Quantum Field Theory

Quantum Field Theory
1
R. Clarkson Dr. D. G. C. McKeon
2
January 13, 2003
1
Notes taken by R. Clarkson for Dr. McKeons Field Theory (Parts I and II) Class.
2
email: dgmckeo2@uwo.ca
2
Contents
1 Constraint Formalism 7
1.1 Principle of Least action: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.2 Hamiltons Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3 Poisson Brackets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.4 Diracs Theory of Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.5 Quantizing a system with constraints . . . . . . . . . . . . . . . . . . . . . . 24
2 Grassmann Variables 27
2.1 Integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.2 Poisson Bracket . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.3 Quantization of the spinning particle. . . . . . . . . . . . . . . . . . . . . . . 32
2.4 General Solution to the free Dirac Equation . . . . . . . . . . . . . . . . . . 50
2.5 Charge Conjugation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
2.6 Majorana Spinors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
2.7 Time Reversal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3 Bargmann-Wigner Equations 61
4 Gauge Symmetry and massless spin one particles 71
4.1 Canonical Hamiltonian Density . . . . . . . . . . . . . . . . . . . . . . . . . 74
5 (2
nd
) Quantization, Spin and Statistics 83
5.1 Harmonic Oscillator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.2 Feynman Propagator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
5.3 Quantizing the Dirac Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
6 Interacting Fields 97
6.1 Gauge Interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
6.2 Heisenberg Picture of Q.M. . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
6.3 Wicks Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
7 Electron-Positron Scattering 105
3
4 CONTENTS
8 Loop Diagrams 109
8.1 Feynman Rules in Momentum Space . . . . . . . . . . . . . . . . . . . . . . 109
8.2 Combinatoric Factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
8.3 Cross Sections From Matrix elements . . . . . . . . . . . . . . . . . . . . . . 115
8.4 Higher order corrections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
8.5 Renormalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
8.6 Regularization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
8.7 Noethers Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
9 Path Integral Quantization 141
9.1 Heisenberg-Dirac . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
9.2 Wave Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
9.3 Free Particle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
9.4 Feynman Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
9.5 Path Integrals for Fermion Fields . . . . . . . . . . . . . . . . . . . . . . . . 160
9.6 Integration over Grassmann Variables . . . . . . . . . . . . . . . . . . . . . . 160
9.7 Gauge Invariance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
10 Quantizing Gauge Theories 169
10.1 Quantum Mechanical Path Integral . . . . . . . . . . . . . . . . . . . . . . . 169
10.2 Gauge Theory Quantization . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
10.3 Feynman Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
10.4 Radiative Corrections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
10.5 Divergences at Higher orders . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
10.5.1 Weinbergs Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
10.6 Renormalization Group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
10.6.1 Eulers Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
10.6.2 Explicit Calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
11 Spontaneous Symmetry Breaking 207
11.1 O(2) Goldstone model: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
11.2 Coleman Weinberg Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . 215
11.3 One loop Eective Potential in
4
model . . . . . . . . . . . . . . . . . . . 216
11.4 Dimensional Regularization . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
11.5 Spontaneous Symmetry Breaking in Gauge Theories . . . . . . . . . . . . . . 225
12 Ward-Takhashi-Slavnov-Taylor Identities 229
12.1 Dimensional Regularization with Spinors . . . . . . . . . . . . . . . . . . . . 234
12.1.1 Spinor Self-Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
12.2 Yang-Mills Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
12.3 BRST Identities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
12.4 Background Field Quantization . . . . . . . . . . . . . . . . . . . . . . . . . 242
CONTENTS 5
13 Anomalies 249
13.1 Anomalies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
14 Instantons 257
14.1 Quantum Mechanical Example . . . . . . . . . . . . . . . . . . . . . . . . . . 257
14.2 Classical Solutions to SU(2) YM eld equations in Euclidean Space . . . . . 261
6 CONTENTS
Chapter 1
Constraint Formalism
1.1 Principle of Least action:
Conguration space:
(q
1
, q
2
, . . . , q
n
)
with q
(i)
(t
i
) & q
(f)
(t
f
) xed, the classical path is the one that minimizes
S
....
action
=
t
f
t
i
L(q
i
(t), q
i
(t))dt (L = Lagrangian) (1.1.1)
i.e. if q
i
(t) = q
classical
i
(t) + q(t).
As S = S() has a minimum at = 0,
dS(0)
d
= 0 (1.1.2)
q(t
i
) = 0
q(t
f
) = 0
7
8 CHAPTER 1. CONSTRAINT FORMALISM
dS(0)
d
= 0
=
t
f
t
i
dt
L
q
i
q
i
+
L
q
i
q
i
Int. by parts
=
t
f
t
i
dt
L
q
i
d
dt
L
q
i
q
i
and, as q
i
is arbitrary,
L
q
i
=
d
dt
L
q
i
(1.1.3)
1.2 Hamiltons Equations
(Legendre transforms)
p
i
=
L
q
i
(1.2.1)
H = p
i
q
i
L(q
i
, q
i
) (1.2.2)
Note that H does not explicitly depend on q
i
. i.e.
H
q
i
= p
i
L
q
i
= 0
Thus H = H(p
i
, q
i
).
Now:
dH =
H
q
i
....
A
dq
i
+
H
p
i
....
B
dp
i
(1.2.3)
= p
i
d q
i
....
cancels
+ q
i
dp
i
L
q
i
dq
i
L
q
i
d q
i
. .. .
cancels
but if we insert (1.2.1) (1.1.3), we get
L
q
i
=
d
dt
(p
i
) = p
i
,
dH = q
i
....
A
dp
i
p
i
....
B
dq
i
(1.2.4)
From (1.2.3) and (1.2.4), we have:
q
i
=
H
p
i
(1.2.5)
p
i
=
H
q
i
(1.2.6)
1.3. POISSON BRACKETS 9
1.3 Poisson Brackets
A = A(q
i
, p
i
) (1.3.1)
B = B(q
i
, p
i
) (1.3.2)
A, B
PB
=
A
q
i
B
p
i
A
p
i
B
q
i
(1.3.3)
So, we have
q
i
= q
i
, H (1.3.4)
p
i
= p
i
, H (1.3.5)
We can generalize this:
d
dt
A(q
i
(t), p
i
(t)) = A, H (1.3.6)
Suppose in dening
p
i
=
L
q
i
(1.3.7)
we cannot solve for q
i
in terms of p
i
. i.e. in
H = p
i
q
i
L(q
i
, q
i
)
ex: With S.H.O.
L =
m
2
q
2
k
2
q
2
p =
L
q
= m q
H = p q L
= p
p
m
m
2
p
m
kq
2
2
=
p
2
2m

kq
2
2
i.e. can solve for q in terms of p here.
Suppose we used the cartesian coords to dene system
L =
m
2
a
2
+
m
2
b
2
+
1
( a b) +
2
(
b +a)
i.e.
(a(t), b(t)) = (cos((t)), sin((t)))
( a(t),

b(t)) = (sin((t))

(t), cos((t))

(t))
(1.2.1)
ex.
a
2
+ b
2
= 1
scale

= const. = 1
a = b /

b = a
Dynamical Variables: a, b,
1
,
2
L
i
=
d
dt
L
i
= 0 =
a b i = 1
b + a i = 2
(
1
,
2
) Lagrangian Multipliers
The trouble comes when we try to pass to Hamiltonian;
p
i
=
L
q
i
p
1
= 0
p
2
= 0
Cannot solve for

i
in terms of p
i
, because these are 2 constraints
i.e. if
p
i
=
L
q
i
(1.3.8)
cannot be solved, then
p
i
q
j
=

2
L
q
i
q
j
cannot be inverted.
In terms of Lagranges equations:
d
dt
L
q
i
(q
i
, q
i
) =
L
q
i
=

2
L
q
i
q
j
q
j
+

2
L
q
i
q
j
q
j
So

2
L
q
i
q
j
. .. .
q
j
=

2
L
q
i
q
j
q
j
+
L
q
i
(1.3.9)
q
i
, q
i
specied at t = t
o
q
i
(t
0
+ t) = q
i
(t
0
)
. .. .
given
+ q
i
(t
0
)
. .. .
given
t +
1
2!
q
i
(t
0
) (t)
2
+
1
3!
d
3
dt
3
q
i
(t
0
)
(t)
3
+ . . .
Can solve for q
i
(t
0
) if we can invert

2
L
q
i
q
j
.
Thus, if a constraint occurs, q
i
(t
0
) cannot be determined from the initial conditions using
1.4. DIRACS THEORY OF CONSTRAINTS 11
Lagranges equations. i.e. from our example,
i
(t
0
+ t) =
i
(t
0
) + . . . +

i
....
(t
0
)
cannot be determined.
1.4 Diracs Theory of Constraints
If p
i
=
L
q
i
implies a constraint
i
(q
i
, p
i
) = 0 (i.e. from our example p
1
= p
2
= 0), then we
can dene
H
0
= p
i
q
i
L(q
i
, q
i
) Constraints hold here (1.4.1)
(So, for our example, (scaling m = 1),
L =
1
2
( a
2
+

b
2
) +
1
( a b) +
2
(
b + a)
p
1
= 0 = p
2
p
a
= a +
1
p
b
=

b +
2
H
0
= p
a
a + p
b
b +
=0
....
p
i
L
= p
a
(p
a
1
) + p
b
(p
b
2
)
1
2
(p
a
1
)
2
+ (p
b
2
)
2
1
[p
a
1
b]
2
[p
b
2
+ a])
The constraints must hold for all t, thus
d
dt
i
(q, p) = 0 (1.4.2)
=
i
, H
PB
0 zero if
i
= 0 (weakly equal to zero)
where H = H
0
+ c
i
i
.
Sept. 15/99
So, we have:
H
0
= p
i
q
i
L(q, q) (1.4.3)
H = H
0
+ c
i
i
(q
i
, p
i
) (1.4.4)
d
dt
i
= [
i
, H
0
+c
i
i
] 0
This consistency condition could lead
to some additional constraints
(1.4.5)
The constraints coming from the denition p
i
=
L
q
i
are called primary constraints.
Additional constraints are called secondary. We could in principle also have tertiary
constraints, etc.. (In practice, tertiary constraints dont arise).
Suppose we have constraints
i
. They can be divided into First class and Second class
constraints.
rst class constraints label
i
(1.4.6)
second class constraints label
i
(1.4.7)
For a rst class constraint,
i
, [
i
,
j
] 0 =
ij
k
x
k
(for all j).
i
is second class if it is not rst class.
We know that
H = H
0
+ c
i
i
= H
0
+ a
i
i
+ b
i
i
d
dt
i
= [
i
, H] 0
Thus
d
dt
i
= [
i
, H
0
+a
j
j
+ b
j
j
]
= [
i
, H
0
] + a
j
[
i
,
j
] +
j
[
i
, a
j
] + b
j
[
i
,
j
] +
j
[
i
, b
j
]
0
[
i
, H
0
]
(true for any a
i
, b
j
)
d
dt
i
= [
i
, H
0
] +
0
. .. .
a
j
[
i
,
j
] +
0
. .. .
j
[
i
, a
j
] +b
j
[
i
,
j
] +
j
....
0
[
i
, b
j
]
0
[
i
, H
0
] + b
j
[
i
,
j
]
this xes b
j
.
Note we have not xed a
i
.
Hence for each rst class constraint there is an arbitrariness in H
0
. To eliminate this ar-
bitrariness we impose extra conditions on the system. (These extra conditions are called
gauge conditions).
We call these gauge conditions
i
(one for each rst class constraint
i
).
Full set of constraints:
i
,
i
,
i
=
i
H = H
0
+ a
i
i
+b
i
i
+ c
i
i
(1.4.8)
Provided
i
,
j
0, then the condition
d
dt
i
0 (1.4.9)
xes a
i
, b
i
, c
i
(All arbitrariness is eliminated).
Dirac Brackets (designed to replace Poisson Brackets so as to eliminate all constraints
from the theory).
Note:
[
i
,
j
] 0 (Could be weakly zero for particular i, j but not in general (overall)).
d
ij
= [
i
,
j
] = [
j
,
i
](Antisymmetric matrix)
det(d
ij
) = 0
Thus i, j must be even. Hence there are always an even number of 2
nd
class constraints.
Now we dene the Dirac Bracket.
[A, B]
= [A, B]
i,j
[A,
i
]d
1
ij
[
j
, B] (1.4.10)
Properties of the Dirac Bracket
1.
[
i
, B]
= [
i
, B]
k,l
[
i
,
k
]
. .. .
d
ik
d
1
kl
. .. .
il
[
l
, B]
= [
i
, B] [
i
, B]
= 0 (1.4.11)
2. We know that
0 = [[A, B], C] + [[B, C], A] + [[C, A], B] (1.4.12)
We can show that
0 = [[A, B]
, C]
+ [[B, C]
, A]
+ [[C, A]
, B]
(1.4.13)
3. If A is some 1
st
class quantity, i.e. if [A,
i
] 0 for any constraint
i
, then
[A, B]
= [A, B]
ij
0
. .. .
[A,
i
] d
1
ij
[
j
, B]
= [A, B] (1.4.14)
Note that if
H = H
0
+a
i
i
+b
i
i
(1.4.15)
then H itself is rst class. Hence,
d(
dt
= [(, H] (1.4.16)
Thus by (3) above,
d(
dt
[(, H]
. (1.4.17)
But [
i
, (]
= 0 by (1). Thus, in H, we can set

i
= 0 before computing [(, H]
.
i.e.
If we want to nd
d(
dt
we can use [(, H]
and take H to be just H = H

0
+ a
i
i
+
=0
....
b
i
i
(Provided we exchange P.B. for Dirac B.).
Thus if we use the Dirac Bracket, we need not determine b
i
. If we include the gauge condition,
we can treat
i
=
i
,
i
,
i
(1.4.18)
as a large set of 2
nd
class constraints, and if
D
ij
=
i
,
j
(1.4.19)
then
[A, B]
= [A, B]
ij
[A,
i
]D
1
ij
[
j
, B] (1.4.20)
Sept. 17/99
So, so far:
L =
1
2
( a
2
+

b
2
a
2
b
2
) +
1
( a b) +
2
(
b + a)
p
a
=
L
a
= a +
1
p
b
=
L
b
=

b +
2
p
1
= p
2
= 0 Constraint
H
0
= p
i
q
i
L
= p
a
a + p
b
b +
0
. .. .
p
1
+
0
. .. .
p
1
2
( a
2
+

b
2
a
2
b
2
) +
1
( a b) +
2
(
b + a)
cant make any sense of this (cant express
i
in terms of p
i
) unless we impose constraints.
= p
a
(p
a
1
) + p
b
(p
b
2
)
1
2
((p
a
1
)
2
+ (p
b
2
)
2
a
2
b
2
) +
1
(p
a
1
b) +
2
(p
b
2
+ a)
=
(p
a
1
)
2
2
+
(p
b
2
)
2
2
+
1
2
(a
2
+ b
2
) +
1
b
2
a
dp
1
dt
= [p
1
, H] 0
p
a
1
b = 0
p
b
2
+ a = 0
Secondary constraints
(tertiary constraints dont arise)
1
= p
2
= p
3
= p
a
1
b
4
= p
b
2
+ a
These are all Second class - i.e. [
1
,
3
] = 1 = [
2
,
4
]
(No rst class constraints no gauge condition).
H = H
0
+ c
i
i
c
i
xed by the condition

i
= [
i
, H] 0
- or could move to Dirac Brackets, and let
i
= 0.
Need:
d
ij
=
0 0 1 0
0 0 0 1
1 0 0 2
0 1 2 0
= [
i
,
j
]
[X, Y ]
= [X, Y ] [X,
i
]d
1
ij
[
j
, Y ]
Can eliminate constraints sequentially instead of all at once (easier).
Eliminate
3
&
4
initially.
(1)
1
= p
a
1
b
(1)
2
= p
b
2
+ a
d
(1)
12
= [
(1)
1
,
(1)
2
]
= 2
d
(1)
ij
=
0 2
2 0
(d
(1)
ij
)
1
=
0
1
2
1
2
0
[X, Y ]
= [X, Y ] [X,
(1)
1
]
1/2
....
d
1
12
[
(1)
2
, Y ] [X,
(1)
2
]
1/2
....
d
1
21
[
(1)
1
, Y ]
Now we need to eliminate
(2)
1
= p
(2)
2
= p
2
d
(2)
ij
= [
(2)
i
,
(2)
j
]
(note *)
d
(2)
12
= 0 [p
1
, p
a
1
b]
1
2
[p
b
2
+a, p
2
]
= [p
1
,
1
]
1
2
[p
2
,
2
]
=
1
2
= d
(2)
21
d
(2)
ij
=
0
1
2
1
2
0
(d
(2)
ij
)
1
=
0 2
2 0
Finally,
[X, Y ]
= [X, Y ]
[X,
(2)
i
]
(d
(2)
ij
)
1
[
(2)
j
, Y ]
We can nally see that

[a, p
a
]
= 1
[b, p
b
]
= 1
and all other fundamental Dirac Brackets are zero. i.e.
[a, p
b
] = 0
as
p
a
1
b = 0
p
b
2
+ a = 0
and
H
0
=
b
2
2
+
(a)
2
2
+
a
2
+ b
2
2
+ (p
a
b)b (p
b
+ a)a
= p
a
b p
b
a
So,
H = p
a
b p
b
a
da
dt
= [a, H]
= [a, p
a
b p
b
a]
= b
db
dt
= [b, H]
= a
If we have gauge conditions & rst class constraints
i
:
H = H
0
+ a
i
i
+
0
....
b
i
i
1
st
stage: get rid of 2
nd
class constraints. Do this by
[ ] [ ]
At this stage,
H = H
0
+ a
i
i
As the a
i
s are not xed,
dA
dt
= [A, H]
[A, H
0
]
+ a
i
[A,
i
]
a
i
Arbitrariness
= 0 must intersect q
i
(t) at one & only one point.
Gribov Ambiguity (to be avoided).
Sept. 21/99
Relativistic Free Particle
S arc length from x
(
i
) to x
(
f
)
= m
dx
dx
d
The m = const. of proportionality g
= (+, , , ).
S = m
i
d
x
2
(1.4.21)
p
=
L
x
=
m x
x
2
(1.4.22)
Constraints
p
=
m
2
x
x
2
0 = p
2
m
2
(1.4.23)
H
0
= p
L
=
m x
x
2
(m
x
2
)
= 0 !
H = H
0
+U
i
i
= (p
2
m
2
) (Pure Constraint!)
x
=
H
p
dx
d
= (2p
) (1.4.24)
This arbitrariness in x
is a reection of the fat that in S, is a freely chosen parameter,

i.e.
S = m
dx
d
dx
d
(1.4.25)
(
t
) d =
d
d
t
d
t
dx
d
=
dx
d
t
d
t
d
(1.4.26)
S = m
d
t
dx
d
t
dx
d
t
(1.4.27)
Now let =
1
2
d
d
. Thus, (insert into (1.4.24))
dx
d
=
d
t
d
p
and equate this with (1.4.26) (1.4.28)

dx
d
t
= p
(1.4.29)
Gauge xing in this case corresponds to a choice of the parameter .
-The formalism of Dirac actually breaks down for gauge choices which are dependent
on time (which in this case means on ).
(Note, we can think of this reparameterization invariance (
t
) as being a form of
dieomorphism invariance in 0 + 1 dimensions, i.e. x
() is a scalar eld moving in 0 + 1

dimensions, and has a so-called tangent space which is four dimensional.
Thus, this is a simpler version of G.R. where we have scalars
t
(x
) moving in 3 + 1
dimensions with the dieomorphism invariance x
(x
t
). Techniques in G.R. & in the
single particle case often overlap.
eq. of motion:
L = m
x
2
0 =
d
d
L
x

L
x
Thus,
d
d
m x
x
2
= 0
m x
x
2
= 0
Now identify with the arc length along the particles trajectory:
ds
2
= dx
dx
If d
2
= ds
2
, then
dx
d
dx
d
= 1 (1.4.30)
x
2
= 1 (1.4.31)
( is called the proper time in this instance).
i.e.
if dx = 0, ds
2
= dt
2
= d
2
.
In this case, the equation of motion becomes
m x
= 0
The corresponding action is
S =
m
2
i
d x
2
absence of

means this is not invariant under (
t
).
d
dt
L
x

L
x
= m x
= 0
H = p
L
where p
=
L
x
= m x
H = p
m
2
2
H =
p
2m
Other gauge choice:
= x
4
= t(breaks Lorentz invariance)
Work directly from the action:
S = m
dt
d
dt
d

dr
d
dr
d
If weve chosen = t, then
S = m
dt
1 v
2
; v =
dr
dt
eq. of motion:
d
dt
L
L
r
= 0
d
dt
mv
1 v
2
= 0
p =
L
r
=
mv
1 v
2
(momentum)
H = p

r L
= p v L
but p
2
=
m
2
v
2
1v
2
p
2
(1 v
2
) = m
2
v
2
p
2
p
2
v
2
= m
2
v
2
p
2
= v
2
( p
2
+ m
2
)
v
2
=
p
2
m
2
+ p
2
1 v
2
= 1
p
2
m
2
+ p
2
=
m
2
+ p
2
p
2
m
2
+ p
2
1 v
2
=
m
2
p
2
+ m
2
Thus,
v =
p
m
2
+ p
2
H = p
p
m
2
+ p
2
(m)
m
m
2
+ p
2
=
p
2
+ m
2
E = numerical value of H =
m
1 v
2
=
mc
2
1
v
2
c
2
Note:
E
2
= p
2
+ m
2
(E
2
p
2
) m
2
= 0
p
= ( p, E)
p
m
2
= 0
Limit m 0
S = m
x
2
0 ??
i.e.
d
dt
mv
1 v
2
0 ???
[v[ 1 as m 0.
Circumvent by introducing a Lagrange multiplier e.
S =
1
2

x
2
e
+ m
2
e
d
e = e() , x
= x
()
m
2
0 is well dened in S. As
d
d
L
e
....
=0
L
e
= 0
So,
x
2
e
2
+ m
2
= 0 e =
x
2
m
2
Thus,
S =
1
2
m x
2
x
2
+m
2
x
2
m
2
= m
x
2
Can see that e eld can be eliminated really just a Lagrange multiplier that insures
S = 0 when m = 0.
Note:
The action
S =
1
2
x
2
e
+ m
2
e
(1.4.32)
is invariant under
+ f() (1.4.33)
x
= x
f() (1.4.34)
e =

fe + ef (1.4.35)
Sept. 22/99
So, for the above system:
L =
1
2
x
2
e
+ m
2
e
eqns of motion:
0 =
x
2
e
2
+m
2
from
d
d
L
e

L
e
= 0
0 =
d
d
e

from
d
d
L
x

L
x
= 0
=
L
x
=
x
e
(no constraint can solve for x
in terms of p
).
p
e
=
L
e
= 0 (primary constraint)
H
0
= p
+
=0
....
p
e
e L
= p
(p
e)
1
2
x
2
e
+ m
2
e
; x
2
= x
= p
2
e
2
=
1
2
e(p
2
m
2
)
p
e
= 0 = [p
e
, H
0
]
p
e
,
1
2
e(p
2
m
2
)
0 =
1
2
(p
2
m
2
) (Secondary Constraint)
Both p
e
= 0 (gauge condition e = 1) and p
2
m
2
= 0 (already discussed) are rst calss.
Note: Remember that
S =
d
4
x

g g
A
(x)

A
(x)
(1.4.36)
action for a scalar eld
A
(x) in 3+1 Dim.
Vierbein (deals with 4-d)
g
= e
a
e
a
g =
det(g
) = [det(e
)]
1
= e
1
In 0 + 1 dimensions
S =
d
1
e
d
d
d
d
d
(

A
)
2
e

d
( x
)
2
e
e Einbein (assoc. with 1 dim)
1.5 Quantizing a system with constraints
[A, B]
PB

1
i
[

A,

B]
commutator (c)
(1.5.1)
ex.
[q, p]
PB
= 1 (1.5.2)
[ q, p]
c
= i (1.5.3)
If there are constraints
i
(q, p) then,
i
( q, p) [ >
phys.
= 0 (1.5.4)
ex. For L =
1
2
x
2
e
+ m
2
e
1
= p
e

2
= p
2
m
2
(2 constraints)
Quantization conditions will be
[x
, p
] = i
(1.5.5)
(+, +, +, +)
re: p
=
L
x
Contrav. = der. of covar.
(p
2
m
2
) [ >
phys
= 0 (Klein-Gordon eq.) (1.5.6)
p
= i

x
(1.5.7)
1.5. QUANTIZING A SYSTEM WITH CONSTRAINTS 25
If (x) =< x[ >
phys
then
i

x
2
m
2
(x) = 0
Classical Motivation for Spin
Brint, deVecchia & How, Nuclear P. 118, pg. 76 (1977)
Chapter 2
Grassmann Variables
(Casa/booni, N.C. 33A)
1
,
2
Grassmann Variables (call if only 1 present), where;
2
=
2
1
(2.0.1)
2
1
=
2
1
= 0 (2.0.2)
ex. (these can only be . . . )
f(x
1
, ) = a(x) + b(x) (2.0.3)
f(x,
1
,
2
) = a(x) + b
i
(x)
i
+
1
2
c(x)
ij
j
(2.0.4)
(
ij
=
ji
)
Calculus
d
d
= 1
d
d
= 0( = const.) (2.0.5)
d
d
1
(
1
2
) =
=1
....
d
1
d
2
2
+
1
=0
. .. .
d
2
d
1
(-ve sign in 2
nd
term because were moving
d
d
1
through to
2
)
=
2
(2.0.6)
Similarly
d
d
2
(
1
2
) =
d
1
d
2
2
+
1
d
2
d
2
=
1
(2.0.7)
27
28 CHAPTER 2. GRASSMANN VARIABLES
2.1 Integration
d
d
d
i.e. Integration & dierentiation are identical (2.1.1)
i.e.
d c = 0 (int. of a constant c ) (2.1.2)
d = 1 (2.1.3)
For example;
d
1
d
2
(
1
2
) =
d
d
1
d
d
2
=
d
d
1
(
1
)
= 1
=
d
2
d
1

1
2
=
d
1
d
2
1
Another example:
d
1
d
2
f(x,
1
,
2
) =
d
1
d
2
a
....
(i)
+ b
i
i
....
(ii)
+
1
2
ij
c
i
(i) 0 (const.) (ii) 0 (

i
int. over
j
gives 0)
= c
d
1
d
2
1
2
1
2
= c(x)
Delta function:
d ()f(x, ) = f(x, 0)
d () [a(x) + b(x)] = a(x)

=() = (2.1.4)
2.1. INTEGRATION 29
The following is an example to demonstrate the properties of dierent kinds of statistical
models. Suppose there are three students:
(T) Tom
(D) Dick
(H) Harry
How many ways can two prizes be awarded to T, D, H?
1. Suppose there are two medals, (distinguishable awards) a Newton medal (N) and a
Shakespear medal (S).
T D H
N S
S N
N S
T D H
S N
N S
S N
T D H
NS
NS
NS
So, there are 9 dierent ways to award 2 distinguishable medals.
2. Two silver dollars (2 medals, indistinguishable)
T D H
$ $
$ $
$ $
T D H
$ $
$ $
$ $
So, there are 6 ways to award 2 indistinguishable medals.
3. Two positions (P) on football team (2 medals, indistinguishable). But! now makes
no sense for one person to receive 2 medals (one player cant have 2 positions).
T D H
P P
P P
P P
There are 3 ways to award to indistinguishable yet distinct medals.
Now call the
prizes particles
students states
1. = Maxwell-Bolzmann statistics
2. = Bose-Einstein
3. = Fermi-Dirac
Sept. 24/99
Returning to our discussion of Grassmann variables, the Lagrangian is now
L = L
(),
a
(), x
(),

a
()
(2.1.5)
where
0 =
a
()
b
(
t
) +
b
(
t
)
a
() (2.1.6)
p
=
L
x
and
a
=
L
a
(2.1.7)
H = H(x
,
a
, p
,
a
) = q
a
....
L (2.1.8)
* - order important
0
d L =
0
d
L
x
+
L
a
+ x
L
x
. .. .
int. by parts
+

a
L
a
. .. .
int. by parts
d
d
L
x
=
L
x
,
d
d
L
a
=
L
a
=
a
H(q, p, , )
q p + q p +

+

q
H
q
p
H
P

H
int. by parts:
=
p q + q p

+

q
H
q
p
H
p

H
And so, for this to be zero, we must have:

p
=
H
q
a
=
H
a
(2.1.9)
q
= +
H
p
a
=
H
a
(2.1.10)
2.2 Poisson Bracket
A(q, p, , ), B(q, p, , )
Note:
Anything Grassmann is odd
Anything else is even
For the order of two quantities:
2.2. POISSON BRACKET 31
Even/Even (doesnt matter)
Even/Odd (doesnt matter)
Odd/Odd Switch the two, you pick up a -ve sign.
[A, B]
PB
=
A
q
B
p

A
p
B
q
+
A
1. A,B even
A
q
B
p

A
p
B
q
+
A
+
A
A
q
B
p
+
B
q
A
p

A
2. A,B Odd
A
q
B
p

B
q
A
p
+
A
+
B
3. A Even, B Odd
With this, we nd that (Not trivial!):
0 = [[A, B] , C] + [[B, C] , A] + [[C, A] , B] (2.2.1)
Look at the spinning particle (Not superparticle).
Simplest action involving Grassmann Variables.
(Brink et. al. NP B118)
L = L(
(),
()) (
(), Grassmann)
=
1
2
e
i
()

()
cf L =
1
2
x
2
2
+ m
2
e
(2.2.2)
Note:
1.

= 0 (why we cant have two

in L).
2. i needed so that L = L
+
i.e. L = (i

) (Re: (AB)
+
= B
+
A
+
)
L
+
= (+i)(

+
) = i

= i

= L
3. The following
+ f() (2.2.3)
=

f (2.2.4)
e =

fe + ef (2.2.5)
f (2.2.6)
is an invariance of S =
d L (reparametrization invariance)
Quantizing this leads to negative norm states in the Hilbert space associated with
0
().
<
0
[
0
> < 0 (Dont want.)
(ex L =
1
2
= (A
0
)
2

A
2
(A
0
= -ve norm state)
We eliminate the unwanted negative norm state by building in an extra symmetry.

L =
1
2
2
e
i
i
e
; () is Grassmann (2.2.7)
1. Add in =

f + f
2. We have another invariance ( = () is Grassmann)
= i
(2.2.8)
e

i
2e
(2.2.9)
e = i (2.2.10)
= 2 (2.2.11)
L =
d
d
(...) He didnt recall (2.2.12)
- Scalar in 0+1 dimensions.

- Spinor in 0+1 dimensions.

e - einbein in 0+1 dimensions.
- gravitino in 0+1 dimensions.
L - analogous to SUGRA L in 3+1 dimensions.
2.3 Quantization of the spinning particle.
=
L
i
2
(2.3.1)
=
i
2
Second Class constraint (2.3.2)
i
2
= 0 (2.3.3)
[
]
PB
= 0 (2.3.4)
2.3. QUANTIZATION OF THE SPINNING PARTICLE. 33
Sept. 28/99
We could proceed with the Dirac formalism. We could also cheat and use the equations of
motion. Recall:
L =
1
2
e
i
i
e
(2.3.5)
= i
(2.3.6)
e

i
2e
(2.3.7)
e = i (2.3.8)
= 2 (2.3.9)
Eq. of motion for e:
d
d
L
e
=
L
e
(2.3.10)
0 =

2
i

(2.3.11)
For :
= 0 eliminates -ve norm states If

0
= v, v
= 0 (2.3.12)
We could have gotten these two equations from the Hamiltonian formalism because these
are the constraint equations that follow from the primary constraints
p
e
=
L
e
= 0 (2.3.13)
p
=
L

= 0 (2.3.14)
These are both rst class constraints gauge conditions.
p
= 0
p
e
= 0

= 0
e = 1
proper time guage conditions. (2.3.15)

The equations of motion for:
:
d
d
=
L
(2.3.16)
0 =
d
d
e

i
(2.3.17)
:
0 = 2

e
(2.3.18)
It is possible to show that the Dirac Brackets are
[
, p
= g
(2.3.19)
[
= ig
Remember that p
=
L
=
i
2
(2.3.20)
Quantization:
1. [ , ]

1
i
[ , ]
(anti)commutator
Thus: (letting = 1)
, p
= ig
= ig
(2.3.21)
+
= g
= g
(2.3.22)
2.
i
[ > = 0 for any constraint
i
.
In the gauge = 0, e = 1
p
=
L
i
2e
. .. .
=0
(2.3.23)
Thus:
2
= 0 p
2
[ > = 0 (2.3.24)
= 0 p

[ > = 0 (2.3.25)
If we let
2
(2.3.26)
and as

= 0 in the Heisenberg picture, then
is a constant and satises the algebra,
= 2g
( , anticommutator). (2.3.27)
If the coordinate representation
() =< [ > (2.3.28)
then
[
, p
] = ig
(2.3.29)
p
= i

and p

[ > = 0 becomes
i
() = 0 (2.3.30)
This is the massless Dirac equation.
For the massive Dirac equation,
S =
1
2
2
e
i

i

e
+ em
2
+ i
5

5
+ im
5
(2.3.31)
Following the same sort of argument as above, we nd,
() = 0
Diracs Approach
Schr odinger Equation: (note rst order time derivative)
i
(x, t)
t
= H(x, p)(x, t) (2.3.32)
p = i (2.3.33)
Klein-Gordon equation (K.G.):
0 =
m
2
(2.3.34)
p
=
2

2
t
2
....
2
nd
order
(c, = 1) (2.3.35)
From Dirac equation
i
t
= ( p + m) (2.3.36)
For to also satisfy the K.G. equation, then
i

t
2
= ( p + m)
2
(2.3.37)
2
t
2
=
1
2
(
i
j
+
j
i
)p
i
p
j
+ 2m(
i
+
i
)p
i
+ m
2
(2.3.38)
Thus
i
,
j
= 2
ij
(2.3.39)
i
, = 0 (2.3.40)
2
= 1 (2.3.41)
=
I 0
0 I
,
i
=
0
i
i
0
(2.3.42)
where I =
1 0
0 1
,
1
=
0 1
1 0
,
2
=
0 i
i 0
,
3
=
1 0
0 1
(2.3.43)
Eddington showed that all representations of
i
, in 4-d are unitarily equivalent.
Inclusion of the electromagnetic eld:
p p e
A
H H + e
A, electromagnetic potentials) (2.3.44)

Thus,
i
t
=
( p e
A) + m+ e
(2.3.45)
Aside:
1. This can be generalized to n dimensions.
dimension of ( , ) is
2
n/2
(n=even)
2
(n1)/2
(n=odd)
2. In even dimensions all ( , ) are unitarily equivalent, but in odd # dimensions there
are 2 sets ( , ), (+ , ) (not unitarily equivalent to each other all sets are
unitarily equivalent to one or the other, not both).
Sept. 29/99
Backtrack
[q, p] = i
[x
i
, p
j
] = i
ij
p
j
= i

x
j
(2.3.46)
i
t
= H (H E) (2.3.47)
E = i

t
(2.3.48)
But also,
x
= (x, t) (c = 1)
p
= ( p, E)
(consistent with (2.3.46),(2.3.48) provided p
= i

x
).
i.e. (Metric (, , , +))
p
= g
= ( p, E) ( p = i , E = i

t
)
= i
= i

x
= i
,

t
t
=
( p e
A) + m+ e
0
0
, =
I 0
0 I
= 4 components
=
(2.3.49)
We had
i
t
=
p
2
2m
Pauli equation
i

t
( p e
A)
2
2m
+e +

B
2m
(2.3.50)
Let be:
=
2

1

2
. .. .
Slowly
varying
=
e
imt
. .. .
(2.3.51)
* - Principle time dependence for slowly moving particles. ( = 1).
i.e.
i
t
= H = E (2.3.52)
e
iEt
= e
imt
in rest frame. (2.3.53)
Thus,
i

t
e
imt
m 0
0 m
e 0
0 e
e
imt
t
+ m
i
t
+ m
+m + e

m + e
(2.3.54)
(Assume

t
, e 0 ( is small) )
Thus from 2
nd
equation,
=

2m
(2.3.55)
Hence:
i
t
=
)
2
2m
+ e
(2.3.56)
Hence,
(
)
2
=
(p eA)
3
(p eA)
1
i(p eA)
2
(p eA)
1
+ i(p eA)
2
(p eA)
3
2
(2.3.57)
But:
j
=
ij
+ i
ijk
k
(2.3.58)
(
i
)
ab
(
i
)
cd
= 2
ad
bc
ab
cd
(2.3.59)
(
)
2
= (
i
j
)
i
j
= (
ij
+ i
ijk
k
)(p eA)
i
(p eA)
j
= ( p e
A)
2
i( p e
A) ( p e
A)
[p
i
, A
j
] = i
A
j
x
i
( p = i)
= ( p e
A)
2
+ e(

A)
= ( p e
A)
2
+ e
B
Conserved Current
i
t
= ( p +m) = (i + m) (2.3.60)
and
(i)
+
t
=
+
(
+
p
+
+
+
m) ; (
+
= ,
+
= )
=
+
( p
+
+ m)
= (+i
+
+
+
m) (2.3.61)
+
(2.3.60)
i
+
t
=
+
(i ) +
+
m (2.3.62)
(2.3.61)
i
+
t
i(
+
)
+
+
m (2.3.63)
Subtract the two
i

t
= i (
+
)

t
(
+
) + (
+
) = 0
i.e. 0 =

t
+
j
where j
= (
+
,
+
)
= (
j, )
= 0 (2.3.64)
+
probability density (2.3.65)
+
probability ux. (2.3.66)
Free particle solution in the frame of reference where p = 0 (rest frame):
i.e. i = 0
= (t)
i

t
(t)
(t)
= m
1
1
1
1
m 0
0 m

(2.3.67)
(t) = e
imt
0
(2.3.68)
(t) = e
+imt
0
(2.3.69)
(t) =
e
imt
0
e
+imt
(2.3.70)
e
+imt
= negative energy (associated with
p
2
+ m
2
)
K.G. equation
( p
2
+p
2
0
m
2
) = 0 (2.3.71)
p
0
=
p
2
+ m
2
(2.3.72)
i
t
= ( p +m) (2.3.73)
Oct. 1/99
=
e
imt
0
1
e
imt
0
2
e
+imt
0
3
e
+imt
0
4
to solve i
t
= ( p + m), perform a Boost. (2.3.74)
Rewrite the Dirac equation in a covariant form.
i
t
= (i + m) (2.3.75)
0 =
i

t
() + i ( ) m
(2.3.76)
Remember that
i
i, i

t
(2.3.77)
Then we have
(i
m) = 0 (2.3.78)
where
0
= = +
0

i
=
i
=
i
As
i
,
j
= 2
ij
(2.3.79)
,
i
= 0 (2.3.80)
2
= 1 (2.3.81)
= 2g
; g
(2.3.82)
Dierent Representations are:
Standard:
0
=
1 0
0 1

i
=
0
i
i
0
(2.3.83)
Chiral Representation:
0
=
0 1
1 0

i
=
0
i
i
0
(2.3.84)
Often, a
is written a.
Lorentz transformation: Either a Boost or a rotation.
Boost in x-direction:
x
t
=
x vt
1 v
2
, t
t
=
vx + t
1 v
2
(2.3.85)
y
t
= y (2.3.86)
z
t
= z (2.3.87)
Rotation (example):
x
t
= xcos() y sin()
y
t
= xsin() + y cos()
We know
g
x
t
x
t
= g
= t
2
x
2
y
2
z
2
(2.3.88)
Linear: x
t
= a
(2.3.89)
g
= g
(2.3.90)
The Lorentz transformation is an O(3,1) transformation.
Under the Lorentz transformation we set
t
(x
t
) = S(a)(x) (S(a) = 4 4 matrix.)
With
0 =

x
t
m
t
(x
t
) (2.3.91)
0 =

x
m
(x) (2.3.92)
(i.e. should have both the same masses and the same
matrices in each equation not a
t
, as in
i
t
x
t
(x
t
) = 0 )
Note rst:
=
x
t
x
x
t
= a
x
t
Hence, (substitute into (2.3.92)),
0 =
x
t
S
1
(a)
t
(x
t
)
multiply on left by S(a).
0 =
S(a)
S
1
(a)
x
t
m
t
(x
t
) (2.3.93)
But change of variables should leave us with an equation in the same form.

