Strings, Branes and - Froste
Strings, Branes and - Froste
Strings, Branes and - Froste
Extra Dimensions
Stefan Förste
Abstract
This review is devoted to strings and branes. Firstly, perturbative string theory is
introduced. The appearance of various types of branes is discussed. These include
orbifold fixed planes, D-branes and orientifold planes. The connection to BPS vacua
of supergravity is presented afterwards. As applications, we outline the role of branes
in string dualities, field theory dualities, the AdS/CFT correspondence and scenarios
where the string scale is at a TeV. Some issues of warped compactifications are also
addressed. These comprise corrections to gravitational interactions as well as the
cosmological constant problem.
Contents
1 Introduction 1
i
CONTENTS ii
4 Applications 138
4.1 String dualities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
4.2 Dualities in Field Theory . . . . . . . . . . . . . . . . . . . . . . . . . 141
4.3 AdS/CFT correspondence . . . . . . . . . . . . . . . . . . . . . . . . . 145
4.3.1 The conjecture . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
4.3.2 Wilson loop computation . . . . . . . . . . . . . . . . . . . . . 149
4.3.2.1 Classical approximation . . . . . . . . . . . . . . . . . 149
4.3.2.2 Stringy corrections . . . . . . . . . . . . . . . . . . . . 154
4.4 Strings at a TeV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
4.4.1 Corrections to Newton’s law . . . . . . . . . . . . . . . . . . . . 165
Introduction
One of the most outstanding problems of theoretical physics is to unify our picture of
electroweak and strong interactions with gravitational interactions. We would like to
view the attraction of masses as appearing due to the exchange of particles (gravitons)
between the masses. In conventional perturbative quantum field theory this is not
possible because the theory of gravity is not renormalizable. A promising candidate
providing a unified picture is string theory. In string theory, gravitons appear together
with the other particles as excitations of a string.
On the other hand, also from an observational point of view gravitational interac-
tions show some essential differences to the other interactions . Masses always attract
each other, and the strength of the gravitational interaction is much weaker than the
electroweak and strong interactions. A way how this difference could enter a theory
is provided by the concept of “branes”. The expression “brane” is derived from mem-
brane and stands for extended objects on which interactions are localized. Assuming
that gravity is the only interaction which is not localized on a brane, the special fea-
tures of gravity can be attributed to properties of the extra dimensions where only
gravity can propagate. (This can be either the size of the extra dimension or some
curvature.)
The brane picture is embedded in a natural way in string theory. Therefore, string
theory has the prospect to unify gravity with the strong and electroweak interactions
while, at the same time, explaining the difference between gravitational and the other
interactions.
This set of notes is organized as follows. In chapter 2, we briefly introduce the
concept of strings and show that quantized closed strings yield the graviton as a string
excitation. We argue that the quantized string lives in a ten dimensional target space.
It is shown that an effective field theory description of strings is given by (higher di-
mensional supersymmetric extensions of) the Einstein Hilbert theory. The concept
1
1. Introduction 2
of compactifying extra dimensions is introduced and special stringy features are em-
phasized. Thereafter, we introduce the orbifold fixed planes as higher dimensional
extended objects where closed string twisted sector excitations are localized. The
quantization of the open string will lead us to the concept of D-branes, branes on
which open string excitations live. We compute the tensions and charges of D-branes
and derive an effective field theory on the world volume of the D-brane. Finally,
perturbative string theory contains orientifold planes as extended objects. These are
branes on which excitations of unoriented closed strings can live. Compactifications
containing orientifold planes and D-branes are candidates for phenomenologically in-
teresting models. We demonstrate the techniques of orientifold compactifications at a
simple example.
In chapter 3, we identify some of the extended objects of chapter 2 as stable
solutions of the effective field theory descriptions of string theory. These will be the
fundamental string and the D-branes. In addition we will find another extended object,
the NS five brane, which cannot be described in perturbative string theory.
Chapter 4 discusses some applications of the properties of branes derived in the
previous chapters. One of the problems of perturbative string theory is that the string
concept does not lead to a unique theory. However, it has been conjectured that all
the consistent string theories are perturbative descriptions of one underlying theory
called M-theory. We discuss how branes fit into this picture. We also present branes
as tools for illustrating duality relations among field theories. Another application, we
are discussing is based on the twofold description of three dimensional D-branes. The
perturbative description leads to an effective conformal field theory (CFT) whereas the
corresponding stable solution to supergravity contains an AdS space geometry. This
observation results in the AdS/CFT correspondence. We present in some detail, how
the AdS/CFT correspondence can be employed to compute Wilson loops in strongly
coupled gauge theories. An application which is of phenomenological interest is the
fact that D-branes allow to construct models in which the string scale is of the order of
a TeV. If such models are realized in nature, they should be discovered experimentally
in the near future.
Chapter 5 is somewhat disconnected from the rest of these notes since it considers
brane models which are not directly constructed from strings. Postulating the existence
of branes on which certain interactions are localized, we present the construction of
models in which the space transverse to the brane is curved. We discuss how an
observer on a brane experiences gravitational interactions. We also make contact to the
AdS/CFT conjecture for a certain model. Also other questions of phenomenological
relevance are addressed. These are the hierarchy problem and the problem of the
cosmological constant. We show how these problems are modified in models containing
1. Introduction 3
branes.
Chapter 6 gives hints for further reading and provides the sources for the current
text.
Our intention is that this review should be self contained and be readable by people
who know some quantum field theory and general relativity. We hope that some people
will enjoy reading one or the other section.
Chapter 2
Perturbative description of
branes
Let us start with the simplest string – the bosonic string. The string moves along a
surface through space and time. This surface is called the worldsheet (in analogy to a
worldline of a point particle). For space and time in which the motion takes place we
will often use the term target space. Let d be the number of target space dimesnions.
The coordinates of the target space are X µ , and the worldsheet is a surface X µ (τ, σ),
where τ and σ are the time and space like variables parameterizing the worldsheet.
String theory is defined by the requirement that the classical motion of the string
should be such that its worldsheet has minimal area. Hence, we choose the action of
the string proportional to the worldsheet. The resulting action is called Nambu Goto
action. It reads
Z
1 √
S=− d2 σ −g. (2.1.1.1)
2πα′
The integral is taken over the parameter space of σ and τ . (We will also use the
notation τ = σ 0 , and σ = σ 1 .) The determinant of the induced metric is called g. The
induced metric depends on the shape of the worldsheet and the shape of the target
space,
4
2. Worldsheet Actions 5
4πα′ δS 1
Tαβ = − √ αβ
= ∂α X µ ∂β Xµ − γαβ γ δγ ∂δ X µ ∂γ Xµ , (2.1.1.4)
−γ δγ 2
where the target space index µ is raised and lowered with Gµν = ηµν . Thus, the γαβ
equation of motion, Tαβ = 0, equates γαβ with the induced metric (2.1.1.2), and the
actions (2.1.1.1) and (2.1.1.3) are at least classically equivalent. If we had just used
covariance as a guiding principle we would have written down a more general expression
for (2.1.1.3). We will do so later. At the moment, (2.1.1.3) with Gµν = ηµν describes
a string propagating in the trivial background. Upon quantization of this theory we
will see that the string produces a spectrum of target space fields. Switching on non
trivial vacua for those target space fields will modify (2.1.1.3). But before quantizing
the theory, we would like to discuss the symmetries and introduce supersymmetric
versions of (2.1.1.3).
First of all, (2.1.1.3) respects the target space symmetries encoded in Gµν . In
our case Gµν = ηµν this is nothing but d dimensional Poincaré invariance. From
the two dimensional point of view, this symmetry corresponds to field redefinitions
in (2.1.1.3). The action is also invariant under two dimensional coordinate changes
(reparametrizations). Further, it is Weyl invariant, i.e. it does not change under
It is this property which makes one dimensional objects special. The two dimensional
coordinate transformations together with the Weyl transformations are sufficient to
transform the worldsheet metric locally to the Minkowski metric,
σ − = τ − σ , and σ + = τ + σ. (2.1.1.7)
2. Worldsheet Actions 6
This invariance is connected to the fact that the trace of the energy momentum tensor
(2.1.1.4) vanishes identically, T+− = 02 . However, the other γαβ equations are not
identically satisfied and provide constraints, supplementing (2.1.1.8),
∂+ ∂− X µ = 0 (2.1.1.11)
where x+ and p+ denote the center of mass position and momentum of the string in
the + direction, respectively. The constraint equations (2.1.1.10) can now be used to
fix
1
X− = √ X0 − X1 (2.1.1.14)
2
In this section we are going to modify the previously discussed bosonic string by
enhancing its two dimensional symmetries. We will start from the gauge fixed ac-
tion (2.1.1.8) which had as residual symmetries two dimensional Poincaré invariance
and conformal coordinate transformations (2.1.1.9).4 A natural extension of Poincaré
invariance is supersymmetry. Therefore, we will study theories which are supersym-
metric from the two dimensional point of view. In order to construct a supersymmet-
ric extension of (2.1.1.8) one should first specify the symmetry group and then use
Noether’s method to build an invariant action. We will be brief and just present the
result,
Z
1 + − µ i µ i µ
S= dσ dσ ∂− X ∂+ Xµ + ψ+ ∂− ψ+µ + ψ− ∂+ ψ−µ , (2.1.1.15)
2πα′ 2 2
where ψ± are two dimensional Majorana-Weyl spinors. To see this, we first note that
!
1 + −
ψ−
iψ+ ∂− ψ+ + iψ− ∂+ ψ− = − (ψ+ , −ψ− ) ρ ∂+ + ρ ∂− , (2.1.1.16)
2 ψ+
where
ρ± = ρ0 ± ρ1 , (2.1.1.17)
with
! !
0 −i 0 i
ρ0 = and ρ1 = . (2.1.1.18)
i 0 i 0
It is easy to check that the above matrices form a two dimensional Clifford algebra,
n o
ρα , ρβ = −2η αβ . (2.1.1.19)
!
ψ−
Also, note that i (ψ+ , −ψ− ) is the Dirac conjugate of for real ψ± , i.e. of the
ψ+
!
ψ−
Majorana spinor . In addition to two dimensional Poincaré invariance and
ψ+
4
Alternatively, we could start from the action (2.1.1.3). This we would modify such that it becomes
locally supersymmetric. Finally, we would fix symmetries in the locally supersymmetric action.
2. Worldsheet Actions 8
µ µ
δX µ = ǭψ µ = iǫ+ ψ− − iǫ− ψ+ , (2.1.1.20)
δψ µ = −iρα ∂α X µ ǫ. (2.1.1.21)
µ
δψ− = −2ǫ+ ∂− X µ , (2.1.1.22)
µ
δψ+ = 2ǫ− ∂+ X µ . (2.1.1.23)
When checking the invariance of (2.1.1.15) under (2.1.1.20), (2.1.1.22), (2.1.1.23) one
should take into account that spinor components are anticommuting, e.g. ǫ+ ψ− =
−ψ− ǫ+ . Since the supersymmetry parameters ǫ± form a non chiral Majorana spinor,
the above symmetry is called (1, 1) supersymmetry. (In the end of this section we will
also discuss the chiral (1, 0) supersymmetry.) To summarize, the action (2.1.1.15) has
the following two dimensional global symmetries: Poincaré invariance and supersym-
metry. The corresponding Noether currents are the energy momentum tensor,
i µ
T++ = ∂+ X µ ∂+ Xµ + ψ+ ∂+ ψ+µ , (2.1.1.24)
2
i µ
T−− = ∂− X µ ∂− Xµ + ψ− ∂− ψ−µ , (2.1.1.25)
2
and the supercurrent
µ
J+ = ψ+ ∂+ Xµ , (2.1.1.26)
µ
J− = ψ− ∂− Xµ . (2.1.1.27)
The vanishing of the trace of the energy momentum tensor T+− ≡ 0 is again a con-
sequence of the invariance under the (local) conformal coordinate transformations
(2.1.1.9). The supercurrent is a spin– 32 object and naively one would expect to get
four independent components. That there are only two non-vanishing components is
a consequence of the fact that the supersymmetries (2.1.1.20), (2.1.1.22), (2.1.1.23)
leave the action invariant also when we allow instead of constant ǫ± for
ǫ− = ǫ− σ + and ǫ+ = ǫ+ σ − , (2.1.1.28)
i.e. they are “partially” local symmetries. Once again, the vanishing of the energy
momentum tensor is an additional constraint on the system. We did not derive this
5
− 1
Under the transformation (2.1.1.9) the spinor components transform as ψ± → σ̃ ±′ 2
ψ± .
2. Worldsheet Actions 9
explicitly here. But it can be easily inferred as follows. In two dimensions the Einstein
tensor vanishes identically. Thus, if we were to couple to two dimensional (Einstein)
gravity, the constraint Tαβ = 0 would correspond to the Einstein equation. Simi-
larly, the supercurrents (2.1.1.26), (2.1.1.27) are constrained to vanish. (If the theory
was coupled to two dimensional supergravity, this would correspond to the gravitino
equations of motion.)
As in the bosonic case we can employ the symmetry (2.1.1.9) to fix
X + = x+ + p+ τ. (2.1.1.29)
(We have written here the target space (light cone) index as µ = + in order to avoid
confusion with the worldsheet spinor indices.) Note, that the gauge fixing condi-
tion (2.1.1.30) is compatible with (2.1.1.29) and the supersymmetry transformations
(2.1.1.20), (2.1.1.21), as (2.1.1.30) implies the supersymmetry transformation
δX + = 0. (2.1.1.31)
The constraints (2.1.1.24), (2.1.1.25), (2.1.1.26), (2.1.1.27) can be solved for X − , and
ψαµ=− (here, α denotes the worldsheet spinor index). Therefore, after fixing the local
symmetries completely we are left with d − 2 free bosons and d − 2 free fermions (from
a two dimensional point of view).
We should note that in the closed string case (periodic boundary conditions in
bosonic directions) we have two choices for boundary conditions on the worldsheet
fermions. Boundary terms appearing in the variation of the action vanish for either
periodic or anti periodic boundary conditions on worldsheet fermions. (Later, we will
call the solutions with antiperiodic fermions Neveu Schwarz (NS) sector and the ones
with periodic boundary conditions Ramond (R) sector.
Going back to (2.1.1.15), we note that alternatively we could have written down
µ
a (1, 0) supersymmetric action by setting the left handed fermions ψ+ = 0. The
supersymmetries are now given by (2.1.1.20) and (2.1.1.22), only. The parameter ǫ−
does not occur anymore, and hence we have reduced the number of supersymmetries by
one half. More generally one can add left handed fermions λA
+ which do not transform
under supersymmetries. Therefore, they do not need to be in the same representation
of the target space Lorentz group as the X µ (therefore the index A instead of µ).
2. Worldsheet Actions 10
this will turn out to be the worldsheet action of the heterotic string. The energy mo-
µ
mentum tensor is as given in (2.1.1.24), (2.1.1.25) with λA
+ replacing ψ+ in (2.1.1.25).
There is only one conserved supercurrent (2.1.1.26).
Finally, we should remark that there are also extended versions of two dimensional
supersymmetry (see for example [456]). We will not be dealing with those in this
review.
In the above we have extended the bosonic string (2.1.1.3) to a superstring from the two
dimensional perspective. We called this worldsheet supersymmetry. Another direction
would be to extend (2.1.1.3) such that the target space Poincaré invariance is enhanced
to target space supersymmetry. This concept leads to the Green Schwarz string. Space
time supersymmetry means that the bosonic coordinates X µ get fermionic partners θ A
(where A labels the number of supersymmetries N ) such that the targetspace becomes
a superspace. In addition to Lorentz symmetry, the supersymmetric extension mixes
fermionic and bosonic coordinates,
δθ A = ǫA , (2.1.1.33)
δθ̄ = ǭA , (2.1.1.34)
µ µ A
δX = iǭΓ θ , (2.1.1.35)
where the global transformation parameter ǫA is a target space spinor and Γµ denotes
a target space Dirac matrix. In order to construct a string action respecting the
symmetries (2.1.1.33) – (2.1.1.35) one tries to replace ∂α X µ by the supersymmetric
combination
This leads to the following proposal for a space time supersymmetric string action
Z
1 √
S1 = − ′
d2 σ −γγ αβ Πµα Πβµ . (2.1.1.37)
4πα
Note that in contrast to the previously discussed worldsheet supersymmetric string,
(2.1.1.37) consists only of bosons when looked at from a two dimensional point of view.
