This action might not be possible to undo. Are you sure you want to continue?
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 JanMay ’09. Please email 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 δfunction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 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 KleinGordon 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 Hydrogenlike 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 KleinGordon 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 Smatrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4.5.3 The Smatrix Expansion . . . . . . . . . . . . . . . . . . . . . . . . . 69
5 Decays 73
5.1 Simple 2body 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 Crosssections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 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
Nonrelativistic 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 antiparticles, the spinstatistics theorem
etc.
1 Preliminaries
Let us introduce the notation we shall be adopting.
1.1 Relativistic Notation
Let us introduce the contravariant 4vector
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 oﬀdiagonal components are zero. Using this, we can deﬁne a covariant 4vector,
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 4vector (i.e. all four components), and
by writing x, only the 3spatial components.
1.2 The Dirac δfunction
As we shall use the δfunction 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 δfunction 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 δfunction return a nonzero
value. Hence, the only contribution from the integral is at x = x
, so that f(x
) is returned.
Thinking about the δfunction as a ﬁltering mechanism can be very useful. For the δfunction
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 δfunction 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 δfunction (1.1). An immediate con
sequence of this is that the δfunction is even, δ(−x) = δ(x). Let us then consider a full
functional argument, δ(g(x)). Now, the δfunction only returns a nonzero 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 δfunction with func
tional argument, into a useable form.
We may have a multidimensional δfunction 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
)δ
(3)
(x −x
).
We can also deﬁne the δfunction 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 4vector. 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 4vector if it transforms as
A
µ
→A
µ
= Λ
ν
µ
A
ν
.
The scalar product of any two 4vectors 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 energymomentum in the 4momentum
p
µ
= (E/c, p)
is a 4vector. 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 4vectors 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 4vector. So, in order that the LHS is a scalar, we must have
that
∂φ
∂x
µ
is a covariant 4vector. 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 4vector
∂
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
Ω
=
_
Ω
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
_
Ω
dV ρ(x, t)
= −
_
Ω
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 4vector.
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 4vector 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 KleinGordon 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 squareroot, 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
Hφ
= 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 KleinGordon (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 KLEINGORDON 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 nonrelativistic case, (1.11), we see that we
cannot interpret ρ as a probability density (as we did do in the nonrelativistic 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 nonrelativistic 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 4vector potential is nonzero, 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 = −
Zα
r
, α =
e
2
4π
=
1
137
. (2.10)
We can solve by comparing with the nonrelativistic 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 KLEINGORDON 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 nonrelativistic 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 nonrelativistic 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 nonrelativistic 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 nonrelativistic result. The
third term gives a splitting for modes with diﬀerent , and is thus the ﬁnesplitting term.
This ﬁnesplitting 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, spinzero particles). Hence, this scalar
KG equation described spin0 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 KLEINGORDON 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 zaxis,
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
∗
dψ
dz
−ψ
dψ
∗
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 KLEINGORDON 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/antiparticle pairs are created at the
barrier, by an interaction with the barrier. Particles move to the left, in the reﬂected wave,
and antiparticles 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 KleinGordon 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
(α
j
β + βα
j
)
∂ψ
∂x
j
+ β
2
m
2
ψ
= −
j
α
2
j
∂
2
ψ
∂x
2
j
−
j=k
(α
j
α
k
+ α
k
α
j
)
∂
2
ψ
∂x
j
∂x
k
−im
j
(α
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
(α
j
α
k
+ α
k
α
j
)
∂
2
ψ
∂x
j
∂x
k
−im
j
(α
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 anticommutation 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
[σ
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 blocknotation 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 nonrelativistic 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 anticommutation 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
iψ
†
∂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
[ψ
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 nonzero. 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 somethingelse).