= S(a)
S
1
(a)a
(2.3.94)
From this we determine S(a)
Suppose the Lorentz transformation is innitesimal:
a
+ w
(2.3.95)
But
g
= g
= g
+ w
)(
+ w
) (2.3.96)
To order w, we see
0 = g
(w
)
or w
= w
(2.3.97)
Now take
S(a) = S(I + w)
= I
i
4
(2.3.98)
Need
. We also have
S
1
(a) = I +
i
4
(2.3.99)
Hence
I
i
4
I +
i
4
+ w
(2.3.100)
must hold to order w.
Consequently,
0 =
+
i
4
(
) w
Hence
i
4
(
) =
1
2
Solution:
=
i
2
[
] (2.3.101)
For a nite Lorentz transformation, let
w
=
1
N
w
(2.3.102)
S(a) = lim
N
1
i
4
N
= exp
i
4
(2.3.103)
and if
x
t
= a
(2.3.104)
then
t
(x
t
) = exp
i
4
(x) (2.3.105)
If
w
= w(I)
w = scalar,
(I)
= 4 4 matrix characterizing the Lorentz transformation.

then for a rotation about the x-axis,
x
t
= x w y (2.3.106)
y
t
= w x + y (2.3.107)
( innitesimal version of reg. rotation.)
In this case,
w
= w
0 1 0 0
+1 0 0 0
0 0 0 0
0 0 0 0
(where rows/columns = [x, y, z, t])

Thus S(a) = exp
i
4
= e
i
w/2
But,
21
=
i
2
(
2
2
)
= i
2
1
= i
0
2
2
0

0
1
1
0
= (1)

3
0
0
3
Thus
S(a) = exp
i
2
w

3
0
0
3
(2.3.108)
If w = 2 (1 revolution)
S(w = 2) = exp

3
0
0
3
;
2
3
= 1
We know,
e
a
= 1 + (a ) +
(a )
2
2!
+ . . .
But (a )
2
=a
2
[
i
j
=
ij
+
ijk
k
]
=
1 +
a
2
2!
+
(a
2
)
2
4!
+. . .
+
a
[a[
[a[ +
[a[
3
3!
+ . . .
= cosh [a[ + sinh [a[

a
[a[
and so S(w = 2) = exp

3
0
0
3
= cosh(i)
. .. .
=1
+sinh(i)
. .. .
=0
3
= 1
Thus if w = 2,
t
(x
t
) = (x), and if w = 4,
t
(x
t
) = +(x) (2.3.109)
All physical quantities need an even # of s.
in 360
0
rotation,

+
Oct. 5/99
So, we have:
Dirac Equation:
0 = (i
m) (2.3.110)
= 2g
; g
= (+) (2.3.111)
exp
i
4
(2.3.112)
+

+
exp
i
4
+
(2.3.113)
with
=
i
2
[
] (2.3.114)
Properties:
0
=
0 1
1 0
= (
0
)
+
i
=
0
i
i
0
= (
i
)
+
0i
=
i
2
[
0
,
i
]
= i
0
i
=
i0
i.e.
(
0i
)
+
= (i
0
i
)
+
= (i)(
+
i

+
0
)
= (i)(
i
)(
0
)
=
0i
ij
=
+
ij
Thus
0
=
(2.3.115)
i.e. = 0, = i
+
0i
0
=
0
(
0i
)
0
= (
0
)
2
0i
=
0i
= i, = j
+
ij
0
=
0
(+
ij
)
0
=
2
0
ij
=
ij
Thus if
=
+
0
then under a Lorentz transformation,
exp
i
4
0
=
+
exp
i
4
0
=
+
exp
+
i
4
0
;
2
0
= 1
=
+
0
. .. .
0
exp
i
4
0
. .. .
exp
i
4

exp
i
4
Hence
U

U
1
U = exp
i
4
Thus, (for ex.)

U
1
U =

(i.e. Lorentz invariant).

Lorentz transformations involving parity (discrete transformations).
(x, y, z, t) (x, y, z, t)
(x)
t
(x
t
) = S(a)(x)
Re. S(a)a
S
1
(a) =
(2.3.116)
x
t
= a
(2.3.117)
a
1
1
1
+1
(2.3.118)
S
i
S
1
=
i
(2.3.119)
S
0
S
1
= +
0
(2.3.120)
Thus
S = e
i
0
= P (Parity operator) (2.3.121)
i.e.
t
(x
t
) = e
i
0
(x) (for parity operation) (2.3.122)
Complete set of matrices (4 4 matrices).
I
(1) (4) (1) (4) (6)

= (16 total) (total linearly independent) (2.3.123)
5
= i
0
3
(2.3.124)
=
i
2
[
] (2.3.125)
,
5
= 0 (anticommute!) (2.3.126)
I =
1 0
0 1
,
0
=
0 1
1 0
,
i
=
0
i
i
0
(2.3.127)
5
= i
0 1
1 0

0
1
1
0

0
2
2
0

0
3
3
0
= i

1
0
0
1

i
1
0
0 i
1
1 0
0 1
(2.3.128)
5
=
0 1
1 0

1 0
0 1
0 1
1 0
(2.3.129)
5
=
0
i
i
0

1 0
0 1
0
i
i
0
(2.3.130)
0i
= i
0
i
= i
0 1
1 0

0
i
i
0
= i

i
0
0
i
(2.3.131)
ij
=
i
2
0
i
i
0

0
j
j
0
= i
ijk

k
0
0
k
(2.3.132)
Note, for example,
= g
+ g
5
(
0123
= +1)
Transformation properties of bilinears in ,

:
scalar
i.e.
S

S
1
5

S
1
5
S
If S = exp
i
4
; [
5
,
] = 0
i.e.
5
i
2
(
)
i
2
(
)
5
= 0
Thus, [
5
, S] = 0
5

....
S
5
S
1
S
=

If S = P = e
i
0
(i.e. if S = parity operator)
5

....
P
0
e
i
5
e
i
0
=
(Picks up -ve Pseudo Scalar ).

So also, one can show that

....
S
a
0

....
P
i

....
P
(The above are vectors)

and also
5

....
S
a
5

....
P
5

....
P
+
(The above are axial vectors)

and

....
S
a
0i

....
P
0i
ij

....
P
+
ij
(The above are Tensors)
0123
= +1 ( Right Handed System)
....
S
a
....
P
(
0123
= 1 Left handed system)
Thus as
5
= i
we have
5
is a pseudotensor
(in tensor ex., above, S eq. would hold, but signs on Ps change).
Oct. 6/99
Useful Identities
= [
. .. .
2g
= 2g
=
1
2
(
) =
1
2
(2g
)
= 2
1
2
(2(4))
= 2
(2.3.133)
= g
+g
5
= g
+ g
....
=0
5
=
= 2
(2.3.134)
So also,
= 4g
(2.3.135)
= 2
(2.3.136)
Tr[
] = 4g
(2.3.137)
Tr[
] = 4(g
+g
) (2.3.138)
Tr[
1
. . .
2n+1
] = Tr[
1
. . .
2n+1
5
] ;
5
5
= 1
= Tr[
5
1
. . .
2n+1
5
] (Using property of traces)
= (1)
2n+1
Tr[
5
1
. . .
2n+1
5
] (Moving
5
through all (2n+1) matrices)
Tr[
1
. . .
2n+1
] = 0 (2.3.139)
Tr[
5
] = 0 (2.3.140)
Tr[
5
] = 4i
(2.3.141)
5
=
5
= i
0
3
= i
0
3
2.4 General Solution to the free Dirac Equation
(i
m) = 0 (2.4.1)
If
0
=
+1 0
0 1
,
i
=
0
i
i
0
=
i
(2.4.2)
then in the rest frame
i.e.

x
i
= 0
=
1
e
imt
0
0
0
. .. .
0
1
+
2
0
e
imt
0
0
. .. .
0
2
+
3
0
0
e
imt
0
. .. .
0
3
+
4
0
0
0
e
imt
. .. .
0
4
(2.4.3)
Using Boosts:
S = exp
i
4
Can express in closed form. (2.4.4)
i
= exp
i
2
0
i
cosh() =
1
1 v
2
, sinh() =
v
1 v
2
(2.4.5)
2.4. GENERAL SOLUTION TO THE FREE DIRAC EQUATION 51
Recall, if
x
t
=
x vt
1 v
2
= cosh( x) sinh( t)
t
t
=
vx + t
1 v
2
= sinh( x) + cosh( t)
then cosh( ) =
1
1 v
2
sinh( ) =
v
1 v
2
But
=
i
2
[
], So,
01
= i
0
1
=
1 0
0 1

0
1
1
0
= i
0
1
1
0
(2.4.6)
S = exp
i
2
= exp
0
1
1
0
(2.4.7)
Recall,
2
1
= 1
0
1
1
0
2
=
1 0
0 1
S = 1

2
0
1
1
0
+
1
2!
0
1
1
0
1
3!
0
1
1
0
3
+
=
cosh
sinh
1
sinh
1
cosh
(Closed form) (2.4.8)

But tanh() = v, and recall the trig identities,
sinh(x + y) = sinh(x) cosh(y) + cosh(x) sinh(y) ; x = y =
sinh(2) = 2 sinh() cosh()
cosh(2) = cosh
2
() + sinh
2
()
then
tanh
=
tanh()
1 +
1 tanh
2
()
=
v
1 +
1 v
2
But p =
mv
1v
2
E =
m
1v
2
,
tanh
=
p
E + m
cosh
E + m
2m
Hence,
S = exp
i
2
01
(x-direction)
=
E + m
2m
1 0 0
p
E+m
0 1
p
E+m
0
0
p
E+m
1 0
p
E+m
0 0 1
(2.4.9)
For a boost in an arbitrary direction:
(p
= p
1
ip
2
) (2.4.10)
S =
E + m
2m
1 0
p
3
E+m
p
E+m
0 1
p
+
E+m
p
3
E+m
p
3
E+m
p
E+m
1 0
p
+
E+m
p
3
E+m
0 1
(2.4.11)
Call the columns of the above matrix W
1
(p), W
2
(p), W
3
(p), W
4
(p). General solutions to
the Dirac equation are:
1
= W
1
(p)e
ipx
2
= W
2
(p)e
ipx
3
= W
3
(p)e
ipx
4
= W
4
(p)e
ipx
(Where imt ip x)
Properties of W
i
(p)
1.
(p
r
m)W
r
(p) = 0 ;
r
=
+1, r = 1, 2
1, r = 3, 4
W(p)(p
r
m) = 0 ; (

W = W
+
0
)
2.

W
r
(p)W
r
(p) =
rr

r
3. (W
r
(p))
+
W
r
(p) =
E
m
rr
4
th
component of a vector.
W
r
W
r
= (W
r
)
+
W
r
= 0
0
0
= 1
= (W
r
)
+
W
r
and
p
m
=
( p, E)
m
4. Completeness:
4
r=1
=
r
W
r
(p)

W
r
(p) =
Oct. 8/99
So we have the following;
W
0
1
=
1
0
0
0
e
imt
, W
0
2
=
1
0
0
0
e
imt
(2.4.12)
W
0
3
=
0
0
1
0
e
imt
, W
0
4
=
0
0
0
1
e
imt
(2.4.13)
Now, boost to a frame where they have momentum p. For example,
W
1
(p) =
1
0
p
z
E+m
p
x
+ip
y
E+m
E + m
2m
e
ipx
(2.4.14)
03
W
0
1,3
= W
0
1,3
(2.4.15)
03
W
0
2,4
= W
0
2,4
(2.4.16)
Where
03
=
1
1
1
1
(2.4.17)
To obtain an eigenstate of

S with eigenvalue 1, we consider
e
i
2
S
W
0
1,2
= W
0,s
1,2
(also for 3, 4) (2.4.18)
where cos =

S n
z
(2.4.19)
Upon boosting W
(0,s)
i
, we obtain,
U(p, u
z
) = W
(u
z
)
1
(p) (2.4.20)
U(p, u
z
) = W
(u
z
)
2
(p) (2.4.21)
V (p, u
z
) = W
(u
z
)
3
(p) (2.4.22)
V (p, u
z
) = W
(u
z
)
4
(p) (2.4.23)
where now as
1. we have,
(p
r
m)W
(s)
r
(p) = 0 (2.4.24)
so (p m)U(p, u
z
) = 0 (2.4.25)
and (p + m)V (p, u
z
) = 0 (2.4.26)
2. and, as

SW
(0,s)
1,3
= W
(0,s)
1,3
(2.4.27)

SW
(0,s)
2,4
= W
(0,3)
2,4
(2.4.28)
in the rest frame, (p
r
= (0, 0, 0, m)),

Su(0, 0, 0, m; s) = u(p, s) (2.4.29)

Sv(0, 0, 0, m; s) = v(p, s) (2.4.30)
For example

Su(0, 0, 0, m; s) = u(p, s)
In general, any spinor which is a solution to the Dirac equation can be described as a linear
superposition of these four eigenstates characterized by
1. p
(p
2
= m
2
) Mass shell condition
2. Sign of p
0
(i.e. in the rest frame,
p
0
m
= 1)
3. Eigenvalue of

S in the rest frame.
Thus,
(x, t) =
d
4
k (k
m
) (2.4.31)
=
d
4
k (k
2
m
2
)
s=
b(k, s)u(k, s) + d
(k, s)v(k, s) (2.4.32)

Note if,
(i m) = 0 (2.4.33)
then (i +m)(i m) = 0 (2.4.34)
( m
2
) = 0 (2.4.35)
Note: =
=
1
2
(
= g
=
2
(
2
+ m
2
) = 0 (K.G. Eqn.) (2.4.36)
i.e., as p
= i
, (p
2
m
2
)(p) = 0 (2.4.37)
Aside:
Recall
dx (cx)f(x) ; y = cx
=
dy
[c[
yf
y
c
=
f(0)
[c[
(cx) =
(x)
[c[
So, note that;
dx (x
2
a
2
)f(x) =
dx ((x a)(x + a)) f(x) (let a > 0) (2.4.38)

=
a+
a
dx ((x a)(2a)) f(x)
+
a+
a
dx ((x +a)(2a)) f(x) (2.4.39)
=
f(a)
[2a[
+
f(a)
[ 2a[
(2.4.40)
=
f(a) + f(a)
[2a[
(2.4.41)
thus (x
2
a
2
) =
(x +a) + (x a)
[2a[
(2.4.42)
Hence for
d
4
k (k
2
m
2
)(k) =
d
3
k
dk
0
(k
2
0
k
2
m
2
)(
k, k
0
) (2.4.43)
=
d
3
k
dk
0
(k
2
0
+
k
2
+m
2
)
2
k
2
+ m
2
+
(k
2
0
k
2
+ m
2
)
2
k
2
+ m
2
(k) (2.4.44)
Thus,
(x, t) =
d
4
k (k
2
m
2
)
s=
b(k, s)u(k, s)(k

0
)e
ikx
+d
(k, s)v(k, s)(k

0
)e
+ikx
(2.4.45)
Integrate over k
0
and insert normalizing factor 2
m
(2)
3
, (E =
k
2
+ m
2
)
+ve energy solution:
+
(x, t) =
d
3
k
(2)
3/2
m
E
s=
b(k, s)e
ikx
u(k, s)(k
0
) (2.4.46)
-ve energy solution:
(x, t) =
d
3
k
(2)
3/2
m
E
s=
d
(k, s)e
ikx
v(k, s)(k
0
) (2.4.47)
Oct 12/99
2.5 Charge Conjugation
Add in an external potential (recall =
)
0 = (i e A m) (x, t) (p
eA
(2.5.1)
i

t
=
( p e
A) + m + e
We can show that there is a wave function

c
(x, t) satisfying,
0 = (i + e A m)
c
(2.5.2)
Taking the complex conjugate of (2.5.1),
0 = (
(i
eA
) m)
= C
0
(i
+ eA
) + m
(C
0
)
1
(C
0
)
If we now set
(c
0
)(
)(C
0
)
1
=
(2.5.3)
and
c
= C
0
(2.5.4)
then we recover (2.5.2). Now, as
0
=
0 1
1 0

i
=
0 i
i
i
i
0
(2.5.5)
we nd that as
0
=
T
(C
0
)
1
=
0
C then (2.5.3) becomes
C
T
C
1
=
(2.5.6)
A solution to this equation is
C = i
2
0
(2.5.7)
= C
= C
1
= C
+
= C
T
(2.5.8)
2.6. MAJORANA SPINORS 57
With this,
(
(i
+eA
) m)
c
= 0 (2.5.9)
Note that and
c
transform in the same way under a Lorentz transformation. If,
t
(x
t
) = S(a)(x) ; S(a) = exp(
i
4

) (2.5.10)
then
t
c
(x
t
) = S(a)
c
(x) (2.5.11)
(2.5.12)
Note that as,
=
+
0
= (
)
T
0
(2.5.13)
then
c
= C

(2.5.14)
If we now set
=
d
3
k
(2)
3/2
s=
b(p, s)u(p, s)e

ipx
+ d
(p, s)v(p, s)e

+ipx
(2.5.15)
then
v
c
(p, s) = C u
T
(p, s) (2.5.16)
u
c
(p, s) = C v
T
(p, s) (2.5.17)
Consequently, charge conjugation takes one from a positive energy solution associated with
a charge e to a negative energy solution associated with a charge e. (You can map one
solution to another by (2.5.16), (2.5.17).
2.6 Majorana Spinors
A Majorana spinor satises an extra condition that
=
c
(Lorentz invariant condition) (2.6.1)
As
c
= C

T
(2.6.2)
=
d
3
k
(2)
3/2
s=
(b
C u
T
e
ipx
+dC v
T
e
ipx
) (2.6.3)
then we see that if satises the Majorana condition, then
b(p, s) = d(p, s) (2.6.4)
Note that in Euclidean space,
c
= C
T
, and if =
c
, then = 0 (only way). i.e. its
not always possible to have Majorana condition.
2.7 Time Reversal
Recall that the Parity operation takes (x, y, z, t) (x, y, z, t)
t
(x
t
) = P(x) (2.7.1)
P = e
i
0
(2.7.2)
With time, the situation is slightly dierent. We must preserve,
i
t
= H =
(ie
A) +m + e
(2.7.3)
and this must be equivalent to (r
t
= r, t
t
= t)
i
t
(t
t
)
t
t
= H
t
t
(t
t
) (2.7.4)
with H
t
=
(ie
A
t
) +m + e
t
(2.7.5)
recall: A
(x) =
dx
t
D
R
(x x
t
)
(x
t
) (2.7.6)
A
= 4j
(2.7.7)
Here,
t
(t
t
) = (t) ,

A
t
(t
t
) =
A(t). In going from t t, we have the same physical

process, but using a camera thats running backwards (sequence of events in reverse order).
A Wigner time transformation takes us from (t) to
t
(t
t
).
t
(t
t
) = T (t) (2.7.8)
where T OT
1
= TO
T
1
(2.7.9)
with 0 = any operator, T is anti-unitary, and T is an ordinary matrix. i.e.
T iT
1
= T(i)T
1
= i (2.7.10)
T = T
(2.7.11)
Starting with (2.7.3),
T i
t
= T
(ie
A) + m + e
T
1
T
. .. .
(2.7.12)
i
T
t
=
(T
T
1
)(ie
A) + T
T
1
m + e
(T
) (2.7.13)
but t = t
t
+i
(T
)
t
=
(T
T
1
)(i+ e
A
t
) + T
T
1
m + e
t
(T) (2.7.14)
This is the same as (2.7.4) provided:
2.7. TIME REVERSAL 59
1.
t
(t
t
) = T
(t)
2. T
T
1
=
3. T
T
1
= +
The solution is:
T = i
1
3
(2.7.15)
= +i
1
3
(2.7.16)
The Dirac equation is invariant under an SO(3, 1) transformation
t
(x
t
) = S(a)(x) (as
well as P, (, T Parity, Charge Conj., time). Under the rotation subgroup SO(3), these are
spinors with s = 1/2.
Oct. 13/99
Chapter 3
Bargmann-Wigner Equations
The Bargman-Wigner elds are elds of higher spin (s = 1, 3/2, 2, 5/2, . . .) that are built up
out of the spin (1/2) elds.
Spin 1/2 : (i
m)
= 0 (3.0.1)
These higher spin elds are associated with wave functions:
2
...
2s
(x) spin s , totally symmetric in these 2s indices (3.0.2)
i.e.
spin 1/2
spin 1 etc.
The wave functions satisfy 2s equations:
0 =
i(
2
...
2s
(3.0.3)
0 =
i(
2
...
2s
(3.0.4)
.
.
.
0 =
i(
2s
2s
2s
2s
2
...
2s
(3.0.5)
In the frame where

x
i
= 0 (rest frame p = 0), we get
0 =
i
0
x
0
m
1
...
i
...
2s
(3.0.6)
with
0
=
1 0
0 1
, the solutions generalize the spin 1/2 solution,

W
1
=
1
0
0
0
e
imt
, W
2
=
0
1
0
0
e
imt
, W
3
=
0
0
1
0
e
+imt
, W
4
=
0
0
0
1
e
+imt
(3.0.7)
When there are 2s indices
+ve energy solutions
61
62 CHAPTER 3. BARGMANN-WIGNER EQUATIONS
(1)
1
1
1
2
. . .
1
2s
=
(1)
1
...
2s
(2)
2
1
1
2
1
3
. . .
1
2s
+
1
1
2
2
1
3
. . .
1
2s
+ . . .
+
1
1
1
2
1
3
. . .
2
2s
=
(2)
1
...
2s
(3)
2
1
2
2
1
3
. . .
1
2s
+
2s
2
symmetrized terms.
(
.
.
.)
(2s+1)
2
1
2
2
2
3
. . .
2
2s
=
(2s+1)
1
...
2s
Similarly, there are (2s + 1) negative energy solutions.
The operator
1
2
operating on
is generalized to
2 (
z
)
1
,
2
2
,...,
2s
2s
= (
z
)
3
. . .
2s
2s
+
1
(
z
)
3
. . .
2s
2s
.
.
.
+
3
. . . (
z
)
2s
2s
(3.0.8)
This acts on
1
...
2s
(1)
= s
(1)
(3.0.9)
(2)
=
s
1
2
(2)
(3.0.10)
(3)
= (s 1)
(3)
(3.0.11)
.
.
.
(2s+2)
= (s)
(2s+2)
(3.0.12)
Thus
(i)
are the (2s + 1) eigenvectors of (
z
). These are the spin eigenstates. Now, lets
examine the spin-one equations (s = 1). Consider
(x) =
(x). We have,
0 =
(3.0.13)
0 =
(3.0.14)
C
1
C =
T
C = i
0
2
= C
= C
1
= C
+
= C
T
(3.0.15)
63
With C, we see that
(
C)
= (
C)
- 4 matrices
(
C)
= (
C)
- 6 matrices
=
i
2
[
For a total of 10 symmetric 4 4 matrices. We can then have,
(x) = mA
(x) (
C)
+
1
2
F
(x) (
C)
(3.0.16)
Substitute this into (3.0.13), (3.0.14), and add;
0 = (i
m)
mA
C)
+
1
2
F
C)
mA
C)
+
1
2
F
C)
(3.0.17)
Now, right multiplying by C
1
(remember that C
T
C
1
=
). This gives,
0 = 2m[
] (
) + i [
2m
2
(3.0.18)
but recall that we know,
[
] = 2i
(3.0.19)
[
] = 2i (g
) (3.0.20)
So,
0 = 2m
2 (g
2m
2
0 = m
[(
) F
] +
2m
2
A
Thus, we get the Proca Equations (Massive spin 1 equations).

F
=
m
2
2
A
(3.0.21)
If we now set m = 0, we recover the free Maxwell equations:
F
t
(3.0.22)
= 0 (3.0.23)
Where, recall that (3.0.22) gives

B = 0 ,

E +

B
t
= 0
and (3.0.23) gives

E = 0 ,

B

E
t
= 0
with, F
ij
=
ijk
B
k
, F
0i
= E
i
.
Oct. 15/99
Transformation of the eld A
(x) under a Lorentz transformation.

c.f.
t
(x
t
) = exp
i
4
(x)

=
i
2
[
]
Since,
= 2g
(+, , , )
= A
+ g
= 2i
. . .
(Aside - if = = 1 , = 2 , = 3,
LS =
12
,
13
i
1
2
, i
1
=
2
3
+
3
2
=
2
,
3
= 2i
23
RS = Ag
11
23
= A
23
A = 2i)
For A
, since it is a vector, under an innitesimal Lorentz transformation, we have,

x
t
= x
) (3.0.24)
A
t
(x
t
) = A
(x)
(x)
= A
(x) +
i
u
(x) (3.0.25)
Where (S
= 2i (g
) (3.0.26)
65
But now,
(S
)(S
) (S
)(S
) (S
(S
(S
(S
= 2i (g
+ g
(3.0.27)
i.e.
A
t
(x
t
) =
exp
i
4
(x
t
) (3.0.28)
for a nite transformation.
- spin 1/2 representation of SO(3, 1) (fundamental rep.)
A
- spin 1 representation of SO(3, 1) (Adjoint representation).

i.e. under a rotation,
exp
where
i
2
,

j
2
= i
ijk
k
2
2
=
1
2
1
2
+ 1
and
A
t
0
(x
t
) = A
0
(x)
A
t
i
(x
t
) =
e
i
ij
A
j
(x)
where
S
1
=
0 0 0
0 0 i
0 i 0
, S
2
=
0 0 i
0 0 0
i 0 0
, S
3
=
0 i 0
i 0 0
0 0 0
(3.0.29)
Note that,
(S
i
)
mn
=
1
2
ijk
(S
k
)
mn
= i
mij
(3.0.30)
[S
i
, S
j
] = i
ijk
S
k
(3.0.31)
S
2
= 1(1 + 1)
= s(s + 1) (3.0.32)
Massless Particles
Dirac equation:
0 = (i
m) (3.0.33)
0
=
0 1
1 0
,
i
=
0
i
i
0
,
5
=
1 0
0 1
0 =
i

t
p
i

t
+ p
, ( p = i) (3.0.34)
If m = 0, then we get the Decoupled (Weyl) equations;
i

t
= p (3.0.35)
i

t
= p (3.0.36)
More generally, if we let p
+
=
1+
5
2
, p
=
1
5
2
where
5
,
= 0 ,
2
5
= 1,
p
+
+ p
= 1
(p
+
)(p
) = 0
(p
+
)
2
= p
+
(p
)
2
= p
So,
p
+
=
0 0
0 1
1 0
0 0
(3.0.37)
p
+
=
(3.0.38)
If
= p
(3.0.39)
then
0 = i
+
m
(3.0.40)
0 = i
m
+
(3.0.41)
For massless particles, we cant go to the rest frame, i.e.
p
2
= (p
0
)
2
p
2
= m
2
= 0
there is no rest frame, so look at the frame where
p
= (0, 0, p, p) ; (p
1
= p
2
= 0 , p
3
= p
0
= p) (3.0.42)
67
In this frame,
p
+
1 0
0 1

1
= p
+
1 0
0 1

1
(3.0.43)
(rst of Weyl equations)
=

1
0
(3.0.44)
(only one state of negative energy)
p
1 0
0 1

1
= p
1 0
0 1

1
(3.0.45)
=
(3.0.46)
(only one state of positive energy)
So, Weyl equations are:
Negative energy solutions:
i

t
= p (3.0.47)
Positive energy solutions:
i

t
= p (3.0.48)
Oct. 19/99
Can we dene spin of a massless particle, seeing that there is no rest frame? - Employ
the Pauli-Lubanski tensor.
S
=
i
2
(3.0.49)
=
i
2
In the rest frame of a massive (m) particle,

k
= (0, 0, 0, m) (3.0.50)
S
=
i
2
m (3.0.51)
(0 in index , , all spatial.)
0123
= +1 (3.0.52)
S
i
=
i
2
ijk
jk
m
= m

i
0
0
i
(3.0.53)
Note:
ij
=
i
2
ijk

k
0
0
k
i
=
0 i
i
i
i
0
Thus, in the rest frame

1
2
S
reduces to
m
2
.
1
2
S
generalizes spin to an arbitrary frame. For a massless particle moving along the
z-axis (with k
2
= 0, k
= (0, 0, k, k)), then we have,

S
1
=
i
2
1
k
=
i
2
(
13
k +
10
k)
(3.0.54)
etc. for S
2
, S
3
, S
4
Now consider the two massless solutions to the Bargmann Wigner equations
W
(1)
2
...
2s
=
1
2
2
2
. . .
2s
2
(3.0.55)
W
(2)
2
...
2s
=
1
3
2
3
. . .
2s
3
(3.0.56)
Generalize the Pauli Lubanski tensor,
(S
1
,...,
2s
2s
= (S
1
(
2
. . .
2s
2s
)
+(S
2
(
3
. . .
2s
2s
)
. . .
+(S
2s
2s
1
. . .
2s1
2s1
(3.0.57)
It can now be shown that
S
1
W
(1)
= 0 S
3
W
(1)
= 2skW
(1)
(3.0.58)
S
2
W
(1)
= 0 S
3
W
(2)
= +2skW
(2)
(3.0.59)
S
1
W
(2)
= 0 S
4
W
(1)
= +2skW
(1)
(3.0.60)
S
2
W
(2)
= 0 S
4
W
(2)
= 2skW
(2)
(3.0.61)
But k
= (0, 0, k, k) and thus

1
2
S
= sk
.
Hence for a massless particle, there are two solutions to the Bargmann-Wigner equations,
and these two solutions correspond to spin states whose eigenvalues are S, and whose
direction is in the direction of k
.
Thus there are only two polarizations for any massless particle of spin S. We lose states
through the introduction of gauge symmetry. ex.
69
massless spin 1/2 - 2 states
massive spin 1/2 - 2
1
2
+ 1 = 2 states
massless spin 1 - 2 states
massive spin 1 - (2s + 1) = 3 states
massless spin 3/2 - 2 states
massive spin 3/2 - (2s + 1) = 5 states
Chapter 4
Gauge Symmetry and massless spin
one particles
Remember that
F
(4.0.1)
= m
2
A
0 as m
2
0 (4.0.2)
If m
2
0, there is the gauge invariance,
A
(4.0.3)
We now have,
= 0 (4.0.4)
F
(4.0.5)
which are derived from the action
S =
1
4
d
4
x F
L =
1
4
F
2
(4.0.6)
(the -ve sign is present to keep the energy +ve, and the magnitude 1/4 is convention). i.e.
S
A
(y)
=
1
4
d
4
s
2F
(y)
=
1
2
d
4
x F
(y)
(
)
=
1
2
d
4
x F
(x)
4
(x y) g
(x)
A
(y)
= g
4
(x y)
(x)
= 0
71
72 CHAPTER 4. GAUGE SYMMETRY AND MASSLESS SPIN ONE PARTICLES
Just as with
L = L(q
i
(t), q
i
(t)) and S =
dt L(q
i
(t), q
i
(t)) (4.0.7)
we have
S =
d
4
x L(A
)
=
dt
d
3
x L
(x, t), A
(x, t),

t
A
(x, t)
. .. .
L
(4.0.8)
q
i
(t) A
(x, t), A
(x, t) (4.0.9)
q
i
(t)

t
A
(x, t) (4.0.10)
and, where previously we had i discrete, we now have a continuum () number of degrees of
freedom. So, just as
p
i
=
L
q
i
(x, t) =
L
t
A
(x, t)
momentum density (4.0.11)
c.f.
=

x
Oct. 20/99
Now, as
A
(x
t
, t)
A
(x, t)
= g
3
(x x
t
) (4.0.12)

(x, t) =
L
1
4
F
=
1
2
F

t
=
0
=
1
2
F
(
0
A
)
F
But,
F
= F
0i
F
0i
+ F
i0
F
i0
+ F
ij
F
ij
+ F
00
F
00
= 2F
0i
F
0i
+ F
ij
F
ij
. .. .
No t dep.
73

= F
0i
(
0
A
)
F
0i
(4.0.13)
F
0i
=
0
A
i
i
A
0
(4.0.14)
Thus,
0
= 0 (4.0.15)
j
= F
0i

(
0
A
)
F
0i
= F
0i

(
0
A
j
)
(
0
A
i
i
A
0
)
= F
0j
(4.0.16)
Remember,
A
A,
(A
0
= , A
i
=

A) (4.0.17)
E =

A
t
=
i
A
0
0
A
i
= +
i
A
0
0
A
i
= F
0i
(4.0.18)

j
= E
j
= F
0j
= +F
0j
(4.0.19)
(4.0.19) is Canonical momentum (Not mechanical momentum). Canonically conjugate to
A
3
. i.e. A
i
has canonically conjugate momentum
j
= E
j
.
A
i
= A
i
A
0
= +A
0
(4.0.20)
0
= 0 (4.0.21)
i.e. A
0
has no conjugate momentum.
Dene a Poison bracket,
[q
i
, p
j
] =
ij
[A, B] =
A
q
i
B
p
i
A
p
i
B
q
i
(4.0.22)
Similarly
[F(A
), G(A
)] =
d
3
x g
F
A
(x, t)
G
(x, t)

F
(x, t)
G
A
(x, t)
(4.0.23)
4.1 Canonical Hamiltonian Density
H =

t
A
L (re. H = p
i
q
i
L) (4.1.1)
=
0
....
=0
0
A
0
+
i
0
A
i
1
4
2F
0i
F
0i
+ F
ij
F
ij
(4.1.2)
(
0
= primary constraint)
H =
i
0
A
i
i
A
0
. .. .
F
0i
+
i
A
0
+
1
2
F
0i
F
0i
+
1
4
F
ij
F
ij
Remember that,
(
B)
i
= (

A)
i
=
ijk
(
j
A
k
)
imn
B
i
=
imn
ijk
j
A
k
imn
ijk
=
mj
nk
mk
nj
So that
imn
B
i
= (
mj
nk
mk
nj
)
j
A
k
=
m
A
n
n
A
m
= F
mn
(4.1.3)
F
mn
=
mnl
B
l
(4.1.4)
Consequently:
H =
i
F
0i
....
i
+
i
i
A
0
+
1
2
F
0i
F
0i
. .. .
i
(
i
)
+
1
4
(
ijk
B
k
)(
ijl
B
l
)
. .. .
2B
i
B
i
=
1
2
i
+
1
2
B
i
B
i
+
i
i
A
0
;
j
= E
j
H =
1
2
(
E
2
+

B
2
) +
i
i
A
0
(4.1.5)
Primary constraint (
0
(x, t) = 0. Thus,
0
= 0 =
0
(x, t), H(y, t)
(4.1.6)
But we have the fundamental poisson bracket
[A
(x, t),
(y, t)] = g
3
(x = y) (c.f. [q
i
, p
j
] =
ij
) (4.1.7)
4.1. CANONICAL HAMILTONIAN DENSITY 75
0
= 0 =
0
(x, t),
i
(y, t)

y
i
A
0
(y, t)
0 =
i
(y, t)

y
i
3
(x y)
0 = +

E(y, t)
3
(x y) (4.1.8)
Secondary constraint

E = 0 (Gausss Law).
Oct. 22/99
So,
L =
1
2
F
(4.1.9)
i
= E
i
= F
0i
(4.1.10)
0
= 0

0
= 0
i
E
i
= 0
First class Constraints (4.1.11)

B
i
=
1
2
ijk
F
jk
(4.1.12)
H
0
=
E
2
+

B
2
2
+ A
0
i
E
i
(4.1.13)
H =
d
3
x H (4.1.14)
H
T
= H
0
+c
i
i
=
1
2
(
E
2
+

B
2
) + A
0
i
E
i
+ c
1
0
+ c
2
i
E
i
; can absorb c
2
A
0
(4.1.15)
A
0
- Lagrange multiplier for the constraint
i
E
i
= 0.
We need two gauge conditions:
Coulomb gauge (usual choice)
A
0
= 0 (4.1.16)
i
A
i
= 0 (4.1.17)
Axial gauge also possible
A
0
= 0 (4.1.18)
A
3
= 0 (say) (4.1.19)
Now, replace Poisson Brackets by Dirac Brackets. The fundamental P.B. is:
A
i
(x, t),
j
(y, t) =
ij
3
(x y) (4.1.20)
i
=
i
. . .
i
(4.1.21)
The Dirac Bracket is then:
A, B
= A, B
i,j
A,
i
d
ij
j
, B (4.1.22)
where d
1
ij
=
i
,
j
So,
0
(x, t), A
0
(y, t) =
3
(x y) (4.1.23)

x
i
i
(x, t),

y
i
A
j
(y, t)
=

x
i
y
i
ij
3
(x y)
=
2
x
3
(x y) (4.1.24)
We now need (
2
)
1
= G(x y). i.e.
2
G(x y) =
3
(x y) =
d
3
k
(2)
3
e
i
k(xy)
(4.1.25)
Write G(x y) =
d
3
k
(2)
3
e
i
k(xy)
g(
k) (Fourier transf.) (4.1.26)

2
G(x y) =
d
3
k
(2)
3
(
k
2
g(
k))
. .. .
=1
e
i
k(xy)
(4.1.27)
Aside:
(x) = 0 (x = 0)
dx (x) = 1
dx
e
ax
2
=
= 1
lim
a
e
ax
2
=
0 if x = 0
if x = 0
Thus,
(x) = lim
a
e
ax
2
So
dk
2
e
ikx
= lim
a0
dk
2
e
ikxak
2
ak
2
is regulator
= lim
a0
dk
2
expa
k
ix
2a
exp
x
2
4a
= lim
a0
dk
t
2
e
a(k
)
2
exp
x
2
4a
= lim
a0
a
1
2
exp
x
2
4a
= (x)
end of aside.
Now, recall (4.1.27)
g(
k) =
1
k
2
(4.1.28)
Thus,
G(x y) =
d
3
k
(2)
3
k
2
e
i
k(xy)
(4.1.29)
Say (x y) is along k
3
axis.
G(x y) =
2
0
d
+1
1
d(cos )

0
dk
(2)
3
k
2
1
k
2
e
ikr cos
; r = [x y[
=
1
(2)
2
1
1
dz

0
dk e
ikrz
; (z = cos )
can extend k integral to () int. over z makes even
=
1
2(2)
2
1
1
dz
dk e
ik(rz)
=
1
4
1
1
dz (rz)
=
1
4r
dz
t
(z
t
) ; z
t
= rz
=
1
4r
(4.1.30)
Thus,
G(x y) =
1
4[x y[
(4.1.31)
One can verify that
1
4[x y[
= 0 , x = y (4.1.32)
and that
d
3
x
2
1
4[x[
=
1
4
d
3
x
1
[x[
by Gausss Law (Radius of Sphere = R)

