Professional Documents
Culture Documents
Poincare Chapters 1&2
Poincare Chapters 1&2
Ave Khamseh
Supervisor: Professor Luigi Del Debbio,
University of Edinburgh
Last updated:
3 Supersymmetry 54
3.1 Extension to graded algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.2 Introducing supersymmetric algebras . . . . . . . . . . . . . . . . . . . . . . . 57
3.3 Irreducible representations of Supersymmetry algebra . . . . . . . . . . . . . 67
3.3.1 Massless case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
3.3.2 Massive case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
iii
Chapter 1
⇤µ⇢ ⇤⌫ ⌘ ⇢ = ⌘ µ⌫ (1.3)
The aim here is to show that these transformations form a group, i.e. we need to prove
the group properties; closure and the existence of the identity element and that of the in-
verse.
1
2 CHAPTER 1. THE POINCARÉ GROUP
¯ µ x0⇢ + āµ = ⇤
x00µ = ⇤ ¯ µ (⇤⇢ x⌫ + a⇢ ) + āµ = (⇤
¯ µ ⇤⇢ ) x ⌫ + ( ⇤
¯ µ x⌫ + āµ ) (1.5)
⇢ ⇢ ⌫ ⇢ ⌫ ⇢
| {z } | {z }
Lorentz transf another translation
When ⇤µ⌫ and ⇤¯ µ⌫ both satisfy the Lorentz condition, so does their product and hence
we have the composition rule for these transformations:
as required.
as required.
To every transformation T , there corresponds a unitary linear operator that acts on vectors
in Hilbert space such that it takes ! U (⇤, a) and obeys the composition rule
U (1, 0) (1.10)
and
1 1
U (⇤ , ⇤ a) (1.11)
respectively.
The whole group of such transformation is known as the inhomogeneous Lorentz group or
Poincaré group . Now we will consider two di↵erent subgroups of this group.
The first one is the group of transformations with no translations present i.e. aµ = 0. This
group is known as the homogeneous Lorentz group or simply the Lorentz group.
1.1. CHARACTERISATION OF THE POINCARÉ GROUP 3
Using the Lorentz condition 1.2, the property of the product of determinants and the fact
that det(⌘) = 1 we have
(det ⇤)2 = 1 ) det ⇤ = ±1 (1.12)
For ⇤µ⌫ and ⇤
¯ µ⌫ we have
¯ 0 =⇤
(⇤⇤) ¯ 0 ⇤0 + ⇤
¯ 0 ⇤1 + ⇤
¯ 0 ⇤2 + ⇤
¯ 03 ⇤ 30 (1.13)
0 0 0 1 0 2 0
⌘µ⌫ ⇤µ0 ⇤⌫0 = ⌘00 = +1 =) (⇤00 )2 = 1 + ⇤i0 ⇤i0 = 1 + ⇤0i ⇤0i (1.14)
We know the identity is where ⇤ = 1 and so in order to preserve continuity we need to take
(⇤00 )2 1 (see Figure 1.1) and det ⇤ = 1. This is to ensure that they have the same sign as
the identity since we want to be able to obtain any Lorentz transformation from the identity
by a continuous change of parameters. Hence it forms a subgroup of Lorentz transformations
know as the proper orthochronous Lorentz group.
4 CHAPTER 1. THE POINCARÉ GROUP
but since ⌘ is symmetric with respect to the change of its two indices
!⇢ + ! ⇢ = 0 ) !⇢ = ! ⇢ (1.18)
i.e. !µ⌫ is a rank-2 antisymmetric tensor in four dimensions and so it has 6 independent
components with zeros on the diagonal. On the other hand, ✏µ is a 4-vector so it has 4
components. Therefore in order to specify a Poincaré transformation we need 10 independent
parameters.
For an infinitesimal Lorentz transformation one can make a Taylor expansion as follows
1
U (1 + !, ✏) = 1 i!µ⌫ M µ⌫ + i✏µ P µ + . . . (1.19)
2
where M µ⌫ and P µ are Hermitian operators so that U is a unitary operator.
Here is a good point to stop and explore these operators in more detail.
1.2. THE POINCARÉ ALGEBRA 5
We can make an infinitesimal change in position and time coordinates of a quantum state
and then Taylor expand as follows
d
(x + ✏, t + ⌧ ) = (x, t) + i✏ · ( ir ) i⌧ (i ) (1.20)
| {z } dt
| {z }
P̂
Ĥ
In other words, the momentum operator and the Hamiltonian correspond to space and time
translations respectively. Putting them together, we form the momentum-energy four vector
P µ which corresponds to translations in space-time. By the same analogy, the angular
momentum operator is the generator of space rotations (See later (1.77) for a more rigorous
proof).
Now consider the general Lorentz transformations from one frame to another i.e. a boost.
We will now prove that boosts are rotations in Minkowski space, i.e. hyperbolic rotations.
Consider a boost in the x direction:
8
>
>
> t0 = (t vxc2
)
>
<x0 = (x vt)
(1.21)
>
>
> y0 = y
>
:z 0 = z
v
Since we are working in natural unit, c = 1 and so we can define = c = v. Therefore the
transformation can be written as follows:
0 01 0 10 1
t 0 0 t
Bx 0 C B 0 0C BxC
C B
B C=B C (1.22)
@y 0 A @ 0 0 1 0A @y A
z0 0 0 0 1 z
Since y and z do not change in this transformation, for simplicity, we may only consider t
and x from now on.
Define the rapidity variable such that tanh = , so
1 1
=p =p = cosh (1.23)
1 1 tanh2
sinh
tanh = ) sinh = (1.24)
cosh
Hence ✓ 0◆ ✓ ◆✓ ◆
t cosh sinh t
= (1.25)
x0 sinh cosh x
6 CHAPTER 1. THE POINCARÉ GROUP
A nice trick: Consider two contracted tensors !µ⌫ M µ⌫ . It is possible to prove that when
!µ⌫ is symmetric (antisymmetric), M µ⌫ has to be symmetric (antisymmetric) as well.
Let
1 1
M µ⌫ = (M µ⌫ + M ⌫µ ) + (M µ⌫ M ⌫µ ) (1.26)
2| {z } 2 | {z }
µ⌫ µ⌫
Symm part M(s) AntiSymm part M(a)
Therefore, going back to equation 1.19 and knowing that ! is antisymmetric, it is immedi-
ately implied that M is antisymmetric with respect to the exchange its two indices and so it
has 6 independent components.
M µ⌫ = M ⌫µ (1.29)
We now know that pµ is the 4-momentum operator since it corresponds to translations. Also,
M µ⌫ is a rotation operator which contains the usual angular momentum 3-vector as well is
the boost 3-vector i.e. 6 independent components, as expected.
1 1 1
U (⇤, a)U (1 + !, ✏)U (⇤, a) = U (⇤(1 + !)⇤ , ⇤✏ ⇤!⇤ a) (1.30)
1 1
U (⇤, a)[ !µ⌫ M µ⌫ + ✏µ P µ ]U 1
(⇤, a) = (⇤!⇤ 1
)⇢ M ⇢ + (⇤✏ ⇤!⇤ 1
a)⇢ P ⇢ (1.31)
2 2
1.2. THE POINCARÉ ALGEBRA 7
U (⇤, a)M µ⌫ U 1
(⇤, a) = ⇤⇢µ ⇤ ⌫ (M ⇢ a⇢ P + a P ⇢ ) (1.33)
U (⇤, a)P µ U 1
(⇤, a) = ⇤⇢µ P ⇢ (1.34)
Note: The equation above is simply stating how the operator P transforms under U , writing
it as a linear combination of operators on the right hand side.
