This action might not be possible to undo. Are you sure you want to continue?
r
X
i
v
:
0
8
1
2
.
4
4
0
8
v
1
[
h
e
p

t
h
]
2
3
D
e
c
2
0
0
8
A solution manual for Polchinski’s String Theory
Matthew Headrick
Martin Fisher School of Physics
Brandeis University
mph@brandeis.edu
Abstract
We present detailed solutions to 81 of the 202 problems in J. Polchinski’s twovolume text
book String Theory.
BRXTH604
CONTENTS 1
Contents
0 Preface 4
1 Chapter 1 5
1.1 Problem 1.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2 Problem 1.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3 Problem 1.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.4 Problem 1.7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.5 Problem 1.9 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2 Chapter 2 11
2.1 Problem 2.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 Problem 2.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.3 Problem 2.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.4 Problem 2.7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.5 Problem 2.9 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.6 Problem 2.11 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.7 Problem 2.13 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.8 Problem 2.15 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.9 Problem 2.17 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3 Chapter 3 21
3.1 Problem 3.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.2 Problem 3.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.3 Problem 3.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.4 Problem 3.4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.5 Problem 3.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.6 Problem 3.7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.7 Problem 3.9 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.8 Problem 3.11 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.9 Problem 3.13 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4 Chapter 4 33
4.1 Problem 4.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5 Chapter 5 35
5.1 Problem 5.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
CONTENTS 2
6 Chapter 6 40
6.1 Problem 6.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
6.2 Problem 6.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
6.3 Problem 6.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
6.4 Problem 6.7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
6.5 Problem 6.9 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
6.6 Problem 6.11 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
6.7 Problem 6.12 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
7 Chapter 7 52
7.1 Problem 7.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
7.2 Problem 7.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
7.3 Problem 7.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
7.4 Problem 7.7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
7.5 Problem 7.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
7.6 Problem 7.9 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
7.7 Problem 7.10 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
7.8 Problem 7.11 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
7.9 Problem 7.13 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
7.10 Problem 7.15 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
8 Chapter 8 69
8.1 Problem 8.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
8.2 Problem 8.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
8.3 Problem 8.4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
8.4 Problem 8.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
8.5 Problem 8.6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
8.6 Problem 8.7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
8.7 Problem 8.9 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
8.8 Problem 8.11 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
9 Appendix A 78
9.1 Problem A.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
9.2 Problem A.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
9.3 Problem A.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
10 Chapter 10 83
10.1 Problem 10.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
10.2 Problem 10.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
CONTENTS 3
10.3 Problem 10.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
10.4 Problem 10.4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
10.5 Problem 10.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
10.6 Problem 10.7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
10.7 Problem 10.10 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
10.8 Problem 10.11 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
10.9 Problem 10.14 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
11 Chapter 11 88
11.1 Problem 11.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
11.2 Problem 11.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
11.3 Problem 11.4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
11.4 Problem 11.7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
11.5 Problem 11.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
11.6 Problem 11.9 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
11.7 Problem 11.12 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
12 Chapter 13 94
12.1 Problem 13.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
12.2 Problem 13.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
12.3 Problem 13.4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
12.4 Problem 13.12 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
13 Chapter 14 103
13.1 Problem 14.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
13.2 Problem 14.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
13.3 Problem 14.6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
14 Chapter 15 108
14.1 Problem 15.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
14.2 Problem 15.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
15 Appendix B 110
15.1 Problem B.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
15.2 Problem B.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
15.3 Problem B.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
0 PREFACE 4
0 Preface
The following pages contain detailed solutions to 81 of the 202 problems in J. Polchinski’s two
volume textbook String Theory [1, 2]. I originally wrote up these solutions while teaching myself
the subject, and then later decided that they may be of some use to others doing the same. These
solutions are the work of myself alone, and carry no endorsement from Polchinski.
I would like to thank R. Britto, S. Minwalla, D. Podolsky, and M. Spradlin for help on these
problems. This work was done while I was a graduate student at Harvard University, and was
supported by an NSF Graduate Research Fellowship.
1 CHAPTER 1 5
1 Chapter 1
1.1 Problem 1.1
(a) We work in the gauge where τ = X
0
. Nonrelativistic motion means
˙
X
i
≡ v
i
≪1. Then
S
pp
= −m
dτ
−
˙
X
µ ˙
X
µ
= −m
dt
1 −v
2
≈
dt (
1
2
mv
2
−m). (1)
(b) Again, we work in the gauge τ = X
0
, and assume
˙
X
i
≡ v
i
≪ 1. Deﬁning u
i
≡ ∂
σ
X
i
, the
induced metric h
ab
= ∂
a
X
µ
∂
b
X
µ
becomes:
¦h
ab
¦ =
−1 +v
2
u v
u v u
2
. (2)
Using the fact that the transverse velocity of the string is
v
T
= v −
u v
u
u, (3)
the NambuGoto Lagrangian can be written:
L = −
1
2πα
′
dσ (−det¦h
ab
¦)
1/2
= −
1
2πα
′
dσ
u
2
(1 −v
2
) + (u v)
2
1/2
≈ −
1
2πα
′
dσ [u[
1 −
1
2
v
2
+
u v
2
2u
2
=
dσ [u[
1
2
ρv
2
T
−L
s
T, (4)
where
ρ =
1
2πα
′
(5)
is the mass per unit length of the string,
L
s
=
dσ [u[ (6)
is its physical length, and
T =
1
2πα
′
= ρ (7)
is its tension.
1 CHAPTER 1 6
1.2 Problem 1.3
It is well known that χ, the Euler characteristic of the surface, is a topological invariant, i.e. does
not depend on the metric. We will prove by explicit computation that, in particular, χ is invariant
under Weyl transformations,
γ
′
ab
= e
2ω(σ,τ)
γ
ab
. (8)
For this we will need the transformation law for the connection coeﬃcients,
Γ
′a
bc
= Γ
a
bc
+∂
b
ωδ
a
c
+∂
c
ωδ
a
b
−∂
d
ωγ
ad
γ
bc
, (9)
and for the curvature scalar,
R
′
= e
−2ω
(R −2∇
a
∂
a
ω). (10)
Since the tangent and normal vectors at the boundary are normalized, they transform as
t
′a
= e
−ω
t
a
, (11)
n
′
a
= e
ω
n
a
. (12)
The curvature of the boundary thus transforms as follows:
k
′
= ±t
′a
n
′
b
(∂
a
t
′b
+ Γ
′b
ac
t
′c
)
= e
−ω
(k ∓t
a
t
a
n
b
∂
d
), (13)
where we have used (9), (11), (12), and the fact that n and t are orthogonal. If the boundary is
timelike then t
a
t
a
= −1 and we must use the upper sign, whereas if it is spacelike then t
a
t
a
= 1
and we must use the lower sign. Hence
k
′
= e
−ω
(k +n
a
∂
a
ω). (14)
Finally, since ds = (−γ
ττ
)
1/2
dτ for a timelike boundary and ds = γ
1/2
σσ
dσ for a spacelike bounday,
ds
′
= ds e
ω
. (15)
Putting all of this together, and applying Stokes theorem, which says that for any vector v
a
,
M
dτ dσ (−γ)
1/2
∇
a
v
a
=
∂M
ds n
a
v
a
, (16)
we ﬁnd the transformation law for χ:
χ
′
=
1
4π
M
dτ dσ (−γ
′
)
1/2
R
′
+
1
2π
∂M
ds
′
k
′
=
1
4π
M
dτ dσ (−γ)
1/2
(R −2∇
a
∂
a
ω) +
1
2π
∂M
ds (k +n
a
∂
a
ω)
= χ. (17)
1 CHAPTER 1 7
1.3 Problem 1.5
For simplicity, let us deﬁne a ≡ (π/2p
+
α
′
l)
1/2
. Then we wish to evaluate
∞
¸
n=1
(n −θ) exp[−(n −θ)ǫa]
= −
d
d(ǫa)
∞
¸
n=1
exp[−(n −θ)ǫa]
= −
d
d(ǫa)
e
θǫa
e
ǫa
−1
= −
d
d(ǫa)
1
ǫa
+θ −
1
2
+
1
12
−
θ
2
+
θ
2
2
ǫa +O(ǫa)
2
=
1
(ǫa)
2
−
1
2
1
6
−θ +θ
2
+O(ǫa). (18)
As expected, the cutoﬀ dependent term is independent of θ; the ﬁnite result is
−
1
2
1
6
−θ +θ
2
. (19)
1.4 Problem 1.7
The mode expansion satisfying the boundary conditions is
X
25
(τ, σ) =
√
2α
′
¸
n
1
n
α
25
n
exp
¸
−
iπncτ
l
sin
πnσ
l
, (20)
where the sum runs over the halfoddintegers, n = 1/2, −1/2, 3/2, −3/2, . . . . Note that there is no
p
25
. Again, Hermiticity of X
25
implies α
25
−n
= (α
25
n
)
†
. Using (1.3.18),
Π
25
(τ, σ) = −
i
√
2α
′
l
¸
n
α
25
n
exp
¸
−
iπncτ
l
sin
πnσ
l
. (21)
We will now determine the commutation relations among the α
25
n
from the equal time com
mutation relations (1.3.24b). Not surprisingly, they will come out the same as for the free string
(1.3.25b). We have:
iδ(σ −σ
′
) = [X
25
(τ, σ), Π
25
(τ, σ)] (22)
= −
i
l
¸
n,n
′
1
n
[α
25
n
, α
25
n
′ ] exp
¸
−
iπ(n +n
′
)cτ
l
sin
πnσ
l
sin
πn
′
σ
′
l
.
Since the LHS does not depend on τ, the coeﬃcient of exp[−iπmcτ/l] on the RHS must vanish for
m = 0:
1
l
¸
n
1
n
[α
25
n
, α
25
m−n
] sin
πnσ
l
sin
π(n −m)σ
′
l
= δ(σ −σ
′
)δ
m,0
. (23)
1 CHAPTER 1 8
Multiplying both sides by sin[πn
′
σ/l] and integrating over σ now yields,
1
2n
[α
25
n
, α
25
m−n
] sin
π(n −m)σ
′
l
+ [α
25
n+m
, α
25
−n
] sin
π(n +m)σ
′
l
(24)
= sin
πnσ
′
l
δ
m,0
,
or,
[α
25
n
, α
25
m−n
] = nδ
m,0
, (25)
as advertised.
The part of the Hamiltonian (1.3.19) contributed by the X
25
oscillators is
l
4πα
′
p
+
l
0
dσ
2πα
′
Π
25
2
+
1
2πα
′
∂
σ
X
25
2
=
1
4α
′
p
+
l
¸
n,n
′
α
25
n
α
25
n
′ exp
¸
−
iπ(n +n
′
)cτ
l
l
0
dσ
−sin
πnσ
l
sin
πn
′
σ
l
+ cos
πnσ
l
cos
πn
′
σ
l
=
1
4α
′
p
+
¸
n
α
25
n
α
25
−n
=
1
4α
′
p
+
∞
¸
n=1/2
α
25
n
α
25
−n
+α
25
−n
α
25
n
=
1
2α
′
p
+
∞
¸
n=1/2
α
25
−n
α
25
n
+
n
2
=
1
2α
′
p
+
¸
∞
¸
n=1/2
α
25
−n
α
25
n
+
1
48
¸
, (26)
where we have used (19) and (25). Thus the mass spectrum (1.3.36) becomes
m
2
= 2p
+
H −p
i
p
i
(i = 2, . . . , 24)
=
1
α
′
N −
15
16
, (27)
where the level spectrum is given in terms of the occupation numbers by
N =
24
¸
i=2
∞
¸
n=1
nN
in
+
∞
¸
n=1/2
nN
25,n
. (28)
The ground state is still a tachyon,
m
2
= −
15
16α
′
. (29)
1 CHAPTER 1 9
The ﬁrst excited state has the lowest X
25
oscillator excited (N
25,1/2
= 1), and is also tachyonic:
m
2
= −
7
16α
′
. (30)
There are no massless states, as the second excited state is already massive:
m
2
=
1
16α
′
. (31)
This state is 24fold degenerate, as it can be reached either by N
i,1
= 1 or by N
25,1/2
= 2. Thus it
is a massive vector with respect to the SO(24,1) Lorentz symmetry preserved by the Dbrane. The
third excited state, with
m
2
=
9
16α
′
, (32)
is 25fold degenerate and corresponds to a vector plus a scalar on the Dbrane—it can be reached
by N
25,1/2
= 1, by N
25,1/2
= 3, or by N
i,1
= 1, N
25,1/2
= 1.
1.5 Problem 1.9
The mode expansion for X
25
respecting the boundary conditions is essentially the same as the
mode expansion (1.4.4), the only diﬀerences being that the ﬁrst two terms are no longer allowed,
and the oscillator label n, rather than running over the nonzero integers, must now run over the
halfoddintegers as it did in Problem 1.7:
X
25
(τ, σ) = (33)
i
α
′
2
¸
n
α
25
n
n
exp
¸
−
2πin(σ +cτ)
l
+
˜ α
25
n
n
exp
¸
2πin(σ −cτ)
l
.
The canonical commutators are the same as for the untwisted closed string, (1.4.6c) and (1.4.6d),
[α
25
m
, α
25
n
] = mδ
m,−n
, (34)
[ ˜ α
25
m
, ˜ α
25
n
] = mδ
m,−n
, (35)
as are the mass formula (1.4.8),
m
2
=
2
α
′
(N +
˜
N +A+
˜
A), (36)
the generator of σtranslations (1.4.10),
P = −
2π
l
(N −
˜
N), (37)
and (therefore) the levelmatching condition (1.4.11),
N =
˜
N. (38)
1 CHAPTER 1 10
However, the level operator N is now slightly diﬀerent,
N =
24
¸
i=2
∞
¸
n=1
α
i
−n
α
i
n
+
∞
¸
n=1/2
α
25
−n
α
25
n
; (39)
in fact, it is the same as the level operator for the open string on a D24brane of Problem 1.7.
The leftmoving level spectrum is therefore given by (28), and similarly for the rightmoving level
operator
˜
N. The zeropoint constants are also the same as in Problem 1.7:
A =
˜
A =
1
2
¸
24
¸
i=2
∞
¸
n=1
n +
∞
¸
n=1/2
n
¸
= −
15
16
. (40)
At a given level N =
˜
N, the occupation numbers N
in
and
˜
N
in
may be chosen independently, so
long as both sets satisfy (28). Therefore the spectrum at that level will consist of the product of
two copies of the Dbrane open string spectrum, and the masssquared of that level (36) will be 4
times the open string masssquared (27). We will have tachyons at levels N = 0 and N = 1/2, with
m
2
= −
15
4α
′
(41)
and
m
2
= −
7
4α
′
, (42)
respectively. The lowest nontachyonic states will again be at level N = 1: a second rank SO(24)
tensor with
m
2
=
1
4α
′
, (43)
which can be decomposed into a scalar, an antisymmetric tensor, and a traceless symmetric tensor.
2 CHAPTER 2 11
2 Chapter 2
2.1 Problem 2.1
(a) For holomorphic test functions f(z),
R
d
2
z ∂
¯
∂ ln [z[
2
f(z) =
R
d
2
z
¯
∂
1
z
f(z)
= −i
∂R
dz
1
z
f(z)
= 2πf(0). (1)
For antiholomorphic test functions f(¯ z),
R
d
2
z ∂
¯
∂ ln [z[
2
f(¯ z) =
R
d
2
z ∂
1
¯ z
f(¯ z)
= i
∂R
d¯ z
1
¯ z
f(¯ z)
= 2πf(0). (2)
(b) We regulate ln[z[
2
by replacing it with ln([z[
2
+ ǫ). This lead to regularizations also of 1/¯ z
and 1/z:
∂
¯
∂ ln([z[
2
+ǫ) = ∂
z
[z[
2
+ǫ
=
¯
∂
¯ z
[z[
2
+ǫ
=
ǫ
([z[
2
+ǫ)
2
. (3)
Working in polar coordinates, consider a general test function f(r, θ), and deﬁne g(r
2
) ≡
dθ f(r, θ).
Then, assuming that g is suﬃciently well behaved at zero and inﬁnity,
d
2
z
ǫ
([z[
2
+ǫ)
2
f(z, ¯ z)
=
∞
0
du
ǫ
(u +ǫ)
2
g(u)
=
−
ǫ
u +ǫ
g(u) +ǫ ln(u +ǫ)g
′
(u)
∞
0
−
∞
0
duǫ ln(u +ǫ)g
′′
(u).
= g(0)
= 2πf(0). (4)
2 CHAPTER 2 12
2.2 Problem 2.3
(a) The leading behavior of the expectation value as z
1
→z
2
is
n
¸
i=1
: e
ik
i
X(z
i
,¯ z
i
)
:
¸
= iC
X
(2π)
D
δ
D
n
¸
i=1
k
i
n
¸
i,j=1
[z
ij
[
α
′
k
i
k
j
= [z
12
[
α
′
k
1
k
2
iC
X
(2π)
D
δ
D
(k
1
+k
2
+
n
¸
i=3
k
i
)
n
¸
i=3
[z
1i
[
α
′
k
1
k
i
[z
2i
[
α
′
k
2
k
i
n
¸
i,j=3
[z
ij
[
α
′
k
i
k
j
≈ [z
12
[
α
′
k
1
k
2
iC
X
(2π)
D
δ
D
(k
1
+k
2
+
n
¸
i=3
k
i
)
n
¸
i=3
[z
2i
[
α
′
(k
1
+k
2
)k
i
n
¸
i,j=3
[z
ij
[
α
′
k
i
k
j
= [z
12
[
α
′
k
1
k
2
: e
i(k
1
+k
2
)X(z
2
,¯ z
2
)
:
n
¸
i=3
: e
ik
i
X(z
i
,¯ z
i
)
:
¸
, (5)
in agreement with (2.2.14).
(b) The z
i
dependence of the expectation value is given by
[z
23
[
α
′
k
2
k
3
[z
12
[
α
′
k
1
k
2
[z
13
[
α
′
k
1
k
3
= [z
23
[
α
′
k
2
k
3
[z
12
[
α
′
k
1
k
2
[z
23
[
α
′
k
1
k
3
1 +
z
12
z
23
α
′
k
1
k
3
(6)
= [z
23
[
α
′
(k
1
+k
2
)k
3
[z
12
[
α
′
k
1
k
2
∞
¸
k=0
Γ(
1
2
α
′
k
1
k
3
+ 1)
k! Γ(
1
2
α
′
k
1
k
3
−k + 1)
z
12
z
23
k
∞
¸
k=0
Γ(
1
2
α
′
k
1
k
3
+ 1)
k! Γ(
1
2
α
′
k
1
k
3
−k + 1)
¯ z
12
¯ z
23
k
.
The radius of convergence of a power series is given by the limit as k →∞ of [a
k
/a
k+1
[, where the
a
k
are the coeﬃcients of the series. In this case, for both of the above power series,
R = lim
k→∞
(k + 1)! Γ(
1
2
α
′
k
1
k
3
−k)
k! Γ(
1
2
α
′
k
1
k
3
−k + 1)
z
23
= [z
23
[. (7)
(c) Consider the interior of the dashed line in ﬁgure 2.1, that is, the set of points z
1
satisfying
[z
12
[ < [z
23
[. (8)
2 CHAPTER 2 13
By equation (2.1.23), the expectation value
': X
µ
(z
1
, ¯ z
1
)X
ν
(z
2
, ¯ z
2
) : /(z
3
, ¯ z
3
)B(z
4
, ¯ z
4
)` (9)
is a harmonic function of z
1
within this region. It can therefore be written as the sum of a
holomorphic and an antiholomorphic function (this statement is true in any simply connected
region). The Taylor expansion of a function that is holomorphic on an open disk (about the center
of the disk), converges on the disk; similarly for an antiholomorphic function. Hence the two Taylor
series on the RHS of (2.2.4) must converge on the disk.
2.3 Problem 2.5
Under the variation of the ﬁelds φ
α
(σ) →φ
α
(σ) +δφ
α
(σ), the variation of the Lagrangian is
δL =
∂L
∂φ
α
δφ
α
+
∂L
∂(∂
a
φ
α
)
∂
a
δφ
α
. (10)
The Lagrangian equations of motion (EulerLagrange equations) are derived by assuming that the
action is stationary under an arbitrary variation δφ
α
(σ) that vanishes at inﬁnity:
0 = δS
=
d
d
σ δL
=
d
d
σ
∂L
∂φ
α
δφ
α
+
∂L
∂(∂
a
φ
α
)
∂
a
δφ
α
=
d
d
σ
∂L
∂φ
α
−∂
a
∂L
∂(∂
a
φ
α
)
δφ
α
(11)
implies
∂L
∂φ
α
−∂
a
∂L
∂(∂
a
φ
α
)
= 0. (12)
Instead of assuming that δφ
α
vanishes at inﬁnity, let us assume that it is a symmetry. In this case,
the variation of the Lagrangian (10) must be a total derivative to insure that the action on bounded
regions varies only by a surface term, thereby not aﬀecting the equations of motion:
δL = ǫ∂
a
/
a
; (13)
/
a
is assumed to be a local function of the ﬁelds and their derivatives, although it is not obvious
how to prove that this can always be arranged. Using (10), (12), and (13),
∂
a
j
a
= 2πi∂
a
∂L
∂(∂
a
φ
α
)
ǫ
−1
δφ
α
−/
a
=
2πi
ǫ
∂L
∂φ
α
δφ
α
+
∂L
∂(∂
a
φ
α
)
∂
a
δφ
α
−δL
= 0. (14)
2 CHAPTER 2 14
If we now vary the ﬁelds by ρ(σ)δφ
α
(σ), where δφ
α
is a symmetry as before but ρ is an arbitrary
function, then the variation of the action will be
δL =
∂L
∂(∂
a
φ
α
)
∂
a
(δφ
α
ρ) +
∂L
∂φ
α
δφ
α
ρ
=
∂L
∂(∂
a
φ
α
)
∂
a
δφ
α
+
∂L
∂φ
α
δφ
α
ρ +
∂L
∂(∂
a
φ
α
)
δφ
α
∂
a
ρ. (15)
Equation (13) must be satisﬁed in the case ρ(σ) is identically 1, so the factor in parentheses must
equal ǫ∂
a
/
a
:
δS =
d
d
σ
ǫ∂
a
/
a
ρ +
∂L
∂(∂
a
φ
α
)
δφ
α
∂
a
ρ
=
d
d
σ
−ǫ/
a
+
∂L
∂(∂
a
φ
α
)
δφ
α
∂
a
ρ
=
ǫ
2πi
d
d
σ j
a
∂
a
ρ, (16)
where we have integrated by parts, assuming that ρ falls oﬀ at inﬁnity. Since δ exp(−S) =
−exp(−S)δS, this agrees with (2.3.4) for the case of ﬂat space, ignoring the transformation of
the measure.
2.4 Problem 2.7
(a) X
µ
:
T(z)X
µ
(0, 0) = −
1
α
′
: ∂X
ν
(z)∂X
ν
(z) : X
µ
(0, 0) ∼
1
z
∂X
µ
(z) ∼
1
z
∂X
µ
(0)
˜
T(¯ z)X
µ
(0, 0) = −
1
α
′
:
¯
∂X
ν
(¯ z)
¯
∂X
ν
(¯ z) : X
µ
(0, 0) ∼
1
¯ z
¯
∂X
µ
(¯ z) ∼
1
¯ z
¯
∂X
µ
(0) (17)
∂X
µ
:
T(z)∂X
µ
(0) ∼
1
z
2
∂X
µ
(z) ∼
1
z
2
∂X
µ
(0) +
1
z
∂
2
X
µ
(0)
˜
T(¯ z)∂X
µ
(0) ∼ 0 (18)
¯
∂X
µ
:
T(z)
¯
∂X
µ
(0) ∼ 0
˜
T(¯ z)
¯
∂X
µ
(0) ∼
1
¯ z
2
¯
∂X
µ
(¯ z) ∼
1
¯ z
2
¯
∂X
µ
(0) +
1
¯ z
¯
∂
2
X
µ
(0) (19)
∂
2
X
µ
:
T(z)∂
2
X
µ
(0) ∼
2
z
3
∂X
µ
(z) ∼
2
z
3
∂X
µ
(0) +
2
z
2
∂
2
X
µ
(0) +
1
z
∂
3
X
µ
(0)
˜
T(¯ z)∂
2
X
µ
(0) ∼ 0 (20)
2 CHAPTER 2 15
: e
ikX
::
T(z) : e
ikX(0,0)
: ∼
α
′
k
2
4z
2
: e
ikX(0,0)
: +
1
z
ik
µ
: ∂X
µ
(z)e
ikX(0,0)
:
∼
α
′
k
2
4z
2
: e
ikX(0,0)
: +
1
z
ik
µ
: ∂X
µ
(0)e
ikX(0,0)
:
˜
T(¯ z) : e
ikX(0,0)
: ∼
α
′
k
2
4¯ z
2
: e
ikX(0,0)
: +
1
¯ z
ik
µ
:
¯
∂X
µ
(¯ z)e
ikX(0,0)
:
∼
α
′
k
2
4¯ z
2
: e
ikX(0,0)
: +
1
¯ z
ik
µ
: ∂X
µ
(0)e
ikX(0,0)
: (21)
(b) In the linear dilaton theory, the energymomentum tensor is
T = −
1
α
′
: ∂X
µ
∂X
µ
: +V
µ
∂
2
X
µ
,
˜
T = −
1
α
′
:
¯
∂X
µ
¯
∂X
µ
: +V
µ
¯
∂
2
X
µ
, (22)
so it suﬃces to calculate the OPEs of the various operators with the terms V
µ
∂
2
X
µ
and V
µ
¯
∂
2
X
µ
and add them to the results found in part (a).
X
µ
:
V
ν
∂
2
X
ν
(z)X
µ
(0, 0) ∼
α
′
V
µ
2z
2
V
ν
¯
∂
2
X
ν
(¯ z)X
µ
(0, 0) ∼
α
′
V
µ
2¯ z
2
(23)
Not only is X
µ
is not a tensor anymore, but it does not even have welldeﬁned weights, because it
is not an eigenstate of rigid transformations.
∂X
µ
:
V
ν
∂
2
X
ν
(z)∂X
µ
(0) ∼
α
′
V
µ
z
3
V
ν
¯
∂
2
X
ν
(¯ z)∂X
µ
(0) ∼ 0 (24)
So ∂X
µ
still has weights (1,0), but it is no longer a tensor operator.
¯
∂X
µ
:
V
ν
∂
2
X
ν
(z)
¯
∂X
µ
(0) ∼ 0
V
ν
¯
∂
2
X
ν
(¯ z)
¯
∂X
µ
(0) ∼
α
′
V
µ
¯ z
3
(25)
Similarly,
¯
∂X
µ
still has weights (0,1), but is no longer a tensor.
∂
2
X
µ
:
V
ν
∂
2
X
ν
(z)∂
2
X
µ
(0) ∼
3α
′
V
µ
z
4
V
ν
¯
∂
2
X
ν
(¯ z)∂
2
X
µ
(0) ∼ 0 (26)
2 CHAPTER 2 16
Nothing changes from the scalar theory: the weights are still (2,0), and ∂
2
X
µ
is still not a tensor.
: e
ikX
::
V
ν
∂
2
X
ν
(z) : e
ikX(0,0)
:∼
iα
′
V k
2z
2
: e
ikX(0,0)
:
V
ν
¯
∂
2
X
ν
(¯ z) : e
ikX(0,0)
:∼
iα
′
V k
2¯ z
2
: e
ikX(0,0)
: (27)
Thus : e
ikX
: is still a tensor, but, curiously, its weights are now complex:
α
′
4
(k
2
+ 2iV k),
α
′
4
(k
2
+ 2iV k)
. (28)
2.5 Problem 2.9
Since we are interested in ﬁnding the central charges of these theories, it is only necessary to
calculate the 1/z
4
terms in the TT OPEs, the rest of the OPE being determined by general
considerations as in equation (2.4.25). In the following, we will therefore drop all terms less singular
than 1/z
4
. For the linear dilaton CFT,
T(z)T(0) =
1
α
′2
: ∂X
µ
(z)∂X
µ
(z) :: ∂X
ν
(0)∂X
ν
(0) :
−
2V
ν
α
′
: ∂X
µ
(z)∂X
µ
(z) : ∂
2
X
ν
(0)
−
2V
µ
α
′
∂
2
X
µ
(z) : ∂X
ν
(0)∂X
ν
(0) : +V
µ
V
ν
∂
2
X
µ
(z)∂
2
X
ν
(0)
∼
D
2z
4
+
3α
′
V
2
z
4
+O
1
z
2
, (29)
so
c = D + 6α
′
V
2
. (30)
Similarly,
˜
T(¯ z)
˜
T(0) =
1
α
′2
:
¯
∂X
µ
(¯ z)
¯
∂X
µ
(¯ z) ::
¯
∂X
ν
(0)
¯
∂X
ν
(0) :
−
2V
ν
α
′
:
¯
∂X
µ
(¯ z)
¯
∂X
µ
(¯ z) :
¯
∂
2
X
ν
(0)
−
2V
µ
α
′
¯
∂
2
X
µ
(¯ z) :
¯
∂X
ν
(0)
¯
∂X
ν
(0) : +V
µ
V
ν
¯
∂
2
X
µ
(¯ z)
¯
∂
2
X
ν
(0)
∼
D
2¯ z
4
+
3α
′
V
2
¯ z
4
+O
1
¯ z
2
, (31)
so
˜ c = D + 6α
′
V
2
. (32)
2 CHAPTER 2 17
For the bc system,
T(z)T(0) = (1 −λ)
2
: ∂b(z)c(z) :: ∂b(0)c(0) :
−λ(1 −λ) : ∂b(z)c(z) :: b(0)∂c(0) :
−λ(1 −λ) : b(z)∂c(z) :: ∂b(0)c(0) :
+λ
2
: b(z)∂c(z) :: b(0)∂c(0) :
∼
−6λ
2
+ 6λ −1
z
4
+O
1
z
2
, (33)
so
c = −12λ
2
+ 12λ −2. (34)
Of course
˜
T(¯ z)
˜
T(0) = 0, so ˜ c = 0.
The βγ system has the same energymomentum tensor and almost the same OPEs as the bc
system. While γ(z)β(0) ∼ 1/z as in the bc system, now β(z)γ(0) ∼ −1/z. Each term in (33)
involved one b(z)c(0) contraction and one c(z)b(0) contraction, so the central charge of the βγ
system is minus that of the bc system:
c = 12λ
2
−12λ + 2. (35)
Of course ˜ c = 0 still.
2.6 Problem 2.11
Assume without loss of generality that m > 1; for m = 0 and m = ±1 the central charge term in
(2.6.19) vanishes, while m < −1 is equivalent to m > 1. Then L
m
annihilates [0; 0`, as do all but
m−1 of the terms in the mode expansion (2.7.6) of L
−m
:
L
−m
[0; 0` =
1
2
m−1
¸
n=1
α
µ
n−m
α
µ(−n)
[0; 0`. (36)
Hence the LHS of (2.6.19), when applied to [0; 0`, yields,
[L
m
, L
−m
][0; 0`
= L
m
L
−m
[0; 0` −L
−m
L
m
[0; 0`
=
1
4
∞
¸
n
′
=−∞
m−1
¸
n=1
α
ν
m−n
′ α
νn
′ α
µ
n−m
α
µ(−n)
[0; 0`
=
1
4
m−1
¸
n=1
∞
¸
n
′
=−∞
(m−n
′
)n
′
η
νµ
η
νµ
δ
n
′
n
+ (m−n
′
)n
′
δ
ν
µ
δ
µ
ν
δ
m−n
′
,n
[0; 0`
=
D
2
m−1
¸
n=1
n(m−n)[0; 0`
=
D
12
m(m
2
−1)[0; 0`. (37)
2 CHAPTER 2 18
Meanwhile, the RHS of (2.6.19) applied to the same state yields,
2mL
0
+
c
12
(m
3
−m)
[0; 0` =
c
12
(m
3
−m)[0; 0`, (38)
so
c = D. (39)
2.7 Problem 2.13
(a) Using (2.7.16) and (2.7.17),
◦
◦b(z)c(z
′
)
◦
◦ =
∞
¸
m,m
′
=−∞
◦
◦b
m
c
m
′
◦
◦
z
m+λ
z
′m
′
+1−λ
=
∞
¸
m,m
′
=−∞
b
m
c
m
′
z
m+λ
z
′m
′
+1−λ
−
∞
¸
m=0
1
z
m+λ
z
′−m+1−λ
= b(z)c(z
′
) −
z
z
′
1−λ
1
z −z
′
. (40)
With (2.5.7),
: b(z)c(z
′
) : −
◦
◦b(z)c(z
′
)
◦
◦ =
1
z −z
′
z
z
′
1−λ
−1
. (41)
(b) By taking the limit of (41) as z
′
→z, we ﬁnd,
: b(z)c(z) : −
◦
◦b(z)c(z)
◦
◦ =
1 −λ
z
. (42)
Using (2.8.14) we have,
N
g
= Q
g
−λ +
1
2
=
1
2πi
dz j
z
−λ +
1
2
= −
1
2πi
dz : b(z)c(z) : −λ +
1
2
= −
1
2πi
dz
◦
◦b(z)c(z)
◦
◦ −
1
2
. (43)
(c) If we rewrite the expansion (2.7.16) of b(z) in the wframe using the tensor transformation
law (2.4.15), we ﬁnd,
b(w) = (∂
z
w)
−λ
b(z)
= (−iz)
λ
∞
¸
m=−∞
b
m
z
m+λ
= e
−πiλ/2
∞
¸
m=−∞
e
imw
b
m
. (44)
2 CHAPTER 2 19
Similarly,
c(w) = e
−πi(1−λ)/2
∞
¸
m=−∞
e
imw
c
m
. (45)
Hence, ignoring ordering,
j
w
(w) = −b(w)c(w)
= i
∞
¸
m,m
′
=−∞
e
i(m+m
′
)w
b
m
c
m
′ , (46)
and
N
g
= −
1
2πi
2π
0
dwj
w
= −
∞
¸
m=−∞
b
m
c
−m
= −
∞
¸
m=−∞
◦
◦b
m
c
−m
◦
◦ −
∞
¸
m=0
1. (47)
The ordering constant is thus determined by the value of the second inﬁnite sum. If we write,
more generally,
¸
∞
m=0
a, then we must regulate the sum in such a way that the divergent part is
independent of a. For instance,
∞
¸
m=0
ae
−ǫa
=
a
1 −e
−ǫa
=
1
ǫ
+
a
2
+O(ǫ); (48)
the ǫindependent part is a/2, so the ordering constant in (47) equals −1/2.
2.8 Problem 2.15
To apply the doubling trick to the ﬁeld X
µ
(z, ¯ z), deﬁne for ℑz < 0,
X
µ
(z, ¯ z) ≡ X
µ
(z
∗
, ¯ z
∗
). (49)
Then
∂
m
X
µ
(z) =
¯
∂
m
X
µ
(¯ z
∗
), (50)
so that in particular for z on the real line,
∂
m
X
µ
(z) =
¯
∂
m
X
µ
(¯ z), (51)
2 CHAPTER 2 20
as can also be seen from the mode expansion (2.7.26). The modes α
µ
m
are deﬁned as integrals over
a semicircle of ∂X
µ
(z) +
¯
∂X
µ
(¯ z), but with the doubling trick the integral can be extended to the
full circle:
α
µ
m
=
2
α
′
dz
2π
z
m
∂X
µ
(z) = −
2
α
′
d¯ z
2π
¯ z
m
¯
∂X
µ
(¯ z). (52)
At this point the derivation proceeds in exactly the same manner as for the closed string treated
in the text. With no operator at the origin, the ﬁelds are holomorphic inside the contour, so with
m positive, the contour integrals (52) vanish, and the state corresponding to the unit operator
“inserted” at the origin must be the ground state [0; 0`:
1(0, 0)
∼
= [0; 0`. (53)
The state α
µ
−m
[0; 0` (m positive) is given by evaluating the integrals (52), with the ﬁelds holomorphic
inside the contours:
α
µ
−m
[0; 0`
∼
=
2
α
′
1/2
i
(m−1)!
∂
m
X
µ
(0) =
2
α
′
1/2
i
(m−1)!
¯
∂
m
X
µ
(0). (54)
Similarly, using the mode expansion (2.7.26), we see that X
µ
(0, 0)[0; 0` = x
µ
[0; 0`, so
x
µ
[0; 0`
∼
= X
µ
(0, 0). (55)
As in the closed string case, the same correspondence applies when these operators act on states
other than the ground state, as long as we normal order the resulting local operator. The result
is therefore exactly the same as (2.8.7a) and (2.8.8) in the text; for example, (2.8.9) continues to
hold.
2.9 Problem 2.17
Take the matrix element of (2.6.19) between '1[ and [1`, with n = −m and m > 1. The LHS yields,
'1[[L
m
, L
−m
][1` = '1[L
†
−m
L
−m
[1`
= L
−m
[1`
2
, (56)
using (2.9.9). Also by (2.9.9), L
0
[1` = 0, so on the RHS we are left with the term
c
12
(m
3
−m)'1[1`. (57)
Hence
c =
12
m
3
−m
L
−m
[1`
2
'1[1`
≥ 0. (58)
3 CHAPTER 3 21
3 Chapter 3
3.1 Problem 3.1
(a) The deﬁnition of the geodesic curvature k of a boundary given in Problem 1.3 is
k = −n
b
t
a
∇
a
t
b
, (1)
where t
a
is the unit tangent vector to the boundary and n
b
is the outward directed unit normal. For
a ﬂat unit disk, R vanishes, while the geodesic curvature of the boundary is 1 (since t
a
∇
a
t
b
= −n
b
).
Hence
χ =
1
2π
2π
0
dθ = 1. (2)
For the unit hemisphere, on the other hand, the boundary is a geodesic, while R = 2. Hence
χ =
1
4π
d
2
σ g
1/2
2 = 1, (3)
in agreement with (2).
(b) If we cut a surface along a closed curve, the two new boundaries will have oppositely directed
normals, so their contributions to the Euler number of the surface will cancel, leaving it unchanged.
The Euler number of the unit sphere is
χ =
1
4π
d
2
σ g
1/2
2 = 2. (4)
If we cut the sphere along b small circles, we will be left with b disks and a sphere with b holes.
The Euler number of the disks is b (from part (a)), so the Euler number of the sphere with b holes
is
χ = 2 −b. (5)
(c) A ﬁnite cylinder has Euler number 0, since we can put on it a globally ﬂat metric for which
the boundaries are geodesics. If we remove from a sphere b + 2g holes, and then join to 2g of the
holes g cylinders, the result will be a sphere with b holes and g handles; its Euler number will be
χ = 2 −b −2g. (6)
3.2 Problem 3.2
(a) This is easiest to show in complex coordinates, where g
zz
= g
¯ z¯ z
= 0. Contracting two indices
of a symmetric tensor with lower indices by g
ab
will pick out the components where one of the
indices is z and the other ¯ z. If the tensor is traceless then all such components must vanish. The
only nonvanishing components are therefore the one with all z indices and the one with all ¯ z
indices.
3 CHAPTER 3 22
(b) Let v
a
1
an
be a traceless symmetric tensor. Deﬁne P
n
by
(P
n
v)
a
1
a
n+1
≡ ∇
(a
1
v
a
2
a
n+1
)
−
n
n + 1
g
(a
1
a
2
∇
[b[
v
b
a
3
a
n+1
)
. (7)
This tensor is symmetric by construction, and it is easy to see that it is also traceless. Indeed,
contracting with g
a
1
a
2
, the ﬁrst term becomes
g
a
1
a
2
∇
(a
1
v
a
2
a
n+1
)
=
2
n + 1
∇
b
v
b
a
3
a
n+1
, (8)
where we have used the symmetry and tracelessness of v, and the second cancels the ﬁrst:
g
a
1
a
2
g
(a
1
a
2
∇
[b[
v
b
a
3
a
n+1
)
=
2
n(n + 1)
g
a
1
a
2
g
a
1
a
2
∇
b
v
b
a
3
a
n+1
+
2(n −1)
n(n + 1)
g
a
1
a
2
g
a
1
a
3
∇
b
v
b
a
2
a
4
a
n+1
=
2
n
∇
b
v
b
a
3
a
n+1
. (9)
(c) For u
a
1
a
n+1
a traceless symmetric tensor, deﬁne P
T
n
by
(P
T
n
u)
a
1
an
≡ −∇
b
u
b
a
1
an
. (10)
This inherits the symmetry and tracelessness of u.
(d)
(u, P
n
v) =
d
2
σ g
1/2
u
a
1
a
n+1
(P
n
v)
a
1
a
n+1
=
d
2
σ g
1/2
u
a
1
a
n+1
∇
a
1
v
a
2
a
n+1
−
n
n + 1
g
a
1
a
2
∇
b
v
b
a
3
a
n+1
= −
d
2
σ g
1/2
∇
a
1
u
a
1
a
n+1
v
a
2
a
n+1
=
d
2
σ g
1/2
(P
T
n
u)
a
2
a
n+1
v
a
2
a
n+1
= (P
T
n
u, v) (11)
3.3 Problem 3.3
(a) The conformal gauge metric in complex coordinates is g
z¯ z
= g
¯ zz
= e
2ω
/2, g
zz
= g
¯ z¯ z
= 0.
Connection coeﬃcients are quickly calculated:
Γ
z
zz
=
1
2
g
z¯ z
(∂
z
g
z¯ z
+∂
z
g
¯ zz
−∂
¯ z
g
zz
)
= 2∂ω, (12)
Γ
¯ z
¯ z¯ z
= 2
¯
∂ω, (13)
3 CHAPTER 3 23
all other coeﬃcients vanishing.
This leads to the following simpliﬁcation in the formula for the covariant derivative:
∇
z
T
a
1
am
b
1
bn
= ∂
z
T
a
1
am
b
1
bn
+
m
¸
i=1
Γ
a
i
zc
T
a
1
cam
b
1
bn
−
n
¸
j=1
Γ
c
zb
j
T
a
1
am
b
1
cbn
=
¸
∂ + 2∂ω
m
¸
i=1
δ
a
i
z
−2∂ω
n
¸
j=1
δ
z
b
j
¸
T
a
1
am
b
1
bn
; (14)
in other words, it counts the diﬀerence between the number of upper z indices and lower z indices,
while ¯ z indices do not enter. Similarly,
∇
¯ z
T
a
1
am
b
1
bn
=
¸¯
∂ + 2
¯
∂ω
m
¸
i=1
δ
a
i
¯ z
−2
¯
∂ω
n
¸
j=1
δ
¯ z
b
j
¸
T
a
1
am
b
1
bn
. (15)
In particular, the covariant derivative with respect to z of a tensor with only ¯ z indices is equal to
its regular derivative, and vice versa:
∇
z
T
¯ z¯ z
¯ z¯ z
= ∂T
¯ z¯ z
¯ z¯ z
,
∇
¯ z
T
zz
zz
=
¯
∂T
zz
zz
. (16)
(b) As shown in problem 3.2(a), the only nonvanishing components of a traceless symmetric
tensor with lowered indices have all them z or all of them ¯ z. If v is an nindex traceless symmetric
tensor, then P
n
v will be an (n + 1)index traceless symmetric tensor, and will therefore have only
two nonzero components:
(P
n
v)
zz
= ∇
z
v
zz
=
1
2
e
2ω
n
∇
z
v
¯ z¯ z
=
1
2
e
2ω
n
∂v
¯ z¯ z
= (∂ −2n∂ω)v
zz
; (17)
(P
n
v)
¯ z¯ z
= (
¯
∂ −2n
¯
∂ω)v
¯ z¯ z
. (18)
Similarly, if u is an (n+1)index traceless symmetric tensor, then P
T
n
u will be an nindex traceless
symmetric tensor, and will have only two nonzero components:
(P
T
n
u)
zz
= −∇
b
u
b
zz
= −2e
−2ω
∇
z
u
¯ zzz
−2e
−2ω
∇
¯ z
u
zzz
= −
1
2
e
2ω
n−1
∂u
¯ z¯ z
¯ z
−2e
−2ω
¯
∂u
zz
= −2e
−2ω
¯
∂u
zz
; (19)
(P
T
n
u)
¯ z¯ z
= −2e
−2ω
∂u
¯ z¯ z
. (20)
3 CHAPTER 3 24
3.4 Problem 3.4
The FaddeevPopov determinant is deﬁned by,
∆
FP
(φ) ≡
[dζ] δ
F
A
(φ
ζ
)
−1
. (21)
By the gauge invariance of the measure [dζ] on the gauge group, this is a gaugeinvariant function.
It can be used to reexpress the gaugeinvariant formulation of the path integral, with arbitrary
gaugeinvariant insertions f(φ), in a gaugeﬁxed way:
1
V
[dφ] e
−S
1
(φ)
f(φ) =
1
V
[dφ] e
−S
1
(φ)
∆
FP
(φ)
[dζ] δ
F
A
(φ
ζ
)
f(φ)
=
1
V
[dζ dφ
ζ
] e
−S
1
(φ
ζ
)
∆
FP
(φ
ζ
)δ
F
A
(φ
ζ
)
f(φ
ζ
)
=
[dφ] e
−S
1
(φ)
∆
FP
(φ)δ
F
A
(φ)
f(φ). (22)
In the second equality we used the gauge invariance of [dφ] e
−S
1
(φ)
and f(φ), and in the third line
we renamed the variable of integration, φ
ζ
→φ.
In the last line of (22), ∆
FP
is evaluated only for φ on the gauge slice, so it suﬃces to ﬁnd an
expression for it that is valid there. Let
ˆ
φ be on the gauge slice (so F
A
(
ˆ
φ) = 0), parametrize the
gauge group near the identity by coordinates ǫ
B
, and deﬁne
δ
B
F
A
(
ˆ
φ) ≡
∂F
A
(
ˆ
φ
ζ
)
∂ǫ
B
ǫ=0
=
∂F
A
∂φ
i
∂
ˆ
φ
ζ
i
∂ǫ
B
ǫ=0
. (23)
If the F
A
are properly behaved (i.e. if they have nonzero and linearly independent gradients at
ˆ
φ), and if there are no gauge transformations that leave
ˆ
φ ﬁxed, then δ
B
F
A
will be a nonsingular
square matrix. If we choose the coordinates ǫ
B
such that [dζ] = [dǫ
B
] locally, then the Faddeev
Popov determinant is precisely the determinant of δ
B
F
A
, and can be represented as a path integral
over ghost ﬁelds:
∆
FP
(
ˆ
φ) =
[dǫ
B
] δ
F
A
(
ˆ
φ
ζ
)
−1
=
[dǫ
B
] δ
δ
B
F
A
(
ˆ
φ)ǫ
B
−1
= det
δ
B
F
A
(
ˆ
φ)
=
[db
A
dc
B
] e
−b
A
δ
B
F
A
(
ˆ
φ)c
B
. (24)
Finally, we can express the delta function appearing in the gaugeﬁxed path integral (22) as a path
integral itself:
δ
F
A
(φ)
=
[dB
A
]e
iB
A
F
A
(φ)
. (25)
3 CHAPTER 3 25
Putting it all together, we obtain (4.2.3):
[dφdb
A
dc
B
dB
A
] e
−S
1
(φ)−b
A
δ
B
F
A
(φ)c
B
+iB
A
F
A
(φ)
f(φ). (26)
3.5 Problem 3.5
For each ﬁeld conﬁguration φ, there is a unique gaugeequivalent conﬁguration
ˆ
φ
F
in the gauge
slice deﬁned by the F
A
, and a unique gauge transformation ζ
F
(φ) that takes
ˆ
φ
F
to φ:
φ =
ˆ
φ
ζ
F
(φ)
F
. (27)
For φ near
ˆ
φ
F
, ζ
F
(φ) will be near the identity and can be parametrized by ǫ
B
F
(φ), the same
coordinates used in the previous problem. For such φ we have
F
A
(φ) = δ
B
F
A
(
ˆ
φ
F
)ǫ
B
F
(φ), (28)
and we can write the factor ∆
F
FP
(φ)δ(F
A
(φ)) appearing in the gaugeﬁxed path integral (22) in
terms of ǫ
B
F
(φ):
∆
F
FP
(φ)δ
F
A
(φ)
= ∆
F
FP
(
ˆ
φ
F
)δ
F
A
(φ)
= det
δ
B
F
A
(
ˆ
φ
F
)
δ
δ
B
F
A
(
ˆ
φ
F
)ǫ
B
F
(φ)
= δ
ǫ
B
F
(φ)
. (29)
Deﬁning ζ
G
(φ) in the same way, we have,
ζ
G
φ
ζ
−1
G
ζ
F
(φ)
= ζ
F
(φ) . (30)
Deﬁning
φ
′
≡ φ
ζ
−1
G
ζ
F
(φ)
, (31)
it follows from (29) that
∆
F
FP
(φ)δ
F
A
(φ)
= ∆
G
FP
(φ
′
)δ
G
A
(φ
′
)
. (32)
It is now straightforward to prove that the gaugeﬁxed path integral is independent of the choice
of gauge:
[dφ] e
−S(φ)
∆
F
FP
(φ)δ
F
A
(φ)
f(φ) =
[dφ
′
]e
−S(φ
′
)
∆
G
FP
(φ
′
)δ
G
A
(φ
′
)
f(φ
′
)
=
[dφ] e
−S(φ)
∆
G
FP
(φ)δ
G
A
(φ)
f(φ).
(33)
In the ﬁrst line we simultaneously used (32) and the gauge invariance of the measure [dφ]e
−S(φ)
and the insertion f(φ); in the second line we renamed the variable of integration from φ
′
to φ.
3 CHAPTER 3 26
3.6 Problem 3.7
Let us begin by expressing (3.4.19) in momentum space, to know what we’re aiming for. The Ricci
scalar, to lowest order in the metric perturbation h
ab
= g
ab
−δ
ab
, is
R ≈ (∂
a
∂
b
−δ
ab
∂
2
)h
ab
. (34)
In momentum space, the Green’s function deﬁned by (3.4.20) is
˜
G(p) ≈ −
1
p
2
(35)
(again to lowest order in h
ab
), so the exponent of (3.4.19) is
−
a
1
8π
d
2
p
(2π)
2
˜
h
ab
(p)
˜
h
cd
(−p)
p
a
p
b
p
c
p
d
p
2
−2δ
ab
p
c
p
d
+δ
ab
δ
cd
p
2
. (36)
To ﬁrst order in h
ab
, the Polyakov action (3.2.3a) is
S
X
=
1
2
d
2
σ
∂
a
X∂
a
X + (
1
2
hδ
ab
−h
ab
)∂
a
X∂
b
X
, (37)
where h ≡ h
aa
(we have set 2πα
′
to 1). We will use dimensional regularization, which breaks
conformal invariance because the graviton trace couples to X when d = 2. The traceless part of h
ab
in d dimensions is h
′
ab
= h
ab
−h/d. This leaves a coupling between h and ∂
a
X∂
a
X with coeﬃcient
1/2 −1/d. The momentumspace vertex for h
′
ab
is
˜
h
′
ab
(p)k
a
(k
b
+p
b
), (38)
while that for h is
−
d −2
2d
˜
h(p)k (k +p). (39)
There are three oneloop diagrams with two external gravitons, depending on whether the gravitons
are traceless or trace.
We begin by dispensing with the hh diagram. In dimensional regularization, divergences in loop
integrals show up as poles in the d plane. Arising as they do in the form of a gamma function,
these are always simple poles. But the diagram is multiplied by two factors of d − 2 from the two
h vertices, so it vanishes when we take d to 2.
The hh
′
ab
diagram is multiplied by only one factor of d−2, so part of it (the divergent part that
would normally be subtracted oﬀ) might survive. It is equal to
−
d −2
4d
d
d
p
(2π)
d
˜
h
′
ab
(p)
˜
h(−p)
d
d
k
(2π)
d
k
a
(k
b
+p
b
)k (k +p)
k
2
(k +p)
2
. (40)
The k integral can be evaluated by the usual tricks:
1
0
dx
d
d
k
(2π)
d
k
a
(k
b
+p
b
)k (k +p)
(k
2
+ 2xp k +xp
2
)
2
(41)
=
1
0
dx
d
d
q
(2π)
d
(q
a
−xp
a
)(q
b
+ (1 −x)p
b
)(q −xp) (q + (1 −x)p)
(q
2
+x(1 −x)p
2
)
2
.
3 CHAPTER 3 27
Discarding terms that vanish due to the tracelessness of h
′
ab
or that are ﬁnite in the limit d → 2
yields
p
a
p
b
1
0
dx(
1
2
−3x −3x
2
)
d
d
q
(2π)
d
q
2
(q
2
+x(1 −x)p
2
)
2
. (42)
The divergent part of the q integral is independent of x, and the x integral vanishes, so this diagram
vanishes as well.
We are left with just the h
′
ab
h
′
cd
diagram, which (including a symmetry factor of 4 for the
identical vertices and identical propagators) equals
1
4
d
d
p
(2π)
d
˜
h
′
ab
(p)
˜
h
′
cd
(−p)
d
d
k
(2π)
d
k
a
(k
b
+p
b
)k
c
(k
d
+p
d
)
k
2
(k +p)
2
. (43)
The usual tricks, plus the symmetry and tracelessness of h
′
ab
, allow us to write the k integral in the
following way:
1
0
dx
d
d
q
(2π)
d
2
d(d+2)
δ
ac
δ
bd
q
4
+
1
d
(1 −2x)
2
δ
ac
p
b
p
d
q
2
+x
2
(1 −x)
2
p
a
p
b
p
c
p
d
(q
2
+x(1 −x)p
2
)
2
. (44)
The q
4
and q
2
terms in the numerator give rise to divergent integrals. Integrating these terms over
q yields
1
8π
1
0
dxΓ(1 −
d
2
)
x(1 −x)p
2
4π
d/2−1
(45)
¸
−
2
d
x(1 −x)δ
ac
δ
bd
p
2
+ (1 −2x)
2
δ
ac
p
b
p
d
.
The divergent part of this is
δ
ac
δ
bd
p
2
−2δ
ac
p
b
p
d
24π(d −2)
. (46)
However, it is a fact that the symmetric part of the product of two symmetric, traceless, 2 2
matrices is proportional to the identity matrix, so the two terms in the numerator are actually equal
after multiplying by
˜
h
′
ab
(p)
˜
h
′
cd
(−p)—we see that dimensional regularization has already discarded
the divergence for us. The ﬁnite part of (45) is (using this trick a second time)
δ
ac
δ
bd
p
2
8π
1
0
dx
¸
−γ −ln
x(1 −x)p
2
4π
(
1
2
−3x + 3x
2
) −x(1 −x)
. (47)
Amazingly, this also vanishes upon performing the x integral. It remains only to perform the
integral for the last term in the numerator of (44), which is convergent at d = 2:
1
0
dx
d
2
q
(2π)
2
x
2
(1 −x)
2
p
a
p
b
p
c
p
d
(q
2
+x(1 −x)p
2
)
2
=
p
a
p
b
p
c
p
d
4πp
2
1
0
dxx(1 −x) =
p
a
p
b
p
c
p
d
24πp
2
. (48)
Plugging this back into (43), we ﬁnd for the 2graviton contribution to the vacuum amplitude,
1
96π
d
2
p
(2π)
2
˜
h
ab
(p)
˜
h
cd
(−p)
p
a
p
b
p
c
p
d
p
2
−δ
ab
p
c
p
d
+
1
4
δ
ab
δ
cd
p
2
. (49)
3 CHAPTER 3 28
This result does not agree with (36), and is furthermore quite peculiar. It is Weyl invariant (since
the trace h decoupled), but not diﬀ invariant. It therefore appears that, instead of a Weyl anomaly,
we have discovered a gravitational anomaly. However, just because dimensional regularization
has (rather amazingly) thrown away the divergent parts of the loop integrals for us, does not
mean that renormalization becomes unnecessary. We must still choose renormalization conditions,
and introduce counterterms to satisfy them. In this case, we will impose diﬀ invariance, which
is more important than Weyl invariance—without it, it would be impossible to couple this CFT
consistently to gravity. Locality in real space demands that the counterterms be of the same form
as the last two terms in the parentheses in (49). We are therefore free to adjust the coeﬃcients
of these two terms in order to achieve diﬀ invariance. Since (36) is manifestly diﬀ invariant, it
is clearly the desired expression, with a
1
taking the value −1/12. (It is worth pointing out that
there is no local counterterm quadratic in h
ab
that one could add that is diﬀ invariant by itself,
and that would therefore have to be ﬁxed by some additional renormalization condition. This is
because diﬀinvariant quantities are constructed out of the Ricci scalar, and
d
2
σR
2
has the wrong
dimension.)
3.7 Problem 3.9
Fix coordinates such that the boundary lies at σ
2
= 0. Following the prescription of problem 2.10
for normal ordering operators in the presence of a boundary, we include in the contraction the
image term:
∆
b
(σ, σ
′
) = ∆(σ, σ
′
) + ∆(σ, σ
′∗
), (50)
where σ
∗
1
= σ
1
, σ
∗
2
= −σ
2
. If σ and σ
′
both lie on the boundary, then the contraction is eﬀectively
doubled:
∆
b
(σ
1
, σ
′
1
) = 2∆
(σ
1
, σ
2
= 0), (σ
′
1
, σ
′
2
= 0)
. (51)
If T is a boundary operator, then the σ
2
and σ
′
2
integrations in the deﬁnition (3.6.5) of [T]
r
can be
done trivially:
[T]
r
= exp
1
2
dσ
1
dσ
′
1
∆
b
(σ
1
, σ
′
1
)
δ
δX
ν
(σ
1
, σ
2
= 0)
δ
δX
ν
(σ
′
1
, σ
′
2
= 0)
T. (52)
Equation (3.6.7) becomes
δ
W
[T]
r
= [δ
W
T]
r
+
1
2
dσ
1
dσ
′
1
δ
W
∆
b
(σ
1
, σ
′
1
)
δ
δX
ν
(σ
1
)
δ
δX
ν
(σ
′
1
)
[T]
r
. (53)
The tachyon vertex operator (3.6.25) is
V
0
= g
o
σ
2
=0
dσ
1
g
1/2
11
(σ
1
)[e
ikX(σ
1
)
]
r
, (54)
3 CHAPTER 3 29
and its Weyl variation (53) is
δ
W
V
0
= g
o
dσ
1
g
1/2
11
(σ
1
) (δω(σ
1
) +δ
W
) [e
ikX(σ
1
)
]
r
= g
o
dσ
1
g
1/2
11
(σ
1
)
δω(σ
1
) −
k
2
2
δ
W
∆
b
(σ
1
, σ
1
)
[e
ikX(σ
1
)
]
r
= (1 −α
′
k
2
)g
o
dσ
1
g
1/2
11
(σ
1
)δω(σ
1
)[e
ikX(σ
1
)
]
r
, (55)
where we have used (3.6.11) in the last equality:
δ
W
∆
b
(σ
1
, σ
1
) = 2δ
W
∆(σ
1
, σ
′
1
) = 2α
′
δω(σ
1
). (56)
Weyl invariance thus requires
k
2
=
1
α
′
. (57)
The photon vertex operator (3.6.26) is
V
1
= −i
g
o
√
2α
′
e
µ
σ
2
=0
dσ
1
[∂
1
X
µ
(σ
1
)e
ikX(σ
1
)
]
r
. (58)
The spacetime gauge equivalence,
V
1
(k, e) = V
1
(k, e +λk), (59)
is clear from the fact that k
µ
∂
1
X
µ
e
ikX
is a total derivative. The expression (58) has no explicit
metric dependence, so the variation of V
1
comes entirely from the variation of the renormalization
contraction:
δ
W
[∂
1
X
µ
(σ
1
)e
ikX(σ
1
)
]
r
=
1
2
dσ
′
1
dσ
′′
1
δ
W
∆
b
(σ
′
1
, σ
′′
1
)
δ
δX
ν
(σ
′
1
)
δ
δX
ν
(σ
′′
1
)
[∂
1
X
µ
(σ
1
)e
ikX(σ
1
)
]
r
= ik
µ
∂
1
δ
W
∆
b
(σ
1
, σ
′′
1
)
σ
′′
1
=σ
1
[e
ikX(σ
1
)
]
r
−
k
2
2
δ
W
∆
b
(σ
1
, σ
1
)[∂
1
X
µ
(σ
1
)e
ikX(σ
1
)
]
r
= iα
′
k
µ
∂
1
δω(σ
1
)[e
ikX(σ
1
)
]
r
−α
′
k
2
δω(σ
1
)[∂
1
X
µ
(σ
1
)e
ikX(σ
1
)
]
r
, (60)
where in the last equality we have used (56) and (3.6.15a):
∂
1
δ
W
∆
b
(σ
1
, σ
′
1
)
σ
′
1
=σ
1
= 2 ∂
1
δ
W
∆(σ
1
, σ
′
1
)
σ
′
1
=σ
1
= α
′
∂
1
δω(σ
1
). (61)
Integration by parts yields
δ
W
V
1
= −i
α
′
2
g
o
(e kk
µ
−k
2
e
µ
)
dσ
1
δω(σ
1
)[∂
1
X
µ
(σ
1
)e
ikX(σ
1
)
]
r
. (62)
For this quantity to vanish for arbitrary δω(σ
1
) requires the vector e kk −k
2
e to vanish. This will
happen if e and k are collinear, but by (59) V
1
vanishes in this case. The other possibility is
k
2
= 0, e k = 0. (63)
3 CHAPTER 3 30
3.8 Problem 3.11
Since we are interested in the H
2
term, let us assume G
µν
to be constant, Φ to vanish, and B
µν
to
be linear in X, implying that
H
ωµν
= 3∂
[ω
B
µν]
(64)
is constant. With these simpliﬁcations, the sigma model action becomes
S
σ
=
1
4πα
′
d
2
σ g
1/2
G
µν
g
ab
∂
a
X
µ
∂
b
X
ν
+i∂
ω
B
µν
ǫ
ab
X
ω
∂
a
X
µ
∂
b
X
ν
=
1
4πα
′
d
2
σ g
1/2
G
µν
g
ab
∂
a
X
µ
∂
b
X
ν
+
i
3
H
ωµν
ǫ
ab
X
ω
∂
a
X
µ
∂
b
X
ν
.
