Sei sulla pagina 1di 125

Lecture Notes for Quantum Field Theory III

Spring 2011
Lecturer: Professor Erick Weinberg
Transcriber: Alexander Chen

July 17, 2011

Contents
1 Lecture 1 3
1.1 Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Group theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2 Lecture 2 7
2.1 Group Representations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

3 Lecture 3 11
3.1 Continued SU (3) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.2 Roots and Weights . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.3 The Algebra of SO(N ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.4 Classification of Root Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

4 Lecture 4 18
4.1 Roots Continued . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
4.2 Exceptional Algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.3 Classification According to Dynkin Diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.4 Weights . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

5 Lecture 5 24
5.1 Spontaneous Symmetry Breaking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

6 Lecture 6 29
6.1 Proof of Goldstone Theorem in General . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
6.2 Sigma Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

7 Lecture 7 33

8 Lecture 8 38

9 Lecture 9 42

1
10 Lecture 10 46

11 Lecture 11 51
11.1 Anomalies Continued . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

12 Lecture 12 55

13 Lecture 13 58
13.1 Grand Unification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

14 Leture 14 62
14.1 Symmetry Breaking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

15 Lecture 15 67
15.1 Solitons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

16 Lecture 16 73
16.1 Kink Solution Continued . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
16.2 Multikink Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

17 Lecture 17 78

18 Lecture 18 83

19 Lecture 19 87

20 Lecture 20 91

21 Lecture 21 95
21.1 Instantons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97

22 Lecture 22 99

23 Lecture 23 103
23.1 Supersymmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104

24 Lecture 24 107
24.1 Wess-Zumino Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
24.2 Notation Transmutation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108

25 Lecture 25 111

26 Lecture 26 115

27 Lecture 27 120

28 Lecture 28 123

2
Quantum Field Theory III Lecture 1

1 Lecture 1
1.1 Structure
We will start with a bit of group theory, and we will talk about spontaneous symmetry broken. Then we
will talk about anomalies and grand unification. Then we will cover solitions & duality instantons. Finally
we wil talk about supersymmetry.

1.2 Group theory


Suppose we have a group G with elements g1 , g2 , . . . with operation g1 g2 = g3 . We require the operation
to be associative, there is an identity and every element has an inverse

gI = Ig = g, gg −1 = g −1 g = I (1.1)

A group can have finite number of elements or infinite. For the infinite case we can have discrete elements
(group of integers) or continuous (real numbers). The most important kind of groups in high energy
physics is Lie groups, which is a continuous group. A group element will be labeled by a number of labels
g(x1 , x2 , . . . ) = g(x). We have the group multiplication law

g(x)g(y) = g(z) (1.2)

where z is a continuous differentiable function of x and y.


We can also think of the Lie group as a manifold. The dimension of the manifold, which is also the
dimension of the Lie group, is equal to the number of parameters necessary to specify each group element.
The manifold can be compact or non-compact. For example, a rotation group is compact, and if you keep
rotation you will come back to the original position. However the Lorentz group is not compact, and you
can boost forever. The Lorentz group will likely be the only non-compact group that we will make reference
to. References can be found in the book by Georgi. Another one written by Gilmore.
Let’s look at some examples. Consider the group SO(N ). This can be defined as the group of N × N
orthogonal matrices with determinant 1, or can be defined as the rotation group in N dimensions. From
N × N we have N 2 parameters, but from orthogonality we have the constraint

Rij Rjk = δik (1.3)

This gives N conditions when i = k and N (N − 1)/2 conditions when i 6= k. So in the end we have
N (N − 1)/2 free parameters. So the group has dimension N (N − 1)/2. Now if we erase S and consider
O(N ). It has the same dimension as SO(N ) but has two disconnected parts. The one containing identity
is the same as SO(N ).
Now we consider U (N ) which is the group of complex N × N unitary matrices. Because it is unitary
we have |det U | = 1. By the same argument as above we can find the dimension of the group to be of
dimension N 2 . Now if we consider SU (N ), then we should have det U = 1. This subtracts one parameter
as the determinant is a continuous variable. So the dimension of SU (N ) is N 2 − 1.
The Lorentz group, which is non-compact, has 3 boosts and 3 rotations. So this has 6 dimensions.
These are the examples we want to consider. But these groups are not all distinct. We can write an
element in U (1) as eiθ , and an element in SO(2) as a matrix with parameter θ. These two groups are
essentially identical
U (1) ∼
= SO(2) (1.4)

3
Quantum Field Theory III Lecture 1

We also have relations between SU (2) and SO(3). Each element in SO(3) corresponds to 2 elements in
SU (2). In general we have the group Spin(N ) which has two to one correspondence with SO(N ).
We also have correspondences between Lie groups and Lie algebras. A Lie algebra is related to the
neighborhood of identity in the Lie group. If we think of the Lie group as a manifold, then the structure
around the identity is almost enough to determine the whole group (apart from some double-cover things).
To do this we think of the group as group of matrices and choose the coordinates so that the identity is
I = (0, 0, . . . ). Near identity we have
X
g =I +i αj Tj + O(α2 ) (1.5)
j

The α’s are the coordinates and Tj are generator matrices. The factor of i is just convention. In SO(3)
the conventional generators will be the angular momenta J1 , J2 , J3 . An important result in the theory of
Lie groups is that any element continuously connected to I can be written as
 X 
g = exp i αj Tj (1.6)

Now let’s get some conditions on the Tj generators. Let’s look at the element

g = eiλTa eiλTb e−iλTa e−iλTb (1.7)

We assume λ  1. By group multiplication law this is an element of the group, so we can write
!
X
g = exp i αk Tk (1.8)
k

Now we need to do some manipulations


 
g = eiλTa e−iλTa eiλTb + [eiλTb , eiλTa ] e−iλTb = I + eiλTa [eiλTb , eiλTa ]eiλTb (1.9)

Now we can expand everything for small λ and keep up to terms of order λ2 , we get

g = I + [iλTb , −iλTa ] + O(λ3 ) = I + λ2 [Tb , Ta ] + O(λ3 ) (1.10)

Now this should be a group element, so the commutator of two generators should be a linear combination
of all the generators
[Ta , Tb ] = if cab Tc (1.11)
It is apparent that f cab = −f cba . If we choose T ’s correctly then fabc will be completely antisymmetric. In
SO(3) for conventional generators we have fabc = εijk . The f ’s are called structure constants.
Now we come back to Lie algebra. A Lie algebra is defined to be a vector space with a bracket operation
satisfying that i[A, B] is inside the algebra for any A, B in the algebra, and we have to satisfy the Jacobian
identity
[A, [B, C]] + [B, [C, A]] + [C, [A, B]] = 0 (1.12)
The Poisson bracket is a form of Lie algebra operation
X  ∂f ∂g ∂f ∂g

{f (q, p), g(q, p)} = − (1.13)
∂qj ∂pj ∂pj ∂qj
j

4
Quantum Field Theory III Lecture 1

It satisfies the Jacobian identity but it is nontrivial to show.


If we consider hermition matrices, in general (AB)† 6= AB, but the commutator
(i[A, B])† = −i[B † , A† ] = i[A, B] (1.14)
So the commutator is inside the Lie algebra. The structure constants determines the whole algebra, so the
algebras for SO(3) and SU (2) are the same.
Let’s consider SO(N ). The elements satisfy RT R = I. So we have
  !
X X X
T
I + i αj Tj + . . .  I + i αk Tk + . . . = I + i αj (Tj + TjT ) + · · · = I (1.15)
j k

So the generators must be N × N antisymmetric matrices. This also applies to O(N ). Same procedure
shows that generators of U (N ) are N × N hermitian matrices. Now for SU (N ) we need a condition on the
determinant
det eiαj Tj = 1 = det [I + iαj Tj + . . . ] = 1 + iαj Tr Tj
 
(1.16)
So the generators are traceless.
We can define direct products of different groups G = H × K, with elements g = (h, k) and product
g1 g2 = (h1 h2 , k1 k2 ). The dimension is obviously the sum of two dimensions. We can choose a basis where
the generators of G is the union of generators of H and generators of K with the two subsets commuting
with each other. Conversely if the set of generators {Tj } can be written as
n o[n o
{Tj } = ta(1) tb(2) (1.17)

with the two subsets mutually commuting then any g can be written as
iαj tj(1) iβk tk(2)
g=e e (1.18)
Then locally G = H × K, G is a direct product. For example we consider SO(4), the generators are
Jij = −Jji . Rotations are associated with a plane. We can define two sets of generators
   
1 1 1 1 1 1
Hi = (J23 + J14 ), (J31 + J24 ), (J12 + J34 ) , Ki = (J23 − J14 ), (J31 − J24 ), (J12 − J34 )
2 2 2 2 2 2
(1.19)
We can check that the Hi commute with Kj and they individually form SO(2) algebra. So the SO(4)
algebra is the same as SU (2) × SU (2) algebra. The correspondence looks like
 
eiα·H , eiβ·K ←→ eiα·H eiβ·K (1.20)

Suppose we rotate by 2π we can have (−I, −I) or (I, I) on the left side, but only I on the right side. So
we have the relation
SU (2) × SU (2)
SO(4) ∼= (1.21)
Z2
Similar procedure can be carried out for the Lorentz group SO(3, 1). Because the commutation relation
for the SO(3, 1) algebra is obtained by just replacing the δij in the commutation relation of SO(4) by gµν ,
the corresponding modification in H and K is just adding i in front of all the J with index 1:
   
1 1 1 1 1 1
Hi = (J23 + iJ14 ), (iJ31 + J24 ), (iJ12 + J34 ) , Ki = (J23 − iJ14 ), (iJ31 − J24 ), (iJ12 − J34 )
2 2 2 2 2 2
(1.22)

5
Quantum Field Theory III Lecture 1

Now because there is a factor of i in front of some of the original generators, we would expect the exponential
j
eiαj H to have some non-periodic behavior in one direction. This is what makes the Lorentz group non-
compact.
A group is simple if there is not subgroup N such that gN g −1 = N . There is no normal subgroups. A
semisimple group is a group without an abelian normal subgroup. A Lie group is simple if it can’t (even
locally) be written as a product H × K. A Lie group is semisimple if it has no U (1) factors. SO(3) is
simple and semisimple. SU (4) ∼ SU (2) × SU (2) is semisimple but not simple. U (N ) ∼ U (1) × SU (N ) is
neither simple nor semisimple.
Any compact Lie group can be build up from simple compact Lie groups either by direct products or
by direct products with some quotients. So if we know the simple compact groups we know all the compact
groups. So we will only care about simple Lie groups. These are classified as SU (N ), SO(N ), Sp(N ), E6 ,
E7 ,. . .
Now we consider group representations. We associate with every group element a matrix D(g) which
satisfies
D(g1 )D(g2 ) = D(g1 g2 ) (1.23)
Two g’s can have the same matrix, so there is aways the trivial representation D(g) = 1 which exists for
any group. The dimension of the representation is the dimension of D which has nothing to do with the
dimension of the group. Matrices correspond to linear transformations, so an r-dimensional representation
corresponds to a set of r objects that mix under these linear transformations.
Two representations are equivalent D(1) (g) ∼
= D(2) (g) if for every g we have

D(1) (g) = SD(2) (g)S −1 (1.24)

and S does not depend on g. This is just like changing basis.


A unitary representation is where all D(g) are unitary matrices. A finite or compact group implies that
all representations are equivalent to a unitary representation. A noncompact group is not so, i.e. we can
find representations not equivalent to unitary ones. The Lorentz group has unitary representations, but
they are infinite dimensional. It also has finite dimensional representations, but they are not unitary.
A reducible representation means that by a change of basis we can write
h i
D(g) = S −1 D(1) (g) ⊕ D(2) (g) S (1.25)

so the representation is equivalent to a block-diagonal one. We can build up all representations this way,
so we will be only interested in irreducible representations. For example, for SO(3) we have vectors which
transform like
Vi −→ Vi0 = Rij Vj (1.26)
But we also have rank-2 tensors which transform like

Tij −→ Tij0 = Rik Rjl Tkl (1.27)

But we can write any tensor into symmetric part and antisymmetric part. A symmetric tensor remains
symmetric under rotation, so does the antisymmetric tensor. It can be further decomposed because trace
is invariant under rotation. So
Tij = T δij + Aij + Sij (1.28)

6
Quantum Field Theory III Lecture 2

2 Lecture 2
2.1 Group Representations
Remember we can associate a matrix D(g) to an element g of the group. We can also have a representation
of the algebra, associating a matrix D(T a ) with a generator T a . But people will write T a for a set of
operators, and ta for another set of operators.
We are usually interested in all the irreducible representations of a group G. This is a very nontrivial
problem. However, we always have

1. Trivial representation: D(g) = 1. Dimension is just 1. This is important because if we have anything
invariant under the group, this is just the representation.

2. Adjoint representation: describing how the generators transform. For example if we have
a b
g −1 eiβa T g = eiαb T (2.1)

and expand it to linear order we have

βa g −1 T a g = αb T b (2.2)

This defines a transformation law. The dimension of this representation is the same as the dimension
of the group. Now let’s assume
g −1 Tk g = ckl T l (2.3)
and expand g to linear order of the generators, then we get
h i
g −1 Tk g = Tk − iαj [Tj , Tk ] = Tk − iαj (if ljk Tl ) = δkl + αj f ljk Tl (2.4)

The coefficient in front of the generators should tell us the matrix representation, so matrix repre-
sentation of the generator is
[D(Tj )]kl = −ifjkl (2.5)

For example let’s consider SO(3). The adjoint representation is just

(Jj )kl = −ijkl (2.6)

Now if we adopt the adjoint representation then we can convert the commutation relation

[Jj , Jk ] = ifjkl Jl (2.7)

into an equation of the f ’s. This will just give us the Jacobian identity.
We usually want to know the dimension of the representations, as well as the properties, i.e. the multiplet
structures. For SU (2) we know the answer. We can label the representations by j = 0, 1/2, 1, . . . . The
dimension is just 2j + 1 = 1, 2, 3, . . . . We label the representations by J 2 and label states by multiplets
of Jz . These are just angular momentum multiplets. We usually are also interested in Clebsh-Gordon
problem, i.e. decomposing tensor product representations into direct sums
M
D(1) (g) ⊗ D(2) (g) = D(i) (g) (2.8)
i

7
Quantum Field Theory III Lecture 2

This is just a generalization of the “addition of angular momenta”. We are interested in tensor methods
which give us complete answers for SU (2), SO(3), SU (3), and smallest representations for SO(N ) and
SU (N ), (except for spinors).
Let’s consider SO(3). We have the trivial representation, with dimension 1. This is just how scalars
transform under rotations. We also have the defining representation D(R) = R, i.e. the matrix is just the
rotation matrix in 3 dimensions. So the dimension of the representation is just D = 3. This is the vector
representation. Now we can also have rank n tensor representations. A rank n tensor transforms as

T i1 i2 ...in −→ T 0i1 ...in = Ri1 j1 . . . Rin jn T j1 ...jn (2.9)

Now note that there are some invariant tensors, for example the Kronecker delta

δ ij −→ δ 0ij = Rik Rjl δ kl = δ ij (2.10)

We also have the completely antisymmetric symbol

ijk −→ 0ijk = Ril Rjm Rkn lmn = ijk det R = ijk (2.11)

These are the only invariant tensors. Let’s use these to consider an arbitrary tensor M ij

M ij δ ij = S, M ij ijk = V k (2.12)

Here S is a rank 0 tensor and V k is a rank 1 tensor. In this way, we can start with any rank 2 tensor and
subtract off these tensors
1   1  
M̃ ij = M ij − δ ij M ab δ ab − ijk M ab abk (2.13)
3 2
Now the first term subtracts the trace, so the new tensor is traceless. The second term subtracts the
antisymmetric part which has 3 components, so this tensor is symmetric and has 5 independent components.
We can carry out the same process for rank 3 tensor M ijk . We can multiply δ ij to get something of
rank 1, and multiply by ijl to get something of rank 2. So what is left must be symmetric in all indices,
and traceless in any pair of indices. This generalizes to rank n. Now we want the number of independent
components. Symmetric in all indices gives us (n + 1)(n + 2)/2 independent entries because we have
3 different indices and we want to separate them into three groups. There are Cn+2 2 places to put the
separators. From the traceless condition we have n(n − 1)/2 conditions. So we do the math
(n + 1)(n + 2) n(n − 1)
− = 2n + 1 (2.14)
2 2
This is what we expect from spherical tensor representations of rank n.
Now we know the irreducible representations, we can consider the Clebsch-Gordon problem. In the
representation D(m) ⊗ D(n) we have some tensor look like S i1 ...im T j1 ...jn . We can also multiply by δ ij
and ijk . These will lower the rank by 2 and 1 respectively. We can multiply repeatedly, but remember
that multiplying two ’s is equivalent to multiply many δ’s. So we can only consider multiplying many
δ’s or multiplying  followed by many δ’s. Suppose m > n. If we multiply by 0, . . . , n δ’s then we get
representations of rank (m + n), (m + n − 2), . . . , (m − n). In the other case we have ranks (m + n −
1), . . . , (m − n + 1). So we have

D(m) ⊗ D(n) = D(m+n) ⊕ D(m+n−1) ⊕ · · · ⊕ D(m−n) (2.15)

8
Quantum Field Theory III Lecture 2

Again this is a result familiar from angular momentum addition.


Now let’s consider SO(N ). The invariant tensors are δ ij and i1 ...iN . For rank 0 we still have the
dimension 1 trivial representation, and for rank 1 we have the N dimensional vector representation. The
rank 2 case we have two parts, the symmetric traceless part with rank N (N +1)/2−1 and the antisymmetric
part with rank N (N − 1)/2. For higher rank we need to consider representations of the permutation group,
and we need to use the Young’s Tableaux.
Usually the rank 2 tensors above are irreducible, but for SO(4) it is. Consider the rank 2 tensor
antisymmetric M ab in dimension 4. We can get the dual of M ab
1
M̃ cd = M ab abcd (2.16)
2
And the double dual is just itself. We can form linear combinations

(M + M̃ )ab , (M − M̃ )ab (2.17)

which are self-dual and anti-self-dual respectively. This reduces the adjoint representation into two 3
dimensional representations. This is like how E and B fields mix, apart from factors of i.
Now we need to consider spinor representations, i.e. representations of the group Spin(N ). The spinor
representation for SO(2N + 1) has dimension 2N and representation for SO(2N + 2) also has dimension
2N . Supersymmetry is the symmetry which relates the tensor representations to the spinor representations,
which are essentially bosonic and fermionic fields. On the left side are the generators of Poincaré group and
the right side are the SUSY generators. But in too high dimensions this magic doesn’t work because the
dimension of spinor representations grows exponentially. So the maximal dimension for SUSY is D = 11.
Now let’s look at SU (3). The trivial representation has dimenison 1. The defining representation is
just D(U ) = U which has 3 complex dimensions. So we can take the conjugate representation D(U ) = U ∗
which also has complex dimension 3. The vectors transform like

Wi −→ Wi0 = Ūi j Wj (2.18)

where Ūi j = (U ij )∗ . This is to be contrasted with the defining representation where vectors transform like

V i −→ V 0i = U ij V j (2.19)

The invariant tensors are ijk , ijk , δji , but not δij and δ ij . An arbitrary representation will have m upper
indices and n lower indices:

T i1 ...imj1 ...jn −→ U i1k1 . . . U imkm Ūj1 l1 . . . Ūjn ln T k1 ...kml1 ...ln (2.20)

This is labeled rank (m, n) representation. Now we can reduce it to (m − 1, n − 1) with δ, and into
(m + 1, n − 2) by ijk or into (m − 2, n + 1) by ijk . The irreducible tensors are those symmetric in all
upper and lower indices, and traceless in any pair of upper and lower indices. So (m, n) representation has
dimension
1
D = (m + 1)(n + 1)(m + n + 2) (2.21)
2
These are all the irreducible representations of SU (3).
Let’s work out the dimensions of various representations, listed in table 1. Note that there are different
representations with the same dimension, so for higher representations we can’t use dimensions to label

9
Quantum Field Theory III Lecture 2

(m, n) Dimension
(0, 0) 1
(1, 0) 3
(0, 1) 3̄
(2, 0) 6
(0, 2) 6̄
(1, 1) 8→ adjoint
(3, 0) 10
(4, 0) 15
(2, 1) 15

Table 1: Various representations of SU (3)

the representations. In QCD the quarks live in the 3 representation and antiquarks in 3̄ representation.
All physically observables live in colorless state which is in the (1, 1) representation. These include the
spin 1/2 baryons and spin 0 mesons. The spin 3/2 baryons live in the 10 representation.
This method generalizes to SU (N ) where we have invariant tensors δji , i1 ...iN , i1 ...iN . We have the
(1, 0) and (0, 1) representations which are dimension N . We also have (2, 0) symmetric and antisymmetric
representations. The (1, 1) traceless representation is the adjoint representation. For N = 2 we have
the 2 dimensional representations (1, 0) and (0, 1), but they are connected by the transformation σ2 . So
defining representation is equivalent to its conjugate. We call this a pseudoreal representation. Because the
invariant tensor is ij , and a tensor of the kind (N, 0) can be made symmetric between all indices. Because
indices take values 1 and 2, there are N + 1 independent components. This is just the representation of
spin N/2.

10
Quantum Field Theory III Lecture 3

3 Lecture 3
3.1 Continued SU (3)
Last time we were considering SU (3). We got the tensor representations (m, n). Let’s do some explicit
examples. Let’s consider the representation 3 ⊗ 3. Any element can be written as V i W j . We can lower
the index by ijk to become Uk , which is in the representation 3̄. The remainder is something symmetric,
which is inside representation (2, 0) = 6. So we have

3 ⊗ 3 = 3̄ ⊕ 6, 3̄ ⊗ 3̄ = 3 ⊕ 6̄ (3.1)

Similarly we can think about 3 ⊗ 3̄. We can contract V i Wj with δji to get a scalar. So we have

3 ⊗ 3̄ = 1 ⊕ 8 (3.2)

Now let’s consider 3 ⊗ 6. An element V i S jk can be contracted with ijl to become something in the 8
representation, and the remainder is something totally symmetric in the three indices which is 10. So we
have
3 ⊗ 6 = 8 ⊕ 10 (3.3)
So we can build upon what we know

3 ⊗ 3 ⊗ 3 = 3 ⊗ (3̄ ⊕ 6) = 3 ⊗ 3̄ ⊕ 3 ⊗ 6 = 1 ⊕ 8 ⊕ 8 ⊕ 10 (3.4)

This has some physical interpretation. If we have a meson, which is formed by a quark and an antiquark,
can be either a singlet or an octet. A baryon which consists of three quarks can be a singlet, an octet or
¯
an dextet. Similarly we can work out 8 ⊗ 8 and 10 ⊗ 10.

3.2 Roots and Weights


Now let’s consider SU (2) again. Remember we can form raising and lowering operators

J± = J1 ± iJ2 (3.5)

and we have
[J3 , J± ] = ±J± (3.6)
So we know J+ increases J3 by 1 an J− decreases J3 by 1. So for any finite representation all the eigenvalues
of J3 are integers or half-integers. For J = 1 we have states J3 = −1, 0, 1. For (J = 1) ⊗ (J = 1)
representation we have

(−1, 0, 1) × (−1, 0, 1) = −2, −1, 0, −1, 0, 1, 0, 1, 2 = (0) + (−1, 0, 1) + (−2, −1, 0, 1, 2) (3.7)

This can be checked directly by the above methods.


Now let’s work out SU (3) in the triplet representation. Let’s define the generators T a = λa /2 where

Tr λa λb = 2δ ab (3.8)

11
Quantum Field Theory III Lecture 3

These λ matrices are called Gell-Mann matrices


     
0 1 0 0 −i 0 1 0 0
1 2 3
λ = 1 0 0 , λ = i
   0 0 , λ = 0 −1
  0 (3.9)
0 0 0 0 0 0 0 0 0
     
0 0 1 0 0 −i 1 0 0
1
λ4 = 0 0 0 , λ5 = 0 0 0  , λ8 = √ 0 1 0 (3.10)
1 0 0 i 0 0 3 0 0 −2
   
0 0 0 0 0 0
λ6 = 0 0 1 , λ7 = 0 0 −i (3.11)
0 1 0 0 i 0

The commutation relations are

[T3 , T1 ± iT2 ] = ± (T1 ± iT2 ) , [T8 , T1 ± iT2 ] = 0 (3.12)

And the other set is



1 3
[T3 , T4 ± iT5 ] = (T4 ± iT5 ) , [T8 , T4 ± iT5 ] = ± (T4 ± iT5 ) (3.13)
2 2
And similar for T6 ± iT7 . So the first set of raising and lower operators raise and lower the eigenvalues

of T3 by 1 and preserve the eigenvalue of T8 . The other two sets change the T3 by 1/2 and T8 by 3/2.
The algebra is called rank 2 because there is maximally two commuting generators. The root diagram is
as shown in figure 3.1.

T8

T4 − iT5 T4 + iT5

T1 + iT2
T1 − iT2 T3

T6 − iT7 T6 + iT7

Figure 3.1: Root diagram for SU (3)

These roots are actually the simultaneous eigenvalues of the generators T3 and T8 in the adjoint repre-
sentation. The roots form a root system, which is defined as follows:

• If β is a root, then −β is also a root

12
Quantum Field Theory III Lecture 3

• Reflection across a plane perpendicular to a root is a symmetry of the root system


Given two simple roots, we can reconstruct the rest by the above two properties. The set of reflection
symmetries of the root system is called the Weyl group. In general for an algebra of rank r, there are r
simple roots. Now we define the simple roots:
• α · β ≤ 0 for α 6= β

• Any root in the root system can be expressed as a sum


X
α= ni αi (3.14)
i

where αi are simple roots and ni are all positive or all negative
The simultaneous eigenvectors of T3 and T8 are just
     
1 0 0
0 , 1 , 0 (3.15)
0 0 1

The eigenvalues can be plotted in a weight diagram as shown in figure 3.2. The different weights can

T8

T3
× ×

Figure 3.2: Weight diagram of the 3 and 3̄ representations of SU (3)

be obtained from one another by applying the root vectors. For the 3̄ representation the eigenvalues are
negative of the eigenvalues of the 3 representation, and we plot them in the same weight diagram with
crosses.
Now let’s look at the relation 3 ⊗ 3̄ = 1 ⊕ 8. The eigenvalues will give us
            
1 1 1 1 1 1 1 1 1 1
, √ , − , √ , 0, − √ ⊗ − , √ ,− − , √ , − 0, − √
2 2 3 2 2 3 3 2 2 3 2 2 3 3
√ √ √ √
=(0, 0) + (1, 0) + (1/2, 3/2) + (−1, 0) + (0, 0) + (−1/2, 3/2) + (−1/2, − 3/2) + (1/2, − 3/2) + (0, 0)
(3.16)

13
Quantum Field Theory III Lecture 3

T8

× ×

×
×
T3
× ×

Figure 3.3: Weight diagram for the 8 representation

The first one is obviously the scalar. The rest eight must be from the 8 representation. Let’s plot the
weights in figure 3.3. Note that there are two weights at the origin. This reflects that there are two
commuting generators, and because 8 is just the adjoint representation, this is just the result that the
action of a generator on another is just zero. The commuting generators form the Cartan subalgebra,
which is the maximal commuting subalgebra of the given Lie algebra.
Let’s look at the generators. First there is an SU (2) subgroup formed by T1 , T2 , T3 . We call this group
I which is also known as the isospin. √We also have the generator T8 which generates the U (1) hypercharge
Y . Usually we define Y to be Y = ( 3/2)T8 . For mesons this is the strangeness and for baryons this is
strangeness plus 1.
Now we can think of group of particles and denote it as (I)Y , where I is the isospin of the representation.
For example the 3 representation will be denoted
 1/3
1
3= + (0)−2/3 (3.17)
2
As a consistency check the total hypercharge must be zero, as the trace of the generators is zero. This is
true because there are 2 states in the first term. We can also write
 −1/3
1
3̄ = + (0)2/3 (3.18)
2
We can also look at 3 ⊗ 3 = 3̄ ⊕ 6, so
"  # "  #
1 1/3 −2/3 1 1/3 −2/3
3⊗3= + (0) × + (0)
2 2
 −1/3  −1/3 (3.19)
1 1
2/3
= (1) + (0) +2/3
+ + (0)−4/3
2 2
We can pick out a 3̄ and write
 −1/3
1
6 = (1) 2/3
+ + (0)−4/3 (3.20)
2

14
Quantum Field Theory III Lecture 3

The weight diagram is as shown in figure 3.4. The weights form a equilateral triangle. Similarly we can

T8

× × ×

T3
× ×

Figure 3.4: Weight diagram for representation 6

work out the expression for 8


 1  −1
01 1
8 = (1) + + + (0)0 (3.21)
2 2
And the weight diagram has been given above in figure 3.3. The double weight√ at the origin is from (0)0
and (1)0 . Note that there are two ways to go from a diagonal state at ( 3/2, 1/2) to the origin using
ladder operators, and if we work out the ladder operators we will see that they don’t commute. The fact
that they don’t commute tells us that there are two different states at the origin.

3.3 The Algebra of SO(N )


Let’s talk about SO(N ). The generators of this group are just angular momenta Jab = −Jba where
a, b = 1, . . . , N . In dimensions higher than 3, we need to associate rotations with the plane of rotation,
hence there are two indices for each generator. The commutation relation of these generators is

[Jab , Jcd ] = i (δac Jbd − δad Jbc − δbc Jad + δbd Jac ) (3.22)

This can be seen from the fact that it must be antisymmetric in ab and cd. And if none of the indices
match then this should be zero. So the only case of nonzero commutator is that one index match while
the other don’t, in which we should have [J23 , J31 ] = iJ12 .
Now we want to find the Cartan subalgebra of commuting generators. First suppose we work in even
dimension SO(2k). We can take J12 , J34 , . . . , J2k−1,2k and by the above commutation rule these generators
commute with each other. There are k generators so the rank of the group is k. Now for SO(2k + 1) we
also have the same set of generators, so the rank is still k.
To be concrete let’s consider SO(5) and the Cartan subalgebra is formed by J12 and J34 . The ladder

15
Quantum Field Theory III Lecture 3

operators are formed as follows

J13 ± iJ23 −→ ∆J12 = ±1 (3.23)


J14 ± iJ24 −→ ∆J12 = ±1 (3.24)
(J13 ± iJ23 ) ± (J14 + iJ24 ) −→ ∆J34 = ±1, ∆J12 = 1 (3.25)

This set is complicated, but the other set is easier to work out

J15 ± iJ25 −→ ∆J12 = ±1, ∆J34 = 0 (3.26)


J34 ± iJ45 −→ ∆J12 = 0, ∆J34 = ±1 (3.27)

So the root diagram is shown in figure 3.5. The simple roots are (−1, 1) and (1, 0). Note for even dimensions

J34

J12

¡++¿

Figure 3.5: Root diagram for SO(5)

we don’t have the Ji5 roots. So we are ready to write down the roots for SO(2k). The roots are

± ei ± ej , where i 6= j = 1, . . . , k (3.28)

The number of roots is 2k 2 − 2k and rank is k. The number of generators is 2k 2 − k which is the dimension
of the group. For SO(2k + 1) the roots are

± ei ± ej , where i 6= j = 1, . . . , k; and ± ei , where i = 1, . . . , k (3.29)

The number of roots is 2k 2 and rank is also k.


Now we generalize to SU (N ). The roots are

± (ei − ej ), where i 6= j = 1, . . . , N (3.30)

The number of roots is N 2 − N and rank is N − 1 so the dimension is N 2 − 1 which is correct. Note
that the root vectors are in N − 1 dimensional space but we wrote them as if they are in N dimensional
space. This is because they are all perpendicular to the vector e1 + e2 + · · · + eN . So they lie in the same
hyperplane.

16
Quantum Field Theory III Lecture 3

3.4 Classification of Root Systems


Now in the root diagram we can start with any root α and construct a SU (2) subalgebra. And any other
roots can be classified in this SU (2) algebra as spin multiplets. The way to do it is just to take the
projection of any β onto the root α, and the projection should satisfy the spin multiplet condition, i.e. the
length should be a half integer. Therefore we require that
α n
β·α=β· |α| = |α| (3.31)
|α| 2

So that
2β · α
=n (3.32)
α2
But equally we can choose β as the SU (2) subalgebra generator, so equivalently we need

2β · α
=p (3.33)
β2
where p is also an integer. The above two identities must be true for any pair of distinct roots α, β in the
root system. Let’s multiply the two expressions to get

4(α · β)2
= 4 cos2 θ = n · p (3.34)
α2 β 2
And if we divide them we get
α2 p
2
= (3.35)
β n
Now from these two conditions, because cos θ < 1, so we get np < 4. We have some possibilities

• np = 0 This implies that α ⊥ β

• np = 1 This implies that cos θ = ±1/2 and |α| = |β|, θ = 60◦ , 120◦ . This is like SU (3)
√ √
• np = 2 Assume n = 1, p = 2, then cos θ = ±1/ 2 and θ = 45◦ , 135◦ and |α| = 2 |β|. This is like
SU (5)
√ √
• np = 3 Then cos θ = ± 3/2 and θ = 30◦ , 150◦ and |α| / |β| = 3. The only example of this is
called the group G2 . We will talk about this on Monday

These are the necessary condition for a root diagram. Now what we need to do is just work out all the
possible solutions to these constraints. It is just a big geometry problem. The answer is that the only
possible root systems are
An , Bn , Cn , Dn , E6 , E7 , E8 , F4 , G2 (3.36)
Where A, B, C, D are called classical groups with n = 1, . . . , ∞ and these are the rank of the algebra. The
rest are called exceptional ones for obvious reasons. Correspondence of these root systems with groups
is that A’s correspond to SU (n + 1), B’s correspond to SO(2n + 1), C’s correspond to Sp(2n) and D’s
correspond to SO(2n).