=
1
4
lim
R=
Sphere
d R
2
n
1
[x[
= 1 (4.1.33)
Thus our Dirac Brackets are
F(x, t), G(y, t)
= F(x, t), G(y, t)
d
3
x
t
d
3
y
t
F(x, t),

x
ti
i
(x
t
, t)
G(x
t
y
t
)

y
t
j
A
j
(y
t
, t), G(y, t)
G(x, t),

x
ti
i
(x
t
, t)
G(x
t
y
t
)

y
t
j
A
j
(y
t
, t), F(y, t)
(4.1.34)
Hence;
A
i
(x, t),
j
(y, t)
ij
(x y)

x
i
x
j
G(x y)
ij
(x y)
. .. .
3
ij
(xy)
2
= G(x y) (4.1.35)
(4.1.36)
Consistency:
x
i
A
i
(x, t),
j
(y, t)
=

x
i
ij

x
i
x
j
1
(x y)
= 0 (4.1.37)
A
i
(x, t),
j
(y, t)
=
3
ij
(x y) (4.1.38)
Oct. 26/99
H
0
=
1
2
(
2
+B
2
) + c
1
0
+ c
2
i
(4.1.39)

A
0
= [A
0
, H]
= [A
0
, c
1
0
] = c
1
(Entirely arbitrary prior to xing gauge) (4.1.40)

A
i
= [A
i
, H]
= [A
i
,
1
2
2
+ c
2
j
] = A
i
i
c
2
(4.1.41)
where the usual gauge invariance in A is recovered;
A
(4.1.42)
c
1
=
0
i
c
2
=
i
The Fourier expansion of A(x, t) in the Coulomb gauge (

i
A
i
= 0) is
A(x, t) =
d
4
k
(2)
4
k
2
2
=1
(k, )[a(k, )e
ikx
+a
(k, )e
+ikx
]
(4.1.43)
1. (k
2
) is the mass shell condition (i.e. ensures that A = 0
= 0 A = 0
if
i
A
i
= 0)
2. (k, ) k = 0 ( = 1, 2) (Ensures
i
A
i
= 0). and
(k, ) (k,
t
) =
(k, 1) (k, 2) =
k
[k[
3. A = A
is real. Thus,
ae
ikx
+ a
e
ikx
occurs.
Note:
dx (x
2
a
2
)f(x) =
a+
a
dx ((x +a)(x a))f(x)
+
a+
a
dx ((x + a)(x a))f(x)
=
a+
a
dx (2a(x a))f(x) +
a+
a
dx (2a(x + a))f(x)
but recall that
dx (cx)f(x) =
f(x)
[c[
=
f(a)
[2a[
+
f(a)
[2a[
(4.1.44)
Hence
(x
2
a
2
) =
(x a) + (x + a)
[2a[
(4.1.45)
A(x, t) =
d
3
k
(2)
3
dk
0
(2)
(k
2
0
k
2
)
2
=1
(k, )[ae
ikx
+a
e
ikx
] (4.1.46)
Rescaling a, a
and setting
k
=
k
2
,
k,
= (k, ), etc., we get
A(x, t) =
1
2
d
3
k
(2)
3/2
1
2
k
2
=1
k,
a
k,
e
i(kx+
k
t)
+ a
k,
e
i(kx
k
t)
+ a
k,
e
+i(kx+
k
t)
+ a
k,
e
+i(kx
k
t)
(4.1.47)
If k k, and if
k,
=
k,
( = 1)
k,
( = 2)
(4.1.48)
k,1
k,2
=
(k)
[k[
(4.1.49)
then
d
3
k
(2)
3/2
1
2
k
2
=1
k,
a
k,
e
i(kx+
k
t)
=
d
3
k
(2)
3/2
1
2
k
2
=1
k,
a
k,
e
+i(kx
k
t)
(4.1.50)
Now set a
k,
=
k,
=1
a
k,
=2
. Thus everything collapses to
A(x, t) =
d
3
k
(2)
3/2
1
2
k
2
=1
k,
a
k,
e
ikx
+ a
k,
e
ikx
(4.1.51)
where k x = k x
k
t
(we need not use lane wave basis; any complete set of functions will do).
Recall the Klein-Gordon equation:
0 = (+ m
2
) (4.1.52)
0 =

2
t
2

2
+ m
2
(4.1.53)
Let =

t

t
= (
2
m
2
)
1
=
1
+
i
m

2
=
1

i
m
(4.1.54)
Let
=
(4.1.55)
i
t
=
(
3
+ i
2
)
(i)
2
2m
+ m
3
. .. .
H
(4.1.56)
2
=
0 i
i 0

3
=
1 0
0 1
(4.1.57)
K.G. equation follows from an action
S =
d
4
x
1
2
(
)(
)
m
2
2

2
. .. .
/
(4.1.58)
Now,
=
L
(
0
)
=
0
(4.1.59)
H = (
0
) L
= (
0
)
1
2
(
0
)
2
1
2
()
2
m
2
=
1
2
2
+ ()
2
+m
2
(4.1.60)
Fundamental P.B.
[(x, t), (y, t)] =
3
(x y) (4.1.61)
(+ m
2
) = j Describes , K, . . . (4.1.62)
i
t
= p2m weve treated as single particle (4.1.63)
i
t
=
p
2
2m
+V
in a classical potential (4.1.64)

Dirac showed that the potential V itself, when its due to an E.M. eld, should be quantized.
Make relativistic by considering Dirac eq.
i
t
= ( p + m) (4.1.65)
the above equation doesnt describe just 1 particle anymore, but in fact now describes
4
this one equation describes particles (spin
1
2
)
if you put cs and s back in, s dont all cancel, oating around
Maxwells equations (Fiddle with Bargmann-Wigner (spin 1), and get these)
= 0 (4.1.66)
F
(4.1.67)
s cancel, if m = 0 classical equations. (2 particles/polarizations). i.e. in order to include
relativity, you end up having to describe more than 1 particle. Wave function describing
single particle actually quantizing wave function itself (2
nd
quantization).
Oct. 27/99
Chapter 5
(2
nd
) Quantization, Spin and Statistics
5.1 Harmonic Oscillator
H =
1
2
(p
2
+
2
q
2
) ; =
k
m
(5.1.1)
Quantize:
[q, p] = i ; ( = 1) (5.1.2)
p = [p, H] =
2
q (5.1.3)
q = [q, H] = p (5.1.4)
Dene:
a =
q + ip
2
; ( annihilation) (5.1.5)
a
=
q ip
2
; (creation) (5.1.6)
Note that:
[a, a
] = 1
[a, a] = [a
, a
] = 0
H =
a +
1
2
and
[H, a
] = a
[H, a] = a
83
84 CHAPTER 5. (2
ND
) QUANTIZATION, SPIN AND STATISTICS
Suppose
H[n` =
n
[n` (5.1.7)
H(a
[n`) = ([H, a
] + a
H)[n`
= (a
[n` + a
n
[n`)
= (
n
+)(a
[n`) (5.1.8)
and similarly H(a[n`) = (
n
)(a[n`) (5.1.9)
More generally,
H(a
m
[n`) = (
n
m)(a
m
[n`) (5.1.10)
For the energy to have a lower bound, there must be a state [0` such that a[0` = 0.
In this case,
H[0` =
a +
1
2
[0` ; a[0` = 0
=
1
2
[0` (Consistent with uncertainty principle).
=
1
2
p
2
+
2
q
2
[0` p q

2
4
(5.1.11)
i.e. Vacuum state [0` is lowest energy state, not nothing.
a
[0` [1`
(a
)
n
[0` [n`
H[n` =
n +
1
2
[n` (n = 0, 1, . . .) (5.1.12)
For 'n[m` =
nm
[n` =
(a
)
n
(n + 1)
[0` (5.1.13)
For our real K.G. eld:
0 =
+ m
2
has a solution
(x) =
d
3
k
(2)
3/2
1
2
k
a
k
e
ikx
+a
k
e
ikx
(5.1.14)
(note that this is only a scalar eld here, dont need polarization vectors
k,
). For a
complex eld, we have the same treatment for real and complex parts.
k x =
k
t k x
with
k
=
k
2
+ m
2
f
k
=
e
ikx
(2)
3
2
k
5.1. HARMONIC OSCILLATOR 85
As
d
3
k
(2)
3
e
ikx
=
3
(x) (5.1.15)
then

d
3
x f
k
(x, t) =
1
2
k
a
k
+a
k
e
2i
k
t
(5.1.16)
Remember that
(x, t) =
(x, t)
t
d
3
x f
k
(x, t) =
i
2
a
k
a
k
e
2i
k
t
(5.1.17)
(5.1.18)
Thus,
a
k
= i
d
3
x
k
(x, t) (
t
(x, t))

t
f
k
(x, t)
(x, t)
. .. .
f

(5.1.19)
If we now quantize
[
(x, t),

(y, t)] = i
3
(x y) ( let = 1) (5.1.20)
then we can show that
[ a(k), a(k
t
)] = 0 (5.1.21)
[ a
(k), a
(k
t
)] = 0 (5.1.22)
[ a(k), a
(k
t
)] =
3
(k k
t
) (5.1.23)
and
H =
d
3
x
1
2
2
+
1
2
()
2
+
m
2
2

2
d
3
x
(k)a(k) +
1
2
(5.1.24)
Harmonic oscillator! (Note the
1
2
term above will diverge at each point, get this vacuum
oscillation
over all points, get .)

States in the Hilbert space of these operators:
[0
k
, 0
k
, 0
k
, . . .` = vacuum [0` (5.1.25)
a
2!
[0` = [0
k
, . . . , 1
l
, . . .` (5.1.26)
86 CHAPTER 5. (2
ND
i.e. the a
l
creates a K.G. particle of momentum l. So,
(a
l
)
2
3!
(a
2!
[0` (5.1.27)
is a state with one K.G. particle of momentum l
t
and two of momentum l. Notation: When
we quantize, the goes to ;
a
k
a
k
(5.1.28)
As it is only the dierence between energy levels that is observed, we take,
: H : normal order of H (5.1.29)
: H : =
d
3
k
k
(a
k
a
k
) (5.1.30)
(excitations above lowest energy state)
As
H =
1
2
d
3
k
k
k
a
k
+ a
k
a
(5.1.31)
: H : is obtained by rewriting H so that all annihilation operators are on the right.
More generally, : (x)(y) : - ordered so that, again, all annihilation operators are on the
right hand side. As,
i
d
3
xf
f
k
=
3
(k k
t
) (5.1.32)
d
3
x a
k
f
k
(5.1.33)
+
=
d
3
x a
k
f
k
(5.1.34)
Oct. 29/99
: (x)(y) : = :
(x) +
+
(x)

(y) +
+
(y)
:
=
(x)
+
(y) +
(y)
+
(x)
. .. .
inversion
+
(x)
(y) +
+
(x)
+
(y)
5.1. HARMONIC OSCILLATOR 87
[(x), (y)] =
(
+
(x) +
(x)), (
+
(y) +
(y))
d
3
k
k
f
k
(x) + a
k
f
k
(x)
d
3
k
t
k
(y) + a
k
f
k
(y)
d
3
k d
3
k
t
(2)
3
2
k

k
a
k
, a
. .. .
kk
e
i(kxk
y)
+
k
, a
k
. .. .
kk
e
i(kxk
y)
d
3
k
(2)
3
(2
k
)
e
ik(xy)
e
+ik(xy)
=
i
(2)
3
d
3
k
k
e
ik(xy)
sin [
k
(x
0
y
0
)]
d
3
k
2
k
d
4
k (k
2
m
2
)
=
d
4
k
(2)
3
(k
2
m
2
)(k
0
)e
k(xy)
where (k
0
) =
+1 k
0
> 0
1 k
0
< 0
= i(x y) (5.1.35)
Properties:
0 =
x
m
2
(x y) expected, as (m
2
) = 0
(x y) = (y x) Reasonable, as [(x), (y)] = [(y), (x)]
Note:
(x y, 0) = 0
i.e.
d
4
k
(2)
3
(k
2
0
k
2
m
2
)(k
0
)e
ik(xy)
= 0
Thus,
(x y) = 0 if (x y)
2
< 0 (5.1.38)
Hence,
[(x), (y)] = 0 if (x y)
2
< 0 (5.1.39)
as required by causality. If x, y are separated temporally, (i.e. within light cone), then the
order of x, y (in time) is the same observers. If they are separated by spatial separation,
(outside light cone), order (time order) not necessarily the same.
88 CHAPTER 5. (2
ND
5.2 Feynman Propagator
'0[T(x)(y)[0` (5.2.1)
Hence,
T(x)(y) = (x
0
y
0
)(x)(y) + (y
0
x
0
)(y)(x) (5.2.2)
(x) =
1 x > 0
0 x < 0
(5.2.3)
This is in fact relativistically invariant.
T(x)(y) = '0[
(x
0
y
0
)(
+
x
+
x
)(
+
y
+
y
) + (y
0
+x
0
)(
+
y
+
y
)(
+
x
+
x
)
[0`
where
+
x

+
(x), etc.
so, in rst term above,
+
x
,
y
0 and in second term,
+
y
,
x
0
a
k
[0` = 0 '0[a
+
k
= 0
= '0[
(x
0
y
0
)
+
y
+ (y
0
x
0
)
y

+
x
[0`
2
a
k
[0` = [1
k
`
'0[
1
2
a
k
= '1
k
[
.
.
.
=
d
3
k
(2)
3
(2
k
)
(x
0
y
0
e
ik(xy)
+ (y
0
x
0
)e
ik(xy)
(5.2.4)
(Note that
(t) =
d
2i
e
it
i
; > 0 (5.2.5)
k
0
(t) =
dk
0
2i
e
itk
0
k
0
i
(5.2.6)
T((x)(y)) = i
d
4
k
(2)
4
e
ik(xy)
k
2
m
2
+ i
(5.2.7)
Note that k
2
0
k
2
m
2
+i poles at k
0
=
k
2
+ m
2
i. Hence for
dk
0
we have the
contour Hence
'0[T(x)(y)[0` = i
d
4
k
(2)
4
e
ik(xy)
k
2
m
2
+ i
= i
F
(x y) (Feynman Propagator) (5.2.8)
5.2. FEYNMAN PROPAGATOR 89
t < 0 no poles, (t) = 0
t > 0 1 pole, residue (t) = 1
Note:
(
x
+ m
2
)
F
(x y) =
4
(x y) (5.2.9)
Remember
(+ m
2
)(x y) = 0 (5.2.10)
We can repeat this procedure for the vector eld. Classical P.B.
[A
i
(x, t),
j
(x, t)] = g
ij
3
(x y) (i, j spatial indices) (5.2.11)
A
i
(x, t),

j
(x, t)
= i
ij
3
(x y) (5.2.12)
90 CHAPTER 5. (2
ND
But, in order to eliminate the constraints, we need the classical Dirac Bracket
A
i
(x, t),
j
(y, t)
ij
3
(x y) (Coulomb gauge) (5.2.13)
which goes to the commutator,
A
i
(x, t),

j
(y, t)
= i
ij
3
(x y) (5.2.14)
Well work with this commutator.
Nov. 2/99
A
i
(x, t) =
d
3
k
(2)
3/2
1
2
k
2
=1
((k, ))
a(k, )e
ikx
+ a
(k, )e
ikx
(5.2.15)
Re: (k, 1) = (k, 1)
(k, 2) = (k, 2)
(k, 1) (k, 2) =
k
[k[
These are appropriate for the Coulomb gauge - the gauge condition is taken care of by the
two conditions with k k, above. Also, once again we have a superposition of harmonic
oscillators making up A(x, t);
[a(k, ), a(k
t
,
t
)] = 0 (5.2.16)
(k, ), a
(k
t
,
t
)
= 0 (5.2.17)
a(k, ), a
(k
t
,
t
)
=
3
(k k
t
)
(5.2.18)
Recall ( = 1)
H =
d
3
x :
1
2
(E
2
+ B
2
) :
=
d
3
k
k
2
=1
a
(k, )a(k, ) (5.2.19)

(i.e. E =
k
N
k
Einstein photoelectric eect)
p =
d
3
x : E B :
= ()
d
3
k k
2
=1
a
(k, )a(k, ) (5.2.20)
p = kN
k
de Broglie) (5.2.21)
Remember that, for a Harmonic Oscillator, if
N = a
a (5.2.22)
5.2. FEYNMAN PROPAGATOR 91
then
N[n` = n[n` (5.2.23)
General expression for A
(x, t) and its time ordered product:

Now A
(x, t) =
d
3
k
(2)
3/2
1
2
k
2
=1
(k, )
(k, )e
ikx
+ a(k, )e
ikx
(5.2.24)
So '0[TA
(x)A
(y)[0` =
d
4
k
(2)
4
e
ik(xy)
k
2
+ i
2
=1
(k, )
(k, )
= iD
F
(x y) (5.2.25)
In the Coulomb gauge, if
(k, ) = (0, (k, )) (k = 0) (5.2.26)

Let
= (1, 0, 0, 0). In frame where
has the above form,
=
k
(k )
2
2
(5.2.27)
Note
k = 0 = 0
k = 0 (5.2.28)
k
(1),
(2),
are 4 orthonormal vectors, all dened with respect to frame where
(k, ) = (0, (k, )).

Now,
2
=1
(k, )
(k, ) = g
(5.2.29)
Reduces to
=
ij
k
i
k
j
k
2
in frame where
= (0, ) (5.2.30)
Hence in the Coulomb gauge,
iD
F
(x y) =
d
4
k
(2)
4
e
ik(xy)
k
2
+ i
+
k
2
+ k
k (k
+k
)
k
2
+ ( k)
2
. .. .
(5.2.31)
is an artifact of using the Coulomb gauge and cannot aect any physical process.
92 CHAPTER 5. (2
ND
5.3 Quantizing the Dirac Field
We have
(x) a
k
, a
(5.3.1)
with states:
[n
k
1
, n
k
2
, . . . , n
k
m
` =
(a
k
1
)
n
k
1
(n
k
1
+ 1)!

(a
k
m
)
n
k
m
(n
k
m
+ 1)!
[0, . . . , 0` (5.3.2)
where n
k
1
, . . . , n
k
m
= 1, 2, . . . , . For electrons, Pauli suggested that no two electrons can
be in the same state.
Fermionic Harmonic Oscillator
H =
b +
1
2
(5.3.3)
b, b =
, b
= 0 (5.3.4)
b, b
= 1 = bb
+ b
b (5.3.5)
where the above are now anti-commutator relations. From this,
[b, H] =
b,
b +
1
2
= (bb
b b
bb) ; bb
= 1 b
b
= (b 2b
bb ; bb +bb = 0 = b, b
= b b
b, b
= b (5.3.6)
If
H[n` =
n
[n` (5.3.7)
H(b[n`) = (bH b)[n`
(where [b, H] = bH Hb = b)
= (
n
)(b[n`) (5.3.8)
So also,
H(b
[n`) = (
n
+ )(b
[n`) (5.3.9)
For the energy of the system to be bounded below,
b[0` = 0 (5.3.10)
5.3. QUANTIZING THE DIRAC FIELD 93
For this state;
H[0` =
b +
1
2
[0`
=

2
[0`
0
=

2
(5.3.11)
Now,
b
[0` = [1` (5.3.12)

H[1` = (b
b + 1/2)[1`
= (b
b + 1/2)b
[0` (b
bb
= b
(bb
+ b
b b
b) = b
)
= (b
+ (1/2)b
)[0`
=
3
2
(b
[0`) =
3
2
[1` (5.3.13)
But (b
)
2
[0` = 0 as b
, b
= 0.
Nov. 3/99
Consider,
H =
b
1
2
(5.3.14)
b, b
= 1
b[0` = 0 , b
[0` = [1` , (b
)
2
[1` = 0
H[0` =
2
[0`
H[1` =

2
[1`
H
t
=
b +
1
2
= H
=
b bb
. .. .
=1
+bb
+
1
2
bb
1
2
(5.3.15)
There is a symmetry between b and b
;
b
[0`
t
= 0 (5.3.16)
b[0` = [1`
t
(5.3.17)
94 CHAPTER 5. (2
ND
Consequently though,
H[0`
t
=
b
1
2
[0`
t
=
b + bb
. .. .
1
bb
1
2
[0`
t
=

2
[0`
t
(5.3.18)
H[1`
t
=
b
1
2
(b[0`
t
)
=
2
[1`
t2
(5.3.19)
[0` [1`
t
(5.3.20)
[1` [0`
t
(5.3.21)
with the operator H, b destroys a state [1` while this operator b becomes a creation operator
for a system with Hamiltonian H
t
.
Now for the electron eld .
(x) =
s=
d
3
p
(2)
3/2
m
E
p
b(p, s)e
ipx
u(p, s) + d
(p, s)e
ipx
v(p, s)
(5.3.22)
(p m)u = 0 = (p + m)v (5.3.23)
s=
u
(p, s) u
(p, s) =
p + m
2m
(5.3.24)
s=
v
(p, s) v
(p, s) =
p + m
2m
(5.3.25)
u(p, s)v(p, s
t
) = 0 (5.3.26)
u(p, s)u(p, s
t
) = v(p, s)v(p, s
t
) =
ss
(5.3.27)
u
(p, s)u(p, s
t
) = v
(p, s)v(p, s
t
) =
E
p
m

ss
(5.3.28)
Quantizing
(i m) = 0 (5.3.29)
This can be derived from the Lagrangian
L =

(i m) (

=
0
) (5.3.30)
5.3. QUANTIZING THE DIRAC FIELD 95
provided and

are treated as being independent.
=
L
(
0
)
=

(i
0
)
= i
(5.3.31)
=
L
(
0

)
= 0 (5.3.32)
Eventually the constraint formalism gives,
(x, t),
(y, t)
quantum

(x, t),
(y, t)
Classical Dirac Bracket

(5.3.33)
The quantization of the E.M. eld gives
(x, t),
(y, t) = i
3
(x y) (5.3.34)
(x, t),
(y, t) =
3
(x y) (5.3.35)
As b, d
are now operators, we obtain,

b(p, s), b
(p
t
, s
t
) =
3
(p p
t
)
ss
d(p, s), d
(p
t
, s
t
) =
3
(p p
t
)
ss
Fermionic creation and

annihilation operators
(5.3.36)
b
(p, s) (b(p, s)) creates (destroys) an electron with energy E

p
= +
p
2
+m
2
and spin s,
momentum p.
d(p, s) (d
(p, s)) creates (destroys) a positron with energy E

p
=
p
2
+ m
2
, spin s and
momentum p. or
d(p, s) (d
(p, s)) annihilates (creates) a positron with energy E

p
= +
p
2
+ m
2
, spin s and
momentum +p. i.e. Now,
: H :=
d
3
x :
t
L
: (5.3.37)
Here :
:=
+
+
+
We nd that
H =
s=
d
3
p E
p
(p, s)b(p, s) +d
(p, s)d(p, s)
Positive Denite Hamiltonian

(5.3.39)
where this is a positive denite Hamiltonian because the b(p, s) destroys an electron of energy
E
p
, and the d(p, s) destroys a positron of momentum E
p
.
If we had quantized using a commutator, we would have gotten;
H =
s=
d
3
p E
p
(p, s)b(p, s) d
(p, s)d(p, s)
(5.3.40)
96 CHAPTER 5. (2
ND
where the negative sign would mean there is no lower bound on the spectrum of H. Consider
the current
j
Vector - eventually this will be the e.m. current. (5.3.41)

: Q :=: j
0
: = :

0
:
= :
:
=
d
3
p
(p, s)b(p, s) d
(p, s)d(p, s)
(5.3.42)
with b
(p, s)b(p, s) the positive contribution from electrons, and d
(p, s)d(p, s) term the neg-

ative contribution from positrons.
Nov. 6/99
(x),

(y) = i(i +m)(x y) (5.3.43)

[(x), (y)] = i(x y)
= 0 if (x y)
2
< 0 (5.3.44)
If we used commutators for
(x, t) and
(y, t) , [
(x),
(y)] = 0 for (x y)
2
< 0. This
would be inconsistent. So also,
'0[T
(x)
[0` = iS
F
(x y) (5.3.45)
S
F
(x y) =
d
4
p
(2)
4
e
ip(xy)
p
2
m
2
+ i
(p + m) (Scalar case) (5.3.46)
Since (p +m)(p m) =p
2
m
2
= p
2
m
2
, (last = due to anti-commutation relations)
S
F
(x y) =
d
4
p
(2)
4
e
ip(xy)
p m + i
(5.3.47)
For Fermi Dirac particles
Ta(t)b(t
t
) = (t t
t
)a(t)b(t
t
) (t
t
t)b(t
t
)a(t) (5.3.48)
Chapter 6
Interacting Fields
L =
1
2
(
)(
)
m
2
2

2
. .. .
Free eld

4
4!
....
Interaction
(6.0.1)
The equation of motion is
2
+ m
2
+

3
6
= 0 (6.0.2)
(Most important type of interaction Gauge).
6.1 Gauge Interaction
Schrodinger equation
i
t
= H complex (6.1.1)
Probability density
(x, t)(x, t) is invariant under a phase change

(x, t) e
i
(x, t) (6.1.2)
where is a constant. If (x, t) e
i(x,t)
(x, t) is an invariant then
is unaltered, but
i
t
= H is changed (6.1.3)
Introduce A
(x, t) and replace
by D
= ieA
, and let A
+
1
e
if e
i
.
i.e.
D
e
i
D
(6.1.4)
A
is the electromagnetic potential; becomes complex.

L = (
)(
) m
2
)
2
(6.1.5)
L = [(
+ ieA
] [(
ieA
) ] m
2
)
2
(6.1.6)
97
98 CHAPTER 6. INTERACTING FIELDS
This is invariant under
e
i
e
i
+
1
e
Also, add in
L =
1
4
F
(6.1.7)
where F
=
i
e
[D
, D
] (f) (f = test function)

=
i
e
[(
ieA
) , (
ieA
)] (f)
= A
,
A
,
(6.1.8)
We can also include spinors
L =

(i (
ieA
m) (6.1.9)
So also there is an interaction between spinors and real scalars:
L =

(i m) +
1
2
(
)(
)

2
2

4
4!
ig

(6.1.10)
where the last term is the yukawa interaction (the i is required for L = L
. (you must have

a complex scalar eld to interact with the electromagnetic eld).
6.2 Heisenberg Picture of Q.M.
A
H
= A
h
(x, t) (Heisenberg operator) (6.2.1)
i

t
A
H
= [A
H
, H] (6.2.2)
c.f.
A
t
= [A, H]
PB

1
i
[ ]
Commutator
(6.2.3)
Solution
A
H
(x, t) = e
iHt
A
H
(x, 0)e
iHt
(6.2.4)
Heisenberg states [`
H
t
[`
H
= 0
. Matrix elements
'[A
H
[`
H
= '[e
iHt
A
H
(x, 0)e
iHt
[`
H
(6.2.5)
6.2. HEISENBERG PICTURE OF Q.M. 99
Let A
H
(x, 0) = A
S
(x) Schrodinger operator
[
s
` = e
iHt
[`
H
= [(t)`
S
(6.2.6)
Now

t
A
s
(x) = 0
i

t
[(t)`
S
= H[(t)`
S
(6.2.7)
Diracs interaction representation,
A
ip
(x, t) = e
iH
0
t
A
S
(x)e
iH
0
t
(6.2.8)
H = H
0
+H
I
= e
iH
0
t
e
iHt
A
H
(x, t)e
iHt
e
iH
0
t
[a(t)` = e
iH
0
t
[a(t)`
S
(6.2.9)
Thus '
H
[A
H
[
H
` = '
S
[A
S
[
S
`
= '
ip
[A
ip
[
ip
` (6.2.10)
We note that
iA
ip
t
= i

t
e
iH
0
t
A
S
e
iH
0
t
= [A
ip
, H
0
] (6.2.11)
(A
ip
(x, t) evolves as if there were no interactions). So, also
i

t
[a(t)`
ip
= i

t
e
iH
0
t
[a(t)`
S
H
0
e
iH
0
t
[a(t)`
S
+e
iH
0
t
i

t
[a(t)`
S
= H
0
e
iH
0
t
[a(t)`
S
+e
iH
0
t
(H
0
+ H
I
) [a(t)`
S
=
e
iH
0
t
H
I
e
iH
0
t
e
iH
0
t
[a(t)`
S
= H
ip
I
[a(t)`
ip
(6.2.12)
Nov. 9/99
Now, drop the ip (were always working in the interaction picture from now on). So,
i

t
[a(t)` = H
I
[a(t)`
i

t
A(t) = [A(t), H
0
]
(6.2.13)
is in the interaction picture. Suppose,
[a(t)` = U(t, t
0
)[a(t
0
)` (6.2.14)
where U(t, t
0
) is an evolution operator. Note:
1. U(t, t) = 1
2. U(t, t
0
)
1
= U(t
0
, t)
3. By (6.2.13),
i
t
U(t, t
0
)[a(t
0
)` = H
I
U(t, t
0
)[a(t
0
)`
Notice [a(t
0
)` has no t dependence
i

t
U(t, t
0
) = H
I
(t)U(t, t
0
) (6.2.15)
Take the Hermitian conjugate
i

t
U
(t, t
0
) = U
(t, t
0
)H
I
(H
I
= H
I
) (6.2.16)
Thus, from
(6.2.15) + (6.2.16)U
(t, t
0
)U(t, t
0
)
= 0 (6.2.17)
and
U
(t, t
0
)U(t, t
0
) = 1 (i.e. = constant, let constant = 1) (6.2.18)
So we have
U
1
(t, t
0
) = U
(t, t
0
) (i.e. U is unitary) (6.2.19)
We can integrate (6.2.15)
U(t, t
0
) = K i
t
t
0
dt
t
H
I
(t
t
)U(t
t
, t
0
) (6.2.20)
As U(t, t) = 1, K = 1. We can iterate this equation,
U(t, t
0
) = 1 i
t
t
0
dt
1
H
I
(t
1
) + (i)
2
t
t
0
dt
1
t
1
t
0
dt
2
H
I
(t
1
)H
I
(t
2
)
+ (i)
3
t
t
0
dt
1
t
1
t
0
dt
2
t
3
t
0
dt
3
H
I
(t
1
)H
I
(t
2
)H
I
(t
3
) + . . . (6.2.21)
This satises (6.2.15). Examine: We can also write as:
(i)
2
2
t
t
0
dt
1
t
1
t
0
dt
2
H
I
(t
1
)H
I
(t
2
) +
t
t
0
dt
2
t
2
t
0
dt
1
H
I
(t
2
)H
I
(t
1
)
(6.2.22)
where the rst region of integration is as in diagram above, and the second region of in-
tegration is the same diagram with the region reected in the t
1
= t
2
line. But this is
also,
=
(1)
2
2
t
t
0
dt
1
t
t
0
dt
2
[TH
I
(t
1
)H
I
(t
2
)] (6.2.23)
6.2. HEISENBERG PICTURE OF Q.M. 101
(i)
2
t
t
0
dt
1
t
1
t
0
dt
2
H
I
(t
1
)H
I
(t
2
)
So also
(i)
3
t
t
0
dt
1
t
1
t
0
dt
2
t
2
t
0
dt
3
H
I
(t
1
)H
I
(t
2
)H
I
(t
3
)
=
(i)
3
3!
t
t
0
dt
1
t
t
0
dt
2
t
t
0
dt
3
TH
I
(t
1
)H
I
(t
2
)H
I
(t
3
)
In general, then,
U(t, t
0
) = 1 + T
(i)
1!
t
t
0
dt
1
H
I
(t
1
) + T
(i)
2
2!
t
t
0
dt
1
t
t
0
dt
2
H
I
(t
1
)H
I
(t
2
)
+ . . . + T
(1)
n
n!
t
t
0
dt
1
. . .
t
t
0
dt
n
H
I
(t
1
) . . . H
I
(t
n
)
= T exp
t
t
0
dt
1
H
I
(t
1
)
(6.2.24)
Consider now a scattering problem:
In this we go from a state [a(t = )) to a state at (a(t = +)[
At t = , we have free particles (Adiabatic Approximation)
At t = , we have a free particle state, [a(t = )), and at t = +, we have a free
particle state [b(t = +)), and what we want to compute is (b()[a()) - This gives the
amplitude for [a()) evolving into [b()). But,
[a()) = U(, ) [a()) (6.2.25)
This means we want,
(b()[ U(, ) [a()) = S
ba
(S-matrix) (6.2.26)
To evaluate this, we need,
(b()[
(i)
n
n!
T
t
t
0
dt
1
t
t
0
dt
2
. . .
t
t
0
dt
n
H
I
(t
1
) . . . H
I
(t
n
) [a()) (6.2.27)
6.3 Wicks Theorem
This converts Time-ordered products to normal-ordered products. (actual theorem works
for any two orderings of operators).
Trivial Case: (n=1)
TH
I
(t
1
) =: H
I
(t
1
) : (6.3.1)
n=2 (let t
1
> t
2
).
TH
I
(t
1
)H
I
(t
2
) = :
H
+
I
(t
1
) + H
I
(t
1
)

H
+
I
(t
2
) + H
I
(t
2
)
:
(lets let H
I
(t
1
) = H
1
, etc.)
= H
+
1
H
+
2
+ H
+
1
H
2
+ H
1
H
+
2
. .. .
Not N.O.d
+H
1
H
2
(6.3.2)
Now,
H
1
H
+
2
=
1
H
+
2
H
+
2
H
1
. .. .
[
H
1
,H
+
2
]
H
+
2
H
1
. .. .
N.O.d
(Bosonic)
H
1
H
+
2
+ H
+
2
H
1
. .. .
|H
1
,H
+
2

H
+
2
H
1
. .. .
N.O.d
(Fermionic)
(6.3.3)
Thus,
TH
1
H
2
=: H
1
H
2
: +
1
, H
+
2
(Bosonic)
1
, H
+
2
(Fermionic)
(6.3.4)
But this (anti)-commutator is a c-number. In general these are just '0[TH
1
H
2
[0`. i.e.
We have (t
1
> t
2
),
1
, H
+
2
= '0[H
1
H
+
2
[0`
= '0[TH
1
H
+
2
[0`
= '0[TH
1
H
2
[0` (6.3.5)
6.3. WICKS THEOREM 103
For both t
1
> t
2
and t
2
> t
1
, we have
TH
I
(t
1
)H
I
(t
2
) =: H
I
(t
1
)H
I
(t
2
) : +'0[TH
I
(t
1
)H
I
(t
2
)[0` (6.3.6)
Nov. 10/99
Inductively, we can show that
T(x
1
) . . . (x
n
) = : (x
1
)(x
2
) . . . (x
n
) :
+
i<j
(0[ T(x
i
)(x
j
) [0) : (x
1
) . . . (x
i1
)(x
i+1
) . . . (x
j1
)(x
j+1
) . . . (x
n
) :
+
all possible 2 pairs

(0[ T(x
i
)(x
j
) [0) (0[ T(x
k
)(x
l
) [0) : (x
1
) . . . (x
n
) :
. .. .
(x
ijkl
) all excluded
+. . .
+
all possible n/2 pairs

(0[ T . . . [0) . . . (0[ T . . . [0) : : (6.3.7)
This is useful, as
(0[ : (x
1
) . . . x
n
: [0) = 0 (6.3.8)
Consider the case where
H
I
=
d
3
x H(x) (6.3.9)
H(x) = i
5
( Pseudo-scalar Yakawa coupling ) (6.3.10)
i.e.
L =
1
2
(
)(
)

2
2

2
+

(i m) i
5
(6.3.11)
(Note that (x, t) = (+x, t)). Hence:
T exp
dt H
I
(t)
= T exp
d
4
xH
I
(x)
= T
1 i(i)
d
4
x (
(x)
5
(x)(x))
+
(+)
2
2!
d
4
x
1
d
4
x
2

(x
1
)
5
(x
1
)(x
1
)
(x
2
)
5
(x
2
)(x
2
)
+ . . .
(6.3.12)
Chapter 7
Electron-Positron Scattering
Now suppose [a()) consists of an electron of momentum p and polarization s, and a
positron of momentum q and polarization t; and suppose [a()) consists of an electron of
momentum p
t
and polarization s
t
, and a positron of momentum q
t
and polarization t
t
.
Thus, we have
[a()) = b
ps
d
qt
[0) [0) = b
ps
d
qt
[a()) (7.0.1)
[b()) = b
[0) (0[ b
p
s
d
q
t
= (b()[ (7.0.2)
Thus,
(b()[ Te
i
d
4
x1(x)
[a())
= (b()[ T
1 +
d
4
x

(x)
5
(x)(x) +

2
2!
d
4
x
1
d
4
x
2
. . .
[a()) (7.0.3)
Upon applying Wicks theorem, the only surviving contributions to this matrix element
will be those terms with destruction operators for e
and e
+
acting on [a()) and creation
operators for e
and e
+
acting on (b()[ . So,
(b()[ Te
i
d
4
x1(x)
[a())
= (b()[a()) +
(b()[
d
4
x :

(x)
5
(x)(x) : [a())
+ (b()[
d
4
x (0[ T

(x)
(x) [0)
: (x) : [a())
= 0 i.e. a
p
[a()) = 0 a
p
destruction op. for (x)
Term of order
2
: The only surviving term.
2
2
d
4
x
1
d
4
x
2
(b()[
(0[ T(x
1
)(x
2
) [0)
. .. .
i
F
(x
1
x
2
)
:

(x
1
)
5
(x
1
)
(x
2
)
5
(x
2
) :
[a())
(7.0.4)
105
106 CHAPTER 7. ELECTRON-POSITRON SCATTERING
The following (
+
,

+
) act on [a()) to give [0) by (7.0.1);
+
=
d
3
p
tt
(2)
3/2
m
E
p
=
b(p
tt
, s
tt
)u(p
tt
, s
tt
)e
ip
x
(7.0.5)
+
=
d
3
p
tt
(2)
3/2
m
E
p
=
d(p
tt
, s
tt
) v(p
tt
, s
tt
)e
ip
x
(7.0.6)
where the
+
represents e
(b operator) and the

+
represents e
+
(d operator). So also, by
(7.0.2),
+
=
d
3
p
tt
(2)
3/2
m
E
p
=
b(p
tt
, s
tt
)u(p
tt
, s
tt
)e
ip
x
(7.0.7)
+
=
d
3
p
tt
(2)
3/2
m
E
p
=
d(p
tt
, s
tt
) v(p
tt
, s
tt
)e
ip
x
(7.0.8)
which act on (b()[ to give (0[ . (

has b
in it e
.)
Nov. 12/99
So, the only surviving term is:
=

2
2
dx
1
dx
2
(0[ T(x
1
)(x
2
) [0) (
5
)
(
5
)
(b()[
(x
1
)
(x
1
)
(x
2
)
+
(x
2
)
+

(x
2
)
(x
2
)
(x
1
)
+
(x
1
)
(x
1
)
(x
2
)
(x
2
)
+
(x
1
)
(x
2
)
(x
1
)
(x
1
)
+
(x
2
)
[a()) (7.0.9)
where the rst two terms give identical contributions, and the last two terms give identical
contributions (the minus sign on the last two terms is due to Fermi-Dirac statistics). ex.
(x) =
s=
d
3
p
m
E
p
d
(p, s)v(p, s)e

ipx
and so on . . .
Recall
[a()) = d
[0)
[0) = d
q
t
b
p
s
[a())
So also,
(0[ T(x
1
)(x
2
) [0) =
i
(2)
4
d
4
k
e
ik(x
1
x
2
)
k
2
2
+ i
(7.0.11)
107
Thus,
S
ba
=
2
d
4
x
1
d
4
x
2
d
4
k
(2)
4
ie
ik(x
1
x
2
)
(k
2
2
+ i)
m
4
E
p
E
q
E
p
E
q
( u(p
t
, s
t
)
5
v(q
t
, t
t
)) ( v(q, t)
5
u(p, s)) e
i(p
x
1
+q
x
1
qx
2
px
2
)
( u(p
t
, s
t
)
5
u(p, s)) ( v(q, t)
5
v(q
t
, t
t
)) e
i(p
x
1
px
1
+q
x
2
qx
2
)
(7.0.12)
Integrate over x
1
and x
2
:
d
4
x
(2)
4
e
ikx
=
4
(k) (7.0.13)
S
ba
= i
2
m
2
E
p
E
q
E
p
E
q
d
4
k
1
k
2
2
i