The action (2.1.1.37) is invariant under global target space supersymmetry, i.e. Lorentz
2. Worldsheet Actions 11
simultaneously. So, we need an additional local symmetry whose gauge fixing will
remove half of the fermions θ A . The symmetry we are looking for is known as κ
symmetry. It exists only in special circumstances. First of all, the number of super-
symmetries should not exceed N = 2 (i.e. A = 1, 2). Then, adding a further term
Z n
1 2 αβ µ 1 1 2 2
S2 = d σ −iǫ ∂ α X θ̄ Γ µ ∂β θ − θ̄ Γ µ ∂β θ
2πα′
o
+ǫαβ θ̄ 1 Γµ ∂α θ 1 θ̄ 2 Γµ ∂β θ 2 (2.1.1.38)
to (2.1.1.37) results in a κ symmetric action. (We will give the explicit transformations
below.) In (2.1.1.38) ǫαβ denotes the two dimensional Levi Civita symbol. If one is
interested in less than N = 2 one can just put the corresponding θ A to zero. The
requirement that adding S2 to the action does not spoil supersymmetry (2.1.1.33) –
(2.1.1.35), leads to further constraints,
(i) d = 3 and θ is Majorana
(ii) d = 4 and θ is Majorana or Weyl
(iii) d = 6 and θ is Weyl
(iv) d = 10 and θ is Majorana-Weyl.
It remains to give the above mentioned κ symmetry transformations explicitly.
By adding S1 and S2 one observes that the kinetic terms for the θ’s (terms with one
derivative acting on a fermion) contain the following projection operators
αβ 1 αβ ǫαβ
P± = γ ±√ . (2.1.1.39)
2 −γ
6
Since the field equations are different for (2.1.1.37) the details of the discussion in the bosonic case
will change. The above frame just gives a rough motivation for a modification of (2.1.1.37) carried
out below.
2. Worldsheet Actions 12
where the worldsheet indices α, β are raised and lowered with respect to the worldsheet
metric γαβ . Now, we are ready to write down the κ transformations,
For a proof that these transformations leave S1 + S2 indeed invariant we refer to[222]
for example.
Once we have established that the number of local symmetries is correct, we can
now proceed to employ those symmetries and reduce the number of degrees of freedom
by gauge fixing. We will go to the light cone gauge in the following. Here, we will
discuss only the most interesting case of d = 10. As usual we use reparametrization
and Weyl invariance to fix γαβ = ηαβ . We can fix κ symmetry (2.1.1.43)–(2.1.1.45) by
the choice
Γ+ θ 1 = Γ+ θ 2 = 0, (2.1.1.46)
where
1
Γ± = √ Γ0 ± Γ9 . (2.1.1.47)
2
This sets half of the components of θ to zero. With the κ fixing condition (2.1.1.46)
the equations of motion for X + and X i (i = 2, . . . , d − 1) turn out to be free field
equations (cf (2.1.1.11)). The reason for this can be easily seen as follows. After
imposing (2.1.1.46), out of the fermionic terms only those containing θ¯A Γ− θ A remain
in the action S1 + S2 . Especially, the terms fourth order in θ A have gone. The above
mentioned terms with Γ− couple to ∂α X + , and hence they will only have influence
on the X − equation (obtained by taking the variation of the action with respect to
X + ). Thus we can again fix the conformal coordinate transformations by the choice
(2.1.1.13). The X − direction is then fixed (up to a constant) by imposing the constraint
of vanishing energy momentum tensor. Since the coupling of bosons and fermions is
2. Quantization of the fundamental string 13
Essentially, we have here two different cases: we take the same SO(8) representation
for both θ’s or we take them mutually different. The first option results in type IIB
theory whereas the second one leads to type IIA.
So, the gauge fixing procedure simplifies the theory substantially. The equations of
motion for the remaining degrees of freedom are just free field equations. For example
for the type IIB theory they read,
∂+ ∂− X i = 0, (2.1.1.50)
∂+ S 1a = 0, (2.1.1.51)
2a
∂− S = 0. (2.1.1.52)
They look almost equivalent to the equations of motion one obtains from the world-
sheet supersymmetric action (2.1.1.15) after eliminating the ± directions by the light
cone gauge. Especially, (2.1.1.51) and (2.1.1.52) have the form of two dimensional
Dirac equations where S 1 and S 2 appear as 2d Majorana-Weyl spinors. An important
difference is however, that in (2.1.1.15) all worldsheet fields transform in the vector
representation of the target space subgroup SO(d − 2).
In the rest of this chapter we will focus only on the worldsheet supersymmetric
formulation. There, target space fermions will appear in the Hilbert space when quan-
tizing the theory. We will come back to the Green Schwarz string only when discussing
type IIB strings living in a non-trivial target space (AdS5 × S 5 ) in section 4.3.
∂+ ∂− X i = 0. (2.1.2.1)
with
µ 1 µ 1 µ − i X 1 µ −2inσ−
XR = x + p σ + α e , (2.1.2.3)
2 2 2 n n
n6=0
1 µ 1 µ + i X 1 µ −2inσ+
XLµ = x + p σ + α̃ e . (2.1.2.4)
2 2 2 n n
n6=0
Here, all σ α dependence is written out explicitly, i.e. xµ , pµ , αµn , and α̃µn are σ α
independent operators. Classically, one can associate xµ with the center of mass
position, pµ with the center of mass momentum and αµn (α̃µn ) with the amplitude
of the n’th right moving (left moving) vibration mode of the string in xµ direction.
Reality of X µ imposes the relations
µ µ
ᵆ µ†
n = α−n and α̃n = α̃−n . (2.1.2.5)
and
h i
Ẋ µ (σ) , X ν σ ′ = −iπδ σ − σ ′ η µν (2.1.2.8)
With this we can translate the canonical commutators (2.1.2.7) and (2.1.2.8) into
commutators of the Fourier coefficients appearing in (2.1.2.3) and (2.1.2.4),
where δn+k is shorthand for δn+k,0 . So far, we did not take into account the constraints
of vanishing energy momentum tensor (2.1.1.10). To do so we go again to the light
cone gauge (2.1.1.13), i.e. set
α+ +
n = α̃n = 0 for n 6= 0. (2.1.2.13)
where a sum over repeated indices i from 2 to d − 1 is understood. The colon denotes
normal ordering to be specified below. We have parameterized the ordering ambiguity
by a constant a. (In principle one could have introduced two constants a, ã. But this
would lead to inconsistencies which we will not discuss here.) Equations (2.1.2.14) and
(2.1.2.15) are not to be read as operator identities but rather as conditions on physical
states which we will construct now. We choose the vacuum as an eigenstate of the pµ
with kµ being an ordinary number. Further, we impose that the vacuum is annihilated
by half of the vibration modes,
The rest of the states can now be constructed by acting with a certain number of
αi−n and α̃i−n (n > 0) on the vacuum. But we still need to impose the constraint
(2.1.1.10). Coming back to (2.1.2.14) and (2.1.2.15) we can now specify what is meant
by the normal ordering. The αik (α̃ik ) with the greater Fourier index k is written to
the right9 . For n 6= 0 (2.1.2.14) and (2.1.2.15) just tell us how any α− −
n or α̃n can be
9
E.g. for k > 0 this implies that : αik αi−k := αi−k αik , i.e. the annihilation operator acts first on a
state.
2. Quantization of the fundamental string 16
expressed in terms of the αik and α̃il . The nontrivial information is contained in the
n = 0 case. It is convenient to rewrite (2.1.2.14) and (2.1.2.15) for n = 0,
The N (Ñ ) are number operators in the sense that they count the number of creation
operators αi−n (α̃i−n ) acting on the vacuum. To be precise, the N (Ñ ) eigenvalue of
a state is this number multiplied by the index n and summed over all different kinds
of creation operators acting on the vacuum (for left and right movers separately).
Interpreting the pµ eigenvalue kµ as the momentum of a particle (2.1.2.18) looks like
a mass shell condition with the mass squared M 2 given by
The second equality in the above equation relates the allowed right moving creation
operators acting on the vacuum to the left moving ones. It is known as the level
matching condition.
For example, the first excited state is
In the following we are going to compute the normal ordering constant a. Requiring
agreement with (2.1.2.23) will give a condition on the dimension of the targetspace
2. Quantization of the fundamental string 17
to be 26. The following calculation may look at some points a bit dodgy when it
comes to computing the exact value of a. So, before starting we should note that the
compelling result will be that a depends on the targetspace dimension. The exact
numerics can be verified by other methods which we will not elaborate on here for
the sake of briefness. We will consider only N since the calculation with Ñ is a very
straightforward modification (just put tildes everywhere). The initial assumption is
that naturally the ordering in quantum expressions would be symmetric, i.e.
∞
1 X
N −a= αi−n αin . (2.1.2.24)
2
n=−∞,n6=0
By comparison with the definition of N (2.1.2.19) and using the commutation relations
(2.1.2.11) we find
∞
d−2X
a=− n. (2.1.2.25)
2
n=1
The above representation is valid for the real part of s being greater than one. The
zeta function, however, can be defined also for complex s with negative real part. This
is done by analytic continuation. The way to make sense out of (2.1.2.25) is now to
replace the infinite sum by the zeta function
d−2 d−2
a=− ζ (−1) = . (2.1.2.27)
2 24
Comparing with (2.1.2.23) we see that we need to take
d = 26 (2.1.2.28)
in order to preserve Lorentz invariance. This result can also be verified in a more rigid
way. Within the present approach one can check that a = 1 and d = 26 are needed
for the target space Lorentz algebra to close. In other approaches, one sees that the
Weyl symmetry becomes anomalous for d 6= 26.
Since N and Ñ are natural numbers we deduce from (2.1.2.21) that the mass
spectrum is an infinite tower starting from M 2 = −8 = −4/α′ and going up in steps of
8 = 4/α′ . The presence of a tachyon (a state with negative mass square) is a problem.
2. Quantization of the fundamental string 18
Spin/~
2
GRAVITON
-4 0 4 8
M 2 α′
It shows that we have looked at the theory in an unstable vacuum. One possibility
that this is not complete nonsense could be that apart from the massterm the tachyon
potential receives higher order corrections (like e.g. a power of four term) with the
opposite sign. Then it would look rather like a Higgs field than a tachyon, and one
would expect some phase transition (tachyon condensation) to occur such that the
final theory is stable. For the moment, however, let us ignore this problem (it will not
occur in the supersymmetric theories to be studied next).
The massless particles are described by (2.1.2.22). The part symmetric in i, j and
traceless corresponds to a targetspace graviton. This is one of the most important
results in string theory. There is a graviton in the spectrum and hence string theory
can give meaning to the concept of quantum gravity. (Since Einstein gravity cannot
be quantized in a straightforward fashion there is a graviton only classically. This
corresponds to the gravitational wave solution of the Einstein equations. The particle
aspect of the graviton is missing without string theory.) The trace-part of (2.1.2.22) is
called dilaton whereas the piece antisymmetric in i, j is simply the antisymmetric ten-
sor field (commonly denoted with B). A schematic summary of the particle spectrum
of the closed bosonic string is drawn in figure 2.1.
As a consistency check one may observe that the massive excitations fit in SO(25)
representations, i.e. they form massive representations of the little group of the Lorentz
group in 26 dimensions.
As we have already mentioned, this theory contains a graviton, which is good since
it gives the prospect of quantizing gravity. On the other hand, there is the tachyon, at
best telling us that we are in the wrong vacuum. (There could be no stable vacuum
at all – for example if the tachyon had a run away potential.) Further, there are no
2. Quantization of the fundamental string 19
target space fermions in the spectrum. So, we would like to keep the graviton but to
get rid of the tachyon and add fermions. We will see that this goal can be achieved
by quantizing the supersymmetric theories.
In this section we are going to quantize the (1,1) worldsheet supersymmetric string.
We will follow the lines of the previous section but need to add some new ingredients.
We start with the action (2.1.1.15). The equations of motion for the bosons X µ are
identical to the bosonic string. So, the mode expansion of the X µ is not altered and
given by (2.1.2.3) and (2.1.2.4). The equations of motion for the fermions are,
µ
∂− ψ+ = 0, (2.1.2.29)
µ
∂+ ψ− = 0. (2.1.2.30)
Further, we need to discuss boundary conditions for the worldsheet fermions. Modulo
the equations of motion (2.1.2.29) and (2.1.2.30) the variation of the action (2.1.1.15)
with respect to the worldsheet fermions turns out to be10
i µ µ π
−ψ+µ δψ+ + ψ−µ δψ− σ=0
. (2.1.2.31)
2π
µ
For the closed string we need to take the variation of ψ+ independent from the one of
µ
ψ− at the boundary (because we do not want the boundary condition to break part
of the supersymmetry (2.1.1.22) and (2.1.1.23)). Hence, the spinor components can
be either periodic or anti-periodic under shifts of σ by π. The first option gives the
Ramond (R) sector. In the R sector the general solution to (2.1.2.29) and (2.1.2.30)
can be written in terms of the following mode expansion
µ
X
ψ− = dµn e−2in(τ −σ) , (2.1.2.32)
n∈Z
µ
X
ψ+ = d˜µn e−2in(τ +σ) . (2.1.2.33)
n∈Z
The other option to solve the boundary condition is to take anti-periodic boundary
conditions. This is called the Neveu Schwarz (NS) sector. In the NS sector the general
solution to the equations of motion (2.1.2.29) and (2.1.2.30) reads11
µ
X
ψ− = bµr e−2ir(τ −σ) , (2.1.2.34)
r∈Z+ 12
µ
X
ψ+ = b̃µr e−2ir(τ +σ) , (2.1.2.35)
r∈Z+ 12
10
Again we put α′ = 21 .
11
The reality (Majorana) condition on the worldsheet spinor components provides relations analo-
gous to (2.1.2.5).
2. Quantization of the fundamental string 20
in the R sectors. Like the bosonic Fourier modes these can be split into creation
operators with negative Fourier index, and annihilation operators with positive Fourier
index. What about zero Fourier index? For the NS sector fermions this does not occur.
The vacuum is always taken to be an eigenstate of the bosonic zero modes where the
eigenvalues are the target space momentum of the state. (This is exactly like in the
bosonic string discussed in the previous section.) The Ramond sector zero modes
form a target space Clifford algebra (cf (2.1.2.39)). This means that the Ramond
sector states form a representation of the d dimensional Clifford algebra, i.e. they are
target space spinors. We will come back to this later. Pairing left and right movers,
there are altogether four different sectors to be discussed: NSNS, NSR, RNS, RR.
In the NSNS sector for example the left and right moving worldsheet fermions have
both anti-periodic boundary conditions. The vacuum in the NSNS sector is defined
via (2.1.2.16), (2.1.2.17) and
We can build states out of this by acting with bosonic left and right moving creation
operators on it. Further, left and right moving fermionic creators from the NS sectors
can act on (2.1.2.40). We should also impose the constraints (2.1.1.24) – (2.1.1.27) on
those states. As before, we do so by going to the light cone gauge
α+ + + +
n = α̃n = br = b̃r = 0. (2.1.2.41)
12
We say NS sectors and not NS sector because there are two of them: a left and a right moving
one.
2. Quantization of the fundamental string 21
Then the constraints can be solved to eliminate the minus directions. The important
information is again in the zero mode of the minus direction. This reads (2.1.2.18)
The expressions for the number operators are modified due to the presence of (NS
sector) worldsheet fermions
∞
X ∞
X
NN S = αi−n αin + rbi−r bir , (2.1.2.43)
n=1 r= 21
and the analogous expression for ÑN S . Its action on states is like in the bosonic
case (see discussion below (2.1.2.20)) taking into account the appearance of fermionic
creation operators. Again, we have put a so far undetermined normal ordering constant
in (2.1.2.42) and taken normal ordered expressions for the number operators. Now,
the first excited state is
Its target space tensor structure is identical to the one of (2.1.2.22). In particular it
forms massless representations of the target space Lorentz symmetry. Thus, Lorentz
covariance implies that
1
aN S = (2.1.2.45)
2
should hold.
