You are on page 1of 97

Ph444 Quantum Field Theory

Winter 2012

Notes
March 7, 2013

Giordon Stark

Contents
1 January 8th, 2013 (Introduction to Fermions)

1.1

Fermions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.2

Group Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.3

Labels Representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

1.4

Symmetries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

2 January 10th, 2013

17

2.1

Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

2.2

Field Operators under the Lorentz Group . . . . . . . . . . . . . . . . . . . . . . . . 18


2.2.1

Motivation through Scalar Fields . . . . . . . . . . . . . . . . . . . . . . . . . 19


1
2

2.3

Fermions, spin

2.4

( 21 , 0) representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

2.5

(0, 21 ) representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

2.6

Lorentz Vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

2.7

1-fermion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

3 Jan. 29th, 2013 Spin-1 Vectors

29

3.1

Photons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

3.2

Lorentz invariance, unitarity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30


3.2.1

Lorentz Boost . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

4 February 5th, 2013


4.1

33

Renormalization of QED . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.1.1

Feynman Rules of Renormalization . . . . . . . . . . . . . . . . . . . . . . . . 36

4.1.2

Renormalization Conditions (on-shell scheme) . . . . . . . . . . . . . . . . . . 36


1

5 February 14th, 2013


5.1

37

Proof . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
5.1.1

Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

5.1.2

Schwinger-Dyson Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

5.1.3

Renormalized QED . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

5.2

General 3-point Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

5.3

Yet Another (not complete) Proof for Ward Identity . . . . . . . . . . . . . . . . . . 42

6 February 19th, 2013

45

6.1

Ward Identities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

6.2

Spontaneous Symmetry Breaking (SSB) . . . . . . . . . . . . . . . . . . . . . . . . . 49

7 February 28th, 2013 Spontaneous Symmetry Breaking (contd)


7.1

52

SO(2) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

8 March 5th, 2013

63

9 March 7th, 2013

68

9.1

Standard Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

10 March 5th, 2013

80

11 March 7th, 2013

85

11.1 Standard Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

List of Figures
1

Discrete symmetry D = 1 no SSB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

QFT II
1

JANUARY 8TH, 2013 (INTRODUCTION TO FERMIONS)

January 8th, 2013 (Introduction to Fermions)

I remember at the very beginning of the last quarter, I sort of draw you some schematic table, that
we can talk about particles, and the related issue in the following way. So you can classify them
by spin. Spin can be 0. What else it can be? 1/2. We are not talking about spin for the particles.
That is a very deep question. I dont know what, I dont know the answer. In principle it can
be as well but usually, maybe we will make a comment about that later. But even with this, it is
not very complete. What is wrong with this list? Every half integer, there is a spin. Why do we
usually stop with these two? The answer is that we only know how to write normal looking up to
spin 2. There is a reason for that, which we know we may not be able to get.
name

scalar

fermion

spin 1

??

graviton

spin

1
2

3
2

renormalization
EFT
quantization
s-matrix
symmetries
Particle data group, there is a particle data group. You Google it. Particle data. Used to be they
mail you a booklet every year. But these days they are all on-line. It is very fast. By the way, a
very good test whether you are any good as a particle physicist is to see how many pages of particle
data book you can understand not meaning just reading the English, but trying to understand the
properties.
There are a lot of confusing issues with quantum theory such as renormalization, EFT, quantization,
S-matrix, symmetries. But now in this quarter, we will mostly try to see what are the new issues
that arise in these categories for spin 1, you see that there are quite a bit of new things. We will also
have a chance to talk about normalization EFT again in the context of these series, just to give you
more examples of that. I think I also mentioned to you last quarter that historic Clee, quantum
4

1.1 Fermions
1 JANUARY 8TH, 2013 (INTRODUCTION TO FERMIONS)
historically quantum field theory is confusing. Is it more confusing than necessary because
unfortunately the first example of quantum field theory, everyone knows what is the first example
of quantum field theory, dynamics, is these particles. All issues tangle together. Historically, our
predecessors have to understand everything all together. That is why it takes a long time. Why
are these things interesting? First of all, they are seen in nature. Most of the particles we know,
elementary particles in nature, spin 1/2, spin one. We might have one elementary particle in spin
0, not so long ago. But before that, everything we know are in these two categories. Even people
suspect there is no spin 0 elementary particles. But we can go through some of those reasons if we
have a chance later. But we still dont know whether it is elementary particle, or not. Most of the
elementary are in these particles. They give you more differentiation of dynamics. We basically
are focusing on that. In this quarter we will basically do Fermian, as I said, and the spin 1. But
mostly, we will do this and that. And there is example, we will do this. We will see how much
those things already are taking up my time. But I do want to in the end, do a little bit of synthesis,
since I do want to talk a little bit about hicks mechanism, in the context of Q.E.D. Not our Q.E.D.
For the Q.E.D. in a different world. Sorry. Either for Q.E.D. in a different work or for Q.E.D. in
the BCS you conduct, inside of superconductor. That is sort of using all the knowledge, use all the
scalar, using the scalar field and using Fermian and Baxter. Also very, BCS series is also a beautiful
demonstration of effective field theory, by the way, if any of you actually read the journals. Lets
begin with fermions.

1.1

Fermions

Were not going to spend too much time because there is not too much deep about fermions and
there is quite a few things deep about scalars, about vectors.
It is very hard to keep track with consistent notation of everything deep. But concept is easy.
What we will do is not to talk about everything about Fermions, is to Ill try to talk about some
basic concept about Fermions, then I will try to basically focus on one special case that we will
need later. That is the Fermions we need which is the probably simplest one. But before I do that,
if you want to ever become a grand master of Fermions, here is something you can, you must have.
There is a review article by this group of people, okay. What else do we want to know? This is

1.1 Fermions
1 JANUARY 8TH, 2013 (INTRODUCTION TO FERMIONS)
a physics report. Its this big. You probably dont want to read it from cover to cover. But you
want to read enough so that you can, this has every single formula you want, that you ever need
about Fermions and also, if you actually do calculation with Fermions, you know that there are 20
different conventions and so on.
H. Dreiner, H. Haber, S. Martin physical report 494 (2010) [pdf]
They have all of them, and teach you how to translate between them, and I dont know whether
you have ever done latex, they have a macrofile that you can use your space time signature, so on,
relapse, the whole thing. It is a different convention, okay? But that is not the main point, why
this, that is one of the main point why this is useful. But the other thing, why this is useful is
these people are widely respected in the community for never make any mistakes, so everything,
anything I wrote on the board is inconsistent with them, they are right. I am wrong. In this
business, negative sign 1/2 is everything. There is nothing too deep about it. But you have to, if
you want to get the right number, you have to have, you have to get those things right.
What Fermion is, what distinguish Fermion and scalar is basically Fermions are representations,
different representation of Lorentz group. Why do we care about Lorentz group? Because its
symmetry. Its a space/time symmetry. Why do we, why is that obvious? Why is Lorentz group
symmetry, used for symmetry? Lorentz group is certainly not symmetry of this classroom. You
are sitting there. Im sitting here. We are certainly not in rotations. So its useful, if you think
about microscopic physics. When you go to a microscopically, we know for a fact that the theory
is, and we talk about the particles and that the scalars are the trivial representation of Lorentz
group. Fermions is the next representation basically. You will see that all these guys falls into
different representations of lower he wants group. It is interesting that nature use lower dimension
representations of loren t z group. As I said we have examples all of this, except of this. This one
we know how to write in theory but we havent observed. Nature has managed to use all of them.
That alone give you a strong motivation to suspect super symmetry is there. Now we are going to
talk more about representations and symmetry. Let me ask, nobody has never seen group theory
before, right? You heard of the word, group.

1.2

1.2

Group Theory

JANUARY 8TH, 2013 (INTRODUCTION TO FERMIONS)

Group Theory
groups symmetries equivalent theories

So groups. Im sure you can look up the definition of group. Im not going to do that. Usually
we think groups are in physics, is a useful concept to handle symmetries. We know symmetry is a
powerful tool in physics. What symmetry says in physics is that there are different series that is
equivalent. Different theories are equivalent. This is the meaning of symmetry. If you think about
it, it is quite a powerful statement. Although, since we brought up in kindergarten getting used to
symmetry you think it is true but it is highly nontrivial. It means the, the difference is, in principle
I can write a theory facing this way and I can write a theory facing this way. The statement that
these two theories are equivalent is actually highly nontrivial statement. It means including all the
quantum corrections, calculate all the loop, no matter what you do, these two theories are the same.
Without doing that very complete calculation, you can already just, if there is a symmetry you can
say those two theories are the same. What is the symmetry, tells me that the theory, this way and
theory that way, is the same? What is the symmetry? That is the rotational symmetry. Typically
three spatial dimensions, for example, symmetries like SO(3), its called, everybody knows what
this means. Special orthogonal group. Rotations. How many parameters in this group? Three.
There is three rotation. For group there are elements, group elements.
Given group elements A, B, AB, we can have generators TA , TB with something like A = eiTA A
(abstractly for SO(3)). In physics, we work with representations of a group. A representation is
something the group acts on. In terms of representations, group elements, its useful to think of them
as being matrices M (A) . Every group corresponds to a matrix acting on some object.
For example, in SO(3) we have a rotational matrix R for every group element. So
~r R~r
where ~r is a 3-dimensional fundamental real representation. In quantum mechanics, we talk about
two components spinors s (2-dimensional complex representation) which acts like
~
r

e i 2 s

1.2 Group Theory


1 JANUARY 8TH, 2013 (INTRODUCTION TO FERMIONS)
For representations, there are also two kinds of representations. Anybody knows what two kinds is?
You can classify them as so called reducible versus ir reducible representations. Its very useful to
talk about irreducible. All kinds of other things like [inaudible] representations, that is not talking
about those things. Irreducible representations, sometimes called irraps. What is a irreducible
representation meaning? Irreducible representation means that lets say I have a five members, it is
useful to think about all of this in terms of some finite representation. If I have certain five members
of this representation, and irreducible representation means I can act a full group on this irreducible
representation, it will go through all five of them. Nothing more, nothing less. If you take all the
elements of the group, acting on this representation, it cannot be a four-dimensional. It has to be
five dimensional. Lets say, if I, so this is irreducible representation, for example. But I cannot
just take acts, for example. Acts. It does not form a representation of the rotational group. But,
and but on the other hand, this three-dimensional vector spin a representation. You dont need
anything more. In principle, you can also a representation. But this is a reducible representation.
In quantum mechanics, we have, angular momentum. We learn angular momentum representations.
The one thing about irreducible representation is that they are all labeled by environs. A different
irreducible representation of a group, reduced, is labeled by environs. The value of those environs
or of that, of that group.
Generate all the reducible representations is a pain lets say SO(3). What are reducible representations of SO(3)? In QM, you will learn angular momentum and spin: L, m along with the
eigenvalues
1
l = 0 ,1
2

2 = l(l + 1),
L

which one definition is to keep ~r2 invariant. So all irreducible for SO(3) is labeled by l(l + 1), the
different values correspond to different irreducible representations.
SO(1,3) Lorentz group keeps x0,2 ~x2 invariant and has 6 generators. J , , = 0, 1, 2, 3 which is

1.2 Group Theory


anti-symmetric under .

JANUARY 8TH, 2013 (INTRODUCTION TO FERMIONS)

[J , J ] = i ([g J ( )] [ ])

1
Li = ijk J jk ,
2

K i = J 0i ,

(i, j, k 1, 2, 3)

(rotations) [Li , Lj ] = iijk Lk ,

(boosts) [K i , K j ] = ijk Lk

[Li , K j ] = iijk K k
It has six generators. Three of them are normal rotation. But there are also three others, what
are those three others? Boost. So a general definition, I can write a general definition of this.
It is orthogonal group. So mu, and then nu, these are generators, from 0 to 3. And this is ante
symmetric under interchange nu and mu. Okay? So there are six of them. Lets just take J nu mu.
J sigma, so algebra is defined by the ante symmetric commutation relations. G ro, J mu sigma
minus mu .... okay. Ante semmette Rickize between these two. Next term is ante symmetrickize
between this set of indices and this set of indices, in the commute eighter, so the next set of terms,
the same thing here. But ro sorry, ante symmetrickize between these two. These are, if you want
defines, the algebra of Lorentz group. If you want to write this in a slightly more familiar notation
... (pause). I can define Li, which is one-half and ki, which is just J0i. I havent done anything
else. Im just calling, take part, take apart this, I should say iJk, is 1, 2, is 1, 2 and 3. There is
special directions. That is the only thing else. There is no J00 because its anti symmetric. Im
just taking different elements of this nu mu matrix and calling it different names. Okay? It is likely
redefine a subset of them. This is just a change, Im changing base of this operators. If you do
this, you observe ..... (pause). Okay? So this is the most familiar part. That is, this is the whole
point of writing it in this way. Okay? This is SO3 group. Epsilon iJk is anti symmetric answer,
which is also the structure constant of SO3. This is SO3, rotation. And this is saying that the
true boost, so these are the boosts so if you actually write the representation out, you can see that
this is boost. The two, the commutation, 2 boost does not commute. The equivalent to a location,
okay? That is just a saying that. And then moreover sorry, yeah, strategically cause this to be
confusing, but this is upper this is a generator. Rotation boost also does not commute. But you
can write it in terms of boosts. The whole point of writing this, this way, is just to understand
how the rotational group is embedded in the Lorentz group. This writing is obviously less coherent
9

1.2 Group Theory


1 JANUARY 8TH, 2013 (INTRODUCTION TO FERMIONS)
to looking it as writing it, the physical meaning of this generators is better. Im sure you are all
familiar with representation. Most [inaudible] representation is Lorentz group, four vectors. Im
sure you can look it up in Jackson.
particles state
Now, what we care about are particles and fields, as a representations of Lorentz group. Okay? So
lets get to that. Right off the bat, its already pretty confusing. You will see there is already the
first confusion is that as I, you have to distinguish between a particle state and the field, okay, in
this case. State is not a field. State is a quantum mechanical state. A field is an operator acting on
the space spanned by those states. Okay? This is one of the main confusing points. A state, you
can write it in, write in for example in some representation such as momentum representation, or
space or position representation, that becomes a wave function. Okay? However, a field operator
is different from that wave function, although sometimes they tend to use the same symbol. That
is why it is confusing. Just keep in mind all of these things. The two Fermion state does not anti
commute with each other. The field operator does anti commute. So for example, okay? So lets
talk about particles first. Particles first. It means lets talk about state first before we go on to
the Fermion field operator. Again having to be careful to distinguish these two things, okay? Now,
Lorentz transformation, we want to build representations of Lorentz group. Okay? That is the
goal. Lorentz transformation on state, okay, and this is a symmetry of our theory.
Now, for our theory, for our theory with symmetry, the symmetry transformation on the state, you
can also represent it as, with a matrix. Okay? That is a representation of the symmetry. So what
is the requirement of the matrix? Yeah, the matrix has to be unitary on a state. Why? Because
quantum mechanics, the inner product of two states has a very important interpretation as the
probability. So if you think something is a symmetry of the theory, it is better, first thing is better
preserve probability. We can not have non[inaudible] interpretations. Lets say one state goes to
some U sorry for Lorentz transformation. Lambda, okay, not to be confused as the color we
talked about last quarter, because some people likes to use lambda. Its on the state, is just state
goes to some state. Okay?

10

1.2 Group Theory


1 JANUARY 8TH, 2013 (INTRODUCTION TO FERMIONS)
We want a Lorentz transformation on a state, so for a Lorentz transformation :
|i U ()|i
where U () is unitary by requirement.
Where this is a matrix, okay? This is a matrix which is unitary. So far I havent talked about
anything about Fermion. But for bosons, for scalars, we already derived what this matrix is.
Its a pure phase. That is obviously unitary. But for Fermions, this can be nontrivial. But
whatever it is, it has to be unitary. Another way, this has to be a unit airy representation of
unitary representation of Lorentz group. Unitary representation of Lorentz group. Off the bat
there is something already confusing because we know Lorentz group, Im stating all this fact,
Lorentz group is noncompact. Meaning that you cannot, the parameter of Lorentz group does not
live in the compact space. Okay? This part, obviously, lives on the compact space, the angles,
rotational angles obviously lives in compact space. But this part doesnt. It can have a, boost
parameter can be as big as you want it to be. This is noncompact. There is a theorem, we cannot
fight with the theorem, there is a theorem saying that for noncompact groups, there is no finite
dimensional unitary representations. Okay? So now there is a problem. So why does it make sense,
this make sense at all? Okay? This means that the representations, irreps(unitary) irreducible
representations are unitary, unitary representations which we know every particle state must be
in, is infinite dimensional, okay? Which you will see is indeed that is the case, they are infinite
dimensional.
Okay? I say that again, okay? Im repeating myself. I say it again. This is on the state. Okay?
When we talk about field operators, that is another set of state. Okay? This is on the state.
Field operator, on the other hand, there is no requirement of field operator be unitary. Okay? In
fact, the field operators field, finite dimensional nonunitary representations of Lorentz group. The
rest of a couple lectures are not really that much about physics, but we have to get through this
to be able to do Fermions. Of course, from this you can already say that you have two Lorentz
transformation, lambda and lambda prime. They are obvious like group theory relations like this.
That is obviously, has to be the case. That is just representations theory from groups.
For two Lorentz transformations , 0 , we have U ()U (0 ) = U (0 ).

11

1.3

1.3

Labels Representation

JANUARY 8TH, 2013 (INTRODUCTION TO FERMIONS)

Labels Representation

Okay. As we said, the different representations, different irreducible representations are labeled by
envariants by invariants under the symmetry transformation. Okay? So, we wanted to, so lets
begin by some state K which is labeled by the momentum. Not that K. Sorry. This is K. This is
momentum. Im sorry about all this notation. But maybe its a good exercise, because the whole
thing about the, the whole difficult thing about the Fermions are conventions, so if you get through
all this, you ... okay? This is momentum, okay? What is the most obvious invariant of Lorentz
group?

(momentum) |ki

k 2 = k k = m2 labels representation

First of all, we know that, so therefore each mass, and the particles with different mass dont
transform into each other. Okay? The particle with certain mass after Lorentz transformation still
have the same mass. This means that the different representations with different mass obviously
does not transform into each other. So all the irreducible representations are labeled by their mass.
They all are just different mass corresponding to different representations. This is one of those
labels, labels representations. There can be additional labels, as you see, I didnt call this irreducible
representation. There are additional labels we can assign to reduce, fully specify ir reducible
representation. How do we do that? I can tell you the result. But there is also a way systematic
way of doing it. Primarily devised by Eugene Wagner which is called induced representation or
commonly known as the little group.

start with n = (m, 0, 0, 0) little group of n is symmetric which leaves n invariant.


The (0, 0, 0) part is a little group of SO(3). Denote elements of this by R. Any vector k is a boost
L(k) : n k. Then
L(k) R L1 (k) leaves k invariant
The last frame also known as the rest frame for momentum. The little group, little group which is
a group of any vector, okay, any full vector, lets say little group of N, is, you know, is the symmetry
12

1.3 Labels Representation


1 JANUARY 8TH, 2013 (INTRODUCTION TO FERMIONS)
which leaves N invariant, okay? Once the symmetry leaves this invariant, the little group of this is
SO3, so little group is SO3. So, writing this way is easy to see this. But a nontrivial statement is
that the, actually, the little group of any arbitrary vector is SO3. Any arbitrary vector with mass
M is SO3. Okay? Lets try to see it. Lets do it here. There is enough space. Any arbitrary vector
in this representation, K, of any vector, K, I can obtain by, from N, by doing a boost. Lets say I
have a boost, lk which takes N to K.
Now, so this little group that keeps N invariant, I can just take any arbitrary element, without
SO3, is called R, R is an element of this little group. And you can see that, so obviously, R is SO3,
which leaves N invariant. But you can also see that the following okay? So, this acting on K
means that this take you back to N, and R doesnt do anything to N. R leaves N invariant. This
takes the N back to K. Is this okay? Is this okay? Okay. But you can show that this is also SO3.
All the elements look like this, this is also SO3. This is verified by doing whatever you want. Okay?
Commutation relations, whatever. So this tells me that the little group for any vector is always
SO3. Is this okay? Which is obvious. This is sort of obvious if you think about it. There is always
three notations that you can do to keep a fallback to invariant. Or said it the other way, you go to
the rest frame of the vector, rotate and then go back. So okay, now, lets think about, this is going
to do something to me eventually I think. (chuckles). Its okay. Now, lets think about, lets think
about how this is acting on vectors. Okay? State. Okay? Lets call a state N. As I said every state
is labeled by the vector, for vector. So this is labeled by this. But Im going to assign it another
indices, because obviously theres additional things that transform under about the states. So this
is what little group acts on. This indices, this individual characteristics about the state that the
little group acts on. This will keep A invariant, and this is what, F is reserved for little group. And
we already argue that the little group is just SO3. We know what, the irreducible representation of
SO3. Maybe Ive already erased it. This is labeled by the total angular momentum, and N. Okay?
Where L can be 0, 1/2, and so on. Okay? At least to start out this vector, we can write, we can
it is pretty clear to me that in this vector space, what rotation really is, representation is. Okay?
Its just that. I can write F or I can classify different offer according to L and M, if you think about
this additional quantum numbers, I can write this in terms of L and N.
where M (R) is a rotational matrix.