( u
p
s

5
v
q
t
)( v
qt
5
u
ps
)
4
(k + p
t
+ q
t
)
4
(k q p)
( u
p
s

5
u
ps
)( v
qt
5
v
q
t
)
4
(k + p
t
p)
4
(k + q
t
q)
= i
2
m
2
E
p
E
q
E
p
E
q
4
(p
t
+ q
t
p q)
( u
p
s

5
v
q
t
)( v
qt
5
u
ps
)
(p
t
+ q
t
)
2
2
+ i
( u
p
s

5
u
ps
)( v
qt
5
v
q
t
)
(p
t
p)
2
2
+ i
(7.0.14)
Pictorial Representation
The relative minus sign in gure 7.0.2 comes from interchanging two fermions. We dont ac-
tually observe the intermediate particles, and so the fact that they are o mass shell doesnt
eect the calculation. After integrating over x
1
and x
2
, and going to momentum space, we
get
108 CHAPTER 7. ELECTRON-POSITRON SCATTERING
Figure 7.0.1: Electron Scattering
Figure 7.0.2: Electron Scattering (Momentum Space, after integra-
tion)
The integral over x
1
, x
2
leads to momentum conservation at each vertex (total momentum
incoming = total outgoing momentum).
Nov 16/99
Chapter 8
Loop Diagrams
8.1 Feynman Rules in Momentum Space
Propagator for spinor
i
p m + i
(Spin 0)
i
k
2
2
+ i
(Spin 0)
ig
+ (gauge dependent)
k
2
+ i
(Spin 1)
109
110 CHAPTER 8. LOOP DIAGRAMS
ie(
(Vertex)
(
5
)
(Vertex)
i
(Vertex)
momentum is conserved at each vertex
for each loop, there is an overall factor of
d
4
k
(2)
4
(This leads to innities).
Now, take an example:
One more photon added:
(i.e. many dierent possibilities)
example:
8.1. FEYNMAN RULES IN MOMENTUM SPACE 111
Figure 8.1.1: we only ever observe the
external particles dont know where
particles in loop are must sum over all
possibilities Uncertainty Principle.
(p + q
t
= q +p
t
). The External lines (spinors) give
m
E
p
V
u(p, s)
. .. .
incoming e
momentum p, spin s
m
E
p
V
u(p, s)
. .. .
outgoing e
m
E
p
V
v(p, s)
. .. .
Outgoing e
+
m
E
p
V
v(p, s)
. .. .
incoming e
+
(8.1.1)
where
1
V
unit probability in a box of volume V .
Factor of (1) for each fermion loop
If we exchange fermions to go from one diagram to another, their relative phase is
(1).
Overall factor of (2)
4
(p
i
p
f
)
For example, e
e
+
e
e
+
(with respective momenta p, q p
t
, q
t
.)
Consider one particular pair of diagrams, which are of order e
4
.
Figure 8.1.2: For Diagram I
(1)
d
4
k
(2)
4
m
E
p
V
u
(p
t
, s
t
)
(ie(
m
E
p
V
u
(p, s)
ig
(p p
t
)
2
+ i
(ie
i
(k+ p p
t
) m + i
(ie
i
k m + i
ig
(p p
t
)
2
+ i
m
E
q
V
v
(q, t)(ie
m
E
q
V
v
(q
t
, t
t
)
= (1)
d
4
k
(2)
4
(2)
4
(p +q p
t
q
t
)
m
2
E
p
E
p
E
q
E
q
1
V
2
[ u(p
t
, s
t
)
u(p, s)]
1
(p p
t
)
2
+ i
2
Tr
1
(k+ p p
t
) m + i
1
k m+ i
v(q, t)
v(q
t
, t
t
) (8.1.2)
Note that (1) the (1) is the fermion loop factor, and (2) the spinor propagator is
,p+m
p
2
m
2
+i

1
,pm+i
Thus II becomes
(1)(1)
d
4
k
(2)
4
. . . (u(p, s) v(q
t
, t
t
)) (8.1.3)
where the second (1) will come from interchanging the spinors.
8.2. COMBINATORIC FACTORS 113
Figure 8.1.3: For Diagram II
8.2 Combinatoric Factors
(Always (+1) in QED, as all particles entering a vertex are distinguishable).
e
(3 distinct elds)
But, not so lucky in other theories - for example
L =
1
2
(
)
2

2
2

2
. .. .
Scalar Propagator
4!
4
(8.2.1)
with the 4! in the
4
term to take into account the indistinguishability of the 4 s. The
Feynman diagrams are:
Consider the one-loop correction to (p)(q) (p
t
)(q
t
), above. There are three possible
diagrams that have the same topology; note that there is no factor of (1) here, though, as
these are scalars (bosons) here.
Nov. 17/99
External Photons; recall, for electrons there was the factor
e
m
E
p
V
u(p, s) (8.2.2)
i
k
2
2
+ i
Scalar Propagator
i
Vertex
Figure 8.2.1: 1-loop correction to
Symmetry Factors
recall the diagram for , gure 8.2.1 above.
As shown in gure 8.2.2, the p leg can attach to either vertex, on any one of the eight dier-
ent legs, then the p
t
leg can attach to the same vertex to any of the 3 remaining legs. After
that the q leg can attach to any of the 4 legs of the remaining vertex, followed by the q
t
leg
attaching to any of the 3 legs of that vertex; nally, one of the remaining legs on the rst
8.3. CROSS SECTIONS FROM MATRIX ELEMENTS 115
(k, )
2
k
V
e
Figure 8.2.2: Symmetry Factor diagram (example)

vertex can attach to either remaining leg on the other vertex, giving in the numerator of the
symmetry factor (8 3) (4 3) (2). In the denominator, we must take into account that
there are 4 possible legs on each vertex (4! 4!), and that we can switch vertices (2!); thus,
the symmetry factor is
(8 3)(4 3)(2)
4! 4! 2!
=
1
2
= Symmetry factor (8.2.3)
8.3 Cross Sections From Matrix elements
Consider e
p e
p , (p = proton). The proton is treated as a heavy electron with

opposite charge. To lowest order,
Figure 8.3.1: Electron-Proton scattering
S
fi
= (a()[b()) =
m
c
f
V
u(p
f
, s
f
)
(ie
m
c
i
V
u(p
i
, s
i
)
ig
(p
f
P
i
)
2
+ i
M
E
f
V
u(P
f
, s
f
)
(+ie
M
E
i
V
u(P
i
, s
i
)
(2)
4
4
(p
i
+ P
i
p
f
P
f
) (8.3.1)
(For e
, we would have to account for the fact that we can exchange e
s (identical
particles). - also note the +
tt
in front of the second e above - this is because the proton
is positive.) The probability of this transition occurring per unit volume of space per unit
time is
[S
fi
[
2
V T
= W
fi
(8.3.2)
Note:
4
(p
i
+ P
i
p
f
P
f
)
2
=
4
(0)
4
(p
i
+ P
i
p
f
P
f
)
As (p) =
dx
(2)
e
ipx
(0) =
dx
(2)

V
2
4
(p
i
+ P
i
p
f
P
f
)
2
=
V T
(2)
4
4
(p
i
+ P
i
p
f
P
f
) (8.3.3)
8.3. CROSS SECTIONS FROM MATRIX ELEMENTS 117
Thus,
W
fi
=
Kinematics
. .. .
(2)
4
4
(p
i
+P
i
p
f
P
f
)
m
2
M
2
V
4
E
i
E
f
c
i
c
f
(ie)(ie)( u
p
f
s
f
u
p
i
,s
i
)
i
(p
f
p
i
)
2
+ i
( u
P
f
s
f
u
P
i
s
i
)
2
. .. .
[M
fi
[
2
(8.3.4)
The number of states in a volume d
3
p
f
, d
3
P
f
is going to be
(V d
3
p
f
)(V d
3
P
f
)W
fi
(8.3.5)
Divide through by J
inc
1
V
, where J
inc
= ux of incident particles = # of particles per unit
area that run by each other per unit time for collinear beams.
J
inc
=
[v
i
V
i
[
V
(8.3.6)
1
V
is the number of target particles per unit volume. (as
i
m
EV
u(p, s) ). The cross section
is thus
d
fi
= V
2
d
3
p
f
d
3
P
f
(W
fi
)
V
J
inc
= d
3
p
f
d
3
P
f
(2)
4
4
(p
i
+ P
i
p
f
P
f
)
m
2
M
2
E
i
E
f
c
i
c
f
1
[v
i
V
i
[
[M
fi
[
2
(8.3.7)
Note that
mM
E
i
c
i
[v
i
V
i
[
=
mM
(p
i
P
i
)
2
m
2
M
2
(Lorentz invariant) (8.3.8)
So also,
d
3
p
(2E
p
)
=
dp
0
(p
2
m
2
)(p
0
)d
3
p also Lorentz invariant (8.3.9)
Thus d is Lorentz invariant.
For unpolarized beams,
Average over the polarizations of the incoming particles
Sum over outgoing polarizations
i.e.
1
2
s
i
1
2
S
i
s
f
S
f
d
fi
(8.3.10)
Remember that,
s=
u
(p, s) u
(p, s) =
p + m
2m
(8.3.11)
s=
v
(p, s) v
(p, s) =
p m
2m
(8.3.12)
Thus,
1
4
s
i
,S
i
s
f
,S
f
[M
fi
[
2
=
1
4
Tr
p
f
+ m
2m

p
i
+ m
2m

Tr
P
f
+M
2M

P
i
+ M
2M

e
4
(p
f
p
i
)
4
(8.3.13)
Recall the relations:
Tr[a b] = 4a b (8.3.14)
Tr[a b c] = 0 (8.3.15)
Tr[a b c d] = 4 (a bc d a cb d + a db c) (8.3.16)
And so
1
4
s
i
,S
i
s
f
,S
f
=
e
4
2m
2
M
2
(p
f
p
i
)
2
P
f
p
f
P
i
p
i
+P
f
p
i
P
i
p
f
m
2
P
f
P
i
M
2
p
f
p
i
+ 2m
2
M
2
(8.3.17)
Nov. 19/99
d =
mM
(p
i
P
i
)
2
m
2
M
2
(2)
4
4
(p
i
+ P
i
p
f
P
f
)
md
3
p
f
(2)
3
c
f
Md
3
P
f
(2)
3
E
f
[M
fi
[
2
(8.3.18)
Consider,
d
3
p
f
= d [p
f
[
2
d[p
f
[
= d [p
f
[
2
c
f
dc
f
(8.3.19)
In the frame of reference where
p
f
= (p
f
, c
f
) (8.3.20)
p
i
= (p
i
, c
i
) (8.3.21)
P
f
P
i
= (0, M) Proton at rest during this scattering (8.3.22)
Note that if p
i
p
f
= [p
i
[[p
f
[ cos , then
q
2
= (p
f
p
i
)
2
= 4c
i
c
f
sin
2
(8.3.23)
8.4. HIGHER ORDER CORRECTIONS 119
Hence,
d =

2
2
m
2
c
i
c
f
sin
4
1 + sin
2
1
2
q
2
M
2
1
(8.3.24)
where =
e
2
4
if q
2
<M
2
. The rst term (the 1) above is the classical Rutherford term.
8.4 Higher order corrections
Figure 8.4.1: Higher order corrections
Figure 8.4.2: Gener-
ates coulomb eld
classically
The loop (solid) is generated as the e
travels - it Polarizes - the e
e
+
created in
vacuum.
The Loop contribution is
Figure 8.4.3: Loop contribution
I = (1)
d
4
q
(2)
4
Tr
(ie
)
i
(k+ q) m + i
(ie
i
q m + i
= e
2
d
4
q
(2)
4
Tr
((k+ q) + m)
(q + m)
((k + q)
2
m
2
)(q
2
m
2
)
= e
2
d
4
q
(2)
4
Tr
(k+ q)
q + m
2
((k + q)
2
m
2
)(q
2
m
2
)
= 4e
2
d
4
q
(2)
4
(k + q)
(k +q) qg
+ (k + q)
+ m
2
g
((k + q)
2
m
2
)(q
2
m
2
)
(8.4.1)
Now, use the relation:
1
ab
=
1
0
dx
1
[ax + b(1 x)]
2
(8.4.2)
example:
dt
t(t + 1)
=
dt
1
0
dx
1
[tx + (t + 1)(1 x)]
2
=
dt
1
0
dx
1
[t + (1 x)]
2
=
1
0
dx
1
(t + (1 x))
+ K
= ln [t + (1 x)]
1
0
+ K
= ln(t) ln(t + 1) + K
(8.4.3)
8.4. HIGHER ORDER CORRECTIONS 121
I = 4e
2
d
4
q
(2)
4
Quadratically divergent
. .. .
(k + q)
+ (k + q)
+ ((k + q) q + m
2
)g
1
0
dx
1
[x((k + q)
2
m
2
) + (1 x)(q
2
m
2
)]
2
= 4e
2
d
4
q
(2)
4
N
1
0
dx
1
[q
2
+ 2k qx
. .. .
(q+kx)
2
k
2
x
2
+k
2
x m
2
]
2
Shift variable of integration (not straightforward) to q
t
= q + kx.
I = 4e
2
d
4
q
t
(2)
4
1
0
dx
1
[q
t2
+ k
2
x(1 x) m
2
]
2
[(k + q
t
xk)
(q
t
xk)
+(k + q
t
xk)
(q
t
xk)
+ ((k + q
t
xk) (q
t
xk) + m
2
)g
We can drop terms odd in q (because its integrating over odd terms over an area symmetric
about origin).
I = 4e
2
d
4
q
(2)
4
1
0
dx
2q
2x(1 x)k
+ (q
2
+ x(1 x)k
2
+ m
2
)g
[q
2
+ k
2
x(1 x) m
2
]
2
where
q
=
1
4
g
q
2
d
4
q f(q
2
)q
= Ag
multiply by g
d
4
q f(q
2
)q
2
= 4A
d
4
q f(q
2
)q
=
g
d
4
q f(q
2
)q
2
1
ab
=
1
0
dx
1
[ax + (1 x)b]
2
(8.4.5)
1
abc
=
1
0
dx
1
[xa + (1 x)b]
2
1
c
take
d
da
of (8.4.5)
1
a
2
b
=
1
0
dy
2y
[ay + (1 y)b]
3
1
abc
=
1
0
dx
1
0
dy
2y
y[ax + (1 x)b] + (1 y)c
3
(8.4.6)
Nov. 23/99
In general (where (x) =
0
dt t
x1
e
t
),
1
D
a
1
1
D
a
2
2
. . . D
a
n
n
=
r(a
1
+ a
2
+ . . . +a
n
)
(a
1
)(a
2
) . . . (a
n
)

1
0
dx
1
1
0
dx
2
. . .
1
0
dx
n
(1 x
1
x
n
)x
a
1
1
1
x
a
n
1
n
[x
1
D
1
+ x
2
D
2
+. . . + x
n
D
n
]
2
(8.4.7)
Recall the loop contribution from last time: gure 8.4.3, with equation
I = 4e
2
d
4
q
(2)
4
1
0
dx
[
1
2
g
q
2
2k
x(1 x) + (q
2
+x(1 x)k
2
+ m
2
)g
[q
2
+x(1 x)k
2
m
2
]
2
(8.4.8)
How do we make sense out of a divergent integral? We rst insert some parameter into the
theory to render this nite in a way consistent with the symmetries of the theory (Regular-
ization). Then absorb this parameter into the quantities that characterize the theory. (mass,
couplings, external wave functions, etc.) (Renormalization).
8.5 Renormalization
Recall the mass on a string problem (see gure 10.4.1):
The Lagrangian for this system is characterized by m, k, x
0
;
L =
1
2
m x
2
1
2
k(x x
0
)
2
(8.5.1)
8.6. REGULARIZATION 123
Now, if we turn on a gravitational eld,
L =
m x
2
2

k(x x
0
)
2
2
+ mgx
=
m
2
x
2
k
2
x x
0
mg
k
2
+
k
2
mg
2
2x
0
mg
k
. .. .
Constant
=
m x
2
2

k
2
(x x
t
0
)
2
+ constant (8.5.2)
x
t
0
= x +
mg
k
x
0
has been renormalized by an amount
mg
k
!
8.6 Regularization
1. Cut-o ()
d
4
q
[q[<
d
4
q (8.6.1)
Inserting the cuto destroys gauge invariance.
S =
d
4
x
1
4
F
2
+

(i m)
(8.6.2)
[x[ >
1
is not gauge invariant

2. Pauli Villars
S S
PV
=
d
4
x
1
4
F
+

(i m) +

(iM)
(8.6.3)
(a) Bosonic spinor
(b) M
(c) Still have gauge invariance
Now we must consider
(gure 8.6.1)
= 4e
2
d
4
q
(2)
4
1
0
dx
1
2
g
q
2
2k
x(1 x) + (q
2
+x(1 x)k
2
+ m
2
)g
[q
2
+ x(1 x)k
2
m
2
]
2
1
2
g
q
2
2k
x(1 x) + (q
2
+x(1 x)k
2
+ M
2
)g
[q
2
+ x(1 x)k
2
M
2
]
2
(8.6.4)
This is Finite! (Divergence reappears as M
2
. The dependence on M
2
disappears when
we renormalize. Combine terms using
1
a
n

1
b
n
= n(a b)
1
0
dx
1
[xa + (1 x)b]
n+1
(8.6.5)
let y = x(a b) + b
= n(a b)
b
a
dy/(a b)
y
n+1
= n
y
n
n
a
b
=
1
a
n

1
b
n
= LHS (8.6.6)
Actually computing the integrals involves the following general integral. In n dimensions
(n = 1, 2, 3, . . .) (in Minkowski space);
d
n
q
(2)
n
(q
2
)
a
(q
2
m
2
)
b
=
i
(4)
n/2
(1)
ab
(m
2
)
n/2+ab
n
2
+ a
b a
n
2
(a)(b)
(8.6.7)
Note:
dq
0
dq
1
. . .
dq
n1
(q
2
)
a
(q
2
m
2
)
b
(8.6.8)
q
2
= q
2
0
q
2
1
q
2
2
. . . q
2
n1
Let q
0
= iq
n
(Wick Rotation - i.e. Cartesian co-ordinates in Minkowski space, giving Eu-
clidean space)
q
E
= q
2
n
+ q
2
1
+ . . . + q
2
n1
So (8.6.8) becomes
= i
dq
n
. . .
dq
n1
= i(1)
ab
d
n
q
e
(q
2
E
)
a
(q
2
E
+ m
2
)
b
(8.6.10)
Figure 8.6.2: Integral around contour is
zero -
dq
0

i
i
dq
0
q
n
= q cos
1
q
1
= q sin
1
cos
2
q
2
= q sin
1
sin
2
cos
3
.
.
.
q
n1
= q sin
1
sin
2
. . . sin
n1
q
2
= q
2
1
+ q
2
2
+ . . . q
2
n1
+ q
2
n
i.e.
d
n
q
E
=
0
dq q
n1
2
0
d
1

0
d
2
sin
2

0
d
3
sin
2
3
. . .

0
d
n1
sin
n2
n1
(8.6.11)
Nov. 24/99
I =
d
n
q
(2)
n
(q
2
)
a
(q
2
m
2
)
b
= i(1)
ab
d
N
q
E
(2)
n
(q
2
E
)
a
(q
2
E
+ m
2
)
b
= i(q)
ab

0
dq q
n1
2
0
d
1

0
d
2
sin(
2
) . . .

0
d
n1
sin
n2
(
n1
)
(q
2
)
a
(q
2
+m
2
)
b
(8.6.12)
But,

0
d sin
m
=
m+1
2
m+2
2

0
dt
t
m1
(t
2
+ a
2
)
n
=
(m)(n m)
(a
2
)
nm
(n)
I =
i
(4)
n/2
(m
2
)
n/2+ab
n
2
+ a
n
2
b a
n
2
(b)
(8.6.13)
This is held to be true for n non-integer valued also. Substitute this into the integral for the
Pauli Villars loop ( gure 8.6.1).
1
a
n

1
b
n
= n(a b)
1
0
dx
1
[xa + (1 x)b]
n+1
(8.6.14)
i.e. Formally, (n = 4, a = 1, b = 2)
d
4
q
(2)
4
q
2
[q
2
+ x(1 x)k
2
m
2
]
2
=
i
(2)
2
m
2
x(1 x)k
2
2+12
(2 + 1)
(2)
(2 1 2)
(2)
(8.6.15)
But (1) is at a pole of (x). ((x) has poles at x = 0, 1, 2, 3, . . .) We can just
let the number of dimensions be n, not 4. We have,
I
PV
=
d
n
k
(2)
n
q
2
[q
2
+ x(1 x)k
2
m
2
]
2
=
i
(4)
n/2
(m
2
x(1 x)k
2
)
n/2+12
n
2
+ 1
2 1
n
2
n
2
(2)
=
i
(4)
n/2
m
2
x(1 x)k
2
n/21

n
2
+ 1
n
2
1
n
2
(2)
Now let = 2
n
2
0 as n 4
n
2
+ 1
=
n
2
n
2
I
PV
=
i
(4)
2
(m
2
x(1 x)k
2
)
+1
(2 )( 1) (8.6.16)
But,
(1 + ) = exp
n=2
(1)
n
n
(n)
n
(8.6.17)
(n) =
k=1
1
k
n
= lim
N
k=1
1
k

N
1
dt
t
= 0.577 . . .
From this,
() =
(1 + )
2
=
1
O() (8.6.18)
a
= e
ln a
= 1 + ( ln a) +

2
ln
2
a
2!
+ . . . (8.6.19)
Thus our integral becomes
I
PV
=
i
(4)
2
(m
2
x(1 x)k
2
)
(1 + ln(4)) (1 (m
2
x(1 x)k
2
))
1

2
1
1
+ . . .
=
2i
(4)
2
(m
2
x(1 x)k
2
)
+ ln(4) ln(m
2
x(1 x)k
2
) +
1
2
+O()
Note that the rst term
1
will be a pole as 0 (other terms nite). This is only a portion

of
Combine this with the integral for
and will see that all terms of order
1
will cancel.
We note that as M
2
m
2
, the combination of these two diagrams (recall gure 8.6.1)
is,
I
(k) =
2i
(k
k
2
)
1
0
dx x(1 x) ln
M
2
m
2
x(1 x)k
2
(8.6.20)
where =
e
2
4c
(note that the ln still diverges as M
2
.) Note that the result is
proportional to k
k
2
.
Now, photon propagator is
i(g
+ gauge dependant parts)

k
2
+ i
(8.6.21)
gauge dependant parts k
. But,
k
= k
(k
k
2
)f(k
2
)
= (k
2
k
k
2
)f(k
2
)
= 0
Thus gauge dependant parts of the photon propagator dont contribute to the physical
process.
This form of I
does not arise if we were to use cutos (i.e. cutos are not gauge
invariant.) Note also that the dependance on M
2
is logarithmic. i.e.

d
4
q[Ag
q
2
+ (Bx(1 x)k
2
+ Cm
2
)g
]
[q
2
+x(1 x)k
2
m
2
]
2
Would expect quadratic divergence M
2
. (M
2
terms all cancelled.) Quadratic divergences
proportional to g
M
2
all cancel. still logarithmically divergent.
Nove 26/99
Elimination of ln M
2
. Remember the photon propagator itself. (
2
some mass scale
introduced arbitrary (important)) Write
I
=
i
3
q
2
g
ln
M
2
1
0
dz z(1 z) ln
m
2
q
2
z(1 x)
=
i
3
q
2
q
2
. .. .
g
T
ln
M
2
+ (q
2
, m
2
2
)
(8.6.22)
= (ie
ig
q
2
+
ig
q
2
iq
2
3
g
T
ln
M
2
+

ig
q
2
ig
q
2
iq
2
3
g
T
ln
M
2
ig
q
2
iq
2
g
T
ln
M
2
ig
q
2
+ . . .
(ie
) (8.6.23)
This is a geometric series, a + ar + ar
2
+ . . . =
a
1r
, [r[ < 1. So,
= (ie
ig
q
2
+
ig
q
2

3
g
T
ln
M
2
+
ig
q
2

3
g
T
ln
M
2
2
+
2
+ . . .
(ie
)
= (ie
ig
T
q
2
1
1 +

3
ln
M
2
+ (q
2
, m
2
,
2
)
(ie
)
=
e
2
1 +

3
ln
M
2
ig
T
q
2
q +
m
3
1+

3
ln
M
2
((q
2
, m
2
,
2
))
(8.6.24)
(neglecting terms of order
2
=
e
2
4
2
). Now, Dene:
e
2
R
=
e
2
1 +

3
ln
M
2
2
e
2
1

3
ln
M
2
(8.6.25)
(Renormalized Charge). Thus, continuing (8.6.24)
= e
2
R
ig
T
q
2
1 +

R
3
(q
2
, m
2
,
2
)
(8.6.26)
We now let M
2
.
Note that there is still a dependence on
2
. (Changes in
2
give rise to corresponding
changes in
R
.
Also, in
g
T
= g
q
2
(ie
R
q
2
(ie
) gives zero provided the ends of the electron legs are on

shell.
Also
(q
2
,
2
, m
2
) = 6
1
0
dz z(1 z) ln
m
2
q
2
z(1 z)
is such that q
2
(q
2
, m
2
,
2
) 0 as q
2
0. Hence,
still has a pole at q
2
= 0. Thus the photon is massless.
In 2-D, if m
2
= 0, (q
2
,
2
) is nite, and (q
2
) =
e
2
q
2
, then the above diagram sum
becomes
=
e
2
g
T
q
2
1
e
2
q
2
=
e
2
g
T
q
2
e
2
8.7. NOETHERS THEOREM 131

Thus the photon develops a mass in 2-D due to the radiative corrections. (e has dimensions
of mass in 2-D). i.e.
S =
d
2
xL
S
....
[]=0
=
d
2
x
....
2
1
4
....
1
A
....
0
2
+

....
1
2
(
....
+1
i e
....
1
A)
(8.6.29)
8.7 Noethers Theorem
For every innitesimal change which leaves the action invariant there exists a conserved cur-
rent which leads to a conserved charge.
Change of type I
L = L(
A
(x),
A
(x)) (8.7.1)
x
x
t
= x
(8.7.2)
A
(x)
t
A
(x) =
A
+
AB
B
(x) (8.7.3)
with
AB
innitesimal. (ex. we had a gauge transformation
(x) e
i(x)
(x) = lim
N
1 +
i
N
N
(x) (1 + i (x)
....
AB
)(x)
Now,
S =
d
4
x L(
A
,
A
)
=
d
4
x L(
t
A
,
t
A
) (8.7.4)
where both L(
A
,
A
), L(
t
A
,
t
A
) have the same functional dependence. (For example,
f(x, y) = x
2
+ y
2
; let x
t
= xcos y sin , y
t
= xsin + y cos , and we still get f(x
t
, y
t
) =
x
t2
+ y
t2
(invariant) - whereas g(x, y) = xy = g(x
t
, y
t
) ).
Nov 30/99
Suppose L is invariant under an innitesimal translation x
.
S =
d
4
x L((x),
(x)) (8.7.5)
=
d
4
x
t
L((x
t
),
t
(x
t
)) (8.7.6)
in the same way as
S =
d
4
x L(x) =
d
4
x
t
L(x
t
) (8.7.7)
d
4
x
t
=
(x
t1
. . . x
t0
)
(x
1
. . . x
0
)
d
4
x = d
4
x
Now,
L = L(x
t
) L(x)
= L(x
) L(x
)
=
L
x
(8.7.8)
or
= L((x
t
),
t
(x
t
)) L((x),
(x)) (8.7.9)
If now,
(x) = (x
t
) (x) = (x
) (x) =
(8.7.10)
then
L = (x)
L
(x)
+ (
(x))
L
(
(x))
=
L
(x)
+

x
L
(
(x))
(8.7.11)
However, we also have the equations of motion.
d
4
xL((x),
(x)) = 0 (8.7.12)
=
cl
+ (8.7.13)
i.e.
d
4
x[L(
cl
+ ,
cl
+
)]
=
d
4
x
L(
cl
,
cl
) +
L(
cl
,
cl
)
+ (
(x))
L(
cl
,
cl
)
(
)
+ . . .
If 0 as x
d
4
x
L(
cl
,
cl
) +(x)
L(
cl
,
cl
)
L(
cl
,
cl
)
(
(x))
. .. .
=0 if
cl
is extremum of S
+. . .
(8.7.14)
Hence as is arbitrary,
0 =
L
L
()
(8.7.15)
on the classical path. Recall (8.7.11),
L =
L
(
+

x
L
(
)
=

x
L
(
(8.7.16)
(8.7.8) and (8.7.16) must be identical. As a result,
L
x
=

x
(g
L)
=

x
L
(
Thus
0 =

L +

x
L
(
. .. .
T
Energy-Momentum Tensor
(8.7.17)
Or,
= 0 (8.7.18)
Thus, there are 4 conserved currents, ( = 1, 2, 3, 0), one for each invariance. This only
holds if that it depends on is a solution to the classical equation. (ex. k
(g
k
2
k
) = 0
always, but k
(g
m
2
k
) = 0 only if k
2
= m
2
. )
There are 4 conserved currents; one fore each direction in which we have an innitesimal
invariance.
For
T
00
= g
00
L +

x
0
L
(
0
)
= L +

t
= H
(H = L + p q)
Now, if
= 0 (8.7.19)
Then,
d
3
x
= 0
0 =
d
3
x
j +

t
j
0
dS j +

t
d
3
x j
0
(x, t) ( by Gauss)
The rst term is 0 if j 0 as [x[ . Let Q(t) =
d
3
x j
0
(x, t).

t
Q(t) = 0 (Charge conservation) (8.7.20)
For = 0,
Q(t) =
d
3
x T
00
(x, t)
=
d
3
x H (8.7.21)
(The charge here is the energy of the eld.) For = i
d
3
xT
0i
(x, t) = P
i
(t) Mechanical momentum of eld (8.7.22)
Innitesimal internal transformations
x
x
t
= x
L = L(
A
(x),
A
(x)) (8.7.23)
Invariance under
A

t
A
=
A
+
AB
B
(8.7.24)
(ex. A
, (1 + i) Gauge transformation.)
S =
d
4
x L(
A
,
A
) =
d
4
x L(
t
A
,
t
A
)
=
d
4
x
L(
A
,
A
) +
AB
B
L
A
+ (
AB
B
)
L
(
A
)
+ . . .
(8.7.25)
Also if
A
satises
/
A
=
/
(
A
)
.
Hence,
0 =
d
4
x
(
AB
B
)
L
(
A
)
+ (
AB
B
)
L
(
A
)
0 =
d
4
x

x
AB
B
L
(
A
)
(8.7.26)
Thus we have a conserved current,
J
AB
B
L
(
A
)
(8.7.27)
ex. if L =

(i A)
1
4
F
2
, then j
.
Dec. 1/99
x
t
= x
t
A
(x) =
A
(x) +
AB
B
(x)
Consider a general transformation in which,
x
t
= x
+ x
(8.7.28)
(x
t
) = (x) + (x) (8.7.29)
ex. consider a rotation of a vector eld A
i
(x
i
) in 2-D.
x
t
= xcos y sin x y for small
y
t
= xsin + y cos x +y
x = y ( small )
y = x ( small )
At the same time,
A
t
x
(x
t
, y
t
) = A
x
(x, y) cos A
y
(x, y) sin
A
t
y
(x
t
, y
t
) = A
x
(x, y) sin +A
y
(x, y) cos
A
x
= A
t
x
(x
t
, y
t
) A
x
(x, y)
= A
t
x
(x
t
, y
t
) A
x
(x
t
, y
t
)
. .. .
+A
x
(x
t
, y
t
) A
x
(x, y)
. .. .
=

A
x
(x
t
, y
t
) + [A
x
(x y, y + x) A
x
(x, y)]
=

A
x
(x
t
, y
t
) +
y

x
+ x

y
A
x
(x, y) (8.7.30)
Invariance of the action S
S =
d
4
x L((x),
(x))
=
d
4
x
t
L(
t
(x
t
),
t
t
(x
t
))
In general, d
4
x
t
= [J[d
4
x,
[J[ =
det
(x
t1
. . . x
t0
)
(x
1
. . . x
0
)
det
x
1
x
1
. . .
x
1
x
0
.
.
.
.
.
.
x
0
x
1
. . .
x
0
x
0
det
1 +
x
1
x
1
x
1
x
2
. . .
x
0
x
1
.
.
.
1 +
x
2
x
2
.
.
.
.
.
.
1 +
x
0
x
0
1 +
x
+O
(x
)
2
(8.7.31)
So also, if we call
L
t
(x
t
) = L(
t
(x
t
),
t
t
(x
t
)) (8.7.32)
L
t
(x
t
) L(x) = L
= L
t
(x
t
) L(x
t
)
. .. .
/(x)
+L(x
t
) L(x)
. .. .
x
L(x)
x
(8.7.33)
Now,
L(x) = L
t
(x
t
) L(x
t
)
= L(
t
(x
t
),
t
t
(x
t
)) L((x
t
),
t
(x
t
))
(Re:
t
(x
t
) (x
t
) =

(x
t
))
= L((x
t
) +

(x
t
),
t
(x
t
) +
t
(x
t
)) L((x
t
),
t
(x
t
))
Hence,
L =

(x
t
)
L
(x
t
)
+
t
(x
t
)
L
(
t
(x
t
))
(8.7.34)
By the equations of motion,
L =

(x
t
)

x
L
(
t
(x
t
))
+ (
t
(x
t
))
L
(
t
(x
t
))
=

x
t
(x
t
)
L
(
t
(x
t
))
(8.7.35)
So, putting together,
0 =
d
4
x
t
L(
t
(x
t
),
t
t
(x
t
))
d
4
x L((x), (x)) ( subs. in (8.7.31), (8.7.33), (8.7.35))
=
d
4
x
1 +
x

L((x),
(x)) +

L + x
L
x
d
4
x L((x),
(x))
To leading order,
0 =
d
4
x
L + x
L
x
+

x
(x)
L
(
(x))
d
4
x
L +

L
(
(x))
As x
and

contain an arbitrary parameter, then
= 0 (8.7.36)
where
j
= x
L +

L
(
)
(8.7.37)
(Noether Current).
Suppose
x
= x
t
x
= A
j
(x)
j
(translation:
A
(x) =
t
A
(x
t
)
A
(x)
= B
Aj
(x)
j
But
A
(x) =
t
A
(x)
A
(x)
=
t
A
(x
t
)
A
(x) +
t
A
(x)
t
A
(x
t
)
=
A
+
t
A
(x)
x
B
Aj
j
A
j
t
A
x
B
Aj
j
A
j
j

x
(can replace
t
because
t
, dier by a term which would lead to an over-all term
2
no good). Thus,
j
= (A
j
j
)L +
B
Aj
j
A
j
A
x
L
(
A
)
(8.7.38)
For each
j
there is a dierent current;
j
k
= A
k
L
B
Ak
A
k
A
x
L
(
A
)
(8.7.39)
(Translations: A
k
= g
k
, B
Ak
= 0. Angular momentum rotations of space-time points
and elds.)
Dec. 3/99
Recall the spin
1
2
particle:
L =

(i m) (8.7.40)
= proton, neutron.
Yukawa postulated that the strong force was mediated by spin-zero pseudo-particles
(mass , V (r)
e
r
r
).
Heisenberg: postulated that the strong force treated the p and n exactly the same way and
these two nucleons were just dierent states of the same particle, and they corresponded
to spin up and spin down states of the electron. (Not really spin but same mathematical
structure, Isosspin).
Isospin doublet (8.7.41)
n
,
p
four component Dirac sponors
=
i
; i = 1, 2 (isospin) , = 1, . . . , 4 ( Dirac).
L =

i
(i m)
i
ig

5
(
ij
)
j
1
2
(
)(
)

2
2

2

4
(
2
)
2
(8.7.42)
with = Pauli matrices, and
1
,
2
,
3
- 3 spin zero pseudo-scalars, pions. This is invariant
under
e
i/2

e
i/2
e
i/2
e
i/2
where = (
1
,
2
,
3
) constant 3 component isovector.
1
,
2
,
3
3 vector particles,
also interact with the nucleons.
Postulate
L =
1
4
(
)
2

2
2

i
i
+

i
+ ( Rest of terms from last L) (8.7.43)
If
i
e
i
i
i
/2
i
e
i
i
i
/2
(8.7.44)
For
i
0
=
1 +
i
i
i
2
1
i
i
i
2
j
+
i
2
. .. .
2i
ijk
k
+O(
2
)
(8.7.45)
Thus,
+
ijk
k
(8.7.46)
Conserved current arises in
L =
1
4

2
2

i
i
(8.7.47)
Here;
0 = L =
L
+
L
(
L
(
+
L
(
L
(
. .. .
j
(8.7.48)
Thus as
L
(
)
=
1
4
(2)(2)
(8.7.49)
Thus,
j
=
f
i
. .. .
(
ijk
k
)
j
k
= f
i
ijk
(8.7.50)
Couple j
k
to A
k
,
L
t
=
/
. .. .
1
4
f
i
f
i

2
2

i
i
+gj
k
+
ijk
k
; A new piece is added to j
k
=
1
4

2
2

i
i
g(
)
ijk
k
(8.7.51)
i.e.
L
new
= g
ijk
(
)
j
k
(8.7.52)
j
new
=
L
new
(
= 2g
ijk
(
j
k
)(
imn
n
) (8.7.53)
Again, as is arbitrary,
j
newi
= 2g
lmn
lji
j
(8.7.54)
L = L
t
+ gj
newk

k
=
1
4
(
)
2

2
2

k
k
g(
)
ijk
+ g(2g
lmn
lji
)(
i
)
fourth term - only depends on , not derivative - need no more terms
=
1
4
F
i
F
i

2
2

i
i
(8.7.55)
where F
i
=
i
2g
ijk
Rescale g by ?
L =
1
4
(
+g
ijk
)
2

2
2

i
i
(8.7.56)
is invariant under
i
+
ijk
k
.
This can be generalized to an SU(N) gauge series.
L =
1
4
(
+ gf
ijk
A
j
A
k
)
2

2
2
A
i
A
i
(8.7.57)
with f
ijk
structure functions for SU(N) group.
L
spin-2
= L(h
) + T
+ T
t
h
+T
tt
h
+ . . . (8.7.58)
h
(x) = h
Dene T
(stress tensor ) from L. Couple

h
L +L
t
T
+ T
t
Sum L
spin-2
(), get
L
spin-2
= R
g (Einstein- Lagrangian) (8.7.59)

If
2
= 0, the gauge invariance becomes local (allows for renormalizability).
(x)
i
(x) +
i
(x)
g
+
ijk
(x)
k
(x) (8.7.60)
Chapter 9
Path Integral Quantization
recall:
9.1 Heisenberg-Dirac
dA(t)
dt
= [A(t), H]
PB
(9.1.1)
=
A
q
i
H
p
i
A
p
i
H
q
i
H
t
= 0
(9.1.2)
With (9.1.1), we can go from a classical to a quantum mechanical (QM) operator,
A(t)