We compute now aN S by first naturally assuming that a symmetrized expression
appears on the rhs of (2.1.2.42). This gives (see also (2.1.2.25))
∞ ∞
d−2 X d−2 X
aN S = − n+ r. (2.1.2.46)
2 2
n=1 1
r= 2
We use again the zeta function regularization to make sense out of (2.1.2.46). For the
second sum the following formula proves useful
∞
X 1
(n + c) = ζ (−1, c) = − 6c2 − 6c + 1 . (2.1.2.47)
n=0
12
(Note, that splitting the lhs of (2.1.2.47) into ζ (−1)+c+cζ (0) gives a different (wrong)
result. This is because we understand the infinite sum as an analytic continuation of
P
a finite one: (n + a)−s with real part of s greater than one. For generic s the above
2. Quantization of the fundamental string 22
d−2
aN S = . (2.1.2.48)
16
We conclude that the critical dimension for the (1, 1) worldsheet supersymmetric string
is
d = 10. (2.1.2.49)
Like in the bosonic string there are more rigid calculations giving the same result.
The massless spectrum from the NSNS sector is identical to the massless spectrum
of the closed bosonic string. Again, we have a tachyon: the NSNS groundstate. Here,
however this can be consistently projected out. This is done by imposing the GSO
(Gliozzi-Scherk-Olive) projection. To specify what this projection does in the NS sector
we introduce fermion number operators F (F̃ ) counting the number of worldsheet
fermionic NS right (left) handed creation operators acting on the vacuum. In addition,
we assign to the right (left) handed NS vacuum an F (F̃ ) eigenvalue of one13 . Now, the
GSO projection is carried out by multiplying states with the GSO projection operator
1 + (−1)F 1 + (−1)F̃
PGSO = . (2.1.2.50)
2 2
Obviously this does not change the first excited state (2.1.2.44) but removes the tachy-
onic NSNS ground state. There are several reasons why this projection is consistent.
At tree level14 for example one may check that the particles which have been projected
out do not reappear as poles in scattering amplitudes. Imposing the GSO projection
becomes even more natural when looking at the one loop level. In the Euclidean
version this means that the worldsheet is a torus. Summing over all possible spin
structures (the periodicities of worldsheet fermions when going around the two cycles
of the torus) leads naturally to the appearance of (2.1.2.50) in the string partition
function [415] (see also [331]). The NSNS spectrum subject to the GSO projection
looks as follows. The number operator (2.1.2.43) is quantized in half-integer steps.
The GSO projection removes half of the states, the groundstate, the first massive
states, the third massive states and so on. The NSNS spectrum of the type II strings
is summarized in figure 2.2.
We have achieved our goal of removing the tachyon from the spectrum while keep-
ing the graviton. We also want to have target space spinors. We will see that those
13 P
This means that we can write F = 1 + r>0 bi−r bir , and an analogous expression for F̃ .
14
The worldsheet has the topology of a cylinder, or a sphere when Wick rotated to the Euclidean
2d signature.
2. Quantization of the fundamental string 23
Spin/~
2
GRAVITON
-2 0 2 4
M 2 α′
Figure 2.2: NS-NS spectrum of the type II string. In comparison to figure 2.1 the
horizontal axis has been stretched by a factor of two.
come by including the R sector into the discussion. The most important issue to be
addressed here is the action of the zero modes on the R groundstate. By going to the
light-cone gauge, we can again eliminate the plus and minus (or the 0 and 1) direc-
tions leaving us with eight15 zero modes for the left and right moving sectors each.
We rearrange these modes into four complex modes
In particular, the DI and DI† are nilpotent. We can now construct the right moving
R vacuum by starting with a state which is annihilated by all the DI ,16
Acting with a DI† on the vacuum changes the Ith minus into a plus, e.g.
Acting once more with D3† will give zero. Acting with D3 on (2.1.2.57) will give back
(2.1.2.56) because of (2.1.2.55). Thus, we have a 24 = 16-fold degenerate vacuum.
This gives an on shell Majorana spinor in ten dimensions. For the left movers the
construction is analogous. (The above method to construct the state is actually an
option to construct (massless) spinor representations when the di0 are identified with
the target space Gamma matrices.) Without further motivation (which is given in the
books and reviews listed in section 6) we state how the GSO projection is performed
in the R sector. First, we define
P
di−n din
(−1)F = 24 d20 d30 d40 d50 d60 d70 d80 d90 (−1) n>0 , (2.1.2.58)
where the factor of 24 has been introduced such that (−)2F = 1, ensuring that
√
(2.1.2.59) defines projection operators. Note also that Γµ = 2dµ0 satisfies the canon-
ically normalized Clifford algebra {Γµ , Γν } = 2η µν . For the groundstate this is just
the chirality operator (the product of all Gamma matrices) in the transverse eight
dimensional space. Now, we multiply the R states by one of the following projection
operators
± 1 ± (−1)F
PGSO = (2.1.2.59)
2
We perform the analogous construction in the left moving R sector. There are essen-
tially two inequivalent options: we take the same sign in (2.1.2.59) for left and right
movers, or different signs. Taking different signs leads to type IIA strings whereas the
option with the same signs is called type IIB. Multiplying the R groundstate with one
of the operators (2.1.2.59) reduces the 16 dimensional Majorana spinor to an eight
dimensional Weyl spinor17 .
To complete the discussion of the R sector we have to combine left and right movers,
i.e. to construct the NSR, RNS, and RR sector of the theory. Let us start with the
NSR sector. The mass shell condition (2.1.2.42) reads now
1
2p+ p− − pi pi = 8 NN S − = 8ÑR , (2.1.2.60)
2
and the analogous expression for the left movers. We have put the normal ordering
constant in the Ramond sector to zero. This can easily be justified by replacing the
17
The two different choices in (2.1.2.59) give either the 8s or the 8c representation of SO(8) men-
tioned in section 2.1.1.3
2. Quantization of the fundamental string 25
half integer modded sum over r by an integer modded one in (2.1.2.46). Level matching
implies that the lowest allowed state in the NSR sector is massless and given by
where ua denotes the eight component Majorana-Weyl spinor comming from the R
ground states surviving the GSO projection. The 64 states contained in (2.1.2.62)
decompose into an eight dimensional and a 56 dimensional representation of the target
space little group SO(8). The 56 dimensional representation gives a gravitino of fixed
chirality, whereas the eight dimensional one gives a dilatino of fixed chirality.
The discussion of the RNS sector goes along the same line giving again a gravitino
and a dilatino either of opposite (to the NSR sector) chiralities corresponding to type
IIA theory, or of the same chiralities when the type IIB GSO projection is imposed.
Finally, in the RR sector the lowest state is obtained by combining the left with
the right moving vacuum. This state is massless due to the normal ordering constant
aR = 0. It has 64 components. The irreducible decompositions of the RR state
depend on whether we have imposed GSO conditions corresponding to type IIA or
type IIB. In the type IIA case the 64 states decompose into an eight dimensional
vector representation and a 56 dimensional representation. Thus in the type IIA
theory, the RR sector gives a massless U (1) one-form gauge potential Aµ and a three-
form gauge potential Cµνρ . In the type IIB theory the 64 splits into a singlet, a 28 and
a 35 dimensional representation of SO(8). This corresponds to a “zero-form” Φ′ , a
′ , and a four-form gauge potential with selfdual field strength C ∗
two-form Bµν µνρσ . The
particle content of the type II theories can be arranged in to N = 2 supermultiplets of
chiral (type IIB) or non-chiral (type IIA) ten dimensional supergravity. The (target
space bosons of the) massless spectrum of the type II strings is summarized in table
2.1.
Since the heterotic string is a bit out of the focus of the present review we will briefly
state the results. The starting point is the action (2.1.1.32). Without the λA + this
looks like the type II theories with the left handed worldsheet fermions removed.
Indeed, this part of the theory leads to the spectrum of the type II theories with
only the NS and R sector. The massless spectrum corresponds to N = 1 chiral
supergravity in ten dimensions. It corresponds to the states (the α̃in are the Fourier
coefficients for the left moving bosons, and the bir for the right moving fermions in the
NS sector)
in the R sector, where we denoted again the GSO projected R vacuum with uα . The
above states must be massless since they form irreducible representations of SO(d−2).
Focusing on the right moving sector we can deduce that the right moving normal
ordering constant must be 21 like in the type II case. Hence, the number of dimensions
(range of µ) is ten. As it stands the above spectrum leads to an anomalous theory.
But there is still the option of switching on the λA
+ . Let us first deduce the number
of additional directions (labeled by A) needed. In the sector where the vacuum is
non degenerate due to the presence of the λA
+ , we know that we need the left moving
normal ordering constant to be one. (Otherwise the states (2.1.2.63) would not be
massless, but still form SO(d − 2) representations.) The vacuum does not receive
further degeneracy in the sector where all of the λ+A have anti-periodic boundary
conditions. In this sector the normal ordering constant is (see also (2.1.2.46)), the
label A stands for anti-periodic
d−2 D
ãA = + , (2.1.2.65)
24 48
where we have called the number of additional directions D (A = 1, . . . , D). The
consistency condition ãA = 1 tells us that there must be 32 additional directions,
D = 32. (2.1.2.66)
Let us first discuss the simplest option, namely that all of the λA
+ have always
identical boundary conditions, either periodic or antiperiodic. In the periodic sector
one easily computes that the normal ordering constant ãP is negative (− 31 ). Hence,
there are no massless states in this sector. In the NS sector we find in addition to
(2.1.2.63) the massless states (denoting with b̃A A
r the Fourier coefficients of λ+ in the
anti-periodic sector)
P sector it removes half of the groundstates (leaving only spinors of definite chirality
with respect to the internal space spanned by the A directions).
Another option is to group the λA+ into two groups of 16 directions. Then we would
naturally split the state (2.1.2.67) into three groups: (120, 1), (1, 120), and (16, 16),
depending on whether A and B in (2.1.2.67) are both in the first half (1, . . . , 16),
both in the second half (17, . . . , 32), or one of them out of the first half and the other
one out of the second half. So far, this gave only a rearrangement of those states.
But now we impose the GSO projection such that only states survive where an even
number of fermionic left moving creators act in each half separately. This removes the
(16, 16) combination. Further, when we split the range of indices into two groups of
16 each, there will be additional massless states. It is simple to check that in the sector
where half of the boundary conditions are periodic and the other half is anti-periodic
(the AP or PA sector), the left moving normal ordering constant vanishes. Hence, the
corresponding ground states give rise to massless fields, provided right moving creation
operators act such that level matching is satisfied. This gives (removing half of those
states by GSO projection) (128, 1) additional massless vectors from the PA sector,
and another (1, 128) from the AP sector. Together with the vectors from the AA
sector this gives an E8 × E8 Yang-Mills field. The R sector state fills in the fermions
needed for N = 1 supersymmetry in ten dimensions. This corresponds to the other
known N = 1 anomaly free field theory.
The bosonic parts of the massless spectra of the consistent closed string theories in
ten dimensions is summarized in table 2.1. We have added the number of supercharges
Q from a target space perspective, and also the number of worldsheet supersymmetries
ψµ , in the NSR formulation.
2. Strings in non-trivial backgrounds 28
where R(2) is the scalar curvature computed from γαβ . Throughout this section we will
consider a Euclidean worlsheet signature. Note, that the dilaton term does not contain
α′ . In general, the theory (2.1.3.1) cannot be quantized in an easy way. The best one
can do is to take a semiclassical approach. Since α′ enters like ~ in ordinary field
theories this will result in a perturbative expansion in α′ . The term with the dilaton
can be viewed as a first order contribution in this expansion. Without this term,
(2.1.3.1) has again three local symmetries: diffeomorphisms and Weyl invariance. The
dilaton term breaks Weyl invariance in general. We will be interested in the question
under which circumstances Weyl invariance remains unbroken in the semiclassically
quantized theory. To answer this, first note that Gµν , Bµν , and Φ can be viewed as
couplings from a two dimensional perspective. Weyl invariance in particular implies
global scale invariance. But scale invariance is related to vanishing beta functions in
field theory. Thus, we will compute the beta functions of Gµν , Bµν and Φ as a power
series in α′ . However, there is a subtlety here. Under field redefinitions (infinitesimal
shifts of X by χ [X]) the couplings change according to
and
Lκ = χρ Bκρ . (2.1.3.6)
2. Strings in non-trivial backgrounds 29
with Vµ being an arbitrary target space vector. It is easy to check that also (2.1.3.1)
possesses the invariance (2.1.3.7). The symmetry (2.1.3.7) can be taken care of by
allowing for arbitrary Lµ in (2.1.3.3). Thus the couplings and hence the beta functions
are not unique. But actually we will be not just interested in vanishing beta functions.
This would ensure only global scale invariance. The requirement of Weyl invariance is
more strict and will fix the arbitrariness.
In order to compute the beta functions, we need to fix the worldsheet diffeomor-
phisms. We leave the explicit form of the fixed metric γαβ unspecified. The gauge
fixing procedure introduces ghosts, the diffeomorphism invariance is replaced by BRST
invariance. The ghost action depends only on the 2d geometry. Therefore, we expect
that the ghosts contribute only to the dilaton beta function. We will not treat them
explicitly but guess their contribution in the end of this section. The semiclassical ap-
proach means that we start from some background string X̄ µ satisfying the equations
of motion. We study the theory of the fluctuations around this background string.
Instead of using the fluctuation in the coordinate field X µ we will take the tangent
vector to the geodesic connecting the background value X̄ µ with the actual value X µ .
This difference is supposed to be small in this approximation. In order to compute
the tangent vectors we connect the background value and the actual position of the
string by a geodesic. The line parameter t is chosen such that at t = 0 we are at the
background position and at t = 1 at the actual position. The geodesic equation is (the
dot denotes the derivative with respect to t),
Note that the target space Christoffel connection Γµνρ depends on X µ . The first non-
trivial effects should come from terms second order in the fluctuations in the action.
(First order terms vanish when the background satisfies the equations of motion.) We
call the tangent vector to the geodesic (at X̄ µ )
where
Further, we may choose local coordinates such that only the constant and the term
linear in t appears in (2.1.3.11) and all higher order terms vanish in a neighborhood
of X̄ µ . (This is done by spanning the local coordinate system by tangent vectors to
geodesics.) The corresponding coordinates are called Riemann normal coordinates. In
these coordinates the Taylor expansion of the various terms in (2.1.3.1) around X̄ µ
takes the following form (up to second order in the fluctuations),
1
∂α X µ = ∂α X̄ µ + Dα ξ µ + Rµ λκν X̄ ξ λ ξ κ ∂α X̄ ν , (2.1.3.13)
3
1
Gµν (X) = Gµν X̄ − Rµρνκ X̄ ξ ρ ξ κ , (2.1.3.14)
3
1
Bµν (X) = Bµν X̄ + Dρ Bµν X̄ ξ ρ + Dλ Dρ Bµν X̄ ξ λ ξ ρ
2
1 λ ρ κ 1 λ
− R ρµκ Bλν X̄ ξ ξ + R ρνκ Bλµ X̄ ξ ρ ξ κ , (2.1.3.15)
6 6
µ 1
Φ (X) = Φ X̄ + Dµ Φ X̄ ξ + Dµ Dν Φ X̄ ξ µ ξ ν , (2.1.3.16)
2
where Dρ denotes the usual covariant derivative in target space, and Rµ νρσ is the
target space Riemann tensor
Note that in the Riemann normal coordinates the contributions quadratic in the
Christoffels vanish. Further, we have defined
Dα ξ µ = ∂α ξ µ + Γµλν ξ λ ∂α X̄ ν . (2.1.3.18)
Collecting everything, one can expand the action (2.1.3.1) in a classical contribution
S0 and a contribution due to fluctuations. There will be no part linear in ξ µ as long
as X̄ µ satisfies the equations of motion. The first non-trivial part is quadratic in the
ξ µ . We denote it by
(2) (2) (2)
S (2) = SG + SB + SΦ , (2.1.3.19)
2. Strings in non-trivial backgrounds 31
with
In what follows, capital latin indices will be raised and lowered with the Minkowski
metric. The normal coordinate expansion is useful not only to get the expressions
(2.1.3.20), (2.1.3.21), (2.1.3.22) in a covariant looking form. An important advantage
of this method is that the functional measure (in a path integral approach) for the
ξ A is the usual translation invariant measure. This will simplify the computation of
the partition function. In order to be able to do the field redefinition (2.1.3.23) in a
meaningfull way we have to ensure that the fluctuations are parameterized by target
space vectors. The tangent vectors to geodesics connecting the background with the
actual value are a natural choice. Before writing down the action in terms of the ξ A ,
we will absorb the first term in (2.1.3.21) in an additional connection in the kinetic
term (the first term in (2.1.3.20)). That can be done by adding and subtracting a
term looking like
∂α X̄ ρ ∂ α X̄ κ Hρ λ µ Hκλν ξ µ ξ ν .