using 1.11 to first order in ! and ✏ we have,
1 1 1 1 1
U (⇤, a) = U (⇤ , ⇤ a) = U ((1 + !) , (1 + !) ✏)
(1.35)
= U (1 !, (1 !)✏) = U (1 !, ✏)
i i
(1 !↵ M ↵ + i✏↵ P ↵ )P µ (1 + ! ⇠ M ⇠ i✏ P )
2 2
i i
=P µ !↵ M ↵ P µ + i✏↵ P ↵ P µ + ! ⇠ P µ M ⇠ i✏ P µ P
2 2 (1.36)
1 1
=P + i( !⇢ M P + ✏⇢ P P + !⇢ P µ M ⇢
µ ⇢ µ ⇢ µ
✏⇢ P µ P ⇢ )
2 2
1
=P µ + i[ !⇢ M ⇢ + i✏⇢ P ⇢ , P µ ]
2
and the RHS
⇤ ⇢µ P ⇢ = ( ⇢
µ
+ !⇢ µ )P ⇢ (1.37)
Setting LHS = RHS,
1
i[ !⇢ M ⇢ + i✏⇢ P ⇢ , P µ ] = !⇢ µ P ⇢ (1.38)
2
8 CHAPTER 1. THE POINCARÉ GROUP
⇤⇢µ ⇤ ⌫ M ⇢ =( µ
⇢ + !⇢ µ )( ⌫
+ ! ⌫ )(M ⇢ ✏⇢ P + ✏ P ⇢ )
µ⌫
=M + !⇢ µ M ⇢⌫ + ! ⌫ M µ ✏µ P ⌫ + ✏⌫ P µ
=M µ⌫ + ⌘ µ !⇢ M ⇢⌫ + ⌘ ⌫⇢ ! ⇢ M µ ✏µ P ⌫ + ✏⌫ P µ
1
=M µ⌫ + [⌘ µ !⇢ M ⇢⌫ + ⌘ µ⇢ ! ⇢ M ⌫ + ⌘ ⌫⇢ ! ⇢ M µ + ⌘ ⌫ !⇢ M µ⇢ ] ✏µ P ⌫ + ✏⌫ P µ
2
1
=M µ⌫ + [⌘ µ !⇢ M ⇢⌫ ⌘ µ⇢ !⇢ M ⌫ ⌘ ⌫⇢ !⇢ M µ + ⌘ ⌫ !⇢ M µ⇢ ] ✏µ P ⌫ + ✏⌫ P µ
2
1
=M + !⇢ [⌘ µ M ⇢⌫ ⌘ µ⇢ M ⌫ ⌘ ⌫⇢ M µ + ⌘ ⌫ M µ⇢ ] ✏µ P ⌫ + ✏⌫ P µ
µ⌫
2
(1.40)
Equating we get,
1 1
i[ !⇢ M ⇢ + ✏⇢ P ⇢ , M µ⌫ ] = !⇢ [⌘ µ M ⇢⌫ ⌘ µ⇢ M ⌫
⌘ ⌫⇢ M µ + ⌘ ⌫ M µ⇢ ] ✏µ P ⌫ + ✏⌫ P µ
2 2
(1.41)
Finally, we extract the coefficients of ! and ✏ and use the fact that M µ⌫ is antisymmetric,
[M µ⌫ , M ⇢ ] = i(⌘ µ⇢ M ⌫ ⌘ µ M ⌫⇢ ⌘ ⌫⇢ M µ + ⌘ ⌫ M µ⇢ ) (1.42)
and
[P ⇢ , M µ⌫ ] = i(⌘ ⇢µ P ⌫ ⌘ ⇢⌫ P µ ) (1.43)
On the other hand, equation 1.38 implies the following,
[P µ , P ⇢ ] = 0 (1.44)
P = P 1, P 2, P 3 (1.45)
J = M 23 , M 31 , M 12 (1.46)
K = M 01 , M 02 , M 03 (1.47)
1.2. THE POINCARÉ ALGEBRA 9
[Ki , Pj ] = iH ij (1.52)
1 1
Wµ = ✏µ⌫⇢ M ⌫⇢ P = ✏µ⌫⇢ P ⌫ M ⇢ (1.55)
2 2
which has the following properties,
Wµ P µ = 0 (1.56)
1 1 1
Wµ P µ = ✏µ⌫⇢ M ⌫⇢ P P µ = ✏µ⌫⇢ M ⌫⇢ P µ P = ✏ ⌫⇢µ M
⌫⇢
P Pµ
2 2 2 (1.57)
1
= ✏µ⌫⇢ M ⌫⇢ P P µ = Wµ P µ
) Wµ P = 0µ
2
[Wµ , P⌫ ] = 0 (1.58)
10 CHAPTER 1. THE POINCARÉ GROUP
Proof:
1 1
[W µ , P ↵ ] = ✏µ⌫⇢ ([P⌫ , P ↵ ]M⇢ + P⌫ [M⇢ , P ↵ ]) = ✏µ⌫⇢ ⌘ ↵ ([P↵ , P ]M⇢ + P⌫ [M⇢ , P ])
2 2
i µ⌫⇢ ↵ i µ⌫↵
= ✏ ⌘ P⌫ (⌘ ⇢ P ⌘ P⇢ ) = (✏ P⌫ P ✏µ⌫⇢↵ P⌫ P⇢ )
2 2
i
= (✏µ⌫↵ P⌫ P + ✏µ ↵⌫ P P⌫ ✏µ⌫⇢↵ P⌫ P⇢ ✏µ⇢⌫↵ P⇢ P⌫ )
4
i
= (✏µ⌫↵ P⌫ P + ✏µ⌫⇢↵ P⌫ P⇢ ✏µ⌫⇢↵ P⌫ P⇢ ✏µ⌫↵ P⌫ P ) = 0
4
(1.59)
Where on the fourth line we have used the fact that P’s commute and then relabelled the
indices for the second and fourth terms. This commutation relation implies that the Pauli-
Lubanski vector is invariant under translations.
Now define
1
M̃µ⌫ = ✏µ⌫⇢ M ⇢ (1.60)
2
So that Wµ = P M̃µ⌫ . Multiplying both sides of equation 1.60 by ✏↵ µ⌫ one gets
⌫
⇣ ⌘
✏↵ µ⌫ M̃µ⌫ = ↵
⇢
↵
⇢ M
⇢
= 2M ↵ (1.61)
where we have used the identity for the four dimensional Levi-Civita tensor. Therefore,
1 ↵
M↵ = ✏ µ⌫
M̃µ⌫ (1.62)
2
Let us determine how M̃µ⌫ transforms under a Lorentz transformation by considering
1 1
[M̃µ⌫ , M↵ ] = ✏[M ⇢ , M↵ ] = ✏µ⌫ ⇢ [M⇢ , M↵ ]
2 2
i ⇢ ⇣ ⇢↵ ⌘
= ✏µ⌫ ⌘ M ⌘ ⇢ M ↵ ⌘ ↵ M ⇢ + ⌘ M ⇢↵
2
i⇣ ⌘ ⇣ ⌘
= ✏µ⌫↵ M ✏µ⌫ M ↵ + ✏µ⌫↵⇢ M⇢ ✏µ⌫ ⇢ M⇢↵ = i ✏µ⌫ ⇢ M⇢ ✏µ⌫ ⇢ M⇢↵
2✓ ◆
1 1 i⇣ ⌘
= i ✏µ⌫ ⇢ ( ✏⇢ ⌧ M̃ ⌧ ) ✏µ⌫ ⇢ ( ✏⇢↵⌧ M̃ ⌧ ) = ✏µ⌫ ⇢ ✏ ⌧ ⇢ ✏µ⌫ ⇢ ✏↵⌧ ⇢ M̃ ⌧
2 2 2
i
= [ ⌘µ (⌘⌫⌧ ⌘↵ ⌘⌫ ⌘⌧ ↵ ) ⌘µ⌧ (⌘⌫ ⌘↵ ⌘⌫ ⌘↵ ) ⌘µ (⌘⌫ ⌘↵⌧ ⌘⌫⌧ ⌘ ↵ ) ↵ $ ]M̃ ⌧
2
i
= [ 2⌘µ M̃⌫↵ ⌘↵ M̃µ⌫ + ⌘⌫ M̃µ↵ ⌘⌫ M̃↵µ + ⌘↵ M̃⌫µ ↵ $ ]
2
= i[⌘µ↵ M̃⌫ ⌘µ M̃⌫↵ + ⌘⌫ M̃µ↵ ⌘⌫↵ M̃µ ]
(1.63)
1.2. THE POINCARÉ ALGEBRA 11
which is the same result as equation 1.42. In other words, M̃µ⌫ behaves like a second rank
tensor under a Lorentz transformation. Note that on the fifth line we have used the identity
for 4-dimensional Levi-Civita tensors:
✏µ⌫ ⇢ ✏ ⌧ ⇢ = ⌘µ (⌘⌫⌧ ⌘↵ ⌘⌫ ⌘⌧ ↵ ) ⌘µ⌧ (⌘⌫ ⌘↵ ⌘⌫ ⌘↵ ) ⌘µ (⌘⌫ ⌘↵⌧ ⌘⌫⌧ ⌘ ↵) (1.64)
Proof:
Proof:
1 1 ⇣ ⌘
[Wµ , W↵ ] = ✏↵ ⌧ [Wµ , M P ⌧ ] = ✏↵ ⌧ [Wµ , M ]P ⌧ + M [Wµ , P ⌧ ]
2 2
1 i
= ✏↵ ⌧ ⌘ ⌘ [Wµ , M⇠ ]P ⌧ =
⇠
✏↵ ⌧ ⌘ ⌘ ⇠ (⌘⇠µ Wk ⌘µ W⇠ ) p⌧
2 2 (1.68)
i⇣ ⌘ i⇣ ⌘
= ✏↵ µ⌧ W ✏↵µ ⌧ W P ⌧ = ✏µ↵ ⌧ W + ✏µ↵ ⌧ W P ⌧
2 2
= i✏µ↵ ⌧ W P ⌧
12 CHAPTER 1. THE POINCARÉ GROUP
1 1
W 2 = W µ Wµ = ✏µ⌫⇢ M⌫⇢ P ✏µ↵ M ↵ P = (✏⌫⇢ µ ✏↵ µ )M⌫⇢ P M ↵ P
4 4
1
= [ ⌫↵ ⇢ ⌫ ⇢
↵
⌫ ⇢
↵
4
⌫ ⇢ ⌫ ⇢ ⌫ ⇢
+ ↵ + ↵ + ↵ ↵ ] M⌫⇢ P M ↵ P
1
= [ M↵ P M P↵ M ↵ P + M ↵ P + M↵ P + M P↵ ]M ↵ P
4
1 1 1
= M ↵ P M ↵ P + M P↵ M ↵ P + M ↵ P M ↵ P
2 2 2
1
= M ↵ P M ↵ P + M P↵ M ↵ P
2
1
= M↵ ⌘ ⌧ P ⌧ M ↵ P + M ⌘ ⌧ ↵ P ⌧ M ↵ P
2
1 h i
= M↵ ⌘⌧ + M ⌘⌧ ↵ M ↵ P ⌧ + i(⌘ ⌧ ↵ P + ⌘ ⌧ P ↵ ) P
2
1 i
= M↵ M ↵ P 2 (M P P + M↵ P ↵ P )
2 2
+M M ↵ P↵ P + i(4M P P + M ↵ P ↵ P )
1
= M↵ M ↵ P 2 + M M ↵ P P↵
2
(1.69)
on the sixth and tenth lines we have made use of the antisymmetry of M and equation 1.43 on
the eighth. W 2 is Lorentz invariant (since it is the length of a 4-vector) and hence commutes
with all the other generators.