17
Quantum Field Theory III Lecture 4

4 Lecture 4
4.1 Roots Continued
Last time we talked about root systems. Remember we have the criterion

4(α · β)2
= np (4.1)
α2 β 2

The only possibilities are
√ the roots are orthogonal, or |α| = |β| when θ = 60◦ , 120◦ , or |α| = 2 |β| when
θ = 45◦ , 135◦ , or |α| = 3 |β| when θ = 30◦ , 150◦ .
Last time we rote down the roots for various groups explicitly
• SU (N ): roots: ±(ei − ej ), i 6= j = 1, 2, . . . , N . This is called AN −1

• SO(2N ): roots: ±ei ± ej , i 6= j = 1, . . . , N . This is called DN

• SO(2N + 1): roots: ±ei ± ej , ±ei , i, j = 1, . . . , N . This is called BN


We can define dual on a root system
α
α −→ α̃ = (4.2)
|α|2
We can have multiplication of the dual

α·β (α̃ · β̃)2 (α · β)2


α̃ · β̃ = , 2 = (4.3)
|α|2 |β|2 |α̃|2 β̃
|α|2 |β|2

So the dual of a root system is also a root system. This is effectively changing the relative lengths of
the roots. We can construct new root systems from known ones by taking the dual. In fact we have the
following root system from the dual of BN :
• Sp(2N ) or Sp(N ): roots: ±ei ± ej , ±2ei , i, j = 1, . . . , N . This is called CN
Note it is sometimes the convention to call Sp(N ) for Sp(2N ) as there is really nothing asPSp(3).
Now let’s look at O(N ). This
P is2 the group of linear transformations that preserves x2i . However
U (N ) is the group preserving |zi | . Now we can generalize to quaternions

q = q0 + iq1 + jq2 + kq3 (4.4)

where i2 = j 2 = k 2 = −1, and ij = −ji = k, jk = i and ki = j. The parameters qi are all real. Note that
the quaternions are not really a field, so we are actually considering a real vector space of dimension 4,
tensored with any other vector space structure. We can generalize the idea of complex conjugation as

q̄ = q0 − iq1 − jq2 − kq3 (4.5)

so that |q|2 = q̄q = q q̄ =


P 2
qi . We can represent
P these as 2 × 2 matrices as ej → iσj . We can define the
group acting on quaternions which preserves |qa |2 under the transformation qa → Mab qb where Mab is a
quaternionic matrix. Because of the above representation we can think of Mab as 2N × 2N complex matrix
plus some constraints. Suppose
Mab = Iab + Oab + . . . (4.6)

18
Quantum Field Theory III Lecture 4

where Oab is small in magnitude. So under the transformation the quaternions will become (check!)

qa0 = qa + Oab qb , q̄a0 = q̄a + q̄b Q̄ab (4.7)

So the length will transform like


X X X
q̄a0 qa0 =

q̄a qa + q̄a Qab qb + q̄a Q̄ba qb + . . . (4.8)
ab

The term in the bracket should vanish, so we require

Q̄ab = −Qba (4.9)

What does this relation tell us? Remember the matrix element
X j
Qab = Q0ab + Qab ej (4.10)
j

When taking conjugate the ej change signs, so we need Q0ab to be N × N real antisymmetric matrix, and
Qjab should be symmetric N × N real matrices. Consider the linear combination

G = I2 ⊗ A + iσ1 ⊗ S1 + iσ2 ⊗ S2 + iσ3 ⊗ S3 (4.11)

and these matrices form a Lie algebra. Another way to define it is the 2N × 2N unitary matrices such that
   
† 0 IN 0 IN
M M= (4.12)
−IN 0 −IN 0

and the linearization is M = I + iH where H has the solution


 
X B
H= (4.13)
B ∗ −X ∗

with X hermitian and B symmetric. This completes our discussion of classic algebras.

4.2 Exceptional Algebras


Let’s look at exceptional cases. We have

• G2 : roots: ±(ei − ej ), ±(2ei − ej − ek ), i, j, k = 1, 2, 3

This has 12 roots and the rank is 2, so the dimension is 14 and the smallest representation has dimension
7. We also have

• F4 : roots: ±ei , ±ei ± ej , 21 (±e1 ± e2 ± e3 ± e4 ), i 6= j = 1, . . . , 4

There are 48 roots and rank is 4, so dimension of the group is 52 and smallest representation has dimension
26.
We now look at the other ones

• E8 : roots: roots of SO(16) plus 12 (±e1 ± · · · ± e8 ) with even number of + signs

19
Quantum Field Theory III Lecture 4

there are 240 roots and rank is 8, so dimension of the group is 248. Smallest representation also has
dimension 248.

• E7 : roots: roots of E8 perpendicular to e1 − e2

The rank is 7 and dimension is 133, with smallest representation 56 dimensional.

• E6 : roots: roots of E7 perpendicular to e2 − e3

This has rank 6 and dimension 78 with smallest representation 27 and 27.¯ Note we don’t have other
exceptional ones. E5 is the same as D5 = so(10).
We can go a step further from quaternions and arrive at octonions A = A0 + 71 Aj ej where e2j = −1
P

and Ā = A0 − 71 Aj ej . Similarly we can define


P

p
Norm(A) = ĀA (4.14)

From complex field to quaternions we lost commutativity and from quaternions to octonions we loose
associativity. It is difficult to do any calculations here. But there is some group that preserves the
multiplication rules for octonions, and that group is just G2 .

4.3 Classification According to Dynkin Diagrams


The rank of an algebra means there are r simple roots. The simple roots have α · β < 0. The simple roots
for the above algebras are

• SU (N ): e1 − e2 , . . . , eN −1 − eN

• SO(2N ): e1 − e2 , . . . , eN −1 − eN , eN −1 + eN

• SO(2N + 1): e1 − e2 , . . . , eN −1 − eN , eN

• Sp(2N ): e1 − e2 , . . . , eN −1 − eN , 2eN

Dynkin found a way to represent these simple roots by diagrams: represent a simple root by a small circlr.
They are separated if they have angle 90◦ , connected if they have angle 120◦ , double connected if 135◦ and
triple connected if 150◦ . An arrow is drawn pointing from the longer root to the shorter one.
Note that there are some equivalence in the algebras. Obviously A1 = B1 = C1 , so su(2) = so(3) =
sp(2). We also have B2 = C2 so so(3) = sp(4). And that A3 = D3 so su(4) = so(6). If we remove any
connection between two roots, then we get a subgroup of the original group. Because the roots of E8 is
the roots of SO(16) plus some extras, so SO(16) is a subgroup of E8 . In fact the 248 representation of E8
is actually
248 = 120 ⊕ 128 (4.15)
where 120 is the adjoint representation of SO(16) and 128 is its spinor representation.

20
Quantum Field Theory III Lecture 4

An : ... E6 :

Bn : ... > E7 :

Cn : ... < E8 :

F4 : >
Dn : ...
G2 : >

Figure 4.1: The Dynkin diagrams for all kinds of simple Lie algebras

× ×

Figure 4.2: Weight diagram for vector representation of SO(2N ) and SO(2N + 1)

4.4 Weights
Given any representation, we can represent the states using weights. For any weigth w, we need to have
2w · α
= n = 0, ±1, ±2, . . . (4.16)
|α|2

for any α in the root system. Let’s look at some representations of SO(2N ) first. The vector representation
has dimension 2N . The weights are just w = ±e1 , ±e2 , . . . , ±eN . So we can draw them in figure 4.2.
Now for SO(2N + 1) the vector representation has dimension 2N + 1. The weights are those from
SO(2N ) plus 0, which is also plotted in figure 4.2 in dot. Note we can go from one weight to another using
the root vectors. Now for any SO(2N ) or SO(2N + 1) we have the vectors 21 (±e1 ± · · · ± eN ) which also
satisfy the condition for weights. These are the spinor representations of SO(X). The dimension for these
representations are 2N . But these are not all irreducible. For SO(2N + 1) the 2N spinor representation
is irreducible because all the weights are connected by root vectors. For SO(2N ) because of the root
structure, representations with even number of + signs can’t go to odd number of + signs. So if N = 2k

21
Quantum Field Theory III Lecture 4

SO(8k) real
SO(8k + 1) real
SO(8k + 2) complex
SO(8k + 3) pseudoreal
SO(8k + 4) pseudoreal (two)
SO(8k + 5) pseudoreal
SO(8k + 6) complex
SO(8k + 7) real

Table 2: Spinor representations for SO(X)

then there are two spinor representations with dimensions 2N −1 , characterized by even + and even − or
odd + and odd −. For N = 2k + 1 we also have two spinor representations of dimensions 2N −1 with even
+ and odd − or odd + and even −, but these are mutual conjugate representations.
For N = 2k we can make the two spinor representations almost real. In fact we have the chart
of situations shown in table 2. We can construct these representations explicitly using the Γ matrices.
Suppose we have Γa where a = 1, . . . , N with anticommutation relations

{Γa , Γb } = 2δab (4.17)

Define Mab = − 2i Γa Γb where a 6= b and these are the generators for SO(N ). For odd N = 2k + 1 we have

Γ 1 = σy ⊗ σz ⊗ · · · ⊗ σz (4.18)
Γ2 = −σx ⊗ σz ⊗ · · · ⊗ σz (4.19)
Γ 3 = I ⊗ σy ⊗ · · · ⊗ σz (4.20)
Γ4 = I ⊗ (−σx ) ⊗ · · · ⊗ σz (4.21)
...
Γ2k−1 = I ⊗ · · · ⊗ I ⊗ σy (4.22)
Γ2k = I ⊗ · · · ⊗ I ⊗ (−σx ) (4.23)
Γ2k+1 = σz ⊗ · · · ⊗ σz (4.24)

It can be checked that these satisfy the anticommutation relations. We have also the following relation

Γ1 Γ2 . . . Γ2k = ik Γ2k+1 , Γ1 Γ2 . . . Γ2k Γ2k+1 = ik (4.25)

All these are 2k × 2k matrices and this is a 2k spinor representation.


How about SO(2k)? We just take Γ1 , . . . , Γ2k . We define

Γ̄ = (−1)k Γ1 . . . Γ2k = Γ2k+1 (4.26)

Note Γ̄2 = I. So states with eigenvalue 1 will be in one representation and with −1 will be in the other
representation. This is in analogue with γ5 in Dirac representation. In this way we can decompose the
representation into two irreducible representations.
Remember M12 = − 2i Γ1 Γ2 which is real. Similarly for M34 , M56 , . . . . Now because for any representa-
tions  ∗ ∗
(D(g))∗ = eiβT = e−iβT (4.27)

22
Quantum Field Theory III Lecture 4

so for real generators there is a conjugate representation. This can also be seen from the Dynkin diagrams.
The An has Z2 symmetry by reflecting across the vertical axis. This means that we have complex repre-
sentations. Bn and Cn have no symmetry. Dn has symmetry across the horizontal, so it has some complex
representations. E6 has the reflection symmetry so it also has complex representations. Now D4 = so(8)
has a triangular symmetry. Its vector and two spinor representations all have dimension 8. The weights
are

• Vector ±ej , j = 1, 2, 3, 4
1 1
• Spinor ± (+ + ++), ± (+ + −−) and permutations
2 2
1
• Spinor’ ± (+ + +−) and permutations
2
Note that these weights transform into each other under rotations in 4D. This is the reason superstring
theory can only live in 10D. We know massless particles in D dimensions has number of states corresponds
to rotation group in D − 2 dimensions, so only in 10D can the different kinds of particles be transformed
into each other. And that’s the only dimension it works.

23
Quantum Field Theory III Lecture 5

5 Lecture 5
5.1 Spontaneous Symmetry Breaking
We are going to talk about spontaneous symmetry breaking. For the moment we talk about global sym-
metries. We use the term symmetry to denote any invariance of L or H. It can be either discrete or
continuous and in the latter case it will lead to a conserved current ∂ µ Jµ = 0.
There are two cases when there is a symmetry, i.e. the ground state is invariant or not. If it is not,
then we say the symmetry is spontaneously broken. But this is terrible terminology because symmetry is
not broken at the lagrangian level. It is not the same as adding a perturbation that actually breaks the
symmetry. Classically if we have a potential like a Mexican hat in 1D, then the ground state sits at either
point a or −a in the bottom, and it is not surprising that the state doesn’t have symmetry. Quantum
mechanically there will be two states |+i and |−i which center at a and −a repspectively. The ground
state and the first excited state are symmetric and antisymmetric linear combinations of the two states,
and the energy difference is proportional to the tunnelling amplitude
 Z a 
−B
p
∆E ∼ e ∼ exp − dx 2m(V − E) (5.1)
−a

So quantum mechanically it looks like we can’t have spontaneous symmetry breaking.


Suppose we have many degrees of freedom with spin up or down. The interaction Hamiltonian is
X
Hint = −k Si Si+1 (5.2)

The more spin we have, the more the tunneling factor is suppressed, and in the limit of infinite spins, we
have two degenerate ground states where either all spin point up or point down. Actually we don’t need
infinity, and 1023 is good enough.
But even if we can suppress the tunnelling constant, we can still take linear combinations. We want
to show that this is not true in field theory. We consider field theory with a discrete set of vacuum states
|ni. We have
P2 |ni = 0 (5.3)
because the vacuum is translational invariant. Consider the vacuum expectation value of some operators
Z
X X
n |A(x)B(y)| n0 = hn |A(x)| mi m |B(y)| n0 + d3 p hn |A(x)| Np i Np |B(y)| n0




m N
X Z X
hn |A(0)| mi m |B(0)| n0 + d3 p e−ip·(x−y) hn |A(0)| Np i Np |B(0)| n0



=
m N
(5.4)

In the limit |x − y| → ∞ the integral term goto zero and we know that the operators A(x) and B(y) are
causally disconnected. This means that the matrices hn |A(0)| mi and hn |B(0)| mi commute. In some basis
these two matrices are diagonal
n |A(x)| n0 ∼ δnn0


(5.5)
So some operator can take you from one vacuum to another.
If there is only one vacuum, in the limit of large separation we should have

h0 |AB| 0i −→ h0 |A| 0i h0 |B| 0i (5.6)

24
Quantum Field Theory III Lecture 5

This is known as cluster decomposition. However if we have degenerate vacuua, we can have cluster
decomposition in the basis where A and B are diagonal, but not when we are working in other bases, as we
will have cross terms in the expectation hVac |A| Vaci hVac |B| Vaci. The fact that cluster decomposition
works in our universe (at least in our labs) shows that our vacuum is not a linear combination.
Now let’s consider field theory with a Lagrangian of a single scalar field

1 m2 2 λ 4
L= (∂ϕ)2 − V (ϕ), V (ϕ) = ϕ + ϕ (5.7)
2 2 4
If m2 > 0 then we have a classical minimum at ϕ = 0. However when m2 = −µ2 < 0 then there are two
minima at r
µ2
ϕ=± = ±ν (5.8)
λ
where ν is called the vacuum expectation value, or VEV. This is classical. Now in quantum case we
quantize it by introducing oscillators. In m2 > 0 case we have
Z  
ϕ(x) = d3 p a† eipx + ae−ipx (5.9)
p
and the Hamiltonian becomes energies of oscillators with ωp = p2 + m2 . Now if m2 < 0 we have
imaginary numbers and will run into problems if we still do things in the same way. We should now choose
ϕ = ν or ϕ = −ν and consider excitations from that VEV. Usually the former is chosen for less minus
signs. We define
ϕ = ν + χ, ∂µ ϕ = ∂µ χ (5.10)
and we have
λ 2 2 λ λ
V = ϕ − ν 2 − ν 4 = λν 2 χ2 + λνχ3 + χ4 + constant (5.11)
4 4 4
Now this is a perfectly fine potential for the field χ with mass

m2χ = 2λν 2 = 2µ2 (5.12)

Now the symmetry of ϕ4 interaction is lost, but again we can detect this because the coefficients of various
terms in the potential are related simply by an algebraic relation.
Now suppose we have √ rotational symmetry. Consider a complex field, or equivalently two real scalar
fields, ϕ = (ϕ1 + iϕ2 ) / 2. The Lagrangian is

L = |∂µ ϕ|2 − V (|ϕ|), V = m2 |ϕ|2 + λ |ϕ|4 (5.13)

This Lagrangian has rotational symmetry SO(2) for rotation on the ϕ1 and ϕ2 plane. Again similar to the
above, if m2 > 0 then we have a symmetric vacuum and ϕ1 , ϕ2 are particles with mass m, and ϕ, ϕ∗ are
particles with charge ±1. Now if m2 = −µ2 < 0 then we have a 3D Mexican hat which gives a degenerate
set of vacuua:
µ2
ϕ21 + ϕ22 = ν 2 = (5.14)
λ
To quantize, we need to choose any one vacuum among the above set. Physically they are all equivalent,
but mathematically for simplicity we choose hϕ1 i = ν and hϕ2 i = 0. We do the same thing again and write

ϕ1 (x) = ν + χ(x), ϕ2 (x) = ψ(x) (5.15)

25
Quantum Field Theory III Lecture 5

The Lagrangian now becomes


1 1
L= (∂µ χ)2 + (∂µ ψ)2 − V (χ, ψ) (5.16)
2 2
where
λ 2 2
V (χ, ψ) = χ + 2χν + ψ 2 + constant
4 (5.17)
 λ 2 2
= λν 2 χ2 + λν χ3 + χψ 2 + χ + ψ2
4

Note now that mχ = 2µ and mψ = 0. It is obvious from the potential because for tangential excitations
we don’t need extra energy to make them. The massless field ψ is called a Goldstone boson.
That was an example of SO(2). Suppose now we have SO(N ). The fields are ϕ1 , . . . , ϕN and the
potential is
m 2 X 2 λ X 2 2
V = ϕj + ϕj (5.18)
2 4
Now when m2 = −µ2 we have the vacuum expectation
P 2
ϕj = ν 2 and we choose hϕj i = νδjN and we have

ϕj (x) = ψj (x), ϕN (x) = ν + χ (5.19)

The potential now becomes


λ X 2 2
V = ψj + χ2 + 2χν = λν 2 χ2 + cubic and quartic terms (5.20)
4
Again the masses are √
mχ = 2µ, mψj = 0 (5.21)
Now there is manifest SO(N − 1) symmetry among the states, so we say that SO(N ) is spontaneously
broken to SO(N − 1). Note this is not accurate because the theory still have SO(N ) symmetry.
Now let’s say we have N − 1 Goldstone Bosons. The number of Goldstone bosons is just

# = dim SO(N ) − dim SO(N − 1) (5.22)

Now this is a general formula for any spontaneously broken symmetry. The statement is as follows. Assume
the potential V (ϕ1 , . . . ) is invariant under some global symmetry G with generators T a , with infinitesimal
transformation
δϕi = i (T a )ij ϕj (5.23)
We also assume the minimum of V is at hϕa i = νa and not all νa are zero. From the first assumption we
know the transformation of the potential
∂V ∂V
δV = δϕi = i (T a )ij ϕj = 0 (5.24)
∂ϕi ∂ϕi
And from the second assumption we know

∂V
=0 (5.25)
∂ϕi ϕi =νi

26
Quantum Field Theory III Lecture 5

Now with these two expressions we differentiate the first equation we get
∂2V ∂V
(T a )ij ϕj + (T a )ik = 0 (5.26)
∂ϕi ∂ϕk ∂ϕi
Now we set ϕi = νi in the above equation and we know immediately that
∂2V

a
= 0, Mki (T a )ij ν j = 0

(T )ij ϕj (5.27)
∂ϕi ∂ϕk ϕ=ν

We can think of the Mki as the mass matrix and the above equation tells us that there are zero eigenvalues.
The number of zeroes correspond to the number of directions that vector ν can be transformed into and
we can easily see
# = dim G − dim H (5.28)
where G is the original and H is the unbroken group. This proves our result in the settings of classical
field theory for scalar fields.
How do we break symmetry from G to H? Let’s consider SO(N ). Any vector can be written as

ϕ = (ν, 0, 0, . . . ) (5.29)

so if we have only one vector then we can only break the symmetry to SO(N − 1). Now suppose we have
ϕ and χ then if they are parallel we can break to SO(N − 1) but if not then we can break into SO(N − 2).
How can we decide if they are parallel? The terms we can put into the potential are
2 2
ϕ2 , χ2 , ϕ2 , χ2 , ϕ2 χ2 , (ϕ · χ)2 (5.30)

Now only the last term is sensitive to the relative direction of ϕ and χ. That term is what we need to
construct a potential which breaks the symmetry to SO(N − 2). Now how do we construct such a potential
and find the minimum? Consider SU (N ) and ϕ in the (1, 1) representation. Its transformation is

ϕij −→ U ik Ūj l ϕkl , ϕ −→ U ϕU † (5.31)

This is like a unitary transformation of bases, so we can find a basis where ϕ is diagonal with eigenvalues
ϕ1 , ϕ2 , . .P
. , ϕN . It is these eigenvalues that decides the symmetry. Note that they can’t all be equal because
we have ϕj = 0. Suppose
ϕ1 = ϕ2 = · · · = ϕN −1 6= ϕN (5.32)
then the group is broken into SU (N − 1) × U (1). If they are all unequal then it is broken into U (1)N −1 .
Now what potential do we have? We can have the traces of ϕ. Let’s consider
2
V = Tr ϕ2 + Tr ϕ3 + Tr ϕ2 + Tr ϕ4 (5.33)

Now the
Psecond and last
P term are sensitive to the differences of eigenvalues. Consider ϕ = νdiag(e1 , . . . , eN )
where ej = 0 and e2j = 1. Then the above potential becomes
X X
V = ν 2 + ν 4 + Aν 3 e3j + Bν 4 e4j (5.34)

We fix ν and minimize the potential with respect to ej ’s. We do this by adding Lagrange multipliers, and
require
α + 2βej + 3Ae2j + 4Be2j = 0 (5.35)

27
Quantum Field Theory III Lecture 5

There are at most three unequal eigenvalues. So there are only three possible broken group:

SU (a) ⊗ SU (b) ⊗ SU (c), SU (a) ⊗ SU (b) ⊗ U (1), SU (N − 2) ⊗ U (1) ⊗ U (1) (5.36)

But we are only halfway done because these are only extrema and we want minima. When there is no
cubic term, we have
   
N +1 N −1
SU (N − 1) ⊗ U (1), or SU (N/2) ⊗ SU (N/2) or SU ⊗ SU (5.37)
2 2

28
Quantum Field Theory III Lecture 6

6 Lecture 6
6.1 Proof of Goldstone Theorem in General
Last time we proved the Goldstone theorem only in the context of classical field theory and scalar fields,
so we want a more general proof. Assume we have some symmetry group G and we want to define an
order parameter f (x) which characterizes the breaking of the symmetry. We assume the symmetry is a
continuous one and it leads to a conserved current
∂µ J µ = 0 (6.1)
J 0 d3 x. We have
R
and we have a conserved charge Q =
[Q, φa (x)] = itab φb (x) (6.2)
where we don’t assume any property for the boson field φa which can even be a composite field. We want
to show that if hφa
i0 6= 0 for some a then there will be Goldstone bosons. We want to look at the vacuum
expectation value 0 [J λ (y), φa (x)] 0 and use the fact that ∂µ J µ = 0.

Let’s start. We want to put in some intermediate states


D E XD E D E
0 [J λ (y), φa (x)] 0 = 0 J λ (y) N hN |φa | 0i − h0 |φa | N i N J λ (y) 0

N
Xh D E D Ei
= eipN ·(x−y) 0 J λ (0) N hN |φa (0)| 0i − e−ipN ·(x−y) h0 |φa (0)| N i N J λ (0) 0

N
Z "
X D E
= d p eip·(x−y)
4
δ(pN − p) 0 J λ (0) N hN |φa (0)| 0i

N
#
X D E
−e−ip·(x−y) δ(pN − p) h0 |φa (0)| N i N J λ (0) 0

N
(6.3)
Now the term inside the summation has one Lorentz index, so it should be a Lorentz vector. The only
Lorentz vector here in the theory is just pλ , so we can write it as
X D E
δ(pN − p) 0 J λ (0) N hN |φa (0)| 0i = ipλ θ(p0 )ρa (p2 ) (6.4)

N

and similarly X D E
δ(pN − p) h0 |φa (0)| N i N J λ (0) 0 = ipλ θ(p0 )ρ̃a (p2 ) (6.5)

N
where ρ̃a is another function. If we choose x0 = y 0 and |x − y| > 0, then we know the commutator should
vanish because of causality. Setting p → −p in the second term, we find ρ̃a (p2 ) = −ρa (p2 ). Because the
function is Lorentz invariant, we know that this is true for all space-time configurations. We can now write
Z
D E ∂ h i
0 [J (y), φa (x)] 0 = − λ d4 p θ(p0 ) eip·(x−y) ρa (p2 ) + e−ip·(x−y) ρ̃a (p2 )
λ
∂y
Z
∂ h i
= − λ d4 p ρa (p2 )θ(p0 ) eip·(x−y) − e−ip·(x−y) (6.6)
∂y
Z Z
∂ h i
= − λ dµ2 ρa (µ2 ) d4 p δ(p2 − µ2 )θ(p0 ) eip·(x−y) − e−ip·(x−y)
∂y

29
Quantum Field Theory III Lecture 6

Now the inner integral looks just like a Green’s function in the free field, and all the information about
interactions is in ρa . We have
Z h i
∆(x − y) = d4 p δ(p2 − µ2 )θ(p0 ) eip·(x−y) − e−ip·(x−y) (6.7)

and we know that it is zero for spacelike separations and nonzero for timelike separations.
Now we use the fact that ∂λ J λ = 0. From the above expression we have
Z Z h i
0 = −y dµ2 ρa (µ2 ) d4 p δ(p2 − µ2 )θ(p0 ) eip·(x−y) − e−ip·(x−y)
Z (6.8)
= dµ2 µ2 ρa (µ2 )∆(x − y)

We know that for any time-like separation ∆(x − y) 6= 0, but the integral is zero, so the interand must be
zero
µ2 ρa (µ2 ) = 0 (6.9)
The obvious possibility is that ρa (µ2 ) = 0. But it can be nonzero when µ2 = 0, so we should have
ρa (µ2 ) ∼ δ(µ2 ).
Let’s go back to the expression for the commutator and set λ = 0 and x0 = y 0
Z Z

0
0 [J (y), φa (x)] 0 = 2i dµ2 ρa (µ2 ) d4 p θ(p0 )δ(p2 − µ2 )p0 eip·(x−y)

Z Z (6.10)
2 2 3 ip·(x−y)
= 2i dµ ρa (µ ) d p e

where in the second line we have integrated the delta function and apply the constraint on p. The integral
on p gives just another delta function. So we get
Z Z Z 
3

0 3 2 2 3
d y 0 [J (y), φa (x)] 0
=i d y dµ ρa (µ )(2π) δ(x − y)
y 0 =x0
Z (6.11)
3 2 2
= i(2π) dµ ρa (µ )

But then the left hand side is just

h0 |[Q, φa (x)]| 0i = itab h0 |φb (x)| 0i = i(2π)3 ca (6.12)

6 0 for some a, then we know that dµ2 ρa (µ2 ) 6= 0 and ρa should be a delta function
R
So when h0 |φa | 0i =

ρa (µ2 ) = ca δ(µ2 ) (6.13)




Now because N δ(pN − p) 0 J λ (0) N hN |φa (0)| 0i = ipλ θ(p0 )ρa (p2 ) where we defined the function ρa ,
P
we know that there will be some state |N i with p2N = 0, so there are massless states. We have roughly the
relation D E
0 J λ N hN |φa | 0i ∼ tab hφb i ∼ ν (6.14)

So this equation tells us that the massless particle couples to the conserved charge according to ν. If we
spontaneously break QED, then we get a massive photon and Goldstone boson is eaten by the gauge boson
to give it mass. This is superconductivity.

30
Quantum Field Theory III Lecture 6

6.2 Sigma Model


Let’s consider the sigma model which was introduced by Gell-mann and Levy. We consider the SU (2) ×
SU (2) = O(4) symmetry. There will be bosons π1 , π2 , π3 , σ and they form an SO(4) vector. The potential
is
V = V (π 2 + σ 2 ) (6.15)
so the potential is invariant under O(4). We write
 
σ + iπ3 iπ1 + π2
S = σI + iπ · τ = (6.16)
iπ1 − π2 σ − iπ3

We have
S † S = (σ 2 + π 2 )I, det S = (σ 2 + π 2 ) (6.17)
The term is transformed as
S −→ U SW † (6.18)
Written in this way, it is apparent to be a SU (2) × SU (2) transformation. Now we write the potential
term as
µ2 λ 2
V = − Tr S † S + Tr S † S (6.19)
4 4
And the kinetic energy
1
K = Tr ∂µ S † ∂ µ S (6.20)
4
We introduce the γ 5 matrix and construct left and right projections using it. We have

ψ̄(i∂/)ψ = ψ̄L (i∂/)ψL + ψ̄R (i∂/)ψR (6.21)

This is the fermion kinetic term. We have ψ̄L ψL = ψ̄R ψR = 0. We now introduce a fermion doublet N
with the additional symmetry

NL −→ U NL , NR −→ W NR =⇒ N̄L −→ N̄L U † , N̄R −→ N̄R W † (6.22)

The fermion Lagrangian is

Lferm = N̄L i∂/NL + N̄R i∂/NR − g(N̄L SNR + N̄R S † NL ) (6.23)

There can’t be any mass term like mN̄R NL because it will violate symmetry. But the interaction term is
symmetric. The interaction term can also be written as

− g σ N̄L NR + N̄R NL + iπ · N̄L τ NR − N̄R τ NL = −g N̄ σ + iπ · τ γ 5 N


    
(6.24)

Now there are two possibilities, if µ2 < 0 then there is manifest symmetry and we have

mπ = mσ = |µ| , mN = 0 (6.25)

The fermion is massless because we are forbidden to add a mass term. Now if µ2 > 0 we have spontaneous
symmetry breaking and
µ2
hπi2 + hσi2 = ν 2 = (6.26)
λ

31
Quantum Field Theory III Lecture 6

So the symmetry is broken SO(4) → SO(3). We need to choose a vacuum and it is convenient to choose
hπi = 0, hσi = ν (6.27)
Now the fermion interaction becomes
LI = −gν N̄ N + . . . (6.28)
and it seems the fermion becomes massive with mass mN = gν. There are obviously three Goldstone
bosons π i and this corresponds
√ to 6 − 3 which is the generators of SO(4) minus those of SO(3). The other
boson now has mass mσ = 2µ.
Originally we have the symmetry S → U SW † . Now what is the symmetry left? If we set U = W , then
we have S → U SU † remaining as the symmtry of S, and indeed this is the symmetry of the vacuum. So
we have
SU (2)L × SU (2)R −→ SU (2)diag (6.29)
We want to work out the Noether currents
2
 
µ 1 µ 1−γ
JL = N̄ γ τ N + bosonic terms (6.30)
2 2
and
1 + γ2
 
1
JµR = N̄ γ µ τ N + bosonic terms (6.31)
2 2
These are the original conserved currents. Now for diagonal SU (2) we have
1
Jµdiag = N̄ γ µ τ N + . . . (6.32)
2
But for Goldstone theorem we are interested in the algebra of broken symmetry, which are generated by
JR − JL . Let U = W † = e−iβ·τ /2 , so the change in the fields are
i i i
δNL = − β · τ NL , δNR = β · τ NR , δN = γ 5 β · τ N (6.33)
2 2 2
The transformation of S is
   
iβ · τ iβ · τ
δS = − S+S − = −iσβ · τ + β · π (6.34)
2 2
And by identifying terms we get
δσ = β · π, δπ = −βσ (6.35)
So the conserved current, by Noether theorem, is just
τ
Jµ = π∂µ σ − σ∂µ π − N̄ γ 5 N (6.36)
2
Now if we expand σ(x) = ν + s(x), then we have
Jµ = ν∂µ π + . . . (6.37)
So if we look at the matrix element appearing in the Goldstone theorem we have
D E D E
0 Jb (x) π (p) = ν 0 ∂ π + . . . π a (p) ∼ νpλ δ ab (1 + . . . )
λ a λ b
(6.38)

So Goldstone theorem works.

32
Quantum Field Theory III Lecture 7

7 Lecture 7
Let’s write down first the σ-model which we introduced last time
1 1 µ2 λ
L = (∂µ σ)2 + (∂µ π)2 + N̄ i∂/N + (σ 2 + π)2 − (σ 2 + π 2 )2 − g N̄ (σ + iπ · τ γ 5 )N (7.1)
2 2 2 4
Now the first thing we want to do is to renormalize. This theory is obviously renormalizable. However
what we want is that renormalization preserves symmetry. We want the counter terms like
1 δλ 2 1
Lct = − δm2 (σ 2 + π 2 ) − (σ + π 2 )2 + δZ[(∂σ)2 + (∂π)2 ] (7.2)
2 4 2
We only want these kinds of counter terms because these have the correct symmetry. It turns out that these
are sufficient. The way to see this is to derive the Ward identities connecting the unwanted divergences to
the above counter terms at the 1-loop level, and generalize to higher loops.
Let’s consider this for a pion. The potential is
λ λ 4
V = (σ 2 + π 2 )2 = π + 2π 2 s2 + s4 + λν(s3 + sπ 2 )

(7.3)
4 4
where ν is vacuum expectation value and σ = ν + s. The corrections to the pion mass is obtained from
looking at the corrections to the pion propagator. First let’s look at the Feynman diagrams from the sigma
model Lagrangian shown in figure 7.1.

1 1 2 2 1 1 1 1

1 1 1 1 s

−6iλ −2iλ −2iλν −2iλ

Figure 7.1: Diagrams corresponding to the sigma model

The corrections to the pion propagator looks like terms shown in figure 7.2, disregarding fermion
contributions.
It can be checked that the terms shown above give a vanishing contribution to the correction to the
pion mass, which is expected because the pion is massless as a Goldstone boson.
Let’s look at Goldstone boson scattering π1 π1 −→ π2 π2 . We have two diagrams at the tree level shown
in figure 7.3.
The tree level contribution to the pion scattering is
2λν 2 p2
 
− 2iλ 1 + 2 = −2iλ (7.4)
p − 2λν 2 p2 − 2λν 2
So when p2 → 0 the amplitude goes to zero. This is also a consequence of the fact that pion is a Goldstone
boson. Let’s write the quadruplet as (π1 , . . . , π4 ). This is a 4-vector and we can write it as a rotation from
a fixed vector
πa (x) = Ra4 w(x) (7.5)

33
Quantum Field Theory III Lecture 7

2, 3

= + +
2, 3

+ + + + ...

Figure 7.2: Corrections to the pion propagator

1 2 1 2

1 2 1 2
i
−2iλ (−2iνλ)2 2
p − 2µ2

Figure 7.3: Tree level scattering diagrams for π1 π1 → π2 π2

where w(x) = (0, 0, 0, w) is a fixed vector. Therefore we expect that


X
πa2 = π 2 + σ 2 = 2
Ra4 w2 = w2 (7.6)
a

The derivative squared is then

(∂πa )2 = [Ra4 ∂w + w∂Ra4 ]2 = (∂w)2 + w2 (∂Ra4 )2 + 2w∂w (Ra4 ∂Ra4 ) (7.7)

But the last term vanishes becase ∂(Ra4 Ra4 ) = 0. So the Lagrangian looks like
2
µ2

1 1 λ
L = (∂w)2 + w2 (∂Ra4 )2 − 2
w − (7.8)
2 2 4 λ

So all the interactions for the pions are hidden in the rotation matrix and it comes in with a derivative.
This means all the scattering amplitude are proportional to the momentum of the pions. If we define
ζi = πi /(π4 + ν) then we have
2ζi 1 − ζ2
Ri4 = , R 44 = (7.9)
1 + ζ2 1 + ζ2
Under the unbroken diagonal SU (2) symmetry we have ζ → α × ζ but under the broken symmetry we
have
δζ = (1 − ζ 2 ) + 2ζ( · ζ) (7.10)

34
Quantum Field Theory III Lecture 7

Suppose we take µ2 → ∞ and λ → ∞ but µ2 /λ fixed, then it is effectively making the Mexican hat
potential steeper, and it is harder and harder to get out of the minimum. In the limit we will have

σ 2 (x) + π 2 (x) −→ ν 2 (7.11)

so in the above parametrization this means w(x) becomes a constant function. Then S = σ +iπ ·τ becomes
ν times a unitary matrix. We can write it as

S = νeiπ(x)·τ /ν = νΣ(x) (7.12)

If we expand, then we will get


 
i 1
νΣ = ν 1 + π · τ − 2 (π · τ )2 + . . .
ν 2ν
(7.13)
1 2
= ν + iπ · τ − π + . . .

