Week 1: Review material 1

Reading material from the books

• Burgess-Moore, Chapter 1

• Weiberg, Chapter 2

1 The irreducible representations of the Poincare

group.
Relativistic systems are those invariant under the Poincare group. If we add
quantum mechanics to the problem, we need to consider how the Poincare
group acts on the Hilbert space of states of the system.
In quantum mechanics continuous symmetries have associated to them
operators that generate infinitesimal transformations. These operators arise
from the quantization of the associated Noether charge of a classical system.
The Poincare group has ten generators: 6 of them associated to the
Lorentz group of transformations, and four more associated to translations
in space and time.
The generators are usually called Mµν or Jµν and Pµ . These are four-
vector indices. Moreover, one can check that these satisfy the following
Lie-algebra relations:

[Mµν , Mρσ ] = iηµρ Mνσ + iηµσ Mρν − (µ ↔ ν) (1)

[Mµν , Pσ ] = iηµσ Pν − iηνσ Pµ (2)
[Pµ , Pν ] = 0 (3)

In the conventions above, all M and P are Hermitian, hence the factors
of i. We also have the convention that ηµν = diag(−1, 1, 1, 1) and it is used to
raise and lower indices. By convention, the operator P 0 is called the energy
operator and it should be positive relative to a stable vacuum.
One can see that P transforms as a four vector under the Lorentz group.
The energy operator is what defines evolution in a quantum system. The
operators that commute with P 0 are of major importance to classify states.
1
David
c Berenstein 2009

1
 From the Lorentz algebra, these are the momentum operators P i ∼ P~
and angular momentum operators J i ∼ 21 ǫijk Jkj .
Usually it is convenient to find a maximally commuting set of operators.
Thus, we will diagonalize all of the P µ operators, to have eigenvalue pµ .
Thus, our states are classified by momentum as follows

P µ |p, σi = pµ |p, σi (4)

We are using the convention that σ is the additional degeneracy of a state,

while we use the convention that the lower case letter is an eigenvalue.
If we perform a Poincare transformation, it can always be exponentiated
from M, P , and moreover we can always write it in the form

U ∼ exp(i/2θµν Mµν ) exp(iaµ Pµ ) (5)

where the translation is taken first. It is also common to denote the the
rotations and boosts by a matrix Λµν , that acts on points in spacetime by
sending xν → (xµ )′ = Λµν xν .
The effect on the operator space of states is to send

P µ → (P ν )′ = (Λνµ )−1 P µ (6)

This is, the transformation should be

U(Λ)P µ U(Λ−1 ) = (Λνµ )−1 P µ (7)

and because it is a symmetry of the quantum system it should be a unitary

transformation.
The equation implies that it sends pµ → (pν )′ = Λνµ pµ for the eigenvalues.
It can also act on the σ labels.
The difference between Λ and Λ−1 is due to the difference between Active
and Passive transformations. The transformations are active on states, but
they should be passive on operators (or viceversa).
Given a four momentum pµ , it is easy to show that p2 = pµ pµ = −m2 is
invariant. One can also check that the operator P µ Pµ commutes with all of
the Lorentz generators. **** Check this as an exercise ****
This property makes it into a Cassimir operator of the Lorentz algebra.
It is called the mass squared operator of a representation. States can be
interpreted as particles only if m2 ≥ 0 and p0 ≥ 0. A state with P = 0 is
called a vacuum.

2
 The reason for this constraint is that if the vacuum is translation invariant
( a property under Lorentz transformations) and if it the lowest energy state
( as required for stability), then all other excitations should have positive
energy. In particular, if one has a particle with m2 < 0, one can show that
there is a Lorentz frame where the vacuum would not be the Lowest energy
state, and hence it would be unstable.
Given this information, we find that p0 > 0 and we have two options,
m > 0 and m2 = 0.
2

In the first case, m2 > 0, we can go to a Lorentz frame where p~ vanishes.