13

1.3

Labels Representation

JANUARY 8TH, 2013 (INTRODUCTION TO FERMIONS)


state

little group

SO(3)

R SO(3)

|n, i

L, m

U (R)|n, i = M (R)|n, i
Okay? Therefore, a, for R, said it the other way, if R is SO3, UR on this state, okay, unitary
representation of this state is just a R for beta R, I can write it as a matrix. And this is just
nothing but the rotational matrix. Im just writing this so that you dont feel this is very mysterious.
This is the rotational matrix you know and love, okay? So spin 1/2 is just the E to the Is sigma
without theta we wrote earlier. Okay? Lets ask ourselves, armed with all this information, lets
ask ourselves what is U lambda? General transformation, on a general state vector. Okay? This
is the last ingredient we need to build irreducible representations. Okay? To do it in a very, how
shall I say, yeah, we will write it in a slightly different way. Okay? So this is the last sort of
nontrivial step. UL Ill write it. Then Ill explain. Okay? So this is, looks very weird. But to
understand why this is the case, all you need to know is the K, lambda K, okay, obviously, this
Lorentz transformation is going to take K to lambda K. Okay?

U ()|k, i = U (L(k))U (L1 (k)L(k))|n, i


boost

rotation

with the Wigner rotation


W (, k) = L1 (k)L(k)
k = L(k)L1 (k)k
= L(k)L1 (k)L(k)n
So you just have to understand, at least this will happen. Lambda K you can write it as L. That is
obvious. This cancels that and that. But I still think its not complicated enough. I want to make
a contact with the invariant factor. Now Im going to call, separate this out, and thats over there.
Now inside, remember that I told you that for two consecutive Lorentz transformation, this must
14

1.3 Labels Representation


1 JANUARY 8TH, 2013 (INTRODUCTION TO FERMIONS)
be true. This is any representation. Using these two, I get that. Is there any questions about that?
(pause). Let me write a little bit further, okay? But so far, whether I write F or not it doesnt
even matter. Lets understand a collection [inaudible] but otherwise there is no problem with this
one. But why do I want to make this thing more complicated than necessary? What you see is
that this is a rotation. Take any vector. This act on vector is just boosted to K. And transform
boost it back. Its, therefore, the end result is a rotation. You can verify again, compute. This is
a rotation, and this is a boost, by definition because I use L to denote boost. Is it okay? So, this
means, why do I want to do this? When I, the first time I see this, I was totally confused. Why are
you bothering to do this? Let me write this as a rotation. Okay? Lambda and K, let me give you
the notation. Lambda K, and this is called a Wagner rotation, Mr. Vigner write it down first. So
this shows any unitary transformation. Unitary transformation of Lorentz group. This is not very
precise mathematical language. Any unitary transformation of Lorentz group can be obtained, can
be obtained from N, alpha by boost times rotation. Wigner rotation first, then do boost.
So any unitary transformation of Lorentz group can be obtained from n, by a boost times a
rotation.
Okay? We already argued that the alpha here is, it can be classified so alpha is just irreps of
SO3. It is a label that, labeled by reduce representations of SO3. Suppose Im already in some
ir reducible representation of SO3, labeled L and M. Labeled by L. Labeled by the total angular
momentum. Rotation is not going to change that. Boost is not going to change that either. Okay?
This means on the general Lorentz vector, any general Lorentz vector, if Im already in the ir
reducible representation for this vector, Im also in the ir reducible representation for any Lorentz,
general Lorentz vector. Boost, you do not change the mass. This basically tells you that there are
two labels of invariant labels of any reducible representation. Boost invariants is M, and rotation
invariants is L.
irreps of SO(3) L. All irreps are m and L (0, 21 , ...). All members of that irrep |n, i boost
times rotation. Infinite dimensional, infinite number of k, k 2 = m2 .
Moreover, how to write all the state vectors in that ir reducible representation. So all state, yeah,
so let me say that again. So all irreps are abled by mass, and L. And all members of that irreducible
representation, can be obtained by starting from NLL and do rotation, do L times sorry, do boost

15

1.4 Symmetries
1 JANUARY 8TH, 2013 (INTRODUCTION TO FERMIONS)
time rotation. In this sense, so you can obtain everybody in the irreducible representation, you can
induce everybody in that irreducible representation by doing a little group rotation, and a boost. So
this is the so-called induced representation method. For us, it is not necessarily absolutely needed
that we have to do this all the time, because irreducible representation is known. But I thought
it would be nice to understand this method at least once. It is very useful, and a nice method.
Okay? So, okay? So all the Lorentz irreducible representations will be NML and L again is a 0,
one-half and so on. This is obviously the Fermion and the 1 is vector and so on. What else? As I
promised, I said that so this is unitary representation. But it must be infinite dimensional. Okay?
Because Lorentz group is noncompact. How do I see its infinite dimensional? Representation for L
is finite dimensional. Okay? Infinite dimensional comes from there, so infinite dimensional comes
from, there is infinite number of K, of momentum K, sorry, momentum K which is set by K score
and N score. There is indeed infinite number of members in this representation. And that infinite
numbers are related to each other by exactly the noncompact part of the group which is the boost.
You can go to any K from N via boost. Is this okay? Okay. So, next time we will focus more on
Fermions. So this is a general representation of Lorentz group. Again, if you havent received the
E-mail, and want to be on the E-mail list, let me know. And send me an E-mail and Ill add on to
it. I will see you on Thursday.

1.4

Symmetries

This is a little bit of a tangent. There are two kinds of symmetries. Kind of symmetry, this kind
of rotational symmetry is one kind. And as we will soon talk about generalization symmetry, and
but we also have concurring, translational, physics theory right here. It is the other end of the
universe is the same, maybe its the same. That is called concurrent symmetry. Translation is also.
But these are called space/time symmetries. So on, so forth. There are other kinds of symmetries.
Another kind of symmetry which we sometimes called a flavor symmetry. For those of you who
learn a little bit of standard model, this is not to be confused as the narrow sense, the actual flavors,
although that flavor is example of this kind of flavor symmetry. But flavor symmetry is some kind
of internal symmetry. An example with complex scalar being
ei

16

1.4 Symmetries
1 JANUARY 8TH, 2013 (INTRODUCTION TO FERMIONS)
which rotates between the real and imaginary component of the field .

17

JANUARY 10TH, 2013

January 10th, 2013

Last time we started to talk about how to build a representation of a Lorentz group using the
little group method. We talked about, for a mass of particle, we can start with a single vector,
a special full vector the little group is the one that keeps this vector invariant [SO(3)]
n = (m, 0, 0, 0)
Therefore, reducible representation of mass of particle has two things one is its momentum which
satisfies k 2 = m2 . There is another label here which is irreducible representation corresponding to
the rotations of SO(3).
|k, i

(L, m = L L)

This irreducible representation is obviously classified by the momentum, angular momentum, and
m (not the mass of the particle, but depends on L). That is a full representation of, so in other
words, the particles, particles in form representations of the Lorentz group, we can distinguish them
by their mass, by their spin and there are infinite number of members in irreducible representation
as they are infinite number of case that satisfy this. So it is infinite number of dimensional representation which is a little bit different from the usual representations where we used to have the
representations of rotations and so on. Those are finite dimensional ones. This is, everything is
consistent, as we said that we need the unitary representations of Lorentz group, the states. The
particle states has to form unitary representations because we want to conserve property probability. On the other hand the Lorentz group is not compact because there is a boost and the
parameter space of boost is not compact. It is a 0 to infinity. Therefore, there is a theorem saying
that you cannot have a finite dimensional unitary representations of a noncompact group. But
this, everything is consistent, this is unitary but its infinite dimensional. This has to be, this needs
to be infinite dimensional because we know particles can come from infinite number of different
momentums. So everything is, is this okay with everyone? This is as group theory will be, okay?
So in this course, I think, it is useful at least to remember a little bit of that. The next thing, so
this is a mass of particle. This applies to any particle. It doesnt, it is not even just Fermion or
anything you want. As we said, so far we know how to write nice looking field theories for these

18

2.1 Example
2 JANUARY 10TH, 2013
kind of spin par. For spin higher than 2, we can write down something but it is not very nice.
3
1
L = 0, , 1, , 2
2
2
massless particle For massless particle, for massless particle there is also a special vector, momentum
vector of massless particle. But its not this one. It does not have a frame, this is the full vector for
vector particle and this one doesnt. What is the special vector. Anybody know? Is what? That
is as special as it can be. You can put this anywhere. But lets put there.
k = (k, 0, 0, k)
where the (0, 0, k) is SO(2) - also a rotational group. Its rotation in two dimensions, not three
dimensions. How many generators are in SO2? Just one. Just one rotation. So this, the irreducible
representations of this guy, is also characterized by L. Okay? And the projections. Okay. The only
thing different is that there is only two projections here. (L, L). Spin up, spin down.

2.1

Example

For a massive particle with spin 21 , we have possible values 12 . For a massive particle with spin
1, we have values 1, 0, 1. For a massless particle with spin

1
2

we have values 21 . For a massless

particle of spin 1, we have values 1. (This last part underlies the Higgs Mechanism the main
difference).

2.2

Field Operators under the Lorentz Group

How does a field operator transform? Under the Lorentz group. As I said, again, last time, its major
difference here, there are two. A is that the field operator does not need to be unitary. It is not a
state. It does not have a direct connection with a probability. It does not need to be, whatever unit
representation the field operators lives in, it does not have to be a unitary representation. These
is that we usually look for only finite dimensional representation of field operators. The reason is
the following. For particles, I said we can have infinite number of possibility, infinite number of
different momentums. We want to have infinite dimensional representation for particle. For field
operators, usually we only have finite number of fields to talk about. We have electrons. We have
19

2.2 Field Operators under the Lorentz Group


2 JANUARY 10TH, 2013
photon. We have infinite number of different protons. Different universe but usually we talk about
finite number of field. But again, you can see that everything is still consistent. Im looking for a
finite number, finite dimensional representation of this group. And it is nonunitary. Again, it is
consistent with the fact that the Lorentz group is not compact.
For general field (do not need the -label for a scalar)
(x)

M 1 () (x)M () = M (1 x)

The field operator is a function of X. Usually the definition of Lorentz transformation is the following. Okay? Transformation, is realized by a matrix. Its a matrix. Lets try to motivate this
a little bit. Lets first imagine, there is a scalar. Okay? There is nothing scalar so this does not
exist. There is at most .... for real scalar there is not even a phase. I must say this.

2.2.1

Motivation through Scalar Fields

So for scalar, there will be just fi X will go to Phi, okay, this is a particular way of
(x) (1 x)
We are just imagining that if I imagine there is some field configuration, here, okay. Im just taking
this to a Lorentz boost, Im just taking it here. Therefore, the new Phi at the particular point will
be the older Phi as if you do inverse on X. Im trying to explain this but you have to think about it
yourself. This does work. This is called the active way of thinking about symmetry transformation.
Transform the field. Another way of doing it is called passive way. You can try to think about
rotating the axis.
My way of thinking about the transformation is that transforming the field (shifting it) is effectively
equivalent to shifting the field operators instead for a specific transformation .

So, lets see what are the possible representations. To do this, we are looking for something else. We
are looking for a finite .... to do this, its useful to recognize the following fact. So again, remember
that we, during the last lecture we write the general formula for the Lorentz transformation on
20

2.2 Field Operators under the Lorentz Group


2 JANUARY 10TH, 2013
generators. Okay? As J, we write them as JN mu but we think it may be more useful to write them
down as the following, use another different base. Writing down them into, putting them into two
class of generators. IJK is from 1, 2, 3. This kind of Greek letter is also 0 to 3. This is a rotation.
This satisfied algebra is 03 and these are the boost. Because there is 1, 2, 3, sometimes we can just
use the normal 3, three dimensional vector notation by putting an arrow.
1
Li = ijk J jk
2
1 ~
~
+ iK)
J~+ = (L
2
[J~+ ,J~ ] = 0

k i = J 0i
1 ~
~
J~ = (L
iK)
2
j
i
k
[J
, J
] = iijk J

This is the thing that always bugs me about group theory and in general mass there seems to be a
lot of tricks going on. There is. You have to be smart to be a mathematician. You have to somehow
stare at these things and notice all these facts. It is not like us, we start with .... ([inaudible] The
merit of doing this, is that you observe that this is zero. You separated the algebra, the algebra
into two parts that actually commute with each other. In other words, the full algebra is that sum
of sub algebra generated by J plus, another sum of sub algebra generated by J minus. Separately, J
plus/minus I, J plus/minus J, the algebra, sub algebra itself is SU2 or SO3. So the two sub algebra
here, so this way means that Lorentz group is well, decomposed into the algebra, the algebra of
Lorentz group can be decomposed into a sum of two sub U2, SU2 sub algebras. Hmm? (pause).
K is not compact. The stuff generated by K is not compact. Whatever a symbol you put it here, I
dont really care. But how you tell that this is what it means, okay, this is what it means.
Lorentz SU(2) SO(2)
So, therefore, all the irreducible representations, all the irreducible representations can be labeled
as a product of this plus and minus. All the irreducible representations are labeled by a quantum
number in plus and a quantum number in minus. That is all I wanted to say. The quantum number
in plus and minus, or in SU2, or SO3, same locally, are easy. Right? So everybody knows its the
Eigen value of 1/2, or Eigen value of, so the Eigen value of SU2, SA plus or minus is labeled by
J plus/minus. Again, this is where this Eigen value is 1 where N is again 1, 2.... okay? 0 as well.

21

2.3 Fermions, spin


That is trivial.

1
2

SU (2)
irreps (j+ , j )

(2j+ + 1)(2j + 1) dimensional

j =

JANUARY 10TH, 2013

n
(n = 1, 2, )
2
i
i
Li = j+
+ j

All the representations can be denoted as J plus, J minus. Okay? That is a product of these
two numbers. As long as I specify the quantum number under plus and minus I specify the
representation. These are all finite dimensional representations. Okay? Again, its different. That
is not, Im not looking for infinite dimensional representations anymore. Okay? So this, the
dimension is dimensional representations. Okay? (pause). The different, it is not here the term.
Lets see how they transform. Very deep. (chuckles). Thank you. Okay. Now, of course, we know
that you can also have a better way of representing this. Well, lets just do it.

(0, 0) : spin 0
1
1
( , 0) : spin (fermion)
2
2
1
1
(0, ) : spin (fermion)
2
2
1 1
( , ) : spin 1 (4 vector)
2 2

2.3

Fermions, spin

1
2

Okay. Now, lets go to talk about Fermions. Lets talk about these two representations. Field
operators in these two representations. (pause). 01, and 10, these are not irreducible. You have
to yes. So, yeah. Okay. Lets do Fermions. Firmiance Fermions. We are almost through,
with Fermions. Okay, so spin 1/2. So these two representations, lets just look at them. This
decomposition is useful, because I can immediately write down the transformation rules, because I
know how to write. I know how to write down the generations in arbitrary number of dimensions, so
thats all known. But before we do that, lets make another sort of general comment. In Fermions,
we are used to think about it, its a column vector, complex column vector, two-dimensional complex
column vector. This is what we usually think about. On the other hand, we know the location of
group is SO3. Whatever locally is SO3, it is rotation. There is a little bit of mystery here, because
22

2.3 Fermions, spin 21


2 JANUARY 10TH, 2013
SO3 we know its real. So why is more convenient to think about of course, you can also say, its
SU2 why its more convenient to think about SU2 than SO3, because SO3 is not single value in spin
1/2 rotations. So as you learn, it is a great mystery why you rotate 360 degrees, get back maybe
one or not one. So SO3 is not single value. But what we will do is to lift SU3 into SU2, so SU2, so
the complex representation, fundamental representation of SU2, that is single value under the spin
1/2 rotation. Thinking about complex representation is always more convenient for Fermions. We
are going to do a very similar thing. In mathematical language, there is another fancy name, its
called universal covering group, use universal covering group. For SO3, we will still want to use a
complex, use this covering group to describe Fermions.

SO(1, 3) (1, ~ ) =

X =

X is hermitian. Det(X) = Lorentz invariant.


We dont want to use literally as it is. This is a Lorentz group. What is that? So again, we can
try to identify it, by the following trick. So, again, so this we want to map it onto some complex,
complex, a group with complex parameters basically. So the trick, again, is to say, consider the
following quantity. Consider the following quantity which I call X, well, sorry. Obviously it has to
do something to do with the poly matrices, let me call it signal mu. This is the normal three signal
matrix and it does identity by two matrix. This form complete base of 2 by 2 is possible complete
2 by 2. (coughing). Matrices. Now, you form the following from it. You form the X, that is a
sigma mu, X mu, this is just some normal space time coordinate. Space/time format. So there are
some things, several things, X is a formation. Okay? Because these guys are permission. And the
determinant of X is just okay? Can just work it out. Therefore, there is a Lorentzian variant
Lorentz invariant. (oh, thanks, thats a nice new word). X, whatever it is, it is a matrix. Its a 2
by 2 matrix. Oh, X. Ki. How do you write X? Those are regular space time 4 vector. Whatever
it is, it is a 4-vector. Okay? This transform as SO3 too SO13. Now, how do I transform this
2 by 2 matrix, that is satisfied these conditions. There is also a set of transformation on these
2 by 2 matrix that satisfy these kind of conditions, which preserve, which makes this Lorentz
value, preserve the determinants, and everything. So the transformation, that thats consistent
with Lorentz transformation, has to satisfy any transformation on, like this, that satisfy the, be
23

2.4 ( 12 , 0) representation
2 JANUARY 10TH, 2013
consistent with Lorentz transformation, that defines a set of transformations.
X 0 = M XM 1

M SL(2, C)

This is the universal cover we are looking for. An analogy in QM is where


x = ~ ~a
i

rotation: e 2 ~ xe 2 ~ SU(2)
~a R()~a
You say that, you say there is a vector, A. You do a rotation. Okay? But what you want to do
is to ask, what is this rotation, how do you translate into SU2? How do you translate into SU2?
What you do is, you form this combination, and ask how does this combination transform another
rotation. This transformation transform as, and this is SU2 rotation. If you remember, okay? You
can show that this, another way of saying, if you show this, you can show that this exercise in
quantum mechanics is that this, that after this, A goes to a rotation by theta of A. It is based on
vector. It is rotated by theta. Anybody knows what blog sphere is? Yeah, but so Im just saying
that this is a common trick of mapping a real rotation into a universal cover group. In quantum
mechanics, the question is how to map SO3 into SU2. What you do is to form this combination,
and do the rotation on this combination. That is SU2. In particular, this is indeed a rotation
because this and that acting on X, in the end, is equivalent to just rotate this vector by theta. This
is the mapping between SU2 rotation and normal space time rotation. Its very similar.

2.4

( 21 , 0) representation

Again this means that its 1/2 under J plus, is 0 under J minus. That is all it means. So the field
operator we call it C. Probably not the best. This is my way. If you have another way, use your
way. The official way is, I dont know what is this. Whatever. Okay? The question, the problem
with this is, another one is zeta which looks like that but never mind. So because of this 1/2 0,
so there is the infinite, that just shows you how much less people try to put to make Fermion
rotation because this is 1/2 0. Because later on we will talk about 0 and 1/2 representation. So
in this representation of course these guys, these are two components. Two components, complex
in general. Okay? So we are labeling the components this by alpha. Alpha is 1 and 2. We only
24

2.4 ( 12 , 0) representation
2 JANUARY 10TH, 2013
are allowed to use alpha, not anything else. Alpha beta maybe. You will see where it is coming
from later. Under the Lorentz transformation, under Lorentz transformation goes to M, M is the
complex 2 by 2 we identified that way. We can identify in, well, you dont even have to go through
the exercise. But if you do, well, that is the 2 by 2 matrix.
M
And we write alpha, beta. Okay. Notice these two are strategically located, and this one is
closer to and this one is further away from M. They have a lot of meanings, because this means
that this, this index is closer to that. In this case it is obvious. In some cases it is not very
obvious. This is not even very, not even very easy because usually beta will be just here. Anyway,
everything has a lot of meanings. You will see what. Okay. Alpha and beta are 1 and 2 obviously.
Immediately, the one thing you want to see for any transformation is what is the invariant tensor
in that transformation.