Now by symmetry the theory is invariant under Σ → U Σ(x)W † . We want a low energy effective action
and want to find invariant terms under the above symmetry that we can put into the action. The terms
without derivatives can include
Tr ΣΣ† = ν 2 (7.14)
But this is not very interesting because it is just a constant. Terms with derivatives can be
h i2
Tr (∂µ Σ)(∂ µ Σ† ), Tr (∂µ Σ)(∂ µ Σ† ) , Tr (∂µ ∂ν Σ)(∂ µ ∂ ν Σ† ), . . . (7.15)

We can then write


ν2 h i2
Leff = Tr (∂µ Σ)(∂ µ Σ† ) + c1 Tr (∂µ Σ)(∂ µ Σ† ) + . . . (7.16)
4
The only relevant term for low energies is the first term.
Let’s think what are the possible vacuum for this theory, where G is broken into H. One possible
vacuum is just φ0 = (0, 0, 0, ν). Then gφ0 is also a valid vacuum. However H is the subgroup where φ0
is invariant and we have hφ0 = φ0 . So the space of vacua is the same as the coset space G/H, which is
exactly the quotient group. Say if G = SO(4) and H = SO(3), then the space of vacua is the space of unit
4-vectors, which is just the sphere S 3 . The effect of our introducing σ is to introduce a coordinate of the
3-sphere, i.e. choosing a point of identity, and obtain a way of spatially disturbing the vauum.
In general for any G/H we can put coordinates on the manifold with φa . The Lagrangian will look like
1
L = gab (x)∂µ φa (x)∂ µ φb (x) + . . . (7.17)
2
The quantity gab can be considered as the metric over the manifold M = G/H. This is what people
nowadays call the nonlinear sigma model.
Let’s add a symmetry breaking term cσ to the original sigma model Lagrangian, where c is a small
positive number. Now the potential is
λ
V = (π + σ 2 − ν 2 )2 − cσ (7.18)
4

35
Quantum Field Theory III Lecture 7

The minimum of this potential is at π = 0 because we want to maximize cσ, and the minimum is at
∂V c
0= = λσ(σ 2 − ν 2 ) − c, σ≈ν+ =w (7.19)
∂σ 2λν 2
If we ignore fermions, the equation of motion for σ is just

σ = −λσ(σ 2 + π 2 − ν 2 ) − c (7.20)

and for π we have


π = −λπ(σ 2 + π 2 − ν 2 ) (7.21)
We can expand the above equation around σ = w +s(x) and get the pion masses due to symmetry breaking
 c 
π = −λπ + ... (7.22)
λν
So the mass is m2π ≈ c/ν = cg/mN . So the mass of the Goldstone boson is proportional to the symmetry
breaking. We can also work out the conserved current

∂µ Jµ = ∂µ (π∂ µ σ − σ∂ µ π) = −cπ = −δL (7.23)

So the current is also proportional to the symmetry breaking. This gives an insight to the real world where
many symmetries are broken to some degree. The currents are related to symmetry by Noether’s theorem,
but they are also dynamical things and they have something to do with the forces in the real world. These
two aspects seem to have little connection, as we can still have currents when they are not conserved.
Bearing these in mind, let’s consider weak interactions in low energy. 80 years ago Fermi wrote down
the β-decay
n −→ p + e− + ν̄e (7.24)
which nowadays people write as
d −→ u + e− + ν̄e (7.25)
The effective Lagrangian is
Leff = GΨ̄[. . . ]ΨΨ̄[. . . ]Ψ (7.26)
where the terms in the brakets could be

(1, γ 5 ) × (1, γ 5 ), (γ µ , γ µ γ 5 ) × (γ µ , γ µ γ 5 ), (σ µν , σ µν γ 5 ) × (σ µν , σ µν γ 5 ) (7.27)

Experiments show that the middle one is correct. And we have


GF
Leff = − √ Jµ† J µ (7.28)
2
where the current has a lepton piece and a hadron piece. The lepton piece is
λ
Jlep = ν̄e γ λ (1 − γ 5 )e + ν̄µ γ λ (1 − γ 5 )µ + . . . (7.29)

From this we could calculate the decay of a muon µ −→ e− + ν̄e + νµ . We now know there is a W boson
connecting the vertices. From this decay process we know experimentally GF = 1.16 × 10−5 (GeV)−2 . Now
how about the hadronic piece? We have the decay processes

π + −→ µ+ + νµ , κ+ −→ µ+ + νe (7.30)

36
Quantum Field Theory III Lecture 7

and these are semi-leptonic. We also have the non-leptonic decay

κ+ −→ π + + π 0 (7.31)

In Fermi theory we can use these processes to calculate the matrix elements of the hadronic current. In
modern days it turns out that the current is closely related to symmetry currents and these give great
insight to strong interactions as well.
Feynman and Gell-mann proposed that Jhadron = V µ + Aµ where V µ is the conserved vector current
and ∂µ V µ ≈ 0. For non-strange particles we can evaluate

0
p(k ) |V µ | n(k) = ūp (k 0 ) gν (q 2 )γ µ + fν (q 2 )iσ µν qν − ih(q 2 )q µ un (k)
 
(7.32)

From leptonic currents we already have GF and we can get numbers here directly. In fact we have
gν (q 2 = 0) = 0.97. Cabibbo pointed out that

Vµ = Vµ(∆s=0) cos θ + Vµ(∆s=1) sin θ (7.33)

where experiments show sin θ ≈ 0.23. This is now understood as the mixing between down and strange
quarks. In the limit where the first term dominates we have gν (0) → 1. To see this we expect that the
integral of the first current over space is just the electric charge. So people proposed that this corresponds
to a symmetry current. Then they proposed the isospins I1 , I2 , and I3 , where

I1 ∼ p̄γ µ n, I2 ∼ n̄γ µ p, I3 ∼ p̄γ µ p − ν̄γ µ n (7.34)

Under this we have B12 , (C12 )∗ and N12 become an I = 1 triplet and they decay to C12 which corresponds
to I = 0.

37
Quantum Field Theory III Lecture 8

8 Lecture 8
We continue our discussion on weak interactions
GF
Jint = √ J µ Jµ† (8.1)
2
We wrote the weak current into two parts

J = Jlep + Jhad , Jlep ∼ ēγ µ (1 − γ 5 )e + . . . , Jhad ∼ V µ + Aµ (8.2)

where V µ is a vector current. It is approximately conserved corresponding to isospin. In fact in modern


theory we know that
V µ ∼ ūγ µ d (8.3)
and the conservation of isospin corresponding to the same-mass approximation of u and d.
We consider the process π −→ µ+νµ . Because this is a semi-leptonic decay we can think of the leptonic
and hadronic currents separately. The hadronic part of the current will be corresponding to the matrix
element
h0 |Aµ (x)| π(k)i = −ik µ e−ik·x fπ (8.4)
If we take divergence of this then we get

∂µ hi = h0 |∂µ Aµ | πi = −k 2 e−ik·x fπ = −fπ m2π e−ik·x (8.5)

Again this is because


√ that the only Lorentz vector is the momentum k µ . From experiment we see that fπ
is 93 MeV or 93 2 MeV or 92 × 2 MeV, depending on which convention we use.
From the above expression we can think of the operator ∂µ Aµ as a pion field

∂µ Aµ ∼ fπ m2π π (8.6)

This is not a trivial statement. The actual requirement is that the above expression is true even when we
take |πi off the mass-shell.
Let’s consider the matrix element for β-decay

iσ µν
 

0 µ
0 µ 5 5 2 5 5 2 µ 5 5 2
p(k ) |A | n(k) = ū(k ) γ γ F1 (q ) + qν γ F2 (q ) + q γ F3 (q ) u(k) (8.7)
2m

The first term is the vector current and experiments show that F15 (q 2 → 0) = qA ∼ 1.24. If we operate
the divergence operator on this element then it is equivalent to multiplying qµ = kµ − kµ0 . So we have the
matrix element of the derivative

0
p(k ) |∂µ Aµ | n(k) = ū(k 0 ) (k
 0
/)γ 5 F15 (q 2 ) + q 2 γ 5 F35 (q 2 ) u(k)

/ −k (8.8)

This expression can be simplified if we observe that

ū(k 0 ) k
/0 − k
/ γ 5 u(k) = ū(k 0 ) k
/0 γ 5 + γ 5 k
/ u(k) = 2mūγ 5 u
 
(8.9)

So we get
p(k 0 ) |∂µ Aµ | n(k) = 2mF15 (q 2 ) + q 2 F35 (q 2 )


(8.10)

38
Quantum Field Theory III Lecture 8

If ∂µ Aµ = 0, i.e. this is a conserved current, we need the second term not to vanish in the limit q 2 → 0. So
we need F3 be the propagator of some massless particle. Then this process is essentially a neutron turns
into a proton while exchanging a pion with the axial current Aµ . The second term will now be −2gπN N fπ
and the element is zero. So we have
mN gA = fπ gπN N (8.11)
Now if the axial current is not conserved, we argued it should be ∂µ Aµ = fπ m2π , then we have

2mgA = fπ m2π hp |π| ni (8.12)

and if we understand this as a decay involving pions, and take the pion propagator to be massive 1/q 2 −m2π ,
then we get the same relation as in equation (8.11). This relation is called the Goldberger-Trieman relation.
It is experimentally tested to be good to 10%.
If we consider a matrix element of a time-ordered operator product

hT {Jµ (x)C(y)}i = θ(x0 − y 0 ) hJCi + θ(y 0 − x0 ) hCJi (8.13)

Now if we take derivative on the whole expression we not only get derivatives on J µ but also contribution
from θ functions, so

∂µ hT {J µ (x)C(y)}i = hT {∂µ J µ (x)C(y)}i + δ(x0 − y 0 ) J 0 (x), C(y)



 
(8.14)

The last commutator has a δ function as coefficient, and it vanishes unless at equal time, so we know it
must be at equal place and we can write it as hδCi δ (4) (x − y). If we plug in ∂µ Aµ ∼ fπ m2π then we can
get expressions about the matrix elements which are very useful. This is usually called current algebra.
We know in our study of sigma model for weak currents it has SU (2) × SU (2) symmetry. But as parity
symmetry is spontaneously broken, the symmetry group is broken down into SU (2)diag . Under this broken
symmetry we have π’s as Goldstone bosons. But π’s are not exactly massless, so we say SU (2) × SU (2) is
“almost” a symmetry, and π’s are “almost” massless. Now if we put in flavors, where ∆s = 0, ±1, then we
have almost SU (3)L × SU (3)R symmetry and it is broken into SU (3)diag which is flavor symmetry. This
is a true symmetry to about 10%. The “almost” Goldstone bosons now are π ± , π 0 , K ± , K 0 , K̄ 0 , η.
Let’s consider now the interaction for QCD
1 a aµν X
L = Fµν F + / − mf ) qf
q̄f (iD (8.15)
4
flavors

where f = u, d, s, c, b, t. The fields qf are SU (3)color triplets, and q̄f are in 3̄ representation. Now suppose
Nf of the mf are equal, then we have a SU (Nf ) symmetry in this theory. The SU (2) isospin symmetry
corresponds to mu ≈ md , and the SU (3) flavor symmetry corresponds to mu ≈ md ≈ ms . Suppose now
Nf of the masses are equal to zero, then we have the SU (Nf )L × SU (Nf )R chiral symmetry.
Now we have a conjecture. In a QCD with Nf massless quarks, the symmetry SU (Nf )L × SU (Nf )R is
spontaneously broken into SU (Nf ). If we ask why, it is somehow in the dynamics of QCD and we don’t
fully understand. So we have

h0 |q̄q| 0i = h0 |q̄L qR | 0i + h0 |q̄R qL | 0i =


6 0 (8.16)

As a result, the quarks acquire masses, just like that in the σ-model.

39
Quantum Field Theory III Lecture 8

Now we have two kinds of quark masses. The parameters mu , md ,. . . are current quark masses, and the
mass 1/3mp is the constituent quark masses. The difficulty is that we can’t isolate a quark and study them.
Obviously the effective masses of the quarks depend on our energy and length scale, so none of the above is
“the physcial mass”. Everything depends on what we are talking about. In low energy 1/3mp ≈ 300 MeV
is a good approximation. If we consider renormalization effects, then the scale ΛQCD when QCD becomes
strong is about ∼ 102 MeV, which is about the same scale as the above mass.
Now the masses mu,d . 10 MeV, so it is a very good approximation compared to the QCD scale that
these quarks are massless, so the SU (2)L × SU (2)R is a pretty good approximate symmetry. The mass of
strange quark ms ∼ 100-200 MeV, which is still relatively small, so the SU (3) × SU (3) symmetry is fairly
good. Now if we move on to charm quark mc ∼ 1.5 GeV and no matter what number we take for ΛQCD
this is large. So we don’t expect chiral SU (4) × SU (4) symmetry.
Now let’s try to write down an effective action for the chiral theory. We define
 
i a a
Σ(x) = exp π λ (8.17)
f

where a runs from 1 to 8. We put in the three lightest quarks and take their masses to zero, and we expect
the πa fields to give the right dynamics of the corresponding particles. The Lagrangian is then

f2  
L= Tr (∂µ Σ) γ µ Σ† + . . . (8.18)
4
Now we add some symmetry breaking, introducing the matrix
 
mu 0 0
M =  0 md 0 (8.19)
0 0 ms

and add to the Lagrangian the following term


 
∆L = −κTr M Σ + Σ† M (8.20)

We want to see what does this term look like. We expand the exponential in Σ, and we get
κ   κ
∆L = 2 Tr M (π a λa )(π b λb ) + · · · = 2 mu [(π · λ)2 ]11 + md [(π · λ)2 ]22 + ms [(π · λ)2 ]33

(8.21)
2f 2f

So we need to know what π · λ is. It is, plugging in our representation for SU (3) we get
 1 
√ π3 + √1 π8 π + K +
√  2 6
π a λa = 2  π− − √12 π3 + √16 π8 K0 

 q  (8.22)
K− K̄ 0 − 23 π8

We expect that π3 and π8 correspond to π 0 and η. If we work out the diagonal elements of the matrix
squared, we can get the expression for ∆L
From the above equation we can see that

m2π± = µ(mu + md ), m2K ± = µ(mu + ms ), m2K 0 = µ(md + ms ) (8.23)

40
Quantum Field Theory III Lecture 8

If we go to a particle data book, we can look up and get

mπ+ = 139.6 MeV, mπ0 = 135.0 MeV, mK + = 493.7 MeV, mK 0 = 497.7 MeV, , mη = 548.8 MeV
(8.24)
Now we can invert the equations above and get
1 2
mπ+ + m2K + − m2K 0

µmu = (8.25)
2
1 2
mπ+ − m2K + + m2K 0

µmd = (8.26)
2
1
−m2π+ + m2K + + m2K 0

µms = (8.27)
2
Now we don’t know the coefficient µ. But we can work out the quotients from the experimental masses
for those mesons.
ms mu
= 24.2, = 0.66 (8.28)
(mu + md )/2 md
There is still terms like aπ32 + 2bπ3 π8 + cπ82 . By diagonalizing the matrix we get the masses for π3 and
π8 . We will get
m2π0 = m2π+ , mη = 562 MeV (8.29)
which is very close to experiment data. Now if we take into consideration EM corrections, then the ratios
above should be corrected to
ms mu
∼ 20, ∼ 0.55 (8.30)
(mu + md )/2 md
From PDG data book the masses are

mu : 1.7 ∼ 3.3 MeV, md : 4.1 ∼ 5.8 MeV, ms : 101+29


−21 MeV (8.31)

Note that these are not directly measurable numbers, but they are extracted from experiment data using
some effective model of interaction, for example our interaction model with M Σ as above.

41
Quantum Field Theory III Lecture 9

9 Lecture 9
Last time we showed that if we just look at weak interactions and currents, strong interaction has very
good SU (2)×SU (2) chiral symmetry, and there is also pretty good approximate SU (3)×SU (3) symmetry.
But now we know the Lagrangian for QCD
X 1 a aµν
LQCD = / − mf )qf − Fµν
q̄f (iD F (9.1)
4
If we have some mf = 0 then there is symmetry

qfL −→ UL qfL , qfR −→ UR qfR (9.2)

However why is the symmetry SU (N ) instead of U (N )? One of the U (1) symmetry just corresponds to
UL = UR ∈ U (1) which just counts the number of u’s and d’s. The other U (1) symmetry UL = UR† seems
to be absent, so it must correspond to a Goldstone boson. So in SU (2) × SU (2) chiral theory we should
have a fourth Goldstone boson in addition to the three pions. Now the mass of π is about 135 ∼ 140 MeV,
and the next light pseudoscalar boson which has zero strangeness is η which has mass 549 MeV, which
is too large to be also a Goldstone bosons. Anything else would be heavier and not plausible. So this is
known as the U (1) problem. It is a long story to a resolution of this problem.
Let’s suppose we have a spontaneously broken gauge theory. First let’s consider gauge propagators,
and we use photon for example. The poton propagator is a series of
The first term is just
λq µ q ν
 
i µν
g + (9.3)
q2 q2
Now the second diagram without the legs is, in general

iΠµν = i(q 2 g µν − q µ q ν )Π(q 2 ) (9.4)

whose form is dictated by current conservation, thus Ward identity. So if we put in the legs we will get

λq α q µ λq ν q β qαqβ
      
i αµ 2 µν µ ν i νβ i αβ
− 2 g + (q g − q q )(iΠ) − 2 g + =− 2 g − 2 Π (9.5)
q q2 q q2 q q

Note that the second term in both of the propagators dot into the middle term to give zero, so they are
irrelevant. We can see the third term will be
qαqβ
 
i αβ
− 2 g − 2 Π2 (9.6)
q q

Now we can sum the series



qαqβ qαqβ
 X
i αβ
− 2 g − 2 Πn + (λ − 1)
q q q2
n=0 (9.7)
qαqβ
 
i 1
=− 2 g αβ − + ...
q q2 1 − Π(q 2 )

42
Quantum Field Theory III Lecture 9

Because this is QED the second term in the propagator does not matter, so it is just 1/q 2 . Any pole in
this expression must come from Π(q 2 ). Say if

µ2
Π(q 2 ) = + f (q 2 ) (9.8)
q2
then the pole will be at µ2 and it will be approximately mass squared.
Let’s consider broken U (1) gauge theory
1 2
L = − Fµν + |Dµ φ|2 − V (|ϕ|) (9.9)
4

and the potential has a minimum at hϕi = ν/ 2. Let’s choose hϕ1 i = ν and hϕ2 i = 0 then we can define
our field as
1
ϕ = √ [ν + χ(x) + iψ(x)] (9.10)
2
then from the potential V there will be no mass term for ψ, but a mass term for χ. The covariant derivative
will be √
2Dµ ϕ = ∂µ χ + i∂µ ψ − eAµ ψ + ieAµ (ν + χ) (9.11)
So the absolute value square is
1h i
|Dµ ϕ| = (∂µ χ − eAµ ψ)2 + (∂µ ψ + eνAµ + eνAµ χ)2 (9.12)
2
Let’s extract from this term the terms quadratic in the fields, and write down the Lagrangian for the
excitation fields ψ and χ
1 2 1 1 1
L = − Fµν + (∂µ χ)2 + (∂µ ψ)2 + e2 ν 2 A2µ + eνAµ ∂µ ψ − V (χ, ψ) + . . . (9.13)
4 2 2 2
Now apart from the usual terms we have two peculiar terms

ie2 ν 2 g µν , eνqµ (9.14)

Now let’s look at contributions to photon self-energy. The sum is


 e2 ν 2
i g µν q 2 − q µ q ν (9.15)
q2
So the correction is in the form of a pole with photon mass m2A = e2 ν 2 . Note here the mass comes from
two very peculiar terms and not from ordinary field mass term.
Now let’s write the field in another form
1
ϕ(x) = √ (ν + ρ(x))eiξ(x)/ν (9.16)
2
For small fields this is essentially the same fields we used before, but things differ in higher order terms.
Now the covariant derivative is
Dµ ϕ = ∂µ ϕ + ieAµ ϕ
(9.17)
    
1 ν+ρ ν+ρ
= √ ∂µ ρ + i∂µ ξ + ieνAµ eiξ/ν
2 ν ν

43
Quantum Field Theory III Lecture 9

So the absolute value square is


 2
2 1 2 1 2 ν+ρ
|Dµ ϕ| = (∂µ ρ) + (∂µ ξ + eνAµ )
2 2 ν
(9.18)
1 2
e ν 2 1  ρ 2
= (∂µ ρ)2 + (Aµ + ∂µ ξ)2 1 +
2 2 eν ν
Now we can redefine the field
1
Bµ = Aµ + ∂µ ξ (9.19)

This is essentially a gauge transformation, so Fµν is not changed. Now the Lagrangian will look like

e2 ν 2 µ 2ρ ρ2
 
1 2 1 2
L = − Fµν + (∂µ ρ) + B Bµ 1 + + 2 − V (ν + ρ) (9.20)
4 2 2 ν ν

Now the field ξ totally disappears, and we have a massive vector field Bµ with mass m2B = e2 ν 2 coupled to
a massive scalar field. If we count the degree of freedom we can see that the degree of freedom is 3 + 1 = 4.
Now if we had a massless vector and two massive scalars we will have degree of freedom 2 + 2 = 4, so the
degree of freedom matches. What happens here is that the Goldstone boson got eaten by the gauge boson.
What we have done above is essentially a gauge transformation. It is like taking a gauge condition,
but instead of requiring ∂µ Aµ = 0 we require ϕ to be real. This is how we eliminated ξ and make the
gauge boson massive. This is why we obtain a mass term which violates gauge invariance: because we have
chosen a gauge.
Now let’s generalize the Yang-Mills. ϕ` are real scalars with vacuum expectation value hϕ` i = ν` , and
we have a gauge group G with action on the ϕ’s

δϕ` = iβ a ta`m ϕm (9.21)

Note ϕ` forms a representation of the gauge group and it might be reducible. Similar to the above we
impose a gauge condition
ϕ` (ita`m νm ) = 0 (9.22)
This is to require that the field is orthogonal to the VEV. Now the covariant derivative is

Dµ ϕ` = ∂µ ϕ` − iAaµ ta`m ϕm = ∂µ ϕ0` − iAaµ ta`m (νm + ϕ0m ) (9.23)

And the square is


h i
|Dµ ϕ` |2 = (∂µ ϕ0` )2 + Aaµ Abµ −ita`m (νm + ϕ0m ) −itb`n (νn + ϕ0n ) − 2i(∂µ ϕ0` )ta`m (νm + ϕ0m )Aaµ

(9.24)

Now we use the gauge condition to simplify the above expression and we can extract the quadratic terms
1 2 1 1
L = − Fµν + (∂µ ϕ0` )2 − Aaµ Abµ (ta`m νm )(tb`n νn ) − V (ϕ0 ) + . . . (9.25)
4 2 2
If we write the mass term for the gauge field as 1/2 Aaµ Abµ (µ2 )ab then the mass matrix is

(µ2 )ab = −(ta`m νm )(tb`n νn ) (9.26)

44
Quantum Field Theory III Lecture 9

Suppose some linear combination of generators ca ta is unbroken, then ca ta`m νm = 0 and this corresponds
to a zero eigenvector of this mass matrix. Conversely if it has a zero eigenvector µ2ab ca cb = 0 then we know
that ca ta`m νm = 0 and this corresponds to an unbroken symmetry.
Again the counting of degree of freedom obviously works. Now we have a bunch of massless or massive
vector bosons and some massive scalars with no Goldstone bosons. This gauge choice is called unitarity
gauge. This is because all the particles appearing in this gauge are physical particles. If we work out the
gauge propagator in this gauge we will get
kµ kν
 
i µν
−g + (9.27)
k 2 − m2V m2V
Note that the second term is problematic and we need to cancel them to get renormalization to work in
loop diagrams. So this gauge is not a very good choice if we work on renormalizability. So let’s work in
Rξ gauge. For simplicity we work in U (1) theory.
1 1
φ = √ (φ1 + iφ2 ) = √ (ν + h(x) + iϕ(x)) (9.28)
2 2
Remember we do perturbation theory by add a gauge-fixing term to the Lagrangian and add a ghost
Lagrangian. The gauge-fixing term for Rξ gauge is
1
G = √ (∂µ Aµ − eνξϕ) (9.29)
ξ
and the ghost Lagrangian will be
 
 µ 2
 2 2 h
Lghost = c̄(δG)c = c̄ −∂µ ∂ − ξe ν(ν + h) c = c̄ −∂ − ξmA (1 + ) c (9.30)
ν
The quadratic terms in the full Lagrangian will be
1 1 1 1 1 2 1 1
L = (∂µ h)2 − m2h h2 + (∂µ ϕ)2 − m2A ξϕ2 − Fµν − (∂µ Aµ )2 + m2A A2µ
2 2 2 2 4 2ξ 2 (9.31)
+ mA Aµ ∂ µ ϕ + eνϕ∂µ Aµ + c̄ −∂ 2 − ξm2A c
 

Now the two terms in the second line will add up to be a total derivative of the gauge fixing term and we
can throw them away. The propagator for the vector boson is
 
i kµ kν
− 2 gµν − 2 (1 − ξ) (9.32)
k − m2A k − ξm2A
The h propagator is just a usual massive propagator, so as ϕ and c. Now when we do perturbation theory
we have three propagators that depend on ξ, namely Aµ , ϕ, c. However ξ is an arbitrary variable and
we expect that all the ξ factors should cancel in our loop calculations, and this is an advantage in our
calculation. This is called renormalizable gauge, because renormalization is easy in this gauge.
If we take the limit ξ → ∞, the vector propagator goes to
kµ kν
 
i µν
− 2 g − (9.33)
k − m2A m2A
The propagators for ϕ and c go to zero, because they are unphysical particles. However the coupling
between c and h goes to infinity. And we have a loop of ghosts and external h legs then the diagram
will give a finite contribution. So this is like unitarity gauge with some funny interactions, i.e. interaction
between arbitrary number of h bosons. These show up in the unitarity gauge if we calculate the path
integral and we will get some δ 4 (0) terms which correspond to these interactions.

45
Quantum Field Theory III Lecture 10

10 Lecture 10
Let’s look at the process π 0 → γ + γ. Before worrying about the composition of pions, the effective
Lagrangian should be
Leff = gπ 0 µναβ F µν F αβ (10.1)
where the coupling constant g has the dimension of inverse mass. The decay rate is just given by the vertex
diagram
m3
Γ = π g2 (10.2)
π
Let’s guess the form of g from the 1-loop correction.The diagram should be like shown in figure 10.1.

γ
f
π
f

f γ

Figure 10.1: 1-loop Correction to Pion Decay

We would guess that


e2 1
g≈ (10.3)
(4π)2 fπ
The e’s are from the photon vertices and the (4π)2 is from the fermion loop. The remaining part should
have dimension of inverse mass, so we guess it would be 1/fπ and from this guess we will get some
Γ ≈ 4 × 1016 sec−1 . Now we recall that π’s are Goldstone bosons, and their interaction with the gauge field
is through derivative terms, and the effective Lagrangian is proportional to the derivative of π

m2π
Leff ∼ ∂ 2 π ∼ (10.4)
m2N

so the decay rate should be Γ ∼ m7π ≈ 2 × 1013 sec−1 because of the above suppression factor. But
the experimental rate is 1.19 × 1016 sec−1 , which is closer to the first one. This is in contrast with our
understanding of Goldstone bosons, so what is going on?
Let’s consider having a single fermion for now and the Lagrangian is

L = ψ̄(iD
/ − m)ψ, Dµ = ∂µ + ieAµ (10.5)

The currents are


J µ = ψ̄γ µ ψ, J5µ = ψ̄γ µ γ 5 ψ (10.6)
where J5µ is the axial current. The equation of motion is

/ − m)ψ = iγ µ ∂µ ψ − eAµ ψ − mψ = 0,
(iD γ µ ∂µ ψ = −ieAµ ψ − imψ (10.7)

The adjoint of the above equation is


∂µ ψ̄γ µ = ieAµ ψ̄ + imψ̄ (10.8)

46
Quantum Field Theory III Lecture 10

From this we can work out the divergence of the current

∂µ J µ = (∂µ ψ̄γ µ )ψ + ψ̄(γ µ ∂µ ψ)


(10.9)
= ieAµ ψ̄γ µ ψ + imψ̄ψ − ieAµ ψ̄γ µ ψ − imψ̄ψ = 0

which is what we expect due to global gauge symmetry. The other divergence is

∂µ J5µ = (∂µ ψ̄γ µ γ 5 )ψ − ψ̄γ 5 (γ µ ∂µ ψ)


= ieAµ ψ̄γ µ γ 5 ψ + ieAµ ψ̄γ 5 γ µ ψ + 2imψ̄γ 5 ψ (10.10)
= 2imψ̄γ 5 ψ

So if the fermion is massless then the divergence of this current also vanishes.
Now let’s say we have some operator j = ψ̄Γψ where Γ is some operator. The vacuum expectation
value of this quantity is



hji = ψ̄Γψ = −Tr ψ ψ̄Γ = −Tr ψ ψ̄Γ (10.11)
where the contraction is just the propagator. Now under interaction the free propagator becomes
1 1

i∂/ − m i∂/ − m − eA
/
(10.12)
1 1 1
= + (eA
/) + ...
i∂/ − m i∂/ − m i∂/ − m

So the graphic representation of the expectation value of the current is shown in figure 10.2, where each
solid line is a free fermion propagator.

A
A
A
Γ× + Γ× + Γ× + Γ× A + ...
A
A

Figure 10.2: Graphic Representation of hji

Now we replace Γ by qµ γ µ γ 5 . Then the above integral should just give us the Fourier transform of the
divergence of the axial current. We want to show that this is indeed as we derived in classical case that it
vanishes for zero fermion mass. We will show that the sum of the above diagrammatic series is just (−2m)
times the same expression with Γ = γ 5 . Diagrammatically the equality is:

/qγ 5 × = −2mγ 5 × (10.13)

where the shaded area means any number of external photon lines. Now let’s prove this equality using
diagrammatical methods.

47
Quantum Field Theory III Lecture 10

The term with zero photon legs vanish because the term is proportional to /q and conservation of
momentum requires q = 0. Also the trace of γ 5 with only one or two γ’s is zero. We need more γ matrices
inside the trace to get a nonzero result. As an illustration we just prove the case when there are two photon
legs, and the other cases are similar, including 1 leg and more legs. The diagram is like shown in figure
10.3, and there are two possible momentum configurations

p+q k1 p + q + k1 k2

/qγ 5 × p + q + k1 + /qγ 5 × p

p k2 p + k1 k1

Figure 10.3: Two-leg Diagram

We introduce a quantity S −1 (p) = 1/(p


/ − m), then we can write /qγ 5 as

/qγ 5 = (p
/ + /q − m)γ 5 − (p
/ − m)γ 5
(10.14)
= γ 5 S(p + q) + S(p)γ 5 − 2mγ 5

where p could be any momentum. So from the first diagram we have

S −1 (p)γ 5 k
/1 S −1 (p + q + k1 )k /1 S −1 (p + q + k1 )k
/2 + γ 5 S(p + q)k /2
(10.15)
−2mS −1 (p)γ 5 S −1 (p + q)k
/1 S −1 (p + q + k1 )k
/2

and from the second diagram we have

S −1 (p)k
/1 S −1 (p + k1 )γ 5 k
/2 + S −1 (p)k
/1 γ 5 S −1 (p + q + k1 )k
/2
(10.16)
−2mS −1 (p)k
/1 S −1 (p + k1 )γ 5 S −1 (p + q + k1 )k
/2

The second line on the two expressions just give the −2m term we want. The second term in the second
diagram cancels with the first term in the first diagram. And we can make the other term cancel if we
note that we are taking trace and integration, and exploit that to move the γ 5 to the end and shift the
integration variable from p to p + q. So we have proved that the equality (10.13) holds.
Let’s check our pion calculation above directly. We do it in 2D first for simplicity. We take

γ 0 = σy , γ 1 = iσx , γ 5 = γ 0 γ 1 = σz (10.17)
ψ+ 
and the two component spinor can be written as ψ = ψ− . Using these components we can write the
kinetic term in the Lagrangian as

/ψ = iψ+ (D0 + D1 )ψ+ + iψ− (D0 − D1 )ψ−


ψ̄iD (10.18)

So if the fermion is massless then we have two decoupled fields. If A = 0 then

(∂0 + ∂1 )ψ+ = 0, ψ+ = ψ+ (t − x) (10.19)

48
Quantum Field Theory III Lecture 10

k+q

/qγ 5 × k

Figure 10.4: The 1-Leg Loop Diagram

so ψ+ are the right moving modes and ψ− are left moving.


Now let’s do the loop integrals. The zero leg loop will vanish due to our previous arguments. The one
leg loop diagram should be proportional to −2m and in the massless limit it should also vanish. But we
can actually evaluate the loop integral explicitly. The diagram is shown in figure
The photon vertex gives a factor of γ ν eAν . We can get a feeling of this diagram by looking at various
terms. In 2D we have
γ µ γ 5 = −µν γ ν (10.20)
So the diagram can be converted to µα iΠαν where Παν is the 1-loop correction to the photon propagator.
Now by Lorentz invariance we know that

iΠαν = P (q 2 )g αν + Q(q 2 )q α q ν (10.21)

and by gauge invariance, i.e. Ward identity we have

qα Παν = 0, P = −q 2 Q (10.22)

Therefore the whole diagram is just

q µ µν Πνα = µν qµ [P g να + Qq ν q α ] ∝ P (q 2 ) (10.23)

In the massless fermion limit we expect the diagram to vanish for any external q, so we will conclude
P (q 2 ) = 0 identically, therefore Q = 0. But we can regularize and calculate Q explicitly and it is actually
Q = −(e2 /π)/q 2 , not zero. There must be something wrong in our previous discussion.
This is because we cheated in our diagrammatic derivation above. We shifted the integration variable
p → p − q in one of the loops, but we can’t do that if the loop integral is divergent. So let’s do it the proper
way now and use some regularization, and evaluate the diagram in figure 10.4. Dimensional regularization
is not very nice here because γ 5 is not defined at dimension d = 2 − . Instead we use Pauli-Villars where
we introduce an extra fermion with a minus sign in kinetic term and mass M and later take M → ∞.

L → L − Ψ(iD
/ − M )Ψ (10.24)

Now we have two fermions which enter the loop and they have different masses and different sign. Previous
argument would lead to

h∂µ J5µ i = 2im hJ5 im − 2iM hJ5 iM 6= 0 in m → 0 limit (10.25)

49
Quantum Field Theory III Lecture 10

The M diagram gives

d2 p
Z  
5 i ν i
(−1) Tr γ γ
(2π)2 /p − M /p + /q − M
d2 p
Z
1 1 5 ν

=− Tr γ (p
/ + M )γ (p
/ + /
q + M )
(2π)2 p2 − M 2 (p + q)2 − M 2
(10.26)
d2 p
Z
1 1
Tr γ 5 (−p / + M )γ ν

=− 2 2 2 2 2
/ + M − /q)(p
(2π) p − M (p + q) − M
d2 p
Z
να 1
= − 2 qα M
(2π) (p − M ) [(p + q)2 − M 2 ]
2 2 2

In the limit of M → ∞ we don’t need to use Feynman parameters and can just do a Wick rotation and
evaluate the integral and get
i e
Diagram = −2να qα M 2
(2iM ) = να qα Aν (q) (10.27)
4πM π
Fourier transform this back we get
e αβ
Diagram =  Fαβ (10.28)

So in 2D we should get
e αβ
∂µ J5µ =  Fαβ (10.29)

This is a gauge invariant result, as expected.
The problem with our previous “proof” of equality 10.13 is that when we have a divergent loop integral
of the sort Z Z
dp f (p, . . . ) = dp f (p + λq, . . . ) + constant (10.30)

So such a diagram is ambiguous up to a constant like Ggµν + Hqµ qν . Shifting the integration variable will
give an extra constant term.
Note also that higher leg diagrams will not give contribution to the anomaly computed above. This
is because the fermion loop is massless, and there is no divergence in those diagrams. Only the divergent
diagram contributes to the anomaly. In 4D it is due to the triangular diagram shown in figure 10.1.