This special Lorentz frame is the center of mass frame. Any element of the
Hilbert space of states with fixed momentum can therefore be transformed
to elements of the Hilbert space in the center of mass frame.
Since the Lorentz transformations also involve rotations, we see that we
still have to consider the spectrum of states under rotations. Since in the
center of mass frame J 1,2,3 commute with P 0 , and the p~ vanishes, the rotations
rotate between different σ labels, but keep the condition p~ = 0. We find that
the standard theory of angular momentum holds.
This group of rotations in the center of mass frame is called the little
group of the representation, and this method of building representations is
called the method of induced representations. Given a representation of the
group of rotations (SO(3)), that is classified by a spin j, and given the value
of m2 > 0 we can build a complete representation of the Lorentz group. It
will have 2j + 1 components (usually classified by jz ), and J 2 = j(j + 1) is
the square of the angular momentum operator.
This is called the spin of the representation. It is a quantum number of
a particle. The state is degenerate.
Notice that we can pick the components of J that are transverse to the
center of mass frame by considering the combinations
1
W µ ∼ ǫµνρσ Mνρ pσ (8)
2
It is obvious that W µ pµ = 0, and in the center of mass, only W 1,2,3 are
non-zero.
When we look more closely and consider the operator
1
W µ ∼ ǫµνρσ Mνρ Pσ (9)
2
we see that this is free of ordering ambiguities: the P and M in the definition

3
commute with each other because of the ǫ symbol. **** Check this as an
exercise ****
Because of this, it is obvious that W · P = 0 as an operator equation.
Also, in the center of mass frame, we find that W i ∼ mJ i and that
therefore W 2 = W µ Wµ ∼ m2 J 2 = m2 j(j + 1).
We can show further that [W 2 , P µ] = 0 and that [W 2 , Mµν ] = 0. ****
Check this as an exercise ****
Thus, the operator combination W 2 is also an invariant of the Poincare
group. This is the one that carries the information on spin of a state.
For single particle states they are classified uniquely by mass and spin
(here spin is in the same sense as in non-relativisiitic mechanics). There can
be additional degeneracies in mass and these would be due to symmetries.
Also, because boosts along the z direction commute with rotations along
z, we see that if start a system at rest and boost it, we keep the angular
momentum along the z axis fixed.
Thus, it is convenient to introduce the projection of spin along the axis
of momentum for a moving particle as a convenience reference on how to get
the state from zero momentum.
This property is called helicity and it depends on the frame of reference.
For massless particles, P 2 = 0, we can only go to a frame where P ∼
(p, p, 0, 0). The little group in this case is called ISO(2) (the standard group
of Euclidean geometry) (see Weinberg for more details).
This has one compact generator of rotations, and two ’translations’. If
single particle states are to have a discrete set of states at fixed momentum,
only the rotations are allowed to have non-trivial eigenvalues on the Hilbert
space.
This rotation is the spin along the axis of momentum and it is still a
valid object, so the notion of helicity remains. In fact, helicity is the only
quantum number that remains and it can not change when we do Lorentz
transformations and it is the same for all states in the representation of the
Lorentz group.
This is the complete set of representations of the Lorentz group that are
used in particle physics.
There is one extra ingredient that is required. This is how we choose the
normalization of the states |p, σi. We choose them so that

hq, σ ′ |p, σi = f (Ep )(2π)3 δ 3 (~p − ~q)δσ,σ′ (10)

4
 We should remember that in non-relativistic situations, we choose f (Ep ) =
1. This is a normalization of the plane wave function so that the density of
particles is fixed and the same for all velocities.
in a relativistic theory, it is convenient to normalize it so that the density
of particles in the rest frame is constant. This means√ that in a moving frame
−1
there will be a Lorentz contraction factor of γ = 1 − v 2 , so the density
gets multiplied by γ. Now E = mγ, so it is better to choose f (E) = 2E.
This differs from the naive density normalization by a factor of m, so it is
more appropriate to talk about this as having normalized energy density per
unit volume (this also works well for massless particles which is why it is
preferred over factors of γ).
This is also the inverse of the normalization of the relativistically invariant
integral over the on-shell momenta, given by

d4 p d3 p
Z Z
2 2 0
(2π)δ(p + m ))θ(p )G(p) ∼ G(p) (11)
(2π)4 (2π)3 2E
p
In G(p) over three momenta, the energy is evaluated on-shell: p0 = + p~2 + m2 .
Notation: integrals over p will always be written with factors of 2π to
the appropriate number of momenta in the denominator, and δ functions in
momentum variables will always carry factors of 2π in the numerator (one
factor of 2π per δ-function).
So this describes the one-particle states in a relativistic quantum field
theory.