12 = 21 = 1

The form invariants, the point of doing upper and lower indices, easier to form invariants. Okay?
You identify, suppose I define something for epsilon alpha beta, where alpha, 12 is negative 21.
Answer is 1. Have to be careful with this matrix. Sometimes people use the other convention. That
is why convention is everything in this business. So
 M M = 
This has alpha beta. It will also transform under this. Lets make it also transform under that.
You can show that epsilon alpha beta M, okay, now alpha is on this side, beta is on this side.
Gamma on this side. This you can show is just a determinant of M, times another epsilon tensor
basically. But then the determinant is 1. Because S over there, SLC, you see it, and so this is
okay? In group theory language, this means that this is a numeric tense or. This is like in Lorentz
transformation with , g .
= 

(M 1 )

= 
= 
= 
=

25

2.5 (0, 21 ) representation


2 JANUARY 10TH, 2013
First you can do this. There is negative sign because these are Fermion field operators. The anti
commute. Okay? The next thing I do is to flip this. This is okay? Now if you see 2 components
Fermion field, you will flip it, the same sign. That is because there is two negative signs. We can
also just write down an expression of that. At least in infinite is malform. Just because we know
the generators in this, acting, how generator is acting on this representation, right? J minus is
trivial. J plus is just to spin 1/2 rotation basically. But with this generator. In this representation,
in this representation L is just normal Paulie Pauli matrices. Okay? On the other hand, the K
is negative I. This I is crucial because K is [inaudible] M alpha beta, in this small transformation
is just this. Sorry. This is velocity. This is the velocity of the boost. Beta is V over C but we
take C as 1 usually. But this is beta. This is again the hermittian part. This is a noncompact
part. You can see that too just from this. If you wonder how a 1/2 spinor, 2 by 2 spinor transform,
it explicitly is like that, for this. Now you can transpose hermittian conjugate, that is something
else.
~ = ~ ,
L
2

2.5

~ = i ~
K
2

M = 1 i

~ ~ ~ ~
,
2
2

~v
~ =
c

(0, 12 ) representation

Now we talk about 0 1/2. Okay? Again, this is just taking the J minus, and switch the ro and
J minus and J plus. You say okay, that is trivial. That is trivial. But usually, there is, because
obviously, you can write down some relations between these two representations. That is why
sometimes it can get confusing. Okay? The fundamental, denoted by something called dagger
just to confuse you. Because its on this, it is Lorentz transformation on the right, if you want,
on this side. And 1/2, so I cannot use alpha anymore, because alpha I used to label the other
representation. Okay? So some genius come up with a representation called a dot representation.
Is it called alpha dot. That is not my fault. I have to tell you what the convention is. If I invented
weel it would be more confusing Weyl. At this point there is no meaning. Im not implying
this is a dagger of anything yet. That is just a notation. This is where I start. On the other
hand, obviously there is a root of this notation, where does this come from? You can show that
this representation and 1/2 0 presentation are conjugate of each. More specifically, which was

26

2.6 Lorentz Vectors


this,

JANUARY 10TH, 2013

= ( )

(M ) = (M ) = M
Now, its very easy to just derive the transformation laws of this, in terms of the transformation
laws on that. So, goes to M star alpha dot, beta dot. And this I can slightly rewrite it to make it
more hermitian conjugate, beta dot, alpha dot. See the subtle difference between these? It is all
very important. Or if you are lazy, write it like this. But you have to be very careful. You have to
understand what this means. Even this stuff sometimes there can be confusion. Some book decide
not to use this, they use bars. Also, you can in terms of this, in terms of this matrix, you can also
work out what is this guy. We formed earlier, we formed earlier how does this transform. Okay?
It just transforms as this close to M dagger on the other side. Therefore, you can figure out how
this transforms, how this guy transforms.
X =

M XM

M 0 0 (M )

() = =
As I said, I havent, so far we havent said anything really really deep. But there is all this
bookkeeping with Fermions you have to go through, because there is two kind of representations,
there is anti commuting itself, epsilon tensor and so on. Obviously, you also have sigma dagger.
This, okay. That is another invariant. You can define epsilon tensor with dot indices as well. This
is just dot. And this is ..... okay? These things are very useful, because so these are first set of
Lorentz invariants you can write down. Lorentz invariant is very important. You want to write
down LaGrangeians. This looks like a mass terms, because as one Fermion times another Fermion.
But we can also have other Lorentz invariants.

2.6

Lorentz Vectors

= (1, ~ )

(1, ~ )
27

2.7 1-fermion
2 JANUARY 10TH, 2013
The obviously, the right thing to do, Im going to write down the right thing to do, is the following:
Now here is the first time we meet with bars, sigma mu bar is one. You can show this transform
is like a vector. This transform, these are Lorentz vectors. So is this one. They transform vector.
That is all I want to say. Okay. Very good. Now we are in a position to write down our first
LaGrangian.

2.7

1-fermion

Okay? Lets have one Fermion. Lets just have that. Okay? Nothing else. This is Lorentz invariant
because that is a Lorentz vector.
1
L = i
m( + )
2
This is a mass of Fermion. Ill write it down. Then Ill make a few comments. You need this.
You need this, because this is the hermitian conjugate of that, first of all. And why isnt Fermion
times Fermion not vanishing? Usually I would say Fermion times, because exclusion principle I
cannot have the same twice because this is different Fermions. There are two component in this,
two component in this. One component of that times second component of that. They are different
Fermions. That is the little bit, it is another confusing thing if you just write them like this. But
usually its much better. That is a minor, lump into this rotation. Okay, several things. Lets
consider a limit, where mass M goes to 0.

limit where mass m 0 symmetry eiF fermion #T hemasstermbreaksthissymmetry.F ermionmassbrea


Remember the difference between this and the scalar mass. What is scalar mass? What does scalar
mass do? Or Phi dagger Phi, whatever it is. This breaks no symmetry. If you met the symmetry
that this breaks, let me know. (chuckles) you met a symmetry that is also consistent with the
kinetic term that this breaks, okay, because you always have to have a kinetic term. Symmetry
is consistent with this, that this breaks. This is very important. This is a fundamental, this is
fundamental importance.
scalar: m2 2

m2
28

()2

2.7 1-fermion
2 JANUARY 10TH, 2013
Second of all, lets say do we know an example of this in nature? Something that we usually the
answer is no. Okay? The Fermion, answer is no in nature, model is neutrino okay. We dont know
whether it, it is consistent with this but we dont know. Electrons allow this. So this set of, this
kind of Fermions are called, because it only has one Fermion, it is its own anti particle these are
called Majorana Fermions.
Im doing the whole Fermions, if you notice, a slightly different from the older textbooks. Old
textbooks start with the same but I assure you Majoranna is much better the two components
you use, the language is much better. The two components, this kind of two components language
and this is a generic thing called a Weyl Fermion (2-component). If you only have one Weyl Fermion
its Majorana, in four dimensions.
Consider 2 species ( 12 , 0) fermions i , i = 1, 2.
L=

X
i=1,2

1
i
m(i i + i i )
2

limit m1 = m2 = m. SO(2) or U(1).


current: J = i(1
2 2
1 )
Flavor symmetry, rotate to each other. There is SO2, or U1, symmetry, rotate these two into each
other. Cosine, sine theta, because again this is invariant. Only in this limit. In general, its not
true. If you follow the procedure, conserve current of this is J mu 1 sigma, 2, so minus, that is the
current of that. This is not a very again, this is SO2 language. This is the SO2. In this case you
will see very explicitly why this SO2. But again, doing this kind of business, SO group is usually
not optimal. It is useful to think about U1 rather than SO2.
1
1
(1 + i2 ), = (1 i2 ) J =

2
2
where J is U(1)Q charge current. This theory looks like QED with electron and positrons.

29

JAN. 29TH, 2013 SPIN-1 VECTORS

Jan. 29th, 2013 Spin-1 Vectors

Spin-1 vectors: photons , W and Z vectors, and gluons. These are called the force carriers. Why?
Can think about two electrons interacting via a photon intermediary. The force carrier refers to
the fact that you can draw this kind of diagram, saying that there are scalars instead of vectors
using the Yukawa theory talked about earlier. There are other vectors, QED vectors, such as
, 0 , a, .
In Effective Field Theory, the cutoff is much larger than the mass of the vectors. In QCD, the
cutoff is approximately the same as the vector, about 1 GeV. Also, last quarter, we said that for
theory that has a cutoff much bigger than the characteristic mass scale of the theory, there is a
pretty nice name to it renormalizable relevant operators are small, suppressed by some large
cutoff.

3.1

Photons

What kind of theory describes vectors? Gauge theory. Lets discuss vectors and focus on photons for
now, E&M. One thing we learned is m = 0. This is spin 1 and for massless spin representation
for Lorentz group, we see that a little group of SO(2) [irreducible representation] works here.
Identify a special vector k = (E, 0, 0, E). This is easiest because mass is zero. Why SO(2)? Two
polarizations ( has 2 physical d.o.f). We can also see this from Maxwells equations - dF = 0 or
F = 0.
F = A A
What is the solution of the wavefunction of the photon? It is in terms of this vector potential with
some polarization vector
A =  eipx
In momentum space (p-space)
p F = p (p  p  ) = p2  (p )p = 0
So there are two kinds of solutions.
30

3.2 Lorentz invariance, unitarity


3 JAN. 29TH, 2013 SPIN-1 VECTORS
p (p)
0

1.  = p2
F = 0 identically longitudinal polarization
2. p2 = 0, (p ) = 0 transverse
We can see that there are two polarization vectors that satisfy this kind of condition. But we can
always add another term because p2 = 0
 = + + +  + p

3.2

Lorentz invariance, unitarity

We always wanted to do things in the manifest Lorentz invariant way so we dont have to check
every time. How do we describe a vector, writing down, writing it as a Lorentz invariant, as
possible. Obviously the way to do it is promote it to a full vector. This usually requires us write
things in terms of A mu. Not A1 , A2 but A . Not just two polarizations. This has 4 degrees of
freedom. But we immediately see there is a problem here we know there are only two physical
degrees of freedom which means that we have a redundant description. There is something about
this theory that needs to tell me that I can remove this redundancy. That is the gauge equivalence,
gauge invariance. Redundant means there are also many equivalent descriptions of the same theory,
because there are two physical variables here.
We know A  eipx so how do we reduce it?
1. impose p  = 0 take 4 d.o.f. 3 d.o.f.
2. physical  =

1
2

(0, 1, i, 0) a quantization, but this is not manifestly Lorentz invariant.

Secretly, there is Lorentz invariant theory, so what is the mechanism? What is the secret of keeping
the theory Lorentz invariant?

3.2.1

Lorentz Boost

  + (p)p

(p ) = 0

We require the theory to be Lorentz invariant which means  and  + (p)p is the same physical
state. It refers to the same particle, just view it in two different Lorentz frames. In a more familiar
31

3.2 Lorentz invariance, unitarity


way, the theory stays the same if

JAN. 29TH, 2013 SPIN-1 VECTORS

A (x) A (x) + (x) A


This is known as the gauge transformation of the photon field. It is also known as a gauge invariance,
((

((
gauge equivalence, (
gauge
symmetry,
gauge redundancy. Speaking roughly, it means the true theory
(((

using a path integral has to be identical


Z

D(A)S [A]

identical

D(A )S [A

Gauge symmetry is not a symmetry. It is a redundant theory. I think I sort of made this point
last quarter but let me make it again. It is very important to understand. We have two, we have
real symmetries, such as global symmetry. One global symmetry is rotational symmetry. I can just
rotate. Rotational symmetry means I have two different theories N. one theory I call this X, this
Y. In another theory I call this X, this Y. These two theories give you the same, if you calculate
the same physical quantity they give you the same result. That is a symmetry. In principle these
are two theories, and thats, symmetry need not hold for any particular physical system. It is a
physical statement, whether I have two theories are different or not. On the other hand, we see
that this is not a symmetry. It means that this is just the same theory. I have infinite way of
describing the same theory. Talking about the same physical degree of freedom.
tl;dr Gauge Redundancy is the best terminology.

We need to do gauge fixing. I implicitly fix the two gauges. This kind of requirement implies strong
constraints on physical observables such as S-matrix. We work with
M =  M ;

invariant   + p

p M = 0 Ward identity (exact)


The Ward identity is exact, nothing about perturbation theory. It is a difference between four
different functions, M0 , M1 , M2 , M3 for arbitrary external momentum. In QFT, exact results are
very rare. Later when we talk about QED, we will prove this identity within that context, but here
you see that it follows directly from gauge invariants.

32

3.2 Lorentz invariance, unitarity


3 JAN. 29TH, 2013 SPIN-1 VECTORS
Lets see what else we can get. Consider a propagator of photons, from .


k
k k
i

:::::::::::::
= D (k) = 2
Ag + B 2
k + i
k
unitarity of S-matrix only 2 d.o.f. on initial & fixed state
Consider a special process [include image with blocks]


k k
i
Ag + B 2
J
M = J D J = J 2
k + i
k
Recall the Optical theorem

Now, we sum over the physical polarizations = and have two parts the left part is 
and the right part is 
. We havent done anything else yet, this is just unitarity. Remember
that the optical theorem relates an imaginary part of the amplitude to a physical amplitude
square with intermediate particle on shell. Therefore, because this is S-matrix element, you
sum only the physical polarizations.
Choose
1
k = E(1, 0, 0, 1),  = (0, 1, i, 0)
2
Define
k = E(1, 0, 0, 1)
33

FEBRUARY 5TH, 2013

so

 
= g +

k k + k k
k k

Using the optical theorem A = 1 (A is fixed)

+k k
k k

kk

is not Lorentz invariant, does not contribute b/c of gauge invariance, b/c k J = 0

(follows from Ward identity)


B is not fixed (not physical) will have to make a gauge choice
End result - propagator is
i
D (k) = 2
k + i


g

k k
+B 2
k

Summing over external state. S-matrix: M =  M .


X

|M|2 =

|M
|2 (
+(
|M3 |2 + |M1 |2 + |M2 |2
 
((0(
M M = g M M = (

(
((

but from before, we saw that since the second part doesnt contribute, we can replace
 
g

X
=

and that we get something physical


|M1 |2 + |M2 |2 = |+ M+ |2 + | M |2
and something that needs to cancel for this statement to make sense. From Ward Identity
k M = 0
and choose k = (E, 0, 0, E) so that
EM0 EM3 = 0

M0 = M3

February 5th, 2013

Okay. Everybody happy with QED so far? Today we will continue with QED. Last time we start
to calculate a scattering process. Let me say again, for the next couple lectures, we complete in
34

4 FEBRUARY 5TH, 2013


parallel with what we did in scalar field. We are going to go through the motions of calculate
some simple process and normalize QED. This is the basic stuff. The only new thing is now we are
dealing with Fermions and vectors. So the complication is more, so the computation is a little more
complicated. You have to deal with 4 by 4 matrices and so on. Otherwise, look at the big review
article by Steve Martin and so on, you would have done the two components. It does not actually
simplify a lot QED, but it does simplify a lot for later if you want to work on Susiea. (?) But for
now we have to deal with, lets go through this fairly standard Q.E.D. stuff. By the way, what Im
going to say is the small subset of chapter 5 in Peskin. There will be homework also doing similar
calculation, but you are strongly encouraged to go through the rest of chapter 5 in Peskin. It is all
fairly straightforward.
We started to calculate the following process last time
e + e +
know, 100 and some dont have 500. Do we know why this is heavier than electron? Is there any
reason why this has to be very heavier? There is no reason. We dont know why. This is called a
flavor problem. It has nothing to do with your food. But these are called flavors. Predecessors are
very good at giving names. These are called flavors. Later on you will learn colors. I think there
are people also, made models that particles, they are called yes. Yes. Those names never caught
on. But mainly because that model doesnt work, so its a technical model, doesnt quite work.
The flavor, the physics of flavor used to be, there is old name that doesnt quite use anymore, called
quantum flavor dynamics. In the physics, colors is called quantum chromo dynamics also known as
QCD. Im trying to remember the word people use for the smell. Okay. Hopefully you dont have
to deal with it.
The nice thing is we only have to deal with one diagram.


1X
1 X e4
|M|2 =
(p0 ) u(p)
u(p) (p0 ) u
(k) (k 0 )
(k 0 ) u(k)
4
4
4
q
 0
 

e4
p me ) (p + me ) Tr (
= 4 Tr (
k + m ) (k 0 m )
4q
where we replaced
u(p)
u(p) p + me
(k 0 )
(k 0 ) k 0 m
35

k 0 = k

FEBRUARY 5TH, 2013

So computing some traces


Tr = 4g

{ , } = 2g I44
+ = 2g

Tr( ) + Tr( ) = 8g

We can also do more complicated traces by running through cyclical permutations (definition of
trace)
Tr( ) = Tr(2g )
= Tr(2g 2g + 2g )
= 4(g g g g + g g )
There are some other things you can say. For example, you can show that the trace of gamma mu,
gamma mu, gamma ro is 0. Or lets begin with a simpler one. First of all, trace of gamma mu is 0.
Explicit verification, or there is nothing that is a number, okay? There is no constant that carries
a mu indices. There is nothing. As I said there is only, the only thing that is a number that carries
over indices is the G mu, mus, are the G mu, mus. Then you can show that the G mu G or gamma
mu gamma mu gamma ro is zero, mostly because we can permute this and that reduce down to
that.
Tr( ) = 0,

Tr( ) = 0,

Tr(odd # ) = 0

Useful identities for -matrices


Back to the big-ass trace before, well quote the result. So for example, the second trace





Tr (
k + m ) (k 0 m ) = 4 k k0 + k k0 g k 0 k + m2

 0



Tr (
p me ) (p + me ) = 4 p0 p + p0 p g p0 p

me 0

Im going to take the limit that the electron mass is zero. Just to simplify my equations. Muons
are much more heavier than electrons. In other words, in order for that process to go through, the
36

4.1 Renormalization of QED


4 FEBRUARY 5TH, 2013
minimum energy, the center of mass energy of those two electrons has to be greater than 2 times
the mass of the muon. This means that |p|, |p0 |  me . Doing the math gives us

1X
8e4 
|M|2 = 4 (pk)(p0 k 0 ) + (pk 0 )(p0 k) + m2 (pp0 )
4
q

4.1

Renormalization of QED
1 0 2
L = (F
) + 0 (i
m0 )0 e0 0 0 A0
4

1/2

1/2

A0 = A z3

0 = z2

3 = z3 1, 2 = z2 1, 1 = z1 1, m = z2 m0 m

1

A 3 (F )2 + (i
2
A
m) e
m ) e1
L = (F )2 + (i
4
4
4.1.1

Feynman Rules of Renormalization

And then there are some rules involved (including the Feynman gauge, photon propagator, full
propagator, fermion propagator, and so forth).

4.1.2

Renormalization Conditions (on-shell scheme)

X
(p = m) = 0


d X
(p)
=0

dp
p=m

m is a pole

Y
(q 2 = 0) = 0

canonical normalization for A

ie (p0 p = q = 0) = ie

37

canonical normalization

FEBRUARY 14TH, 2013

February 14th, 2013

Ward Identitiy QED (Ward-Takahasha Identity)




 + (p2 )p

 M

p M = 0

I remind you that the one identity, we mean that, so what we said is, its crucial to ensure that
this is the piece that ensures the series of Lorentzian variant, and, so sorry, its consistent to have
a gauge equivalence, and Lorentzian variants at the same time. So, you can take a different point
of view, as we said, at the beginning, if you just insist on, Im going to talk about a quantity thats
... if you want to start out with something, like this, and under Lorentz transformation, it goes
like this. It goes to like times some alpha, in general. Polarization vector has this kind of boost.
This can be a arbitrary function of P in principle, scatter function. And because S matrix element
is proportional to this, therefore, Lorentzian variance has to be consistent with this picture. I
have to require this is 0. Okay? This is the version of Lorentz, one identity. This is the gauge
transformation. That is the gauge transformation on the photon field. In this way of saying that,
the fact that the theory is the equivalent under this kind of, this set of gauge transformations,
is essential in guarantee the Lorentzian variants. It means that this must be zero. Okay? So,
this sounds like, this sounds very reasonable. This almost is a drirvation derivation, sometimes
you can think of this as a dare ivation if you think gauge symmetry, this is really fundamental.
And thats fine. That is probably the quickest way to see that this has to be true. If you buy
this, that the theory is gauge equivalents. But to be very strictly speaking, this is a statement of
just equivalence between the gauge transformation and the one identity. These two statements are
equivalent. But it is not really a proof of this identity itself. In particular, in the following sense,
we know that we havent shown that the theory, if you just compute these things, it actually satisfy
this property. Okay. You have just shown that these two statements are equivalent.
We have to fix the gauge, and we do this by introducing a term like this into the Lagrangian

( A)2
LG.f.
2

Let me repeat that proof. It is interesting proof. You should go through it. Yeah. You mean
whether this term is, has a counter term or not? What is the question? You add this term from
the beginning. One of the things we will show is that this term is not renormalized. That is a
38

5.1 Proof
5 FEBRUARY 14TH, 2013
consequence that we will show. Okay? That is part of the result. After we are done, similarly you
will see, okay. That is what we do. We are going to start with LaGrangian gauge fixing term, and
show that that is the case. Okay? Without using explicit gauge symmetry, without using gauge
symmetry.

5.1

Proof

S-matrix

hf |ii = i

5.1.1

dx eikx 2x (2xi + Mi2 )h|T A (x) |i

Example

For example, if I have 5 masses interacting somehow, then I can expect


hf |ii (p21 m21 )(p22 m22 ) h|T 1 2 |i
with the large vacuum state component at the end
h|T 1 2 |i

5.1.2

1
1
p21 m21 p22 m22

Schwinger-Dyson Equation
n

h|T

X
S
a1 (x1 ) an (xn )|i = i
h|T a1 (x1 ) aai (4) (x xi ) a an (xn )|i
a
i=1

which contains the action S with respect to any variation of any fields. (Note differentiation
follows the Feynman rules weve defined before along with the fourier transformation **** Ask
about functional differentiation)
S
= 0 classical equation of motion
a

39

5.2 General 3-point Function


From above, combining everything, we have 2A = J+ .
Z

d4 x d4 xi eikx eiki xi (x xi )f (x, xi ) =

FEBRUARY 14TH, 2013

d4 x ei(k+ki )x f (x, xi = x)

= f (k + ki )

to contribute to S-matrix, needs to have form


1
1
k 2 ki2 m2i

5.1.3

Renormalized QED


E.O.M. Z3 2A = z1 j where j = e
hf |ii =  i

z1
z3

d4 x eikx h|j (x) |i

=  M

M = 0 (Ward identity)

So
h|T j (x)(x1 ) (xn )|i = i

h|T (x1 ) (4) (x xi ) (xn )|i

with the conserved current symmetry = . This is a statement of global symmetry and does
not depend on whether your gauge is fixed or not. Therefore, we can make the replacement
k M =

d4 x eikx h|T j |i = 0

This is the Ward Identity.