A(t) (QM operator) (9.1.3)
d

A(t)
dt
=
1
i
A(t),

H
=
1
i
A

H

H

A
[ , ]
PB

1
i
[ , ] (9.1.4)
i

A
t
=
A,

H
(9.1.5)
(Note: will drop now, as we will deal with operators only from now on.) The solution
to this is,
A(t) = e
iHt/
A(0)e
iHt/
(9.1.6)
= A
H
(t) ; (Heisenberg Variable - Time dependent) (9.1.7)
141
142 CHAPTER 9. PATH INTEGRAL QUANTIZATION
State vectors [
H
` are time independent
'
H
[A
H
(t)[
H
` = '
H
[e
iHt/
. .. .
/
s
[
A(0) e
iHt/
[
H
`
. .. .
[
s
)
(9.1.8)
We go to Schrodinger state vectors by;
[
s
(t)` = e
iHt/
[
H
` (9.1.9)
i
d
dt
[
s
(t)` = H[
s
(t)` Schrodinger eq. (9.1.10)
'
s
(t)[ = '
H
[e
iHt/
(9.1.11)
A
s
= A(0) time independent. (9.1.12)
Consider eigenstates of the Heisenberg operators.
A
H
(t)[a(t)` = a[a(t)` (9.1.13)
e
iHt/
A(0)e
iHt/
[a(t)` = a[a(t)`
A(0)
e
iHt/
[a(t)`
= a
e
iHt/
[a(t)` (9.1.14)
Let t = 0,
A(0)[a(0)` = a[a(0)` (9.1.15)
Hence
[a(t)` = e
iHt/
[a(0)` (9.1.16a)
'a(t)[ = 'a(0)[e
iHt/
(9.1.16b)
9.2 Wave Functions
Schrodinger wave function
(q, t) = 'q[(t)`
s
(9.2.1)
where i
d
dt
[
s
(t)` = H[
s
(t)`, [(t)` = e
iHt/
[(0)`. Thus
(q, t) = 'q[e
iHt/
. .. .
(9.1.16b)
[(0)`
= 'q
H
(t)[(0)` (9.2.2)
(here Q(t)[q
H
(t)` = q[q
H
(t)` because we have a complete set of eigenfunctions).
Note:
dq(t) [q
H
(t)`'q
H
(t)[ = 1 for any t (9.2.3)
9.2. WAVE FUNCTIONS 143
This is the Completeness Relation. Let 0 < t
1
< t
2
< . . . < t
n
< t .
(q, t) =
dq
t
(0) 'q
H
(t)[q
t
H
(0)`
. .. .
K(t,0)
propagator
'q
t
H
(0)[(0)`
. .. .
(q,0)=/q[(0))
(9.2.4)
For
K
1
(t, 0) = 'q
H
(t)[q
t
H
(0)`
=
'q
H
(t)[q
Hn
(t
n
)`'q
Hn
(t
n
)[q
H(n1)
(t
n1
)`'q
H(n1)
(t
n1
)[ . . . [q
H1
(t
1
)`
'q
H1
(t
1
)[q
t
(0)`dq
n
(t
n
) dq
n1
(t
n1
) . . . dq
1
(t
1
) (9.2.5)
Consider one part of above:
'q
H(i+1)
(t
i+1
)[q
Hi
(t
i
)` = 'q
s(i+1)
(0)[e
iHt
i+1
/
e
+iHt
i
/
[q
si
(0)`
= 'q
s(i+1)
(0)[e
iH(t
i+1
t
i
)/
[q
si
(0)`
(9.2.6)
Let = t
i+1
t
i
. If 0, then,
'q
H(i+1)
(t
i+1
)[q
Hi
(t
i
)` 'q
s(i+1)
(0)[
1
iH
[q
si
(0)` (9.2.7)
Remember
[q, p] = i (9.2.8)
Thus,
< q, p >=
e
ipq/
2
(9.2.9)
i.e.
'q[ p[` = i

q
'q[` (9.2.10)
which is satised if [` = [p`. So,
'q
s(i+1)
(0)[
1
iH
[q
i
(0)` =
dp
i+1
dp
i
'q
i+1
(0)[p
i+1
`
. .. .
e
ip
i+1
q
i+1
/
2
'p
i+1
[
1
iH
[p
i
` 'p
i
[q
i
(0)`
. .. .
e
ip
i
q
i
/
2
(9.2.11)
i.e. used
'q[ p[(t)` = i

q
'q[(t)` i

q
(q, t) (9.2.12)
'q[ [ q, p] [(t)` = i'q[(t)`
= q'q[ p[(t)` 'q[ p q[(t)`
= q
i

q
'q[(t)`
i

q
'q[ q[(t)`
i

q
i

q
'q[(t)`
= i'q[(t)` (9.2.13)
[ q, p] = i (9.2.14)
'q[ p[(t)` =
i

q
+ f(q)
'q[(t)` (9.2.15)
See Dirac - f(q) can be absorbed into the phase of (q, t).
Jan. 13/2000
i.e.
i

q
+ f(q)
exp
q
f(q
t
)dq
t
(q, t) =
i

q
(q, t)
exp
q
f(q
t
)dq
t
(9.2.16)
Re: K(q
t
, t
t
; q, t),
(q
t
, t
t
) =
dq K(q
t
, t
t
; q, t)(q, t) (9.2.17)
K(q
t
, t
t
; q, t) = '(q
t
, t
t
)
H
[(q, t)
H
`
=
dq
1
. . . dq
n
'(q
t
, t
t
)
H
[(q
n
, t
n
)
H
`'(q
n
t
n
)
H
[(q
n1
t
n1
)
H
` . . . '(q
1
t
1
)
H
[(q, t)
H
`
(9.2.18)
'(q
i+1
t
i+1
)
H
[(q
i
t
i
)
H
` = 'q
i+1
s
[e
iH(t
i
1
t
i
)/
[q
i
s
`
'q
i+1
s
[
1
i
H( q, p)
[q
i
s
`
=
dp
i
'q
i+1
[
1
i
H( q, p)
[p
i
`'p
i
[q
i
` (9.2.19)
'q[ p[p` = i

q
'q[p` (9.2.20)
'q[p` = Ke
ipq/
dp 'q
t
[p`'p[q` =
dq e
ip(q
q)/
= (q
t
q) (9.2.21)
K =
1
2
(9.2.22)
9.2. WAVE FUNCTIONS 145
'(q
i+1
t
i+1
)
H
[(q
i
t
i
)
H
` =
dp
i
2
'q
i+1
[
1
i
H( q, p)
[p
i
`e
ip
i
q
i
/
'q
i+1
[H( q, p)[p
i
` H(q
i+1
, p
i
)'q
i+1
[p
i
`
( Have to worry about ordering of p

t
s, q
t
s.)
'(q
i+1
t
i+1
)
H
[(q
i
t
i
)
H
` =
dp
i
2
e
ip
i
(q
i+1
q
i
)/
1
i
H(q
i+1
, p
i
)
(9.2.23)
(The H term is important in curved space). Thus, K(q
t
, t
t
; q, t) is;
K(q
t
, t
t
; q, t) =
dq
1
. . . dq
n
dp
0
2
. . .
dp
n
2
e
ip
0
(q
1
q)/
e
ip
1
(q
2
q
1
)/
. . . e
ip
n
(q
q
n
)/
1
i
H(q
1
, p
0
)
1
i
H(q
2
, p
1
)
. . .
1
i
H(q
t
, p
n
)
dq
1
. . . dq
n
dp
0
2
. . .
dp
n
2
exp
p
0
(q
1
q
0
) H(q
1
, p
0
)
+ p
1
(q
2
q
1
) H(q
2
, p
1
) + . . . + p
n
(q
t
q
n
) H(q
t
, p
n
)
Let
q
i+1
q
i
q(t
i
). Thus,
K(q
t
, t
t
; q, t) =
dq
1
. . . dq
n
dp
0
2
. . .
dp
n
2
exp
i=0
Riemann Sum
. .. .
[p
i
( q
i
(t
i
)) H(q
i+1
, p
i
)]
(9.2.24)
See gure 9.2.1. As 0,
Figure 9.2.1: Particle path, where the q
n+1
=
q
t
, q
0
= q are xed, and the other qs (and ps)
are all integrated over.
K
q(t
)=q
q(t)=q
Tq(t)
Tp(t) exp
t
dt [p(t) q(t) H(q(t), p(t))]
(9.2.25)
Which is the Feynman Path Integral (Dirac/Feynman Papers in a collection edited by
Schwinger ).
Suppose,
H(q, p) =
p
2
2m
+ V (q) (9.2.26)
i.e.
dp
i
2
e
i
p
i
(q
i+1
q
i
)
p
2
i
2m
(9.2.27)
- this is a Gaussian
Recall:
dx e
axbx
2
=
e
b
(
x+
a
2b
)
2
+
a
2
4b
=
b
e
a
2
/4b
(9.2.28)
Ignore i in exponent of integral by putting in a convergence factor. i.e.,
dp
i
2
exp
p
i
(q
i+1
q
i
)
p
2
i
2m
+ ip
2
(let 0)
=
1
2

i
2m
exp
i(q
i+1
q
i
)
2
4
i
2m
m
2i
exp
im(q
i+1
q
i
)
2
2
(9.2.29)
Plug this in p
i
s ,
K(q
t
, t
t
; q, t) =
dq
1
. . . dq
n
m
2i
n+1
. .. .
N
exp
i=0
m(q
i+1
q
i
)
2
2
V (q
i+1
)
= N
dq
1
. . . dq
n
exp
i=0
m
2
q
i+1
q
i
2
. .. .
( q
i
(t
i
))
2
V (q
i+1
)
L(q, q) =
q
2
2m
V (q) Dim. of Energy = mass
l
2
t
2
K(q
t
, t
t
; q, t) = N
q(t
)=q
q(t)=q
dq
1
. . . dq
n
exp
t
dt L(q(t), q(t))
(9.2.30)
9.3. FREE PARTICLE 147
Jan. 17/2000
As 0, the dominant contribution when integrating over all paths comes from the path
that minimizes
S =
t
d L(q, q) (9.2.31)
i.e.
t
d L(q, q) = 0 (9.2.32)
(Hamiltons Principle of Least Action).
9.3 Free Particle
For a free particle, we mean V = 0.
'q
t
, t
t
[q, t` = N
q(t
)=q
q(t)=q
Tq() exp
t
d L( q(), q())
( = m = 1)
= N
q(t
)=q
q(t)=q
Tq() exp
t
d
q
2
()
2
(9.3.1)
But recall;
'(q
t
t
t
)
H
[(qt)
H
` =
'q
t
[e
i

Ht
e
+i

Ht
[q`
; free particle H =
p
2
2
=
dp 'q
t
[e
i p
2
t
/2
[p`'p[e
i p
2
t/2
[q`
=
dp e
ip
2
(tt
)/2
'q
t
[p`
. .. .
e
iq
2
'p[q`
....
e
iqp
2
=
dp
2
e
ip
2
(tt
)/2
e
ip(q
q)p
2
(9.3.2)
where 0
+
(inserted so integral converges). Now, recall the Gaussian:
dx e
ax
2
bx
=
dx e
a
x+
b
2a
2
e
b
2
/4a
=
a
e
b
2
/4a
(9.3.3)
Here we have a = +
i(t
t)
2
, b = i(q
t
q). So,
'(q
t
, t
t
)
H
[(q, t)
H
` =
1
2

+
i(t
t)
2
exp
(q
t
q)
2
4
+
i(t
t)
2
=
1
2i(t
t
t)
exp
i(q
t
q)
2
2(t
t
t)
(9.3.4)
Remember that,
(q
t
, t
t
) =
dq K(q
t
, t
t
; q, t)(q, t) (9.3.5)
where
K(q
t
, t
t
; q, t) = '(q
t
, t
t
)
H
[(q, t)
H
` (9.3.6)
but also
i

t
(q, t) =
1
2

2
q
2
(q, t) (9.3.7)
Hence,
i

t
t
K(q
t
, t
t
; q, t) =
1
2
2
q
t2
K(q
t
, t
t
; q, t) (9.3.8)
And
lim
t
t
K(q
t
, t
t
; q, t) = (q
t
q)(t
t
t) (9.3.9)
If we start putting in potentials, its awfully dicult to work out, especially when you put
in boundary conditions. Its easier to consider vacuum to vacuum transitions (in order to
eliminate having to consider q(t) = q and q(t
t
) = q
t
i.e. the B.C.s). Examine,
'(q
t
t
t
)
H
[(qt)
H
`
J
=
q(t
)=q
q(t)=q
Tq(t) e
i
t
d [L( q,q)+J()q()]
(9.3.10)
(where J = source) Look at, (now dropping subscript H always in Heisenberg eigen-
states);
'(Q
t
, T
t
)
H
[(Q, T)
H
`
J
=
dq
t
dq 'Q
t
T
t
[q
t
t
t
`
J
'q
t
t
t
[qt`
J
'qt[QT`
J
(9.3.11)
Now
'Q
t
T
t
[q
t
t
t
` = 'Q
t
[e
i

HT
e
i

Ht
[q
t
`
If

H[n` = E
n
[n`
=
n
'Q
t
[e
iE
n
T
[n`'n[e
iE
n
t
[q
t
`
If 'q[n` =
n
(q)
=
n
e
iE
n
(T
n
(Q
t
)
n
(q
t
) (9.3.12)
and similarly for 'q, t[Q, T`
J
.
'Q
t
T
t
[QT` =
dq dq
t
n,n
e
iE
n
(T
)
e
iE
n
(tT)
n
(Q
t
)
n
(q
t
)
n
(Q)
n
(q)'q
t
t
t
[qt`
J
(9.3.13)
Let T
t
i, T +i The vacuum becomes the dominant contribution. (i.e. only
the vacuum state [0` survives - projects only the vacuum state) - holds for any t, t
t
.
'Q
t
T
t
[QT` =
dq dq
t
e
iE
0
(t
+t)
. .. .
0
(Q
t
)
0
(Q)e
iE
0
(T
T)
0
(q
t
)'q
t
t
t
[qt`
0
(q)
. .. .
(9.3.14)
( * involves integral over q, q
t
). If
0
(q)e
iE
0
t
=
0
(q, t),
dq dq
t

0
(q
t
, t
t
)'q
t
t
t
[qt`
J
0
(q, t) = lim
T
i
T+i
'Q
t
T
t
[QT`
J
0
(Q, T)
0
(Q
t
, T
t
)
(9.3.15)
= '0, t
t
[0, t` (9.3.16)
(with 'Q
t
T
t
[QT`
J
= Path integral, and
0
(Q, T)
0
(Q
t
, T
t
) = Normal factor N). So,
'0, [0, `
J
= lim
T
i
T+i
N
Q(T
)=Q
Q(T)=Q
TQ() exp
T
d
L(

Q(), Q()) + J()Q()
(9.3.17)
Jan. 19/2000
Consider now,
'q
t
, t
t
[ q(t
n
) q(t
n1
) . . . q(t
1
)[q, t` = 'q
t
, t
t
[T q(t
1
) q(t
2
) . . . q(t
n
)[q, t` (9.3.18)
where T is the time ordering operator (largest times to the left), and where t
t
> t
n
> . . . >
t
1
> t. Proceed as we did with the QM Path Integral. i.e.
'q
t
t
t
[ q(t
n
) . . . q(t
1
)'qt[ =
dq
n
. . . dq
1
'q
t
, t
t
[
q
n
[q
n
t
n
)
. .. .
q(t
n
)[q
n
t
n
`'q
n
t
n
[
q
n1
[q
n1
t
n1
)
. .. .
q(t
n1
)[q
n1
t
n1
`
'q
n1
t
nq
[ . . .
q
1
[q
1
t
1
)
. .. .
q(t
1
)[q
1
t
1
`'q
1
t
1
[qt` (9.3.19)
Thus, just as
'q
t
t
t
[qt` =
q(t
)=q
q(t)=q
Tq(t) exp
t
d L( q(), q())
(9.3.20)
in this case we obtain
'q
t
t
t
[T q(t
1
) . . . q(t
n
)[qt` =
q(t
)=q
q(t)=q
Tq(t) q(t
1
)q(t
2
) . . . q(t
n
) exp
t
d L( q(), q())
(9.3.21)
(where t
t
t
i
t i = 1, . . . , n). Add J()q() to L:
'q
t
t
t
[T q(t
1
) . . . q(t
n
)[qt` =
q(t
)=q
q(t)=q
Tq(t) q(t
1
) . . . q(t
n
) exp
t
d [L( q(), d()) +J()q()]
J=0
=

iJ(t
1
)
. . .

iJ(t
n
)
q(t
)=q
q(t)=q
Tq(t) exp
t
d [L( q(), q())
+J()q()]
J=0
(9.3.22)
Thus,
'0, [Tq(t
1
) . . . (t
n
)[0, ` =

iJ(t
1
)
. . .

iJ(t
n
)
lim
T
i
Ti
N'Q
t
T
t
[QT` (9.3.23)
With Q = Q
t
= 0,
'0, [Tq(t
1
) . . . q(t
n
)[0, `
=

iJ(t
1
)
. . .

iJ(t
n
)
N lim
T
i
Ti
Q(T
)=0
Q(T)=0
TQ(T) exp
....
T
d
L(

Q(), Q())
+ J()Q()
(9.3.24)
Oscillates?
L =
Q
2
2
V (Q) =
1
2
d
d
Q
2
V (Q)

ic
0
dx e
ix
ix = y
= i
+
0
dy d
y
(well dened) (9.3.25)
Now let = i
E
, (E = Euclidean). So, we get
'0, [Tq(t
1
) . . . q(t
n
)[0, `
=
1
i
n

J(t
1
)
. . .

J(t
n
)
N
lim
T
T
E
q()=0
q()=0
Tq(t) exp
E
T
E
d
E
Q
2
2
V (Q) + J()Q()
J()=0
(9.3.26)
The same procedure can be applied in QFT. Here,
'0[T(x
1
) . . . (x
n
)[0` =
1
i
n
J(x
1
)
. . .

J(x
n
)
N
T(x) e
d
4
x
E
[/
E
(
,)+J(x)(x)]
J=0
(9.3.27)
(where (x) 0 as x .) If
L
M
(
, ) =
1
2
(
0
)
2
()
2
. .. .
A
m
2

4!
4
(9.3.28)
where A = (
)(
) (+, , , ) (Bj. and Drell), or = (
)(
) (, +, +, +).
(and L
M
is the Minkowski lagrangian). And,
L
E
=
1
2
(
4
)
2
()
2
m
2

4!
4
(9.3.29)
with x
0
....
(M)
= ix
4
....
(E)
x
2
E
= (x
4
)
2
+ (x)
2
.
Let
Z
E
(J) = N
Texp
d
4
x
E
1
2
(
E
)
2
2
2

4
4!
+J(x)(x)
(9.3.30)
(Generating Functional) - (call the mass now). So,
'0[T

(x
1
) . . .

(x
n
)[0` =
1
i
n
J(x
1
)
. . .

J(x
n
)
Z
E
(J)
J=0
(9.3.31)
Jan. 21/2000
Consider,
dx e
a
2
x
2
+jxx
4
j=0
=
dx
n=0
(x
4
)
n
n!
e
a
2
x
2
+jx
j=0
=
n=0
()
n
n!
d
dj
4n
dx e
a
2
x
2
+jx
j=0
=
n=0
()
n
n!
d
4n
dj
4n
dx exp
a
2
x
j
2a
2
2
+
j
2
4a
2
j=0
Let x
t
= x
j
2a
2
=
n=0
()
n
n!
d
4n
dj
4n
dx
t
e
a
2
x
2
. .. .
a
e
j
2
/(4a
2
)
j=0
=
n=0
()
n
n!
d
4n
dj
4n
e
j
2
/(4a
2
)
j=0
=
1 +

1!
d
4
dj
4
j=0
+
()
2
2!
d
8
dj
8
j=0
+ . . .
e
j
2
/(4a
2
)
=
1 +
1!
3
4a
2
+ . . .
(9.3.32)
Note:
d
dj
e
j
2
/(4a
2
)
=
2j
4a
2
e
j
2
/(4a
2
)
d
2
dj
2
e
j
2
/(4a
2
)
=
2
4a
2
+
2j
4a
2
e
j
2
/(4a
2
)
. . . and so
d
4
dj
4
e
j
2
/(4a
2
)
j=0
=
8
16a
4
+
4
16a
4
=
3
4a
4
Thus,
dx e
a
2
x
2
+jxx
4
=
1 +
1!
d
4
dj
4
j=0
+
()
2
2!
d
8
dj
8
j=0
+ . . .
e
j
2
/(4a
2
)
(9.3.33)
with the () representing a dot (vertex), and
1
a
2
being the loops. (see gure 9.3.1) For

Z(J) =
T exp
d
4
x
1
2
(
) (
)

2
2

2

4
4!
+ J
k=0
4!
d
4
x

4
J
4
(x)
T exp
d
4
x
1
2
+ J
(9.3.34)
Figure 9.3.1: Simple Feynman Diagram
example
But now,
d
4
x
1
2
+ J
(Now, complete square)

=
d
4
x
1
2
+J
1
2
J
(9.3.35)
Let
t
= +
J
So,
Z(J) =
k=0
4!
d
4
x

4
J
4
(x)
T
t
(x) e
d
4
x
1
2
2
)
1
2
J
(9.3.36)
Here,
1
2
(
2
2
)
1
(
2
2
)
= 1. So,
1
2
= G(x, y) (
2
x
2
)G(x, y) =
4
(x y) (9.3.37)
Thus,
J
1
2
J =
d
4
x d
4
y J(x)G(x, y)J(y) (9.3.38)
Also, if,
G(x, y) =
d
4
k
(2)
4
g(k)e
ik(xy)
(9.3.39)
Then using (9.3.37) ,
d
4
k
(2)
4
e
ik(xy)
g(k) =
d
4
k
(2)
4
e
ik(xy)
. .. .
(xy)
(k
2
2
)g(k) = 1
G(x, y) =
d
4
k
(2)
4
e
ik(xy)
(k
2
2
)
(Non-singular integral) (9.3.40)
Note that this provides meaning of the last term in the exponent of Z(J). (also, k
2
=
k
2
1
+ k
2
2
+ k
2
3
+ k
2
4
).
For the other part of the exponent,
T e
d
4
x
1
2
(
2
2
)
(9.3.41)
Consider,
d
n
x e
xMx
; M
T
= M (M
= M)
O
1
MO = T =
1
.
.
.
(
i
= eigenvalues of )
O
1
= O
T
Let
y = O
1
x (9.3.42)
d
n
y = det
O
1
d
n
x
= (1)d
n
x (9.3.43)
(det(O) = Product of eigenvalues of O). So,
d
n
x e
x
T
Mx
d
n
y e
yDy
=
dy
1
e
1
y
2
1
dy
2
e
2
y
2
2
. . .
dy
n
e
n
y
2
n
=
2
. . .
n
=

n/2
det
1/2
()
(9.3.44)
So,
T
t
e
d
4
x
[
1
2
2
)
]
= Ndet
1/2
(
2
2
) (9.3.45)
and so,
Z
0
(J) = Ndet
1/2
(
2
2
)e
1
2
d
4
x d
4
y J(x)G(x,y)J(y)
(9.3.46)
Thus,
Z(J) =
n=0
()
n
n!
d
4
w

4
J
4
(w)
n
Ndet
1/2
(
2
2
)e
+
1
2
d
4
x d
4
y J(x)

G(x,y)J(y)
(9.3.47)
where

G(x, y) =
d
4
k
(2)
4
e
ik(xy)
k
2
+
2
= G(x, y).
Jan. 24/2000
Recall;
Z[J] =
Texp
d
4
x
1
2
(
)
2
1
2

4
4!
+ J
(9.3.48)
Z
0
[J] = Ne
(1/2)
d
4
xd
4
yJ(x)G(x,y)J(y)
( = 0) (9.3.49)
(i)
n
'0[T

(x
1
) . . .

(x
n
)[0`
=

J(x
1
)
. . .

J(x
n
)
Z[J]
J=0
=

J(x
1
)
. . .

J(x
n
)
j=0
1
(j!)
4!
d
4
z

4
J
4
(z)
j
Ne
(1/2)
d
4
xd
4
y J(x)J(y)G(x,y)
J=0
(9.3.50)
For example, say n = 4 (and let

J
1
=

J(x
1
)
, etc.),
(i)
n
'0[T

(x
1
) . . .

(x
4
)[0`
=

J
1
. . .

J
4
1 +
4!
d
4
z

4
J
4
z
+
1
2!
4!
d
4
z
1
4
J
4
z
1
dz
2
4
J
4
z
2
+. . .
exp
1
2
d
4
xd
4
yJ(x)J(y)G(x, y)
J=0
= [G(x
1
x
2
)G(x
3
x
4
) +G(x
1
x
3
)G(x
2
x
4
) +G(x
1
x
4
)G(x
2
x
3
)]
. .. .
()
0
+ ()
G(0) [ ] +
d
4
z G(x
1
z)G(x
2
z)G(x
3
z)G(x
4
z) + . . .
+ . . .
or, diagrammatically, this is
=
+()
+ + +
+ + + +
+ ()
2
+
+ + + . . . + + + . . .
+ . . . (9.3.51)
where,
G(x, y)
()
(9.3.52)
9.4. FEYNMAN RULES 157
9.4 Feynman Rules
= (9.4.1)
= G(x y) (9.4.2)
Integrate over all internal points z
Take into account symmetry factors and number of diagrams of a given topology
Z
0
[J] = Ne
(1/2)
d
4
x d
4
y J(x)G(xy)J(y)
(9.4.3)
Z[J] =
j
1
j!
4!

4
J
4
z
d
4
z
j
Z
0
[J] (9.4.4)
It can be shown (c.f. Cheng and Li) . . . Just as

J(x
1
)

J(x
n
)
Z[J]
gives rise to all N

point Feynman diagrams, so also for the connected diagrams,
(i)
n
'0[T

(x
1
) . . .

(x
n
)[0`
connected
=

J
1
. . .

J
n
W[J] (9.4.5)
where
W[J] = ln (Z[J]) (9.4.6)
For example, in n = 4,
(i)
4
'0[T

(x
1
) . . .

(x
4
)[0`
connected
=
+
+ + + + . . .
+ . . . (9.4.7)
(i.e. we dont get diagrams with separate disconnected parts) (*) and disconnected
diagrams can be shown to contribute only to the phase of the Greens Function (G.F.).
We thus only need the connected G.F. to calculate physical cross sections. Futhermore,
if
(x) =
W[J]
J(x)
(9.4.8)
and
[] = W[J]
d
4
x J(x)(x) (9.4.9)
then
(x
1
)
. . .

(x
n
)
[] (9.4.10)
gives rise to the one-particle irreducible (1PI) connected Greens Functions. i.e. cant cut
diagrams into 2 parts by cutting a single internal line. For example,
Not 1PI
is 1PI
1PI
Not 1PI
This is useful in calculations because we can compute each 1PI part and connect the two to
get non-1PI parts.
For example,
= ()
2
d
4
z
1
d
4
z
2
G(x
1
z
1
)G(x
2
z
2
)G
3
(z
1
z
2
) (9.4.11)
# = 1 (one such diagram). Symmetry factor (# of ways we can connect these lines)
S =
1
6
. (Recall, for the symmetry factor: for the numerator, x
1
can go to 8 points (four
points per vertex), then x
2
can go to 4 points (four points on the second vertex), and then
one of the remaining three lines on one of the vertices can go to 3 possible lines on the other
vertex, then one of the remaining two lines can go to 2 possible lines on the other, and the
Figure 9.4.1: Momentum Space dia-
gram - note momentum conserved at
each vertex
nal line can go to the 1 line on the other vertex - for the denominator, there are two 4!
factors (four points per vertex, two vertices), and we can switch the vertices, so there is also
a 2!; thus,
S =
8 4 3 2 1
4! 4! 2!
=
1
6
Note that the 4! in

4!

4
per vertex is accounted for in the symmetry factor, as is the
1
2!
that
would arise from the sum (see (9.4.4))).
Convert to p-space,
G(x y) =
d
4
k
(2)
4
e
ik(xy)
k
2
+
2
(9.4.12)
In momentum space
for
the number of integrals number of loops.

'0[T

(x
1
)
(x
2
)[0` =
1
6
()
2
d
4
p
(2)
4
e
ip(x
1
x
2
)
1
(p
2
+m
2
)
2
d
4
k
1
d
4
k
2
1
k
2
1
+m
2
1
k
2
2
+ m
2
1
(k
1
+ k
2
+ p)
2
+ m
2
(9.4.13)
(see gure 9.4.1)
In Minkowski space, we have the same diagrams, with,
= i (9.4.14)
= i(x y) =
d
4
k
(2)
4
e
ik(xy)
k
2
2
+ i
(9.4.15)
Note: k
2
= k
2
0
k
2
, pole at k
0
=
k
2
+
2
.
'0[T

(x
1
) . . .

(x
n
)[0`
Minkowski
=

iJ(x
1
)
. . .

iJ(x
n
)
Z
m
[J] (9.4.16)
9.5 Path Integrals for Fermion Fields
L =

(i m) (9.5.1)
Canonical Quantization ,

= i( ). In the path integral, and

are Grassmann
Variables. i.e.
Z(, ) =
TT
exp
d
4
x
L + +

(9.5.2)
Its important that ,

are treated independently. (Note, though, if =
c
=

C

T
(ma-
jorana), then ,

are no longer independent).
9.6 Integration over Grassmann Variables
Recall,
2
=
2
1
(
2
i
= 0) (9.6.1)
d
d
a
(
m
n
) =
d
m
d
a
n
+
m
d
n
d
a
=
ma
na
m
(9.6.2)
If we have
1
. . .
n
, then
p(x,
1
,
2
, . . . ,
n
) = p
0
(x) + p
i
(x)
i
+ p
ij
j
+ . . . + p
i
1
,...,i
n
(x)
i
, . . . ,
i
n
(9.6.3)
9.6. INTEGRATION OVER GRASSMANN VARIABLES 161
(note no (n + 1) term due to
2
i
= 0) where p
ij
= p
ji
, p
ijk
= p
jik
, etc.. We have,
p(x, ) = p
0
+ p
1
d p(x, ) =
d [p
0
+ p
1
]
=
d [p(x, + )] ( a displacement) (9.6.4)

For this to hold,
d = 1 (9.6.5)
(1)d = 0 (9.6.6)
Thus, (9.6.4) is
=
d [p
0
+p
1
( +)] (9.6.7)
Thus,
d
d
d
(i.e. are the same operator). Also, note that the only term that will survive
in the integral will be the one with n s (all other terms wont have enough s). So,
d
1
. . . d
n
[p(x,
1
, . . . ,
n
)] =
d
n
. . . d
1
[p
0
+p
i
i
+ p
ij
j
+ . . . + p
i
1
,...i
n
i
1
. . .
i
n
]
=
d
n
. . . d
1
p
i
1
...i
n
i
1
. . .
i
n
=
i
1
...i
n
p
i
1
...i
n
. .. .
n! terms
= n!p
123...n
(9.6.8)
(the
i
1
...i
n
is present because order of s may not be i
1
, i
2
, . . .). If we now change variables
of integration,
i
= a
ij
j
(9.6.9)
then,
d
i
p(x,
i
) =
i
p(x,

i
) (expect)
=
i
p
i
1
...i
n
i
1
. . .

i
n
=
i
(p
i
1
...i
n
)a
i
1
j
1
. . . a
i
n
j
n
j
1
. . .
j
n
(9.6.10)
=
d
1
. . . d
n
(p
i
1
...i
n
)
i
1
. . .
i
n
(9.6.11)
and (9.6.10) must be equal to (9.6.11). Thus,
d
1
. . . d
n
p
1...n
1
. . .
n
=
1
. . . d
n

i
1
...i
n
a
i
1
j
1
. . . a
i
n
j
n
p
1...n
1
. . .
n
(9.6.12)
Hence
d
1
. . . d
n
= d
1
. . . d
n
i
1
...i
n
a
i
1
j
1
. . . a
i
n
j
n
. .. .
=det(a)
= det(a)d
1
. . . d
n
i.e. d
n
= det
1
. . .

n
)
(
1
. . .
n
)
. .. .
d
n
(9.6.13)
where * is the reciprocal of the normal jacobian; recall
d
n
x f(x) =
d
n
x det
(x
1
. . . x
n
)
( x
1
. . . x
n
)
f( x) (9.6.14)
i.e. d
n
x = det
(x
1
. . . x
n
)
( x
1
. . . x
n
)
d
n
x (9.6.15)
Also note,
d
1
. . . d
n
d
1
. . . d
n
e
i
C
ij
j
(expand exponential)
=
d
1
. . . d
n
d
1
. . . d
1
n!
(
i
C
ij
j
)
n
=
1
n!
i
1
...i
n
j
1
...j
n
C
i
1
j
1
. . . C
i
n
j
n
=
1
n!
(n! det(C))
= det(C) (9.6.16)
Recall,
d
1
. . . d
n
e
i
C
ij
j
= det
1/2
(C)
but
dx
1
. . . dx
n
d x
1
. . . d x
n
e
x
i
C
ij
x
j
If
i
> 0 for C
ij
then
=

n
detC
x/x
t
s distinct, not
det(C)
9.6. INTEGRATION OVER GRASSMANN VARIABLES 163

For n = 2,
d
1
d
2
exp
[
1
,
2
]
0
0

1
d
1
d
2
1 + [
1
,
2
]
0
0

1
d
1
d
2
[
1
1
]
=
= det
(1/2)
0
0
(9.6.17)
Hence, for,
Z(, ) =
TT
exp
d
4
x
(p m) + +

TT
exp
d
4
x
1
p m
(p m)
1
p m
1
p m
( recall p i)
=
TT
exp
d
4
x
(p m) +
1
p m
( = (x
))
= det(p m)
. .. .
normalization
exp
d
4
x
1
p m
(9.6.18)
Recall:
Z[J] =
1
det
(1/2)
(p
2
m
2
)
exp
i
2
dx J
1
p
2
m
2
Only fallout from having to take into account that , are Grassmann.
i
p m+ i
Propagator (9.6.20)
Closed loop of fermions factor of (1) (9.6.21)
(9.6.22)
Jan. 27/2000
9.7 Gauge Invariance
(1
st
Principle) U(1)
S =
d
4
x
1
2
(
2
+
2
)
d
4
x
1
2
)
2
( 1 degree of freedom ) (9.7.1)

If complex,
S =
d
4
x
)(
)
2
( 2 degrees of freedom
, ) (9.7.2)
We demand
e
i(x)
(9.7.3)
be an invariance.
L = (
)(
) m
2
(9.7.4)
Invariant under a global transformation,
e
i
( const.) (9.7.5)
If = (x),
L =
1
4
(
)
2
+ [(
+ ieA
] [(
ieA
) ] m
2
(9.7.6)
where
A
+
1
e
(9.7.7)
D
= ieA
covariant derivative (9.7.8)

D
UD
(9.7.9)
(D
) U(D
)
= U(ie)F
; F
(9.7.10)
Advantage of looking at E.M. this way can generalize,
p
n
= N Heisenberg postulate (9.7.11)

N = N
i
(9.7.12)
where is the Dirac index, and i = 1, 2 (1 , 2 . The invariance is of the form
(forgetting Dirac index);
N
i
U
ij
()N
j
(9.7.13)
9.7. GAUGE INVARIANCE 165
where
U
ij
() =
e
i/2
ij
(9.7.14)
and
=
1
2
0 1
1 0
0 i
i 0
1 0
0 1
(9.7.15)
N
i
(i
ij
m
ij
) N
j
is globally SU(2) invariant. Suppose = (x
). Add in A
ij
=
A
a
(
a
)
ij
. So,

i
2
g A
N ; A = A
a
(
a
)
ij
=

N
i
ij
i
2
gA
a
(
a
)
ij
m
ij
N
j
(9.7.16)
If N UN,

N

NU
1
, then,
Ni

ig
2
A
N

NiU
1

ig
2
A
UN
=

N
U
1
ig
2
U
1
A
UN
= i

NU
1
(UN)
. .. .
(
U) + (U)
ig
2
A
N
= i

N [
ig
2
A
t
N (9.7.17)
(Notation: note that (
A) = A
,
i.e. , derivative wrt. .) where,
U
1
(
U)
ig
2
U
1
A
U =
igA
t
2
U
1
U
,
ig
2
U
1
A
U =
ig
2
A
t
(9.7.18)
For
a
0,
U
ij

ij
+ i
a
(
a
)
ij
(9.7.19)
U
1
= 1 i (9.7.20)
ig
2
A
t
= (1 i )
i
,
ig
2
(1 i )
(1 + i )
= i
,
ig
2
i A
+ i A
(9.7.21)
But,
A A =
a
A
b
= i
abc
a
A
b
c
( =
1
2
a)
= i A (9.7.22)
So,
ig
2
A
t
= i
,
ig
2
i(2i)( A
(9.7.23)
Hence,
ig
2
A
t
= i
a
,
+ i
g
2
a
A
ta
=
2
g
a
,
abc
A
b
c
(9.7.24)
In other words, we have 3 photon elds, 1 for each spin matrix. (A
+
1
e
).
Consider three scalars
a
(a = 1, 2, 3).
L =
1
2
a
)(
a
) m
2
(9.7.25)
For an invariance under
a
a
+ g
abc
c
(x), for
a
0, then,
L =
ab
+ g
acb
A
c
ap
+ g
aqp
A
q
m
2
a
(9.7.26)
with
A
ta
= A
a
a
+ g
abc
A
b
c
(9.7.27)
(again, SU(2), but SU(2) rep. that is isomorphic to SU(3)). Dene:
D
ab
ab
+ g
amb
A
m
(9.7.28)
(Analogous to D =
ieA
do the same thing D
= ieF
),
L = (D
am
m
)(D
an
n
) m
2
a
(9.7.29)
A
ta
= A
a
+D
ab

b
(9.7.30)
ta
=
a
+ g
abc
c
(9.7.31)
Now,
D
ab
(A) (
ab
+ g
abc
c
)(D
bm
m
) (9.7.32)
under the gauge transformations.
Consider,
(D
ab
D
bc
D
ab
D
bc
(9.7.32) (
ab
+ g
abc
c
)
(D
bm
D
mn
D
bm
D
mn
)
n
(9.7.33)
9.7. GAUGE INVARIANCE 167
But now,
D
ab
D
bc
D
ab
D
bc
= (
ab
+g
apb
A
p
)(
bc
+ g
bqc
A
q
)
(
ab
+ g
apb
A
p
)(
bc
+ g
bqc
A
q
)
= g
abc
A
b
) (
A
b
) + g
bmn
A
m
A
n
(c.f.
abc
bmn
=
am
cn
+
an
cm
)
= g
abc
F
b
(9.7.34)
F
(U(1) case) (9.7.35)

F
a
(
ab
+
abc
c
)F
b
(9.7.36)
Lagrangian for A
a
L =
1
4
F
a
F
a
(9.7.37)
(Gauge invariant).
Jan. 31/2000
Chapter 10
Quantizing Gauge Theories
We could look at canonical quantization A
a
(x); Dene:
a
=
L
(
A
a
)
(10.0.1)
Identify constraints
insert gauge conditions for each of the 1
st
class constraints
Form Dirac Brackets
From these, determine commutators
Problems with this in practice are;
1. Manifest Lorentz invariance is lost (Important)
2. For the Coulomb Gauge (for example), where
i
A
a
i
(x, t) = 0, in forming the Dirac
Brackets will involve
i
(D
ab
1
(10.0.2)
where D
ab
=
i
ab
+ g
apb
A
p
i
. .. .
( * part drops out in abelian case).