50
Quantum Field Theory III Lecture 11

11 Lecture 11
11.1 Anomalies Continued
Today we introduce two other ways to compute the anomaly. Let’s take n dimensional space-time. The
spatial integral of the divergence of a current is
Z Z
n−1
x ∂µ J = dx ∂0 J 0 + ∂i J i = ∂0 Q
µ
 
d (11.1)

We have the number operators for left and right hand fermions
Z Z
† †
NR = ψ+ ψ+ , NL = ψ− ψ− (11.2)

and the charge is just Q = NR − NL , and we computed the anomaly


Z
d e
(NR − NL ) = dn−1 xαβ Fαβ (11.3)
dt 2π
If we choose the gauge A0 = 0, then it becomes
Z
d e
(NR − NL ) = ∂t dn−1 xA1 (11.4)
dt π
Let’s define the Wilson loop. We know if we have a loop integral of a gauge potential, and we do a
gauge transformation A → A − ∂α/e, then the integral
I I
dl · A → dl · A + α(L) − α(0) (11.5)
C C

So if the loop is closed this is invariant. However if the integral sits in the exponential, then this is
measurable, so we have the Aharonov-Bohm effect. Suppose we have 1 dimensional space with length L
and periodic boundary condition, then we can think of the integral as
 Z L   Z L 
exp −ie dl · A → exp −ie dl · A − iα(L) + iα(0) (11.6)
0 0

Because the difference of α(L) and α(0) sits in the exponential, it can be a multiple of 2π and Rthe single-
valuedness of the exponential will not produce a difference. The same happens if we increase dxA1 by
2π/e, and it should not give a difference again because of the exponential. But now the time derivative of
NR − NL will change and we do have a physical difference. So what happened?
Suppose A1 is a constant and the fermion mass is zero, the Hamiltonian is
† †
H = −iψ+ (∂1 + ieA1 )ψ+ − iψ− (∂1 + ieA1 )ψ− (11.7)

The energy eigenfunctions are



ψ = eikn x , kn =
n (11.8)
L
and the energies for right handed and left handed fermions are
 
2π 2π
ψ+ : En = n + eA1 ψ− : En = − n + eA1 (11.9)
L L

51
Quantum Field Theory III Lecture 11

R
So if we change dxA1 by 2π/e it is the same as shifting the energy levels of ψ+ up by one and ψ− down
by one. If we use the Dirac sea picture with all the negative energy states occupied, then this is effectively
creating a right handed particle and leaving a left handed hole. This changes the overall chirality of the
vacuum.
The axial current is ψ̄(x)γ µ γ 5 ψ(x). Note that this evaluates two fields at the same point, but physically
this produces some problem. We want to separate them by an infinitesimal amount . The first trial is
  µ 5  
ψ̄ x + γ γ ψ x− (11.10)
2 2
But this is not gauge invariant. The way we fix it is
   µ 5 −ie R dl·A  
ψ̄ x + γ γ e ψ x− (11.11)
2 2
where the integration is from x − /2 to x + /2. In this way the divergence of the axial current is
h   µ 5   µ 5 i
∂µ J5µ = lim ieAµ x + ψ̄γ γ ψ − ieAµ x − ψ̄γ γ ψ − ie∂µ (ν Aν )ψ̄γ µ γ 5 ψ + O(2 )
→0 2 2
= lim ψ̄γ µ γ 5 ψ [ie (ν ∂ν Aµ + ν ∂µ Aν )] (11.12)
= lim ieν Fµν ψ̄γ µ γ 5 ψ

This term is first order in , but we will see that ψ̄γ µ γ 5 ψ is singular and is order 1/. We know the
expectation value

µ 5
ψ̄γ γ ψ = − Tr(ψ ψ̄γ µ γ 5 )


(11.13)
The contraction is the propagator and we Wick rotate it to get
d2 k −ik·(y−z) ik
Z
/
ψ(y)ψ̄(z) = 2
e
(2π) k2
Z 2 (11.14)
W ick d kE 1 −ik·(y−z)
−−−→ −γ µ ∂µ 2 e
(2π)2 kE

The integral is the propagator in 2D and we claim it is ∆ = ln(y − z)2 . By matching against the box
operator we can get
i i (y − z)µ
∆= ln(y − z)2 , ∂µ ∆ = (11.15)
4π 2π (y − z)2
Plug this in, we get
i  α 
∂µ J5µ = lim − 2 Tr γ α γ µ γ 5 (ieν Fνµ )

→0 2π  (11.16)
e  α ν 
= lim Fνµ Tr(γ α γ µ γ 5 )
2π 2
Now the trace is just 2εαµ , and in the integration we can replace the α ν by gαν 2 /2, so in the end we get
e
h∂µ J5µ i = Fνµ νµ (11.17)

This is exactly the same result we derived last time using Pauli-Villars on the loop diagram. However in
4D this point splitting is harder because the anomaly comes from the triangular diagram and we need to
separate three points.

52
Quantum Field Theory III Lecture 11

So let’s do yet another method of calculatingR the anomaly.


Q We use path integrals. Remember in the
path integral we have the fermion field measure [dψ][dψ̄] ∼ x dψ(x)ψ̄(x). Let’s expand these functions
by eigenfunctions of the Dirac operator

iD
/φm = λm φm , −iDµ φ̂m γ µ = λm φ̂m (11.18)

and we have X X
ψ(x) = am φm (x), ψ̄ = âm φ̂m (x) (11.19)
Q
And the integration measure becomes m dam dâm . Under a gauge transformation we have
Z
ψ → (1 + iα(x)γ 5 + . . . )ψ, am → dxφ†m (1 + iαγ 5 )φn an = (δmn + cmn )an (11.20)

where the coefficients Z


cmn = i dx α(x)φ†m γ 5 φn (11.21)

Plug this into the integration measure, we essentially have a Jacobian factor

[dψ][dψ̄] → J −2 [dψ][dψ̄], J = det (I + C) (11.22)

Now because detM = exp(tr ln M ), so we have

det (I + C) = exp (tr ln(I + C))


(11.23)
≈ exp trC

If we also expand J −2 then the change in the action is actually −2tr(C). What is this trace? It is
XZ
tr C = i dx αφ†n γ 5 φn (11.24)
n

The trace of γ 5 is zero, but this is a divergent sum. So we need to regularize it. Let’s do it this way
XZ
/)2 /M 2
tr C = i dx φ†n γ 5 e(iD φn
n
2 2
X
=i φ†n (x)γ 5 eλn /M φn (x) (11.25)
n
Z D E
continuum /)2 /M 2
−−−−−−−→ dx tr x γ 5 e(−iD x

/)2
and in the end we take M → ∞. Now we evaluate (iD

/)2 = −γ µ Dµ γ ν Dν
(iD
1 1
= − {γ µ , γ ν } Dµ Dν − [γ µ , γ ν ] Dµ Dν (11.26)
2 2
2 e µν
= −D − σ Fµν
2

53
Quantum Field Theory III Lecture 11

So the integrand we need to evaluate is

e2 (σ · F )2
 
D E D
−D2 /M 2 −(e/2)σ·F/M 2 5
E
−∂ 2 /M 2 5 eσ·F
x e e γ x = x e x tr γ 1 − + + ... (11.27)
2 M2 8 M4

The expectation value can be evaluated in Fourier space, so we get


d
dd k k2 /M 2 dd k −k2 /M 2
E Z Z 
D 2 2 M
x e−∂ /M x = e =i e =i √ (11.28)

(2π)d (2π)d 2 π

Now if d = 2, the trace of γ 5 vanishes, the M 2 in the above expression cancels the 1/M 2 term in the series
above and we have
e
− 2tr C = − µν Fµν (11.29)

and this gives our anomaly. Now in 4D we can carry out the same procedure, the trace of γ 5 with 2 γ
matrices is zero, which prevents the M 2 term from blowing up. The 1/M 4 term will give us the anomaly
which is
e2 αβµν
∂µ J5µ = −  Fαβ Fµν (11.30)
16π 2
This applies to all the anomalies of even dimensions and in odd dimensions there is no anomaly. We could
have taken other regulators than e−(D /)2 /M 2 . All we need is that f (0) = 1 and f (∞) = 0, so we could have

taken the regulator to be


1 M2
2 2
= (11.31)
1 + (D /) /M /) + M 2
(D 2

Now let’s go back to the π mesons. The change in the effective Lagrangian when we do a transformation
is what gives the divergence of the axial current. We need to add to the effective Lagrangian which cancels
this change, such that  2 
e
δ (∆Leff ) = −C εαβµν Fµν Fαβ (11.32)
16π 2
Now the field tensor is invariant under gauge transformation, and recall the sigma model we have δπ =
−σ = −fπ , so our change in the effective Lagrangian should be

Ce2 π
∆Leff = εF F = gπεF F (11.33)
16π 2 fπ
Let’s evaluate the loop correction due to this term. We need to add up all the quark loop contributions.
The quarks will give factors
 2    2  
2 1 1 1
u loop:Nc , d loop:Nc − − (11.34)
3 2 3 2

So in the end we get


 2
Nc
Γ= 1.11 × 1016 sec−1 (11.35)
3
and at Nc = 3 it is very close to the experiment.

54
Quantum Field Theory III Lecture 12

12 Lecture 12
Let’s generalize our anomaly discussion. We work in 4D and consider a general time-ordered product
hT (j1 j2 j3 )i. There will be two triangular diagram corresponding to two orientations of the fermion loop.
We can write a current in left and right handed parts
5 5
   
L µ 1−γ R µ 1+γ
ja = ψ̄i γ (Ta )ij ψj , ja = ψ̄i γ (Ta )ij ψj (12.1)
2 2

Putting this into the graphs, we get a factor of tr T a T b T c + T a T c T b from the complete fermion loop.


From the fermion propagators we will also have a factor of


5 1 ± γ5 5
     
µ 1±γ ν λ 1±γ
γ (k
/ + m)γ (p
/ + m)γ (q
/ + m) (12.2)
2 2 2

We know the anomaly comes from the term that do not depend on mass, because there is anomaly even
for massless fermions. But the trace of that term vanishes unless all are (1 + γ 5 ) or all are (1 − γ 5 ).
Let’s define  n o  n o
Aabc
L = tr TL
a
T b
L L, T c
, Aabc
R = tr TR
a
T b
R R, T c
(12.3)

The anomaly will be proportional to the factor Aabc abc


L − AR . Remember our anomaly came from the axial
current, and it contributes to the pion decay. It doesn’t matter if the current is not conserved. Now if we
consider gauge current and want to prove renormalizability, we need Ward identity which says qµ Πµν = 0.
But the following diagram will give us an anomaly, which will give us problems in renormalization. In
order to prove that all the triangular diagrams cancel, we need to see when is the above quantity AL − AR
zero.
Let’s consider a real representation D(g) of the gauge group. Then the generators will be imaginary
and hermitian, and we have T T = −T . We have

tr (Ta {Tb , Tc }) = tr TcT TbT TaT + TbT TcT TaT = −tr (Ta {Tc , Tb }) = 0

(12.4)

So we have the anomaly A = 0 for any real representation. Suppose the representation is pseudoreal, but
then it is just equivalent to a real representation by a similarity transformation, and the above result is still
true. So we only have anomalies when we have a complex representation. The groups that have complex
representations are SU (N ), SO(4N + 2), E6 , U (1). The generators of SO(N ) are Trs which transform like
2-tensors, so the trace term tr (Trs {Ttu , Tvw }) will transform as a 6-tensor. But it is antisymmetric under
r ↔ s or t ↔ u or v ↔ w. And it must be symmetric under (rs) ↔ (tu) or (rs) ↔ (vw) or (vw) ↔ (tu). We
need something like rstuvw , which only exists in SO(6). But SO(6) = SU (4). The exceptional group E6 is
anomaly-free. So we only need to consider groups SU (N ) or U (1) and they are just groups arise in particle
physics. For SU (N ) the generators are closed under anticommutation, so we can use anticommutation to
characterize the group n o
T a , T b = idabc T c (12.5)

and the constants dabc are totally antisymmetric. We can take trace and get
 n o
2idabd = tr T d T a , T b (12.6)

55
Quantum Field Theory III Lecture 12

and we can evaluate, for example d888 in SU (3), and it is proportional to tr(λ38 ) and it is nonzero. So not
all these constants are zero.
In an anomaly-free theory we need Aabc abc
L = AR . One way to have this is that we have a “vector
like” theory, therefore jL = jR . Or we can just work on anomaly-free groups, or only real and pseudoreal
representations. Or we could have “lucky” cancellation in the theory so that there is no anomaly.
Let’s look at the standard model with gauge group SU (3) × SU (2) × U (1), with charge Q = I3 + Y /2.
We can look at the fields listed in table 3. These are all the particles in one family.

SU (3) SU (2) Y Q
(u, d)L 3 2 1/3 (2/3, −1/3)
uR 3 1 4/3 2/3
dR 3 1 -2/3 −1/3
(ν, e)L 1 2 -1 (0, −1)
eR 1 1 -2 −1
νR 1 1 0 0

Table 3: Representations in Standard Model

This table looks completely random. Let’s see what kind of triangles we can have. The triangle with
three SU (3) currents will be vector-like and is anomaly-free. The triangle with two SU (3) vertices is
possibly problematic. The triangle with only one SU (3) vertex will vanish, because the SU (3) generators
all have vanishing trace.
The triangle with one vertex being SU (2) will also vanish due to the same reasoning as above. So the
possible anomalies are
1. SU (3) × SU (3) × Y
In this case the diagram will be proportional to tr (Y {Ti , Tj }) ∼ δij tr Y , which only lies in the quark
sector. So here we should have
   
X 1 1 X 4 2
AL ∼ Y =3 + , AR ∼ Y =3 − (12.7)
3 3 3 3
left q right q

So we have AL − AR = 0 and there is no anomaly.


2. SU (2) × SU (2) × Y
In this case the diagram is proportional to tr (Y {τa /2, τb /2}) = δab tr Y /2 which lies only in the
doublets. So we have
 
X 1 1
AR = 0, AL ∼ Y =3 + + (−1 − 1) = 0 (12.8)
3 3
left particles

So there is again no anomaly.


3. Y × Y × Y
This will be just the sum of Y 3 . So we have
 3  3  3
1 3 16 4 2 64 8 72 16
AL = 6 + 2(−1) = − , AR = 3 +3 − + (−2)3 = − − =− (12.9)
3 9 3 3 9 9 9 9
So again AL − AR = 0.

56
Quantum Field Theory III Lecture 12

This tells us phenomenologically that quarks must come as doublets, and we need cancellation between
quarks and leptons. So if we have more families of quarks we need more families of leptons and in particular
neutrinos. If we consider the decay of Z bosons it goes to all quark-antiquark pairs or lepton-antilepton
pairs. We can’t measure the rate into neutrino pairs, but we can first calculate the total decay width which
is verified by theory. We can calculate the total decay width and measure the total width and the part
that is missing must come from the neutrinos. The result actually fits 3 neutrino species. So if we have an
additional family of neutrinos then its mass must exceed half of the Z mass so that it won’t show up in
the above described process. This also puts a strong limit on anything else that couple to Z.
Suppose we have a theory with gauge group G = SU (3)color × g where g is the electro-weak group
which contains SU (2) × U (1) as a subgroup, and is simple. Because particles fall into representations of
g, suppose q and leptons fall into different representations of g, but we know that g is simple, and the
generators are traceless including the charge, but we know that the lepton charges do not add up to zero,
neither do the quarks, so we either need extra particles or they actually are in the same representation.
Now let’s be ambitious and try to construct a grand unification theory with gauge group G which
contains SU (3) × SU (2) × U (1) as a subgroup. We want G to be a simple group to have only one coupling
constant. The representation of G should be the direct product of H which are subgroups and we want
all known particles to fit into H multiplets without adding many more new particles. We want this to be
able to explain all the gauge theories, and this will automatically give us quantization of electric charges.
But if we only have a simple G then we should have a single coupling constant which we know is wrong
experimentally. But it could happen if we take into consideration the running of couplings and say the
theories unite at a sufficiently high energy scale.
Let’s write down the standard model fields. We have uL which annihilate left handed quarks and create
right handed antiquarks. We also have uR which does the opposite thing. In a grand unified theory we need
to put these into the same multiplet, but they have the wrong spinor structure. So we need the charge
conjugated fields ucR and ucL . We can work in Majorana representation of the Dirac γ matrices where
they are all imaginary, so the fields are real and we have ψ c = ψ ∗ . But in any other representation we
have ψ c = Cψ ∗ where C is the charge conjugation operator. We use that to define the charge conjugated
fields so that ucL will annihilate right handed antiquarks and create left handed quarks, and ucR doing the
opposite. In terms of particle structure we have ucL ↔ ūR and ucR ↔ ūL .

57
Quantum Field Theory III Lecture 13

13 Lecture 13
13.1 Grand Unification
Let’s write down all the particles in the Standard Model.

Particles Rerepesentation
(u, d)L (3, 2)1/3
dcL (3̄, 1)2/3
ucL (3̄, 1)−4/3
(ν, e)L (1, 2)−1
ecL (1, 1)2

Table 4: Standard Model Particles

If we have a grand unification theory, the electric charge should be a generator, therefore should be
traceless, so the sum of the electric charges of all particles should vanish, if we are to group them into
a single multiplet. Now if we have G breaks down to SU (3) × SU (2) × U (1) which has rank 4, then we
need the rank of G larger than 4. Because dcL and dL transform differently, we need G to have complex
representations. Remember the groups that have complex representations are SU (N ), SO(4N + 2), E6 .
For rank 4 we have SU (5). For rank 5 we have SU (6) or SO(10). And for rank 6 we have SU (7) and E6 .
We have obviously SU (7) ⊃ SU (6) ⊃ SU (5). But less obviously we have E6 ⊃ SO(10) ⊃ SU (5), which
can be seen from Dynkin diagrams shown in figure 13.1.

⊃ ⊃

E6 SO(10) SU (5)

Figure 13.1: Dynkin Diagrams for E6 , SO(10), and SU (5)

There will be anomaly problems if we use SU (N ), of course, but we can have anomaly problems as
long as the total anomaly cancel in the end. So let’s first consider SU (5). The way SU (5) breaks into
the SM gauge group is kind of obvious, because we can write the 5 × 5 matrix in SU (5) can be put into
diagonal 3 × 3 and 2 × 2 matrices and the U (1) generator comes from the matrix diag(1, 1, 1, −3/2, −3/2).
Now the representation should be
M
(rep)SU (5) = (SU (3), SU (2))Y (13.1)

So we can write down the representations

• 5: (3, 1)−2a/3 ⊕ (1, 2)a

• 5̄: (3̄, 1)2a/3 ⊕ (1, 2)−a

• 10 = (5 ⊗ 5)asym : (3̄, 1)−4a/3 ⊕ (1, 1)2a ⊕ (3, 2)a/2

58
Quantum Field Theory III Lecture 13

• (5 ⊗ 5)sym : (6, 1)−4a/3 ⊕ (1, 3)2a ⊕ (3, 2)a/3

• (5 ⊗ 5̄ − trace): (8, 1)0 ⊕ (1, 3)0 ⊕ (1, 1)0 ⊕ (3, 2)−5a/3 ⊕ (3̄, 2)5a/3

The last two representations will introduce new particles and we don’t want them. There is no (3, 1)
particle in our theory so we should also discard 5. Now in the rest we can just choose a = 1 and we recover
all the particles in the standard model with no less and no extra. The 5̄ representation correspond to dcL
and (ν, e)L , and the 10 = (5 × 5)asym representation correspond to ucL , ecL and (u, d)L . Looks pretty good.
Let’s write down the multiplets explicitly. The 5̄ representation field will be

5̄ = ψi = (dc1 , dc2 , dc3 , e, −ν)L (13.2)

and the 10 representation will be

uc3 −uc2 −u1 −d1


 
0
−uc3 0 uc1 −u2 −d2 
ij 1  c c

10 = χ = √  u2 −u1 0 −u3 −d3  (13.3)
2 u

u2 u3 0

−e+ 
1
d1 d2 d3 e+ 0

If we had a right-handed neutrino νR it will transform as a singlet under SU (5). Things look good so far.
Now let’s worryabout anomalies. The anomaly A will be proportional to dabc which is proportional to
the trace tr(T a T b , T c ). Now in SU (5) the generator T 24 is analogous to the T 8 in SU (3). Its matrix
representation in the fundamental representation is
1
T 24 = √ diag(1, 1, 1, 1, −4) (13.4)
10
The trace of this generator cubed is just
h 3 i 1 60
tr T 24 = (1 + 1 + 1 + 1 − 64) = − 3/2 (13.5)
103/2 10
√ √
So in the fundamental representation 5 we have −60/( 10)3 and in 5̄√we have +60/( 10)3 . In 10 repre-
sentation for ab a5
√ elements like χ where a, b = 1, . . . , 4 we have Tab = 2/ 10, while for elements χ we have
T24 = −3/ 10. So in the end the sum will be
 3  3
h 3 i 2 3 60
tr T 24 =6× √ + 4 × −√ = − 3/2 (13.6)
10 10 10

So a miracle happens, that there is no anomaly if the fields are from 5̄ and 10. In general for SU (N ) and
the antisymmetric product of p fundamental representations, the anomaly will be
n o  (N − 3)!(N − 2p)
tr T a , T b T c = A(R)dabc , A(R) = (13.7)
(N − p − 1)!(p − 1)!
So all the anomalies will cancel out if they cancel for one set of indices abc.
Now let’s consider the gauge bosons for SU (5). These fit into

24 = (8, 1)0 ⊕ (1, 3)0 ⊕ (1, 1)0 ⊕ (3, 2)−5a/3 ⊕ (3̄, 2)5a/3 (13.8)

59
Quantum Field Theory III Lecture 13

The first part (8, 1)0 is just the gluons. The second and third will give us W ± , W 0 , and C 0 and the linear
combination of the latter two will give us Z 0 and γ. The fourth term can be called X 4/3 and Y 1/3 and
the fifth part is just the antiparticles of these X̄ and Ȳ . Presumably there will be some Higgs mechanism
which breaks the symmetry group to the standard model and these particles X and Y obtain a very high
mass.
Let’s write the gauge bosons in 5 × 5 matrices V µ . The gauge field will be Vaµ T a where we normalize
the generators by trT a T b = 2δ ab . We write the gauge field as the following matrix
 
X1 Y1
 gluons X2 Y2 
1  
Vµ = √  X3 Y3 (13.9)
 
2 √ 
W+
 
X̄1 X̄2 X̄3 W3 / 2 √

Ȳ1 Ȳ2 Ȳ3 W− W3 / 2

In the lower right corner we have the matrix



1 √3 B W −/ 2
!
2 W3 + 2 15 3
√ = W · τ /2 + √ I (13.10)
W +/ 2 − 12 W3 + √3 B
2 15
2 15

Let’s now worry about coupling constants. The SU (5) coupling constant is g5 , then we can fix the
SU (3) coupling constant to also be g = g5 . If we multiply the above expression by g5 then we get

g0 3
= √ g5 (13.11)
2 2 15
where g 0 /2 is the hypercharge√coupling constant in standard model. So we have the weak mixing angle
should be tan θW = g 0 /g = 3/ 15. But experimentally we have sin2 θW = 3/8 which contradicts with the
prediction.
Now let’s just ignore it and move on. Let’s consider couplings between gauge bosons and 5 and 5̄.
Because 5̄ has one index down and 5 has one index up, so V should have one index up and one down, so
we have the gauge coupling term like
g ψ̄ i γ µ (Vµ )j i ψj (13.12)
The terms with X have indices a4 and terms with Y have indices a5. So we will look at couplings like

g ψ̄ a γ µ (Vµ )4a ψ4 (13.13)

This will give us a coupling between d¯ and X and e, so the term will look like d¯cL X a eL and it gives us a
process
X 4/3 −→ e+ + d¯ (13.14)
Similarly we have a process
Y 1/3 −→ ν + d¯ (13.15)
Now let’s look at the 10 representation. We have the coupling term look like

χ̄ij γ µ (Vµ )j k χik (13.16)

60
Quantum Field Theory III Lecture 13

If we want X boson then we again look at a4 term which gives us

χ̄ba γ µ (Vµ )a4 χb4 =⇒ uXu (13.17)

So this will be a process


X 4/3 −→ u + u, or X 4/3 −→ d¯ + e+ (13.18)
Similarly we can work out the processes for Y bosons

Y 1/3 −→ e+ + ū, or Y 1/3 −→ u + d (13.19)

Now let’s look at quantum numbers. For the X −4/3 decay it can go to e− d which have charge −4/3, baryon
number 1/3 and lepton number 1, so we have B − L = −2/3. It can also go to ūū which have charge
−4/3, baryon number −2/3 and lepton number 0, which also have B − L = −2/3. Similarly we can work
out Y −1/3 decay. Now it is obvious that baryon number is not conserved because X can’t have a definite
baryon number. But now the difference of baryon number and lepton number is conserved, because both
X and Y can be assigned this quantity B − L = −2/3.
We can draw graphs using the above to give us processes that violate baryon number conservation.
These diagram will allow for processes like

p −→ e+ + π 0 , p −→ ν̄ + π + (13.20)

And this is just one generation, and we can also have p → e+ K 0 etc.
The matrix element of these kind of processes will be proportional to g 2 /MX 2 . So the decay rate which

is proportional to the square of this will be proportional to (g 2 /MX


2 )2 , times some mass to the fifth power

to get the dimension right. The lifetime τ will be 1/Γ and the kinematic mass should be proton mass. So
we should expect
4 
1 5 MX 4 mπ,µ 5
    
1 MX
τ= ∼ 4 ∼ τπ,µ (13.21)
Γ g mp MW mp
We can get from here the bound on MX . The lifetime of proton have a bound of τ & 8.2 × 1033 yr. To
detect this kind of possible decay, we either wait for such a long time looking at one proton, or we assemble
a large body of protons and isolate it well enough to see the decay.

61
Quantum Field Theory III Lecture 14

14 Leture 14
Let’s continue to look at baryon number violation. Let’s consider graphs with external legs all standard
model particles. What do we need to get baryon number violation? Recall left-handed quarks have baryon
number 1/3 and right-handed quarks have baryon number −1/3. However we need also to get a color
singlet in the process, so we can have
• (3, 2)1/3 , (3, 2)1/3 , (3, 1)−2/3
• (3̄, 1)2/3 , (3̄, 1)2/3 , (3̄, 1)−4/3
But these are fermions, and we can’t have an interaction between three fermions in standard model, so
we need at least 4 particles. But the amplitude will then be proportional to ψ 4 ∼ (1/M )2 where M is
whatever mass associated with the intermediate particles.
How about lepton number? We can have
• (1, 2)−1 , (1, 2)−1 , (1, 1)2
• (1, 1)2 , (3̄, 1)−4/3 , (3, 1)−2/3
• (1, 2)−1 , (3, 2)1/3 , (3̄, 1)2/3
Again we have similar suppresion as above. But now if we are in supersymmetry then every fermion has
a boson partner with the same quantum numbers, and we don’t have the suppression.
Let’s now consider the change of B − L. We can have the 4 external legs as

(3, 2)1/3 , (3, 1)4/3 , (3.1)−2/3 , (1, 2)1 (14.1)

The total B − L for this process is 2. So in grand unification theory, B − L can be violated.

14.1 Symmetry Breaking


Let’s now consider how the symmetry SU (5) is broken into the standard model gauge group. Presumably
this is due to some Higgs mechanism, and there is a Higgs field Φ with VEV at the GUT scale. However
we also know the breaking of the electroweak due to Higgs of VEV at about 250 GeV. We know that the
latter Higgs gives masses to know fermions, but the previous Higgs boson should better not give fermion
masses.
The expectation value of Φ should be in the SU (5) representation (1, 1)0 because it is a scalar. The
smallest representation containing this is the adjoint representation 24. Let’s write the Φ in terms of 5 × 5
traceless hermitian matrices, and then the Higgs potential can be
µ2 a  b
V (Φ) = − Tr Φ2 + Tr Φ2 + Tr Φ4 (14.2)
2 4 4
Only the first term is sensitive to the distribution of eigenvalues. For b > 0 we can take the vacuum
expectation value as  
1
 1 
 
hΦi = v 
 1 
 (14.3)
 −3/2 
−3/2

62
Quantum Field Theory III Lecture 14

This breaks the group into SU (3)×SU (2)×U (1). The above matrix is a block matrix, and the off diagonal
blocks are absorbed into the gauge bosons X and Y to give them masses m2X = m2Y = 25g 2 v 2 /8. The
remaining Higgs bosons are (8, 1)0 , (1, 3)0 , and (1, 1)0 and they should have masses about the same as mX
and mY .
The electroweak Higgs is in representation (1, 2)1 which is contained in the representation 5. But it is
also contained in 45 which is the tensor Hcab satisfying

Hcab = −Hcba , Haab = 0 (14.4)

We can consider it as χab ψc minus a trace piece. Now between 5 and 45 we should definitely first try 5.
We write the components as
H = (h1 , h2 , h3 , h+ , h0 ) (14.5)
The last two should be the Weinberg-Salam doublet and the first three should be an SU (3) triplet. So far
the triplet has not been seen, but neither is the doublet. The potential should be
 2
V (H) = −m2 H † H + λ H † H (14.6)

with minimum at H † H = m2 /2λ which should be around (250 GeV)2 . Now suppose

hHi = (0, 0, 0, 0, w/ 2) (14.7)

Now we have some problem. How do we know that the expectation value sits in the SU (2) part instead of
the SU (3) part? This will also be a symmetry breaking into SU (4) × U (1), which introduces 9 Goldstone
bosons, of which 3 are eaten by W and Z, and the remaining 6 gain mass due to the previous symmetry
breaking with masses at the order of O(w2 /v 2 )w which are very light. These can’t be so light but we didn’t
see them.
Now let’s consider loops. The loops due to X and Y will give us terms like

V (H, Φ) = αH † H Tr Φ2 + βH † Φ2 H (14.8)

Now if β > 0 then we expect the nonvanishing part of H should be in the SU (3) part, and if β < 0 we
expect it in the SU (2) part. So if we take β < 0 then this resolves the first problem above. Now let’s
minimize the whole potential. Because H breaks the SU (2) symmetry, we will expect

hΦi = v diag(1, 1, 1, −3/2 − /2, −3/2 + /2), H = (0, 0, 0, 0, w/ 2)T (14.9)

Then the value of the potential is

m2 2 λ 4 α 2
   
15 2 β 9 2
V (H) = − w + w + w v + w2 v + O() + V (v) (14.10)
2 4 2 2 2 4
If we differentiate with respect to w and then divide by w then we get the minimum to be at
   
2 1 2 15 9
w = m − α + β v2 (14.11)
λ 2 4

Now w2 and v 2 are at scales of about 16 orders of magnitude different. It is possible but very difficult
to tune the parameters, because we also have the loop corrections to all orders. This is one version of

63
Quantum Field Theory III Lecture 14

the Hierarchy problem. Supersymmetry is a partial solution to this in the sense that with SUSY if we
fix the tree level parameters then we can fix all the loop corrections. But somehow if we worked out the
parameters, then the masses of the Higgs bosons from the SU (3) sector of H will be very heavy, and this
resolves the second problem above.
Let’s now look at fermion masses. The Dirac mass term is proportional to ψ̄R ψL + ψ̄L ψR . Note the
charge conjugated field is

ψLc = CψR =⇒ ψ̄R = (ψLc )T (C −1 )T γ 0 (14.12)
So we can write the mass term as proportional to ψLc ψL . Now the fermion mass terms should come from
(5̄ ⊕ 10) ⊗ (5̄ ⊕ 10). Recall that 5̄ = (dc , e− ), and 10 = (u, d, uc , e+ ). So we expect that 10 ⊗ 10 gives
us u mass, 10 ⊗ 5̄ gives us d and e masses and 5̄ ⊗ 5̄ should not give us anything. Now let’s work the
representations out and we have

¯ ⊕ 15,
5̄ ⊗ 5̄ = 10 ¯ 5̄ ⊗ 10 = 5 ⊕ 45, ¯ ⊗ 50
10 ⊗ 10 = 5̄ ⊗ 45 (14.13)

Note that none of these have a singlet, so we can’t put in a bare mass term by hand. However we can get
mass terms from the Higgs particles, corresponding to 5 or 45 and complex conjugate. Let’s try the 5 first
with expectation value at the fifth component. We have nothing at 5̄ ⊗ 5̄. The 5̄ ⊗ 10 gives us something
like
w
GD ψ̄jc χjk Hk −→ GD ψ̄jc χj5 √ (14.14)
2
Now the fifth column of χ is just d and e+ so from this term we get md and me . From this we know
me = md , and similar for all the families. This is a nice relation except that it is not true. Now let’s
consider 10 ⊗ 10 and we get the term like
w
GV abcde χab χcd H e −→ GV abcd5 χab χcd √ (14.15)
2
This gives us the u mass.
Now we need to deal with the problem that me = md . First we can take into account that the couplings
run. But we can also take the Higgs to be in 45 instead of 5, but then we loose some predictability because
there are many more things to fix.
Enough with masses, so let’s come to couplings. The most obvious problem is that in SU (5) we only
have one coupling constant. Note that the loops only depend on particles with mass lower than the energy
scale. So we expect that for energies larger than the GUT scale, all the couplings are the same and they
run together. But at some lower scale the couplings divide into 3 different ones. Now we can figure out
how the couplings run at lower scales and extrapolate them to see if they meet at a high scale. Now some
data. At mZ we have
−1
αEM = 127.906 ± 0.019, sin2 θW = 0.2312 ± 0.0002, α3 = 0.1187 ± . . . (14.16)

So that gives us about


α3−1 = 8.425, α2−1 = 29.57, α3−1 = 59.00 (14.17)
where α2 = αEM / sin2 θW and α1 = αEM / cos2 θW . Now at one loop order we have

dgj
µ = bj gj3 (14.18)

64
Quantum Field Theory III Lecture 14

and at this order they don’t mix. So we can write

dαj 1 dαj−1 ∂αj−1


µ = bj g 4 = 8παj2 bj , µ = −8πbj = (14.19)
dµ 2π j dµ ∂ ln µ
So at one loop order these run in straight lines, and we just need to extend them from their initial value
at mZ to see if they meet. Now for U (1) we have
!
1 4 X 2 1 X 2
8πb = qj + qj (14.20)
2π 3 3
fermions scalars
p
where qj = 3/20 Y . For SU (N ) coupling it is
!
1 11 4 X 1 X
8πb = − N+ T (R) + T (R) (14.21)
2π 3 3 3
fermions scalars

where T (R) depends on representations and


(
1/2 for fundamental rep.
T (R) = (14.22)
N for adjoint rep.