5.2

General 3-point Function

insert a picture of a general three point function with j (x), p0 , p, (z),


(y) and with q

40

5.2

General 3-point Function

G (q, p0 , p) = iz1

FEBRUARY 14TH, 2013

d4 x d4 y d4 z exp(ip0 y iqx ipz)h|T j (x)(y)(z)|i




= (2)4 (q + p p0 ) S(p0 ) ie (p0 , p S(p)

S(p) =

i
P
p m (p)




q G = i(2)4 (q + p p0 ) S(p0 ) (ieq ) S(p)
Z

= z1 dx dy dz e x h|T j (x)(y)(z)|i
But we can also see that this one is taking, again, the expression upstairs, its just so. Again, I
dont write the exponential. TJ mu, okay? That is just that. Im taking the derivative with respect
to X. Derivative with respect to X. Okay. Now we say, great. Why is that great? Because this is 0.
Okay? Unfortunately, no, that is not 0. Okay? We will have to be careful with this. Okay?
So, because now, Im not talking about the S matrix anymore. Okay? Im talking about, Im talking
about actually a matrix element. Now you do have to worry about all those Delta functions. Im
not doing S matrix anymore. It is not S matrix. Its a greens function. So this is, this is again to
conserve the current of our U1 symmetry.
= z2 j
J = z2 e

conserved current

Lets explicitly write those Delta functions out. Okay? Okay, again Im using Schwinger-Dyson
equation. I dont know which board its on. I think I erased it. But that is just the SchwingerDyson equation for conserve the current. Okay? Im just writing it out. There are two Delta
functions corresponding to when X and Y gets closer, and, or X and Z gets closer.

U(1) symm. z2 h|j (x)(y)(z)|


= e 4 (x y)h|T (y)(z)|i

+ e 4 (x z)h|T (y)(z)|i
Okay? Again, so its these two point functions, obviously are connected to the propagators. These
are just the propagator, two propagator. So psiY sidebar Z, these are just the fourier transform of

41

5.2 General 3-point Function


the full propagator.

h|T (y)(z)|i
=

FEBRUARY 14TH, 2013

d4 k ik(zy)
e
S(k)
(2)4

This SK is just the, this propagator here. Okay? Now Im going to put all these things together.
So I have this, okay? I have this. But I did not use this, provided in terms of the sum of two,
two-point functions. Two, two point functions are just a propagator. Let me put everything in,
and then write it more carefully. The end result for that is that I have this, E5P, C1 over C2.
Okay? So C1 I inherited from here, and C2, I inherit from here. This is important, C1 and C2,
normalization constants. Q, ESP minus ESP prime. Okay? These are, again these are the two,
two-point functions, which is just a full propagator.


z1
(2)4 4 (q + p p0 ) eS(p) eS(p0 )
z2


= q S(p0 )(ie )S(p)

q G (q, p0 , p) =

where the last line was derived from drawing diagrams and looking at the diagram. The reason
I wanted to do that, in the end there is a relation between this three-point function and those
two-point functions. So you see this is already quite nontrivial. Right? This means that ward
identity, or more specifically, the conservation of a global U1 symmetry, the global, exist in global
U1 symmetry. There is a nontrivial relation between a general three point function, three point
vertex function, and the full two point functions. Again, this is exact. Okay? Let me, when we
went through all this, not because we cannot calculate one loop to verify this but because this
is exactly wrong. The only thing that we relied on are just this existence of symmetry, and the
Schwinger-Dyson equations which are all exactly right.
These are 4 by 4 matrices. Let me multiply from this side S inverse from P and from this side
S inverse P. This side S inverse P and from this side S inverse P. Okay? Just to make it looks
nicer.
iq (p0 = p + q, p) =


z1
eS 1 (p0 ) eS 1 (p)
z2

But before we even do that, lets just say this is, behold, 1 over U, exactly Delta function. Lets
see how to use it. As PQ goes to 0, so of course you say this is 0. Until you notice the other side
is also 0. Okay? If you say this means 0 equals 0 you havent graduated yet. This means that you
42

5.3 Yet Another (not complete) Proof for Ward Identity


5 FEBRUARY 14TH, 2013
need to expand both sides, to the second order, and see if there is, there is a nontrivial result, once
you obtain the result on both sides.
and then take limit as q 0.
(q = 0) =

z1 = z2

Yeah, you can divide both of them by Q. Either way you want to do it, there is a expansion you can
do, that is what Im going to say. Okay? We know that origin of the table expansion, this is 0, this
is gamma mu. This is our renormalization symmetry. That is just the condition we choose.
Under this condition, after you do your tailor expansion which is trivial, when those things are
getting very close by its trivial, you will see that the nontrivial result you get is the factor in front
is what? Here you are using this. You are using that. You are using everything. Okay? This is
our proof that these two normalization constant is the same, as we demonstrated at the one loop
order already, Delta 1 is the same as Delta 2 but this is the exact proof.

5.3

Yet Another (not complete) Proof for Ward Identity

In fact, Im not going to give you the full proof. But just show you how this thing will go, and in
the one intermediate step we will actually obtain another useful result. Okay? This is even more
like, even more direct in a sense you will see. We know that everything is specified by some Z
function. For QED partition function can be written down as like this. The passage, I havent told
you how to do past integral in Fermions and we are not going to use it here. Lets just pretend we
can do it.
Z
Z[J] =

 Z


( A)2
4

DADD exp i d x Lgauge-inv


+ J A +
2

That is not the conserve current. We need to introduce os strom from Fermions. Im not going
to use that so let me just dot dot dot. Okay? If you want to calculate correction functions with
Fermions [inaudible] only thing annoying is that the source term is familiar in your case. But we
havent talked about variables. We are not going to use it. Okay. Thats that. But that is as ab
initial yo as it can be, there is nothing more general than starting from that. Now lets do gauge
transformation. Lets transform. Just be done with it. Now let me write. Suppress all the indices.
43

5.3 Yet Another (not complete) Proof for Ward Identity


Imagine these are some differential forms.
A = B

= ei 0 ,

DA = DB

FEBRUARY 14TH, 2013

D = D 0

We are changing variables from A B and so on, these measures are not invariant. Were doing
a phase shift.
Z
Z[J] =

 Z

1
0
0
2
0

DBD D exp i Lgauge-inv (B, ) (B 2) + J(B )


2

From that I derive this DB, psi bar ... again, gauge invariant, that doesnt change, because its
gauge invariant. I can just call the field new names. That is what I mean by gauge invariant. The
second term changes under this transformation. The second term changes, to alpha .. there is a
derivative acting on that. Becomes a box. That, and J, the third term changes too, by the way.
So J.E minus D alpha. Okay? So everything changes. Lets do some manipulation. First I wanted
to, just to be confusing, I will relabel B as A. I just go back. I want to compare with the original
LaGrangian basically. It is just a name.
relabel B A, 0
Z
Z[J] =




 Z

 Z

1
1
(A)2



+ J 2 A 
(2)
exp i J
DADD exp i Lgauage-inv (A, )
2

2


So all of this, Im just coming from, Im squaring this and recombining the form. Box alpha square,
so the whole thing, the whole thing times another exponential, I alpha, I alpha D mu J. Okay?
And Im not asking you to follow every single step. But you see the gist of this. Im just taking the
function and do a gauge, go from there. Not make any assumptions whatsoever. This is a constant.
Its not a constant but it is a constant as far as past integral is concerned. We know any constant
in past integral doesnt matter, I just get rid of. In principle I can write a overall constant in front.
But its cumbersome. Why dont I just assume thats one. Okay? So lets see.
So, lets see what this means to us. This means if you compare this with the original LaGrangian,
this actually means CJ is C, the same Z but only a shift is J. This looks like a shift on the source
term. Okay? Nothing, nothing fancy. Im just shifting the source term.
 R

1

Z[J] = Z J 2 ei J

44

5.3 Yet Another (not complete) Proof for Ward Identity


5 FEBRUARY 14TH, 2013
Okay? We know that this is the generating function of all correlation functions, W is the generating
function. That is why overall constant in that doesnt matter, okay? This is the generating function,
or constant is this plus a constant but doesnt depend on J, so it doesnt matter.
Z[J] = eiW [J]
Taking this derivative with respect to J, you get all the graing functions but this means that WJ
is just WJ1 over ... that is why Im always keeping this J around

 Z
1
W [J] = W J 2 J

45

FEBRUARY 19TH, 2013

February 19th, 2013

Anybody tried homework already? Easy? Trivial, right? It is only QED calculations. Later has
loops. But that is only QED loops. The first time is always hard, you have to do it at least
once. The second time, you begin to do actual, it will be harder. Then you discover people dont
really do it. The software help you do it. You have to write software, because all the procedures
is pretty standard. The real difficult part is not to write down a loop, is to actually evaluate it.
The real process has hundreds of diagrams. You talk about QCD process, for example. You can
actually make a living doing that. Today we will wrap up ward identity and move on to the next
subject.

6.1

Ward Identities

Let me briefly recap what we did last time. I remind you, the last time, the goal is to start out with
a gauge fixed Lagrangian, and demonstrate ward identity. Last time, close to the end we embarked
on another different approach, trying to demonstrate explicitly ward identity is true.
We started out with the partition function the grandaddy of everything if we know it, we know
the whole theory. All the correlation functions can be obtained by differentiation with respect to
this current, for each field you have a current. Lets just write it formally, gauge invariant (GI =
gauge invarient, GF = gauge fixing)
 Z

Z
4

Z[J] = DADD exp i d x (LGI (A) + LGF (A) + J A + )


The gauge fixing here is (as a function of A)
LGF =

(A)2
2

Now we would like to start out with this lagrangian and do gauge transformation. Gauge transformation is doing this
A = B (x), = ei 0
DD = D 0 D0

DA = DB
LGI (A) = LGI (B)

relabel B A

46

LGI (A)

6.1 Ward Identities


6 FEBRUARY 19TH, 2013
This is just a shift, and as far as functional integral is concerned, a constant shift. The gauge
invariant part is obviously invariant. Lets rewrite it in terms of the new variables
 Z



Z
R

1 ( A)2
1
1


Z[J] = DADD exp i LGI (A)
+ J 2(x) A 
(2) ei J
2 2

2

Lets write the corrections. So always in this square bracket, J mu ... alpha X, A mu, okay, that
is one term. The other one is box alpha square. Okay? This one is a constant. Its a constant so
it goes away. This means that this one is not, it does not have anything that will integrate over,
or integrate over A. It does not depend on J. It is not something that we care about in the end.
It is just a constant. So also known as doesnt exist, okay? So we can see, the overall constant,
past integral doesnt matter. The whole thing, so this also times yet another exponential I can
A independent exponential looks like that. This one is not quite a constant because it depends on
J. It is a constant as far as the functional integral is concerned. But we want to keep it because it
depends on J. Thats it. That is the result. I havent made any assumptions. I just did a dumb
gauge transformation.
The new J looks like
1
J J0 = J 2(x)

So now you can also write this, you recognize this is just the statement that Z[J], the first part is
just old Lagrangian. The first part is the old Lagrangian. This one is, plus this term is just old
Z[J].
Z[J] = Z[J 0 ]ei

And recall from before that


Z[J] = eiW [J]
which can be called the Ward identity - to write Z[J] is a more explicit function. The usefulness
of W [J] is the same as Z[J] where if we differentiate with respect to the J (currents), we obtain
correlation functions. But we can see that
0

W [J] = W [J ]

47

6.1 Ward Identities


So lets do this for very small

6
Z

W [J] = W [J]

W [J]
1
+ +
d x 2

J (x)
4

FEBRUARY 19TH, 2013

d4 x J (x)

So since alpha is arbitrary, this guy is arbitrary. Okay? The same thing you do with classical
mechanics variation of principles. You make a variation and the fact of the derivation the rest of
it is 0. So this kills that. You get our first identity, which is box minus plus mu J mu is 0. Okay?
Is this okay? Is everything okay? Lets put like X back. This is all X. Everything is with respect
to X. Okay? You can call this ward identity too. You can call many things ward identity. Lets
take another variation. Lets take another functional derivative out of this equation. Okay? So
basically because this is identical is 0, lets take another functional derivative of it. Okay? Of JY.
Just to be, so lets be careful, lets put labels on these things. Now, this gives you 1 over ... let
me be careful, box X, D square, Delta square WJ. Mu Y is the same. Im moving this term to the
other side. Okay? Partial X mu, Delta X minus 1. This is still partial X. Mu becomes mu because
Im taking derivative with Delta J mu. Im just using Delta J mu X, Delta J mu Y, the Delta mu,
mu, Delta X minus Y basically. That is all I have done. The reason I wanted to do that is this
looks very much like a 2 point function already. This part is just a 2 point function. Remember
this is how we derived our propagators. Take the two functional derivative of generating functional
W.



1
W [J]

2
J (x) = 0
J (y)
J (x)
1
2 W [J]
x 2x
= x (x y)

J (x)J (y)
This gives you a 2 point correlation function. Which is related to the full propagator. So in
particular, full propagator, two point function, is defined by just this. Again we did this with
scalar field last quarter without the mu and mu. But that is not, there is nothing new really with
mu. At J equal 0, this is our 2 point function. Okay? This is our 2 point function. Now, what does
this mean? I can take this definition, put it in here, and what does this mean? This means that 1
over XKC partial mu X mu box X, by the way, this all just means that Im taking derivative with
respect to X. Mu, mu, X minus Y, D mu X Delta X minus Y. Okay? Or 1 over KCP mu P square
Delta mu, mu P is P mu.

2 W [J]
= (x y)
J (x)J (y) J=0
48

6.1

Ward Identities

FEBRUARY 19TH, 2013

1
2x (x y) = x (x y)
x
or
1 2
p p (p) = p ***

So lets see, I can get rid of this already. That equation means, so okay, lets start out from another
point of view. Okay? What is the general form of mu new? Just general form. General form,
of mu and new, of Delta new mu. We know it has something to do with the momentum. The
momentum will enter, we know between mu and mu, the things that can give you mu and mu
are the momentum and G mu mu. Using that, you can just argue that the most general form of
whatever it is, is the following. Its just this. Okay? Thats it. Okay?



p p
p p
1
= 2
g 2
F (p2 ) + 2 G(p2 )
derive this from an effective lagrangian
p
p
p
Let me call it, give it a symbol. Applying this to that gives you, lets see P mu, P mu, cancels that,
so it gives you that P mu over KC. Sorry. Let me just .... (pause). Yes. Times GP square gives
P mu. Okay? Im just using that identity on this one. This means P square, although in general
is a function, you think it can be a function of P square based on Lorentz invariants, but its not.
It is just a constant. Okay? Exact ward identity implication this term is just a constant which is
KC, period. Nothing okay? All right. This means that for exactly, exact 2 point function, you
can always write this form. This is always KC. It is just the gauge fixing parameter. No matter
what you do. Okay? I mean independent of renormalization, and independent of anything. That
is why also we decompose it this way. This is exactly just the gauge fixing term basically, showing
up itself in propagators. Okay? We can also look at it a little bit carefully.
***

p
G(p2 ) = p

G(p2 ) =

So this, you see that I can derive this from this effective LaGrangian. I can derive this from
an effective action, actually. But effective LaGrangian. How do I derive this propagator from
LaGrangian? It is coming from the LaGrangian of this form. P square plus P mu, P mu, A mu, P,
F minus 1P square and A mu. Okay? Starting out with this LaGrangian, you ask yourself what is
the propagator of A? You get that. Okay? Here, so now this is momentum space. You may not feel
comfortable with momentum space because we write it in LaGrangian position space. So but this,
the first term, the first term in position space is just, so the first term is just the following. F mu,
49

6.2 Spontaneous Symmetry Breaking (SSB)


6 FEBRUARY 19TH, 2013
mu, F nu mu, now it looks a little bit more familiar, right? Times some function, some function.
You feel very bad about this? (chuckles).
1
F 1 (p2 ) 1
Lfull = A (p)(g p2 + p p )A (p)
+ A (p)p p A (p)G1 (p2 )
2
3
2
In position space, it would look something like
Z
1
d4 x F f (1 + f (2)) full quantum corrections
4
That is why I have this here. Its all quantum corrections in it already. Im not doing any although
this looks familiar with your initial LaGrangian to begin with, but it is not. It has all the quantum
corrections already. But even at this LaGrangian level, the gauge fixing term is still this. Okay?
The conclusion is that, that is not renormalized. That is what this ward identity tells us. Even if
you write it exact operator, you invert it, you get your full LaGrangian, the gauge fixing term is
still this. The F nu mu square term on the other hand gets renormalized by this function. Its a
highly nonintuitive statement. I want to emphasize that. It is not intuitive, because you are used
to thinking about the fact that once you open the Pandora box, everything goes. So you start to
calculate corrections, everything will correct, is corrected. It include the counter term of everything.
But we dont ever include counter term for the gauge fixing term. That is why just because its
not renormalized, dictated by ward identity. In other words, another way of saying it is that the
gauge fixing term, something is renormalized, means that something is not physical. There is no,
the actual value of KC has no physical consequence. Otherwise it has to be renormalized. Okay?
That is another way of saying it. From hindsight, it is not that surprising. None of the physical
quantities you calculate depends on KC. Even if you want to set up a renormalizing condition,
based on physical quantities, to determine KC, there is no such a condition you can write down.
Okay? Because none of the physical quantity depends on KC.

6.2

Spontaneous Symmetry Breaking (SSB)

It is not a very good word, but that is the technical term. We will see why its not a very good
word. SSB. Basically, the idea of spontaneous symmetry breaking as opposed to explicit symmetry
breaking, so you have spontaneous symmetry and explicit symmetry breaking. Explicit means
there is no symmetry. But why do you call it a explicit symmetry breaking if there is no symmetry
50

6.2 Spontaneous Symmetry Breaking (SSB)


6 FEBRUARY 19TH, 2013
to begin with? It is sometimes useful to think symmetry breaking that is small, okay, you have
approximate symmetry with a few small breaking effect. But that can be explicit breaking. So that
is explicit symmetry breaking. Spontaneous symmetry breaking is completely a totally different
beast. Historically there have been a lot of confusion about this again. But it is basically a
statement of symmetry of the dynamics, versus symmetry of the grand state. Okay? You will see
that even this is not a very good description. But lets go on. This is the easiest way of getting
to this concept. Okay? Even this is not completely right. The most famous, everybody use that
example is a physicist inside a magnet, it is a small physicist inside a big magnet. Okay? Whatever,
lets consider a magnet. Lets consider a magnet. A magnet, a perfect place to use quantum field
theory. Not even Lorentz invariance. But it has many, a good model of magnet consisted of many
spins. Many, many spins, it can be dipoles, going up, going down, going up, going down. A lot of
dipoles inside. This is called a feral magnet, by the way. The Hamiltonian, to some leading order,
to some approximation, lets do a simple model, okay? Lets only consider the nearest neighbor
interaction, which is approximation. Is it a good approximation. Hamiltonian can be written down
as something like this. There is interacting spins basically. But it doesnt even matter actually for
this argument, whether its nearest neighbor or not.
H =

~si ~sj

full 3D rotation symmetry, SO(3)

i6=j

Lets consider the grand state of this theory. Lets consider the grand state, lowest energy level.
Ground state, the ground state means that obviously, you want to minimize H. Okay? There can
be many very complicated configurations. But the one thats actually really minimize the local
ground state is everything is aligned. Everything is parallel. That is where the product is biggest
because there is a negative sign and this is a ground state. This is magnet. Here is a physicist
inside a magnet. A small physicist in the big magnet. What he will see is that all the spins are
aligned. This is the ground state. What he would conclude from that observation is that there is
no rotational symmetry. The quantum, the state does not have a rotational symmetry. Right? You
just pick out a direction. It could also pick up this direction. It could also pick up that direction.
That also has some deep meaning to it, by the way. But nevertheless, right now lets all agree
that it will pick out a direction. It will point towards some direction. That is the ground state.
Okay? That ground state, consider there is only a 2D rotational symmetry around that axis, SO2.
This is the symmetry that preserve the ground state. That is example that ground state does not
51