10.1 Quantum Mechanical Path Integral
Use the Q.M Path Integral (QMPI).
L =
1
4
F
a
(A)F
a
(A) + J
a
A
a
(10.1.1)
where for now, we will eliminate the second term J
a
A
a
purely classical L. Note that we
have full gauge invariance A
a
A
a
a
+ g
abc
A
b
c
. Consider,
Z[J] =
TA
a
exp
d
4
x
L + J
a
A
a
(10.1.2)
169
170 CHAPTER 10. QUANTIZING GAUGE THEORIES
Z[J] =
TA
a
exp
d
4
x
1
4
/
0
. .. .
A
a
A
a
2
+2g
A
a
A
a
abc
A
b
A
c
+g
2
amn
A
p
A
q
A
m
A
n
+ J
a
A
a
(10.1.3)
but, L
0
=
1
2
A
a
(
2
g
) A
a
, and,
TA
a
exp
d
4
x
1
2
A
a
(
2
g
)A
a
+ J
a
A
a
det
(1/2)
ab
(g
(10.1.4)
and,
2
g
) = 0 (10.1.5)
is an eigenvector with vanishing eigenvalue! Hence,

det
2
g
= 0 (10.1.6)
(Dirac put in
1
2
[
A
a
]
2
into (10.1.2), (in exponential integral) eliminates
but
he found inconsistencies). To see the Faddeev-Popov procedure, consider,
d
n
xe
x
T
Nx
=

n/2
det
1/2
(N) (10.1.7)
det(N) =
1
. . .
n
dx
1
. . . dx
n
e
x
2
1
1
. . . e
x
2
n
Only if
i
> 0i (10.1.8)
Suppose,
N m(x)
0
= 0 (10.1.9)
m(x)
0
eigenvector corresponding to eigenvalue zero. Note that,
d
n
y (Ay + b)f(y) (10.1.10)
Let z = Ay + b, so d
n
y =
1
detA
d
n
z. So,
d
n
y (Ay + b)f(y) =
1
detA
d
n
z
n
(z)f(A
1
(z b))
=
f(b)
det(A)
(10.1.11)
Thus,
(Ay + b) =
(y + A
1
b)
det(A)
(10.1.12)
10.1. QUANTUM MECHANICAL PATH INTEGRAL 171
(c.f. 1-D (ax) =
1
[a[
(x)). Consider
1 =
d
0
[L(x + m(x)
0
)] det(L m(x)) (10.1.13)
Insert this into
d
n
x e
x
T
N x
=
d
n
x
d
0
[L(x m(x)
0
)] det [L m(x)] e
x
T
N x
(10.1.14)
Let x x m(x)
0
. Then,
d
n
x e
x
T
N x
=
d
n
x
d
0
(L x) det [L m(x)] e
x
T
N x
(note no change in e
x
T
N x
because N(m
0
) = 0)
=
d
0
d
n
x (L x) det (L m(x)) e
x
T
N x
(10.1.15)
(The
d
0
contains the occurring in the integral over the eigenvector with vanishing
eigenvalue). Note:
1. det [L m(x)] =
dcdc e
c
T
(L m(x))c
(where c, c Grassmann vectors).
2. (L x) = lim
0
e
(L x)
2
So,
d
n
xe
x
T
N x
=
d
0
d
n
x lim
0
e
x
T
N x
dc dc e
c
T
L m(x)c(L x)
2
=
d
0
lim
0
. .. .
absorb into Normalizing factor
d
n
x dc dc e
[
x
T
(
N+c
T
L
)
x+c
T
L m(x)c
]
(10.1.16)
Choose L so that (N + L
T
L) is invertible (i.e. (N + L
T
L) has no vanishing eigen-
value).
c
T
L m(x)c cancels the contribution of eigenvectors with vanishing eigenvalues (ghost
elds).
Feb. 2/2000
The condition is N(m(x)
0
) = 0
1 =
d
0
(L(x m(x)
0
))det(L m(x))
x x m(x)
0
=
d
0
d
n
x (L x)
. .. .
A
det(L m(x))
. .. .
B
e
x
T
N x
A lim
e
(L x)
2
/
B
dcd c e
c
T
L m(x)c
10.2 Gauge Theory Quantization
Recall the path integral
TA
a
(x)e
d
4
x
(F
a
(x))
2
4
+J
a
A
a
(10.2.1)
If J = 0,
A
a
A
a
+ D
ab
(A)
b
(x) (10.2.2)
(Analogy x A
a
(x) from above
With D
ab
(A) =
ab
+g
apb
A
b
Insert a factor of 1.
1 =
d
a
(x) (
....
L
(A
a
(x)
. .. .
x
+D
ab
(A)
b
. .. .
m(x)
0
))det(
D
ab
(A)
. .. .
Lm(x)
) (10.2.3)
=
d
a
(x) (
(x))det(
D
ab
(A))
Where,
A
(x) = U
1
()(A
a
)U() ; U()
ab
+
apb
p
A
a
+ D
ab
(A)
b
Let A
A
a
(reverse gauge transf. )

1 =
d
a
dA
a
A
a
)det(
D
ab
(A)e
i
d
4
x
(
1
4
(F
a
)
2
)
= lim
0
d
a
. .. .
Absorb into
Normalization
dA
a
dc
a
d c
a
exp
d
4
x
1
4
F
a
F
a
1
2
(
A
a
)
2
c
a
D
ab
(A)c
b
(10.2.4)
Recall that the cs are grassmann. Also, note that, in the last term, D
ab
(A) =
ab
+
g
apb
A
p
, and so ghost elds are important (ghosts are free elds) Mathematical constructs,
Longitudinal mode of vector particle. Note that in the U(1) case (gauge group), the Ghost
Lagrangian is c
c = c
c.
10.3 Feynman Rules
Here are the Feynman Rules.
i
ab
(g
k
2
)
k
2
+
k
k
4
Where the last term is the dependance - (longitudinal part of propagator).
i
ab
k
2
g
abc
p
g
abc
(p q)
+ (q r)
+ (r p)

ig
2

f
abc
f
cde
(g
)
f
abe
f
bde
(g
)
f
ade
f
cbe
(g
Feb. 3/2000
10.4 Radiative Corrections
L =
1
2
(
)
2
m
2
2

2

4
4!
(10.4.1)
and recall the Feynman rules are,
i
p
2
m
2
i
Recall that, in general,
d
n
(2)
n
(
2
)
a
(
2
m
2
)
b
=
i(1)
ab
(4)
n/2
a +
n
2
a b
n
2
(m
2
)
ba+n/2
(b)
n
2
(10.4.2)
d
n
(2)
n
i
(
2
m
2
)
=
i(1)
(4)
n/2

1
n
2
. .. .
pole at n=4
(m
2
)
n/21
(10.4.3)
10.4. RADIATIVE CORRECTIONS 175
Also, in general,
1
a
=
( + )
()()
1
0
dx
x
1
(1 x)
1
[xa + (1 x)b]
+
(10.4.4)
The 1PI contribution to the 2-pt. function is
=
i
p
2
m
2
+
i
p
2
m
2
1
2
(i)
d
n
(2)
n
i
(
2
m
2
)
i
p
2
m
2
(10.4.5)
Which can be easily integrated. Another loop integral is (dropping external leg contribution
for now);
=
(i)
2
6
d
n
k d
n
(2)
2n
1
(k
2
m
2
)(
2
m
2
)[(p k )
2
m
2
]
=
(i)
2
6
d
n
k d
n
(2)
2n
1
(
2
m
2
)
1
0
dx
1
[(1 x)(k
2
m
2
) + x((p k )
2
m
2
)]
2
=
(i)
2
6
d
n
k d
n
(2)
2n
1
(
2
m
2
)
1
0
dx
1
[k
2
2xk(p ) + x(p )
2
m
2
]
2
=
(i)
2
6
d
n
k d
n
(2)
2n
1
(
2
m
2
)
1
0
dx
1
[(k x(p )
. .. .
k
)
2
+ x(1 x)(p )
2
m
2
]
=
(i)
2
6
d
n
k
t
d
n
(2)
2n
1
(
2
m
2
)
1
0
dx
1
[k
t2
+ x(1 x)(p )
2
m
2
]
2
The integral over k
t
can be evaluated; recall (10.4.2) (Note: General integral correct in
(10.4.2), but check signs on as and bs).
=
(i)
2
6
1
0
dx
d
n
(2)
n
i
(4)
n/2
2
n
2
(
2
m
2
)(m
2
x(1 x)(p )
2
)
2n/2
=
(i)
2
6
1
0
dx
d
n
(2)
n
i
2
n
2
(4)
n/2
1
(
2
m
2
)
(x(1 x))
2+n/2
1
( p)
2
m
2
x(1x)
2n/2
Now, recall (10.4.4).
=
(i)
2
6
1
0
dx
d
n
(2)
n
i
2
n
2
(4)
n/2
(x(1 x))
2+n/2

1 + 2
n
2
(1)
2
n
2

dy
y
11
(1 y)
2n/21
y(
2
m
2
) + (1 y)
( p)
2
m
2
x(1x)
3n/2
=
1
0
dx
1
0
dy
d
n
(2)
n
i
3
n
2
(x(1 x))
2+n/2
(1 y)
1n/2
(4)
n/2
2
+ y(1 y)p
2
m
2
y
m
2
(1y)
x(1x)
3n/2
=
1
0
dx
1
0
dy
i
(4)
n/2
3
n
2
(x(1 x))
2+n/2
(1 y)
1n/2
i
(4)
n/2
(1)
3+n/2
3
n
2

n
2
y(1 y)p
2
+ m
2
y +
m
2
(1 y)
x(1 x)
n/2+n/23
=
1
0
dx
1
0
dy
i
(4)
n/2
3
n
2
(3 n)(1)
5+n
(x(1 x))
2+n/2
(1 y)
1n/2
y(1 y)p
2
+ m
2
y +
m
2
(1 y)
x(1 x)
n3
Now, the integrals over x, y are problems. But, if m
2
0,
=
i
(4)
n/2
1
0
dx
1
0
dy
3
n
2
(3 n)((1)
n
)(x(1 x))
2+n/2
(1 y)
1n/2
(1)
n3
y(1 y)p
2
n3
=
i
(4)
n/2
1
0
dx
1
0
dy
3
n
2
(3 n)(x(1 x))
2+n/2
y
n3
(1 y)
n/22
(p
2
)
n3
Now, using (10.4.4), we have a = b = 1, so
=
i
(4)
n/2
3
n
2
(3 n)
. .. .
Pole at
n=4
n
2
1
(n 2)
(n 2)
n
2
1
3n
2
3
(p
2
)
n3
The (p
2
)
n3
can be expanded as (1 + (n 3) ln(p
2
) + . . .), and recall,
() =
1

E
+ ln(4) + . . . (10.4.6)
Thus, the 2-loop integral is
=
A
2
+
Bln(p
2
) + C
+O() (10.4.7)
(ln(p
2
) is singular if p
2
= m
2
= 0). At one loop,
d
n
(2)
n
1
(
2
m
2
)
=
A
2
+ B (10.4.8)
B is independent of p
2
(this is a uke).
Feb. 7/2000
= 2
+ 2i
5
(10.4.9)
where
=
i
2
[
],
= 2g
, and
5
=
1
4!
(10.4.10)
Let
L =
1
2
(
0
)
2

2
0
2
0
2

0
4
4!
(10.4.11)
= i(p)
= + + + . . .
= Sum of all 2pt 1PI diagrams.
Let
'0[T
0
(x)
0
(y)[0`
connected
=
=
i
p
2
2
0
+ i
+
i
p
2
2
0
+ i
(i)
i
p
2
2
0
+ i
+
i
p
2
2
0
+ i
(i)
i
p
2
2
0
+ i
(i)
i
p
2
2
0
+ i
+ . . .
= a +ar + ar
2
+. . .
=
a
1 r
2
=
i
p
2
2
0
+i
1 (i)
i
p
2
2
0
+i
=
i
p
2
2
0
(p
2
) + i
= i(p
2
) (10.4.12)
Eliminate the divergence (at least to one loop order).
Follow treatment by Lurie
on-shell Renormalization scheme
2
is the mass of the eld (x) (p
2
=
2
on
shell)
(p
2
) = (
2
)
. .. .
When p
2
=
2
diverges
+(p
2
2
)
t
(
2
)
. .. .
log
divergence
+Converging terms
. .. .
c
(p
2
)
(10.4.13)
where
t
(p
2
) =
d
dp
2
(p
2
). Thus,
i(p
2
) =
i
p
2
2
0
(
2
)
. .. .
(p
2
2
)
t
(
2
)
c
(p
2
)
(10.4.14)
* - let
2
=
2
0
+(
2
) (
2
nite,
2
0
, (
2
) diverge if n = 4). This is mass renormalization,
eliminates quadratic divergences. (c.f. recall gure 10.4.1, from previous notes from QFTI).
Thus,
Figure 10.4.1: Recall mass on string
example. x
0
gets renormalized by
m
g
. x
t
0
= x
0
+
mg
k
i(p
2
) =
i
(p
2
2
)(1
t
(
2
))
c
(p
2
)
Pole occurs at p
2
=
2
=
i
1
1
(
2
)
p
2

c
(p
2
)
1
(
2
)
1
1
(
2
)
p
2
c
(p
2
)
to lowest order in
0
(10.4.15)
Let
Z
=
1
1
t
(
2
)
1 +
t
(
2
) (log divergent) (10.4.16)
i(p
2
) =
iZ
p
2
c
(p
2
)
= F.T.'0[T
0
(x)
0
(y)[0`
connected
(10.4.17)
(F.T. = Fourier Transf.). Now let
(x) = Z
1/2

0
(x) (10.4.18)
(re-scaling of
0
eld in L). Hence:
F.T.'0[T(x)(y)[0` =
i
p
2
c
(p
2
)
(Finite) (10.4.19)
What diagrams do diverge?
The supercial degree of divergence = D. For any diagram,
(g)
#
d
4
1
. . . d
4
L
1
2
1
2
. . .
1
2
L
2
; (g
#
) coupling const.
= 4L 2I (only if g is dimensionless) (10.4.20)
L = # of loops
I = # of internal lines
V = # of vertices
= '0[T
1
. . .
4
[0` (10.4.21)
4V = 2I +E (E = # of external legs) (10.4.22)
The 4 is in front of the V because each vertex has 4 lines going out. Check: (gure 10.4.2)
In gure 10.4.2, L = 2, I = 3, V = 2. Thus,
Figure 10.4.2: E = 2, I = 3, V = 2
4(2) = 2(3) + 2
8 = 6 + 2 Check (10.4.23)
Also,
L = I (V 1)
= I V + 1 (10.4.24)
Hence,
D = 4L 2I
4V = 2I + E 2I = 4V E 4I 8V = 2E
D = 4L (4V E)
= 4(I V + 1) 4V + E
= 4I 8V + 4 + E
= 2E + 4 +E
D = 4 E Depends only on E!! (10.4.25)
Thus, D = 4 E as is dimensionless.
D = 4 if (E = 0) not possible
D = 3 if (E = 1) not possible , L even in . There must be a symmetry,
E = 1
D = 2 if (E = 2) Quadratic divergence + Logarithmic
D = 0 if (E = 4) Log divergence
D < 0 if (E > 4) Converges (Higher pt. diagrams are ok).
Now, consider for example a
p
theory: Use the following scalar eld theory in eld theory
in four dimensions
L =
1
2
()
2
1
2
m
2
p
(10.4.26)
(where p 3 is a positive integer) to explain the dierence between theories that are: renor-
malizable, nonrenormalizable and super-renormalizable. (Simply consider the (supercial)
degree of divergence of the Feynman diagrams).
The supercial degree of divergence is given by
D = 4L 2I (10.4.27)
where
L = # of loops
I = # of internal lines
E = # of external lines
V = # of vertices
Each loop momentum k has a volume element d
4
k associated with it, and each (scalar)
internal line is associated with a propagator, which for large [ k[ behaves like [ k[
2
. Now,
each vertex has p lines emerging from it, and each internal line removes two of these, so
E = pV 2I (10.4.28)
and the number of loops is given by
L = I V + 1 (10.4.29)
(only V 1 conservation constraints, due to overall momentum conservation). So, subs.
(10.4.29) into (10.4.27),
D = 4(I V + 1) 2I
= 4I 4V + 4 2I
= 2I 4V + 4 ; Subs. in (10.4.28)
= pV E 4V + 4
= 4 E + (p 4)V (10.4.30)
The conditions are:
If D < 0 super-renormalizable
if D = 0 renormalizable
if D > 0 non-renormalizable
So, as can be seen from (10.4.30), if p < 4, then D will become negative very quickly as V
increases, so p = 3 theories would be super-renormalizable; for p = 4, we get the usual
4
theory, which is renormalizable; and for p > 4, the nal term will fast outweigh the rst two,
D will be positive as more and more vertices are added, and the theory will be (supercially)
non-renormalizable.
Feb. 9/2000
Recall:
F.T.'0[T
0
(x)
0
(y)[0` =
iZ
p
2
c
(p
2
)
(10.4.31)
where
c
(
2
) = 0 (10.4.32)
0
= Z
1/2
(10.4.33)
F.T.'0[T(x)(y)[0` =
i
p
2
c
(p
2
)
(10.4.34)
The four-point function:
F.T.'0[T
0
(x
1
) . . .
0
(x
4
)[0` =
= i
0
+ (t) + (s) + (u) + . . . (10.4.35)
=
1
2
(i
0
)
2
d
n
(2)
n
2
0
i
( + p
1
p
3
)
2
2
0
(p
1
p
3
)
2
= (t) (10.4.36)
(recall the Mandelstem Variables:)
s = (p
1
+p
2
)
2
t = (p
1
p
3
)
2
u = (p
1
p
4
)
2
These variables satisfy:
s + t + u = p
2
1
2p
1
p
3
+ p
2
3
+ p
2
1
+ 2p
1
p
2
+ p
2
2
+p
2
1
2p
1
p
4
+ p
2
4
But p
1
+ p
1
= p
3
+p
4
, p
2
i
=
2
= 6
2
+ 2(p
1
p
3
+ p
1
p
2
p
1
p
4
)
= 6
2
+ 2[p
1
p
3
+ p
1
p
2
p
1
(p
1
+ p
2
p
3
)]
= 6
2
2p
2
1
= 4
2
(10.4.37)
The truncated Fourier transform of '0[
0
(x
1
) . . .
0
(x
4
)[0` means the external legs are
removed. Hence,
Truncated F.T.'0[
0
(x
1
) . . .
0
(x
4
)[0` = i
0
+ (s) + (t) + (u) (10.4.38)
where
(s)
d
4
dx
x
log divergent (10.4.39)
(c.f. D = 4 E . . . E = 4 D = 0). Somehow must get rid of divergences in (s), (t),
(u). Expand (s), (t), (u) about s =
4
2
3
, t =
4
2
3
, u =
4
2
3
.
Truncated F.T.'0[
0
(x
1
) . . .
0
(x
4
)[0` = i
0
+ 3
4
2
3
+

(s) +

(t) +

(u)
. .. .
Finite
Now let iZ
1

0
= i
0
+ 3
4
2
3
= iZ
1

0
+

(s) +

(t) +

(u) (10.4.40)
What we want to compute is
'0[T(x
1
) . . . (x
4
)[0` ( are Renormalized elds) (10.4.41)
But
(x) = Z
1/2

0
(x) (10.4.42)
Recall
F.T.'0[T
0
(x)
0
(y)[0` =
iZ
p
2
c
(
2
)
(10.4.43)
Now:
G
0
(x
1
, . . . , x
4
) = F.T.'0[
0
(x
1
) . . .
0
(x
4
)[0`
=
=
(where the circle in the line in the second diagram =
iZ
p
2
i
2
.
0
(x
1
, . . . , x
n
) = Truncated
n-pt. unrenormalized G.F.
G(x
1
, . . . , x
n
) = '0[T(x
1
) . . . (x
n
)[0`
= Z
n/2
'0[T
0
(x
1
) . . .
0
(x
n
)[0`
= Z
n/2
G
0
(x
1
. . . x
n
) (10.4.44)
But then
(x
1
, . . . , x
n
) = Z
n/2

0
(x
1
, . . . , x
n
) (10.4.45)
We have
F.T.
0
(x
1
, . . . , x
n
) = iZ
1

0
+

(s) +

(t) +

(u) (10.4.46)
Thus,
F.T. (x
1
, . . . , x
n
) = Z
2
iZ
1

0
+
C(
2
0
)
. .. .
(s) +

(t) +

(u)
(10.4.47)
(recall that Z
2
1 +

0
2
2
). To order
2
0
,
F.T. (x
1
, . . . , x
4
) iZ
2
Z
1

0
+

(s) +

(t) +

(u) (10.4.48)
10.5. DIVERGENCES AT HIGHER ORDERS 185
Let = Z
2
Z
1

0
(Renormalized Coupling - where Z
2
arises from 2 pt. function, and Z

1
arises from 4 pt. function - Z

2
, Z
1
(cancel)). So,
F.T. (x
1
, . . . , x
4
) = i +

(s) +

(t) +

(u) (Finite) (10.4.49)
aside
F.T.'0[T(x
1
)(x
2
)[0` =
i
p
2
c
(p
2
)
= F.T. G(x
1
, x
2
) (10.4.50)
F.T. (x
1
, x
2
) =
p
2
c
(p
2
)
i
(10.4.51)
end aside
10.5 Divergences at Higher orders
10.5.1 Weinbergs Theorem
A Feynman integral is convergent if its degree of divergence is negative and the degree
of divergence of any integral associated with a subdiagram is also negative. Recall:
4
:
D = 4 E
Figure 10.5.1:
Convergence
Figure 10.5.2:
Subdiagram is
divergent:
whole diagram is
divergent
In gure 10.5.3,
1. All subdiagrams have D < 0
Figure 10.5.3:
2. D < 0 for entire diagram
W.Th. Overall diagram convergent! (i.e. no unanticipated divergences will appear).
Thus, all divergences can be eliminated by an iterative procedure (i.e. rst eliminate
divergences in subdiagrams, and then in the diagrams as a whole (if there are any)).
Feb. 10/2000
Consider
L =
1
2
(
)
2
m
2
2

2

6
6!
(4 D) (10.5.1)
What goes wrong? Consider:
d
4
(2)
4
2
m
2
1
( + p)
2
m
2
divergent (10.5.2)
We can add in
8
8!
to L to absorb this divergence. Once we have
8
, we can examine the
12-point function:
d
4
(2)
4
2
m
2
1
( + p)
2
m
2
(10.5.3)
We now need
12
12
12!
in order to absorb this new divergence. This goes on and on - in total,
there will be an innite number of vertices to absorb all of these divergences.
Consider a diagram with
n
i
# of vertices of type i
b
i
# of Bosons in the i
th
vertex
f
i
# of Fermions in the i
th
vertex
d
i
# of derivatives in the i
th
vertex
B # of external Bosons
I
B
# of internal Bosons
F # of external Fermions
I
F
# of internal Fermions
L # of loops
D degree of divergence
b = 1
f = 2
d = 1
b = 1
f = 2
d = 1
(where F
= (
)) So,
2I
B
+ B =
i
n
i
b
i
2I
F
+ F =
i
n
i
f
i
# of loops L = I
B
+ I
F

i
n
i
+ 1
(where the 1 is present in L because one of the -functions is superuous).
D = 4L 2I
B
I
F
+
i
n
i
d
i
Now eliminate L, I
B
, I
F
.
D = 4 B
3
2
F +
i
n
i
i
(10.5.5)
where
i
= b
i
+
3
2
f
i
+d
i
4.
The i
th
vertex:
L
I
= g
d
4
x
....
p
4
1
. . .
f
i
. .. .
p
3f
i
/2
1
. . .
d
i
. .. .
p
d
i
1
. . .
b
i
. .. .
p
b
i
(10.5.6)
This means that the dimension of g must be p
i
in order to keep overall dimensionless.
(recall:
d
4
x
....
p
4
....
p
3/2
....
p
+1
....
p
3/2
dimensionless). With a dimensionless coupling constant,
i
= 0 and
D = 4 B
3
2
F (10.5.8)
and then there are only a certain number of divergent diagrams.
For
i
> 0, the degree of divergence increases with more vertices of type n
i
(BAD). (ex.
6
in 4 dimensions). NON-RENORMALIZABLE THEORY!
i
< 0; ex.
3
in 4 dimensions -
i
= 3 4 = 1 < 0 Only a nite # of divergent
diagrams. In fact:
are the only fundamental divergent diagrams. SUPERRENORMALIZABLE.
This (above) treatment is a bit primitive - we are only counting powers of momenta in
diagrams to determine divergence.
In Gauge theories, there is a cancellation of divergences between 2 dierent diagrams.
(See for example Figure 10.5.4)
Figure 10.5.4:
Example of gauge
theories where
(here) the diver-
gences in last 3
diagrams cancel
For example
L =

(p m) +
1
2
(
)
2
M
2
2

2
g

4!
4
(10.5.9)
(with g dimensionless). There is still a diagram that is divergent:
d
4
()
4
We require
4!
4
term to absorb this. The following diagram
(
)
2
(10.5.11)
is convergent.
d
4
For massive vectors, the lagrangian contribution is

L =
1
4
(
)
2
m
2
2
V
(10.5.12)
Note that the last term isnt gauge invariant, and so were stuck with a longitudinal polar-
ization (
m
2
) in the propagator; The propagator is:
m
2
k
2
m
2
+ i
(10.5.13)
The integral goes like (k
2
)
0
constant. This causes divergences.
L
I
=

+ g(
)
4
(10.5.14)
Then we could have
d
4
()
4
(Higgs Mechanism)
10.6 Renormalization Group
There is an arbitrariness inherent in renormalization. ex. Self-energy in
4
:
i(p) =
i
p
2
2
0
(p
2
)
= (10.6.1)
If were just interested in eliminating the innities, we could expand about some point
2
.
Let
(p
2
) = (
2
)
. .. .
Quadratic
divergence
+(p
2
2
)
t
(
2
)
. .. .
log
divergence
1
2!
(p
2
2
)
(
2
)
. .. .
=
c
(p
2
,
2
)
(converges)
+. . .
=
1
p
2
2
0
+ (
2
) (p
2
2
)
t
(
2
)
c
(p
2
,
2
)
Wave function renormalization is done rst:
=
i
(1
(
2
))
p
2
2
0
+(
2
)
2
1
(
2
)

c
(p
2
,
2
)
(1
t
(
2
))
. .. .
c
(p
2
,
2
)
Let
2
=

2
0
+ (
2
)
2
1
t
(
2
)
i(p) =
iZ
p
2
c
(p
2
,
2
)
; Z
=
1
1
t
(
2
)
1 +
t
(
2
) (10.6.2)
where arbitrary. Note
1.
2
no longer the physical mass of (
c
(
2
,
2
) = 0,
c
(
2
,
2
) = 0).
2
p
is the
physical mass, given by
2
p
c
(
2
p
,
2
) = 0 (10.6.3)
2. How does something physical depend on ?
See New methods for the renormalization group, J.C. Collins and A.J. Mac-
farlane, Phys. Rev. D, Vol 10, Number 4, 15 August 1974.
Feb. 14/2000
Momentum subtraction is dicult to work out in practice.
10.6. RENORMALIZATION GROUP 191
Minimal Subtraction in conjunction with dimensional regularization is the easiest way to
renormalize. c.f. Collins and MacFarlane.
L =
1
2
(
0
)
2
m
2
0
2

2
0

0
4!
4
0
(10.6.4)
L
(2)
=
1
2
(
0
)
2
m
2
2

2
0
(m renormalized) (10.6.5)
L
I
=
(m
2
0
m
2
)
2

2
0
. .. .
0
4!
4
0
(10.6.6)
Again, we have,
F.T.'0[T
0
0
[0`
Truncated
=
i(p
2
) =
1
2
(i
0
)
4n
....
d
n
(2)
n
i
(
2
m
2
)
+
i
2
(m
2
m
2
0
) (10.6.7)
* thus
0
=
42n+n
=
4n
. Keep in mind, were working in n dimensions. The Action
is:
S =
d
n
x
....
a
1
2
(
....
b
0
)
2

0
4!
4
0
(10.6.8)
Where the dimensions are: [a] =
n
, [b] =
+2
. Since the dimension of the action must be
zero, this means that the dimension of [
0
]
(2+n)/2
.
Note: as expected, if n = 4,
0
is dimensionless.
i(p
2
) =
1
2
(i
0
)
4n
d
n
(2)
n
i
(
2
m
2
)
+
i
2
(m
2
m
2
0
)
(p) =

4n
0
2(4)
n/2
1
n
2
(m
2
)
1+n/2
+
1
2
(m
2
m
2
0
)
=

0
m
2
(4)
n/2
1
n
2
m
2
2+n/2
+
1
2
(m
2
m
2
0
)
Let n 4
=

0
m
2
(4)
n/22+2
2
n
2
1
n
2
m
2
2+n/2
+
1
2
(m
2
m
2
0
)
Let = 2
n
2
( 0 as n 4).
=

0
m
2
(4)
2
(1 + ln(4) + . . .)
. .. .
(4)
n/2+2

E
+O()
. .. .
(
2
n
2
)
(1)
1 + +O(
2
)
. .. .
(
1
n
2
)
1
1 ln
m
2
+O()
. .. .
m
2
2+n/2
+
1
2
(m
2
m
2
0
)
=

0
m
2
(4)
2
(1)
+ ln(4)
E
ln
m
2
+
1
2
(m
2
m
2
0
) +O()
Let m
2
0
= m
2
1
2
0
(4)
2
.
=

0
(4)
2
ln(4)
E
ln
m
2
= Finite (10.6.10)
(where the ln
m
2
is a residual dependence on renormalization).

We could also have absorbed the ln(4),
E
in to m
2
0
, or
2
. Let
t2
=
2
e
E
ln(4)
.
ln
m
2
= ln
m
2
t2
+ ln
e
(
E
ln(4))
(10.6.11)
So wed get
(p) =

0
(4)
2
ln
m
2
t2
(10.6.12)
( modied minimal subtraction: gets rid of ln(4) and
E
.)
So also, for the four-point function:
. .. .
i
0
4+n
+
. .. .
absorb terms of
1
into
0
(10.6.13)
Net Result:
4+n
= +
=1
a
(, m, )
(n 4)
(10.6.14)
m
0
= m
1 +
=1
b
(, m, )
(n 4)
= mZ
(10.6.15)
Z
= 1 +
=1
c
(, m, )
(n 4)
(10.6.16)
Note: a
, b
, c
in minimal subtraction are independent of m

2
and
2
in practice.
Note also:
a
i
() =
j=i+1
a
ij
i = # of powers of
1
n4
j = # of loops
(10.6.17)
i.e.
. .. .
2
#
(n4)
+
. .. .
#
(n4)
2
+
#
(n4)
3
(10.6.18)
(and the sum goes higher, etc.). So also,
b
i
=
j=i+1
b
ij
j
(10.6.19)
c
i
=
j=i+1
c
ij
j
(10.6.20)
Note that

0
4
0
4!

0
occurs in Feynman rules, but isnt nite. ex: see gure 10.6.1.
Figure 10.6.1: Overall divergence
comes only partially from the dia-
gram
0
also has poles due to cal-
culated graph at 1-loop order.
Net result:
R
(p, m, , ) = lim
n4
R
(p, m(n), (n), , n) (10.6.21)
But now,
R
(p, m(n), (n), , n) =

p,
n4
, n
m
0
Z
1
m
((
0
)) , , n
= Z
N/2

0
(p
i
, m
0
,
0
, n) (10.6.22)
(N is the # of external legs - Note dependence vanishes. cant be dependent on
L.H.S.)
Feb. 16/2000
or
R
= Z
(
0
n4
, n)
0
(p,
0
(n), m
0
(n), n) (10.6.23)
with Z
= Z
N/2
d
d
0
= 0
=
d
d
lim
n4
Z
1
R
(p, (
0
n4
, n), m(
0
n4
, n), , n)
0 =
+Z
m
m
Z
1
m
m

Z
Z
1
R
Now set
()
m
m Z
m
m
Z
1
m
Z
1
, and let n 4.
0 =
+ ()

m
m

m

(10.6.24)
Note: as is nite, ,
m
and
are nite.
Suppose we just look at the engineering dimension (depends only on the # of external
legs), D
, of = (p, , m, ).
D
= 2
D
= +2
D
= 0
Consider p
= p
0
(p
0
reference momentum).
10.6.1 Eulers Theorem
f(x, y) =
D
f(x, y)
d
d
at = 1 (10.6.25)
x

x
+ y

y
f = D
D1
F (10.6.26)
From this, if = (p
0
, , m, ), then
+ m

m
+

= D
(10.6.27)
Now, eliminate

from (10.6.24), (10.6.27):

0 =
()

+ (1 +
m
)m

m
+ (
(p
0
, , m, ) (10.6.28)
Can trade in dependence on mass scale parameter for dependence on scale in front
of momentum p
0
.
By rescaling momentum (changing ) will eectively change value of coupling con-
stant/mass parameter (, m).
() =
,
m
= Z
m
Z
1
m
.
Formal Solution
(p
0
, , m, ) =
D
exp
(
t
))
d
t
(p
0
,
(), m(), ) (10.6.29)

with
()
= (
())
m()
1 +
m
(
())
m()
and with boundary conditions
(1) = , m(1) = m.
For example, consider the 2-pt. function:
+
. .. .
A
+
. .. .
B
+
+ + +. . .
. .. .
C
+. . .
The net result of renormalization will be:
A =
A
1
ln
p
2
+ B
1
(10.6.30)
B = A
2
ln
2
p
2
+ B
2
ln
p
2
+C
2
(10.6.31)
C = A
3
ln
3
p
2
+ B
3
ln
2
p
2
+ C
3
ln
p
2
+ D
3
(10.6.32)
were just taking the leading order terms, (A
t
s), all terms subsumed into

(). Thus,
()
= (
())
=
1
2
....
lowest
order
+
2
3
+
3
4
+ . . . (10.6.33)
For perturbation theory to make sense,

() must be small.
If,
lim
() = 0 (10.6.34)
we have asymptotic freedom (Yang-Mills Theory)
If
lim
0
() = 0 (10.6.35)
we have infrared freedom (almost everything else).
Feb. 17/2000
Recall
()
= (
()) (10.6.36)
m()
1 +
m
(
())
m() (10.6.37)
In Dimensional Regularization
n4
= +
=1
a
()
(n 4)
(10.6.38)
The
n4
was introduced so that is dimensionless.
0

4n
() =

Now, how do we get ()?
() = x
0
+ x
1
(n 4) + x
2
(n 4)
2
+. . . (10.6.39)
(
0
n4
) =
....
()
=1
a
()
(n 4)
(10.6.40)
Subs. in ()
(n 4)
0
n4
. .. .
subs. (10.6.38)
=
x
0
+ x
1
(n 4) +x
2
(n 4)
2
+ . . .
1 +
=1
a
t
()
(n 4)
(n 4)
=1
a
()
(n 4)
= [x
0
+x
1
(n 4) + . . .]
1 +
=1
a
t
()
(n 4)
Thus,
(n 4) + a
1
+
a
2
(n 4)
+
a
3
(n 4)
2
+. . . =
x
0
+ x
1
(n 4) + x
2
(n 4)
2
+ . . .
+
+
x
0
a
t
1
(n 4)
+x
1
a
t
1
+ x
2
a
t
1
(n 4) + . . .
x
0
a
t
2
(n 4)
2
+
x
1
a
t
2
(n 4)
+ x
2
a
t
2
+ . . .
+ . . .
We set x
2
= x
3
= x
4
= . . . = 0.
(n 4) + a
1
+
a
2
(n 4)
+ . . . = x
1
(n 4) + (x
0
+ x
1
a
t
1
)
= +(x
0
a
t
1
+ x
1
a
t
2
)
1
(n 4)
+(x
0
a
t
2
+ x
1
a
t
3
)
1
(n 4)
2
(10.6.41)
Matching powers of (n 4):
x
1
=
x
0
+ x
1
a
t
1
= a
1
x
1
=
x
0
= a
1
a
t
1
() = [a
1
a
t
1
] + (n 4)
x
0
a
t
1
+ x
1
a
t
2
= a
2
.
.
.
x
0
a
t
+ x + 1a
t
+1
constraint eqs that x a

2
, a
3
, . . . , a
in terms of a
1
(Note as n 4, () = (a
1
a
t
1
)). Thus,
n4
=
=1
a
()
(n 4)
(10.6.42)
only a
1
is needed for ()
a
1
xes a
2
, a
3
, . . .
Remember that
4
in 4-D has the form:
a
1
() = a
12
....
From 1-loop
diagrams
2
+ a
13
....
2-loop
3
+ a
14
....
3-loop
4
+ . . . (10.6.43)
Recall that:
a
j=+1
a
j
j
(10.6.44)
ex: for
a
2
= x + 0a
t
1
+ x
1
a
t
2
a
23
3
+ a
24
4
+ . . . =
a
12
2
+a
13
3
+ . . .
d
d
a
12
2
+ a
13
3
+ . . .
d
d
a
12
2
+a
13
3
+ . . .
d
d
a
23
3
+ a
24
4
+. . .
Here, a
2j
can be solved for in terms of a
1j
. Similarly, we can show that, if
Z
m
= 1 +
=1
b
(n 4)
, Z
= 1 +
=1
c
(n 4)
(10.6.45)
Then (***)
m
() =
b
t
1
m
(10.6.46)
= c
t
1
(10.6.47)
(c.f. Collins & MacFarlane)
10.6.2 Explicit Calculations
L =
1
2
(
)
2
1
2
m
2
0

0
4!
4
(10.6.48)
L
(2)
=
1
2
(
)
2
1
2
m
2
2
(10.6.49)
L
I
=
1
2
(m
2
0
m
2
)
2
4
4!
(10.6.50)
Two point function
. .. .
A
+
. .. .
B
+
. .. .
0
(1+one contribution
to
0
)
+
+ +
(* - because it is itself renormalized). We nd that, to 1-loop order,

S
1
F
=
A
. .. .
(p
2
m
2
)
B
. .. .
m
2
b
11
(n 4)
1-loop contrib.
. .. .
m
2
2
16
2
(n 4)
+ nite +(two-loop) (10.6.51)
i.e.
m
2
0
= m
2
1 +
=1
b
()
(n 4)
(10.6.52)
(we just have to take lowest order term. (b
2
() = b
11
+ b
12
2
+ . . .). S
1
F
will be nite
if we choose

b
11
=

16
2
(10.6.53)
Note, no divergence p
2
(could be, in principle, but it just doesnt occur in
4
theory).
Z
= 1 to one-loop order
For the 4-pt. function
=
. .. .
A
+ + (two-loop)
=
0
+

2
0
(n 4)
3
16
2
+ (nite) (10.6.54)
but,
0
= +
a
12
2
(n 4)
+ . . .
= +
a
12
2
(n 4)
+

2
(n 4)
3
16
2
+. . .
arrange for pole from A and from 1-loops to cancel
a
12
=
3
16
2
(10.6.55)
Feb. 28/2000
Recall:
+
= m
2
b
11
(n 4)
+
b
12
2
(n 4)
+ . . .
. . .
S
1
F
= p
2
m
2
1 +
b
11
(n 4)
m
2
16
2
(n 4)
+ nite +O(
2
) (10.6.56)
where b
11
=
1
16
2
, Z = 1.
= +O(
3
)
= +

2
a
12
(n 4)
+
3
2
(n 4)16
2
+ nite +O(
3
) (10.6.57)
with a
12
=
3
16
2
. Higher order contributions to the 2-pt. function are:
=
1
2
1
(4)
n/2
1
n
2
(m
2
)
(n/21)
4+n
+
a
12
2
(n 4)
= O() +
m
2
32
2
2a
12
(n 4)
2
+
2 + ( 1)a
12
(n 4)
+ nite
(10.6.58)
(where the vertex of the loop with the propagator is
0
= +
a
12
2
(n4)
).
=

84n
4(2)
2n
d
n
k d
n
1
(k
2
m
2
)
1
(
2
m
2
)
2
=
m
2
2
(16
2
)
2
1
(n 4)
2
+
(4 2)m
2
2
(32
2
)
2
(n 4)
+ nite (10.6.59)
= O() +

2
m
2
a
11
16
2
1
(n 4)
2
+
b
11
2
m
2
32
2
(n 4)
+ nite (10.6.60)
Recall L =
1
2
(m
2
0
m
2
)
2
+ . . .
(where the cross at the top of the loop is due to the mass insertion).
= ( Collins )
=
m
2
2
(16
2
)
2
1
(n 4)
2
+

2
(16
2
)
2
(n 4)
p
2
12
....
m
2
2
+ ( 1)m
2
+ nite (10.6.61)
( wave function renormalization)
All together (to 2-loop order - only showing two loop order here):
S
1
p
= p
2
1

2
12(16
2
)
2
1
(n 4)
m
2
1 +

2
(n 4)
b
12
1
2(16
2
)
2
. .. .
would think
b
12
=
1
2(16
2
)
2
+

2
(n 4)
2
b
22
2
(16
2
)
2
. .. .
and here
b
22
=
2
(16
2
)
2
(10.6.62)
First!! Remember that
FT '0[T[0`
Z
(p
2
m
2
nite )
(10.6.63)
Thus,
S
1
= Z
1
p
2
m
2
nite
(10.6.64)
Thus,
Z
1
= 1

2
12(16
2
)
2
1
(n 4)
(10.6.65)
So,
S
1
= Z
1
p
2
m
2
nite
1

2
12(16
2
)
2
1
(n 4)
p
2
m
2
1 +

2
(n 4)
b
12
1
2(16
2
)
2
+
1
12(16
2
)
2
+

2
(n 4)
2
b
22
2
(16
2
)
2
(10.6.66)
Thus,
Z
= 1 +

2
12(16
2
)
1
(n 4)
(10.6.67)
b
22
=
2
(16
2
)
2
b
12
=
5
12(16
2
)
2
m
2
0
= m
2
1 +
1
(n 4)

16
2
+
5
2
12(16
2
)
2
+
2
2
(16
2
)
2
1
(n 4)
2
+O(
3
)
(10.6.68)
0
=
4n

3
2
16
2
1
(n 4)
+O(
3
)
(10.6.69)
Thus,
a
1
() =
3
2
16
2
+O(
3
) (10.6.70)
() =
a
1
()
=
3
2
16
2
(10.6.71)
Thus the running coupling is the solution to:
()
= (
())
0
d
t
t
=

()
)
(10.6.72)
originally
+ ()

(p, , ) = 0(?)
(p
0
, , ) (p
0
, ,
())
+ ()

?
....
= 0
(10.6.73)
The boundary conditions are:

(
0
) = . So,
ln (
t
)
0
=

()
2
16
2
ln
=
16
2
3
()
3
16
2
ln
=
1
()
+
1
() =
1
1

3
16
2
ln
() =

1
3
16
2
ln
as ( unstable )
0
+
as 0
+
( stable *)
(10.6.74)
i.e. perturbation theory acceptable for small momenta.
Thus, as 0
+
, perturbation theory reliable.
()
= (
()) (10.6.75)
The plot of this can be seen in Figure 10.6.2. If we include higher orders, we get Figure
10.6.2.
Mar. 1/2000
()
=
3
16
2
2
()

()
t
()
t2
()
=
3
16
2
0
d
t
t
(10.6.76)
() =

1
3
16
2
ln
0
(10.6.77)
Figure 10.6.2: Beta function
plot
Figure 10.6.3: Beta function
plot 2 - note = 0!! - depends
on where you start. (
()
xed
as )
B.C.