Now we can plug everything from standard model into the above equations and work out
4 22 4 1 19 4 1 41
16π 2 b3 = −11 + Ng = −7, 16π 2 b2 = − + Ng + = − , 16π 2 b1 = Ng + = (14.23)
3 3 3 6 6 3 10 10
We can write  
MU
αj−1 = 8πbj ln −1
+ αU (14.24)
µ
where U means at GUT scale. Because any two lines intersect, so let’s do them in pairs. We get
• From 1,2: MU = 1.0 × 1013 GeV, ln MU /MZ = 25.46

• From 1,3: MU = 2.6 × 1014 GeV, ln MU /MZ = 28.68

• From 2,3: MU = 1.0 × 1017 GeV, ln MU /MZ = 34.66


So they don’t meet. There are things that we can change. We can also consider the 2-loop and higher
loop corrections. It might also be that they are not turned of at exactly the GUT scale. Or there are
more particles in between mZ and GUT. But let’s consider supersymmetry. We have the MSSM which is
the minimal supersymmetric standard model. We add a superpartner for all the standard model particles.
So for each qL ,qR ,lL , and lR we add one scalar, so we multiply the coefficient of Ng by 3/2 in the above
formulae. We also add a Weyl fermion for each gauge boson, and we add 2N/3 to each number. For the
Higgs boson, now we need two Higgs doublets instead of 1, and they both have fermion partners. So we
have the 1/3 Higgs contribution turning into 2. Now the formulae become
3 33
16π 2 b1 = 2Ng + = , 16π 2 b2 = −6 + 2Ng + 1 = 1, 16π 2 b3 = −9 + 2Ng = 3 (14.25)
5 5
Now the same number calculation will give us

65
Quantum Field Theory III Lecture 14

• From 1,2: MU = 2.2 × 1016 GeV, ln MU /MZ = 33.13

• From 1,3: MU = 2.2 × 1016 GeV, ln MU /MZ = 33.10

• From 2,3: MU = 2.4 × 1016 GeV, ln MU /MZ = 33.22

So these do seem to meet! Of course we still have problems like fermion masses, but this is a strong
indication.

66
Quantum Field Theory III Lecture 15

15 Lecture 15
Let’s say something about SO(10). We know that in SU (5) the standard model fits into 5̄ ⊕ 10(⊕1). In
SO(10) we know that it contains SU (5), in two ways. If we look at the Dynkin diagram and remove a
node,
P 2 we can get SU (5). AnotherPway 2to see it is that SU (10) is the transformation that fixes the sum
xi , but for SU (5) it conserves |zi | which also have 10 components but it is more restrictive.
Let’s look at the correspondence. The 10 representation of SO(10) corresponds to 5 ⊕ 5̄, but it is real
and we don’t want it. So we need to look at the spinor representation. There are 5 commuting generators
so the spinor representation should be written as (±1/2, ±1/2, ±1/2, ±1/2, ±1/2). But these are reducible
into representations of odd number of + and even number of +. These are 16 and 16 respectively.
Let’s look at 16. The representation with one + has 5 permutations, the one with three + has 10
permutations, and the one with 5 + has only one case. Under the generators of SU (5) these do not mix
with each other, so it reduces into 1 ⊕ 5̄ ⊕ 10. The bar or no bar on 5 doesn’t matter, just a matter of
choice. We can see that 10 is the 10 we want from anomaly considerations. So all the standard model
particle fits into 16 exactly, which is economical and attractive.
Now we need to break this symmetry. We can imagine
SO(10) −→ SU (5) × U (1) −→ SU (5) −→ . . . (15.1)
But we can also imagine
SO(10) −→ SO(6) × SO(4) = SU (4) × SU (2)L × SU (2)R (15.2)
where SU (2)L is the weak coupling that breaks chirality. This looks like we have deeper chiral symmetry
which is broken in standard model. Among the infinite amount of papers that are generated, there are
also more ways to break the group down.

15.1 Solitons
Now we change subject completely and talk about solitons. Solitons are localized solutions of classical field
equations. By localized we mean that if it starts at some finite region in space it stays there. Originally
they were required to retain their form even after scattering. This is the soliton in mathematical sense
occurring in the solution in nonlinear differential equations.
Let’s start with some classical field theory in 1 + 1 dimensions. The Lagrangian is
1 m2 2 λ 4 λ 4 λ 2
L = (∂µ ϕ)2 − V (ϕ), V (ϕ) = − ϕ + ϕ + v = (ϕ − v 2 )2 (15.3)
2 2 4 4 4
p
where we defined v = m2 /λ. The minima of the potential are at v and −v. We√know from symmetry
breaking that there are two vacuum states and elementary excitation has mass 2m. If you are in a
vacuum state and are asked which vacuum you are in, there is no way to tell, because physics looks the
same.
The classical field equation is
∂ 2 ϕ ∂ 2 ϕ ∂V
− + =0 (15.4)
∂t2 ∂x2 ∂ϕ
The static solution is just when the time derivative is zero, and it is equivalent to minimizing the potential
energy Z  
1 02
U [ϕ(x)] = dx ϕ + V (ϕ) (15.5)
2

67
Quantum Field Theory III Lecture 15

Now we need to look for a solution. It had better have finite energy and go to one vacuum or another
at infinity. Let’s consider a configuration which connects one vacuum to the other at different infinities.
Now if we start from this configuration and do a variation continuously to lower the energy, we will hit
a minimum, which must be different from either vacuum solution. So we should have a solution which
connects two vacuua and we call this a “kink”. Now mathematicians will say the space of functions is not
compact and we can’t use this argument, but it turns out that here it is fine.
Let’s multiply the equation by ϕ0
 
0 00 0 ∂V d 1 02
0 = −ϕ ϕ + ϕ = − ϕ +V (15.6)
∂ϕ dx 2

So the term in brackets is independent of x, and we can shift it to be zero, which we already did when
introducing v. So we have
dϕ √ dϕ
= ± 2V , √ = dx (15.7)
dx 2V
and we can integrate the equation. Let’s call the point when ϕ = 0 by x0 , and integrate from there
r Z ϕ
dϕ0
Z x
2
= dx0 = x − x0 (15.8)
λ 0 v 2 − ϕ2 x0

If we do the integral on the left then we will find that


 
m
ϕ = v tanh √ (x − x0 ) (15.9)
2

The size of the kink is on the order of 1/m which can be seen from the equation directly. So ϕ0 is on
the order of m2 v 2 ∼ m4 /λ. The classical energy of this solution is
√ √ !
1 m4 m3 2 2 m3 2 2 m2
 
Ecl = Mcl ∼ ∼ = = m (15.10)
m λ λ 3 λ 3 λ

The term in the final bracket is dimensionless in 1 + 1 dimensions. In the limit of weak coupling
√ we have
λ/m2  1, so this mass, or energy, is much much larger than the elementary excitations 2m.
Now we can consider solutions of many kinks. But we need to have the above integral relation which
tells us that we can only go to vacuum at x → ∞. But we can have time-dependent solution, which are
kinks moving around. When they collide a kink will dissipate with an antikink. The number of kinks
minus the number of antikinks will be conserved and we can define the topological current
1 µν
Jµ =  ∂ν ϕ (15.11)
2v
This is obviously conserved because of the  factor. We can similarly define the topological charge
Z Z
0 1 1
Qtop = dx J = dx 01 ∂1 ϕ = [ϕ(∞) − ϕ(−∞)] = ±1 (15.12)
2v 2v
But this is not due to a symmetry!

68
Quantum Field Theory III Lecture 15

Another example which will be important is


" √ ! #
m4 λ
V =− cos ϕ −1 (15.13)
λ m

The classical field equation is


√ !
m3 λ
0 = ϕ + √ sin ϕ (15.14)
λ m
Some people call this “Sine-Gordon” equation. This is a classical theory where there is solition solution.
Let’s look at the vacuum solutions. It occurs at minimum of the potential, which is at
2πm
ϕ = √ N = Nv (15.15)
λ
where N is an integer. So we have many vacuum solutions. If we consider ϕ ≈ 0 then we have

m4 λ2 4 m2 2 λ 4
 
λ 2
V =− − 2ϕ + ϕ + . . . = ϕ − ϕ + O(ϕ6 ) (15.16)
λ 2m 4!m4 2 4!
so the elementary excitation around ϕ = 0 has mass m. Now if we proceed as before, we can find the kink
solution which connects the vacuum N to N + 1 as
2v  
ϕkink = N v + tan−1 em(x−x0 ) (15.17)
π
with classical mass Mcl = 8m3 /λ.
All these discussions are classical, but we know that there is nothing as classical field theory in nature.
So question is when is the classical theory like above meaningful? Suppose we want to consider the kink
as a particle and think of its position, then we need its Compton wavelength to be much less than the size
of the particle
1 λ 1
λc ∼ ∼ 3  (15.18)
Mcl m m
so this is true in weak coupling. What about the value of the field? How do we measure it? We can’t
measure the value of a field at some point, because it will oscillate wildly. Let’s define a smeared ϕL (x)
over a smeared length L. We will expect that
 (d−2)/2
1
(∆ϕ) ∼ (15.19)
L
where d is the space-time dimension and when d = 2 it does not depend on L, up to logarithm terms. So
at very small distances we expect large fluctuations, but it should be smooth when L is comparable to the
characteristic size of the classical solution
m
(∆ϕL )quantum  (ϕ0 L) ∼ v = √  1 (15.20)
λ
So now let’s do quantum field theory. We do QFT about a vacuum using

ϕ(x, t) = ϕvac + η(x, t) (15.21)

69
Quantum Field Theory III Lecture 15

and we write the Lagrangian as a classical piece plus a quadratic piece and an interaction piece

L = Lclass + Lquad + Lint (15.22)

The first is just the vacuum energy. The second is


Z  
1 2 1 00 2
L = dx (∂µ η) − V (ϕvac )η (15.23)
2 2
We find the normal modes of η using the classical equation of motion and write η(x, t) in a sum of modes
X
η(x, t) = [cj (t)fj (x) + h.c.] (15.24)
j

where fj are classical functions and cj are quantum mechanical operators. Then the Hamiltonian will look
like that of harmonic oscillators and the energy is
X 1 
E = Ecl + + nj ωj + . . . (15.25)
2
j

Even if nj = 0 for all j we have divergence when we add 1/2 for sufficiently many times. So we add a term
X1
− δE = ωj (15.26)
2
j

to the Lagrangian to cancel this zero point energy. There is also divergence due to loop corrections so we
add a counterterm
1
Lct = (δm)2 ϕ2 (15.27)
2
Now we write ϕ(x, t) = ϕkink + η(x, t). This is equivalent to taking the classical kink as a background.
Now the normal modes are solutions to the equation
d2
 
00
− 2 + V (ϕkink ) fj (x) = (ωjkink )2 fj (x) (15.28)
dx
and again we have divergences in the energy. But we have already added the vacuum energy and the loop
counterterms and we don’t have any freedom left in our theory to fix these divergences. These divergences
had better cancel.
Let’s try the kink solution  
m
ϕkink = v tanh √ (x − x0 ) (15.29)
2
so the potential derivative is
 
00 2 3
V (ϕkink ) = m 2 − √ (15.30)
cosh2 (m(x − x0 )/ 2)
So the classical normal mode equation looks like a Schrödinger equation with a potential like the above.
Note that the ϕkink satisfies
∂V ∂2V
ϕ00kink − =0 =⇒ ϕ000 0
kink − ϕkink =0 (15.31)
∂ϕ ∂ϕ2

70
Quantum Field Theory III Lecture 15

So this tells us that ϕ0kink is a zero-mode solution with ω 2 = 0. This is the lowest energy solution to the
Schrödinger equation and it has no nodes. All the other solutions should have higher energies and more
nodes. There is another node which has ω12 = 3m2 /2 and the function is

sinh(m(x − x0 )/ 2)
f1 = √ (15.32)
cosh2 (m(x − x0 )/ 2)

After that there is continuum with ω = k 2 + m2 and

    
ikx 2 2 m(x − x0 ) 2 2 m(x − x0 )
fk = e 3m tanh √ − m − 2k − 3 2imk tanh √ (15.33)
2 2
We are really interested in the behavior at x → ±∞ which is
p
fk → 4(m2 − k 2 )2 + 18m2 k 2 ei[kx±δ(k)/2] (15.34)

where √ !
−1 3 2 mk
δ(k) = 2 tan (15.35)
2 k 2 − m2

We need to find the branch of δ so we define δ(0+ ) = 2π and δ(+∞) = 0. Note we have δ(−k) = −δ(k)
and there is a discontinuity at 0.
Now the energy of the kink is
 X  Z
1 kink 1 X vac 1 2
dx ϕ2kink − ϕ2vac

Ekink = Mcl + ωj − ωj + δm (15.36)
2 2 2

The sums are badly divergent. Let’s define the world to have length a and use periodic boundary conditions,
so we have a cut off on momentum. We also assume that space has only N points, which means there are
only N modes. Now in vacuum instead of continuum of k we should have kn a = 2πn where n = ±1, ±2, . . .
for the vacuum solutions. But for the kinks we have kn a + δ(kn ) = 2πn where n = ±2, ±3, . . . . This is
because for the kink we have two discrete states below the continuum. Now we should have
 
kink vac 1 vac 1
kn − kn = − δ(kn ) + O (15.37)
a a2

Now the vacuum zero point energies become


N q
!
√ √
r 
1 1 3 X
kink 2 2
p
vac 2 2
(0 − 2m) + m − 2m + (kn ) + 2m − (kn ) + 2m
2 2 2
n=2
! (15.38)
√ √
r
vac δ(k )
 
1 1 3 Xp −k n n 2
= (0 − 2m) + m − 2m + (knvac )2 + 2m2 + O(1/a )
2 2 2 [(kn )2 + 2m2 ]a

Now if we take the continuum limit we have instead of the sum


Z Z ∞  
1 kδ(k) 1 d p 2 2
p
2 2

− dk √ =− dk [ k + 2m δ(k)] − k + 2m (15.39)
2π k 2 + m2 2π 0 dk dk

71
Quantum Field Theory III Lecture 15

And the vacuum energy is


√ √ √
X 3 2
Z
1 6 3 2 √
ω→− m dk √ − m− m + 2m (15.40)
2π 2
k + 2m2 6 2π

Now we have got rid of most of the divergences but we still have a logarithmic divergence from the above
integral. It is cancelled by the counterterm. So in the end we have
√ √ √ !
2 2 m3
 
6 3 2 λ
Mkink = + − m+O m (15.41)
3 λ 12 2π m2

Now we are just quantizing things at the kink background, instead of the kink itself. But let’s look at the
spectrum. The first excited state is like a one particle state, and the continuum is like particles scattering
off the kink. However the ω = 0 mode doesn’t correspond to any excitations. It turns out that this is the
mode correspond to the position of the kink.

72
Quantum Field Theory III Lecture 16

16 Lecture 16
16.1 Kink Solution Continued
Suppose we have some classical solution
X
ϕ(x, t) = ϕkink (x) + cn (t)fn (x) (16.1)
n

where fn (x) are normal modes about the kink solution


∂2
 
00
− 2 + V (ϕkink ) fn = ωn fn (16.2)
∂x
we always have the zero mode form the equation of motion of the kink itself
∂ϕkink
ω = 0 −→ f (x) ∼ (16.3)
∂x
If we consider the perturbation around the kink, the Lagrangian can be written as
L = Lclassical + Lquad + Lint (16.4)
and the energy will be  
X 1
Ekink = Ecl + ~ω nk + + ... (16.5)
2
k
Last time we found that the ω = 0 mode will give us infrared divergence. The first thing we need to
realize is that the zero mode is not an oscillator. We introduce a collective coordinate z(t) and expand the
solution as X
ϕ(x, t) = ϕkink (x − z(t)) + ck (t)fk (x − z(t)) (16.6)
k6=0

Up to now we haven’t specified the solution. For ϕ4 theory we had that


 
m
ϕkink (x) = v tanh √ (x − z(t)) (16.7)
2
so z(t) now represents the position of the kink. This essentially absorbs the zero mode into the definition
of the coordinate. We now need to work out the Lagrangian in terms of the collective coordinate. The
original Lagrangian is " #
1 2 1 dϕ 2
Z  
L = dx ϕ̇ − − V (ϕ) (16.8)
2 2 dx
Now ϕ̇ is
∂ X ∂fk

ϕ̇(x, t) = −ż ϕkink + ċk fk (x − z(t)) − żck (16.9)
∂x ∂x
k6=0

The Lagrangian gets split into different parts. The part without z and ż is the same as before, a sum of
simple harmonic oscillators, except that we don’t have the zero mode, but we didn’t have the zero mode
anyway because ω = 0. The term with ż 2 will look like
Z  2
1 2 ∂ϕkink
dx ż (16.10)
2 ∂x

73
Quantum Field Theory III Lecture 16

And we still have the terms with ċk ż which is


XZ ∂ϕkink (x − z)
ż ċk fk (x − z) = 0 (16.11)
∂x
k6=0

this is zero because the zero mode is orthogonal to the higher modes and the integral vanishes. So the only
extra term is from the ż 2 . But remember from the classical energy we have
Z  2 Z
1 ∂ϕ 1
dx = dx V (ϕ) = Mcl (16.12)
2 ∂x 2

So the extra term is essentially Mcl ż 2 /2. So now the Hamiltonian is

1 X P2
H = Mcl ż 2 + (S.H.O) + Mcl = + ... (16.13)
2 2Mcl
Now the states are labelled with the momentum of the zero mode P and the oscillators nj , so the energies
will be
P2 X 1 P2
 X p X
E = Mcl + + + nj ωj + · · · = Mkink + + nj ωj + · · · = Mkink + P 2 + nj ωj + . . .
2Mcl 2 2Mcl
j j j
(16.14)
Now we can do perturbation theory for the zero mode. Similarly we can generalize to systems with many
zero modes or zero modes with internal quantum numbers and internal charges.

16.2 Multikink Solutions


Let’s look at sine-Gordon equation where there are many kinks going around and we can write down the
solutions analytically. Remember the sine-Gordon kink solutions are
2v h i
ϕ(x) = N v + tan−1 em(x−x0 ) (16.15)
π
The is a kink that takes the N v vacuum to the (N + 1)v vacuum because ϕ(x → −∞) = N v whereas
ϕ(x → ∞) = (N + 1)v. The time-dependent two-kink solution can be written as
(    −1 )
2v −1 mx mut
ϕ(x, t) = tan u sinh √ cosh √ (16.16)
π 1 − u2 1 − u2

The limiting behavior is ϕ → ±v when x → ±∞. The energy is

16m3 1 2Mkink
E= √ =√ (16.17)
λ 1−u2 1 − u2
which is what we expect from two moving kinks with no interacting energies between them. The kink-
antikink solution is
(    −1 )
2v −1 1 mut mx
ϕ(x, t) = tan sinh √ cosh √ (16.18)
π u 1 − u2 1 − u2

74
Quantum Field Theory III Lecture 16

So now the limiting behavior is ϕ → 0 as x → ±∞ and the energy is the same as above. Now what we can
do is take the velocities to be imaginary and take u → is and the solution will look like
(    −1 )
2v 1 mst mx
ϕ(x, t) = tan−1 sin √ cosh √ (16.19)
π s 1 + s2 1 + s2

Now the energy is


2Mkink
E=√ (16.20)
1 + s2
The solution is periodic in t because of the sine function, and the period is
√ "  #1/2
2π 1 + s2 16m3 2π 2

τ= , E= 1− (16.21)
ms λ mτ

This solution is called the Doublet, or the Breather. Quantum mechanically we expect the period to be
quantized, and we expect some quantum states. From Bohr-Sommerfeld quantization we have
I I I
p dq = pq̇ dt = (H + L) dt = 2πN (16.22)

We can extend this to multiple degrees of freedom. Ref. Phys. Rev. D 10, 4114 (1974) by Dashen, Hass-
lacher, and Nero. Applying their method to our sine-Gordon case we have

S(τ ) + M τ (M ) = 2πN (16.23)


H
where S = L dt. From the expressions above we can work out S(τ ) in closed form
( )
32πm3
 
2π 1
S(τ ) = cos−1 −p (16.24)
λ mτ (mτ /2π)2 − 1

But notice that the second term in the bracket is just M τ (M ), so we just have

16m3
   
−1 2π 2π λ
Nπ = cos , = cos N (16.25)
λ mτ mτ 16m3
and the quantized energy becomes
16m3
 
λ
EN = sin N (16.26)
λ 16m3
This
P is the leading order. The next order correction comes from summing over the zero-point energies
~ω/2. The mass of the kink gets replaced by
−1
8m3

8m 8m λ
0 λ
Mkink = = → 0 , γ = 2 1− (16.27)
λ γ γ m 8πm2
And substituting this into our energy expression above we have the energy to first order correction

N γ0
 
16m
MN = 0 sin (16.28)
γ 16

75
Quantum Field Theory III Lecture 16

And this turns out to be the only correction, just as the first correction to the harmonic oscillator is exact.
Let’s consider the weak coupling limit λ/m2  1, and we have

16m γ 0
   
0 λ λ
γ = 2 1+ + . . . , M1 = 0 + . . . = m + O(λ2 ) (16.29)
m 8πm2 γ 16

This mass is equal to the elementary particle, and actually the N = 1 state is just the elementary particle
ϕ. The higher masses are " #
 2
1 3 λ
Mn = M1 n − (n − n) + ... (16.30)
6 16m2
So this looks like n elementary particles bind together, and the second term is just the binding energy.
Now because the energy is periodic, we need to look at n < 8π/γ 0 or we will be overcounting. At large
coupling limit we have  
Nπ 1
MN = 2Mkink sin (16.31)
2 (8πm2 /λ − 1)
So our requirement on n is now translated into N < 8πm2 /λ − 1. So we have two possibilities

• If λ/m2 > 4π there is no breathers and there is no elementary ϕ.

• If 2π < λ/m2 < 4π we have Mkink < M1 = Mϕ .

Now if we are near the λ/m2 = 4π point we have


λ 1 π 1 π π
= , = = − 2δ (16.32)
4πm2 1 + δ/π 2 (8πm2 /λ − 1) 2 + 4δ/π 2
So the mass of the breather is
4δ 3
 
2
M1 = Mkink 2−δ + + ... (16.33)
π
Note there is no breathers if δ < 0.
Now let’s change topic and consider the Thirring Model. The Lagrangian is
g
L = ψ̄(i∂/ − M )ψ − (ψ̄γ µ ψ)2 (16.34)
2
Because there is a mass, we call it the Massive Thirring Model. The spectrum of the theory contains
fermions and antifermions with mass M . If the coupling is positive g > 0 we also have bound states with

4g 3
 
2
M1 = M 2 − g + + ... (16.35)
π

This is exactly the same formula as our soliton energy above. Let’s wildly suppose there is some corre-
spondence here with g ↔ δ. Then we have the correspondence between coupling constants
λ 1
2
←→ (16.36)
4πm 1 + g/π
and weak sine-Gordon model corresponds to strong massive Thirring model, and strong sine-Gordon corre-
sponds to weak massive Thirring, and too strong sine-Gordon will correspond to negative g. The particles

76
Quantum Field Theory III Lecture 16

will also have correspondence. The kinks and anti-kinks correspond to ψ and ψ̄ in Thirring model. The
elementary excitation ϕ will correspond to ψ ψ̄ bound states and higher breathers will correspond to higher
bound states.
Remember the current in the soliton theory we have something like ∂ϕ. Coleman showed that the
correspondence is
√ √ !
λ µν µ m4 λ
 ∂ν ϕ ←→ −ψ̄γ ψ, cos ϕ ←→ −M ψ̄ψ (16.37)
2πm λ m
This suggests that the two theory are actually the same theory written in two different forms. This suggests
that solitons are just different ways to consider particles in different coupling regimes. This is one form of
duality, and it is very useful if you want to solve strong interaction.
Now let’s make the correspondence more explicit. Let’s write the γ matrices as
   
0 0 1 1 0 1
γ = , γ = (16.38)
1 0 −1 0
ψ1

and the fermion fields ψ = ψ2 are

2πi x
 Z 

ψ1 = C : exp − ϕ(x) − dz ϕ̇(x) : (16.39)
2 β −∞
2πi x
 Z 

ψ2 = −iC : exp ϕ(x) − dz ϕ̇(x) : (16.40)
2 β −∞

We want to have {ψ1 (x, t), ψ1 (y, t)} = 0 when x 6= y. Let’s work that out

ψ1 (x)ψ1 (y) = C 2 eA(x) eA(y) , ψ1 (y)ψ1 (x) = C 2 eA(y) eA(x) (16.41)

but remember we have eB eC = eC eB e[B,C] if [B, C] is a complex number. So now we need to work out the
commutator of the exponents. Assume y > x then
 Z y 
[A(x), A(y)] = −π ϕ(x), dw ϕ̇(w) = −iπ (16.42)
−∞

So we have ψ1 (x)ψ1 (y) = −ψ1 (y)ψ(x). So indeed the statistics match. This is a correspondence between
fermionic theory and bosonic theory, so this is also called bosonization.
What if there is a kink-kink bound state? The solution should look like
(    −1 )
2v −1 mx mut
tan u sinh √ cosh √ (16.43)
π 1 − u2 1 − u2

Now we can’t replace u → is because there is an extra i in the argument of inverse tangent and there will
not be breather solutions.

77
Quantum Field Theory III Lecture 17

17 Lecture 17
Let’s get over one dimensions and go to 2 + 1 dimensions. In one dimension we have two vacua and we
argued that if the solution goes to one vacuum in one direction and the other vacuum in the other direction
then there exists some solution. Now in two dimensions we have a circle in the infinity. Suppose we have
a complex scalar field ϕ = ϕ1 + iϕ2 = ρ(x)eiα(x) and take the Lagrangian as

1 λ 2 
L= |∂µ ϕ|2 − |ϕ| − v 2 (17.1)
2 4
The vacua must be |ϕ| = v. If we take the vacuum to be ϕ = veiβ with β constant then there is one
Goldstone boson and there is a remaining massive particle.
For each β we have a different vacuum. Now suppose we have some configuration which goes to a
outward vacuum in the infinity, then we ask whether we can continuously deform it into some vacuum
solution. But because it is radially outwards, there will be some discontinuity if we deform it into one
direction. So we expect some solitons. Let’s define the quantity
I
1
N [C] = dl · ∇α = n (17.2)

For a vacuum obviously we have N = 0 for any closed curve C. Now in the case where the configuration at
infinity is radially outwards, we have N = 1. This number is known as the winding number, or vorticity.
Now if we continuously shrink the curve C to very small, the integer must vary continuously, and the
only way is for it to remain constant. But when it shrinks to zero then N must come to zero, but that
contradicts with continuity, so we need some place where ϕ = 0 so that α is not well defined. But there are
two kinds of ϕ = 0 with α increases clockwise, or anticlockwise. These are called zeros and antizeros, and
the winding number is just the number of zeros minus the number of antizeros. Now for static solutions we
can classify them according to N . For N = 0 it is a vacuum. For N = 1 it is a vortex, and the configuration
looks like figure 17.1 (a). For N = −1 it is an antivortex and the configuration looks like figure 17.1 (b).
The visualization is obtained by drawing the components of ϕ as if they are the two components of a
vector.

(a) Vortex (b) Antivortex

Figure 17.1: Vortex and Antivortex configurations

78
Quantum Field Theory III Lecture 17

So now let’s look for solutions. From the Lagrangian we can get the equation of motion. Let’s try an
ansatz, and assume the field to be like
ϕ(x) = f (r)eiθ (17.3)
This usually does not work. But if we have some symmetry of the Lagrangian, then we should try some
ansatz with the same symmetry, then we have some hope. Usually when there is some symmetry G of L,
then we try the most general ansatz consistent with G and plug it into L and get
Z ( )
δL
S[χ] = d4 x L(χ̄) +

(δ1 χ + δ2 χ) (17.4)
δχ χ̄

where δ1 χ is G invariant and δ2 χ is not. If we choose the ansatz χ to be G invariant then the second term
is zero, so we just need the first variation to be zero. So in the above specific example we substitute it
into the Lagrangian to get L(f ). Note that the ansatz is not invariant under rotation because θ picks up
an additional constant. But it is invariant if we do a rotation and then a U (1) phase rotation. Also it is
invariant if we complex conjugate it and replace θ with −θ. So we need f (r) to be real.
The Lagrangian now is
Z  
1 0 2 1 2 λ 2 2 2
− S = 2π dr r (f ) + 2 f − (v − f ) (17.5)
2 2r 4

so the equation of motion is


d 1 λr 2 1 1 λ
− (rf 0 ) + f + (v − f 2 ) = f 00 + f 0 − 2 f − (v 2 − f 2 )f (17.6)
dr r 2 r r 2
There is no closed analytic solution. But we can see what a solution look like. In order for the energy to
be finite, it should go to a vacuum solution at infinity, so f (∞) = v. We also require f (0) = 0 because
at the origin θ is not well defined. So from the equation the asymptotic forms are f ∼ r near origin and
f ∼ v − e−mφ r near r = ∞.
Now let’s work out the energy of the solution. We have

f2
(∇ϕ)2 = (∇f )2 + (17.7)
r2
but its integration to infinity diverges logarithmically. Now there is a theorem coming to rescue, which is
Derrick’s theorem. It states the following. For scalar fields ϕa for any numbers of a and if the Lagrangian
is
1
L = Gab ∂µ ϕa ∂ µ ϕb − V (ϕ) (17.8)
2
and require V (ϕ) = 0 at its minimum without loss of generality. The energy of a static solution will just
be Z
E[ϕ] = dD x Gab ∂j ϕa ∂ j ϕb + V (ϕ) = IK [ϕ] + IV [ϕ]
 
(17.9)

A static solution must be the minimum point of the energy. Now suppose ϕ̄(x) is a solution then define
fλ (x) = ϕ̄(λx) then we can pick out one line in the configuration space by plugging this into the energy
expression
E(λ) = E[fλ (x)] (17.10)

79
Quantum Field Theory III Lecture 17

Because ϕ̄ is a solution so λ = 1 must be a stationary point, so we need



∂E
=0 (17.11)
∂λ λ=1

But we can just work out that



−D 2−D ∂E
E(λ) = λ IV [ϕ̄] + λ IK [ϕ̄], = −DIV [ϕ̄] + (2 − D)IK [ϕ̄] (17.12)
∂λ λ=1

When D = 1 we need to have IK [ϕ̄] = IV [ϕ̄] which is exactly what we have in the kink. When D = 2 we
need −2IV [ϕ̄] = 0 so the field must everywhere be at vacuum. And for D ≥ 3 we don’t have any solution.
Now let’s try to put in some more structure. Let’s try gauge theory
1 2 1
L = − Fµν + (Dµ ϕ)† (Dµ ϕ) − V (ϕ) (17.13)
4 2
and carry out the same thing with ϕ → fλ (x) = ϕ̄(λx) and Aj → gjλ (x) = λĀj (λx). The λ in front of A
is because A appears in the covariant derivative. Assume in the static case and that A0 = 0 then we call
the energy of the gauge field as IF . The energy function will be

4−D 2−D −D ∂E
E(λ) = λ IF + λ IK + λ IV , = (4 − D)IF − DIV + (2 − D)IK (17.14)
∂λ λ=1

So at D = 2 we need IF = IV to have a solution. Note that this doesn’t tell us there is a solution, just
doesn’t contradict. For D = 3 we need IF − IK − 3IV = 0. For D = 4 then we need IK = IV = 0 and we
only have vacuum solutions. For D > 4 there is no solution.
Let’s follow this lesson and introduce gauge field. The Lagrangian now is
1 2 1 λ
L = − Fµν + |Dµ ϕ|2 − (|ϕ|2 − v 2 )2 (17.15)
4 2 4
Because of the symmetry breaking the gauge field becomes massive with mass mA = ev. Under gauge
transformation we have
ϕ → eieΛ(x) ϕ, Aµ → Aµ − ∂µ Λ (17.16)
So now let’s look at the winding number (17.2). Under a gauge transformation we will get
I
1
N [C] → N [C] + dl · ∇Λ = N [C] (17.17)
2π C

if the theory is gauge invariant. Now let’s look at energy. The gradient of ϕ at r → ∞ is

Dϕ = [∇ρ + i(∇α − eA)ρ] eiα (17.18)

The minus sign is from the metric which lowers Aµ to Aµ . At infinity the first term should go to zero, and
we want the terms in the bracket to cancel. So what we want is at large r
1
A ≈ ∇α (17.19)
e

80
Quantum Field Theory III Lecture 17

so the winding number becomes


I Z
1 e
N [C] = dl · A = dS · B (17.20)
2πe C 2π
This tells us that the magnetic flux is quantized. But now we are dealing with classical theory and why is
there quantization? If we put ~ back carefully we will see that there is ~ in the coupling constant e and in
taking ~ → 0 we will recover continuity.
Now let’s try the ansatz
a(evr)
ϕ = veiθ f (evr), Aj = jk x̂k , A0 = 0 (17.21)
er
the energy then is
"  2 #
v 2 df 2 v 2 (1 − a)2 2 λv 4
Z  
2 1 da 2 2
E= d x + + f + (1 − f )
2e2 r2 dr 2 dr 2 r2 4
 0 2
(1 − a2 ) 2
Z 
2 (a ) 0 2 λ 2 2 (17.22)
= πv du u + (f ) + f + 2 (1 − f )
u2 u2 2e
πmϕ 2  
λ h mϕ i
= F 2
∼ mϕ F
2λ e 2λ
where we defined u = evr and F is of order 1 or 10 at most. So the excitation energy is much larger than
the mass of the elementary particle in the weak coupling limit.
The equations of motion are

1 0 1 0 (1 − a)2 λ
a00 − a + (1 − a)f 2 = 0, f 00 + f − f + 2 (1 − f 2 )f (17.23)
u u u2 e
with boundary conditions
f (∞) = a(∞) = 1, f (0) = a(0) = 0 (17.24)
Again there is no analytic solution. Let’s look at large distance behavior. At r → ∞ we have

1 − a ∼ e−u = e−evr = e−mA r , v − ϕ ∼ e−mϕ r (17.25)

but the later is under the assumption λ/e2 < 2.


We can try an ansatz of vorticity of n
a(evr)
ϕ = veinθ f (evr), Aj = jk x̂k , A0 = 0 (17.26)
er
and boundary conditions
f (∞) = 1, a(∞) = n, f (0) = a(0) = 0 (17.27)
This is a solution like n vortices staying together at the origin, and that is a minimum of the symmetric
configurations. Now we want to ask whether this configuration is stable. We will find if 2λ/e2 > 1, which
means mϕ > mA then the solution is unstable. And if 2λ/e2 < 1 then it is stable. Why is that? Let’s
consider two widely separated vortex solutions. Their interaction comes from the exponential tails. But
here are two kinds of exponential tails as we have shown. The vector force gives a repulsive force and the

81
Quantum Field Theory III Lecture 17

scalar gives a attractive force. The dominating force is whatever decays slower, so when mϕ < mA we have
an attractive force and stable configuration. However what if mϕ = mA ? In this case there is a solution for
any n positions of vortices and the energy is independent of these positions. If we move the vortices there
are zero modes associated with them. This is related to a superconductor. We can model a superconductor
as follows. We use |ϕ| to represent the density of Cooper pairs. We can think of a superconductor as a
nonrelativistic version of a spontaneously broken U (1) gauge theory. If we put a superconductor inside
a magnetic field the field can’t penetrate the superconductor, or they can penetrate it in flux tubes. In
superconductors the parameters are fixed by the properties of the materials. But there are two types.
Type I superconductors are when mA > mϕ and vortices attract. In type II superconductors we have on
the other hand mA < mϕ so the vortices repel and they form a lattice.