6.2 Spontaneous Symmetry Breaking (SSB)


6 FEBRUARY 19TH, 2013
have the same symmetry as the dynamics of the system. The dynamics of the system is this. I
think there is a theorem saying that the converse is actually true. If ground state is preserved some
symmetry, the dynamics must have a bigger symmetry, contains that symmetry. But not this way,
okay? Ground state can have much less symmetry than the dynamics of the system. Everybody
have heard analogy of you are sitting around a table, oh, okay. So I dont know. So its, imagine
you are sitting around a table with your, I guess its I dont know, with your silverware. Now
imagine setup for that, is it unclear whether its on your left-hand side or on your right-hand side.
It is completely symmetrical, your plate in the middle, your silverware on the both sides. I dont
know western food very well. So I may not do this right. But lets just imagine there is such a
configuration. Now dynamically its completely ambiguous. You should do this. Ill do that. Unless
you prefer your right hand. Lets imagine that we have a, what is that word? There is a word you
can use. Yeah. Lets imagine that everybody is that. Like that, okay. Now, dynamically there is
no preference. There is a complete refraction symmetry lets say going that way and going that
way. It only up to one guy to make a mental decision, due to a small perturbation in his brain that
he wants to use this hand. The analogy is quite exact. Now you break the symmetry. Everybody,
okay, so now the configuration in the end, you end up with fork in your hand and start eating, is
that state breaks the dynamics of the system you begin with.
Consider real scalar field
1

V () = 2 2 + 4
2
4

1
L = ()2 V ()
2

Z2 symmetry

ground state vacuum


and other random things that Ill have to look in notes for at this point

52

FEBRUARY 28TH, 2013 SPONTANEOUS SYMMETRY BREAKING (CONTD)

February 28th, 2013 Spontaneous Symmetry Breaking (contd)

Talked about the Nambu-Goldstone theorem, with the Goldstone m = 0 which can be seen by
symmetry or under the shift + v.
Then we started to do a more slightly general analysis U(1) to see this breaking
a a + a ()

linear function of fields

Symm: V (a ) = V (a + a ()) a () V()


=0
a

V
=0
Vacc:
a
a
a
=0

a ()
b





2 V ()
V ()
a
+ () a b
=0
a ab =ab
ab =ab
0
0
{z
} |
{z
}

=0

=0

So from this I can already derive that an equation, so I expand this out and take the derivatives
and derive the equation. So this actually so lets just write it down. This means delta A phi D
phi, phi A is zero. Identical is zero, I just explain. Because alpha is arbitrary, and and I take
another derivative out of that. Okay? And set everything to its vacuum value. So the this will
be so Im just repeating the last part of our lecture last time. So lets take the derivative over
35 B, some other field strands, and D phi D phi A, but I wanted to set this into this (inaudible)
vacuum value, okay? And theres another term which you keep (maybe it was three phi B back
there). But taking the second derivative with phi AB equals phi AB zero, okay? And thats zero.
Okay? So there are there are in general two solutions, and well, first of all lets observe, this is
zero, because this is zero, because my condition that this is a vacuum, so this is this is my ground
state. Now, then, therefore this has to be zero, and there are two cases where this can be zero. A
is is delta A phi well, error at the vacuum zero, okay? And this means the symmetry this
means no SSB in phi A. Theres no spontaneous symmetry in that field, even you give it a value
it doesnt break anything, doesnt break symmetry. Or theres no vac, either way. Okay? B so
thats a possibility. So that but thats just original paste, whatever my ground state is, doesnt
break symmetry, okay?
a ( = 0 ) = 0 No SSB in a

53

a (

FEBRUARY 28TH, 2013 SPONTANEOUS SYMMETRY BREAKING (CONTD)


2V

= 0 a () Goldstone Mode
= 0 ) (a )(
b)
ab
ab
=0

Now, for the case that ground state actually breaks some of the symmetry for those field values that
actually breaks the symmetry, you are end up with and let me write down this equation again.
Whatever and D square D phi A, D phi B. Okay? So then you ask, whats the importance
of this one? You realize that for the potential, in general you can write them down as V phi
zero, where everything is plus of you just Taylor expand, doing a Taylor expansion around its
ground state. So this is a constant, and the first (inaudible) terms all vanish, because again, my
condition. Then the secondary (inaudible) terms are like this. Okay? Theres small fluctuations.
Its exactly that. Okay? They are higher order terms, but this term is exactly the mass matrix,
okay? This now its a constant because I set all the fields into their vacuum value, so that this
is the mass matrix. Okay? Mass of the scaler field. Its a matrix in general, just again, I said
like last time, where were used to the fact that if we have a coupled system, harmonic oscillators,
okay, so the frequency is in the generic base is a matrix, okay? And they (inaudible)ize it, you
find eigen values, eigen (inaudible). So this is just like an eigenvalue equation and this tells me
that for everything for every value thats actually for every field of variation thats been broken
by giving fields (inaudible), there is this combination, which is which is a serial eigenvalue state,
which is a ghost (inaudible) mode. So this is the ghostal mode (I dont know what hes saying
there. (maybe not serial).
1
V () = V (0 ) + a b
2



2V

+
a b ab =ab
0

so in general in a spontaneous symmetry breaking (inaudible), you consider a co-sat, co-sat, so


whats the symbol? So the original the original global symmetry, you call it a G, global symmetry.
You call it a G. And after a symmetry breaking, there is still an unbroken subgroup thats possible,
which we call H, okay? And then the goldstone mold lives in this co-sat. Okay? Which is a fancier
way for saying that for every broken generator can associate a goldstone mode. Okay? Theres a
slightly theres a much fancier formalism to talk about this called CCWZ (inaudible). Okay. But
well see whether we have a chance to get to that or not. Probably maybe not. Okay?
For an original global symmetry G, we have the Goldstone unbroken H so the Goldstone number is
~1 S
~2 which has SO(3)/SO(2)
G/H. For example, if you have the Hamiltonian of a magnet H = S
2 Goldstones.
54

7.1

7.1

SO(2)7

FEBRUARY 28TH, 2013 SPONTANEOUS SYMMETRY BREAKING (CONTD)

SO(2)

So lets lets just try to see. So this may be a little bit more abstract. Lets try to see this in this
U1 example. Lets try to see how this works. Lets try to write things this way. Okay. So lets try
to do this for the U1 example. Lets erase this co-sat stuff. So let lets lets not talk about the
U1 but instead talking about SO 2, okay? Its just to for for a change. Okay? So we argue that
this is equivalent talk about the U1 rotating by a complex face or just SO 2 (inaudible) in those
two. So phi complex field or will be like this. I can equal it just just think about the two real
fields, okay? In SO2 rotation, two real fields are rotated by this, okay.
Lets start with = 1 + i2 .

0 =

cos sin
sin

cos

1
2

For small
0 = +

{z

()

1
2

Now we know theres a Goldstone and that that is actually telling me what the Goldstone mode
is. Okay? If I set so that delta phi zero is where the Goldstone mode is, if I set the phi to its
vacuum value. Okay? So lets see what this is.
vac: 10 = v, 20 = 0. So

(0 ) =

0
0

So first of all lets recall that the last time I said that the existence of phase transition does depend
on dimensions. Okay, does depend on dimensions. So last time I already said it. In the case of
discrete symmetry breaking, okay, there is also a degenerative vacuum but theres no Goldstone
mode, but even that case you can say that the D equals to 1, there is no excess speed, okay? In
one-dimensional, okay? Because one dimension thats just quantum mechanics. And we argue in
quantum mechanics the ground state is not a symmetry breaking state. In quantum mechanics
theres something like this, minus the ground state is plus this is the ground state period, in
55

7.1 SO(2)7 FEBRUARY 28TH, 2013 SPONTANEOUS SYMMETRY BREAKING (CONTD)


quantum mechanics. But the one (inaudible) is quantum mechanics, as we said last, okay? So
discrete (inaudible) symmetry breaking only happens when the (inaudible) greater than 1. Okay?
So so the icing model has a discrete symmetry breaking, for example.

Figure 1: Discrete symmetry D = 1 no SSB

Now, lets talk about continuous. So we are worried about the fluctuation like that, okay? From
point to point is you can have a large fluctuation. So lets again just take some simple example.
Lets I parameterize my field this way. I can parameterize my field this way. My worry is that
the value of my vacuum can fluctuate from place to place. My pi can be at different values, okay?
If thats the case, then theres in the (inaudible) phase transition theres no long-range order. I
can not talk about a single phase. Where I am pi is phi or where you are pi is 1, so I dont know
we are not in the same phase. What happens is really there is no meaningful way of talking about
phase transition. There is no other parameter in that case. Okay?
= vei/v
Consider a space volume with size R ( v on length scale R; v/R)
Z
S

dD x ()2 RD

v2
v 2 RD2
R2

So the infinite wisdom of calculating action is that the thats actually whats characterized the
contribution of a certain field configuration to the pass integral, its proportionate to the action.
Okay? So in particular is this. Okay? Thats your weight, okay? And by the way, this means
that, you know, when H (inaudible) R goes to zero, exactly you want to minimize the action, okay?
You want to minimize the action. Thats exactly where the classical limit is. I assume youre all
familiar with that story. But I can plug this in ( I guess limit). This means HR V square, R, D
56

7.1 SO(2)7 FEBRUARY 28TH, 2013 SPONTANEOUS SYMMETRY BREAKING (CONTD)


minus 2. Okay? So what so what you want to what you worry about is that any fluctuation
can destroy long-range order, okay? So long-range spontaneous imagery is supposed to happen
in all the in a big space time volume. Otherwise, you dont even have a right to talk about it.
Long-range order. So Im borrowing a lot of words from phase transition and (inaudible) physics
because they are already the same. So long-range means, you know, R is much, much bigger than
1 over V. Okay. Much, much bigger than 1 over V. So you see that this is suppressed. Okay? So
fluctuation let me write it carefully fluctuation suppressed for D greater than 2. Okay? So the
answer is that if you live in a space time dimension greater than 2, you dont worry about those
things. Okay. It makes a lot of sense to say pi is just fixated, just fixed at the where it is. You
dont the fluctuation in pi is highly suppressed at the large at the large distance scales, and
therefore it doesnt matter. Okay? All right.
S

e ~ e

v 2 D2
R
~

long range order R 

1
v

Fluctuation suppressed for D > 2. For D 2 no SSB of continuous Goldstone symmetry.


Of course there is the converse of this story. What happens if D equals to less than or equal to,
okay? No spontaneous imagery (inaudible), okay (I guess imagery). I can keep writing continuous
global symmetry. Because the fluctuation is big. Its not suppressed. In fact, its enhanced, going
to allow larger distances. Okay? So because of this quantum fluctuation theres no and there
is a slightly fancier way of saying the same thing. Okay? A slightly fancier way of saying the
same thing is that a measure of the quantum fluctuation in the field is the expectation value of a
2 point function, okay? If you if you measure the expectation value of a one point function
is expectation value. The variance of the expression value is square, basically. Thats how much
things going around, and if you do this (expression not expectation I guess) in momentum space,
its like this. Okay? You see when D less than or equal to 2, there is infrared divergence. Okay?
Means that when in dimension less than 2, the long wavelength mode is very, very important.
Thats another fancy way of saying theres long-range fluctuations, thats being very important,
okay? But its basically just this, okay? If youre a little bit nervous about this, seeing the two-point
functions. Okay. But they are basically talking about the same thing.
Z
hi

dD k
k2

D 2 1R divergence

57

7.1 SO(2)7 FEBRUARY 28TH, 2013 SPONTANEOUS SYMMETRY BREAKING (CONTD)


So here I the expansion parameter is really H. Hbar here is very important because that actually
characterizes the size of the (inaudible) fluctuation, okay, you know, even in D even in DD equals
1, if hbar zero, this is always suppressed, okay? If hbar goes to zero attention youre shutting off
the quantum fluctuation, okay? So (quantum fluctuation). There is a completely analogous there
is a ah.
eH/T

D 2 thermal fluctuation No SSB

Colemans theorem. Mermin-Wagner-Hohenberg.


Okay. Now lets lets promote so again, I remind you that we have said repeatedly that there is
a degeneracy of vacuum and there is Goldstone mode. Say those words in the quantum mechanical
way, okay? Lets say those words.
Conserved current

J = 0

Z
Q=

d3 xJ 0 (x)

dQ
=0
dt

[Q, H] = 0

Consider the ground state |i and undergo a transformation | 0 i = eiQ |i. Since [Q, H] = 0 then
|i, | 0 i are degenerate in energy.
Okay, let me see. Remember, we always start with a conserved global symmetry. A conserved
global symmetry means there is a conserved current, meaning that equals zero modulo (inaudible)
terms. So for conserve current, I can define something, well call it a conserved charge. Okay. And,
you know, the fact that the current is conserved means that the DQ/DP is zero. Okay? Okay. Now,
another way of saying this is a conserved current is that this is zero, okay? Thats just another way
of saying this is a conserved charge, okay. Now I consider consider ground state sign, lets just
call it the sign. And now lets consider transformation of ground state. Okay? Which is E to the
IQ alpha (inaudible), okay? This is the generator of the symmetry, which is the charge, and thats
a small, and lets call this a side prime, but because QH is zero, so because QH is zero, zero, you
can show that the xi and the xi prime degenerate. In energy (zi or xi) so this is a statement you
have degeneracy of vacuum. And this is not a very good argument. Its very easy to go through,
but its not a very how should I say, its not thats the way you see it on textbook all the time,
I think, and its not very good in the sense that I have just argued, you know, this doesnt even

58

7.1 SO(2)7 FEBRUARY 28TH, 2013 SPONTANEOUS SYMMETRY BREAKING (CONTD)


depend on this is a ground state. Doesnt even depend on spontaneous (inaudible), looks like. So
its not a very good argument.
Lets talk about something manifestly symmetric assume Q exists.
[Q, (x)] = i(x)
Q|0i = 0
So now, lets calculate this quantity
ih0|[Q, (x)]|0i = h0|(x)|0i = 0
which we have in this vacuum. Therefore, we know that:

h0|(x)|0i =
6 0 SSB where we

transform + . So if there is a global symmetry, I can still construct a conservative


current. So lets just write down the Ward identity
x h0|T J (x)(y)|0i = i (4) (x y)h0||0i
Consider
G (x y) = h0|T J (x)(y)|0i
Z
d4 p ip(xy) e
G (p)
=
e
(2)4
Now combine these two statements and get
Z
G (x y) =

d4 p
(ip )G (p)eip(xy)
(2)4
Z
d4 p
= h0||0i
(i)eiP (xy)
(2)4

and note that h0|(x)|0i = h0|eiP x (0)eiP x |0i = h0|(0)|0i.


Okay. Now its time, lets give you this already tells you something about this. Right? So for
example, this tells me this tells me that the P mu well, P mu, P (inaudible) mu, P, is just this
matrix element. Because the rest of them are the same. Okay.
And this leads to
e (p) = h0||0i
p G

e (p) = p g(p2 )
G
59

7.1 SO(2)7 FEBRUARY 28TH, 2013 SPONTANEOUS SYMMETRY BREAKING (CONTD)


Now I can also say something about this. Okay, in general, because this carries mill indices, and
theres only one thing that carries mill indices here which is P itself. You have to be able to write
this as a P mu times some scaler functional P (I guess it was P mu? Mill indices?) So now I
combine this. How do I combine this? I can solve for GP square, so this means that the P square,
GP square is zero delta phi zero and GP square also known well okay, I wont repeat the trivial
step, but you solve GP square plug (inaudible) back. This is the P mu. P square, okay? This is
that. Of course this is nothing if theres no spontaneous symmetry breaking, is zero, but there is
something if spontaneous symmetry is (inaudible) down here, that we say spontaneous symmetry
means that means that this, so spontaneous symmetry breaking means this is nonzero, okay? So
this is actually a legitimate greens function, not the trivial one. If theres no axis B this is a trivial
one, this is zero.
p2 g(p2 ) = h0||0i

e (p) = p h0||0i
G
p2 | {z }

pole at p2 = 0

SSB:6=0

Moreover, I claim this proves the Goldstone theorem. Why? Anybody knows why? So whats the
meaning of I have a zero mass particle? It means that some greens function there is a physical
pull at the mass equals zero (maybe hes saying Lagrangian??)
So here is zero pull. Okay, this is a formal way of proof there is a massless particle, that you can
construct any green function you want. You identify the post, okay, or the physical post should
show up in the green function, in some greens functions. They dont have to show up in all greens
function, any all of they hope, bub at least if there is a mass if theres a physical particle with
mass, you should be able to come up with a greens function, where the Poe equal to the mass, okay?
And we came up with a green function and we find our Poe. Thats basically it, and it (inaudible)
depends on the fact that this is known zero (or nonzero). Is it okay? This is very formal. It doesnt
depend on the fact that the have to go around the Mexican hat potential and so on and so forth.
Okay? I dont know whether you found this is very comforting, but I think that the shift symmetry
argument is pretty good already in the beginning, but this is just in case you wonder whether its
a theorem or not, this is a theorem.

60

7.1

SO(2)7

FEBRUARY 28TH, 2013 SPONTANEOUS SYMMETRY BREAKING (CONTD)

h0|J (x)|pi = if p eipx


So, in fact, we can say a few things one more thing. Lets just say one more thing about base
based on this set of equation we have our data (inaudible) now. Okay? Lets say something about
the matrix element of J mu. Between that and the momentum P state. Okay? Lets say something
we can say one thing about this matrix element, turns up. Okay? That again, carries mu, right,
this momentum matrix (inaudible) carries mu, and the only thing that carries mu is here, okay?
And so this is (inaudible) P mu, and you can show that this under space/time boost, under space
time translation it translates a state with momentum P. So therefore it has to have something like
this. Okay? The argument I just said is a little bit quick, but its correct. So and it is the most
general form you can write down. Of course we need a proportional constant, which we call F,
okay? Its just some constant. And and just to be devious, usually thats the convention, okay?
But once you get used to this, this is the thing that everybody use, in (inaudible) theory and so
on, thats the convention everybody uses, and you hate the other conventions. You hate the people
that dont use your convention. Yeah. Okay?
But anyway, dealing with being able to deal with, you know, different conventions is a test of
whether you have (inaudible) to become a theoretical physicist or not. You have to deal with
different conventions. Okay? So theres nothing to complain.
So recall our what the definition of G mu, and which is just a time order correlation function
between this and phi. So I can so I can write it out and using using translational variance and
insert a set of complete momentum states, okay, I can write the whole thing out. So here is the
result. Its time ordered so it depends on which whether X happen first or Y happen first. So there
is zeta function X zero minus Y zero. (zeta I guess) okay? So just Im (inaudible) Im putting
P in between. Okay? So basically Im inserting inserting the following comment. Im inserting
this quantity into this formula (quantity, not comment). So and using my parameterizations, so
Im inserting things here. Okay? Using that formula. So thats P mu Im just going to write
one of them. P phi zero again use use something very similar to what I did there with the
with the space translation. Start and the plus another one was X zero Y zero interchange,
basically. Okay? And that one actually has Im just trying to remember that da, da, da, da,
61

7.1 SO(2)7 FEBRUARY 28TH, 2013 SPONTANEOUS SYMMETRY BREAKING (CONTD)


WIP (inaudible). Okay. Okay?
G (x y) = h0|T J (x)(y)|0i
Z
d3 p
0
0
(if )p hp|(0)|0ieip(xy) + (x0 y 0 ) (+if ) eip(xy)
= (x y )
(2)3 2Ep

Z

i
d4 p
= if h0|(0)|pi i
eip(xy)
4
2
x
(2) p + i
Z
d4 p p
=
f h0|(0)|pieip(xy)
(2)4 p2
So it takes a little bit of insight, but it doesnt take very long insight to show its similar
similar to things that we go through and we derive the Cline Gordon propagator basically in the
first quarter to show that this is the following. Of course, I mean, these things all factor out, so
Id write it out first, and the next step is a little bit you have to convince yourself a little bit, but
its its the same that we did with with our D 4 P (writing on board). Its nothing but just
that. Okay? Again, this is a very this is identical steps. We went through identical steps when
we derived (inaudible) and operator. You can go back to the easier because there are actually two
Poes you complete. Depends on which one is bigger, you each completed counter above or counter
below, exactly get these two terms. Okay?
And if you look at the work weve shown before, you can get that
f

h0|(0)|pi
| {z }

= h0||0i

=1 (normalization)

f = h0||0i
So, in fact, this is usually normalized to 1. You can check with our normalization. This matrix
element is 1. So in this case F is just a okay? So this is the same. Let me just let me just
say that in word so you know Im not going through this scene just to have fun. So there is
a matrix element, okay, of the symmetry current, pitching vacuum and the warm particle state,
okay? Thats what this is (or one particle). Roughly speaking this is the probability amplitude
of getting a one particle state on a vacuum from the symmetry current itself. Okay? Statement
No. 1. And that matrix element modulo this general piece, that matrix element has a strength,
okay, which is (inaudible), okay? And the F this is completely general but this F is exactly the
same as the strength of spontaneous symmetry breaking, okay? Remember, this again, this is
62

7.1 SO(2)7 FEBRUARY 28TH, 2013 SPONTANEOUS SYMMETRY BREAKING (CONTD)


the signal of strength of spontaneous symmetry breaking. So this setting in a more mundane word,
this is the (inaudible), okay? And the way that the that symmetry symmetry current acting
on vacuum is the probability of pulling out a particle by doing a symmetry transformation on the
vacuum is proportional to the strength of symmetry breaking. Okay? The probability of taking out
a pion from vacuum is proportional to F. Proportional to Vaf. So this is usually the VAF. Okay?
In some very different context this can be very different things, in spontaneous imagery breaking,
in another model this is called the hicks VAF. Okay, this is the miks VAF. And in QCD this is
called the pion K constant, but its the its some VAF that breaks cierl spontaneous symmetry
breaking, the global symmetry is the chi rale symmetry. This is the doing this kind of formalism
you dont refer to the hicks field. You dont have to refer to the hicks field, and in QCD its not
clear whether there is a hicks field or not, but everything else (or fixed) everything I said still goes
through because it just relies it doesnt rely on the my initial picture of Mexican hat potential.
Has to have a hicks field. But it doesnt depend on that at all. Thats just thats just a cute
example, simplest example I can write down, but everything else goes through without relying on
the skater field, okay?

63

MARCH 5TH, 2013

March 5th, 2013

Okay. So far, we have been talking about spontaneous [inaudible] just with the scalar field. Its
see so called spontaneous symmetry breaking of global symmetry. Although I think Ive told you
also that gauge symmetry is not a symmetry. That is the only symmetry. Okay. That is always a
spontaneous, although you frequently heard a word of spontaneous symmetry, spontaneous breaking
of gauge symmetry, but that is a misnomer. So that is gauge symmetry is, I think I told you that
gauge symmetry is not really a symmetry. But we are going to use that word anyways, consistent, so
commonly used. Today we are going to talk about what happened if we gauge the global symmetry.
Remember still our canonical example is a complex scalar field.
V () = 2 + ( )2
and when 2 < 0 you will have this Mexican hat potential. And there is a radial mode. Just to
remind you, we decided the distance away from the center, called v for vacuum expectation variable
of this field
|| 1
hi = v =
2

m2h = 2v 2

This is what we have done so far with this scalar field. Now I want to put a photon into this system.
Okay? For the gauge field into this system, and so I will introduce a gauge field. A mu. Again, you
want gauge field. You want vector field lets say, U1 gauge field. There is a gauge transformation.
Im going to gauge transformation associated with this view 1.
1
A A (x)
g
And I will also say has a charge +1, like an electric charge. So transforms ei(x) . The
Gauge transformation is just that, A and with charge +1.
Kinetic Terms
1 2
( )
2
Kinetic terms looks like in original theory, okay? Now this is not quite invariant under that. Okay?
Lets call this gauge transformation. Okay. But as we already learn in QED the trick is write some

64

8 MARCH 5TH, 2013


covariant derivatives. Whenever I say g, its gauge coupling; or if there is a charge, I use q.
D = ( + igA )

D gauge trans ei(x) D


hat is the benefits of including the well, by putting them together basically. So therefore, its
clear to how to write a LaGrangian that is invariant under this gauge transformation which is this.
Okay? That is our LaGrangian
1
L = F F + |D |2 V ()
4
So, okay. Everything looks pretty innocent at this point. Our LaGrangian, lets not worry about
the gauge fixing and anything like that, is just that. This is a kinetic term because I introduced the
A field. I have a kinetic term for this guy, so D mu ph iSquare over, then Im getting to potential.
This is called a boolean hicks model. Everything has a name. It is very important. It is knowing the
name is more important than sometimes knowing the exactly the physics. Because you immediately
know what people are talking about, okay? So that is our LaGrangian. Now you have to go work
at it. Lets try to do it. So to proceed, we will obviously start with this point, okay? Ill expand my
field around this point, because I want it to be in the broken phase. So, lets see, in broken phase,
what I will have, well, there are different parameterization, but the parameterization we have been
using is the following. By the way, there is an equal probability or in contrary parameterization
with square root and without square root too. Anyway, that is completely, it is very annoying. It
is annoying if you talk about standard model. You can talk about with the vacuum value of the
field, whether with or without square root. With square root 2 it is 100, it is the same as top mass.
With square root 2 it is 250 something DV. If you want to write a paper about why its the same
as top mass you have to be aware this is a convention dependent statement.
Broken Phase
1
i
(x) = (v + h(x))e
2

(x)
v

I already have two degrees of freedom to parameterize. There is an angular variable that takes you
around the circle and there is this radial mode, which we call h. Now we have to work it out a

65

MARCH 5TH, 2013

little bit on the board


(x)

ei v
D =
2

i
h + i + ( )h + ig(v + h)A
v

normal looking kinetic terms

|D |2 =

mass term and Higgs field

z
}|
{
z
}|
{
1
1
1 2
2
2
2 2
(h) + ()
+
g (v + h) A
2
2
2
1 1
1
+ 2 ()2 h2 + ()2 h
4v
2v
1 g
+ g()(v + h)A + ()h(v + h)A
2v
|
{z
}
goes away by gauge transformation

Under Gauge transformation + v(x). So we gauge away ( = 0; unitary gauge).


1
g2
L = (h)2 + (v + h)2 A2
2
2
1
g2v2 2
= (h)2 +
A
2
| 2{z }
=

And the g 2 v and

g2
2

m2
A A2
2

+g 2 vhA2 +

g2 2 2
A h
2

mass term

terms are fixed by D. I want to in particular draw your attention to these

things. Okay? All these coefficient are fixed. All these things are fixed. Basically, by the form of
this, like this. All of these are completely fixed. Okay? So, you know, in other words, in order to see
whether, so, okay, let me first say a few oh, in order to see whether, just in order to test whether
this scheme that we have drawn up so far, to give gauge a mass is correct or not, it is not enough
to just see that. Okay? It is far from enough to just see that, just see a Higgs boson as I would say,
Ill comment later on. But it is very, you know, very important to also measure these couplings
because these couplings are precisely related to each other and related to the mass. That is why
its important even if, even as you discover some boson, it may not be the Higgs boson, you wanted
to measure these couplings. So you want to measure the coupling with a Higgs, with two As, and
you want to measure the coupling with the Higgs, two Higgs with two As. Okay? This is U1, but
you must have read a newspaper for SU2 version of this, for some reason that is more important.
And this is more or less measured already. This is almost impossible to measure. But that is the
prediction, even measuring this is already a big confirmation of how this works. You must know
what these things are called. This is called Higgs mechanism. I will clarify the terminology further
later. This is Higgs mechanism. But, this week there is a big meeting in morian, everybody know
where it is? I dont. It is somewhere in Europe. It is a ski resort. Doing physics is great. You go
66

Coulomb

Phases

2 > 0, no SSB

mA =0
2 polarizations: =1

charged
2 massive scalars

2 < 00

mA 6=0
3 polarizations: =1,0

= would-be Goldstone
h= radial mode

MARCH 5TH, 2013

Table 1: Phases

to ski resorts. And having meetings, and that is usually some place where big experimental result
gets announced. I was just told that the official name now for that meeting at least, of everybody
else is going to call it Higgs boson but for that meeting it is going to be called BEH boson (Brout,
Englert, Higgs).
And now we proceed to the dropping of the names.
So anyway, but those days, people, when you write a paper, you have to, it was even before my
time, so it was when you write a paper, you call up your friend say Im going to send you a preprint.
Months later the preprint arrived and so on, so forth. These days, if you have anything in your
head, you put it on Facebook first.
Everything is uniquely normalized of course. But Im just going to do the classic. Actually Im
not going to normalize it. That is going to take another three weeks to do, introducing goldstone
which we havent even talked about. Im just going to do the classical, mostly classical part of the
story. Okay? This is called the Higgs mechanism. Maybe you have been slightly wondering what
exactly is going on. This is nice. But when Peter Higgs likes to give a talk, Ive seen one of his
talks, his talks, is my life as a boson. Have you seen that? You should go to see the talk. It is
interesting. He said when he first wrote that paper, he went to Princeton and Harvard and MIT to
give talks. Everybody assured him he is completely wrong, because the existence of goldstone after
symmetry breaking has been proved by some, I dont know, called C style algebra or whatever,
some axiomatic quantum field theory. So lets understand a little bit better, just in that vein. Of
course, axiomatic quantum field theory doesnt understand in this. That is almost totally useless
endeavor. Okay. Lets try to understand why exactly we get a mass. Or why do we get a mass?
This is a completely correct but you may want to, may want me to say it in a more deeper sounding
way to get convinced. Here is a slightly deeper way of saying it.
I have goldstone and I have radial mode. But this guy obviously becomes our H in the end. But this
67

8 MARCH 5TH, 2013


guy has come here actually, become one of the polarizations. Usually I said it becomes longitudinal
polarization but that is slightly gauge dependent that statement. What these three polarization is
coming from is the taking a goldstone mode and taking this to form of three polarizations. You can
see this as I again, I want to, I can gain away the Pi. And it becomes part of A, basically. There is
no, there is not a invariant distinction between Pi and A basically. When you shift the Pi, A shift.
Ill write it down. As you can see, any time you shift A to some, alpha Pi also shift. There is no,
with A around, there is no invariant distinction between Pi and A. They just become one of the
same state basically. This is the official statement of Higgs mechanism. The goldstone is again the
terminology is eaten by the gauge boson. Gauge boson become massive and the goldstone here is
not called goldstone anymore. This is called a would-be goldstone. If there is no A, it is its own
goldstone but if there is A, when there is A, this is not an independent observable physical state.
And these are also not independent observable physical state. They combine into one massive
vector basically. That is the official statement of Higgs mechanism. This is the Higgs boson. But
it is not, you see that it is actually just playing a peripheral role in this case. It is not completely
we will argue that, why, its actually not necessary, Higgs boson here.

68

MARCH 7TH, 2013

March 7th, 2013

Today we wrap up a few remaining issues on symmetry breaking with U1.


We already said that the gauge U1, in gauge symmetry, gauge symmetry breaking, the story is that
you are going from A mu, massless, with no symmetry breaking. This is massless. The polarization
is plus/minus 1. Another complex scalar field with two degree of freedom, okay. But then you
go through a phase where there is symmetry breaking, this becomes massive. The correlation is
plus/minus 1 and 0. On the other hand, we said that in, after spontaneous symmetry breaking,
the salear vector changes as well. It goes to give you a Goldstone, and another scalar. We said
this is not observable, directly observable, because its becoming part of the massive vector, with
three polarizations. This is called eaton. Thats basically I think where we are last time. Does this
picture bother anybody? We went through two ways of convincing ourselves, we just do the straight
LaGrangian way, and we did a more formal sounding way, talking about poles or propagators. Poles,
the second one is actually very close to actually prove a theorem. The first one is showing you
what is going on. For your understanding of standard model its important, particle physics. Lets
talk about Fermions in connection with spontaneous symmetry breaking. Maybe lets not do this.
Lets do, lets briefly remark on some more complicated spontaneous symmetry breaking pattern
first. How about I have a U1, still have a U1 but now I have multiple Higgs. Now I have multiple
field, phi1 with charge Q1 and phi2 with charge Q2 and so on. How about I have that? Lets
say every one of them gets a V1. If any of these values is nonzero, U1 symmetry is spontaneously
broken. There is actually for each one of them, in general, for each one of them, there is a U1 global
symmetry as well. If I just do this, now, the question is, what happened to the theory, after Im
breaking all those things. Okay? After I give all these things values. To be specific, lets consider
two. The theory will be ... D mu phi1 is IGQ1A mu, 1 and so on. Okay? Write this down, this is
going to be homework. (chuckles). They just go do this, basically. Generally this has value, that
means there is Vphi1, phi2 as well, which Im not writing down at this moment. But you know
what is going to happen. Roughly speaking you know what is going to happen. Roughly, you know.
What is going to happen is that there is going to be a (coughing). Sorry. Of course the gauge
boson will become massive. What is going to happen, so you can do the whole thing and look at
the mass and so on. What is going to happen, A, you are going to have MA. The mass is, the

69

9 MARCH 7TH, 2013


gauge boson is going to have a mass. Exercise to find out what the mass is, okay? Its the sum of
the values and so on, so forth. But it is going to have a mass because the symmetry is broken. The
question is what happened to the rest of the spectrum? I start out here. These are two complex
scalar field. This is 4-degree of freedom.

U (1) :

4 d.o.f.

1 : Q1 , 2 : Q2
h1 i = v1

h2 i = v2

so
|D 1 |2 +|D 2 |2 + V (1 , 2 )D 1 |{z}
= ( + igQ1 A )1
| {z }
A=1,0

mA

You know one of them is gone. One of them which is whatever a Goldstone is going to be going
to the, is going to the mass, the massive A vector field. That massive vector field has to have one
Goldstone, has to eat one Goldstone to become massive because otherwise it doesnt have three
polarizations, so its just wrong. But then the rest of the three, still remaining, you still have three.
The other three scalars. This theory has three extra scalars. Without going through the whole
exercise, the exercise itself is pretty simple, multiply things out and look at the mass and so on and
the couple. Now, what if I also do another exercise, suppose 2U1. 1U2. I have two U1 symmetry.
I have two gauge bosons. A1 and A2. Suppose I have that now, suppose I have a charge, a particle
that is charged under both. I can have that too, right? I can have a particle charge under 2U1
symmetries, and with Q1 and Q2. Suppose I have that. So this sometimes can be summarized in
the diagram like this. It is called U11, U12 and 5 I guess. Now, depends on what field are you in,
this diagram have different names. This diagram has different names. In some fields its called this.
This is actually called a minimum moose, longer lattice, quiver, but anyway, so Im just preparing
you for something. These are the sides and this is called the link.
U (1)1 ) U (1)2
| {z } | {z }
A1

: Q1 , Q2

A2

and you can imagine that is the link between the two sites U (1)1 , U (2)2 .
D = ( + iQ1 g1 A1 + iQ2 g2 A2 )
70

MARCH 7TH, 2013

and has 2 d.o.f. hi = v would-be 1 Goldstone.


But its important to give them good names. So that they can be remembered. So anyway, string
theory, deeper in physics, this is called a quiver. Symmetric H theories. Anyway, suppose I have a
quiver, a moose like this. So what do I, what is the LaGrangian? It is actually very simple. Those
names are sometimes scary. But it is actually simple. D mu Phi is partial mu. Now you have to
include all the gauge field here. Because the Phi is a charged under two gauge fields. It has to be
locally gauge, has to be gauge under the transformation of the, there are two independent gauge
transformation. So the current derivative has to involve all of them. Q1 in general, there is a G1A
mu plus IQ2G2A mu 12, not using, choosing very illuminating notation, but this is what you do.
The LaGrangian is, again that. You can open this up. And do our usual thing. You can open this
up and look at the masses and so on, look at the mass of gauge bosons and so on. But you already
know what is going to happen in the end. Qualitatively, what is going to happen in the end, you
can work out the details, or you will work out the details, is that there are two degree of freedom
here. Okay? Only one of them shift under the gauge symmetry. After, again after SSB, so after
SSB, only one Goldstone, would be Goldstone, only one would be Goldstone. Okay? Therefore,
you know that what is going to happen is that in the end, there is going to be one unbroken, one
massless vector. Okay? It is going to be one massive vector, plus one massless vector. Because
there is only one Goldstone, it can only supply one of the vectors with its longitudinal mode. Just
by counting the degree of freedom, so, yeah. But this massive one is not any one of these. This
massless one is not any one of this. They are linear combinations. You have to work out the linear
combination. Basically, all of them are just linear combinations of A1 and A2. Okay? That you
have to, you open this up, square, there is a mass matrix, Again this is similar to what you do with
couple harmonic oscillator. You have to diagrammize the frequency mate I can to find the Eigen
mode. There are two Eigen mode. One of them is massive vector. You know there can only be
one because there is only one Goldstone to be eaten. Is this okay? Both of these two situations,
these two situations happen, this happens in standard model. A similar story happen here in the
standard model, similar story to that happens in EMSSM. Lets also talk about the Carol Fermions.
Do I want to talk about Fermions here or later? Lets talk about it here. Lets remember what
the Fermion mass is. Fermion mass is something like that. But we know that, you still remember
Fermion part at the beginning of this quarter, we can write it as a, left-handed or right-handed

71

MARCH 7TH, 2013

component. Okay?


=

L
R




= R
, L

= m(L R + R
L )

So that, okay? That is that, okay? That is just that. Fermion mass is mass. This okay? Im
multiplying them out. Okay? Lorentz invariance, this basically is fixed by Lorentz invariants.
Lorentz invariants basically is saying that left-handed and right-handed for massive Fermions are
not good quantum numbers, because they mix. That is basically what Lorentz invariants is telling
us for massive Fermions. On the other hand, there could be situations that those two charged under
different gauge symmetry. That would carry different charges under gauge symmetry. Okay?
Under U (1) Under U1, under U1, there can be two situations, okay? One is that Q left is the same
as Q right. The left hand Fermion and right-handed Fermion has the same charge under this U1.
So, ah, I should actually say something slightly different. Now, lets look at the kinetic terms. Lets
look at the kinetic terms. The kinetic terms looks like this, looks like this.

( + iQR A )R
( + iQL A )L + R
L

This is 2 by 2 poly matrices basically Pauli. It doesnt matter for our argument but these are
kinetic term. This is gauge invariant, kinetic terms. Gauge invariants does not tell me anything
about QL versus QR, okay, of the kinetic terms. There is no limit. I can just set this whatever I
want. There are two situations. Case one is QL equals QR. This is called a vector like, Fermions. It
doesnt mean the Fermion becomes a vector. It just means that the left hand and right hand have
the same charge. This means that it is invariant. That just means that this term is invariant, if QL
is the same as QR, and the charge of this, rotation of this is opposite of the rotation of that.
invariant
i) QL = QR vector-like m
 not invariant
m

ii) QL 6= QR chiral 
This is gauge invariant. Then there is another situation with QL not the same as QR. Okay? In
this case, this is called the kierl, the Fermion is called the curial. In that case, it is not invariant.
That cannot exist. A term like that cannot exi For kierl Fermion does not have mass basically,
that is the statement. Just because you can assign left-handed and right hand Fermions, different
charges. Lets consider an extreme case.
72

9.1 Standard Model


Consider QL = 1, QR = 0

MARCH 7TH, 2013

y(L R ) + y R
L )

hi = v
yv (L R ) + h.c.
|{z}
m

= m L R + h.c.
QL is 1, QR is 0. That is one of those cases. There is nothing, you cannot write a mass term
basically. Okay? Look at this mass term, only QL carries charge. U1 rotation is certainly not
gaiblg gauge invariant. So now you wonder is there any way to give those Fermion masses,
because after all we live in a world with Fermion masses, where Fermion has masses. Maybe you
say, okay, maybe the Fermion world is all vector like, but it is not. Unfortunately its not. We
will talk about it too. Suppose I give you a pair of Fermions, how do you give them mass? The
trick is again there is a spontaneous symmetry break. It is going through spontaneous symmetry
break.
Consider , Q = 1, hi = v.
= (v + h) yhL R + h.c. =

9.1

m
hL R
v

Standard Model

So, are you all, you have all seen standard model at some point? Particles and interactions? Not
see it but know about it. Standard model, so the story we just said about U1 is important for
understanding the most intriguing part of, not the most intriguing but at least the most, one of
the most interesting part of the standard model. Standard model are controlled by gauge similar
trees symmetries.
gauge symmetry
SU (3)C

SU (3)C SU (2)L U (1)Y


strong interactions 8 gluons

quarks are 3: u,d,c,s,t,b

SU (2)L U (1)Y
SU (2)L

doublet

u
d L

73

which mirrors UR , dR singlet

9.1 Standard Model


doublet

9

scalar: H =

h
H0

SSB

h
v + h + ia


Y =

MARCH 7TH, 2013

1
2

SU (2)L U (1)Y
which gives 3 massive particles W , Z and a massless photon. We have 3-Goldstones
yu H QL uR + yD H c QL dR
|{z}
(udLL )
actually which way is the left-handed. Lets call this left-handed. Lets call this right-handed. Okay.
They carry different charge, if you want, under SU2 gauge interaction in the most extreme way.
One of them couples to them, one of them dont. Okay? But this too also interchange each other
under parity. Okay? So means that if you imagine you put a mirror here, this goes to that. Okay?
So if you put a mirror here, this just goes to that. Okay? Gauge interaction SU2L violate parity, so
this is parity violation. Is there anything wrong, violating parity? Nothing. It can be violated. It
is violated in nature. Nothing, nothing goes horribly wrong. For a long time we thought it would
be, parity is obviously a symmetry but its not. But again very similar to what I said here, very
similar to what I said here. You see the statement I just made is not completely consistent with
Lorentz invariants, of massive Fermions at least, because we know from experiment these things
gets, all have mass. By the way this is called opt because usually we write it on top, I dont know,
top down, the term strange is a little bit, I dont know why, maybe strange that they dont think
it should be there, and charm is just try to make a good name. I dont know. But this is not
consistent with Lorentz invariants because for massive Fermions, so there is a very intuitive way of
understanding why mass makes two polarizations, because for massive particle there is a rest frame.
Once you get into the rest room you lost a direction, there is no direction. In the rest room you can
pick any access of your polarization vector, and obviously left-handed can become right-handed.
Right-handed can become left-handed. Its your choice. In the massless case it makes sense because
there is no rest room. You are always moving in that direction. You always use that direction as
your reference, polarization. For massless Fermions there is a direction. We also know that this
symmetry is broken. The good thing is, so at this point in order to accommodate massive Fermions
you need to do one of those two things. Either break the symmetry or break Lorentz symmetry.
Lorentz symmetry we are not breaking. We are not breaking this one. But we know this is broken.
74

9.1 Standard Model


9 MARCH 7TH, 2013
The gauge bosons are massive. Its broken by a doublet. It can be broken by a doublet, scalar
doublet. This is again a scalar. H is a have to write it like this, H0. Just in a very suggestive
form, and this under spontaneous symmetry breaking, you can write this as an H plus, V plus H
plus IA. So that is how you do standard model doublet. And, okay. On the other hand, the H also
charges under this U1. So we havent talked about U1 yet. H also is charged under this U1. So YH
is 1/2, according to normalization, that is the usual normalization we use. So thats that. So this H
when it gets verve here, first of all you may ask me why am I giving vev to this guy, why am I doing
this, why am I not giving to it that guy or whatever it is? (vev?) The answer is it doesnt matter.
You can put it wherever you want, you can give vev to every component. It is similar to the U1
case, the valve can go any direction and the end result is the same. All those are vacuum, all those
vacuum are equivalent as we have said, along the circle. Similarly, you can take this to where SU2
rotation, give vev to anywhere you want, but all those vacuums are completely equivalent. Very
similar to the U1 case, okay? This breaks both SU2, cross U1Y. It breaks both of them. Okay?
On the other hand, there are only well, the global symmetry breaking is SU2 down to nothing.
There are only three Goldstones which is this and this. Its again very similar to the U1 case. Its
not everything in the scalar, it can be a Goldstone. Okay? In this case, these and that can, gold
stones. These and that. This is a complex scalar so there are two degree of freedom. Between two
of them there is one of them. There are only three Goldstones and how many generators are here?
How many gauge bosons are here? SU3 has one, how many, SU2 has how many? Three. Its SO(3)
basically, just three rotations. There are four generators. Three Goldstones. Okay? There is one
remain massless. Very similar to what I just erased, similar to what I just erased. Okay? So what
is going to happen is going to be three massive gauge bosons, and they all have names. They are W
plus/minus and the Z. And one massless. So because, again, which one is, what is the massless one
here? Its the photon. That is the symmetry breaking. Again, but again, its not exactly this one
doesnt become the photon, and this one is not exactly these three, and it is a linear combination.
In particular, Z and the photon are linear combination of the tau 3 here, and this U1 basically.
But lets not work it out. It is going to take us too long. But so yeah, these are, Im mentioning
this because this is continuous, it is a continuous story that we have talked about. You see there
is nothing to it. All you have to do is write down the current derivative and diegriz it, basically.
(?) With this H obviously we can write masses now for the Fermions. With this H I can start

75

9.1 Standard Model


9 MARCH 7TH, 2013
to write YHQLUR, call this YU plus YD, H, conjugate, QLDR. Doesnt quite matter what this
is. But you see these are the mass terms. Okay? This is, again, this is just my shorthand way
of writing down ULDL. These are the mass terms of the standard model quarks basically. They
have different Yukawa couplings in general. They dont have the same. We all know that there are
more than one quarks. Right? In general, these are 3 by 3 matrices. They are three quarks, three
flavor of quarks. They are [inaudible] too. To understand that, that is called a flavor problem. We
dont know really how to derive these Yukawa matrices from first principles, and its unlikely we
are going to know any time soon. Okay. That is basically the standard model. Everything what
I said has already been discovered, by the way. Im not making any of this story up. Lets take a
step back. Since we have been talking about Higgs mechanism and so on, lets take a step back,
lets try to understand Higgs mechanism slightly better, what does it really mean. For that lets go
back to the U1 story. Lets from a bottom up point of view, lets take a bottom up point of view.
Lets suppose Im very dumb. I dont know about lets just say I observe a massive gauge boson.
Okay?
m2
1
F F + A A A
4
2

unitary gauge

Ill write a mass term like that. Okay? You say wait, that is all wrong, right? It is not gauge
symmetry. We all know we need gauge symmetry, blah blah blah, but actually we dont. First of
all we dont. Lets say we do, but the usual story tells you that now gauge symmetry forbid this
term. That is why I need to introduce, that is why I need to do this whole Higgs boson mechanism
that write a cowritten derivative, give you a mass, that part of the story is wrong. The reason is
that, the reason this is wrong is that this is a secret to the gauge symmetric.
in general gauge Okay? Its not gauge symmetric when you fix to a unitary gauge which is not
such a big surprise. But suppose lets work in the slightly more general gauge. In general, in a
general gauge I can write this as this following form. You see this is completely gauge invariant.
Remember in gauge transformation A mu goes to A mu D mu, alpha we already said there is no
need to realize Pi, Pi shift as of this. Under this gauge transformation this is completely invariant,
complete gauge invariant.
m2 
2
1
F F + A A +
4
2
v

76

9.1 Standard Model


with gauge transformation

A A + ,

MARCH 7TH, 2013

On the other hand we know these two are completely [inaudible] to each other, there is nothing
different. Using gauge transformation I can gauge only Pi it become this, become unitary gauge.
Therefore, this is actually gauge invariant. There is nothing wrong with that. Gauge invariance
does not prevent you to write down a mass term. It is completely fine. The only thing you have to
remember is now there are three degree of freedom. But that is the real physical statement, between
massive and massless gauge boaston is not whether its gauge symmetry or not. Its whether there
is three degree of freedom or not. That what Im trying to get at. This is a real physical statement
is that. Goldstone is everything. The eaten Goldstone captures the essence of the physics really.
This also tells you that the LaGrangian sometimes can be very deceiving. Although we are getting
really used to doing everything with LaGrangian, but LaGrangian sometimes can tell you, teach
you a very wrong lessons, because it can hide things. You can hide things with LaGrangian.
You can do horribly wrong things with LaGrangian. If something looks horribly wrong, it can be
completely fine. Okay? So for that, lets just digress. Let me give you an example of something
that looks horribly wrong but is completely fine. Lets actually take this example. Lets take
this one. If I square it out, it goes like, Im going to be using shorthand, but I think its pretty
understandable what these things are. Its A square plus 2 Pi, 2D Pi, A plus, Im actually setting,
by the way, I think Im not being careful with gauge couplings, Im setting gauge coupling to one
in this case.

1 2 1 2 2
F mA A + 2A + ()2
| {z }
4
2
2

But general lesson is still there. This part of the story is a little bit schematic. That, okay. So this
is also known as a negative Pi box Pi after integrated by part. Now the next thing you observe
is that this is a quadratic in Pi. This is a quadratic in Pi. Nothing goes beyond that. When
something is a quadratic in a field, we dont even need to do preservation theory. We can just do
exactly its a gaussian, if its quadratic its gaussian. I can integrate this out basically. I can
perform D Pi exactly. Integrate out. In fact, I dont need to do that. I can use the equation motion

77

9.1 Standard Model


9 MARCH 7TH, 2013
of Pi to integrate up, because something quadratic, there is nothing wrong with it.


m2A
1
F
L= F 1+
4
2
Now I can integrate, I can substitute, integrate means I substitute everything I see a box Pi, I
substitute with DA. Okay? With everything, as every time we see Pi, I substitute so I make this
the following substitution. Probably having a heart attack with this one box. But thats okay.
The thing we end up with is the following. Again, Im not being careful with the factors. That
is the LaGrangian. So if I write it this LaGrangian to begin with you will kill me, because its
like, what is this? This is not local. But we know secretly its just that. That is completely fine.
Anyway, its not a very important for our course. But this is just an example of, LaGrangian can
fool you. If you are writing down this LaGrangian you think you are doing cutting edge research
in particle physics, you are not. You are just doing that. Okay? Okay. Now, this does bring up
an interesting point. But because now I completely trashed gauge symmetry, I told you early on
it is not symmetry and so on, but you see that I can make anything that doesnt look like gauge
symmetric gauge symmetry. Anything is gauge symmetric. Just have to replace A, A, replace with
that and Im done. The whole thing is gauge symmetry. Gauge symmetrical. Now you say now
we dont have roots anymore. I can write down everything I want, anything I want to write. Now
I can write down anything I want to, so all hell breaks through. Again, Im coming back to this
LaGrangian. I already told you that this term is fine. Okay?
m2
1
L = F 2 + A A2 + a(A)2 + bA4
4
2
Now you ask why stop here. Im going to keep going. Im going to add this one. Nothing wrong.
Because you do add something like this as a gauge fixing but this is not gauge fixing, Im not even
doing that at this moment. This is a physical coupling Im adding here. Okay? I can do this as well.
You dont have dimensional terms but maybe those things are suppressed by some energy scalar.
But these are renormalizable looking. Nothing quite wrong. This is the case, where actually, we
are keeping track with the Pis are very important. You keep track Pis are wrong actually teach
you a lot, okay is this so lets try to do that. You notice that the deep, my rule is to replace every
A with that.
A = (A + ) = A + 2


(A)2 = a (A)2 + (2)2 +
78

9.1 Standard Model


9 MARCH 7TH, 2013
Every single A with that, okay? Im going to do that. Okay? Im going to do that. Again
schematically, okay? Lets imagine I redefine the Pis with V inside of Pi now, Im just not going to
keep all these things around, but you can put all the factors in and the story is the same. Okay?
This one is DA plus box Pi. You say whats wrong with that? You can put it back in. And so DA
square is A, DA square plus box Pi square plus ... okay? Now, but also A square contribute a term
like this. So now lets put these two terms together, and A square, okay? You say what is wrong
with this story? Its a little bit wrong.

m2A
1
2 + a(2)2 2 2
2
p (p m02 )

Because, well, lets go do it. We know that this is completely fine. This is just a kinetic term for
the Goldstone. Goldstone needs to have kinetic terms. Anything that can legitimately be called
a Cal man field needs a kinetic term is fine but this is not fine with these two. With these two
now you can solve it. If I use this to write down my propagator, this is going to be a P square in
the propagator there is going to be a P to the force in the propagator. Therefore, the propagator
will have the form of P square, P square minus P prime square, something like that. That is my
propagator. There is a pole in P square equals 0, that is fine. Thats Goldstone. But there is
another pole. Okay. Once you have P to the fourth there is another pole. There is another pole
on the order of MA square over A. This is certainly unphysical. There is something wrong with
this theory, a theory just like that. Because you are introducing a spurious poles, this will be,
okay?
pole

m2A
unphysical
a

a must be small

To make this theory healthy, A must be, be small so this pole is above your cutoff. You can put this
in. But there is a limit, how big this A is. This kind of story tells you that Pi teach us a lot. Keep
the Pis around, they really teach you a lot about what is allowed to have. Not gauge symmetry
itself. But you have the Pis around, tells you how healthy a theory is. Okay? So lets do SU2
version of it, of Pi, SU2 version. Lets do SU2 version. Imagine everything here SU2, lets imagine
all these things are SU2s. Lets define a few things. A will be sigma AA. This is SU2 generators.
Pauli matrices. The appropriate quantity now, we want to have is, another is the, instead of E to
the I Pi, we will have E to the I sigma A Pi A over V, which I will call E to the I Pi over V, but

79

9.1 Standard Model


the Pi is sigma A Pi A. This is the generators, SU2 generators.
A = a Aa

U = ei

a a
v

= ei v

= aa





()2 2
2
2
L = v trace |U | + trace () +
+
v2
2

80

MARCH 7TH, 2013

10

10

MARCH 5TH, 2013

March 5th, 2013

Okay. So far, we have been talking about spontaneous [inaudible] just with the scalar field. Its
see so called spontaneous symmetry breaking of global symmetry. Although I think Ive told you
also that gauge symmetry is not a symmetry. That is the only symmetry. Okay. That is always a
spontaneous, although you frequently heard a word of spontaneous symmetry, spontaneous breaking
of gauge symmetry, but that is a misnomer. So that is gauge symmetry is, I think I told you that
gauge symmetry is not really a symmetry. But we are going to use that word anyways, consistent, so
commonly used. Today we are going to talk about what happened if we gauge the global symmetry.
Remember still our canonical example is a complex scalar field.
V () = 2 + ( )2
and when 2 < 0 you will have this Mexican hat potential. And there is a radial mode. Just to
remind you, we decided the distance away from the center, called v for vacuum expectation variable
of this field
|| 1
hi = v =
2

m2h = 2v 2

This is what we have done so far with this scalar field. Now I want to put a photon into this system.
Okay? For the gauge field into this system, and so I will introduce a gauge field. A mu. Again, you
want gauge field. You want vector field lets say, U1 gauge field. There is a gauge transformation.
Im going to gauge transformation associated with this view 1.
1
A A (x)
g
And I will also say has a charge +1, like an electric charge. So transforms ei(x) . The
Gauge transformation is just that, A and with charge +1.
Kinetic Terms
1 2
( )
2
Kinetic terms looks like in original theory, okay? Now this is not quite invariant under that. Okay?
Lets call this gauge transformation. Okay. But as we already learn in QED the trick is write some

81

10 MARCH 5TH, 2013


covariant derivatives. Whenever I say g, its gauge coupling; or if there is a charge, I use q.
D = ( + igA )

D gauge trans ei(x) D


hat is the benefits of including the well, by putting them together basically. So therefore, its
clear to how to write a LaGrangian that is invariant under this gauge transformation which is this.
Okay? That is our LaGrangian
1
L = F F + |D |2 V ()
4
So, okay. Everything looks pretty innocent at this point. Our LaGrangian, lets not worry about
the gauge fixing and anything like that, is just that. This is a kinetic term because I introduced the
A field. I have a kinetic term for this guy, so D mu ph iSquare over, then Im getting to potential.
This is called a boolean hicks model. Everything has a name. It is very important. It is knowing the
name is more important than sometimes knowing the exactly the physics. Because you immediately
know what people are talking about, okay? So that is our LaGrangian. Now you have to go work
at it. Lets try to do it. So to proceed, we will obviously start with this point, okay? Ill expand my
field around this point, because I want it to be in the broken phase. So, lets see, in broken phase,
what I will have, well, there are different parameterization, but the parameterization we have been
using is the following. By the way, there is an equal probability or in contrary parameterization
with square root and without square root too. Anyway, that is completely, it is very annoying. It
is annoying if you talk about standard model. You can talk about with the vacuum value of the
field, whether with or without square root. With square root 2 it is 100, it is the same as top mass.
With square root 2 it is 250 something DV. If you want to write a paper about why its the same
as top mass you have to be aware this is a convention dependent statement.
Broken Phase
1
i
(x) = (v + h(x))e
2

(x)
v

I already have two degrees of freedom to parameterize. There is an angular variable that takes you
around the circle and there is this radial mode, which we call h. Now we have to work it out a

82

10

MARCH 5TH, 2013

little bit on the board


(x)

ei v
D =
2

i
h + i + ( )h + ig(v + h)A
v

normal looking kinetic terms

|D |2 =

mass term and Higgs field

z
}|
{
z
}|
{
1
1
1 2
2
2
2 2
(h) + ()
+
g (v + h) A
2
2
2
1 1
1
+ 2 ()2 h2 + ()2 h
4v
2v
1 g
+ g()(v + h)A + ()h(v + h)A
2v
|
{z
}
goes away by gauge transformation

Under Gauge transformation + v(x). So we gauge away ( = 0; unitary gauge).


1
g2
L = (h)2 + (v + h)2 A2
2
2
1
g2v2 2
= (h)2 +
A
2
| 2{z }
=

And the g 2 v and

g2
2

m2
A A2
2

+g 2 vhA2 +

g2 2 2
A h
2

mass term

terms are fixed by D. I want to in particular draw your attention to these

things. Okay? All these coefficient are fixed. All these things are fixed. Basically, by the form of
this, like this. All of these are completely fixed. Okay? So, you know, in other words, in order to see
whether, so, okay, let me first say a few oh, in order to see whether, just in order to test whether
this scheme that we have drawn up so far, to give gauge a mass is correct or not, it is not enough
to just see that. Okay? It is far from enough to just see that, just see a Higgs boson as I would say,
Ill comment later on. But it is very, you know, very important to also measure these couplings
because these couplings are precisely related to each other and related to the mass. That is why
its important even if, even as you discover some boson, it may not be the Higgs boson, you wanted
to measure these couplings. So you want to measure the coupling with a Higgs, with two As, and
you want to measure the coupling with the Higgs, two Higgs with two As. Okay? This is U1, but
you must have read a newspaper for SU2 version of this, for some reason that is more important.
And this is more or less measured already. This is almost impossible to measure. But that is the
prediction, even measuring this is already a big confirmation of how this works. You must know
what these things are called. This is called Higgs mechanism. I will clarify the terminology further
later. This is Higgs mechanism. But, this week there is a big meeting in morian, everybody know
where it is? I dont. It is somewhere in Europe. It is a ski resort. Doing physics is great. You go
83

10

Coulomb

Phases

2 > 0, no SSB

mA =0
2 polarizations: =1

charged
2 massive scalars

2 < 00

mA 6=0
3 polarizations: =1,0

= would-be Goldstone
h= radial mode

MARCH 5TH, 2013

Table 2: Phases

to ski resorts. And having meetings, and that is usually some place where big experimental result
gets announced. I was just told that the official name now for that meeting at least, of everybody
else is going to call it Higgs boson but for that meeting it is going to be called BEH boson (Brout,
Englert, Higgs).
And now we proceed to the dropping of the names.
So anyway, but those days, people, when you write a paper, you have to, it was even before my
time, so it was when you write a paper, you call up your friend say Im going to send you a preprint.
Months later the preprint arrived and so on, so forth. These days, if you have anything in your
head, you put it on Facebook first.
Everything is uniquely normalized of course. But Im just going to do the classic. Actually Im
not going to normalize it. That is going to take another three weeks to do, introducing goldstone
which we havent even talked about. Im just going to do the classical, mostly classical part of the
story. Okay? This is called the Higgs mechanism. Maybe you have been slightly wondering what
exactly is going on. This is nice. But when Peter Higgs likes to give a talk, Ive seen one of his
talks, his talks, is my life as a boson. Have you seen that? You should go to see the talk. It is
interesting. He said when he first wrote that paper, he went to Princeton and Harvard and MIT to
give talks. Everybody assured him he is completely wrong, because the existence of goldstone after
symmetry breaking has been proved by some, I dont know, called C style algebra or whatever,
some axiomatic quantum field theory. So lets understand a little bit better, just in that vein. Of
course, axiomatic quantum field theory doesnt understand in this. That is almost totally useless
endeavor. Okay. Lets try to understand why exactly we get a mass. Or why do we get a mass?
This is a completely correct but you may want to, may want me to say it in a more deeper sounding
way to get convinced. Here is a slightly deeper way of saying it.
I have goldstone and I have radial mode. But this guy obviously becomes our H in the end. But this
84

10 MARCH 5TH, 2013


guy has come here actually, become one of the polarizations. Usually I said it becomes longitudinal
polarization but that is slightly gauge dependent that statement. What these three polarization is
coming from is the taking a goldstone mode and taking this to form of three polarizations. You can
see this as I again, I want to, I can gain away the Pi. And it becomes part of A, basically. There is
no, there is not a invariant distinction between Pi and A basically. When you shift the Pi, A shift.
Ill write it down. As you can see, any time you shift A to some, alpha Pi also shift. There is no,
with A around, there is no invariant distinction between Pi and A. They just become one of the
same state basically. This is the official statement of Higgs mechanism. The goldstone is again the
terminology is eaten by the gauge boson. Gauge boson become massive and the goldstone here is
not called goldstone anymore. This is called a would-be goldstone. If there is no A, it is its own
goldstone but if there is A, when there is A, this is not an independent observable physical state.
And these are also not independent observable physical state. They combine into one massive
vector basically. That is the official statement of Higgs mechanism. This is the Higgs boson. But
it is not, you see that it is actually just playing a peripheral role in this case. It is not completely
we will argue that, why, its actually not necessary, Higgs boson here.

85

11

11

MARCH 7TH, 2013

March 7th, 2013

Today we wrap up a few remaining issues on symmetry breaking with U1.


We already said that the gauge U1, in gauge symmetry, gauge symmetry breaking, the story is that
you are going from A mu, massless, with no symmetry breaking. This is massless. The polarization
is plus/minus 1. Another complex scalar field with two degree of freedom, okay. But then you
go through a phase where there is symmetry breaking, this becomes massive. The correlation is
plus/minus 1 and 0. On the other hand, we said that in, after spontaneous symmetry breaking,
the salear vector changes as well. It goes to give you a Goldstone, and another scalar. We said
this is not observable, directly observable, because its becoming part of the massive vector, with
three polarizations. This is called eaton. Thats basically I think where we are last time. Does this
picture bother anybody? We went through two ways of convincing ourselves, we just do the straight
LaGrangian way, and we did a more formal sounding way, talking about poles or propagators. Poles,
the second one is actually very close to actually prove a theorem. The first one is showing you
what is going on. For your understanding of standard model its important, particle physics. Lets
talk about Fermions in connection with spontaneous symmetry breaking. Maybe lets not do this.
Lets do, lets briefly remark on some more complicated spontaneous symmetry breaking pattern
first. How about I have a U1, still have a U1 but now I have multiple Higgs. Now I have multiple
field, phi1 with charge Q1 and phi2 with charge Q2 and so on. How about I have that? Lets
say every one of them gets a V1. If any of these values is nonzero, U1 symmetry is spontaneously
broken. There is actually for each one of them, in general, for each one of them, there is a U1 global
symmetry as well. If I just do this, now, the question is, what happened to the theory, after Im
breaking all those things. Okay? After I give all these things values. To be specific, lets consider
two. The theory will be ... D mu phi1 is IGQ1A mu, 1 and so on. Okay? Write this down, this is
going to be homework. (chuckles). They just go do this, basically. Generally this has value, that
means there is Vphi1, phi2 as well, which Im not writing down at this moment. But you know
what is going to happen. Roughly speaking you know what is going to happen. Roughly, you know.
What is going to happen is that there is going to be a (coughing). Sorry. Of course the gauge
boson will become massive. What is going to happen, so you can do the whole thing and look at
the mass and so on. What is going to happen, A, you are going to have MA. The mass is, the

86

11 MARCH 7TH, 2013


gauge boson is going to have a mass. Exercise to find out what the mass is, okay? Its the sum of
the values and so on, so forth. But it is going to have a mass because the symmetry is broken. The
question is what happened to the rest of the spectrum? I start out here. These are two complex
scalar field. This is 4-degree of freedom.

U (1) :

4 d.o.f.

1 : Q1 , 2 : Q2
h1 i = v1

h2 i = v2

so
|D 1 |2 +|D 2 |2 + V (1 , 2 )D 1 |{z}
= ( + igQ1 A )1
| {z }
A=1,0

mA

You know one of them is gone. One of them which is whatever a Goldstone is going to be going
to the, is going to the mass, the massive A vector field. That massive vector field has to have one
Goldstone, has to eat one Goldstone to become massive because otherwise it doesnt have three
polarizations, so its just wrong. But then the rest of the three, still remaining, you still have three.
The other three scalars. This theory has three extra scalars. Without going through the whole
exercise, the exercise itself is pretty simple, multiply things out and look at the mass and so on and
the couple. Now, what if I also do another exercise, suppose 2U1. 1U2. I have two U1 symmetry.
I have two gauge bosons. A1 and A2. Suppose I have that now, suppose I have a charge, a particle
that is charged under both. I can have that too, right? I can have a particle charge under 2U1
symmetries, and with Q1 and Q2. Suppose I have that. So this sometimes can be summarized in
the diagram like this. It is called U11, U12 and 5 I guess. Now, depends on what field are you in,
this diagram have different names. This diagram has different names. In some fields its called this.
This is actually called a minimum moose, longer lattice, quiver, but anyway, so Im just preparing
you for something. These are the sides and this is called the link.
U (1)1 ) U (1)2
| {z } | {z }
A1

: Q1 , Q2

A2

and you can imagine that is the link between the two sites U (1)1 , U (2)2 .
D = ( + iQ1 g1 A1 + iQ2 g2 A2 )
87

11

MARCH 7TH, 2013

and has 2 d.o.f. hi = v would-be 1 Goldstone.


But its important to give them good names. So that they can be remembered. So anyway, string
theory, deeper in physics, this is called a quiver. Symmetric H theories. Anyway, suppose I have a
quiver, a moose like this. So what do I, what is the LaGrangian? It is actually very simple. Those
names are sometimes scary. But it is actually simple. D mu Phi is partial mu. Now you have to
include all the gauge field here. Because the Phi is a charged under two gauge fields. It has to be
locally gauge, has to be gauge under the transformation of the, there are two independent gauge
transformation. So the current derivative has to involve all of them. Q1 in general, there is a G1A
mu plus IQ2G2A mu 12, not using, choosing very illuminating notation, but this is what you do.
The LaGrangian is, again that. You can open this up. And do our usual thing. You can open this
up and look at the masses and so on, look at the mass of gauge bosons and so on. But you already
know what is going to happen in the end. Qualitatively, what is going to happen in the end, you
can work out the details, or you will work out the details, is that there are two degree of freedom
here. Okay? Only one of them shift under the gauge symmetry. After, again after SSB, so after
SSB, only one Goldstone, would be Goldstone, only one would be Goldstone. Okay? Therefore,
you know that what is going to happen is that in the end, there is going to be one unbroken, one
massless vector. Okay? It is going to be one massive vector, plus one massless vector. Because
there is only one Goldstone, it can only supply one of the vectors with its longitudinal mode. Just
by counting the degree of freedom, so, yeah. But this massive one is not any one of these. This
massless one is not any one of this. They are linear combinations. You have to work out the linear
combination. Basically, all of them are just linear combinations of A1 and A2. Okay? That you
have to, you open this up, square, there is a mass matrix, Again this is similar to what you do with
couple harmonic oscillator. You have to diagrammize the frequency mate I can to find the Eigen
mode. There are two Eigen mode. One of them is massive vector. You know there can only be
one because there is only one Goldstone to be eaten. Is this okay? Both of these two situations,
these two situations happen, this happens in standard model. A similar story happen here in the
standard model, similar story to that happens in EMSSM. Lets also talk about the Carol Fermions.
Do I want to talk about Fermions here or later? Lets talk about it here. Lets remember what
the Fermion mass is. Fermion mass is something like that. But we know that, you still remember
Fermion part at the beginning of this quarter, we can write it as a, left-handed or right-handed

88

11

MARCH 7TH, 2013

component. Okay?


=

L
R




= R
, L

= m(L R + R
L )

So that, okay? That is that, okay? That is just that. Fermion mass is mass. This okay? Im
multiplying them out. Okay? Lorentz invariance, this basically is fixed by Lorentz invariants.
Lorentz invariants basically is saying that left-handed and right-handed for massive Fermions are
not good quantum numbers, because they mix. That is basically what Lorentz invariants is telling
us for massive Fermions. On the other hand, there could be situations that those two charged under
different gauge symmetry. That would carry different charges under gauge symmetry. Okay?
Under U (1) Under U1, under U1, there can be two situations, okay? One is that Q left is the same
as Q right. The left hand Fermion and right-handed Fermion has the same charge under this U1.
So, ah, I should actually say something slightly different. Now, lets look at the kinetic terms. Lets
look at the kinetic terms. The kinetic terms looks like this, looks like this.

( + iQR A )R
( + iQL A )L + R
L

This is 2 by 2 poly matrices basically Pauli. It doesnt matter for our argument but these are
kinetic term. This is gauge invariant, kinetic terms. Gauge invariants does not tell me anything
about QL versus QR, okay, of the kinetic terms. There is no limit. I can just set this whatever I
want. There are two situations. Case one is QL equals QR. This is called a vector like, Fermions. It
doesnt mean the Fermion becomes a vector. It just means that the left hand and right hand have
the same charge. This means that it is invariant. That just means that this term is invariant, if QL
is the same as QR, and the charge of this, rotation of this is opposite of the rotation of that.
invariant
i) QL = QR vector-like m
 not invariant
m

ii) QL 6= QR chiral 
This is gauge invariant. Then there is another situation with QL not the same as QR. Okay? In
this case, this is called the kierl, the Fermion is called the curial. In that case, it is not invariant.
That cannot exist. A term like that cannot exi For kierl Fermion does not have mass basically,
that is the statement. Just because you can assign left-handed and right hand Fermions, different
charges. Lets consider an extreme case.
89

11.1 Standard Model


Consider QL = 1, QR = 0

11

MARCH 7TH, 2013

y(L R ) + y R
L )

hi = v
yv (L R ) + h.c.
|{z}
m

= m L R + h.c.
QL is 1, QR is 0. That is one of those cases. There is nothing, you cannot write a mass term
basically. Okay? Look at this mass term, only QL carries charge. U1 rotation is certainly not
gaiblg gauge invariant. So now you wonder is there any way to give those Fermion masses,
because after all we live in a world with Fermion masses, where Fermion has masses. Maybe you
say, okay, maybe the Fermion world is all vector like, but it is not. Unfortunately its not. We
will talk about it too. Suppose I give you a pair of Fermions, how do you give them mass? The
trick is again there is a spontaneous symmetry break. It is going through spontaneous symmetry
break.
Consider , Q = 1, hi = v.
= (v + h) yhL R + h.c. =

11.1

m
hL R
v

Standard Model

So, are you all, you have all seen standard model at some point? Particles and interactions? Not
see it but know about it. Standard model, so the story we just said about U1 is important for
understanding the most intriguing part of, not the most intriguing but at least the most, one of
the most interesting part of the standard model. Standard model are controlled by gauge similar
trees symmetries.
gauge symmetry
SU (3)C

SU (3)C SU (2)L U (1)Y


strong interactions 8 gluons

quarks are 3: u,d,c,s,t,b

SU (2)L U (1)Y
SU (2)L

doublet

u
d L

90

which mirrors UR , dR singlet

11.1 Standard Model


doublet

11

scalar: H =

h
H0

SSB

h
v + h + ia


Y =

MARCH 7TH, 2013

1
2

SU (2)L U (1)Y
which gives 3 massive particles W , Z and a massless photon. We have 3-Goldstones
yu H QL uR + yD H c QL dR
|{z}
(udLL )
actually which way is the left-handed. Lets call this left-handed. Lets call this right-handed. Okay.
They carry different charge, if you want, under SU2 gauge interaction in the most extreme way.
One of them couples to them, one of them dont. Okay? But this too also interchange each other
under parity. Okay? So means that if you imagine you put a mirror here, this goes to that. Okay?
So if you put a mirror here, this just goes to that. Okay? Gauge interaction SU2L violate parity, so
this is parity violation. Is there anything wrong, violating parity? Nothing. It can be violated. It
is violated in nature. Nothing, nothing goes horribly wrong. For a long time we thought it would
be, parity is obviously a symmetry but its not. But again very similar to what I said here, very
similar to what I said here. You see the statement I just made is not completely consistent with
Lorentz invariants, of massive Fermions at least, because we know from experiment these things
gets, all have mass. By the way this is called opt because usually we write it on top, I dont know,
top down, the term strange is a little bit, I dont know why, maybe strange that they dont think
it should be there, and charm is just try to make a good name. I dont know. But this is not
consistent with Lorentz invariants because for massive Fermions, so there is a very intuitive way of
understanding why mass makes two polarizations, because for massive particle there is a rest frame.
Once you get into the rest room you lost a direction, there is no direction. In the rest room you can
pick any access of your polarization vector, and obviously left-handed can become right-handed.
Right-handed can become left-handed. Its your choice. In the massless case it makes sense because
there is no rest room. You are always moving in that direction. You always use that direction as
your reference, polarization. For massless Fermions there is a direction. We also know that this
symmetry is broken. The good thing is, so at this point in order to accommodate massive Fermions
you need to do one of those two things. Either break the symmetry or break Lorentz symmetry.
Lorentz symmetry we are not breaking. We are not breaking this one. But we know this is broken.
91

11.1 Standard Model


11 MARCH 7TH, 2013
The gauge bosons are massive. Its broken by a doublet. It can be broken by a doublet, scalar
doublet. This is again a scalar. H is a have to write it like this, H0. Just in a very suggestive
form, and this under spontaneous symmetry breaking, you can write this as an H plus, V plus H
plus IA. So that is how you do standard model doublet. And, okay. On the other hand, the H also
charges under this U1. So we havent talked about U1 yet. H also is charged under this U1. So YH
is 1/2, according to normalization, that is the usual normalization we use. So thats that. So this H
when it gets verve here, first of all you may ask me why am I giving vev to this guy, why am I doing
this, why am I not giving to it that guy or whatever it is? (vev?) The answer is it doesnt matter.
You can put it wherever you want, you can give vev to every component. It is similar to the U1
case, the valve can go any direction and the end result is the same. All those are vacuum, all those
vacuum are equivalent as we have said, along the circle. Similarly, you can take this to where SU2
rotation, give vev to anywhere you want, but all those vacuums are completely equivalent. Very
similar to the U1 case, okay? This breaks both SU2, cross U1Y. It breaks both of them. Okay?
On the other hand, there are only well, the global symmetry breaking is SU2 down to nothing.
There are only three Goldstones which is this and this. Its again very similar to the U1 case. Its
not everything in the scalar, it can be a Goldstone. Okay? In this case, these and that can, gold
stones. These and that. This is a complex scalar so there are two degree of freedom. Between two
of them there is one of them. There are only three Goldstones and how many generators are here?
How many gauge bosons are here? SU3 has one, how many, SU2 has how many? Three. Its SO(3)
basically, just three rotations. There are four generators. Three Goldstones. Okay? There is one
remain massless. Very similar to what I just erased, similar to what I just erased. Okay? So what
is going to happen is going to be three massive gauge bosons, and they all have names. They are W
plus/minus and the Z. And one massless. So because, again, which one is, what is the massless one
here? Its the photon. That is the symmetry breaking. Again, but again, its not exactly this one
doesnt become the photon, and this one is not exactly these three, and it is a linear combination.
In particular, Z and the photon are linear combination of the tau 3 here, and this U1 basically.
But lets not work it out. It is going to take us too long. But so yeah, these are, Im mentioning
this because this is continuous, it is a continuous story that we have talked about. You see there
is nothing to it. All you have to do is write down the current derivative and diegriz it, basically.
(?) With this H obviously we can write masses now for the Fermions. With this H I can start

92

11.1 Standard Model


11 MARCH 7TH, 2013
to write YHQLUR, call this YU plus YD, H, conjugate, QLDR. Doesnt quite matter what this
is. But you see these are the mass terms. Okay? This is, again, this is just my shorthand way
of writing down ULDL. These are the mass terms of the standard model quarks basically. They
have different Yukawa couplings in general. They dont have the same. We all know that there are
more than one quarks. Right? In general, these are 3 by 3 matrices. They are three quarks, three
flavor of quarks. They are [inaudible] too. To understand that, that is called a flavor problem. We
dont know really how to derive these Yukawa matrices from first principles, and its unlikely we
are going to know any time soon. Okay. That is basically the standard model. Everything what
I said has already been discovered, by the way. Im not making any of this story up. Lets take a
step back. Since we have been talking about Higgs mechanism and so on, lets take a step back,
lets try to understand Higgs mechanism slightly better, what does it really mean. For that lets go
back to the U1 story. Lets from a bottom up point of view, lets take a bottom up point of view.
Lets suppose Im very dumb. I dont know about lets just say I observe a massive gauge boson.
Okay?
m2
1
F F + A A A
4
2

unitary gauge

Ill write a mass term like that. Okay? You say wait, that is all wrong, right? It is not gauge
symmetry. We all know we need gauge symmetry, blah blah blah, but actually we dont. First of
all we dont. Lets say we do, but the usual story tells you that now gauge symmetry forbid this
term. That is why I need to introduce, that is why I need to do this whole Higgs boson mechanism
that write a cowritten derivative, give you a mass, that part of the story is wrong. The reason is
that, the reason this is wrong is that this is a secret to the gauge symmetric.
in general gauge Okay? Its not gauge symmetric when you fix to a unitary gauge which is not
such a big surprise. But suppose lets work in the slightly more general gauge. In general, in a
general gauge I can write this as this following form. You see this is completely gauge invariant.
Remember in gauge transformation A mu goes to A mu D mu, alpha we already said there is no
need to realize Pi, Pi shift as of this. Under this gauge transformation this is completely invariant,
complete gauge invariant.
m2 
2
1
F F + A A +
4
2
v

93

11.1 Standard Model


with gauge transformation

11

A A + ,

MARCH 7TH, 2013

On the other hand we know these two are completely [inaudible] to each other, there is nothing
different. Using gauge transformation I can gauge only Pi it become this, become unitary gauge.
Therefore, this is actually gauge invariant. There is nothing wrong with that. Gauge invariance
does not prevent you to write down a mass term. It is completely fine. The only thing you have to
remember is now there are three degree of freedom. But that is the real physical statement, between
massive and massless gauge boaston is not whether its gauge symmetry or not. Its whether there
is three degree of freedom or not. That what Im trying to get at. This is a real physical statement
is that. Goldstone is everything. The eaten Goldstone captures the essence of the physics really.
This also tells you that the LaGrangian sometimes can be very deceiving. Although we are getting
really used to doing everything with LaGrangian, but LaGrangian sometimes can tell you, teach
you a very wrong lessons, because it can hide things. You can hide things with LaGrangian.
You can do horribly wrong things with LaGrangian. If something looks horribly wrong, it can be
completely fine. Okay? So for that, lets just digress. Let me give you an example of something
that looks horribly wrong but is completely fine. Lets actually take this example. Lets take
this one. If I square it out, it goes like, Im going to be using shorthand, but I think its pretty
understandable what these things are. Its A square plus 2 Pi, 2D Pi, A plus, Im actually setting,
by the way, I think Im not being careful with gauge couplings, Im setting gauge coupling to one
in this case.

1 2 1 2 2
F mA A + 2A + ()2
| {z }
4
2
2

But general lesson is still there. This part of the story is a little bit schematic. That, okay. So this
is also known as a negative Pi box Pi after integrated by part. Now the next thing you observe
is that this is a quadratic in Pi. This is a quadratic in Pi. Nothing goes beyond that. When
something is a quadratic in a field, we dont even need to do preservation theory. We can just do
exactly its a gaussian, if its quadratic its gaussian. I can integrate this out basically. I can
perform D Pi exactly. Integrate out. In fact, I dont need to do that. I can use the equation motion

94

11.1 Standard Model


11 MARCH 7TH, 2013
of Pi to integrate up, because something quadratic, there is nothing wrong with it.


m2A
1
F
L= F 1+
4
2
Now I can integrate, I can substitute, integrate means I substitute everything I see a box Pi, I
substitute with DA. Okay? With everything, as every time we see Pi, I substitute so I make this
the following substitution. Probably having a heart attack with this one box. But thats okay.
The thing we end up with is the following. Again, Im not being careful with the factors. That
is the LaGrangian. So if I write it this LaGrangian to begin with you will kill me, because its
like, what is this? This is not local. But we know secretly its just that. That is completely fine.
Anyway, its not a very important for our course. But this is just an example of, LaGrangian can
fool you. If you are writing down this LaGrangian you think you are doing cutting edge research
in particle physics, you are not. You are just doing that. Okay? Okay. Now, this does bring up
an interesting point. But because now I completely trashed gauge symmetry, I told you early on
it is not symmetry and so on, but you see that I can make anything that doesnt look like gauge
symmetric gauge symmetry. Anything is gauge symmetric. Just have to replace A, A, replace with
that and Im done. The whole thing is gauge symmetry. Gauge symmetrical. Now you say now
we dont have roots anymore. I can write down everything I want, anything I want to write. Now
I can write down anything I want to, so all hell breaks through. Again, Im coming back to this
LaGrangian. I already told you that this term is fine. Okay?
m2
1
L = F 2 + A A2 + a(A)2 + bA4
4
2
Now you ask why stop here. Im going to keep going. Im going to add this one. Nothing wrong.
Because you do add something like this as a gauge fixing but this is not gauge fixing, Im not even
doing that at this moment. This is a physical coupling Im adding here. Okay? I can do this as well.
You dont have dimensional terms but maybe those things are suppressed by some energy scalar.
But these are renormalizable looking. Nothing quite wrong. This is the case, where actually, we
are keeping track with the Pis are very important. You keep track Pis are wrong actually teach
you a lot, okay is this so lets try to do that. You notice that the deep, my rule is to replace every
A with that.
A = (A + ) = A + 2


(A)2 = a (A)2 + (2)2 +
95

11.1 Standard Model


11 MARCH 7TH, 2013
Every single A with that, okay? Im going to do that. Okay? Im going to do that. Again
schematically, okay? Lets imagine I redefine the Pis with V inside of Pi now, Im just not going to
keep all these things around, but you can put all the factors in and the story is the same. Okay?
This one is DA plus box Pi. You say whats wrong with that? You can put it back in. And so DA
square is A, DA square plus box Pi square plus ... okay? Now, but also A square contribute a term
like this. So now lets put these two terms together, and A square, okay? You say what is wrong
with this story? Its a little bit wrong.

m2A
1
2 + a(2)2 2 2
2
p (p m02 )

Because, well, lets go do it. We know that this is completely fine. This is just a kinetic term for
the Goldstone. Goldstone needs to have kinetic terms. Anything that can legitimately be called
a Cal man field needs a kinetic term is fine but this is not fine with these two. With these two
now you can solve it. If I use this to write down my propagator, this is going to be a P square in
the propagator there is going to be a P to the force in the propagator. Therefore, the propagator
will have the form of P square, P square minus P prime square, something like that. That is my
propagator. There is a pole in P square equals 0, that is fine. Thats Goldstone. But there is
another pole. Okay. Once you have P to the fourth there is another pole. There is another pole
on the order of MA square over A. This is certainly unphysical. There is something wrong with
this theory, a theory just like that. Because you are introducing a spurious poles, this will be,
okay?
pole

m2A
unphysical
a

a must be small

To make this theory healthy, A must be, be small so this pole is above your cutoff. You can put this
in. But there is a limit, how big this A is. This kind of story tells you that Pi teach us a lot. Keep
the Pis around, they really teach you a lot about what is allowed to have. Not gauge symmetry
itself. But you have the Pis around, tells you how healthy a theory is. Okay? So lets do SU2
version of it, of Pi, SU2 version. Lets do SU2 version. Imagine everything here SU2, lets imagine
all these things are SU2s. Lets define a few things. A will be sigma AA. This is SU2 generators.
Pauli matrices. The appropriate quantity now, we want to have is, another is the, instead of E to
the I Pi, we will have E to the I sigma A Pi A over V, which I will call E to the I Pi over V, but

96

11.1 Standard Model


the Pi is sigma A Pi A. This is the generators, SU2 generators.
A = a Aa

U = ei

a a
v

= ei v

11

= aa





()2 2
2
2
L = v trace |U | + trace () +
+
v2
2

97

MARCH 7TH, 2013

You might also like