(
0
) = .
We often let = at
0
= . i.e.

() = . Thus,

()
d
t
t2
=
3
16
2
ln(
t
)
()
=
3
16
2
ln
=
3
16
2
ln
() =
16
2
3 ln
(10.6.78)
Note that:
0,

() 0
+
,

() 0
(unacceptable)
,

() perturbation breaks down.
In QCD
L =
1
4
F
a
2
+

(i g
a
A
a
) +
iegT
a
A
a
2
ij

j
(10.6.79)
(g) =
g
3
16
2
11
3
C
2
(v) +
4
3
C
2
(F) +
1
3
C
2
(S)
(10.6.80)
T
a
, T
b
= c
abc
T
c
(10.6.81)
Figure 10.6.4: Beta func-
tion plot for pure Yang-
Mills.
Figure 10.6.5: Beta func-
tion plot for pure Yang-
Mills - two-loop order.
Then,
C
2
(v)
ab
= c
amn
c
bmn
= n for SU(n) (10.6.82)
ab
C
2
(F) = Tr(
a
b
)
=
1
2
for quarks in QCD (10.6.83)
C
2
(S)
ab
= Tr(T
a
T
b
)
=
1
2
for complex scalars in the fundamental rep. for SU(3) (10.6.84)
For pure SU(n) Yang-Mills theory,
(g) =
11
3
n
g
3
16
2
(10.6.85)
whose graph looks like Figure 10.6.4 Thus,
d g()
d
=
11
3
n g
3
()
16
2
(10.6.86)
d g
d
< 0 as asymptotic freedom (10.6.87)
g()
dg
t
g
t3
=
11n
3(4)
2
d
t
t
(10.6.88)
(B.C. g() = )
1
2g
t2
g()
=
11n
3(4)
2
ln
g
2
=
3(4)
2
22nln
, g
2
0
< , g
2
() < 0 (Not acceptable)
(10.6.89)
For N
f
avours of quarks, (see Figure 10.6.5)
(g) =
g
3
16
2

11
3
SU(3)
....
(3) +
4
3
1
2
N
f
. .. .
11+
2N
f
3
<0 if N
f
<16
1
2
avours
(10.6.90)
Chapter 11
Spontaneous Symmetry Breaking
4
4
(the subscript is the number of dimensions).
L =
1
2
(
)
2
m
2
2
2

4
4!
(11.0.1)
Symmetry: (must be respected. ex: if we have a 5 pt. function Greens functions
are negative, must be 0).
H =
t
L (11.0.2)
=
L
t

t
=
1
2
2
+ ()
2
. .. .
Kinetic part
+
m
2
2

2
+

4
4!
. .. .
Potential V ()
(11.0.3)
Lowest Energy state:
KE
0
= 0 if = const.
(11.0.4)
Energy = V ()
=
m
2
2
2
+

4
4!
; > 0 for E to be bounded below (11.0.5)
There are two cases: we can have a positive mass squared (m
2
> 0 - Figure 11.0.1) or a
negative mass squared (m
2
< 0 - Figure 11.0.2). For the rst case,
m
2
2
2
+

4
4!
(11.0.6)
For the second, we get
dV
d
= m62 +

3
6
= 0 if
= 0
=
6m
2
(2 possible vacua) (11.0.7)

207
208 CHAPTER 11. SPONTANEOUS SYMMETRY BREAKING
1
0
1
2
3
4
5
y
3 2 1 1 2 3
phi
Figure 11.0.1: Positive
mass (m
2
> 0) plot.
2
1
1
2
3
4
y
4 2 2 4
phi
Figure 11.0.2: Negative
mass (m
2
< 0) plot
Lets consider excitations above
0
= +
6m
2
.
=
t
+
6m
2
(11.0.8)
The Lagrangian for
t
is:
L =
1
2
[
(
t
+
0
)]
2
m
2
2
(
t
+
0
)
2

4!
(
t
+
0
)
4
=
1
2
(
t
)
2
m
2
2
t2
+ 2
t
0
. .. .
+
2
0

4!
t4
+ 4
t3
0
+ 6
t2
2
0
+ 4
t
3
0
. .. .
+
4
0
t
s cancel using
0
denition
=
1
2
(
t
)
2
m
2
2

t2

4!
t4
+ 4
t3
6m
2
+ 6
t2
6m
2
+ const
=
1
2
(
t
)
2
+
m
2
2
+
3m
2
2
t2

4!
t4

3!
t3
6m
2
=
1
2
(
t
)
2
1
2
(2m
2
)
t2

3!
6m
2

t3
....
4!
t4
(11.0.9)
mass of
t
is (2m
2
) > 0.
breaks the symmetry. This can be seen in Figure 11.0.2 - the bottom of the
two wells will have ground states [0
`, [0
+
` - this ground state doesnt respect symmetry.
For example,
F

F
1
= (suppose)
F[0` = [0` if (m
2
> 0)
F[0
+
` = [0
` if (m
2
< 0)
209
[0
)+[0
+
)
2
symmetry restored
.
Mar. 2/2000
Heisenberg (1-D) Ferromagnet:
1 2 3 4 5 6 7 8

(11.0.10)
(where = +1 for positions (1 + 2), = 1 for positions (3 + 4),. . . ).
H =
i+1
(11.0.11)
=
(+1)
2
+ (+1)(1) + (1)
2
+ . . .
There is an up-down symmetry in this Hamiltonian (no preference for up/down).

The lowest energy states are all up or all down.
Generalize to an O(N) model.
= (
1
,
2
, . . . ,
N
) (11.0.12)
L is invariant under
R (R
T
= R
1
) (11.0.13)
R is a global orthogonal matrix.
L =
1
2
(
)(
)
m
2
2

4!
( )
2
(11.0.14)
The potential for this Lagrangian is
Figure 11.0.3: O(N) model
for m
2
> 0.
Figure 11.0.4: O(N) model
(m
2
< 0) plot Mexican
Hat potential
V () =
m
2
2

2
+

4!
(()
2
)
2
(11.0.15)
(where > 0 for stability). This can be plotted - see gures 11.0.3 (for m
2
> 0) and 11.0.4
(for m
2
< 0).
V
i
= m
2
i
+

4!
(2
2
)(2
i
)
=
m
2
+

6
i
= 0 at
i
= 0 or at [[ =
6m
2
(11.0.16)
Suppose,
0
= V
i1
( in the one direction)
= (V +
t
1
,
t
2
, . . . ,
t
N
)
L =
1
2
(
t
)(
t
)
m
2
2
(V +
t
1
)
2
+
t
2
+. . . +
t
N
. .. .
=V
2
+2V
1
+
2
1
+...
=V
2
+2V
1
+
2
4!
(V +
t
1
)
2
+
t2
2
+ . . . +
t2
N
2
=
1
2
(
t
)
2
m
2
2
V
2
+ 2V
t
1
+
t2

4!
V
2
+ 2V
t
1
+
t2
2
=
1
2
(
t
)
2

4!
2V
2
t2
+ 4V
2
t2
1
+ interaction terms
=
1
2
(
t
)
2
m
2
2

t2

4!
6m
2
2
t2
+ 4
t2
1
=
1
2
(
t
)
2
m
2
2

t2
+ m
2
1
2
t2
+
t2
1
=
1
2
(
t
)
2
1
2
(2m
2
)
t2
1
(11.0.17)
Thus

t
1
mass (2m
2
)

t
2
, . . . ,
t
N
massless (N 1 massless Goldstone Bosons).
In general, if a system of N scalars has an invariance under a group G, and if the ground
state of the theory has an invariance under a group G
t
G, then there is a massless excitation
above this ground state for each generator of G
t
.
Not realized in Nature.
Here we have
G O(N) (Orthogonal group, N-dimensional) (11.0.18)
(see gures 11.0.5,11.0.6). Figure 11.0.7 No massless scalars (spin 0) particles observed
211
Figure 11.0.5: So G
t
O(N 1)
Figure 11.0.6: Excitation
where you just roll around
bottom of well (Correspond
to Goldstone Bosons)
Figure 11.0.7: m
2
> 0 - (No massless
excitations).
in nature.
L =
1
4
F
2
m
2
2
A
2
(Massive spin-1 particle) (11.0.19)

iD
=
i
m
2
k
2
m
2
( whole term is transverse) (11.0.20)
where the
k
m
2
term comes from the longitudinal polarization (Destroys Renormalizability)
- the gauge invariance is ruined by m
2
.
In the Higgs Mechanism, the massless Goldstone Bosons are absorbed by the longitudinal
polarization of the massless vectors to which they are coupled to give massive vector particles.
However, gauge invariance is not sacriced, so renormalizability is retained. The massive
scalars left over are all the Higgs particles.
Mar. 6/2000
11.1 O(2) Goldstone model:
L =
1
2
m
2
2

i

4!
(
i
i
)
2
(11.1.1)
i
= (
1
,
2
)
cos sin
sin cos

1
Let =

1
+i
2
2
,
=

1
i
2
2
.
L = (
)(
) m
2

6
(
)
2
(11.1.2)
e
i
Local gauge invariance

e
i(x)
(11.1.3)
Introduce a gauge eld
L =
1
4
F
+ [( + ieA
] [(
ieA
)] m
2
g(
)
2
(11.1.4)
If m
2
< 0 - (see gure 11.1.1).
Figure 11.1.1: m
2
< 0, V () = m
2
+ g(
)
2
V
= 0 [[ =
m
2
2g
(11.1.5)
11.1. O(2) GOLDSTONE MODEL: 213
Choose
min
=
m
2
2g
(Real) (11.1.6)
=
m
2
2g
+
t
=
V
2
+
t
(11.1.7)
Now,
L =
1
4
F
2
+ ieA
2
+
t

(
ieA
2
+
t
m
2
2
+
t
2
+
t
2
+
t
2
+
t
2
=
1
4
F
2
+ [(
+ ieA
)
t
] [(
ieA
)
t
] +
V
2
[(ieA
t
) (
t
)(ieA
)]
+
V
2
2
e
2
A
. .. .
m
2
t
+
V
2
(
t
+
t
) +
V
2
2
V
2
2
+
V
2
(
t
+
t
) +
t
2
(Note that it would appear, from the * term, that A
has a mass V
2
e
2
). Now, let
D
ieA
.
L =
1
4
F
2
+ (D
t
)
(D
t
) +
ieV
2
[A
t
(
t
)A
t
] +
e
2
V
2
2
A
m
2
t
g
V
2
2
(
t
+
t
)
2
+ 2
V
2
2
t
. .. .
appears that Im(
)
is massless
+
2V
2
(
t
+
t
)
t
t
+ (
t
t
)
2
+ const.(11.1.8)
To determine the actual degrees of freedom, set
(x) =
(x) +V
e
i(x)/V
(11.1.9)
where (x), (x) are real. Then, let
v
(x) = e
i/V
(x) (11.1.10)
B
= A
1
e
(x)
V
(11.1.11)
where (11.1.10,11.1.11) are a gauge transformation. Then,
v
(x) =
(x) + V
2
(11.1.12)
Now we nd that
L =
1
2
(
)
2
1
2
(2m
2
)
2
1
4
(
)
2
+
1
2
(eV )
2
B
+
1
2
e
2
B
(
2
+ 2V ) gV
2
g
4
4
(11.1.13)
(x) disappeared (Goldstone boson has vanished)
(x) mass (2m
2
) Higgs
B
massive vector (3 polarizations) - mass (eV )

2
There are still only 3 parameters that characterize the theory, , e, g. (recall V =
m
2
2g
).
This is the form of the Lagrangian in the U (unitary) gauge - physical degrees of
freedom are apparent.
Renormalizability is only manifest in other gauges (f-gauges) (Note, though, that since
the original theory was renormalizable, this one is also).
In general, for a non-Abelian Field,
L =
1
2
i
+ igT
a
ij
A
a
j
)(
i
igT
b
il
A
b
l
)
V (
i
)
1
4
F
a
F
a
(11.1.14)
If we say
n # of scalar elds
i
(2n if
i
complex).
and if
i
= 0 if
i
=
V
i
2
= 0 (i = 1, . . . , n
t
) (n
t
< n)
then there are then n
t
massive Higgs and (n n
t
) massive vector particles.
Electroweak SU(2) U(1) model
i
i = 1, 2 (Complex doublet of SU(2)) (11.1.15)
(n = 4),
n
t
= 1 One Higgs particle
4 vectors, (4-1) = 3 are massive, and 1 is massless (photon).
Mar. 8/2000
11.2. COLEMAN WEINBERG MECHANISM 215
11.2 Coleman Weinberg Mechanism
L =
1
2
(
)
2
1
2
m
2

4!
4
(11.2.1)
V =
m
2
2

2
+

4!
4
V
tot
= +
. .. .
Classical
+
+ + . . .
. .. .
One-loop
+
+ . . .
. .. .
Two-loop
= Sum of 1PI (irreducible) diagrams with an arbitrary number of external
legs where momenta = 0
Remember (Generator function of all diagrams)
Z[J] =
d e
(i/)
dx (/+J)
(11.2.2)
= e
iW(J)/
(11.2.3)
where W(J) is the generating functional of all connected diagrams.
c
(x) =
W(J)
J(x)
Legendre Transformation (11.2.4)
Renormalized (
c
) = W(J)
dx J(x)
c
(x) (11.2.5)
(where
c
is the 1PI generating functional). Dene the eective potential to be
V () =
n=0
1
n!
(0, . . . , 0)
n
(11.2.6)
We rst of all renormalize ()
We next dene
m
2
R
=
d
2
V
d
2
=
(11.2.7)
R
=
d
4
V
d
4
(11.2.8)
(m
R
= Renormalized mass).
Thus,
V = V (,
R
, m
2
R
) (11.2.9)
We can now determine if there is Spontaneous Symmetry breaking.
Note that even if V
cl
has
min
= 0 (i.e. no spontaneous symmetry breakdown), it is
conceivable that V
tot
does generate the breakdown. If there is spontaneous symmetry and
the Higgs mechanism has generated massive vectors, then there will be no divergences in
this theory with vectors.
Heuristic argument for the renormalizability of models with the Higgs mechanism for
generating massive vectors.
11.3 One loop Eective Potential in
4
model
+ +
c
constant eld (external momenta = 0)
First loop:
d
4
k
(2)
4
i
(k
2
m
2
+ i)
n
th
Loop: S
n
d
4
k
(2)
4
(i)
n
i
(k
2
m
2
+ i)
n
(S
n
Symmetry factor)
2n legs
(2n)!
(2n)2
n
11.3. ONE LOOP EFFECTIVE POTENTIAL IN
4
MODEL 217
(where, in the denominator, the 2n is because we can start anywhere in the circle, and the
2
n
is because the legs are interchangeable). Thus,
(0, . . . , 0
. .. .
2n zero
momenta
) =
(2n)!
2
n
(2n)
d
4
k
(2)
4

k
2
m
2
+ i
n
(11.3.2)
n = 0 (constant term) (11.3.3)
Therefore,
V () =
n=1
1
(2n)!
(0, . . . , 0)
2n
c
=
n=1
1
(2n)!
(2n)!
2
n
(2n)
d
4
k

2
c
k
2
m
2
+ i
n
=
d
4
k
(2)
4
1
2
n=1
1
n
2
c
2
k
2
m
2
+ i
n
(n = 1, 2diverges)
=
d
4
k
(2)
4
1
2
ln
2
c
2
k
2
m
2
+ i
(as ln(1 x) = x +
x
2
2
+
x
3
3
+. . .)
=
d
4
k
(2)
4
1
2
ln
k
2
m
2

2
c
2
ln
k
2
m
2
+ i
. .. .
constant
(11.3.4)
The minus sign means we must still renormalize. Note that the second (constant) term is a
problem in G.R. -
d
4
x

g
k Cosmological constant = 0 actually.

Alternative Derivation
Z(J) =
d exp
d
4
x (L() +J())
(11.3.5)
Let (x) =
c
+ h(x) = (constant eld) + (quantum uctuations about
c
).
Z(J) =
dhexp
[L(
c
+h) + Jh] d
4
x
(11.3.6)
Consider only diagrams with external
c
s and are 1PI to get V (
c
). But,
L = L(
c
+ h)
=
1
2
(
h)
2
m
2
2
(
c
+h)
2

4!
(
2
c
+ 2
c
h + h
2
)
2
=
1
2
(
h)
2
m
2
2
(
2
c
+ 2
c
h + h
2
)

4!
4
c
+ 4
3
c
h + 6
2
c
h
2
+O(h
3
)

The only terms that will contribute to V (
c
) at one-loop order are O(h
2
). i.e. One-Loop
1
2
(
h)
2
m
2
2
h
2
(< hh >) (11.3.7)
2
c
h
2
(11.3.8)
One loop
Two-loop
h
3
, (11.3.9)
Then,
Z =
dhexp
d
4
x
1
2
(
h)
2
m
2
2
h
2
2
c
h
2
dh exp
d
4
x
1
2
ln
p
2
m
2

2
c
2
= det
1/2
p
2
+ m
2
+

2
c
2
(11.3.10)
Mar. 9/2000
V () = i
d
4
k
(2)
4
1
2
n=1
1
n
1
2
2
k
2
m
2
+ i
=
i
2
d
4
k
(2)
4
ln
k
2
m
2
1
2
2
+ i
(11.3.11)
11.3. ONE LOOP EFFECTIVE POTENTIAL IN
4
MODEL 219
Recall we can get the same thing from:
Z(J) =
dhexp
d
4
x
1
2
h(x)
p
2
m
2

2
c
2
h(x)
= det
1/2
p
2
m
2

2
c
2
(1-loop 1PI contribution to Z if

c
is a constant)
= e
iW
(11.3.12)
where we dene
iW = ln
det
1/2
p
2
m
2

2
c
2
W =
i
2
ln
det
p
2
m
2

2
c
2
= (
c
)
dx
c
J
. .. .
=0
(11.3.13)
Thus,
(
c
) =
i
2
ln
det
p
2
m
2

2
c
2
(11.3.14)
Note:
ln(det(O)) = ln(det(U
OU)) ; O = O
= ln
det
1
.
.
.
= ln (
1
. . .
n
)
= ln(
1
) + . . . + ln(
n
) (11.3.15)
and
Tr(ln(O)) = Tr(ln(U
OU))
= Tr
ln
1
.
.
.
= ln(
1
) + . . . + ln(
n
) (11.3.16)
Then,
Tr(ln(O)) = ln(det(O)) (11.3.17)
So
(
c
) =
i
2
ln
det
p
2
m
2

2
c
2
=
i
2
Tr
ln
p
2
m
2

2
c
2
=
i
2
dp 'p[ ln
p
2
m
2

2
c
2
[p`
=
i
2
d
4
k
(2)
4
ln
k
2
m
2

2
c
2
(11.3.18)
Now, to compute V (
c
) in closed form.
11.4 Dimensional Regularization
V (
c
) =
i
2
d
n
k
(2)
2
n
ln
k
2
m
2

2
c
2!
+ i
(11.4.1)
ln(a + i) ln(b + i) =

0
d(it)
it
e
i(b+ia)t
e
i(a+i)
(11.4.2)
i.e.:
d
da
ln(a + i) =
d
da

0
d(it)
it
e
i(a+i)t
(1)
1
a + i
=

0
d(it) e
i(a+i)t
=
e
i(a+i)t
a + i
0
=
1
a + i
(11.4.3)
f
L
(a, b) = f
R
(a, b) + k (11.4.4)
f
L
a
=
f
R
a
(11.4.5)
f
L
b
=
f
R
b
(11.4.6)
k
a
=
k
b
= 0 (11.4.7)
11.4. DIMENSIONAL REGULARIZATION 221
V (
c
) =
i
2
d
n
k
(2)
n
ln
k
2
m
2

2
c
2
+ i
ln
k
2
m
2
+ i
=
i
2
d
n
k
(2)
n
d(it)
it
e
i(k
2
m
2
+i)t
e
i
k
2
m
2
2
c
2
+i
=
i
2
d
n
k
(2)
n

0
d(it)
it
e
i
k
2
m
2
2
c
2
+i
t
(11.4.8)
Will often
see written
1
H + i
=

0
d(it)
it
e
i(H+i)t
Not true strictly speaking

(11.4.2) is true
However,
a
s
(s) =

0
dt t
s1
e
at
(11.4.9)
Hence,
d
n
k
(2)
n
e
i(k
2
+i)t
. .. .
No angular comp.
=
i
(4it)
n/2
(11.4.10)
Thus, one loop
V (
c
) =
i
2
d
4
x (0 . . . 0)
n
n!
=
i
2

0
d(it)
it
i
(4it)
n/2
e
i
m
2
2
c
2
+i
t
Let it = u
=
1
2

0
du
u
1+n/2
1
(4)
n/2
e
m
2
+
2
c
2
u
=
1
2(4)
n/2
n
2
m
2
+

2
c
2
n/2
(11.4.11)
This diverges as n 4 (
n
2
(2) diverges at negative integers). So, let

= 2
n
2
.
n
2
= (2 + )
=
(1 + )
2 +
=
1
(2 +)
()
(1 + )
((x + 1) = x(x))
() =
1

E
+ . . . (11.4.12)
V (
c
) =
1
2
1
(4)
2
1
2
1
1

E
+. . .
m
2
+

2
c
2
2
=
1
2
1
(4)
2
1 + ln(4) +O(
2
)

1
2 3 + . . .

1

E
+ . . .
m
2
+

2
c
2
1 ln
m
2
+

2
c
2
+. . .
=
1
4
1
(4)
2
m
2
+

2
c
2
+
3
2

E
+ ln(4) ln
m
2
+

2
c
2
+O()
(11.4.13)
So,
V
TOT
=
m
2
2
c
2
. .. .
+

2
c
4
....
+
(1)
64
2
m
4
+ m
2
2
c
. .. .
+

2
4
c
4
. .. .
ln
m
2
+

2
c
2
+ ln(4)
E
+
3
2
+O()
(11.4.14)
So, to one loop order,
m
2
R
2
=
m
2
2
+
1
64
2
m
2
(11.4.15)
R
4!
=

4!
+

2
4
1
64
2
(11.4.16)
+ + + + + ldots
V
nite
Tot
=

2
4
c
256
2
ln
2
c
+k
(11.4.18)
(lim m
2
0). Thus, as m
2
0,
V
1 loop
(
c
) =

2
4
c
256
2
ln
2
c
+k
(11.4.19)
V
Tot
=

4
c
4!
+

2
4
c
256
2
ln
2
c
+k
(11.4.20)
(See Figure 11.4.1)
11.4. DIMENSIONAL REGULARIZATION 223
1
0
1
2
3
4
5
y
3 2 1 1 2 3
phi
Figure 11.4.1: V
c
(
c
) =

4
c
4!
dV
Tot
(
c
)
d
c
=

3
c
6
+

2
256
2
4
3
c
ln
2
c
+k
+
4
c
1
=
3
c
6
+

2
256
2
4 ln
2
c
+ k
+ 1
=
3
c
6
+

2
64
2
ln
2
c
t2
= 0
if
c
= 0
or
2
c
=
t2
e
(64
2
/6)
=
t2
e
(32
2
/3)
(11.4.21)
Changes in are compensated for by changing .
Mar. 13/2000
Recall:
V
tot
=

4
4!
+

2
256
2
f
4
ln
f
2
+k
=
f
4
4!
+

2
f
4
256
2
ln
f
2
t2
(f =
c
from last time)
V
tot
f
= 4f
3
4!
+

2
256
2
ln
f
2
+

2
f
tt
256
2
1
f
0 = f
3
6
+
4
2
256
2
ln
f
2
+

2
256
2

If f = 0, or
ln
f
2
6

2
256
2
256
2
4
2
=
42
2
3
2
4

1
4
=
21
1
3
2
2

1
4
=
64
2
6

1
4
=
32
2
3

1
4
ln
f
2
t2
=
32
2
3
f
2
min
=
t2
exp
32
2
3
(11.4.22)
At this value of f
min
, we have,
V
TREE
<V
1loop
(11.4.23)
Thus we are beyond the region where perturbation theory can be trusted.
Note: For scalar electrodynamics (See Coleman and Weinberg).
L =
1
4
F
+ (D
(D
) m
2
)
2
(11.4.24)
V
Classical
= m
2
+(
)
2
, V
Tot
= V
cl
+V
1loop
= V
tot
(m
2
, , e, f)
D
+ ieA
Here,
V
Tot
f
= 0 at f = f
min
(11.4.25)
This xes in terms of e, but introduces a new parameter f
min
into the resting eective
potential.
Dimensional Transmutation (Coleman and Weinberg).
V
tot
to higher loop order
+ +
+. . . +
+ . . .
+ . . .
+ . . . (11.4.26)
11.5. SPONTANEOUS SYMMETRY BREAKING IN GAUGE THEORIES 225
The loop expansion is an expansion in powers of .
L =
1
2
(
)
2
m
2
2

2

4
4!
(11.4.27)
Z =
De
(i/)
d
4
x /
(11.4.28)
Vertices
1
Propagator
2
+ m
2
1
=

k
2
m
2
. .. .
Mom. space
For a given diagram
()
IV
I # of internal lines
V # of vertices
(11.4.29)
But in a given diagram,
L = # of loops
= I (V 1)
L = I V + 1 (11.4.30)
(Where the V is present because the -function at each vertex imposes a restriction, but
the (1) in the second line is because one of the -functions is superuous due to overall
-function).
Thus, a given diagram
()
L1
(11.4.31)
So, for example, the expression (11.4.26) will have the following powers of :
+ . . . + () +
+ . . . (11.4.32)
11.5 Spontaneous Symmetry Breaking in Gauge The-
ories
Choice of gauge
L =
1
4
F
2
+ (D
(D
) m
2
)
2
(11.5.1)
(x) = V +
t
(x) (V = V

)
The kinetic energy for scalar:
K.E. = [
t
ieA
(V +
t
)] [
t
+ ieA
(V +
t
)]
=
t
t
+ e
2
V
2
A
+ ie(
t
)V A
. .. .
leads to mixed propagator
+O(
3
) (11.5.2)
L
gf
=
1
2
(A)
2
(gf = gauge xing) (11.5.3)
(Feynman gauge). We can eliminate the mixed propagator by working in the R
gauge.
L
gf
=
1
2
( A ieV (
t
t
))
2
(11.5.4)
If
t
=
1
+ i
2
L
gf
=
1
2
( A 2eV
2
)
2
(11.5.5)
(recall previously, =
V +
e
i/
2V
, where the imaginary (exponential) part is eliminated
by gauge transf.).
L
(2)
+L
gf
=
1
4
(F
)
2
+ (
1
)
2
+ (
2
)
2
+e
2
V
2
A
From (11.5.2)
+ 2eV A
2
)
1
2
( A)
2
+ 2m(
)
2
2(eV )
2
2
2
+ 2m
2
2
1
(11.5.6)
The A
2
propagator has disappeared.
1
Propagator
i
k
2
+ 2m
2
(11.5.7)
2
Propagator
i
k
2
m
2
M
2
=
(eV )
2
2
(11.5.8)
A
Propagator
i
(1)
k
2
M
2
k
2
M
2
(11.5.9)
Mar. 15/2000
Recall:
(x) = V +
1
+ i
2
(11.5.10)
11.5. SPONTANEOUS SYMMETRY BREAKING IN GAUGE THEORIES 227
i
1
=
i
k
2
2
2
(11.5.11)
i
2
=
i
k
2
M
2
(11.5.12)
i
=
i
(1)
k
2
M
2
k
2
M
2
(11.5.13)
R
gauge
L
gf
=
1
2
( A + M
2
)
2
(11.5.14)
In the limit 0
i
1
=
i
k
2
2
2
(Higgs Field) (11.5.15)
i
2
=
i
k
2
(Serves to act as the longitudinal mode of the vector)(11.5.16)
i
=
i
k
2
k
2
M
2
(Purely Transverse) (11.5.17)
This is the renormalized (R) gauge.
If ,
i
1
=
i
k
2
2
2
(11.5.18)
i
2
= 0 (11.5.19)
i
=
i
M
2
k
2
M
2
1
4
F
m
2
2
A
(11.5.20)
Unitary (U) gauge.
= 1 (Feynman-tHooft gauge).
i
1
=
i
k
2
2
2
(11.5.21)
i
2
=
i
k
2
M
2
(11.5.22)
i
=
ig
k
2
M
2
(11.5.23)
Easy to calculate with.
Note that Physical eects are independent of .
Chapter 12
Ward-Takhashi-Slavnov-Taylor
Identities
The WTST identities are relations between dierent Greens Functions that follow from
gauge invariance. (Well look at Slavnovs approach).
I(a, b) =
d
3
x e
ax
2
+bx
(12.0.1)
If
x x + x ( 0)
x
2
x
2
Let
x = x
t
+ x
t
(12.0.2a)
d
3
x = d
3
x
t
, x
t2
= x
2
(12.0.2b)
So,
I(a, b) =
d
3
x
t
e
ax
2
+b(x
+x
)
(Re: 0)
d
3
x
t
e
ax
2
+bx
(1 +b x
t
) (12.0.3)
(relabel x
t
x, and note that the exponent times the 1 in the last line = I(a, b).)
0 =
d
3
x e
ax
2
+bx
(b x)
0 = b
d
3
x xe
ax
2
+bx
0 =

b
a
ijk
b
j
d
3
x x
k
e
ax
2
+b
m
x
m
0 =
d
3
x(
iak
x
k
+
ijk
b
j
x
k
x
a
) e
ax
2
+bx
(12.0.4)
229
230 CHAPTER 12. WARD-TAKHASHI-SLAVNOV-TAYLOR IDENTITIES
Perform analogous operation in the functional integral.
L =
1
4
F
+

(i e A m) + J
+ +

1
2
( A)
2
(12.0.5)
Z(J
, , ) =
d d
dA
e
i
dx
(
/
cl
+JA+ +
1
2
(A)
2
)
(12.0.6)
Use transformation analogous to (12.0.2a,12.0.2b).
A
+
1
e
(12.0.7a)
e
i
(12.0.7b)

e
i
(12.0.7c)
L
cl
L
cl
(analogous to x
t2
x
2
) (12.0.7d)
dA
d d
dA
d d
(Note demonstrating this NOT trivial) (12.0.7e)

Thus,
Z(J, , ) =
dA
d d
exp
dx
L
cl
+ J (A + d) + e
i
+

e
i
1
2
(A +
2
)
2
(12.0.8)
For 0 (i.e. expanding exponent)
=
dA
d d
exp
dx
L
cl
+ J A + +

1
2
( A)
2
1 +
dx
(x)(x)(x) i (x)(x)(x)
( A(x))(
2
(x) + iJ
(x)
(x))
. .. .
=0
Let
(x) =
d
4
yD
F
(x y)(y) (12.0.9)
2
(x) = (x) (12.0.10)
231
Thus,
0 =
dA
d d
e
i
dx
[
/
cl
+
+ +JA
1
2
(A)
2
]
d
4
x
(x)(x)D
F
(x y)
i (x)(x)D
F
(x y) + iJ
(x)

x
D
F
(x y)
i
( A(x))
4
(x y)
0 =
d
4
x

i(x)
(x)D
F
(x y) i (x)

i (x)
D
F
(x y)
i

x

iJ
(x)
4
(x y) +J
(x)

x
D
F
(x y)
Z(J
, , )

i (x)
. Acto on this with

iJ
(x)
and then let J
= = = 0.
0 =
d
4
x
(i)

x

iJ
(x)
iJ
(x)
(x y)Z[J, 0, 0] +
i

x
D
F
(x y)Z[0, 0, 0]
(12.0.11)
Mar. 16/2000
From last time:
Z =
dA
d d
exp
dx
L
cl
+ J A +

+
1
2
( A)
2
(12.0.12)
A
+
1
e
=

(i e A m)
e
i

e
i
(12.0.13)
0 =
dA
dd
e
i
dx
(
/
cl
+JA+
+
1
2
(A)
2
)
1
e
J + i
( A)
1
e
(x) =
d
4
yD
F
(x y)(y)
0 =
dA
dd
dx
1
e
J
(x)

x
D
F
(x y) + i
(x)
(x)D
F
(x y)
i
(x)
(x)D
F
(x y)
1
e
(x)(x y)
e
i
dx (/
cl
+...)
0 =
1
e
J
(x)

x
D
F
(x y) + i

i
(x)
(x)D
F
(x y) i
(x)

i
(x)
D
F
(x y)
1
e
iJ
(x)
(x y)
Z
by

J
(z)
and let J = = = 0
0 =
dx
1
e
g
(x z)

x
D
F
(x y)
1
e
4
(x y)

iJ
(x)
(z)
Z[J, 0, 0]
J=0
But,
'0[TA
(x)A
(y)[0` =

iJ
(x)
iJ
(y)
Z[J, 0, 0]
J=0
0 =
1
d
4
x
i'0[TA
(x)A
(y)[0`
4
(x y)
+ (x z)

x
D
F
(x y)
0 =
i
'0[TA
(y)A
(z)[0` +

z
D
F
(z y)
0 =
i
'0[TA
(y)A
(z)[0` +

y
D
F
(y z) (12.0.15)
Thus if
'0[TA
(y)A
(z)[0` =
d
4
k
(2)
4
e
ik(yz)
i
(k) (12.0.16)
D
F
(y z) =
d
4
k
(2)
4
e
ik(yz)
k
2
+ i
(12.0.17)
Hence
0 =
k
(k) +
(ik
)
k
2
(12.0.18)
233
'0[TA
(y)A
(z)[0`
FT
= + +
+ + . . .
= i
(k) (12.0.19)
Here,
=
i
(1)
k
2
k
2
+ i
(12.0.20)
Hence,
0 =
k
(1)
k
2
k
2
+ i
+
ik
k
2
but k
k
2
= 0. Thus,
0 = k
+ + + . . .
And so,
+ +. . .
must be proportional to
g
k
2
(12.0.23)
i.e.
+ + . . . = (k
2
g
)(k
2
)
= k
2
g
1
(k
2
) k
2
(k
2
) +m
2
3
(k
2
)g
Must have
1
(k
2
) =
2
(k
2
)
3
(k
2
) = 0
Thus, all divergences are proportional to (g
k
2
k
). No divergences proportional to
g
alone.
If,
L
cl
=
1
4
F
2
=
1
4
(k
(k) k
(k))
2
=
1
2
A
(k
2
g
)A
divergences and L
cl
k
2
g
.
Thus all divergences in the vacuum polarization can be removed by a wave function
renormalization.
A
bare
= Z
1/2
3
A
Renormalized
(12.0.24)
i.e. if
1
=
2
,
3
= 0,
=
k
2
g
log div.
....
1
. .. .
divergence can be absorbed into
A
(g
k
2
k
)A
+(
2
+
1
)
log div.
....
k
. .. .
need A
+
quadratic
div.
. .. .
m
2
3
g
. .. .
m
2
R
A
(12.0.25)

1
,
2
Log. div.