82
Quantum Field Theory III Lecture 18

18 Lecture 18
Last time we consider solitons in 2D. We need to develop a language which describes what we want more
efficiently. Let’s consider a manifold M which is the set of vacua and φ0 is a vacuum. Let the gauge group
be G and it spontaneously breaks down to H. So if φ0 is a vacuum then gφ0 is a vacuum, and hφ0 = φ0 .
In particular we have
gφ0 = ghφ0 (18.1)
So the manifold is essentially M = G/H. Let’s consider 2D, then the infinity is essentially a circle S 1 . If
we going around the circle in the soliton solution, then this gives us a loop in the space of vacua. The
argument that a vortex solution can’t be deformed continuously into a vacuum solution is that a simple
loop in the vacua manifold can’t be continuously deformed into a point.
The mathematical language which describes this is homotopy. Let’s consider a space M and a point
x0 in M .

x0

Figure 18.1: Space M with a part cut out

We define a loop as a map f (s) : [0, 1] −→ M such that f (0) = f (1) = x0 . Suppose we have two loops
f (s) and g(s) we can define the continuous deform k(s, t) where 0 ≤ s, t ≤ 1 such that k(s, 0) = f (s) and
k(s, t) = g(s). If such a continuous k exists then we say f and g are homotopic at x0 . We can define the
product of two loops f ◦ g as (
f (2s) 0 ≤ s ≤ 1/2
f ◦ g(s) = (18.2)
g(2s − 1) 1/2 ≤ s ≤ 1
We can define the identity as I(s) = x0 and the inverse of f can be found as f −1 (s) = f (1 − s), the loop
traversed inversely. In order to consider this as a group, instead of considering loops we consider homotopy
classes of loops and call [f ] the set of loops homotopic to f . Then the group operations are

[f ] ◦ [g] = [f ◦ g], [f ]−1 = [f −1 ], i = [I] (18.3)

This group is π1 (M, x0 ) which is either called the fundamental group of M at x0 , or first homotopy group
of M at x0 . If M is connected, then the group does not depend on x0 , however obviously if it has multiple

83
Quantum Field Theory III Lecture 18

connected pieces then the group will depend on x0 . If π(M, x0 ) = 1 then we say the manifold is simply
connected.
Let’s consider a few examples. Suppose we have two dimensional Euclidean space. It is connected so we
just write π1 (R2 ). Obviously it is trivial because we can shrink everything. Same for π1 (Rn ). Now suppose
M = R2 − disk as our example above. If a loop does not go around the disk then it can be shrunk into a
point. Another observation is that loops that circle the disk different number of times can’t be deformed
into each other, and their product goes around the disk the n1 + n2 times. So the group is π1 (M ) = Z.
Similarly we have π1 (S 1 ) = Z.
How about S 2 ? We can punch a hole on it and stretch it to R2 and obviously we can shrink any loop
there. So π1 (S 2 ) = 0. This argument goes for any n > 2.
Now consider the space with two holes. Let a be a loop goes around one hole once, and b the other
hole once. The loop aba−1 b−1 can’t be contracted to a point, however aa−1 bb−1 can. So the fundamental
group for this space is non-Abelian. Now we are interested in the fundamental group of Lie groups, and
for any Lie group G we have π1 (G) is Abelian. Let’s consider SU (2) and SO(3) and this is the place where
π1 becomes important. The group SU (2) is the group of 2 × 2 unitary matrices with determinant 1. We
can write the group element as

u = a0 + ia · σ, a20 + a21 + a22 + a23 = 1 (18.4)

From the constraint we know that SU (2) ∼ = S 3 so that π1 (SU (2)) = 0. Now consider SO(3). It is described
by axis of rotation n̂ and angle ϕ. But we have the additional identification (n̂, π) = (−n̂, π). We can think
of this as a 3D ball of radius π, with antipodal points identified. There are two kinds of loops. One is
just inside the ball like an ordinary loop. The other is a loop that goes to the boundary and by antipodal
identification goes to the other end of the ball and comes back. These loops can’t be deformed into each
other. However if the loop hits boundary twice we can move the two points closer and closer so they meet
and we reduce to the first case. So loops with even number of jumps are contractable and loops with odd
number of jumps is not so
π1 (SO(3)) = Z2 (18.5)
What is the relation between SO(3) and SU (2)? For (n̂, ϕ) in SO(3) we can identify the element
exp[i(n̂ · σ)ϕ/2. Now if we change ϕ → ϕ + 2π then there is no effect on rotation, but in SU (2) we
change U → −U . So the full range of rotation in SO(3) is 2π and in SU (2) is 4π. We can interpret this
using the center of the group, which is the set of elements that commute with all elements. In SU (2) the
center is {I, −I} = Z2 . Starting from SU (2) we can define an equivalence g ∼= −g and this equivalence
preserves group multiplication because −I commutes with everything. Now the group that we get from
this equivalence is just SO(3)
SU (2)/Z2 = SO(3) (18.6)
We can also think of this as S 3 with antipodal points identified, which will just be a hemisphere with
the antipodal points on the boundary identified, which is exactly the 3D ball with antipodal points on
boundary identified.
Now from any Lie algebra we can get a unique simply connected group. We call this the universal
covering group. Suppose G is the universal covering group and K the center of G. We could take G/K
and get some other group. However we can also take K as some subgroup of the center. Now G/K is
another Lie group which is not simply connected. In fact

π1 (G/K) = K (18.7)

84
Quantum Field Theory III Lecture 18

This can be seen if we consider loops from point g. For each element z in the center, the curve that
connects g to zg is a loop in the quotient group, and it is different from the curve connecting g to z 0 g. For
SO(N ) the universal covering group is Spin(N ). It is always the case that

SO(N ) = Spin(N )/Z2 (18.8)

If we look at the center of Spin(N ), then we have



Z2 ,
 N odd
Center(Spin(N )) = Z2 × Z2 , N = 4k (18.9)

Z4 , N = 4k + 2

This is because SO(N ) has nontrivial center for even N , when −I also has determinant 1. However SU (N )
has center ZN which are elements ei2πk/N I. Note that U (1) ∼= S 1 so that π1 (U (1)) = Z. The group Sp(N )
has center Z2 , E6 has center Z3 and E7 has center Z2 .
Now remember SU (2) has integer and half integer spin representations, and SO(3) only have integer
spin representations. In general the universal covering group has all the representations, and G/K has
some subset of the representations. So in SU (3) we can define some “triality” which is the number of
upper indices minus lower indices mod 3. So the representation 3 has triality 1, where 3̄ has triality −1,
and the 8 representation has triality 0. This is because SU (3) has center Z3 . For SU (N ) we can do this
and classify the representations into N subsets.
Now let’s get back to solitons. The loops beginning and ending at a point φ0 corresponds to loops in
the space of vacuum G/H. These objects are almost identical, but not quite. In our deformation there is
no reason we keep the point φ0 fixed. So we define the free homotopy. Now when we have two holes and
the fundamental group is nonabelian, then free homotopy is essentially making the group abelian. Now
we want to relate the group structure to topological charge. Now if we have two vortices in the space,
the topological charge of one plus the other is just going around the product of two loops. Here we want
abelian because we don’t want the order of going around to matter.
Consider the example for U (1) completely broken by complex ϕ. The manifold for vacua is U (1) and
π1 (M ) = Z. So the charge is just integers. Now consider SO(N ) broken into SO(N − 1) the space of vacua
is
M = SO(N )/SO(N − 1) ∼ = S N −1 (18.10)
so π1 (M ) = 0. Now suppose the group is SO(3) and we have two scalar field which are in the vector
representation of SO(3). The potential being

λϕ 2 λχ 2
V = (ϕ − vϕ2 )2 + (χ − vχ2 )2 + g(ϕ · χ)2 (18.11)
4 4
If g > 0 then ϕ is perpendicular to χ so SO(3) is completely broken. Then π1 (M ) = π1 (SO(3)) = Z2 . So
what do the solutions look like? Possible solutions look like
a(r)
ϕ = (0, 0, vϕ ), χ = (cos θ, sin θ, 0)f (r), Aj = jk x̂k (0, 0, 1) (18.12)
r
or we can have
χ = (0, 0, vχ ), ϕ = (cos θ, sin θ, 0)f (r), Aj = . . . (18.13)

85
Quantum Field Theory III Lecture 18

Now the fundamental group being Z2 means that there is one kind of solution which is topologically trivial
and deformable to a vacuum solution. This class of solution is just

a(r)
ϕ = (0, 0, vϕ ), χ = (cos 2θ, sin 2θ, 0)f (r), Aj = jk x̂k (0, 0, 1) (18.14)
r
What is the other class? Suppose vϕ  vχ , then we can think of the breaking as SO(3) to SO(2) = U (1)
and then broken by χ. But the above solution is stable because for it to deform into the vacuum we need
to deform ϕ and it involves a lot of energy. So topology isn’t everything.
Now let’s consider Weinberg-Salam model. The group is G = gSU (2) × g 0 U (1) with fields ϕ and the
vacuum is ϕ† ϕ = v 2 /2. The space of vacua is S 3 and π1 (S 3 ) = 0. So there should be no vortices. However
if g = 0 then SU (2) is global and U (1) is local, then we have vortices in U (1) section. These are called
semi-local string and they are actually stable. Now suppose g  g 0 then there is only large finite energy
barrier against decay of these strings. However what is “small” enough? It is a numerical question, and
it turns out that the “small” is smaller than observed number, so there is no such things in electroweak
theory.
Now if there is a π1 (M ) then there is obviously π2 . The groups πn (M ) are maps from S n to M . When
we consider solitons in 2D then π1 is what to look at. However when we consider solitons in 3D then π2 (M )
is what to look at. We do this by converting S 2 into a square with all sides identified. We define a “loop”
as
f (s, t) : [0, 1] × [0, 1] −→ M, f (0, t) = f (1, t) = f (s, 0) = f (s, 1) = x0 (18.15)
We define homotopy by the existence of a continuous map k(s, t, u) which connects f and g. The product
of two loops is roughly as before
(
f (s, 2t), 0 ≤ t ≤ 1/2
f ◦ g(s, t) = (18.16)
g(s, 2t − 1), 1/2 ≤ t ≤ 1

Note this can always be continuously deformed to g ◦ f , so π2 (M ) is always abelian. Similarly πn (M ) is


abelian for n > 2. Let’s state some results. In general we have

πn (S n ) = Zn , πn (S 1 ) = 0, πn (S m ) = 0 for m > n (18.17)

86
Quantum Field Theory III Lecture 19

19 Lecture 19
Last time we considered the group π2 (M ) which consists of maps from S 2 to M . We considered them as
mapping the square into M but with the sides of the square identified. Remember π1 (S 1 ) = Z and we
can visualize it as winding the circle n times. We also have π2 (S 2 ) but now it is much more difficult to
visualize. Let’s see how can we do it. Consider a unit vectors ê(r), and it is equivalent to a point on S 2 if
we are in 3D. Now let’s consider the unit vector at infinity ê(r = ∞, θ, ϕ). We define the integral
Z
1
N= dSi ijk abc êa (∂j ê)b (∂k ê)c
8π r=∞
Z (19.1)
1
= ijk dSi ê · (∂j ê) × (∂k ê)

Now if we make an infinitesimal change ê → ê + v such that ê · v = 0, and we also have ê · ∂j ê. Now the
change in N is Z
ijk
δN = dSi [2ê · (∂j ê) × (∂k v) + v · (∂j ê) × (∂k ê)] = 0 (19.2)

The second term vanishes because all three are orthogonal to ê, and the first term vanishes because it is a
total derivative, and S 2 doesn’t have a boundary. So the integral is quantized. If we take

ê = (sin θ cos nϕ, sin θ sin nϕ, cos θ) (19.3)

then the integral is just N = n. Now if we consider vortices and antivortices then N is just the number of
vortices minus the number of antivortices, so π2 (S 2 ) = Z.
We need more than this. Let G be some compact and connected Lie group, then a theorem says that
π2 (G) = 0. This is a corollary of a theorem due to Cartan. There is no easy proof for the theorem. For π1
it is easy to figure out the group from the manifold, but it is not easy for π2 . Suppose G is broken into H
by φ then the vacuum manifold is M = G/H and φ(s, t) = g(s, t)φ0 where φ0 is a point in M . Now let’s
draw loops on the sphere S 2 and let t be the distance along the loop, varying from 0 to 1, and s labeling
the loop, also varying from 0 to 1, and s = 0 and s = 1 loops are just points. We choose g(0, 0) = I. So for
the s = 0 loop we have g = I, but for the s = 1 loop it might not be g = I because the only requirement
is for g(1, 1)φ0 = φ0 , so g(s = 1) ∈ H. Therefore we have reduced the problem of finding φ(s, t) which are
elements of π2 (G/H) to finding h(s) which are loops in π1 (H).
Suppose φ1 (s, t) and φ2 (s, t) correspond to the same h(s), then g3 (s, t) = g2 (s, t)g1−1 (s, t) has I on all
boundary, so is an element of π2 (G). But π2 (G) is trivial, so g3 is homotopic to identity, and φ1 is homotopic
to φ2 . Therefore the elements of π2 (G/H) and π1 (H) are in 1-1 correspondence if G is simply-connected.
But for any group we can consider their universal cover. So the problem is solved, and

π2 (G/H) = π1 (H) (19.4)

Now consider SU (2) broken into U (1) by a particle φ. Then π2 (G/H) = π1 (U (1)) = Z. We can also say
that at infinity φ should have fixed magnitude but varying direction so this is equivalent to π2 (S 2 ) = Z. If
we identify the U (1) as the electromagnetic U (1) then the solitons we get here will have magnetic charge.
Magnetic charges are defined analogously as electric charge such that
QM r̂ r̂
B= =g 2 (19.5)
4π r2 r

87
Quantum Field Theory III Lecture 19

Now if we have magnetic charge, then we can’t write it as B = ∇ × A. But we know from quantum
mechanics that A are physical objects, so we can’t throw them away. Let’s define
g
Ai = −ij3 r̂j (19.6)
z+r
where in spherical coordinates we have

Aϕ = −g(cos θ − 1) (19.7)

Now we have problem with z = −r so we have problem with the minus z-axis. We call this vector potential
AI . Except for this singularity the curl of this potential is the monopole magnetic field. We can also define

AII
ϕ = −g(cos θ + 1) (19.8)

and now the singularity is on the positive z-axis. This potential also gives the magnetic monopole field,
and these are related by a gauge transformation

AIϕ − AII
ϕ = (2gφ) (19.9)
∂ϕ
The gauge transformation is singular on the whole z-axis, and with choice of coordinates we can make it
lie on any line, or even any curve. This singularity is called the Dirac string. Now in classical physics A is
not an observable, and we require that this singularity do not matter. Under a gauge transformation the
wave function changes like H
ψ → ψeiq dl·A (19.10)
Now the integral can go around nothing or around a string, depending on our choice of the position of the
string. So we require that around a string the integral should give a multiple of 2π, so

iqg(2π)(1 − cos θ) = 2nπ (19.11)

So we need to require qg = n/2. This needs to be true for any q and any g, so this can only be true if we
have some quantization condition on electric and magnetic charges

q = kqmin , g = mgmin (19.12)

and qmin gmin = 1/2. Now there are exceptions. If we have Dyons which are particles with both electric
and magnetic charges, and for these the constraint is

q1 g2 − q2 g1 = n/2 (19.13)

Now if we consider quarks, then we need to consider if there is a counterpart for “magnetic color charge”.
Then we need
qg + qSU (3) gSU (3) = n/2 (19.14)
These are the Dirac quantization condition.
There is another derivation. Consider the force law
dv r̂
m = qv × B = gqv × 2 (19.15)
dt r

88
Quantum Field Theory III Lecture 19

Now let’s look at angular momentum


d dr̂
(r × mv) = r × F = gq (19.16)
dt dt
So angular momentum is not conserved, despite spherical symmetry. However if we define the angular
momentum as
L = r × mv − gqr̂ (19.17)
then this is conserved, which we should have if we have spherical symmetry. But this has some extra piece
which is along r̂, so the magnitude of angular momentum must be

|L| ≥ qg (19.18)

But we know the quantization of angular momentum


n~ n
L · r̂ = =⇒ gq = (19.19)
2 2
Now there is a third derivation which makes the mathematicians happier. Suppose our monopole is at
the origin and we divide our space into two overlapping regions I and II. Now we can define our potential
in region I to be AI and in region II to be AII and they are both nonsingular, and we require that on the
overlap AI and AII are gauge-equivalent. This is analogous to defining a manifold where we have several
coordinate patches and on overlap we require the appropriate transformation law. So on overlap we have

AI − AII = ∇Λ = 2gϕ (19.20)

But we also need eiqΛ to be well defined, or single-valued, so we have the quantization condition

4πgq = 2nπ =⇒ gq = n/2 (19.21)

Let’s come back to field theory. Our Lagrangian here is


1 1 2 λ
L = (Dϕi )2 − Fµν − (ϕ2 − v 2 )2 (19.22)
2 4 4
and because
√ of√symmetry breaking we have massless photon, massive W ± with mass ev and massive Higgs
m = 2λv = 2µ. In the limit r → ∞ we should have |ϕ| → v and Dϕ̂ → 0. So if we take the integral
and it should be Z
1
ijk dSi ϕ̂ · Dj ϕ̂ × Dk ϕ̂ = 0 (19.23)

Now we can put in the covariant derivative and write
Z
1
ijk dSi −eϕ̂ · ∂j ϕ̂ × (Ak × ϕ̂) − (j ↔ k) + e2 ϕ̂ · (Aj × ϕ̂) × (Ak × ϕ̂)
 
0=N+ (19.24)

The first two terms give −eϕ̂ · (∂j Ak − ∂k Aj ), whereas the third term gives

e2 ϕ̂ · [Ak (ϕ · Aj × ϕ̇) − ϕ̂(A × ϕ̂) · Ak ] = −e2 ϕ̂ · (Aj × Ak ) (19.25)

So in the end we have Z Z


e e
0=N− ijk dSi ϕ̂ · Fjk = N − dSi ϕ̂ · Bi (19.26)
8π 4π

89
Quantum Field Theory III Lecture 19

Here Bi is not the magnetic field, but at large distances its component along the unbroken ϕ̂ is indeed the
magnetic field. So the integral is magnetic flux, and we have quantization condition

QM = N, eg = N (19.27)
e
Here we have integer instead of a half integer because we have T = 1 representation as our Higgs field. If
we used a isospin doublet then we will find eg/2 = N/2 which is the same thing as we had.
Now let’s look at solutions. We have the Hedgehog solution which looks like
 
a a a n 1 − u(r)
ϕ = r̂ h(r), Ai = ian r̂ , Aa0 = 0 (19.28)
er

and the boundary conditions are

h(0) = 0, h(∞) = v, u(0) = 1, u(∞) = 0 (19.29)

Plug that into the Lagrangian we have the equations of motion

2 2u2 h u(u2 − 1)
h00 + h0 − 2 + λ(v 2 − h2 )h, 0 = u00 − − e2 uh (19.30)
r r r2
Plug this ansatz into the magnetic field we get

1 − u2 u0
 
Bia = r̂i r̂a 2
− (δia − r̂a r̂i ) (19.31)
er er

So at infinity it is a monopole field with magnetic charge 4π/e. To get the mass of this, let’s do an
argument. Say the particle is localized in a core of radius R. The mass is the sum of energy in the core
and energy due to Coulomb potential in the exterior region. Let’s say the energy density inside the core is
constant so
4πR3
Z
1 4π 2 4 3 2π
M= ρcore + d3 r B 2 ∼ e v R + 2 (19.32)
3 2 3 e R
Now we can adjust the core size to minimize the energy, and setting ∂E/∂R = 0 will give us
1 1 1
R4 ∼ , R∼ = (19.33)
e4 v 4 ev mW
Now plug this back into the mass we have
4π 4π
M∼ mW ∼ v (19.34)
e2 e
Now if we calculate it instead of approximating, we have the actual value

M= vf (λ/e2 ) (19.35)
e

90
Quantum Field Theory III Lecture 20

20 Lecture 20
Let’s come back the ansatz we wrote down last time
 
a a 1 − u(r)
ϕ = r̂ h(r), Aai = iam r̂ m
, A0 = 0 (20.1)
er

This is the Hedgehog ansatz. Suppose we twist the field everywhere to point to one direction, then
there will be a singular line where we can’t make this rotation continuous. But nevertheless let’s do the
transformation and we will get
r̂ 1
ϕa = δ a3 h(r), A3i = ai = −iy3 (20.2)
er+z
We can write
1 u(r)
Wi = √ (A1i + iA2i ) = vi (20.3)
2 er
where
i h i 1 h i i
v1 = − √ 1 − eiφ cos φ(1 − cos θ) , v2 = √ 1 + eiφ sin φ(1 − cos θ) , v3 = √ eiφ sin θ (20.4)
2 2 2

Now ϕ3 is the neutral Higgs, Ai the E-M field and the Wi are the charged vector boson. However there is
still a freedom
(Wi1 + iWi2 ) → eiα (Wi1 + iWi2 ) (20.5)
So α can be thought of as a collective coordinate. Suppose α = ωt. If we have some charged scalar particle
ψ, then the U (1) charge is just Z
QU (1) = i d3 x ψ̇ ∗ ψ − ψ ∗ ψ̇ (20.6)

Similar is true for Wi , so they have U (1) charge. The fact that α = ωt combined with Gauss’s law tells
us that A0 6= 0. We can make a time-dependent gauge transformation to change α into a constant, or
equivalently make the solution into a static one. Or, as historically was done by Julia and Zee, we can put
in an ansatz
Aa0 = r̂a j(r) (20.7)
Now this gives us something with both electric charge and magnetic charge, therefore is called Dyon.
Now can we get solutions with higher magnetic charge QM ? For vortices it is easy, we just replaced eiθ
by einθ and just has n times the magnetic charge. However here in an SU (2) theory we can’t even write
down a spherically symmetric ansatz. Now forget about fields for the moment and consider expansions in
Jackson. The spin 0 field can just be expanded in Ylm . The spin 1 field should be expanded like r̂Ylm ,
∇Ylm , etc. We need to expand in vector spherical harmonics Ylm . We introduce the so-called monopole
harmonics Zjm . The smallest angular momentum j is the product of electric and magnetic charge k = qg.
For spin 1 we have vector monopole harmonics Zjm where the smallest angular momentum is j = k − 1.
So if we want to write down a spherically symmetric configuration for the vector field, we can only have
qg = 1 which is just the solution we wrote down. For higher QM there is no spherically symmetric solution
because the expansion starts at j = 1.

91
Quantum Field Theory III Lecture 20

Let’s consider SU (3) with φa in the adjoint representation. Let the vacuum expectation value be
 
2b
hφi =  −b  (20.8)
−b

so the group is broken into SU (2)×U (1). So we have π2 (G/H) = π1 (H) = Z. So there should be monopole
solutions. Now suppose ϕ sits in the symmetric representation 6, and ϕ transforms as ϕ → U ϕU T . The
vacuum expectation value is  
a
hϕi =  a  (20.9)
a
Now what is the unbroken group? It is generated by the generators λ2 , λ5 and λ7
     
0 −i 0 0 0 i 0 0 0
λ2 =  i 0 0 , λ5 =  0 0 0 , λ7 = 0 0 −i (20.10)
0 0 0 −i 0 0 0 i 0

But these are just generators of SO(3), not SU (2). So this breaks the group into SO(3). But π2 (G/H) =
π1 (H) = Z2 . So now we can have magnetic monopole, but the monopole and anti-monopole are equivalent.
If we also introduce ψ as a triplet representation with vacuum expectation (0, 0, b), then the only thing
leaving both invariant is U (1), and the gauge group breaks into SO(2) = U (1). Now the second homotopy
group is Z and we have monopoles have any charge.
But supose a  b. The ϕ first breaks the group down to SO(3), and we have Z2 monopoles with mass

M∼ a (20.11)
g
Then we have ψ breaking the rest of the group into U (1). If we just consider low energy physics and do
not know about SO(3) or SU (3), then the Z2 monopole just becomes the n = 1 monopole. But we also
have n = 2 monopole, and it is due to the twisting of ψ with mass

M2 ∼ bM (20.12)
g
The higher charge monopoles are lighter than the lowest charge monopole.
Now let’s consider SU (2) × U (1) → U (1). It is broken by the Higgs doublet φ = φφ12 with φ† φ = v 2 /2.


The vacuum structure is π2 (G/H) = π2 (S 3 ) = 0 so Weinberg-Salam model does not have monopoles. But
let’s consider some grand unified theory with gauge group G. We want G to be a simple group, and we are
also free to assume it being the covering group which is simply connected. We know that finally it breaks
down to SU (3) × U (1), so the vacuum structure is

π2 (G/H) = π1 (H) = π1 (SU (3) × U (1)) = Z (20.13)

So any grand unified theory contains monopoles. What are the masses? Suppose the unified group is
SU (5). The group first breaks into SU (3) × SU (2) × U (1) by field Φ, and the monopole has mass

M∼ hΦi ∼ 1015 GeV (20.14)
g

92
Quantum Field Theory III Lecture 20

If the unified group is Spin(10) then let’s say it first breaks to Spin(6)×Spin(4)/Z2 . The second homotopy
group for the vacuum is going to be Z2 . So we get Z2 monopoles with large mass scale as above in the
SU (5) case. But then this group must break down to SU (3) × U (1). Suppose it achieves this by breaking
into SU (3) × U (1) × SO(4) by field χ. Then the Z monopole due to this breaking will have charge n.
Similar to the discussion above we have an n = 2 charged monopole

M2 ∼ hχi  M1 (20.15)
g
Now physics is an experimental science, and the question is whether we have seen these particles. The
simple answer is no. But are we supposed to see these monopoles? How are we going to see these objects?
The most possible place to see them is in the early universe. We assume the universe is homogeneous and
isotropic, and describe it using the Roberson-Walker metric

dr2
 
2 2 2 2 2 2 2

ds = dt − a (t) + r dθ + sin θdφ (20.16)
1 − kr2

where if k = 1 the slices of constant time are S 3 and we have closed universe. If k = 0 then slices are R3
and we have flat universe. For k = 1 the slices are hyperbloid and we have open universe. The physical
distance between two comoving objects is

`phy = a(t)`coord (20.17)

We define the Hubble constant H = ȧ/a. If we plug the metric into the Einstein equations we have
 2
ȧ 8π k
= 2
ρ− 2 (20.18)
a 3Mp a

where ρ is the energy density. For matter we assume P = 0 and it goes like ρ ∼ 1/a3 . For radiation or
photons we have P = ρ/3 and it goes like ρ ∼ 1/a4 . For dark energy it is like a cosmological constant and
it goes like ρ ∼ constant. We are now at dark energy dominated regime, and in early universe we were
once in the radiation dominated regime. It is in the radiation dominated regime that we want to look at.
Approximately we know that k ≈ 0. So in early universe we just take k = 0 and radiation domination and

ρrad ∼ nT 4 (20.19)

where n is the number of “massless” degrees of freedom. The entropy density S is S ∼ nT 3 . When the
universe is expanding, we assume it is an adiabatic process, which means that a3 S is a constant, or aT is
a stepwise constant. If we plug into the Friedman equation the radiation dominance we get a ∼ t1/2 , and
from adiabaticity we have
Mp 1/2
 
T ∼ (20.20)
t
Let’s go back to the metric. Suppose we send out a signal at r = 0 and t = t0 . We ask where it will be
at a later time. For light signal we have ds2 = 0 so dr/dt = 1/a. The physical distance it travels is
Z t
1
a(t)r(t) = a(t) dt0 (20.21)
t0 a(t0 )

93
Quantum Field Theory III Lecture 20

The horizon distance, which is the furthest distance anything can travel since t = 0 is just
Z t
1 Mp
dH (t) = a(t) dt0 0 ∼ 2t ∼ 2 (20.22)
0 a(t ) T

Let’s ask what happens when the world becomes hotter. Consider a ferromagnet. For zero temperature
the rotational symmetry is spontaneously broken by the ground state, but if we heat it up the symmetry is
restored. Same happens for a crystal which breaks translational symmetry. If we heat it up it will melt and
we restore translational symmetry. Now consider a field theory at T = 0 with some spontaneous symmetry
potential V (φ). Symmetry breaking gives the masses mψ ∼ Gφ and mA ∼ gφ. For T = 0 the equilibrium
state is the state with lowest energy, but for finite T the equilibrium state is the state with lowest free
energy, and we need to minimize F . Let’s consider uniform φ and evaluate F = Veff (φ, T ). For ideal gass
with mass M  T then the free energy is

F ∼ −T 4 + M 2 T 2 + · · · ∼ −T 4 + G2 φ2 T 2 + . . . (20.23)

Now the condition M  T is equivalent to φ being small. Near φ = 0 the effective potential is
 2 
µ 1 2
4
Veff = −T + − + σT φ2 + . . . , σ ∼ g 2 + G2 + λ (20.24)
2 2

So when T becomes big the sign of the φ2 term becomes positive and we recovered the symmetry. Now
this is only to second order, and it was calculated in detail to higher orders by Dolan and Jackiw in 1974.
In general we have two possibilities. When we lower the energy we can go from an ordinary potential to a
spontaneous symmetry breaking potential across Tc , so we have a second order phase transition. However
we could still have a minimum at φ = 0 at Tc and we have a first order phase transition, which is like the
super-cooling of water.
In our universe, when the temperature was at the order of T ∼ 102 MeV we have chiral symmetry
and the quarks are not confined. When the temperature was at 102 GeV the electroweak symmetry was
restored. At higher temperature higher symmetry could be restored, so at 1016 GeV we could have some
GUT symmetry restored.

94
Quantum Field Theory III Lecture 21

21 Lecture 21
Last time we talked about that long ago in the universe when the temperature is at the GUT scale
Tc ∼ 1016 GeV the symmetry was restored. It is necessary to have symmetry breaking to have monopoles.
Now let’s say the symmetry group is broken from G to H and we have monopoles in G/H for T < Tc .
Symmetry breaking only says they are possible, but will they happen? In G we have φ = 0 whereas in H
we have φ 6= 0. Now when we cool a ferromagnet down from above the Curie temperature, we don’t get
a giant ferromagnet, but we get domains, with different vacuum in each domain. Same thing happens for
the universe. Let’s say there is some characteristic distance between domain walls which is ξ. This is a
hard question. If we cool a ferromagnet very slowly we should be able to get fairly large domains. So we
need to know the rate at which the universe was cooling. However we have a causality bound from the
horizon distance. Simply from causality we should have
Mp
ξ < dH (Tc ) ∼ (21.1)
Tc2

So what will happen? If we break discrete symmetry, this will give us domain walls. If we have
π1 (G/H) 6= 0 and think of a complex scalar field ϕ. The vacuum expectation value of ϕ in various domains
will start to smooth out, unless there is a kind of defect, and we have strings, or vortices. Similarly if
π2 (G/H) 6= 0 then we will have monopoles. The density of strings will be inversely proportional to the
dimension of domains, times the probability of having a configuration like a vortex, which is not terribly
small, say about 0.1. The density would be about
2
Tc2

1 1
P ∼ (21.2)
ξ2 10 Mp

and similar for density of monopoles, except that the power is 3.


Now is this true? If we have a domain wall what will happen for the matter around it? In field theory
the energy-momentum tensor for a scalar field is

Tµν = ∂µ ϕ∂ν ϕ − gµν L(ϕ) (21.3)

Now let’s say the domain wall is along the x-y plane. Then we have
 
1 1 02 1
T00 = ϕ02 + V, T11 = T22 = − ϕ + V , T33 = ϕ02 − V (21.4)
2 2 2

So P ∼ −ρ and there is negative pressure. So you are pushed away from the wall, or the things accelerate
away from the wall. However if the energy density of walls dominates now, then it clearly contradicts with
observation. We need ρwall < ρmatter , ρΛ . If σ is the energy density of the wall then we should have

σ . (30 MeV)3 (21.5)

But σ ∼ m3 /λ so the walls come from some low energy physics which we have not discovered yet, but it
is unlikely.
Now how about strings? For walls we have two directions of negative pressure, but for string we only
have one. Away from the string the space is flat, but near it the space looks like a cone. This will give
gravitational lensing. Strings could exist, but they are constrained to be very scarce.

95
Quantum Field Theory III Lecture 21

What about monopoles? Their initial density should be


3
1 Tc2

nint ∼ (21.6)
10 Mp
Our argument last time provides compelling evidence that they exist. But monopoles and antimonopoles
can annihilate. For magnetic monopoles in very high temperature and dense plasma they will loose energy
by interaction with the charged particles, and they become captured and annihilate. When the temperature
becomes less they still can loose energy by Bremsstralung-like process. However, we have expansion of the
universe. We can account for it by defining r = nM /S. What we get is
(
nM rinit ∼ (Tc2 /Mp )3
rnow = = min (21.7)
S r∗ ∼ 10−11 (Mmon /1017 GeV)

From the monopole mass we should have r . 10−26 m17 . We can also have bound from the flux. It can be
done from the magnetic field in galaxies and Parker did it to get r . 10−26 . Also we have direct searches
which yield r . 10−27 . These are inconsistent with the above limits. What are the possibilities? First is
that there is no GUT. But GUT is too nice to throw away. The second possibility is inflation. If monopoles
have initial density like r ∼ 10−12 then it will be diluted by a factor of inflation. And it is also diluted by
reheating. This is originally why Guth came up with the idea of inflation.
Let’s come back to field theory. Consider the theory where SU (2) → U (1) with some Lagrangian
1 a2 1 µ2 λ λv 2
L = − Fµν + (Dµ ϕa )2 + ϕa2 − (ϕ2 )2 − (21.8)
4 2 2 4 4
Remember we put in the ansatz and got equations on h and u which can’t be solved analytically. What
we want to do is take λ → 0 and µ2 → 0 but fix v 2 = µ2 /λ. In this case we can solve the equations and get
1 evr
h = v coth(evr) , u= (21.9)
er sinh(evr)
Now
q the mass can be analytically solved to be M = 4πv/e = QM v. For Dyon mass we will have M =
Q2M + Q2E v. This limit is called the BPS limit.
Let’s look at the energy in the static case and A0 = 0
Z   Z  
3 1 a 2 1 a 2 3 1 a a 2 a a
E= d x (F ) + (Di ϕ ) = d x (B − Di ϕ ) + Bi Di ϕ (21.10)
4 ij 2 2 i
Now the second term can be integrated by parts and we can get
Z Z
d x Bi · Di ϕ = d3 x [∂i (Bi · ϕ) − (Di Bi ) · ϕ]
3
(21.11)

But the second term is zero because Di Bi = 0. The first term will become a boundary term which can be
evaluated as vQM . So the energy will become
Z  
3 1 a a 2
E = vQM + d x (B − Di ϕ ) (21.12)
2 i
Now minimizing this energy we will get a solution, and obviously we have a solution when Bia = Di ϕa .
This reduces the equation tremendously.
Let’s look at the particles in BPS limit

96
Quantum Field Theory III Lecture 21

1. Photons γ have m = 0, QE = QM = 0

2. Higgs field φ has m = 0, QE = QM = 0

3. Gauge boson W ± have m = ev, QE = ±e, and QM = 0

4. Monopoles have m = 4πv/e, QE = 0 and QM = ±4π/e

This suggests we have some duality between a theory with elementary W and monopole solitons, and a
theory with elementary monopoles and soliton W ’s. But there is a problem of spin. The W have spin 1
but our monopoles are spherically symmetric. But let’s say we add a triplet spinor field ψ. We expand it
in normal modes, and for any modes with ω > 0 we have modes with ω < 0 which are just particle and
antiparticle modes. However there are modes with ω = 0. With bosons we know this mode corresponds
to some collective coordinate. But what does it do here? If we call the creation operator of the zero mode
by c† , then |0i and c† |0i are degenerate. Now if we have two zero modes then there are 4 different states,
and two have spin 0 and two have spin 1/2. But this is not quite correct yet because we need spin 1. Now
if we add two spin-1/2 fermions then we have 4 zero modes in total and 16 degenerate vacuum states. We
have 3 spin 1 states, 8 spin 1/2 states, and 5 spin 0 states.
Now let’s count. The elementary particles in our theory are ψ1 , ψ2 which are spin-1/2 particles. We
have W ± , γ which are spin 1 bosons. Then there are 6 spinless bosons. This actually fits into the N = 4
super-Yang-Mills theory where there is Aµ , ψ1 , ψ2 and 6 scalar fields. And it is believed in this theory
that the above duality is true. In the SUSY theory if we demand that some supersymmetry is preserved,
then we will get the same equation B = Dϕ.