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
]
[σ
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
[α
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 4component 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 4component 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 anticommutation 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 nonrelativistic 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 antiparticle. 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
antiparticle
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
iβ
∂ψ
∂t
= (−iβα ∇+ m) ψ,
writing βα ∇ as
βα ∇ = βα
i
∂
∂x
i
,
then we see that the Dirac equation looks like
iβ
∂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
iγ
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 4vector. 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 γmatrix to still be a γmatrix, 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
⇒
_
iγ
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
_
iγ
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 xaxis. 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
Pψ
(+)
= ηψ
(+)
, Pψ
(−)
= −ηψ
(−)
.
Therefore, we see that both the positive and negative energy solutions are eigenstates of
parity P, but particles and antiparticles 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,
Pψ
(+)
p,s
(x) = N
_
χ
s
σ·p
E+m
χ
s
_
e
−i(Et
+p·x
)
,
= ψ
(+)
−p,s
(x
)
Pψ
(−)
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 antiparticles.
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 nonrelativistic weakﬁeld 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 socalled gfactor (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 antiparticles 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 socalled 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 antiquarks
to attract (note, we want antiquarks to attract as well, as we would like to be able to allow
antiprotons to exist).
We shall set m
q
= 0 (which is a good approximation for u, dquarks). 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 Hydrogenlike Atoms
Now consider a system with
S(r) = 0, V (r) = −
Zα
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 = −
mα
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 nonrelativistic case, the spinorbit 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 energyinformation is
3.7 Particles in a Spherical Potential 45
nonrelativistic
∆
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 Hatom. In the nonrelativistic case, the spinorbit
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 nonrelativistic 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
nonrelativistic 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
2π
+ . . .
_
,
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 manybody 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
√
2ω
n
L
_
a
n
e
−i(ωnt−knz)
+ a
∗
n
e
i(ωnt−knz)
¸
. (4.2)
The prefactor 1/
√
2ω
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
√
2ω
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 noncommuting
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 zeropoint 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 zdirection 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 anticommutators) are bosons.
4.1.2 Field Operators
Recall (4.2),
φ(t, z) =
∞
n=−∞
1
√
2ω
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 KleinGordon equation, for a massless particle,
φ = ∂
µ
∂
µ
φ = 0,
4.2 The KleinGordon 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 energymomentum 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 KleinGordon Field
Let us consider quantising ﬁelds that satisfy the KleinGordon 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 =
2π
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 componentbycomponent 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 onehalf (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 KleinGordon 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 zeropoint 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 4momentum 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 spin0 particles, with 4momentum
(E(p), p) and mass m; and correspondingly, a(p) and c(p) both destroy spin0 particles,
with 4momentum (E(p), p) and mass m. Then, we may ask, what is the diﬀerence between
a and c. To answer this, consider the conserved 4current 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 antiparticles with charge Q = −q; and a destroys particles with charge Q = +q, c
destroys antiparticles 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 nonHermitian 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 equaltime 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 KleinGordon ﬁeld, we looked at ﬁelds which permit particles
(and antiparticles) 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 singleparticle state, in n = p, s and [n, n
) is a twoparticle state.
Now, by the Pauli exclusion principle (which states that no two fermions can be in the
same quantum state; and that fermions are antisymmetric), 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,
iψ
†
∂ψ
∂t
= ψ
†
(−iα ∇+ βm) ψ,
and integrating,
i
_
d
3
xψ
†
∂ψ
∂t
=
_
d
3
xψ
†
(−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 antiparticle 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 timeordered) product is
T
_
φ(x)φ
†
(0)
_
= θ(t)φ(x)φ
†
(0) + θ(−t)φ
†
(0)φ(x), (4.42)
and the θfunction 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 θfunction 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 pointsource KleinGordon 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 KleinGordon equation. That is, we can use the
Green function G
F
(which is the solution to the above pointsource 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 δfunction 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: Lowerhalfplane In this case, we close the contour in the lower half plane, and
let the radius of the semicircle go to inﬁnity. There is no contribution from the semicircle,
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 (anticlockwise) 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
lowerhalf 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
2ω
k
.
Notice that the minussign 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: Upperhalfplane In this case, we close the contour in the upperhalfplane, 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 freeﬁeld results in the
Heisenberg picture hold for interacting ﬁelds, in the interaction picture.
68 4 QUANTUM FIELDS
4.5.2 The Smatrix
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 Smatrix 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 freeparticle 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 Smatrix, (4.64). Thus, we can write the ﬁnal state as a
linear sum of free particle states,
[ψ(∞))
I
= [ψ
f
)
I
=
i
o
ﬁ
[ψ
i
)
I
.
We can write the ﬁnal state like this, as we have stated that the Smatrix 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
ﬁ
= ¸ψ
f
[ S [ψ
i
) , (4.65)
which occurs with probability
P
fi
= [o
ﬁ
[
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 Smatrix 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
ﬁ
describe the amplitude for a particular initial and ﬁnal state.
4.5 Interactions 69
The Unitarity of the Smatrix 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 Smatrix
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
†
[ψ
f
) ¸ψ
f
[ S [ψ
i
)
= ¸ψ
i
[ S
†
S [ψ
i
)
= ¸ψ
i
[ψ
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 Smatrix 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 Smatrix,
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 timeordered 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 upperlimit 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 4position vector x,
H
I
(x) = H
i
(φ
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 Smatrix expansion.
72 4 QUANTUM FIELDS
73
5 Decays
We shall illustrate calculations by considering examples.
5.1 Simple 2body 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 2body 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)
ﬁ
= −i
_
d
4
x
¸
f = τ
+
τ
−
¸
¸
H
I
(x)
¸
¸
i = H
0
_
,
which denotes the amplitude for a decay of H
0
→τ
+
+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 overbar 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
ﬁ
= −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 2body Decay: φ
3
75
give nonzero 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
ﬁ
= −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
Kroneckerdelta 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
ﬁ
=
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)/
ﬁ
, (5.10)
where we have deﬁned the Feynman amplitude
/
ﬁ
= −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
ﬁ
= normalisation energymomentum conservation dynamics.
The initial three factors involving energy are the normalisation, and will be common to all
processes. The second term with the deltafunction is infact the statement of conservation
of energymomentum. The ﬁnal term /
ﬁ
is the only term that depends on the speciﬁc
interaction, and is where all dynamical information is kept. We can represent /
ﬁ
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
ﬁ
[
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
[/
ﬁ
[
2
,
where we are in the ﬁnite volume/time limit, to allow us to have a meaning in squaring the
deltafunction. Now we have squared the deltafunction, we can take the limit V, T →∞, so
that the energymomentum conservation term becomes just
V Tδ
k+k
,p
−→(2π)
4
δ
(4)
(k + k
−p),
and the two phasespace 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
2ω
p
s,s
_
d
3
k
(2π)
3
2E
k
_
d
3
k
(2π)
3
2E
k
(2π)
4
δ
(4)
(k + k
−p) [/
ﬁ
[
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
[/
ﬁ
[
2
= g
2
s,s
[¯ u
s
(k)v
s
(k
)[
2
.
So, to begin,
s,s
[/
ﬁ
[
2
=
/
ﬁ
/
∗
ﬁ
= 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
)γ
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
[/
ﬁ
[
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
[/
ﬁ
[
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
[/
ﬁ
[
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
[/
ﬁ
[
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 γmatrices 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
[/
ﬁ
[
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
[/
ﬁ
[
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 restframe of the Higgs boson, P
µ
= (m
H
, 0). Therefore, conser
vation of 4momentum 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
[/
ﬁ
[
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 phasespace integral (5.12), in the restframe 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 deltafunction 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 nonzero 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 deltafunction, 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 deltafunction 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 restframe 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
8π
.
We can use g from (5.3) to see that
Γ
_
H
0
→τ
+
τ
−
_
= /m
H
m
2
τ
, / ≡
(
8
√
2π
. (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 /
ﬁ
, 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 σﬁeld is uncharged, which is modeled
by being Hermitian (i.e. we use a and a
†
). In contrast, the φﬁeld is charged, modeled by
not being Hermitian, so we use b and d
†
.
We need to ﬁrst compute o
ﬁ
,
o
ﬁ
= −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 nonzero 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
ﬁ
, so that
o
ﬁ
= −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)
/
ﬁ
,
84 5 DECAYS
where we have deﬁned the Feynman amplitude to be
/
ﬁ
≡ −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
ﬁ
;
o
ﬁ
=
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)/
ﬁ
.
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
ﬁ
[
2
T
.
To be able to compute [o
ﬁ
[
2
, (infact, more precisely, to compute the square of the delta
function), we go into the ﬁnitebox 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
(δ
k+k
,p
)
2
,
but
(δ
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
ﬁ
[
2
=
1
(2V E
p
)
1
(2V ω
k
)
1
(2V ω
k
)
V
2
T
2
δ
k+k
,p
[/
ﬁ
[
2
.
Let us cancel oﬀ one of the factors of V , and divide by T, so that
[o
ﬁ
[
2
T
=
1
2E
p
1
(2V ω
k
)
1
(2V ω
k
)
V Tδ
k+k
,p
[/
ﬁ
[
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
ﬁ
[
2
T
=
1
2E
p
1
(2V ω
k
)
1
(2V ω
k
)
(2π)
4
δ
(4)
(k + k
−p)[/
ﬁ
[
2
.
To compute the decay rate, we must sum this expression over all possible ﬁnal states, so that
Γ =
f
[o
ﬁ
[
2
T
=
k,k
1
2E
p
1
(2V ω
k
)
1
(2V ω
k
)
(2π)
4
δ
(4)
(k + k
−p)[/
ﬁ
[
2
.
To perform the summation, we take the inﬁnitebox limit, so that
k
1
V
−→
_
d
3
k
(2π)
3
.
Hence,
Γ =
1
2E
p
_
d
3
k
(2π)
3
2ω
k
_
d
3
k
(2π)
3
2ω
k
(2π)
4
δ
(4)
(k + k
−p)[/
ﬁ
[
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 σparticle at rest. This
will allow us to compute the deltafunction.
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 4dimensional deltafunction, thus
δ
(4)
(k + k
−p) = δ(ω
k
+ ω
k
−E
p
)δ
(3)
(k +k
−p) .
In the rest frame of the decaying particle, the second 3D deltafunction just becomes δ
(3)
(k+
k
). Hence, using this, the decay rate becomes
Γ =
1
2E
p
_
d
3
k
(2π)
3
2ω
k
_
d
3
k
(2π)
3
2ω
k
_
(2π)
4
δ(ω
k
+ ω
k
−E
p
)δ
(3)
(k +k
)[/
ﬁ
[
2
_
.
The ﬁrst d
3
k
integral will only return a nonzero value if k
= −k, due to the 3D delta
function. Hence, after doing this integral, we have
Γ =
1
2E
p
_
d
3
k
(2π)
3
2ω
k
1
(2π)
3
2ω
−k
(2π)
4
δ(ω
k
+ ω
−k
−E
p
)[/
ﬁ
[
2
.
86 5 DECAYS
By inspection of ω
k
=
_
k
2
+ m
2
φ
, we see that ω
−k
= ω
k
. Hence, the remaining deltafunction
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 deltafunction useable, we must get it into the form δ(k − k
0
), so that upon
integration, only k = k
0
will give a nonzero 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 deltafunction, 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 deltafunction in the decay rate, to see that
Γ =
1
2E
p
_
d
3
k
(2π)
3
4ω
2
k
1
(2π)
3
(2π)
4
ω
k
0
2[k
0
[
δ ([k[ −[k
0
[) [/
ﬁ
[
2
.
Let us rearrange slightly,
Γ =
1
2E
p
ω
k
0
2[k
0
[
_
d
3
k
(2π)
2
4ω
2
k
δ ([k[ −[k
0
[) [/
ﬁ
[
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
4π
(2π)
2
4ω
2
k
δ ([k[ −[k
0
[) [/
ﬁ
[
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
[/
ﬁ
[
2
=
1
2E
p
[k
0
[
8πω
k
0
[/
ﬁ
[
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
8π
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
[/
ﬁ
[
2
.
Due to the simple form of our /
ﬁ
= −ig, we easily have
[/
ﬁ
[
2
= g
2
.
Hence, we have computed that the complete expression for the decay rate of a scalar σparticle
to a pair of scalar charged φparticles 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 Smatrix 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
)ψ
(+)
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
)ψ
(+)
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
ﬁ
= (−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 unbarred 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)
ﬁ
=
_
4
i=1
1
(2V E
i
)
1/2
_
(2π)
4
δ
(4)
(p
3
+ p
4
−p
1
−p
2
) /
ﬁ
, (6.5)
where we have deﬁned the Feynman amplitude to be
/
ﬁ
= (−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 energymomentum 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) structuredterms 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 4momentum 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 deltafunctions
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 doublediagram 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 nonidentical 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 anticommutation relations (which reﬂects the Pauli principle). The full
invariant amplitude is
/
ﬁ
= /
ﬁ
direct
+/
ﬁ
exchange
,
where
/
ﬁ
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
/
ﬁ
exchange
(p
1
p
2
p
3
p
4
) = −/
ﬁ
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 antiparticle 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 antiparticle scattering.
6.3 Scattering Crosssections
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 crosssection 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 crosssection has the units of an area.
For decays, we have that the transition rate is deﬁned as
transition rate = w
fi
=
[o
ﬁ
[
2
T
, (6.7)
for some unique ﬁnal state f.
Now, the ﬂux for colinear 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 Crosssections 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 crosssection is
dσ =
1
ﬂux
transition rate element size (6.8)
=
1
f
[o
ﬁ
[
2
T
V
(2π)
3
d
3
p
1
V
(2π)
3
d
3
p
2
.
We know o
ﬁ
, 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 deltafunction),
o
ﬁ
=
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
/
ﬁ
.
Hence, we write the elemental crosssection
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
[/
ﬁ
[
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
[/
ﬁ
[
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 boxsize limit. Hence, using this deﬁnition, (6.9)
becomes
dσ =
1
4E
1
E
2
[v
1
−v
2
[
[/
ﬁ
[
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
2µ
)
2
−m
2
1
m
2
2
.
The ﬂux factor F and invariant phase space dQ are both Lorentz invariant, as is /
ﬁ
. Hence,
dσ (6.11) is Lorentz invariant under colinear 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 deltafunction, 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 zerothcomponent of the deltafunction,
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 deltafunction 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 crosssection is
dσ
dΩ
cm
=
[/
ﬁ
[
2
64π
2
(E
1
+ E
2
)
2
. (6.13)
To go further, we have to evaluate [/
ﬁ
[
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 crosssection 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 Smatrix element for scattering from [i) to [f) is
o
ﬁ
= −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
ﬁ
= −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
ﬁ
= −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
ﬁ
= −igV (q)¯ u
s
(p
)u
s
(p)Tδ
E,E
1
2V
√
EE
.
Now, the transition rate is just
w
fi
=
[o
ﬁ
[
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
Tδ
2
E,E
= Tδ
E,E
= (2π)δ(E −E
),
so that (6.17) becomes
w
fi
= [/[
2
2π
δ(E −E
)
V
2
,
where
[/[
2
≡
g
2
4EE
[¯ u
s
(p
)u
s
(p)[
2
[V (q)[
2
.
To calculate the crosssection, 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 crosssection is
dσ =
V
v
[/[
2
2π
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 4vector). So,
d
3
p
= p
2
dp
dΩ
= p
E
dE
dΩ
.
Hence, inserting this into the elemental crosssection (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
crosssection is
dσ
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 crosssection is proportional to the
modulussquared of the Fourier transform of the scattering potential – this is very similar to
the Born approximation in nonrelativistic 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
nonrelativistic 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 oneparticle 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 /
ﬁ
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 antiparticle.
We also use the convention that a solid line denotes a fermion, and a dashed line a boson.
All factors are pretty selfexplanatory; but it is worth noting that m denotes the mass of
the exchange boson, and q the 4momentum 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)
antifermion
incoming ¯ v
s
(k)
outgoing v
s
(k)
antifermion
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 δfunction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 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 KleinGordon 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 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . Hydrogenlike 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 KleinGordon Field . . . . . . Interactions . .1. . . . . . . . .1 4. . . . . . .2 CONTENTS Parity . . . . . . . . . . . . . .5. . . . . . 4. . . . . . . . . . . . . . . . . . . . . . . . . Scattering Crosssections . . . . . . . . . . . .1 5. . . .3. . . . . . . . . . . . . . . . . . . . . .7 3. . .iv 3. . . . . . . . . . . . . . . . . . . . . . . . The Smatrix Expansion . . The Smatrix . . . . . . . . . . . . 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 2body 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 spinstatistics theorem etc. x). The latter needs the full formalism of quantum ﬁeld theory. T > mc2 .1 Nonrelativistic 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 4vector 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 antiparticles. −1. the latter is able to produce particles. 2 2m Under this. −1. where all oﬀdiagonal components are zero. we can deﬁne a covariant 4vector. 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 3spatial components. Only when x = x does the δfunction return a nonzero value. For the δfunction to be used directly in this way. If instead we have a functional argument.2 The Dirac δfunction As we shall use the δfunction 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 δfunction as a ﬁltering mechanism can be very useful. Suppose we have. by writing x. one can think about the δfunction 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 4vector (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 δfunction with functional argument. Hence. We may have a multidimensional δfunction 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 δfunction 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 δfunction 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 δfunction (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 δfunction only returns a nonzero 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 δfunction 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 4vector. 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 4vectors 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 energymomentum in the 4momentum pµ = (E/c. we must have ∂φ that ∂xµ is a covariant 4vector. 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 4vector. 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 4vector 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 4vector. All equations written consistently in terms of 4vectors 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 4vector ∂ 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 4vector.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 4vector 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 ∂ ∂ ∗ ψ2 = 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 ∂ ψ2 = ∂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 squareroot.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 KleinGordon Equation So. This is the KleinGordon (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 nonrelativistic 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 nonrelativistic 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 KLEINGORDON EQUATION E = − p2 + m2 < −m. Hence. we see that we cannot interpret ρ as a probability density (as we did do in the nonrelativistic 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 nonrelativistic 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 4vector potential is nonzero. 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 KLEINGORDON EQUATION So.16) − (Zα)2 . 2n 2 nr = 0. . . Notice that (2. upon comparison of this with the nonrelativistic version (2. Hence. after putting in the central Coulomb potential (2. the energy levels of the nonrelativistic 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 nonrelativistic equations.
Notice that in this regime. The ﬁrst term is the rest mass of the electron. Hence. spinzero 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 nonrelativistic result.17). 2 so that for = 0. ..2. and is thus the ﬁnesplitting term. n = n. and thus cannot be used as a description of the Hydrogen atom.. this scalar KG equation described spin0 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 ﬁnesplitting 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 = ip . 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 zaxis.1 The Klein Paradox T =E−m 2 THE KLEINGORDON 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 C2 . jR = −2pB2 . 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 = 2pA2 . 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 KLEINGORDON 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 antiparticles 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/antiparticle 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 KleinGordon 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 anticommutation 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 nonrelativistic 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 blocknotation 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 anticommutation 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 somethingelse).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 nonzero. 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 4component 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 4component 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 anticommutation 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 nonrelativistic limit p → 0 (and the energy E is of the order the rest mass m). which looks like an antiparticle. 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 antiparticle 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 4vector. 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 γmatrix to still be a γmatrix. (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 xaxis. = 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 antiparticles 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 antiparticles. 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 nonrelativistic weakﬁeld 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 antiparticles 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 socalled gfactor (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 antiprotons to exist). and interact weakly at r 1fm (their socalled asymptotic freedom). dr we can substitute the second. df + Eg = 0. dquarks). 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 antiquarks 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 antiquarks 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 nonrelativistic 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 spinorbit 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 energyinformation is .2 Hydrogenlike 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 nonrelativistic ∆ 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 nonrelativistic 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 nonrelativistic case.2: The spectrum of splittings in the Hatom. 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 nonrelativistic 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 spinorbit 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 manybody 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 noncommuting 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 zeropoint energy. n (4. and energy by n units of ωn . We will make a guess – and conﬁrm later – that the momentum operator in the zdirection 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 anticommutators) are bosons. ∂φ . we have the (veriﬁable) L ∂φ ∂φ † kn an an = − dz .1. a† ] = 0. by inspection. this is just the invariant KleinGordon 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 KleinGordon 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 KleinGordon Field Let us consider quantising ﬁelds that satisfy the KleinGordon 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 energymomentum 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 componentbycomponent 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 onehalf (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 KleinGordon 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 4momentum 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 zeropoint 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 nonHermitian 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 antiparticles with charge Q = −q. Now. what we see. and a. consider the conserved 4current for the KG equation. with 4momentum (E(p).e no c(p)type). is that a† (p) and c† (p) both create spin0 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 spin0 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 antiparticles with charge Q = −q. Q. with 4momentum (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 equaltime 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 KleinGordon 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 antiparticles) 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 twoparticle 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 antisymmetric). 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 singleparticle 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 antiparticle 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 timeordered) 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 θfunction 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 θfunction 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 δfunction 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 pointsource KleinGordon 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 KleinGordon equation. we can use the Green function GF (which is the solution to the above pointsource 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 lowerhalf plane. and let the radius of the semicircle 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 semicircle.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: Lowerhalfplane 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 (anticlockwise) closed path. That is. do the integral. 2ωk Notice that the minussign 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 upperhalfplane. 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 = Hψ(t) S .e. t < 0: Upperhalfplane 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 freeﬁeld 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 Smatrix as a “mixing matrix”. − τ ) = U (∞.66) (4.5. which is a freeparticle state (i. At ﬁnite t. The initial state (i. We deﬁne the Smatrix 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 = Sψt=−∞ I . −∞). t0 ) = 1. at t = −∞) is just ψ(−∞) I = ψi I .e. as we have stated that the Smatrix 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 Smatrix. 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 Smatrix 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 Smatrix Expansion Now. appears within the integral itself. t0 ). For example. Pf i = f f (4.5 Interactions The Unitarity of the Smatrix 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 Smatrix 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 upperlimit 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 Smatrix. . . t t 4 QUANTUM FIELDS U (t. (4. ∞ (4. we have an expression which is of the timeordered 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 Smatrix expansion. . the expressions (4. (4.5 Interactions where we may have multiple ﬁelds which are functions of the 4position 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 2body 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 2body 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 overbar 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 2body Decay: φ3 75 give nonzero 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 Kroneckerdelta 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 deltafunction is infact the statement of conservation of energymomentum. 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 × energymomentum 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 00 = 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 deltafunction. s Mf i = −ig¯s (k)vs (k ) u k . and the two phasespace terms 1 V −→ k d3 k .5.s 1 1 T 2 V 2 δk+k . Now we have squared the deltafunction.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 energymomentum 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 γmatrices 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 deltafunction 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 restframe of the Higgs.16) to write I= d3 k δ (2E(k) − mH ) . g2 s. 2 (5.s s. conservation of 4momentum 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 phasespace integral (5.17) into this expression. ﬁlters out k = −k as the only nonzero contribution. We now specialise to the restframe 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= πk0  . mH k0 ≡ m2 H − m2 . we have that (5. . if we assume no angular dependence. I = = 4πk2 E(k) dk δ k − 2 4Ek 2k 4πk0  . Now. using these results.2 Decay Rates & Lifetimes and then E(k) = k2 + m2 .19) To evaluate the deltafunction. τ 4 m2 H − m2 τ 4 Then. the argument of the deltafunction in (5.19) reads I= d3 k E(k) δ k − 2 4Ek 2k 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 πk0  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 restframe 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 nonzero 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 φﬁeld 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 σﬁeld 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 ﬁnitebox 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 deltafunction. The ﬁrst d3 k integral will only return a nonzero value if k = −k. we can always split up the 4dimensional deltafunction. We now specialise to the rest frame of the decaying particle. due to the 3D 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 3D deltafunction just becomes δ (3) (k+ k ). we will compute the decay rate of a σparticle 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 inﬁnitebox limit.
given our deltafunction. 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 deltafunction in the decay rate. dk k2 4π δ (k − k0 ) Mﬁ 2 . φ 5 DECAYS k2 + m2 . Hence. Hence. only k = k0 will give a nonzero contribution. φ 4 2k . To do this. so that Γ = 1 ωk0 k0 2 π Mﬁ 2 2ω2 2Ep 2k0  (2π) k0 1 k0  = Mﬁ 2 . f (x0 ) f (x0 ) = 0. so that upon integration. 3 4ω 2 (2π)3 (2π) k 2k0  1 ω k0 2Ep 2k0  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 2k0  dk k2 4π. the remaining deltafunction 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 . 2k0  Therefore. so that d3 k −→ Hence. To make this deltafunction 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 σparticle to a pair of scalar charged φparticles 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 Smatrix 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 unbarred 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 energymomentum 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 4momentum 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 deltafunctions 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) structuredterms 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 doublediagram 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 nonidentical particle scatterings. 6. Figure 6.
We could also have particle antiparticle 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 anticommutation 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 crosssection σ is deﬁned as the transition rate into a given set of ﬁnal states per unit ﬂux of initial particles.7) . the ﬂux for colinear 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 antiparticle scattering.3 Scattering Crosssections 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 crosssection 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 crosssection 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 Crosssections 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 deltafunction).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 crosssection 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 boxsize 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 deltafunction. 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 colinear 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 zerothcomponent of the deltafunction.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 deltafunction 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 crosssection 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 crosssection 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 Smatrix 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 = M2 2π where M2 ≡
δ(E − E ) , V2
g2 ¯s (p )us (p)2 V (q)2 . u 4EE To calculate the crosssection, 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 crosssection is V δ(E − E ) V d3 p 2 dσ = M 2π v V2 (2π)3 d3 p 1 M2 2πδ(E − E ) = . v (2π)3 E 2 = p2 + m2 , and hence where p ≡ p (i.e. not the 4vector). So, d3 p = p 2 dp dΩ = p E dE dΩ . Hence, inserting this into the elemental crosssection (6.18), we see that 1 1 dσ = M2 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 crosssection is dσ Ep = M2 , (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 crosssection is proportional to the modulussquared of the Fourier transform of the scattering potential – this is very similar to the Born approximation in nonrelativistic quantum mechanics. M2 ≡ 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 nonrelativistic 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 oneparticle 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 4momentum 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 antiparticle. and do not include any QED. All factors are pretty selfexplanatory. 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 antifermion vs (k) incoming antifermion vs (k) ¯ exchange boson i q 2 −m2 +i Figure 7.