3
quadratic
Recall:
= (1)(ie)
2
d
4
(2)
4
Tr
i( + m)
2
m
2
+ i
i(+ k + m)
( + k)
2
m
2
+ i
(12.0.26)
= appears to be quadratic divergent (12.0.27)
Gauge invariance overrides simple power counting arguments.
Quadratic divergence disappears
Automatically satised in dimensional Regularization.
Mar. 20/2000
12.1 Dimensional Regularization with Spinors
d
n
k
(2)
n
(k
2
)
a
(k
2
m
2
)
b
=
i
(4)
n/2
(1)
ab
(m
2
)
(n/2)+ab
n
2
+a
b a
n
2
n
2
(b)
(12.1.1)
12.1. DIMENSIONAL REGULARIZATION WITH SPINORS 235
with spinors,
= 2g
; g
= n (12.1.2)
2g

= 2
= 2
1
2
(
= 2
1
2
(2g
)
= (2 n)
(12.1.3a)
n = 4
= 2
(12.1.3b)
12.1.1 Spinor Self-Energy
0 =

iJ
(y)
+ ie
dx

(x)

i
(x)
D
F
(x y) + ie
dx (x)

i(x)
D
F
(x y)
+
dx J
(x)

x
D
F
(x y)
Z[J
, (x),

(x)] (12.1.4)
Take

(w)
(z)
of (12.1.4) =

= J
= 0. We get,
'0[TA
(x)
(y)(z)[0` = e
D
(0)
F
(x y) D
(0)
F
(x z)
'0[T

(y)(z)[0`
. .. .
S
F
(yz)
(12.1.5)
(Note that D
(0)
F
(x y) = (x y)). In F.T. space, if
'0[TA
(x)
(y)(z)[0` = e
dx
t
dy
t
dz
t
D
F
(x x
t
)S
F
(y y
t
)
(x
t
, y
t
, z
t
)S
F
(z
t
z) (12.1.6)
D
F
(x x
t
) =

x
D
(0)
F
(x x
t
)
amputating external legs:
=
(p p
t
)
(p, p
t
) =
S
1
F
(p
t
)
S
1
F
(p)
In the limit p p
t
,
(p, p) =

p
S
1
F
(p) Ward Identity (12.1.7)
At tree level, let
S
1
F
= p m
S
1
F
=
at tree level
(trivial at tree level to get Ward identity). One loop:
d
4
k
(2)
4
(ie
i
p+ k m
(ie
i
p
t
k m
(ie
ig
k
2
i
(12.1.8)
S
F
=
=
d
4
k
(2)
4
(ie
i
p+ k m
(ie
ig
k
2
i
(12.1.9)
Bj. and Drell
12.1. DIMENSIONAL REGULARIZATION WITH SPINORS 237
(One loop order, evident (12.1.7) holds also).
In general, as p m,
S
F
(p)
Z
2
p m
c
(p)
(12.1.10)
c
(p) 0 as p
2
m
2
And vertex function,

u(p)
(p, p)u(p) = Z
1
1
u(p)
u(p) (12.1.11)
Thus our Ward identity gives Z
1
= Z
2
.
Tells us, if we compute a self-energy diagram,
+ +
Bare
= Z
1/2
2

Renormalized
, A
Bare
= Z
1/2
3
A
Ren.
(12.1.13)
And look at all vertex functions,
Z
1
: + + (12.1.14)
(Z
1
parameterizes s in diagrams).
e
Bare
=
Z
1
Z
1/2
3
Z
2
e
Renormalized
(12.1.15)
L =
1
4
A
B
A
B
2
+

p m
B
e
B
A
B
B
=
1
4
Z
3
A
R
A
R
2
+Z
2

p m
R
+ m
Z
1
Z
1/2
3
Z
2
e
R
(Z
1/2
3
A
R
)Z
2

R
(12.1.16)
(What we would expect from renormalization)
We also have
Z
1
= Z
2
(12.1.17)
e
B
=
Z
1
Z
1/2
3
Z
2
e
R
= Z
1/2
3
e
R
(12.1.18)
Suppose we have 2 fermions (e
, proton):
L =
1
4
A
B
A
B
2
+

p m
B
e
B
A
B
B
+

p M
B
B
e
B
A
B
B
. .. .
e
B
for proton
(12.1.19)
Renormalization of proton will be dierent. e
B
for proton Ward identity insures Z
1
= Z
2
.
(Lurie).
i.e. e
B
same for e
, proton, Ward identity ensures e

R
will be same Z
1
for e
= Z
1
for
p, but Z
2
for e
= Z
2
for p either.
This has the eect that if 2 dierent charged elds have the same bare charge, then they
will have same renormalized charge.
12.2 Yang-Mills Theory
Mar. 22/2000
Z(J
) =
TA
F
(A
) exp
d
4
x
L
Y M

1
2
( A)
2
+ J A
(12.2.1)
F
(A) = det( D(A)) , D
ab
ab
+ gf
apb
A
p
(12.2.2)
1
F
(A) =
d (A
) , A
a
= A
a
+ D
ab
(A)
b
(12.2.3)
1
F
(A
) =
d ( A
) (12.2.4)
But, A
= A
and
d integrate over all gauge transformations.
1
F
(A
) =
d
tt
( A
)
=
1
F
(A
) (12.2.5)
Thus,
F
(A
) =
f
(A
) (12.2.6)
In the functional integral for Z, let,
A
+ D
(12.2.7)
DA
DA
= DA
(12.2.8)
Thus,
Z[J] =
TA
F
(A
) exp
dx
L
Y M

1
2
( A
)
2
+ J A
(12.2.9)
12.2. YANG-MILLS THEORY 239
For small ,
Z[J] =
TA
F
(A) exp
dx
L
Y M

1
2
( A
)
2
+ J A
1 + 2
1
2
( A)( D) + J (D)
0 =
TA
F
(A)e
i
dx
[
/
Y M
1
2
(A)
2
+JA
]
d
4
x
A
a
(x))
b
. .. .
(
D
ab
(A)
b
(x))
+ J
a
(x)D
ab
(A)
b
(x)
. .. .
M
bc
(12.2.10)
If
D
ab
(A)M
bc
(x, y) =
(x y)
ac
(12.2.11)
and
a
(x) =
d
4
y M
ab
(x, y)
b
(y) (12.2.12)

D
ab
b
(x) =
a
(x)
Then, as A
a
(x)
1
i
J
a
(x)
, we have,
0 =
d
4
x
iJ
a
(x)
a
(x)
d
4
x J
a
(x)D
ab

iJ
(M
bc
)
1
(x, y)
. .. .
Messy
b
(y)
Z
(12.2.13)
From the resulting Ward identities, (WTST identities), it was proven that the coupling
constant g could be renormalized by considering any vertex into which it enters, and the
same result would arise.
i.e.
+ + + . . . (A A A vertex)
+ + . . . (A A A A vertex)
+ + . . . (c c A vertex)
+ + . . . (

A vertex)
If g
B
= Z
g
g
R
, Z
g
is the same for all 4 vertices Gauge invariance of Lagrangian has not
been altered. Renormalization doesnt break gauge invariance.
12.3 BRST Identities
WSTS Were derived by using a local, linear transformation on A
a
that left L
Y M
and
F
(A) invariant.
BRST identities involve global non-linear transformations which leave L
Y M
+L
ghost
+L
gf
invariant (gf = gauge xing).
Z[J
a
,
a
,
a
] =
TA
a
Tc
a
T c
a
exp
d
4
x
1
4
(F
a
(A))
2
c
a
D
ab
(A)
c
b
1
2
( A
a
)
2
+ J
a
A
a
+ c
a
a
+
a
c
a
(12.3.1)
Remarkably,
L
eff
= L
Y M
+L
gh
+L
gf
(12.3.2)
12.3. BRST IDENTITIES 241
is invariant under
A
a
A
a
+ D
ab
(A)c
b
(, c Grassmann) (12.3.3)
c
a
c
a
1
2
f
abc
c
b
c
c
(12.3.4)
c
a
c
a
+
1
( A
a
) (12.3.5)
(The Proof involves the Jacobi identity).
To derive the BRST identities, let A A+(Dc),etc., in the functional integral, and to
leading order in ,
Z[J] =
dA dc d c exp
dx (L
eff
+ J A + c + c)
1 +
d
4
x
J
a
D
ab
(A)c
b
+
1
( A
a
)c
a
+
1
2

a
f
abc
c
b
c
c
0 =
d
4
x
J
a
(x)D
ab

iJ

i
b
(x)
+
1
iJ
a

i
a
(x)
1
2
f
abc

a
(x)

i
b
(x)

i
c
(x)
Z[j, , ] (12.3.6)
This can be streamlined:
BRST
Any Greens Fn.
. .. .
'0[TO
1
. . . O
N
[0` = 0 (12.3.7)
Consider
0 =
BRST
'0[TA
a
(x) c
b
(y)[0`
0 = '0[T
D
ac
(A)c
c
(x) c
b
(y)
[0`
+'0[
TA
a
(x)
1
A
b
(y)
[0`
1
'0[TA
a
(x)A
b
(y)[0` = '0[TD
ab
(A)c
c
(x) c
b
(y)[0` (12.3.8)
Typical of BRST identity Relates two Greens Functions.
Note: '0[TA
a
(x)A
b
(y)[0` is transverse only if = 0 (Landau gauge).

Mar. 23/2000
12.4 Background Field Quantization
L =
1
2
(
)
2
m
2
2

2

4!
4
(12.4.1)
Z[j] =
d exp
d
4
x (L() + J)
(12.4.2)
write (x) = f(x) +h(x), (f(x) = Classical Background, and h(x) = Quantum Correction).
Z(f, J) =
Th exp
d
4
x (L(f + h) +Jh)
(12.4.3)
L =
1
2
(
(f + h))
2
m
2
2
(f +h)
2

4!
(f + h)
4
=
1
2
(
f)
2
m
2
2
f
2

4!
f
4
. .. .
Classical Lagrangian
+ f(. . .)
. .. .
Can be
neglected
+
1
2
(
h)
2
m
2
h
2
2
. .. .
A

4!
(6f
2
h
2
)
. .. .
B
+
4!
(4fh
3
....
C
+ h
4
....
D
) (12.4.4)
With
A = =
i
p
2
m
2
B = 4 - pt.
C =
D =
(12.4.5)
12.4. BACKGROUND FIELD QUANTIZATION 243
2
1
1
2
3
4
y
4 2 2 4
phi
Figure 12.4.1: = V +
t
V background - (Just
some const. previously). Where V () =
m
2
2
2
+

4
4!
Aside: Recall the Figure 12.4.1:
If we consider only diagrams with external elds, we will reproduce Z(J). Let h hf
in

Z.
Z(f, J) =
Th exp
dx (L(h) + J(h f))
= e
i
dx Jf
Z(J) (12.4.6)
W(J) = i ln(Z(J))
W(f, J) = i ln(

Z(f, J))
Generating functional for connected diagrams (12.4.7)

Thus,
W(f, J) = i ln
e
i
d
4
x Jf
Z(J)
= i ln (Z(J)) i ln
e
i
d
4
x Jf
= W(J)
d
4
x Jf (12.4.8)

W(J)
J
(12.4.9)
() = W(J)
d
4
x J (12.4.10)
= Generating functional for 1PI diagrams
So also,

W(J)
J
(12.4.11)
() =

W(J)
d
4
x J
(12.4.12)
=

J
dx Jf
= f
=

+ f (12.4.13)
So also
(f,

) =

W(f, J)
dx J
= W(J)
d
4
x Jf
d
4
x J
= W(J)
d
4
x
= () (12.4.14)
Hence,
() =

(

f,

)
or
(f +

) =

(f,

)
if

= 0
(f) =

(f, 0) (12.4.15)
()

(
)
L = L()
If we make the identication f, we can ignore the diagrams on the left in

column,
and get the diagram on the right.
Applying this to YM theory,
L =
1
4
(F
(V ))
2
(12.4.16)
Z[J] =
dV
F
exp
dx (L +L
gf
+ JA)
(12.4.17)
V
a
(x) = A
a
(x) + Q
a
(x)
Z[A
a
, J
a
] =
TQ exp
dx (L(A + Q) +L
gf
(A + Q) + J Q)
(12.4.18)
V
a
= D
ab
(V )
b
=
a
+ gf
abc
V
b
c
This can become:
(i)
A
a
= 0
Q
a
a
+ gf
abc
(A
b
+ Q
b
)
c
Choose L
gf
so that this gauge invariance is bro-
ken.
(ii)
A
a
a
+ gf
abc
A
a
c
Q
a
= gf
abc
Q
b
Leave this intact.

Homerkemp gauge,
L
gf
=
1
2
(D
ab
(A)Q
b
)
2
(12.4.19)
Leaves type (ii) gauge invariance unbroken.
(A
a
,

Q
a
) = (A
a
+

A
a
) (12.4.20)
in same way as in scalar case

(f,

) = (f + ).
Let

Q = 0. Thus,
(A
a
) =

(A
a
, 0)
. .. .
(12.4.21)
* - Invariant under A
a
A
a
+ D
ab
(A)
b
. Thus,
(A
a
) =

(Tr
(F
a
(A))
2
+ Tr
(F
a
(A))
3
) (12.4.22)
Mar. 27/2000
Recall: if we have
L
gf
=
1
2
(D
ab
(A)Q
b
)
2
(12.4.23)
then Z[A
a
] invariant under A
a
= D
ab
(A)
b
. This gauge xing (L
gf
) breaks the invari-
ance.
A
a
= 0 (12.4.24)
Q
a
= D
ab
(A + Q)
b
(12.4.25)
If we follow through with the standard way of deriving the ghost Lagrangian, we nd
that,
L
ghost
= c
a
D
ab
(A)D
bc
(A + Q)c
c
(12.4.26)
(where c, c are Grassmann ghost elds). Z depends only on F
a
(A)F
a
(A), etc. (i.e. another

would be
abc
F
a
F
b
F
c
must construct Z out of gauge invariant quantities).

But divergences, by simple power counting, can arise only in 2, 3 and 4 point functions.
quadratic divergent
d
4
k
k
k
2
k
2
(12.4.27)
(k
, k
at each vertex)
Linear
d
4
k
k
k
2
k
2
k
2
(12.4.28)
log (12.4.29)
+ log (12.4.30)
The divergences in all three diagrams can only serve to renormalize
F
a
F
a
F
a
A
a
A
a
+g
abc
A
b
A
c
(12.4.31)
We can get divergences in (12.4.28,12.4.29) diagrams automatically if we know the diver-
gences in (12.4.27). Hence, the divergences in Z(A) are of the form,
j=1
a
j
j
F
a
(A)F
a
(A) (12.4.32)
ex:
abc
F
a
F
b
F
c

abc
A
a
A
a
A
a
A
a
A
a
A
a
ppp
d
4
k
1
k
2
k
2
k
2
Not divergent -
no renormalization req.
(12.4.33)
F
a
F
a
one power of external momentum associated with it.
d
4
k
k
k
2
k
2
k
2
p
. .. .
log div.
Could have
d
4
k
k
k
2
k
2
k
2
. .. .
linearly div.
(12.4.34)
From the two point function,
A
a Bare
= Z
1/2
A
A
a Ren.
(12.4.35)
From the three point function,
g
Bare
= Z
g
g
Ren.
(12.4.36)
Thus Z[A] has a contribution
A
a Bare
A
a Bare
+ g
Bare
abc
A
b Bare
A
c Bare
2
=
Z
1/2
A
A
a Ren.
A
a Ren.
+Z
1/2
A
Z
g
g
Ren.
abc
A
b Ren.
A
c Ren.
2
(12.4.37)
In order for this to remain gauge invariant,
Z
1/2
A
Z
g
= 1 (12.4.38)
So, Z
1/2
A
and Z
g
are not independent in background eld quantization!
But Z
A
can be determined from the 2 point function alone.
Note also - the relation between 2, 3 and 4 point functions implied by the gauge invariance
of Z(A) is the Ward identities in the context of background eld quantization.
The Feynman rules are:
For Z
A
to one loop order
Have the divergent parts

ig
2
ab
c
2
(4)
2
10
3
(p
2
g
) +
ig
2
(4)
2
ab
c
2
1
3
(p
2
g
) (12.4.39)
(g) =
11
3
c
2
(4)
2
g
3
(12.4.40)
(Details: See Abbott - Nuclear Physics 1981).
Chapter 13
Anomalies
A theory has an anomaly if a symmetry of the classical Lagrangian is broken by quantum
eects.
Most important one: Chiral anomaly for massless fermions.
L =
1
4
F
(A)F
(A)
1
4
F
(B)F
(B) +

(i A B
5
) (13.0.1)
There are two invariances in this Lagrangian,
1. (x) e
i(x)
(x) provided A
(x)
2. (x) e
i
5
(x)
5
(x) provided B
5
(x)
i.e.
F
(B) =
(B
5
....
(B
5
. .. .
) (13.0.2)
* - cancel, provided derivatives commute with each other.
(x) e
i
5
(x)
5
(x) (13.0.3)
(x)
e
i
5
(x)
5
=
e
i
5
(x)
5
(13.0.4)
e
i
5
(x)
5
0
=

e
i
5
(x)
5
(13.0.5)
Thus,
(i B
5
)

e
i
5
5
[i ( B
5
)
5
] e
i
5
e
i
5
e
i
5
5
[i ( + i
5
5
) ( B
5
)
5
]
=

(i B
5
) (13.0.6)
Note:
e
i
5
5

e
i
5
=

(13.0.7)
Chiral gauge invariance is not respected by the mass term in a Lagrangian.
249
250 CHAPTER 13. ANOMALIES
In Euclidean space,
= 2
(13.0.8)
So

=
(not

=
0
as in Minkowski), and so

(i B
5
) is not chirally gauge
invariant, but

is.
Mar. 29/2000
L =
1
4
F
2
(A)
1
4
F
2
(B) +

(i e A g B) (13.0.9)
e
ie(x)
A
(13.0.10)
e
ig
5
(x)
B
5
(13.0.11)
(x) = 0 j
(13.0.12)
5
(x) = 0 j
5
=

5
(13.0.13)
The Greens Functions are external in the elds A
, B
. i.e.
G
1
,...
n
,
1
,...,
m
(x
1
. . . x
n
, y
1
. . . y
m
) = '0[TA
1
(x
1
) . . . A
n
(x
n
)B
1
(y
1
) . . . B
m
(y
m
)[0` (13.0.14)
G
x
i
= 0 =
G
y
j
(13.0.15)
Remembera CA
C
1
= A
n is even , CB
C
1
= +B
m is even or odd.
Examine
'0[TA
(x
1
)A
(x
2
)B
(y)[0` (13.0.16)
In momentum space at one-loop order,
+
or can draw
(13.0.17)
R
(p, q) = (1)(ie)
2
(ig)
d
4
k
(2)
4
Tr
(i
5
)
i(k+ p)
(k + q)
2
(i
)
i k
k
2
(i
)
i(k p)
(k p)
2
+(i
5
)
i(k+ p)
(k + p)
2
(i
i k
k
2
(i
)
i(k q)
(k q)
2
(13.0.18)
251
(the rst term in the trace refers to the second diagram, and the second term to the rst
diagram). By using
a b c = a b c a c b + b c a + i
0123
= +1 ,
5
= i
0
can show that the two traces are identical. Examine: q
(p, q). Note:

q
(k+ q)
(k + q)
2
k
k
2
= q
1
(k+ q)
1
k
=
1
(k+ q)
((k+ q) k)
1
k
=
1
k
(k+ q)
1
k
=
1
k

1
(k+ q)
(13.0.19)
We nally obtain
q
(p, q) = e
2
g
d
4
k
(2)
4
Tr
5
1
k
1
(k p)

5
1
(k+ q)
1
(k p)
5
1
(k+ p)
1
k
+
5
1
(k+ p)
1
(k q)
(13.0.20)
Now, k k + p in the rst term in the trace, and k k + p q in the second term -
q
(p, q)
?
= 0 upon this shifting of variables.
But, it is quadratically divergent (actually just linearly divergent because of
ap-
pearing upon taking the trace), so care must be taken when shifting variables. ex.
Figure 13.0.1: Plot of tan
1
(x + a)
and tan
1
(x)
I =
dx
tan
1
(x + a) tan
1
(x)
?
= 0(shifting x x a)
but (see gure 13.0.1)
= lim
dx
tan
1
(x + a) tan
1
(x)

Or
tan
1
(x + a) = tan
1
(x) +a
d
dx
tan
1
(x) +
a
2
2!
d
2
dx
2
tan
1
(x) + . . . (13.0.21)
Substitute this back into I
I =
dx
tan
1
(x) + a
d
dx
tan
1
(x) +
a
2
2!
d
2
dx
2
tan
1
(x) + . . . tan
1
(x)
a tan
1
(x) +
a
2
2!
d
dx
tan
1
(x) +
a
3
3!
d
2
dx
2
tan
1
(x) + . . .
= a
+ 0 + . . .
= a (13.0.22)
In general, if
d
4
k f(k
2
) is linearly divergent, then,
d
4
k
f((k + a)
2
) f(k
2
)
= a
d
4
k

k
f(k
2
)
= i
d
k
k
3
a k
n
f(k
2
)
k
2
by Gauss
= 2
2
ia
kf(k
2
)
k
2
(13.0.23)
(
d
k
= 2
2
in 4-d, recall in 3-d,
d
k
= 4). Go back to
R
(a) = e
2
g
d
4
k
(2)
4
Tr
5
1
(k+ a
1
+ q)
1
(k+ a
1
)
1
(k+ a
1
p)
5
1
(k+ a
2
+ p)
1
(k+ a
2
)
1
(k+ a
2
q)
(a) R
(0) = e
2
g
(2
2
i)
(2)
4
lim
k
2
Tr
1
k
2
k
5
1
(k+ q)
1
k
1
(k p)
+ a
2
k
2
k
5
1
(k+ p)
1
k
1
(k q)
.
.
.
=
e
2
g
(a
1
a
2
)
(13.0.24)
a
1
a
2
= Ap
+ Bq
(13.0.25)
R
(a) = R
(0) +
e
2
g
(Ap
+ Bq
) (13.0.26)
So, (by same analogy), by surface terms,
q
(p, q) =
e
2
g
(13.0.27)
253
(does not appear to be = 0).
Mar. 30/2000
Recall:
R
(a) R
(0) =
e
2
g
(Ap
+ Bq
)
q
(0) = e
2
g
(2
2
i)
(2)
4
lim
k
2
(p
)k
2
k
Tr
k
4
p
(p q)
k
2
k
Tr
k
4
(q
=
e
2
g
(unsure of sign - see (13.0.34)) (13.0.28)

In the same way,
p
(0) =
e
2
g
(13.0.29)
And lastly the axial vertex,
(p + q)
(0) = 0 (13.0.30)
Maybe can choose A, B so we get conservation of momentum at all 3 vertices.
q
(a) = q
(0) +
e
2
q
(Ap
+ Bq
=
e
2
g
Ap
=
e
2
g
(q
)(1 + A)
(13.0.31)
So also,
p
(a) = p
(0) +
e
2
g
(Ap
+Bq
=
e
2
g
+
e
2
g
B
=
e
2
g
(1 B) (unsure of signs - see (13.0.34)) (13.0.32)

And,
(p + q)
(a) = (p + q)
(a) +
e
2
g
(Ap
+ Bq
= 0 +
e
2
g
(Ap
+ Bq
)
=
e
2
g
(A B)p
(13.0.33)
It should be: If
q
(a) = 0
p
(a) = 0
(13.0.34)
then
(p + q)
(a) =
e
2
g
2
2
(13.0.35)
So, if we do have conservation of vector current, then axial current not conserved. i.e. If,
= 0 (13.0.36)
then
s
= e
2
(A)F
(A)
(4)
2
(13.0.37)
(Note: anomaly independent of mass). This is the modication of j
5
= 0 that is required
to account for (p + q)
= 0.
5
= 0 (13.0.38)
Note: The presence of the anomaly accounts for the decay of the
0
(
0
). i.e. If
were massive,

5
= 2m
=

[
5
] +
But,
(i A B
5
) m = 0
=
1
i
[( A+ B
5
) + m] (13.0.39)
Similarly,
=
1
i
( A
+ B
5
) + m
=
1
i
( A B
5
) + m
(13.0.40)
13.1. ANOMALIES 255
5
(pseudoscalar)
=
1
2m
=
1
2m
j
5
(13.0.41)
0 if j
5
= 0 (13.0.42)
(i.e. says
0
cannot go to
0
0
).
But the anomaly says this 0 if
j
5
=
e
2
(4)
2
(13.0.43)
If j
5
couples to a vector B
, and j
5
is not conserved because of the anomaly, then renormal-
izability is lost.
1
(i )
1
+

2
(i )
2
g

1
B
+ g

2
B
; B
5
(13.0.44)
Anomalous contributions to
5
cancel in this case, and renormalizability is no longer a
problem.
In the standard model, having equal numbers of quarks and leptons ensures the cancel-
lation of anomalies.
Apr. 3.2000
13.1 Anomalies
1. Adler Bardeen theorem Anomalies are exclusively one-loop eects. i.e.
'0[T V
(x)V
(y)
. .. .
Transverse
A
(z)
. .. .
Anomaly
[0` (13.1.1)
. .. .
A
+ +
. .. .
B
(13.1.2)
5
=
e
2
16
2
(13.1.3)
All of the anomaly can be shown to be from the A graph only. No anomalous con-
tribution comes from B, though they are needed in order to ensure conservation of
momentum at all three vertices. So, no
e
4
(4)
4
correction for B.
2.
Z(J, K) =
TA
TB
TT
exp
d
4
x [L
cl
+L
gf
+ A J + B J
5
]
(13.1.4)
L
cl
=

(p me A g B
5
)
1
4
F
2
(A)
1
4
F
2
(B)
A
, e
ie(x)
(13.1.5)
B
5
, e
ig
5
. .. .
(13.1.6)
* - this gauge transformation does aect measure there is a non-trivial Jacobian
associated with this gauge transformation.
(Fujikawa showed that this Jacobian gives the anomalous term in the WTST identity
for the axial current.)
anomaly arises in trying to make this gauge transformation in this innite measure
(TT
) integrating over all possibilities.

Chapter 14
Instantons
Instantons are generated in a 2 stage process.
1. Make Wick rotation to Euclidean space
2. Solve resulting equations of motion (Instanton)
14.1 Quantum Mechanical Example
ex: (see gure 14.1.1)
Figure 14.1.1: Classically, if in the
lowest energy state at the bottom
(one of the V (q) = 0 q
0
posi-
tions), particle would just sit there.
V (q) = (q
2
q
2
0
)
2
(14.1.1)
Vacuum tunnelling
[q(t )` = [q
0
` (14.1.2)
[q(t )` = [+q
0
` (14.1.3)
Compute transition (Q.M.) Path integral.
'q
f
[e
iHt
[q
i
` (14.1.4)
257
258 CHAPTER 14. INSTANTONS
is the amplitude for being in state [q
i
` at t = 0 and being in state [q
f
` at t = t.
Path integral:
'q
f
[e
iHt
[q
i
` =
q(t)=q
f
q(0)=q
i
Tq(t) exp
t
0
d L(q())
(14.1.5)
L(q()) =
1
2
q
2
() V (q())
=
1
2
q
2
() (q
2
q
2
0
)
2
Let i.
'q
f
[e
iHt
[q
i
` =
q(t)=q
f
q(0)=q
i
Tq(t) exp
t
0
d
q
2
2
(q
2
q
2
0
)
2
q(t)=q
f
q(0)=q
i
Tq(t) exp
S
E
(14.1.6)
S
E
=
t
0
d
q
2
2
+ (q
2
q
2
0
)
2
= S
Euclidean
Dominant path: path that satises
S
E
q()
= 0 with the appropriate boundary conditions.
Now then, if the
t
i
q(t) q
0
t
f
+ q(t) +q
0
The solution to the equations of motion that satises the above is
q
cl
(t) = q
0
tanh(
2q
0
t) (14.1.7)
The more general solution is
q
cl
(t) = q
o
tanh(
2q
0
(t t
0
....
)) (14.1.8)
* - localized at an instant t
0
instanton.
We will now consider
'q(+) = +q
0
[q() = q
0
` =
T(q) e
[S
E
(q
cl
+q)]
(14.1.9)
q = 0 at t expand about q
cl
.
'q(+) +q
0
[q() q
0
` =
T(q) exp
S
E
(q
cl
)
(q())
2
D
2
S
E
(Dq())
2
+ . . .
= e
S
E
(q
cl
)
[1 +O()] (14.1.10)
here S
E
(q
cl
) = 4
2
q
3
0
3
= e
S
E
(q
cl
)
1 + det
D
2
S(q
cl
)
(Dq())
2
+. . .
(14.1.11)
14.1. QUANTUM MECHANICAL EXAMPLE 259
Suppose we have 2 spinors X, Y . Consider two functions f
0
and f
1
.
f
i
: X Y (i = 0, 1) (14.1.12)
f
0
(x), f
1
(x) are homotopic if there exists a function F(x, t) such that F(x, 0) = f
0
(x) and
F(x, 1) = f
1
(x) and F(x, t) is continuous in t in (0 t 1).
Suppose we took
X : 0 < 2 (14.1.13)
Y : Z; Z ( , [Z[ = 1 (14.1.14)
Consider
f
() = e
i(+)
(14.1.15)
f
() = e
i(+)
(14.1.16)
Homotopic F(x, t) = e
i(+t+(1t))
Homotopy connecting f
, f
. But can also write

two functions not homotopic.
f
() = e
i(+)
(14.1.17)
f
() = e
i(2+)
(14.1.18)
Not homotopic (still functions that map X Y , but cant continuously deform one into the
other).
We have in general dierent classes of functions characterized by integers n. These are
of the form,
f
n
() = e
i(n+)
; n = 1, 2, . . . (sign indicates direction of winding.) (14.1.19)
In general, if two functions have dierent ns, Not homotopic.
n Pontryagin index (Winding number)
n = i
2
0
d
f
t
()
f()
(14.1.20)
Apr. 5/2000
Recall:
X : [0, 2)
Y : Z, [Z[ = 1
F
(n)
() = e
in
n = 1, 2, . . .
n = i
2
0
d
f
t
()
f()
(counts # of times each pt. in X mapped onto Y)
Figure 14.1.2: X :(, )
We can let X be (, ) provided + and are mapped on to the same point. (See
gure (14.1.2)). Consider,
f(x) = exp
ix
x
2
+
2
(14.1.21)
This maps < x < onto Z where [Z[ = 1.
n =
1
2
dx
f
t
(x)
f(x)
(Winding number for this mapping) (14.1.22)
Two functions with dierent winding numbers have dierent homotopies.
Consider the mapping from X = S
3
to Y = SU(2). i.e. S
3
Set of all points with
x
2
1
+ x
2
2
+ x
2
3
+ x
2
0
= 1 = x
2
0
+ x
2
(4-d sphere).
SU(2) set of 2 2 matrices U with UU
= 1.
f(x
0
, x) = x
0
+ ix (14.1.23)
(x
0
+ ix )
(x
0
+ ix ) = (x
0
ix )(x
0
+ ix )
= 1 (if x
2
0
+x
2
= 1)
Winding number for this mapping is,
n =
1
24
2
d
1
d
2
d
3
Tr
ijk
A
i
A
j
A
k
(14.1.24)
Where
i
are the polar angles characterizing S
3
and A
i
= f
1
()

i
f()
(Colemans lectures on Instantons).
We now extend X to be the entire three dimensional Euclidean space with all points at
identied with a single point.
Here,
f
1
(x) = exp
ix
x
2
0
+ x
2
+
2
(14.1.25)
14.2. CLASSICAL SOLUTIONS TOSU(2) YMFIELDEQUATIONS INEUCLIDEANSPACE261
is such a mapping. The winding number is
n =
1
24
2
d
3
x
ijk
Tr A
i
A
j
A
k
(14.1.26)
where
A
i
(x) = f
1

x
i
f (14.1.27)
Recall: Gauge theories,
A
(x) U
1
(A
(x) +
)U (14.1.28)
(if A
originally 0 would get pure gauge function very similar to above).

14.2 Classical Solutions to SU(2) YM eld equations in
Euclidean Space
S
Y M
=
d
4
x
1
4
F
a
(A)F
a
(A)
(14.2.1)
F
i
A
i
A
i
+
ijk
A
j
A
k
Equations of motion:
D
ij
(A)F
j
(A) = 0 D
ij
ij
+
imj
A
m
(14.2.2)
Bit of a short-cut to solve this (not in every case). Bianchi Identity.
F
i
=
1
2
F
i
(Dual) (14.2.3)
D
ij
F
i
= 0
Thus, if
F
i
= F
i
(14.2.4)
(ex: in E.M. would be analogous to requiring E = B) then,
D
ij
F
j
= D
ij
F
j
= 0 (14.2.5)
Let
A
i
m
= (
imk
mi
0
) ln[f(x
0
, x
m
)] (14.2.6)
A
i
0
=
i
ln[f(x
0
, x
m
)] (14.2.7)
(m spacial index, i group index). If F = F, then
f
1
2
f = 0 (14.2.8)
f =
n
i=1
2
i
(x x
i
)
2
(n Instanton solution)
Size
i
position x
i
Perform a gauge transformation on the one instanton solution.
A
t
= U
1
(A
)U (14.2.9)
with
U = exp
ix
x
2
+
2
(14.2.10)
tan() =
x
0
x
2
+
2
n
2
(14.2.11)
(n is integer, not the n from above). Note, as
[x[ , A
ti
0
0 (not easy to show)
x
0
+ , A
ti
m

1
n
x
m
n
(14.2.12)
x
0
, A
ti
m

1
n1
x
m
n1
where
n
=
exp
ix
x
2
+
2
n
.
Figure 14.2.1: I, II, III are all 3-
d spaces. In all 3 spaces, A
i
m

U
1
m
U (i.e. A
i
m
is pure gauge func-
tion).
To show this, consider F
(A) ; F
0 as [x
[ . Consider U as a mapping from

3-d Euclidean space onto SU(2).
n =
1
24
2
d
3
x
ijk
Tr A
i
A
j
A
k
(14.2.13)
A
i
= U
1

x
i
U
In region III (gure 14.2.1);
n
III
=
1
24
2
III
d
3
x
ijk
Tr A
i
A
j
A
k
where i, j, k has one index 0 because were on one side of sphere

= 0 (on III) (14.2.14)
But, from (14.2.12),
n
I
=
1
24
2
I
d
3
x
ijk
Tr A
i
A
j
A
k
= n ((i, j, k) are a mixture of 1,2,3 (no 0) in integral) (14.2.15)

and
n
II
=
1
24
2
II
d
3
x
ijk
Tr A
i
A
j
A
k
= n 1 (14.2.16)
See gure 14.2.2.
Figure 14.2.2:
Figure 14.2.3:
Figure 14.2.4: q(t) =
q
0
tan(
2q
0
t) instanton
interpolated from +.
Apr. 6/2000
For completeness, One-instanton:
A
(x) =
x
2
x
2
+
2
; =
x
4
ix
x
2
(14.2.17)

1
(pure gauge) as [x[

m =
1
24
2
d
3
x
ijk
Tr(A
i
A
j
A
k
) ; A
i
= f
1
i
f (14.2.18)
Note:
K
= 4
Tr
+
2
3
A
(not gauge invariant) (14.2.19)

. . . satises
= 2Tr F
; F
=
1
2
(14.2.20)
Thus,
Tr
d
4
x F
d
4
1
2
=
1
2
d
3
(By Gauss Theorem) (14.2.21)

But,
A
as [x[ (14.2.22)
If A
=
1
, then,
K
=
4
3
Tr
(
1
)(
1
)(
1
(14.2.23)
Thus,
Tr
d
4
x F
=
1
2
d
3
4
3
Tr
(
1
)(
1
)(
1
. .. .
m (above)
(14.2.24)
Tr
d
4
x F
=
1
2
4
3
24
2
m
= 16
2
m (14.2.25)
m is instanton #.
Sectors of the vacuum are labelled by the winding number n.
[n` (14.2.26)
Perform a gauge transformation
T
1
[n` = [n + 1` (14.2.27)
But,
[T
1
, H] = 0 (14.2.28)
as H is gauge invariant. Thus the vacuum must be an eigenstate of T
1
.
Consider the vacuum.
[` =
+
n=
e
in
[n` (14.2.29)
i.e.
T
1
[` =
n=
e
in
[n + 1`
= e
i
n=
e
i(n+1)
[n + 1`
= e
i
[` (14.2.30)
Thus we have a vacuum labelled by this parameter .
Vacuum-vacuum transition.
'
t
[e
iHt
[` = (
t
)'
t
[e
iHt
[` (14.2.31)
=
m=
n=
'm[e
im
e
iHt
e
in
[n`
=
m,n
e
i(nm
)
'm[e
iHt
[n`
=
t
from (14.2.31)
=
=
e
i
B
.
TA
e
i
d
4
x /
Y M
(m = n +) (14.2.32)
([/ = A

1
n

n
at t = ] n, [B = A

1
m

m
at t = +] m)
The dominant instanton for this transition has instanton number .
=
1
24
2
Tr
d
4
x F
(14.2.33)
'
t
[e
iHt
[` =
m
n
TA
exp
d
4
x
L
Y M
+

24
2
Tr F
. .. .
(Violates Parity, CP)
(14.2.34)
L
Y M
=
1
4
F
leads to eective term in action, which leads to the Parity, CP violation. < 10
9
(electric dipole moment of neutron).

Quantum Field Theory - R. Clarkson, D. McKeon

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Quantum Field Theory - R. Clarkson, D. McKeon

Uploaded by

Copyright:

Available Formats

Cannot solve for

= 0 by (1). Thus, in H, we can set

and take H to be just H = H

cant make any sense of this (cant express

We can nally see that

is a reection of the fat that in S, is a freely chosen parameter,

and equate this with (1.4.26) (1.4.28)

() is a scalar eld moving in 0 + 1

Contrav. = der. of covar.

d c = 0 (int. of a constant c ) (2.1.2)

(i) 0 (const.) (ii) 0 (

d () [a(x) + b(x)] = a(x)

And so, for this to be zero, we must have:

= 0 (why we cant have two

We eliminate the unwanted negative norm state by building in an extra symmetry.

- Spinor in 0+1 dimensions.

Second Class constraint (2.3.2)

= 0 eliminates -ve norm states If

proper time guage conditions. (2.3.15)

= 0 in the Heisenberg picture, then

is a constant and satises the algebra,

A, electromagnetic potentials) (2.3.44)

matrices in each equation not a

= 4 4 matrix characterizing the Lorentz transformation.

(where rows/columns = [x, y, z, t])

= cosh [a[ + sinh [a[

Thus, (for ex.)

(i.e. Lorentz invariant).

(1) (4) (1) (4) (6)

(Picks up -ve Pseudo Scalar ).

(The above are vectors)

(The above are axial vectors)

(The above are Tensors)

Can express in closed form. (2.4.4)

(Closed form) (2.4.8)

(k, s)v(k, s) (2.4.32)

dx ((x a)(x + a)) f(x) (let a > 0) (2.4.38)

b(k, s)u(k, s)(k

(k, s)v(k, s)(k

We can show that there is a wave function

b(p, s)u(p, s)e

(p, s)v(p, s)e

A(t). In going from t t, we have the same physical

, the solutions generalize the spin 1/2 solution,

For a total of 10 symmetric 4 4 matrices. We can then have,

Thus, we get the Proca Equations (Massive spin 1 equations).

(x) under a Lorentz transformation.

, since it is a vector, under an innitesimal Lorentz transformation, we have,

- spin 1 representation of SO(3, 1) (Adjoint representation).

In the rest frame of a massive (m) particle,

Thus, in the rest frame

= (0, 0, k, k)), then we have,

= (0, 0, k, k) and thus

First class Constraints (4.1.11)

k) (Fourier transf.) (4.1.26)

by Gausss Law (Radius of Sphere = R)

= F(x, t), G(y, t)

The Fourier expansion of A(x, t) in the Coulomb gauge (

in a classical potential (4.1.64)

over all points, get .)

(k, )a(k, ) (5.2.19)

(k, )a(k, ) (5.2.20)

(x, t) and its time ordered product: