1 Free real scalar field

The Hamiltonian is
� �
3 1 � �
H = d xH = d3 x p(x)2 + (∇φ)2 + m2 φ2
2
Let us expand both φ and p in Fourier series:
� �
d3 p ip·x d3 p
φ(t, x) = φ̃(t, x)e , p(t, x) = p̃(t, x)eip·x .
2ω(p) 2ω(p)
�
where ω(p) = p2 + m2 . Then:
� � �
1 d3 p 2 2 2
H= |p̃(p)| + |φ̃(p)| ω(p) .
2 (2π)3 (2ω(p))2

This is a Hamiltonian for an infinite collection of harmonic oscillators labeled

by p ∈ R3 and energy ω(p). Introduce creation-annihilation operators:

p(p) − iφ(p) p† (p) + iφ† (p)

a(p) = √ , a† (p) = √ .
ω(p) 2 ω(p) 2

Then: � � �
d3 p † 1
H= ω(p) a (p)a(p) +
(2π)3 2ω(p) 2
The last term in parentheses can be dropped (divergent vacuum energy).
The operators a, a† satisfy:

[a(p), a† (p� )] = (2π)3 2ω(p)δ 3 (p − p� ), [a(p), a(p� )] = 0.

The expression 2ω(p)δ 3 (p − p� ) is Lorenz-invariant, so this is a natural nor-

malization of creation-annihilation operators in a relativistic theory.
So, as expected, the free scalar field describes noninteracting spinless
�
bosonic particles with a relativistic energy-momentum relation E(p) = p2 + m2 .

2 Free complex scalar field

Commutation relations:

1
 [φ(t, x), p(t, y)] = iδ 3 (x − y),
[φ(t, x)† , p(t, y)† ] = iδ 3 (x − y),
[φ(t, x)† , p(t, y)] = 0,
[φ(t, x), p(t, y)† ] = 0,
[φ(t, x), φ(t, y)] = 0,
[φ(t, x)† , φ(t, y)† ] = 0,
[φ(t, x), φ(t, y)† ] = 0,
[p(t, x), p(t, y)] = 0,
[p(t, x)† , p(t, y)† ] = 0,
[p(t, x), p(t, y)† ] = 0.

Here p = φ̇† , p† = φ̇.

Hamiltonian:
�
� �
H= d3 x pp† + ∂i φ† ∂i φ + m2 φ† φ .

Let us show that these equations� describe the bosonic Fock space for
relativistic particles (with Ep = p2 + m2 ). Let us Fourier transform the
scalar field φ: �
d3 p
φ(t, x) = 3
φ̃(t, p)eip·x .
2Ep (2π)
The Klein-Gordon equation

(∂02 − ∇2 + m2 )φ = 0

gives an ordinary differential equation for φ̃(t, p):

∂ 2 φ̃
= −(p2 + m2 )φ̃.
∂t2
The general solution is

φ̃(t, p) = e−iEp t a(p) + eiEp t c(p).

2
It will be convenient to rename c(p) = b(−p)† . Then
�
d3 p � ip·x �
φ(t, x) = 3
ap e + b(p)† e−ip·x .
2Ep (2π)

Similarly �
† d3 p � ip·x † −ip·x
�
φ(t, x) = b p e + a(p) e .
2Ep (2π)3
We can invert these formulas and express a, b, a† , b† in terms of φ, φ̇ and φ† , φ̇† .
(This is an exercise). Then the commutation relations of a, a† , b, b† turn out

[a(p), a† (q)] = (2π)3 2Ep δ 3 (p − q), (1)

† 3 3
[b(p), b (q)] = (2π) 2Ep δ (p − q), (2)

with all other commutators vanishing. Thus it is natural to postulate the

existence of the vacuum state |0�, annihilated by all a(p) and b(p). Then
the Hilbert space is the bosonic Fock space built on the sum of two copies
of L2 (R3 ). Why two copies? We expected only one! Resolution: we have an
additional quantum number which distinguishes b-particles from a-particles.
The b-particles are actually anti-particles of a-particles! (see below).
Hamiltonian becomes
�
1 d3 p � �
H= 3
Ep a† (p)a(p) + a(p)a† (p) + b† (p)b(p) + b(p)b† (p) .
2 2Ep (2π)

Let us normal-order it:

�
−3
H = V (2π) d3 pEp + . . .

Thus the vacuum energy density is divergent. If we cut off the integral at

|p| = Λ, we find
Λ4
E0 = 2
8π
This is the simplest example of an ultraviolet divergence.