(65)
In the second line we have used the fact that ǫ
ab
X
ω
∂
a
X
µ
∂
b
X
ν
is totally antisymmetric in ω, µ, ν
(up to integration by parts) to antisymmetrize ∂
ω
B
µν
.
Working in conformal gauge on the worldsheet and transforming to complex coordinates,
g
1/2
g
ab
∂
a
X
µ
∂
b
X
ν
= 4∂X
(µ
¯
∂X
ν)
, (66)
g
1/2
ǫ
ab
∂
a
X
µ
∂
b
X
ν
= −4i∂X
[µ
¯
∂X
ν]
, (67)
d
2
σ =
1
2
d
2
z, (68)
the action becomes
S
σ
= S
f
+S
i
, (69)
S
f
=
1
2πα
′
G
µν
d
2
z ∂X
µ
¯
∂X
ν
, (70)
S
i
=
1
6πα
′
H
ωµν
d
2
z X
ω
∂X
µ
¯
∂X
ν
, (71)
where we have split it into the action for a free CFT and an interaction term. The path integral is
now
'. . . `
σ
= 'e
−S
i
. . . `
f
= '. . . `
f
−'S
i
. . . `
f
+
1
2
'S
2
i
. . . `
f
+ , (72)
where ' `
f
is the path integral calculated using only the free action (70). The Weyl variation of the
ﬁrst term gives rise to the D−26 Weyl anomaly calculated in section 3.4, while that of the second
gives rise to the term in β
B
µν
that is linear in H (3.7.13b). It is the Weyl variation of the third term,
quadratic in H, that we are interested in, and in particular the part proportional to
d
2
z ': ∂X
µ
¯
∂X
ν
: . . . `
f
, (73)
3 CHAPTER 3 31
whose coeﬃcient gives the H
2
term in β
G
µν
. This third term is
1
2
'S
2
i
. . . `
f
=
1
2(6πα
′
)
2
H
ωµν
H
ω
′
µ
′
ν
′ (74)
d
2
zd
2
z
′
': X
ω
(z, ¯ z)∂X
µ
(z)
¯
∂X
ν
(¯ z) :: X
ω
′
(z
′
, ¯ z
′
)∂
′
X
µ
′
(z
′
)
¯
∂
′
X
ν
′
(¯ z
′
) : . . . `
f
,
where we have normalordered the interaction vertices. The Weyl variation of this integral will come
from the singular part of the OPE when z and z
′
approach each other. Terms in the OPE containing
exactly two X ﬁelds (which will yield (73) after the z
′
integration is performed) are obtained by
performing two crosscontractions. There are 18 diﬀerent pairs of crosscontractions one can apply
to the integrand of (74), but, since they can all be obtained from each other by integration by parts
and permuting the indices ω, µ, ν, they all give the same result. The contraction derived from the
free action (70) is
X
µ
(z, ¯ z)X
ν
(z
′
, ¯ z
′
) =: X
µ
(z, ¯ z)X
ν
(z
′
, ¯ z
′
) : −
α
′
2
G
µν
ln[z −z
′
[
2
, (75)
so, picking a representative pair of crosscontractions, the part of (74) we are interested in is
18
2(6πα
′
)
2
H
ωµν
H
ω
′
µ
′
ν
′
d
2
zd
2
z
′
−
α
′
2
G
ωµ
′
∂
′
ln [z −z
′
[
2
−
α
′
2
G
νω
′
¯
∂ ln [z −z
′
[
2
': ∂X
µ
(z)
¯
∂
′
X
ν
′
(¯ z
′
) : . . . `
f
= −
1
16π
2
H
µλω
H
ν
λω
d
2
zd
2
z
′
1
[z
′
−z[
2
': ∂X
µ
(z)
¯
∂
′
X
ν
(¯ z
′
) : . . . `
f
. (76)
The Weyl variation of this term comes from cutting oﬀ the logarithmically divergent integral of
[z
′
−z[
−2
near z
′
= z, so we can drop the less singular terms coming from the Taylor expansion of
¯
∂
′
X
ν
(¯ z
′
):
−
1
16π
2
H
µλω
H
ν
λω
d
2
z ': ∂X
µ
(z)
¯
∂X
ν
(¯ z) : . . . `
f
d
2
z
′
1
[z
′
−z[
2
. (77)
The diﬀinvariant distance between z
′
and z is (for short distances) e
ω(z)
[z
′
−z[, so a diﬀinvariant
cutoﬀ would be at [z
′
− z[ = ǫe
−ω(z)
. The Weyldependent part of the second integral of (77) is
then
d
2
z
′
1
[z
′
−z[
2
∼ −2π ln(ǫe
−ω(z)
) = −2π ln ǫ + 2πω(z), (78)
and the Weyl variation of (76) is
−
1
8π
H
µλω
H
ν
λω
d
2
z δω(z)': ∂X
µ
(z)
¯
∂X
ν
(¯ z) : . . . `
f
. (79)
Using (66) and (68), and the fact that the diﬀerence between ' `
σ
and ' `
f
involves higher powers
of H (see (72)) which we can neglect, we can write this as
−
1
16π
H
µλω
H
ν
λω
d
2
σ g
1/2
δωg
ab
': ∂
a
X
µ
∂
b
X
ν
: . . . `
σ
. (80)
3 CHAPTER 3 32
This is of the form of (3.4.6), with
T
′a
a
=
1
8
H
µλω
H
ν
λω
g
ab
∂
a
X
µ
∂
b
X
ν
(81)
being the contribution of this term to the stress tensor. According to (3.7.12), T
′a
a
in turn con
tributes the following term to β
G
µν
:
−
α
′
4
H
µλω
H
ν
λω
. (82)
3.9 Problem 3.13
If the dilaton Φ is constant and D = d+3, then the equations of motion (3.7.15) become, to leading
order in α
′
,
R
µν
−
1
4
H
µλω
H
ν
λω
= 0, (83)
∇
ω
H
ωµν
= 0, (84)
d −23
α
′
−
1
4
H
µνλ
H
µνλ
= 0. (85)
Letting i, j, k be indices on the 3sphere and α, β, γ be indices on the ﬂat ddimensional spacetime,
we apply the ansatz
H
ijk
= hǫ
ijk
, (86)
where h is a constant and ǫ is the volume form on the sphere, with all other components vanishing.
(Note that this form for H cannot be obtained as the exterior derivative of a nonsingular gauge
ﬁeld B; B must have a Diractype singularity somewhere on the sphere.) Equation (84) is then
immediately satisﬁed, because the volume form is always covariantly constant on a manifold, so
∇
i
H
ijk
= 0, and all other components vanish trivially. Since ǫ
ijk
ǫ
ijk
= 6, equation (85) ﬁxes h in
terms of d:
h
2
=
2(d −23)
3α
′
, (87)
implying that there are solutions only for d > 23. The Ricci tensor on a 3sphere of radius r is
given by
R
ij
=
2
r
2
G
ij
. (88)
Similarly,
ǫ
ikl
ǫ
j
kl
= 2G
ij
. (89)
Most components of equation (83) vanish trivially, but those for which both indices are on the
sphere ﬁx r in terms of h:
r
2
=
4
h
2
=
6α
′
d −23
. (90)
4 CHAPTER 4 33
4 Chapter 4
4.1 Problem 4.1
To begin, let us recall the spectrum of the open string at level N = 2 in lightcone quantization.
In representations of SO(D −2), we had a symmetric rank 2 tensor,
f
ij
α
i
−1
α
j
−1
[0; k`, (1)
and a vector,
e
i
α
i
−2
[0; k`. (2)
Together, they make up the traceless symmetric rank 2 tensor representation of SO(D−1), whose
dimension is D(D −1)/2 −1. This is what we expect to ﬁnd.
In the OCQ, the general state at level 2 is
[f, e; k` =
f
µν
α
µ
−1
α
ν
−1
+e
µ
α
µ
−2
[0; k`, (3)
a total of D(D + 1)/2 +D states. Its norm is
'e, f; k[e, f; k
′
` = '0; k[
f
∗
ρσ
α
ρ
1
α
σ
1
+e
∗
ρ
α
ρ
2
f
µν
α
µ
−1
α
ν
−1
+e
µ
α
µ
−2
[0; k
′
`
= 2
f
∗
µν
f
µν
+e
∗
µ
e
µ
'0; k[0; k
′
`. (4)
The terms in the mode expansion of the Virasoro generator relevant here are as follows:
L
0
= α
′
p
2
+α
−1
α
1
+α
−2
α
2
+ (5)
L
1
=
√
2α
′
p α
1
+α
−1
α
2
+ (6)
L
2
=
√
2α
′
p α
2
+
1
2
α
1
α
1
+ (7)
L
−1
=
√
2α
′
p α
−1
+α
−2
α
1
+ (8)
L
−2
=
√
2α
′
p α
−2
+
1
2
α
−1
α
−1
+ . (9)
As in the cases of the tachyon and photon, the L
0
condition yields the massshell condition:
0 = (L
0
−1)[f, e; k`
= (α
′
k
2
+ 1)[f, e; k`, (10)
or m
2
= 1/α
′
, the same as in the lightcone quantization. Since the particle is massive, we can
go to its rest frame for simplicity: k
0
= 1/
√
α
′
, k
i
= 0. The L
1
condition ﬁxes e in terms of f,
removing D degrees of freedom:
0 = L
1
[f, e; k`
= 2
√
2α
′
f
µν
k
ν
+e
µ
α
µ
−1
[0; k`, (11)
4 CHAPTER 4 34
implying
e
µ
=
√
2f
0µ
. (12)
The L
2
condition adds one more constraint:
0 = L
2
[f, e; k`
=
2
√
2α
′
k
µ
e
µ
+f
µ
µ
[0; k`. (13)
Using (12), this implies
f
ii
= 5f
00
, (14)
where f
ii
is the trace on the spacelike part of f.
There are D + 1 independent spurious states at this level:
[g, γ; k` =
L
−1
g
µ
α
µ
−1
+L
−2
γ
[0; k` (15)
=
√
2α
′
g
(µ
k
ν)
+
γ
2
η
µν
α
µ
−1
α
ν
−1
[0; k` +
g
µ
+
√
2α
′
γk
µ
α
µ
−2
[0; k`.
These states are physical and therefore null for g
0
= γ = 0. Removing these D−1 states from the
spectrum leaves D(D−1)/2 states, the extra one with respect to the lightcone quantization being
the SO(D−1) scalar,
f
ij
= fδ
ij
, f
00
=
D −1
5
f, e
0
=
√
2(D −1)
5
f, (16)
with all other components zero. (States with vanishing f
00
must be traceless by (14), and this is
the unique state satisfying (12) and (14) that is orthogonal to all of these.) The norm of this state
is proportional to
f
∗
µν
f
µν
+e
∗
µ
e
µ
=
(D −1)(26 −D)f
2
25
, (17)
positive for D < 26 and negative for D > 26. In the case D = 26, this state is spurious, corre
sponding to (15) with γ = 2f, g
0
= 3
√
2f. Removing it from the spectrum leaves us with the
states f
ij
, f
ii
= 0, e = 0—precisely the traceless symmetric rank 2 tensor of SO(25) we found in
the lightcone quantization.
5 CHAPTER 5 35
5 Chapter 5
5.1 Problem 5.1
(a) Our starting point is the following formal expression for the path integral:
Z(X
0
, X
1
) =
X(0)=X
0
X(1)=X
1
[dXde]
V
diﬀ
exp (−S
m
[X, e]) , (1)
where the action for the “matter” ﬁelds X
µ
is
S
m
[X, e] =
1
2
1
0
dτ e
e
−1
∂X
µ
e
−1
∂X
µ
+m
2
(2)
(where ∂ ≡ d/dτ). We have ﬁxed the coordinate range for τ to be [0,1]. Coordinate diﬀeomorphisms
ζ : [0, 1] →[0, 1], under which the X
µ
are scalars,
X
µζ
(τ
ζ
) = X
µ
(τ), (3)
and the einbein e is a “covector,”
e
ζ
(τ
ζ
) = e(τ)
dτ
dτ
ζ
, (4)
leave the action (2) invariant. V
diﬀ
is the volume of this group of diﬀeomorphisms. The e integral
in (1) runs over positive functions on [0,1], and the integral
l ≡
1
0
dτ e (5)
is diﬀeomorphism invariant and therefore a modulus; the moduli space is (0, ∞).
In order to make sense of the functional integrals in (1) we will need to deﬁne an inner product
on the space of functions on [0,1], which will induce measures on the relevant function spaces. This
inner product will depend on the einbein e in a way that is uniquely determined by the following
two constraints: (1) the inner product must be diﬀeomorphism invariant; (2) it must depend on
e(τ) only locally, in other words, it must be of the form
(f, g)
e
=
1
0
dτ h(e(τ))f(τ)g(τ), (6)
for some function h. As we will see, these conditions will be necessary to allow us to regularize
the inﬁnite products that will arise in carrying out the functional integrals in (1), and then to
renormalize them by introducing a counterterm action, in a way that respects the symmetries of
the action (2). For f and g scalars, the inner product satisfying these two conditions is
(f, g)
e
≡
1
0
dτ efg. (7)
5 CHAPTER 5 36
We can express the matter action (2) using this inner product:
S
m
[X, e] =
1
2
(e
−1
∂X
µ
, e
−1
∂X
µ
)
e
+
lm
2
2
. (8)
We now wish to express the path integral (1) in a slightly less formal way by choosing a ﬁducial
einbein e
l
for each point l in the moduli space, and replacing the integral over einbeins by an
integral over the moduli space times a FaddeevPopov determinant ∆
FP
[e
l
]. Deﬁning ∆
FP
by
1 = ∆
FP
[e]
∞
0
dl
[dζ] δ[e −e
ζ
l
], (9)
we indeed have, by the usual sequence of formal manipulations,
Z(X
0
, X
1
) =
∞
0
dl
X(0)=X
0
X(1)=X
1
[dX] ∆
FP
[e
l
] exp (−S
m
[X, e
l
]) . (10)
To calculate the FaddeevPopov determinant (9) at the point e = e
l
, we expand e about e
l
for small
diﬀeomorphisms ζ and small changes in the modulus:
e
l
−e
ζ
l+δl
= ∂γ −
de
l
dl
δl, (11)
where γ is a scalar function parametrizing small diﬀeomorphisms: τ
ζ
= τ + e
−1
γ; to respect the
ﬁxed coordinate range, γ must vanish at 0 and 1. Since the change (11) is, like e, a covector, we
will for simplicity multiply it by e
−1
l
in order to have a scalar, and then bring into play our inner
product (7) in order to express the delta functional in (9) as an integral over scalar functions β:
∆
−1
FP
[e
l
] =
dδl[dγdβ] exp
2πi(β, e
−1
l
∂γ −e
−1
l
de
l
dl
δl)
e
l
(12)
The integral is inverted by replacing the bosonic variables δl, γ, and β by Grassman variables ξ, c,
and b:
∆
FP
[e
l
] =
dξ[dcdb] exp
1
4π
(b, e
−1
l
∂c −e
−1
l
de
l
dl
ξ)
e
l
=
[dcdb]
1
4π
(b, e
−1
l
de
l
dl
)
e
l
exp
1
4π
(b, e
−1
l
∂c)
e
l
. (13)
We can now write the path integral (10) in a more explicit form:
Z(X
0
, X
1
) =
∞
0
dl
X(0)=X
0
X(1)=X
1
[dX]
c(0)=c(1)=0
[dcdb]
1
4π
(b, e
−1
l
de
l
dl
)
e
l
(14)
exp (−S
g
[b, c, e
l
] −S
m
[X, e
l
]) ,
where
S
g
[b, c, e
l
] = −
1
4π
(b, e
−1
l
∂c)
e
l
. (15)
5 CHAPTER 5 37
(b) At this point it becomes convenient to work in a speciﬁc gauge, the simplest being
e
l
(τ) = l. (16)
Then the inner product (7) becomes simply
(f, g)
l
= l
1
0
dτ fg. (17)
In order to evaluate the FaddeevPopov determinant (13), let us decompose b and c into nor
malized eigenfunctions of the operator
∆ = −(e
−1
l
∂)
2
= −l
−2
∂
2
: (18)
b(τ) =
b
0
√
l
+
2
l
∞
¸
j=1
b
j
cos(πjτ), (19)
c(τ) =
2
l
∞
¸
j=1
c
j
sin(πjτ), (20)
with eigenvalues
ν
j
=
π
2
j
2
l
2
. (21)
The ghost action (15) becomes
S
g
(b
j
, c
j
, l) = −
1
4l
∞
¸
j=1
jb
j
c
j
. (22)
The zero mode b
0
does not enter into the action, but it is singled out by the insertion appearing in
front of the exponential in (13):
1
4π
(b, e
−1
l
de
l
dl
)
e
l
=
b
0
4π
√
l
. (23)
The FaddeevPopov determinant is, ﬁnally,
∆
FP
(l) =
∞
¸
j=0
db
j
∞
¸
j=1
dc
j
b
0
4π
√
l
exp
¸
1
4l
∞
¸
j=1
jb
j
c
j
¸
=
1
4π
√
l
∞
¸
j=1
j
4l
=
1
4π
√
l
det
′
∆
16π
2
1/2
, (24)
the prime on the determinant denoting omission of the zero eigenvalue.
5 CHAPTER 5 38
(c) Let us decompose X
µ
(τ) into a part which obeys the classical equations of motion,
X
µ
cl
(τ) = X
0
+ (X
1
−X
0
)τ, (25)
plus quantum ﬂuctuations; the ﬂuctuations vanish at 0 and 1, and can therefore be decomposed
into the same normalized eigenfunctions of ∆ as c was (20):
X
µ
(τ) = X
µ
cl
(τ) +
2
l
∞
¸
j=1
x
µ
j
sin(πjτ). (26)
The matter action (8) becomes
S
m
(X
0
, X
1
, x
j
) =
(X
1
−X
0
)
2
2l
+
π
2
l
2
∞
¸
j=1
j
2
x
2
j
+
lm
2
2
, (27)
and the matter part of the path integral (10)
X(0)=X
0
X(1)=X
1
[dX] exp (−S
m
[X, e
l
])
= exp
−
(X
1
−X
0
)
2
2l
−
lm
2
2
D
¸
µ=1
∞
¸
j=1
dx
µ
j
exp
¸
−
π
2
l
2
∞
¸
j=1
j
2
x
2
j
¸
= exp
−
(X
1
−X
0
)
2
2l
−
lm
2
2
det
′
∆
π
−D/2
, (28)
where we have conveniently chosen to work in a Euclidean spacetime in order to make all of the
Gaussian integrals convergent.
(d) Putting together the results (10), (24), and (28), and dropping the irrelevant constant factors
multiplying the operator ∆ in the inﬁnitedimensional determinants, we have:
Z(X
0
, X
1
) =
∞
0
dl
1
4π
√
l
exp
−
(X
1
−X
0
)
2
2l
−
lm
2
2
det
′
∆
(1−D)/2
. (29)
We will regularize the determinant of ∆ in the same way as it is done in Appendix A.1, by dividing
by the determinant of the operator ∆ + Ω
2
:
det
′
∆
det
′
(∆ + Ω
2
)
=
∞
¸
j=1
π
2
j
2
π
2
j
2
+ Ω
2
l
2
=
Ωl
sinhΩl
∼ 2Ωl exp (−Ωl) , (30)
5 CHAPTER 5 39
where the last line is the asymptotic expansion for large Ω. The path integral (29) becomes
Z(X
0
, X
1
) (31)
=
1
4π(2Ω)
(D−1)/2
∞
0
dl l
−D/2
exp
−
(X
1
−X
0
)
2
2l
−
l(m
2
−(D −1)Ω)
2
.
The inverse divergence due to the factor of Ω
(1−D)/2
in front of the integral can be dealt with by a
ﬁeld renormalization, but since we will not concern ourselves with the overall normalization of the
path integral we will simply drop all of the factors that appear in front. The divergence coming
from the Ω term in the exponent can be cancelled by a (diﬀeomorphism invariant) counterterm in
the action,
S
ct
=
1
0
dτ eA = lA (32)
The mass m is renormalized by what is left over after the cancellation of inﬁnities,
m
2
phys
= m
2
−(D −1)Ω −2A, (33)
but for simplicity we will assume that a renormalization condition has been chosen that sets m
phys
=
m.
We can now proceed to the integration over moduli space:
Z(X
0
, X
1
) =
∞
0
dl l
−D/2
exp
−
(X
1
−X
0
)
2
2l
−
lm
2
2
. (34)
The integral is most easily done after passing to momentum space:
˜
Z(k) ≡
d
D
X exp (ik X) Z(0, X)
=
∞
0
dl l
−D/2
exp
−
lm
2
2
d
D
X exp
ik X −
X
2
2l
=
π
2
D/2
∞
0
dl exp
−
l(k
2
+m
2
)
2
=
π
2
D/2
2
k
2
+m
2
; (35)
neglecting the constant factors, this is precisely the momentum space scalar propagator.
6 CHAPTER 6 40
6 Chapter 6
6.1 Problem 6.1
In terms of u = 1/z, (6.2.31) is
δ
d
(
¸
k
i
)
¸
i<j
1
u
i
−
1
u
j
α
′
k
i
k
j
= δ
d
(
¸
k
i
)
¸
i<j
[u
ij
[
α
′
k
i
k
j
[u
i
u
j
[
−α
′
k
i
k
j
= δ
d
(
¸
k
i
)
¸
i<j
[u
ij
[
α
′
k
i
k
j
¸
i
[u
i
[
α
′
k
2
i
. (1)
Since this is an expectation value of closedstring tachyon vertex operators, α
′
k
2
i
= 4 and the
expectation value is smooth at u
i
= 0.
6.2 Problem 6.3
For any n ≥ 2 numbers z
i
, we have
n
¸
i=1
(−1)
i
¸
j<k
j,k=i
z
jk
=
1 1 z
1
z
2
1
z
n−2
1
1 1 z
2
z
2
2
z
n−2
2
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1 1 z
n
z
2
n
z
n−2
n
= 0. (2)
The minor of the matrix with respect to the ﬁrst entry in the ith row is a Vandermonde matrix for
the other z
j
, so its determinant provides the ith term in the sum. Specializing to the case n = 5,
relabeling z
5
by z
′
1
, and dividing by z
11
′ z
21
′ z
31
′ z
41
′ yields
−
z
23
z
24
z
34
z
11
′
+
z
13
z
14
z
34
z
21
′
−
z
12
z
14
z
24
z
31
′
+
z
12
z
13
z
23
z
41
′
=
z
12
z
13
z
14
z
23
z
24
z
34
z
11
′ z
21
′ z
31
′ z
41
′
, (3)
which is what we are required to prove.
6.3 Problem 6.5
(a) We have
I(s, t) =
Γ(−1 −α
′
s)Γ(−1 −α
′
t)
Γ(−2 −α
′
s −α
′
t)
, (4)
so the pole at α
′
s = J − 1 arises from the ﬁrst gamma function in the numerator. The residue of
Γ(z) at z a nonpositive integer is (−1)
z
/Γ(1 −z), so the residue of I(s, t) is
(−1)
J
Γ(−1 −α
′
t)
Γ(J + 1)Γ(−1 −J −α
′
t)
=
1
J!
(2 +α
′
t)(3 +α
′
t) (J + 1 +α
′
t), (5)
a polynomial of degree J in t. Using
s +t +u = −
4
α
′
, (6)
6 CHAPTER 6 41
once s is ﬁxed at (J −1)/α
′
, t can be expressed in terms of t −u:
t =
t −u
2
−
J + 3
2α
′
, (7)
so (5) is also a polynomial of degree J in t −u.
(b) The momentum of the intermediate state in the s channel is k
1
+ k
2
= −(k
3
+ k
4
), so in its
rest frame we have
k
i
1
= −k
i
2
, k
i
3
= −k
i
4
, k
0
1
= k
0
2
= −k
0
3
= −k
0
4
=
√
s
2
. (8)
Specializing to the case where all the external particles are tachyons (k
2
= 1/α
′
) and the interme
diate state is at level 2 (s = 1/α
′
), we further have
k
i
1
k
i
1
= k
i
2
k
i
2
= k
i
3
k
i
3
= k
i
4
k
i
4
=
5
4α
′
. (9)
It also determines t in terms of k
i
1
k
i
3
:
t = −(k
1
+k
3
)
2
= −
5
2α
′
−2k
i
1
k
i
3
. (10)
Using (5), the residue of the pole in I(s, t) at α
′
s = 1 is
1
2
(2 +α
′
t)(3 +α
′
t) = −
1
8
+ 2α
′2
(k
i
1
k
i
3
)
2
. (11)
The operator that projects matrices onto multiples of the identity matrix in D−1 dimensional
space is
P
0
ij,kl
= δ
ij
1
D −1
δ
kl
, (12)
while the one that projects them onto traceless symmetric matrices is
P
2
ij,kl
=
1
2
(δ
ik
δ
jl
+δ
il
δ
jk
) −P
0
ij,kl
. (13)
Inserting the linear combination β
0
P
0
+β
2
P
2
between the matrices k
i
1
k
j
1
and k
k
3
k
l
3
yields
(β
0
−β
2
)
k
i
1
k
i
1
k
j
3
k
j
3
D −1
+β
2
(k
i
1
k
i
3
)
2
= (β
0
−β
2
)
25
16α
′2
(D −1)
+β
2
(k
i
1
k
i
3
)
2
. (14)
Comparison with (11) reveals
β
0
=
2α
′2
(26 −D)
25
, β
2
= 2α
′
, (15)
so that, as promised, β
0
is positive, zero, or negative depending on whether D is less than, equal
to, or greater than 26.
6 CHAPTER 6 42
What does all this have to do with the open string spectrum at level 2? The amplitude I has
a pole in s wherever s equals the masssquared of an open string state, allowing the intermediate
state in the s channel to go on shell. The residue of this pole can be written schematically as
'f[S
¸
o
[o`'o[
'o[o`
S[i`, (16)
where the sum is taken over open string states at level J = α
′
s + 1 with momentum equal to that
of the initial and ﬁnal states; we have not assumed that the intermediate states are normalized, to
allow for the possibility that some of them might have negative norm. More speciﬁcally,
[i` = [0; k
1
`[0; k
2
`, 'f[ = '0; −k
3
['0; −k
4
[. (17)
The open string spectrum at level 2 was worked out as a function of D in problem 4.1. For any D
it includes D(D − 1)/2 − 1 positivenorm states transforming in the spin 2 representation of the
little group SO(D−1). Working in the rest frame of such a state, the Smatrix elements involved
in (16) are ﬁxed by SO(D−1) invariance and (anti)linearity in the polarization matrix a:
'a[S[0; k
1
`[0; k
2
` ∝ a
∗
ij
k
i
1
k
j
2
, '0; −k
3
['0; −k
4
[S[a` ∝ a
kl
k
k
3
k
l
4
. (18)
Summing over an orthonormal basis in the space of symmetric traceless matrices yields the contri
bution of these states to (16):
¸
a
a
∗
ij
k
i
1
k
j
2
a
kl
k
k
3
k
l
4
= k
i
1
k
j
1
P
2
ij,kl
k
k
3
k
l
3
. (19)
This explains the positive value of β
2
found in (15).
For D = 26, there is another state [b` in the spectrum whose norm is positive for D < 26 and
negative for D > 26. Since this state is an SO(D − 1) scalar, the Smatrix elements connecting it
to the initial and ﬁnal states are given by:
'b[S[0; k
1
`[0; k
2
` ∝ δ
ij
k
i
1
k
j
2
, '0; −k
3
['0; −k
4
[S[b` ∝ δ
kl
k
k
3
k
l
4
. (20)
Its contribution to (16) is therefore clearly a positive multiple of
k
i
1
k
j
1
P
0
ij,kl
k
k
3
k
l
3
(21)
if D < 26, and a negative multiple if D > 26.
6.4 Problem 6.7
(a) The X path integral (6.5.11) follows immediately from (6.2.36), where
v
µ
(y
1
) = −2iα
′
k
µ
2
y
12
−
k
µ
3
y
13
(22)
6 CHAPTER 6 43
(we leave out the contraction between
˙
X
µ
(y
1
) and e
ik
1
X(y
1
)
because their product is already renor
malized in the path integral). Momentum conservation, k
1
+ k
2
+ k
3
= 0, and the mass shell
conditions imply
0 = k
2
1
= (k
2
+k
3
)
2
=
2
α
′
+ 2k
2
k
3
, (23)
1
α
′
= k
2
3
= (k
1
+k
2
)
2
=
1
α
′
+ 2k
1
k
2
, (24)
1
α
′
= k
2
2
= (k
1
+k
3
)
2
=
1
α
′
+ 2k
1
k
3
, (25)
so that 2α
′
k
2
k
3
= −2, while 2α
′
k
1
k
2
= 2α
′
k
1
k
3
= 0. Equation (6.5.11) therefore simpliﬁes to
⋆
⋆
˙
X
µ
e
ik
1
X
(y
1
)
⋆
⋆
⋆
⋆e
ik
2
X
(y
2
)
⋆
⋆
⋆
⋆e
ik
3
X
(y
3
)
⋆
⋆
D
2
(26)
= 2α
′
C
X
D
2
(2π)
26
δ
26
(k
1
+k
2
+k
3
)
1
y
2
23
k
µ
2
y
12
+
k
µ
3
y
13
.
The ghost path integral is given by (6.3.2):
'c(y
1
)c(y
2
)c(y
3
)`
D
2
= C
g
D
2
y
12
y
13
y
23
. (27)
Putting these together with (6.5.10) and using (6.4.14),
α
′
g
2
o
e
−λ
C
X
D
2
C
g
D
2
= 1, (28)
yields
S
D
2
(k
1
, a
1
, e
1
; k
2
, a
2
; k
3
, a
3
) (29)
= −2ig
′
o
(2π)
26
δ
26
(k
1
+k
2
+k
3
)
y
13
e k
2
+y
12
e k
3
y
23
Tr(λ
a
1
λ
a
2
λ
a
3
) + (2 ↔3).
Momentum conservation and the physical state condition e
1
k
1
= 0 imply
e
1
k
2
= −e
1
k
3
=
1
2
e
1
k
23
, (30)
so
S
D
2
(k
1
, a
1
, e
1
; k
2
, a
2
; k
3
, a
3
) (31)
= −ig
′
0
(2π)
26
δ
26
(k
1
+k
2
+k
3
)e
1
k
23
Tr(λ
a
1
[λ
a
2
, λ
a
3
]),
in agreement with (6.5.12).
6 CHAPTER 6 44
(b) Using equations (6.4.17), (6.4.20), and (6.5.6), we see that the fourtachyon amplitude near
s = 0 is given by
S
D
2
(k
1
, a
1
; k
2
, a
2
; k
3
, a
3
; k
4
, a
4
)
=
ig
2
o
α
′
(2π)
26
δ
26
(
¸
i
k
i
)
Tr(λ
a
1
λ
a
2
λ
a
4
λ
a
3
+λ
a
1
λ
a
3
λ
a
4
λ
a
2
−λ
a
1
λ
a
2
λ
a
3
λ
a
4
−λ
a
1
λ
a
4
λ
a
3
λ
a
2
)
u −t
2s
= −
ig
2
o
2α
′
(2π)
26
δ
26
(
¸
i
k
i
)Tr([λ
a
1
, λ
a
2
][λ
a
3
, λ
a
4
])
u −t
s
. (32)
We can calculate the same quantity using unitarity. By analogy with equation (6.4.13), the
4tachyon amplitude near the pole at s = 0 has the form
S
D
2
(k
1
, a
1
; k
2
, a
2
; k
3
, a
3
; k
4
, a
4
)
= i
d
26
k
(2π)
26
¸
a,e
S
D
2
(−k, a, e; k
1
, a
1
; k
2
, a
2
)S
D
2
(k, a, e; k
3
, a
3
; k
4
, a
4
)
−k
2
+iǫ
= i
d
26
k
(2π)
26
¸
a,e
1
−k
2
+iǫ
(−i)g
′
0
(2π)
26
δ
26
(k
1
+k
2
−k)e k
12
Tr(λ
a
[λ
a
1
, λ
a
2
])
(−i)g
′
0
(2π)
26
δ
26
(k +k
3
+k
4
)e k
34
Tr(λ
a
[λ
a
3
, λ
a
4
])
= −ig
′2
o
(2π)
26
δ
26
(
¸
i
k
i
)
¸
a
Tr(λ
a
[λ
a
1
, λ
a
2
])Tr(λ
a
[λ
a
3
, λ
a
4
])
¸
e
e k
12
e k
34
s +iǫ
= −ig
′2
o
(2π)
26
δ
26
(
¸
i
k
i
)Tr([λ
a
1
, λ
a
2
][λ
a
3
, λ
a
4
])
u −t
s +iǫ
. (33)
In the second equality we have substituted equation (31) (or (6.5.12)). The polarization vector e
is summed over an orthonormal basis of (spacelike) vectors obeying the physical state condition
ek = 0, which after the integration over k in the third equality becomes e(k
1
+k
2
) = e(k
3
+k
4
) = 0.
If we choose one of the vectors in this basis to be e
′
= k
12
/[k
12
[, then none of the others will
contribute to the sum in the second to last line, which becomes,
¸
e
e k
12
e k
34
= k
12
k
34
= u −t. (34)
In the last equality of (33) we have also applied equation (6.5.9). Comparing (32) and (33), we see
that
g
′
o
=
g
o
√
2α
′
, (35)
in agreement with (6.5.14).
This result conﬁrms the normalization of the photon vertex operator as written in equation
(3.6.26). The stateoperator mapping gives the same normalization: in problem 2.9, we saw that
6 CHAPTER 6 45
the photon vertex operator was
e
µ
α
µ
−1
[0; 0`
∼
= i
2
α
′
e
µ
⋆
⋆∂X
µ
e
ikX⋆
⋆ = i
2
α
′
e
µ
⋆
⋆
¯
∂X
µ
e
ikX⋆
⋆. (36)
Since the boundary is along the σ
1
axis, the derivative can be written using (2.1.3):
˙
X = ∂
1
X = (∂ +
¯
∂)X = 2∂X. (37)
Hence the vertex operator is
i
√
2α
′
e
µ
⋆
⋆
˙
X
µ
e
ikX⋆
⋆, (38)
which, after multiplying by the factor −g
o
and integrating over the position on the boundary, agrees
with (3.6.26).
6.5 Problem 6.9
(a) There are six cyclic orderings of the four vertex operators on the boundary of the disk,
illustrated in ﬁgure 6.2. Consider ﬁrst the ordering (3, 4, 1, 2) shown in ﬁgure 6.2(a). If we ﬁx the
positions of the vertex operators for gauge bosons 1, 2, and 3, with
−∞< y
3
< y
1
< y
2
< ∞, (39)
then we must integrate the position of the fourth gauge boson vertex operator from y
3
to y
1
. The
contribution this ordering makes to the amplitude is
e
−λ
g
4
o
(2α
′
)
−2
Tr(λ
a
3
λ
a
4
λ
a
1
λ
a
2
)e
1
µ
1
e
2
µ
2
e
3
µ
3
e
4
µ
4
y
1
y
3
dy
4
⋆
⋆c
1
˙
X
µ
3
e
ik
3
X
(y
3
)
⋆
⋆
⋆
⋆
˙
X
µ
4
e
ik
4
X
(y
4
)
⋆
⋆
⋆
⋆c
1
˙
X
µ
1
e
ik
1
X
(y
1
)
⋆
⋆
⋆
⋆c
1
˙
X
µ
2
e
ik
2
X
(y
2
)
⋆
⋆
(40)
= e
−λ
g
4
o
(2α
′
)
−2
Tr(λ
a
3
λ
a
4
λ
a
1
λ
a
2
)e
1
µ
1
e
2
µ
2
e
3
µ
3
e
4
µ
4
iC
X
D
2
C
g
D
2
(2π)
26
δ
26
(
¸
i
k
i
)
[y
12
[
2α
′
k
1
k
2
+1
[y
13
[
2α
′
k
1
k
3
+1
[y
23
[
2α
′
k
2
k
3
+1
y
1
y
3
dy
4
[y
14
[
2α
′
k
1
k
4
[y
24
[
2α
′
k
2
k
4
[y
34
[
2α
′
k
3
k
4
'[v
µ
3
(y
3
) +q
µ
3
(y
3
)][v
µ
4
(y
4
) +q
µ
4
(y
4
)]
[v
µ
1
(y
1
) +q
µ
1
(y
1
)][v
µ
2
(y
2
) +q
µ
2
(y
2
)]` .
The v
µ
that appear in the path integral in the last two lines are linear in the momenta; for instance
v
µ
(y
3
) = −2iα
′
(k
µ
1
y
−1
31
+k
µ
2
y
−1
32
+k
µ
4
y
−1
34
). (41)
They therefore contribute terms in which the polarization vectors e
i
are dotted with the momenta.
Since we are looking only for terms in which the e
i
appear in the particular combination e
1
e
2
e
3
e
4
,
6 CHAPTER 6 46
we can neglect the v
µ
. The terms we are looking for arise from the contraction of the q
µ
with each
other. Speciﬁcally, the singular part of the OPE of q
µ
(y) with q
ν
(y
′
) is
−2α
′
(y −y
′
)
−2
η
µν
, (42)
so the combination e
1
e
2
e
3
e
4
arises from the contractions of q
µ
1
(y
1
) with q
µ
2
(y
2
) and q
µ
3
(y
3
)
with q
µ
4
(y
4
):
ig
2
o
α
′−1
Tr(λ
a
3
λ
a
4
λ
a
1
λ
a
2
)e
1
e
2
e
3
e
4
(2π)
26
δ
26
(
¸
i
k
i
)
[y
12
[
2α
′
k
1
k
2
−1
[y
13
[
2α
′
k
1
k
3
+1
[y
23
[
2α
′
k
2
k
3
+1
y
1
y
3
dy
4
[y
14
[
2α
′
k
1
k
4
[y
24
[
2α
′
k
2
k
4
[y
34
[
2α
′
k
3
k
4
−2
. (43)
We can choose y
1
, y
2
, and y
3
as we like, so long as we obey (39), and the above expression simpliﬁes
if we take the limit y
2
→ ∞ while keeping y
1
and y
3
ﬁxed. Then [y
12
[ ∼ [y
23
[ ∼ y
2
and (since
y
3
< y
4
< y
1
) [y
24
[ ∼ y
2
as well. Making these substitutions above, y
2
appears with a total power
of
2α
′
k
1
k
2
−1 + 2α
′
k
2
k
3
+ 1 + 2α
′
k
2
k
4
= 2α
′
(k
1
+k
3
+k
4
) k
2
= −2α
′
k
2
2
= 0. (44)
We can simplify further by setting y
3
= 0 and y
1
= 1. Since s = −2k
3
k
4
and u = −2k
1
k
4
, the
integral above reduces to:
1
0
dy
4
(1 −y
4
)
−α
′
u
y
−α
′
s−2
4
= B(−α
′
u + 1, −α
′
s −1). (45)
If we now consider a diﬀerent cyclic ordering of the vertex operators, we can still ﬁx y
1
, y
2
,
and y
3
while integrating over y
4
. Equation (43) will remain the same, with two exceptions: the
order of the λ
a
matrices appearing in the trace, and the limits of integration on y
4
, will change
to reﬂect the new order. The limits of integration will be whatever positions immediately precede
and succede y
4
, while the position that is opposite y
4
will be taken to inﬁnity. It can easily be seen
that the trick that allowed us to take y
2
to inﬁnity (equation (44)) works equally well for y
1
or y
3
.
The lower and upper limits of integration can be ﬁxed at 0 and 1 respectively as before, and the
resulting integral over y
4
will once again give a beta function. However, since diﬀerent factors in
the integrand of (43) survive for diﬀerent orderings, the beta function will have diﬀerent arguments
in each case. Putting together the results from the six cyclic orderings, we ﬁnd that the part of the
6 CHAPTER 6 47
four gauge boson amplitude proportional to e
1
e
2
e
3
e
4
is
ig
2
o
α
′
e
1
e
2
e
3
e
4
(2π)
26
δ
26
(
¸
i
k
i
) (46)
Tr(λ
a
1
λ
a
2
λ
a
4
λ
a
3
+λ
a
1
λ
a
3
λ
a
4
λ
a
2
)B(−α
′
t + 1, −α
′
s −1)
+Tr(λ
a
1
λ
a
3
λ
a
2
λ
a
4
+λ
a
1
λ
a
4
λ
a
2
λ
a
3
)B(−α
′
t + 1, −α
′
u + 1)
+Tr(λ
a
1
λ
a
2
λ
a
3
λ
a
4
+λ
a
1
λ
a
4
λ
a
3
λ
a
2
)B(−α
′
u + 1, −α
′
s −1)
.
(b) There are four treelevel diagrams that contribute to fourboson scattering in YangMills
theory: the schannel, the tchannel, the uchannel, and the fourpoint vertex. The fourpoint
vertex diagram (which is independent of momenta) includes the following term proportional to
e
1
e
2
e
3
e
4
:
−ig
′2
o
e
1
e
2
e
3
e
4
(2π)
26
δ
26
(
¸
i
k
i
)
¸
e
(f
a
1
a
3
e
f
a
2
a
4
e
+f
a
1
a
4
e
f
a
2
a
3
e
) (47)
(see Peskin and Schroeder, equation A.12). The YangMills coupling is
g
′
o
= (2α
′
)
−1/2
g
o
(48)
(6.5.14), and the f
abc
are the gauge group structure constants:
f
abc
= Tr
[λ
a
, λ
b
]λ
c
. (49)
We can therefore rewrite (47) in the following form:
−
ig
2
o
α
′
e
1
e
2
e
3
e
4
(2π)
26
δ
26
(
¸
i
k
i
)Tr
λ
a
1
λ
a
3
λ
a
2
λ
a
4
+λ
a
1
λ
a
4
λ
a
2
λ
a
3
(50)
−
1
2
(λ
a
1
λ
a
2
λ
a
4
λ
a
3
+λ
a
1
λ
a
3
λ
a
4
λ
a
2
+λ
a
1
λ
a
2
λ
a
3
λ
a
4
+λ
a
1
λ
a
4
λ
a
3
λ
a
2
)
.
Of the three diagrams that contain threepoint vertices, only the schannel diagram contains a term
proportional to e
1
e
2
e
3
e
4
. It is
−ig
′2
o
e
1
e
2
e
3
e
4
(2π)
26
δ
26
(
¸
i
k
i
)
u −t
s
¸
e
f
a
1
a
2
e
f
a
3
a
4
e
= −
ig
2
o
α
′
e
1
e
2
e
3
e
4
(2π)
26
δ
26
(
¸
i
k
i
)
t −u
2s
(51)
Tr (λ
a
1
λ
a
2
λ
a
3
λ
a
4
+λ
a
1
λ
a
4
λ
a
3
λ
a
2
−λ
a
1
λ
a
2
λ
a
4
λ
a
3
−λ
a
1
λ
a
3
λ
a
4
λ
a
2
) .
Combining (50) and (51) and using
s +t +u = 0, (52)
6 CHAPTER 6 48
we obtain, for the part of the fourboson amplitude proportional to e
1
e
2
e
3
e
4
, at tree level,
−
ig
2
o
α
′
e
1
e
2
e
3
e
4
(2π)
26
δ
26
(
¸
i
k
i
) (53)
¸
Tr(λ
a
1
λ
a
2
λ
a
4
λ
a
3
+λ
a
1
λ
a
3
λ
a
4
λ
a
2
)
−1 −
t
s
+Tr(λ
a
1
λ
a
3
λ
a
2
λ
a
4
+λ
a
1
λ
a
4
λ
a
2
λ
a
3
)
+Tr(λ
a
1
λ
a
2
λ
a
3
λ
a
4
+λ
a
1
λ
a
4
λ
a
3
λ
a
2
)
−1 −
u
s
.
This is intentionally written in a form suggestively similar to equation (46). It is clear that (46)
reduces to (53) (up to an overall sign) if we take the limit α
′
→ 0 with s, t, and u ﬁxed, since in
that limit
B(−α
′
t + 1, −α
′
s −1) ≈ −1 −
t
s
,
B(−α
′
t + 1, −α
′
u + 1) ≈ 1, (54)
B(−α
′
u + 1, −α
′
s −1) ≈ −1 −
u
s
.
Thus this single string theory diagram reproduces, at momenta small compared to the string scale,
the sum of the ﬁeld theory Feynman diagrams.
6.6 Problem 6.11
(a) The X path integral is given by (6.2.19):
: ∂X
µ
¯
∂X
ν
e
ik
1
X
(z
1
, ¯ z
1
) :: e
ik
2
X
(z
2
, ¯ z
2
) :: e
ik
3
X
(z
3
, ¯ z
3
) :
S
2
= −iC
X
S
2
α
′2
4
(2π)
26
δ
26
(k
1
+k
2
+k
3
) (55)
[z
12
[
α
′
k
1
k
2
[z
13
[
α
′
k
1
k
3
[z
23
[
α
′
k
2
k
3
k
µ
2
z
12
+
k
µ
3
z
13
k
ν
2
¯ z
12
+
k
ν
3
¯ z
13
.
The ghost path integral is given by (6.3.4):
': c(z
1
)˜ c(¯ z
1
) :: c(z
2
)˜ c(¯ z
2
) :: c(z
3
)˜ c(¯ z
3
) :`
S
2
= C
g
S
2
[z
12
[
2
[z
13
[
2
[z
23
[
2
. (56)
The momentumconserving delta function and the mass shell conditions k
2
1
= 0, k
2
2
= k
2
3
= 4/α
′
imply
k
1
k
2
= k
1
k
3
= 0, k
2
k
3
= −4/α
′
. (57)
Using (57), the transversality of the polarization tensor,
e
1µν
k
µ
1
= 0, (58)
6 CHAPTER 6 49
and the result (6.6.8),
C
S
2
≡ e
−2λ
C
X
S
2
C
g
S
2
=
8π
α
′
g
2
c
, (59)
we can put together the full amplitude for two closedstring tachyons and one massless closed string
on the sphere:
S
S
2
(k
1
, e
1
; k
2
; k
3
)
= g
2
c
g
′
c
e
−2λ
e
1µν
: ˜ cc∂X
µ
¯
∂X
ν
e
ik
1
X
(z
1
, ¯ z
1
) :: ˜ cce
ik
2
X
(z
2
, ¯ z
3
) :: ˜ cce
ik
2
X
(z
2
, ¯ z
3
) :
S
2
= −iC
S
2
g
2
c
g
′
c
α
′2
4
(2π)
26
δ
26
(k
1
+k
2
+k
3
)
e
1µν
[z
12
[
2
[z
13
[
2
[z
23
[
2
k
µ
2
z
12
+
k
µ
3
z
13
k
ν
2
¯ z
12
+
k
ν
3
¯ z
13
= −i2πα
′
g
′
c
(2π)
26
δ
26
(k
1
+k
2
+k
3
)
e
1µν
[z
12
[
2
[z
13
[
2
[z
23
[
2
k
µ
23
2z
12
−
k
µ
23
2z
13
k
ν
23
2¯ z
12
−
k
ν
23
2¯ z
13
= −i
πα
′
2
g
′
c
(2π)
26
δ
26
(k
1
+k
2
+k
3
)e
1µν
k
µ
23
k
ν
23
. (60)
(b) Let us calculate the amplitude for massless closed string exchange between closed string
tachyons (this is a treelevel ﬁeld theory calculation but for the vertices we will use the amplitude
calculated above in string theory). We will restrict ourselves to the schannel diagram, because we
are interested in comparing the result with the pole at s = 0 in the VirasoroShapiro amplitude.
Here the propagator for the massless intermediate string provides the pole at s = 0:
−i(2π)
26
δ
26
(
¸
k
i
)
π
2
α
′2
g
′2
c
4s
¸
e
e
µν
k
µ
12
k
ν
12
e
µ
′
ν
′ k
µ
′
34
k
ν
′
34
. (61)
Here e is summed over an orthonormal basis of symmetric polarization tensors obeying the condition
e
µν
(k
µ
1
+k
µ
2
) = 0. We could choose as one element of that basis the tensor
e
µν
=
k
12µ
k
12ν
k
2
12
, (62)
which obeys the transversality condition by virtue of the fact that k
2
1
= k
2
2
. With this choice, none
of the other elements of the basis would contribute to the sum, which reduces to
(k
12
k
34
)
2
= (u −t)
2
. (63)
The amplitude (61) is thus
−i(2π)
26
δ
26
(
¸
k
i
)
π
2
α
′2
g
′2
c
4
(u −t)
2
s
. (64)
6 CHAPTER 6 50
Now, the VirasoroShapiro amplitude is
i(2π)
26
δ
26
(
¸
k
i
)
16π
2
g
2
c
α
′
Γ(a)Γ(b)Γ(c)
Γ(a +b)Γ(a +c)Γ(b +c)
, (65)
where
a = −1 −
α
′
s
4
, b = −1 −
α
′
t
4
, c = −1 −
α
′
u
4
. (66)
The pole at s = 0 arises from the factor of Γ(a), which is, to lowest order in s,
Γ(a) ≈
4
α
′
s
. (67)
To lowest order in s the other gamma functions simplify to
Γ(b)Γ(c)
Γ(a +b)Γ(a +c)Γ(b +c)
≈
Γ(b)Γ(c)
Γ(b −1)Γ(c −1)Γ(2)
= (b −1)(c −1)
= −
α
′2
64
(u −t)
2
. (68)
Thus the part of the amplitude we are interested in is
−i(2π)
26
δ
26
(
¸
k
i
)π
2
g
2
c
(u −t)
2
s
. (69)
Comparison with (64) shows that
g
c
=
α
′
g
′
c
2
. (70)
(c) In Einstein frame the tachyon kinetic term decouples from the dilaton, as the tachyon action
(6.6.16) becomes
S
T
= −
1
2
d
26
x(−
˜
G)
1/2
˜
G
µν
∂
µ
T∂
ν
T −
4
α
′
e
˜
Φ/6
T
2
. (71)
If we write a metric perturbation in the following form,
˜
G
µν
= η
µν
+ 2κe
µν
f, (72)
where e
µν
e
µν
= 1, then the kinetic term for f will be canonically normalized. To lowest order in f
and T, the interaction Lagrangian is
L
int
= κe
µν
f∂
µ
T∂
ν
T +κe
µ
µ
f
2
α
′
T
2
−
1
2
∂
µ
T∂
µ
T
(73)
(from now on all indices are raised and lowered with η
µν
). The second term, proportional to the
trace of e, makes a vanishing contribution to the amplitude onshell:
−iκe
µ
µ
4
α
′
+k
2
k
3
(2π)
26
δ
26
(k
1
+k
2
+k
3
) = 0. (74)
6 CHAPTER 6 51
(The trace of the massless closed string polarization tensor e used in the string calculations of parts
(a) and (b) above represents the dilaton, not the trace of the (Einstein frame) graviton, which can
always be gauged away.) The amplitude from the ﬁrst term of (73) is
2iκe
µν
k
µ
2
k
ν
3
(2π)
26
δ
26
(k
1
+k
2
+k
3
) = −i
κ
2
e
µν
k
µ
23
k
ν
23
(2π)
26
δ
26
(k
1
+k
2
+k
3
), (75)
where we have used the transversality of the graviton polarization e
µν
k
µ
1
= 0. Comparison with
the amplitude (6.6.14) shows that
κ = πα
′
g
′
c
. (76)
6.7 Problem 6.12
We can use the three CKVs of the upper halfplane to ﬁx the position z of the closedstring vertex
operator and the position y
1
of one of the upperstring vertex operators. We integrate over the
position y
2
of the unﬁxed openstring vertex operator:
S
D
2
(k
1
, k
2
, k
3
)
= g
c
g
2
o
e
−λ
dy
2
: c˜ ce
ik
1
X
(z, ¯ z) :
⋆
⋆c
1
e
ik
2
X
(y
1
)
⋆
⋆
⋆
⋆e
ik
3
X
(y
2
)
⋆
⋆
D
1
= g
c
g
2
o
e
−λ
C
g
D
2
[z −y
1
[[¯ z −y
1
[[z − ¯ z[iC
X
D
2
(2π)
26
δ
26
(k
1
+k
2
+k
3
)
[z − ¯ z[
α
′
k
2
1
/2
[z −y
1
[
2α
′
k
1
k
2
dy
2
[z −y
2
[
2α
′
k
1
k
3
[y
1
−y
2
[
2α
′
k
2
k
3
= iC
D
2
g
c
g
2
o
(2π)
26
δ
26
(k
1
+k
2
+k
3
)
[z −y
1
[
−2
[z − ¯ z[
3
dy
2
[z −y
2
[
−4
[y
1
−y
2
[
2
. (77)
The very last line is equal to 4π, independent of z and y
1
(as it should be), as can be calculated
by contour integration in the complex plane. Taking into account (6.4.14), the result is
4πig
c
α
′
(2π)
26
δ
26
(k
1
+k
2
+k
3
). (78)
7 CHAPTER 7 52
7 Chapter 7
7.1 Problem 7.1
Equation (6.2.13) applied to the case of the torus tells us
n
¸
i=1
: e
ik
i
X(w
i
, ¯ w
i
)
:
¸
T2
=
iC
X
T
2
(τ)(2π)
d
δ
d
(
¸
k
i
) exp
¸
−
¸
i<j
k
i
k
j
G
′
(w
i
, w
j
) −
1
2
¸
i
k
2
i
G
′
r
(w
i
, w
i
)
¸
. (1)
G
′
is the Green’s function (7.2.3),
G
′
(w
i
, w
j
) = −
α
′
2
ln
ϑ
1
w
ij
2π
[τ
2
+α
′
(Imw
ij
)
2
4πτ
2
+k(τ), (2)
while G
′
r
is the renormalized Green’s function, deﬁned by (6.2.15),
G
′
r
(w
i
, w
j
) = G
′
(w
i
, w
j
) +
α
′
2
ln [w
ij
[
2
, (3)
designed to be ﬁnite in the limit w
j
→w
i
:
lim
w
j
→w
i
G
′
r
(w
i
, w
j
) = −
α
′
2
ln
lim
w
ij
→0
ϑ
1
w
ij
2π
[τ
w
ij
2
+k(τ)
= −
α
′
2
ln
∂
ν
ϑ
1
(ν = 0[τ)
2π
2
+k(τ). (4)
The argument of the exponential in (1) is thus
−
¸
i<j
k
i
k
j
G
′
(w
i
, w
j
) −
1
2
¸
i
k
2
i
G
′
r
(w
i
, w
i
)
=
¸
i<j
α
′
k
i
k
j
ln
ϑ
1
w
ij
2π
[τ
−
(Imw
ij
)
2
4πτ
2
+
α
′
2
ln
∂
ν
ϑ
1
(0[τ)
2π
¸
i
k
2
i
−
1
2
k(τ)
¸
i,j
k
i
k
j
=
¸
i<j
α
′
k
i
k
j
ln
2π
∂
ν
ϑ
1
(0[τ)
ϑ
1
w
ij
2π
[τ
−
(Imw
ij
)
2
4πτ
2
, (5)
where in the second equality we have used the overall momentumconserving delta function, which
implies
¸
i,j
k
i
k
j
= 0. Plugging this into (1) yields (7.2.4).
Under the modular transformation τ → τ
′
= −1/τ the coordinate w is mapped to w
′
= w/τ.
The weights of the vertex operator : exp(ik
i
X(w
i
, ¯ w
i
)) : are (2.4.17)
h =
˜
h =
α
′
k
2
i
4
(6)
7 CHAPTER 7 53
so, according to (2.4.13), the product of vertex operators on the LHS of (1) transforms to
n
¸
i=1
: e
ik
i
X(w
′
i
, ¯ w
′
i
)
:
¸
T
2
= [τ[
P
i
α
′
k
2
i
/2
n
¸
i=1
: e
ik
i
X(w
i
, ¯ w
i
)
:
¸
T
2
. (7)
On the RHS of (1), the vacuum amplitude
C
X
T
2
= (4πα
′
τ
2
)
−d/2
[η(τ)[
−2d
(8)
is invariant, since
τ
′
2
=
τ
2
[τ[
2
, (9)
and (7.2.4b)
η(τ
′
) = (−iτ)
1/2
η(τ). (10)
According to (7.2.40d),
ϑ
1
w
′
ij
2π
[τ
′
= −i(−iτ)
1/2
exp
iw
2
ij
4πτ
ϑ
1
w
ij
2π
[τ
(11)
and
∂
ν
ϑ
1
(0[τ
′
) = (−iτ)
3/2
∂
ν
ϑ
1
(0[τ), (12)
so
ln
2π
∂
ν
ϑ
1
(0[τ
′
)
ϑ
1
w
′
ij
2π
[τ
′
= ln
2π
τ∂
ν
ϑ
1
(0[τ)
ϑ
1
w
ij
2π
[τ
−Im
w
2
ij
4πτ
. (13)
The second term on the right is equal to
Im
w
2
ij
4πτ
=
1
4π[τ[
2
2τ
1
Imw
ij
Re w
ij
−τ
2
(Re w
ij
)
2
+τ
2
(Imw
ij
)
2
, (14)
and it cancels the change in the last term on the RHS of (5):
(Imw
′
ij
)
2
4πτ
′
2
=
1
4πτ
2
[τ[
2
−2τ
1
τ
2
Imw
ij
Re w
ij
+τ
2
2
(Re w
ij
)
2
+τ
2
1
(Imw
ij
)
2
. (15)
The only change, then, is the new factor of [τ[
−1
in the logarithm on the RHS of (13), which gets
taken to the power
¸
i<j
α
′
k
i
k
j
= −
1
2
¸
i
α
′
k
2
i
; (16)
we ﬁnally arrive, as expected, at the transformation law (7).
7 CHAPTER 7 54
7.2 Problem 7.3
As usual it is convenient to solve Poisson’s equation (6.2.8) in momentum space. Because the torus
is compact, the momentum space is a lattice, generated by the complex numbers k
a
= 1 − iτ
1
/τ
2
and k
b
= i/τ
2
. The Laplacian in momentum space is
−[k[
2
= −[n
a
k
a
+n
b
k
b
[
2
= −
[n
b
−n
a
τ[
2
τ
2
2
≡ −ω
2
nan
b
. (17)
In terms of the normalized Fourier components
X
nan
b
(w) =
1
2πτ
1/2
2
e
i(naka+n
b
k
b
)w
(18)
(where the dot product means, as usual, A B ≡ Re AB
∗
), the Green’s function in real space is
(6.2.7)
G
′
(w, w
′
) =
¸
(na,n
b
),=(0,0)
2πα
′
ω
2
nan
b
X
nan
b
(w)
∗
X
nan
b
(w
′
). (19)
Rather than show that (19) is equal to (7.2.3) (or (2)) directly, we will show that (7.2.3) has the
correct Fourier coeﬃcients, that is,
T
2
d
2
wX
00
(w)G
′
(w, w
′
) = 0, (20)
T
2
d
2
wX
nan
b
(w)G
′
(w, w
′
) =
2πα
′
ω
2
nan
b
X
nan
b
(w
′
), (n
a
, n
b
) = (0, 0). (21)
The windependent part of the Green’s function is left as the unspeciﬁed constant k(τ) in (7.2.3),
which is adjusted (as a function of τ) to satisfy equation (20). To prove (21), we ﬁrst divide both
sides by X
nan
b
(w
′
), and use the fact that G
′
depends only on the diﬀerence w − w
′
to shift the
variable of integration:
T
2
d
2
we
i(naka+n
b
k
b
)w
G
′
(w, 0) =
2πα
′
ω
2
nan
b
. (22)
To evaluate the LHS, let us use coordinates x and y on the torus deﬁned by w = 2π(x +yτ). The
Jacobian for this change of coordinates is (2π)
2
τ
2
, so we have
(2π)
2
τ
2
1
0
dy
1
0
dxe
2πi(nax+n
b
y)
¸
−
α
′
2
ln [ϑ
1
(x +yτ[τ)[
2
+α
′
πy
2
τ
2
. (23)
Using the inﬁniteproduct representation of the theta function (7.2.38d), we can write,
ln ϑ
1
(x +yτ[τ) = K(τ) −iπ(x +yτ)
+
∞
¸
m=0
ln
1 −e
2πi(x+(y+m)τ)
+
∞
¸
m=1
ln
1 −e
−2πi(x+(y−m)τ)
, (24)
7 CHAPTER 7 55
where we’ve collected the terms that are independent of x and y into the function K(τ), which
drops out of the integral. Expression (23) thus becomes
−2π
2
α
′
τ
2
1
0
dy
1
0
dxe
2πi(nax+n
b
y)
¸
2πτ
2
y(1 −y)
+
∞
¸
m=0
ln
1 −e
2πi(x+(y+m)τ)
2
+
∞
¸
m=1
ln
1 −e
−2πi(x+(y−m)τ)
2
¸
. (25)
The integration of the ﬁrst term in the brackets is straightforward:
1
0
dx
1
0
dy e
2πi(nax+n
b
y)
2πτ
2
y(1 −y) =
−
τ
2
πn
2
b
, n
a
= 0
0, n
a
= 0
. (26)
To integrate the logarithms, we ﬁrst convert the inﬁnite sums into inﬁnite regions of integration in
y (using the periodicity of the ﬁrst factor under y →y + 1):
1
0
dy e
2πi(nax+n
b
y)
¸
∞
¸
m=0
ln
1 −e
2πi(x+(y+m)τ)
2
+
∞
¸
m=1
ln
1 −e
−2πi(x+(y−m)τ)
2
¸
=
∞
0
dy e
2πi(nax+n
b
y)
ln
1 −e
2πi(x+yτ)
2
+
0
−∞
dy e
2πi(nax+n
b
y)
ln
1 −e
−2πi(x+yτ)
2
. (27)
The x integral can now be performed by separating the logarithms into their holomorphic and
antiholomorphic pieces and Taylor expanding. For example,
1
0
dxe
2πi(nax+n
b
y)
ln
1 −e
2πi(x+yτ)
= −
∞
¸
n=1
1
n
1
0
dxe
2πi((na+n)x+(n
b
+nτ)y)
=
1
na
e
2πi(n
b
−naτ)y
, n
a
< 0
0, n
a
≥ 0
. (28)
The y integral is now straightforward:
∞
0
dy
1
n
a
e
2πi(n
b
−naτ)y
= −
1
2πin
a
(n
b
−n
a
τ)
. (29)
7 CHAPTER 7 56
Similarly,
∞
0
dy
1
0
dxe
2πi(nax+n
b
y)
ln
1 −e
−2πi(x+y¯ τ)
=
1
2πina(n
b
−na¯ τ)
, n
a
> 0
0, n
a
≤ 0
, (30)
0
−∞
dy
1
0
dxe
2πi(nax+n
b
y)
ln
1 −e
−2πi(x+yτ)
=
−
1
2πina(n
b
−naτ)
, n
a
> 0
0, n
a
≤ 0
, (31)
0
−∞
dy
1
0
dxe
2πi(nax+n
b
y)
ln
1 −e
2πi(x+y¯ τ)
=
1
2πina(n
b
−na¯ τ)
, n
a
< 0
0, n
a
≥ 0
. (32)
Expressions (26) and (29)–(32) can be added up to give a single expression valid for any sign of n
a
:
−
τ
2
π[n
b
−n
a
τ[
2
; (33)
multiplying by the prefactor −2π
2
α
′
τ
2
in front of the integral in (25) indeed yields precisely the
RHS of (22), which is what we were trying to prove.
7.3 Problem 7.5
In each case we hold ν ﬁxed while taking τ to its appropriate limit.
(a) When Imτ →∞, q ≡ exp(2πiτ) →0, and it is clear from either the inﬁnite sum expressions
(7.2.37) or the inﬁnite product expressions (7.2.38) that in this limit
ϑ
00
(ν, τ) →1, (34)
ϑ
10
(ν, τ) →1, (35)
ϑ
01
(ν, τ) →0, (36)
ϑ
11
(ν, τ) →0. (37)
Note that all of these limits are independent of ν.
(b) Inverting the modular transformation (7.2.40a), we have
ϑ
00
(ν, τ) = (−iτ)
−1/2
e
−πiν
2
/τ
ϑ
00
(ν/τ, −1/τ)
= (−iτ)
−1/2
∞
¸
n=−∞
e
−πi(ν−n)
2
/τ
. (38)
When we take τ to 0 along the imaginary axis, each term in the series will go either to 0 (if
Re(ν −n)
2
> 0) or to inﬁnity (if Re(ν −n)
2
≤ 0). Since diﬀerent terms in the series cannot cancel
for arbitrary τ, the theta function can go to 0 only if every term in the series does so:
∀n ∈ Z, Re(ν −n)
2
> 0; (39)
7 CHAPTER 7 57
otherwise it will diverge. For [ Re ν[ ≤ 1/2, condition (39) is equivalent to
[ Imν[ < [ Re ν[; (40)
in general, for [ Re ν −n[ ≤ 1/2, the theta function goes to 0 if
[ Imν[ < [ Re ν −n[. (41)
Since ϑ
01
(ν, τ) = ϑ
00
(ν + 1/2, τ), the region in which it goes to 0 in the limit τ → 0 is simply
shifted by 1/2 compared to the case treated above.
For ϑ
10
, the story is the same as for ϑ
00
, since
ϑ
10
(ν, τ) = (−iτ)
−1/2
e
−πiν
2
/τ
ϑ
01
(ν/τ, −1/τ)
= (−iτ)
−1/2
∞
¸
n=−∞
(−1)
n
e
−πi(ν−n)
2
/τ
; (42)
the sum will again go to 0 where (39) is obeyed, and inﬁnity elsewhere.
Finally, ϑ
11
goes to 0 in the same region as ϑ
01
, since the sum is the same as (42) except over
the halfoddintegers,
ϑ
11
(ν, τ) = i(−iτ)
−1/2
e
−πiν
2
/τ
ϑ
11
(ν/τ, −1/τ)
= (−iτ)
−1/2
∞
¸
n=−∞
(−1)
n
e
−πi(ν−n+1/2)
2
/τ
, (43)
so the region (39) is shifted by 1/2.
(c) According to (7.2.39) and (7.2.40), under the modular transformations
τ
′
= τ + 1, (44)
τ
′
= −
1
τ
, (45)
the theta functions are exchanged with each other and multiplied by factors that are ﬁnite as long
as ν and τ are ﬁnite. Also, under (45) ν is transformed to
ν
′
=
ν
τ
. (46)
We are considering limits where τ approaches some nonzero real value τ
0
along a path parallel
to the imaginary axis, in other words, we set τ = τ
0
+ iǫ and take ǫ → 0
+
. The property of
approaching the real axis along a path parallel to the imaginary axis is preserved by the modular
transformations (to ﬁrst order in ǫ):
τ
′
= τ
0
+ 1 +iǫ, (47)
τ
′
= −
1
τ
0
+i
ǫ
τ
2
0
+O(ǫ
2
), (48)
7 CHAPTER 7 58
under (44) and (45) respectively. By a sequence of transformations (44) and (45) one can reach
any rational limit point τ
0
starting with τ
0
= 0, the case considered in part (b) above. During
these transformations (which always begin with (44)), the region (39), in which ϑ
00
goes to 0, will
repeatedly be shifted by 1/2 and rescaled by τ
0
(under (46)). (Note that the limiting value, being
either 0 or inﬁnity, is insensitive to the ﬁnite prefactors involved in the transformations (7.2.39) and
(7.2.40).) It is easy to see that this cumulative sequence of rescalings will telescope into a single
rescaling by a factor q, where p/q is the ﬁnal value of τ
0
in reduced form.
As for the case when τ
0
is irrational, I can only conjecture that the theta functions diverge
(almost) everywhere on the ν plane in that limit.
7.4 Problem 7.7
The expectation value for ﬁxed open string tachyon vertex operators on the boundary of the cylinder
is very similar to the corresponding formula (7.2.4) for closed string tachyon vertex operators on
the torus. The major diﬀerence comes from the fact that the Green’s function is doubled. The
method of images gives the Green’s function for the cylinder in terms of that for the torus (7.2.3):
G
′
C
2
(w, w
′
) = G
′
T
2
(w, w
′
) +G
′
T
2
(w, −¯ w
′
). (49)
However, since the boundary of C
2
is given by those points that are invariant under the involution
w → −¯ w, the two terms on the RHS above are equal if either w or w
′
is on the boundary. The
renormalized selfcontractions are also doubled, so we have
n
¸
i=1
⋆
⋆e
ik
i
X(w
i
, ¯ w
i
)⋆
⋆
¸
C
2
=
iC
X
C
2
(t)(2π)
26
δ
26
(
¸
i
k
i
)
¸
i<j
η(it)
−3
ϑ
1
w
ij
2π
[it
exp
¸
−
(Imw
ij
)
2
4πt
2α
′
k
i
k
j
. (50)
The boundary of the cylinder breaks into two connected components, and the vertex operator
positions w
i
must be integrated over both components. If there are ChanPaton factors, then the
integrand will include two traces, one for each component of the boundary, and the order of the
factors in each trace will be the order of the operators on the corresponding component. We will
denote the product of these two traces T(w
1
, . . . , w
n
), and of course it will also depend implicitly
on the ChanPaton factors themselves λ
a
i
. Borrowing from the cylinder vacuum amplitude given
7 CHAPTER 7 59
in (7.4.1), we can write the ntachyon amplitude as
S
C
2
(k
1
, a
1
; . . . ; k
n
, a
n
) =
i(2π)
26
δ
26
(
¸
i
k
i
)g
n
o
∞
0
dt
2t
(8π
2
α
′
t)
−13
η(it)
−24
n
¸
i=1
∂C
2
dw
i
T(w
1
, . . . , w
n
)
¸
i<j
η(it)
−3
ϑ
1
w
ij
2π
[it
exp
¸
−
(Imw
ij
)
2
4πt
2α
′
k
i
k
j
. (51)
7.5 Problem 7.8
In this problem we consider the part of the amplitude (51) in which the ﬁrst m ≥ 2 vertex operators
are on one of the cylinder’s boundaries, and the other n − m ≥ 2 are on the other one. For
concreteness let us put the ﬁrst set on the boundary at Re w = 0 and the second set on the
boundary at Re w = π, and then double the amplitude (51). Since we will be focusing on the
region of the moduli space where t is very small, it is convenient to scale the vertex operator
positions with t, so
w
i
=
2πix
i
t, i = 1, . . . , m
π + 2πix
i
t, i = m+ 1, . . . , n
. (52)
The x
i
run from 0 to 1 independent of t, allowing us to change of the order of integration in (51).
Using (16) and the mass shell condition α
′
k
2
i
= 1 we can write the part of the amplitude we’re
interested in as follows:
S
′
(k
1
, a
1
; . . . ; k
n
, a
n
) = i(2π)
26
δ
26
(
¸
i
k
i
)g
n
o
2
−13
(2π)
n−26
α
′−13
¸
i
1
0
dx
i
T
1
(x
1
, . . . , x
m
)T
2
(x
m+1
, . . . , x
n
)
∞
0
dt t
n−14
η(it)
3n−24
¸
i<j
ϑ
1,2
(ix
ij
t[it) exp(−πx
2
ij
t)
2α
′
k
i
k
j
. (53)
In the last product the type of theta function to use depends on whether the vertex operators
i and j are on the same boundary (in which case ϑ
1
) or on opposite boundaries (in which case
ϑ
2
(ix
ij
t[it) = −ϑ
1
(ix
ij
t − 1/2[it)). Concentrating now on the last line of (53), let us apply the
modular transformations (7.2.40b), (7.2.40d), and (7.2.44b), and change the variable of integration
to u = 1/t:
∞
0
duη(iu)
3n−24
¸
i<j
[ϑ
1,4
(x
ij
[iu)[
2α
′
k
i
k
j
. (54)
For large u, each of the factors in the integrand of (54) can be written as a fractional power
of q ≡ e
−2πu
times a power series in q (with coeﬃcients that may depend on the x
ij
; see (7.2.37),
7 CHAPTER 7 60
(7.2.38), and (7.2.43)):
η(iu) = q
1/24
(1 −q +. . . ); (55)
ϑ
1
(x[iu) = 2q
1/8
sin(πx) (1 −(1 + 2 cos(2πx))q +. . . ) ; (56)
ϑ
4
(x[iu) =
∞
¸
k=−∞
(−1)
k
q
k
2
/2
e
2πikx
= 1 −2 cos(2πx)q
1/2
+. . . . (57)
For η and ϑ
1
this power series involves only integer powers of q, whereas for ϑ
4
it mixes integer
and halfinteger powers. Substituting (55)(57) into (54) yields
¸
i<j
i≃j
[2 sin πx
ij
[
2α
′
k
i
k
j
∞
0
duq
n/8−1+α
′
P
i<j
i≃j
k
i
k
j
/4
(1 +. . . )
=
¸
i<j
i≃j
[2 sin πx
ij
[
2α
′
k
i
k
j
∞
0
duq
−1−α
′
s/4
(1 +. . . ) (58)
The symbol i ≃ j means that the sum or product is only over pairs of vertex operators on the same
boundary of the cylinder. We obtain the second line from the ﬁrst by using
¸
i<j
i≃j
k
i
k
j
=
¸
i<j
k
i
k
j
−
m
¸
i=1
n
¸
j=m+1
k
i
k
j
= −
1
2
n
¸
i=1
k
2
i
−
m
¸
i=1
k
i
¸
n
¸
j=m+1
k
j
¸
= −
n
2α
′
−s, (59)
where s = −(
¸
n
i=1
k
i
)
2
is the mass squared of the intermediate state propagating along the long
cylinder. (The two n’s we have cancelled against each other in (58) both come from α
′
¸
i
k
2
i
, so
in fact we could have done without the mass shell condition in the derivation.) Each power of q
appearing in the power series (1 + . . . ) in (58) will produce, upon performing the u integration, a
pole in s:
∞
0
duq
−1−α
′
s/4+k
=
2
π(4k −4 −α
′
s)
. (60)
Since every (nonnegative) integer power k appears in the expansion of (54), we have the entire
sequence of closed string masses at s = 4(k −1)/α
′
.
What about the halfinteger powers of q that appear in the expansion of ϑ
4
, (57)? The poles
from these terms in fact vanish, but only after integrating over the vertex operator positions. To
see this, consider the eﬀect of uniformly translating all the vertex operator positions on just one of
the boundaries by an amount y: x
i
→ x
i
+ y, i = 1, . . . , m. This translation changes the relative
7 CHAPTER 7 61
position x
ij
only if the vertex operators i and j are on opposite boundaries; it thus leaves the
ChanPaton factors T
1
and T
2
in (53) invariant, as well as all the factors in the integrand of (54)
except those involving ϑ
4
; those become
m
¸
i=1
n
¸
j=m+1
[ϑ
4
(x
ij
+y[iu)[
2α
′
k
i
k
j
. (61)
Expanding this out using (57), each term will be of the form
cq
P
l
k
2
l
/2
e
2πiy
P
l
k
l
, (62)
where the k
l
are integers and the coeﬃcient c depends on the x
i
and the momenta k
i
. Since y is
eﬀectively integrated over when one integrates over the vertex operator positions, only the terms
for which
¸
l
k
l
= 0 will survive. This condition implies that
¸
l
k
2
l
must be even, so only integer
powers of q produce poles in the amplitude.
7.6 Problem 7.9
We wish to consider the result of the previous problem in the simplest case, when n = 4, m = 2,
and there are no ChanPaton indices. We are interested in particular in the ﬁrst pole, at s =
−4/α
′
, where the intermediate closed string goes on shell as a tachyon. Neglecting the “. . . ” and
approximating the exponents 2α
′
k
1
k
2
= 2α
′
k
3
k
4
= −α
′
s −2 by 2, (58) becomes
−sin
2
(πx
12
) sin
2
(πx
34
)
32
π(α
′
s + 4)
. (63)
After integrating over the positions x
i
, the amplitude (53) is
S
′
(k
1
, . . . , k
4
) = −i(2π)
26
δ
26
(
¸
i
k
i
)g
4
o
2
−9
(2π)
−23
α
′−14
1
s + 4/α
′
. (64)
In Problem 6.12, we calculated the threepoint vertex for two openstring tachyons to go to a
closedstring tachyon:
4πig
c
α
′
(2π)
26
δ
26
(k
1
+k
2
+k
3
). (65)
We can reproduce the schannel pole of (64) by simply writing down the Feynman diagram using
this vertex and the closedstring tachyon propagator, i/(s + 4/α
′
):
−i(2π)
26
δ
26
(
¸
i
k
i
)g
2
c
4(2π)
2
α
′−2
1
s + 4/α
′
. (66)
We can now compare the residues of the poles in (64) and (66) to obtain the following relation
between g
c
and g
o
(note that, since the vertex (65) is valid only onshell, it is only appropriate to
compare the residues of the poles at s = −4/α
′
in (64) and (66), not the detailed dependence on s
away from the pole):
g
2
o
g
c
= 2
11/2
(2π)
25/2
α
′6
. (67)
This is in agreement with (8.7.28).
7 CHAPTER 7 62
7.7 Problem 7.10
When the gauge group is a product of U(n
i
) factors, the generators are block diagonal. A treelevel
diagram is proportional to Tr(λ
a
1
λ
a
l
), and vanishes unless all the ChanPaton factors are in the
same block. Unitarity requires
Tr(λ
a
1
λ
a
i
λ
a
i+1
λ
a
l
) = Tr(λ
a
1
λ
a
i
λ
e
) Tr(λ
e
λ
a
i+1
λ
a
l
), (68)
which is an identity for any product of U(n
i
)s. (The LHS vanishes if all the λ
a
are not in the same
block, and equals the usual U(n) value if the are. The RHS has the same property, because the
λ
e
must be in the same block both with the ﬁrst group of λ
a
s and with the second group for a
nonzero result.)
Now consider the cylinder with two vertex operators on each boundary, with ChanPaton fac
tors λ
a
and λ
b
on one boundary and λ
c
and λ
d
on the other. This amplitude is proportional to
Tr(λ
a
λ
b
) Tr(λ
c
λ
d
), i.e. we only need the two ChanPaton factors in each pair to be in the same
block. However, if we make a unitary cut in the open string channel, then the cylinder becomes
a disk with an open string propagator connecting two points on the boundary. If on the inter
mediate state we sum only over blockdiagonal generators, then this amplitude will vanish unless
λ
a
, λ
b
, λ
c
, λ
d
are all in the same block. For this to be consistent there can only be one block, i.e.
the gauge group must be simple.
7.8 Problem 7.11
We will be using the expression (7.3.9) for the pointparticle vacuum amplitude to obtain a gen
eralized version of the cylinder vacuum amplitude (7.4.1). Since (7.3.9) is an integral over the
circle modulus l, whereas (7.4.1) is an integral over the cylinder modulus t, we need to know the
relationship between these quantities. The modulus l is deﬁned with respect to the pointparticle
action (5.1.1), which (after choosing the analog of unit gauge for the einbein e) is
1
2
l
0
dτ
(∂
τ
x)
2
+m
2
. (69)
The Polyakov action in unit gauge (2.1.1), on a cylinder with modulus t, is
1
4πα
′
π
0
dw
1
2πt
0
dw
2
(∂
1
X)
2
+ (∂
2
X)
2
. (70)
Decomposing X(w
1
, w
2
) into its centerofmass motion x(w
2
) and its internal state y(w
1
, w
2
) (with
dw
1
y = 0), the Polyakov action becomes
1
4α
′
2πt
0
dw
2
(∂
2
x)
2
+
1
π
π
0
dw
1
(∂
1
y)
2
+ (∂
2
y)
2
. (71)
7 CHAPTER 7 63
We can equate (69) and (71) by making the identiﬁcations τ = 2α
′
w
2
, l = 4πα
′
t, and
m
2
=
1
4πα
′2
π
0
dw
1
(∂
1
y)
2
+ (∂
2
y)
2
. (72)
Using the relation l = 4πα
′
t to translate between the circle and the cylinder moduli, we can
now sum the circle vacuum amplitude (7.3.9) over the open string spectrum, obtaining the second
line of (7.4.1):
Z
C
2
= iV
D
∞
0
dt
2t
(8π
2
α
′
t)
−D/2
¸
i∈1
⊥
o
e
−2πtα
′
m
2
i
. (73)
Taking a spectrum with D
′
net sets of oscillators and a ground state at m
2
= −1/α
′
, the sum is
evaluated in the usual way (with q = e
−2πt
):
¸
i∈1
⊥
o
q
α
′
m
2
i
=
¸
D
′
¸
i=1
∞
¸
n=1
∞
¸
N
in
=0
¸
q
−1+
P
∞
n=1
nN
in
= q
−1
D
′
¸
i=1
∞
¸
n=1
¸
∞
¸
N
in
=0
q
nN
in
¸
= q
−1
D
′
¸
i=1
∞
¸
n=1
1
1 −q
n
= q
D
′
/24−1
η(it)
−D
′
. (74)
As in Problem 7.8 above, we are interested in studying the propagation of closed string modes
along long, thin cylinders, which corresponds to the region t ≪1. So let us change the variable of
integration to u = 1/t,
Z
C
2
=
iV
D
2(8π
2
α
′
)
D/2
∞
0
dt t
−D/2−1
e
−2πt(D
′
/24−1)
η(it)
−D
′
=
iV
D
2(8π
2
α
′
)
D/2
∞
0
duu
(D−D
′
−2)/2
e
−2π(D
′
/24−1)/u
η(iu)
−D
′
, (75)
and add a factor of e
−πuα
′
k
2
/2
to the integrand, where k is the momentum ﬂowing along the cylinder:
∞
0
duu
(D−D
′
−2)/2
e
−2π(D
′
/24−1)/u−πuα
′
k
2
/2
η(iu)
−D
′
. (76)
Now we can expand the etafunction for large u using the product representation (7.2.43). Each
term will be of the form
∞
0
duu
(D−D
′
−2)/2
e
−2π(D
′
/24−1)/u−2πu(α
′
k
2
/4+m−D
′
24)
, (77)
where m is a nonnegative integer.
7 CHAPTER 7 64
In the case D = 26, D
′
= 24, (77) reveals the expected series of closedstring poles, as found in
Problem 7.8. If D
′
= 24, the integral yields a modiﬁed Bessel function,
∞
0
duu
c−1
e
−a/u−b/u
= 2b
−c/2
a
c/2
K
c
2
√
ab
, (78)
which has a branch cut along the negative real axis, hence the constraint D
′
= 24. When D
′
= 24,
(77) simpliﬁes to
∞
0
duu
(D−26)/2
e
−2πu(α
′
k
2
/4+m−1)
=
Γ
D−24
2
(2π(α
′
k
2
/4 +m−1))
(D−24)/2
. (79)
For D odd there is a branch cut. For D even but less than 26, the gamma function is inﬁnite;
even if one employs a “minimal subtraction” scheme to remove the inﬁnity, the remainder has a
logarithmic branch cut in k
2
. For even D ≥ 26 there is indeed a pole, but this pole is simple (as
one expects for a particle propagator) only for D = 26.
7.9 Problem 7.13
We follow the same steps in calculating the vacuum Klein bottle amplitude as Polchinski does in
calculating the vacuum torus amplitude, starting with ﬁnding the scalar partition function:
Z
X
(t) = '1`
X,K
2
= Tr(Ωe
−2πtH
)
= q
−13/6
Tr(Ωq
L
0
+
˜
L
0
). (80)
The operator Ω implements the orientationreversing boundary condition in the Euclidean time
(σ
2
) direction of the Klein bottle. Since Ω switches leftmovers and rightmovers, it is convenient
to work with states and operators that have deﬁnite properties under orientation reversal. We
therefore deﬁne the raising and lowering operators α
±
m
= (α
m
± ˜ α
m
)/
√
2, so that α
+
m
commutes
with Ω, while α
−
m
anticommutes with it (we have suppressed the spacetime index µ). These are
normalized to have the usual commutation relations, and can be used to build up the spectrum of
the closed string in the usual way. The ground states [0; k` of the closed string are invariant under
orientation reversal. The Ωq
L
0
+
˜
L
0
eigenvalue of a ground state is q
α
′
k
2
/2
; each raising operator α
+
−m
multiplies that eigenvalue by q
m
, while each raising operator α
−
−m
multiplies it by −q
m
. Summing
over all combinations of such operators, the partition function is
Z
X
(t) = q
−13/6
V
26
d
26
k
(2π)
26
q
α
′
k
2
/2
∞
¸
m=1
(1 −q
m
)
−1
(1 +q
m
)
−1
= iV
26
(4π
2
α
′
t)
−13
η(2it)
−26
. (81)
7 CHAPTER 7 65
For bc path integrals on the Klein bottle, it is again convenient to introduce raising and lowering
operators with deﬁnite properties under orientation reversal: b
±
m
= (b
m
±
˜
b
m
)/
√
2, c
±
m
= (c
m
±
˜ c
m
)/
√
2. The particular path integral we will be interested in is
c
±
m
b
+
0
K
2
= q
13/6
Tr
(−1)
F
Ωc
±
m
b
+
0
q
L
0
+
˜
L
0
. (82)
Build up the bc spectrum by starting with a ground state [ ↓↓` that is annihilated by b
±
m
for m ≥ 0
and c
±
m
for m > 0, and acting on it with the raising operators b
±
m
(m < 0) and c
±
m
(m ≤ 0).
The operator (−1)
F
Ωq
L
0
+
˜
L
0
is diagonal in this basis, whereas the operator c
±
m
b
+
0
takes basis states
to other basis states (if it does not annihilate them). Therefore the trace (81) vanishes. The
only exception is the case of the operator c
+
0
b
+
0
, which is diagonal in this basis; it projects onto
the subspace of states that are built up from c
+
0
[ ↓↓`. This state has eigenvalue −q
−2
under
(−1)
F
Ωc
+
0
b
+
0
q
L
0
+
˜
L
0
. Acting with c
−
0
does not change this eigenvalue; acting with b
+
−m
or c
+
−m
(m > 0) multiplies it by −q
m
, and with b
−
−m
or c
−
−m
by q
m
. We thus have
c
+
0
b
+
0
K
2
= 2q
1/6
∞
¸
m=1
(1 −q
m
)
2
(1 +q
m
)
2
= 2η(2it)
2
. (83)
(We have taken the absolute value of the result.)
The Klein bottle has only one CKV, which translates in the σ
2
direction. Let us temporarily
include an arbitrary number of vertex operators in the path integral, and ﬁx the σ
2
coordinate of
the ﬁrst one 1
1
. According to (5.3.9), the amplitude is
S =
∞
0
dt
4
dσ
1
1
c
2
1
1
(σ
1
1
, σ
2
1
)
1
4π
(b, ∂
t
ˆ g)
n
¸
i=2
d
2
w
i
2
1
i
(w
i
, ¯ w
i
)
¸
K
2
. (84)
The overall factor of 1/4 is from the discrete symmetries of the Klein bottle, with 1/2 from w → ¯ w
and 1/2 from w →−w.
To evaluate the b insertion, let us temporarily ﬁx the coordinate region at that for t = t
0
and
let the metric vary with t:
ˆ g(t) =
1 0
0 t
2
/t
2
0
. (85)
Then
∂
t
ˆ g(t
0
) =
0 0
0 2/t
0
(86)
and
1
4π
(b, ∂
t
ˆ g) =
1
4π
dσ
1
dσ
2
2
t
b
22
= −
dσ
1
(b
ww
+b
¯ w ¯ w
)
= 2π(b
0
+
˜
b
0
)
= 2
√
2πb
+
0
. (87)
7 CHAPTER 7 66
If we expand the c insertion in the path integral in terms of the c
m
and ˜ c
m
,
c
2
(σ
1
, σ
2
) =
1
2
c(z)
z
+
˜ c(¯ z)
¯ z
=
1
2
¸
m
c
m
z
m
+
˜ c
m
¯ z
m
, (88)
then, as we saw above, only the m = 0 term, which is c
+
0
/
√
2, will contribute to the ghost path
integral. This allows us to factor the c ghost out of the integal over the ﬁrst vertex operator position
in (83), and put all the vertex operators on the same footing:
S =
∞
0
dt
4t
c
+
0
b
+
0
n
¸
i=1
d
2
w
i
2
1
i
(w
i
, ¯ w
i
)
¸
K
2
. (89)
We can now extrapolate to the case where there are no vertex operators simply by setting n = 0
above. Using (80) and (82), this gives
Z
K
2
= iV
26
∞
0
dt
2t
(4π
2
α
′
t)
−13
η(2it)
−24
. (90)
This is oﬀ from Polchinski’s result (7.4.15) by a factor of 2.
7.10 Problem 7.15
(a) If the σ
2
coordinate is periodically identiﬁed (with period 2π), then a crosscap at σ
1
= 0
implies the identiﬁcation
(σ
1
, σ
2
)
∼
= (−σ
1
, σ
2
+π). (91)
This means the following boundary conditions on the scalar and ghost ﬁelds:
∂
1
X
µ
(0, σ
2
) = −∂
1
X
µ
(0, σ
2
+π), ∂
2
X
µ
(0, σ
2
) = ∂
2
X
µ
(0, σ
2
+π), (92)
c
1
(0, σ
2
) = −c
1
(0, σ
2
+π), c
2
(0, σ
2
) = c
2
(0, σ
2
+π), (93)
b
12
(0, σ
2
) = −b
12
(0, σ
2
+π), b
11
(0, σ
2
) = b
11
(0, σ
2
+π). (94)
These imply the following conditions on the modes at σ
1
= 0:
α
µ
n
+ (−1)
n
˜ α
µ
−n
= c
n
+ (−1)
n
˜ c
−n
= b
n
−(−1)
n
˜
b
−n
= 0 (95)
(for all n). The state corresponding to the crosscap is then
[C` ∝ exp
¸
−
∞
¸
n=1
(−1)
n
1
n
α
−n
˜ α
−n
+b
−n
˜ c
−n
+
˜
b
−n
c
−n
¸
(c
0
+ ˜ c
0
)[0; 0; ↓↓`. (96)
7 CHAPTER 7 67
(b) The Klein bottle vacuum amplitude is
∞
0
ds 'C[c
0
b
0
e
−s(L
0
+
˜
L
0
)
[C`. (97)
Since the raising and lowering operators for diﬀerent oscillators commute with each other (or, in
the case of the ghost oscillators, commute in pairs), we can factorize the integrand into a separate
amplitude for each oscillator:
e
2s
'↓↓ [(b
0
+
˜
b
0
)c
0
b
0
(c
0
+ ˜ c
0
)[ ↓↓`'0[e
−sα
′
p
2
/2
[0`
∞
¸
n=1
'0[e
−(−1)
n
cn
˜
bn
e
−sn(b
−n
cn+˜ c
−n
˜
bn)
e
−(−1)
n
b
−n
˜ c
−n
[0`
'0[e
−(−1)
n
˜ cnbn
e
−sn(
˜
b
−n
˜ cn+c
−n
bn)
e
−(−1)
n˜
b
−n
c
−n
[0`
25
¸
µ=0
'0[e
−(−1)
n
˜ α
µ
n
αnµ/n
e
−s(α
µ
−n
αnµ+˜ α
µ
−n
˜ αnµ)
e
−(−1)
n
α
µ
−n
˜ α
−nµ
/n
[0`
¸
(98)
(no summation over µ in the last line). We have used the expressions (4.3.17) for the adjoints of
the raising and lowering operators. The zeromode amplitudes are independent of s, so we won’t
bother with them. The exponentials of the ghost raising operators truncate after the second term:
e
−(−1)
n
b
−n
˜ c
−n
[0` = [0` −(−1)
n
b
−n
˜ c
−n
[0`, (99)
'0[e
−(−1)
n
cn
˜
bn
= '0[ −(−1)
n
'0[c
n
˜
b
n
. (100)
Both terms in (99) are eigenstates of b
−n
c
n
+ ˜ c
−n
˜
b
n
, with eigenvalues of 0 and 2 respectively. The
ﬁrst ghost amplitude is thus:
'0[ −(−1)
n
'0[c
n
˜
b
n
[0` −e
−2sn
(−1)
n
b
−n
˜ c
−n
[0`
= 1 −e
−2sn
. (101)
The second ghost amplitude gives the same result. To evaluate the scalar amplitudes, we must
expand out the [C` exponential:
e
−(−1)
n
α
µ
−n
˜ αnµ/n
[0` =
∞
¸
m=0
1
m!n
m
(−1)
(n+1)m
(α
µ
−n
˜ α
nµ
)
m
[0`. (102)
Each term in the series is an eigenstate of α
µ
−n
α
nµ
+ ˜ α
µ
−n
˜ α
nµ
, with eigenvalue 2nm, and each term
has unit norm, so the amplitude is
∞
¸
m=0
e
−2snm
=
1
1 −e
−2sn
. (103)
Finally, we ﬁnd that the total amplitude (98) is proportional to
e
2s
∞
¸
n=1
(1 −e
−2sn
)
−24
= η(is/π)
−24
. (104)
7 CHAPTER 7 68
This is the sdependent part of the Klein bottle vacuum amplitude, and agrees with the integrand
of (7.4.19).
The vacuum amplitude for the M¨obius strip is
∞
0
ds'B[c
0
b
0
e
−s(L
0
+
˜
L
0
)
[C`. (105)
The only diﬀerence from the above analysis is the absence of the factor (−1)
n
multiplying the bras.
Thus the ghost amplitude (101) becomes instead
'0[ −'0[c
n
˜
b
n
[0` −e
−2sn
(−1)
n
b
−n
˜ c
−n
[0`
= 1 −(−1)
n
e
−2sn
, (106)
while the scalar amplitude (103) becomes
∞
¸
m=0
(−1)
nm
e
−2snm
=
1
1 −(−1)
n
e
−2sn
. (107)
The total amplitude is then
e
2s
∞
¸
n=1
1 −(−1)
n
e
−2sn
−24
= e
2s
∞
¸
n=1
1 +e
−4s(n−1/2)
−24
∞
¸
n=1
1 −e
−4sn
−24
= ϑ
00
(0, 2is/π)
−12
η(2is/π)
−12
, (108)
in agreement with the integrand of (7.4.23).
8 CHAPTER 8 69
8 Chapter 8
8.1 Problem 8.1
(a) The spatial worldsheet coordinate σ
1
should be chosen in the range −π < σ
1
< π for (8.2.21)
to work. In other words, deﬁne σ
1
= −Imln z (with the branch cut for the logarithm on the
negative real axis). The only nonzero commutators involved are
[x
L
, p
L
] = i, [α
m
, α
n
] = mδ
m,−n
. (1)
Hence
[X
L
(z
1
), X
L
(z
2
)]
=−i
α
′
2
ln z
2
[x
L
, p
L
] −i
α
′
2
ln z
1
[p
L
, x
L
] −
α
′
2
¸
m,n, =0
[α
m
, α
n
]
mnz
m
1
z
n
2
=
α
′
2
¸
lnz
2
−ln z
1
−
¸
n,=0
1
n
z
1
z
2
n
¸
=
α
′
2
ln z
2
−ln z
1
+ ln
1 −
z
1
z
2
−ln
1 −
z
2
z
1
=
α
′
2
ln z
2
−ln z
1
+ ln
1 −
z
1
z
2
1 −
z
2
z
1
=
α
′
2
ln z
2
−ln z
1
+ ln
−
z
1
z
2
=
iα
′
2
σ
1
1
−σ
1
2
+ (σ
1
2
−σ
1
1
±π)
. (2)
Because of where we have chosen to put the branch cut for the logarithm, the quantity in the inner
parentheses must be between −π and π. The upper sign is therefore chosen if σ
1
1
> σ
1
2
, and the
lower otherwise. (Note that the fourth equality is legitimate because the arguments of both 1 −
z
1
z
2
and 1 −
z
2
z
1
are in the range (−π/2, π/2).)
(b) Inspection of the above derivation shows that
[X
R
(z
1
), X
R
(z
2
)] = −
πiα
′
2
sign(σ
1
1
−σ
1
2
). (3)
The CBH formula tells us that, for two operators A and B whose commutator is a scalar,
e
A
e
B
= e
[A,B]
e
B
e
A
. (4)
In passing 1
k
L
k
R
(z, ¯ z) through 1
k
′
L
k
′
R
(z
′
, ¯ z
′
), (4) will give signs from several sources. The factors
from the cocyles commuting past the operators e
ik
L
x
L
+ik
R
x
R
and e
ik
′
L
x
L
+ik
′
R
x
R
are given in (8.2.23).
This is cancelled by the factor from commuting the normal ordered exponentials past each other:
e
−(k
L
k
′
L
−k
R
k
′
R
)πiα
′
sign(σ
1
−σ
1
′
)/2
= (−1)
nw
′
+n
′
w
. (5)
8 CHAPTER 8 70
8.2 Problem 8.3
(a) In the sigmamodel action, we can separate X
25
from the other scalars, which we call X
µ
:
S
σ
=
1
4πα
′
M
d
2
σ g
1/2
g
ab
G
µν
+iǫ
ab
B
µν
∂
a
X
µ
∂
b
X
ν
+2
g
ab
G
25µ
+iǫ
ab
B
25µ
∂
a
X
25
∂
b
X
µ
+g
ab
G
25,25
∂
a
X
25
∂
b
X
25
. (6)
(Since we won’t calculate the shift in the dilaton, we’re setting aside the relevant term in the
action.)
(b) We can gauge the X
25
translational symmetry by introducing a worldsheet gauge ﬁeld A
a
:
S
′
σ
=
1
4πα
′
M
d
2
σ g
1/2
g
ab
G
µν
+iǫ
ab
B
µν
∂
a
X
µ
∂
b
X
ν
+ 2
g
ab
G
25µ
+iǫ
ab
B
25µ
(∂
a
X
25
+A
a
)∂
b
X
µ
+g
ab
G
25,25
(∂
a
X
25
+A
a
)(∂
b
X
25
+A
b
)
. (7)
This action is invariant under X
25
→X
25
+λ, A
a
→A
a
−∂
a
λ. In fact, it’s consistent to allow the
gauge parameter λ(σ) to be periodic with the same periodicity as X
25
(making the gauge group a
compact U(1)). This periodicity will be necessary later, to allow us to unwind the string.
(c) We now add a Lagrange multiplier term to the action,
S
′′
σ
=
1
4πα
′
M
d
2
σ g
1/2
g
ab
G
µν
+iǫ
ab
B
µν
∂
a
X
µ
∂
b
X
ν
+ 2
g
ab
G
25µ
+iǫ
ab
B
25µ
(∂
a
X
25
+A
a
)∂
b
X
µ
+g
ab
G
25,25
(∂
a
X
25
+A
a
)(∂
b
X
25
+A
b
) +iφǫ
ab
(∂
a
A
b
−∂
b
A
a
)
. (8)
Integrating over φ forces ǫ
ab
∂
a
A
b
to vanish, which on a topologically trivial worldsheet means that
A
a
is gaugeequivalent to 0, bringing us back to the action (6). Of course, there’s not much point
in making X
25
periodic on a topologically trivial worldsheet, and on a nontrivial worldsheet the
gauge ﬁeld may have nonzero holonomies around closed loops. In order for these to be multiples
of 2πR (and therefore removable by a gauge transformation), φ must also be periodic (with period
2π/R). For details, see Rocek and Verlinde (1992).
(d) Any X
25
conﬁguration is gauge equivalent to X
25
= 0, this condition leaving no additional
gauge degrees of freedom. The action, after performing an integration by parts (and ignoring the
8 CHAPTER 8 71
holonomy issue) is
S
′′′
σ
=
1
4πα
′
M
d
2
σ g
1/2
g
ab
G
µν
+iǫ
ab
B
µν
∂
a
X
µ
∂
b
X
ν
+G
25,25
g
ab
A
a
A
b
+ 2
G
25µ
g
ab
∂
b
X
µ
+iB
25µ
ǫ
ab
∂
b
X
µ
+iǫ
ab
∂
b
φ
A
a
. (9)
We can complete the square on A
a
,
S
′′′
σ
=
1
4πα
′
M
d
2
σ g
1/2
g
ab
G
µν
+iǫ
ab
B
µν
∂
a
X
µ
∂
b
X
ν
+G
25,25
g
ab
(A
a
+G
−1
25,25
B
a
)(A
b
+G
−1
25,25
B
b
) −G
−1
25,25
g
ab
B
a
B
b
, (10)
where B
a
= G
25µ
g
ab
∂
b
X
µ
+iB
25µ
ǫ
ab
∂
b
X
µ
+iǫ
ab
∂
b
φ. Integrating over A
a
, and ignoring the result,
and using the fact that in two dimensions g
ac
ǫ
ab
ǫ
cd
= g
bd
, we have
S
σ
=
1
4πα
′
M
d
2
σ g
1/2
g
ab
G
′
µν
+iǫ
ab
B
′
µν
∂
a
X
µ
∂
b
X
ν
+2
g
ab
G
′
25µ
+iǫ
ab
B
′
25µ
∂
a
φ∂
b
X
µ
+g
ab
G
′
25,25
∂
a
φ∂
b
φ
, (11)
where
G
′
µν
= G
µν
−G
−1
25,25
G
25µ
G
25ν
+G
−1
25,25
B
25µ
B
25ν
,
B
′
µν
= B
µν
−G
−1
25,25
G
25µ
B
25ν
+G
−1
25,25
B
25µ
G
25ν
,
G
′
25µ
= G
−1
25,25
B
25µ
,
B
′
25µ
= G
−1
25,25
G
25µ
,
G
′
25,25
= G
−1
25,25
. (12)
Two features are clearly what we expect to occur upon Tduality: the inversion of G
25,25
, and the
exchange of B
25µ
with G
25µ
, reﬂecting the fact that winding states, which couple to the former,
are exchanged with compact momentum states, which couple to the latter.
8.3 Problem 8.4
The generalization to k dimensions of the Poisson resummation formula (8.2.10) is
¸
n∈Z
k
exp (−πa
mn
n
m
n
n
+ 2πib
n
n
n
) = (det a
mn
)
1/2
¸
m∈Z
k
exp (−πa
mn
(m
m
−b
m
)(m
n
−b
n
)) , (13)
where a
mn
is a symmetric matrix and a
mn
is its inverse. This can be proven by induction using
(8.2.10). The Virasoro generators for the compactiﬁed Xs are
L
0
=
1
4α
′
v
2
L
+
∞
¸
n=1
α
−n
α
n
, (14)
˜
L
0
=
1
4α
′
v
2
R
+
∞
¸
n=1
˜ α
−n
˜ α
n
, (15)
8 CHAPTER 8 72
where products of vectors are taken with the metric G
mn
. The partition function is
(q¯ q)
−1/24
Tr
q
L
0
¯ q
˜
L
0
= [η(τ)[
−2k
¸
n,w
1
∈Z
k
exp
−
πτ
2
α
′
(v
2
+R
2
w
2
1
) + 2πiτ
1
n w
1
. (16)
Using (13) we now get
V
k
Z
X
(τ)
k
¸
w
1
,w
2
∈Z
k
exp
¸
−
πR
2
α
′
τ
2
[w
2
−τw
1
[
2
−2πib
mn
w
n
1
w
m
2
. (17)
This includes the expected phase factor (8.2.12)—but unfortunately with the wrong sign! The
volume factor V
k
= R
k
(det G
mn
)
1/2
comes from the integral over the zeromode.
8.4 Problem 8.5
(a) With l ≡ (n/r +mr/2, n/r −mr/2), we have
l ◦ l
′
= nm
′
+n
′
m. (18)
Evenness of the lattice is obvious. The dual lattice is generated by the vectors (n, m) = (1, 0) and
(0, 1), which also generate the original lattice; hence it is selfdual.
(b) With
l ≡
1
√
2α
′
(v +wR, v −wR), (19)
one can easily calculate
l ◦ l
′
= n w
′
+n
′
w (20)
(in particular, B
mn
drops out). Again, at this point it is more or less obvious that the lattice is
even and selfdual.
8.5 Problem 8.6
The metric (8.4.37) can be written
G
mn
=
α
′
ρ
2
R
2
M
mn
(τ), M(τ) =
1
τ
2
1 τ
1
τ
1
[τ[
2
. (21)
Thus
G
mn
∂
µ
G
np
= ρ
−1
2
∂
µ
ρ
2
δ
m
p
+ (M
−1
∂
µ
M)
m
p
. (22)
The decoupling between τ and ρ is due to the fact that the determinant of M is constant (in fact
it’s 1), so that M
−1
∂
µ
M is traceless:
G
mn
G
pq
∂
µ
G
mp
∂
µ
G
nq
=
2∂
µ
ρ
2
∂
µ
ρ
2
ρ
2
2
+ Tr(M
−1
∂
µ
M)
2
. (23)
8 CHAPTER 8 73
With a little algebra the second term can be shown to equal 2∂
µ
τ∂
µ
¯ τ/τ
2
2
. Meanwhile, the antisym
metry of B implies that G
mn
G
pq
∂
µ
B
mp
∂
µ
B
nq
essentially calculates the determinant of the inverse
metric det G
mn
= (R
2
/α
′
ρ
2
)
2
, multiplying it by 2∂
µ
B
24,25
∂
µ
B
24,25
= 2(α
′
/R
2
)
2
∂
µ
ρ
1
∂
µ
ρ
1
. Adding
this to (23) we arrive at (twice) (8.4.39).
If τ and ρ are both imaginary, then B = 0 and the torus is rectangular with proper radii
R
24
=
G
24,24
R =
α
′
ρ
2
τ
2
, (24)
R
25
=
G
25,25
R =
α
′
ρ
2
τ
2
. (25)
Clearly switching ρ and τ is a Tduality on X
24
, while ρ → −1/ρ is Tduality on both X
24
and
X
25
combined with X
24
↔X
25
.
8.6 Problem 8.7
(a) Since we have already done this problem for the case p = 25 in problem 6.9(a), we can simply
adapt the result from that problem (equation (46) in the solutions to chapter 6) to general p. (In
this case, the ChanPaton factors are trivial, and we must include contributions from all three
combinations of polarizations.) The open string coupling g
o,p
depends on p, and we can compute
it either by Tduality or by comparing to the lowenergy action (8.7.2).
Due to the Dirichlet boundary conditions, there is no zero mode in the path integral and
therefore no momentumconserving delta function in those directions. Except for this fact, the
threetachyon and Veneziano amplitudes calculated in section 6.4 go through unchanged, so we
have
C
D
2
,p
=
1
α
′
g
2
o,p
. (26)
The fourripple amplitude is
S =
2ig
2
o,p
α
′
(2π)
p+1
δ
p+1
(
¸
i
k
i
)
(e
1
e
2
e
3
e
4
F(t, u) +e
1
e
3
e
2
e
4
F(s, u) +e
1
e
4
e
2
e
3
F(s, t)) , (27)
where
F(x, y) ≡
B(−α
′
x + 1, α
′
x +α
′
y −1) +B(−α
′
y + 1, α
′
x +α
′
y −1) +B(−α
′
x + 1, −α
′
y + 1). (28)
If we had instead obtained this amplitude by Tdualizing the answer to problem 6.9(a), using
the fact that κ, and therefore g
2
o,25
, transform according to (8.3.30), we would have found the same
result with g
2
o,p
replaced with g
2
o,25
/(2π
√
α
′
)
25−p
, so we ﬁnd
g
o,p
=
g
o,25
(2π
√
α
′
)
(25−p)/2
. (29)
8 CHAPTER 8 74
(b) To examine the Regge limit, let us rewrite the amplitude (27) in the following way:
S =
2ig
2
o,p
α
′
(2π)
p+1
δ
p+1
(
¸
i
k
i
)
1 −cos πα
′
t + tan
πα
′
s
2
sinπα
′
t
e
1
e
2
e
3
e
4
Γ(α
′
s +α
′
t + 1)Γ(−α
′
t + 1)
Γ(α
′
s + 2)
+e
1
e
3
e
2
e
4
Γ(α
′
s +α
′
t + 1)Γ(−α
′
t −1)
Γ(α
′
s)
+e
1
e
4
e
2
e
3
Γ(α
′
s +α
′
t −1)Γ(−α
′
t + 1)
Γ(α
′
s)
. (30)
The factor with the sines and cosines gives a pole wherever α
′
s is an odd integer, while the last
factor gives the overall behavior in the limit s →∞. In that limit the coeﬃcients of e
1
e
2
e
3
e
4
and
e
1
e
4
e
2
e
3
go like s
α
′
t−1
Γ(−α
′
t +1), while the coeﬃcient of e
1
e
3
e
2
e
4
goes like s
α
′
t+1
Γ(−α
′
t −1),
and therefore dominates (unless e
1
e
3
or e
2
e
4
vanishes). We thus have Regge behavior.
For hard scattering, the amplitude (27) has the same exponential falloﬀ (6.4.19) as the Veneziano
amplitude, since the only diﬀerences are shifts of 2 in the arguments of some of the gamma functions,
which will not aﬀect their asymptotic behavior.
(c) Expanding F(x, y) for small α
′
, ﬁxing x and y, we ﬁnd (with some assistance from Mathe
matica) that the leading term is quadratic:
F(x, y) = −
π
2
α
′
2
2
xy +O(α
′
3
). (31)
Hence the low energy limit of the amplitude (27) is
S ≈ −iπ
2
α
′
g
2
o,p
(2π)
p+1
δ
p+1
(
¸
i
k
i
) (e
1
e
2
e
3
e
4
tu +e
1
e
3
e
2
e
4
su +e
1
e
4
e
2
e
3
st) . (32)
The Dbrane is embedded in a ﬂat spacetime, G
µν
= η
µν
, with vanishing B and F ﬁelds and
constant dilaton. We use a coordinate system on the brane ξ
a
= X
a
, a = 0, . . . , p, so the induced
metric is
G
ab
= η
ab
+∂
a
X
m
∂
b
X
m
, (33)
where the ﬁelds X
m
, m = p+1, . . . , 25, are the ﬂuctuations in the transverse position of the brane,
whose scattering amplitude we wish to ﬁnd. Expanding the action (8.7.2) to quartic order in the
ﬂuctuations, we can use the formula
det(I +A) = 1 + Tr A +
1
2
(Tr A)
2
−
1
2
Tr A
2
+O(A
3
), (34)
to ﬁnd
S
p
= −τ
p
d
p+1
ξ
1 +
1
2
η
ab
∂
a
X
m
∂
b
X
m
+
1
8
(η
ab
η
cd
−2η
ac
η
bd
)∂
a
X
m
∂
b
X
m
∂
c
X
n
∂
d
X
n
. (35)
8 CHAPTER 8 76
upper halfplane is by ﬁxing z
2
on the positive imaginary axis and integrating z
1
from 0 to z
2
. The
amplitude is
S = g
2
c
e
−λ
z
2
0
dz
1
: c
1
e
i(k
1
+q
1
)X
(z
1
, ¯ z
1
) :: c˜ ce
i(k
2
+q
2
)X
(z
2
, ¯ z
2
) :
D
2
,p
, (40)
where, as in problem 8.9, k and q represent the momenta parallel and perpendicular to the Dbrane
respectively. The ghost path integral is
1
2
(c(z
1
) + ˜ c(¯ z
1
))c(z
2
)˜ c(¯ z
2
)
D
2
=
C
g
D
2
2
((z
1
−z
2
)(z
1
− ¯ z
2
) + (¯ z
1
−z
2
)(¯ z
1
− ¯ z
2
)) (z
2
− ¯ z
2
)
= 2C
g
D
2
(z
1
−z
2
)(z
1
+z
2
)z
2
. (41)
To evaluate the X path integral we use the result of problem 8.9(a):
: e
i(k
1
+q
1
)X(z
1
,¯ z
1
)
:: e
i(k
2
+q
2
)X(z
2
,¯ z
2
)
:
D
2
,p
= iC
X
D
2
,p
(2π)
p+1
δ
p+1
(k
1
+k
2
)
[z
1
− ¯ z
1
[
α
′
(k
2
1
−q
2
1
)/2
[z
2
− ¯ z
2
[
α
′
(k
2
2
−q
2
2
)/2
[z
1
−z
2
[
α
′
(k
1
k
2
+q
1
q
2
)
[z
1
− ¯ z
2
[
α
′
(k
1
k
2
−q
1
q
2
)
= iC
X
D
2
,p
(2π)
p+1
δ
p+1
(k
1
+k
2
)
2
2α
′
k
2
−4
[z
1
[
α
′
k
2
−2
[z
2
[
α
′
k
2
−2
[z
1
−z
2
[
−α
′
s/2−4
[z
1
+z
2
[
α
′
s/2−2k
2
+4
. (42)
We have used the kinematic relations k
1
+k
2
= 0 and (k
1
+ q
1
)
2
= (k
2
+ q
2
)
2
= 4/α
′
, and deﬁned
the parameters
k
2
≡ k
2
1
= k
2
2
= 4 −q
2
1
= 4 −q
2
2
,
s ≡ −(q
1
+q
2
)
2
. (43)
So we have
S = −g
2
c
C
D
2
,p
(2π)
p+1
δ
p+1
(k
1
+k
2
)
2
2α
′
k
2
−3
[z
2
[
α
′
k
2
−1
z
2
0
dz
1
[z
1
[
α
′
k
2
−2
[z
1
−z
2
[
−α
′
s/2−3
[z
1
+z
2
[
α
′
s/2−2k
2
+5
= −ig
2
c
C
D
2
,p
(2π)
p+1
δ
p+1
(k
1
+k
2
)2
2α
′
k
2
−3
1
0
dxx
α
′
k
2
−2
(1 −x)
−α
′
s/2−3
(1 +x)
α
′
s/2−2k
2
+5
= −i
π
3/2
(2π
√
α
′
)
11−p
g
c
32
(2π)
p+1
δ
p+1
(k
1
+k
2
)B(α
′
k
2
−1, −α
′
s/4 −1). (44)
In the last line we have used (26), (29), and (8.7.28) to calculate g
2
c
C
D
2
,p
.
(b) In the Regge limit we increase the scattering energy, k
2
→ −∞, while holding ﬁxed the
momentum transfer s. As usual, the beta function in the amplitude give us Regge behavior:
S ∼ (−k
2
)
α
′
s/4+1
Γ(−α
′
s/4 −1). (45)
In the hard scattering limit we again take k
2
→−∞, but this time ﬁxing k
2
/s. As in the Veneziano
amplitude, the beta function gives exponential behavior in this limit.
8 CHAPTER 8 77
(c) The beta function in the amplitude has poles at α
′
k
2
= 1, 0, −1, . . . , representing onshell
intermediate open strings on the Dbrane: the closed string is absorbed and then later reemitted
by the Dbrane. These poles come from the region of the integral in (44) where z
1
approaches the
boundary.
Note that there are also poles at s = 0 and s = −4 (the poles at positive s are kinematically
forbidden), representing an onshell intermediate closed string: the tachyon “decays” into another
tachyon and either a massless or a tachyonic closed string, and the latter is then absorbed (com
pletely, without producing open strings) by the Dbrane. These poles come from the region of the
integral in (44) where z
1
approaches z
2
.
9 APPENDIX A 78
9 Appendix A
9.1 Problem A.1
(a) We proceed by the same method as in the example on pages 339341. Our orthonormal basis
for the periodic functions on [0, U] will be:
f
0
(u) =
1
√
U
,
f
j
(u) =
2
U
cos
2πju
U
, (1)
g
j
(u) =
2
U
sin
2πju
U
,
where j runs over the positive integers. These are eigenfunctions of ∆ = −∂
2
u
+ω
2
with eigenvalues
λ
j
=
2πj
U
2
+ω
2
. (2)
Hence (neglecting the counterterm action)
Tr exp(−
ˆ
HU) =
[dq]
P
exp(−S
E
)
=
det
P
∆
2π
−1/2
=
2π
λ
0
∞
¸
j=1
2π
λ
j
=
√
2π
ω
∞
¸
j=1
2πU
2
4π
2
j
2
+ω
2
U
2
. (3)
This inﬁnite product vanishes, so we regulate it by dividing by the same determinant with ω →Ω:
Ω
ω
∞
¸
j=1
1 +
ΩU
2πj
2
1 +
ωU
2πj
2
=
sinh
1
2
ΩU
sinh
1
2
ωU
. (4)
For large Ω this becomes
e
ΩU/2
2 sinh
1
2
ωU
; (5)
the divergence can easily be cancelled with a counterterm Lagrangian L
ct
= Ω/2, giving
Tr exp(−
ˆ
HU) =
1
2 sinh
1
2
ωU
. (6)
The eigenvalues of
ˆ
H are simply
E
i
= (i +
1
2
)ω, (7)
9 APPENDIX A 79
for nonnegative integer i, so using (A.1.32) gives
Tr exp(−
ˆ
HU) =
∞
¸
i=0
exp(−E
i
U)
= e
−ωU/2
∞
¸
i=0
e
−iωU
=
1
2 sinh
1
2
ωU
. (8)
To be honest, the overall normalization of (6) must be obtained by comparison with this result.
(b) Our basis for the antiperiodic functions on [0, U] consists of the eigenfunctions of ∆,
f
j
(u) =
2
U
cos
2π(j +
1
2
)u
U
, (9)
g
j
(u) =
2
U
sin
2π(j +
1
2
)u
U
,
where the j are nonnegative integers, with eigenvalues
λ
j
=
2π(j +
1
2
)
U
2
+ω
2
. (10)
Before including the counterterm and regulating, we have
Tr
exp(−
ˆ
HU)
ˆ
R
=
∞
¸
j=0
2πU
2
(2π(j +
1
2
))
2
+ (ωU)
2
. (11)
After including the counterterm action and dividing by the regulator, this becomes
e
−LctU
∞
¸
j=0
1 +
ΩU
2π(j+
1
2
)
2
1 +
ωU
2π(j+
1
2
)
2
= e
−LctU
cosh
1
2
ΩU
cosh
1
2
ωU
∼
e
(Ω/2−Lct)U
2 cosh
1
2
ωU
, (12)
so the answer is
Tr
exp(−
ˆ
HU)
ˆ
R
=
1
2 cosh
1
2
ωU
. (13)
This result can easily be reproduced by summing over eigenstates of
ˆ
H with weight (−1)
R
, since
the even i eigenstates are also even under reﬂection, and odd i eigenstates odd under reﬂection:
Tr
exp(−
ˆ
HU)
ˆ
R
=
∞
¸
i=0
(−1)
i
e
−(j+1/2)ωU
=
1
2 cosh
1
2
ωU
. (14)
9 APPENDIX A 80
9.2 Problem A.3
The action can be written
S =
1
4πα
′
d
2
σX(−∂
2
1
−∂
2
2
+m
2
)X
=
1
2
d
2
σX∆X, (15)
∆ =
1
2πα
′
(−∂
2
1
−∂
2
2
+m
2
). (16)
The periodic eigenfunctions of ∆ can be given in a basis of products of periodic eigenfunction of
−∂
2
1
on σ
1
with periodic eigenfunctions of −∂
2
2
on σ
2
:
F
jk
(σ
1
, σ
2
) = f
j
(σ
1
)g
k
(σ
2
), (17)
f
0
(σ
1
) =
1
√
2π
,
f
j
(σ
1
) =
1
√
π
sin
cos
¸
jσ
1
, j = 1, 2, . . . ,
g
0
(σ
2
) =
1
√
T
,
g
k
(σ
2
) =
1
√
2T
sin
cos
¸
kσ
2
, k = 1, 2, . . . .
The eigenvalues are
λ
jk
=
1
2πα
′
j
2
+
2πk
T
2
+m
2
, (18)
with multiplicity n
j
n
k
, where n
0
= 1 and n
i
= 2 (i = 1, 2, . . . ). The path integral is
det
P
∆
2π
−1/2
=
∞
¸
j,k=0
2π
λ
jk
n
j
n
k
/2
=
∞
¸
j=0
∞
¸
k=0
2π
λ
jk
n
k
/2
n
j
=
∞
¸
j=0
1
2 sinh
1
2
j
2
+m
2
T
n
j
, (19)
where in the last step we have used the result of problem A.1. The inﬁnite product vanishes; it
would have to be regulated and a counterterm introduced to extract the ﬁnite part.
9.3 Problem A.5
We assume that the Hamiltonian for this system is
H = mχψ. (20)
9 APPENDIX A 81
The periodic trace is
Tr
(−1)
ˆ
F
exp(−
ˆ
HU)
=
dψ'ψ, U[ψ, 0`
E
=
[dχdψ] exp
¸
U
0
du(−χ∂
u
ψ −H)
=
[dχdψ] exp
¸
U
0
duχ∆ψ
, (21)
where
∆ = −∂
u
−m. (22)
The periodic eigenfunctions of ∆ on [0, U] are
f
j
(u) =
1
√
U
e
2πiju/U
, (23)
while the eigenfunctions of ∆
T
= ∂
u
−m are
g
j
(u) =
1
√
U
e
−2πiju/U
, (24)
with j running over the integers. Their eigenvalues are
λ
j
= −
2πij
U
−m
, (25)
so the trace becomes
[dχdψ] exp
¸
U
0
duχ∆ψ
=
∞
¸
j=−∞
λ
j
=
∞
¸
j=−∞
−
2πij
U
+m
= −m
∞
¸
j=1
2πj
U
2
+m
2
. (26)
This is essentially the inverse of the inﬁnite product that was considered in problem A.1(a) (eq.
(3)). Regulating and renormalizing in the same manner as in that problem yields:
Tr
(−1)
ˆ
F
exp(−
ˆ
HU)
= 2 sinh
1
2
mU. (27)
This answer can very easily be checked by explicit calculation of the LHS. The Hamiltonian operator
can be obtained from the classical Hamiltonian by ﬁrst antisymmetrizing on χ and ψ:
H = mχψ =
1
2
m(χψ −ψχ) (28)
9 APPENDIX A 82
−→
ˆ
H =
1
2
m( ˆ χ
ˆ
ψ −
ˆ
ψˆ χ). (29)
Now
ˆ
H[ ↑` =
1
2
[ ↑`, (30)
ˆ
H[ ↓` = −
1
2
[ ↓`,
so, using A.2.22,
Tr
(−1)
ˆ
F
exp(−
ˆ
HU)
= e
mU/2
−e
−mU/2
, (31)
in agreement with (27).
The antiperiodic trace is calculated in the same way, the only diﬀerence being that the index
j runs over the halfintegers rather than the integers in order to make the eigenfunctions (23) and
(24) antiperiodic. Eq. (26) becomes
[dχdψ] exp
¸
U
0
duχ∆ψ
=
∞
¸
j=1/2,3/2,...
2πj
U
2
+m
2
. (32)
When we regulate the product, it becomes
¸
j
2πj
U
2
+m
2
2πj
U
2
+M
2
=
¸
j
1 +
mU
2πj
2
¸
j
1 +
MU
2πj
2
=
cosh
1
2
mU
cosh
1
2
MU
. (33)
With the same counterterm Lagrangian as before to cancel the divergence in the denominator as
M →∞, we are simply left with
Tr exp(−
ˆ
HU) = 2 cosh
1
2
mU, (34)
the same as would be found using (30).
10 CHAPTER 10 83
10 Chapter 10
10.1 Problem 10.1
(a) The OPEs are:
T
F
(z)X
µ
(0) ∼ −i
α
′
2
ψ
µ
(0)
z
, T
F
(z)ψ
µ
(0) ∼ i
α
′
2
∂X
µ
(0)
z
. (1)
(b) The result follows trivially from (2.3.11) and the above OPEs.
10.2 Problem 10.2
(a) We have
δ
η
1
δ
η
2
X = δ
η
1
(η
2
ψ +η
∗
2
˜
ψ) = −η
2
η
1
∂X −η
∗
2
η
∗
1
¯
∂X, (2)
so, using the anticommutativity of the η
i
,
[δ
η
1
, δ
η
2
]X = 2η
1
η
2
∂X + 2η
∗
1
η
∗
2
¯
∂X = δ
v
X (3)
(see (2.4.7)). For ψ we must apply the equation of motion ∂
˜
ψ = 0:
δ
η
1
δ
η
2
ψ = δ
η
1
(−η
2
∂X) = −η
2
∂(η
1
ψ +η
∗
1
˜
ψ) = −η
2
η
1
∂ψ −η
2
∂η
1
ψ, (4)
so
[δ
η
1
, δ
η
2
]ψ = −v∂ψ −
1
2
∂v ψ = δ
v
ψ, (5)
the second term correctly reproducing the weight of ψ. The
˜
ψ transformation works out similarly.
(b) For X:
δ
η
δ
v
X = −vη∂ψ −v∂η ψ −v
∗
η
∗
¯
∂
˜
ψ −v
∗
(∂η)
∗
˜
ψ, (6)
δ
v
δ
η
X = −vη∂ψ −
1
2
η∂vψ −v
∗
η
∗
¯
∂
˜
ψ −
1
2
η
∗
(∂v)
∗
˜
ψ, (7)
so
[δ
η
, δ
v
]X = δ
η
′ X, (8)
where
η
′
= −v∂η +
1
2
∂v η. (9)
For ψ,
δ
η
δ
v
ψ = v∂η∂X +vη∂
2
X +
1
2
η∂v∂X, (10)
δ
v
δ
η
ψ = ηv∂
2
X +η∂v∂X, (11)
so
[δ
η
, δ
v
]ψ = δ
η
′ ψ, (12)
and similarly for
˜
ψ.
10 CHAPTER 10 84
10.3 Problem 10.3
(a) Since the OPE of T
X
B
= −(1/α
′
)∂X
µ
∂X
µ
and T
ψ
B
= −(1/2)ψ
µ
∂ψ
µ
is nonsingular,
T
B
(z)T
B
(0) ∼ T
X
B
(z)T
X
B
(0) +T
ψ
B
(z)T
ψ
B
(z), (13)
which does indeed reproduce (10.1.13a). We then have
T
B
(z)T
F
(0)
i
2/α
′
∼
1
z
2
∂X
µ
(z)ψ
µ
(0) +
1
2z
∂
µ
(z)∂ψ
µ
(z)∂X
µ
(0) +
1
2z
2
ψ
µ
(z)∂X
µ
(0)
∼
3
2z
2
ψ
µ
∂X
µ
(0) +ψ
µ
∂
2
X
µ
(0) +
1
z
∂ψ
µ
∂X
µ
(0),
T
F
(z)T
F
(0) ∼
D
z
3
−
2
α
′
z
∂X
µ
(z)∂X
µ
(0) +
1
z
2
ψ
µ
(z)ψ
µ
(0).
(14)
(b) Again, there is no need to check the T
B
T
B
OPE, since that is simply the sum of the X part
and the ψ part. The new terms in T
B
and T
F
add two singular terms to their OPE, namely
V
µ
∂
2
X
µ
(z)i
2
α
′
ψ
ν
∂X
ν
(0) +
1
2
ψ
µ
∂ψ
µ
(z)i
√
2α
′
V
ν
∂ψ
ν
(0)
∼
i
√
2α
′
z
3
V
µ
ψ
µ
(0) −i
2
α
1
z
2
V
µ
∂ψ
µ
(z) −
i
√
2α
z
3
V
µ
ψ
µ
(z)
∼ −3i
α
′
2
1
z
2
V
µ
∂ψ
µ
(0) −
i
√
2α
′
z
V
µ
∂
2
ψ
µ
(0),
(15)
which are precisely the extra terms we expect on the right hand side of (10.1.1b). The new terms
in the T
F
T
F
OPE are
2ψ
µ
∂X
µ
(z)V
ν
∂ψ
ν
(0) + 2V
µ
∂ψ
µ
(z)ψ
ν
∂X
ν
(0) −2α
′
V
µ
∂ψ
µ
(z)V
ν
∂ψ
ν
(0)
∼
2
z
2
V
µ
∂X
µ
(z) −
2
z
2
V
µ
∂X
µ
(0) +
4α
′
V
2
z
3
∼
2
z
V
µ
∂
2
X
µ
(0) +
4α
′
V
2
z
3
.
(16)
10.4 Problem 10.4
The [L
m
, L
n
] commutator (10.2.11a) is as in the bosonic case. The current associated with the
charge ¦G
r
, G
s
¦ is, according to (2.6.14),
Res
z
1
→z
2
z
r+1/2
1
T
F
(z
1
)z
s+1/2
2
T
F
(z
2
) = Res
z
12
→0
z
r+s+1
2
1 +
z
12
z
2
r+1/2
2c
3z
3
12
+
2
z
12
T
B
(z
2
)
=
(4r
2
−1)c
12
z
r+s−1
2
+ 2z
r+s+1
2
T
B
(z
2
).
(17)
10 CHAPTER 10 85
¦G
r
, G
s
¦ is in turn the residue of this expression in z
2
, which is easily seen to equal the RHS of
(10.2.11b). Similarly, for [L
m
, G
r
] we have
Res
z
1
→z
2
z
m+1
1
T
B
(z
1
)z
r+1/2
2
T
F
(z
2
)
= Res
z
12
→0
z
r+m+3/2
2
1 +
z
12
z
2
m+1
3
2z
2
12
T
F
(z
2
) +
1
z
12
∂T
F
(z
2
)
=
3(m + 1)
2
z
r+m+1/2
2
T
F
(z
2
) +z
r+m+3/2
2
∂T
F
(z
2
).
(18)
The residue in z
2
of this is
3(m + 1)
2
G
r+m
−
r +m+
3
2
G
r+m
, (19)
in agreement with (10.2.11c).
10.5 Problem 10.5
Let us denote by c
B
the central charge appearing in the T
B
T
B
OPE, and by c
F
that appearing in
the T
F
T
F
OPE. One of the Jacobi identies for the superconformal generators is, using (10.2.11),
0 = [L
m
, ¦G
r
, G
s
¦] +¦G
r
, [G
s
, L
m
]¦ −¦G
s
, [L
m
, G
r
]¦
=
1
6
c
B
(m
3
−m) +
c
F
4
(2s −m)(4r
2
−1) + (2r −m)(4s
2
−1)
δ
m+r+s,0
=
1
6
(c
B
−c
F
)(m
3
−m)δ
m+r+s,0
.
(20)
Hence c
B
= c
F
.
10.6 Problem 10.7
Taking (for example) z
1
to inﬁnity while holding the other z
i
ﬁxed at ﬁnite values, the expectation
value (10.3.7) goes like z
−1
1
. Since e
iǫ
1
H(z
1
)
is a tensor of weight 1/2, transforming to the u = 1/z
1
frame this expectation value becomes constant, O(1), in the limit u → 0. This is the correct
behavior—the only poles and zeroes of the expectation value should be at the positions of the other
operators. If we now consider some other function, with exactly the same poles and zeroes and
behavior as z
1
→ ∞, the ratio between this function and the one given in (10.3.7) would have to
be an entire function which approaches a constant as z
1
→ ∞. But the only such function is a
constant, so (applying the same argument to the dependence on all the z
i
) the expression in (10.3.7)
is unique up to a constant.
10 CHAPTER 10 86
10.7 Problem 10.10
On the bosonic side, the energy eigenvalue in terms of the momentum k
L
and oscillator occupation
numbers N
n
is
L
0
=
1
2
k
2
L
+
∞
¸
n=1
nN
n
. (21)
On the fermionic side, there are two sets of oscillators, generated by the ﬁelds ψ and
¯
ψ, and the
energy is
L
0
=
∞
¸
n=1
n −
1
2
(N
n
+
¯
N
n
). (22)
We will denote states on the bosonic side by (k
L
, N
1
, N
2
), and on the fermionic side by (N
1
+
¯
N
1
, N
2
+
¯
N
2
, N
3
+
¯
N
3
). We won’t need any higher oscillators for this problem. On the fermionic
side N
n
+
¯
N
n
can take the values 0, 1, or 2, with degeneracy 1, 2, and 1 respectively. Here are the
states with L
0
= 0, 1/2, . . . , 5/2 on each side:
L
0
(k
L
, N
1
, N
2
) (N
1
+
¯
N
1
, N
2
+
¯
N
2
, N
3
+
¯
N
3
)
0 (0, 0, 0) (0, 0, 0)
1/2 (±1, 0, 0) (1, 0, 0)
1 (0, 1, 0) (2, 0, 0)
3/2 (±1, 1, 0) (0, 1, 0)
2 (±2, 0, 0) (1, 1, 0)
(0, 2, 0)
(0, 0, 1)
5/2 (±1, 2, 0) (2, 1, 0)
(±1, 0, 1) (0, 0, 1)
(23)
10.8 Problem 10.11
We will use the ﬁrst of the suggested methods. The OPE we need is
: e
iH(z)
:: e
inH(0)
:= z
n
: e
i(n+1)H(0)
: +O(z
n+1
). (24)
Hence
ψ(z)F
n
(0) = z
n
F
n+1
(0) +O(z
n+1
). (25)
It’s easy to see that this is satisﬁed by
F
n
=:
n−1
¸
i=0
1
i!
∂
i
ψ : . (26)
The OPE of ψ(z) with F
n
(0) is nonsingular, but as we Taylor expand ψ(z), all the terms vanish
until the nth one because ψ is fermionic. F
n
and e
inH
obviously have the same fermion number
10 CHAPTER 10 87
n. Their dimensions work out nicely: for e
inH
we have n
2
/2; for F
n
we have n ψs and n(n −1)/2
derivatives, for a total of n/2 +n(n −1)/2 = n
2
/2.
For e
−inH
we obviously just replace ψ with
¯
ψ.
10.9 Problem 10.14
We start with the NS sector. The most general massless state is
[ψ` = (e ψ
−1/2
+fβ
−1/2
+gγ
−1/2
)[0; k`
NS
, (27)
where k
2
= 0 (by the L
0
condition) and [0; k`
NS
is annihilated by b
0
. The BRST charge acting on
[ψ` is
Q
B
[ψ` = (c
0
L
0
+γ
−1/2
G
m
1/2
+γ
1/2
G
m
−1/2
)[ψ`
=
√
2α
′
(e kγ
−1/2
+fk ψ
−1/2
)[0; k`
NS
. (28)
The L
0
term of course vanishes, along with many others we have not indicated. For [ψ` to be
closed requires e k = f = 0. Furthermore exactness of (28) implies g
∼
= g +
√
2α
′
e
′
k for any e
′
(so we might as well set g = 0), while e
∼
= e +
√
2α
′
f
′
k for any f
′
. We are left with the 8 transverse
polarizations of a massless vector.
The R case is even easier: all of the work is done by the constraint (10.5.26), and none by the
BRST operator. The general massless state is
[ψ` = [u; k`
R
, (29)
where u is a 10dimensional Dirac spinor, and k
2
= 0. [ψ` is deﬁned to be annihilated by b
0
and
β
0
, which implies that it is annihilated by G
g
0
. According to (10.5.26), this in turn implies that it
is annihilated by G
m
0
, which is exactly the OCQ condition. The only thing left to check is that all
the states satisfying these conditions are BRST closed, which follows more or less trivially from all
the above conditions. Finally, since they are all closed, none of them can be exact.
11 CHAPTER 11 88
11 Chapter 11
11.1 Problem 11.1
In order to establish the normalizations, we ﬁrst calculate the T
+
F
T
−
F
OPE:
e
+i
√
3H(z)
e
−i
√
3H(0)
∼
1
z
3
+
i
√
3∂H(0)
z
2
−
3∂H(0)∂H(0)
2z
+
i2
√
3∂
2
H(0)
z
. (1)
Since the third term is supposed to be 2T
B
(0)/z, where
T
B
= −
1
2
∂H∂H, (2)
we need
T
±
F
=
2
3
e
±i
√
3H
. (3)
It follows from the ﬁrst term that c = 1 (as we already knew), and from the second that
j =
i
√
3
∂H. (4)
It is now straightforward to verify each of the OPEs in turn. The T
B
T
±
F
and T
B
j OPEs are
from Chapter 2, and show that T
±
F
and j have weight
3
2
and 1 respectively. The fact that the T
±
F
T
±
F
OPE is nonsingular was also shown in Chapter 2. For the jT
±
F
OPE we have
j(z)T
±
F
(0) = ±
2
3
e
±i
√
3H(0)
z
. (5)
Finally,
j(z)j(0) =
1
3z
2
. (6)
11.2 Problem 11.3
See the last paragraph of section 11.2.
11.3 Problem 11.4
(a) Dividing the 32 leftmoving fermions into two groups of 16, the untwisted theory contains 8
sectors:
¦(+ + +), (− −+), (+ −−), (−+−)¦ ¦(NS, NS, NS), (R, R, R)¦, (7)
where the ﬁrst symbol in each triplet corresponds to the ﬁrst 16 leftmoving fermions, the second
to the second 16, and the third to the 8 rightmoving fermions. If we now twist by exp(πiF
1
), we
project out (− − +) and (− + −), but project in the twisted sectors (R, NS, NS) and (NS, R, R).
We again have 8 sectors:
¦(+ + +), (+−−)¦ ¦(NS, NS, NS), (R, R, R), (R, NS, NS), (NS, R, R)¦, (8)
11 CHAPTER 11 89
Let us ﬁnd the massless spacetime bosons ﬁrst, to establish the gauge group. These will have right
movers in the NS+ sector, so there are two possibilities, (NS+, NS+, NS+) and (R+, NS+, NS+),
and the states are easily enumerated (primes refer to the second set of leftmoving fermions):
α
i
−1
˜
ψ
j
−1/2
[0
NS
, 0
NS
, 0
NS
` graviton, dilaton, antisymmetric tensor;
λ
A
′
−1/2
λ
B
′
−1/2
˜
ψ
i
−1/2
[0
NS
, 0
NS
, 0
NS
` adjoint of SO(16)
′
;
λ
A
−1/2
λ
B
−1/2
˜
ψ
i
−1/2
[0
NS
, 0
NS
, 0
NS
` adjoint of SO(16);
˜
ψ
i
−1/2
[u
R
, 0
NS
, 0
NS
` 128 of SO(16). (9)
The last two sets combine to form an adjoint of E
8
, so the gauge group is E
8
SO(16). There are
similarly two sectors containing massless spacetime fermions, (NS+, R+, R+) and (NS+, R−, R−)
(the (R, R, R) states are all massive due to the positive normalordering constant for the leftmoving
R fermions); these will give respectively (8, 1, 128) and (8
′
, 1, 128
′
) of SO(16)
spin
E
8
SO(16).
Finally, the tachyon must be in (NS+, NS−, NS−); the only states are λ
A
′
−1/2
[0
NS
, 0
NS
, 0
NS
`, which
transform as (1, 1, 8
v
).
(b) Dividing the leftmoving fermions into four groups of 8, if we twist the above theory by the
total fermion number of the ﬁrst and third groups, we get a total of 32 sectors:
¦(+ + + + +), (− −−−+), (+ + +−−), (− −−+−)¦
¦(NS, NS, NS, NS, NS), (NS, R, R, NS, NS), (R, NS, R, NS, NS), (R, R, NS, NS, NS),
(NS, NS, R, R, R), (NS, R, NS, R, R), (R, NS, NS, R, R), (R, R, R, R, R)¦. (10)
Again we begin by listing the massless spacetime bosons, together with their SO(8)
spin
SO(8)
1
SO(8)
2
SO(8)
3
SO(8)
4
quantum numbers:
(NS+, NS+, NS+, NS+, NS+) :
α
i
−1
˜
ψ
j
−1/2
[0
NS
, 0
NS
, 0
NS
, 0
NS
, 0
NS
` (1 ⊕28 ⊕35, 1, 1, 1, 1)
λ
A
4
−1/2
λ
B
4
−1/2
˜
ψ
i
−1/2
[0
NS
, 0
NS
, 0
NS
, 0
NS
, 0
NS
` (8
v
, 1, 1, 1, 28)
λ
A
1
−1/2
λ
B
1
−1/2
˜
ψ
i
−1/2
[0
NS
, 0
NS
, 0
NS
, 0
NS
, 0
NS
` (8
v
, 28, 1, 1, 1)
λ
A
2
−1/2
λ
B
2
−1/2
˜
ψ
i
−1/2
[0
NS
, 0
NS
, 0
NS
, 0
NS
, 0
NS
` (8
v
, 1, 28, 1, 1)
λ
A
3
−1/2
λ
B
3
−1/2
˜
ψ
i
−1/2
[0
NS
, 0
NS
, 0
NS
, 0
NS
, 0
NS
` (8
v
, 1, 1, 28, 1)
(NS+, R+, R+, NS+, NS+) :
˜
ψ
i
−1/2
[0
NS
, u
R
, v
R
, 0
NS
, 0
NS
` (8
v
, 1, 8, 8, 1)
(R+, NS+, R+, NS+, NS+) :
˜
ψ
i
−1/2
[u
R
, 0
NS
, v
R
, 0
NS
, 0
NS
` (8
v
, 8, 1, 8, 1)
(R+, R+, NS+, NS+, NS+) :
˜
ψ
i
−1/2
[u
R
, v
R
, 0
NS
, 0
NS
, 0
NS
` (8
v
, 8, 8, 1, 1). (11)
11 CHAPTER 11 90
The ﬁrst set of states gives the dilaton, antisymmetric tensor, and graviton. The next set gives the
SO(8)
4
gauge bosons. The rest combine to give gauge bosons of SO(24), once we perform triality
rotations on SO(8)
1,2,3
to turn the bispinors into bivectors.
The massless spacetime fermions are:
(R+, NS+, NS+, R+, R+) : [u
R
, 0
NS
, 0
NS
, v
R
, w
R
` (8, 8, 1, 1, 8)
(NS+, R+, NS+, R+, R+) : [0
NS
, u
R
, 0
NS
, v
R
, w
R
` (8, 1, 8, 1, 8)
(NS+, NS+, R+, R+, R+) : [0
NS
, 0
NS
, u
R
, v
R
, w
R
` (8, 1, 1, 8, 8). (12)
The triality rotation again turns the SO(8)
1,2,3
spinors into vectors, which combine into an SO(24)
vector.
Finally, the tachyon is an SO(8)
4
vector but is neutral under SO(24):
(NS+, NS+, NS+, NS−, NS−) : λ
A
4
−1/2
[0
NS
, 0
NS
, 0
NS
, 0
NS
` (1, 1, 1, 1, 8
v
). (13)
11.4 Problem 11.7
We wish to show that the state j
a
−1
j
a
−1
[0` corresponds to the operator : jj(0) :. Since j
a
−1
[0` clearly
corresponds to j
a
(0), and j
a
−1
=
dz/(2πi)j
a
(z)/z, we have
: jj(0) : =
dz
2πi
j
a
(z)j
a
(0)
z
, (14)
where the contour goes around the origin. The contour integral picks out the z
0
term in the Laurent
expansion of j
a
(z)j
a
(0), which is precisely (11.5.18).
We will now use this to prove the ﬁrst line of (11.5.20), with z
1
= 0 and z
3
= z. Using the
Laurent expansion (11.5.2) we have
: jj(0) : j
c
(z)
∼
=
∞
¸
m=−∞
1
z
m+1
j
c
m
j
a
−1
j
a
−1
[0`. (15)
Only three terms in this sum are potentially interesting; the rest are either nonsingular (for m < 0)
or zero (for m > 2, since the total level of the state would be negative). In fact, the term with
m = 2 must also vanish, since (being at level 0) it can only be proportional to the ground state [0`,
leaving no room for the free Lie algebra index on (15); this can also be checked explicitly. For the
other two terms we have:
j
c
0
j
a
−1
j
a
−1
[0` = if
cab
(j
a
−1
j
b
−1
+j
b
−1
j
a
−1
)[0` = 0, (16)
j
c
1
j
a
−1
j
a
−1
[0` = (
ˆ
kδ
ca
+if
cab
j
b
0
+j
a
−1
j
c
1
)j
a
−1
[0`
= (2
ˆ
kj
c
−1
−f
cab
f
bad
j
d
−1
)[0`
= (k +h(g))ψ
2
j
c
−1
[0`. (17)
11 CHAPTER 11 91
The RHS of (17) clearly corresponds to the RHS of (11.5.20).
To check the TT OPE (11.5.24), we employ the same strategy, using the Laurent coeﬃcients
(11.5.26). The OPE will be the operator corresponding to
1
(k +h(g))ψ
2
1
z
4
L
2
+
1
z
3
L
1
+
1
z
2
L
0
+
1
z
L
−1
j
a
−1
j
a
−1
[0`; (18)
terms with L
m
are nonsingular for m < −1 and vanish for m > 2. Life is made much easier by the
fact that j
b
m
j
a
−1
j
a
−1
[0` = 0 for m = 0 and m > 1, and the value for m = 1 is given by (17) above.
Thus:
L
2
j
a
−1
j
a
−1
[0` =
1
(k +h(g))ψ
2
j
b
1
j
b
1
j
a
−1
j
a
−1
[0` = j
b
1
j
b
−1
[0` =
ˆ
kdim(g)[0`,
L
1
j
a
−1
j
a
−1
[0` = 0,
L
0
j
a
−1
j
a
−1
[0` =
2
(k +h(g))ψ
2
j
b
−1
j
b
1
j
a
−1
j
a
−1
[0` = 2j
b
−1
j
b
−1
[0`,
L
−1
j
a
−1
j
a
−1
[0` =
2
(k +h(g))ψ
2
j
b
−2
j
b
1
j
a
−1
j
a
−1
[0` = 2j
b
−2
j
b
−1
[0`. (19)
All of these states are easily translated back into operators, the only slightly nontrivial one being
the last. From the Laurent expansion we see that the state corresponding to ∂T
s
B
(0) is indeed
L
−3
[0` = 2/((k +h(g))ψ
2
)j
b
−2
j
b
−1
[0`. Thus we have precisely the OPE (11.5.24).
11.5 Problem 11.8
The operator : jj(0) : is deﬁned to be the z
0
term in the Laurent expansion of j
a
(z)j
a
(0). First let
us calculate the contribution from a single current iλ
A
λ
B
(A = B):
iλ
A
(z)λ
B
(z)iλ
A
(0)λ
B
(0) = λ
A
(z)λ
A
(0)λ
B
(z)λ
B
(0) (no sum)
=: λ
A
(z)λ
A
(0)λ
B
(z)λ
B
(0) : +
1
z
: λ
A
(z)λ
A
(0) : +
1
z
: λ
B
(z)λ
B
(0) : +
1
z
2
. (20)
Clearly the order z
0
term is : ∂λ
A
λ
A
: + : ∂λ
B
λ
B
:. Summing over all A and B with B = A double
counts the currents, so we divide by 2:
: jj : = (n −1) : ∂λ
A
λ
A
: (sum). (21)
Finally, we have k = 1, h(SO(n)) = n −2, and ψ
2
= 2, so
T
s
B
=
1
2
: ∂λ
A
λ
A
: (sum). (22)
11 CHAPTER 11 92
11.6 Problem 11.9
In the notation of (11.6.5) and (11.6.6), the lattice Γ is Γ
22,6
, i.e. the set of points of the form
(n
1
, . . . , n
28
) or (n
1
+
1
2
, . . . , n
28
+
1
2
),
¸
i
n
i
∈ 2Z (23)
for any integers n
i
. It will be convenient to divide Γ into two sublattices, Γ = Γ
1
∪ Γ
2
where
Γ
1
= ¦(n
1
, . . . , n
28
) :
¸
n
i
∈ 2Z¦,
Γ
2
= Γ
1
+l
0
, l
0
≡ (
1
2
, . . . ,
1
2
). (24)
Evenness of l ∈ Γ
1
follows from the fact that the number of odd n
i
must be even, implying
l ◦ l =
22
¸
i=1
n
2
i
−
28
¸
i=23
n
2
i
∈ 2Z. (25)
Evenness of l +l
0
∈ Γ
2
follows from the same fact:
(l +l
0
) ◦ (l +l
0
) = l ◦ l + 2l
0
◦ l +l
0
◦ l
0
= l ◦ l +
22
¸
i=1
n
i
−
28
¸
i=23
n
i
+ 4 ∈ 2Z. (26)
Evenness implies integrality, so Γ ⊂ Γ
∗
. It’s easy to see that the dual lattice to Γ
1
is Γ
∗
1
=
Z
28
∪ (Z
28
+ l
0
) ⊃ Γ. But to be dual for example to l
0
requires a vector to have an even number
of odd n
i
s, so Γ
∗
= Γ.
To ﬁnd the gauge bosons we need to ﬁnd the lattice vectors satisfying (11.6.15). But these
are obviously the root vectors of SO(44). In addition there are the 22 gauge bosons with vertex
operators ∂X
m
˜
ψ
µ
, providing the Cartan generators to ﬁll out the adjoint representation of SO(44),
and the 6 gauge bosons with vertex operators ∂X
µ
˜
ψ
m
, generating U(1)
6
.
11.7 Problem 11.12
It seems to me that both the statement of the problem and the derivation of the Hagedorn tem
perature for the bosonic string (Vol. I, pp. 32021) are misleading. Equation (7.3.20) is not the
correct one to use to ﬁnd the asymptotic density of states of a string theory, since it does not take
into account the level matching constraint in the physical spectrum. It’s essentially a matter of
luck that Polchinski ends up with the right Hagedorn temperature, (9.8.13).
Taking into account level matching we have n(m) = n
L
(m)n
R
(m). To ﬁnd n
L
and n
R
we
treat the leftmoving and rightmoving CFTs separately, and include only the physical parts of the
spectrum, that is, neglect the ghosts and the timelike and longitudinal oscillators. Then we have,
as in (9.8.11),
¸
m
2
n
L
(m)e
−πα
′
m
2
l/2
∼ e
πc/(12l)
, (27)
11 CHAPTER 11 93
implying
n
L
(m) ∼ e
πm
√
α
′
c/6
. (28)
Similarly for the rightmovers, giving
n(m) ∼ e
πm
√
α
′
/6(
√
c+
√
˜ c)
. (29)
The Hagedorn temperature is then given by
T
−1
H
= π
√
α
′
c
6
+
˜ c
6
. (30)
For the type I and II strings, this gives
T
H
=
1
2π
√
2α
′
, (31)
while for the heterotic theories we have
T
H
=
1
π
√
α
′
(1 −
1
√
2
). (32)
The Hagedorn temperature for an open type I string is the same, (31), as for the closed string,
since n
open
(m) = n
L
(2m).
There is an interesting point that we glossed over above. The RHS of (27) is obtained by doing
a modular transformation l → 1/l on the torus partition function with τ = il. Then for small l
only the lowestlying state in the theory contributes. We implicitly took the lowestlying state to
be the vacuum, corresponding to the unit operator. However, that state is projected out by the
GSO projection in all of the above theories, else it would give rise to a tachyon. So should we
really consider it? To see that we should, let us derive (27) more carefully, for example in the case
of the leftmovers of the type II string. The GSO projection there is (−1)
F
= −1, where F is the
worldsheet fermion number of the transverse fermions (not including the ghosts). The partition
function from which we will extract n
L
(m), the number of projectedin states, is
Z(il) =
¸
i∈R,NS
q
h
i
−c/24
1
2
(1 −(−1)
F
i
) =
¸
m
2
n
L
(m)e
−πα
′
m
2
l/2
. (33)
This partition function corresponds to a path integral on the torus in which we sum over all four
spin structures, with minus signs when the fermions are periodic in the σ
2
(“time”) direction. Upon
doing the modular integral, this minus sign corresponds to giving a minus sign to R sector states.
But the sum on R and NS sectors in (33) means that we now project onto states with (−1)
F
= 1:
Z(il) =
¸
i∈R,NS
e
−2π/l(h
i
−c/24)
(−1)
α
i
1
2
(1 + (−1)
F
i
) ∼ e
πc/(12l)
. (34)
We see that the ground state, with h = 0, is indeed projected in, and therefore dominates in the
limit l →0.
12 CHAPTER 13 94
12 Chapter 13
12.1 Problem 13.2
An open string with ends attached to Dpbranes is Tdual to an open type I string. An open string
with both ends attached to the same Dpbrane and zero winding number is Tdual to an open type
I string with zero momentum in the 9 −p dualized directions, and ChanPaton factor of the form
t
a
=
1
√
2
¸
0 i
−i 0
¸
⊗diag(1, 0, . . . , 0). (1)
Four open strings attached to the same Dpbrane are Tdual to four open type I strings with zero
momentum in the 9 − p dualized directions and the same ChanPaton factor (1). The scattering
amplitude for four gauge boson open string states was calculated in section 12.4. Using
Tr(t
a
)
4
=
1
2
, (2)
the result (12.4.22) becomes in this case
S(k
i
, e
i
) = (3)
−8ig
2
YM
α
′2
(2π)
10
δ
10
(
¸
i
k
i
)K(k
i
, e
i
)
Γ(−α
′
s)Γ(−α
′
u)
Γ(1 −α
′
s −α
′
u)
+ 2 permutations
.
The kinematic factor K is written in three diﬀerent ways in (12.4.25) and (12.4.26), and we won’t
bother to reproduce it here. Since the momenta k
i
all have vanishing components in the 9 − p
dualized directions, the amplitude becomes,
−8ig
2
(p+1),YM
α
′2
V
2
9−p
(2π)
p+1
δ
p+1
(
¸
i
k
i
) (4)
K(k
i
, e
i
)
Γ(−α
′
s)Γ(−α
′
u)
Γ(1 −α
′
s −α
′
u)
+ 2 permutations
,
where V
9−p
is the volume of the transverse space, and we have used (13.3.29). But in order to
get the proper (p +1)dimensional scattering amplitude, we must renormalize the wave function of
each string (which is spread out uniformly in the transverse space) by a factor of
V
9−p
:
S
′
(k
i
, e
i
) = −8ig
2
(p+1),YM
α
′2
(2π)
p+1
δ
p+1
(
¸
i
k
i
) (5)
K(k
i
, e
i
)
Γ(−α
′
s)Γ(−α
′
u)
Γ(1 −α
′
s −α
′
u)
+ 2 permutations
.
Finally, using (13.3.30) and (13.3.28), we can write the dimensionally reduced type I YangMills
coupling g
(p+1),YM
, in terms of the coupling g
Dp
on the brane:
g
2
(p+1),YM
= g
2
Dp,SO(32)
= 2g
2
Dp
, (6)
12 CHAPTER 13 95
so
S
′
(k
i
, e
i
) = −16ig
2
Dp
α
′2
(2π)
p+1
δ
p+1
(
¸
i
k
i
) (7)
K(k
i
, e
i
)
Γ(−α
′
s)Γ(−α
′
u)
Γ(1 −α
′
s −α
′
u)
+ 2 permutations
.
(One can also use (13.3.25) to write g
Dp
in terms of the string coupling g.)
12.2 Problem 13.3
(a) By equations (B.1.8) and (B.1.10),
Γ
2a
Γ
2a+1
= −2iS
a
, (8)
where a = 1, 2, 3, 4, so
β
2a
β
2a+1
= 2iS
a
. (9)
If we deﬁne
β ≡ β
1
β
2
β
3
, (10)
and label the D4branes extended in the (6,7,8,9), (4,5,8,9), and (4,5,6,7) directions by the subscripts
2, 3, and 4 respectively, then
β
⊥
2
= β
1
β
2
β
3
β
4
β
5
= 2iS
2
β (11)
and similarly
β
⊥
3
= 2iS
3
β, β
⊥
4
= 2iS
4
β. (12)
The supersymmetries preserved by brane a (a = 2, 3, 4) are
Q
s
+ (β
⊥
a
˜
Q)
s
= Q
s
+ 2is
a
(β
˜
Q)
s
, (13)
so for a supersymmetry to be unbroken by all three branes simply requires s
2
= s
3
= s
4
. Taking
into account the chirality condition Γ = +1 on Q
s
, there are four unbroken supersymmetries:
Q
(+++++)
+i(β
˜
Q)
(+++++)
,
Q
(−−+++)
+i(β
˜
Q)
(−−+++)
,
Q
(+−−−−)
−i(β
˜
Q)
(+−−−−)
,
Q
(−+−−−)
−i(β
˜
Q)
(−+−−−)
. (14)
12 CHAPTER 13 96
(b) For the D0brane,
β
⊥
D0
= β
1
β
2
β
3
β
4
β
5
β
6
β
7
β
8
β
9
= −8iS
2
S
3
S
4
β, (15)
so the unbroken supersymmetries are of the form,
Q
s
+ (β
⊥
D0
˜
Q)
s
= Q
s
−8is
2
s
3
s
4
(β
˜
Q)
s
. (16)
The signs in all four previously unbroken supersymmetries (14) are just wrong to remain unbroken
by the D0brane, so that this conﬁguration preserves no supersymmetry.
(c) Let us make a brane scan of the original conﬁguration:
1 2 3 4 5 6 7 8 9
D4
2
D D D D D N N N N
D4
3
D D D N N D D N N
D4
4
D D D N N N N D D
D0 D D D D D D D D D
There are nine distinct Tdualities that can be performed in the 4, 5, 6, 7, 8, and 9 directions, up
to the symmetries 4 ↔ 5, 6 ↔ 7, 8 ↔ 9, and (45) ↔ (67) ↔ (89). They result in the following
brane content:
Tdualized directions (p
1
, p
2
, p
3
, p
4
)
4 (5, 3, 3, 1)
4, 5 (6, 2, 2, 2)
4, 6 (4, 4, 2, 2)
4, 5, 6 (5, 3, 1, 3)
4, 5, 6, 7 (4, 4, 0, 4)
4, 6, 8 (3, 3, 3, 3)
4, 5, 6, 8 (4, 2, 2, 4)
4, 5, 6, 7, 8 (3, 3, 1, 5)
4, 5, 6, 7, 8, 9 (2, 2, 2, 6)
Further Tduality in one, two, or all three of the 1, 2, and 3 directions will turn any of these
conﬁgurations into (p
1
+ 1, p
2
+ 1, p
3
+ 1, p
4
+ 1), (p
1
+ 2, p
2
+ 2, p
3
+ 2, p
4
+ 2), and (p
1
+ 3, p
2
+
3, p
3
+ 3, p
4
+ 3) respectively.
Tdualizing at general angles to the coordinate axes will result in combinations of the above
conﬁgurations for the directions involved, with the smallerdimensional brane in each column being
replaced by a magnetic ﬁeld on the largerdimensional brane.
12.3 Problem 13.4
(a) Let the D2brane be extended in the 8 and 9 directions, and let it be separated from the
D0brane in the 1 direction by a distance y. Tdualizing this conﬁguration in the 2, 4, 6, and 8
12 CHAPTER 13 97
directions yields 2 D4branes, the ﬁrst (from the D2brane) extended in the 2, 4, 6, and 9 directions,
and the second (from the D0brane) in the 2, 4, 6, and 8 directions. This is in the class of D4brane
conﬁgurations studied in section 13.4; in our case the angles deﬁned there take the values
φ
1
= φ
2
= φ
3
= 0, φ
4
=
π
2
, (17)
and therefore (according to (13.4.22)),
φ
′
1
= φ
′
4
=
π
4
, φ
′
2
= φ
′
3
= −
π
4
. (18)
For the three directions in which the D4branes are parallel, we must make the substitution (13.4.25)
(without the exponential factor, since we have chosen the separation between the D0 and D2branes
to vanish in those directions, but with a factor −i, to make the potential real and attractive). The
result is
V (y) = −
∞
0
dt
t
(8π
2
α
′
t)
−1/2
exp
−
ty
2
2πα
′
iϑ
4
11
(it/4, it)
ϑ
11
(it/2, it)η
9
(it)
. (19)
Alternatively, using the modular transformations (7.4.44b) and (13.4.18b),
V (y) = −
1
√
8π
2
α
′
∞
0
dt t
3/2
exp
−
ty
2
2πα
′
ϑ
4
11
(1/4, i/t)
ϑ
11
(1/2, i/t)η
9
(i/t)
. (20)
(b) For the ﬁeld theory calculation we lean heavily on the similar calculation done in section 8.7,
adapting it to D = 10. Polchinski employs the shifted dilaton
˜
Φ = Φ −Φ
0
, (21)
whose expectation value vanishes, and the Einstein metric
˜
G = e
−
˜
Φ/2
G; (22)
their propagators are given in (8.7.23):
'
˜
Φ
˜
Φ(k)` = −
2iκ
2
k
2
, (23)
'h
µν
h
σρ
(k)` = −
2iκ
2
k
2
η
µσ
η
νρ
+η
µρ
η
νσ
−
1
4
η
µν
η
σρ
, (24)
where h =
˜
G−η. The Dbrane action (13.3.14) expanded for small values of
˜
Φ and h is
S
p
= −τ
p
d
p+1
ξ
p −3
4
˜
Φ +
1
2
h
a
a
, (25)
where the trace on h is taken over directions tangent to the brane. From the point of view of the
supergravity, the Dbrane is thus a source for
˜
Φ,
J
˜
Φ,p
(X) =
3 −p
4
τ
p
δ
9−p
(X
⊥
−X
′
⊥
), (26)
12 CHAPTER 13 98
and for h,
J
µν
h,p
(X) = −
1
2
τ
p
e
µν
p
δ
9−p
(X
⊥
−X
′
⊥
), (27)
where X
′
⊥
is the position of the brane in the transverse coordinates, and e
µν
p
is η
µν
in the directions
parallel to the brane and 0 otherwise. In momentum space the sources are
˜
J
˜
Φ,p
(k) =
3 −p
4
τ
p
(2π)
p+1
δ
p+1
(k

)e
ik
⊥
X
′
⊥
, (28)
˜
J
µν
h,p
(k) = −
1
2
τ
p
e
µν
p
(2π)
p+1
δ
p+1
(k

)e
ik
⊥
X
′
⊥
. (29)
Between the D0brane, located at the origin of space, and the D2brane, extended in the 8 and 9
directions and located in the other directions at the point
(X
′
1
, X
′
2
, X
′
3
, X
′
4
, X
′
5
, X
′
6
, X
′
7
) = (y, 0, 0, 0, 0, 0, 0), (30)
the amplitude for dilaton exchange is
/
˜
Φ
= −
d
10
k
(2π)
10
˜
J
˜
Φ,0
(k)'
˜
Φ
˜
Φ(k)`
˜
J
˜
Φ,2
(−k)
= i
3
8
τ
0
τ
2
κ
2
d
10
k
(2π)
10
2πδ(k
0
)
1
k
2
(2π)
3
δ
3
(k
0
, k
8
, k
9
)e
ik
1
y
= iT
3
8
τ
0
τ
2
κ
2
d
7
k
(2π)
7
e
iky
k
2
= iT
3
8
τ
0
τ
2
κ
2
G
7
(y), (31)
where G
7
is the 7dimensional massless scalar Green function. We divide the amplitude by −iT to
obtain the static potential due to dilaton exchange:
V
˜
Φ
(y) = −
3
8
τ
0
τ
2
κ
2
G
7
(y). (32)
The calculation for the graviton exchange is similar, the only diﬀerence being that the numerical
factor (1/4)2(3/4) is replaced by
1
2
e
µν
0
2
η
µσ
η
νρ
+η
µρ
η
νσ
−
1
4
η
µν
η
σρ
1
2
e
σρ
2
=
5
8
. (33)
The total potential between the D0brane and D2brane is therefore
V (y) = τ
0
τ
2
κ
2
G
7
(y) = −π(4π
2
α
′
)
2
G
7
(y), (34)
where we have applied (13.3.4). As expected, gravitation and dilaton exchange are both attractive
forces.
In the largey limit of (20), the integrand becomes very small except where t is very small. The
ratio of modular functions involved in the integrand is in fact ﬁnite in the limit t →0,
lim
t→0
ϑ
4
11
(1/4, i/t)
ϑ
11
(1/2, i/t)η
9
(i/t)
= 2, (35)
12 CHAPTER 13 99
(according to Mathematica), so that the ﬁrst term in the asymptotic expansion of the potential in
1/y is
V (r) ≈ −
1
√
2π
2
α
′
∞
0
dt t
3/2
exp
−
ty
2
2πα
′
= −π
−1/2
(2πα
′
)
2
Γ(
5
2
)y
−5
= −π(4π
2
α
′
)
2
G
7
(y), (36)
in agreement with (34).
12.4 Problem 13.12
The tension of the (p
i
, q
i
)string is (13.6.3)
τ
(p
i
,q
i
)
=
p
2
i
+q
2
i
/g
2
2πα
′
. (37)
Let the three strings sit in the (X
1
, X
2
) plane. If the angle string i makes with the X
1
axis is θ
i
,
then the force it exerts on the junction point is
(F
1
i
, F
2
i
) =
1
2πα
′
(cos θ
i
p
2
i
+q
2
i
/g
2
, sin θ
i
p
2
i
+q
2
i
/g
2
). (38)
If we orient each string at the angle
cos θ
i
=
p
i
p
2
i
+q
2
i
/g
2
, sin θ
i
=
q
i
/g
p
2
i
+q
2
i
/g
2
, (39)
then the total force exerted on the junction point is
1
2πα
′
3
¸
i=1
(p
i
, q
i
/g), (40)
which vanishes if
¸
p
i
=
¸
q
i
= 0. This is the unique stable conﬁguration, up to rotations and
reﬂections.
The supersymmetry algebra for a static (p, q) string extended in the X
i
direction is (13.6.1)
1
2L
¸
Q
α
˜
Q
α
¸
,
Q
†
β
˜
Q
†
β
¸
= τ
(p,q)
δ
αβ
¸
1 0
0 1
¸
+
(Γ
0
Γ
i
)
αβ
2πα
′
¸
p q/g
q/g −p
¸
. (41)
Deﬁning
u ≡
p
p
2
+q
2
/g
2
, U ≡
1
√
2
¸
√
1 +u
√
1 −u
−
√
1 −u
√
1 +u
¸
, (42)
12 CHAPTER 13 100
we can use U to diagonalize the matrix on the RHS of (41):
1
2Lτ
(p,q)
U
¸
Q
α
˜
Q
α
¸
,
Q
†
β
˜
Qβ
†
U
T
¸
=
¸
(I
16
+ Γ
0
Γ
i
)
αβ
0
0 (I
16
−Γ
0
Γ
i
)
αβ
¸
. (43)
The top row of this 2 2 matrix equation tells us that, in a basis in spinor space in which Γ
0
Γ
i
is
diagonal, the supersymmetry generator
√
1 +uQ
α
+
√
1 −u
˜
Q
α
(44)
annihilates this state if (I
16
+ Γ
0
Γ
i
)
αα
= 0. We can use (I
16
− Γ
0
Γ
i
) to project onto this eight
dimensional subspace, yielding eight supersymmetries that leave this state invariant:
(I
16
−Γ
0
Γ
i
)(
√
1 +uQ+
√
1 −u
˜
Q)
α
. (45)
The other eight unbroken supersymmetries are given by the bottom row of (43), after projecting
onto the subspace annihilated by (I
16
−Γ
0
Γ
i
):
(I
16
+ Γ
0
Γ
i
)(−
√
1 −uQ +
√
1 +u
˜
Q)
α
. (46)
Now let us suppose that the string is aligned in the direction (39), which depends on p and q.
We will show that eight of the sixteen unbroken supersymmetries do not depend on p or q, and
therefore any conﬁguration of (p, q) strings that all obey (39) will leave these eight unbroken. If
the string is aligned in the direction (39), then
Γ
i
=
p
p
2
+q
2
/g
2
Γ
1
+
q/g
p
2
+q
2
/g
2
Γ
2
= uΓ
1
+
1 −u
2
Γ
2
. (47)
Our ﬁrst set of unbroken supersymmetries (45) becomes
(I
16
−uΓ
0
Γ
1
−
1 −u
2
Γ
0
Γ
2
)(
√
1 +uQ+
√
1 −u
˜
Q)
α
. (48)
We work in a basis of eigenspinors of the operators S
a
deﬁned in (B.1.10). In this basis Γ
0
Γ
1
= 2S
0
,
while
Γ
0
Γ
2
=
¸
0 1
−1 0
¸
⊗
¸
0 −1
1 0
¸
⊗I
2
⊗I
2
⊗I
2
. (49)
We can divide the sixteen values of the spinor index α into four groups of four according to the
12 CHAPTER 13 101
eigenvalues of S
0
and S
1
:
(I
16
−uΓ
0
Γ
1
−
1 −u
2
Γ
0
Γ
2
)(
√
1 +uQ+
√
1 −u
˜
Q)
(++s
2
s
3
s
4
)
(50)
= (1 −u)(
√
1 +uQ+
√
1 −u
˜
Q)
(++s
2
s
3
s
4
)
+
1 −u
2
(
√
1 +uQ+
√
1 −u
˜
Q)
(−−s
2
s
3
s
4
)
,
(I
16
−uΓ
0
Γ
1
−
1 −u
2
Γ
0
Γ
2
)(
√
1 +uQ+
√
1 −u
˜
Q)
(++s
2
s
3
s
4
)
(51)
= (1 +u)(
√
1 +uQ+
√
1 −u
˜
Q)
(−−s
2
s
3
s
4
)
+
1 −u
2
(
√
1 +uQ+
√
1 −u
˜
Q)
(++s
2
s
3
s
4
)
,
(I
16
−uΓ
0
Γ
1
−
1 −u
2
Γ
0
Γ
2
)(
√
1 +uQ+
√
1 −u
˜
Q)
(+−s
2
s
3
s
4
)
(52)
= (1 −u)(
√
1 +uQ+
√
1 −u
˜
Q)
(+−s
2
s
3
s
4
)
−
1 −u
2
(
√
1 +uQ+
√
1 −u
˜
Q)
(−+s
2
s
3
s
4
)
,
(I
16
−uΓ
0
Γ
1
−
1 −u
2
Γ
0
Γ
2
)(
√
1 +uQ+
√
1 −u
˜
Q)
(−+s
2
s
3
s
4
)
(53)
= (1 +u)(
√
1 +uQ+
√
1 −u
˜
Q)
(−+s
2
s
3
s
4
)
−
1 −u
2
(
√
1 +uQ+
√
1 −u
˜
Q)
(+−s
2
s
3
s
4
)
.
(The indexing by s
2
, s
3
, s
4
is somewhat redundant, since the chirality condition on both Q and
˜
Q
implies the restriction 8s
2
s
3
s
4
= 1 in the case of (50) and (51), and 8s
2
s
3
s
4
= −1 in the case of
(52) and (53).) It easy to see that (50) and (51) diﬀer only by a factor of
(1 −u)/(1 +u), and
(52) and (53) similarly by a factor of −
(1 −u)/(1 +u), so (50) and (52) alone are suﬃcient to
describe the eight independent supersymmetry generators in this sector. In the other sector, given
by (46), there is a similar repetition of generators, and the eight independent generators are
(I
16
+uΓ
0
Γ
1
+
1 −u
2
Γ
0
Γ
2
)(−
√
1 −uQ+
√
1 +u
˜
Q)
(++s
2
s
3
s
4
)
(54)
= (1 +u)(−
√
1 −uQ+
√
1 +u
˜
Q)
(++s
2
s
3
s
4
)
−
1 −u
2
(−
√
1 −uQ+
√
1 +u
˜
Q)
(−−s
2
s
3
s
4
)
,
(I
16
+uΓ
0
Γ
1
+
1 −u
2
Γ
0
Γ
2
)(−
√
1 −uQ+
√
1 +u
˜
Q)
(+−s
2
s
3
s
4
)
(55)
= (1 +u)(−
√
1 −uQ+
√
1 +u
˜
Q)
(+−s
2
s
3
s
4
)
+
1 −u
2
(−
√
1 −uQ+
√
1 +u
˜
Q)
(−+s
2
s
3
s
4
)
.
Are there linear combinations of the generators (50), (52), (54), and (55) that are independent of u,
and therefore unbroken no matter what the values of p and q? Indeed, by dividing (50) by 2
√
1 −u
and (54) by 2
√
1 +u and adding them, we come up with four such generators:
˜
Q
(++s
2
s
3
s
4
)
+Q
(−−s
2
s
3
s
4
)
. (56)
12 CHAPTER 13 102
Four more are found by dividing (52) by 2
√
1 −u and (55) by
√
1 +u:
˜
Q
(+−s
2
s
3
s
4
)
−Q
(−+s
2
s
3
s
4
)
. (57)
As promised, one quarter of the original supersymmetries leave the entire conﬁguration described
in the ﬁrst paragraph of this solution invariant.
13 CHAPTER 14 103
13 Chapter 14
13.1 Problem 14.1
The excitation on the Fstring will carry some energy (per unit length) p
0
, and momentum (per unit
length) in the 1direction p
1
. Since the string excitations move at the speed of light, leftmoving
excitation have p
0
= −p
1
, while rightmoving excitations have p
0
= p
1
. The supersymmetry algebra
(13.2.9) for this string is similar to (13.6.1), with additional terms for the excitation:
1
2L
¸
Q
α
˜
Q
α
¸
,
Q
†
β
˜
Q
†
β
¸
(1)
=
1
2πα
′
¸
(δ + Γ
0
Γ
1
)
αβ
0
0 (δ −Γ
0
Γ
1
)
αβ
¸
+
p
0
δ
αβ
+p
1
(Γ
0
Γ
1
)
αβ
¸
1 0
0 1
¸
.
The ﬁrst term on the RHS vanishes for those supersymmetries preserved by the unexcited Fstring,
namely Qs for which Γ
0
Γ
1
= −1 and
˜
Qs for which Γ
0
Γ
1
= 1. The second term thus also vanishes
(making the state BPS) for the Qs if the excitation is leftmoving, and for the
˜
Qs if the excitation
is rightmoving.
For the Dstring the story is almost the same, except that the ﬁrst term above is diﬀerent:
1
2L
¸
Q
α
˜
Q
α
¸
,
Q
†
β
˜
Q
†
β
¸
(2)
=
1
2πα
′
g
¸
δ
αβ
(Γ
0
Γ
1
)
αβ
(Γ
0
Γ
1
)
αβ
δ
αβ
¸
+
p
0
δ
αβ
+p
1
(Γ
0
Γ
1
)
αβ
¸
1 0
0 1
¸
.
When diagonalized, the ﬁrst term yields the usual preserved supersymmetries, of the form Q
α
+
(β
⊥
˜
Q)
α
. When 1 − (Γ
0
Γ
1
)
αα
= 0 this supersymmetry is also preserved by the second term if the
excitation is leftmoving; when 1 + (Γ
0
Γ
1
)
αα
= 0 it is preserved if the excitation is rightmoving.
Either way, the state is BPS.
13.2 Problem 14.2
The supergravity solution for two static parallel NS5branes is given in (14.1.15) and (14.1.17):
e
2Φ
= g
2
+
Q
1
2π
2
(x
m
−x
m
1
)
2
+
Q
2
2π
2
(x
m
−x
m
2
)
2
,
G
mn
= g
−1
e
2Φ
δ
mn
, G
µν
= gη
µν
,
H
mnp
= −ǫ
mnp
q
∂
q
Φ, (3)
where µ, ν = 0, . . . , 5 and m, n = 6, . . . , 9 are the parallel and transverse directions respectively, and
the branes are located in the transverse space at x
m
1
and x
m
2
. (We have altered (14.1.15a) slightly
in order to make (3) Sdual to the Dbrane solution (14.8.1).) A Dstring stretched between the
13 CHAPTER 14 104
two branes at any given excitation level is a point particle with respect to the 5+1 dimensional
Poincar´e symmetry of the parallel dimensions. In other words, if we make an ansatz for the solution
of the form
X
µ
= X
µ
(τ), X
m
= X
m
(σ), (4)
then, after performing the integral over σ in the Dstring action, we should obtain the pointparticle
action (1.2.2) in 5+1 dimensions,
S
pp
= −m
dτ
−∂
τ
X
µ
∂
τ
X
µ
, (5)
where m is the mass of the solution X
m
(σ) with respect to the 5+1 dimensional Poincar´e symmetry.
Assuming that the gauge ﬁeld is not excited, with this ansatz the Dstring action (13.3.14)
factorizes:
S
D1
= −
1
2πα
′
dτdσ e
−Φ
−det(G
ab
+B
ab
)
= −
1
2πα
′
dτdσ e
−Φ
−
G
µν
∂
τ
X
µ
∂
τ
X
ν
(G
µn
+B
µn
)∂
τ
X
µ
∂
σ
X
n
(G
mν
+B
mν
)∂
σ
X
m
∂
τ
X
ν
G
mn
∂
σ
X
m
∂
σ
X
n
= −
1
2πα
′
dσ e
−Φ
∂
σ
X
m
∂
σ
X
m
dτ
−∂
τ
X
µ
∂
τ
X
µ
. (6)
In the last equality we have used the fact that neither the metric nor the twoform potential in the
solution (3) have mixed µn components. Comparison with (5) shows that
m =
g
−1/2
2πα
′
dσ [∂
σ
X
m
[, (7)
where the integrand is the coordinate (not the proper) line element in this coordinate system. The
ground state is therefore a straight line connecting the two branes:
m =
g
−1/2
[x
m
2
−x
m
1
[
2πα
′
. (8)
As explained above, this mass is deﬁned with respect the geometry of the parallel directions,
and it is only in string frame that the parallel metric G
µν
is independent of the transverse position.
We can nonetheless deﬁne an Einsteinframe mass m
E
with respect to G
µν
at [x
m
[ = ∞, and it
is this mass that transforms simply under Sduality. (Here we are using the deﬁnition (14.1.7) of
the Einstein frame, G
E
= e
−Φ/2
G, which is slightly diﬀerent from the one used in volume I and in
Problem 14.6 below, where G
E
= e
−
˜
Φ/2
G.) From the deﬁnition (5) of the mass,
m
E
= g
1/4
m, (9)
13 CHAPTER 14 105
so (7) becomes
m
E
=
g
−1/4
2πα
′
dσ [∂
σ
X
m
[. (10)
We can calculate the mass of an Fstring stretched between two D5branes in two diﬀerent
pictures: we can use the black 5brane supergravity solution (14.8.1) and do a calculation similar
to the one above, or we can consider the Fstring to be stretched between two elementary D5branes
embedded in ﬂat spacetime. In the ﬁrst calculation, the Sduality is manifest at every step, since
the NS5brane and the black 5brane solutions are related by Sduality, as are the Dstring and F
string actions. The second calculation yields the same answer, and is much easier: since the tension
of the Fstring is 1/2πα
′
, and in ﬂat spacetime (G
µν
= η
µν
) its proper length and coordinate length
are the same, its total energy is
m =
1
2πα
′
dσ [∂
σ
X
m
[. (11)
Its Einsteinframe mass is then
m
E
=
g
1/4
2πα
′
dσ [∂
σ
X
m
[, (12)
which indeed agrees with (10) under g →1/g.
13.3 Problem 14.6
To ﬁnd the expectation values of the dilaton and graviton in the low energy ﬁeld theory, we add to
the action a source term
S
′
=
d
10
X
K
˜
Φ
˜
Φ +K
µν
h
h
µν
, (13)
and take functional derivatives of the partition function Z[K
˜
Φ
, K
h
] with respect to K
˜
Φ
and K
h
.
For the Dbrane, which is a real, physical source for the ﬁelds, we also include the sources J
˜
Φ
and
J
h
, calculated in problem 13.4(b) (see (26) and (27) of that solution):
J
˜
Φ
(X) =
3 −p
4
τ
p
δ
9−p
(X
⊥
), (14)
J
µν
h
(X) = −
1
2
τ
p
e
µν
p
δ
9−p
(X
⊥
). (15)
Recall that h is the perturbation in the Einsteinframe metric,
˜
G = η + h, and that e
µν
equals
η
µν
for µ, ν parallel to the brane and zero otherwise. (We have put the brane at the origin, so
that X
′
⊥
= 0.) Since the dilaton decouples from the Einsteinframe graviton, we can calculate the
partition functions Z[K
˜
Φ
] and Z[K
h
] separately. Using the propagator (23),
Z[K
˜
Φ
] = −Z[0]
d
10
k
(2π)
10
˜
J
˜
Φ
(−k)'
˜
Φ
˜
Φ(k)`
˜
K
˜
Φ
(k)
= Z[0]
3 −p
2
iκ
2
τ
p
d
9−p
k
⊥
(2π)
9−p
1
k
2
⊥
˜
K
˜
Φ
(k
⊥
, k

= 0)
= Z[0]
3 −p
2
iκ
2
τ
p
d
9−p
X
⊥
G
9−p
(X
⊥
)
d
p+1
X

K
˜
Φ
(X), (16)
13 CHAPTER 14 106
so that
'
˜
Φ(X)` =
1
iZ[0]
δZ[K
˜
Φ
]
δK
˜
Φ
(X)
=
3 −p
2
κ
2
τ
p
G
9−p
(X
⊥
). (17)
Using (13.3.22), (13.3.23), and the positionspace expression for G
d
, this becomes
'
˜
Φ(X)` =
3 −p
4
(4π)
(5−p)/2
Γ(
7 −p
2
)gα
′(7−p)/2
r
p−7
=
3 −p
4
ρ
7−p
r
7−p
, (18)
where ρ
7−p
is as deﬁned in (14.8.2b) with Q = 1. Hence
e
2/Φ)
≈ g
2
(1 + 2'
˜
Φ`)
≈ g
2
1 +
ρ
7−p
r
7−p
(3−p)/2
, (19)
in agreement with (14.8.1b) (corrected by a factor of g
2
).
The graviton calculation is very similar. Using the propagator (24),
Z[K
h
] = −Z[0]
d
10
X
(2π)
10
˜
J
µν
h
(−k)'h
µν
h
ρσ
(k)`
˜
K
ρσ
h
(k)
= Z[0]
p + 1
8
η
µν
−e
µν
2iκ
2
τ
p
d
9−p
k
⊥
(2π)
10
1
k
2
⊥
˜
K
µν
h
(k
⊥
, k

= 0)
= Z[0]
p + 1
8
η
µν
−e
µν
2iκ
2
τ
p
(20)
d
9−p
X
⊥
G
9−p
(X
⊥
)
d
p+1
X

K
µν
h
(X),
so
'h
µν
(X)` =
p + 1
8
η
µν
−e
µν
2κ
2
τ
p
G
9−p
(X
⊥
)
=
p + 1
8
η
µν
−e
µν
ρ
7−p
r
7−p
. (21)
Hence for µ, ν aligned along the brane,
'
˜
G
µν
` ≈
1 +
ρ
7−p
r
7−p
(p−7)/8
η
µν
, (22)
while for m, n transverse to the brane,
'
˜
G
mn
` ≈
1 +
ρ
7−p
r
7−p
(p+1)/8
δ
mn
. (23)
13 CHAPTER 14 107
The string frame metric, G = e
˜
Φ/2
˜
G, is therefore
'G
µν
` ≈
1 +
ρ
7−p
r
7−p
−1/2
η
µν
, (24)
'G
mn
` ≈
1 +
ρ
7−p
r
7−p
1/2
δ
mn
, (25)
in agreement with (14.8.1).
14 CHAPTER 15 108
14 Chapter 15
14.1 Problem 15.1
The matrix of inner products is
´
3
= 'h[
L
3
1
L
1
L
2
L
3
¸
¸
¸
L
3
−1
L
−2
L
−1
L
−3
[h` (1)
=
24h(h + 1)(2h + 1) 12h(3h + 1) 24h
12h(3h + 1) h(8h + 8 +c) 10h
24h 10h 6h + 2c
¸
¸
¸. (2)
This matches the Kac formula, with
K
3
= 2304. (3)
14.2 Problem 15.3
Let’s begin by recording some useful symmetry relations of the operator product coeﬃcient with
lower indices, derived from the deﬁnition (6.7.13) and (6.7.14),
c
ijk
=
/
′
i
(∞, ∞)/
j
(1, 1)/
k
(0, 0)
S
2
. (4)
The following relations then hold, with the sign of the coeﬃcient depending on the statistics of the
operators:
c
ijk
= ±(−1)
h
j
+
˜
h
j
c
kji
if /
j
is primary (5)
c
ijk
= ±(−1)
h
i
+h
j
+h
k
+
˜
h
i
+
˜
h
j
+
˜
h
k
c
ikj
if /
i
is primary (6)
c
ijk
= ±(−1)
h
k
+
˜
h
k
c
jik
if /
i
, /
j
, /
k
are primary. (7)
Actually, in the above “primary” may be weakened to “quasiprimary” (meaning annihilated by
L
1
, rather than L
n
for all n > 0, and therefore transforming as a tensor under PSL(2, C) rather
than general local conformal transformations), but Polchinski does not seem to ﬁnd the notion of
quasiprimary operator interesting or useful.
Armed with these symmetries (and in particular relation (5)), but glibly ignoring phase factors
as Polchinski does, we would like to claim that the correct form for (15.2.7) should be as follows
(we haven’t bothered to raise the index):
c
ik,
˜
k¦,mn
= lim
zn→∞
zm→1
z
2hn
n
¯ z
2
˜
hn
n
L
−k¦
˜
L
−
˜
k¦
'O
n
(z
n
, ¯ z
n
)O
m
(z
m
, ¯ z
m
)O
i
(0, 0)`
S
2
. (8)
We would also like to claim that the LHS of (15.2.9) should read
O
′
l
(∞, ∞)O
j
(1, 1)O
m
(z, ¯ z)O
n
(0, 0)
S
2
. (9)
14 CHAPTER 15 109
We are now ready to solve the problem. The case N = 0 is trivial, since by the deﬁnition
(15.2.8),
β
i¦
mn
= 1, (10)
so the coeﬃcient of z
−hm−hn+h
i
in T
jl
mn
(i[z) is 1. For N = 1 there is again only one operator,
L
−1
O
i
. We have
β
i1¦
mn
=
1
2h
i
(h
i
+h
m
−h
n
). (11)
Thus the coeﬃcient of z
−hm−hn+h
i
+1
is
1
2h
i
(h
i
+h
m
−h
n
)(h
i
+h
j
−h
l
). (12)
15 APPENDIX B 110
15 Appendix B
15.1 Problem B.1
Under a change of spinor representation basis, Γ
µ
→ UΓ
µ
U
−1
, B
1
, B
2
, and C, all transform the
same way:
B
1
→U
∗
B
1
U
−1
, B
2
→U
∗
B
2
U
−1
, C →U
∗
CU
−1
. (1)
The invariance of the following equations under this change of basis is more or less trivial:
(B.1.17) (the deﬁnition of B
1
and B
2
):
U
∗
B
1
U
−1
UΓ
µ
U
−1
(U
∗
B
1
U
−1
)
−1
= U
∗
B
1
Γ
µ
B
−1
1
U
T
= (−1)
k
U
∗
Γ
µ∗
U
T
= (−1)
k
(UΓ
µ
U
−1
)
∗
, (2)
U
∗
B
2
U
−1
UΓ
µ
U
−1
(U
∗
B
2
U
−1
)
−1
= U
∗
B
2
Γ
µ
B
−1
2
U
T
= (−1)
k+1
U
∗
Γ
µ∗
U
T
= (−1)
k+1
(UΓ
µ
U
−1
)
∗
. (3)
(B.1.18), using the fact that Σ
µν
transforms the same way as Γ
µ
:
U
∗
BU
−1
UΣ
µν
U
−1
UB
−1
U
T
= U
∗
BΣ
µν
B
−1
U
T
= −U
∗
Σ
µν∗
U
T
= −(UΣ
µν
U
−1
)
∗
. (4)
The invariance of (B.1.19) is the same as that of (B.1.17).
(B.1.21):
UB
∗
1
U
T
U
∗
B
1
U
−1
= UB
∗
1
B
1
U
−1
= (−1)
k(k+1)/2
UU
−1
= (−1)
k(k+1)/2
, (5)
UB
∗
2
U
T
U
∗
B
2
U
−1
= UB
∗
2
B
2
U
−1
= (−1)
k(k−1)/2
UU
−1
= (−1)
k(k−1)/2
. (6)
(B.1.24) (the deﬁnition of C):
U
∗
CU
−1
UΓ
µ
U
−1
UC
−1
U
T
= U
∗
CΓ
µ
C
−1
U
T
= −U
∗
Γ
µT
U
T
= −(UΓ
µ
U
−1
)
T
. (7)
(B.1.25): all three sides clearly transform by multiplying on the left by U and on the right by
U
−1
.
(B.1.27):
U
∗
BU
−1
UΓ
0
U
−1
= U
∗
BΓ
0
U
−1
= U
∗
CU
−1
. (8)
15 APPENDIX B 111
We will determine the relation between B and B
T
in the ζ
(s)
basis, where (for d = 2k + 2) B
1
and B
2
are deﬁned by equation (B.1.16):
B
1
= Γ
3
Γ
5
Γ
d−1
, B
2
= ΓB
1
. (9)
Since all of the Γ’s that enter into this product are antisymmetric in this basis (since they are
Hermitian and imaginary), we have
B
T
1
= (Γ
3
Γ
5
Γ
d−1
)
T
= (−1)
k
Γ
d−1
Γ
d−3
Γ
3
= (−1)
k(k+1)/2
Γ
3
Γ
5
Γ
d−1
= (−1)
k(k+1)/2
B
1
. (10)
Using (10), the fact that Γ is real and symmetric in this basis, and (B.1.19), we ﬁnd
B
T
2
= (ΓB
1
)
T
= B
T
1
Γ
T
= (−1)
k(k+1)/2
B
1
Γ
= (−1)
k(k+1)/2+k
Γ
∗
B
1
= (−1)
k(k−1)/2
B
2
. (11)
Since B
1
is used when k = 0, 3 (mod 4), and B
2
is used when k = 0, 1 (mod 4), so in any dimension
in which one can impose a Majorana condition we have
B
T
= B. (12)
When k is even, C = B
1
Γ
0
, so, using the fact that in this basis Γ
0
is real and antisymmetric,
and (B.1.17), we ﬁnd
C
T
= Γ
0T
B
T
1
= (−1)
k(k+1)/2+1
Γ
0∗
B
1
= (−1)
k/2+1
B
1
Γ
0
= (−1)
k/2+1
C. (13)
On the other hand, if k is odd, then C = B
2
Γ
0
, and
C
T
= Γ
0T
B
T
2
= (−1)
k(k−1)/2+1
Γ
0∗
B
2
= (−1)
(k+1)/2
B
2
Γ
0
= (−1)
(k+1)/2
C. (14)
That all of these relations are invariant under change of basis follows directly from the trans
formation law (1), since B
T
and C
T
transform the same way as B and C.
15 APPENDIX B 112
15.2 Problem B.3
The decomposition of SO(1,3) spinor representations under the subgroup SO(1,1)SO(2) is de
scribed most simply in terms of Weyl representations: one positive chirality spinor, ζ
++
, transforms
as a positive chirality Weyl spinor under both SO(1,1) and SO(2), while the other, ζ
−−
, transforms
as a negative chirality Weyl spinor under both (see (B.1.44a)). Let us therefore use the Weylspinor
description of the 4 real supercharges of d = 4, ^ = 1 supersymmetry. The supersymmetry algebra
is (B.2.1a):
¦Q
++
, Q
†
++
¦ = 2(P
0
−P
1
),
¦Q
−−
, Q
†
−−
¦ = 2(P
0
+P
1
), (15)
¦Q
++
, Q
†
−−
¦ = 2(P
2
+iP
3
).
We must now decompose these Weyl spinors into MajoranaWeyl spinors. Deﬁne
Q
1
L
=
1
2
(Q
++
+Q
†
++
), Q
2
L
=
1
2i
(Q
++
−Q
†
++
), (16)
Q
1
R
=
1
2
(Q
−−
+Q
†
−−
), Q
2
R
=
1
2i
(Q
−−
−Q
†
−−
). (17)
(The subscripts L and R signify that the respective supercharges have positive and negative SO(1,1)
chirality.) Hence, for instance,
¦Q
1
L
, Q
1
L
¦ =
1
4
¦Q
++
, Q
++
¦ +¦Q
†
++
, Q
†
++
¦ + 2¦Q
++
, Q
†
++
¦
. (18)
The last term is given by the algebra (14), but what do we do with the ﬁrst two terms? The d = 4
algebra has a U(1) Rsymmetry under which Q
++
and Q
−−
are both multiplied by the same phase.
In order for the d = 2 algebra to inherit that symmetry, we must assume that
Q
2
++
= Q
2
−−
= ¦Q
++
, Q
−−
¦ = 0. (19)
The d = 2 algebra is then
¦Q
A
L
, Q
B
L
¦ = δ
AB
(P
0
−P
1
), (20)
¦Q
A
R
, Q
B
R
¦ = δ
AB
(P
0
+P
1
), (21)
¦Q
A
L
, Q
B
R
¦ = Z
AB
, (22)
where
Z =
¸
P
2
−P
3
P
3
P
2
¸
. (23)
The central charges are thus the KaluzaKlein momenta associated with the reduced dimensions.
If these momenta are 0 (as in dimensional reduction in the strict sense), then the algebra possesses
a further Rsymmetry, namely the SO(2) of rotations in the 23 plane. We saw at the beginning
that Q
++
is positively charged and Q
−−
negatively charged under this symmetry. This symmetry
and the original U(1) Rsymmetry of the d = 4 algebra can be recombined into two independent
SO(2) Rsymmetry groups of the Q
A
L
and Q
A
R
pairs of supercharges.
15 APPENDIX B 113
15.3 Problem B.5
Unfortunately, it appears that some kind of fudge will be necessary to get this to work out correctly.
There may be an error lurking in the book. The candidate fudges are: (1) The vector multiplet is
8
v
+8
′
, not 8
v
+8 (this is suggested by the second sentence of Section B.6). (2) The supercharges
of the N = 1 theory are in the 16
′
, not 16, of SO(9,1). (3) The frame is one in which k
0
= k
1
, not
k
0
= k
1
as purportedly used in the book. We will arbitrarily choose fudge #1, although it’s hard
to see how this can ﬁt into the analysis of the type I spectrum in Chapter 10.
The 8
v
states have helicities (±1, 0, 0, 0), (0, ±1, 0, 0), (0, 0, ±1, 0), and (0, 0, 0, ±1). The 8
′
states
have helicities ±(−
1
2
, +
1
2
, +
1
2
, +
1
2
), ±(+
1
2
, −
1
2
, +
1
2
, +
1
2
), ±(+
1
2
, +
1
2
, −
1
2
, +
1
2
), and ±(+
1
2
, +
1
2
, +
1
2
, −
1
2
).
In a frame in which k
0
= k
1
, the supersymmetry algebra is
¦Q
α
, Q
†
β
¦ = 2P
µ
(Γ
µ
Γ
0
)
αβ
= −2k
0
(1 + 2S
0
)
αβ
, (24)
so that supercharges with s
0
= −
1
2
annihilate all states. The supercharges with s
0
= +
1
2
form
an 8 representation of the SO(8) little group. Since the operator B switches the sign of all
the helicities s
1
, . . . , s
4
, the Majorana condition pairs these eight supercharges into four indepen
dent sets of fermionic raising and lowering operators. Let Q
(+
1
2
,+
1
2
,+
1
2
,+
1
2
,+
1
2
)
, Q
(+
1
2
,+
1
2
,+
1
2
,−
1
2
,−
1
2
)
,
Q
(+
1
2
,+
1
2
,−
1
2
,+
1
2
,−
1
2
)
, and Q
(+
1
2
,+
1
2
,−
1
2
,−
1
2
,+
1
2
)
be the raising operators. We can obtain all sixteen of
the states in 8
v
+8
′
by starting with the state (−1, 0, 0, 0) and acting on it with all possible com
binations of these four operators. The four states in the 8
′
with s
1
= −
1
2
are obtained by acting
with a single operator. Acting with a second operator yields the six states in the 8
v
with s
1
= 0. A
third operator gives s
1
= +
1
2
, the other four states of the 8
′
. Finally, acting with all four operators
yields the last state of the 8
v
, (+1, 0, 0, 0).
REFERENCES 114
References
[1] J. Polchinski, String Theory, Vol. 1: Introduction to the Bosonic String. Cambridge
University Press, 1998.
[2] J. Polchinski, String Theory, Vol. 2: Superstring Theory and Beyond. Cambridge University
Press, 1998.
CONTENTS
1
Contents
0 Preface 1 Chapter 1 1.1 Problem 1.2 Problem 1.3 Problem 1.4 Problem 1.5 Problem 2 Chapter 2 2.1 Problem 2.2 Problem 2.3 Problem 2.4 Problem 2.5 Problem 2.6 Problem 2.7 Problem 2.8 Problem 2.9 Problem 3 Chapter 3 3.1 Problem 3.2 Problem 3.3 Problem 3.4 Problem 3.5 Problem 3.6 Problem 3.7 Problem 3.8 Problem 3.9 Problem 4 5 5 6 7 7 9 11 11 12 13 14 16 17 18 19 20 21 21 21 22 24 25 26 28 30 32
1.1 1.3 1.5 1.7 1.9
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
2.1 . 2.3 . 2.5 . 2.7 . 2.9 . 2.11 2.13 2.15 2.17
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
3.1 . 3.2 . 3.3 . 3.4 . 3.5 . 3.7 . 3.9 . 3.11 3.13
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
4 Chapter 4 33 4.1 Problem 4.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 5 Chapter 5 35 5.1 Problem 5.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
CONTENTS
2 40 40 40 40 42 45 48 51 52 52 54 56 58 59 61 62 62 64 66 69 69 70 71 72 72 73 75 75
6 Chapter 6 6.1 Problem 6.2 Problem 6.3 Problem 6.4 Problem 6.5 Problem 6.6 Problem 6.7 Problem 7 Chapter 7 7.1 Problem 7.2 Problem 7.3 Problem 7.4 Problem 7.5 Problem 7.6 Problem 7.7 Problem 7.8 Problem 7.9 Problem 7.10 Problem 8 Chapter 8 8.1 Problem 8.2 Problem 8.3 Problem 8.4 Problem 8.5 Problem 8.6 Problem 8.7 Problem 8.8 Problem
6.1 . 6.3 . 6.5 . 6.7 . 6.9 . 6.11 6.12
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
7.1 . 7.3 . 7.5 . 7.7 . 7.8 . 7.9 . 7.10 7.11 7.13 7.15
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
8.1 . 8.3 . 8.4 . 8.5 . 8.6 . 8.7 . 8.9 . 8.11
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
9 Appendix A 9.1 Problem A.1 9.2 Problem A.3 9.3 Problem A.5
78 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
10 Chapter 10 83 10.1 Problem 10.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83 10.2 Problem 10.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
. . . . . . . . . . . .2 . 113 . . . . . . . . .7 . . . . . .4 . . . . . . . . .4 10. . . . .1 Problem 14. . . . .3 Problem B. . . . . . . . . .3 Problem 12. . . .2 Problem 12. . . . . . . . . .2 . . . . . . . . . . . . 15. . . 15. . . . . . 13 Chapter 14 103 13. . .9 Problem Problem Problem Problem Problem Problem Problem 11 Chapter 11 11. . . . . . . . . . . . . . . . . . . . . . . . . . . . 10. . . . . . . . . . . . . . . . . . . . . 110 . . . . . . . . . . . . . . . . . . . . . .6 10. . . . . . . . . . . . . . . . . . . . . . . . . . . 13. . . . . . . . . . . . . . . . .12 .11 . . . . . . . . . . . . . . . . . . . . . . . . . . .4 . .4 Problem 11. . . . . . . . . . . . . . . . . . . .14 .10 . . . . . .8 . . . . . . . . . . . . . . . . . . . . . . . .5 Problem 11. . . . .2 Problem B. . . . . . . . . . . . . . . . .5 . . . . . .9 . . . . . . . . . . . . . . . . . . . . . .1 . . . . . . . . . . . . . . . . . .1 Problem 15. . . 105 14 Chapter 15 108 14. . .1 . . . . . . . . . . . . . . . 108 15 Appendix B 15. . . 10. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112 . . . . . . . . 11. . . . . . . . . . . . . 11. . 10. . . .3 . . . . . . . . . 84 84 85 85 86 86 87 88 88 88 88 90 91 92 92 94 94 95 96 99 10. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 Problem 12. . . . . . . . . . . . . . . 11. . . .1 . . . . 108 14. . . . . . . . . . . . . . . . . . . .3 Problem 11. . . . . . . . . . . . . . . . . . . . . . 103 13. . . . . .1 Problem 11. . . . . . . . . . . .8 10. . . . . . . 13. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .CONTENTS 3 10. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Problem B. .6 Problem 11. . . . . . . . . . . .7 Problem 12 Chapter 13 12. .3 10. . 11. .7 . . . . . . . . . . . . .7 10. . . . . . . . . . . . . . . . . . . 11. 10. . . . . . . . . . . . . . . . . . . . . . .3 . . . . . . . . . . . .12 . . . . . . . . . . . . . . . . . . 10. . . . . . . . . . . . . . . . . . . . . . . . . .3 . . . . . . . .5 10. . 110 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 Problem 15. . . . . . . . . . . . .3 . . . . . . . . . . . . . .4 Problem 11. . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 Problem 11. . . . . . 13. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .6 . . . . . . . . .3 Problem 14. . .2 Problem 14. . . . . . .5 . . . . . . . . . . . . . . . . . . . . . . . . 13. . . . . . . . . . . .1 . . . . . . . . . . . . . . . .3 . . . . . . . . . . . . . . . . . . . . . . . . 10. . . . . . . . . . . . . 103 13. . .
and carry no endorsement from Polchinski. Polchinski’s twovolume textbook String Theory [1. Podolsky. 2]. These solutions are the work of myself alone. S. and M. I originally wrote up these solutions while teaching myself the subject. Spradlin for help on these problems. This work was done while I was a graduate student at Harvard University. I would like to thank R. Minwalla. . and then later decided that they may be of some use to others doing the same. Britto. D.0 PREFACE 4 0 Preface The following pages contain detailed solutions to 81 of the 202 problems in J. and was supported by an NSF Graduate Research Fellowship.
1 ˙ (a) We work in the gauge where τ = X 0 . and assume X i ≡ v i ≪ 1. 2 ˙ (b) Again. Nonrelativistic motion means X i ≡ v i ≪ 1. (2) Using the fact that the transverse velocity of the string is vT = v − the NambuGoto Lagrangian can be written: L=− 1 2πα′ 1 =− 2πα′ 1 ≈− 2πα′ = where ρ= is the mass per unit length of the string.1 CHAPTER 1 5 1 1. we work in the gauge τ = X 0 . the induced metric hab = ∂a X µ ∂b Xµ becomes: {hab } = −1 + v 2 u · v u·v u2 u·v u. dσ u 1 =ρ 2πα′ (6) dσ (− det{hab })1/2 dσ u2 (1 − v 2 ) + (u · v)2 u · v2 1 dσ u 1 − v 2 + 2 2u2 (4) 1/2 (3) 1 2 dσ u ρvT − Ls T. Deﬁning ui ≡ ∂σ X i . Then Spp = −m = −m ≈ dτ dt ˙ ˙ −X µ Xµ 1 − v2 (1) 1 dt ( mv 2 − m). Ls = is its physical length. u . and T = is its tension.1 Chapter 1 Problem 1. 2 1 2πα′ (5) (7) .
a The curvature of the boundary thus transforms as follows: ′ k′ = ±t′a nb (∂a t′b + Γ′b t′c ) ac (10) (11) (12) = e−ω (k ∓ ta ta nb ∂d ). (11). whereas if it is spacelike then ta ta = 1 and we must use the lower sign. 1/2 (14) Finally. (13) where we have used (9). χ is invariant under Weyl transformations. bc bc (9) and for the curvature scalar. ∂M (15) (16) M we ﬁnd the transformation law for χ: χ′ = 1 4π 1 = 4π = χ. they transform as t′a = e−ω ta . We will prove by explicit computation that. Since the tangent and normal vectors at the boundary are normalized.τ ) γab . and applying Stokes theorem. a a Γ′a = Γa + ∂b ωδc + ∂c ωδb − ∂d ωγ ad γbc . in particular.e. R′ = e−2ω (R − 2∇a ∂ a ω). Hence k′ = e−ω (k + na ∂a ω). Putting all of this together. does not depend on the metric. ds′ = ds eω . n′ = eω na . i. dτ dσ (−γ ′ )1/2 R′ + M 1 2π ds′ k′ ∂M dτ dσ (−γ) M 1/2 (R − 2∇a ∂ a ω) + 1 2π ds (k + na ∂a ω) ∂M (17) . the Euler characteristic of the surface. which says that for any vector v a . (8) For this we will need the transformation law for the connection coeﬃcients.3 It is well known that χ. ′ γab = e2ω(σ. since ds = (−γτ τ )1/2 dτ for a timelike boundary and ds = γσσ dσ for a spacelike bounday.1 CHAPTER 1 6 1. is a topological invariant. dτ dσ (−γ)1/2 ∇a v a = ds na va . If the boundary is timelike then ta ta = −1 and we must use the upper sign.2 Problem 1. and the fact that n and t are orthogonal. (12).
−3/2.0 . the coeﬃcient of exp[−iπmcτ /l] on the RHS must vanish for m = 0: 1 1 25 25 πnσ π(n − m)σ ′ [αn . . Π25 (τ. the ﬁnite result is − 1 2 1 − θ + θ2 . 6 (19) 1. Using (1.3. σ)] 1 25 25 iπ(n + n′ )cτ πn′ σ ′ πnσ i [αn .1 CHAPTER 1 7 1. Not surprisingly. sin n l l (20) where the sum runs over the halfoddintegers. (23) l n n l l . Hermiticity of X 25 implies α25 = (α25 )† . Then we wish to evaluate ∞ n=1 (n − θ) exp[−(n − θ)ǫa] =− =− d exp[−(n − θ)ǫa] d(ǫa) n=1 ∞ d eθǫa d(ǫa) eǫa − 1 1 1 1 θ θ2 d +θ− + − + =− d(ǫa) ǫa 2 12 2 2 1 1 1 = − θ + θ 2 + O(ǫa). Note that there is no p25 . they will come out the same as for the free string (1. σ) = − √ 2α′ l n α25 exp − n iπncτ πnσ .4 Problem 1. sin l l (21) We will now determine the commutation relations among the α25 from the equal time comn mutation relations (1.24b). sin =− l ′n l l l n. 3/2.18). . αn′ ] exp − sin . n = 1/2. −n n i Π25 (τ. the cutoﬀ dependent term is independent of θ. σ) = √ 2α′ n iπncτ πnσ 1 25 αn exp − . Again.7 The mode expansion satisfying the boundary conditions is X 25 (τ.3.3. −1/2. αm−n ] sin sin = δ(σ − σ ′ )δm.25b).n (22) Since the LHS does not depend on τ . − 2 (ǫa) 2 6 ǫa + O(ǫa)2 (18) As expected. . We have: iδ(σ − σ ′ ) = [X 25 (τ. let us deﬁne a ≡ (π/2p+ α′ l)1/2 . σ).5 For simplicity. .3 Problem 1.
n . 48 (26) = where we have used (19) and (25). n m−n as advertised.1 CHAPTER 1 Multiplying both sides by sin[πn′ σ/l] and integrating over σ now yields. . . ∞ nNin + ∞ n=1/2 nN25. 1 2n [α25 . . [α25 .0 . α25 ] sin n m−n π(n + m)σ ′ π(n − m)σ ′ 25 + [α25 . α25 ] = nδm.3. 24) 1 15 = ′ N− . Thus the mass spectrum (1. α−n ] sin n+m l l = sin or. The part of the Hamiltonian (1. l (25) dσ 0 2πα′ Π25 = 2 + 1 ∂σ X 25 2πα′ 2 1 ′ p+ l 4α × n.19) contributed by the X 25 oscillators is l 4πα′ p+ l 8 (24) πnσ ′ δm. (28) i=2 n=1 m2 = − 15 . .3. α 16 where the level spectrum is given in terms of the occupation numbers by 24 1 2α′ p+ ∞ n=1/2 (27) N= The ground state is still a tachyon.36) becomes m2 = 2p+ H − pi pi (i = 2.0 . 16α′ (29) .n′ l α25 α25 exp − n n′ − sin iπ(n + n′ )cτ l dσ 0 πn′ σ πnσ πn′ σ πnσ sin + cos cos l l l l 1 = 4α′ p+ = 1 4α′ p+ 1 2α′ p+ n ∞ α25 α25 n −n 25 α25 α25 + α25 αn n −n −n = n=1/2 ∞ n=1/2 α25 α25 + −n n α25 α25 + −n n n 2 1 .
4. as the second excited state is already massive: m2 = 1 . l (37) 2 ˜ ˜ (N + N + A + A). The third excited state.6d). must now run over the halfoddintegers as it did in Problem 1.7: X 25 (τ. 1.4.1) Lorentz symmetry preserved by the Dbrane. with 9 m2 = .4. 16α′ 9 (30) There are no massless states. [α25 .4.−n . (1.1/2 = 3. and the oscillator label n.4).4.1/2 = 1).8). Thus it is a massive vector with respect to the SO(24.−n .4. rather than running over the nonzero integers. (34) (35) and (therefore) the levelmatching condition (1. m n [˜ 25 . as it can be reached either by Ni. (32) 16α′ is 25fold degenerate and corresponds to a vector plus a scalar on the Dbrane—it can be reached by N25. and is also tachyonic: m2 = − 7 .1/2 = 1.10).9 The mode expansion for X 25 respecting the boundary conditions is essentially the same as the mode expansion (1. m2 = the generator of σtranslations (1. P =− 2π ˜ (N − N ). by N25.1 = 1.5 Problem 1.11). σ) = i α′ 2 α25 2πin(σ − cτ ) 2πin(σ + cτ ) α25 ˜ n exp − + n exp n l n l . 16α′ (31) This state is 24fold degenerate. ˜ N = N.1/2 = 1. or by Ni.1/2 = 2. α′ (36) = mδm. N25.1 CHAPTER 1 The ﬁrst excited state has the lowest X 25 oscillator excited (N25.6c) and (1. α25 ] αm ˜ n as are the mass formula (1. (38) . the only diﬀerences being that the ﬁrst two terms are no longer allowed. α25 ] = mδm. (33) n The canonical commutators are the same as for the untwisted closed string.1 = 1 or by N25.
an antisymmetric tensor. 16 (40) ˜ ˜ At a given level N = N .7: ∞ 24 ∞ 1 ˜ A=A= n n+ 2 n=1 i=2 n=1/2 15 =− . the level operator N is now slightly diﬀerent. (42) 4α′ respectively. it is the same as the level operator for the open string on a D24brane of Problem 1. The zeropoint constants are also the same as in Problem 1. the occupation numbers Nin and Nin may be chosen independently. so long as both sets satisfy (28). The lowest nontachyonic states will again be at level N = 1: a second rank SO(24) tensor with 1 m2 = . (43) 4α′ which can be decomposed into a scalar. We will have tachyons at levels N = 0 and N = 1/2. The leftmoving level spectrum is therefore given by (28).1 CHAPTER 1 10 However. . and a traceless symmetric tensor.7. with m2 = − and m2 = − 15 4α′ (41) 7 . 24 N= ∞ i=2 n=1 i αi αn + −n ∞ n=1/2 α25 α25 . −n n (39) in fact. and the masssquared of that level (36) will be 4 times the open string masssquared (27). Therefore the spectrum at that level will consist of the product of two copies of the Dbrane open string spectrum. and similarly for the rightmoving level ˜ operator N .
z ) ¯ (z2 + ǫ)2 = 0 ∞ dθ f (r. consider a general test function f (r. ¯ d2z ∂ ∂ ln z2 f (z) = ¯1 d2z ∂ f (z) z R 1 dz f (z) = −i z ∂R = 2πf (0). assuming that g is suﬃciently well behaved at zero and inﬁnity. This lead to regularizations also of 1/¯ z and 1/z: z ¯ ǫ ¯ ¯ z ∂ ∂ ln(z2 + ǫ) = ∂ 2 =∂ 2 = . R (2) (b) We regulate ln z2 by replacing it with ln(z2 + ǫ).2 CHAPTER 2 11 2 2. d2z ǫ f (z. θ). θ). R (1) For antiholomorphic test functions f (¯). − ∞ 0 du ǫ ln(u + ǫ)g′′ (u). and deﬁne g(r 2 ) ≡ Then.1 (a) For holomorphic test functions f (z).1 Chapter 2 Problem 2. z ¯ d2z ∂ ∂ ln z2 f (¯) = z 1 d2z ∂ f (¯) z z ¯ R 1 z d¯ f (¯) z =i z ¯ ∂R = 2πf (0). (3) z + ǫ z + ǫ (z2 + ǫ)2 Working in polar coordinates. (4) . du ǫ g(u) (u + ǫ)2 ∞ 0 ǫ g(u) + ǫ ln(u + ǫ)g′ (u) u+ǫ = g(0) = − = 2πf (0).
the set of points z1 satisfying z12  < z23 .2 Problem 2.14).2 CHAPTER 2 12 2.3 (a) The leading behavior of the expectation value as z1 → z2 is n z : eiki ·X(zi .¯2 ) : i=3 z : eiki ·X(zi . R = lim 1 (k + 1)! Γ( 2 α′ k1 · k3 − k) z23 k! Γ( 1 α′ k1 · k3 − k + 1) 2 k→∞ = z23 .j=3 = z12 α k1 ·k2 ′ z : ei(k1 +k2 )·X(z2 .j=1 zij α ki ·kj n ′ = z12 α k1 ·k2 iC X (2π)D δD (k1 + k2 + n ki ) i=3 n × ′ i=3 n z1i α k1 ·ki z2i α k2 ·ki n ′ ′ i. that is. In this case. (b) The zi dependence of the expectation value is given by z23 α k2 ·k3 z12 α k1 ·k2 z13 α k1 ·k3 = z23  α′ k2 ·k3 ′ ′ ′ ′ z12  α′ k1 ·k2 z23  α′ k1 ·k3 ∞ z12 1+ z23 α′ k1 ·k3 k (6) z12 z23 = z23 α (k1 +k2 )·k3 z12 α k1 ·k2 × ′ Γ( 1 α′ k1 · k3 + 1) 2 1 ′ k! Γ( 2 α k1 · k3 − k + 1) k=0 Γ( 1 α′ k1 · k3 + 1) 2 1 k! Γ( 2 α′ k1 · k3 − k + 1) k=0 ∞ z12 ¯ z23 ¯ k . in agreement with (2. The radius of convergence of a power series is given by the limit as k → ∞ of ak /ak+1 . (c) (7) Consider the interior of the dashed line in ﬁgure 2.j=3 n zij α ki ·kj zij α ki ·kj (5) ′ ′ ≈ z12 α k1 ·k2 iC X (2π)D δD (k1 + k2 + n ki ) i=3 i=3 z2i α (k1 +k2 )·ki ′ i.¯i ) : .1.¯i ) : i=1 n n = iC X (2π)D δD i=1 ′ ki i. where the ak are the coeﬃcients of the series. (8) .2. for both of the above power series.
z3 )B(z4 . thereby not aﬀecting the equations of motion: δL = ǫ∂a Ka . 2. The Taylor expansion of a function that is holomorphic on an open disk (about the center of the disk). ∂φα ∂(∂a φα ) (12) Instead of assuming that δφα vanishes at inﬁnity. let us assume that it is a symmetry. It can therefore be written as the sum of a holomorphic and an antiholomorphic function (this statement is true in any simply connected region). converges on the disk. = ∂L ǫ−1 δφα − Ka ∂(∂a φα ) ∂L ∂L ∂a δφα − δL δφα + ∂φα ∂(∂a φα ) (14) . z4 ) ¯ ¯ ¯ ¯ (9) is a harmonic function of z1 within this region. ∂φα ∂(∂a φα ) (10) The Lagrangian equations of motion (EulerLagrange equations) are derived by assuming that the action is stationary under an arbitrary variation δφα (σ) that vanishes at inﬁnity: 0 = δS = = = implies ddσ δL ddσ ddσ ∂L ∂L δφα + ∂a δφα ∂φα ∂(∂a φα ) ∂L ∂L − ∂a δφα ∂φα ∂(∂a φα ) (11) ∂L ∂L − ∂a = 0.5 Under the variation of the ﬁelds φα (σ) → φα (σ) + δφα (σ). z2 ) : A(z3 .4) must converge on the disk. ∂a j a = 2πi∂a 2πi ǫ = 0. the variation of the Lagrangian (10) must be a total derivative to insure that the action on bounded regions varies only by a surface term. the expectation value : X µ (z1 . although it is not obvious how to prove that this can always be arranged. (13) Ka is assumed to be a local function of the ﬁelds and their derivatives. z1 )X ν (z2 .23). similarly for an antiholomorphic function. and (13). Using (10).3 Problem 2. the variation of the Lagrangian is δL = ∂L ∂L δφα + ∂a δφα . (12).2.1. Hence the two Taylor series on the RHS of (2. In this case.2 CHAPTER 2 13 By equation (2.
7 (a) X µ : T (z)X µ (0.2 CHAPTER 2 14 If we now vary the ﬁelds by ρ(σ)δφα (σ).4 Problem 2. 2. where δφα is a symmetry as before but ρ is an arbitrary function. 0) = − 1 1 : ∂X ν (z)∂Xν (z) : X µ (0. so the factor in parentheses must equal ǫ∂a Ka : δS = = = ddσ ddσ ǫ 2πi ǫ∂a Ka ρ + ∂L δφα ∂a ρ ∂(∂a φα ) ∂L δφα ∂a ρ −ǫKa + ∂(∂a φα ) (16) ddσ j a ∂a ρ. then the variation of the action will be δL = ∂L ∂L ∂a (δφα ρ) + δφα ρ ∂(∂a φα ) ∂φα ∂L ∂L ∂L = ∂a δφα + δφα ρ + δφα ∂a ρ.3. this agrees with (2. 0) ∼ ∂X µ (¯) ∼ z ¯ z z α z ¯ 1 ∂X µ (0) z 1¯ µ ∂X (0) z ¯ (17) ∂X µ : 1 1 1 ∂X µ (z) ∼ 2 ∂X µ (0) + ∂ 2 X µ (0) 2 z z z ˜(¯)∂X µ (0) ∼ 0 T z T (z)∂X µ (0) ∼ ¯ ∂X µ : ¯ T (z)∂X µ (0) ∼ 0 1 ¯ 1 ¯ 1¯ ˜ z ¯ T (¯)∂X µ (0) ∼ 2 ∂X µ (¯) ∼ 2 ∂X µ (0) + ∂ 2 X µ (0) z z ¯ z ¯ z ¯ ∂2X µ: T (z)∂ 2 X µ (0) ∼ 2 2 1 2 ∂X µ (z) ∼ 3 ∂X µ (0) + 2 ∂ 2 X µ (0) + ∂ 3 X µ (0) 3 z z z z ˜(¯)∂ 2 X µ (0) ∼ 0 T z (18) (19) (20) . ∂(∂a φα ) ∂φα ∂(∂a φα ) (15) Equation (13) must be satisﬁed in the case ρ(σ) is identically 1. 0) ∼ ∂X µ (z) ∼ α′ z 1 ¯ 1¯ ˜ z T (¯)X µ (0. 0) = − ′ : ∂X ν (¯)∂Xν (¯) : X µ (0. assuming that ρ falls oﬀ at inﬁnity. where we have integrated by parts. Since δ exp(−S) = − exp(−S)δS. ignoring the transformation of the measure.4) for the case of ﬂat space.
α′ 1 ¯ ¯ ¯ ˜ T = − ′ : ∂X µ ∂Xµ : +Vµ ∂ 2 X µ . because it is not an eigenstate of rigid transformations. 0) ∼ (23) Not only is X µ is not a tensor anymore.0) : z : ∂X µ (0)eik·X(0. ∂2X µ: 3α′ V µ z4 ¯2 X ν (¯)∂ 2 X µ (0) ∼ 0 Vν ∂ z Vν ∂ 2 X ν (z)∂ 2 X µ (0) ∼ (25) (26) .0). ∂X µ : Vν ∂ 2 X ν (z)∂X µ (0) ∼ α′ V µ z3 ¯2 X ν (¯)∂X µ (0) ∼ 0 Vν ∂ z (24) So ∂X µ still has weights (1. 0) ∼ z 2¯2 z Vν ∂ 2 X ν (z)X µ (0.0) : ∼ α′ k2 4z 2 α′ k2 ∼ 4z 2 α′ k2 :∼ 4¯2 z α′ k2 ∼ 4¯2 z 1 : eik·X(0.2 CHAPTER 2 : eik·X :: T (z) : eik·X(0.0) : ¯ : ∂X µ (¯)eik·X(0.0) : + ikµ z 1 : eik·X(0.1).0) : : ∂X µ (0)eik·X(0. the energymomentum tensor is 1 : ∂X µ ∂Xµ : +Vµ ∂ 2 X µ . ¯ ∂X µ : ¯ Vν ∂ 2 X ν (z)∂X µ (0) ∼ 0 α′ V µ ¯ Vν ∂ 2 X ν (¯)∂X µ (0) ∼ z ¯ z3 ¯ ¯ Similarly. ∂X µ still has weights (0. but it does not even have welldeﬁned weights. but is no longer a tensor.0) : 15 ˜ z T (¯) : eik·X(0.0) : + ikµ z ¯ : ∂X µ (z)eik·X(0. α T =− (22) ¯ so it suﬃces to calculate the OPEs of the various operators with the terms Vµ ∂ 2 X µ and Vµ ∂ 2 X µ and add them to the results found in part (a).0) : + ikµ z 1 : eik·X(0.0) : + ikµ z ¯ 1 : eik·X(0.0) (21) (b) In the linear dilaton theory. but it is no longer a tensor operator. X µ: α′ V µ 2z 2 α′ V µ ¯ Vν ∂ 2 X ν (¯)X µ (0.
but.0) :∼ ¯ Vν ∂ 2 X ν (¯) : eik·X(0.0).5 Problem 2. Similarly. and ∂ 2 X µ is still not a tensor.9 Since we are interested in ﬁnding the central charges of these theories. For the linear dilaton CFT.0) :e : 2z 2 iα′ V · k ik·X(0.4. ˜ (32) (29) so (30) (31) . In the following. 4 2z z z2 c = D + 6α′ V 2 . the rest of the OPE being determined by general considerations as in equation (2. 4 2¯ z z ¯ z2 ¯ so c = D + 6α′ V 2 . it is only necessary to calculate the 1/z 4 terms in the T T OPEs. T (z)T (0) = 1 : ∂X µ (z)∂Xµ (z) :: ∂X ν (0)∂Xν (0) : α′2 2Vν − ′ : ∂X µ (z)∂Xµ (z) : ∂ 2 X ν (0) α 2Vµ 2 µ − ′ ∂ X (z) : ∂X ν (0)∂Xν (0) : +Vµ Vν ∂ 2 X µ (z)∂ 2 X ν (0) α D 1 3α′ V 2 ∼ 4+ +O .2 CHAPTER 2 16 Nothing changes from the scalar theory: the weights are still (2.0) z iα′ V · k ik·X(0. 4 4 (28) 2. its weights are now complex: α′ 2 α′ (k + 2iV · k). we will therefore drop all terms less singular than 1/z 4 . curiously.0) :∼ :e : 2¯2 z (27) Thus : eik·X : is still a tensor. (k2 + 2iV · k) .25). : eik·X :: Vν ∂ 2 X ν (z) : eik·X(0. 1 ¯ ¯ ¯ ˜ z ˜ z ¯ z T (¯)T (0) = ′2 : ∂X µ (¯)∂Xµ (¯) :: ∂X ν (0)∂Xν (0) : α 2Vν ¯ z ¯ z ¯ − ′ : ∂X µ (¯)∂Xµ (¯) : ∂ 2 X ν (0) α 2Vµ ¯ ¯ ¯ z ¯ z ¯ − ′ ∂ 2 X µ (¯) : ∂X ν (0)∂Xν (0) : +Vµ Vν ∂ 2 X µ (¯)∂ 2 X ν (0) α D 1 3α′ V 2 ∼ 4+ +O .
(34) − λ(1 − λ) : b(z)∂c(z) :: ∂b(0)c(0) : (33) 2.19) vanishes. ˜ (35) c = −12λ2 + 12λ − 2.6 Problem 2. while m < −1 is equivalent to m > 1. 0 = 1 2 m−1 n=1 αµ αµ(−n) 0. While γ(z)β(0) ∼ 1/z as in the bc system.19).11 Assume without loss of generality that m > 1. ˜ The βγ system has the same energymomentum tensor and almost the same OPEs as the bc system. as do all but m − 1 of the terms in the mode expansion (2. 0 . T (z)T (0) = (1 − λ)2 : ∂b(z)c(z) :: ∂b(0)c(0) : − λ(1 − λ) : ∂b(z)c(z) :: b(0)∂c(0) : + λ2 : b(z)∂c(z) :: b(0)∂c(0) : 1 −6λ2 + 6λ − 1 +O . 0 . yields. 0 . 12 . 0 (37) D m(m2 − 1)0. 0 . L−m ]0. so c = 0. when applied to 0. for m = 0 and m = ±1 the central charge term in (2.n 0. 0 = Lm L−m 0. Then Lm annihilates 0. 0 = 1 4 1 4 ∞ m−1 n′ =−∞ n=1 m−1 ∞ αν ′ ανn′ αµ αµ(−n) 0. Of course c = 0 still. so the central charge of the βγ system is minus that of the bc system: c = 12λ2 − 12λ + 2. 0 − L−m Lm 0.6. Each term in (33) involved one b(z)c(0) contraction and one c(z)b(0) contraction.6) of L−m : L−m 0. [Lm .7. 0 m−n n−m ν µ (m − n′ )n′ η νµ ηνµ δn′ n + (m − n′ )n′ δµ δν δm−n′ . 0 = n=1 n′ =−∞ m−1 n=1 = = D 2 n(m − n)0. ∼ 4 z z2 so ˜ z ˜ Of course T (¯)T (0) = 0.2 CHAPTER 2 17 For the bc system. now β(z)γ(0) ∼ −1/z.6. n−m (36) Hence the LHS of (2.
6.16) of b(z) in the wframe using the tensor transformation law (2.4. 0 = (m3 − m)0.19) applied to the same state yields. 12 12 so c = D. ◦ ◦ 2πi 2 (43) (c) If we rewrite the expansion (2.2 CHAPTER 2 18 Meanwhile. the RHS of (2. we ﬁnd.7 Problem 2. z (42) 1 1 dz jz − λ + 2πi 2 1 1 =− dz : b(z)c(z) : −λ + 2πi 2 1 1 =− dz ◦ b(z)c(z)◦ − . 0 .m′ =−∞ ◦ bm cm′ ◦ m+λ z ′m′ +1−λ z ◦ ◦ 1 bm cm′ − m+λ z ′−m+1−λ m+λ z ′m′ +1−λ z z m=0 z z′ 1−λ ∞ = b(z)c(z ′ ) − With (2.16) and (2. z − z′ z z′ 1−λ (40) : b(z)c(z ′ ) : −◦ b(z)c(z ′ )◦ = ◦ ◦ 1 z − z′ −1 .7. : b(z)c(z) : −◦ b(z)c(z)◦ = ◦ ◦ Using (2. c c 2mL0 + (m3 − m) 0. we ﬁnd. N g = Qg − λ + = 1 2 1−λ . (41) (b) By taking the limit of (41) as z ′ → z.7.17). 1 .15).14) we have.m′ =−∞ ∞ m.8. (38) (39) 2. b(w) = (∂z w)−λ b(z) = (−iz)λ = e−πiλ/2 ∞ m=−∞ ∞ m=−∞ bm m+λ z eimw bm .5.13 (a) Using (2. (44) .7).7. ◦ ◦ b(z)c(z ′ )◦ = ◦ = ∞ m.
jw (w) = −b(w)c(w) =i and Ng = − =− =− 1 2πi ∞ m=−∞ ∞ m=−∞ 2π ∞ −πi(1−λ)/2 ∞ m=−∞ eimw cm . z ) ≡ X µ (z ∗ . z ). For instance. ∞ a.m′ =−∞ dw jw 0 bm c−m ◦ ◦ bm c−m ◦ − ◦ ∞ m=0 1. ǫ 2 a 1 − e−ǫa (48) the ǫindependent part is a/2.2 CHAPTER 2 19 Similarly. (47) The ordering constant is thus determined by the value of the second inﬁnite sum. z so that in particular for z on the real line. ∞ m=0 ae−ǫa = = 1 a + + O(ǫ). ¯ ¯ Then ¯ ∂ m X µ (z) = ∂ m X µ (¯∗ ). z ∗ ). deﬁne for ℑz < 0. then we must regulate the sum in such a way that the divergent part is m=0 independent of a. ¯ X µ (z. If we write.8 Problem 2. ignoring ordering. ′ (46) m. ¯ ∂ m X µ (z) = ∂ m X µ (¯). c(w) = e Hence. so the ordering constant in (47) equals −1/2. z (51) (50) (49) . 2.15 To apply the doubling trick to the ﬁeld X µ (z. more generally. (45) ei(m+m )w bm cm′ .
9.7. so on the RHS we are left with the term c (m3 − m) 11 .19) between 1 and 1 . 0 . (58) . 0). for example. L0 1 = 0.9).8. 12 Hence c= 12 L−m 1 m3 − m 11 2 (57) ≥ 0. with the ﬁelds holomorphic −m inside the contours: αµ 0. the contour integrals (52) vanish. 0) ∼ 0.6.8. but with the doubling trick the integral can be extended to the ¯ a semicircle of ∂X z full circle: dz m d¯ m ¯ µ z 2 2 z ∂X µ (z) = − z ∂X (¯). (2. = (55) As in the closed string case.8) in the text. L−m ]1 = 1L† L−m 1 −m = L−m 1 2 . as long as we normal order the resulting local operator. and the state corresponding to the unit operator “inserted” at the origin must be the ground state 0.9 Problem 2.26). with n = −m and m > 1.7. 0 = xµ 0. 0 ∼ X µ (0. ¯ z (52) αµ = m ′ ′ α 2π α 2π At this point the derivation proceeds in exactly the same manner as for the closed string treated in the text. (56) using (2. The result is therefore exactly the same as (2.8. With no operator at the origin. 0 .9). 1[Lm . Also by (2. 0)0.17 Take the matrix element of (2. (m − 1)! (54) Similarly. 0 : 1(0. = (53) The state αµ 0. we see that X µ (0.7a) and (2. 0 ∼ = −m 2 α′ 1/2 i ∂ m X µ (0) = (m − 1)! 2 α′ 1/2 i ¯ ∂ m X µ (0).9. so with m positive. the same correspondence applies when these operators act on states other than the ground state.2 CHAPTER 2 20 as can also be seen from the mode expansion (2. the ﬁelds are holomorphic inside the contour. using the mode expansion (2. so xµ 0. 0 (m positive) is given by evaluating the integrals (52).9) continues to hold.26). 2. The modes αµ are deﬁned as integrals over m µ (z) + ∂X µ (¯). The LHS yields.
3 CHAPTER 3
21
3
3.1
Chapter 3
Problem 3.1
(a) The deﬁnition of the geodesic curvature k of a boundary given in Problem 1.3 is k = −nb ta ∇a tb , (1)
where ta is the unit tangent vector to the boundary and nb is the outward directed unit normal. For a ﬂat unit disk, R vanishes, while the geodesic curvature of the boundary is 1 (since ta ∇a tb = −nb ). Hence 2π 1 χ= dθ = 1. (2) 2π 0 For the unit hemisphere, on the other hand, the boundary is a geodesic, while R = 2. Hence χ= in agreement with (2). (b) If we cut a surface along a closed curve, the two new boundaries will have oppositely directed normals, so their contributions to the Euler number of the surface will cancel, leaving it unchanged. The Euler number of the unit sphere is χ= 1 4π d2σ g1/2 2 = 2. (4) 1 4π d2σ g1/2 2 = 1, (3)
If we cut the sphere along b small circles, we will be left with b disks and a sphere with b holes. The Euler number of the disks is b (from part (a)), so the Euler number of the sphere with b holes is χ = 2 − b. (5) (c) A ﬁnite cylinder has Euler number 0, since we can put on it a globally ﬂat metric for which the boundaries are geodesics. If we remove from a sphere b + 2g holes, and then join to 2g of the holes g cylinders, the result will be a sphere with b holes and g handles; its Euler number will be χ = 2 − b − 2g. (6)
3.2
Problem 3.2
¯¯ (a) This is easiest to show in complex coordinates, where gzz = gz z = 0. Contracting two indices of a symmetric tensor with lower indices by gab will pick out the components where one of the indices is z and the other z . If the tensor is traceless then all such components must vanish. The ¯ only nonvanishing components are therefore the one with all z indices and the one with all z ¯ indices.
3 CHAPTER 3
22
(b) Let va1 ···an be a traceless symmetric tensor. Deﬁne Pn by (Pn v)a1 ···an+1 ≡ ∇(a1 va2 ···an+1 ) − n g ∇ vb . n + 1 (a1 a2 b a3 ···an+1 ) (7)
This tensor is symmetric by construction, and it is easy to see that it is also traceless. Indeed, contracting with ga1 a2 , the ﬁrst term becomes ga1 a2 ∇(a1 va2 ···an+1 ) = 2 ∇ vb , n + 1 b a3 ···an+1 (8)
where we have used the symmetry and tracelessness of v, and the second cancels the ﬁrst: ga1 a2 g(a1 a2 ∇b v b a3 ···an+1 ) =
2(n − 1) a1 a2 2 ga1 a2 ga1 a2 ∇b v b a3 ···an+1 + g ga1 a3 ∇b v b a2 a4 ···an+1 n(n + 1) n(n + 1) 2 = ∇b v b a3 ···an+1 . n
(9)
(c)
T For ua1 ···an+1 a traceless symmetric tensor, deﬁne Pn by T (Pn u)a1 ···an ≡ −∇b ub a1 ···an .
(10)
This inherits the symmetry and tracelessness of u. (d) (u, Pn v) = = =− = d2σ g1/2 ua1 ···an+1 (Pn v)a1 ···an+1 d2σ g1/2 ua1 ···an+1 ∇a1 va2 ···an+1 − n ga a ∇ v b n + 1 1 2 b a3 ···an+1
d2σ g1/2 ∇a1 ua1 ···an+1 va2 ···an+1
T d2σ g1/2 (Pn u)a2 ···an+1 va2 ···an+1
T = (Pn u, v)
(11)
3.3
Problem 3.3
(a) The conformal gauge metric in complex coordinates is gz z = gz z = e2ω /2, gzz = gz z = 0. ¯ ¯ ¯¯ Connection coeﬃcients are quickly calculated: Γz = zz
¯ Γz z z¯ ¯
1 zz g ¯(∂z gz z + ∂z gz z − ∂z gzz ) ¯ ¯ ¯ 2 = 2∂ω, ¯ = 2∂ω,
(12) (13)
3 CHAPTER 3
23
all other coeﬃcients vanishing. This leads to the following simpliﬁcation in the formula for the covariant derivative:
m ···a ···a ∇z Tba1···bnm = ∂z Tba1···bnm + 1 1 i=1 ···c···a Γai Tba1···bn m − zc 1 n a δz i − 2∂ω j=1 n j=1 ···am Γc j Tba1···c···bn zb 1
in other words, it counts the diﬀerence between the number of upper z indices and lower z indices, while z indices do not enter. Similarly, ¯
m n ···a ∇z Tba1···bnm ¯ 1
= ∂ + 2∂ω
m i=1
···a z δbj Tba1···bnm ; 1
(14)
In particular, the covariant derivative with respect to z of a tensor with only z indices is equal to ¯ its regular derivative, and vice versa:
z ···¯ z z ···¯ z ∇z Tz¯···¯ = ∂Tz¯···¯ , ¯ z ¯ z ¯ ∇z T z···z = ∂T z···z . ¯ z···z z···z
¯ ¯ = ∂ + 2∂ω
a δz i ¯
i=1
¯ − 2∂ω
j=1
···a z ¯ δbj Tba1···bnm . 1
(15)
(16)
(b) As shown in problem 3.2(a), the only nonvanishing components of a traceless symmetric tensor with lowered indices have all them z or all of them z . If v is an nindex traceless symmetric ¯ tensor, then Pn v will be an (n + 1)index traceless symmetric tensor, and will therefore have only two nonzero components: (Pn v)z···z = ∇z vz···z 1 2ω n ¯ z e ∇z v z ···¯ = 2 1 2ω n z ···¯ = ∂v ¯ z e 2 = (∂ − 2n∂ω)vz···z ; ¯ ¯ (Pn v)z ···¯ = (∂ − 2n∂ω)vz ···¯. ¯ z ¯ z
(17) (18)
T Similarly, if u is an (n + 1)index traceless symmetric tensor, then Pn u will be an nindex traceless symmetric tensor, and will have only two nonzero components: T (Pn u)z···z = −∇b ub z···z
T (Pn u)z ···¯ = −2e−2ω ∂uz ···¯. ¯ z ¯ z
1 2ω n−1 ¯ z ¯ e ∂uz z ···¯ − 2e−2ω ∂uz···z ¯ 2 ¯ = −2e−2ω ∂uz···z ; =−
= −2e−2ω ∇z uz z···z − 2e−2ω ∇z uzz···z ¯ ¯
(19) (20)
with arbitrary gaugeinvariant insertions f (φ). and deﬁne gauge group near the identity by coordinates ǫ ˆ ∂F A (φζ ) ˆ δB F A (φ) ≡ B ∂ǫ FA = ǫ=0 ˆζ ∂F A ∂ φi ∂φi ∂ǫB . (21) By the gauge invariance of the measure [dζ] on the gauge group. It can be used to reexpress the gaugeinvariant formulation of the path integral. in a gaugeﬁxed way: 1 V [dφ] e−S1 (φ) f (φ) = 1 V 1 = V = [dφ] e−S1 (φ) ∆FP (φ) ζ [dζ] δ F A (φζ ) f (φ) [dζ dφζ ] e−S1 (φ ) ∆FP (φζ )δ F A (φζ ) f (φζ ) [dφ] e−S1 (φ) ∆FP (φ)δ F A (φ) f (φ).4 The FaddeevPopov determinant is deﬁned by. φζ → φ. and in the third line we renamed the variable of integration. ǫ=0 (23) If the are properly behaved (i. In the last line of (22). we can express the delta function appearing in the gaugeﬁxed path integral (22) as a path integral itself: δ F A (φ) = [dBA ]eiBA F A (φ) . parametrize the B . if they have nonzero and linearly independent gradients at ˆ ˆ φ).3 CHAPTER 3 24 3. then the FaddeevPopov determinant is precisely the determinant of δB F A . and if there are no gauge transformations that leave φ ﬁxed. this is a gaugeinvariant function.e. Let φ be on the gauge slice (so F A (φ) = 0). then δB F A will be a nonsingular square matrix. (22) In the second equality we used the gauge invariance of [dφ] e−S1 (φ) and f (φ). (24) Finally. so it suﬃces to ﬁnd an ˆ ˆ expression for it that is valid there.4 Problem 3. (25) . If we choose the coordinates ǫB such that [dζ] = [dǫB ] locally. ∆FP (φ) ≡ [dζ] δ F A (φζ ) −1 . and can be represented as a path integral over ghost ﬁelds: ˆ ∆FP (φ) = = ˆ [dǫB ] δ F A (φζ ) −1 −1 ˆ [dǫB ] δ δB F A (φ)ǫB ˆ = det δB F A (φ) = [dbA dcB ] e−bA δB F A (φ)cB ˆ . ∆FP is evaluated only for φ on the gauge slice.
For such φ we have ˆ F A (φ) = δB F A (φF )ǫB (φ). (31) It is now straightforward to prove that the gaugeﬁxed path integral is independent of the choice of gauge: F [dφ] e−S(φ) ∆FP (φ)δ F A (φ) f (φ) = [dφ′ ]e−S(φ ) ∆G (φ′ )δ GA (φ′ ) f (φ′ ) FP [dφ] e−S(φ) ∆G (φ)δ GA (φ) f (φ). . F Deﬁning ζG (φ) in the same way. we obtain (4. (30) ζF (φ) . in the second line we renamed the variable of integration from φ′ to φ. FP FP (32) −1 −1 (29) ζF (φ) = ζF (φ) . FP (33) ′ = In the ﬁrst line we simultaneously used (32) and the gauge invariance of the measure [dφ]e−S(φ) and the insertion f (φ). ζG φζG Deﬁning φ′ ≡ φζG it follows from (29) that ∆F (φ)δ F A (φ) = ∆G (φ′ )δ GA (φ′ ) . and a unique gauge transformation ζF (φ) that takes φF to φ: ˆζ (φ) φ = φFF . we have. (26) 3.3 CHAPTER 3 25 Putting it all together.3): [dφ dbA dcB dBA ] e−S1 (φ)−bA δB F A (φ)cB +iB F A (φ) A f (φ). F (28) and we can write the factor ∆F (φ)δ(F A (φ)) appearing in the gaugeﬁxed path integral (22) in FP terms of ǫB (φ): F F ˆ ∆F (φ)δ F A (φ) = ∆FP (φF )δ F A (φ) FP ˆ ˆ = det δB F A (φF ) δ δB F A (φF )ǫB (φ) F = δ ǫB (φ) . (27) ˆ For φ near φF .5 ˆ For each ﬁeld conﬁguration φ. ζF (φ) will be near the identity and can be parametrized by ǫB (φ).5 Problem 3. there is a unique gaugeequivalent conﬁguration φF in the gauge ˆ slice deﬁned by the F A .2. the same F coordinates used in the previous problem.
dx = (2π)d (q 2 + x(1 − x)p2 )2 0 (41) .20) is 1 ˜ G(p) ≈ − 2 p (again to lowest order in hab ). But the diagram is multiplied by two factors of d − 2 from the two h vertices.19) is − a1 8π d2p ˜ ˜ hab (p)hcd (−p) (2π)2 1 2 pa pb pc pd − 2δab pc pd + δab δcd p2 . In momentum space.19) in momentum space. to know what we’re aiming for. The traceless part of hab in d dimensions is h′ = hab − h/d. The Ricci scalar. (39) 2d There are three oneloop diagrams with two external gravitons. the Polyakov action (3. depending on whether the gravitons are traceless or trace. is R ≈ (∂a ∂b − δab ∂ 2 )hab . divergences in loop integrals show up as poles in the d plane. so it vanishes when we take d to 2.4. to lowest order in the metric perturbation hab = gab − δab .4. ab while that for h is − (38) d − 2˜ h(p)k · (k + p).3a) is SX = 1 d2σ ∂a X∂a X + ( hδab − hab )∂a X∂b X .2. the Green’s function deﬁned by (3. The momentumspace vertex for h′ is ab ˜ h′ (p)ka (kb + pb ). so the exponent of (3. so part of it (the divergent part that ab would normally be subtracted oﬀ) might survive. p2 (36) (35) (34) To ﬁrst order in hab .7 Let us begin by expressing (3.4. This leaves a coupling between h and ∂a X∂a X with coeﬃcient ab 1/2 − 1/d. In dimensional regularization. Arising as they do in the form of a gamma function. which breaks conformal invariance because the graviton trace couples to X when d = 2. (2π)d k2 (k + p)2 (40) The k integral can be evaluated by the usual tricks: dx 0 dd k ka (kb + pb )k · (k + p) (2π)d (k2 + 2xp · k + xp2 )2 1 dd q (qa − xpa )(qb + (1 − x)pb )(q − xp) · (q + (1 − x)p) .3 CHAPTER 3 26 3.6 Problem 3. We begin by dispensing with the hh diagram. We will use dimensional regularization. these are always simple poles. It is equal to − 1 d−2 4d dd p ˜ ′ ˜ h (p)h(−p) (2π)d ab dd k ka (kb + pb )k · (k + p) . 2 (37) where h ≡ haa (we have set 2πα′ to 1). The hh′ diagram is multiplied by only one factor of d − 2.
24π(d − 2) (46) dx 0 −γ − ln x(1 − x)p2 4π 1 ( − 3x + 3x2 ) − x(1 − x) . so the two terms in the numerator are actually equal ˜ ˜′ after multiplying by h′ (p)hcd (−p)—we see that dimensional regularization has already discarded ab the divergence for us. and the x integral vanishes. 2 × 2 matrices is proportional to the identity matrix. we ﬁnd for the 2graviton contribution to the vacuum amplitude. It remains only to perform the integral for the last term in the numerator of (44). plus the symmetry and tracelessness of h′ . d The divergent part of this is However. (2π)d k2 (k + p)2 (43) Discarding terms that vanish due to the tracelessness of h′ or that are ﬁnite in the limit d → 2 ab yields 1 q2 dd q 1 . (42) dx ( − 3x − 3x2 ) pa pb 2 (2π)d (q 2 + x(1 − x)p2 )2 0 The usual tricks. it is a fact that the symmetric part of the product of two symmetric.3 CHAPTER 3 27 The divergent part of the q integral is independent of x. which is convergent at d = 2: 1 dx 0 d2 q x2 (1 − x)2 pa pb pc pd pa pb pc pd = (2π)2 (q 2 + x(1 − x)p2 )2 4πp2 d2p ˜ ˜ hab (p)hcd (−p) (2π)2 1 0 dx x(1 − x) = pa pb pc pd . (44) The q 4 and q 2 terms in the numerator give rise to divergent integrals. allow us to write the k integral in the ab following way: 1 dx 0 dd q (2π)d 2 4 d(d+2) δac δbd q 1 + d (1 − 2x)2 δac pb pd q 2 + x2 (1 − x)2 pa pb pc pd (q 2 + x(1 − x)p2 )2 . so this diagram vanishes as well. 24πp2 (48) Plugging this back into (43). traceless. this also vanishes upon performing the x integral. p2 4 (49) . which (including a symmetry factor of 4 for the ab cd identical vertices and identical propagators) equals 1 4 dd p ˜ ′ ˜ h (p)h′ (−p) cd (2π)d ab dd k ka (kb + pb )kc (kd + pd ) . The ﬁnite part of (45) is (using this trick a second time) δac δbd p2 8π 1 δac δbd p2 − 2δac pb pd . We are left with just the h′ h′ diagram. 2 (47) Amazingly. 1 96π pa pb pc pd 1 − δab pc pd + δab δcd p2 . Integrating these terms over q yields 1 8π 1 0 d dx Γ(1 − ) 2 x(1 − x)p2 4π d/2−1 (45) 2 × − x(1 − x)δac δbd p2 + (1 − 2x)2 δac pb pd .
σ2 = 0) F. it is clearly the desired expression. σ2 = 0) δXν (σ1 . We must still choose renormalization conditions. just because dimensional regularization has (rather amazingly) thrown away the divergent parts of the loop integrals for us. σ1 ) δ δ ′ [F]r .7) becomes 1 2 ′ ′ dσ1 dσ1 ∆b (σ1 . and d2 σR2 has the wrong dimension. In this case. which is more important than Weyl invariance—without it. σ ′∗ ). instead of a Weyl anomaly.9 Fix coordinates such that the boundary lies at σ2 = 0. δX ν (σ1 ) δXν (σ1 ) (53) The tachyon vertex operator (3. It is Weyl invariant (since the trace h decoupled). then the contraction is eﬀectively doubled: ′ ′ ′ ∆b (σ1 . σ2 = −σ2 .) 3. Following the prescription of problem 2.7 Problem 3. We are therefore free to adjust the coeﬃcients of these two terms in order to achieve diﬀ invariance.5) of [F]r can be done trivially: [F]r = exp Equation (3. (50) ∗ ∗ where σ1 = σ1 . we will impose diﬀ invariance. σ ′ ) + ∆(σ. σ2 = 0). σ2 = 0) . Since (36) is manifestly diﬀ invariant. This is because diﬀinvariant quantities are constructed out of the Ricci scalar. it would be impossible to couple this CFT consistently to gravity. (52) δW [F]r = [δW F]r + 1 2 ′ ′ dσ1 dσ1 δW ∆b (σ1 . However. Locality in real space demands that the counterterms be of the same form as the last two terms in the parentheses in (49). and that would therefore have to be ﬁxed by some additional renormalization condition. (σ1 . σ ′ ) = ∆(σ. and is furthermore quite peculiar.10 for normal ordering operators in the presence of a boundary. but not diﬀ invariant. does not mean that renormalization becomes unnecessary. then the σ2 and σ2 integrations in the deﬁnition (3. (It is worth pointing out that there is no local counterterm quadratic in hab that one could add that is diﬀ invariant by itself. σ1 ) δ δ ′ ′ δX ν (σ1 . If σ and σ ′ both lie on the boundary. σ1 ) = 2∆ (σ1 .6.6. and introduce counterterms to satisfy them. we include in the contraction the image term: ∆b (σ. we have discovered a gravitational anomaly.3 CHAPTER 3 28 This result does not agree with (36).6. 1/2 (54) . with a1 taking the value −1/12. (51) ′ If F is a boundary operator. It therefore appears that.25) is V0 = go σ2 =0 dσ1 g11 (σ1 )[eik·X(σ1 ) ]r .
1 . but by (59) V1 vanishes in this case. α′ (57) dσ1 [∂1 X µ (σ1 )eik·X(σ1 ) ]r . − where in the last equality we have used (56) and (3.26) is go eµ V1 = −i √ 2α′ The spacetime gauge equivalence. e) = V1 (k. σ1 ) ′ σ1 =σ1 (60) ′ = 2 ∂1 δW ∆(σ1 . (62) For this quantity to vanish for arbitrary δω(σ1 ) requires the vector e · kk − k2 e to vanish. σ1 )[∂1 X µ (σ1 )eik·X(σ1 ) ]r 2 = iα′ kµ ∂1 δω(σ1 )[eik·X(σ1 ) ]r − α′ k2 δω(σ1 )[∂1 X µ (σ1 )eik·X(σ1 ) ]r . (63) . so the variation of V1 comes entirely from the variation of the renormalization contraction: δW [∂1 X µ (σ1 )eik·X(σ1 ) ]r 1 δ δ µ ik·X(σ1 ) ′ ′′ ′ ′′ = ]r dσ1 dσ1 δW ∆b (σ1 . where we have used (3. The other possibility is k2 = 0. This will happen if e and k are collinear. σ2 =0 (58) V1 (k.3 CHAPTER 3 29 and its Weyl variation (53) is δW V0 = go = go dσ1 g11 (σ1 ) (δω(σ1 ) + δW ) [eik·X(σ1 ) ]r dσ1 g11 (σ1 ) δω(σ1 ) − 1/2 1/2 1/2 k2 δW ∆b (σ1 . e · k = 0.11) in the last equality: ′ δW ∆b (σ1 . σ1 ) [eik·X(σ1 ) ]r 2 (55) = (1 − α′ k2 )go dσ1 g11 (σ1 )δω(σ1 )[eik·X(σ1 ) ]r .15a): ′ ∂1 δW ∆b (σ1 . σ1 ) ′ σ1 =σ1 = α′ ∂1 δω(σ1 ).6. e + λk).6. σ1 ) ν ′ ′′ [∂1 X (σ1 )e 2 δX (σ1 ) δXν (σ1 ) ′′ = ikµ ∂1 δW ∆b (σ1 .6. (61) Integration by parts yields δW V1 = −i α′ go (e · kkµ − k2 eµ ) 2 dσ1 δω(σ1 )[∂1 X µ (σ1 )eik·X(σ1 ) ]r . σ1 ) ′′ σ1 =σ1 [eik·X(σ1 ) ]r k2 δW ∆b (σ1 . (56) Weyl invariance thus requires k2 = The photon vertex operator (3. σ1 ) = 2α′ δω(σ1 ). σ1 ) = 2δW ∆(σ1 . The expression (58) has no explicit metric dependence. (59) is clear from the fact that kµ ∂1 X µ eik·X is a total derivative.
. 6πα′ (69) (70) (71) where we have split it into the action for a free CFT and an interaction term. . (73) . Φ to vanish...13b). quadratic in H. ¯ g1/2 gab ∂a X µ ∂b X ν = 4∂X (µ ∂X ν) . (72) where f is the path integral calculated using only the free action (70). The path integral is now .4. that we are interested in. f f f − Si .3 CHAPTER 3 30 3. 2 i f + ··· . and Bµν to be linear in X. let us assume Gµν to be constant.7. the sigma model action becomes Sσ = 1 4πα′ 1 = 4πα′ d2σ g1/2 Gµν gab ∂a X µ ∂b X ν + i∂ω Bµν ǫab X ω ∂a X µ ∂b X ν i d2σ g1/2 Gµν gab ∂a X µ ∂b X ν + Hωµν ǫab X ω ∂a X µ ∂b X ν 3 . implying that Hωµν = 3∂[ω Bµν] (64) is constant.. (65) In the second line we have used the fact that ǫab X ω ∂a X µ ∂b X ν is totally antisymmetric in ω. and in particular the part proportional to ¯ d2z : ∂X µ ∂X ν : . 2πα′ 1 ¯ Si = Hωµν d2z X ω ∂X µ ∂X ν . It is the Weyl variation of the third term. while that of the second B gives rise to the term in βµν that is linear in H (3. 2 (66) (67) (68) the action becomes Sσ = Sf + Si . ¯ g1/2 ǫab ∂a X µ ∂b X ν = −4i∂X [µ ∂X ν] .11 Since we are interested in the H 2 term. Working in conformal gauge on the worldsheet and transforming to complex coordinates. 1 d2σ = d2z. With these simpliﬁcations. ν (up to integration by parts) to antisymmetrize ∂ω Bµν .. µ. + 1 2 S . = . .8 Problem 3. . . 1 ¯ Sf = Gµν d2z ∂X µ ∂X ν . f .. σ = e−Si . .. The Weyl variation of the ﬁrst term gives rise to the D − 26 Weyl anomaly calculated in section 3.
z ′ )∂ ′ X µ (z ′ )∂ ′ X ν (¯′ ) : . picking a representative pair of crosscontractions. f . f d2z ′ ′ . f . The Weyldependent part of the second integral of (77) is then 1 ∼ −2π ln(ǫe−ω(z) ) = −2π ln ǫ + 2πω(z). (80) . z )X ν (z ′ . (75) 2 so. The Weyl variation of this integral will come from the singular part of the OPE when z and z ′ approach each other. since they can all be obtained from each other by integration by parts and permuting the indices ω. but. z σ (79) involves higher powers Using (66) and (68). This third term is 31 1 2 S . z )X ν (z ′ . z d2zd2z ′ ′ z − z2 ¯ × : ∂X µ (z)∂ ′ X ν (¯′ ) : .3 CHAPTER 3 G whose coeﬃcient gives the H 2 term in βµν . . .. . . 2 i × f = 1 Hωµν Hω′ µ′ ν ′ 2(6πα′ )2 ′ ′ ′ (74) ¯ ¯ d2zd2z ′ : X ω (z. . . we can write this as − 1 Hµλω Hν λω 16π and f d2σ g1/2 δωgab : ∂a X µ ∂b X ν : . . . and the fact that the diﬀerence between of H (see (72)) which we can neglect. Terms in the OPE containing exactly two X ﬁelds (which will yield (73) after the z ′ integration is performed) are obtained by performing two crosscontractions. . so a diﬀinvariant cutoﬀ would be at z ′ − z = ǫe−ω(z) . σ. z ′ f (76) The diﬀinvariant distance between z ′ and z is (for short distances) eω(z) z ′ − z. z ′ ) =: X µ (z.. (78) d2z ′ ′ z − z2 and the Weyl variation of (76) is − 1 Hµλω Hν λω 8π ¯ d2z δω(z) : ∂X µ (z)∂X ν (¯) : . . (77) z 2 µλω ν 16π z − z2 1 ¯ : ∂X µ (z)∂ ′ X ν (¯′ ) : . ν. ¯ z ¯ z where we have normalordered the interaction vertices. they all give the same result. The contraction derived from the free action (70) is α′ µν G ln z − z ′ 2 . so we can drop the less singular terms coming from the Taylor expansion of ¯ ∂ ′ X ν (¯′ ): z 1 1 ¯ − H H λω d2z : ∂X µ (z)∂X ν (¯) : . f . There are 18 diﬀerent pairs of crosscontractions one can apply to the integrand of (74). . z )∂X µ (z)∂X ν (¯) :: X ω (z ′ . z ′ ) : − ¯ ¯ ¯ ¯ 18 Hωµν Hω′ µ′ ν ′ 2(6πα′ )2 α′ × d2zd2z ′ − 2 1 Hµλω Hν λω =− 16π 2 Gωµ ∂ ′ ln z − z ′ 2 − ′ α′ 2 ′ ¯ Gνω ∂ ln z − z ′ 2 The Weyl variation of this term comes from cutting oﬀ the logarithmically divergent integral of z ′ − z−2 near z ′ = z. . µ. the part of (74) we are interested in is X µ (z.
13 If the dilaton Φ is constant and D = d+ 3. (88) r Similarly. (90) h d − 23 . The Ricci tensor on a 3sphere of radius r is given by 2 Rij = 2 Gij . According to (3. k be indices on the 3sphere and α. α′ 4 (83) (84) (85) Letting i. because the volume form is always covariantly constant on a manifold. (Note that this form for H cannot be obtained as the exterior derivative of a nonsingular gauge ﬁeld B. β.6). equation (85) ﬁxes h in terms of d: 2(d − 23) h2 = . γ be indices on the ﬂat ddimensional spacetime. then the equations of motion (3. to leading order in α′ .) Equation (84) is then immediately satisﬁed. ǫikl ǫj kl = 2Gij . T ′a a in turn conG tributes the following term to βµν : α′ − Hµλω Hν λω . but those for which both indices are on the sphere ﬁx r in terms of h: 4 6α′ r2 = 2 = .3 CHAPTER 3 32 This is of the form of (3. B must have a Diractype singularity somewhere on the sphere.7.4. (89) Most components of equation (83) vanish trivially. j. with all other components vanishing. and all other components vanish trivially. (86) where h is a constant and ǫ is the volume form on the sphere. d − 23 1 − Hµνλ H µνλ = 0. 4 ∇ω Hωµν = 0. (82) 4 3.7. (87) 3α′ implying that there are solutions only for d > 23. Since ǫijk ǫijk = 6.15) become. so ∇i Hijk = 0. we apply the ansatz Hijk = hǫijk .9 Problem 3. 1 Rµν − Hµλω Hν λω = 0. with T ′a a = 1 Hµλω Hν λω gab ∂a X µ ∂b X ν 8 (81) being the contribution of this term to the stress tensor.12).
the same as in the lightcone quantization. we had a symmetric rank 2 tensor. −1 −1 and a vector. k0. k′ = 0. (1) (2) Together. they make up the traceless symmetric rank 2 tensor representation of SO(D − 1). Since the particle is massive. k . i ei α−2 0. e. ki = 0. −1 (11) . fµν αµ αν + eµ αµ 0. fij αi αj 0. let us recall the spectrum of the open string at level N = 2 in lightcone quantization. 2 As in the cases of the tachyon and photon. ke. The L1 condition ﬁxes e in terms of f . k′ .1 To begin. k . k . In representations of SO(D − 2). we can √ go to its rest frame for simplicity: k0 = 1/ α′ . f . Its norm is ∗ e. the general state at level 2 is f. (10) or m2 = 1/α′ . In the OCQ. k . k √ =2 2α′ fµν kν + eµ αµ 0. e. f . the L0 condition yields the massshell condition: 0 = (L0 − 1)f. e. −1 −1 −2 a total of D(D + 1)/2 + D states.4 CHAPTER 4 33 4 4. whose dimension is D(D − 1)/2 − 1. e. This is what we expect to ﬁnd. k fρσ αρ ασ + e∗ αρ ρ 2 1 1 ∗ ∗ = 2 fµν f µν + eµ eµ (3) 0. removing D degrees of freedom: 0 = L1 f. k (5) (6) (7) (8) (9) = (α′ k2 + 1)f.1 Chapter 4 Problem 4. k′ −1 −1 −2 (4) The terms in the mode expansion of the Virasoro generator relevant here are as follows: L0 = α′ p2 + α−1 · α1 + α−2 · α2 + · · · √ L1 = 2α′ p · α1 + α−1 · α2 + · · · √ 1 L2 = 2α′ p · α2 + α1 · α1 + · · · 2 √ ′p · α L−1 = 2α −1 + α−2 · α1 + · · · √ 1 L−2 = 2α′ p · α−2 + α−1 · α−1 + · · · . k = fµν αµ αν + eµ αµ 0. k .
this implies fii = 5f00 . corre√ sponding to (15) with γ = 2f . √ 2(D − 1) D−1 fij = f δij . g0 = 3 2f . (17) 25 positive for D < 26 and negative for D > 26. fii = 0.) The norm of this state is proportional to (D − 1)(26 − D)f 2 ∗ ∗ fµν f µν + eµ eµ = . γ. (12) 0 = L2 f.4 CHAPTER 4 34 implying eµ = The L2 condition adds one more constraint: √ 2f0µ . k . (States with vanishing f00 must be traceless by (14). Removing it from the spectrum leaves us with the states fij . (16) 5 5 with all other components zero. e0 = f. Removing these D − 1 states from the spectrum leaves D(D − 1)/2 states. e. k . and this is the unique state satisfying (12) and (14) that is orthogonal to all of these. k −1 √ √ γ = 2α′ g(µ kν) + ηµν αµ αν 0. f00 = f. There are D + 1 independent spurious states at this level: g. In the case D = 26. . −1 −1 −2 2 (13) (14) (15) These states are physical and therefore null for g0 = γ = 0. where fii is the trace on the spacelike part of f . Using (12). this state is spurious. e = 0—precisely the traceless symmetric rank 2 tensor of SO(25) we found in the lightcone quantization. k + gµ + 2α′ γkµ αµ 0. k √ µ = 2 2α′ kµ eµ + fµ 0. the extra one with respect to the lightcone quantization being the SO(D − 1) scalar. k = L−1 gµ αµ + L−2 γ 0.
5 CHAPTER 5
35
5
5.1
Chapter 5
Problem 5.1
(a) Our starting point is the following formal expression for the path integral: Z(X0 , X1 ) =
X(0)=X0 X(1)=X1
[dXde] exp (−Sm [X, e]) , Vdiﬀ
(1)
where the action for the “matter” ﬁelds X µ is Sm [X, e] = 1 2
1 0
dτ e e−1 ∂X µ e−1 ∂Xµ + m2
(2)
(where ∂ ≡ d/dτ ). We have ﬁxed the coordinate range for τ to be [0,1]. Coordinate diﬀeomorphisms ζ : [0, 1] → [0, 1], under which the X µ are scalars, X µζ (τ ζ ) = X µ (τ ), and the einbein e is a “covector,” (3)
dτ , (4) dτ ζ leave the action (2) invariant. Vdiﬀ is the volume of this group of diﬀeomorphisms. The e integral in (1) runs over positive functions on [0,1], and the integral eζ (τ ζ ) = e(τ )
1
l≡
dτ e
0
(5)
is diﬀeomorphism invariant and therefore a modulus; the moduli space is (0, ∞). In order to make sense of the functional integrals in (1) we will need to deﬁne an inner product on the space of functions on [0,1], which will induce measures on the relevant function spaces. This inner product will depend on the einbein e in a way that is uniquely determined by the following two constraints: (1) the inner product must be diﬀeomorphism invariant; (2) it must depend on e(τ ) only locally, in other words, it must be of the form
1
(f, g)e =
0
dτ h(e(τ ))f (τ )g(τ ),
(6)
for some function h. As we will see, these conditions will be necessary to allow us to regularize the inﬁnite products that will arise in carrying out the functional integrals in (1), and then to renormalize them by introducing a counterterm action, in a way that respects the symmetries of the action (2). For f and g scalars, the inner product satisfying these two conditions is
1
(f, g)e ≡
dτ ef g.
0
(7)
5 CHAPTER 5
36
We can express the matter action (2) using this inner product: lm2 1 . Sm [X, e] = (e−1 ∂X µ , e−1 ∂Xµ )e + 2 2 (8)
We now wish to express the path integral (1) in a slightly less formal way by choosing a ﬁducial einbein el for each point l in the moduli space, and replacing the integral over einbeins by an integral over the moduli space times a FaddeevPopov determinant ∆FP [el ]. Deﬁning ∆FP by 1 = ∆FP [e]
0 ∞
dl
[dζ] δ[e − eζ ], l
(9)
we indeed have, by the usual sequence of formal manipulations, Z(X0 , X1 ) =
0 ∞
dl
X(0)=X0 X(1)=X1
[dX] ∆FP [el ] exp (−Sm [X, el ]) .
(10)
To calculate the FaddeevPopov determinant (9) at the point e = el , we expand e about el for small diﬀeomorphisms ζ and small changes in the modulus: el − eζ = ∂γ − l+δl del δl, dl (11)
where γ is a scalar function parametrizing small diﬀeomorphisms: τ ζ = τ + e−1 γ; to respect the ﬁxed coordinate range, γ must vanish at 0 and 1. Since the change (11) is, like e, a covector, we will for simplicity multiply it by e−1 in order to have a scalar, and then bring into play our inner l product (7) in order to express the delta functional in (9) as an integral over scalar functions β: ∆−1 [el ] = FP dδl[dγdβ] exp 2πi(β, e−1 ∂γ − e−1 l l del δl)el dl (12)
The integral is inverted by replacing the bosonic variables δl, γ, and β by Grassman variables ξ, c, and b: ∆FP [el ] = = 1 del (b, e−1 ∂c − e−1 ξ)el l l 4π dl 1 1 del [dcdb] (b, e−1 )el exp (b, e−1 ∂c)el l l 4π dl 4π dξ[dcdb] exp
.
(13)
We can now write the path integral (10) in a more explicit form: Z(X0 , X1 ) =
0 ∞
dl
X(0)=X0 X(1)=X1
[dX]
c(0)=c(1)=0
[dcdb]
1 del (b, e−1 )e l 4π dl l
(14)
× exp (−Sg [b, c, el ] − Sm [X, el ]) , where Sg [b, c, el ] = − 1 (b, e−1 ∂c)el . l 4π (15)
5 CHAPTER 5
37
(b) At this point it becomes convenient to work in a speciﬁc gauge, the simplest being el (τ ) = l. Then the inner product (7) becomes simply
1
(16)
(f, g)l = l
0
dτ f g.
(17)
In order to evaluate the FaddeevPopov determinant (13), let us decompose b and c into normalized eigenfunctions of the operator ∆ = −(e−1 ∂)2 = −l−2 ∂ 2 : l b0 b(τ ) = √ + l c(τ ) = with eigenvalues νj = The ghost action (15) becomes 1 Sg (bj , cj , l) = − 4l
∞ j=1
(18)
2 l
∞ j=1
bj cos(πjτ ),
(19) (20)
2 l
∞ j=1
cj sin(πjτ ),
π2 j 2 . l2
(21)
jbj cj .
(22)
The zero mode b0 does not enter into the action, but it is singled out by the insertion appearing in front of the exponential in (13): 1 b0 del (23) (b, e−1 )el = √ . l 4π dl 4π l The FaddeevPopov determinant is, ﬁnally, ∆FP (l) = =
∞ j=0 ∞
dbj
∞ j=1
dcj
j 1 √ 4π l j=1 4l 1 √ det′ 4π l ∆ 16π 2
b0 1 √ exp 4l 4π l
∞ j=1
jbj cj
1/2
=
,
(24)
the prime on the determinant denoting omission of the zero eigenvalue.
X1 ) = 0 ∞ dl (X − X0 )2 lm2 1 √ exp − 1 − 2l 2 4π l det′ ∆ (1−D)/2 . (24). (d) Putting together the results (10). xj ) = + 2 2l l and the matter part of the path integral (10) X(0)=X0 X(1)=X1 ∞ j=1 j 2 x2 + j lm2 . where we have conveniently chosen to work in a Euclidean spacetime in order to make all of the Gaussian integrals convergent. µ Xcl (τ ) = X0 + (X1 − X0 )τ. (30) . j (26) The matter action (8) becomes (X1 − X0 )2 π 2 Sm (X0 . we have: Z(X0 . (29) We will regularize the determinant of ∆ in the same way as it is done in Appendix A.5 CHAPTER 5 Let us decompose X µ (τ ) into a part which obeys the classical equations of motion. and can therefore be decomposed into the same normalized eigenfunctions of ∆ as c was (20): µ X µ (τ ) = Xcl (τ ) + 2 l ∞ j=1 xµ sin(πjτ ). the ﬂuctuations vanish at 0 and 1.1. X1 . by dividing by the determinant of the operator ∆ + Ω2 : det′ ∆ = det′ (∆ + Ω2 ) = ∞ j=1 π2 j 2 π 2 j 2 + Ω2 l 2 Ωl sinh Ωl ∼ 2Ωl exp (−Ωl) . 2 (27) [dX] exp (−Sm [X. 38 (c) (25) plus quantum ﬂuctuations. and dropping the irrelevant constant factors multiplying the operator ∆ in the inﬁnitedimensional determinants. el ]) (X1 − X0 )2 lm2 = exp − − 2l 2 = exp − (X1 − X0 2l )2 − lm2 2 D ∞ µ=1 j=1 det′ ∆ π π2 dxµ exp − 2 j l −D/2 ∞ j=1 j 2 x2 j (28) . and (28).
(34) The integral is most easily done after passing to momentum space: ˜ Z(k) ≡ = 0 dDX exp (ik · X) Z(0. k 2 + m2 (35) neglecting the constant factors. 1 Sct = 0 dτ eA = lA (32) The mass m is renormalized by what is left over after the cancellation of inﬁnities. phys (33) but for simplicity we will assume that a renormalization condition has been chosen that sets mphys = m. We can now proceed to the integration over moduli space: Z(X0 . X) ∞ dl l−D/2 exp − D/2 = = π 2 π 2 D/2 lm2 X2 dDX exp ik · X − 2 2l ∞ 2 + m2 ) l(k dl exp − 2 0 2 . this is precisely the momentum space scalar propagator. The inverse divergence due to the factor of Ω(1−D)/2 in front of the integral can be dealt with by a ﬁeld renormalization. m2 = m2 − (D − 1)Ω − 2A. X1 ) = 1 4π(2Ω)(D−1)/2 ∞ 0 (31) dl l−D/2 exp − (X1 − X0 2l )2 − l(m2 − (D − 1)Ω) 2 . X1 ) = 0 ∞ dl l−D/2 exp − (X1 − X0 )2 lm2 − 2l 2 . but since we will not concern ourselves with the overall normalization of the path integral we will simply drop all of the factors that appear in front. The divergence coming from the Ω term in the exponent can be cancelled by a (diﬀeomorphism invariant) counterterm in the action. . The path integral (29) becomes Z(X0 .5 CHAPTER 5 39 where the last line is the asymptotic expansion for large Ω.
so its determinant provides the ith term in the sum.2. 2 z1 2 z2 . . . . 2 n−2 1 1 zn zn · · · zn 1 1 . t) = (4) so the pole at α′ s = J − 1 arises from the ﬁrst gamma function in the numerator.6 CHAPTER 6 40 6 6. . α′ ki = 4 and the expectation value is smooth at ui = 0. .1 In terms of u = 1/z. ′ 2 (1) 2 Since this is an expectation value of closedstring tachyon vertex operators. ′ relabeling z5 by z1 . .3 For any n ≥ 2 numbers zi . we have n (−1)i i=1 j<k j. (6..5 Γ(−1 − α′ s)Γ(−1 − α′ t) . α′ (6) (5) . The residue of Γ(z) at z a nonpositive integer is (−1)z /Γ(1 − z). Specializing to the case n = 5. .1 Chapter 6 Problem 6. Γ(−2 − α′ s − α′ t) (a) We have I(s. .31) is δ ( d ki ) i<j 1 1 − ui uj α′ ki ·kj = δd ( = δd ( ki ) i<j uij α ki ·kj ui uj −α ki ·kj uij α ki ·kj ′ ′ ′ ki ) i<j i ui α ki . . and dividing by z11′ z21′ z31′ z41′ yields − z12 z13 z14 z23 z24 z34 z23 z24 z34 z13 z14 z34 z12 z14 z24 z12 z13 z23 . (2) The minor of the matrix with respect to the ﬁrst entry in the ith row is a Vandermonde matrix for the other zj .3 Problem 6. . . Using s+t+u = − 4 . n−2 · · · z1 n−2 · · · z2 = 0. ′ t) Γ(J + 1)Γ(−1 − J − α J! a polynomial of degree J in t.k=i zjk 1 1 = . 6. + − + = z11′ z21′ z31′ z41′ z11′ z21′ z31′ z41′ (3) which is what we are required to prove. z1 z2 . . . 6. t) is (−1)J Γ(−1 − α′ t) 1 = (2 + α′ t)(3 + α′ t) · · · (J + 1 + α′ t). so the residue of I(s.2 Problem 6.
kl . equal to. as promised. the residue of the pole in I(s.6 CHAPTER 6 once s is ﬁxed at (J − 1)/α′ . zero.kl = δij D−1 while the one that projects them onto traceless symmetric matrices is 2 Pij. 2 2α′ 41 (7) so (5) is also a polynomial of degree J in t − u. t can be expressed in terms of t − u: t= t−u J +3 − . k1 = k2 = −k3 = −k4 = . so in its rest frame we have √ s i i i i 0 0 0 0 k1 = −k2 . 4α′ (9) t = −(k1 + k3 )2 5 i i = − ′ − 2k1 k3 . 2 (13) i j k l Inserting the linear combination β0 P 0 + β2 P 2 between the matrices k1 k1 and k3 k3 yields (β0 − β2 ) i i j j k1 k1 k3 k3 25 i i i i + β2 (k1 k3 )2 = (β0 − β2 ) + β2 (k1 k3 )2 . 25 β2 = 2α′ . ′2 (D − 1) D−1 16α (14) Comparison with (11) reveals β0 = 2α′2 (26 − D) . (8) 2 Specializing to the case where all the external particles are tachyons (k2 = 1/α′ ) and the intermediate state is at level 2 (s = 1/α′ ). 2 8 (10) (11) The operator that projects matrices onto multiples of the identity matrix in D − 1 dimensional space is 1 0 δkl . (15) so that. we further have i i i i i i i i k1 k1 = k2 k2 = k3 k3 = k4 k4 = i i It also determines t in terms of k1 k3 : 5 . . k3 = −k4 . t) at α′ s = 1 is 1 1 i i (2 + α′ t)(3 + α′ t) = − + 2α′2 (k1 k3 )2 . or negative depending on whether D is less than.kl = 1 0 (δik δjl + δil δjk ) − Pij. or greater than 26. 2α Using (5). β0 is positive. (12) Pij. (b) The momentum of the intermediate state in the s channel is k1 + k2 = −(k3 + k4 ).
there is another state b in the spectrum whose norm is positive for D < 26 and negative for D > 26. k1 0. For any D it includes D(D − 1)/2 − 1 positivenorm states transforming in the spin 2 representation of the little group SO(D − 1).kl k3 k3 (21) if D < 26. (18) Summing over an orthonormal basis in the space of symmetric traceless matrices yields the contribution of these states to (16): k l i j k l i j 2 a∗ k1 k2 akl k3 k4 = k1 k1 Pij. where v µ (y1 ) = −2iα′ µ k2 kµ − 3 y12 y13 (22) . More speciﬁcally. allowing the intermediate state in the s channel to go on shell. the Smatrix elements involved in (16) are ﬁxed by SO(D − 1) invariance and (anti)linearity in the polarization matrix a: i j aS0. ij a (19) This explains the positive value of β2 found in (15). −k4 . ij k l 0. to allow for the possibility that some of them might have negative norm. k2 . k2 ∝ δij k1 k2 . and a negative multiple if D > 26. −k3  0.11) follows immediately from (6. k2 ∝ a∗ k1 k2 .36). (20) Its contribution to (16) is therefore clearly a positive multiple of i j 0 k l k1 k1 Pij.6 CHAPTER 6 42 What does all this have to do with the open string spectrum at level 2? The amplitude I has a pole in s wherever s equals the masssquared of an open string state. Since this state is an SO(D − 1) scalar. The residue of this pole can be written schematically as f S o o oo Si . k1 0. −k4 Sb ∝ δkl k3 k4 . f  = 0. 6. i = 0.7 (a) The X path integral (6. k1 0.5.1. (16) o where the sum is taken over open string states at level J = α′ s + 1 with momentum equal to that of the initial and ﬁnal states. −k3  0. the Smatrix elements connecting it to the initial and ﬁnal states are given by: i j bS0. For D = 26. −k4 Sa ∝ akl k3 k4 . k l 0. Working in the rest frame of such a state.2.4 Problem 6. −k3  0. we have not assumed that the intermediate states are normalized.kl k3 k3 . (17) The open string spectrum at level 2 was worked out as a function of D in problem 4.
2 so SD2 (k1 . a2 . k3 .11) therefore simpliﬁes to ⋆ ⋆ ˙ X µ eik1 ·X (y1 )⋆ ⋆ eik2 ·X (y2 )⋆ ⋆ eik3 ·X (y3 )⋆ ⋆⋆ ⋆⋆ ⋆ D2 (26) 1 2 y23 µ kµ k2 + 3 y12 y13 X = 2α′ CD2 (2π)26 δ26 (k1 + k2 + k3 ) .10) and using (6.2): c(y1 )c(y2 )c(y3 ) D2 g = CD2 y12 y13 y23 . k1 + k2 + k3 = 0. ′ −ig0 (2π)26 δ26 (k1 (30) (31) + k2 + k3 )e1 · k23 Tr(λ [λ . ′ α α 1 1 2 = k2 = (k1 + k3 )2 = ′ + 2k1 · k3 . k2 . g 2 X α′ go e−λ CD2 CD2 = 1. λ ]).5. α′ 1 1 2 = k3 = (k1 + k2 )2 = ′ + 2k1 · k2 . a3 ) ′ = −2igo (2π)26 δ26 (k1 + k2 + k3 ) (29) y13 e · k2 + y12 e · k3 Tr(λa1 λa2 λa3 ) + (2 ↔ 3). The ghost path integral is given by (6. while 2α′ k1 · k2 = 2α′ k1 · k3 = 0.12). and the mass shell conditions imply 2 + 2k2 · k3 .6 CHAPTER 6 43 ˙ (we leave out the contraction between X µ (y1 ) and eik1 ·X(y1 ) because their product is already renormalized in the path integral). y23 Momentum conservation and the physical state condition e1 · k1 = 0 imply 1 e1 · k2 = −e1 · k3 = e1 · k23 .5. e1 . (28) yields SD2 (k1 . (27) Putting these together with (6. a2 .3. k3 . ′ α α 2 0 = k1 = (k2 + k3 )2 = (23) (24) (25) so that 2α′ k2 · k3 = −2. a1 . a1 . k2 .4. Equation (6. e1 . Momentum conservation.5. a3 ) = in agreement with (6. a1 a2 a3 .14).
a3 . a1 . e. e.6 CHAPTER 6 44 (b) Using equations (6. Comparing (32) and (33). which after the integration over k in the third equality becomes e·(k1 +k2 ) = e·(k3 +k4 ) = 0. the 4tachyon amplitude near the pole at s = 0 has the form SD2 (k1 . a2 . s We can calculate the same quantity using unitarity.5. a. λa2 ][λa3 .e −k2 + iǫ =i d26k (2π)26 a. k2 .20). k2 .5. a2 )SD2 (k. k4 . e · k12 e · k34 = k12 · k34 = u − t. λa2 ]) −k2 + iǫ ′ ×(−i)g0 (2π)26 δ26 (k + k3 + k4 )e · k34 Tr(λa [λa3 . a3 . λa2 ][λa3 . k3 . a1 . The stateoperator mapping gives the same normalization: in problem 2.6). we saw that . a4 ) d26k SD2 (−k. (34) e In the last equality of (33) we have also applied equation (6. k4 . a3 . This result conﬁrms the normalization of the photon vertex operator as written in equation (3.4. k3 . λa4 ]) e ki ) s + iǫ a ′2 = −igo (2π)26 δ26 ( ′2 = −igo (2π)26 δ26 ( i ki )Tr([λa1 .5. λa4 ]) i u−t .5. λa4 ]) e · k12 e · k34 Tr(λa [λa1 . then none of the others will contribute to the sum in the second to last line. k4 . and (6. we see that the fourtachyon amplitude near s = 0 is given by SD2 (k1 .26). λa4 ]) i u−t . a2 . a4 ) =i (2π)26 a. a4 ) ig2 = o (2π)26 δ26 ( ki ) α′ i ×Tr(λa1 λa2 λa4 λa3 + λa1 λa3 λa4 λa2 − λa1 λa2 λa3 λa4 − λa1 λa4 λa3 λa2 ) =− 2 igo (2π)26 δ26 ( 2α′ u−t 2s (32) ki )Tr([λa1 . k3 . which becomes. k1 .9).12)). we see that go ′ go = √ . If we choose one of the vectors in this basis to be e′ = k12 /k12 . λa2 ])Tr(λa [λa3 .e 1 ′ (−i)g0 (2π)26 δ26 (k1 + k2 − k)e · k12 Tr(λa [λa1 . (35) 2α′ in agreement with (6.6. k2 .14). a1 . (6.17).4. By analogy with equation (6.4.9. a. s + iǫ (33) In the second equality we have substituted equation (31) (or (6.13). The polarization vector e is summed over an orthonormal basis of (spacelike) vectors obeying the physical state condition e·k = 0.
The contribution this ordering makes to the amplitude is 4 e−λ go (2α′ )−2 Tr(λa3 λa4 λa1 λa2 )e1 1 e2 2 e3 3 e4 4 µ µ µ µ y1 × dy4 y3 ⋆ ⋆ ˙ ˙ c1 X µ3 eik3 ·X (y3 )⋆ ⋆ X µ4 eik4 ·X (y4 )⋆ ⋆⋆ ⋆ ˙ ˙ × ⋆ c1 X µ1 eik1 ·X (y1 )⋆ ⋆ c1 X µ2 eik2 ·X (y2 )⋆ ⋆ ⋆⋆ ⋆ (40) ki ) i g 4 X = e−λ go (2α′ )−2 Tr(λa3 λa4 λa1 λa2 )e1 1 e2 2 e3 3 e4 4 iCD2 CD2 (2π)26 δ26 ( µ µ µ µ ×y12 2α k1 ·k2 +1 y13 2α k1 ·k3+1 y23 2α k2 ·k3 +1 y1 ′ ′ ′ × y3 dy4 y14 2α k1 ·k4 y24 2α k2 ·k4 y34 2α k3 ·k4 × [v µ3 (y3 ) + q µ3 (y3 )][v µ4 (y4 ) + q µ4 (y4 )] × [v µ1 (y1 ) + q µ1 (y1 )][v µ2 (y2 ) + q µ2 (y2 )] .2(a). Since we are looking only for terms in which the ei appear in the particular combination e1 ·e2 e3 ·e4 . for instance µ −1 µ −1 µ −1 v µ (y3 ) = −2iα′ (k1 y31 + k2 y32 + k4 y34 ). 4. α′ (36) Since the boundary is along the σ 1 axis. .26). Hence the vertex operator is (37) i ˙ eµ ⋆ X µ eik·X ⋆ . the derivative can be written using (2. 2) shown in ﬁgure 6. 2. 1. agrees with (3. illustrated in ﬁgure 6.1. ′ ′ ′ The v µ that appear in the path integral in the last two lines are linear in the momenta.9 (a) There are six cyclic orderings of the four vertex operators on the boundary of the disk. after multiplying by the factor −go and integrating over the position on the boundary. (41) They therefore contribute terms in which the polarization vectors ei are dotted with the momenta. Consider ﬁrst the ordering (3. 0 ∼ i = −1 2 ⋆ eµ ⋆ ∂X µ eik·X ⋆ = i ⋆ α′ 2 ⋆ ¯ µ ik·X ⋆ eµ ⋆ ∂X e ⋆.3): ¯ ˙ X = ∂1 X = (∂ + ∂)X = 2∂X.2. with − ∞ < y3 < y1 < y2 < ∞. If we ﬁx the positions of the vertex operators for gauge bosons 1. and 3.5 Problem 6.6. (39) then we must integrate the position of the fourth gauge boson vertex operator from y3 to y1 . (38) ⋆ ⋆ ′ 2α which.6 CHAPTER 6 45 the photon vertex operator was eµ αµ 0. √ 6.
y2 . so long as we obey (39). y2 appears with a total power of 2α′ k1 · k2 − 1 + 2α′ k2 · k3 + 1 + 2α′ k2 · k4 = 2α′ (k1 + k3 + k4 ) · k2 2 = −2α′ k2 = 0. will change to reﬂect the new order. we ﬁnd that the part of the . the beta function will have diﬀerent arguments in each case. It can easily be seen that the trick that allowed us to take y2 to inﬁnity (equation (44)) works equally well for y1 or y3 . However. The limits of integration will be whatever positions immediately precede and succede y4 .6 CHAPTER 6 46 we can neglect the v µ . ′ ′ (43) We can choose y1 . Speciﬁcally. and the limits of integration on y4 . Since s = −2k3 · k4 and u = −2k1 · k4 . Making these substitutions above. and the resulting integral over y4 will once again give a beta function. Then y12  ∼ y23  ∼ y2 and (since y3 < y4 < y1 ) y24  ∼ y2 as well. and y3 as we like. Putting together the results from the six cyclic orderings. The terms we are looking for arise from the contraction of the q µ with each other. while the position that is opposite y4 will be taken to inﬁnity. the singular part of the OPE of q µ (y) with q ν (y ′ ) is − 2α′ (y − y ′ )−2 η µν . and y3 while integrating over y4 . with two exceptions: the order of the λa matrices appearing in the trace. (44) We can simplify further by setting y3 = 0 and y1 = 1. and the above expression simpliﬁes if we take the limit y2 → ∞ while keeping y1 and y3 ﬁxed. The lower and upper limits of integration can be ﬁxed at 0 and 1 respectively as before. we can still ﬁx y1 . the integral above reduces to: 1 0 −α dy4 (1 − y4 )−α u y4 s−2 = B(−α′ u + 1. since diﬀerent factors in the integrand of (43) survive for diﬀerent orderings. −α′ s − 1). (42) so the combination e1 · e2 e3 · e4 arises from the contractions of q µ1 (y1 ) with q µ2 (y2 ) and q µ3 (y3 ) with q µ4 (y4 ): 2 igo α′−1 Tr(λa3 λa4 λa1 λa2 )e1 · e2 e3 · e4 (2π)26 δ26 ( ×y12 2α k1 ·k2 −1 y13 2α k1 ·k3 +1 y23  y1 ′ ′ ki ) i 2α′ k2 ·k3 +1 ′ × y3 dy4 y14 2α k1 ·k4 y24 2α k2 ·k4 y34 2α k3 ·k4 −2 . Equation (43) will remain the same. ′ ′ (45) If we now consider a diﬀerent cyclic ordering of the vertex operators. y2 .
6 CHAPTER 6 four gauge boson amplitude proportional to e1 · e2 e3 · e4 is 2 igo e1 · e2 e3 · e4 (2π)26 δ26 ( α′ 47 ki ) i (46) × Tr(λa1 λa2 λa4 λa3 + λa1 λa3 λa4 λa2 )B(−α′ t + 1.5.14). We can therefore rewrite (47) in the following form: − 2 igo e1 · e2 e3 · e4 (2π)26 δ26 ( α′ (49) ki )Tr λa1 λa3 λa2 λa4 + λa1 λa4 λa2 λa3 i (50) 1 − (λa1 λa2 λa4 λa3 + λa1 λa3 λa4 λa2 + λa1 λa2 λa3 λa4 + λa1 λa4 λa3 λa2 ) . −α′ s − 1) +Tr(λa1 λa3 λa2 λa4 + λa1 λa4 λa2 λa3 )B(−α′ t + 1. λb ]λc . It is ′2 −igo e1 · e2 e3 · e4 (2π)26 δ26 ( ki ) i u−t s ki ) f a1 a2 e f a3 a4 e e = ig2 − o e1 α′ · e2 e3 · e4 (2π)26 δ26 ( ×Tr (λa1 λa2 λa3 λa4 + λa1 λ λ λ Combining (50) and (51) and using i a4 a3 a2 t−u 2s − λa1 λa2 λa4 λa3 − λa1 λa3 λa4 λa2 ) . (52) . the uchannel. −α′ u + 1) +Tr(λa1 λa2 λa3 λa4 + λa1 λa4 λa3 λa2 )B(−α′ u + 1. and the fourpoint vertex. (51) s + t + u = 0. (b) There are four treelevel diagrams that contribute to fourboson scattering in YangMills theory: the schannel. and the f abc are the gauge group structure constants: f abc = Tr [λa . −α′ s − 1) . 2 Of the three diagrams that contain threepoint vertices. equation A. the tchannel. only the schannel diagram contains a term proportional to e1 · e2 e3 · e4 . The YangMills coupling is ′ go = (2α′ )−1/2 go (48) (6.12). The fourpoint vertex diagram (which is independent of momenta) includes the following term proportional to e1 · e2 e3 · e4 : ′2 − igo e1 · e2 e3 · e4 (2π)26 δ26 ( ki ) i e (f a1 a3 e f a2 a4 e + f a1 a4 e f a2 a3 e ) (47) (see Peskin and Schroeder.
−α′ u + 1) ≈ 1.6 Problem 6. It is clear that (46) reduces to (53) (up to an overall sign) if we take the limit α′ → 0 with s. z1 ) :: eik2 ·X (z2 .2. × Tr(λa1 λa2 λa4 λa3 + λa1 λa3 λa4 λa2 ) −1 − +Tr(λa1 λa3 λa2 λa4 + λa1 λa4 λa2 λa3 ) +Tr(λa1 λa2 λa3 λa4 + λa1 λa4 λa3 λa2 ) −1 − This is intentionally written in a form suggestively similar to equation (46). −α′ s − 1) ≈ −1 − . at momenta small compared to the string scale. since in that limit t B(−α′ t + 1. s (54) Thus this single string theory diagram reproduces. −α′ s − 1) ≈ −1 − .3.19): ¯ : ∂X µ ∂X ν eik1 ·X (z1 . u B(−α′ u + 1. 6. z2 ) :: eik3 ·X (z3 . the transversality of the polarization tensor.6 CHAPTER 6 we obtain. the sum of the ﬁeld theory Feynman diagrams. g = CS2 z12 2 z13 2 z23 2 . k2 = k3 = 4/α′ imply k1 · k2 = k1 · k3 = 0.11 (a) The X path integral is given by (6. k2 · k3 = −4/α′ .4): : c(z1 )˜(¯1 ) :: c(z2 )˜(¯2 ) :: c(z3 )˜(¯3 ) : cz cz cz S2 ′ ′ ′ . s B(−α′ t + 1. − 2 igo e1 · e2 e3 · e4 (2π)26 δ26 ( α′ 48 ki ) i (53) t s u s . and u ﬁxed. (56) 2 2 2 The momentumconserving delta function and the mass shell conditions k1 = 0. z3 ) : ¯ ¯ ¯ ′2 X α (2π)26 δ26 (k1 + k2 + k3 ) = −iCS2 4 S2 (55) µ kµ k2 + 3 z12 z13 ν k2 kν + 3 z12 z13 ¯ ¯ ×z12 α k1 ·k2 z13 α k1 ·k3 z23 α k2 ·k3 The ghost path integral is given by (6. µ e1µν k1 = 0. for the part of the fourboson amplitude proportional to e1 · e2 e3 · e4 . (58) . at tree level. t. (57) Using (57).
e ′ ′ (61) Here e is summed over an orthonormal basis of symmetric polarization tensors obeying the condition µ µ eµν (k1 + k2 ) = 0. We will restrict ourselves to the schannel diagram. 2 k12 (62) 2 2 which obeys the transversality condition by virtue of the fact that k1 = k2 . 2 c (60) (b) Let us calculate the amplitude for massless closed string exchange between closed string tachyons (this is a treelevel ﬁeld theory calculation but for the vertices we will use the amplitude calculated above in string theory). k2 . 2 α′ gc (59) we can put together the full amplitude for two closedstring tachyons and one massless closed string on the sphere: SS2 (k1 . k3 ) 2 ′ = gc gc e−2λ e1µν ¯ × : cc∂X µ ∂X ν eik1 ·X (z1 . 4 s (63) (64) . none of the other elements of the basis would contribute to the sum. X g CS2 ≡ e−2λ CS2 CS2 = 8π . With this choice.6 CHAPTER 6 49 and the result (6. Here the propagator for the massless intermediate string provides the pole at s = 0: − i(2π)26 δ26 ( ki ) ′2 π 2 α′2 gc 4s µ ν µ ν eµν k12 k12 eµ′ ν ′ k34 k34 . which reduces to (k12 · k34 )2 = (u − t)2 . We could choose as one element of that basis the tensor eµν = k12µ k12ν . The amplitude (61) is thus − i(2π)26 δ26 ( ki ) ′2 π 2 α′2 gc (u − t)2 . z3 ) : ˜ ¯ ˜ ¯ ˜ ¯ ′2 2 ′α = −iCS2 gc gc (2π)26 δ26 (k1 + k2 + k3 ) 4 z12 2 z13 2 ×e1µν z23 2 S2 µ k2 kµ + 3 z12 z13 ν k2 kν + 3 z12 z13 ¯ ¯ ν k23 kν − 23 2¯12 2¯13 z z ′ = −i2πα′ gc (2π)26 δ26 (k1 + k2 + k3 ) µ kµ z12 2 z13 2 k23 − 23 ×e1µν z23 2 2z12 2z13 πα′ ′ µ ν = −i g (2π)26 δ26 (k1 + k2 + k3 )e1µν k23 k23 . e1 . z3 ) :: cceik2 ·X (z2 .8). because we are interested in comparing the result with the pole at s = 0 in the VirasoroShapiro amplitude.6. z1 ) :: cceik2 ·X (z2 .
˜ Gµν = ηµν + 2κeµν f. α′ s (67) To lowest order in s the other gamma functions simplify to Γ(b)Γ(c) Γ(b)Γ(c) ≈ Γ(a + b)Γ(a + c)Γ(b + c) Γ(b − 1)Γ(c − 1)Γ(2) = (b − 1)(c − 1) α′2 (u − t)2 . =− 64 Thus the part of the amplitude we are interested in is − i(2π)26 δ26 ( Comparison with (64) shows that gc = 2 ki )π 2 gc (68) (u − t)2 . s (69) ′ α′ gc . 2 (70) (c) In Einstein frame the tachyon kinetic term decouples from the dilaton. 4 c = −1 − α′ u . To lowest order in f and T .6. the VirasoroShapiro amplitude is i(2π)26 δ26 ( where a = −1 − α′ s . which is. then the kinetic term for f will be canonically normalized. α′ (74) . (71) 2 α If we write a metric perturbation in the following form. 4 (66) ki ) 2 16π 2 gc Γ(a)Γ(b)Γ(c) . makes a vanishing contribution to the amplitude onshell: − iκeµ µ 4 + k2 · k3 (2π)26 δ26 (k1 + k2 + k3 ) = 0. the interaction Lagrangian is Lint = κeµν f ∂µ T ∂ν T + κeµ f µ 2 2 1 µ T − ∂ T ∂µ T α′ 2 (73) (from now on all indices are raised and lowered with ηµν ). to lowest order in s.6 CHAPTER 6 50 Now. 4 b = −1 − α′ t . (72) where eµν eµν = 1. proportional to the trace of e. as the tachyon action (6. Γ(a) ≈ 4 .16) becomes 1 4 ˜ ˜ ˜ ST = − d26x (−G)1/2 Gµν ∂µ T ∂ν T − ′ eΦ/6 T 2 . ′ α Γ(a + b)Γ(a + c)Γ(b + c) (65) The pole at s = 0 arises from the factor of Γ(a). The second term.
7 Problem 6. Comparison with the amplitude (6.14) shows that ′ κ = πα′ gc . k3 ) 2 = gc go e−λ c ¯ ⋆ dy2 : c˜eik1 ·X (z. which can always be gauged away. k2 . z ) : ⋆ c1 eik2 ·X (y1 )⋆ ⋆ eik3 ·X (y2 )⋆ ⋆⋆ ⋆ − y1 ¯ − y1 z − z ′ D1 = g 2 gc go e−λ CD2 z ′ 2 z iCD2 (2π)26 δ26 (k1 ¯ X ′ + k2 + k3 ) ′ ×z − z α k1 /2 z − y1 2α k1 ·k2 ¯ 2 = iCD2 gc go (2π)26 δ26 (k1 + k2 + k3 ) dy2 z − y2 2α k1 ·k3 y1 − y2 2α k2 ·k3 (77) ×z − y1 −2 z − z 3 ¯ dy2 z − y2 −4 y1 − y2 2 .12 We can use the three CKVs of the upper halfplane to ﬁx the position z of the closedstring vertex operator and the position y1 of one of the upperstring vertex operators. independent of z and y1 (as it should be). α′ (78) .14). The very last line is equal to 4π. as can be calculated by contour integration in the complex plane. 2 (75) µ where we have used the transversality of the graviton polarization eµν k1 = 0.6. We integrate over the position y2 of the unﬁxed openstring vertex operator: SD2 (k1 . not the trace of the (Einstein frame) graviton.6 CHAPTER 6 51 (The trace of the massless closed string polarization tensor e used in the string calculations of parts (a) and (b) above represents the dilaton. the result is 4πigc (2π)26 δ26 (k1 + k2 + k3 ).4. (76) 6. Taking into account (6.) The amplitude from the ﬁrst term of (73) is κ µ ν µ ν 2iκeµν k2 k3 (2π)26 δ26 (k1 + k2 + k3 ) = −i eµν k23 k23 (2π)26 δ26 (k1 + k2 + k3 ).
wj ) − = i<j 2 ki G′ (wi . (5) = i<j α′ ki · kj ln where in the second equality we have used the overall momentumconserving delta function. wj ) − i 2 ki G′ (wi .1 Chapter 7 Problem 7.2.2.3). (4) The argument of the exponential in (1) is thus − ki · kj G′ (wi . wi ) . wj ) = G′ (wi . wi )) : are (2. which implies i. wj ) + r designed to be ﬁnite in the limit wj → wi : lim G′ (wi .j ki · kj wij 2π ϑ1 τ ∂ν ϑ1 (0τ ) 2π − (Im wij )2 4πτ2 .1 Equation (6.7 CHAPTER 7 52 7 7. wj ) = − ki ) exp − i<j ki · kj G′ (wi .2. deﬁned by (6. The weights of the vertex operator : exp(iki · X(wi .4. 4πτ2 (2) while G′ is the renormalized Green’s function. r G′ (wi .wi ) : i=1 T2 X iCT 2 (τ )(2π)d δd ( = 1 2 G′ is the Green’s function (7.2. wi ) r i i<j α′ ki · kj ln ϑ1 1 − k(τ ) 2 wij τ 2π − (Im wij )2 4πτ2 + ∂ν ϑ1 (0τ ) α′ ln 2 2π 2 ki i i.4).15). Under the modular transformation τ → τ ′ = −1/τ the coordinate w is mapped to w′ = w/τ . G′ (wi . wj ) wj →wi r ij ϑ1 2π τ α′ = − ln lim wij →0 2 wij α′ ln wij 2 . (1) r wij α′ ln ϑ1 τ 2 2π 2 + α′ (Im wij )2 + k(τ ).17) ¯ 2 α′ ki ˜ h=h= 4 (6) . 2 w 2 (3) + k(τ ) 2 =− 1 2 ∂ν ϑ1 (ν = 0τ ) α′ ln 2 2π + k(τ ).13) applied to the case of the torus tells us n ¯ : eiki ·X(wi .j ki · kj = 0. Plugging this into (1) yields (7.
τ 2 (9) η(τ ′ ) = (−iτ )1/2 η(τ ).13). According to (7. the vacuum amplitude X CT 2 = (4πα′ τ2 )−d/2 η(τ )−2d (8) is invariant. then. ′ 4πτ2 4πτ2 τ 2 (15) The only change.7 CHAPTER 7 53 so. according to (2. 4πτ 2 (14) and it cancels the change in the last term on the RHS of (5): ′ (Im wij )2 1 2 2 = −2τ1 τ2 Im wij Re wij + τ2 (Re wij )2 + τ1 (Im wij )2 . so ln 2π ϑ1 ∂ν ϑ1 (0τ ′ ) ′ wij ′ τ 2π ′ wij ′ τ 2π (10) = −i(−iτ )1/2 exp 2 iwij 4πτ ϑ1 wij τ 2π (11) (12) 2 wij 4πτ = ln wij 2π ϑ1 τ τ ∂ν ϑ1 (0τ ) 2π − Im . (7) On the RHS of (1).wi ) : i=1 T2 .4b) τ2 .2. . (16) α′ ki · kj = − 2 i<j i we ﬁnally arrive.40d).2.4. is the new factor of τ −1 in the logarithm on the RHS of (13). the product of vertex operators on the LHS of (1) transforms to n n :e i=1 ′ ¯′ iki ·X(wi . since ′ τ2 = and (7. which gets taken to the power 1 2 α′ ki . ϑ1 and ∂ν ϑ1 (0τ ′ ) = (−iτ )3/2 ∂ν ϑ1 (0τ ). (13) The second term on the right is equal to Im 2 wij 4πτ = 1 2τ1 Im wij Re wij − τ2 (Re wij )2 + τ2 (Im wij )2 . at the transformation law (7).wi ) : T2 = τ  P i 2 α′ ki /2 ¯ : eiki ·X(wi . as expected.
2.2.3) has the correct Fourier coeﬃcients. that is.2. so we have (2π)2 τ2 0 1 1 dy 0 dx e2πi(na x+nby) − α′ ln ϑ1 (x + yτ τ )2 + α′ πy 2 τ2 . (22) ω na nb T2 To evaluate the LHS. The Laplacian in momentum space is − k2 = −na ka + nb kb 2 = − In terms of the normalized Fourier components Xna nb (w) = 1 1/2 2πτ2 nb − na τ 2 2 ≡ −ωna nb . w′ ) = Xn n (w)∗ Xna nb (w′ ). 2 ω na nb a b (na . 2πα′ Xn n (w′ ). we can write.3).2. as usual. the Green’s function in real space is (6. 0) = 2 . 0). we will show that (7. nb ) = (0.8) in momentum space. (19) 2 ω na nb a b (na . and use the fact that G′ depends only on the diﬀerence w − w′ to shift the variable of integration: 2πα′ d2w ei(na ka +nb kb )·w G′ (w. w′ ) = 0. (24) .38d). w′ ) = The windependent part of the Green’s function is left as the unspeciﬁed constant k(τ ) in (7. A · B ≡ Re AB ∗ ). (20) (21) T2 T2 d2w Xna nb (w)G′ (w. d2w X00 (w)G′ (w. the momentum space is a lattice. ln ϑ1 (x + yτ τ ) = K(τ ) − iπ(x + yτ ) + ∞ m=0 ln 1 − e 2πi(x+(y+m)τ ) + ∞ m=1 ln 1 − e−2πi(x+(y−m)τ ) . we ﬁrst divide both sides by Xna nb (w′ ). which is adjusted (as a function of τ ) to satisfy equation (20). generated by the complex numbers ka = 1 − iτ1 /τ2 and kb = i/τ2 .3) (or (2)) directly.2 Problem 7.0) Rather than show that (19) is equal to (7. Because the torus is compact.3 As usual it is convenient to solve Poisson’s equation (6. 2 τ2 (17) ei(na ka +nb kb )·w (18) (where the dot product means. The Jacobian for this change of coordinates is (2π)2 τ2 .7) 2πα′ G′ (w.7 CHAPTER 7 54 7. To prove (21). let us use coordinates x and y on the torus deﬁned by w = 2π(x + yτ ). 2 (23) Using the inﬁniteproduct representation of the theta function (7.nb )=(0.2.2.
0. For example. we ﬁrst convert the inﬁnite sums into inﬁnite regions of integration in y (using the periodicity of the ﬁrst factor under y → y + 1): 1 0 The integration of the ﬁrst term in the brackets is straightforward: − τ22 . which drops out of the integral. (28) The y integral is now straightforward: ∞ 0 0. Expression (23) thus becomes − 2π 2 α′ τ2 1 1 dy 0 0 dx e2πi(na x+nby) 2πτ2 y(1 − y) + ∞ m=0 ln 1 − e 2πi(x+(y+m)τ ) 2 + ∞ m=1 ln 1 − e−2πi(x+(y−m)τ ) 2 . 0 0 na = 0 (26) dy e2πi(na x+nby) × ∞ m=0 ln 1 − e 2 2πi(x+(y+m)τ ) 2 + ∞ m=1 ln 1 − e−2πi(x+(y−m)τ ) 2 = 0 ∞ dy e2πi(na x+nb y) ln 1 − e2πi(x+yτ ) 0 + −∞ dy e2πi(na x+nb y) ln 1 − e−2πi(x+yτ ) . na 2πina (nb − na τ ) (29) . na 1 n 1 0 dx e2πi((na +n)x+(nb +nτ )y) na < 0 na ≥ 0 . dy 1 1 2πi(nb −na τ )y e =− . (25) To integrate the logarithms. 1 0 dx e2πi(na x+nb y) ln 1 − e2πi(x+yτ ) = − = ∞ n=1 1 e2πi(nb −na τ )y . (27) 2 The x integral can now be performed by separating the logarithms into their holomorphic and antiholomorphic pieces and Taylor expanding.7 CHAPTER 7 55 where we’ve collected the terms that are independent of x and y into the function K(τ ). na = 0 1 1 πnb 2πi(na x+nb y) dy e 2πτ2 y(1 − y) = dx .
ϑ01 (ν. 0.38) that in this limit ϑ00 (ν. na ≤ 0 1 ¯ . na > 0 . 0. Since diﬀerent terms in the series cannot cancel for arbitrary τ . (38) When we take τ to 0 along the imaginary axis. we have ϑ00 (ν. na < 0 = 2πina (nb −na τ ) . na ≥ 0 1 2πina (nb −na τ ) . = 0. ϑ10 (ν. which is what we were trying to prove.3 Problem 7. ¯ na > 0 (30) (31) (32) multiplying by the prefactor −2π 2 α′ τ2 in front of the integral in (25) indeed yields precisely the RHS of (22). the theta function can go to 0 only if every term in the series does so: ∀n ∈ Z.7 CHAPTER 7 56 Similarly. Re(ν − n)2 > 0. ∞ 0 0 1 dy 0 1 τ dx e2πi(na x+nb y) ln 1 − e−2πi(x+y¯) = dy −∞ 0 0 dx e2πi(na x+nb y) ln 1 − e−2πi(x+yτ ) 1 τ dx e2πi(na x+nb y) ln 1 − e2πi(x+y¯) dy −∞ 0 Expressions (26) and (29)–(32) can be added up to give a single expression valid for any sign of na : − τ2 . na ≤ 0 1 − 2πina (nb −na τ ) . −1/τ ) 2 /τ e−πi(ν−n) . and it is clear from either the inﬁnite sum expressions (7. ϑ11 (ν.40a).5 In each case we hold ν ﬁxed while taking τ to its appropriate limit. πnb − na τ 2 (33) . τ ) = (−iτ )−1/2 e−πiν = (−iτ )−1/2 ∞ n=−∞ 2 /τ (34) (35) (36) (37) ϑ00 (ν/τ.2. 7.2. τ ) → 0. τ ) → 1. (b) Inverting the modular transformation (7. each term in the series will go either to 0 (if Re(ν − n)2 > 0) or to inﬁnity (if Re(ν − n)2 ≤ 0). Note that all of these limits are independent of ν. τ ) → 0. q ≡ exp(2πiτ ) → 0.2. (a) When Im τ → ∞. (39) . τ ) → 1.37) or the inﬁnite product expressions (7.
7 CHAPTER 7 otherwise it will diverge. τ0 τ0 (47) (48) . 1 τ′ = − . −1/τ ) 2 /τ (−1)n e−πi(ν−n+1/2) . τ (46) We are considering limits where τ approaches some nonzero real value τ0 along a path parallel to the imaginary axis. (c) According to (7. we set τ = τ0 + iǫ and take ǫ → 0+ .40). in other words. For  Re ν ≤ 1/2. the theta function goes to 0 if  Im ν <  Re ν − n. The property of approaching the real axis along a path parallel to the imaginary axis is preserved by the modular transformations (to ﬁrst order in ǫ): τ ′ = τ0 + 1 + iǫ. and inﬁnity elsewhere. 57 (40) (41) Since ϑ01 (ν. (43) n=−∞ the theta functions are exchanged with each other and multiplied by factors that are ﬁnite as long as ν and τ are ﬁnite.39) and (7. for  Re ν − n ≤ 1/2. τ ) = ϑ00 (ν + 1/2. the region in which it goes to 0 in the limit τ → 0 is simply shifted by 1/2 compared to the case treated above. τ ) = (−iτ )−1/2 e−πiν = (−iτ )−1/2 ∞ 2 /τ ϑ01 (ν/τ. (42) n=−∞ the sum will again go to 0 where (39) is obeyed. For ϑ10 . τ ).2. in general. τ (44) (45) ∞ 2 /τ ϑ11 (ν/τ. condition (39) is equivalent to  Im ν <  Re ν.2. under (45) ν is transformed to ν′ = ν . the story is the same as for ϑ00 . since the sum is the same as (42) except over the halfoddintegers. Also. τ ) = i(−iτ )−1/2 e−πiν = (−iτ )−1/2 so the region (39) is shifted by 1/2. ϑ11 (ν. under the modular transformations τ ′ = τ + 1. Finally. ϑ11 goes to 0 in the same region as ϑ01 . ǫ 1 τ ′ = − + i 2 + O(ǫ2 ). −1/τ ) 2 /τ (−1)n e−πi(ν−n) . since ϑ10 (ν.
By a sequence of transformations (44) and (45) one can reach any rational limit point τ0 starting with τ0 = 0. since the boundary of C2 is given by those points that are invariant under the involution w → −w. being either 0 or inﬁnity. so we have n ⋆ ⋆ ¯ eiki ·X(wi . If there are ChanPaton factors. wn ). the case considered in part (b) above. in which ϑ00 goes to 0. w′ ) + G′ 2 (w.40). The ¯ renormalized selfcontractions are also doubled.7 The expectation value for ﬁxed open string tachyon vertex operators on the boundary of the cylinder is very similar to the corresponding formula (7. where p/q is the ﬁnal value of τ0 in reduced form. and of course it will also depend implicitly on the ChanPaton factors themselves λai . one for each component of the boundary. and the vertex operator positions wi must be integrated over both components. I can only conjecture that the theta functions diverge (almost) everywhere on the ν plane in that limit. the two terms on the RHS above are equal if either w or w′ is on the boundary. Borrowing from the cylinder vacuum amplitude given . will repeatedly be shifted by 1/2 and rescaled by τ0 (under (46)). 7.7 CHAPTER 7 58 under (44) and (45) respectively. (Note that the limiting value. w′ ) = G′ 2 (w. C T T (49) However. We will denote the product of these two traces T (w1 .2. and the order of the factors in each trace will be the order of the operators on the corresponding component. then the integrand will include two traces.4 Problem 7.wi ) ⋆ ⋆ C2 = −3 i=1 X iCC2 (t)(2π)26 δ26 ( i ki ) i<j η(it) ϑ1 wij (Im wij )2 it exp − 2π 4πt 2α′ ki ·kj .2.2. .2. . −w′ ). As for the case when τ0 is irrational. (50) The boundary of the cylinder breaks into two connected components.39) and (7. is insensitive to the ﬁnite prefactors involved in the transformations (7. During these transformations (which always begin with (44)). .4) for closed string tachyon vertex operators on the torus. The major diﬀerence comes from the fact that the Green’s function is doubled. the region (39).) It is easy to see that this cumulative sequence of rescalings will telescope into a single rescaling by a factor q. .3): ¯ G′ 2 (w. The method of images gives the Green’s function for the cylinder in terms of that for the torus (7.
xn ) i 0 × ∞ 0 dt tn−14 η(it)3n−24 i<j ϑ1.2. For concreteness let us put the ﬁrst set on the boundary at Re w = 0 and the second set on the boundary at Re w = π. (53) In the last product the type of theta function to use depends on whether the vertex operators i and j are on the same boundary (in which case ϑ1 ) or on opposite boundaries (in which case ϑ2 (ixij tit) = −ϑ1 (ixij t − 1/2it)). . a1 . it is convenient to scale the vertex operator positions with t. and the other n − m ≥ 2 are on the other one. and then double the amplitude (51). kn .4 (xij iu)2α ki ·kj . . . kn . . (51) 7. . . . (7. Concentrating now on the last line of (53). i = m + 1.8 In this problem we consider the part of the amplitude (51) in which the ﬁrst m ≥ 2 vertex operators are on one of the cylinder’s boundaries. . . and (7. . . a1 . xm )T2 (xm+1 .5 Problem 7.37). 2 Using (16) and the mass shell condition α′ ki = 1 we can write the part of the amplitude we’re interested in as follows: S ′ (k1 . . wn ) n dwi i=1 −3 ∂C2 η(it) i<j ϑ1 wij (Im wij )2 it exp − 2π 4πt 2α′ ki ·kj . an ) = i(2π)26 δ26 ( i 1 n ki )go 2−13 (2π)n−26 α′−13 × dxi T1 (x1 . . . . ′ (54) For large u. .2 (ixij tit) exp(−πx2 t) ij 2α′ ki ·kj .40b).4. allowing us to change of the order of integration in (51).40d). we can write the ntachyon amplitude as SC2 (k1 . . . . Since we will be focusing on the region of the moduli space where t is very small.1). m i wi = .7 CHAPTER 7 59 in (7. (52) π + 2πix t. each of the factors in the integrand of (54) can be written as a fractional power of q ≡ e−2πu times a power series in q (with coeﬃcients that may depend on the xij . an ) = i(2π) δ ( i 26 26 n ki )go ∞ 0 dt (8π 2 α′ t)−13 η(it)−24 2t × T (w1 . and change the variable of integration to u = 1/t: ∞ du η(iu)3n−24 i<j 0 ϑ1.44b). let us apply the modular transformations (7. .2. . . . see (7. . so 2πix t. i = 1. .2. . . n i The xi run from 0 to 1 independent of t.2. . .
and (7. (60) π(4k − 4 − α′ s) 0 Since every (nonnegative) integer power k appears in the expansion of (54). ∞ (55) (56) (57) ϑ1 (xiu) = 2q 1/8 sin(πx) (1 − (1 + 2 cos(2πx))q + .43)): η(iu) = q 1/24 (1 − q + .2. We obtain the second line from the ﬁrst by using m i<j i≃j n ki · kj = i<j ki · kj − n i=1 2 ki − i=1 j=m+1 m ki · kj n j=m+1 1 =− 2 =− ki i=1 n − s. . 2α′ · kj (59) where s = −( n ki )2 is the mass squared of the intermediate state propagating along the long i=1 2 cylinder.) Each power of q appearing in the power series (1 + . . ) ∞ 0 i<j i≃j = i<j i≃j 2 sin πxij 2α ki ·kj du q −1−α s/4 (1 + . . upon performing the u integration.38). For η and ϑ1 this power series involves only integer powers of q. To see this. .2. ). . . . Substituting (55)(57) into (54) yields 2 sin πxij  2α′ ki ·kj 0 ′ ∞ n/8−1+α′ P du q i<j i≃j ki ·kj /4 (1 + . m. so in fact we could have done without the mass shell condition in the derivation. i = 1. we have the entire sequence of closed string masses at s = 4(k − 1)/α′ . . ) in (58) will produce. ϑ4 (xiu) = (−1)k q k 2 /2 k=−∞ e2πikx = 1 − 2 cos(2πx)q 1/2 + . . . (The two n’s we have cancelled against each other in (58) both come from α′ i ki . ) . consider the eﬀect of uniformly translating all the vertex operator positions on just one of the boundaries by an amount y: xi → xi + y. but only after integrating over the vertex operator positions.7 CHAPTER 7 60 (7. . (57)? The poles from these terms in fact vanish. This translation changes the relative . a pole in s: ∞ 2 ′ du q −1−α s/4+k = . ) (58) ′ The symbol i ≃ j means that the sum or product is only over pairs of vertex operators on the same boundary of the cylinder. . . . . What about the halfinteger powers of q that appear in the expansion of ϑ4 . whereas for ϑ4 it mixes integer and halfinteger powers. . .
We are interested in particular in the ﬁrst pole. (62) where the kl are integers and the coeﬃcient c depends on the xi and the momenta ki . . not the detailed dependence on s away from the pole): 2 go = 211/2 (2π)25/2 α′6 . each term will be of the form cq P l 2 kl /2 2πiy e P l kl . since the vertex (65) is valid only onshell. . Since y is eﬀectively integrated over when one integrates over the vertex operator positions. at s = −4/α′ . Neglecting the “. s + 4/α′ (64) In Problem 6. (63) ′ s + 4) π(α After integrating over the positions xi . so only integer powers of q produce poles in the amplitude. when n = 4. (67) gc This is in agreement with (8. m = 2. i/(s + 4/α′ ): 1 2 − i(2π)26 δ26 ( ki )gc 4(2π)2 α′−2 . we calculated the threepoint vertex for two openstring tachyons to go to a closedstring tachyon: 4πigc (2π)26 δ26 (k1 + k2 + k3 ). where the intermediate closed string goes on shell as a tachyon. 7. only the terms for which l kl = 0 will survive. k4 ) = −i(2π)26 δ26 ( 4 ki )go 2−9 (2π)−23 α′−14 i 1 .9 We wish to consider the result of the previous problem in the simplest case. . it is only appropriate to compare the residues of the poles at s = −4/α′ in (64) and (66). ′ (61) Expanding this out using (57). . and there are no ChanPaton indices. . ” and approximating the exponents 2α′ k1 · k2 = 2α′ k3 · k4 = −α′ s − 2 by 2.28).7. . the amplitude (53) is S ′ (k1 .7 CHAPTER 7 61 position xij only if the vertex operators i and j are on opposite boundaries.12. . This condition implies that l kl2 must be even. (58) becomes 32 − sin2 (πx12 ) sin2 (πx34 ) . as well as all the factors in the integrand of (54) except those involving ϑ4 . those become m n i=1 j=m+1 ϑ4 (xij + yiu)2α ki ·kj . (66) s + 4/α′ i We can now compare the residues of the poles in (64) and (66) to obtain the following relation between gc and go (note that. it thus leaves the ChanPaton factors T1 and T2 in (53) invariant. (65) α′ We can reproduce the schannel pole of (64) by simply writing down the Feynman diagram using this vertex and the closedstring tachyon propagator.6 Problem 7.
(70) Decomposing X(w1 .e. i. However. then the cylinder becomes a disk with an open string propagator connecting two points on the boundary. w2 ) (with dw1 y = 0). (69) The Polyakov action in unit gauge (2.10 When the gauge group is a product of U (ni ) factors. which (after choosing the analog of unit gauge for the einbein e) is 1 2 l 0 dτ (∂τ x)2 + m2 . (The LHS vanishes if all the λa are not in the same block. For this to be consistent there can only be one block. we only need the two ChanPaton factors in each pair to be in the same block.9) is an integral over the circle modulus l.4. The modulus l is deﬁned with respect to the pointparticle action (5. the generators are block diagonal. because the λe must be in the same block both with the ﬁrst group of λa s and with the second group for a nonzero result. λb . on a cylinder with modulus t. Since (7. If on the intermediate state we sum only over blockdiagonal generators. (71) . This amplitude is proportional to Tr(λa λb ) Tr(λc λd ). and vanishes unless all the ChanPaton factors are in the same block.1) is an integral over the cylinder modulus t. is 1 4πα′ π 2πt dw1 0 0 dw2 (∂1 X)2 + (∂2 X)2 . and equals the usual U (n) value if the are.) Now consider the cylinder with two vertex operators on each boundary.1). w2 ) into its centerofmass motion x(w2 ) and its internal state y(w1 .11 We will be using the expression (7.1.7 Problem 7. then this amplitude will vanish unless λa . λd are all in the same block.9) for the pointparticle vacuum amplitude to obtain a generalized version of the cylinder vacuum amplitude (7. i.7 CHAPTER 7 62 7. 7.3. whereas (7. The RHS has the same property. with ChanPaton factors λa and λb on one boundary and λc and λd on the other. the Polyakov action becomes 1 4α′ 2πt 0 dw2 (∂2 x)2 + 1 π π 0 dw1 (∂1 y)2 + (∂2 y)2 .8 Problem 7. if we make a unitary cut in the open string channel. the gauge group must be simple. A treelevel diagram is proportional to Tr(λa1 · · · λal ). we need to know the relationship between these quantities. Unitarity requires Tr(λa1 · · · λai λai+1 · · · λal ) = Tr(λa1 · · · λai λe ) Tr(λe λai+1 · · · λal ).e. (68) which is an identity for any product of U (ni )s.1).1.3.4.1). λc .
we are interested in studying the propagation of closed string modes along long.1): ∞ dt ′ 2 ZC2 = iVD (73) e−2πtα mi . thin cylinders. which corresponds to the region t ≪ 1.8 above. ′ (75) and add a factor of e−πuα k to the integrand. the sum is evaluated in the usual way (with q = e−2πt ): ′ D i∈H⊥ o q α mi = ′ 2 ∞ ∞ i=1 n=1 Nin =0 D′ ∞ = q −1 D′ i=1 n=1 q −1+ ∞ P∞ n=1 nNin Nin =0 = q −1 i=1 1 1 − qn n=1 η(it)−D . and m2 = 1 4πα′2 π 0 63 dw1 (∂1 y)2 + (∂2 y)2 . . (8π 2 α′ t)−D/2 2t 0 ⊥ i∈Ho Taking a spectrum with D′ net sets of oscillators and a ground state at m2 = −1/α′ .43). (77) where m is a nonnegative integer. (72) Using the relation l = 4πα′ t to translate between the circle and the cylinder moduli.7 CHAPTER 7 We can equate (69) and (71) by making the identiﬁcations τ = 2α′ w2 . ZC2 = iVD 2(8π 2 α′ )D/2 iVD = 2(8π 2 α′ )D/2 ′ 2 /2 ∞ 0 ∞ 0 dt t−D/2−1 e−2πt(D /24−1) η(it)−D ′ ′ ′ ′ du u(D−D −2)/2 e−2π(D /24−1)/u η(iu)−D . Each term will be of the form ∞ 0 du u(D−D −2)/2 e−2π(D /24−1)/u−2πu(α k ′ ′ ′ 2 /4+m−D ′ 24) . l = 4πα′ t.2. ′ (76) Now we can expand the etafunction for large u using the product representation (7.4. where k is the momentum ﬂowing along the cylinder: ′ ′ ′ 2 /2 ∞ 0 du u(D−D −2)/2 e−2π(D /24−1)/u−πuα k η(iu)−D . So let us change the variable of integration to u = 1/t.9) over the open string spectrum. obtaining the second line of (7.3. ′ ∞ q nNin =q D ′ /24−1 (74) As in Problem 7. we can now sum the circle vacuum amplitude (7.
9 Problem 7. the partition function is ZX (t) = q −13/6 V26 d26 k α′ k2 /2 (1 − q m )−1 (1 + q m )−1 q (2π)26 m=1 (81) ∞ = iV26 (4π 2 α′ t)−13 η(2it)−26 . (77) simpliﬁes to ∞ 0 du u (D−26)/2 −2πu(α′ k 2 /4+m−1) e = Γ D−24 2 (2π(α′ k2 /4 + m − 1))(D−24)/2 . For even D ≥ 26 there is indeed a pole. ∞ 0 √ du uc−1 e−a/u−b/u = 2b−c/2 ac/2 Kc 2 ab . D ′ = 24. . (78) which has a branch cut along the negative real axis. Summing multiplies that eigenvalue by q −m over all combinations of such operators. When D ′ = 24. but this pole is simple (as one expects for a particle propagator) only for D = 26. The ground states 0. starting with ﬁnding the scalar partition function: ZX (t) = 1 X.8. it is convenient to work with states and operators that have deﬁnite properties under orientation reversal. while αm normalized to have the usual commutation relations. (79) For D odd there is a branch cut. Since Ω switches leftmovers and rightmovers. k of the closed string are invariant under ′ 2 ˜ orientation reversal. the gamma function is inﬁnite. For D even but less than 26.13 We follow the same steps in calculating the vacuum Klein bottle amplitude as Polchinski does in calculating the vacuum torus amplitude. hence the constraint D ′ = 24.K2 = Tr(Ωe−2πtH ) = q −13/6 Tr(Ωq L0 +L0 ). each raising operator α+ −m m . so that α+ commutes ˜ m m − anticommutes with it (we have suppressed the spacetime index µ). while each raising operator α− multiplies it by −q m . We √ therefore deﬁne the raising and lowering operators α± = (αm ± αm )/ 2. and can be used to build up the spectrum of the closed string in the usual way. If D ′ = 24.7 CHAPTER 7 64 In the case D = 26. even if one employs a “minimal subtraction” scheme to remove the inﬁnity. The Ωq L0 +L0 eigenvalue of a ground state is q α k /2 . These are with Ω. 7. the remainder has a logarithmic branch cut in k2 . ˜ (80) The operator Ω implements the orientationreversing boundary condition in the Euclidean time (σ 2 ) direction of the Klein bottle. the integral yields a modiﬁed Bessel function. (77) reveals the expected series of closedstring poles. as found in Problem 7.
c± = (cm ± b m m √ cm )/ 2. We thus have −m −m c+ b+ 0 0 K2 = 2q 1/6 ∞ m=1 (1 − q m )2 (1 + q m )2 = 2η(2it)2 . wi ) ¯ 2 . which is diagonal in this basis. m m ˜0 ± + The operator (−1)F Ωq L0 +L is diagonal in this basis.3. ∂t g ) ˆ 4π n i=2 d2 wi Vi (wi . σ1 ) 1 (b.9). This state has eigenvalue −q −2 under 0 ˜ (−1)F Ωc+ b+ q L0 +L0 . Therefore the trace (81) vanishes. and with b− or c− by q m .7 CHAPTER 7 65 For bc path integrals on the Klein bottle. ∂t g ) = ˆ 4π 4π =− 2 dσ 1 dσ 2 b22 t dσ 1 (bww + bww ) ¯¯ 0 0 0 2/t0 (86) = 2π(b0 + ˜0 ) b √ + = 2 2πb0 . Let us temporarily include an arbitrary number of vertex operators in the path integral. m ˜ (82) ± Build up the bc spectrum by starting with a ground state  ↓↓ that is annihilated by bm for m ≥ 0 ± and c± for m > 0. and ﬁx the σ 2 coordinate of the ﬁrst one V1 . which translates in the σ 2 direction. it projects onto 0 0 the subspace of states that are built up from c+  ↓↓ . Acting with c− does not change this eigenvalue.) The Klein bottle has only one CKV. K2 (84) The overall factor of 1/4 is from the discrete symmetries of the Klein bottle. with 1/2 from w → w ¯ and 1/2 from w → −w. (83) (We have taken the absolute value of the result. let us temporarily ﬁx the coordinate region at that for t = t0 and let the metric vary with t: 1 0 g(t) = ˆ . the amplitude is S= 0 ∞ dt 4 1 1 2 dσ1 c2 V1 (σ1 . whereas the operator cm b0 takes basis states to other basis states (if it does not annihilate them). The only exception is the case of the operator c+ b+ . To evaluate the b insertion. it is again convenient to introduce raising and lowering √ operators with deﬁnite properties under orientation reversal: b± = (bm ± ˜m )/ 2. (87) . (85) 0 t2 /t2 0 Then ∂t g(t0 ) = ˆ and 1 1 (b. According to (5. The particular path integral we will be interested in is ˜ c± b+ m 0 K2 + = q 13/6 Tr (−1)F Ωc± b0 q L0 +L0 . acting with b+ or c+ −m −m 0 0 0 (m > 0) multiplies it by −q m . and acting on it with the raising operators b± (m < 0) and cm (m ≤ 0).
and put all the vertex operators on the same footing: S= 0 ∞ dt 4t n c+ b+ 0 0 i=1 d2 wi Vi (wi . wi ) ¯ 2 . σ 2 + π). This allows us to factor the c ghost out of the integal over the ﬁrst vertex operator position in (83). σ ) = −c (0. which is c+ / 2. K2 (89) We can now extrapolate to the case where there are no vertex operators simply by setting n = 0 above. σ + π). σ 2 ) ∼ (−σ 1 . c (0.15 (a) If the σ 2 coordinate is periodically identiﬁed (with period 2π). σ 2 ) = −b12 (0.7 CHAPTER 7 66 If we expand the c insertion in the path integral in terms of the cm and cm . only the m = 0 term. σ 2 + π). The state corresponding to the crosscap is then C ∝ exp − ∞ (95) (−1)n n=1 1 α−n · α−n + b−n c−n + ˜−n c−n ˜ ˜ b n (c0 + c0 )0. 2t (90) This is oﬀ from Polchinski’s result (7. σ 2 ) = 1 2 ˜z c(z) c(¯) + z z ¯ = 1 2 cm cm ˜ + m m z z ¯ . this gives ZK2 = iV26 0 ∞ dt (4π 2 α′ t)−13 η(2it)−24 .10 Problem 7. (88) m √ then. σ 2 + π). b12 (0. σ 2 ) = −∂1 X µ (0. 1 2 1 2 ∂2 X µ (0. σ 2 ) = ∂2 X µ (0.15) by a factor of 2. will contribute to the ghost path 0 integral. σ 2 ) = b11 (0. σ 2 + π). c (0. 2 2 2 2 (92) (93) (94) These imply the following conditions on the modes at σ 1 = 0: ˜ b αµ + (−1)n αµ = cn + (−1)n c−n = bn − (−1)n˜−n = 0 ˜ −n n (for all n). 0. then a crosscap at σ 1 = 0 implies the identiﬁcation (σ 1 . ˜ (96) . ↓↓ . σ ) = c (0. Using (80) and (82). ˜ c2 (σ 1 . as we saw above.4. b11 (0. (91) = This means the following boundary conditions on the scalar and ghost ﬁelds: ∂1 X µ (0. σ 2 + π). 7. σ + π).
(104) . we ﬁnd that the total amplitude (98) is proportional to e 2s ∞ n=1 (1 − e−2sn )−24 = η(is/π)−24 . b Both terms in (99) are eigenstates of b−n cn + c−n˜n . with eigenvalues of 0 and 2 respectively.17) for the adjoints of the raising and lowering operators. ˜ (97) Since the raising and lowering operators for diﬀerent oscillators commute with each other (or. so we won’t bother with them. (101) The second ghost amplitude gives the same result. We have used the expressions (4. The zeromode amplitudes are independent of s. The ˜ b ﬁrst ghost amplitude is thus: 0 − (−1)n 0cn˜n b 0 − e−2sn (−1)n b−n c−n 0 ˜ = 1 − e−2sn . with eigenvalue 2nm. To evaluate the scalar amplitudes. ˜ ˜ (99) (100) = 0 − (−1)n 0cn˜n . so the amplitude is ∞ m=0 e−2snm = 1 .7 CHAPTER 7 67 (b) The Klein bottle vacuum amplitude is ∞ 0 ds Cc0 b0 e−s(L0 +L0 ) C . we must expand out the C exponential: e −(−1)n αµ αnµ /n −n ˜ 0 = ∞ m=0 1 (−1)(n+1)m (αµ αnµ )m 0 . we can factorize the integrand into a separate amplitude for each oscillator: ′ 2 e2s ↓↓ (b0 + ˜0 )c0 b0 (c0 + c0 ) ↓↓ 0e−sα p /2 0 b ˜ × ∞ 0e−(−1) nc n bn ˜ c e−sn(b−n cn +˜−n bn ) e−(−1) ˜ nb ˜ −n c−n n=1 0 n˜ b−n c−n × 0e−(−1) 25 nc b ˜n n ˜ e−sn(b−n cn +c−n bn ) e−(−1) ˜ 0 −(−1)n αµ α−nµ /n −n ˜ × 0e µ=0 −(−1)n αµ αnµ /n ˜n e −s(αµ αnµ +αµ αnµ ) ˜ −n ˜ −n e (no summation over µ in the last line). 1 − e−2sn (103) Finally. in the case of the ghost oscillators. commute in pairs). and each term ˜−n ˜ −n has unit norm. −n ˜ m!nm (102) Each term in the series is an eigenstate of αµ αnµ + αµ αnµ .3. The exponentials of the ghost raising operators truncate after the second term: e−(−1) nb ˜ −n c−n nc 0 (98) 0e−(−1) n bn 0 = 0 − (−1)n b−n c−n 0 .
−12 η(2is/π) −12 . 1 − (−1)n e−2sn (107) The total amplitude is then e 2s ∞ n=1 1 − (−1) e n −2sn −24 =e 2s ∞ n=1 1+e −4s(n−1/2) −24 ∞ n=1 1 − e−4sn −24 = ϑ00 (0. (106) while the scalar amplitude (103) becomes ∞ m=0 (−1)nm e−2snm = 1 . The vacuum amplitude for the M¨bius strip is o ∞ 0 ds Bc0 b0 e−s(L0 +L0 ) C .4. ˜ (105) The only diﬀerence from the above analysis is the absence of the factor (−1)n multiplying the bras. 2is/π) in agreement with the integrand of (7. (108) .7 CHAPTER 7 68 This is the sdependent part of the Klein bottle vacuum amplitude. and agrees with the integrand of (7. Thus the ghost amplitude (101) becomes instead 0 − 0cn˜n b 0 − e−2sn (−1)n b−n c−n 0 ˜ = 1 − (−1)n e−2sn .4.19).23).
2 The CBH formula tells us that. the quantity in the inner 1 1 parentheses must be between −π and π.1 (a) The spatial worldsheet coordinate σ 1 should be chosen in the range −π < σ 1 < π for (8.−n . This is cancelled by the factor from commuting the normal ordered exponentials past each other: e−(kL kL −kR kR )πiα ′ ′ ′ sign(σ1 −σ1 )/2 ′ = (−1)nw +n w . αn ] = mδm.) z1 = (b) Inspection of the above derivation shows that πiα′ 1 1 sign(σ1 − σ2 ). XR (z2 )] = − eA eB = e[A. The factors ¯ ¯ ′ ′ from the cocyles commuting past the operators eikL xL +ikR xR and eikL xL +ikR xR are given in (8.2. (2) 2 Because of where we have chosen to put the branch cut for the logarithm. αn ] m n mnz1 z2 = = α′ 2 ln z2 − ln z1 + ln 1 − ln z2 − ln z1 + ln z1 z2 − ln 1 − z2 z1 α′ 2 1− z1 α′ ln z2 − ln z1 + ln − 2 z2 ′ iα 1 1 1 1 = σ1 − σ2 + (σ2 − σ1 ± π) . (3) 1− z1 z2 z2 z1 (4) ′ ′ In passing VkL kR (z.23).n=0 [αm .1 Chapter 8 Problem 8. (1) m. xL ] − 2 2 2 α′ 1 z1 n = ln z2 − ln z1 − 2 n z2 n=0 [αm . The only nonzero commutators involved are [xL . Hence [XL (z1 ). ′ ′ (5) . The upper sign is therefore chosen if σ1 > σ2 . π/2). z ) through VkL kR (z ′ . (Note that the fourth equality is legitimate because the arguments of both 1 − z1 z2 and 1 − z2 are in the range (−π/2.2.8 CHAPTER 8 69 8 8. pL ] − i ln z1 [pL . (4) will give signs from several sources.B] eB eA . z ′ ). for two operators A and B whose commutator is a scalar.21) to work. pL ] = i. In other words. XL (z2 )] =−i α′ α′ α′ ln z2 [xL . [XR (z1 ). deﬁne σ 1 = − Im ln z (with the branch cut for the logarithm on the negative real axis). and the lower otherwise.
we can separate X 25 from the other scalars. Aa → Aa − ∂a λ. see Rocek and Verlinde (1992). we’re setting aside the relevant term in the action. it’s consistent to allow the gauge parameter λ(σ) to be periodic with the same periodicity as X 25 (making the gauge group a compact U(1)). which on a topologically trivial worldsheet means that Aa is gaugeequivalent to 0. after performing an integration by parts (and ignoring the . φ must also be periodic (with period 2π/R).2 Problem 8.3 (a) In the sigmamodel action.) (b) We can gauge the X 25 translational symmetry by introducing a worldsheet gauge ﬁeld Aa : ′ Sσ = 1 4πα′ d2σ g1/2 M gab Gµν + iǫab Bµν ∂a X µ ∂b X ν + 2 gab G25µ + iǫab B25µ (∂a X 25 + Aa )∂b X µ + gab G25. there’s not much point in making X 25 periodic on a topologically trivial worldsheet. (8) Integrating over φ forces ǫab ∂a Ab to vanish. For details. In fact. which we call X µ : Sσ = 1 4πα′ d2σ g1/2 M gab Gµν + iǫab Bµν ∂a X µ ∂b X ν +2 gab G25µ + iǫab B25µ ∂a X 25 ∂b X µ + gab G25. and on a nontrivial worldsheet the gauge ﬁeld may have nonzero holonomies around closed loops. This periodicity will be necessary later. (6) (Since we won’t calculate the shift in the dilaton. (7) This action is invariant under X 25 → X 25 + λ.25 (∂a X 25 + Aa )(∂b X 25 + Ab ) . Of course. bringing us back to the action (6). this condition leaving no additional gauge degrees of freedom. In order for these to be multiples of 2πR (and therefore removable by a gauge transformation). 1 4πα′ d2σ g1/2 M ′′ Sσ = gab Gµν + iǫab Bµν ∂a X µ ∂b X ν + 2 gab G25µ + iǫab B25µ (∂a X 25 + Aa )∂b X µ + gab G25.25 ∂a X 25 ∂b X 25 .8 CHAPTER 8 70 8. (c) We now add a Lagrange multiplier term to the action. The action.25 (∂a X 25 + Aa )(∂b X 25 + Ab ) + iφǫab (∂a Ab − ∂b Aa ) . to allow us to unwind the string. (d) Any X 25 conﬁguration is gauge equivalent to X 25 = 0.
10). and using the fact that in two dimensions gac ǫab ǫcd = gbd . µν 25. we have Sσ = 1 4πα′ d2σ g1/2 M ′ gab G′ + iǫab Bµν ∂a X µ ∂b X ν µν ′ +2 gab G′ + iǫab B25µ ∂a φ∂b X µ + gab G′ 25. which couple to the latter.2.25 . The Virasoro generators for the compactiﬁed Xs are L0 = 1 2 v + 4α′ L ∞ n=1 ∞ n=1 amn α−n · αn .2. which couple to the former. 25.25 . reﬂecting the fact that winding states. and the exchange of B25µ with G25µ . ′′′ Sσ = 1 4πα′ d2σ g1/2 M gab Gµν + iǫab Bµν ∂a X µ ∂b X ν + G25.25 = G25.25 ′ B25µ = G−1 G25µ . ′ Bµν = Bµν − G−1 G25µ B25ν + G−1 B25µ G25ν .25 25. are exchanged with compact momentum states. This can be proven by induction using (8. 8.25 G′ = G−1 B25µ . α−n · αn . 25µ 25. (11) 25µ where G′ = Gµν − G−1 G25µ G25ν + G−1 B25µ B25ν . and ignoring the result.3 Problem 8. (10) 25.25 gab Aa Ab + 2 G25µ gab ∂b X µ + iB25µ ǫab ∂b X µ + iǫab ∂b φ Aa . ˜ ˜ (14) (15) 1 2 ˜ v + L0 = 4α′ R . 25. Integrating over Aa .25 where B a = G25µ gab ∂b X µ + iB25µ ǫab ∂b X µ + iǫab ∂b φ.8 CHAPTER 8 71 holonomy issue) is ′′′ Sσ = 1 4πα′ d2σ g1/2 M gab Gµν + iǫab Bµν ∂a X µ ∂b X ν + G25. (13) where is a symmetric matrix and amn is its inverse.25 ∂a φ∂b φ .25 (12) Two features are clearly what we expect to occur upon Tduality: the inversion of G25.25 25. (9) We can complete the square on Aa .4 The generalization to k dimensions of the Poisson resummation formula (8.10) is exp (−πamn nm nn + 2πibn nn ) = (det amn )1/2 n∈Z k m∈Z k exp (−πamn (mm − bm )(mn − bn )) .25 25.25 gab (Aa + G−1 Ba )(Ab + G−1 Bb ) − G−1 gab Ba Bb .25 −1 G′ 25.25 25.
12)—but unfortunately with the wrong sign! The volume factor Vk = Rk (det Gmn )1/2 comes from the integral over the zeromode.w1 ∈Z k exp − πτ2 2 2 (v + R2 w1 ) + 2πiτ1 n · w1 .5 Problem 8. hence it is selfdual.5 (a) With l ≡ (n/r + mr/2. at this point it is more or less obvious that the lattice is even and selfdual. so that M −1 ∂µ M is traceless: Gmn Gpq ∂µ Gmp ∂ µ Gnq = 2∂µ ρ2 ∂ µ ρ2 + Tr(M −1 ∂µ M )2 . 2α′ l ◦ l ′ = n · w ′ + n′ · w (19) one can easily calculate (20) (in particular. v − wR). (18) Evenness of the lattice is obvious. 8. ρ2 2 (23) . 8. which also generate the original lattice. The dual lattice is generated by the vectors (n. (b) With 1 l≡ √ (v + wR. n/r − mr/2).8 CHAPTER 8 72 where products of vectors are taken with the metric Gmn .2. Again.w2 ∈Z k ˜ n. we have l ◦ l′ = nm′ + n′ m. The partition function is (q q )−1/24 Tr q L0 q L0 = η(τ )−2k ¯ ¯ Using (13) we now get Vk ZX (τ )k w1 . α′ τ2 (17) This includes the expected phase factor (8. 2 (22) α′ ρ2 Mmn (τ ). Bmn drops out). 0) and (0.4 Problem 8. (21) The decoupling between τ and ρ is due to the fact that the determinant of M is constant (in fact it’s 1). α′ (16) exp − πR2 n m w2 − τ w1 2 − 2πibmn w1 w2 . m) = (1. 1). R2 M (τ ) = 1 τ2 1 τ1 τ1 τ 2 .4.6 The metric (8.37) can be written Gmn = Thus Gmn ∂µ Gnp = ρ−1 ∂µ ρ2 δm p + (M −1 ∂µ M )m p .
25 /(2π α′ )25−p . u) + e1 · e4 e2 · e3 F (s.25 = 2(α′ /R2 )2 ∂µ ρ1 ∂ µ ρ1 . so we have 1 CD2 . If τ and ρ are both imaginary. t)) . −α′ y + 1). (27) where F (x.7. Except for this fact. using 2 the fact that κ.25 ∂ µ B24. u) + e1 · e3 e2 · e4 F (s.p The fourripple amplitude is S= 2 2igo.p replaced with go. (28) If we had instead obtained this amplitude by Tdualizing the answer to problem 6. multiplying it by 2∂µ B24.3.p depends on p. while ρ → −1/ρ is Tduality on both X 24 and X 25 combined with X 24 ↔ X 25 .9(a).24 R = G25. τ2 α′ ρ2 τ2 .25 . Meanwhile. and therefore go. we would have found the same √ 2 2 result with go. α′ x + α′ y − 1) + B(−α′ x + 1. (2π α′ )(25−p)/2 √ (29) .p = ′ 2 .7 (a) Since we have already done this problem for the case p = 25 in problem 6. (24) (25) Clearly switching ρ and τ is a Tduality on X 24 .39).25 . so we ﬁnd go.) The open string coupling go. 8. (In this case.30).p (2π)p+1 δp+1 ( α′ ki ) i × (e1 · e2 e3 · e4 F (t.4 go through unchanged. Due to the Dirichlet boundary conditions.9(a). the antisym¯ 2 metry of B implies that Gmn Gpq ∂µ Bmp ∂ µ Bnq essentially calculates the determinant of the inverse metric det Gmn = (R2 /α′ ρ2 )2 . and we must include contributions from all three combinations of polarizations. the ChanPaton factors are trivial. transform according to (8.6 Problem 8. α′ x + α′ y − 1) + B(−α′ y + 1. the threetachyon and Veneziano amplitudes calculated in section 6. (26) α go.8 CHAPTER 8 73 With a little algebra the second term can be shown to equal 2∂µ τ ∂ µ τ /τ2 .4. then B = 0 and the torus is rectangular with proper radii R24 = R25 = G24.2).25 R = α′ ρ2 . y) ≡ B(−α′ x + 1. we can simply adapt the result from that problem (equation (46) in the solutions to chapter 6) to general p.p = go. Adding this to (23) we arrive at (twice) (8. there is no zero mode in the path integral and therefore no momentumconserving delta function in those directions. and we can compute it either by Tduality or by comparing to the lowenergy action (8.
. 25. 2 8 (35) (34) . with vanishing B and F ﬁelds and constant dilaton. which will not aﬀect their asymptotic behavior. . and therefore dominates (unless e1 · e3 or e2 · e4 vanishes). . (33) where the ﬁelds X m . a = 0. are the ﬂuctuations in the transverse position of the brane. We use a coordinate system on the brane ξ a = X a . In that limit the coeﬃcients of e1 · e2 e3 · e4 and ′ ′ e1 · e4 e2 · e3 go like sα t−1 Γ(−α′ t + 1). ﬁxing x and y. .4. (c) Expanding F (x. the amplitude (27) has the same exponential falloﬀ (6. while the last factor gives the overall behavior in the limit s → ∞. Expanding the action (8. while the coeﬃcient of e1 · e3 e2 · e4 goes like sα t+1 Γ(−α′ t − 1). We thus have Regge behavior. m = p + 1. we can use the formula 1 1 det(I + A) = 1 + Tr A + (Tr A)2 − Tr A2 + O(A3 ). 2 2 to ﬁnd Sp = −τp 1 1 dp+1ξ 1 + η ab ∂a X m ∂b X m + (η ab η cd − 2η ac η bd )∂a X m ∂b X m ∂c X n ∂d X n .p (2π)p+1 δp+1 ( α′ i ki ) 1 − cos πα′ t + tan πα′ s sin πα′ t 2 Γ(α′ s + α′ t + 1)Γ(−α′ t − 1) Γ(α′ s + α′ t + 1)Γ(−α′ t + 1) + e1 · e3 e2 · e4 × e1 · e2 e3 · e4 Γ(α′ s + 2) Γ(α′ s) ′ s + α′ t − 1)Γ(−α′ t + 1) Γ(α + e1 · e4 e2 · e3 . . For hard scattering. .19) as the Veneziano amplitude. . since the only diﬀerences are shifts of 2 in the arguments of some of the gamma functions.p (2π)p+1 δp+1 ( i ki ) (e1 · e2 e3 · e4 tu + e1 · e3 e2 · e4 su + e1 · e4 e2 · e3 st) . . p. so the induced metric is Gab = ηab + ∂a X m ∂b X m . y) = − π 2 α′ 2 3 xy + O(α′ ). whose scattering amplitude we wish to ﬁnd. y) for small α′ . we ﬁnd (with some assistance from Mathematica) that the leading term is quadratic: F (x. (30) Γ(α′ s) The factor with the sines and cosines gives a pole wherever α′ s is an odd integer.7. let us rewrite the amplitude (27) in the following way: S= 2 2igo. 2 (31) Hence the low energy limit of the amplitude (27) is 2 S ≈ −iπ 2 α′ go. Gµν = ηµν .8 CHAPTER 8 74 (b) To examine the Regge limit. (32) The Dbrane is embedded in a ﬂat spacetime.2) to quartic order in the ﬂuctuations.
.
As in the Veneziano amplitude. s ≡ −(q1 + q2 )2 . as in problem 8. z2 ) : D2 .p (2π)p+1 δp+1 (k1 + k2 )2 dx xα k −2 (1 − x)−α s/2−3 (1 + x)α s/2−2k 0 √ π 3/2 (2π α′ )11−p gc (2π)p+1 δp+1 (k1 + k2 )B(α′ k2 − 1.p = X iCD2 .¯2 ) : D2 .9(a): z z : ei(k1 +q1)·X(z1 .8 CHAPTER 8 76 upper halfplane is by ﬁxing z2 on the positive imaginary axis and integrating z1 from 0 to z2 . As usual.28) to calculate gc CD2 .p (2π)p+1 δp+1 (k1 + k2 ) z2 0 2α′ k 2 −3 × 22α k ′ 2 −3 z2 α k ′ 2 −1 dz1 z1 α k 1 ′ 2 −2 z1 − z2 −α s/2−3 z1 + z2 α s/2−2k ′ ′ ′ ′ 2 +5 2 = −igc CD2 . (41) To evaluate the X path integral we use the result of problem 8. and deﬁned the parameters 2 2 2 2 k2 ≡ k1 = k2 = 4 − q1 = 4 − q2 . and (8.p (2π)p+1 δp+1 (k1 + k2 ) ′ 2 2 ′ 2 2 ′ ′ X = iCD2 . −α′ s/4 − 1). but this time ﬁxing k2 /s. = −i 32 2 In the last line we have used (26). The ghost path integral is 1 (c(z1 ) + c(¯1 ))c(z2 )˜(¯2 ) ˜z cz 2 = D2 g CD2 ((z1 − z2 )(z1 − z2 ) + (¯1 − z2 )(¯1 − z2 )) (z2 − z2 ) ¯ z z ¯ ¯ 2 g = 2CD2 (z1 − z2 )(z1 + z2 )z2 . ′ 2 2 +5 (44) (b) In the Regge limit we increase the scattering energy. while holding ﬁxed the momentum transfer s. (43) So we have 2 S = −gc CD2 .p (2π)p+1 δp+1 (k1 + k2 ) × z1 − z1 α (k1 −q1 )/2 z2 − z2 α (k2 −q2 )/2 z1 − z2 α (k1 ·k2 +q1 ·q2 ) z1 − z2 α (k1 ·k2 −q1 ·q2) ¯ ¯ ¯ ′ 2 −4 × 22α k z1 α k ′ 2 −2 z2 α k ′ 2 −2 z1 − z2 −α s/2−4 z1 + z2 α s/2−2k ′ ′ 2 +4 . z1 ) :: c˜ei(k2 +q2 )·X (z2 .9.7. .p . The amplitude is 2 S = gc e−λ z2 0 ¯ c ¯ dz1 : c1 ei(k1 +q1 )·X (z1 . the beta function in the amplitude give us Regge behavior: S ∼ (−k2 )α s/4+1 Γ(−α′ s/4 − 1).p .¯1 ) :: ei(k2 +q2 )·X(z2 . k and q represent the momenta parallel and perpendicular to the Dbrane respectively. the beta function gives exponential behavior in this limit. (29). (42) We have used the kinematic relations k1 + k2 = 0 and (k1 + q1 )2 = (k2 + q2 )2 = 4/α′ . ′ (45) In the hard scattering limit we again take k2 → −∞. k2 → −∞. (40) where.
. Note that there are also poles at s = 0 and s = −4 (the poles at positive s are kinematically forbidden). without producing open strings) by the Dbrane. −1. representing an onshell intermediate closed string: the tachyon “decays” into another tachyon and either a massless or a tachyonic closed string. and the latter is then absorbed (completely. These poles come from the region of the integral in (44) where z1 approaches z2 . .8 CHAPTER 8 77 (c) The beta function in the amplitude has poles at α′ k2 = 1. These poles come from the region of the integral in (44) where z1 approaches the boundary. . . 0. . representing onshell intermediate open strings on the Dbrane: the closed string is absorbed and then later reemitted by the Dbrane.
Our orthonormal basis for the periodic functions on [0. so we regulate it by dividing by the same determinant with ω → Ω: Ω ω For large Ω this becomes eΩU/2 .9 APPENDIX A 78 9 9. U ] will be: 1 f0 (u) = √ .1 (a) We proceed by the same method as in the example on pages 339341.1 Appendix A Problem A. 1 2 sinh 2 ωU the divergence can easily be cancelled with a counterterm Lagrangian Lct = Ω/2. 4π 2 j 2 + ω 2 U 2 (3) √ 2π = ω This inﬁnite product vanishes. These are eigenfunctions of ∆ = −∂u + ω 2 with eigenvalues λj = Hence (neglecting the counterterm action) ˆ Tr exp(−HU ) = = = 2πj U 2 + ω2. gj (u) = U U (1) 2 where j runs over the positive integers. U 2πju 2 cos . 2 sinh 1 ωU 2 (6) (5) ∞ j=1 1+ 1+ ΩU 2πj ωU 2πj 2 2 = sinh 1 ΩU 2 . 2 (7) . sinh 1 ωU 2 (4) 1 Ei = (i + )ω. giving ˆ Tr exp(−HU ) = ˆ The eigenvalues of H are simply 1 . (2) [dq]P exp(−SE ) det P 2π λ0 ∆ 2π −1/2 ∞ j=1 ∞ j=1 2π λj 2πU 2 . fj (u) = U U 2πju 2 sin .
32) gives ˆ Tr exp(−HU ) = ∞ i=0 exp(−Ei U ) ∞ i=0 =e −ωU/2 e−iωU (8) 1 . so using (A. since the even i eigenstates are also even under reﬂection. and odd i eigenstates odd under reﬂection: ˆ ˆ Tr exp(−HU )R = = ∞ i=0 (−1)i e−(j+1/2)ωU (14) 1 . (10) Before including the counterterm and regulating. with eigenvalues fj (u) = λj = 1 2π(j + 2 ) U 2 (9) + ω2 . 2π(j + 1 )u 2 2 cos . U U 2π(j + 1 )u 2 2 gj (u) = sin . U U where the j are nonnegative integers. (b) Our basis for the antiperiodic functions on [0. U ] consists of the eigenfunctions of ∆. we have ˆ ˆ Tr exp(−HU )R = 2πU 2 . 1 2 cosh 2 ωU (13) ˆ This result can easily be reproduced by summing over eigenstates of H with weight (−1)R . = 1 2 sinh 2 ωU To be honest. 1 2 cosh 2 ωU (12) 1 . this becomes e−Lct U ∞ j=0 1+ 1+ ΩU 1 2π(j+ 2 ) ωU 1 2π(j+ 2 ) = e−Lct U 1 cosh 2 ΩU cosh 1 ωU 2 ∼ so the answer is ˆ ˆ Tr exp(−HU )R = e(Ω/2−Lct )U . (2π(j + 1 ))2 + (ωU )2 2 j=0 2 2 ∞ (11) After including the counterterm action and dividing by the regulator. 1 2 cosh 2 ωU .1. the overall normalization of (6) must be obtained by comparison with this result.9 APPENDIX A 79 for nonnegative integer i.
). .1. + m2 . (16) 2πα′ The periodic eigenfunctions of ∆ can be given in a basis of products of periodic eigenfunction of 2 2 −∂1 on σ1 with periodic eigenfunctions of −∂2 on σ2 : Fjk (σ1 . .2 Problem A. σ2 ) = fj (σ1 )gk (σ2 ).k=0 2π λjk ∞ k=0 nj nk /2 = ∞ j=0 2π λjk 1 nk /2 nj = ∞ j=0 1 2 sinh 2 nj j 2 + m2 T . The path integral is ∆ det P 2π −1/2 = ∞ j. . . (20) . The inﬁnite product vanishes. . . .3 1 2 2 d2 σX(−∂1 − ∂2 + m2 )X 4πα′ 1 = d2 σX∆X. 2π 1 fj (σ1 ) = √ π 1 g0 (σ2 ) = √ . j = 1.3 Problem A. . k = 1.5 We assume that the Hamiltonian for this system is H = mχψ. . T 1 gk (σ2 ) = √ 2T The eigenvalues are λjk = 1 2πα′ j2 + 2πk T 2 (17) sin cos jσ1 . 1 f0 (σ1 ) = √ . . sin cos kσ2 . 2. 2. where n0 = 1 and ni = 2 (i = 1. (19) where in the last step we have used the result of problem A. . (18) with multiplicity nj nk .9 APPENDIX A 80 9. 2. 2 ∆= The action can be written S = (15) 1 2 2 (−∂1 − ∂2 + m2 ). it would have to be regulated and a counterterm introduced to extract the ﬁnite part. 9.
U ] are 1 fj (u) = √ e2πiju/U .1(a) (eq. (21) [dχdψ] exp 0 (22) (23) (24) 2πij −m . U with j running over the integers. Regulating and renormalizing in the same manner as in that problem yields: 1 ˆ ˆ Tr (−1)F exp(−HU ) = 2 sinh mU. (26) This is essentially the inverse of the inﬁnite product that was considered in problem A. U (25) [dχdψ] exp 0 duχ∆ψ = = ∞ j=−∞ ∞ j=−∞ λj − 2πij +m U 2πj U 2 = −m ∞ j=1 + m2 .9 APPENDIX A 81 The periodic trace is ˆ Tr (−1)F exp(−HU ) = = = where ∆ = −∂u − m. U while the eigenfunctions of ∆T = ∂u − m are 1 gj (u) = √ e−2πiju/U . The periodic eigenfunctions of ∆ on [0. Their eigenvalues are λj = − so the trace becomes U ˆ dψ ψ. 0 [dχdψ] exp E U 0 U du(−χ∂u ψ − H) duχ∆ψ . U ψ. (3)). 2 (27) This answer can very easily be checked by explicit calculation of the LHS. The Hamiltonian operator can be obtained from the classical Hamiltonian by ﬁrst antisymmetrizing on χ and ψ: H = mχψ = 1 m(χψ − ψχ) 2 (28) .
cosh 1 M U 2 (33) With the same counterterm Lagrangian as before to cancel the divergence in the denominator as M → ∞.2. 2 the same as would be found using (30).. (32) When we regulate the product.3/2. it becomes 2πj U j 2πj U 2 2 + m2 = + M2 = j 1+ 1+ mU 2πj MU 2πj 2 2 j cosh 1 mU 2 ..9 APPENDIX A 1 ˆ −→ H = m(χψ − ψ χ). the only diﬀerence being that the index j runs over the halfintegers rather than the integers in order to make the eigenfunctions (23) and (24) antiperiodic. 2 so. (34) . ˆ ˆ ˆˆ 2 Now 1 ˆ H ↑ =  ↑ . ˆ Tr (−1)F exp(−HU ) = emU/2 − e−mU/2 . Eq.. using A. The antiperiodic trace is calculated in the same way.22. we are simply left with 1 ˆ Tr exp(−HU ) = 2 cosh mU. 2 1 ˆ H ↓ = −  ↓ . 2πj U 2 + m2 . (26) becomes U [dχdψ] exp 0 duχ∆ψ = ∞ j=1/2. ˆ 82 (29) (30) (31) in agreement with (27).
where For ψ. δv ]ψ = δη′ ψ. (12) (10) (11) 1 η ′ = −v∂η + ∂v η. 2 2 (6) (7) (8) (9) so [δη .3.2 ∗˜ ∗ ∗¯ δη1 δη2 X = δη1 (η2 ψ + η2 ψ) = −η2 η1 ∂X − η2 η1 ∂X. (5) 2 ˜ the second term correctly reproducing the weight of ψ. so [δη .2 Problem 10. using the anticommutativity of the ηi . 1 δη δv ψ = v∂η∂X + vη∂ 2 X + η∂v∂X.1 Chapter 10 Problem 10. 2 . (a) We have (2) so.1 (a) The OPEs are: TF (z)X µ (0) ∼ −i α′ ψ µ (0) . ∗ ∗¯ [δη1 . (b) For X: ¯˜ ˜ δη δv X = −vη∂ψ − v∂η ψ − v ∗ η ∗ ∂ ψ − v ∗ (∂η)∗ ψ.10 CHAPTER 10 83 10 10. 1 ¯˜ 1 ˜ δv δη X = −vη∂ψ − η∂vψ − v ∗ η ∗ ∂ ψ − η ∗ (∂v)∗ ψ. For ψ we must apply the equation of motion ∂ ψ = 0: ∗˜ δη1 δη2 ψ = δη1 (−η2 ∂X) = −η2 ∂(η1 ψ + η1 ψ) = −η2 η1 ∂ψ − η2 ∂η1 ψ. 2 z TF (z)ψ µ (0) ∼ i α′ ∂X µ (0) . The ψ transformation works out similarly. 2 δv δη ψ = ηv∂ 2 X + η∂v∂X. δη2 ]ψ = −v∂ψ − ∂v ψ = δv ψ. 10. δη2 ]X = 2η1 η2 ∂X + 2η1 η2 ∂X = δv X (3) ˜ (see (2.7)). (4) so 1 [δη1 . 2 z (1) (b) The result follows trivially from (2.11) and the above OPEs. ˜ and similarly for ψ.4. δv ]X = δη′ X.
ψ ψ X X TB (z)TB (0) ∼ TB (z)TB (0) + TB (z)TB (z). since that is simply the sum of the X part and the ψ part.3 ψ X (a) Since the OPE of TB = −(1/α′ )∂X µ ∂Xµ and TB = −(1/2)ψ µ ∂ψµ is nonsingular. 2z z 2 1 µ D TF (z)TF (0) ∼ 3 − ′ ∂Xµ (z)∂X µ (0) + 2 ψ (z)ψµ (0). Ln ] commutator (10.11a) is as in the bosonic case. = 12 (17) .4 Problem 10.13a). (13) which does indeed reproduce (10. there is no need to check the TB TB OPE. namely Vµ ∂ 2 X µ (z)i √ 1 2 ν ψ ∂Xν (0) + ψ µ ∂ψµ (z)i 2α′ Vν ∂ψ ν (0) α′ 2 √ √ i 2α′ i 2α 2 1 µ µ ∼ Vµ ψ (0) − i Vµ ∂ψ (z) − Vµ ψ µ (z) z3 α z2 z3 √ α′ 1 i 2α′ µ ∼ −3i Vµ ∂ψ (0) − Vµ ∂ 2 ψ µ (0).1b). The current associated with the charge {Gr . The new terms in TB and TF add two singular terms to their OPE. We then have 1 TB (z)TF (0) 1 1 ∼ 2 ∂X µ (z)ψµ (0) + ∂ µ (z)∂ψ µ (z)∂Xµ (0) + 2 ψ µ (z)∂Xµ (0) ′ z 2z 2z i 2/α 3 1 ∼ 2 ψ µ ∂Xµ (0) + ψ µ ∂ 2 Xµ (0) + ∂ψ µ ∂Xµ (0). z αz z (14) (b) Again. according to (2. Gs } is. Resz1 →z2 z1 r+1/2 TF (z1 )z2 s+1/2 r+s+1 TF (z2 ) = Resz12 →0 z2 1+ z12 z2 r+1/2 2 2c 3 + z TB (z2 ) 3z12 12 (4r 2 − 1)c r+s−1 r+s+1 z2 + 2z2 TB (z2 ). The new terms in the TF TF OPE are 2ψ µ ∂Xµ (z)Vν ∂ψ ν (0) + 2Vµ ∂ψ µ (z)ψ ν ∂Xν (0) − 2α′ Vµ ∂ψ µ (z)Vν ∂ψ ν (0) ∼ 2 4α′ V 2 2 Vµ ∂X µ (z) − 2 Vµ ∂X µ (0) + z2 z z3 ′V 2 2 4α ∼ Vµ ∂ 2 X µ (0) + .14).10 CHAPTER 10 84 10. 2 z2 z (15) which are precisely the extra terms we expect on the right hand side of (10.1.3 Problem 10.1.4 The [Lm .6. z z3 (16) 10.2.
in the limit u → 0.11c). so (applying the same argument to the dependence on all the zi ) the expression in (10.3. Gr+m . 0 = [Lm . [Lm .7) goes like z1 .7) would have to be an entire function which approaches a constant as z1 → ∞. Gs } is in turn the residue of this expression in z2 . Gr ] we have m+1 Resz1 →z2 z1 TB (z1 )z2 r+1/2 TF (z2 ) = Resz12 →0 z2 = r+m+3/2 1+ z12 z2 m+1 3 1 2 TF (z2 ) + z ∂TF (z2 ) 2z12 12 3(m + 1) r+m+1/2 r+m+3/2 z2 TF (z2 ) + z2 ∂TF (z2 ).0 (20) 10. transforming to the u = 1/z1 frame this expectation value becomes constant.11b).3. which is easily seen to equal the RHS of (10.2.5 Problem 10.10 CHAPTER 10 85 {Gr . . But the only such function is a constant.6 Problem 10. {Gr . This is the correct behavior—the only poles and zeroes of the expectation value should be at the positions of the other operators.11). One of the Jacobi identies for the superconformal generators is. using (10. Similarly.5 Let us denote by cB the central charge appearing in the TB TB OPE. the expectation −1 value (10. If we now consider some other function. 2 (18) The residue in z2 of this is 3 3(m + 1) Gr+m − r + m + 2 2 in agreement with (10. and by cF that appearing in the TF TF OPE.7) is unique up to a constant. (19) 10. the ratio between this function and the one given in (10. Gr ]} cF 1 = cB (m3 − m) + (2s − m)(4r 2 − 1) + (2r − m)(4s2 − 1) 6 4 1 = (cB − cF )(m3 − m)δm+r+s. O(1). Gs }] + {Gr . for [Lm . with exactly the same poles and zeroes and behavior as z1 → ∞.0 .2. 6 Hence cB = cF . δm+r+s. Since eiǫ1 H(z1 ) is a tensor of weight 1/2. [Gs .7 Taking (for example) z1 to inﬁnity while holding the other zi ﬁxed at ﬁnite values. Lm ]} − {Gs .3.2.
0) (0. 0) (±1. but as we Taylor expand ψ(z). 0) (1. 1. 0. 0. 1. N2 ) (N1 + N1 . N1 .10 On the bosonic side. 2. 1. 0) (23) 5/2 (2. and 1 respectively. 0) (±1. (22) 2 n=1 We will denote states on the bosonic side by (kL . N3 + N3 ). i! (26) The OPE of ψ(z) with Fn (0) is nonsingular. 0. N2 + N2 . N1 . On the fermionic ¯ side Nn + Nn can take the values 0. Here are the states with L0 = 0. all the terms vanish until the nth one because ψ is fermionic. 0. 5/2 on each side: ¯ ¯ ¯ L0 (kL . 1/2. 0. 0) (±1. N3 + N3 ) 0 1/2 1 3/2 2 (0. 0. Fn and einH obviously have the same fermion number . generated by the ﬁelds ψ and ψ. 2. there are two sets of oscillators. Hence ψ(z)Fn (0) = z n Fn+1 (0) + O(z n+1 ). 0) (0.11 We will use the ﬁrst of the suggested methods. and the energy is ∞ 1 ¯ L0 = n− (Nn + Nn ). the energy eigenvalue in terms of the momentum kL and oscillator occupation numbers Nn is ∞ 1 2 nNn . 0. . or 2. 0) (0. 0) (0. We won’t need any higher oscillators for this problem. 0) (1. 1. 0) (2.8 Problem 10. with degeneracy 1. 0. and on the fermionic side by (N1 + ¯ ¯ ¯ N1 . . 1.10 CHAPTER 10 86 10. 1) 10. 1. N2 + N2 . 0) (±2. 2.7 Problem 10. . 0) (0. The OPE we need is : eiH(z) :: einH(0) := z n : ei(n+1)H(0) : +O(z n+1 ). 1) (0. 0. It’s easy to see that this is satisﬁed by n−1 (24) (25) Fn =: i=0 1 i ∂ψ:. N2 ). (21) L0 = kL + 2 n=1 ¯ On the fermionic side. 1) (±1. .
Their dimensions work out nicely: for einH we have n2 /2.14 We start with the NS sector. We are left with the 8 transverse = polarizations of a massless vector. while e ∼ e + 2α′ f ′ k for any f ′ . (29) where u is a 10dimensional Dirac spinor. (27) is annihilated by b0 . The general massless state is ψ = u.26). The most general massless state is ψ = (e · ψ−1/2 + f β−1/2 + gγ−1/2 )0. and none by the BRST operator. The BRST charge acting on QB ψ = (c0 L0 + γ−1/2 Gm + γ1/2 Gm )ψ 1/2 −1/2 √ ′ (e · kγ = 2α −1/2 + f k · ψ−1/2 )0. which is exactly the OCQ condition. . For ψ to be √ closed requires e · k = f = 0.5. Furthermore exactness of (28) implies g ∼ g + 2α′ e′ · k for any e′ = √ (so we might as well set g = 0). 10. k R. k NS . k where k2 = 0 (by the L0 condition) and 0. since they are all closed. for a total of n/2 + n(n − 1)/2 = n2 /2. which implies that it is annihilated by Gg . (28) The L0 term of course vanishes.10 CHAPTER 10 87 n. According to (10. which follows more or less trivially from all the above conditions. Finally. The only thing left to check is that all 0 the states satisfying these conditions are BRST closed. this in turn implies that it 0 is annihilated by Gm . k ψ is NS NS . ¯ For e−inH we obviously just replace ψ with ψ.26). along with many others we have not indicated. The R case is even easier: all of the work is done by the constraint (10. and k2 = 0. for Fn we have n ψs and n(n − 1)/2 derivatives.9 Problem 10.5. ψ is deﬁned to be annihilated by b0 and β0 . none of them can be exact.
NS.3 See the last paragraph of section 11. 11.1 + − In order to establish the normalizations.2. (R. e z z2 2z z (1) Since the third term is supposed to be 2TB (0)/z. (R. (+ − −). R)}. (7) where the ﬁrst symbol in each triplet corresponds to the ﬁrst 16 leftmoving fermions. R). R. NS. R). (− + −)} × {(NS. NS). The TB TF and TB j OPEs are ± ± ± from Chapter 2. R. R. NS.2 Problem 11. the second to the second 16. R.11 CHAPTER 11 88 11 11. The fact that the TF TF 2 ± OPE is nonsingular was also shown in Chapter 2. the untwisted theory contains 8 sectors: {(+ + +). 2 2 ±i√3H e . (R. (− − +). 3 It follows from the ﬁrst term that c = 1 (as we already knew).4 (a) Dividing the 32 leftmoving fermions into two groups of 16. 3 z 1 . NS. For the jTF OPE we have ± j(z)TF (0) =± 2 e±i 3H(0) . where 1 TB = − ∂H∂H. j(z)j(0) = (6) 11. we project out (− − +) and (− + −). we ﬁrst calculate the TF TF OPE: √ √ √ √ 1 i 3∂H(0) 3∂H(0)∂H(0) i2 3∂ 2 H(0) +i 3H(z) −i 3H(0) e ∼ 3+ − + . (NS. If we now twist by exp(πiF1 ). but project in the twisted sectors (R. NS). We again have 8 sectors: {(+ + +). and the third to the 8 rightmoving fermions. NS) and (NS. and from the second that ± TF = (2) we need (3) i j = √ ∂H. NS). (8) .1 Chapter 11 Problem 11. 3z 2 √ (5) Finally. 3 (4) ± It is now straightforward to verify each of the OPEs in turn. (+ − −)} × {(NS. R)}.3 Problem 11. and show that TF and j have weight 3 and 1 respectively.
R. 0NS . 1. 8. 0NS . NS+) and (R+. 1) (8v . NS+. R−. 0NS −1 ′ graviton. the only states are λA 0NS . 1) (8v . 1) (8v . NS+. NS−. 8. NS+. 28) (8v . together with their SO(8)spin × SO(8)1 × SO(8)2 × SO(8)3 × SO(8)4 quantum numbers: (NS+. R. 1. 0NS . 0NS . 0NS . 0NS . 1) (8v . 0NS . 0NS . 0NS . 28. These will have rightmovers in the NS+ sector. 0NS . NS. 1) (8v . NS+) : ˜i ψ−1/2 uR . (R. NS+) : (R+. R). NS. NS. NS). these will give respectively (8. 0NS −1/2 −1/2 (NS+. (NS. 28. R). adjoint of SO(16)′ . 1. R+. NS). 0NS . 1. 0NS −1/2 ˜i λA λB ψ−1/2 0NS . 0NS . 0NS . NS. NS+. R+. NS+. R+. vR . 0NS . NS+) : ˜i λA4 λB4 ψ−1/2 0NS . 0NS . R) states are all massive due to the positive normalordering constant for the leftmoving R fermions). so there are two possibilities. 1. NS+. 0NS . NS. R)}. 0NS −1/2 −1/2 ′ ˜i ψ−1/2 uR . 1. 0NS . R+. (+ + + − −). R. dilaton. R. 0NS The last two sets combine to form an adjoint of E8 . R. (NS+. 8. R. 1. 8. (R. 0NS . to establish the gauge group. 1. (NS+. 0NS −1 (1 ⊕ 28 ⊕ 35. NS+. (9) B ˜i λA λ−1/2 ψ−1/2 0NS . R+. NS+. 0NS . R. R. 0NS . 1. so the gauge group is E8 × SO(16). (11) ˜i λA3 λB3 ψ−1/2 0NS . There are similarly two sectors containing massless spacetime fermions. 28. vR . NS. antisymmetric tensor. 128′ ) of SO(16)spin × E8 × SO(16). R. 1. 1. 1. 8. 0NS . we get a total of 32 sectors: {(+ + + + +).11 CHAPTER 11 89 Let us ﬁnd the massless spacetime bosons ﬁrst. 0NS . the tachyon must be in (NS+. 1. NS−). NS. NS+. (10) Again we begin by listing the massless spacetime bosons. NS. NS). (b) Dividing the leftmoving fermions into four groups of 8. R−) (the (R. 0NS . NS+) : (R+. uR . 1. R). NS. 0NS ˜i λA2 λB2 ψ−1/2 0NS . NS). if we twist the above theory by the total fermion number of the ﬁrst and third groups. NS+). (R. R. 8v ). NS. 1. 1) (8v . NS. 0NS . NS+. 1. R+) and (NS+. 0NS ˜i ψ−1/2 0NS . 0NS . R. vR . 128 of SO(16). (NS. 0NS . R. (NS. which −1/2 transform as (1. 0NS −1/2 −1/2 ˜i λA1 λB1 ψ−1/2 0NS . 0NS −1/2 −1/2 ˜j αi ψ−1/2 0NS . adjoint of SO(16). (− − − − +). and the states are easily enumerated (primes refer to the second set of leftmoving fermions): ˜j αi ψ−1/2 0NS . 1). NS. 0NS . 128) and (8′ . (− − − + −)}× {(NS. ′ Finally. R. 8. 1. 0NS −1/2 −1/2 ˜i ψ−1/2 uR . 1. (R.
20). 0NS . 0NS . vR .2. R+. 1. this can also be checked explicitly. R+) : (NS+. 0NS . 0NS . 8) (8. 1. 8v ). 2πi z (14) where the contour goes around the origin. Finally. NS+. 0NS . For the other two terms we have: c a a a b b a j0 j−1 j−1 0 = if cab (j−1 j−1 + j−1 j−1 )0 = 0. (17) . NS−. since (being at level 0) it can only be proportional to the ground state 0 . 1. NS+.5. 0NS . R+. the rest are either nonsingular (for m < 0) or zero (for m > 2. R+) : uR . In fact.2) we have : jj(0) : j (z) ∼ = c ∞ m=−∞ 1 c a a j j j 0 m+1 m −1 −1 z . uR . 8) (8. leaving no room for the free Lie algebra index on (15). and graviton. once we perform triality rotations on SO(8)1. vR . 8). NS+.5. the term with m = 2 must also vanish. NS−) : λA4 0NS . which is precisely (11. with z1 = 0 and z3 = z. The contour integral picks out the z 0 term in the Laurent expansion of j a (z)j a (0).4 Problem 11.2. 1. NS+. vR . (15) Only three terms in this sum are potentially interesting. (16) c a a b a c a ˆ j1 j−1 j−1 0 = (kδca + if cab j0 + j−1 j1 )j−1 0 ˆ = (2kj c − f cab f bad j d )0 −1 = (k + c h(g))ψ 2 j−1 0 −1 . 8. NS+. antisymmetric tensor.18). since the total level of the state would be negative). uR . wR 0NS . which combine into an SO(24) vector.3 to turn the bispinors into bivectors.5. We will now use this to prove the ﬁrst line of (11. and j−1 = dz/(2πi)j a (z)/z. The rest combine to give gauge bosons of SO(24). 8. 0NS −1/2 (1. Since j−1 0 clearly a corresponds to j a (0). 1. R+) : (NS+. 1. wR (8. Using the Laurent expansion (11. we have : jj(0) : = dz j a (z)j a (0) . (12) The triality rotation again turns the SO(8)1. R+. R+. 1. (13) 11. the tachyon is an SO(8)4 vector but is neutral under SO(24): (NS+. wR 0NS .7 a a a We wish to show that the state j−1 j−1 0 corresponds to the operator : jj(0) :. 1. 8. 1. R+.11 CHAPTER 11 90 The ﬁrst set of states gives the dilaton. The next set gives the SO(8)4 gauge bosons.3 spinors into vectors. The massless spacetime fermions are: (R+. NS+.
and the value for m = 1 is given by (17) above. (20) z z z Clearly the order z 0 term is : ∂λA λA : + : ∂λB λB :. From the Laurent expansion we see that the state corresponding to ∂TB (0) is indeed b b L−3 0 = 2/((k + h(g))ψ 2 )j−2 j−1 0 .5. The OPE will be the operator corresponding to 1 (k + h(g))ψ 2 1 1 1 1 a a L + 3 L1 + 2 L0 + L−1 j−1 j−1 0 . and ψ 2 = 2. (22) . using the Laurent coeﬃcients (11. so s TB = (sum).5. (k + h(g))ψ 2 1 1 −1 −1 2 b b j b j b j a j a 0 = 2j−1 j−1 0 . Summing over all A and B with B = A double counts the currents.20). so we divide by 2: : jj : = (n − 1) : ∂λA λA : Finally. the only slightly nontrivial one being s the last. Thus we have precisely the OPE (11. First let us calculate the contribution from a single current iλA λB (A = B): iλA (z)λB (z)iλA (0)λB (0) = λA (z)λA (0)λB (z)λB (0) (no sum) 1 1 1 = : λA (z)λA (0)λB (z)λB (0) : + : λA (z)λA (0) : + : λB (z)λB (0) : + 2 . (k + h(g))ψ 2 −1 1 −1 −1 2 b b a a j b j b j a j a 0 = 2j−2 j−1 0 . Thus: a a L2 j−1 j−1 0 = 1 b b ˆ j b j b j a j a 0 = j1 j−1 0 = kdim(g)0 .8 The operator : jj(0) : is deﬁned to be the z 0 term in the Laurent expansion of j a (z)j a (0).11 CHAPTER 11 91 The RHS of (17) clearly corresponds to the RHS of (11. 11. h(SO(n)) = n − 2.24).24). To check the T T OPE (11.5. we employ the same strategy. Life is made much easier by the b a a fact that jm j−1 j−1 0 = 0 for m = 0 and m > 1. L−1 j−1 j−1 0 = (k + h(g))ψ 2 −2 1 −1 −1 a a L0 j−1 j−1 0 = a a L1 j−1 j−1 0 = 0.5 Problem 11.26). (21) 1 : ∂λA λA : 2 (sum). we have k = 1. (19) All of these states are easily translated back into operators.5. 4 2 z z z z (18) terms with Lm are nonsingular for m < −1 and vanish for m > 2.
5) and (11. 32021) are misleading.9 In the notation of (11. (24) l◦l = i=1 n2 − i i=23 n2 ∈ 2Z. ˜ and the 6 gauge bosons with vertex operators ∂X µ ψ m . . . It’s easy to see that the dual lattice to Γ1 is Γ∗ = 1 28 ∪ (Z 28 + l ) ⊃ Γ. the set of points of the form (n1 . that is. ′ 2 nL (m)e−πα m l/2 ∼ eπc/(12l) .8. . generating U (1)6 . providing the Cartan generators to ﬁll out the adjoint representation of SO(44). i (25) Evenness of l + l0 ∈ Γ2 follows from the same fact: (l + l0 ) ◦ (l + l0 ) = l ◦ l + 2l0 ◦ l + l0 ◦ l0 = l ◦ l + 22 i=1 28 ni − i=23 ni + 4 ∈ 2Z. .13).7 Problem 11. But these are obviously the root vectors of SO(44). . . .11 CHAPTER 11 92 11. Γ = Γ1 ∪ Γ2 where 1 1 l0 ≡ ( .15). 2 2 (23) ni ∈ 2Z for any integers ni . implying Γ2 = Γ1 + l0 . Taking into account level matching we have n(m) = nL (m)nR (m). i. To ﬁnd nL and nR we treat the leftmoving and rightmoving CFTs separately. . (27) m2 . It will be convenient to divide Γ into two sublattices.6). the lattice Γ is Γ22. In addition there are the 22 gauge bosons with vertex ˜ operators ∂X m ψ µ .11). 2 2 Evenness of l ∈ Γ1 follows from the fact that the number of odd ni must be even. n28 ) or i 1 1 (n1 + . n28 ) : ni ∈ 2Z}. n28 + ). But to be dual for example to l requires a vector to have an even number Z 0 0 ∗ = Γ. (26) Evenness implies integrality. .20) is not the correct one to use to ﬁnd the asymptotic density of states of a string theory. of odd ni s. . . (9.6. I. . Then we have. so Γ To ﬁnd the gauge bosons we need to ﬁnd the lattice vectors satisfying (11. Equation (7.6 .12 It seems to me that both the statement of the problem and the derivation of the Hagedorn temperature for the bosonic string (Vol.8. and include only the physical parts of the spectrum.3. as in (9. so Γ ⊂ Γ∗ . since it does not take into account the level matching constraint in the physical spectrum. . . neglect the ghosts and the timelike and longitudinal oscillators. ). pp. 22 28 Γ1 = {(n1 . It’s essentially a matter of luck that Polchinski ends up with the right Hagedorn temperature. .6. 11.6 Problem 11.6. .e.
The partition function from which we will extract nL (m). 2 (34) We see that the ground state. There is an interesting point that we glossed over above. But the sum on R and NS sectors in (33) means that we now project onto states with (−1)F = 1: Z(il) = i∈R. (33) This partition function corresponds to a path integral on the torus in which we sum over all four spin structures. with h = 0. (31). the number of projectedin states. this gives TH = while for the heterotic theories we have 1 1 TH = √ (1 − √ ). and therefore dominates in the limit l → 0. else it would give rise to a tachyon. is indeed projected in. is Z(il) = i∈R. We implicitly took the lowestlying state to be the vacuum. as for the closed string. with minus signs when the fermions are periodic in the σ2 (“time”) direction. Then for small l only the lowestlying state in the theory contributes. So should we really consider it? To see that we should. since nopen (m) = nL (2m). The GSO projection there is (−1)F = −1. where F is the worldsheet fermion number of the transverse fermions (not including the ghosts). Upon doing the modular integral.11 CHAPTER 11 93 implying Similarly for the rightmovers. let us derive (27) more carefully.NS 1 q hi −c/24 (1 − (−1)Fi ) = 2 nL (m)e−πα m m2 ′ 2 l/2 . . However.NS 1 e−2π/l(hi −c/24) (−1)αi (1 + (−1)Fi ) ∼ eπc/(12l) . that state is projected out by the GSO projection in all of the above theories. corresponding to the unit operator. this minus sign corresponds to giving a minus sign to R sector states. The RHS of (27) is obtained by doing a modular transformation l → 1/l on the torus partition function with τ = il. (28) (29) The Hagedorn temperature is then given by √ −1 TH = π α′ For the type I and II strings. 2 π α′ (32) 2π 2α′ 1 √ . (31) c + 6 c ˜ 6 . giving √ ′ nL (m) ∼ eπm α c/6 . (30) The Hagedorn temperature for an open type I string is the same. √ ′ √ √ ˜ n(m) ∼ eπm α /6( c+ c) . for example in the case of the leftmovers of the type II string.
Γ(1 − α′ s − α′ u) Finally. and we have used (13. ei ) Γ(−α′ s)Γ(−α′ u) + 2 permutations . . ei ) Γ(−α′ s)Γ(−α′ u) + 2 permutations . Since the momenta ki all have vanishing components in the 9 − p dualized directions. .3. the amplitude becomes.29). −i 0 (1) Four open strings attached to the same Dpbrane are Tdual to four open type I strings with zero momentum in the 9 − p dualized directions and the same ChanPaton factor (1). Γ(1 − α′ s − α′ u) where V9−p is the volume of the transverse space. An open string with both ends attached to the same Dpbrane and zero winding number is Tdual to an open type I string with zero momentum in the 9 − p dualized directions. The scattering amplitude for four gauge boson open string states was calculated in section 12. ei ) i Γ(−α′ s)Γ(−α′ u) Γ(1 − α′ s − α′ u) + 2 permutations . 2 2 −8ig(p+1).YM α′2 V9−p (2π)p+1 δp+1 ( ki ) i (4) ×K(ki . and we won’t bother to reproduce it here. 0). we must renormalize the wave function of each string (which is spread out uniformly in the transverse space) by a factor of V9−p : 2 S ′ (ki .YM α′2 (2π)p+1 δp+1 ( ki ) i (5) ×K(ki .25) and (12.3. ei ) = −8ig(p+1). 0. using (13.SO(32) = 2gDp .1 Chapter 13 Problem 13. Using Tr(ta )4 = the result (12.YM .4.4. .4. we can write the dimensionally reduced type I YangMills coupling g(p+1).28).4. (6) .12 CHAPTER 13 94 12 12.YM = gDp. ei ) = 2 −8igYM α′2 (2π)10 δ10 ( 1 . But in order to get the proper (p + 1)dimensional scattering amplitude.2 An open string with ends attached to Dpbranes is Tdual to an open type I string. and ChanPaton factor of the form 1 ta = √ 2 0 i ⊗ diag(1. . The kinematic factor K is written in three diﬀerent ways in (12. 2 (2) (3) ki )K(ki .3. in terms of the coupling gDp on the brane: 2 2 2 g(p+1).30) and (13.26).22) becomes in this case S(ki .
(10) (9) (8) and label the D4branes extended in the (6.3.5. then ⊥ (11) β2 = β 1 β 2 β 3 β 4 β 5 = 2iS2 β and similarly ⊥ β3 = 2iS3 β.12 CHAPTER 13 95 so 2 S ′ (ki .25) to write gDp in terms of the string coupling g.1. there are four unbroken supersymmetries: ˜ Q(+++++) + i(β Q)(+++++) . ˜ Q(−+−−−) − i(β Q)(−+−−−) .8. Taking into account the chirality condition Γ = +1 on Qs .10).) 12. where a = 1. ⊥ β4 = 2iS4 β. so β 2a β 2a+1 = 2iSa .8) and (B. and (4. (13) so for a supersymmetry to be unbroken by all three branes simply requires s2 = s3 = s4 . Γ2a Γ2a+1 = −2iSa . 2. (4.9). 4) are ⊥˜ ˜ Qs + (βa Q)s = Qs + 2isa (β Q)s . 3.7) directions by the subscripts 2. 3.9). ˜ Q(+−−−−) − i(β Q)(+−−−−) .2 Problem 13. 3. (12) The supersymmetries preserved by brane a (a = 2. (14) . If we deﬁne β ≡ β1β2β3. ei ) = −16igDp α′2 (2π)p+1 δp+1 ( ki ) i (7) ×K(ki .6.3 (a) By equations (B. and 4 respectively. Γ(1 − α′ s − α′ u) (One can also use (13.8. ˜ Q(−−+++) + i(β Q)(−−+++) .7. ei ) Γ(−α′ s)Γ(−α′ u) + 2 permutations .1. 4.5.
6) Further Tduality in one. 6 (5. 3. 8 (3. 5. p4 + 3) respectively. 1) 4. 2) 4. They result in the following brane content: Tdualized directions (p1 . 6. 6. p3 . 3. 4) 4.4 (a) Let the D2brane be extended in the 8 and 9 directions. 12. Tdualizing at general angles to the coordinate axes will result in combinations of the above conﬁgurations for the directions involved. 2. 0. 5. and (45) ↔ (67) ↔ (89). (16) The signs in all four previously unbroken supersymmetries (14) are just wrong to remain unbroken by the D0brane. 7. 5. 3. p3 + 1.3 Problem 13. p2 + 2. 8 (4. and 8 . 5) 4. 3) 4. or all three of the 1. 2. 6 ↔ 7. 5 (6. so that this conﬁguration preserves no supersymmetry. ⊥ βD0 = β 1 β 2 β 3 β 4 β 5 β 6 β 7 β 8 β 9 = −8iS2 S3 S4 β. 4. 3. 8 (3. p2 + 3. 8. 7 (4. 9 (2. p3 + 2. 4. 4. p4 ) 4 (5. 6. p4 + 2). 8 ↔ 9. 7. and 3 directions will turn any of these conﬁgurations into (p1 + 1. two. 2. 1. 2. p3 + 3. 6. 6. up to the symmetries 4 ↔ 5. and 9 directions. 6. p4 + 1). 7. 4) 4. with the smallerdimensional brane in each column being replaced by a magnetic ﬁeld on the largerdimensional brane. 2. (c) Let us make a brane scan of the original conﬁguration: 1 2 3 4 5 6 7 8 9 D42 D D D D D N N N N D43 D D D N N D D N N D44 D D D N N N N D D D0 D D D D D D D D D There are nine distinct Tdualities that can be performed in the 4. ⊥ ˜ ˜ Qs + (βD0 Q)s = Qs − 8is2 s3 s4 (β Q)s . 5. 2) 4. 3. 3. and (p1 + 3. 2. and let it be separated from the D0brane in the 1 direction by a distance y. (15) so the unbroken supersymmetries are of the form. p2 .12 CHAPTER 13 96 (b) For the D0brane. 5. 6 (4. 2. 1. 8. Tdualizing this conﬁguration in the 2. 3) 4. (p1 + 2. 2. p2 + 1. 5. 6.
4.4.25) (without the exponential factor.4. 2 3 4 (18) φ4 = π . 2 (17) For the three directions in which the D4branes are parallel.22)). k 4 ˜ ˜ where h = G − η.14) expanded for small values of Φ and h is Sp = −τp dp+1ξ p−3˜ 1 a Φ + ha . 6. (21) (22) their propagators are given in (8. 4. the ﬁrst (from the D2brane) extended in the 2.4. JΦ.7. From the point of view of the ˜ supergravity. since we have chosen the separation between the D0. φ′ = φ′ = 1 4 π . i/t) 11 .p (X) = ˜ 3 − p 9−p ′ τp δ (X⊥ − X⊥ ).18b). the Dbrane is thus a source for Φ.7. 4 π φ′ = φ′ = − . ϑ11 (1/2. i/t)η 9 (i/t) (20) (b) For the ﬁeld theory calculation we lean heavily on the similar calculation done in section 8.44b) and (13. but with a factor −i. it) ty 2 11 V (y) = − (8π 2 α′ t)−1/2 exp − .3. and the second (from the D0brane) in the 2. Polchinski employs the shifted dilaton ˜ Φ = Φ − Φ0 . k 2iκ2 1 hµν hσρ (k) = − 2 ηµσ ηνρ + ηµρ ηνσ − ηµν ησρ . it)η 9 (it) 0 Alternatively. whose expectation value vanishes.23): 2iκ ˜˜ ΦΦ(k) = − 2 . (19) t 2πα′ ϑ11 (it/2. 4. and 9 directions. and therefore (according to (13. The result is ∞ dt iϑ4 (it/4. V (y) = − √ 1 8π 2 α′ 0 ∞ dt t3/2 exp − ty 2 2πα′ ϑ4 (1/4. and 8 directions. and the Einstein metric ˜ ˜ G = e−Φ/2 G. This is in the class of D4brane conﬁgurations studied in section 13. to make the potential real and attractive).4.12 CHAPTER 13 97 directions yields 2 D4branes. adapting it to D = 10. 4 (26) . we must make the substitution (13. The Dbrane action (13.and D2branes to vanish in those directions. using the modular transformations (7. 4 2 (25) 2 (23) (24) where the trace on h is taken over directions tangent to the brane. in our case the angles deﬁned there take the values φ1 = φ2 = φ3 = 0. 6.
gravitation and dilaton exchange are both attractive forces. k8 . 1 µν ′ Jh. p 2 (28) (29) Between the D0brane. X2 .12 CHAPTER 13 98 and for h.p (k) = − τp eµν (2π)p+1 δp+1 (k )eik⊥ ·X⊥ .4). ˜ 8 (32) The calculation for the graviton exchange is similar. We divide the amplitude by −iT to obtain the static potential due to dilaton exchange: 3 VΦ (y) = − τ0 τ2 κ2 G7 (y). the only diﬀerence being that the numerical factor (1/4)2(3/4) is replaced by 1 µν 1 e0 2 ηµσ ηνρ + ηµρ ηνσ − ηµν ησρ 2 4 1 σρ 5 e = . located at the origin of space. As expected. i/t)η 9 (i/t) lim (35) . 0. (30) the amplitude for dilaton exchange is AΦ = − ˜ d10k ˜ ˜˜ ˜˜ J ˜ (k) ΦΦ(k) JΦ. k9 )eik1 y 8 (2π)10 k d7k eiky 3 = iT τ0 τ2 κ2 8 (2π)7 k2 3 = iT τ0 τ2 κ2 G7 (y). (27) 2 p ′ where X⊥ is the position of the brane in the transverse coordinates.p (k) = 4 1 ′ ˜µν Jh. the integrand becomes very small except where t is very small. (34) where we have applied (13. i/t) 11 = 2. 0. X4 . In the largey limit of (20). X5 . X3 . JΦ. ϑ4 (1/4. and the D2brane. The ratio of modular functions involved in the integrand is in fact ﬁnite in the limit t → 0.p (X) = − τp eµν δ9−p (X⊥ − X⊥ ).3. t→0 ϑ11 (1/2.0 3 1 d10k = i τ0 τ2 κ2 2πδ(k0 ) 2 (2π)3 δ3 (k0 . extended in the 8 and 9 directions and located in the other directions at the point ′ ′ ′ ′ ′ ′ ′ (X1 . and eµν is η µν in the directions p parallel to the brane and 0 otherwise. 0). 8 (31) where G7 is the 7dimensional massless scalar Green function.2 (−k) (2π)10 Φ. 0. In momentum space the sources are ′ 3−p ˜˜ τp (2π)p+1 δp+1 (k )eik⊥ ·X⊥ . 0. 0. X7 ) = (y. 2 2 8 (33) The total potential between the D0brane and D2brane is therefore V (y) = τ0 τ2 κ2 G7 (y) = −π(4π 2 α′ )2 G7 (y). X6 .
1) 1 2L Deﬁning u≡ p p2 + q 2 /g2 . i=1 (40) which vanishes if pi = qi = 0. qi )string is (13. (36) in agreement with (34).6. i (38) If we orient each string at the angle cos θi = pi 2 p2 + qi /g2 i . 12. qi /g). sin θi = qi /g 2 p2 + qi /g2 i . Q† Q† β β ˜ Qα = τ(p.3) τ(pi .6. Fi2 ) = 1 (cos θi 2πα′ 2 p2 + qi /g2 . The supersymmetry algebra for a static (p. (37) Let the three strings sit in the (X 1 . so that the ﬁrst term in the asymptotic expansion of the potential in 1/y is 1 V (r) ≈ − √ 2π 2 α′ = −π −1/2 ∞ 0 dt t3/2 exp − ty 2 2πα′ 5 (2πα′ )2 Γ( )y −5 2 2 ′ 2 = −π(4π α ) G7 (y). If the angle string i makes with the X 1 axis is θi .qi ) = 2 p2 + qi /g2 i 2πα′ . q) string extended in the X i direction is (13. (39) then the total force exerted on the junction point is 1 2πα′ 3 (pi . up to rotations and reﬂections.q) δαβ 1 0 (Γ0 Γi )αβ + 2πα′ 0 1 p q/g .4 Problem 13. q/g −p (41) √ √ 1+u 1−u . X 2 ) plane. sin θi i 2 p2 + qi /g2 ).12 The tension of the (pi . √ √ − 1−u 1+u (42) . 1 U≡√ 2 Qα ˜ . then the force it exerts on the junction point is (Fi1 .12 CHAPTER 13 99 (according to Mathematica). This is the unique stable conﬁguration.
(47) Our ﬁrst set of unbroken supersymmetries (45) becomes (I16 − uΓ0 Γ1 − √ √ ˜ 1 − u2 Γ0 Γ2 )( 1 + uQ + 1 − uQ) α . In this basis Γ0 Γ1 = 2S0 .12 CHAPTER 13 100 we can use U to diagonalize the matrix on the RHS of (41): 1 2Lτ(p. (43) The top row of this 2 × 2 matrix equation tells us that. yielding eight supersymmetries that leave this state invariant: √ √ ˜ (I16 − Γ0 Γi )( 1 + uQ + 1 − uQ) α . in a basis in spinor space in which Γ0 Γi is diagonal.1. after projecting onto the subspace annihilated by (I16 − Γ0 Γi ): √ √ ˜ (I16 + Γ0 Γi )(− 1 − uQ + 1 + uQ) α . (46) Now let us suppose that the string is aligned in the direction (39). q) strings that all obey (39) will leave these eight unbroken. Q† Qβ † U T β ˜ Qα = (I16 + Γ0 Γi )αβ 0 0 (I16 − Γ0 Γi )αβ .q) U Qα ˜ .10). while 0 1 0 −1 Γ0 Γ2 = ⊗ ⊗ I2 ⊗ I2 ⊗ I2 . (45) The other eight unbroken supersymmetries are given by the bottom row of (43). the supersymmetry generator √ √ ˜ 1 + uQα + 1 − uQα (44) annihilates this state if (I16 + Γ0 Γi )αα = 0. If the string is aligned in the direction (39). We can use (I16 − Γ0 Γi ) to project onto this eightdimensional subspace. which depends on p and q. (48) We work in a basis of eigenspinors of the operators Sa deﬁned in (B. then Γi = p p2 + q 2 /g2 Γ1 + q/g p2 + q 2 /g2 Γ2 = uΓ1 + 1 − u2 Γ 2 . We will show that eight of the sixteen unbroken supersymmetries do not depend on p or q. (49) −1 0 1 0 We can divide the sixteen values of the spinor index α into four groups of four according to the . and therefore any conﬁguration of (p.
√ and therefore unbroken no matter what the values of p and q? Indeed. there is a similar repetition of generators. by dividing (50) by 2 1 − u √ and (54) by 2 1 + u and adding them. √ √ ˜ 1 − u2 Γ0 Γ2 )( 1 + uQ + 1 − uQ) (++s2 s3 s4 ) √ √ ˜ = (1 + u)( 1 + uQ + 1 − uQ)(−−s2 s3 s4 ) √ √ ˜ + 1 − u2 ( 1 + uQ + 1 − uQ)(++s2 s3 s4 ) . (54). √ √ ˜ 1 − u2 Γ0 Γ2 )( 1 + uQ + 1 − uQ) (+−s2 s3 s4 ) √ √ ˜ = (1 − u)( 1 + uQ + 1 − uQ)(+−s2 s3 s4 ) √ √ ˜ − 1 − u2 ( 1 + uQ + 1 − uQ)(−+s2 s3 s4 ) . (50) (I16 − uΓ0 Γ1 − (51) (I16 − uΓ0 Γ1 − (52) (I16 − uΓ0 Γ1 − (53) ˜ (The indexing by s2 . and (52) and (53) similarly by a factor of − (1 − u)/(1 + u). since the chirality condition on both Q and Q implies the restriction 8s2 s3 s4 = 1 in the case of (50) and (51). given by (46). (54) (I16 + uΓ0 Γ1 + (55) Are there linear combinations of the generators (50).) It easy to see that (50) and (51) diﬀer only by a factor of (1 − u)/(1 + u). so (50) and (52) alone are suﬃcient to describe the eight independent supersymmetry generators in this sector. √ √ ˜ 1 − u2 Γ0 Γ2 )( 1 + uQ + 1 − uQ) (−+s2 s3 s4 ) √ √ ˜ = (1 + u)( 1 + uQ + 1 − uQ)(−+s2 s3 s4 ) √ √ ˜ − 1 − u2 ( 1 + uQ + 1 − uQ)(+−s2 s3 s4 ) . s4 is somewhat redundant. and the eight independent generators are (I16 + uΓ0 Γ1 + √ √ ˜ 1 − u2 Γ0 Γ2 )(− 1 − uQ + 1 + uQ) (++s2 s3 s4 ) √ √ ˜ = (1 + u)(− 1 − uQ + 1 + uQ)(++s2 s3 s4 ) √ √ ˜ − 1 − u2 (− 1 − uQ + 1 + uQ)(−−s2 s3 s4 ) . (56) . we come up with four such generators: ˜ Q(++s2 s3 s4 ) + Q(−−s2 s3 s4 ) . In the other sector. s3 .12 CHAPTER 13 101 eigenvalues of S0 and S1 : (I16 − uΓ0 Γ1 − √ √ ˜ 1 − u2 Γ0 Γ2 )( 1 + uQ + 1 − uQ) (++s2 s3 s4 ) √ √ ˜ = (1 − u)( 1 + uQ + 1 − uQ)(++s2 s3 s4 ) √ √ ˜ + 1 − u2 ( 1 + uQ + 1 − uQ)(−−s2 s3 s4 ) . √ √ ˜ 1 − u2 Γ0 Γ2 )(− 1 − uQ + 1 + uQ) (+−s2 s3 s4 ) √ √ ˜ = (1 + u)(− 1 − uQ + 1 + uQ)(+−s2 s3 s4 ) √ √ ˜ + 1 − u2 (− 1 − uQ + 1 + uQ)(−+s2 s3 s4 ) . (52). and 8s2 s3 s4 = −1 in the case of (52) and (53). and (55) that are independent of u.
.12 CHAPTER 13 102 √ √ Four more are found by dividing (52) by 2 1 − u and (55) by 1 + u: ˜ Q(+−s2 s3 s4 ) − Q(−+s2 s3 s4 ) . (57) As promised. one quarter of the original supersymmetries leave the entire conﬁguration described in the ﬁrst paragraph of this solution invariant.
) A Dstring stretched between the .8. . 5 and m. with additional terms for the excitation: 1 2L Qα ˜ . .17): e2Φ = g2 + 2π 2 (xm Q1 Q2 m )2 + 2π 2 (xm − xm )2 . Since the string excitations move at the speed of light. . . the ﬁrst term yields the usual preserved supersymmetries.1. − x1 2 Gµν = gηµν . n = 6. . The supersymmetry algebra (13. except that the ﬁrst term above is diﬀerent: 1 2L Qα ˜ . and momentum (per unit length) in the 1direction p1 . while rightmoving excitations have p0 = p1 .2 The supergravity solution for two static parallel NS5branes is given in (14. of the form Qα + ˜ (β ⊥ Q)α . Either way. . Q† Q† β β ˜ Qα = 1 2πα′ g δαβ (Γ0 Γ1 )αβ (Γ0 Γ1 )αβ δαβ + p0 δαβ + p1 (Γ0 Γ1 )αβ 1 0 . and for the Qs if the excitation is rightmoving. (We have altered (14. 0 1 (2) When diagonalized. the state is BPS.13 CHAPTER 14 103 13 13.15a) slightly 2 1 in order to make (3) Sdual to the Dbrane solution (14. 9 are the parallel and transverse directions respectively.15) and (14. leftmoving excitation have p0 = −p1 . (3) Gmn = g−1 e2Φ δmn .1.1 The excitation on the Fstring will carry some energy (per unit length) p0 .1 Chapter 14 Problem 14. For the Dstring the story is almost the same.2.6. .9) for this string is similar to (13. Q† Q† β β ˜α Q = 1 2πα′ 0 (δ + Γ0 Γ1 )αβ 0 (δ − Γ0 Γ1 )αβ + p0 δαβ + p1 (Γ0 Γ1 )αβ 1 0 . The second term thus also vanishes ˜ (making the state BPS) for the Qs if the excitation is leftmoving.1). when 1 + (Γ0 Γ1 )αα = 0 it is preserved if the excitation is rightmoving. 0 1 (1) The ﬁrst term on the RHS vanishes for those supersymmetries preserved by the unexcited Fstring. and the branes are located in the transverse space at xm and xm . When 1 − (Γ0 Γ1 )αα = 0 this supersymmetry is also preserved by the second term if the excitation is leftmoving.1. Hmnp = −ǫmnp q ∂q Φ. . ν = 0. ˜ namely Qs for which Γ0 Γ1 = −1 and Qs for which Γ0 Γ1 = 1.2 Problem 14. 13. where µ.1).
3.6 below. with this ansatz the Dstring action (13. which is slightly diﬀerent from the one used in volume I and in ˜ Problem 14. and it is this mass that transforms simply under Sduality. and it is only in string frame that the parallel metric Gµν is independent of the transverse position. In other words. after performing the integral over σ in the Dstring action. Spp = −m dτ −∂τ X µ ∂τ Xµ . (9) .13 CHAPTER 14 104 two branes at any given excitation level is a point particle with respect to the 5+1 dimensional Poincar´ symmetry of the parallel dimensions.7) of the Einstein frame. (5) where m is the mass of the solution X m (σ) with respect to the 5+1 dimensional Poincar´ symmetry. if we make an ansatz for the solution e of the form X µ = X µ (τ ). mE = g1/4 m. where GE = e−Φ/2 G. Comparison with (5) shows that m= g−1/2 2πα′ dσ ∂σ X m . (Here we are using the deﬁnition (14.14) factorizes: SD1 = − 1 2πα′ 1 =− 2πα′ dτ dσ e−Φ dτ dσ e−Φ × =− 1 2πα′ − Gµν ∂τ X µ ∂τ X ν (Gµn + Bµn )∂τ X µ ∂σ X n (Gmν + Bmν )∂σ X m ∂τ X ν Gmn ∂σ X m ∂σ X n ∂σ X m ∂σ Xm dτ −∂τ X µ ∂τ Xµ .2) in 5+1 dimensions. (7) where the integrand is the coordinate (not the proper) line element in this coordinate system. e Assuming that the gauge ﬁeld is not excited.2. X m = X m (σ). we should obtain the pointparticle action (1.) From the deﬁnition (5) of the mass. (6) − det(Gab + Bab ) dσ e−Φ In the last equality we have used the fact that neither the metric nor the twoform potential in the solution (3) have mixed µn components. We can nonetheless deﬁne an Einsteinframe mass mE with respect to Gµν at xm  = ∞. The ground state is therefore a straight line connecting the two branes: m= g−1/2 xm − xm  2 1 . (4) then. 2πα′ (8) As explained above. GE = e−Φ/2 G. this mass is deﬁned with respect the geometry of the parallel directions.1.
1) and do a calculation similar to the one above. since the NS5brane and the black 5brane solutions are related by Sduality.6 To ﬁnd the expectation values of the dilaton and graviton in the low energy ﬁeld theory. calculated in problem 13. the Sduality is manifest at every step. ˜ ˜ For the Dbrane. and in ﬂat spacetime (Gµν = ηµν ) its proper length and coordinate length are the same. ν parallel to the brane and zero otherwise. = Z[0] ˜ 2 (16) .13 CHAPTER 14 105 so (7) becomes g−1/4 dσ ∂σ X m . which is a real. (11) 2πα′ Its Einsteinframe mass is then g1/4 dσ ∂σ X m . Kh ] with respect to KΦ and Kh . 2 p JΦ (X) = ˜ (14) (15) ˜ Recall that h is the perturbation in the Einsteinframe metric. (We have put the brane at the origin. or we can consider the Fstring to be stretched between two elementary D5branes embedded in ﬂat spacetime. In the ﬁrst calculation. k = 0) 2 (2π)9−p k⊥ Φ 3−p 2 iκ τp d9−pX⊥ G9−p (X⊥ ) dp+1X KΦ (X). physical source for the ﬁelds. Using the propagator (23).3 Problem 14. ˜ Z[KΦ ] = −Z[0] ˜ d10k ˜ ˜˜ ˜˜ J ˜ (−k) ΦΦ(k) KΦ (k) (2π)10 Φ d9−pk⊥ 1 ˜ 3−p 2 iκ τp = Z[0] 2 K ˜ (k⊥ . (12) mE = 2πα′ which indeed agrees with (10) under g → 1/g. The second calculation yields the same answer. 4 1 µν Jh (X) = − τp eµν δ9−p (X⊥ ). so η ′ that X⊥ = 0.8.4(b) (see (26) and (27) of that solution): 3 − p 9−p τp δ (X⊥ ). we also include the sources JΦ and ˜ Jh . its total energy is 1 m= dσ ∂σ X m . as are the Dstring and Fstring actions. and is much easier: since the tension of the Fstring is 1/2πα′ . we can calculate the partition functions Z[KΦ ] and Z[Kh ] separately. we add to the action a source term µν ˜ (13) S ′ = d10X KΦ Φ + Kh hµν . mE = 13.) Since the dilaton decouples from the Einsteinframe graviton. G = η + h. (10) 2πα′ We can calculate the mass of an Fstring stretched between two D5branes in two diﬀerent pictures: we can use the black 5brane supergravity solution (14. ˜ and take functional derivatives of the partition function Z[KΦ . and that eµν equals µν for µ.
8.23). n transverse to the brane. (22) δmn . k = 0) 8 (2π)10 k⊥ h p+1 ηµν − eµν 2iκ2 τp 8 × so hµν (X) = = Hence for µ.3.13 CHAPTER 14 106 so that ˜ Φ(X) = 1 δZ[KΦ ] ˜ iZ[0] δKΦ (X) ˜ 3−p 2 κ τp G9−p (X⊥ ). = 2 (17) Using (13. (13. this becomes 3−p 7−p ˜ Φ(X) = (4π)(5−p)/2 Γ( )gα′(7−p)/2 r p−7 4 2 3 − p ρ7−p .22).1b) (corrected by a factor of g2 ).8. Z[Kh ] = −Z[0] = Z[0] = Z[0] d10X ˜µν ˜ ρσ J (−k) hµν hρσ (k) Kh (k) 10 h (2π) p+1 d9−pk⊥ 1 ˜ µν ηµν − eµν 2iκ2 τp 2 K (k⊥ . ˜ Gµν ≈ while for m. ν aligned along the brane. p+1 ηµν − eµν 8 p+1 ηµν − eµν 8 2κ2 τp G9−p (X⊥ ) ρ7−p .3. Using the propagator (24). (19) in agreement with (14. and the positionspace expression for Gd . ˜ Gmn ≈ 1+ ρ7−p r 7−p (p+1)/8 (20) d9−pX⊥ G9−p (X⊥ ) µν dp+1X Kh (X). Hence ˜ e2 Φ ≈ g2 (1 + 2 Φ ) ≈ g2 1 + (18) ρ7−p r 7−p (3−p)/2 .2b) with Q = 1. The graviton calculation is very similar. (23) . = 4 r 7−p where ρ7−p is as deﬁned in (14. r 7−p (21) 1+ ρ7−p r 7−p (p−7)/8 ηµν .
1).13 CHAPTER 14 ˜ 107 ˜ The string frame metric. (24) (25) 1/2 δmn . G = eΦ/2 G. is therefore Gµν ≈ Gmn ≈ in agreement with (14. 1+ 1+ ρ7−p r 7−p ρ7−p r 7−p −1/2 ηµν .8. .
7) should be as follows (we haven’t bothered to raise the index): 2h 2˜ ˜ ˜ ¯ ¯ ¯ ci{k. cijk = A′ (∞. 0) ˜ →∞ zm →1 n S2 . ∞)Oj (1.7. but glibly ignoring phase factors as Polchinski does. 0) ¯ S2 .7.14 CHAPTER 15 108 14 14. Ak are primary. with (3) 14. Aj . 1)Ak (0. in the above “primary” may be weakened to “quasiprimary” (meaning annihilated by L1 . we would like to claim that the correct form for (15. (4) The following relations then hold. 24h 10h 6h + 2c K3 = 2304. (1) (2) This matches the Kac formula.14). but Polchinski does not seem to ﬁnd the notion of quasiprimary operator interesting or useful.1 Chapter 15 Problem 15. derived from the deﬁnition (6. ∞)Aj (1.2 Problem 15. (8) We would also like to claim that the LHS of (15.k}.3 Let’s begin by recording some useful symmetry relations of the operator product coeﬃcient with lower indices. zn )Om (zm . Armed with these symmetries (and in particular relation (5)). 1)Om (z.mn = z lim zn n znhn L−{k} L−{k} On (zn . (5) (6) (7) cjik Actually.13) and (6. C) rather than general local conformal transformations). zm )Oi (0. and therefore transforming as a tensor under P SL(2.9) should read Ol′ (∞. 0) i S2 . z )On (0. with the sign of the coeﬃcient depending on the statistics of the operators: cijk = ±(−1)hj +hj ckji cijk = ±(−1) cijk = ±(−1) ˜ h k +h k ˜ ˜ ˜ ˜ hi +hj +hk +hi +hj +hk if Aj is primary cikj if Ai is primary if Ai . rather than Ln for all n > 0. (9) .2.2.1 The matrix of inner products is L3 1 M3 = h L1 L2 L3 L−2 L−1 L−3 h −1 L3 24h(h + 1)(2h + 1) 12h(3h + 1) 24h = 12h(3h + 1) h(8h + 8 + c) 10h .
(10) jl so the coeﬃcient of z −hm −hn +hi in Fmn (iz) is 1. L−1 · Oi . 2hi (12) .14 CHAPTER 15 109 We are now ready to solve the problem. i{} βmn = 1. We have 1 i{1} βmn = (hi + hm − hn ). For N = 1 there is again only one operator.2. (11) 2hi Thus the coeﬃcient of z −hm −hn +hi +1 is 1 (hi + hm − hn )(hi + hj − hl ).8). since by the deﬁnition (15. The case N = 0 is trivial.
1.24) (the deﬁnition of C): U ∗ CU −1 U Γµ U −1 U C −1 U T = U ∗ CΓµ C −1 U T = −U ∗ ΓµT U T = −(U Γµ U −1 )T . using the fact that Σµν transforms the same way as Γµ : U ∗ BU −1 U Σµν U −1 U B −1 U T = U ∗ BΣµν B −1 U T = −U ∗ Σµν∗ U T = −(U Σµν U −1 )∗ . −1 U ∗ B2 U −1 U Γµ U −1 (U ∗ B2 U −1 )−1 = U ∗ B2 Γµ B2 U T (2) = (−1)k+1 U ∗ Γµ∗ U T = (−1)k+1 (U Γµ U −1 )∗ .21): ∗ ∗ U B1 U T U ∗ B1 U −1 = U B1 B1 U −1 = (−1)k(k+1)/2 U U −1 = (−1)k(k+1)/2 .1 Appendix B Problem B.25): all three sides clearly transform by multiplying on the left by U and on the right by U −1 .1. (B. (B. B2 → U ∗ B2 U −1 . ∗ U B2 U T U ∗ B2 U −1 (5) (6) = ∗ U B2 B2 U −1 = (−1) k(k−1)/2 UU −1 = (−1) k(k−1)/2 .17) (the deﬁnition of B1 and B2 ): −1 U ∗ B1 U −1 U Γµ U −1 (U ∗ B1 U −1 )−1 = U ∗ B1 Γµ B1 U T = (−1)k U ∗ Γµ∗ U T = (−1)k (U Γµ U −1 )∗ .18).1.27): U ∗ BU −1 U Γ0 U −1 = U ∗ BΓ0 U −1 = U ∗ CU −1 .15 APPENDIX B 110 15 15. C → U ∗ CU −1 . (B. (8) . and C. (7) (B. (B.1. B2 .1.17). Γµ → U Γµ U −1 .19) is the same as that of (B.1 Under a change of spinor representation basis. B1 . (4) (3) The invariance of (B.1.1.1. all transform the same way: B1 → U ∗ B1 U −1 . (1) The invariance of the following equations under this change of basis is more or less trivial: (B.
1. since B T and C T transform the same way as B and C. and B2 is used when k = 0. 3 (mod 4). we have T B1 = (Γ3 Γ5 · · · Γd−1 )T = (−1)k(k+1)/2 B1 .16): B1 = Γ3 Γ5 · · · Γd−1 . where (for d = 2k + 2) B1 and B2 are deﬁned by equation (B.15 APPENDIX B 111 We will determine the relation between B and B T in the ζ (s) basis. On the other hand.1. we ﬁnd T C T = Γ0T B1 = (−1)k(k+1)/2+1 Γ0∗ B1 = (−1)k/2+1 B1 Γ0 = (−1)k/2+1 C. if k is odd.1. (12) When k is even. (11) Since B1 is used when k = 0. the fact that Γ is real and symmetric in this basis. and (B. we ﬁnd T B2 = (ΓB1 )T T = B1 ΓT = (−1)k(k+1)/2 B1 Γ = (−1)k(k+1)/2+k Γ∗ B1 = (−1)k(k−1)/2 B2 . then C = B2 Γ0 .19). and (B. .17). so in any dimension in which one can impose a Majorana condition we have B T = B. 1 (mod 4). (9) Since all of the Γ’s that enter into this product are antisymmetric in this basis (since they are Hermitian and imaginary). = (−1)k(k+1)/2 Γ3 Γ5 · · · Γd−1 (10) = (−1)k Γd−1 Γd−3 · · · Γ3 Using (10). using the fact that in this basis Γ0 is real and antisymmetric. and T C T = Γ0T B2 (13) = (−1)k(k−1)/2+1 Γ0∗ B2 = (−1)(k+1)/2 B2 Γ0 = (−1)(k+1)/2 C. C = B1 Γ0 . B2 = ΓB1 . so. (14) That all of these relations are invariant under change of basis follows directly from the transformation law (1).
L L =Z AB (19) (20) (21) (22) {QA . The supersymmetry algebra is (B. Q† } + 2{Q++ .) Hence.2. We must now decompose these Weyl spinors into MajoranaWeyl spinors. while the other. QB } = δAB (P 0 + P 1 ). transforms as a positive chirality Weyl spinor under both SO(1. Q++ } + {Q† .3) spinor representations under the subgroup SO(1. P 2 −P 3 . 1 (18) {Q++ . {Q1 . (17) R R −− −− 2 2i (The subscripts L and R signify that the respective supercharges have positive and negative SO(1. If these momenta are 0 (as in dimensional reduction in the strict sense).1a): {Q−− . Q2 = (Q−− − Q† ).2 Problem B. ζ ++ . Q† } .1)×SO(2) is described most simply in terms of Weyl representations: one positive chirality spinor. Let us therefore use the Weylspinor description of the 4 real supercharges of d = 4. Q† } −− = 2(P + iP ). but what do we do with the ﬁrst two terms? The d = 4 algebra has a U(1) Rsymmetry under which Q++ and Q−− are both multiplied by the same phase. P3 P2 (23) The central charges are thus the KaluzaKlein momenta associated with the reduced dimensions. QB } = δAB (P 0 − P 1 ). This symmetry and the original U(1) Rsymmetry of the d = 4 algebra can be recombined into two independent SO(2) Rsymmetry groups of the QA and QA pairs of supercharges. we must assume that Q2 = Q2 = {Q++ .1) and SO(2). Deﬁne 1 1 Q2 = (Q++ − Q† ). namely the SO(2) of rotations in the 23 plane. Q1 } = L L ++ ++ ++ 4 The last term is given by the algebra (14).44a)).3 The decomposition of SO(1. −− 2 3 † {Q++ .1) chirality. Q++ } = 2(P 0 − P 1 ). L R . We saw at the beginning that Q++ is positively charged and Q−− negatively charged under this symmetry. transforms as a negative chirality Weyl spinor under both (see (B. L L ++ ++ 2 2i 1 1 Q1 = (Q−− + Q† ). N = 1 supersymmetry. ++ −− The d = 2 algebra is then {QA . for instance.15 APPENDIX B 112 15. (15) {Q++ . Q−− } = 0. then the algebra possesses a further Rsymmetry. QB } L R where Z= {QA .1. ζ −− . Q† } = 2(P 0 + P 1 ). In order for the d = 2 algebra to inherit that symmetry. R R . (16) Q1 = (Q++ + Q† ).
2 2 2 2 2 2 2 2 2 2 Q(+ 1 . 0. 0). Since the operator B switches the sign of all the helicities s1 . 0. + 2 ).+ 1 . 0).+ 1 ) be the raising operators.6). ±1. 0. The 8′ states 1 1 1 1 have helicities ±(− 2 . not 8v + 8 (this is suggested by the second sentence of Section B. . and Q(+ 1 .1).15 APPENDIX B 113 15. (0. not 16. . it appears that some kind of fudge will be necessary to get this to work out correctly. There may be an error lurking in the book. .+ 1 . The supercharges with s0 = + 2 form an 8 representation of the SO(8) little group. and ±(+ 2 . Let Q(+ 1 . − 1 ). and (0. + 1 . ±1. + 1 .− 1 . 0) and acting on it with all possible combinations of these four operators. ±1). − 1 . . + 1 .+ 1 . + 1 ). The four states in the 8′ with s1 = − 1 are obtained by acting 2 with a single operator.5 Unfortunately. 0). 0. + 1 . Q† } = 2Pµ (Γµ Γ0 )αβ = −2k0 (1 + 2S0 )αβ . s4 . A 1 third operator gives s1 = + 2 .− 1 . 2 2 2 2 2 2 2 2 2 2 2 2 In a frame in which k0 = k1 . β (24) 1 1 so that supercharges with s0 = − 2 annihilate all states. We can obtain all sixteen of 2 2 2 2 2 2 2 2 2 2 the states in 8v + 8′ by starting with the state (−1. + 1 ). (3) The frame is one in which k0 = k1 . although it’s hard to see how this can ﬁt into the analysis of the type I spectrum in Chapter 10. The 8v states have helicities (±1. + 1 . not k0 = k1 as purportedly used in the book. Acting with a second operator yields the six states in the 8v with s1 = 0. Q(+ 1 .+ 1 . ±(+ 1 .3 Problem B. 0). 0. the other four states of the 8′ .+ 1 .− 1 .+ 1 . (+1. ±(+ 1 .+ 1 . 0. 0. − 1 . acting with all four operators yields the last state of the 8v . + 2 . 0.+ 1 ) . of SO(9. (2) The supercharges of the N = 1 theory are in the 16′ . Finally. (0.− 1 . the supersymmetry algebra is {Qα . 0. . We will arbitrarily choose fudge #1.− 1 ) . The candidate fudges are: (1) The vector multiplet is 8v + 8′ .− 1 ) . 0. the Majorana condition pairs these eight supercharges into four independent sets of fermionic raising and lowering operators.+ 1 .
Vol. . [2] J. Cambridge University Press. 1998. Polchinski. Polchinski. Vol. String Theory. 1: Introduction to the Bosonic String. 1998.REFERENCES 114 References [1] J. String Theory. 2: Superstring Theory and Beyond. Cambridge University Press.