We define the covariant derivative on ξ A by plugging (2.1.3.23) into (2.1.3.18) and
introducing an additional connection
i ǫα β
Dα ξ A = Dα ξ A + √ ∂β X̄ ρ EµA Hρ µ ν EB
ν B
ξ , (2.1.3.26)
2 −γ
where Dα ξ A corresponds to the contribution from (2.1.3.18). The part of the action
quadratic in fluctuations finally takes the form
Z
2 √
1
(2) αβ A A B
S =− d σ γ γ Dα ξ Dβ ξ A + MAB ξ ξ , (2.1.3.27)
4πα′
2. Strings in non-trivial backgrounds 32
ǫαβ
MAB = γ αβ ∂α X̄ µ ∂β X̄ ν GµνAB + i √ ∂α X̄ µ ∂β X̄ ν BµνAB + α′ R(2) FAB . (2.1.3.28)
γ
following way,
α′
Gµν = Gren
µν − log µ2 /ǫ GµνA A , (2.1.3.34)
2
α ′
ren
Bµν = Bµν − log µ2 /ǫ BµνA A , (2.1.3.35)
2
ren 1 2
d + cg ′ A
Φ = Φ − log µ /ǫ − + α FA , (2.1.3.36)
2 6
where G, B and F are as defined in (2.1.3.29), (2.1.3.30) and (2.1.3.31) but now in
terms of the renormalized couplings. Further, we have included a possible contribution
of the diffeomorphism fixing ghosts. Their action depends only on the intrinsic two
dimensional geometry and neither on the embedding in the target space nor on the
form of the background fields Gµν , Bµν and Φ. Therefore, the ghosts can contribute
only a constant renormalization of the dilaton Φ which we have parameterized by cg
in (2.1.3.36). The beta functions can be computed by taking the derivative of the
renormalized couplings with respect to log µ using the µ independence of the bare
couplings. Up to order α′ this leads to (they are all expressed in terms of renormalized
quantities and we supress the corresponding superscript in the following)
(G) ′ 1 λρ
βµν = α Rµν − Hµ Hνλρ , (2.1.3.37)
4
(B) α′ λ
βµν = D Hλµν , (2.1.3.38)
2
d + cg α′
β (Φ) = − + D 2 Φ. (2.1.3.39)
6 2
Because of the ambiguities related to the field redifinitions (2.1.3.2) – (2.1.3.4) we
cannot just set the beta functions to zero but only deduce that the model is Weyl
invariant (in first approximation) if
(G) (G)
β̄µν = βµν + D(µ Mν) = 0, (2.1.3.40)
(B) (B) 1
β̄µν = βµν + Hµν λ Mλ + ∂[µ Lν] = 0, (2.1.3.41)
2
1
β̄ (Φ) = β (Φ) + ∂µ Φ M µ = 0 (2.1.3.42)
2
The vectors Mµ and Lµ are not fixed by just checking for global scale invariance. In
order to compute them we would need to impose (local) Weyl invariance. This could
be done by computing the expectation value of the trace of the energy momentum ten-
sor. Instead of doing so, we will choose a rather indirect way of fixing the ambiguity.
Implicitly, we will be using a theorem stating that β̄ (Φ) is constant if (2.1.3.40) and
(2.1.3.41) are satisfied[111]. In other words this means that up to a constant contribu-
tion (2.1.3.42) should be an integrability condition for the other two equations. Before
2. Strings in non-trivial backgrounds 34
deriving such a condition we need to study of which form the vectors Mµ and Lµ could
be at the given order in α′ . We want to express a vector in terms of our background
fields Gµν , Φ and Bµν . The field Bµν should enter only via the gauge invariant field
strength Hµνλ (since we have performed partial integegrations in S (2) such that the
beta functions come out in a gauge invariant form). With this information it is easy
to check that the only option we have is (a is some constant)
Mµ = a∂µ Φ , Lµ = 0, (2.1.3.43)
where we do not consider a gradient contribution in Lµ since this would not be relevant.
The next step is to take the divergence of (2.1.3.40). Using the Bianchi identity (i.e.
the vanishing divergence of the Einstein tensor), the identity
1
H ρλµ Dρ Hλµν = Dν Hµλρ H µλρ ,
6
and equations (2.1.3.40) and (2.1.3.41) one obtains
a 2 α′ 2 a2 2
Dν D Φ + H + ′ (∂Φ) = 0, (2.1.3.44)
2 12 2α
where we have defined
Without the H 2 term, (2.1.3.44) and (2.1.3.46) would be the same. The H 2 term in
the dilaton beta function is actually missing in our computation since we took into
account only one loop contributions. Any counterterm in the action leading to an
order α′ contribution in the dilaton beta function should be linear in α′ . Since, at tree
level the B field enters with a factor α1′ , the H 2 term in β (Φ) corresponds to a two loop
contribution. In our implicit approach we obtained this term (and in fact all order α′
terms in β (Φ) ) without doing a two loop analysis.
We were not able to fix the value of the constant a, however. This is because
it could be absorbed in a rescaling of the field Φ. But this would change the ratio
of the constant contribution to the dilaton beta function to the other contributions.
Therefore, a is not arbitrary. The constant a can be fixed for example by studying
models with trivial metric and B field and a linear dilaton. These models are much
easier to treat than the generic one. The result is
a = 2α′ . (2.1.3.47)
2. Strings in non-trivial backgrounds 35
Let us discuss the case of trivial metric and vanishing B field a bit further. For a
linear dilaton, the β̄ (G) and β̄ (B) vanish identically. According to the previously stated
theorem (and to our result) the β̄ (Φ) function is constant in this case. Models with
that feature are known as conformal field theories. The constant dilaton β̄ function
is related to an anomaly of the transformation of the energy momentum tensor under
conformal coordinate changes (while keeping the worldsheet metric fixed). If we fix
the worldsheet metric to be the Minkowski metric the anomalous transformation of
the energy momentum tensor with respect to (2.1.1.9) reads
+ 2
dσ c
T̃σ̃+ σ̃+ = +
Tσ+ σ+ σ + + S σ̃ + , σ + (2.1.3.48)
dσ̃ 12
where the second term denotes the Schwarz derivative
z ′ z ′′′ − 32 (z ′′ )2
S (w, z) = , (2.1.3.49)
(z ′ )2
where z is a function of w and the primes denote derivatives. The Schwarz derivative
has the following chain rule
2
∂v
S (w (v (z)) , z) = S (v, z) + S (w, v) . (2.1.3.50)
∂w
The transformation law (2.1.3.48) is the most general possibility such that associa-
tivity holds. An analogous consideration applies to T−− . Now, from (2.1.3.48) one
can deduce part of the operator product expansion (OPE) of two energy momentum
tensors. To this end, one considers infinitesimal transformations and uses the fact that
they are generated by T++ . One obtains the following OPE
c/2 2
Tσ̃+ σ̃+ σ̃ + Tσ+ σ+ σ + = 4 + 2 Tσ+ σ+ σ
+
(σ̃ +
−σ )+ +
(σ̃ − σ ) +
1 +
+ + ∂σ + Tσ + σ + σ + ... , (2.1.3.51)
σ̃ − σ +
where the dots stand for terms which are regular for σ̃ + = σ + . For linear dilaton
backgrounds this OPE can easily be computed directly19 , leading to (2.1.3.47).
It remains to fix the contribution coming from the gauge fixing ghosts cg . This
can of course be calculated directly[375, 376]. Here, we will guess it correctly, instead.
From our discussion of the quantized bosonic string in the light cone gauge in 2.1.2 we
remember that the classical Lorentz covariance was preserved in d = 26. Comparing
with (2.1.3.48) we observe that our gauge fixing procedure was justified only if c = 0.
Since, we did not have a linear dilaton background there, this can happen only if
cg = −26. (2.1.3.52)
19
One should first compute Tαβ by varying the action with respect to γαβ , and gauge fix γαβ = ηαβ
afterwards.
2. Perturbative expansion and effective actions 36
+ +...
Figure 2.3: Perturbative expansion of the four point function in a string computation
power series
∞
X
A= gs2n+2 A(n) . (2.1.4.1)
n=0
This power series can be truncated after the first contributions as long as gs ≪ 1. It
remains to specify what gs is. To this end, we first observe that the power of gs in the
expansion terms in (2.1.4.1) is nothing but minus the Euler number of a worldsheet
with n handles and four boundaries. From (2.1.3.1) we know that the dilaton Φ
couples to the Euler density of the worldsheet. This follows immediately from the
Gauss-Bonnet theorem
Z
1 √
d2 σ γR(2) = χ ≡ 2 (1 − n) − b, (2.1.4.2)
4π
where n is the number of handles and b is the number of boundaries of the two
dimensional worldsheet. (The calculation of the scattering amplitudes is performed
after the worldsheet signature has been Wick rotated to the Euclidian one.) Thus, one
can identify
gs = ehΦi , (2.1.4.3)
where hΦi denotes a constant vacuum expectation value (VEV) of the dilaton. (Re-
member from the previous section that a constant contribution to Φ was not fixed
by the conformal invariance conditions. This is true for all string theories as can be
easily seen by noticing that a constant shift in Φ shifts the action (2.1.3.1) by a con-
stant.) Therefore, the string coupling is an arbitrary parameter in the perturbative
approach to string theory. It is only restricted by the consistency requirement that
the perturbative expansion in figure 2.3 should not break down, i.e. gs ≪ 1.
There is also a second approximation in the computation of the scattering ampli-
tudes. Since the massive string states have masses of the order of the Planck mass,
they are “integrated out”. This means that we are interested in effects below the
Planck scale where those fields do not propagate. The effective field theory actions
2. Perturbative expansion and effective actions 38
contain only the massless modes. For consistency, one should then restrict to processes
where the momentum transfer is p2 ≪ 1/α′ .
The (bosonic part of the) effective Lagrangians with at most two derivatives and
only the massless fields turn out to be
where Suniv does not depend on which of the superstring theories we are looking at,
and Smodel is model dependent. The universal sector has as bosonic fields the metric,
the dilaton, and the B field. The corresponding action is
Z
1 10
√ −2Φ 2 1 2
Suniv = 2 d x −Ge R + 4 (∂Φ) − H , (2.1.4.5)
2κ 12
where κ2 ∼ (α′ )4 is the ten dimensional gravitational constant. Note that the set
of equations of motion obtained from this action coincides with the conformal invari-
ance conditions (2.1.3.53)– (2.1.3.55), with the difference that for the superstring the
constant contribution in the dilaton equation vanishes for d = 10.
For type II strings there are additional contributions giving the kinetic terms for
the RR gauge forms and Chern-Simons interactions,
Z
II 1 X 1
Smodel = − 2 d10 x 2
Fp+2 , (2.1.4.6)
2κ p
2(p + 2)!
where Fp+2 is the fieldstrength of a p + 1 form RR gauge field (plus –in some cases–
additional contributions which we will discuss later),
The number p is the spatial extension of an object which couples electrically to the
p + 2 form gauge field. In the worldvolume action of the corresponding p dimensional
object this coupling is
Z
dp+1 σ iǫα1 ···αp+1 ∂α1 X µ1 · · · ∂αp+1 X µp+1 Aµ1 ···µp+1 . (2.1.4.8)
From expression (2.1.3.1) we observe that the fundamental string is electrically charged
under the NSNS B field. We will meet objects which are charged under the RR gauge
forms when discussing D-branes in section 2.3. For the type IIA theory we have
2. Perturbative expansion and effective actions 39
In the type IIA theory the definition (2.1.4.7) is modified for the four form field strength
F4 = dA3 + A1 ∧ H, (2.1.4.10)
leading to a non standard Bianchi identity for the four form field strength
dF4 = F2 ∧ H. (2.1.4.11)
F3 = dA2 − A0 ∧ H, (2.1.4.13)
1 1
F5 = dA4 − √ A2 ∧ H + √ B ∧ F3 . (2.1.4.14)
3 3
The Chern-Simons interaction for the type IIB theory is
Z
IIB 9
SCS = − 2 A4 ∧ H ∧ F3 . (2.1.4.15)
4κ
The five form field strength has to be selfdual. This is not encoded in the action
(2.1.4.6) but has to be added as an additional constraint,
F5 = ⋆F5 . (2.1.4.16)
In the heterotic string we have gauge fields transforming in the adjoint of SO(32)
or E8 × E8 . Their field strength is defined as (we assign mass dimension one to the
√
gauge fields A – this is related to a α′ rescaling of A)
F = dA + A ∧ A. (2.1.4.17)
20
This can be done by adding the Bianchi identity dFp+2 = d (· · · ) with a Lagrange multiplier to
the action and integrating out Ap+1 . Because of covariance the Lagrange multiplier is a 7 − p form
and its field strength is an 8 − p form.
2. Perturbative expansion and effective actions 40
α′
H = dB − (ωY − ωL ) . (2.1.4.18)
4
The Yang-Mills Chern-Simons form ωY is
2
ωY = trA ∧ dA + trA ∧ A ∧ A, (2.1.4.19)
3
where A is the gauge connection of either E8 × E8 or SO(32). The modification
(2.1.4.18) implies that the B field transforms under gauge transformations and under
local Lorentz rotations in a non-trivial way such that H is gauge invariant. The
Yang-Mills Chern-Simons form has the property that its exterior derivative gives the
instanton density (in a four dimensional subspace with Euclidean signature),
If we take the ten dimensional geometry to consist of a direct product of a six dimen-
sional non compact and a four dimensional compact space (with Euclidean signature)
the modification (2.1.4.18) implies restrictions on the allowed gauge bundles on the
four dimensional compact space. The integration of dH over a compact space should
vanish. It follows that the Euler number of this space must be equal to the instanton
number of the gauge bundle.
In addition to the universal piece (2.1.4.5), the heterotic action contains a gauge
kinetic term and also the Green-Schwarz term which ensures anomaly cancellation
heterotic
Smodel = Sgauge + SGS , (2.1.4.23)
with
Z
1 α′
Sgauge = − d10 xe−2Φ trF 2 , (2.1.4.24)
2κ2 8
21
We present the effective action for the heterotic string just for completeness, more details on
differential geometry and anomaly cancelation in the context of the effective heterotic theory can be
found e.g. in [223].
2. Perturbative expansion and effective actions 41
where again, the trace is taken over the gauge group (E8 × E8 or SO(32)). The
Green-Schwarz term is
Z
8πi
SGS = ′ B ∧ X8 , (2.1.4.25)
α
with (here, a power is meant with respect to the wedge product, e.g. F 4 ≡ F ∧F ∧F ∧F )
1 1 1 5 4 1 2 2
1 2 2 1 4 1 2 2
X8 = trF − trF − trF trR + trR + trR
2 (2π)6 48 4 8 8 8 32
(2.1.4.26)
and similar relations for the other RR field strengths. (If the terms denoted by dots in
(2.1.4.7) contain other RR field strengths additional Φ derivatives will be picked up.
But no relative power of eΦ will appear, since those terms always contain one RR field
strength or potential and an NSNS field strength or potential. The NSNS fields are
not rescaled.) After this rescaling all terms in the type II thoeries are of the structure
Since the rescaled (primed) fields correspond to the actual string excitations, the
effective type II actions given here contain only tree level contributions.
As (implicitly) stated above, one loop contributions are multiplied by gs2 and hence
enter the effective action with a factor of e−2Φ+2Φ = 1. In the type II examples, we
have seen that due to field redefinitions this correspondence may be changed. In
the heterotic case, however, there is no field redefinition such that all the terms in
the effective action are multiplied by the same power of eΦ . Indeed, the appearance
2. Toroidal Compactification and T-duality 42
of the Green-Schwarz term corresponds to a torus amplitude from the string theory
perspective. We excluded also higher orders in α′ which would lead to higher derivative
terms and contributions with massive string excitations.22 As long as the string scale
is much shorter (in length) than the scale of the process we are interested in those
terms can be neglected.
y ≡ y + 2πR. (2.1.5.2)
Solutions to (2.1.5.1) have to respect the periodicity (2.1.5.2). Therefore, the y depen-
dent part of ϕ can be expanded into a Fourier series of periodic functions. Focusing
22
An exception is the ωL correction in (2.1.4.18) and the Green Schwarz term. They can be deduced
by using supersymmetry and anomaly cancellation.
2. Toroidal Compactification and T-duality 43
with integer n, i.e. the momentum in the fifth direction is quantized. Plugging (2.1.5.3)
back into (2.1.5.1) leads to
η µν ∂µ ∂ν − m2n ϕn (xµ ) = 0, (2.1.5.4)
with
n
mn = , (2.1.5.5)
R
i.e. the nth Fourier mode leads in the effective four dimensional description to a Klein-
Gordon field with mass (2.1.5.5). Since the general solution of (2.1.5.1) is a super-
position of all Fourier modes the four dimensional description contains an infinite
Kaluza-Klein tower of massive four dimensional fields (depending only on the xµ ).
There are two limits to be discussed. The decompactification limit is R → ∞. In
this case all the Kaluza-Klein masses (2.1.5.5) vanish. The four dimensional descrip-
tion breaks down. The other limit is R → 0 (or the compactification radius becomes
much shorter than the experimental distance resolution). In this case, the KK masses
(2.1.5.5) become infinite except for n = 0. Only the massless mode survives and no
trace from the fifth dimension is left. This picture is very different in string theories
as we will see now.
Even though the bosonic string is inconsistent because it contains a tachyon, we will
first study the compactification of the bosonic string on a circle. The essential stringy
properties will be visible in this toy model. We compactify the 26th coordinate (the
25th spatial direction),
In the point particle limit string theory is just quantum mechanics of a free relativistic
25
particle. The plane wave solution contains the factor eip25 x where p25 is the center of
mass momentum in the 25th direction. This wave function should be periodic under
(2.1.5.6). This leads to a quantization condition for the center of mass momentum in
the compact direction
n
p25 = , (2.1.5.7)
R
2. Toroidal Compactification and T-duality 44
with integer n (the momentum number). So far, everything is analogous to the free
scalar field discussed above. The new stringy property arises by observing that the
string can wind around the compact direction. Technically, this means that the peri-
odic boundary condition for the closed string is modified
where the integer m denotes the winding number. With this ingredients the mode
expansions (2.1.2.3) and (2.1.2.4) are
25 1 25 n i X 1 25 −2ikσ−
XR = x + − mR σ − + α e , (2.1.5.9)
2 2R 2 k k
k6=0
1 25 n i X 1 25 −2ikσ+
XL25 = x + + mR σ + + α̃ e . (2.1.5.10)
2 2R 2 k k
k6=0
Taking into account the compact direction, the mass shell condition has to be modified
in a straightforward way,
24
X
pµ pµ = −M 2 . (2.1.5.11)
µ=0
N − Ñ = nm. (2.1.5.13)
Thus, for zero winding and momentum number the spectrum coincides with the spec-
trum of the uncompactified string (see section 2.1.2). In the massless sector we have
again a graviton, antisymmetric tensor and dilaton which are obtained from the state
αi−1 α̃25 25 i
−1 |ki , α−1 α̃−1 |ki (2.1.5.15)
are target space vectors. They correspond to gauge fields of a U (1) × U (1) gauge
symmetry. Finally, the state
α25 25
−1 α̃−1 |ki (2.1.5.16)
2. Toroidal Compactification and T-duality 45
n m N Ñ
1 1 1 0
-1 -1 1 0
1 -1 0 1
-1 1 0 1
2 0 0 0
-2 0 0 0
0 2 0 0
0 -2 0 0
Table 2.2: Each line in this table gives a configuration of winding, momentum and
occupation numbers leading to massless states at R2 = 12 .
Each of the first four states in table 2.2 contains one creator. This gives four
additional massless vectors (if the creator points into a non-compact direction) and four
massless scalars (if the creator points into the 25th direction). The latter four states
in table 2.2 correspond to massless scalars. Together with (2.1.5.15) and (2.1.5.16) we
have six vectors and nine scalars. The vectors combine into an SU (2) × SU (2) gauge
field whereas the scalars form a (3, 3) representation. For the special value (2.1.5.18)
the gauge group U (1) × U (1) is enhanced to the non-abelian group SU (2) × SU (2).
The rank of the gauge group is not changed.
An immediate question is: what is so special about (2.1.5.18)? To answer this, we
rewrite (2.1.5.12) in a suggestive way
n2
M 2 = 4N + 4Ñ − 8 + + 4m2 R2 , (2.1.5.19)
R2
23
Later, in section 2.2.1, we will also discus the case R2 = 2, where less states become massless.
2. Toroidal Compactification and T-duality 46
where we already aplied (2.1.5.13). We observe that the spectrum is invariant under
α′
n ↔ m and R ↔ . (2.1.5.20)
R
Recall that in the previous equations we have set α′ = 1/2. The symmetry (2.1.5.20)
is called T-duality. Winding and momentum numbers are interchanged and simulta-
neously the compactification radius is inverted. If R takes the value (2.1.5.18), the
spectrum is invariant under interchanging winding with momentum. This radius is
called the selfdual radius. Because of the symmetry (2.1.5.20) we can restrict the
compactifications to radii equal or larger than (2.1.5.18). This is an important differ-
ence to the point particle discussed in the previous section. To make this difference
clearer let us take the compactification radius to zero. All the Kaluza-Klein momenta
diverge and only states with n = 0 survive. This is similar to the point particle case.
On the other hand, all winding states degenerate. In order to make sense out of
this situation one can apply the T-duality tranformation (2.1.5.20). But then R = 0
leads to the decompactification limit and we are back at the 26 dimensional string.
Therefore, in string theory there are always traces of compact dimensions left.
Compactifying the string on a D dimensional torus, the above considerations lead
to a ZD
2 symmetry in a straightforward way. However, combining the T-duality along
circles with basis redefinitions of the torus lattice and integer shifts in the internal B
field leads to an enhancement of the T-duality group to SO (D, D, Z).
In this section we will argue that the above described T-dualiy is also a symmetry for
non-trivial background configurations. We closely follow[392]. Our starting point is
the non-linear sigma model (2.1.3.1). Compactification of one target space dimension
is possible if the sigma model is invariant under constant shifts in this direction. For
the first term in (2.1.3.1) this implies that the tangent to the compactified direction
is a Killing vector. The second term is invariant provided that the Lie derivative of
Bµν in the Killing direction is an exact two-form. For the last term to be invaraint
the Lie derivative of the dilaton Φ must vanish. We now choose coordinates such that
the isometry is represented by a translation in the d − 1 direction
We call the other coordinates X i . The previously mentioned conditions on Bµν and
Φ imply that those fields are independent of X d−1 (up to gauge transformations).
The next step is to gauge the symmetry (2.1.5.21) and to “undo” this by constrain-
ing the gauge fields to be of pure gauge. The constraint is implemented with the help
2. Toroidal Compactification and T-duality 47
of a Lagrange multiplier λ which finally will replace X d−1 in the T-dual model. We
introduce two dimensional gauge fields Aα changing under (2.1.5.21) as
Aα → Aα − ∂α c, (2.1.5.22)
and replace
Integrating over λ will result in the constraint of vanshing field strength for the Aα
which in turn imposes
Aα = ∂α ϕ, (2.1.5.25)
with ϕ being a worldsheet scalar. Shifting X d−1 by ϕ gives back the original sigma
model (2.1.3.1). Thus, adding (2.1.5.24) does not change anything. However, there
is a subtlety here. Compactifying the d − 1 direction means that we identify X d−1
with X d−1 + 2π (this time we put the compactification radius into the target space
metric). In order to be able to absorb ϕ into X µ , ϕ should respect the same periodicity.
This can be ensured as follows. We continue the worldsheet to Euclidean signature
and study the sigma model for a torus worldsheet. Then we can assign two winding
numbers (corresponding to the two cycles of the torus) to the Lagrange multiplier
λ. Summing over these winding numbers (in a path integral approach) will impose
the required periodicity on the gauge fields Aα . Going through the details of this
prescription leads to the conclusion that the λ “direction” is compact λ ≡ λ + 2π.
Instead of integrating out λ (to check that we did not change the model) we can
integrate out Aα . (Since Aα is not a propagating field this can be done by solving the
equations of motion. As well, one can integrate out Aα in a path integral, which is
Gaussian.) This procedure leads us to a dual model
Z √
1 2 αβ µ ν αβ µ ν
S = − d σ −γγ G̃ ∂
µν α X̃ ∂β X̃ + ǫ B̃ ∂
µν α X̃ ∂β X̃
4πα′
Z
1 √
− d2 σ −γ Φ̃R(2) . (2.1.5.26)
4π
24
For the Minkowskian worldsheet signature the i in front of the Bµν coupling in (2.1.3.1) is replaced
by one.
2. Toroidal Compactification and T-duality 48
n o
The set of dual coordinates is X̃ µ = λ, X i , and the dual background fields are,
1 Bd−1,i
G̃d−1,d−1 = , G̃d−1,i = , (2.1.5.27)
Gd−1,d−1 Gd−1,d−1
To find the dual expression for the dilaton is a bit more complicated. One can compute
Φ̃ in a perturbative way. To this end, one requires that Φ̃ is such that the conformal in-
variance conditions (2.1.3.53), (2.1.3.54), (2.1.3.55) are satisfied whenever the original
background satisfies them. This leads to the following formula for the dual dilaton
√ p
e−2Φ −G = e−2Φ̃ −G̃. (2.1.5.30)
From a path integral perspective the dilaton transformation can be motivated as fol-
lows. The path integral measure for the X µ is covariant with respect to the metric
Gµν . In the dual model one would naturally use a measure which is covariant with re-
spect to the dual metric G̃µν . The change of the measure introduces a Jacobian which
leads to (2.1.5.30). To our knowledge this is a rather qualitative statement which is
difficult to prove explicitly.
To make contact with the simple case discussed in the previous section we should
take Gd−1,d−1 = R2 /α′ , Gij = ηij , Bµν = 0 and Φ = const. Then the T-duality
formulæ (especially (2.1.5.27)) imply that the compactification radius is inverted. The
dilaton receives a constant shift and nothing else changes. This dilaton shift was not
visible in the discussion of the previous section. On the other hand, in the present
section we did neither see that T-duality interchanges winding with momentum nor
that there is an enhancement of gauge symmetry at the selfdual radius, because we
did not study the spectrum of the general string theory.
while keeping the original prescription of combining left with right movers. Carrying
this prescription over to the fermionic sector we observe that in the right moving
Ramond sector (see (2.1.2.58) for the definition)
This in turn implies that the T-duality transformation takes us from the type IIA
to type IIB string and vice versa. Hence, T-duality is not a symmetry in type II
superstrings but relates the type IIA string with type IIB. (This is true, whenever we
perform T-duality in an odd number of directions.) The type IIA string compactified
on a circle with radius R is equivalent to the type IIB string compactified on a circle
with radius α′ /R. This is consistent with the observation that the massless spectra of
circle compactified type IIA and type IIB theories are identical as depicted in table
2.3, where µ, ν = 0, . . . , 8.
NS-NS R-R
IIA Gµν , Bµν , Φ, Gµ9 , Bµ9 Aµ , A9 , Cµνρ , Cµν9
IIB Gµν , Bµν , Φ, Gµ9 , Bµ9 ′ , Φ′ , C
Bµ9 ′
µνρ9 , Bµν
possibility that the non-abelian gauge fields can take constant vacuum expectation
values (vev) in the Cartan subalgebra of the gauge group. The fact that (at least) one
of the ten directions has been compactified is important here. Otherwise, a constant
vev could be gauged away. To see this explicitly, let us assume that the gauge field
vev is (proportional to a generator in the Cartan subalgebra)
Θ x9 Θ x9 Θ
A9 = = e− R ∂9 e R , (2.1.5.33)
R
where the second part of the equation shows that a constant vev is a pure gauge
configuration. However, in the compact case we have identified x9 with x9 + 2πR and
hence only gauge transformations which are periodic under this shift are allowed. This
implies that the Wilson line (2.1.5.33) can be gauged away only if Θ is an integer. From
this discussion it follows that generically the gauge group is broken to U (1)16 in the
compactification process. In addition there are the (abelian) Kaluza-Klein gauge fields
G9µ and B9µ corresponding to a U (1) × U (1) gauge symmetry. Thus, generically there
is a U (1)18 gauge symmetry in the circle compactified heterotic string. Depending
on the moduli (Wilson lines and compactification radius) there are special points of
stringy gauge group enhancement.
It can be proven that the E8 × E8 heterotic string and the SO(32) string are
continuously connected in moduli space once they have been compactified to nine
dimensions. This can be shown by observing that for a certain configuration of Wilson
lines (where the gauge group is broken to SO(16) × SO(16) in either theory) T-
duality maps the two compactifications on each other. (For details see e.g. Polchinski’s
book[371].) All other compactifications can be reached by continuously changing the
moduli. Including the original ten dimensional theories as decompactification limits
we see that the two different heterotic strings belong to the same set of theories sitting
at different corners in moduli space. For completeness we should mention that for
compactifications of the heterotic string on a D dimensional torus one finds the T-
duality group SO (16 + D, D, Z).
00
11
11
00
00
11
00
11 000
111
11
00
00
11
111
000
000
111 ≡
00
11
11
00
00
11
00
11
Figure 2.4: The interval as an orbicircle. The fixed points (black dots) form the ends
of the interval.
group. (We will consider Z2 as such a group. It leaves an arbitrary lattice invariant.)
There are some points or —when combined with the other directions— planes which
are invariant under the discrete group. These are the orbifold fixed planes. They
present singularities in the compact part of the space time. String theory gives a
physical meaning to orbifold fixed planes. We will see that certain string excitations
(particles or gauge fields from the target space perspective) are confined to be located
at the orbifold fixed planes. By looking at an example where the orbifold can be
reached as a singular limit of a smooth manifold we will see that for string theory the
singular nature of this limit is not “visible”. Instead of discussing the general setups
for orbifold compactifications we will present two examples: the bosonic string on an
orbicircle and the type IIB string on T 4 /Z2 . We hope that this will demonstrate the
general idea with a minimal amount of technical complications. For more details (and
also orbifold compactifications of the heterotic string) we recommend the review[354].
which is invariant under the orbifold projection x25 → −x25 . Since in the bosonic
string the right moving sector is identical to the left moving one, we first write down
the right moving states only. The result is collected in table 2.4 (i = 2, . . . , 24; the
zeroth and first direction are eliminated by the light-cone gauge).
Now we need to combine left with right movers such that the resulting state is
invariant under the Z2 . This is the case for the product of the vector with the vector
and the scalar with the scalar. Thus, we obtain the metric Gij , the antisymmetric
tensor Bij and the dilaton Φ. The additional U (1) vectors Gi25 and Bi25 are projected
out in contrast to section 2.1.5.2. The combination of the scalar from the left moving
sector with the scalar from the right moving sector yields a target space scalar G25 25 .