2 Fock space: raising and lowering operators

Having described the single particle states, we can now consider multiparticle
states. These are described by taking the Hilbert space to be given by

H = ⊕∞
i=0 Hi (12)

and where the Hilbert space of states H0 = |0i is called the vacuum and has
a single basis element. It is normalized to one h0|0i = 1.
The space H∞ is the single particle set of states (it looks as above with
various masses and particle species. It can include composite particles (a
hydrogen atom).

5
 In perturbation theory we expect that interactions are weak between par-
ticles, so an initial condition with some fixed number of particles stays ap-
proximately that way for a long time until we get an interaction and particles
are produced/destroyed. Thus in the free field limit we can describe a col-
lection of many particles with the single particle states as building blocks.
The states with N particles are schematically a product ⊗N i=1 H1 where
identical particles can be exchanged and the answer does not change.
The particles are classified into bosons and fermions according to spin
(the spin-statistics theorem). Thus a two particle state can be thought of as
being described by

|p1 , σ1 ; p2 , σ2 i ≃ ±|p2 , σ2 ; p1 , σ1 i (13)

where the minus sign applies only if the two particles are fermions. In the
convention above, the σ might includes the identity of the particle. We would
normalize these with the product of the norms of the one-particle states, with
statistics taken into account:

hq1 , τ1 ; q2 , τ2 |p1 , σ1 ; p2 , σ2 i = (2E1 )(2π)3 δ 3 (p1 − q1 )δσ1 ,τ1 (14)

×(2E2 )(2π)3 δ 3 (p2 − q2 )δσ2 ,τ2 (15)
± [(q1 , τ1 ) ↔ (q2 , τ2 )] (16)

It is convenient to introduce raising and lowering operators for particles:

operators that create single particle states. They are defined by

a† (p, σ)|0i = |p, σi (17)

And more schematically

a† (p, σ)|αi = |p, σ; αi ∼ |p, σi ⊗ |αi (18)

for |αi ∈ H>0 .

The lowering operators a(p, σ) are the adjoint of a† (σ). Notice that

h0|a† |αi = 0, (19)

for all |αi, so that if we take the adjoint of this equation we find that

a|0i = 0 (20)

is necessarily zero (it is orthogonal to every other state).

6
 Exercise: Show that if a† (q, σ) is fermionic, then (a† )2 (q, σ)|αi = 0
(Calculate the norm of the state).
It is easy to show that the following commutation relations are satisfied

a(p, σ)a† (q, σ ′ ) ∓ a† (q, σ ′ )a(p, σ) = 2Ep (2π)3 δ 3 (p − q) (21)

The minus sign is if any of a, a† are bosons and the plus sign is for fermions
only. **** Check this as an exercise ****
The Hilbert space H is called a Fock space.
It can be proved by induction on the number of particles of a state that
any operator acting on H can be written as a Taylor series in the raising
and lowering operators with all lowering operators to the right of the raising
operators.
In the free theory, we have that momentum is additive
M
X
µ
P |p1 , σ1 ; . . . ; pN , σN i = pi |p1 , σ1 ; . . . ; pN , σN i (22)
i=1

and that the number operator N̂ has the representation

N̂ |p1 , σ1 ; . . . ; pN , σN i = N|p1 , σ1 ; . . . ; pN , σN i (23)

Thus it is easy to show that

[N, a† ] = a† (24)

and that similarly [N, a] = a. This defines the operator N.

The number operator is given by the following polynomial in the raising
and lowering operators
3
N = σ (2π)d3p2Ep a† (p, σ)a(p, σ)
P R
(25)
3 3
= σ,σ′ (2π)d 3p2Ep (2π)d 3q2Eq a† (p, σ)a(q, σ ′)(2π)3 (2Eq )δ 3 (p − q)δσ,σ′ (26)
P R R

**** Check this as an exercise ****

Also check that
d3 p
Z
µ
P = 3
pµ a† (p, σ)a(p, σ) (27)
(2π) Ep
and that the corresponding operators transform as in equation 7

7
 Notice that in the definitions above we have very pedantically included
all the δ functions and integrals with the correct Lorentz factors and factors
of (2π)3 .
The Lorentz transformations are encoded into U(Λ)|p, σi ∼ |Λp, Λσi,
because we chose an invariant normalization of the density in the rest frame,
and so
U(Λ)a† (p, σ)U −1 (Λ) = a† (Λp, Λσ) (28)
You should check that this makes P µ into a four vector in the Lorentz sense.