21.1 Instantons
Let’s come to instantons. Let’s consider the vacuum of Yang-Mills theory. We write
1
Aµ = Aaµ T a , tr(T a T b ) = δab (21.13)
2
The Lagrangian is then
1 2
L = − trFµν (21.14)
2
Let’s choose a gauge where A0 = 0. The Lagrangian will then look like
1
L = trȦ2j − trFij2 (21.15)
2
However this gives us three “Maxwell” equations, but not Gauss’s law. So we need to impose that Di Ȧi = 0.
Now we ask what is the vacuum? The classical vacuum is where Fij = 0. This doesn’t mean that Aj = 0.
This means that
i
Aj = − G−1 ∂j G (21.16)
g
where G is some element of the gauge group. Assume that G → I when r → ∞. Note that G(x) is a map
from R3 to the gauge group. At infinity we have a boundary like S 3 , and we can classify the theory using
the group π3 (G). In the case of SU (2) we have

π3 (SU (2)) = π3 (S 3 ) = Z (21.17)

97
Quantum Field Theory III Lecture 21

In fact this holds for any compact Lie group. Now we define something like the winding number
Z
1
d3 x tr G−1 ∂i G G−1 ∂j G G−1 ∂k G

N [G] = 2
ijk (21.18)
24π
and this is an integer. We can check that

G(x) = eir̂a ·σa f (r) , f (0) = −π, f (∞) = 0 (21.19)

will have number equation to 1. We also have

N [G1 G2 ] = N [G1 ] + N [G2 ] (21.20)

We define the current


g 2 µναβ
 
µ 2ig
jA = 2 tr Aν ∂α Aβ + Aν Aα Aβ
8π 3
(21.21)
g 2 µναβ
 
2ig
=  tr Aν Fαβ − Aν Aα Aβ
16π 2 3

The divergence of this current is


g2
µ
∂µ jA = trFµν F̃ µν (21.22)
16π 2
which is what occurs in anomalies. Now that we have a current we can as well have a charge which is

g 2 ijk
Z Z  
3 0 3 2ig
QA = d xjA =  d x tr Ai Fjk − Ai Aj Ak (21.23)
16π 2 3

Now if A is pure gauge then we have


Z  
1 ijk 2
3
tr G−1 ∂i G G−1 ∂j G G−1 ∂k G
 
QA =  d x (21.24)
16π 2 3

and this is just the winding number of G! Now if we integrate the divergence of the current from G1 to
G2 then we have

g2
Z Z
4 µ
d x ∂µ jA = Q(t2 ) − Q(t1 ) =⇒ d4 x trFµν F̃ µν = N [G2 ] − N [G1 ] (21.25)
16π 2

98
Quantum Field Theory III Lecture 22

22 Lecture 22
Last time we looked at A0 = 0 gauge and looked at classical vacuum. In vacuum we have Fµν = 0 but
that doesn’t mean Aµ must be zero. It must be a pure gauge

i
Aj = − G−1 ∂j G (22.1)
g

We require that G → I as x → ∞ so the vacua correspond to elements of π3 (G/H) = Z. The generalization


of winding number is Z
1
d3 x tr G−1 ∂i G G−1 ∂j G G−1 ∂k G

N [G] = 2
ijk (22.2)
24π
We defined the current
µ
jA ∼ tr (F A − AAA) (22.3)
This current is not gauge invariant, but the divergence

µ g2
∂µ j A = Fµν F̃ µν (22.4)
16π 2
and it is gauge invariant. If the field strength vanishes, then the charge is just
Z
QA = d3 x jA 0
= N [G] (22.5)

If Aµ corresponds to a pure gauge G1 at time t1 and G2 at time t2 , then integrating the divergence over
from t1 to t2 will just give the difference between winding number.
Let’s think about quantum mechanics. One problem with A0 = 0 is that we need to impose Gauss’s
law J (x) = Dj (∂0 Aj ) = 0. However from the other equations of motion we have ∂0 J (x) = 0. Now let’s
look at it quantum mechanically. Suppose we have some infinitesimal gauge transformation Λ(x) which go
to 0 as r → ∞. Let’s look at the quantity
Z Z
d x Λ(x)J (x) = − d3 x (Dj Λ(x))Ȧj
3
(22.6)

we are allowed to do integration by parts because of the boundary condition. Now Ȧj is just the momentum
Πj , so Z Z
d3 x Λ(x)J (x) = − d3 x (Dj Λ(x)) Πj (22.7)

Therefore we have [Ak , JΛ ] ∼ Dk Λ(x). This kind of gauge transformation will not change the winding
number because it goes to zero at infinity. Requiring that J vanish means that the value does not does not
depend on the position inside one vacuum with definite N . So we can define vacuum states |ni inside the
classical space of vacua N = n. Classically we can’t go from one vacuum to another, but we can quantum
mechanically.
Let’s define a transformation T |ψi = |ψ 0 i. We define the wave function of ψ 0 to be

ψ 0 [Aµ (x)] = ψ[Agµ (x)] (22.8)

99
Quantum Field Theory III Lecture 22

where g is the group element which takes N = 1 vacuum to the N = 2 vacuum. However the Hamiltonian
is gauge-invariant, so T commutes with the Hamiltonian, and we should be able to get energy eigenstates
which are also eigenstates of T . Let’s define a vacuum state

X
|θi = einθ |ni (22.9)
n=−∞

Then the operator T acting on this state is just

T |θi = eiθ |θi (22.10)

Note also that θ is periodic with period 2π. Let’s find the amplitude of going from |θi to |θ0 i. It is just

0 −iHt X imθ0 −inθ
−iHt
θ e θ = e e m e n
m,n
X 0
(22.11)
eim(θ −θ) [ei(m−n)θ m e−iHt n ]


=
m,n

Now the Hamiltonian is gauge invariant, so the term in square bracket can only depend on m − n, so we
can write

0 −iHt X imθ0 X −ikθ
−iHt X
0 = 2πδ(θ − θ0 ) eikθ k e−iHt 0


θ e θ = e e k e (22.12)
m k k

So if we start with a state with some θ then we can’t get to a state with another θ. Now let’s look at the
matrix element, then we can write it as a path integral
Z
0 = [dAµ ]ei d4 x L

−iHt R
k e (22.13)

We can absorb the factor eikθ into the path integral by adding a term to the Lagrangian

θg 2
∆L = trFµν F̃ µν (22.14)
16π 2
Note that F F̃ is a total divergence, and classically total divergence doesn’t contribute anything.
Why do we choose A0 = 0 gauge which does not fix the potential? For example if we choose the axial
gauge A3 = 0 then classically Fµν = 0 will fix Aµ = 0 and there is only one vacuum. But now even if there
is ambiguity in the vacuum state, gauge transformation should not change the physics. In that case we
can still define some transformation which takes the one point back to itself, and we can get similar results
as here. Let’s consider the classical Lagrangian
I 2
L= α̇ − g cos α (22.15)
2
This is just the Lagrangian of a pendulum, or a periodic potential. The pendulum picture is the axial
gauge picture, and the periodic potential picture is the A0 = 0 gauge picture. Now if we add a term θα̇/2π,
then classically it won’t make any difference, but quantum mechanically it will. Let’s just consider
I 2 θ
L= α̇ + α̇ (22.16)
2 2π

100
Quantum Field Theory III Lecture 22

then the momentum is


θ 2
 
θ 1
p = I α̇ + , En = n− (22.17)
2π 2I 2π
So in quantum mechanics this factor definitely has an effect.
Now let’s come back at field theory. Let’s look at how to tunnel from one vacuum to another in quantum
mechanics. If the potential has two minima at q0 and qf , then WKB tells us that the tunnelling amplitude
is R qf √
T ∼ e−J = e− q0 dq 2(V (q)−E) (22.18)
If we have many degrees of freedom. Suppose q0 is surround by a potential barrier and the tunnelling
amplitude for getting outside is calculated by a prescription
1. Calculate Z p
J= ds 2[V (q(s)) − V (q0 )] (22.19)

along all the paths.

2. Find the path that minimizes J and that path gives the correct exponential factor. All the other
paths are exponentially suppressed.
R p
Now recall minimizing the quantity ds R2(E p − V ) is equivalent to minimizing the action, and if we
screw up the signs, minimizing the quantity ds 2(V − E) is equivalent to minimizing the Euclideanized
action. So to find the classical tunnelling path we just need to solve the Euclidean equations of motion.
Now we have
s "   #
dqj 2 1 dq 2
Z Z  Z
p ds
ds 2(V − E) = dτ = dτ + (V − E) = SE (22.20)
dτ dτ 2 dτ

The solution to this problem is called an instanton if the initial point has energy E = 0. A solution
describes the classical path in the Euclidean space, which translates to a time-dependent configuration
connecting two vacua.
Now let’s solve the equation. The Euclidean action for us is
Z
1
SE = d4 x trFµν 2
(22.21)
2
with fixed change in winding number
16π 2
Z
d4 x F F̃ = k (22.22)
g2
The Euclidean field equation is Da Fab = 0. However we can rewrite the action as
Z   Z  
1 2 1 2 1 1
SE = tr d4 x Fµν + F̃µν = tr d4 x (Fab ∓ F̃ab )2 ± Fab F̃ab
4 4 4 2
2 Z    (22.23)
8π 1  2
= ± 2 k + tr d4 x Fab ∓ F̃ab
g 4

Now if we take k > 0 then we take the upper sign, and F = F̃ is definitely a solution because SE takes a
minimum there. Similarly if k < 0 then we take the lower sign, and the solution is anti-selfdual, F = −F̃ .

101
Quantum Field Theory III Lecture 22

Consider gauge group SU (2) and define


a a a a
ηij = η̄ij = aij , ηi4 = −η̄i4 = δai (22.24)
then η is self-dual and η̄ is anti-selfdual. If we try
Aaµ = ηµν
a
xν f (x2 ) (22.25)
then we find a solution which is
1 a (x − z)ν
Aaµ = − ηµν (22.26)
g (x − z)2 + λ2
where z can be interpreted as the position of the instanton, and λ can be thought of the “size” of the
instanton. At large distances Aµ falls like 1/x. However, in fact
a 2 a λ2 1
Fµν = ηµν 2 2 2
∼ 4 (22.27)
g [(x − z) + λ ] x
However, ‘t Hooft tried something else
1 a
Aaµ = η̄ ∂ν ln ρ(x) (22.28)
g µν
If we plug this in, we will get ρ/ρ = 0 and the solution to this is
k
X λ2j
ρ=1+ (22.29)
(x − zj )2
j=1

Again we can think of z and λ as positions and scales of k instantons. In particular if k = 1 then we can
get
−2λ2 xν
 
a 1 a 1
Aµ = η̄µν 2 2 2
∼ 3 (22.30)
g x (x + λ ) x
and the integral for winding number comes from the pole at origin.
In general there are 8k − 3 parameter solutions, where the 8 comes from 4 position parameters, 1 scale
parameter and 3 parameters for SU (2) orientation. Any other group can be considered by embedding
SU (2) into that group.
Now we have instantons with an amplitude ∼ exp −8π/g 2 , but we know couplings run, so which


g should we use? Also we need to sum over the ways of tunnelling, and this is summing over all the
parameters for the instanton. We also want to get the leading corrections from the stationary points.
Again ‘t Hooft carried out the calculation and found that the tunnelling factor of instantons should be
corrected to
2 2 d4 zdλ 1 −8π2 /g2 −22 ln µ0 λ/3
e−8π /g → e e (22.31)
λ5 g 8
This tells us how the coupling runs
8π 2
 
2 1 1 22
→ 8π − ln(µ0 λ) (22.32)
g2 g 2 8π 2 3
Now the λ part of the factor comes from the log factor and the 1/λ5 factor
1
dλ 5 e22 ln λ/3 = dλ λ22/3−5 (22.33)
λ
So for λ → 0 which is short-distance and high energy limit, and instantons don’t have any effect. In the
limit λ → ∞ the integral seems to diverge, and we can’t calculate it.

102
Quantum Field Theory III Lecture 23

23 Lecture 23
Remember we had the additional term in the Lagrangian
θg 2
∆L = trFµν F̃ µν (23.1)
16π 2
Note this term violates parity and violates time-reversal invariance. If we believe that CP T is conserved,
it means that this violates CP , which is a good thing because we know experimentally that CP is violated.
We know this from the amplitude of KL → π + π − and KS → π + π − . We also know from the neutron
electric dipole moment. The natural size of the dipole moment is e · (10−13 cm). However experimentally
it is less than 10−12 of this number. This suppression is due to CP violation and also because it is small
in the first place. So in addition to the CKM matrix we still need θ to be small, such that θ . 10−9 . This
used to be called the “strong CP -problem” and used to be an important fine-tuning problems, until the
occur of cosmological constant problem.
One solution of the problem is due to Peccei and Quinn. We note that fermions also contribute to this
θ. The chiral rotation on fermions will correct θ to θeff . The Peccei-Quinn symmetry is slightly broken, so
we have some almost-Goldstone bosons, which are called axions. This tunes θeff towards 0. We can define
ψL and ψR and currents
1 1
jLµ = ψ̄L γ µ ψL = η̄γ µ (1 − γ 5 )ψ, µ
jR = ψ̄R γ µ ψR = ψ̄γ µ (1 + γ 5 )ψ (23.2)
2 2
and we have QL = NL and QR = NR . The anomaly tells us that
Nf 2 Nf 2
∂µ jLµ = − g trF F̃ , µ
∂µ jR = g trF F̃ (23.3)
16π 2 16π 2
where Nf is the number of fermions. And the divergence of the axial current is
2Nf 2
∂µ j5µ = g trF F̃ (23.4)
16π 2
Now if we define the current
2Nf µ
J µ = j5µ −
j (23.5)
16π 2 A
will have divergence 0. In A0 = 0 gauge we will have a conserved charge

Q = NR − NL − 2Nf (vacuum #) (23.6)

This tells us that when we change from |ni vacuum to |n + 1i vacuum then NR and NL shift accordingly.
The fermion spectrum at |ni should be the same as that in |n + 1i, but if we go from |ni to |n + 1i through
some other configuration, then we will see that all the R fermions shift up by one level and all the L
fermions shift down by one level. If we start with a fermion vacuum where all E > 0 are empty and E < 0
states are filled, then afterwards one E > 0 R state will be filled, and one E < 0 L state will be empty, and
this is like creating an R particle and an L anti-particle. And this happens for every flavor. The change in
effective Lagrangian is like
 
Z Nf
dλ −8π2 /g2 (λ) Y
i i i j
∆Leff ∼ e ψ̄R ψL + ψ̄Li ψR
i 
∼ det(ψ̄R ψL ) (23.7)
λ5
i=1

103
Quantum Field Theory III Lecture 23

This term is not U (Nf )×U (Nf ) invariant. We had the U (1) problem in QCD because our chiral Lagrangian
was U (Nf ) × U (Nf ) invariant, but when we realize there are instantons then the above term explicitly
breaks to symmetry down to SU (Nf ) × SU (Nf ) and we don’t have the Goldstone boson which is not there
in the first place.
Now consider the Weinberg-Salam model which is SU (2) × U (1). The vacuum is

i
Aµ = − G−1 ∂µ G, φ = G hφ0 i (23.8)
g
If we just look at the gauge field, there is instantons. The Higgs field will prevent the instanton to be big
and the size will be
1 1
λ∼ ∼ (23.9)
hφi mW
Now what we should look at is anomalies. Remember we can’t have any anomalies for pure gauge inter-
actions, but we can have anomaly for baryon current coupled to two SU (2) currents through the triangle
diagram, similar for the lepton current. If we consider quarks and electron/neutrinos, then we will find that
B + L current is anomalous, and when we shift vacuum in principle we can somehow annihilate baryons
and create leptons, for example
p + n −→ e+ + ν (23.10)
The rate of the process should be
2 /g 2
rate ∼ e−16π W ∼ 10−176 (23.11)
Now the observed universe has ∼ 1078 baryons. The age of the universe is about 1040 seconds, and the
product is 10118 which is still much too small when multiplied by the above rate. However we know that in
early universe when there was enough energy then we could easily tunnel. So this process potentially will
dilute the baryon number. But remember B − L is not anomalous, so if we start with some baryon-lepton
number asymmetry then we are ok.

23.1 Supersymmetry
Let’s come to supersymmetry. Supersymmetry is a symmetry mixing bosons with fermions. It is a symme-
try mixing particles, so it is an internal symmetry, but it also mixes spins, so it mixes angular momentum
and Lorentz structure. There was a theorem by Coleman and Mandula stating that there is no non-
trivial mixing of Poincaré and internal symmetries. This means that the Poincaré generators will always
commute with the internal generators. However their proof only made use of commutators instead of
anticommutators. So for supersymmetry we need to make use of anticommutators.
We will start by using 4-component language for fermions. Remember charge conjugation

ψ → −γ 0 Cψ ∗ (23.12)

where
Cγµ C −1 = −γµT , CC † = I (23.13)
Remember complex scalar field corresponds to charged bosons, and real scalar field corrsponds to neutral
bosons. Similarly we want to do the same for fermions. In other words charged fermions will have double

104
Quantum Field Theory III Lecture 23

the degree of freedom. For neutral fermions we will have the Majorana basis where γ matrices are imaginary
and the Dirac equation is real. In Majorana basis we have
         
0 0 σy 1 iσz 0 2 0 −σy 3 −iσx 0 5 σy 0
γ = ,γ = ,γ = ,γ = ,γ = (23.14)
σy 0 0 iσz σy 0 0 −iσx 0 −σy

Note γ 0 and γ 5 are antisymmetric while others are symmetric.


We want to look at the quantity ᾱΓβ where α and β are Majorana spinors, which are real, and Γ is a
combination of γ matrices. Now we have
 
† 0 † T 0 T
ᾱΓβ = α γ Γβ = − β Γ (γ ) α = β̂ Γ̃α (23.15)

˜ = B̃ Ã. Now for γ µ we need to evaluate γ̃ µ , and we can verify


where Γ̃ = γ 0 ΓT γ 0 . So we can write AB
that
γ̃ µ = −γ µ (23.16)
Similarly we have
σ̃ µν = −σ µν , γ̃ 5 = γ 5 (23.17)
Therefore we have (
β̄Γα, Γ = I, γ 5 , γ µ γ 5
ᾱΓβ = (23.18)
−β̄Γα, Γ = γ µ , σ µν
We can write down the Lagrangian for Majorana fields just like
1
L = ψ̄(i∂/ − m)ψ (23.19)
2
but now ψ̄ can’t be varied independent of ψ.
The simplest supersymmetric model is called Wess-Zumino model where we have a Majorana field ψ
which has 4 real degrees of freedom, and we have 4 scalar fields A, B, F, G. In terms of particles we have
2 fermions states which are spin up and down, and two bosons states because two of the scalar fields are
just auxiliary fields with no dynamics. We will now write down the supersymmetry transformation with
infinitesimal Majorana spinor α

δA = iᾱψ, δB = iᾱγ 5 ψ, δF = −ᾱ∂/ψ, G = iᾱγ 5 ∂/ψ (23.20)


δψ = ∂/(A + iBγ 5 )α − iF α − Gγ5 α (23.21)

Let’s figure out how the kinetic term varies under this transformation
 
i i
(δ ψ̄∂/ψ + ψ̄γ µ ∂µ δψ

δ ψ̄∂/ψ =
2 2
i
−(∂µ ψ̄)γ µ δψ + ψ̄γ µ ∂µ δψ

=
2 (23.22)
= ∂µ {} + iψ̄γ µ ∂µ δψ
= ∂µ {} + iψ̄ ∂ 2 A + iBγ 5 α − i∂/F α − ∂/Gα
  

= ∂µ {} + iᾱ ∂ 2 (A + iBγ 5 ) + iγ µ ∂µ F − ∂ µ γ 5 ∂µ G ψ
 

105
Quantum Field Theory III Lecture 23

where in the last line the derivatives also acts on ᾱ. Now we can do an integration by parts again and put
derivatives on ψ
 
i
ψ̄∂/ψ = ∂µ {} + iᾱ −(∂µ A)(∂ µ ψ) − i(∂µ B)γ 5 (∂ µ ψ) − iγ µ F ∂µ ψ + Gγ µ γ 5 ∂µ ψ
 
δ
2
= ∂µ {} − (∂µ A)(∂ µ δA) − (∂µ B)(∂ µ δB) − F δF − GδG (23.23)
1
= −δ (∂A)2 + (∂B)2 + F 2 + G2

2
Therefore we have the invariant kinetic term
1 1 1 1 1
L1 = (∂µ A)2 + (∂µ B)2 + F 2 + G2 + ψ̄∂/ψ (23.24)
2 2 2 2 2
Also we can check that the following terms are invariant
1
L2 = ψ̄ψ + AF − BG, L3 = ψ̄(A − iBγ 5 )ψ + (A2 − B 2 )F − 2ABG (23.25)
2
What we need is the following Fierz identity

(ᾱψ)(ψ̄ψ) = (ᾱγ 5 ψ)(ψ̄γ 5 ψ) (23.26)

So our Lagrangian will be


L = L1 − mL2 + λL3 (23.27)
If λ = 0 then the equations of motion are

A = −mF, B = mG, F = mA, G = −mB, (∂/ − m)ψ = 0 (23.28)

We can eliminate F and G in terms of A and B, so the equations of motion for A and B are

( + m2 )A = 0, ( + m2 )B = 0 (23.29)

Now if λ 6= 0 then the equations of motion are more complicated

A = −mF + 2λAF − 2λBG + λψ̄ψ, B = mG − 2λBF − 2λAG − iλψ̄γ 5 ψ (23.30)


2 2
F = mA − λ(A − B ), G = −mB + 2λAB (23.31)

If we substitute F and G back, it is equivalent to having a potentail for A and B


1 1
V = m2 (A2 + B 2 ) − mλA(A2 + B 2 ) + λ2 (A2 + B 2 )2 (23.32)
2 2
Let’s look at the vertices of the theory
We have scalar couplings which are like square of the fermion-scalar coupling, and the cubic coupling
is not independent. Now if we look at the loop correction to the scalar mass
If we work out the coefficients we will see that the Λ2 divergences automatically cancel, due to the
fermion loop corrections and scalar loop corrections. The phenomenological motivation for supersymmetry
is that this can solve the hierarchy problem, because the radiative corrections are not so important here.

106
Quantum Field Theory III Lecture 24

24 Lecture 24
24.1 Wess-Zumino Model
Let’s get back to Wess-Zumino model. Recall the SUSY transformation for our bunch of fields
δA = iᾱψ, δB = iᾱγ 5 ψ, δF = −ᾱ∂/ψ, G = iᾱγ 5 ∂/ψ (24.1)
5
δψ = ∂/(A + iBγ )α − iF α − Gγ5 α (24.2)
We wrote this down in an ad hoc way. Now let’s look at the double transformation δβ δα Φ − δα δβ Φ where
Φ is one of the fields. Let’s consider Φ = A then we can work out
δβ δα A = iᾱψ + iβ̄(ψ + δα ψ), δα δβ A = iβ̄ψ + iᾱ(ψ + δβ ψ) (24.3)
So the difference will be
(δβ δα − δα δβ ) A = iβ̄δα ψ − iᾱδβ π = 2iβ̄γ µ α∂µ A (24.4)
Recall the traslation of the field Φ is written as
δ Φ = i [µ Pµ , Φ] = µ ∂µ Φ (24.5)
which has the same structure as the above. If we write an operator which is the supercharge
δα Φ = i [ᾱQ, Φ] (24.6)
then we can write
    
(δβ δα − δα δβ )Φ = − β̄Q, [ᾱQ, Φ] + ᾱQ, β̄Q, Φ
   (24.7)
= − β̄Q, Q̄α , Φ
So if we compare this with the above transformation we can find this looks like a translation with µ =
2iβ̄γ µ α, so we can write
  
β̄Q, Q̄α = β̄a Qa Q̄b αb − Q̄b αb β̄a Qa = β̄ Qa Q̄b + Q̄b Qa αb = −iPµ (24.8)
So we can write  µ
Qa , Q̄b = 2γab Pµ (24.9)
Now this is a quantum mechanical operator equality. We can decompose the bar and write just the
conjugate of Qb and get n o
Qa , Q†b = 2 γ µ γ 0 ac Pµ

(24.10)
Let’s take trace on the above equation and set a = c, we get
n o
Qa , Q†a = 2 γ µ γ 0 aa Pµ = 8P 0 = 8H

(24.11)

Because we have an operator times its conjugate on the left hand side, it must be positive definite, so we
have H ≥ 0. In particular if
1D E
hHi = Qa Q†a + Q†a Qa = 0 (24.12)
8
then the state that we are taking expectation must satisfy Qa |αi = 0 and Q†a |αi = 0. So this state must
be invariant under SUSY. Equivalently if we have a SUSY invariant state, then we have E = 0. This is
very nice in that we get rid of the infinite vacuum energy, but we know that the cosmological constant is
nonzero. This is solved by supergravity where we introduce coupling to gravity.

107
Quantum Field Theory III Lecture 24

24.2 Notation Transmutation


So far we have used the 4-component Majorana spinors. We’d like to change into the 2-component spinor
representation which is more convenient. Remember the representation of the Lorentz group labelled by
(j1 , j2 ). Lorentz vectors are in (1/2, 1/2) representation and an antisymmetric tensor is in (1, 0) ⊕ (0, 1)
representation. A Dirac spinor is in the representation (1/2, 0) ⊕ (0, 1/2). To see this more clearly let’s use
the chiral basis where
     
5 I 0 0 0 −I i 0 σ
γ = , γ = , γ = (24.13)
0 −I −I 0 −σ 0

The charge conjugation matrices are

0 −iσ 2
   
−iσ2 0 0
C= , γ C= (24.14)
0 iσ2 iσ 2 0

A Dirac spinor is written as


iσ2 χ∗
   
η c
Ψ= , Ψ = (24.15)
χ −iσ2 η ∗
And Majorana spinors are just where η and χ are related. Now we write the upper components in one
spinor ηα where α = 1, 2 and we write the lower components in another spinor χα̇ where α̇ = 1, 2. The dot
means they are in different representations. We can raise and lower indices on the 2-component spinors by
αβ which satisfies 12 = −21 = 1. A Lorentz transformation will bring

η α → (η 0 )α = M αβ η β , (η α )∗ → (M αβ )∗ (η β )∗ (24.16)

But remember conjugation changes the representation, so we write (η α )∗ = η̄ α̇ . Therefore we have the
transformation law
ψα → Mα β ψβ , ψ̄α̇ → (M ∗ )α̇β̇ ψ̄β̇ (24.17)
Let’s construct Lorentz invariants out of the spinors. We can write

ψ α ηα = ψ 1 η1 + ψ 2 η2 = η 1 ψ1 + η 2 ψ2 = ψη = ηψ (24.18)

So whenever we write ψη we mean the former spinor has upper index. Similarly we have

ψ̄ η̄ = η̄ ψ̄ = ψ̄α̇ η̄ α̇ (24.19)

We write the four Pauli matrices as

σαmα̇ = (−I, σ)αα̇ , σ̄ mα̇α = (−I, −σ ∗ ) = α̇β̇ αβ σβmβ̇ (24.20)

The supercharges are Majorana spinors, so they carry spin indices. We write Qα and Q̄α̇ as the four
generators with the following relation
n o n o
Qα , Q̄β̇ = 2σαmβ̇ Pm , {Qα , Qβ } = 0, Q̄α̇ , Q̄β̇ = 0 (24.21)

Consider p2 < 0 or m > 0 and work in the rest frame where pm = (m, 0, 0, 0) and pm = (−m, 0, 0, 0).
In this case we have n o
Qα , Q̄β̇ = 2M δαβ̇ (24.22)

108
Quantum Field Theory III Lecture 24

So let’s define
1 1
aα = √ Qα , a†α = √ Q̄α̇ (24.23)
2M 2M
Then these a and a† have the algebra of fermionic creation and annihilation operators.
Let’s suppose we have some state |Ωi which has the property that

aα |Ωi = 0, for α = 1, 2 (24.24)

Then we can construct two states


a†1 |Ωi , a†2 |Ωi (24.25)
which have jz ± 1/2, where jz is the angular momentum that we start with, and we can also create

a†1 a†2 |Ωi (24.26)

which has jz again. So starting from any such |Ωi we have a multiplet of 4 states. If |Ωi has jz = 0 then
we get a state with j = 1/2 and two jz = 0 states. If |Ωi = 1/2 then we have two jz = 1/2 states and
jz = 0, 1 states. Putting them together we can get two Majorana spinors, one spin 0 scalar and one spin 1
field.
But what if we have zero mass? Then we don’t have rest frame and we have to have pm = (−E, 0, 0, E).
In this case we have  
n o 1 0
Qα , Q̄β̇ = 4E (24.27)
0 0 αβ̇
n o
Therefore the difference is that Q2 , Q̄2̇ = 0 = Q2 , Q†2 . So we have to have Q2 = 0. Then we can


define
1 1
a = √ Q1 , a† = √ Q̄1 (24.28)
2 E 2 E
So we only have one kind of creation and annihilation operators and the multiplet has 2 states with helicity
λ and λ + 1/2.
We can have more general algebra, which is the extended SUSY. We write the operators QL α and Q̄β̇
L

where L = 1, 2, . . . , N . In a simple basis, we can work out the algebra of Q and Q̄ with Pµ and Mµν , and
we will get the most general possible algebra
n o
QL M
= 2σαmβ̇ Pµ δ LM ,
 L M
α , Q̄β̇
Qα , Qβ = αβ Z LM (24.29)

The second relation is new. We must have Z LM = −Z M L and it must commute with all Q, Q̄, Pµ , Mµν and
all the internal symmetries. We call this the central charge generator. Let’s look at what kind of multiplets
we can have. For massive particles and N -SUSY, assuming Z = 0, we have 4N supercharges and 2N of
them raise jz by 1/2 and the other 2N lower jz by 1/2. Similar to above we can define 2N creation and
2N annihilation operators, and these give us 22N states. If we start with a state with jz = m and lower it
2N times, then we get
jz = m, m − 1/2, . . . , m − N (24.30)
Obviously we have bound on N and m. If we have N = 2 then we have to start with jz = 1 and go from
1 to −1. If we have N = 4 then we have to start with jz = 2. Now if we have massless M = 0 case

109
Quantum Field Theory III Lecture 24

and N -SUSY and Z = 0. 2N charges will vanish like above, and the remaining 2N will separate into N
annihilation and N creation, and we have 2N states.
Now if we look at N = 2 SUSY. There are 16 massive states in the multiplet. Suppose jzmax = 1, we
have 5 kinds of state with different number of creation operators in front of |Ωi, with numbers 1,4,6,4,1.
These sum up to give one j = 1 multiplet, 4 Majorana j = 1/2, and 5 j = 0 scalars. If we count it there
will be 8 bosonic states and 8 fermionic states. For massless case we have 4 states in the multiplet. If
we start with λ = 1/2 then we have λ = 1/2, 0, 0, −1/2. This is called a “hyper multiplet”. If we start
with λ = 1 then we get 1, 1/2, 1/2, 0 but in order to have CP T invariance we need another multiplet
λ = −1, −1/2, −1/2, 0, and these combined is called a “vector supermultiplet”.
Now let’s consider phenomenology. If we put things in massless multiplets, then gauge bosons need
fermionic partners in the adjoint representation. The fermions we observe will be in the hyper multiplet
representation but here positive and negative helicity states transform the same. We need to somehow
break this also. So N = 2 SUSY is no good for phenomenology.
Let’s consider N = 4 SUSY. The massive states is a 256 multiplet. The massless states is a 16 multiplet
and we have one j = 1 and 4 j = 1/2 and 6 j = 0 and they are all in the adjoint representation. This is
the simplest Yang-Mills theory because all the divergence cancel, and couplings do not run.
For N higher than 4 we need higher spin particles. So for supergravity if we want to study gravitons
we can take N = 8. Now N = 8 is also a remarkable theory, because people show that here the theory is
renormalizable.
Now let’s consider Z 6= 0 case. For N = 2 we have Z 12 = −Z 21 = 2Z then the algebra becomes
n o
QL M
= 2σαmβ̇ Pµ δ LM ,
 L M
α , Q̄ β̇
Qα , Qβ = 2αβ LM Z (24.31)

We define
1 h i 1h 1 i
aα = √ Q1α + α̇β̇ Q̄2β̇ , bα = Qα − α̇β̇ Q̄2β̇ (24.32)
2 2
and conjugate defined similarly. Then we can work out the algebra of a and b
n o n o
{a, a} = {a, b} = 0, aα , a†β = 2M δαβ̇ − 2Zδαβ̇ , aα , b†β = 0 (24.33)

and for b and b† it is plus in the middle equation. Now if M = ±Z then one of the anticommutators
will vanish, so either a = 0 or b = 0, so it is like a massless multiplet in size, which are called “short”
supermultiplets. We can also check from the algebra and positive-definiteness that M 2 ≥ Z 2 . So Z gives
a lower bound on M .
We want to ask the question what does it mean by having a classical solution which is invariant under
SUSY. For example when we have a solution φ and A which are nonzero. We can add fermionic field ψ = 0
and scalars, and require that δφ = δA = 0 and the δψ = 0 equation gives us a constraint identical to the
BPS theory.