The other Lorentz invariant entity is the mass operator P 2 = P µ Pµ which again, commutes
with all generators. Therefore, W 2 and P 2 are the Casimir operators for the Poincaré algebra.
Now, we let N ! 1 and expand the right hand side of the equation to first order in a using
equation 1.19 to get
i
U (1, a) = lim [1 + P µ aµ ]N (1.72)
N !1 N
1.2. THE POINCARÉ ALGEBRA 13
Hence the operator for pure translations acting on the Hilbert space is
µa
U (1, a) = eiP µ
(1.73)
and so for an infinitesimal rotation we do a Taylor expansion using the formula below for
small a,
X1
1
(x + a) = (a · r)n (x) = exp(a · r) (x) (1.76)
n!
n=0
Therefore,
U (R(✓), 0) = eiJ·✓ (1.79)
Pµ p, = pµ p, (1.80)
In general, it may be possible for C to be made block diagonal so that each block furnishes
a representation of the Poincaré group.Therefore, our goal from now on is to find the irre-
ducible representations of the Poincaré group.
V̄+ = {q µ : q 0 0, q 2 0} (1.84)
1.3. CONNECTION WITH ONE-PARTICLE STATES 15
Then
q ⌫ 2 V̄+ ) q 0µ = ⇤µ⌫ q ⌫ 2 V̄+ , 8 ⇤µ⌫ 2 SO(3, 1) (1.85)
Proof:
1 ⌧
q 0µ qµ0 = ⇤µ⌫ q ⌫ ⇤µ⌧ q⌧ = (⇤ ) µ ⇤ µ⌫ q ⌫ q ⌧ = ⌧
⌫q
⌫
q⌧ = q ⌧ q⌧ = q 2 0 (1.86)
and so the second condition is satisfied. As for the first one,
but since 0
p we are dealing with proper orthochronous subgroup we must have ⇤ 0 0 so
0
⇤ 0 > (⇤ 0 )0 2 2 0
1. Also, for light-like 4-vector invariant length q = (q ) 2 2
|q| and given
q0 0 by equation 1.84, it is implied that q 0 |q|. Therefore, 0 q 00 and so the first
condition is also satisfied. Q.E.D
Note that p2 and the sign of p0 (for p2 0) are left invariant by all proper orthochronous
Lorentz transformations. So one can choose a standard 4-momentum k µ and apply a Lorentz
boost such that
pµ = Lµ⌫ (p)k µ (1.89)
and define states with momentum p
k ! p ! ⇤p ! k (1.92)
16 CHAPTER 1. THE POINCARÉ GROUP
i.e. this transformation is a member of the subgroup of the homogeneous Lorentz group
consisting of Lorentz transformations W µ⌫ that leave the standard 4-momentum invariant:
W µ⌫ k ⌫ = k µ (1.93)
Since W leaves k invariant, the only degree of freedom that can change is . Also, for W̄ , W 2
little group, we have
X
D 0 (W̄ W ) k, 0 = U (W̄ W ) k, = U (W̄ )U (W ) k,
0
X X
=U (W̄ ) D 00 (W ) k, 00 = D 00 (W )[U (W̄ ) k, 00 ] (1.95)
00 00
XX
= D 00 (W )D 0 00 (W̄ ) k, 0
0 00
This is precisely the definition of the representation of a group. So it can be said that D(W )
furnish a representation of the little group.
So for a particular element of the little group
1
W (⇤, p) ⌘ L (⇤p)⇤L(p) (1.97)
Table 1.1 summarises di↵erent classes of 4-momenta and the corresponding little group ele-
ment.
We will now briefly discuss each case. The connection between the little group and the Pauli-
Lubanski vector defined in section 1.55 and a more detailed explanation of cases (a) and (c)
will be presented in the next two sections.
• Case (a): Since we are in the rest frame of a massive particle, all ordinary three
dimensional rotations leave k µ invariant. Therefore the little group is all rotations in
the three spacial dimensions i.e. SO(3).
• Case (b): This is not a physical state because the energy is negative. Since state of
lowest energy is always more favourable, we keep decreasing the energy down to 1
which is not physical. So we define the vacuum state to be at zero. The rotation group
is again, SO(3) for the same reason as above.
• Case (c): Covers particles of mass zero, since the mass operator p2 = 0. The energy p0
is more than zero. Therefore it represents a physical state. As well as that, the little
group is the group of Euclidean Geometry consisting of rotations and translations in
two dimensions, i.e. ISO(2).
18 CHAPTER 1. THE POINCARÉ GROUP
• Case (d): Again, it does not represent a physical state since the energy p0 < 0.
• Case (e): This is the space-like case and so it does not represent a physical state. The
little group is the group of rotations that that keep the third direction fixed i.e. 2+1
dimensional rotations keeping the third axis fixed SO(1, 2)
• Case (f): This is a physical state and it corresponds to the vacuum. The standard
4-vector is at the origin and so the all rotations in space-time leave k µ invariant. Hence
the little group is SO(1, 3).
Before we proceed to building the representations of the Poincaré group for the massive (a)
and massless (c) cases, let us consider the normalisation of the states.
We choose the states with standard momentum k µ to be orthonormal. Form quantum me-
chanics, we know that the scalar product of the eigenstates of a Hermitian operator with
eigenvalues k0 and k is
h k0 , 0 | k, i = 3 (k0 k) 0 (1.99)
Therefore, using 1.94 and 1.98 , it can be said that the representations of the little group are
unitary:
D† (W ) = D 1 (W ) (1.100)
However, for arbitrary momenta the scalar product is
1.90
h p0 , 0 | p, i =h p0 , 0 |N (p)U (L(p)) k, i = N (p)hU † (L(p)) p0 , 0 | k, i
1 1
=N (p)hU (L(p)) p0 , |
0 k, i = N (p)hU (L (p)) p0 , 0 | k, i
1.98 N ⇤ (p0 ) X
= N (p) h D 00 0 (W (L 1
(p), p0 )) L 1 (p)p0 , 00 | k, i
N ⇤ (L 1 (p)p0 )
00
N ⇤ (p0 ) X ⇤ (1.101)
=N (p) D 00 0 (W (L 1 (p), p0 ))h k0 , 00 | k, i
N ⇤ (k 0 ) 00
1.99
X
= N (p)N ⇤ (p0 ) D⇤ 00 0 (W (L 1 (p), p0 )) 00 (k0 k)
00
where on the fifth line we take N ⇤ (k) = 1 and on the sixth line the fact that the delta function
only picks up the term = 00 from the sum.
For p = p0 the transformation W (L 1 (p), p) = 1 and the scalar product of the two states is
The remaining task now is to find the proportionality factor relating (k0 k) and (p0 p).