Since the groundstate is Z2 invariant, the tachyon will survive the projection. If we are
at the selfdual radius, there might be additional massless states (without imposing Z2
invariance these are listed in table 2.2). The action of the Z2 takes winding number
to minus winding number and momentum number to minus momentum number as
can be seen from the mode expansion (2.1.5.9), (2.1.5.10). This means that we can
keep only invariant superpositions of states. From the first four entries in table 2.2
we obtain two additional massless vectors. These arise as follows. We add the first
state to the second state of the listing and act with αi−1 , or we add the third to the
fourth state and act with α̃i −1 . We can also subtract the second from the first state
and act with α25 25
−1 , or we subtract the fourth from the third state and act with α̃−1 .
This gives two massless scalars. Adding the fifth to the sixth entry and the seventh to
the eighth, we obtain two more scalars at the selfdual radius. This looks very unusual.
Since we do not have any U (1) gauge fields away from the selfdual radius, now also the
rank of the gauge group is enhanced at the selfdual radius. There are also additional
tachyons at the selfdual radius. These are the two states which are obtained by adding
the n = 0, m = 1 vacuum to the n = 0, m = −1 vacuum. The second state is the
same with m and n interchanged. These two additional tachyons have mass squared
M 2 = −6, as can be easily computed from (2.1.5.19).
Now, we come to a new feature which is unique to string theory. There are addi-
tional twisted sector states. These states are periodic under shifting σ by π only up
to a (non-trivial) Z2 transformation. In our case this implies for the string that its
center of mass position has to be located at a fixed plane and that the integer Fourier
2. The bosonic string on an orbicircle 53
modes are replaced by half-integer ones in the 25th direction. In the twisted sector we
need to compute the groundstate energy. This can be done by first modifying equation
(2.1.2.46) in a straightforward way
∞ ∞
23 X 1X
atwisted = − n− r. (2.2.1.1)
2 2 1
n=1 r= 2
15
Regularizing this expression according to the prescription (2.1.2.47) gives a = 16 .
This implies that the groundstate is tachyonic and also that there is no massless state
coming from this twisted ground state. There is one more tachyonic state at the first
level in the twisted sector. This is obtained by acting with α25 α̃25 . Collecting the
−1 −1 2 2
results, we obtain one tachyon with M 2 = − 15 2 7
2 and one with M = − 2 at each fixed
plane. Altogether, there are four tachyons (and states with positive mass squared)
located at the fixed planes.
The singular nature of the fixed points does not raise any problem in string theory.
It introduces twisted sector states which result in additional particles which are located
at the orbifold-24-planes in target space.
It is interesting to note that the orbifold at the selfdual radius is equivalent to the
toroidally compactified bosonic string at twice the selfdual radius.27 For a detailed
disussion of this equivalence we refer to Polchinski’s book[371]. Here, we will only
check that the light (tachyonic and massless) spectra coincide. Obviously, the gauge
groups U (1) × U (1) are the same. For the bosonic string on the circle (with R2 = 2)
these are the off-diagonal metric and B-field components Gi25 , Bi25 , whereas for the
orbicircle compactification at R2 = 12 these come from states with non-trivial winding
and momenta as discussed above. It remains to identify the four additional massless
scalars and the two tachyons (found in the non-trivial winding-momenta sector) of the
orbicircle compactification (at selfdual radius) and the four additional tachyons from
the twisted sector in the circle compactification. Here, the special choice R2 = 2 for
the circle compactification comes into the game. With (2.1.5.19) evaluated at R2 = 2
we find exactly these missing states. At first, there are four massless scalars: the vacua
with m = 0 and n = ±4, or m = ±1 and n = 0. The two tachyons with M 2 = −6 are
obtained from the two vacua with m = 0 and n = ±2. The two twisted sector tachyons
with M 2 = − 15
2 correspond to the two vacua with m = 0 and n = ±1. The other two
twisted sector tachyons with M 2 = − 27 can be identified in the circle compactification
as the vacua with m = 0 and n = ±3.
The equivalence of the S 1 /Z2 compactification at the selfdual radius and the S 1
compactification at twice the selfdual radius shows that the moduli spaces of both
27
Alternatively, we could use the T-dual version at half the selfdual radius.
2. Type IIB on T 4 /Z2 54
x9
x8
compactifications are connected at this point. This feature has a stringy origin. From
the target space perspective this is quite surprising. A field theory on 24 + 1 dimen-
sional Minkowski space times an interval with certain fields constrained to live at the
endpoints of the interval is smoothly connected to a field theory on 24 + 1 dimensional
Minkowski space times a circle with all fields living in the whole space. However, due
to the tachyons both vacua are unstable. In the next section we will see that similar
things happen for the superstring which does not have tachyons in its spectrum.
This Z2 action leaves the four points indicated by black dots in figure 2.5 times the
2. Type IIB on T 4 /Z2 55
four points in the first torus invariant. Thus, we obtain sixteen orbifold five-planes. At
first, we construct the untwisted spectrum. Since in the type IIB case the left moving
sector is identical to the right moving one, we first write down the right moving states,
only. The result is collected in table 2.5. We choose the GSO projection such that
—in the notation of (2.1.2.56)— states with an odd number of minus signs survive.
The projection (2.2.2.1) can be viewed as rotations by 180◦ in the eight-nine plane
and simultaneously in the six-seven plane. This is useful for the identification of the
behavior of the R-sector under Z2 transformations. We consider only states which lead
to massless particles when combined with the left movers.
In the NSNS sector, we can combine the left moving vector with the right moving
one leading to the six dimensional graviton Gij , the antisymmetric tensor Bij and the
dilaton Φ. Further, we can combine scalars from the left moving sector with scalars
from the right moving one. This gives sixteen massless scalars corresponding to Gab
and Bab where the indices a, b are internal, i.e. a, b = 6, . . . , 9. The target space vectors
Gia and Bia are projected out. In the RR sector, we can combine the chiral spinor with
the chiral one, and the anti-chiral with the anti-chiral one. This leads to 32 massless
(on-shell) degrees of freedom in the RR sector. Tensoring a chiral spinor with a chiral
spinor gives a selfdual two-form potential (3 on shell components) and a scalar. The
tensor product of two antichiral spinors yields an anti-selfdual two-form potential and
a scalar. We can perform four of those combinations, each. With the notation of table
2.1 the RR states can be identified as follows:
!
∗ ∗ 4
• Cijkl (or Cabcd ) gives = 1 degree of freedom (one scalar),
4
∗
• Cabij gives 3 anti-selfdual two-forms and 3 selfdual two-forms (18 degrees of
freedom),
′ gives a two-form (6 degrees of freedom),
• Bij
′ gives six scalars,
• Bab
2. Type IIB on T 4 /Z2 56
• Φ′ gives a scalar .
All other fields from the RR sector are projected out. Fermionic degrees of freedom
are obtained from the NSR and RNS sector. Combing the vector with the anti-chiral
spinors gives four times28 a (2, 2) ⊗ (2, 1) = (3, 2) ⊕ (1, 2) representation of the six
dimensional little group SO (4) = SU (2) × SU (2).29 Therefore, this tensor product
provides us with four chiral gravitini and four chiral fermions.
Combining the NS sector scalars with the chiral R sector spinors gives 16 chiral
spinors. From the existence of the four chiral gravitini we can guess that the resulting
low energy effective field theory has N = 4 chiral supersymmetry in six dimensions.
(For a collection of supersymmetries in various dimension see[399].)
Before checking that also the rest of the massless states fit into supersymmetric
multiplets we should construct the twisted sector states. The construction does not
depend on the location of the fixed plane. Therefore, we restrict the construction
to one plane and multiply the result by 16. In the twisted sector, the NS fermions
with an index corresponding to a compact dimension are integer modded whereas the
R sector fermions are half integer modded. Now, there are NS sector zero modes
forming a four dimensional Clifford algebra. The twisted NS ground state is two-fold
degenerate after imposing the GSO projection. (We modify the notation of (2.1.2.56)
in a straightforward way. Since we have only two creators and two anihilators, the
twisted NS groundstate has two entries. Performing the GSO projection means that
we keep those states with an odd number of minus signs.) In the twisted R sector
we do not have zero modes in the compact direction. This lifts some of the vacuum
degeneracy as compared to the untwisted sector. The twisted R ground state is labeled
only by the first two entries. Again, we keep only states with an odd number of minus
signs. In order to deduce the masses of the states in the twisted NS sector, we observe
from (2.1.2.46) (and its regularisation) that replacing four integer moded bosons by
4 4 5
half integer moded ones changes the normal ordering constant by − 24 − 16 = − 12 .
Changing the modding of four worldsheet fermions from half-integer to integer gives
4 4 1
another shift of − 16 + 24 = − 12 . Thus, we arrive at
1
atwisted
NS = auntwisted
NS − = 0. (2.2.2.2)
2
The twisted NS sector groundstate is massless. The R sector groundstate is always
massless, since fermions have the same modding as bosons. The analogon of table
2.5 for the twisted sector is table 2.6. Since all the twisted sector groundstates are
28
One factor of two arises because the NSR and RNS sector yield such a tensor product, each. The
second factor of two is due to the two anti-chiral spinors in table 2.5.
29
A vector is in the (2, 2), an anti-chiral spinor in the (2, 1), a chiral spinor in the (1, 2), a selfdual
twoform in the (3, 1) and an anti-selfdual twoform in the (1, 3).
2. Comparison with type IIB on K3 57
s3 = −s4 + 2 scalars
R: |s1 s2 i
s1 = −s2 , + 1 (chiral) spinors
invariant under the Z2 we can form all possible left-right tensor products. Multiplying
with 16 (the number of fixed planes) we obtain 64 scalars from the NSNS sector. The
RR sector leads to 16 anti-selfdual two-forms and 16 scalars. The RNS and NSR sector
give rise to 64 chiral spinors.
After we have obtained the full massless spectrum of the type IIB string on T 4 /Z2
we can fit it into super-multiplets of N = 4 chiral supergravity in six dimensions. The
possible supermultiplets are the gravity multiplet and the tensor multiplet. The gravity
multiplet contains the graviton and four chiral gravitini from the untwisted sector. In
addition, five selfdual two-forms are in the gravity multiplet. A tensor multiplet is
made out of an anti-selfdual two-form, five scalars and four chiral fermions. The five
selfdual two-forms in the gravity multiplet we take from Bij , Bij ′ and C ∗ . After
ijab
filling the gravity multiplet, we are left with 21 anti-selfdual two-forms, 105 scalars
and 84 chiral fermions. Thus, the remaining degrees of freedom fit into 21 tensor
multiplets.
To summarize, the massless spectrum of the type IIB string on T 4 /Z2 consists of
one gravity multiplet and 21 tensor multiplets of D = 6 chiral N = 4 supersymmetry.
Some of the degrees of freedom are confined to live on the orbifold-5-planes which
fill the 5 + 1 dimensional non-compact space but are located in the four dimensional
compact space. In the next section we will argue, that this setup is smoothly connected
to compactifications without orbifold-5-planes.
SU (n) holonomy. This means that (for K3) going around closed (non-contractable)
curves induces an SU (2) transformation. In a toroidal compactification we split the
ten dimensional spinor into a couple of lower dimensional spinors. The possible values
of the internal spinor indices count the number of resulting lower dimensional spinors.
In a torus compactification, each value of the internal indices gives rise to a massless
spinor. This is because the internal homogenous Dirac equation has always a solution
– any constant spinor. If instead of a torus with trivial holonomy we compactify on
K3 with SU (2) holonomy, only spinors which do not transform under the holonomy
group give rise to massless six dimensional spinors. This removes half of the internal
components and thus breaks half of the supersymmetry. Indeed, all K3 compactifi-
cations yield the same massless spectra. This is a consequence of the fact that the
number of zero-modes (of Laplace and Dirac operators) does not change as we move
from one K3 to another one. The number of zero modes of the Laplace operators30
are usualy listed in Hodge diamonds. The Hodge diamond for K3 is
1
0 0
1 20 1 . (2.2.3.1)
0 0
1
In the following we explain (roughly) how to read (2.2.3.1). The K3 is a complex
manifold. Therefore, we can choose complex coordinates (and so we do). Then a
tensor can have a couple of holomorphic indices and a couple of anti-holomorphic
indices. In other words, there are (p, q) forms on K3, where p corresponds to the
number of holomorphic indices and q to the number of anti-holomorphic ones. Since
the complex dimension of K3 is two, the values of p, q can be zero, one or two. We
denote the number of zero modes of a (p, q) form with h(p,q) . These (Hodge) numbers
are arranged into a Hodge diamond as follows31
h0,0
h1,0 h0,1
h2,0 h1,1 h0,2 . (2.2.3.2)
h2,1 h1,2
h2,2
From (2.2.3.1) we can deduce that an object which represents a zero or a four form
in the internal space, has one zero mode. Such an object gives rise to one massless
30
There are several different Laplace operators whose form depends on the tensor structure of the
object they act on.
31
The symmetry of (2.2.3.1) is not accidental. The vertical symmetry is related to Hodge duality
and the horizontal one to interchanging holomorphic with anti-holomorphic coordinates.
2. Comparison with type IIB on K3 59
The Z2 action is taken into account by removing forms which are odd under the Z2 .
2. Comparison with type IIB on K3 60
1
0 0
1 4 1 . (2.2.3.4)
0 0
1
′ and three scalars Φ, Φ′ , C ∗ .
In six dimensions we obtain two twoforms Bij and Bij ijkl
The internal metric components Gab yield ten scalars. (Note, that constant rescal-
ings of the coordinates would change the range of those coordinates, and hence are
non-trivial deformations equivalent to a constant change of the corresponding metric
components.) From the metric deformations we obtain 48 less scalars than in the K3
compactification. It remains to take into account the twoforms on T 4 /Z2 : Bab , Bab
′
2.3 D-branes
In this section we will present another kind of extended objects resulting from string
theory — the D-branes. They are different from the previously studied orbifold planes.
D-branes can exist also in uncompactified theories. They are dynamical objects, i.e.
they interact with each other and can move independent of the size of some compact
space. (The orbifold planes could move only if we changed the size or shape of the
compact manifold.) When we discussed the fundamental string we did not consider
the possibility of open strings. We will catch up on that in the following.
δX µ ∂σ Xµ |πσ=0 = 0. (2.3.1.2)
For the closed string we have solved this equation by relating the values at σ = 0 with
the ones at σ = π. This procedure was local because we took the string to join to a
closed string at σ = π. Now, we proceed differently by not correlating the two ends of
the string, i.e.
Let us focus on the boundary at σ = 0. We have two choices to satisfy the boundary
condition. If —for i = 0, . . . , p— we allow for free varying ends (δX i arbitrary at the
boundary) we obtain Neumann boundary conditions at those ends32
∂σ X i = 0 , i = 0, . . . , p. (2.3.1.4)
For the remaining d − p − 1 coordinates X a we choose to freeze the end of the string
— the variation vanishes at the boundary. Hence, in those directions the end of the
string is confined to some constant position. The resulting boundary conditions are
Dirichlet conditions (ca is a constant vector),
X a = ca , a = p + 1, . . . , d. (2.3.1.5)
32
Here, we use i to label Neumann directions. After fixing the light cone gauge we take i = 2, . . . , p.
2. Open strings 62
The end of the open string defines a surface which extends along p + 1 dimensions and
is located in d − p − 1 dimensions. This object is called D-brane, where the “D” refers
to the Dirichlet boundary condition specifying its position. If we choose identical
boundary conditions for the other end of the open string (at σ = π), we describe
an open string starting and ending on the same D-brane. For different boundary
conditions the open string stretches between two different D-branes. The Neumann
conditions imply that no momentum can flow out of the ends of the open string. In the
Dirichlet directions momentum can leave the string through its end — it is absorbed
by the D-brane. The target space Lorentz group is broken to SO(p, 1).