110
Quantum Field Theory III Lecture 25

25 Lecture 25
Let’s write down the Wess-Zumino model in two-component language

A, B −→ complex A (25.1)
F, G −→ complex F (25.2)
√ √
δA = 2ηψ = 2η α ψα (25.3)
√ √
δψ = i 2σαmβ̇ η̄ β̇ ∂m A + 2ηa F (25.4)

δF = i 2η̄α̇ σ̄ mα̇β ∂m ψβ (25.5)

The Lagrangians are

L0 = i∂m ψα σ̄ mα̇β ψβ + A∗ A + F ∗ F (25.6)


 
∗ ∗ 1 1
Lm = m AF + A F − ψψ − ψ̄ ψ̄ (25.7)
2 2

We define a superspace as follows. We have xm which are 4 real bosonic coordinates, and θα and θ̄α̇
are 4 real fermionic coordinates, or Grassmann numbers. We know that Pm momentum are translations in
real space xm , so somehow Q is translation in θ coordinates. However this is not quite true, because the
anticommutator of Q will give us Pm . So let’s write
∂ ∂
Qα = + Aα , Q̄β̇ = − + Bβ̇ (25.8)
∂θα ∂ θ̄β̇
Now we ancitipate that  
∂ ∂
, =0 (25.9)
∂θ ∂ θ̄
So the anticommutation between Q and Q̄ is
n o ∂ ∂
Qα , Q̄β̇ = α Bβ̇ − Aα = 2iσαmβ̇ ∂m (25.10)
∂θ ∂ θ̄β̇
Now to make this equality hold we can choose the following A and B

Aα = −iσαmβ̇ θ̄β̇ ∂m , Bβ̇ = iθα σαmβ̇ ∂m (25.11)

This constitutes a representation of the SUSY algebra in the superspace


∂ ∂
Qα = − iσαmβ̇ θ̄β̇ ∂m , Q̄β̇ = − + iθα σαmβ̇ ∂m (25.12)
∂θα ∂ θ̄β̇
Now we want to write down some functions on superspace Φ(x, θ, θ̄), which are called superfields.
We’d like to define some derivatives on this function. However just like in gauge theory, we want to have
the derivative transform as the original field δ(DΦ) = D(δΦ). If the transformation is generated by the
operator ηQ + η̄ Q̄, the we would like [ηQ + η̄ Q̄, D] = 0. So we want

{Qα , Dβ } = 0 = Q̄α̇ , Dβ (25.13)

111
Quantum Field Theory III Lecture 25

Naively taking Dα = ∂/∂θα will obviously not work. However what can work is

∂ ∂
Dβ = + iσβnβ̇ θ̄β̇ ∂n , D̄β̇ = − − iθα σαnβ̇ ∂n (25.14)
∂θβ ∂ θ̄β̇
So the D’s are just like the Q’s except that some signs are flipped. Now obviously

Dα , D̄α̇ = −2iσαmα̇ ∂m

(25.15)

And by construction
   
{D, D} = D, D̄ = {D, Q} = D, Q̄ = D̄, Q = D̄, Q̄ = 0 (25.16)

We have exactly one fermionic variable in each term in D, so we have

D(AB) = (DA)B ± A(DB) (25.17)

where plus or minus depends on whether A is bosonic or fermionic. Now suppose we define

y m = xm + iθα σαmα̇ θ̄α̇ (25.18)

So we can evaluate
D̄α̇ y m = iθα σαmα̇ − iθα σαmα̇ = 0 (25.19)
Now let’s look at superfields. We can write a function f (x, θ, θ̄) as

f (x, θ, θ̄) = f (x) + θφ(x) + θ̄χ̄(x) + θθm(x) + θ̄θ̄n(x) + θσ m θ̄vm (x) + (θθ)θ̄λ̄(x) + (θ̄θ̄)θψ̄(x) + (θθ)(θ̄θ̄)d(x)
(25.20)
3 3
We don’t have any more terms because θ = θ̄ = 0. If we count the number of fields here, it is exactly
16. Now we want to find some sets of these which only mix among themselves. One possibility is to define
chiral superfields, which satisfies
D̄α̇ Φ = 0 (25.21)
From what we have above, we would know that

Φ = Φ(y m , θ) (25.22)

Now what can Φ look like? We can write



Φ(y, θ) = A(y) + 2θα ψα (y) + θα θθ F (y)
1 √ 
= A(x) + ∂m A(x)(iθσ m θ̄) + i∂m ∂n A(x)(iθσ m θ̄)(iθσ n θ̄) + 2θ ψ(x) + ∂m ψ(iθσ m θ̄) + θθF (x)

2
(25.23)

Let’s simplify this. Remember θα θβ ∝ αβ . We can evaluate

θ1 θ2 = −θ1 θ1 = −θ2 θ1 = −θ2 θ2 (25.24)

So we have
1
θα θβ = − αβ θθ (25.25)
2

112
Quantum Field Theory III Lecture 25

and similarly
1
θ̄α̇ θ̄β̇ = α̇β̇ θ̄θ̄ (25.26)
2
With more work we can get more identities
1
θα ∂m ψα (θσ m θ̄) = − θθ(∂m ψσ m θ̄) (25.27)
2
1
(θσ m θ̄)(θσ n θ̄) = − η mn (θθ)(θ̄θ̄) (25.28)
2
1
(θη)(θχ) = − (θθ)(ηχ) (25.29)
2
ησ m χ̄ = −χ̄σ̄ m η (25.30)

We can use these to simplify the superfield expression



√ m 2 i
Φ(y, θ) = A(x) + 2θψ(x) + θθF (x) + i∂m A(θσ θ̄) − θθ(∂m ψσ m θ̄) + (A)(θθ)θ̄θ̄ (25.31)
2 4
And the conjugate field is

∗ ∗ i ∗ ∗
√ 2
m
Φ (y, θ) = A − iθσ θ̄∂m A − θθθ̄θ̄A + 2θ̄ψ̄ − θ̄θ̄σ m ∂m ψ̄ + θ̄θ̄F ∗ (25.32)
4 2
Now what does SUSY transformation look like? We can just use δΦ = (ηQ + η̄ Q̄)Φ, so we have
√ √ √ √
δA = 2ηψ, δF = i 2η̄σ̄ m ∂m ψ, δψ = 2ηF + i 2(σ m η̄)∂m A (25.33)

Note that apart from a constant 2 these are the same as the SUSY transformation we introduced.
Now we are interested in the Lagrangian that we can construct. Note that

δf (Φ, Φ† )|θθθ̄θ̄ = ∂m (), δg(Φ)|θθ = ∂m () (25.34)

So we can construct the invariant Lagrangian

L = f (Φ, Φ† ) + g(Φ) + h(Φ† ) (25.35)

Note the Grassmann variable integration is identical to differentiation, so we can write


Z Z Z
L = d2 θ d2 θ̄ f (Φ, Φ† ) + d2 θ g(Φ) + d2 θ̄ h(Φ† ) (25.36)

where f is called the Kahler potential, and g is called the superpotential. Let’s consider one specific
example of f schematically

Φ† Φ ∼ A∗ A + ψ̄∂ψ + F ∗ F + (∂ ψ̄)ψ + A∗ A + (∂A∗ )(∂A) (25.37)

θθθ̄θ̄

If we go through all the identities carefully we will get



Φ† Φ = A∗ A + i∂m ψ̄σ m ψ + F ∗ F + ∂m () (25.38)

θθθ̄θ̄

113
Quantum Field Theory III Lecture 25

This is the most general kinetic term we can get. Now let’s look at the potential term g. Assume it is a
polynomial, then we can write

1 √ √ ∂ 2 g

∂g
g(Φ)|θθ = θθFj + ( 2θψj )( 2θψk ) (25.39)
∂Φj 2 ∂Φj ∂Φk Φ=A

Conventionally we write g(Φ) as W (A) so we have

∂W 1 ∂2W
g(Φ)|θθ = Fj − ψj ψk (25.40)
∂Aj 2 ∂Aj ∂Ak

Now the total Lagrangian is like


∂W ∂W ∗
L = Fj Fj∗ + Fj + Fj∗ + (A, ψ terms) (25.41)
∂Aj ∂A∗j

So the equations of motion for F and F ∗ are


∂W ∂W ∗
0 = Fj + = Fj∗ + (25.42)
∂Aj ∂A∗j

The potential for Aj then will be like


X ∂W 2

V (Aj ) =
∂Aj (25.43)
j

We can see that the potential of the scalar fields are related to the Yukawa couplings. Let’s say that
W = bΦ + cΦ2 + dΦ3 , then we can always shift to get rid of the linear term, and write
m 2 λ 3
W = Φ + Φ (25.44)
2 3
Then the potential for A will become

∂W 2

2 2 ∗ 2 2 4
∂A = m |A| + 2mλRe (A A ) + λ |A| (25.45)

This is the potential in Wess-Zumino model.

114
Quantum Field Theory III Lecture 26

26 Lecture 26
Let’s write down the superfield (without worrying about factors of i or 2)

Φ = A(y) + θψ(y) + θθF (y) = A(x) + θσ θ̄∂A + θθθ̄θ̄A + θψ + θθ(∂ψ)σ θ̄ + θθF (26.1)

Recall that the superfield is conveniently defined to satisfy D̄Φ = 0. The Lagrangian is written as

L = K(Φ† , Φ) + W (Φ)|θθ + W † (Φ† )

θθ θ̄ θ̄ θ̄θ̄
Z Z Z (26.2)
= dθ2 dθ̄2 K + dθ2 W + dθ̄2 W †

In the Wess-Zumino model the kinetic term and superpotential is


1 1
W = mΦ2 + λΦ3 , K = Φ† Φ (26.3)
2 3
Let’s consider the transformation

θ → eiα θ, Φ → e2iα/3 Φ, Φ† → e−2iα/3 Φ† , d2 θ → e−2iα d2 θ (26.4)

Then if we take m = 0 then the Lagrangian is invariant under this transformation. The component fields
transform as
A → e2iα/3 A, ψ → e−iα/3 ψ, F → e−4iα/3 F (26.5)
R 2 2
This symmetry is called R-symmetry, and if we require this symmetry then terms like d θ Φ will not
occur in renormalization because it has to obey this symmetry.
Now let’s consider vector superfields with the constraint V = V †
i i
V = c(x) + iθχ − iθ̄χ̄ + θθ(M + iN ) − θ̄θ̄(M − iN )
2 2 (26.6)
m 1
− θσ θ̄vm + iθθθ̄λ̄ − iθ̄θ̄θλ + θθθ̄θ̄D
2
where c, M, N, D are real scalar fields, vm is a vector field, and χ and λ are Majorana spinor fields. Now
vm is a candidate for gauge field. Because a SUSY transformation is written as ηQ + η̄ Q̄ the condition
V = V † is maintained, so these fields remain real. We want to write down the transformation of the gauge
field. We claim that
V → V + Φ + Φ† (26.7)
is a legitimate transformation. The component fields transform as

c → c + A + A∗ , χ → χ − i 2ψ, M + iN → M + iN − 2iF, vm → vm − i∂m (A − A∗ ), (26.8)
1 1
λ → λ − √ σ m ∂m ψ, D → D + (A + A∗ ) (26.9)
2 2
Now we can choose the real part of A and ψ and F to make c, χ, and M + iN vanish. This gauge is called
the Wess-Zumino gauge. However we still have the freedom in A − A∗ which is the imaginary part of A.
So in this gauge the field is written as

V = θσ θ̄vm + θθθ̄λ̄ + θ̄θ̄θλ + θθθ̄θ̄D (26.10)

115
Quantum Field Theory III Lecture 26

So if we just change the field by A − A∗ then we will have an ordinary gauge transformation on vm .
Let’s write a superfield as
1
Wα = − D̄β̇ D̄β̇ Dα V (26.11)
4
Then we immediately know that this is a chiral field because D̄β Wα = 0. Now we want to know how it
transforms under gauge transformation. We find that
1 1
Wα → Wα − D̄D̄Dα Φ = Wα + D̄β̇ D̄β̇ Dα Φ (26.12)
4 4
Note that D̄Φ = 0 so the last term is just like an anticommutator acting on Φ, so we have
1
Wα → Wα + D̄β̇ (−2iσαmβ̇ ∂m Φ) = Wα (26.13)
4
So this field is gauge invariant, and that is reason we wrote it down in the first place. Its component form
is
i
Wα = −iλα (y) + θα D(y) − (σ m σ̄ n )αβ θβ (∂m vn (y) − ∂n vm (y)) + θθσαmα̇ ∂m λ̄α̇ (y) (26.14)
2
The Lagrangian for V will then be
Z Z 
1 2 α 2 α̇
LV = d θ W Wα |θθ + d θ W̄α̇ W̄ |θ̄θ̄
4
(26.15)
1 2 1 mn m
= D − v vmn − iλσ ∂m λ̄
2 4
We can also add some mass term, but it will not be gauge invariant, and we need to use something like
the Higgs mechanism. We will not do it here.
Now let’s consider the transformation of the matter fields

Φ → e−iqΛ Φ, Φ† → Φ† eiqΛ , V → V + i(Λ − Λ† ) (26.16)

We claim that the term Φ† eqV Φ is gauge invariant, indeed


† †
Φ† eqV Φ → Φ† eiqΛ eq(V +iΛ−iΛ ) e−iqΛ Φ = Φ† eqV Φ (26.17)

And this does not depend on the Wess-Zumino gauge at all. Now the super potential is a polynomial in
Φ, so a term like Φi Φj Φk will be invariant if qi + qj + qk = 0. A natural invariant Lagrangian is

1 α 
L= W Wα |θθ + W̄β̇ W̄ β̇ |θ̄θ̄ + Φ†j eqj Vj Φj |θθθ̄θ̄ + W (Φj )|θθ + W † (Φ†j )|θ̄θ̄ + κV |θθθ̄θ̄ (26.18)
4
The last term is called the Fayet-Iliopoulos term, and it is invariant because the D term is invariant under
pure gauge transformation. Now let’s look at the second term

Φ† eqV Φ = Φ† Φ + qΦ† V Φ + q 2 Φ† V 2 Φ (26.19)

The first term is just the ordinary kinetic term. The second term will give us the interaction like

qΦ† V Φ ∼ vA∗ ∂A + v(∂A∗ )A + vm ψ ∗ σ m ψ + A∗ λψ + Aλ̄ψ̄ + D |A|2 (26.20)

116
Quantum Field Theory III Lecture 26

The third term is more restrictive because the only term surviving V 2 will be just vm
2 , so it will give a term

like
q 2 Φ† V 2 Φ ∼ vm v m |A|2 (26.21)
This is the last term in the ordinary coupling of gauge field to a charged scalar. So the only new interaction
is that between A, λ, and ψ.
Now let’s do supersymmetric QED. We have the gauge field Aµ and the Dirac field ψ which has 2
Majorana spinors. Let’s count the number of real degree of freedom. The Aµ has 4 − 1 = 3 real fields, and
it has 2 degrees of freedom, and ψ has 8 real fields, and 4 degrees of freedom. Now we need to put these
into superfields, so we need to introduce λ called the photino, which is a Majorana field with 2 physical
degrees of freedom. We also need an auxiliary field with no physical degree of freedom. Now for the Dirac
field because we have 2 Majorana fields we need two superfields Φ+ and Φ− . They have bosonic partners
A+ and A− which are called selectrons, and there are 4 real fields with 4 physical degrees of freedom. We
also have the auxiliary fields F+ and F− with no physical degree of freedom.
Now recall the Lagrangian of the theory (26.18). It will become
1  
L= ( W α Wα |θθ + h.c.) + Φ†+ eeV Φ+ + Φ†− eeV Φ− +m Φ+ Φ− |θθ + Φ†+ Φ†− +κ V |θθθ̄θ̄ (26.22)

4 θθθ̄θ̄ θθθ̄θ̄ θ̄θ̄

The gauge transformation will read


Φ± → e∓ieΛ Φ± (26.23)
We claim that there is R-symmetry here

θ → eiα θ, V → V, Φ± → eiq± α Φ± (26.24)

and if we require q+ +q− = 2 then the Lagrangian is R-invariant. For example we could choose q+ = q− = 1,
then we will have
λ → e−iα λ, A± → eiα A± , ψ± → ψ± , vm → vm (26.25)
Under this transformation the physical fields do not transform, and the supersymmetric fields gets trans-
formed. This symmetry requires that all the supersymmetric particles have to appear or annihilate in
pairs. Phenomelogically it means that if supersymmetric particles were created in early universe they will
decay into the lightest particle, and the annihilation rate of these lightest particles will be smaller and
smaller when the universe expands and finally freeze in. So they are a candidate for dark matter particles.
We can write out the Lagrangian in component form

1 2 e   X  ∂W 
2 2 2 2
L = D + |F+ | + |F− | + D |A+ | + |A− | + Fj + h.c. + κD + . . . (26.26)
2 2 ∂Aj
j=±

From here we can write out the equations of motion for the auxiliary fields
e 
D+ |A+ |2 + |A− |2 + κ = 0, F±∗ + mA∓ = 0 (26.27)
2
So the effective potential for A will become

1 1 1 e2  2
V = κ2 + (m2 + eκ) |A+ |2 + (m2 − eκ) |A− |2 + |A+ |2 − |A− |2 (26.28)
2 2 2 8

117
Quantum Field Theory III Lecture 26

Suppose κ = 0 then the minimum of this potential will be at A+ = A− = 0 and there is no spontaneous
symmetry breaking for the gauge field of U (1). Now if |κ| > 2m2 /e then there will be spontaneous
symmetry breaking of U (1).
That is Abelian gauge theory. Now let’s consider non-Abelian case
1
V = Tija Va , Wα = − D̄D̄e−V Dα eV (26.29)
4
where the transformation law is

eV → e−iΛ eV eiΛ (26.30)
If we go through the same computation as above for the effective potential of A then we will get
1 ∂W
V = |Fi |2 + |Da |2 , Fj∗ = , Da = ga A∗i Tija Aj (26.31)
2ga2 ∂Aj

There is no κ term now because V is not invariant under global gauge transformation. However the rest
is just like what we had above.
However we don’t have supersymmetry. It has to be spontaneously broken. Let’s just assume for now
that the potential is
X ∂W 2

V =
∂Aj (26.32)
j

If there is only one A field then V = 0 if ∂W/∂A = 0 and we have an unbroken vacuum. But we can’t
find a function of complex variable that its derivative is nonzero everywhere, so there is always unbroken
supersymmetric vacuum if there is only one A. So now let’s consider the O’Raifeartaigh model where

W = λΦ0 + mΦ1 Φ2 + gΦ0 Φ21 (26.33)

where λ, m, g are real and positive. Then the potential will becomes

∂W 2 ∂W 2 ∂W 2

V = + +

∂A0 ∂A1 ∂A2 (26.34)
2
= λ + gA21 + |mA2 + 2gA0 A1 |2 + m2 |A1 |2

Assuming all parameters are nonzero, then let’s find the minimum of V . We have ∂V /∂A2 = 0 which is
A2 = −(2g/m)A0 A1 . Let’s set A21 = −x where x > 0, then the potential is just

V = (λ − gx)2 + m2 x (26.35)

Then minimizing this is to find where ∂V /∂x = 0, which is where

m2 − 2g(λ − gx) = (m2 − 2gλ) + 2g 2 x (26.36)

If m2 > 2gλ then the minimum is at x = 0 so that A1 = A2 = 0 and A0 is anything we want. If m2 < 2gλ
then the minimum is at
2gλ − m2
x= (26.37)
2g 2

118
Quantum Field Theory III Lecture 26

then we have s s
2gλ − m2 2ig 2gλ − m2
A1 = ±i , A2 = ∓ A0 (26.38)
2g 2 m 2g 2
and A0 again can be arbitrary. This is no problem as we can have family of degenerate vacuum. The
important thing for broken supersymmetry is that ∂W/∂A is nowhere vanishing. Let’s come back to the
SUSY QED
1h i2  
V = e(|A+ |2 − |A− |2 ) + 2κ + m2 |A+ |2 + |A− |2 (26.39)
8
So we have a constraint on κ if we want a broken supersymmetry.
Let’s call Mij = ∂ 2 W/∂Ai ∂Aj , then at minimum of V we should have

∂W 2 2
 
∂ ∼ ∂ W ∂W
0= (26.40)
∂Ai ∂Aj ∂Ai ∂Aj ∂Aj

But we know that ∂W/∂A never vanish, so the mass matrix Mij has to have zero eigenvalue, so we have
massless fermion. This is analogous to the Goldstone boson, and we call this a Goldstino. Just like in
Higgs mechanism the Goldstone boson combines with the gauge boson to give a massive vector particle,
the Goldstino in supergravity combines with the spin 3/2 gravitino to give massive gravitinos.

119
Quantum Field Theory III Lecture 27

27 Lecture 27
Let’s get to the breaking of supersymmetry. If we have a standard Lagrangian L. Then
X X
(−1)F m2j + 2 DA (tr tA ) = 0 (27.1)
j A

The second term is summing P over all the


PU (1) symmetries, and the trace means the sum of U (1) charges.
2 2
The first term is actually bosons mj − fermions mj , while keeping track of number of spins. This identity
is true at tree level, but loops will give us corrections. P
Let’s
P consider standard model. The only U (1) is just QEM or QY , but we know in SM that QY =
0 = QEM . So the second term in the above identity vanishes in SM. In standard model we have as
bosons the γ, g, W ± , Z, and 6 × 3 quarks, e, µ, τ and 3 ν’s. Let’s consider partners. For the bosons we need
15 Majorana partners in total. For fermions we have 24 Dirac fermions, and we need 96 scalars as their
partners. By the above identity we need
X X
m2j = m2j (27.2)
boson fermion

However if we have supersymmetry then we the sum of boson masses will much exceed the sum of fermion
masses, and there is nothing we can do about. This suggests that there is a hidden sector in supersymmetric
standard model, and also we have supersymmetry breaking. Now suppose we don’t know this, then our
Lagrangian will have some SUSY terms, and some SUSY violating terms. If we didn’t know about the
Higgs particles, the gauge-invariant terms will have mass dimension 4, and gauge-violating terms will have
dimension less than 4, for example the mass term for gauge bosons. The idea is the same in SUSY. We will
write SUSY parts in the Lagrangian, and write the SUSY violating terms which are softer, or with mass
dimension less than 4. This is analogous to doing SU (2) × U (1) symmetry without the Higgs, where we
just introduce W masses by hand in the form of soft gauge-violating terms, with the mass as the parameter
of the theory.
Let’s try to group the particles in a minimal set. We first write down the vector superfields

1. Ga : gluons gµa and gluinos g̃ a .

2. W i : Wµi and W̃ i .

3. B: Bµ and B̃ which are called winos, binos, zinos, photinos together with the W̃ i .

Now we also have the chiral superfields

1. Q: 3 under SU (3), 2 under SU (2) and Y = 1/3; they correspond to (u, d)L and (ũL , d˜L )

2. U : 3̄ under SU (3), 1 under SU (2) and Y = −4/3; they correspond to ucL and ũ∗R

3. D: 3̄ under SU (3), 1 under SU (2) and Y = 2/3; they correspond to dcL and d˜∗R

4. L: 1 under SU (3), 1 under SU (2) and Y = −1; they correspond to (ν, e)L and (ν̃L , ẽL )

5. E c : 1 under SU (3), 1 under SU (2) and Y = 2; they correspond to ecR and ẽ∗R

6. H1 : 1 under SU (3), 2 under SU (2) and Y = −1; they are Higgs H1 and Higgsinos h̃1

120
Quantum Field Theory III Lecture 27

7. H2 : 1 under SU (3), 2 under SU (2) and Y = +1; they are Higgs H2 and Higgsinos h̃2

The fact that we have two Higgs fields is due to the cancellation of anomalies. However now we have 8
real degrees of freedom and 3 of which disappear because of Higgs mechanism to give masses to W ’s, but
we have 5 particles left.
Now let’s look at the superpotentials. We need to hypercharge to vanish. The two field couplings will
be like
µH1 H2 + µ0a La H2 (27.3)
Let’s look at three fields. the terms that work are

ij λL H1i Lj E c + λD H1i Qj Dc + λν H2i Qj U c + λ1 Li Lj E c + λ2 Li Qj Dc + λ3 U c Dc Dc


 
(27.4)

The first three terms have vanishing baryon number and lepton number, however the last three terms
violate baryon and lepton numbers. So we could have a process like

u + d −→ d˜ −→ ν̄ + d¯ (27.5)

Because we have a squark in between, if the mass of squark is very large then it is okay, but if that is
the case the there is no point in having supersymmetry. So we have to have λ2 and λ3 very small. We
introduce R-parity transformation where θ → −θ and θ̄ → −θ̄ and Φ → ±Φ. We choose the parity to be
+1 for vector superfields and H1 and H2 , and −1 for other fields. This means that for the known particles
we have parity +1 and the unknown particles have parity −1. Now if we require this symmetry, then the
last three terms will go away. Similarly the term µ0a La H2 will go away as well. Now we have 19 parameters
because all the λ parameters above are matrices in the flavor space.
This is not the whole story yet. We need to put in the SUSY violating terms. Now we can’t write in
terms of superfields because SUSY is violated. The possible terms are
     
Lsoft = −m21 |H1 |2 − m22 |H2 |2 + Bµij H1j H2j + c.c. + MQ̃ 2
ũ∗L ũL + d˜∗L d˜L + MŨ2 (ũ∗R ũR ) + Md2˜ d˜∗R d˜R
1h ¯ i
+ ML̃2 (ẽ∗L ẽL + ν̃L∗ ν̃L ) + Me2 (ẽ∗R ẽR ) + ¯i w̃i + M3 g̃g̃
M1 b̃b̃ + M2 w̃ ¯
h 2 i
+ ij AD H1 Q̃ dR + AU H2 Q̃ ũR + AE H1i L̃j e∗R + h.c.
i j ˜∗ i j ∗

(27.6)

Now we have a lot more parameters! But this is just the most general form and it might well be that the
various parameters are related to each other. We now need to patch this to the real world. The first thing
we want is that SU (2) × U (1) is broken, so both of the Higgses should acquire vacuum expectation values.
However the other fields should have vanishing vacuum expectation values. We want to have minimum at
 0    +  
h1 v1 h2 0
H1 = → , H2 = → (27.7)
h−
1 0 h 0
2 v 2

and their ratio is tan β = v2 /v1 = vu /vd . Now the supersymmetric part of the Higgs potential will be like
 
V (H1 , H2 ) = |µ|2 |H1 |2 + |H2 |2 + (DA )2 terms
  g 02  2 g 2  2 (27.8)
= |µ|2 |H1 |2 + |H2 |2 + |H2 |2 − |H1 |2 + H1i∗ τija H1j + H2i∗ τija H2j
8 8

121
Quantum Field Theory III Lecture 27

where g 0 sector come from the U (1) part and the g sector come from the SU (2) part. This is fine and
preserves SUSY and SU (2) × U (1). Now we add the soft part which breaks symmetry
      g 2 + g 02  2 g 2 2
VHiggs = |µ|2 + m21 |H1 |2 + |µ|2 + m22 |H2 |2 −µBij H1i H2j + c.c. + |H1 |2 − |H2 |2 + H1† H2

8 2
(27.9)
Now the terms that know about relative orientation between H1 and H2 are the third and last terms. We
want our previous vacuum expectation value structure, and that is good because when they are orthogonal
the third term is the most negative and the last term is zero. The result we get out of these is that
1 2
m2Z = g + g 02 v12 + v22 , m2A = 2 |µ|2 + m21 + m22
 
tan β = v2 /v1 , (27.10)
2
Remember we have 5 remaining fields, and we will call them h± , h, H, A. Two of them are charged and
three of them not. h and H are neutral scalars and A is a neutral pseudoscalar.
We can get a constraint on the parameters by requiring that V has a lower bound. The constraint is

m2A > 2µB > 0 (27.11)

We can also require that H1 = H2 will correspond to a local maximum, so we will get

(µB)2 > µ2 + m21 µ2 + m22


 
(27.12)

If we work out masses we will get that


1
µB = m2A sin2 β, m22 − m21 = (m2Z + m2A ) cos2 β (27.13)
2
and the masses of the heavy and light neutral scalar Higgses are
2
   
MH 1
q
2 2 2 2 2 2 2
= mA + mZ ± (mA − mZ ) + 4mA mZ sin 2β (27.14)
Mh2 2

Now the mass of the light Higgs will be lighter than the minimum of mA and mZ , whereas the heavy guy
will be heavier than the maximum of these two. If we do more we can find that

Mh2 < m2Z cos2 2β (27.15)

This is obviously a problem because any Higgs mass less than the Z boson mass has been experimentally
ruled out.
Let’s look at loop corrections. We have four H legs with a fermion running around the look, or a
scalar running around the loop. If there is supersymmetry the two loops will cancel, but now we broke the
supersymmetry. The largest correction will come from the top quark t and the correction should be about
!
Mt4 GF Mt̃2
∆∼ ln (27.16)
sin2 β Mt2

Now this estimate varies with time. Steve Weinberg said in 2000 that for tan β > 10 and Mt̃ ∼ 300 GeV
then Mh . 110 GeV. The Particle Data Group at 2007 gave that for Mt̃ . 2 TeV then Mh . 135 GeV.
But whatever estimate, the light Higgs mass is tied to the Z mass.

122
Quantum Field Theory III Lecture 28

28 Lecture 28
Recall the superpotential of the supersymmetric Higgs field can be written as

W (Φ) = µH1 H2 + λL HLE + λD HQD + λU HQU (28.1)

The soft symmetry breaking terms are


     
Lsoft = −m21 |H1 |2 − m22 |H2 |2 + Bµij H1j H2j + c.c. + MQ̃ 2
ũ∗L ũL + d˜∗L d˜L + MŨ2 (ũ∗R ũR ) + Md2˜ d˜∗R d˜R
1h ¯ i
+ ML̃2 (ẽ∗L ẽL + ν̃L∗ ν̃L ) + Me2 (ẽ∗R ẽR ) + ¯i w̃i + M3 g̃g̃
M1 b̃b̃ + M2 w̃ ¯
h 2 i
+ ij AD H1 Q̃ dR + AU H2 Q̃ ũR + AE H1i L̃j e∗R + h.c.
i j ˜∗ i j ∗

(28.2)

Now there are problems, first of which is that the Higgs mass terms are negative. This can be resolved
by requiring the couplings to run, and we need to work very hard to get only these two masses to run
to negative region, but not any other masses. The other problem, which is more important is that this
Lagrangian sets a new scale in our theory which is µ. And we need a natural way to set this scale at some
reasonable level.
We can look at the K 0 and K̄ 0 mass difference by working out the loop correction to the K propagator,
which is just a four-leg quark diagram. We can have a diagram with W as the intermediate particles
If we work out the amplitude carefully it will be

4 (m2c − m2u )
gW sin2 θc cos2 θc (28.3)
m2W

and this is more or less the right number.


Now we have a whole bunch of new particles we can have squarks and gluinos as the intermediate
particles
We will get an amplitude like
" #
Vsi Vdi∗ ∆Mi2 M̃g
4
gQCD < 1.5 × 10−3 (28.4)
M̃g 2 100 GeV

where the bound is due to experiment.


Another issue is that we will have processes like µ → eγ. We can have this process with intermediate
particles W̃ , Z̃ and ẽ, µ̃, τ̃ , ν̃. The branching ratio is about
 4
MW
α ∼ 10−6 (28.5)
MSUSY

This is because we take MSUSY ∼ 1 TeV. But the experimental bound is 1.2 × 10−11 . Now we can resolve
this by looking at the masses of the involved particles and look for cancellations. Anyway we need to
work much harder. Another issue is CP -violation. So even if we started out with more than a hundred
parameters, a huge part of the parameter space has been ruled out by these kinds of considerations.
Let’s look at the mediation between the standard model sector and the hidden sector which we don’t
know. We will be looking at messengers between the standard model and the hidden sector. Now these

123
Quantum Field Theory III Lecture 28

messengers will break the supersymmetry and give masses to the superpartners. If we look at the messenger
correction to the gaugino mass, the main contribution is like the following diagram
The gaugino mass correction can be worked out

(a) gS2 X X
Mgluino = C3k Mk , C3k = tr(ta2 ) (28.6)
16π 2 a
k

and similar for the wino and bino


g2 X g 02 X
Mwino = C2k Mk , Mzino = C1k Mk (28.7)
16π 2 16π 2
Now because the gluino masses are decided by the strong interaction scale, it has the highest mass. The
gluinos are the easiest to find because they couple to strong interaction. Now if the experiment can’t find
gluinos, it does not rule out the existence of winos and binos. In any case we get Mgaugino ∼ αΛ where Λ
is some messenger scale.
Now we can also look at the squark masses. According to the same reasoning we will have
2
Msquark ∼ α 2 Λ2 (28.8)

Now if the squarks are any good at solving the hierarchy problem, then they must have masses around
100 GeV. This translates to a messenger scale of about 30 TeV.
Let’s look at gravitino. The W boson mass in standard model is MW ∼ gv. Now in analogy we have
the mass scale of the spin-3/2 partner of graviton
√ 2
 
Λ
M3/2 ∼ GΛ ∼ Λ (28.9)
Mp
So if Λ = 30 TeV then the gravitino mass will be M3/2 ∼ 0.1 eV.
This is only one possibility of mediating between the standard model and the hidden sector. Another
mediation can be due to gravity. If this is the case then the soft broken superpartner masses will be
√ √
Msoft ∼ GΛ2 , M3/2 ∼ GΛ2 (28.10)

So the gravitino should have similar mass to the superpartner masses. Now the problem with this is that
in early universe the annihilate rate of these particles is too low, and it is easy to have too many of these
particles left now. And now that the superpartners obtain mass from gravitational coupling, then it is
universal and there is no way to solve the flavor problem, and also we need CP -violation.
Let’s get back to field theory. We wrote the Lagrangian of the supersymmetric theory as
1
L = Φ† e−V Φ + 2Re [ W (Φ)|θθ ] + 2 Re [ W α Wα |θθ ] (28.11)

θθθ̄θ̄ 2g
Now when we do renormalization we integrate out the high energy parts, and this Lagrangian changes.
We rewrite the Lagrangian as
1
L# = Φ† e−V Φ + 2Re [Y W (Φ)|θθ ] + Re [X W α Wα |θθ ] (28.12)

θθθ̄θ̄ 2
so if Y = 1 and X = 1/g 2 then this will reduce to the above Lagrangian. Now this has symmetries. If we
change X → X + iα, then because the imaginary part of W α Wα is proportional to F F̃ , it will not change

124
Quantum Field Theory III Lecture 28

Field R charge
θ 1
θ̄ −1
V, Φ, X 0
Y 2
Wα 1

the Lagrangian in perturbation theory. It only has effect in instantons. So this is indeed a symmetry in
perturbation theory. Now we also have R-symmetry. The R charges of the various fields are:
Now when we do loop corrections, the Lagrangian will be change into

L#λ = Aλ (Φ, Φ† , V, X, X † , Y, Y † , . . . )θθθ̄θ̄ + 2Re [Bλ (Φ, Wα , X, Y )θθ ] (28.13)

Here B is a chiral superfield and it does not depend on Φ† . Now due to R charge we should write Bλ as
the following form
Bλ = Y fλ (Φ, X) + W α Wα h(Φ, X) (28.14)
but due to X → X + iα symmetry we can conclude that

fλ = fλ (Φ), hλ = cλ X + kλ (Φ) (28.15)

In the limit of weak coupling we have Y → 0 and X → ∞. Now the interactions all come from the first
term in the Lagrangian, and it has equal number of Φ and Φ† , but we only have Φ in kλ , so we have to
have kλ equal to a constant.
If we look at Feynman diagrams the number of loops is

L = I − V + 1 = IW + IΦ − VW − VΦ (28.16)

If there is no external Φ’s then we have IΦ = VΦ and the graphs will be proportional to X N where
N = 1 − L. For N = 1 there is no loop and we have tree diagrams. Then cλ = 1. For N = 0 these are
1-loop diagrams then we have corrections due to kλ . So under loop corrections the coupling constant will
change according to 1-loop β-functions, and the Kahler potential can be changed, but no other. Specifically
W is not changed. So at tree level there is no SUSY breaking if we have ∂W/∂Φ = 0 has no solution,
and we know that this statement is true for all orders of perturbation theory. But non-perturbatively the
instantons can break SUSY because we can add F F̃ terms to the Lagrangian and the above X → X + iα
symmetry doesn’t work.

125

Potrebbero piacerti anche