1.3. CONNECTION WITH ONE-PARTICLE STATES 19
Consider the integral of an arbitrary scalar function f (p) over 4-momenta with p2 = M 2 0
and p0 > 0 i.e. (a) and (c),
Z Z
4 2 2
d p (p M )✓(p )f (p) = d3 p dp0 ((p0 )
2 0
p2 M 2 )✓(p0 )f (p0 , p)
Z p
2
= d3 p dp0 ((p0 ) ( p2 + M 2 )2 )✓(p0 )f (p0 , p)
Z p p
1
= d3 p dp0 p [ (p0 p2 + M 2 ) + (p0 + p2 + M 2 )]✓(p0 )f (p0 , p) (1.103)
2 p2 + M 2
Z p
1
= d3 p dp0 p (p0 p2 + M 2 )f (p0 , p)
2
2 p +M 2
Z p
1
= d3 p p f ( p2 + M 2 , p)
2 p2 + M 2
is a Lorentz invariant. Finally, the step function ✓(p0 ) also remains invariant under Lorentz
transformations since the sign of p0 > 0 does not change.
p2 M 2 is known as the mass shell and the invariant volume element is
d3 p
p (1.106)
p2 + M 2
We know that the LHS is invariant and so must be the RHS. We also know that the volume
element is invariant, and so the term in the brackets i.e.
p
p02 + M 2 3 (p0 p) = p0 3 (p0 p) (1.108)
does not change from going from one frame to another, then
p0 3
(p0 p) = k 0 3
(k0 k) (1.109)
In the two sections that follow, we will derive the representations of Poincaré group for
massive and massless cases (a,c) and find the connection between the corresponding little
group and the Pauli-Lubanski vector.
W0 = 0 ; Wi = MJi (1.112)
i.e. the spacial components of W are proportional to J i which are in turn the generators of
the little group SO(3).
Acting with Si (or more generally Ji ) on this orthonormal set of vectors, results in a linear
combination of {|s, 0 i} vectors, i.e what we have here is an irreducible representation which
is 2s + 1 dimensional.
In other words, the unitary representation of the little group SO(3) can be broken up into a
(j)
direct sum of the irreducible unitary representations D 0 (R).
These representations are built up using matrices for infinitesimal rotations Rik = ik + ⇥ik
where ⇥ik is antisymmetric with respect to the exchange of its two indices.
(s) i (s)
D 0 (1 + ⇥) = 0 ⇥ik (Sik ) 0 (1.114)
2
where W (⇤, p) = L 1 (⇤p)⇤L(p) is known as the Wigner rotation. In order to calculate this
rotation, one can apply a ’standard boost’ L(p)
0 p p p 1
p̂1 2 1 p̂2 2 1 p̂3 2 1
B p C
Bp̂ 2 1 1+( 1)p̂21 ( 1)p̂1 p̂2 ( 1)p̂1 p̂3 C
L(P ) = B 1 p 2 C (1.123)
@p̂2 p 1 ( 1)p̂2 p̂1 1 + ( 1)p̂22 ( 1)p̂2 p̂3 A
p̂3 2 1 ( 1)p̂3 p̂1 ( 1)p̂3 p̂2 1 + ( 1)p̂23
22 CHAPTER 1. THE POINCARÉ GROUP
q
pi 2 +M 2
such that p̂i =|p| with = p M . (For a more detailed discussion of how L(p) is actually
derived, see the proof below.)
Note that when ⇤µ⌫ is an arbitrary 3-dimensional rotation R, the Wigner rotation W (⇤, p) =
R, 8p.
Proof: We would like to go from a frame where the particle has momentum k, i.e. the rest
frame, to a frame where the particle is viewed to have momentum p. We first make a boost,
say in the third direction:
0 p 1
0 0 2 1
B 0 1 0 0 C
B(|p|) = B
@
C
A (1.124)
p 0 0 1 0
2 1 0 0
1
p
where we have used the fact that =p 2
) 2 1= .
1
However, in order to get to the more general form of L we need to apply a 3-dimensional
spacial rotation that takes us from the third direction into the direction of p̂ according to
Figure 1.3,
1.3. CONNECTION WITH ONE-PARTICLE STATES 23
1
L(p) = R(p̂)B(|p|)R (p̂) (1.125)
ẑ⇥p
i.e. a rotation by some angle ✓ about axis n, normal to the plane of z and p̂ with n = |p|
ẑ·p
and cos ✓ = |p| .
W (R, p) =L 1 (Rp)RL(p)
⇥ 1
⇤ 1 1
= R(Rp̂)B(|p|)R (Rp̂) RR(p̂)B(|p|)R (p̂)
(1.127)
1 1 1
=R(Rp̂)B (|p|) R (Rp̂)RR(p̂) B(|p|)R (p̂)
| {z }
R(✓)
Note that R 1 (Rp̂)RR(p̂) takes the 3-axis into the direction of p̂ then to Rp̂ and finally
back to the 3-axis. Since the 3-axis remains fixed under this transformation, it must be
equivalent to a rotation by some angle ✓ about the 3-axis i.e.
0 1
1 0 0 0
B0 cos ✓ sin ✓ 0C
R 1 (Rp̂)RR(p̂) = R(✓) = B@0
C (1.128)
sin ✓ cos ✓ 0A
0 0 0 1
It can easily be verified that R(✓) commutes with B(|p|) which implies that
1 1
W (R, p) =R(Rp̂)B (|p|)R(✓)B(|p|)R (p̂)
1.128
(1.129)
1
=R(Rp̂)R(✓)R (p̂) = R
Hence
W (R, p) = R (1.130)
Q.E.D
24 CHAPTER 1. THE POINCARÉ GROUP
Therefore, it can be concluded that under rotations, the states of a moving massive par-
ticle have the same transformations as that of the non-relativistic case. This can also be
extended to multi-particle states.
(W t)µ = (1 + ⇠, ↵, , ⇠) (1.131)
↵2 + 2
⇠= (1.132)
2
Therefore, it can be said that W µ⌫ acting on t has the same e↵ect as the Lorentz transfor-
mation S µ⌫ (↵, ), where S is
0 1
1+⇠ ↵ ⇠
B ↵ 1 0 ↵C
S µ⌫ (↵, ) = B
@
C
A (1.133)
0 1
⇠ ↵ 1 ⇠
But the question is how this matrix is actually derived. From equation 1.131, it can easily
be observed that the first column of such a 4⇥4 matrix must be as above. Also when S acts
on k it must leave it invariant and hence the last column. Now it remains to find the other
two columns. We will use the fact that S is a Lorentz transformation
⌘µ⌫ S µ⇢ S ⌫ = ⌘ ⇢ (1.134)
and as a result we may impose two conditions. The first one is that the ‘length’ of each
column must be invariant and equal to 1. The second one is that the columns must be
orthonormal to each other. We may also choose the block in the middle to be the identity
and calculate the rest of the entires based on that. So are task is to find x, y, u, v
1.3. CONNECTION WITH ONE-PARTICLE STATES 25
0 1
1 + ⇠ x? u? ⇠
B ↵ 1 0 ↵C
S µ⌫ (↵, ) = B
@
C
A (1.135)
0 1
⇠ y? v? 1 ⇠
For example, one can try ⇢ = = 2 Then, using the first condition
1 =S 02 S 02 S 12 S 12 S 22 S 22 S 32 S 32
(1.136)
=u2 02 12 v 2 ) u = ±v
Similarly by taking ⇢ = = 1 one gets x = ±y. Imposing the second condition and setting
⇢ = 2, = 3 and then ⇢ = 1, = 3 we get x = y = ↵ and u = v = . Hence the result.
We mentioned that acting on t, this matrix has the same ’e↵ect’ as W . This does not mean
that they are equal but it does imply that S 1 (↵, )W is a Lorentz transformation that
leaves the vector (1, 0, 0, 0) invariant. Therefore, it can only be a pure rotation. As shown
above, S µ⌫ also leaves (1, 0, 0, 1) invariant, i.e. the 3-axis remains fixed and so S 1 (↵, )W
must be a rotation by an arbitrary angle ✓ about the 3-axis
1
S (↵, )W = R(✓) (1.137)
where 0 1
1 0 0 0
B0 cos ✓ sin ✓ 0C
Rµ⌫ (✓) = B
@0
C (1.138)
sin ✓ cos ✓ 0A
0 0 0 1
From equation 1.137, it can be concluded that the most general element of the little group
has the form
W (✓, ↵, ) = S(↵, )R(✓) (1.139)
The aim now is to identify this group. It can be checked that for those transformations with
✓ = 0 we have R = 1 and
↵, ¯)S(↵, ) = S(¯
S(¯ ↵ + ↵, ¯ + ) (1.140)
and for those with ↵ = = 0 ) S = 1,
¯
R(✓)R(✓) = R(✓¯ + ✓) (1.141)
which is also additive. These transformations forming the two subgroups above are Abelian
i.e. their elements commute with each other. As well as that, this first subgroup mentioned
26 CHAPTER 1. THE POINCARÉ GROUP
is invariant such that its elements are transformed into other elements of the same subgroup
by any element of the group
1
R(✓)S(↵, )R (✓) = S(↵ cos ✓ + sin ✓, ↵ sin ✓ + cos ✓) (1.142)
which can easily be verified by simple matrix multiplication.
