pn . n n −n 4 where we have used |k−n | = kn and ω−n = ωn . our analysis has been completely classical in nature.5) [pn . n 2 so that upon rearrangement and insertion into the total energy (4.3). pm ] = [qn . a† ] = 0. an . So far. and an .3). we promote qn .e. pn to Hermitian operators. To quantise the theory. and therefore commute. replacement of quantities with non-commuting operators). 2m 2 to see that (4. and hence.4) To find the Hamiltonian of this theory (i. a† = δnm . a∗ to the an . qm ] = −iδnm .1 Quantum Mechanics of a String In a very similar fashion. we compute the potential energy. 2 n We can then compare this with the usual expression for the energy of a simple harmonic oscillator. Notice that we have not been careful with allowing the coefficients an to commute – this is because they are purely classical in nature. and we find H= n ωn † an an + an a† . we repeat the procedure that resulted in (4. (4. 1 2 L 49 dz 0 ∂φ ∂z 2 = n ωn an a−n + 2a∗ an + a∗ a∗ . p2 1 H= + ω 2 x2 . a† operators. we could introduce the real variables. but being careful with ordering.6) . We then add the potential to the kinetic energy to find the total energy. qm ] = 0. an + a∗ qn ≡ √ n . 2ωn pn ≡ −i ω (an − a∗ ) . H= n ω n a∗ an . n m (4.3) is actually equivalent to an infinite discrete sum over modes of the harmonic oscillator.3) To make this look more familiar.

1 Phonons [H.8) Let us consider the state |ψ . Nn ! Nn (4. Therefore. |{Nn } = n a† √n |0 . C] = A[B. Similarly. We define the ground state of a system to be the state with lowest possible energy. Therefore. and suppose that the Hamiltonian acting upon the state gives the energy. n The second equality follows after using the first of the commutation relations (4. n n 4. and an decreases the energy by ωn . we say that a† increases the n energy of a system by ωn . by applying a† to the ground n state as many times as we wish.11) . (4. Now. with eigenvalue E + ωn . H |ψ = E |ψ . C] + [A.7) The identity [AB.9). (4. a† ] = 1. any attempt to lower the energy with an results in zero. denoted by |0 . consider H a† |ψ n = Ha† |ψ n = ωn a† + a† H |ψ n n = (E + ωn ) a† |ψ .1. (4. Hence. one can compute that Han |ψ = (E − ωn ) an |ψ .10) We can then construct excited states from the ground state. we see that a† is an eigenstate of the Hamiltonian. n H= n 4 QUANTUM FIELDS ωn a† an + n 1 2 . as it is the state with lowest energy. an |0 = 0. a† ] = ωn a† . an ] = −ωn an . ∀n. C]B is very useful in computing [H. and n that an is an eigenstate with eigenvalue E − ωn .9) (4.50 or. using the commutator [an .

E −→ E − EZP . We say that a† creates a particle.12) = 1. we see that a† increases. as usual. Thus. redefining the Hamiltonian. 2 n=−∞ ∞ this sum is over an infinite range. with analogy with (4. momenta by units of kn .1 Quantum Mechanics of a String so that we will have Nn quanta in the mode n. and an decreases. an ] = −kn an . That is. The particle is associated with the wavefunction e−i(ωn t−kn z) . what about momentum. and hence gives an infinite value. H −→ H = n ωn a† an . this infinite energy is not a problem: we measure energies relative to some value – we measure energy differences. we say that phonons are the quanta of elastic waves.13) Recall that kn = 2π/λn . equivalently. by redefining the energy. in this definition. By (4. a† ] = kn a† . So. n n [Pz .9). Just as we say that photons n are the quanta of electromagnetic waves. E= n Nn ωn . we can easily see that [Pz .4. the zero-point energy. n (4. and energy by n units of ωn . We will make a guess – and confirm later – that the momentum operator in the z-direction is Pz = n kn a† an . so that we recover the de Broglie relation P = k. n In practice. This has the interpretation of looking like a state with Nn particles (or phonons) with energy ωn . H |0 = EZP = 1 ωn . or. Hence. this is not the probability amplitude. Notice that in this expression for the energy. This infinity can be “taken care of”. and an destroys a particle. as opposed to absolute values. but. as waves carry momentum. the action of a† upon a system. The states |ψ are the probability amplitudes. we have an infinity. The energy eigenvalues are then given by ∞ 51 E ({Nn }) = n=−∞ Nn + 1 2 ωn . . the energy is just the sum over the occupation number of each mode. is to create a massless particle with n ωn = kn .

1. ∂t (4.5). So. rather than anti-commutators) are bosons. ∂φ . we have the (verifiable) L ∂φ ∂φ † kn an an = − dz .1. a† ] = 0. by inspection. this is just the invariant Klein-Gordon equation. kn . we can easily verify that [Pz . Hence. we thus have [H. for example. n 2ωn L where we have now promoted the coefficients to operators.9). . Pz = ∂t ∂z 0 n and. Following from the commutators (4. |kn . . with identical properties kn . n 2L and therefore. ωn . ∂z (4.15) ∂ 2φ ∂ 2φ = 2. the states are symmetric under interchange of particles. That is.13).2). ∂t2 ∂z where φ is now a Lorentz scalar. m n n m where the second equality follows from the commutator [a† . the operator is time dependent.14) which is the Heisenberg equation of motion. φ(t. for a massless particle. so that Nn = 0. since. 2. Therefore. the field φ is a field operator. (4. φ] = −i ∂φ . Infact. φ] = i Now. we have H in terms of φ. Hence. φ] = n ωn −an e−i(ωn t−kn z) + a† ei(ωn t−kn z) . . consider some wave equation. it all stems from the usage of commutators. but the states |{Nn } are independent of time. is allowed. but states do not. . we are in the Heisenberg picture of quantum mechanics – operators evolve. 4. using the commutators (4. km = a† a† |0 = a† a† |0 = |km .2 Field Operators ∞ Recall (4. φ = ∂µ ∂ µ φ = 0. and using a unit speed of light. Hence. the particles described by this formalism (infact. [H. We can also place n m any number of particles in a given state. z) = √ n=−∞ 1 an e−i(ωn t−kn z) + a† ei(ωn t−kn z) .52 4 QUANTUM FIELDS Finally.

Pµ ]. ∂z ∂φ = [φ.2 The Klein-Gordon Field 53 in (1 + 1)-dimensions. ∂t to hold. The momentum and energy are given by p= 2π (nx . what about the Hamiltonian? The essential thing is that we require the Heisenberg equation of motion. px = pµ xµ = Et − p · x.1). We shall work in a box. H] = −[H. c† are operators. φ]. So. . with volume V = L3 . We have that the coefficients a. so that δpp = δp1 p1 δp2 p2 δp3 p3 .2 The Klein-Gordon Field Let us consider quantising fields that satisfy the Klein-Gordon equation. ny . at least). Hence. H]. (4. So. as we do not want φ = φ† (in general. ∂φ i = [φ.14) and (4. Consider the Hamiltonian for a 1D classical massless real string.16) which is in accordance with (4. ∂t i ∂φ = −[φ. i ∂φ = [φ.17) We have that a(p) = c(p). −z). ∂xµ (4. (4. 4. the normal modes are discrete (as a finite sized box). so that the field operator is φ(x) = p 1 a(p)e−ipx + c† (p)eipx 2V Ep (4. We shall consider complex scalar fields. and the invariant generalisation of the Heisenberg equation of motion is i that is. let us say that H = E must then be a component of the energy-momentum vector P µ = (H.4. and are of the form e±ipx . L H= 0 dz 1 2 ∂φ ∂t 2 + 1 2 ∂φ ∂z 2 . L Ep ≡ E(p) = p2 + m2 . and we assume the usual commutation relations a(p). nz ) . Pz ). Pz ].15). All other commutators vanish. c† (p ) .18) This holds on a component-by-component basis. a† (p ) = δpp = c(p). where xµ = (t. z) and xµ = (t. H] . c. a† . it has two components. after noting that [φ. under periodic boundaries.

54 4 QUANTUM FIELDS Notice that we have appropriately chosen units to make the propagation speed unity. the prefactor of φ† φ). is simple.21) .p pi pi 2V Ep Ep a† (p)a(p )eipx e−ip x −a† (p)c† (p )eipx eip x −c(p)a(p )e−ipx e−ip x +c(p)c† (p )e−ipx eip x . The generalisation of this.e.20) d3 x φ† · ( φ) = p p2 a† (p)c† (−p) + c(p)a(−p) Ep E−p .−p .19) Notice where the “mass term” comes in (i. (4. also notice that there is no factor of one-half (this is because we treat φ and φ† independently). So. H= d3 x ∂φ† ∂φ + ( φ)† · ( φ) + m2 φ† φ . to a massive complex 3D field. (4. We then expand this Hamiltonian by inserting our field operator (4. and hence. we have ∂φ† ∂φ = ∂t ∂t Ep Ep 2V Ep Ep a† (p)a(p )eipx e−ip x −a† (p)c† (p )eipx eip x −c(p)a(p )e−ipx e−ip x +c(p)c† (p )e−ipx eip x .p d3 xeipx eip x = V δp. ∂t ∂t (4. φ† · ( φ) = p.17). using the orthogonality condition. p. We will make use of the orthogonality relations d3 xe−ipx eip x = V δpp . d3 x ∂φ† ∂φ = ∂t ∂t Ep † a (p)a(p) + c(p)c† (p) 2 Ep E−p a† (p)c† (−p) + c(p)a(−p) 2 Ep E−p p a† (p)a(p) + c(p)c† (p) Ep 2 p − 1 2 + .

except H is replaced by P and Ep by p. Hence. and we have a set of commutators exactly like (4. H.22).4.24). [H. c† (p) = E(p)c† (p).23) Thus.2 The Klein-Gordon Field We can also compute that d3 x m2 φ† φ = 1 2 + Now. (4.18).20). a† (p) = E(p)a† (p). a(p)] = −E(p)a(p). (4. we have that the 4-momentum is P µ = (H. That is. we can write that the Hamiltonian is H = = p d3 x ∂φ† ∂φ + ( φ)† · ( φ) + m2 φ† φ ∂t ∂t E p a† a + c † c + 1 . where P= p p a† a + c † c . where in the last step we used the commutator (4. this is like two infinite sets of simple harmonic oscillators – this corresponds to the two degrees of freedom in a complex field. [H.21) and (4. Hence. . the Hamiltonian is H = p E(p) a† (p)a(p) + c† (p)c(p) + 1 . we have a factor of Ep which cancels from the second term of (4. a† (p) = pa† (p). is consistent with the KG equation of motion. P). with the assumed commutators. using these last two results. By covariance of the theory. We shall ignore the zero-point energy in what follows. And hence. H] ∂t holds.24) which are similar to (4.22) 1 a† (p)c† (−p) + c(p)a(−p) Ep E−p which can be used to see that upon adding (4. we can use these to show that i ∂φ = [φ.9). H. using (4. our formalism of a quantised complex scalar field. Thus. for example P. we can then easily compute the following commutation relations. (4. c(p)] = −E(p)c(p).18). notice that Ep = p2 + m2 ⇒ Ep = E−p . m2 p 55 1 a† (p)a(p) + c(p)c† (p) Ep .

p) and mass m. we may ask. a(p)] = −qa(p). what is the difference between a and c. we require non-Hermitian fields.26) We can then insert the fields. Q = = iq d3 x j 0 (x) d3 x φ† ∂φ ∂φ† − φ . we see that the first term is the number operator for particles with charge q. Hence. c (p) = −qc (p).27) Hence. and correspondingly. c† decrease the charge by q. given (4. (4. c destroys anti-particles with charge Q = −q. Now. what we see. and a. consider the conserved 4-current for the KG equation. with 4-momentum (E(p).e no c(p)-type). is that a† (p) and c† (p) both create spin-0 particles. and expand just as we did for the Hamiltonian (and we still ignore the vacuum energy part). It is interesting to notice that if the field φ were Hermitian. with Hermitian fields being uncharged. to have charge. p) and mass m. a† (p) = qa† (p).56 4 QUANTUM FIELDS So. † † [Q. ∂t ∂t (4. where we have merely inserted a constant q. Q. Then. . and the second term is the number operator for particles with charge −q. j µ (x) = iq φ† (∂ µ φ) − ∂ µ φ† φ . putting this all together. To answer this. Hence. we say that a† creates particles with charge Q = +q. corresponding to this. c(p)] = qc(p). [Q. a(p) and c(p) both destroy spin-0 particles.28) Therefore. We can now use this conserved charge to construct a set of commutators (which we derive either directly or by analogy – they are effectively trivial to see.24)). and we find Q=q p a† (p)a(p) − c† (p)c(p) .25) (4. so that we only had a(p)-type coefficients (i. then the charge Q is zero. c† creates anti-particles with charge Q = −q. Q. with 4-momentum (E(p). is a conserved charge. (4. and a destroys particles with charge Q = +q. we can see that a† and c increase the charge by q.

the final terms which need integrating. unless they are at the same place. x ) = −iδ (3) (x − x ). p which one can simplify using commutation relations. the charge Q is found by multiplying the above by iq.p (4. one can compute the equal-time commutator. that the format of an operator.3 The Dirac Field In deriving the charge operator Q. x). we also want fields that allow particles . these infact cancel due to the symmetry of the summation. are the second and third terms of both (4. when we get the modes to have destruction on the right and creation on the left.3 The Dirac Field In our previous discussion on the Klein-Gordon field. ∂φ (t.30). to be inkeeping with the literature. 57 p.29) and (4. integrating the first and fourth terms of both expressions (where the second expression is subtracted from the first). It is worth noting.30) Then. one computes the quantities φ† ∂φ = −i ∂t Ep 2V Ep Ep a† (p)a(p )eipx e−ip x −a† (p)c† (p )eipx eip x +c(p)a(p )e−ipx e−ip x −c(p)c† (p )e−ipx eip x . Then. is called “time ordered”.4. we looked at fields which permit particles (and anti-particles) which are spinless. to −i p a† (p)a(p) − c† (p)c(p) − 1 . In a very similar fashion to the above. φ(t.29) p. ∂φ† φ=i ∂t Ep 2V Ep Ep a† (p)a(p )eipx e−ip x +a† (p)c† (p )eipx eip x −c(p)a(p )e−ipx e−ip x −c(p)c† (p )e−ipx eip x .p (4. one finds − i 2 a† (p)a(p) + a† (p)a(p) − c(p)c† (p) − c(p)c† (p) . Naturally. Hence. 4. ∂t which is essentially the statement that one cannot know the position and velocity of a field at the same time.

1 Fermions Suppose we have some operators b† and bn which create and destroy spin. b† = 0. n n so that two fermions cannot be created in the same state. n m (4. but are consistent with the anticommutation relation. we require |n. bn |0 b† |0 m † † bn bn |0 = 0. We use these anticommutation relations instead of the commutation relations we used for bosonic fields. n . in n = p. in the n 2 state n = p. . this and (4. So.2 The Complex Dirac Field ψ(x) Recall that the Dirac equation reads i ∂ψ = (−iα · ∂t + βm) ψ. b† = δnm . Therefore. n = b† b† |0 = 0. by the Pauli exclusion principle (which states that no two fermions can be in the same quantum state. 4. the obvious choice is some Dirac field (given that the Dirac equation permitted particles with spin). = |n . a given value of momentum and spin). m {bn .e. ∀n. bn . = |n. using anticommutation relations. s (i. and see if the answers are consistent. n is a two-particle state. b† n n We also require that = b† b† + b† b† = 0. n n n n (4.33) for the creation and destruction operators. 4. n = −b† b† |0 .31) mean that the b† operators are inconsistent with the usual commutation n relation [b† . That is. we must ask: does this work? To answer this.1 fermions. and that fermions are anti-symmetric). Hence.32) |n.58 4 QUANTUM FIELDS which have spin.3. b† ] = 0. Now. this suggests that we should quantise fermionic fields using anticommutation relations.31) where |n is a single-particle state. n = − |n . we shall need to quantise a theory. s and |n. n n b† . n n Now.3. bm } = b† . (4.

p 1 bs (p)us (p)e−ipx + d† (p)vs (p)eipx . (4.34) into this Hamiltonian.37) Notice that this Hamiltonian is just the expectation value of the thing we “used to” call the Hamiltonian (i. Now we have a field operator. s s (4. we specify the anticommutation relations (4. alternatively. d† (p ) . s s with all other anticommutators vanishing. in this case.36). with destruction operator bs . one finds that H= s. 2Ep V with the spinors being orthogonal. H]. 2Ep V which follows from (3. and a set of anticommutation relations. ψ(x) = s. upon substitution of the field (4. we s must specify a set of commutation relations. ∂t + βm. i ∂ψ = [ψ. is to integrate this over all space. is the energy). So.33). s 59 vs (p)eipx .3 The Dirac Field and has plane wave solutions. with creation operator d† . as we want to describe fermions.p E(p) b† (p)bs (p) + d† (p)ds (p) .4. b† (p ) = δss δpp = ds (p). Hence. we suppose that the Hamiltonian in the quantum Dirac field theory is H= d3 x ψ † (−iα · + βm) ψ. (4. s 2Ep V (4. the general solution of the Dirac equation can be written as a sum over modes. and the second term being a negative frequency solution.35) The Hamiltonian for the Dirac equation was H = −iα · Hence. To quantise this field operator properly. frequency – solution. a good “guess”. so that bs (p). Therefore. † u† (p)us (p) = 2Ep δss = vs (p)vs (p). Specifically. we want to check if the Hamiltonian (after being expanded in terms of the field operators modes) satisfies the Heisenberg equation of motion.34) The first term (will have) the interpretation of being a positive energy – or.38) .e.36) (4. us (p)e−ipx .

we can compute the commutator of this charge operator with the creation operators.p b† (p)bs (p) − d† (p)ds (p) . C} − {A. Hence. iψ † and integrating. is to recall that the Dirac equation is i ∂ψ = (−iα · ∂t ∂ψ = ψ † (−iα · ∂t + βm) ψ. satisfies the Heisenberg equation of motion (4. d† (p) = E(p)d† (p). and s s bs (p).60 4 QUANTUM FIELDS after ignoring the vacuum energy. if we insert our field expansion (4.36). we find that Q=q s. ∂ψ = d3 xψ † (−iα · + βm) ψ. b† (p) = E(p)b† (p).37). (4. s s H. Now. b† (p) = qb† (p). these can be used to show that the Hamiltonian (4. so that upon contraction with ψ † from the left.39) Hence.34).34). bs (p)] = −E(p)bs (p). one finds H. s s [H. ds (p) destroying particles of energy E and momentum p. d† (p) = −qd† (p). recall that the conserved charge for the Dirac equation was ρ = ψ † (x)ψ(x). Therefore. we now see that we have the interpretation of b† (p). Q. i d3 xψ † We can use the easily provable identity [AB. d† (p) creating particles of energy E and momentum p. the conserved charge operator for the quantised Dirac field is Q=q d3 x ψ † (x)ψ(x). Q. + βm) ψ. C}B to find the commutators of the Hamiltonian with the creation and destruction operators (to infact verify their interpretation). Then. Then. s s s s . and it is easier to compute the LHS. [H. s s (4. in order that the RHS is true.40) And hence. The easiest way to show that this is the Hamiltonian. ∂t we have on the RHS the Hamiltonian (4. Thus. where we have merely inserted a multiplicative charge q. C] = A{B.37) with the field quantised as (4. ds (p)] = −E(p)ds (p). the LHS must also be true.

s s Therefore. energy E and momentum p. and bs (p) destroys that particle. what the Dyson product does. Now.42) (4.43) (4.p 2V 1 Ep Ep a(p)a† (p )e−ipx + a(p)c(p )e−ipx +c† (p)a† (p )eipx + c† (p)c(p )eipx . and φ† (x) creates particles. φ(x)φ† (0) = p. 4. Hence. we still need a way of propagating from one place to another. s have energies E(p) = + p2 + m2 .41) Basically. . if t < 0.1 particle with charge Q = s 2 +q. That is. t < 0. The crucial quantity which does this is the Feynman propagator. notice that neither the particle or anti-particle have negative energies any more. It is worth noting again that the vacuum state (which is the state with lowest energy) satisfies bs (p) |0 = ds (p) |0 = 0. relative to the vacuum. energy E and momentum p. where the Dyson (also called the time-ordered) product is T φ(x)φ† (0) = θ(t)φ(x)φ† (0) + θ(−t)φ† (0)φ(x). then GF represents the creation of a particle at t = 0 and subsequent destruction at time t. we have the interpretation that b† (p) creates a spin. We also have that d† (p) s creates a spin. p. φ(x) destroys particles. And also that single particle states b† (p) |0 . We define it for the complex KG field. ∀s.4 The Feynman Propagator GF (x) So far we have operators which create and destroy particles. This allows us to see what the difference is in the particles that b† and d† create. then an antiparticle is created at t and destroyed at time t = 0. GF (x) = −i 0| T φ(x)φ† (0) |0 . is to ensure that particles are created before they are annihilated. s d† (p) |0 . but. if t > 0.1 antiparticle with charge Q = −q. which we summarise below. and ds (p) 2 destroys that antiparticle.4 The Feynman Propagator GF (x) 61 and obvious equivalents for the destruction operators. Similarly. and the θ-function is defined θ(t) = 1 0 t > 0. it ensures causality is not violated.4. (4.

The arrow is sometimes emitted for bosons. The direction of the arrow only denotes whether the moving object is a particle or antiparticle.1: Representations of the Feynman propagator. (4.p 2V 2V 1 0| a(p)a† (p ) |0 e−ipx Ep Ep 1 δpp e−ipx Ep Ep = p. t < 0. . 2V Ep Hence. We can write this using the θ-function as GF (x) = −i p θ(t) eipx e−ipx + θ(−t) 2Ep V 2Ep V . 0| φ(x)φ† (0) |0 = p. 2V Ep eipx . (4.41) in its two limits is GF = −i p e−ipx . 2V Ep t > 0.46) These can be represented by diagrams. 2V Ep Similarly. x O x t x t x O t>0 t<0 Figure 4.p = p e−ipx . 0| φ† (0)φ(x) |0 = p eipx .62 4 QUANTUM FIELDS so that the only term to contribute upon contracting with the vacuum state is the first.45) GF = −i p where px = pµ xµ = Et − p · x.44) (4. see f4. we can use these to see that the Feynman propagator (4.1.

−i 0| θ(−t)φ† (0) so that their sum is just the Feynman propagator (4. using dθ(t) = δ(t). −i 0| θ(t) ∂t ∂t Now. Now. dt (4. dt So. ∂ 2 GF ∂φ(x) † ∂ 2 φ(x) † = −i 0| δ(t) φ (0) |0 − i 0| θ(t) φ (0) |0 ∂t2 ∂t ∂t2 ∂φ(x) ∂ 2 φ(x) +i 0| δ(t)φ† (0) |0 − i 0| θ(−t)φ† (0) |0 ∂t ∂t2 ∂φ(x) † ∂ 2 φ(x) † = −iδ(t) 0| . Hence. the first term is zero as φ(t. ∂GF (x) ∂φ(x) † ∂φ(x) = −i 0| θ(t) φ (0) + θ(−t)φ† (0) |0 . Hence. which we can use in the last two terms. φ† (t. . recall the KG equation.41). − m2 ) 0| θ(−t)φ+ (0)φ(x) |0 . Let us differentiate the propagator (4.47) − m2 )φ.4 The Feynman Propagator GF (x) 63 Another way of evaluating the propagator is as a single term.41) with respect to t. ∂ 2 GF ∂φ(x) † = −iδ(t) 0| . −i 0| θ(−t)φ (0) ∂t2 where we have noticed the appearance of the commutator. ∂GF (x) = −iδ(t) 0| φ(x)φ† (0) − φ† (0)φ(x) |0 ∂t ∂φ(x) ∂φ(x) † φ (0) + θ(−t)φ† (0) |0 . ∂ 2φ =( ∂t2 ∂ 2 φ(x) † φ (0) |0 ∂t2 ∂ 2 φ(x) |0 ∂t2 2 dθ(−t) = −δ(t). x). explicitly. φ (0) |0 + ( 2 ∂t ∂t 2 − m2 )GF . x ) = 0. −i 0| θ(t) = −i 0| θ(t)( = −i( = −i( 2 2 2 − m2 )φ(x)φ+ (0) |0 − m2 ) 0| θ(t)φ(x)φ+ (0) |0 .4. φ (0) |0 − i 0| θ(t) φ (0) |0 ∂t ∂t2 ∂ 2 φ(x) † |0 . ∂t ∂t ∂t Let us differentiate again. So.

so that upon substitution of these transforms into (4.50) by Fourier transforms. we have d4 k ˜ −k 2 + m2 e−ikx GF (k) = − (2π)4 d4 k −ikx e . φ (0) |0 . x) † . . This can also be derived by differentiating (4. upon substitution of this into (4. φ (0) |0 . x ) = −iδ (3) (x − x ). To find the Green function GF (x). ∂t We can show. So. where ρ(x) is an arbitrary source. we have GF (x) = δ (4) (x) = d4 k −ikx ˜ e GF (k) (2π)4 d4 k −ikx e . That is.48) 2 + m2 GF = −iδ(t) 0| ∂φ(x) † . and its solution is the Feynman propagator.51) noting that a δ-function can be written as the Fourier tranform of unity.46). (2π)4 Thus. we solve (4.48). One should recall that such solutions are Green functions. as the integrals are identical.49) This is a point-source Klein-Gordon equation. we see that ∂ µ ∂µ + m2 GF (x) = −δ (4) (x). taking the far right term over to the LHS. ∂t Hence. ˜ GF (k) −k 2 + m2 = −1. ∂t (4.64 4 QUANTUM FIELDS Therefore. Now. (4.50) (4. ∂ µ ∂µ e−ikx = −k 2 e−ikx . we equate the integrands. (2π)4 (4. we see that we have ∂2 − ∂t2 which is just ∂ µ ∂µ + m2 GF = −iδ(t) 0| ∂φ(x) † . and that Green functions can be integrated over a charge distribution to give the solution for the inhomogeneous Klein-Gordon equation. we can use the Green function GF (which is the solution to the above point-source equation) to solve ∂ µ ∂µ + m2 φ(x) = −ρ(x). by simply expanding in terms of modes. φ (t.50). that ∂φ(t.

we use the Feynman prescription. we only have a pole at k 0 = ωk − i . e−ik t . 65 1 . for our contour which is closed in the lower-half plane. and let the radius of the semi-circle go to infinity.51). . (4. This pole is on the real axis. this expression has a singularity at k 2 = m2 . However.52) This then gives us an expression for the Green function.4 The Feynman Propagator GF (x) and hence. √ singularity : k 0 = ±ωk ≡ ± k2 + m2 . we see the complex k 0 plane. with poles. we close the contour in the lower half plane. We separate the cases t < 0 and t > 0. Thus. and that we have set = 0. (2π)4 k 2 − m2 + i > 0. d4 k e−ikx . and at the end of any calculations. by integrating over it. with poles at k 0 = ± (ωk − i ) . With reference to Figure (4.54) is e−iωk t I = −2πi . Therefore. we modify the Green function (4. ˜ GF (k) = GF (x) = k 2 ≡ k µ kµ . There is no contribution from the semi-circle.53) perform the integral. (2π)4 k 2 − m2 (4.54) by contour integration. then put everything back as it was. which is to displace the pole from the real axis. t > 0: Lower-half-plane In this case. once we have performed the integral. and the contour we shall integrate over. set component. is just 2πi multiplied by the sum of the residues enclosed by the path. dk 0 2 I= 2 (k ) − ωk + i −∞ ∞ 0 = 0. We evaluate the time- 0 ωk = √ k2 + m2 . we have an expression which allows us to solve any inhomogeneous KG equation. to get around this problem. k 2 − m2 We then substitute this back into (4. as k 0 < 0.2). over an (anti-clockwise) closed path. That is. do the integral. 2ωk Notice that the minus-sign is due to our contour being clockwise. We do the contour integral by noting that Cauchy’s theorem states that the integral of any analytic function.52) into GF (x) = d4 k e−ikx . (4. the value of the integral (4.4.

this is the same as the causal result.44).1 Interactions The Interaction Picture So far. we can write down (4. in k 0 -space. (4.2: The complex plane.53). using these results.54) over the contour shown. 4. only now we have moved to continuous momentum. t > 0. i (2π)3 i GF (x) = − (2π)3 GF (x) = − e−i(Et−p·x) . we have two ways of describing the time evolution of a quantum system: the Schrodinger and Heisenberg pictures. we close the contour in the upper-half-plane. in the two regimes. We shall leave propagators for the moment. Thus.5 4. Hence. This is the t > 0 case.56) Notice that these are the same as (4.66 Im(k 0 ) −ωk + i 4 QUANTUM FIELDS ωk − i Re(k 0) Figure 4. only the states evolve and not the operators. 2Ep ei(Et−p·x) d3 p . with the poles marked on. so that the only contributing pole is at k 0 = −ωk + i . We evaluate (4.5. where the states obey i ∂ |ψ(t) ∂t S = H|ψ(t) S .e. t < 0: Upper-half-plane In this case.45).55) (4. In the Schrodinger picture. 2Ep d3 p (4. t < 0. and discuss interactions. . when we discuss exchange bosons). We shall come back to them when we need a term which allows the creation and destruction of a particle within a process (i.

and redefine a few terms. in the interaction picture. To define the picture. we see that by writing (4. S ψ(t)| AS |ψ(t) S =S ψ(0)| eiHt AS e−iHt |ψ(0) S . we use a third picture. and all time evolution is carried out by the operators. differentiating.59) If we have that H = H0 . −i dAH = [H. our free-field results in the Heisenberg picture hold for interacting fields. Hence. Hence. we can use this to find the expectation value of an operator. AH (t) ≡ eiHt AS e−iHt . the interaction picture. we write the Hamiltonian as a sum of a free Hamiltonian and an interaction Hamiltonian.60) (4. If H = H0 + HI . (4. Simply by differentiating the second of (4. (4. Hence. H = H0 + HI . Thus. What we have done. the state is constant. then the Heisenberg and interaction pictures are identical. AI (t)] . dt (4. |ψ H ≡ |ψ(0) S . we get i where HI (t) = e−iH0 t HI.57) so that the expected value becomes just S ψ(t)| AS |ψ(t) S =H ψ| AH (t)|ψ H .61) AI (t) = eiH0 t AS e−iH0 t . then we have an intermediate picture in which operators (which includes field operators) evolve according to the free Hamiltonian H0 .57).S eiH0 t . (4. dt −i dAI = [H0 . AH (t)] .58) For perturbation theory.57). We can take this.62) d |ψ(t) I = HI (t)|ψ(t) I . dt (4. we can derive the Heisenberg equation of motion. . We also define the interaction picture states and operators as |ψ(t) I = eiH0 t |ψ(t) S .4.5 Interactions The formal solution to this is just |ψ(t) S 67 = e−iHt |ψ(0) S . In the Heisenberg picture. we have preserved the form of the expectation value of an operator. is to change the object that is evolving in time: from the state to the operator.

2 2 τ →∞ (4. |ψ(∞) I = |ψf I = i Sfi |ψi I .64). and is a complicated state of interacting particles. One can think about the S-matrix as a “mixing matrix”. − τ ) = U (∞.66) (4.5. which is a free-particle state (i. At finite t. The initial state (i. We define the S-matrix via S = lim U ( τ . and summarise the physics of a set of interactions. (4. t0 ) = HI (t)U (t. where i d U (t. is just |ψ(∞) I = S|ψt=−∞ I . −∞). t0 ) = 1. at t = −∞) is just |ψ(−∞) I = |ψi I .e. as we have stated that the S-matrix ends a long time after the particles have finished interacting. at t = ∞).65) These matrix elements are the only observables of the system. after using the definition of the S-matrix. dt (4.e.68 4. it is useful to write |ψ(t) I = U (t. which occurs with probability Pf i = |Sfi |2 . which describes the transitions from the initial state that took place to give the final state. we can easily see that the amplitude to end up in the final state f is Sfi = ψf | S |ψi . Those familiar with statistical physics can think about this matrix as being like the Markov chain transition matrix.64) Hence. . we can write the final state as a linear sum of free particle states.e. The matrix elements Sfi describe the amplitude for a particular initial and final state. we start long before particles get close enough to interact. The final state (i. Thus. t0 ). t0 ) |ψ(t0 ) . We can write the final state like this. the state looks like |ψ(t) I = U (t.2 The S-matrix 4 QUANTUM FIELDS For time evolution. (4. an eigenfunction of the free Hamiltonian H0 ). −∞)|ψi I . and finish longer after particles have finished interacting. Hence.63) under the boundary condition U (t0 .

t0 ) = U (t0 .68) under the boundary condition U (t0 . by inserting the formal solution into itself. is that the thing we are trying to solve for. Now. we can write |ψ(t) I S 69 = = = = = eiH0 t |ψ(t) S eiH0 t e−iHt |ψ(t0 ) S e−H0 t e−iHt eiHt0 |ψ(t0 ) eiH0 t e−iH(t−t0 ) |ψ(t0 ) I U (t.5. (4. Hence. The consequence of this. U −1 (t.3 The S-matrix Expansion Now. appears within the integral itself. t0 ). For example. Pf i = f f (4.5 Interactions The Unitarity of the S-matrix Recall that |ψ(t) I = eiH0 t |ψ(t) from the first of (4. taking the limit t → ∞ and t0 → −∞.67) | ψf | S |ψi |2 ψi | S † |ψf f = = = ψf | S |ψi ψi | S † S |ψi ψi |ψi . we can form an iterative solution. t0 ) = 1. we arrive at the statement that the S-matrix is unitary. Hence. The reason we say “formal”.4. is that probabilities are conserved. t t U (t. we see that if a state is normalised. U (t. t0 )|ψ(t0 ) I . from (4. S † S = 1. I Now. we have the formal solution t U (t. t) = U † . t0 ). we thus have that U is unitary. t0 ) = 1 − i t0 dt1 HI (t1 ) + (−i)2 t0 dt1 HI (t1 ) t1 t0 dt2 HI (t2 )U (t2 .60). as H and H0 are both Hermitian. The third equality follows from the second due to the completeness of eigenfunctions. 4. then the normalisation is preserved. Hence. t0 ) = 1 − i t0 dt1 HI (t1 )U (t1 .63). t0 ). .

HI (tn )} . t1 > t2 t2 > t1 ⇒ ⇒ HI (t1 )HI (t2 ). the expression still is not explicitly covariant.70) Now. and t0 → −∞. −∞) = S. .71) We have set the upper-limit on the second integral to ∞ as we notice that the whole t1 − t2 plane will be covered by such a Dyson product. S (2) (−i)2 = 2 (−i)2 dt2 HI (t2 ) + dt1 HI (t1 ) 2 −∞ −∞ ∞ t1 ∞ dt2 HI (t2 ) −∞ t2 −∞ dt1 HI (t1 ). if we take t → ∞. HI (t) = d3 x HI . the “earlier” Hamiltonian appears to the right of the “later” Hamiltonian. .69) t1 −∞ S (2) = (−i)2 −∞ dt1 HI (t1 ) dt2 HI (t2 ). ∞ S (1) = −i −∞ dt1 HI (t1 ). . −∞ −∞ dtn T {HI (t1 ) . We now note that in field theory the Hamiltonian H is the spatial integral of the Hamiltonian density H. U (∞. We can then notice that in the both terms. (4.70 We can imagine doing this iterative insertion many times. . S (n) (−i)n = n! ∞ ∞ dt1 . and for example. as the case was before we wrote the symmetric sum. We can notice that in the first term. S = S (0) + S (1) + S (2) + . for the nth -order term. . . t0 ) = 1 − i t0 dt1 HI (t1 ) + (−i) 2 t0 dt1 HI (t1 ) t1 t0 dt2 HI (t2 ) + . So. by definition. where S (0) = 1 denotes no interaction. it is not covariant: we have singled out the time coordinate. t2 < t1 and in the second term. Hence. More generally. this becomes the expansion of the S-matrix. . . t t 4 QUANTUM FIELDS U (t. (4. ∞ (4. we have an expression which is of the time-ordered form (or Dyson product). (4. Hence. Let us rewrite S (2) in a more symmetric form by interchanging t1 and t2 . HI (t2 )HI (t1 ). we write this as a sum of terms of incrementing orders. hence. . . t1 < t2 .72) This now does not single out a specific time value. . S (2) = (−i)2 2 ∞ ∞ dt1 −∞ −∞ dt2 T {HI (t1 )HI (t2 )} . as the expression stands. but does single out time from space.

Hence. .4. . rather than the Hamiltonian. HI (xn )} .70) can be written in a covariant form. .69) and (4. . d4 xn T {HI (x1 ) . d 4 x1 d4 x2 T {HI (x1 )HI (x2 )} . HI (x) = Hi (φi . 71 Finally.75) d 4 x1 . using the Hamiltonian density. ∂µ φi ) . S (1) = (−i) S (2) = S (n) (−i)2 2 (−i)n = n! d4 x HI (x). we have an explicitly covariant form of the S-matrix expansion. . the expressions (4. (4.5 Interactions where we may have multiple fields which are functions of the 4-position vector x.74) (4.73) (4.

72 4 QUANTUM FIELDS .

The dimensionless coupling parameter g is g = mτ G √ 2 1/2 . s τ− τ+ Figure 5. The choice of interaction term is effectively “the problem”. Thus. s ) + (k. (5. the decay can be written (p. s). we have that φ(x) is a Hermitian scalar field. in the standard model. (5.s H0 p k. k. G ≡ 1. Hence. As H 0 is an uncharged boson.73 5 Decays We shall illustrate calculations by considering examples.e. H 0 −→ τ + + τ − . the tauons τ ± are spin. we have that ψ(x) is a Dirac spinor field. with mass mτ ≈ 1.1) 1 The scalar uncharged Higgs boson H 0 has zero spin.8GeV/c2 . s). (5. (p.1: Schematic of the 2-body decay (5. fermions). ¯ Notice that ψ(x)ψ(x) is a Lorentz scalar density. 5.3) . as τ ± are charged particles with spin.1 Simple 2-body Decay: φ3 Let us consider an example of a Higgs boson decaying to two tauons. above. different interaction terms produce different physics. Now.2) where N denotes normal (or time) ordering. we are considering a φ3 -theory. the interaction is of the form ¯ HI (x) = gN ψ(x)ψ(x)φ(x) . a given interaction term is “the theory” we are investigating. In terms of momentum and spin.1). 0) −→ (k .16 × 10−5 GeV−2 .2 leptons (i.

each of these terms correspond to a different process. if we were to plug in the expressions for the field expansions. 1/2 p (2V ωp ) so that φ(+) destroys and φ(−) creates netural scalar bosons. The neutral scalar field has its expansion being φ(x) = φ(+) (x) + φ(−) (x) 1 = a(p)e−ipx + a† (p)eipx .74 5 DECAYS where G is Fermi’s coupling constant. which would denote particle creation. s | HI |p . τ + .s (2V Ep ) so that ψ (+) destroys τ − and ψ (−) creates τ + . As the field is a Dirac field. p2 + m2 . s. such as f = τ + τ − H 0 | HI (x) |i = 0 . The final state similarly is |f = τ − . to first order in HI . which denotes the amplitude for a decay of H 0 → τ + + τ − . s 1/2 p. We are interested therefore in 2 body decay : k. we also note that ¯ ¯ ψ (−) creates τ − . with given momenta. k. |i = H 0 . and ψ (+) destroys τ + . the matrix element is Sfi = −ig ¯ d4 x f | N ψ(x)ψ(x)φ(x) |i . we would have 8 terms. so that ¯ ¯ ¯ ψ(x) = ψ (−) (x) + ψ (+) (x). we need to evaluate the matrix element Sfi = −i (1) d4 x f = τ + τ − HI (x) i = H 0 . So. where ωp = ω(p) = The charged fermionic field expansion is ψ(x) = ψ (+) (x) + ψ (−) (x) 1 = bs (p)us (p)e−ipx + d† (p)vs (p)eipx . but will only . s. Now. k s = b† (k)d† (k ) |0 . where an over-bar denotes the Dirac adjoint. which we can create from the vacuum. k . p = a† (p) |0 . s s So. with momentum p. To first order. and expand out. our initial state is a single Higgs boson. In fact. Notice that other matrix elements are possible.

s| ψ (−) (x) = 1 (2V Ek ) 1 1/2 us (k)eikx 0| . 1 (2V ωp )1/2 1 a(p)e−ipx . s = 1 (2V Ek ) 1/2 us (k)e−ikx |0 .1 Simple 2-body Decay: φ3 75 give non-zero contribution if they correspond to the process we are currently dealing with (due to the orthonormality of states. Now. it is the adjoint of this term which appears in (5. the second creates a τ + and the third destroys a H 0 . The term on the far RHS is φ(+) (x) |i but. doing so. So.4). let us treat each field in turn. we can write a(p )a† (p) = a† (p)a(p )+δpp . φ(+) (x) |i = (2V ωp )1/2 p The first bracketed term gives zero upon action with the vacuum state. hence. 1 e−ip x a† (p)a(p ) + δpp |0 . and our specific choice of initial and final states). One can see that the only contributing term is Sfi = −ig ¯ d4 x f | ψ (−) (x)ψ (−) (x)φ(+) (x) |i . (5. and hence φ(+) (x) |i = p (2V ωp ) 1/2 a(p )e−ip x a† (p) |0 . (5. Now.5) In an entirely analogous manner. Hence.6) And the final term can be seen by analogy again. one can easily see that ψ (+) (x) |k. ¯ (5. k . ¯ k. The Kronecker-delta term simply filters out p = p in the summation to leave φ(+) (x) |i = 1 (2V ωp ) 1/2 e−ipx |0 . s | ψ (−) (x) = (2V Ek ) v 1/2 s (k )eik x 0| .5. taking the adjoint. (5. as a(p) |0 = 0.7) .4) where the first term creates a τ − . φ(+) (x) = p = φ(+) (x) |p = φ(+) (x)a† (p) |0 . by the usual commutation relation.

See Figure (5. we will have that d4 x eix(k+k −p) = V T δk+k . Now.3) to see the whole amplitude for a given graph. The final term Mfi is the only term that depends on the specific interaction.10). putting together (5. The second term with the delta-function is infact the statement of conservation of energy-momentum. we have the integral over space of this quantity: this will only have the effect on the exponential. and is where all dynamical information is kept. So. T f f .8) and (5. we see that Sfi = 1 (2V Ek ) 1/2 (5. we see that ¯ f | ψ (−) ψ (−) φ(+) |i = 1 1 1 (2V Ek )1/2 (2V Ek )1/2 (2V ωp )1/2 ×eix(k+k −p) us (k)vs (k ). (5.76 Hence. We can represent Mfi by a Feynman graph. putting together (5. (5. and will be common to all processes.9) 1 (2V Ek ) 1/2 1 (2V ωp ) 1/2 (2π)4 δ (4) (k + k − p)Mfi .2 Decay Rates & Lifetimes We now want to calculate the measured decay rate for a given process.8) where we have been careful to maintain the order of the spinors. in (5. u Let us now consider the three types of term in (5. If we are not in the limit of an infinite volume/time box.7). its structure is Sfi = normalisation × energy-momentum conservation × dynamics. as in Figure (5. The initial three factors involving energy are the normalisation.4).2) for the assignment of mathematical factors to specific features of a Feynman graph.4). ¯ 5 DECAYS (5.6) and (5.9) in (5.1).5). The probability of decay per unit time is |Sfi |2 Γ= Γf i = . d4 x eix(k+k −p) = (2π)4 δ (4) (k + k − p).10) where we have defined the Feynman amplitude Mfi = −ig¯s (k)vs (k ). See Figure (5. (5. and noting that 0|0 = 1.p .11) 5.

T → ∞.2: The assignment of factors to the elements of a Feynman graph. k. to allow us to have a meaning in squaring the delta-function. s Mf i = −ig¯s (k)vs (k ) u k . and the two phase-space terms 1 V −→ k d3 k .5.s 1 1 T 2 V 2 δk+k . Now we have squared the delta-function.3: The Feynman amplitude for the given graph.2 Decay Rates & Lifetimes 77 vertex factor −ig outgoing fermion us (k) ¯ outgoing antifermion vs (k) Figure 5.k . inserting (5. (2π)3 . 3ω E E T 8V p k k where we are in the finite volume/time limit.10). we can take the limit V. so that the energy-momentum conservation term becomes just V T δk+k . Hence.p |Mfi |2 .s Figure 5. we see that Γ= k.s.p −→ (2π)4 δ (4) (k + k − p).

Let us do the spin sum first. u Hence. u† (k)γ 0 v(k ) Therefore.s |¯s (k)vs (k )| . u 2 So. all factors that related to the volume and time of the box disappear. hence. this is just two scalars multiplied together. as they are implied by the argument of the spinor. and can hence be commuted at will. so that we see the Dirac adjoint appear again.s = v (k )u(k). u† (k)γ 0 v(k ) † † † ∗ † = v † (k )γ 0† u(k).s s. γ 0† = γ 0 . We must evaluate s. ¯ u ∗ = g2 where we drop the s. we can treat complex conjugation exactly as Hermitian conjugation. u v = g2 . to begin. as is the term directly before it).s ua (k)va (k )¯b (k )ub (k) ¯ v ub (k)¯a (k)va (k )¯b (k ). |Mfi |2 = s. ¯ Let us then write this with indices. we thus have (¯(k)v(k )) = u† (k)γ 0 v(k ) . † but. ¯ u(k)v(k )¯(k )u(k).s Mfi M∗ fi u(k)v(k ) (¯(k)v(k )) .78 5 DECAYS Thus. we need to evaluate the spin sum and do the phase space integrals. (¯(k)v(k )) = (¯(k)v(k )) . so that all terms involved are numbers. notice that the bracketed term is a scalar (indeed.12) To continue. Therefore. and swapping the order of all terms). Now.s |Mfi |2 = g 2 s. taking the Hermitian conjugate (which has the effect of conjugating everything. s . 3 2E (2π) k (5. |Mfi |2 = g 2 s. u u ¯ If we recall that the Dirac adjoint was defined to be A = A† γ 0 . |Mfi |2 = g 2 s. a term such as u(k)v(k ) is a scalar. leaving us with Γ= 1 2ωp d3 k (2π)3 2Ek d3 k (2π)4 δ (4) (k + k − p) |Mfi |2 . ¯ v Now.

/) where we first note that k µ is not a matrix. and secondly that the γ-matrices are traceless.5. due to the cyclicality of the trace. we see that |Mfi |2 = 4g 2 g µν kµ kν − m2 .14) To evaluate the first term. /k τ s. /k Hence. /k /) / τ (5. by noting Tr (k = Tr (γµ k µ ) = k µ Tr (γµ ) = 0. we write k / = γµ γν k µ k ν = γν γµ k ν k µ . Hence. Therefore.2 Decay Rates & Lifetimes We now use two identities of the spinors. Tr (k / ) = 4g µν kµ kν . ub (k)¯a (k) = (k + mτ )ba . τ s. we write Tr (γ µ γ ν ) + Tr (γ ν γ µ ) = 2g µν Tr (14 ).s 79 va (k )¯b (k ) = (k − mτ )ab . Tr (γ µ γ ν ) = g µν Tr (14 ) = 4g µν . γ ν } = 2g µν 14 (where we must remember that the LHS of this is a matrix equation. using these two results.13) reduces down to |Mfi |2 = g 2 Tr (k / ) − 4m2 . using this in (5.s . Hence. we see that (5. the LHS is just 2Tr (γ µ γ ν ). so that |Mfi |2 = g 2 Tr [(k + mτ ) (k − mτ )] / / s. Now.13) Now the traces of the middle two terms are zero.s = g 2 Tr (k / ) − mτ Tr (k + mτ Tr (k ) − m2 Tr (14 ) . /k We now use {γ µ . an expression with indices arranged so is just a trace. and hence so must the RHS – therefore the introduction of the unit matrix).14). v / Now. / / s. u / where k ≡ γµ k µ . and hence / |Mfi |2 = g 2 (k + mτ )ba (k − mτ )ab . Therefore.s (5. The last term in (5.13) is fairly obviously just Tr (14 ) = 4.

s 5 DECAYS |¯s (k)vs (k )| = 4g 2 kk − m2 . whilst evaluating the d3 k -integral. we see that m2 kk = H − m2 . Hence. 4Ek E−k We now use the second of (5. Hence. and we have also split the delta-function into its space and time parts. so that I= d3 k δ (E(k) + E(−k) − mH ) . E 2 (k) = k2 + m2 τ ⇒ k2 = E 2 (k) − m2 τ 2 mH = − m2 . Therefore.s |¯s (k)vs (k )| = 8g 2 u 2 m2 H − m2 . 2 4Ek .15) can be written kk = k µ kµ = E 2 (k) + k2 = 2k2 + m2 . noting that the kk term in (5. in the rest-frame of the Higgs.16) to write I= d3 k δ (2E(k) − mH ) . g2 s. 2 (5.s s. conservation of 4-momentum implies that k = −k . The δ (3) (k + k ) term. u τ 2 (5.80 and therefore. τ 2 which we can insert into (5. τ 4 E(k) = E(k ) = mH . we must compute d3 k d3 k I= δ (E(k) + E(k ) − mH ) δ 3 (k + k ). P µ = (mH . inserting (5.15) where kk = k µ kµ . τ 4 (5. 0).17) Therefore. 2Ek 2Ek where we have ignored the factors of 2π (to be put back in later).16) (5.12).15) to see that |Mfi |2 = g 2 s. So.18) Let us now do the phase-space integral (5.17) into this expression. filters out k = −k as the only non-zero contribution. We now specialise to the rest-frame of the Higgs boson. τ where we have taken careful note of the change in sign.

δ (f (x)) = x0 δ(x − x0 ) . using the second of (5. or. using this result and (5. τ 4 Hence. we use the usual relation for a functional argument. this cancels down to I= π|k0 | . mH k0 ≡ m2 H − m2 . we have that (5. . if we assume no angular dependence. I = = 4π|k|2 E(|k|) d|k| δ |k| − 2 4Ek 2|k| 4π|k0 | . Now. using these results.2 Decay Rates & Lifetimes and then E(k) = k2 + m2 .19) To evaluate the delta-function. τ 4 m2 H − m2 τ 4 Then. the argument of the delta-function in (5.19) reads I= d3 k E(|k|) δ |k| − 2 4Ek 2|k| m2 H − m2 τ 4 . |f (x0 )| f (x0 ) = 0. at k0 is f (k0 ) = 2k . 8E(k0 ) k0 ≡ m2 H − m2 .18). E(k) Therefore. τ 2 4Ek 81 (5.19) is zero when k = k0 = m2 H − m2 . τ 4 and the value of the derivative of the argument.20) − m2 τ πm2 H .12) becomes Γ = = 1 1 8g 2 2mH (2π)2 g2 m2 H 4 m2 H − m2 τ 4 3/2 π|k0 | mH (5. to see that τ I= d3 k δ 2 k2 + m2 − mH .16).5. we see that the decay rate (5.

to give the decay rate Γ H0 → + − = KmH m2 . ¯ q ¯ As the top t quark has the highest mass. 8π We can use g from (5. consider σ −→ φ− + φ+ p −→ k + k . 8 2π 5 DECAYS mτ (which is a (5. the dominant decay channel is H 0 → tt.82 after using ωp = mH in the rest-frame of the Higgs.21) Therefore. (5. the general decay rate Γ and the decay rate of a stationary particle.3) to see that Γ H 0 → τ + τ − = KmH m2 . Ep = p2 + m2 . g 2 mH Γ H 0 → τ +τ − = . of a neutral scalar particle decaying into two charged scalars. We will compute the matrix element Mfi . φ . 1/2 (2V Ep ) 1 b(k)e−ikx + d† (k)eikx . Now. 5. where the field expansions are σ(x) = p 1 a(p)e−ipx + a† (p)eipx . in the limit mH very good approximation). τ G K≡ √ . as there are 3 colour states that could be “decayed into”. So. we have found that the decay rate for the process H 0 → τ + τ − increases with the mass of the Higgs particle (infact. as there is a larger phase space to “decay into”). hence increasing the phase space further. 1/2 (2V ωk ) φ(x) = k and px = pµ xµ = Et − p · x. we have to multiply the answer by 3. τ − . We can somewhat generalise this decay rate to any lepton = e− . The interaction Hamiltonian we are considering is HI = gφ† (x)φ(x)σ(x). µ− . σ ωk = k2 + m2 . it will increase for the larger product mass as well.22) For quarks.3 Example: Decay of a Scalar to two Scalars Let us consider a second example. Γ H 0 → q q = 3KmH m2 .

|i = a† (p) |0 . The only non-zero terms in the field expansion can then be deduced by requiring those which “null out” those operators which are already present. so we use b and d† . as that will destroy the particle the already present a† (p)-term has created. Hence. so that Sfi = −i = d4 x g f | φ† φσ |i d4 x 1 1 1 eix(k+k −p) Mfi . we also need d† (k ). In contrast. to leave just g f | φ† φσ |i = g 1 1 1 eix(k+k −p) . |f = φ− . k. d4 x f | HI (x) |i . 1/2 (2V ω )1/2 (2V ω )1/2 (2V Ep ) k k . the φ-field is charged. φ+ . the integral of this is just Sfi . k . 1/2 (2V ω )1/2 (2V ω )1/2 (2V Ep ) k k Thus. using our interaction Hamiltonian. in which case we pick up a factor of eikx as well. we use a and a† ). We get the b† (k)-term by conjugating the field expansion. The term from σ will be a(p)e−ipx . Sfi = −i where the initial and final states are just |i = |σ. modeled by not being Hermitian. We can get d† (k ) simply from the field expansion as it stands (also picking up a factor of eik x ). we have (also inserting the factors of energy and volume we have neglected to mention thus far). we want to compute g f | φ† φσ |i = g 0| d(k )b(k)φ† φσa† (p) |0 . g f | φ† φσ |i = g 0| d(k )b(k)b† (k)d† (k )a(p)a† (p) |0 1 1 1 × 1/2 (2V ω )1/2 (2V ω )1/2 (2V Ep ) k k ×eix(k+k −p) . which is modeled by being Hermitian (i. The entire “state” part contracts to unity (by construction). These states can be created from the vacuum via application of the creation operators. |f = b† (k)d† (k ) |0 .3 Example: Decay of a Scalar to two Scalars 83 The important features of these expansions are: the σ-field is uncharged.and b† (k)-terms to create the particles that are destroyed by the already present d(k )b(k)-terms.e. We need to first compute Sfi . Then. p .5. Similarly.

T 2Ep (2V ωk ) (2V ωk ) . so that |Sfi |2 1 1 1 = V T δk+k . upon computing the square of the matrix element.p . we go into the finite-box limit. so that we can do the integral in Sfi .p )2 . (2π)4 δ (4) (k + k − p) 2 = V 2 T 2 (δk+k .84 where we have defined the Feynman amplitude to be Mfi ≡ −ig. Let us now compute the decay rate.p .p )2 = δk+k . 5 DECAYS This rather simple amplitude is just because every particle involved is scalar. to compute the square of the deltafunction). Hence. (2V Ep ) (2V ωk ) (2V ωk ) Let us cancel off one of the factors of V . We now use the property that d4 x eix(k+k −p) = (2π)4 δ (4) (k + k − p).p |Mfi |2 . we perform a summation over all possible final states. Hence. 2 = V 2 T 2 δk+k . |Sfi |2 . and divide by T .p |Mfi |2 . Sfi = 1 1 1 (2π)4 δ (4) (k + k − p)Mfi . we easily see that |Sfi |2 = 1 1 1 V 2 T 2 δk+k . (infact. To do so. 1/2 (2V ω )1/2 (2V ω )1/2 (2V Ep ) k k Hence. so that (2π)4 δ (4) (k + k − p) = V T δk+k . we have computed an expression for the matrix element for neutral scalar decay into charged scalar particles. Γ= T f To be able to compute |Sfi |2 . more precisely. and hence we only have a contribution from the vertex itself. (δk+k .p . squaring this: (2π)4 δ (4) (k + k − p) but So.

so that Ep = p2 + m2 = mσ . 3 2ω (2π)3 2ω (2π) k −k . so that Γ = f |Sfi |2 T 1 1 1 (2π)4 δ (4) (k + k − p)|Mfi |2 . This will allow us to compute the delta-function. The first d3 k -integral will only return a non-zero value if k = −k. we can always split up the 4-dimensional delta-function. We now specialise to the rest frame of the decaying particle. due to the 3-D deltafunction. 2Ep (2V ωk ) (2V ωk ) = k. we have Γ= 1 2Ep d3 k 1 (2π)4 δ(ωk + ω−k − Ep )|Mfi |2 . so that |Sfi |2 1 1 1 = (2π)4 δ (4) (k + k − p)|Mfi |2 . we have p = 0. thus δ (4) (k + k − p) = δ(ωk + ωk − Ep )δ (3) (k + k − p) . after doing this integral. Hence. that is. using this. (2π)3 2ωk This is a very general expression for the decay rate. In this frame. (2π)3 k Hence. Hence. In the rest frame of the decaying particle. Γ= 1 2Ep d3 k (2π)3 2ωk d3 k (2π)4 δ (4) (k + k − p)|Mfi |2 .p → (2π)4 δ (4) (k + k − p). the decay rate becomes Γ = 1 2Ep d3 k (2π)3 2ωk d3 k (2π)3 2ωk (2π)4 δ(ωk + ωk − Ep )δ (3) (k + k )|Mfi |2 . T 2Ep (2V ωk ) (2V ωk ) 85 To compute the decay rate. we must sum this expression over all possible final states. the second 3-D delta-function just becomes δ (3) (k+ k ). we will compute the decay rate of a σ-particle at rest. so that 1 −→ V d3 k .3 Example: Decay of a Scalar to two Scalars and we now send V T δk+k .k To perform the summation.5. σ Now. we take the infinite-box limit.

given our delta-function. Γ= We now make the integral spherically symmetric.86 By inspection of ωk = becomes just Inserting ωk just gives δ(2ωk − mσ ) = δ 2 k2 + m2 − mσ . using this. we can insert this expression for the delta-function in the decay rate. d|k| |k|2 4π δ (|k| − |k0 |) |Mfi |2 . φ 5 DECAYS k2 + m2 . Hence. Hence. only k = k0 will give a non-zero contribution. φ 4 2|k| . To do this. so that Γ = 1 ωk0 |k0 |2 π |Mfi |2 2ω2 2Ep 2|k0 | (2π) k0 1 |k0 | = |Mfi |2 . |f (x0 )| f (x0 ) = 0. so that upon integration. 3 4ω 2 (2π)3 (2π) k 2|k0 | 1 ω k0 2Ep 2|k0 | d3 k δ (|k| − |k0 |) |Mfi |2 . 2Ep 8πωk0 . we must get it into the form δ(k − k0 ). ωk ωk0 δ (|k| − |k0 |) . 2 4ω 2 (2π) k Let us rearrange slightly. we use the property δ(f (x)) = x0 δ(x − x0 ) . 2 4ω 2 (2π) k where integration will now just filter out |k| = |k0 |. we see that ω−k = ωk . we see that its argument is zero when |k| = |k0 | = We also see that f (|k|) = Hence. Γ= 1 ωk0 2Ep 2|k0 | d|k| |k|2 4π. the remaining delta-function φ δ(2ωk − Ep ) = δ(2ωk − mσ ). to see that δ 2 k2 + m2 − mσ = φ Γ= 1 2Ep d3 k 1 ωk (2π)4 0 δ (|k| − |k0 |) |Mfi |2 . 2|k0 | Therefore. so that d3 k −→ Hence. To make this delta-function useable. m2 σ − m2 .

φ 4 Due to the simple form of our Mfi = −ig. we can use this in the decay rate.5. each fermion has two degrees of freedom). we have computed that the complete expression for the decay rate of a scalar σ-particle to a pair of scalar charged φ-particles is Γ σ → φ+ φ− = g2 8πm2 σ m2 σ − m2 . we easily have |Mfi |2 = g 2 . m2 σ − m2 . so that Ep = mσ . τ πm2 4 σ We can now make a comparison of the two expressions. Secondly.23) Let us recall the equivalent expression we derived for a scalar decaying to two fermions. . Now. to see that Γ= 1 1 4πm2 2 σ m2 σ − m2 |Mfi |2 . the scalar decay rate has a factor 1/8 suppression relative to the fermionic. This is a general feature of fermionic decays. (5. This is because there are more fermionic states in the phase space to decay to. First.3 Example: Decay of a Scalar to two Scalars 87 We can tidy this expression up slightly. 0 φ where we defined k2 = 0 Hence. |k0 | 1 2 = 8πωk0 8π mσ Therefore. φ 4 ωk0 = Thus. 3/2 g2 m2 σ Γ(σ → τ + τ − ) = − m2 . One of the differences is that the bracketed factor is raised to a different power.20). by recalling that we are in the frame where the decaying particle is at rest. we note that the overall structure of the decay rate is the same. we have that 2 ωk0 = k2 + m2 . than in the scalar phase space (that is. Hence. 2 m2 σ − m2 . φ 4 mσ . φ 4 (5.

88 5 DECAYS .

s So that we can describe the process.s τ. which create electrons/tauons and destroy electrons/tauons. ψτ (x ). s2 ) −→ e− (p3 . S (2) = (−i)2 2 d4 x d4 x f | T (HI (x)HI (x )) |i . e. p1 . e. the matrix element to be computed is S (2) = (−i)2 d4 x d4 x ge gτ (6. s1 = (+) (+) ¯(−) ¯(−) × f | ψe (x)ψτ (x )ψe (x)ψτ (x )T (φ(x)φ(x )) |i . s1 ) + τ − (p2 . s4 ). Hence.s1 (p1 )e−ip1 x 2V E(p1 ) |0 . we have two types of lepton. (6. τ − and e− . ψe (x) and ψτ (x ).89 6 6. (6.s and the final state |f = b† 3 (p3 )b† 4 (p4 ) |0 . notice that we have electron destruction/creation at a different place to tauon destruction/creation.1) Hence. where we have multiplied by 2 as the choice of which is x and which x is arbitrary. Therefore. (+) ψe (x) e− . we obviously need two electron and two tauon field operators. to ¯ ¯ HI = gτ ψτ (x)ψτ (x)φ(x) + ge ψe (x)ψe (x)φ(x). s3 ) + τ − (p4 . So. we need the operators ψe (x).3) The fields act upon the initial and final states in an identical manner to our discussion on decays. so that for example. The lowest order term in the S-matrix expansion that can do this is the second order.1 Scattering Processes An Outline of Particle Exchange Let us extend our previous interaction Hamiltonian. the term we want is (+) (+) ¯(−) ¯(−) 2 f | ψe (x)ψτ (x )ψe (x)ψτ (x ) |i . which can scatter by boson exchange in the process e− (p1 . ue.s τ. . the initial state is just |i = b† 1 (p1 )b† 2 (p2 ) |0 .2) (+) (+) ¯(−) ¯(−) From this.

s2 τ− p4. denoting their interaction happens at a different place to the electrons interaction.s2 (p2 ) u u 4 × i=1 1 (2V Ei )1/2 d4 x d4 x eip3 x e−ip1 x eip4 x e−ip2 x (6. .s3 (p3 )ue. Therefore.1: Feynman graph for e− τ − → e− τ − scattering. These terms perhaps become a little more understandable if we consider Figure (6. s4 τ− Figure 6.s1 (p1 )(−igτ )¯τ. s3 × 0| T (φ(x)φ(x )) |0 .3).90 6 SCATTERING PROCESSES Then. The final term is the Feynman propagator.4). with an un-barred spinor on the far right.e. this Feynman propagator term describes the exchange of a “virtual” H 0 .1).s1 (p1 ) can be thought about as first destroying an electron (i. we end up with Sfi = (−ige )¯e. Notice that the exponential factors (and these are the only spatial terms) for the tauons have primes on them. First.4) e− p3. integrating over x only concerns the terms d4 xeip3 x e−ip1 x e−iqx = (2π)4 δ (4) (p3 − p1 − q) . e− p1 . The exponential factors eip3 x e−ip1 x can be thought about equivalently. 4 2 − m2 + i (2π) q H We now recall that we interpreted 0| T (φ(x)φ(x )) |0 as creating a particle at x and destroying at x .s3 (p3 )ue. The spinor factors ue. plugging all of the relevant terms into (6. ¯ a particle). can be written in a variety of ways: 0| T (φ(x)φ(x )) |0 = iGF (x − x ) = i d4 q −iq(x−x ) e . s1 p2. we can begin to do integrals.4). Substituting the propagator into (6. with a barred spinor. then an electron created after the destruction. which as we saw in Section (4.s4 (p4 )uτ.

d4 q (2π)4 which simply gives (2π)4 δ (4) (p3 + p4 − p1 − p2 ) q2 i . all initial states on the left. normalisation times energy-momentum conservation times the dynamic parts. 91 Hence. (6. which we understand to be present when we conventionally draw the third. we would get a single diagram out. We can note that the factors in Figure (6. We can represent the amplitude (6. and the number of connections. we have that (6. and only the connections (i. are no terms like ψ We should also note that 4-momentum is conserved at each vertex (there are two vertices in Figure (6. It is interesting to note that if we had derived the Feynman rules without the time ordering.5) has an identical structure to the matrix element we derived for decays. u − m2 + i H (6. we should note that the Feynman amplitude represents two timeorderings. − m2 + i H q = p 3 − p1 = p 4 − p2 .5) where we have defined the Feynman amplitude to be Mfi = (−ige )¯e. something which came about due to delta-functions in the integral.s4 (p4 )uτ. which comes from the Feynman propagator allowing for x or x to come first (i.6) Notice that (6. We should also note that at each vertex.4) becomes 4 (2) Sfi = i=1 1 (2V Ei )1/2 (2π)4 δ (4) (p3 + p4 − p1 − p2 ) Mfi . Finally.s2 (p2 ). Figure (6. This highlights the fact that is dosent matter how the diagrams are drawn.1 An Outline of Particle Exchange and integrating over x only d4 x ei(p4 −p2 +q)x = (2π)4 δ (4) (p4 − p2 + q) .e. Finally. integrating over q.s1 (p1 ) u q2 i (−igτ )¯τ. ¯ This conservation stems from the ψe (x)ψe (x) structured-terms in the Hamiltonian. .6. and final on the right). there is no reason which to choose to be first). charge and lepton numbers are conserved (where lepton numbers stay within their own species).e. Figure (6.s3 (p3 )ue.2)).3) has some examples.6) by a Feynman diagram. (2π)4 δ (4) (p3 − p1 − q) (2π)4 δ (4) (p4 − p2 + q) q2 i − m2 + i H .4) has the two equivalent diagrams (the first two). as q = p3 − p1 = p2 − p4 . which would not be Lorentz invariant (the time ordered double-diagram is Lorentz invariant). putting all of this together. there ¯τ (x)ψe (x) (which would create a τ − and destroy a e− ).2). except we now have a factor for the virtual exchange boson. as in Figure (6.2) are very similar to those we found for the decay.

p2 (−igτ ) uτ.3: Examples of charge and lepton conserving diagrams.92 e− . as well as invalid diagrams.s1 (p1) (−ige) 6 SCATTERING PROCESSES e− . p4 Figure 6. We can now have that when the particles are created. e− e− τ− τ− (a) Valid diagrams. e− τ− e− e+ (b) Invalid diagrams. There are now two possibilities.s2 (p2) τ − .s4 (p4) ¯ τ − . p1 ue. such as τ − (p1 ) + τ − (p2 ) −→ τ − (p3 ) + τ − (p4 ).2: Feynman graph for e− τ − → e− τ − scattering. p3 ue. We could discuss identical particle scatterings as well. . a τ − can be created with either momentum p3 or p4 . with Feynman rules.s3 (p3) ¯ q i q 2 −m2 +i H uτ.2 Other Scattering Processes We have discussed non-identical particle scatterings. 6. Figure 6.

We could also have particle anti-particle scattering. 1 3 1 3 2 4 2 4 Figure 6. and Mfi exchange (p1 p2 p3 p4 ) = −Mfi direct (p1 p2 p4 p3 ). everything is conserved. τ − + τ + −→ τ − + τ + .2 Other Scattering Processes x ≡ x x x ≡ x x 93 Figure 6. The diagram on the left is the direct diagram.6). If we dealt with bosons. The minus sign is simply because we are exchanging identical fermions. where Mfi direct (p1 p2 p3 p4 ) = (−ig)¯(p3 )u(p1 ) u as before. as in Figure (6. we would find a relative minus sign from the anti-commutation relations (which reflects the Pauli principle). but without the minus sign (as bosonic field commute). where. The full invariant amplitude is Mfi = Mfi direct + Mfi exchange .5: The equivalent diagrams represented by the time ordering of the Feynman propagator for scattering of identical particles. one would have the exact same thing. on the right the exchange term. u q 2 − m2 + i H .4: The equivalent diagrams represented by the time ordering of the Feynman propagator. If we do the calculation and keep track of fermion field order.6. i (−ig)¯(p4 )u(p2 ).

but they may have different momenta and spin. where the same particles come out as go in. [σ] = For decays. T −1 = L2 . Hence.7: Schematic of elastic scattering. Let us consider elastic scattering. Now. in terms of dimensions. we have that the transition rate is defined as transition rate = wf i = for some unique final state f .94 u u ¯ 6 SCATTERING PROCESSES v ¯ v Figure 6. The cross-section σ is defined as the transition rate into a given set of final states per unit flux of initial particles.7) . the flux for co-linear collisions (i. with 180◦ between them) is flux ≡ f = 1 |v1 − v2 |. V |Sfi |2 . T (6. p1 p1 p2 p2 Figure 6. 6. see Figure (6.6: The two diagrams corresponding to particle anti-particle scattering.3 Scattering Cross-sections We now outline how to compute quantities we can actually observe. where the colliding particles hit each other head on. T −1 L−2 Hence. the cross-section has the units of an area.7).e.

p1 +p2 d3 p 2 d3 p1 (2π)3 2E1 (2π)3 2E2 d3 p1 d3 p2 4 (4) = (2π) δ (p1 + p2 − p1 − p2 ) . (2π)3 (2π)3 which we clean up using dQ ≡ T V δp1 +p2 .p1 +p2 2V Ei 2V Ei V V d3 p1 d3 p2 3 3 (2π) (2π) V 1 1 1 1 1 = T 2 V 2 δp1 +p2 . 3 (2π) (2π)3 so that the corresponding elemental cross-section is dσ = 1 × transition rate × element size flux V 1 |Sfi |2 V = d3 p1 d3 p 2 .8) We know Sfi . and vi are the velocities of the colliding particles.9) (6.3 Scattering Cross-sections 95 where V is some volume. by adapting (6.5) to be in the finite volume/time limit (which we do such that 2 we can use δij = δij .9) becomes dσ = 1 |Mfi |2 dQ. rather than having to compute the square of the delta-function).6. V V d3 p1 d3 p2 . (2π)3 2E1 (2π)3 2E2 (6.p1 +p2 |v1 − v2 | T 2V E1 2V E1 2V E2 2V E2 V V ×|Mfi |2 d3 p1 d3 p2 . we write the elemental cross-section V 1 dσ = |v1 − v2 | T ×|Mfi |2 2 i=1 1 1 T 2 V 2 δp1 +p2 . 2 Sfi = i=1 1 1 T V δp1 +p2 .p1 +p2 Mfi .11) .10) where we have now taken the infinite box-size limit. 3 f T (2π) (2π)3 (6. Let us consider an element of the final phase space. 4E1 E2 |v1 − v2 | (6. Hence. This is simply the statement that flux is the density times relative speed. 1/2 (2V E )1/2 (2V Ei ) i Hence. (6. using this definition.

One must note that the angles are frame dependent. Energy conservation requires that |p1 | = |p1 |. The six from having two d3 p’s (each has three variables) are constrained by the delta-function. Now. p1 θ. Thus.12) can be written as F = 4E1 E2 |p1 | |p2 | + E1 E2 = 4(E1 + E2 )|p1 | We can rewrite the phase space factor (6.11) is Lorentz invariant under co-linear transformations. . F ≡ 4E1 E2 |v1 − v2 |. as we have the same particles come out that go in.1 Centre of Momentum Frame (6. eliminating four of them.8). where p1 + p2 = 0 and thus p1 + p2 = 0. See Figure (6.10) the invariant phase space factor. dσ (6.10) by integrating over redundant variables. we see that we can write a “modified” flux factor.12) Let us now go into the centre of momentum frame. upon comparison of this with our prototype (6.96 6 SCATTERING PROCESSES We call (6. Notice that dQ (6. 6.8: Colinear scattering in the centre of momentum frame. δ(E1 + E2 − E1 − E2 ) = δ m2 + |p1 |2 + 1 m2 + |p2 |2 − E1 − E2 2 = δ(f (|p|)). the remaining two are usually chosen to be the polar angle θ. as is Mfi . Hence.10) has only two independent scalar variables. φ p1 p2 = −p1 p2 = −p1 Figure 6. 1 2 1 The flux factor F and invariant phase space dQ are both Lorentz invariant. The flux factor (6. Notice that we take take out the zeroth-component of the delta-function.8).3. This can also be written as F = 4 (pµ p2µ )2 − m2 m2 . φ relative to the beam direction of particle 1 (say).

6.4 Scattering From a Classical Field

97

Then, we rewrite this delta-function using the usual method for the argument of the function, δ(E1 + E2 − E1 − E2 ) = Hence, doing this, we find E1 E2 δ(|p1 | − |p1 |). (E1 + E2 )|p1 |

|p1 | 1 dΩcm , 2 4(E + E ) (2π) 1 2 so that the differential cross-section is dQ = dσ |Mfi |2 = . dΩcm 64π 2 (E1 + E2 )2 (6.13)

To go further, we have to evaluate |Mfi |2 , which will depend upon spins and momentum. If we “dont care about” spins (i.e. polarisation), then we average over initial spins and sum over final spins – this is done if a detector does not pick up different polarisations. Finally, if we want the total cross-section we integrate over the solid angles θ, φ.

6.4

Scattering From a Classical Field

Here we shall consider scattering from a static classical field – classical means that the field is so “big” that one can ignore quantum fluctuations. This could be the case, for example, if electrons are scattered from a large magnetic field, or the Coulomb field of a heavy nucleus. Let us recall the previous interaction Hamiltonian, for the fermionic fields scattering from a bosonic field, ¯ HI = gN ψ(x)ψ(x)φ(x) . Suppose we take our scalar field φ(x) to now be static, and let us denote it by V (x). Then, the first order S-matrix element for scattering from |i to |f is Sfi = −ig ¯ d4 x f | N ψ(x)ψ(x) |i V (x). (6.14)

Now, suppose we have the initial and final states being |i = |p, s , |f = |p , s , then, the only terms in the field expansion of (6.14) to contribute are Sfi = −ig Now, we know that e−ipx us (p) √ |0 , 2EV eip x us (p ) ¯ ¯ p , s | ψ (−) (x) = 0| √ . 2EV ψ (+) (x) |p, s = ¯ d4 x f | ψ (−) (x)ψ (+) (x) |i V (x).

98 Hence, we use these in the matrix element to easily obtain Sfi = −ig d4 x V (x)¯s (p )us (p) u

6 SCATTERING PROCESSES

eix(p −p) √ . 2V EE

(6.15)

The dx0 integral is easy to compute, as the only term to contribute is the exponential, dx0 ex
0 (E

−E)

= T δE,E = (2π)δ(E − E ).

The only terms in the dxi integral are the potential and exponential, which form the integral V (q) ≡ d3 x V (x)e−ix·(p −p) , (6.16)

where we have introduced the Fourier transform of the scattering potential, V (q), where q ≡ p − p. So, using these, (6.15) becomes Sfi = −igV (q)¯s (p )us (p)T δE,E u Now, the transition rate is just wf i = |Sfi |2 T 2 T 2 δE,E g2 |¯s (p )us (p)|2 |V (q)|2 . u = 2 EE 4V T 1 √ . 2V EE

(6.17)

We now take advantage of
2 T δE,E = T δE,E = (2π)δ(E − E ),

so that (6.17) becomes wf i = |M|2 2π where |M|2 ≡

δ(E − E ) , V2

g2 |¯s (p )us (p)|2 |V (q)|2 . u 4EE To calculate the cross-section, we need both the flux and density of states. The flux is just f = ρv = 1 v, V

where v is the difference in velocities. The density of (final) states is simply V d3 p . (2π)3

6.4 Scattering From a Classical Field Hence, following the same structure as (6.8), the elemental cross-section is V δ(E − E ) V d3 p 2 dσ = |M| 2π v V2 (2π)3 d3 p 1 |M|2 2πδ(E − E ) = . v (2π)3 E 2 = |p|2 + m2 , and hence where p ≡ |p| (i.e. not the 4-vector). So, d3 p = p 2 dp dΩ = p E dE dΩ . Hence, inserting this into the elemental cross-section (6.18), we see that 1 1 dσ = |M|2 2πδ(E − E ) p E dE dΩ . v (2π)3 p dE = , dp E

99

(6.18)

Now, let us use a trick to convert d3 p into an integral over energy. We can easily recall that

If we integrate over E now, we simply filter out E = E and p = p , so that the differential cross-section is dσ Ep = |M|2 , (6.19) 2v dΩ (2π) where g2 |¯s (p )us (p)|2 |V (q)|2 . u (6.20) 4EE Basically, one of the points of this, is that the differential cross-section is proportional to the modulus-squared of the Fourier transform of the scattering potential – this is very similar to the Born approximation in non-relativistic quantum mechanics. |M|2 ≡ If we consider the low energy limit, then the spinors reduce √ χs , us (p → 0) = 2m 0 so that us (p )us (p) = 2mδss . ¯

Thus, in the low energy limit, the spins of the outgoing particles is the same as that of the incoming particles. That is, spin is conserved; in which case the result reduces to the non-relativistic Born approximation obtained from the Schrodinger equation. We can also get the same result from our previous completely quantised calculation of e− τ − scattering by one-particle exchange; where we make V (q) proportional to the propagator. That is, the scattering potential is proportional to the Fourier transform of the propagator of the exchanged particle.

100 6 SCATTERING PROCESSES .

and q the 4-momentum carried by the exchange boson. but it is worth noting that m denotes the mass of the exchange boson.101 7 Feynman Rules The Feynman rules can be used to construct the invariant Feynman amplitude Mfi for a given process. When drawing Feynman graphs. These only apply up to second order perturbation theory. . One write a vertex factor for every vertex in the diagram. weak or QCD interactions. and a dashed line a boson. we use the convention that the arrow only denotes whether the “particle” described is a particle or anti-particle. and do not include any QED. All factors are pretty self-explanatory. We also use the convention that a solid line denotes a fermion.

.1: Summary of the Feynman rules we have derived.102 7 FEYNMAN RULES vertex factor −ig outgoing fermion us (k) ¯ incoming fermion us (k) outgoing anti-fermion vs (k) incoming anti-fermion vs (k) ¯ exchange boson i q 2 −m2 +i Figure 7.

Sign up to vote on this title
UsefulNot useful