Varying the action with respect to the worldsheet fermions results in the same
equations of motions as in the closed string case and in the boundary conditions
µ µ π
−ψ+µ δψ+ + ψ−µ δψ− σ=0
= 0. (2.3.1.6)
In the closed string case we have solved this by assigning either periodic or anti-
periodic boundary conditions to the worldsheet fermions. Since now the ends of the
string are separated in the target space this would imply some non-locality. Therefore,
we impose the boundary conditions (2.3.1.6) at each end separately
µ µ µ µ
−ψ+µ δψ+ + ψ−µ δψ− σ=0
= −ψ+µ δψ+ + ψ−µ δψ− σ=π
= 0. (2.3.1.7)
Let us focus again on the boundary at σ = 0. We can solve (2.3.1.7) by one of the
options
µ µ
ψ+ = ±ψ− at σ = 0. (2.3.1.8)
However, there is a correlation with the bosonic boundary conditions via worldsheet
supersymmetry. To be specific, we choose the plus sign for Neumann conditions
i i
ψ+ = ψ− at σ = 0. (2.3.1.9)
ǫ+ = −ǫ− at σ = 0, (2.3.1.10)
which implies that due to the boundary (at least) half of the worldsheet supersymmetry
is broken. (If we had started with (1, 0) worldsheet supersymmetry —as we did for the
heterotic string— the boundary would break worldsheet supersymmetry completely.)
In order to ensure that not all of the supersymmetry is broken we have to choose
2. Open strings 63
a a
ψ+ = −ψ− at σ = 0. (2.3.1.11)
We could also interchange the fermionic boundary conditions in Dirichlet and Neu-
mann directions. Then another worldsheet supersymmetry would survive. There is
no physical difference between the two choices. Nevertheless it is important, that we
take the boundary conditions in the Neumann directions to be “opposite” to the ones
in Dirichlet directions. One may also check that the open string action is invariant
under the worldsheet supersymmetries (2.1.1.20), (2.1.1.22) and (2.1.1.23) provided
that the worldsheet fermions satisfy the boundary conditions (2.3.1.9), (2.3.1.11) and
(2.3.1.10) is fulfilled. (Partial integration introduces boundary integrals which vanish
if these additional constraints hold.) Recall also that the functional form of the world-
sheet supersymmetry parameter is restricted by the chirality conditions (2.1.1.28).
In the following we are going to discuss the boundary conditions at the other end
of the open string at σ = π. Going back to the closed string we deduce from (2.1.1.22)
that for anti-periodic supersymmetry parameter ǫ+ the fermions are anti-periodic for
periodic bosons and vice versa. This means that anti-periodic ǫ+ belongs to the
NS sector and periodic ones to the R sector. From the discussion of the boundary
conditions at σ = 0 we infer that the supersymmetry parameter has to satisfy one of
the following conditions,
ǫ+ = ±ǫ− at σ = π. (2.3.1.12)
The sign in the second line of (2.3.1.13) is correlated to the sign in (2.3.1.12) by the
requirement of continuity at σ = π. Hence, ε is (anti)-periodic for the lower (upper)
sign in (2.3.1.12) (taking into account the sign in (2.3.1.10)).
Now, let us perform this doubling trick also on the worldsheet bosons and fermions.
For the bosons it is useful to rewrite the boundary conditions. Dirichlet boundary
conditions mean that
∂+ X a = −∂− X a at σ + − σ − = 0. (2.3.1.14)
2. Open strings 64
∂+ X i = ∂− X i at σ + − σ − = 0. (2.3.1.15)
The next step is to specify the boundary conditions at σ = π. After having done this,
one can perform the doubling trick, i.e. define a boson ∂X µ on the interval 0 ≤ σ < 2π
analogously to the definition of ε in (2.3.1.13) where ∂± X take the role of ǫ± . As the
reader can easily verify, the outcome is that ∂X µ is periodic whenever we have chosen
the same type of boundary conditions at the two ends of the string in the xµ direction.
The corresponding open string sectors are called DD (NN) according to the choice of
Dirichlet (Neumann) boundary conditions at the two ends. For an opposite choice of
boundary conditions (ND or DN strings) ∂X µ will turn out to be anti-periodic. In
analogy to the closed string we call the sector with periodic ε R sector and the one
with anti-periodic ε NS sector. For DD or NN strings this implies that in the NS
sector we take the boundary conditions at σ = π to be
i i a a
ψ+ = −ψ− , ψ+ = ψ− at σ = π. (2.3.1.16)
i i a a
ψ+ = ψ− , ψ+ = −ψ− at σ = π (2.3.1.17)
and obtain periodic boundary conditions. In the above we have used that, for example,
in the R-sector periodicity of ε implies that the boundary conditions of ǫ± at σ = π
are identical to the ones at σ = 0. Plugging this back into the supersymmetry trans-
formations (2.1.1.22) and (2.1.1.23) evaluated at σ = π and taking into account the
boundary conditions for the bosons, we obtain the boundary conditions of the world-
sheet fermions at σ = π. This in turn determines the periodicity of Ψµ . Performing
the same procedure for ND or DN boundary conditions, one finds that Ψµ is periodic
in the NS sector and anti-periodic in the R sector whenever xµ is a direction with ND
or DN boundary conditions. The ND or DN directions are somewhat similar to the
twisted sectors we met when discussing Z2 orbifolds.
At first, we will consider only the case of a single D-brane. This means that we can
have only DD or NN boundary conditions depending on whether we are looking at a
direction transverse or longitudinal to the D-brane. Then ∂X µ will always be periodic
and Ψµ (anti-)periodic in the (NS) R sector.
2. Open strings 65
The quantization of the open string is very similar to the closed superstring. In
the following we will point out the differences. At first, we solve the equations of
motion again by taking a superposition of left-moving and right-moving fields. For
the bosons, these are given in (2.1.2.3) and (2.1.2.4). The boundary conditions relate
± ±
this two solutions. (In addition, we need to replace e−2inσ → e−inσ .)33 In Neumann
directions, they impose
For Dirichlet directions, we obtain a similar relation and constraints on the zero modes,
i 1 X i −irσ+
ψ+ = √ br e , (2.3.1.23)
2 1
r∈Z+ 2
a 1 X a −irσ−
ψ− = √ br e , (2.3.1.24)
2 1
r∈Z+ 2
a 1 X a −rσ+
ψ+ = −√ br e . (2.3.1.25)
2 1
r∈Z+ 2
The next step is to eliminate two directions by performing the light cone gauge. We
take this to be the time-like (Neumann) direction and a space-like Neumann direction,
which we choose to be x1 .34 For the open string we have only an NS sector and
an R sector. Since the right movers are not independent from the left movers, the
right and left moving sectors do not decouple anymore. The constraints that the
expressions (2.1.1.24) – (2.1.1.27) vanish are not all independent. The zero mode part
of vanishing energy momentum tensor again yields the mass shell condition (since the
mode expansions differ by factors of two, there is a difference of a factor of four as
compared to the closed string (2.1.2.42)),
M 2 = 2 (N − a) , (2.3.1.30)
where a is the normal ordering constant and the number operator N is defined in
(2.1.2.43) for the NS sector and in (2.1.2.61) for the R sector. In the NS sector the
GSO projection operator is as in (2.1.2.50) with the second factor removed. The lowest
GSO invariant states in the NS sector are
where we have indicated again the momentum eigenvalue of the vacuum by k. The
first set of these states transforms in the vector representation of SO(p − 1) – the little
group of the unbroken Lorentz group. Hence, this state should be massless leading to
the consistency condition35
1
aN S = . (2.3.1.32)
2
Like in the closed string, this translates into a condition on the number of target-space
dimensions
d = 10. (2.3.1.33)
The first states in (2.3.1.31) (with label i) form a U (1) gauge field. The states with
label a are scalars transforming in the adjoint of U (1) (here, this appears just as a
pompous way of saying that they are neutral under U (1), however, later we will discuss
34
We do not consider an open string ending on a D0 brane, here. As in the case of the closed
string, it is useful to take Lorentz invariance as a guiding principle for a consistent quantization. For
a D0 brane, the Lorentz group is broken down to time reparameterizations which is too small for our
purposes. Later, we will see that we can obtain the D0 brane by T-dualizing a higher dimensional
D-brane.
35
In principle, we could combine the first states in (2.3.1.31) with one of the second states in order
to form a massive vector representation as long as p < d − 1. However, later we will see that the case
p = d − 1 is related by T-duality to the other cases. With this additional ingredient it follows that
the states in (2.3.1.31) must be massless.
2. Open strings 67
Spin/~
1
PHOTON
0 1
M 2 α′
non-abelian gauge groups where those fields are adjoints rather than singlets). Since
the center of mass position of the open string is confined to be within the world volume
of the D-brane, all the open string states correspond to target-space particles which
are confined to live on the D-brane. The NS mass-spectrum is depicted in figure 2.6.
The construction of the R sector vacuum state goes along the same lines as in
the closed string. The ten dimensional Majorana spinor decomposes into a couple of
spinors with respect to the unbroken Lorentz group SO(p, 1). We impose the GSO
projection by multiplying the states with one of the projection operators defined in
(2.1.2.59). The sign is a matter of convention. The R-vacuum is massless on its own.
It leads to target space spinors providing all fermionic degrees of freedom which are
needed to obtain the maximal rigid supersymmetry in p + 1 dimensions.36
In the following sections we will investigate systems with more than one D-brane.
This will lead to non-abelian field theories on a stack of D-branes. But before doing
so, we will briefly discuss the possible D-brane setups which are in agreement with
supersymmetry.
At first, consider the case that we have an open string with an odd number of ND
directions. Thus, we will have an odd number of directions where the worldsheet
36
The maximally possible amount of supersymmetry differs for rigid and local supersymmetry. In
rigid supersymmetry, the highest occurring spin should not exceed one, whereas in locally supersym-
metric field theories, spin two fields (the gravitons) are allowed. In 3 + 1 dimensions, the maximal
supersymmetry is N = 4 (N = 8) for rigid (local) supersymmetry. From this one can deduce the
maximally allowed amount of supersymmetry in higher dimensions by viewing the 3 + 1 dimensional
theory as a toroidally compactified higher dimensional theory.
2. Open strings 68
fermions have zero-modes. For example in the R sector, the zero-modes form a Clifford
algebra in p + 1 = odd dimensions. The representation of this algebra by the R ground
state will be irreducible (there is no notion of chirality in odd dimensions). Therefore,
we cannot perform the GSO projection on those states. The theory will not lead to
target-space supersymmetry.
Let us now discuss the case of an even number of ND directions, taken to be 8− 2n.
Then the GSO projection operator on the R sector ground state will be of the form
From this we deduce that the GSO projection is possible only if the number of ND
directions is an integer multiple of four. This means for example that, if a lower
dimensional D-brane lives inside the worldvolume of a higher dimensional D-brane,
the higher dimensional D-brane has to extend in four or eight more directions. We
could have deduced this result faster by noting that (2.3.1.34) defines a projection
operator only if n is a multiple of four since otherwise the second term in (2.3.1.34)
squares to −1.
In this section we will discuss sets of parallel Dp-branes.37 First, let us have a look
at two parallel Dp-branes which are separated by a vector δca in the transverse space.
(Later we will see that the distance between parallel D-branes is a modulus, i.e. any
value is consistent.) From strings ending with both ends on the same D-brane we
obtain the same spectrum as discussed in the section 2.3.1.2. In particular, we obtain
a U (1) × U (1) gauge symmetry where the corresponding gauge fields live on the first
brane for the first U (1) and on the second brane for the second U (1) factor.
37
Recall that the worldvolume of a Dp-brane has p space like and one time like dimension.
2. Open strings 69
In addition, we have strings stretching between the two branes. There are two
such strings with opposite orientations. As compared to section 2.3.1.2, only the
mode expansion for the bosons in Dirichlet directions is modified. The string starting
on the brane at ca and ending on the brane at ca + δca has the mode expansion
δca X1
X a = ca + σ− αa e−inτ sin nσ. (2.3.1.38)
π n n
n6=0
where |12i and |21i denote the NS vacua for strings stretched between the two D-
branes and we have dropped the zero mode momentum eigenvalues k in the notation.
In this sector, the lightest states form a vector and d−(p+2) scalars. (Note that, in the
light cone gauge, we have to combine one of the transverse excitations (2.3.1.42) with
the longitudinals (2.3.1.41) in order to obtain a massive vector.) The R sector states
provide the fermions needed to fill up supermultiplets. (The amount of supersymmetry
is the same as in the single brane case.)
2. D-brane interactions 70
Now, take the inter-brane distance to zero. We obtain two massless vectors and
2 (d − p − 1) massless scalars. Together with the massless fields coming from strings
ending on identical branes, the vectors combine into a U (2) gauge field, and the scalars
combine into d − p − 1 scalars transforming in the adjoint of U (2). (One can split
U (2) into a diagonal U (1) times an SU (2). All fields are SU (2) adjoints and neutral
under U (1).) Moving the D-branes apart from each other can be viewed as a Higgs
mechanism from the target space perspective. The amount of supersymmetry leads
to flat directions for the scalars in the adjoint of U (2). This means that a scalar can
have some non zero vev breaking U (2) to U (1) × U (1). With the given amount of
supersymmetry, all massless fields transform in the same representation of the gauge
group as the vector bosons (viz. in the adjoint). Therefore, the Higgs mechanism can
only work for non-abelian gauge groups. Our charge assignments of the open strings
stretched between two different D-branes thus lead to a consistent picture.
An economic way of studying systems with N parallel D-branes is to replace all
the different sectors corresponding to the possibilities of strings stretching among the
N D-branes by one matrix valued state
The off-diagonal elements correspond to massive vectors and scalars. Open string
sectors with opposite orientation have opposite charges under U (1)N , i.e.
Di Dj
Figure 2.7: D-brane Di and D-brane Dj talking to each other by exchanging a closed
string.
it can escape into the bulk of the target space. In particular, it may reach another
D-brane by which it is absorbed. The absorption process is inverse to the emission
process. The closed string hits the D-brane where it can split into an open string
which is constrained to live on the second D-brane. D-branes talk to each other by
exchanging closed strings. In figure 2.7 we have drawn such a process. In order to
make contact to conventions of the standard reviews on D-brane physics we take the
closed string twice as long as the open string (see also the footnote 33). This implies
± ±
that in the closed string mode expansions we replace e−2inσ → e−inσ .
We will compute the process depicted in figure 2.7 in Euclidean worldsheet signa-
ture. The range of the worldsheet coordinates is
σ σ
Dj
Di
Figure 2.8: D-brane Di and D-brane Dj talking to each other by a pair of virtual open
strings stretching between them.
µ µ
ψ± (τ + 2π, σ) = (−)F ψ± (τ, σ) . (2.3.2.3)
Thus, a (−)F insertion (canceling the (−)F in (2.3.2.3)) corresponds to closed string
RR sector exchange whereas no (−)F insertion yields the closed string NSNS sector
exchange. From the open string perspective there is no exchange of NSR or RNS
sector closed strings between the D-branes. In the picture 2.7, the D-brane appears
as a boundary state of the closed string.38 This boundary state is a superposition
of an NSNS sector state and an RR sector state. There are no NSR or RNS sector
contributions. This can be explained by the fact that the D-brane is a target space
boson. It is specified by the target space vector ca in (2.3.1.5) and hence transforms
as a vector and not as a spinor under rotations in the space transverse to the brane.
We will not present the details of the boundary state formalism here, and recommend
38
It will turn out that the closed strings which are exchanged in figure 2.7 are type II superstrings.
2. D-brane interactions 73
the review [195] instead. To be slightly more specific let us just present the defining
equation for a boundary state in closed string theory (as usual we label the Neumann
directions by i = 0, . . . , p and the Dirichlet directions by a = p + 1, . . . , 9)
This relates the right moving and left moving bosonic excitations the boundary state
can carry. Applying the worldsheet supersymmetry transformations (2.1.1.22) and
(2.1.1.23) on (2.3.2.4) and requiring that there is a combination of the two supersym-
metries which annihilates the boundary state tells us that the boundary state should
have the same number of right moving and left moving fermionic excitations. Also,
when ǫ+ is taken to be (anti)-periodic then ǫ− should have the same periodicity. The
boundary state cannot be excited by NS fermions in, say, the right moving sector and
R fermions in the left moving sector. It has only an NSNS and an RR sector.