Therefore the product of any group elements can be calculated using the above equations.
These multiplication rules can be identified as those of the little group ISO(2) i.e. the group
of Euclidean Geometry consisting of rotations (by angle ✓) and translations (by vector (↵, ))
in two dimensions.
Here is good point to stop and turn our attention to the Pauli-Lubanksi vector.
Starting from the standard 4-momentum k µ = (, 0, 0, ) and the Pauli-Lubanski vector we
have,
1
W0 = ✏0µ⌫3 M µ⌫ k 3 = M 12 = J3 = W3 (1.143)
2
W1 = (M 23 + M 20 ) = (J1 + K2 ) (1.144)
W2 = (M 31 + M 01 ) = (J2 K1 ) (1.145)
Groups with no invariant Abelian subgroups are called semi-simple groups. They have certain
simple properties, hence the name. ISO(2), like the Poincaré group, is not semi-simple. We
will reproduce the Lie Algebra 1.146 to 1.148 using the general little group element W and
infinitesimal transformations.
W (✓, ↵, )µ⌫ = µ
⌫ + ! µ⌫ , (1.149)
On the other hand, W (✓, ↵, ) = S(↵, )R(✓) so using 1.133 and 1.138 and making the
approximation cos(✓) ⇡ 1, sin(✓) ⇡ ✓ and ⇠ = 0 for infinitesimal ✓, ↵, , we get
0 1
0 ↵ 0
B ↵ 0 ✓ ↵C
!µ⌫ = B@
C
A (1.150)
✓ 0
0 ↵ 0
1.3. CONNECTION WITH ONE-PARTICLE STATES 27
which is antisymmetric and has 6 independent components. and so the corresponding oper-
ator
U (W (✓, ↵, )) = 1 i↵A i B i✓J3 (1.151)
where
A= J 13 + J 10 = J2 + K1 (1.152)
B= J 23 + J 20 = J 1 + K2 (1.153)
are Hermitian operators. Then using equations 1.48 to 1.50
A k,a,b =a k,a,b
(1.157)
B k,a,b =b k,a,b
Consider the transformation of A and B, by observing equation 1.142 and remembering that
the coefficients of A and B in equation 1.151 are ↵ and respectively, we can find the
transformation of A and B to be as follows
1
U [R(✓)]AU [R(✓)] = A cos ✓ B sin ✓
1
(1.158)
U [R(✓)]BU [R(✓)] = A sin ✓ + B cos ✓
This can also be derived directly by using the matrix form of J3 , A and B, i.e.
0 1
0 0 0 0
B0 0 i 0C
J3 = B@0 i 0 0A
C (1.159)
0 0 0 0
0 1 0 1
0 0 0 0 0 i 0 0
B0 0 0 iC B i 0 0 0C
A=B
@0
C+B
A @
C (1.160)
0 0 0 0 0 0 0A
0 i 0 0 0 0 0 0
| {z } | {z }
J2 K1
28 CHAPTER 1. THE POINCARÉ GROUP
Similarly 0 1 0 1
0 0 0 0 0 0 i 0
B0 0 0 0C B 0 0 0 0C
B=B
@0
C+B
A @
C (1.161)
0 0 i i 0 0 0A
0 0 i 0 0 0 0 0
| {z } | {z }
J1 K2
0 10 10 1
1 0 0 0 0 i 0 0 1 0 0 0
B0 cos ✓ sin ✓ 0C B i 0 0 iC B0 cos ✓ sin ✓ 0C
B CB CB C = A cos ✓ B sin ✓
@0 sin ✓ cos ✓ 0 @
A 0 0 0 0A @0 sin ✓ cos ✓ 0A
0 0 0 1 0 i 0 0 0 0 0 1
(1.162)
as expected. The calculation is similar for B.
Define ✓ ⌘U 1 (R(✓))
k,a,b k,a,b . Then for A, we have
1
U [R(✓)]AU [R(✓)] k,a,b = (A cos ✓ B sin ✓) k,a,b )
✓
U [R(✓)]A k,a,b = (a cos ✓ b sin ✓) k,a,b ) (1.163)
✓ ✓
A k,a,b = (a cos ✓ b sin ✓) k,a,b
Similarly
✓ ✓
B k,a,b = (a sin ✓ + b cos ✓) k,a,b (1.164)
The problem is that if one finds one set of non-zero eigenvalues of A and B, one finds a whole
continuum. However, this is in contrast with the observations that suggest massless particles
have no continuous degree of freedom. As a result other restrictions need to be imposed.
We must require the physical states to be eigenvectors of A and B with eigenvalues a=b=0.
A k, =B k, =0 (1.165)
Note that here we have labeled the states by k, since a and b are now fixed.
Now, in order to distinguish these states from each other, we need to consider the eigenvalues
of the remaining generator J3 .
J3 k, = k, (1.166)
Since the standard 4-momentum k = (1, 0, 0, 1), the 3-momentum vector k = (0, 0, 1) is in the
the 3-direction which means that is the component of angular momentum in the direction
of motion and it is called helicity.
1.3. CONNECTION WITH ONE-PARTICLE STATES 29
Therefore, it can be said that the states are identified by the momentum in the 3-direction
and the angular momentum in the direction of motion.
i(↵A+ B)
U (S(↵, )) = e (1.167)
and for finite ✓
iJ3 ✓
U (R(✓)) = e (1.168)
The little group element W can then be written as follows using equation 1.139
i(↵A+ B) i✓J3 i✓
U (W ) k, =e e k, =e k, (1.169)
where we have used the fact that the A and B have zero eigenvalues when acting on .
Hence, according to equation 1.94
i✓
D 0 (W ) = e 0 (1.170)
Then, using equations 1.98 and 1.111 the Lorentz transformation rule for a massless particle
with helicity is given by
s
(⇤p)0 i ✓(⇤,p)
U (⇤) p, = e ⇤p, (1.171)
p0
The allowed values of are restricted to integers and half integers due to topological consid-
erations. In order to calculate the little group element W we need to set a standard Lorentz
transformation that takes k µ = (, 0, 0, ) ! pµ . We choose it such that it has the following
form
L(p) = R(p̂)B(|p|/)R 1 (p̂) (1.173)
with B(u) being a pure boost in the 3-direction:
0 2 1
(u + 1)/2u 0 0 (u2 1)/2u
B 0 1 0 0 C
B(u) ⌘ B@
C
A (1.174)
0 0 1 0
(u2 1)/2u 0 2
0 (u + 1)/2u
30 CHAPTER 1. THE POINCARÉ GROUP
One can check that this is matrix a a Lorentz transformation (boost) by comparing it with
the general boost along the 3-axis
0 1
0 0
B 0 1 0 0 C
B C (1.175)
@ 0 0 1 0 A
0 0
2 2
where we must have 2 2 2 = 1. In our case, we may set u +1 ⌘ and u 2u 1 ⌘ , then
2u
2 2
it is easy to verify that the required relation ( u 2u+1 )2 ( u 2u 1 )2 = 1 holds as required by a
Lorentz transformation.
On the other hand R(p̂) is a pure rotation that takes the 3-axis into the direction of motion
p̂. For example in polar coordinates
then taking R(p̂) to be a rotation by angle ✓ about the 2-axis, taking (0, 0, 1) ! (sin ✓, 0, cos ✓),
followed by a rotation about the 3-axis by angle :
i J3 i✓J2
U (R(p̂)) = e e (1.177)
with 0 ✓ ⇡ and 0 < 2⇡. Note that the reason we give U (R(p̂)) rather than R(p̂) is
that shifting ✓ or by 2⇡, leads to the same result for R(p̂) but picks up a minus sign for
U (R(p̂)) when it acts on states with half-integer spins. Since 1.177 is a rotation that takes
the 3-axis into direction p̂, any other choice of R(p̂) will be di↵erent from this one by at
most an initial rotation about the 3-axis (because it keeps this axis fixed) which means that
one-particle states merely di↵er by a phase factor.
It is very important to note that helicity is Lorentz invariant i.e. a massless particle of
helicity, say, looks the same (aside from its momentum) in all inertial reference frames.
The reason lies in equation 1.171. It can be observed that remains invariant. This is in
contrast with the massive case 1.122 were ’s mix under the Lorentz transformation and
hence it is not invariant and it transforms among the 2s + 1 possible values.
Chapter 2
The aim of the chapter is to construct Lorentz-invariant field theories using fields of spin 0,
1/2 and 1 in connection with the representation of the Lorentz group. We will then discuss
spinor fields in more detail and learn how to manipulate spinor indices.
1 1
U (⇤)'(x)U (⇤) = '(⇤ x) (2.1)
U (⇤)@ µ '(x)U 1
(⇤) = ⇤µ⇢ @¯⇢ '(⇤ 1
x) (2.2)
where we put a bar at the top of @ to indicated that it is the derivative with respect the
argument x̄ = ⇤ 1 x
This implies that a vector field Aµ and a tensor field B µ⌫ transform as
U (⇤)A⇢ (x)U 1
(⇤) = ⇤µ⇢ A⇢ (⇤ 1
x) (2.3)
U (⇤)B µ⌫ (x)U 1
(⇤) = ⇤µ⇢ ⇤⌫ B ⇢ (⇤ 1
x) (2.4)
respectively.
Note that the symmetry and antisymmetry properties of B µ⌫ are preserved under Lorentz
31
32 CHAPTER 2. FIELDS WITH SPIN 0, 1/2, 1
transformations.
Proof: Let B µ⌫ = B ⌫µ then
We define the trace of B to be T (x) ⌘ ⌘µ⌫ B µ⌫ . This entity transforms like a scalar field
under Lorentz:
U (⇤)T (x)U 1 (⇤) = T (⇤ 1 x) (2.6)
1
B µ⌫ (x) = Aµ⌫ (x) + S µ⌫ (x) + ⌘ µ⌫ T (x) (2.8)
4
where Aµ⌫ is an antisymmetric tensor and hence traceless, S µ⌫ is symmetric and traceless
i.e. taking ⌘µ⌫ S µ⌫ = 0. It is important to notice that the fields Aµ⌫ , S µ⌫ and T do not mix
with one another under Lorentz transformations. This is because we just proved that the
symmetry properties are preserved under Lorentz transformations and T is a scalar while the
other two are tensors.
The main question now is that whether this can be further decomposed into smaller irre-
ducible representations. We will come back to this question later on in this chapter.
Consider a field 'A (x) where A is a Lorentz index. It transforms under Lorentz accord-
ing to
U (⇤)'A (x)U 1 (⇤) = LAB (⇤)'B (⇤ 1 x) (2.9)
where LAB (⇤) is a finite dimensional matrix that depends on ⇤ so that it results in a linear
combination on the RHS of the equation. The group composition rule is obeyed
Therefore, LAB (⇤) furnish the representation of the homogeneous Lorentz group, as seen
before. Again, for and infinitesimal transformation ⇤µ⌫ = µ⌫ + ! µ⌫
i
U (1 + !) = 1 !µ⌫ M µ⌫ (2.11)
2
2.1. REPRESENTATIONS OF THE LORENTZ GROUP USING SU (2) ⇥ SU (2) 33
with M µ⌫ being the generators of the Lorentz group and !µ⌫ being antisymmetric. As shown
in the first chapter, the Lie algebra of the Lorentz group is specified by
[M µ⌫ , M ⇢ ] = i(⌘ µ⇢ M ⌫ ⌘ µ M ⌫⇢ ⌘ ⌫⇢ M µ + ⌘ ⌫ M µ⇢ ) (2.12)
From which the relations (1.48 to 1.50) between the angular momenta Ji ⌘ 12 ✏ijk M jk and
boosts Ki ⌘ M 0i are derived.
For an infinitesimal trasformation,
i
LAB (1 + !) = A
B
!µ⌫ (S µ⌫ )AB (2.13)
2
Then, equation 2.9 implies
where Lµ⌫ = i(xµ @ ⌫ x⌫ @ µ ). Note that when µ and ⌫ are the spacial coordinates, Lµ⌫
corresponds to orbital angular momentum.
1 1
!µ⌫ x⌫ @ µ = [ !µ⌫ x⌫ @ µ !⌫µ xµ @ ⌫ ] = [ x⌫ @ µ + xµ @ ⌫ ] !µ⌫ (2.17)
2 2
Therefore, comparing the coefficients of ! on the left and the right hand sides of the equation,
we have
1
['A (x), M µ⌫ ] = (xµ @ ⌫ x⌫ @ µ )'A (x) (S µ⌫ )AB 'B (x) (2.18)
i
µ⌫
Substituting for L gives the result. Q.E.D
It is not hard to show that both the di↵erential operator and matrices S µ⌫ satisfy the same
34 CHAPTER 2. FIELDS WITH SPIN 0, 1/2, 1
then
[Lµ⌫ , L⇢ ] = [xµ @ ⌫ x ⌫ @ µ , x⇢ @ x @⇢]
= (⌘ ⌫⇢ xµ @ ⌘ µ ⇢ ⌫
x @ ⌘ µ⇢ x⌫ @ + ⌘ ⌫ x⇢ @ µ ⌘ ⌫ xµ @ ⇢ + ⌘ ⇢µ x @ ⌫ + ⌘ µ x⌫ @ ⇢ ⌘ ⇢⌫ x @ µ )
= {⌘ ⌫⇢ (xµ @ x @µ) ⌘ µ
(x⇢ @ ⌫ x⌫ @ ⇢ ) ⌘ µ⇢ (x⌫ @ x @ ⌫ ) + ⌘ ⌫ (x⇢ @ µ xµ @ ⇢ )}
= i{⌘ ⌫⇢ Lµ ⌘ µ
L⇢⌫ ⌘ µ⇢ L⌫ + ⌘ ⌫ L⇢µ } = i(⌘ µ⇢ L⌫ ⌘ µ L⌫⇢ + ⌘ ⌫ Lµ⇢ ⌘ ⌫⇢ Lµ )
(2.20)
We know that the algebra 1.48 to 1.50 form the Lie algebra of SO(3). Consider equation
1.48, from quantum mechanics we know that we can find three (2j + 1) ⇥ (2j + 1) hermitian
matrices corresponding the the angular momenta in each direction. For example the matrix
of J3 is a diagonal matrix corresponding to an infinitesimal rotation by some angle ✓ about
the third axis. The eigenvalues of J3 are m = j, j + 1, ..., j. The allowed values of j are
0, 1/2, 1, ... . For a finite rotation, and some j (as seen in the previous chapter) we have,
d(j) (✓)m0 m = eim✓ m0 m =) d(j) (2⇡) = ei(2⇡m) = cos(2⇡m) = ( 1)2m = ( 1)2j (2.21)
1
Ni† ⌘ (Ji + iKi ) (2.23)
2
and so In other words,
Ji = Ni + Ni† (2.24)
and
Ki = i(Ni Ni† ) (2.25)
2.1. REPRESENTATIONS OF THE LORENTZ GROUP USING SU (2) ⇥ SU (2) 35
Equations 1.48 to 1.50 then imply that the two operators satisfy the following
integer values of j. This implies that the first two candidates are not allowed for a 4-vector
whose space components are a 3-vector. The ( 12 , 12 ) representation contains j = 0, 1 and so
it is the only correct 4-vector representation.
where k , k = 1, 2, 3 are the familiar Pauli matrices, which are in fact the infinitesimal
generators of SU (2):
✓ ◆ ✓ ◆ ✓ ◆ ✓ ◆
1 0 0 1 0 i 1 0
0 = , 1 = , 2 = , 3 = (2.41)
0 1 1 0 i 0 0 1
Note that the indices a, b run form 1 to 2 and identify each element of the 2 ⇥ 2 matrix,
whereas i, j indices in S ij are merely there to label the type of rotation we are consid-
ering. For example,(SL12 ) means rotation in the 1-2 plane, i.e. about the 3-axis. Then
(SL12 )ab = 12 ✏12k k = 12 3 . Then, e.g. (SL12 )11 = + 12 and so on.
Now we need to identify the matrices for the boost operator Kk = M 0k in the ( 12 , 0) repre-
sentation. We also know that in this representation, Nk† is zero since n0 = 0. Equations 2.24
and 2.25 then imply that Kk = iJk and so
1
(SL0k )ab = i k (2.42)
2
As we have already mentioned, hermitian conjugation swaps the two SU (2) Lie algebras.
Hence if one takes the hermitian conjugate of a field a (x) in the ( 12 , 0) representation, it
will result in a field in the (0, 12 ) representation. This field is called the right-handed spinor
(Weyl) field. We will adopt the notation
† †
[ a (x)] = ȧ (x) (2.43)
with Rȧḃ being a matrix in the (0, 12 ) representation. The group composition rule satisfied
by these matrices is
Rȧḃ (⇤0 )Rḃċ (⇤) = Rȧċ (⇤0 ⇤) (2.45)
For an infinites rents transformation ⇤µ⌫ = µ
⌫ + ! µ⌫ we have, just as before,
i µ⌫ ḃ
Rȧḃ (1 + !) = ȧ
ḃ
!µ⌫ (SR )ȧ (2.46)
2
µ⌫ ḃ ⌫µ ḃ
where (SR )ȧ = (SR )ȧ are 2 ⇥ 2 matrices that obey the same commutation relation as
the generators M µ⌫ . Then
† µ⌫ ḃ †
[M µ⌫ , ȧ (0)] = (SR )ȧ ḃ
(0) (2.47)
taking the hermitian conjugate, knowing that M µ⌫ is a hermitian operator, we get
µ⌫ µ⌫ ḃ ⇤
[ a (0), M ] = [(SR )ȧ ] b (0) (2.48)
38 CHAPTER 2. FIELDS WITH SPIN 0, 1/2, 1
Comparing with [M µ⌫ , a (0)] = (SLµ⌫ )ab b (0) derived before, implies that
µ⌫ ḃ
(SR )ȧ = [(SLµ⌫ )ab ]⇤ (2.49)
Let us now examine the properties of spinor indices in more detail. For this purpose, consider
a field Cab (x) which carries two spinor indices. Under a Lorentz transformation
1
U (⇤)Cab U (⇤) = Lac Lbd Ccd (⇤ 1
x) (2.50)
The question is whether the four components of Cab can be broken into smaller part that do
no mix under the Lorentz transformations. To address this question we need to refer back
to quantum mechanics. We know that two particles, each with spin 12 , can be in a state of
total spin zero or one. Then
• Total spin j=0: The antisymmetric combination
1
|0, 0i = p (| "#i | #"i) (2.51)
2
• Total spin j=1: The symmetric combination
8
>
>
>
> |1, 1i = | ##i
>
>
<
|1, 0i = p12 (| "#i + | #"i) (2.52)
>
>
>
>
>
>
: |1, 1i = | ""i
In other words, 12 ⌦ 12 = 0A 1S . Note that the dimension of the left hand side is equal to
2 ⇥ 2 = 4 which is the same as that on the right hand side i.e. 1 + 3 = 4. Therefore, for the
Lorentz group we have
1 1
( , 0) ⌦ ( , 0) = (0, 0)A (1, 0)S (2.53)
| 2 {z 2 } | {z }
dim = 1 + 3 = 4
dim = 2 ⇥ 2 = 4
Hence Cab (x) can be written as
where we have used fact that D(x) transforms as a scalar and ✏ is a constant.
On the other hand, using equation 2.50 and then 2.54 itself, we have
In precisely the same way that we use ⌘µ⌫ and ⌘ µ⌫ to raise and lower vector indices, one can
use ✏ab and ✏ab to raise and lower left-handed spinor indices. We define
✓ ◆
ab 0 1
✏ = = ✏ab (2.59)
1 0
Then
✏ab ✏bc = a
c
(2.60)
✏ab ✏bc = a
c (2.61)
One can easily check these by simple matrix multiplication. So we may define
a
(x) ⌘ ✏ab b (x) (2.62)
As well as that,
b
a = ✏ab = ✏ab ✏bc c = a
c
c (2.63)
Equation 2.62 may also be written as
a
= ✏ab b = ✏ba b = b✏
ba
= b✏
ab
(2.64)
Using the same method, one can find an invariant symbol ✏ȧḃ = ✏ḃȧ for the second SU(2)
factor, i.e. for a right-handed spinor field
1 1
(0, ) ⌦ (0, ) = (0, 0)A (0, 1)S (2.66)
2 2
40 CHAPTER 2. FIELDS WITH SPIN 0, 1/2, 1
We now have ✓ ◆
ȧḃ 0 1
✏ = = ✏ȧḃ (2.67)
1 0
The same relations presented above hold if we replace all the undotted indices with the dot-
ted ones.
Let Aaȧ (x) be a field carrying one undotted and one dotted index and so it belongs to the
( 12 , 12 ) representation. But we also said that ( 12 , 12 ) representation corresponds to the vector
representation. The aim is to now write the field in the ‘more natural’ vector representation
Aµ . For this purpose, we need to find a mapping which can be written as
µ
Aaȧ (x) = aȧ Aµ (x) (2.68)
where aµȧ is an invariant symbol. To understand why such a symbol must exist consider the
an object Taµȧ that carries a Lorentz index µ as well as two left and right handed indices a, ȧ.
It transforms under Lorentz according to
so that
1 1 1 1
( , 0) ⌦ (0, ) ⌦ ( , ) = (0, 0) ... (2.70)
2 2 2 2
and so on the RHS we have the (n, n0 ) = (0, 0) which corresponds to a singlet and so it is
invariant i.e. when acting with N and N † on such a state, we get zero. We take aµȧ to be
the invariant part of Taµȧ . It turns out to be possible to choose aµȧ to be
µ
aȧ = (I, ~ ) (2.71)
but we need to check that this is indeed a viable choice and consistent with the conventions
for (SLµ⌫ )ab and (SR
µ⌫ ḃ
)ȧ . Consider a combination of invariant symbols e.g. ⌘µ⌫ aµȧ b⌫ḃ . This
whole object must also be invariant and because it carries two dotted and undotted indices,
we would expect it to be proportional to ✏ab ✏ȧḃ .
✓ ◆2 ✓ ◆2 ✓ ◆2 ✓ ◆2 ✓ ◆
1 0 0 1 0 i 1 0 2 0
⌘µ⌫ aµȧ b⌫ḃ = = (2.72)
0 1 1 0 i 0 0 1 0 2
as required. If this turned out not to be correct, 2.71 would not a a suitable choice for such
a symbol. In a similar fashion, we must require ✏ab ✏ȧḃ aµȧ b⌫ḃ to be proportional to ⌘ µ⌫ :
µ ⌫
✏ab ✏ȧḃ aȧ bḃ = 2⌘ µ⌫ (2.75)
The proof of this equation and a more detailed discussion on the properties of ✏ab and its
relation with S µ⌫ will be presented in the next section.
We can generalise the existence of such invariant symbols i.e. whenever the direction product
of a representation includes a scalar (singlet), there exists an invariant symbol. For example
from
1 1 1 1
( , ) ⌦ ( , ) = (0, 0)S (0, 1)A (1, 0)A (1, 1)S (2.76)
2 2 2 2
we can conclude the existence of ⌘µ⌫ since we know it is invariant under Lorentz transforma-
tions and it carries two vector indices. Another example is the four dimensional Levi-Civita
symbol which is deduced using
1 1 1 1 1 1 1 1
( , ) ⌦ ( , ) ⌦ ( , ) ⌦ ( , ) = (0, 0)A ... (2.77)
2 2 2 2 2 2 2 2
the symbol ✏µ⌫⇢ carries four vector indices and is antisymmetric on exchange of any pair
of indices. It is normalised in such a way that ✏0123 = +1 and is invariant. This is because
⇤µ↵ ⇤⌫ ⇤⇢ ⇤ ⌧ ✏↵ ⌧ is antisymmetric with respect to exchange of any pair of indices that
are not contracted hence it must be proportional to ✏µ⌫⇢ .
This is a good point to refer back to the question raised in the previous section on whether
or not a field carrying two vector indices B µ⌫ could be further decomposed into yet smaller
irreducible representations, where we found B µ⌫ to be equal to
1
B µ⌫ (x) = Aµ⌫ (x) + S µ⌫ (x) + ⌘ µ⌫ T (x) (2.78)
4
We will now compare this with equation 2.76. T (x) is a scalar, so it correspond to the
(0, 0) representation. S µ⌫ corresponds to (1, 1)S since it is symmetric. Also, because it is
traceless we conclude that it has 10 1 = 9 independent components which is the same as the
dimensionality of the (1, 1) representation. The antisymmetric field Aµ⌫ (x) must therefore
correspond to (0, 1)A (1, 0)A . However, a field in the (0, 1)A representation carries two
symmetric undotted indices. On the other hand, a field in the (1, 0)A representation carries
two symmetric dotted indices and is the hermitian conjugate of the previous field.Our aim
is now to determine a mapping similar to 2.68 for Aµ⌫ (x) in terms of a field Gab and its
hermitian conjugate G†ȧḃ .
The generators SLµ⌫ and SR µ⌫
provide such a mapping. Note that since Pauli matrices are
traceless, equations 2.40 and 2.42 imply that (SLµ⌫ )aa = 0. We also have
(SLµ⌫ )aa = ✏ab (SLµ⌫ )ab = 0 (2.79)
42 CHAPTER 2. FIELDS WITH SPIN 0, 1/2, 1
Therefore,
(SLµ⌫ )ab = (SLµ⌫ )ba (2.81)
is symmetric on the exchange of its two spinor indices. A similar method can be used to
µ⌫ µ⌫
show that (SR )ȧḃ = (SR )ḃȧ . According to equations 2.40 and 2.42,
1 i i
(SL01 )ab = i 1 = ✏231 1 = i(SL23 )ab ) (SLµ⌫ )ab = + ✏µ⌫⇢ (SL⇢ )ab (2.82)
2 2 2
where the factor of 1/2 is required since S and ✏ are antisymmetric with respect to exchange
of indices ⇢ and . Taking the conjugate and using equation 2.49 we also have
µ⌫ ḃ i µ⌫⇢
(SR )ȧ = ✏ (SR⇢ )ȧḃ (2.83)
2
Now, consider Gab (x) to be a field in the (1, 0) representation which can be mapped into a
self-dual antisymmetric tensor:
i
Gµ⌫ (x) = + ✏µ⌫⇢ G⇢ (x) (2.85)
2
Taking the hermitian conjugate and applying equation 2.49 we get
µ⌫ ȧḃ †
G†µ⌫ (x) = (SR ) Gȧḃ (x) (2.86)
i µ⌫⇢ †
G†µ⌫ (x) = ✏ G⇢ (x) (2.87)
2
From which one can extract the self-dual and anti-self-dual parts of the tensor field Aµ⌫
1 i
Gµ⌫ (x) = Aµ⌫ + ✏µ⌫⇢ A⇢ (2.88)
2 4
1 i µ⌫⇢
G†µ⌫ (x) = Aµ⌫ ✏ A⇢ (2.89)
2 4
2.3. MANIPULATING SPINOR INDICES 43
Let us now find more informations about matrices (SLµ⌫ )aa and (SR
µ⌫ ḃ
)ȧ . First, we use the the
µ
invariance of the symbols ✏ab , ✏ȧḃ and aȧ .
and so the object in the bracket must be equal to zero. Hence (SLµ⌫ )ab = (SLµ⌫ )ba i.e. it
is symmetric with respect to the spinor indices. Note that this result was derived in the
µ⌫ µ⌫
previous section using a di↵erent method. Similarly (SR )ȧḃ = (SR )ḃȧ . Now, consider aµȧ
for which we have
⇢ ⇢ b ḃ ⌧
aȧ = ⇤ ⌧ L(⇤)a R(⇤)ȧ bḃ (2.95)
So for infinitesimal transformations one has
i
⇤ ⇢⌧ = ⇢
⌧ !µ⌫ (Svµ⌫ )⇢⌧ (2.96)
2
i
Lab (1 + !) = a
b
!µ⌫ (SLµ⌫ )ab (2.97)
2
i µ⌫ ḃ
Rȧḃ (1 + !) = ȧ
ḃ
!µ⌫ (SR )ȧ (2.98)
2
with
(Svµ⌫ )⇢⌧ ⌘ i(⌘ µ⇢ ⌫
⌧ ⌘ ⌫⇢ µ
⌧) (2.99)
The proof of equation 2.96 is presented below:
1 1
⇤ ⇢⌧ = ⇢
⌧ + !µ⌫ (⌘ µ⇢ ⌫
⌧ ⌘ ⌫⇢ µ
⌧) = ⇢
⌧ + (!µ⌧ ⌘ µ⇢ !⌧ ⌫ ⌘ ⌫⇢ )
2 2 (2.100)
⇢ 1
= ⌧ + (!µ⌧ ⌘ µ⇢ !⌧ µ ⌘ µ⇢ ) = ⇢
⌧ + ! ⇢⌧
2
as expected. We can now substitute equations 2.96 to 2.99 into equation 2.95
✓ ◆✓ ◆✓ ◆
⇢ ⇢ 1 µ⇢ ⌫ ⌫⇢ µ b i µ⌫ b ḃ i µ⌫ ḃ ⌧
aȧ = ⌧ + !µ⌫ (⌘ ⌧ ⌘ ⌧) a !µ⌫ (SL )a ȧ !µ⌫ (SR )ȧ bḃ
2 2 2
✓ ◆
⇢ b ḃ i ⇢ b µ⌫ ḃ i ⇢ ḃ µ⌫ b 1 µ⇢ ⌫ ⌫⇢ µ b ḃ ⌧
= ⌧ a ȧ !µ⌫ ⌧ a (SR )ȧ !µ⌫ ⌧ ȧ (SL )a + !µ⌫ (⌘ ⌧ ⌘ ⌧ ) a ȧ bḃ
2 2 2
(2.101)
(⌘ µ⇢ ⌫
⌧ ⌘ ⌫⇢ µ ⌧
⌧ ) aȧ i(SLµ⌫ )ab ⇢
bȧ
µ⌫ ḃ
i(SR )ȧ ⇢
aḃ
=0 (2.102)
µ⌫
Now multiply the equation by ✏ȧċ and use the fact that ✏ȧċ (SR )ȧċ = 0 and ✏ȧċ ✏ȧċ = 2 to
get
i
(SLµ⌫ )ac = ✏ȧċ cµċ a⌫ȧ ⌫ µ
cċ aȧ (2.105)
4
Similarly, multiplying equation 2.104 by ✏ac results in
µ⌫ i µ ⌫ ⌫ µ
(SR )ȧċ = ✏ac ( aȧ cċ aȧ cċ ) (2.106)
4
Define
µ
¯ µȧa ⌘ ✏ab ✏ȧḃ bḃ
(2.107)
Numerically,
¯ µȧa = (I, ~ ) (2.108)
Proof:
First note that ✓ ◆ ✓ ◆
ab 0 1 0 i
✏ = =i =i 2 (2.109)
1 0 i 0
Using the above result we have
h iȧa
µ ḃȧ aȧ T
¯ µȧa = ✏ab bḃ
✏ = [(i 2 )(
µ
)(i 2 )] = ( 2
µ
2) (2.110)
Try µ = 1:
✓ ◆✓ ◆✓ ◆ T ✓ ◆
1 0 i 0 1 0 i 0 1
¯ = = = 1 (2.111)
i 0 1 0 i 0 1 0
The method is similar for the other components. Hence, ¯ µȧa = (I, ~ ). Q.E.D
Multiplying equations 2.105 and 2.106 by ✏bc and ✏ḃċ respectively and using ¯ µȧa we get
i ⇣ µ ⌫ ȧb ⌘ i
(SLµ⌫ )ab = aȧ ¯ ⌫ µȧb
a ȧ ¯ = ( µ ¯⌫ ⌫ µ b
¯ )a (2.112)
4 4
and
µ⌫ ḃ i ⇣ µ ⌫ ḃa ⌫ µḃa
⌘ i ⇣ ⌫ ḃa µ
⌘
(SR )ȧ = aȧ ¯ aȧ ¯ = ¯ aȧ ¯ µḃa ⌫
aȧ
4 4 (2.113)
µ⌫ ḃ i µ ⌫
) (SR ) ȧ = (¯ ¯ ⌫ µ )ḃȧ
4
We will now adopt the following convention, a missing pair of contracted undotted indices,
and a missing pair of contracted dotted indices will be written as
a † † † †ȧ
= a and = ȧ (2.114)
46 CHAPTER 2. FIELDS WITH SPIN 0, 1/2, 1
respectively.
We also know that the Weyl fields describe spin- 12 particles i.e. fermions. Hence the corre-
sponding fields anticommute rather than commute:
a
a = ✏ab ✏ac b
c
= ✏ba ✏ac b
c
= b
c b
c
= b
b
= a
a
(2.116)
Q.E.D
also † † = † † by 2.117. Note that when the indices are suppressed, a left-handed field is
always written without a dagger and a right handed one is written with a dagger.
To understand these conventions more, let us take the hermitian conjugate of the follow-
ing
¯ = ȧ† ¯ µȧc c
† µ
(2.119)
Its transformations under Lorentz is
h i h i
U (⇤) † ¯ µ U 1
(⇤) = ⇤µ⌫ † ⌫
¯ (2.120)
where we have used the fact that ¯ µ = (I, ~ ) is hermitian. Note the way induces are summer
on each side of the equation. Also remember that since we are dealing with it fields here,
and are the operators.
Bibliography
[1] Steven Weinberg, The quantum theory of fields: Foundations. volume 1 Cambridge
University Press, Cambridge, 2005.
[2] Mark Allen Srednicki, Quantum field theory, Cambridge University Press, Cambridge,
2007
[3] Julius Wess and Jonathan Bagger, Supersymmetry and supergravity, Princeton University
Press, Princeton, N.J., 2nd ed., rev. and expanded edition, 1992.
[5] Wu-Ki Tung, Group theory in physics, World Scientific, Philadelphia, 1985.
[6] Steven Weinberg, The quantum theory of fields: Supersymmetry. volume 3 Cambridge
University Press, Cambridge, 2005.
83