3 Noether’s theorem
(Reading: section 22, pp. 132-135).

3
 Noether’s theorem says that for every continuous symmetry of the ac-
tion there is a current jµ (vector-valued function made of fields and their
derivatives) which satisfies
∂µ j µ = 0.
This implies that �
Q= d3 xj 0 (t, x)

is time-independent. I.e. it is a conserved charge. In the Hamiltonian for-

malism this is expressed as Q, H = 0, which upon quantization becomes
[Q, H] = 0.
Let us derive the Noether theorem for a theory of scalar fields with a
Lagrangian L(φa ). Suppose the infinitesimal symmetry transformation is
given by
δφa = � · v a (φ).
Consider now the same transformation, but with � a function of x. Since the
action is of first order in derivatives of φ, the variation of the action must be
of the form �
δS = d4 xj µ ∂µ �,

for some j µ independent of �. But on equations of motion this must vanish,

for arbitrary �. Therefore ∂µ j µ = 0.
Let us apply this procedure to the complex scalar field φ and the trans-
formation
δφ = i�φ, δφ† = −i�φ† .
The variation of the action is
�
� �
δS = i d4 x∂µ � −φ† ∂ µ φ + ∂ µ φ† φ .

Hence the current is � �

jµ = −i φ† ∂µ φ − ∂µ φ† φ .
What is the meaning of the corresponding charge, in terms of particles?
�
d3 k � † �
Q= 3
a (k)a(k) − b† (k)b(k) .
(2π) 2Ek

I.e. it is the number of particles minus the number of anti-particles.

4
 Let me consider another example: translational symmetry. Here

δφ = �µ ∂µ φ.

Note that here � has a vector index. Thus we expect

�
δS = d4 x∂ν �µ Tµν

for some tensor T . (It is called the stress-energy tensor). Let us determine
T . For constant � we have
�
δS = d4 x�µ ∂µ L.

This indeed vanishes for constant � (by integration by parts), but does not
vanish for nonconstant �. But for nonconstant � we also get other terms in
the variation:
� � �
4 µ ν ∂L
δS = d x −∂µ � L + ∂µ � ∂ν φ .
∂∂µ φ

Hence
∂L
Tνµ = ∂ν φ − δνµ L.
∂∂µ φ
For the free scalar field, we get

Tνµ = −∂ µ φ† ∂ν φ + ∂ µ φ∂ν φ† − δνµ L.

For example:
T00 = ∂0 φ† ∂0 φ + ∇φ† ∇φ + m2 φ† φ.
The corresponding “charge” is the energy (i.e. the Hamiltonian). Similarly,

Ti0 = ∂0 φ† ∂i φ + ∂i φ† ∂0 φ.

The corresponding charge is minus the momentum. Indeed, after expressing

in terms of a and b get
� �
d3 k � �
3 0
d xTi = − 3
ki a† (k)a(k) + b† (k)b(k) .
(2π) 2Ek

5
 Starting from a symmetry, one can get a conserved charge. Conversely,
starting from a conserved charge Q, one can try to get a symmetry transfor-
mation, by letting
δF = {Q, F }.
Then δH = 0, and δ commutes with time translations.
One can show directly that Q is the generator of symmetry transforma-
tions: �
Q = − d3 xpi δφi , {Q, φj } = δφj .

In quantum theory:
[Q, φj ] = −iδφj .
A finite transformation is

φ → U −1 φU, U = exp(−itQ).

In relativistic field theory, we are interested in translations and Lorenz

transformations. Together they form Poincare group:

x → Λx + a.
�
Generators of translations are momenta Pµ = d3 xTµ0 . Lorenz transforma-
tions act by
φ� (x) = φ(Λ−1 x).
Infinitesimal transformation Λ = 1 + ω gives
12
δφ = (xµ ∂ ν − xν ∂ µ )φ.
ω µν
We can achieve this by letting
�
� �
M = d3 x xµ T 0ν − xν T 0µ .
µν

This suggests that the conserved current for the Lorenz transformations is

Lρµν = xµ T ρν − xν T ρµ .

It is conserved because T µν = T νµ .

6
 It is interesting to compute Poisson brackets or commutator of all these
generators. For example:

[M µν , M ρσ ] = i(g µρ M νσ − (µ ↔ ν)) − (ρ ↔ σ).

This algebra characterizes infinitesimal Lorenz transformations. Infinitesimal

rotations are
1
Ji = �ijk M jk ,
2
i0
infinitesimal boosts are Ki = M . In terms of J and K we have

[Ji , Jj ] = i�ijk Jk , [Ji , Kj ] = i�ijk Kj , [Ki , Kj ] = −i�ijk Jk .

The other commutators are

[P µ , M ρσ ] = i(g µσ P ρ − (ρ ↔ σ)).

4 The spin-statistics relation

Let us compute the commutator of φ(x) and φ(y) (in the real case). It
vanishes outside the light-cone.
Now let us try to construct a similar theory based on the fermionic Fock
space. The anticommutator comes out to be nonvanishing outside the light-
cone,so this is unacceptable.

5 Scattering theory
First: �
1 1 2iωt †
d3 xe−ikx φ(x) = a(k) + e a (−k),
2ω 2ω
�
i i
d3 xe−ikx ∂0 φ = − a(k) + e2iωt a† (−k).
2 2
Hence � �
↔
−ikx
a(k) = 3
d xe (i∂0 φ + ωφ) = i d3 xe−ikx ∂ 0 φ.