Instead of the non-standard range for the open string worldsheet coordinates we
would like to have the standard range
σ
In order to achieve this we redefine τ → τ t and σ → 2l . Under this redefinition,
the Hamiltonian (which is obtained by integrating the τ τ component of the energy
momentum tensor over σ) transforms according to
H → 2lt2 H. (2.3.2.6)
Further, we want the time evolution operator when going once around the annulus
to transform as (2πt should be identified with the worldsheet time it takes the open
string to travel around the annulus once)
1 1 1 dH −ǫ
− log det H = − tr log H = lim tr =
2 Z2 2 ǫ→0
dǫ
∞
d dt ǫ −2πtH 1 1 ′
tr lim ǫ te + log 2π − Γ (1) . (2.3.2.9)
ǫ→0 dǫ 0 2t 2 2
2. D-brane interactions 74
At a formal level this expression is correct. However, the next step is to take the
limit of ǫ → 0 before performing the integral over t. This would be allowed only if
the integral were converging. This is not the case in most of the applications (for
example, the integral diverges if H is just a number and no trace is taken). But the
error done is some unknown additive constant contribution which is not of interest for
1
us39 . Together with this unknown constant we also drop the 2 (log 2π − Γ′ (1)) and
obtain for the amplitude in figure 2.8 (reinstalling α′ )
Z ∞
dt ′
Str e−2πα tH (2.3.2.10)
0 2t
(here we have replaced the trace by a supertrace. It refers to target space supersym-
metry, i.e. the trace receives an additional minus sign for target space spinors. The
integral over t is usually regulated by a UV cutoff near t = 0.) The expression (2.3.2.10)
has also an intuitive interpretation. The supertrace describes a process where a pair of
open strings is created from the vacuum, then propagates for a time 2πt and annihi-
lates. This corresponds to the diagram drawn in figure 2.8. Further, we integrate over
all possible moduli t of the annulus with the measure dt 2
2t . The Hamiltonian is p + M
2
α′ H = α′ H0 + HB + HN S/R , (2.3.2.14)
39
In our case the trace is actually a supertrace which vanishes when taken over a constant. However,
since the corresponding series does not converge absolutely the result depends on the ordering in which
we take the trace.
2. D-brane interactions 75
with
y2
H 0 = p2 + , (2.3.2.15)
4α′2 π 2 !
8 ∞
X X 1
HB = αi−n αin − , (2.3.2.16)
24
i=1 n=1
8 ∞
X X 1
rbi−r bir − ,
HN S = (2.3.2.17)
48
i=1 1
r= 2
8 ∞
!
X X 1
HR = ndi−n din + . (2.3.2.18)
24
i=1 n=1
An additional minus sign in the R sector contribution is due to the fact that we take
the supertrace with respect to space time supersymmetry (R-sector states are space
time fermions). The superscript GSO indicates that the trace is taken over GSO even
states. We will clarify this point later.
The trace over the zero modes is
Z
dp+1 k −2πtα′ k2 − ty2′ p+1 − ty2
2 ′ − 2
tr ZERO = 2Vp+1 e 2πα = 2Vp+1 8π α t e 2πα′ , (2.3.2.19)
MODES (2π)p+1
where the factor of two counts the possible orientations of the open string traveling
through the annulus. The factor Vp+1 denotes formally the worldvolume of the parallel
Dp branes. It arises due to the normalization of states with continuous momentum
(hp|pi = δ(p+1) (0) = Vp+1 / (2π)p+1 ). To express oscillator traces, it is useful to define
the following set of functions40
∞ ∞
1 Y 1 √ Y
f1 (q) = q 12 1 − q 2n , f2 (q) = q 12 2 1 + q 2n ,
n=1 n=1
∞
Y ∞
Y
1 1
f3 (q) = q − 24 1 + q 2n−1 , f4 (q) = q − 24 1 − q 2n−1 . (2.3.2.20)
n=1 n=1
In order to translate the open string calculation back to the closed string process
(figure 2.7) we will make use of the modular transformation properties,
π √ π π
f1 e− s = sf1 e−πs , f3 e− s = f3 e−πs , f2 e− s = f4 e−πs . (2.3.2.22)
40
These are related to the Jacobi theta functions, which are also often used in the literature.
2. D-brane interactions 76
Next, we are going to compute the trace over the worldsheet bosons. The sum over the
coordinate label i in (2.3.2.16) can be written in front of the exponential as a product.
Nothing depends explicitly on the direction i, therefore this gives a power of eight to
the result for a single bosonic direction. The second sum can be decomposed into a
level part and an occupation number part, giving the result
∞ ∞
!8
πt Y X
trBOSONS e−2πtHB = e 12 e−2πtlk . (2.3.2.23)
l=1 k=0
Here, l denotes the level of a creator α−l acting on the ground state and k is the
occupation number (the number of times this creation operator acts). The sum over
k is just a geometric series, and we obtain the result
1
trBOSONS e−2πtHB = . (2.3.2.24)
f18 (e−πt )
The calculation of the traces over the fermionic sectors is similar. Let us just point
out the differences. First of all, we have to take the trace only over GSO even states.
This is done by inserting the GSO projection operator into the trace and summing
over all states
1 1
trGSO (· · · ) = tr (· · · ) + tr (−)F · · · . (2.3.2.25)
2 2
The second difference – as compared to the bosonic calculation – is that for world-
sheet fermions the occupation number can be only zero or one (since the creators
anticommute). The NS trace without the (−)F insertion comes out to be
1 1
trNS e−2πtHNS = f38 e−πt . (2.3.2.26)
2 2
Since the NS vacuum is GSO odd we assign an additional minus to states with even
and zero occupation if (−)F is inserted into the trace,
1 1
trNS (−)F e−2πtHNS = − f48 e−πt . (2.3.2.27)
2 2
For the R-sector trace without the (−)F insertion one obtains
1 1
− trR e−2πtHR = − f28 e−πt , (2.3.2.28)
2 2
where the 16-fold degeneracy of the R vacuum has been taken into account by the
√
factor of 2 in the definition of f2 (2.3.2.20). The R sector trace with a (−)F insertion
vanishes identically. Half of the R sector groundstates have eigenvalue +1 whereas the
other half has eigenvalue −1. Adding up all the results and using the identity (2.3.2.21)
we find that the net result for the annulus amplitude vanishes. This implies also that
2. D-brane interactions 77
R
Sector l→∞ dl×
− p+1 p−9 y2
RR - 12 4α′ π 2 2
Vp+1 l 2 e− 4πα′ l
− p+1 p−9 y2
NSNS 1
2 4α′ π 2 2
Vp+1 l 2 e− 4πα′ l
Table 2.7: Contributions of massless RR and NSNS sector closed strings to figure 2.7.
the closed string diagram 2.7 vanishes, and a hasty interpretation of this would lead
to the conclusion that D-branes do not interact (at least not via the exchange of
closed strings). However, as we will argue now, this is not the case. The situation
rather is that repulsive and attractive interactions average up to zero. In order to
see this41 , let us translate the annulus result back to the tree channel. Further, we
would like to filter out the contributions of massless closed string excitations. To this
end, we replace t in terms of l using (2.3.2.8), and afterwards apply (2.3.2.22). The
contribution of massless closed string excitations is obtained by focusing on the leading
contribution in the l → ∞ limit. (Massless interactions have infinite range whereas
the interactions carried by massive bosons have finite range.) We collect the result
of this straightforward calculation in table 2.7. We separate closed RR contributions
from NSNS sector contributions. The former ones correspond to the (−)F insertion
and the latter to the 1 insertion in the annulus amplitude.
From table 2.7 we deduce that interactions carried by closed strings in the RR
sector cancel interactions mediated by closed strings in the NSNS sector. One can
take the diagram 2.7 to the field theory limit. In that limit the ‘hose’ connecting
the two D-branes becomes particle propagators (lines). In the NSNS sector we find
propagators for the metric (fluctuations), the dilaton and the anti-symmetric tensor
Bµν . The D-branes appear as source terms for those fields. The NSNS contribution to
diagram 2.7 tells us the strength of this coupling. In particular, it yields the strength
of the gravitational coupling which is given by the tension Tp . A detailed analysis of
the effective field theory and comparison with table 2.7 leads to the result [371]
π 2 ′ 3−p −2Φ0
Tp2 = 4π α e , (2.3.2.29)
κ2
41
Already in the annulus computation, there are signs for such a cancellation. For the result, the
minus sign in front of the R sector contribution was essential. Since R sector states are target space
fermions, this indicates that the result is due to target space supersymmetry.
2. D-brane actions 78
where κ is the gravitational coupling in the effective theory (see section 2.1.4) and
Φ0 denotes the constant vev of the dilaton. Even though we did not derive this
explicitly here, let us make a few comments to motivate the expression qualitatively.
In a field theory calculation the propagator of an NSNS field is accompanied by a
power of κ2 e2Φ0 . With Tp defined as in (2.3.2.29) the κ and Φ0 dependence drop
out. Since κ2 ∼ (α′ )4 , the mass dimension of Tp is correct. (The α′ dependence
in the string calculation yields agreement with (2.3.2.29) after substituting for the
integration variable l such that the α′ dependence in the exponent vanishes.) Further,
the exchange of massless particles should lead to a Coulomb interaction in field theory.
For the interaction between the two D-branes this means that the potential should be
given by the distance to the power of two minus the number of transverse dimensions
V ∼ y p−7 . (2.3.2.30)
In the string result (table 2.7), we can extract the y dependence after rescaling the
integration parameter l such that the y dependence in the exponent disappears. The
result agrees with (2.3.2.30). In order to fix the numerical coefficient, one needs to do
a more detailed analysis of the field theory calculation. More details on this can be
found in Polchinski’s book[371].
The second line in table 2.7 tells us that and how string RR fields couple to the
brane. We find the same Coulomb law as for the gravitational interaction. The RR
field should be a p + 1 form. The p = even branes interact via closed type IIA strings
and the p = odd branes via closed type IIB strings. The value of the RR contribution
is exactly minus the value of the NSNS contribution. This provides us with the RR
charge of the D-brane42
where we have taken into account the dilaton dependent RR field redefinition per-
formed in section 2.1.4. The signs are undetermined at this level.
In this section we will modify the calculation presented in section 2.1.3 such that
it accommodates open string excitations. We perform a Wick rotation such that the
worldsheet is of Euclidean signature. Further, we map the parameter space of the open
string worldsheet from a strip to the upper half plane via the conformal transformation
z = eτ +iσ = z 1 + iz 2 . (2.3.3.1)
The discussion is performed for bosonic strings but later we will also give the result for
superstrings. Since the open string couples naturally to closed strings (via joining its
ends) we also switch on non-trivial closed string modes. These are the metric Gµν and
the antisymmetric tensor Bµν . We will take those background fields to be constant.
For the open string we take Neumann boundary conditions in all directions at first.
(For the superstring this is not consistent with RR conservation. At the moment we
will ignore this problem and return to it later.) Later, we will discuss that T-duality
maps Neumann to Dirichlet boundary conditions. Hence, the restriction to Neumann
boundary conditions will not result in a loss of generality. At first, we consider a single
brane setup. The massless open string excitation is now a U (1) gauge field Aµ . We
restrict ourself to the special case that the U (1) field strength Fµν is slowly varying, i.e.
we neglect contributions containing second or higher derivatives of Fµν . Under these
conditions we will be able to perform the calculation to all orders in α′ (in difference
to section 2.1.3). The nonlinear sigma model reads
Z
1 1
S = ′
d2 z ∂α Xµ ∂ α X µ + iǫαβ Bµν ∂α X µ ∂β X ν
2πα 2
Z
1 µ
+i dz Aµ ∂1 X . (2.3.3.2)
z 2 =0
Here, A has been rescaled such that α′ appears as an overall factor in front of the action.
The target space indices µ,ν are raised and lowered with the constant background
metric Gµν . The worldsheet metric is taken to be the identity in the z α coordinates
(2.3.3.1).43 Using Stoke’s theorem the term with the constant B field can be rewritten
as a surface integral
Z Z
2 αβ µ ν
d zǫ Bµν ∂α X ∂β X = 2 dz 1 Bµν X µ ∂1 X ν . (2.3.3.3)
z 2 =0
The term with the B-field can be absorbed into a redefinition of the gauge field Aµ
and we can put it to zero without loss of generality. (It can be recovered by replacing
F → F − 2B.)
43
In principle this choice introduces gauge fixing ghosts as stated in section 2.1.3. Since their effect
is not altered by the presence of the boundary we will not discuss the ghosts here. We should, however,
mention that there are technical subtleties when taking into account the dilaton in worldsheets with
boundaries, see e.g. [49]. (Recall that the ghosts contribute to the dilaton beta function.)
2. D-brane actions 80
∂z ∂z̄ X̄ µ = 0, (2.3.3.4)
∂2 X̄ µ + iFν µ ∂1 X̄ ν |z 2 =0 = 0. (2.3.3.5)
The presence of the U (1) gauge field Aµ results in inhomogeneous Neumann boundary
conditions. Since we have restricted ourselves to the case where the target space
metric is constant, the background field expansion simplifies in comparison to the
computation of section 2.1.3. The fields can simply be Taylor expanded. Neglecting
second and higher derivatives of Fµν , the background field expansion terminates at
the third order in the fluctuations,
Z
1 1
S X̄ + ξ = S X̄ + d2 z ∂ α ξµ ∂α ξ µ
2πα′ 2
Z
i 1 1 ν µ 1 ν λ µ 1 ν λ µ
+ dz Fµν ξ ∂1 ξ + ∂ν Fµλ ξ ξ ∂1 X̄ + ∂ν Fµλ ξ ξ ∂1 ξ .(2.3.3.6)
2πα′ z 2 =0 2 2 3
Since we have chosen the worldsheet metric to be the identity (and the geodesic curva-
ture of the boundary to vanish) a suitable technique to integrate out the fluctuations
ξ µ is given by a Feynman diagrammatic approach. This means that we split the action
into a free and an interacting piece. The free piece determines the propagator whereas
the interacting one leads to vertices. As the free part of the action we take
Z Z
1 2 α µ i
Sf ree = d z∂ ξ ∂
µ α ξ + dz 1 Fµν ξ ν ∂1 ξ µ . (2.3.3.7)
4πα′ 4πα′ z 2 =0
Hence, the interacting part is given by the rest
Z
i 1 1 ν λ µ 1 ν λ µ
Sint = dz ∂ F
ν µλ ξ ξ ∂1 X̄ + ∂ F
ν µλ ξ ξ ∂1 ξ . (2.3.3.8)
2πα′ z 2 =0 2 3
In order to compute the propagator we have to invert the two dimensional Laplacian
and to satisfy the (inhomogeneous Neumann) boundary conditions arising from the
variation of Sf ree with respect to ξ µ (with free varying ends of the ξ µ ),
1 ′
′
✷∆ µν z, z = −δ z − z Gµν , (2.3.3.9)
2πα′
∂2 ∆µν z, z ′ + iFµ λ ∂1 ∆λν z, z ′ 2
= 0. (2.3.3.10)
|z =0
This boundary value problem can be solved (for example by borrowing the method of
mirror charges from electro statics) with the result
∆µν z, z ′ =
|z − z ′ | −1 µν 2 z − z̄ ′
′
−α Gµν log + Ĝ log z − z̄ ′ +θ µν
log .(2.3.3.11)
|z − z̄ ′ | z̄ − z ′
2. D-brane actions 81
µ ν
∆µν (z, z ′ )
′
z z
ν λ
1 µ
2 ∂ν Fµλ ∂1 X̄
ν µ λ
1
3 ∂ν Fµλ · (∂1 ·)·
Figure 2.9: Feynman rules: The dotted line denotes the worldsheet boundary. The
slash on the leg means that a derivative acts on the corresponding leg.
Here, we have introduced the following matrices (an index S(A) stands for (anti-
)symmetrization and Gµν are the components of the inverted target space metric G−1 ,
as usual)
µν µν µν
−1 1 1 1
Ĝ = = G , (2.3.3.12)
G+F S G+F G−F
Ĝ = Gµν − F G−1 F µν , (2.3.3.13)
µν
µν µν
1 1 1
θ µν = = F . (2.3.3.14)
G+F A G+F G−F
The interaction piece Sint gives rise to two vertices —one with two and one with
three legs— located at the boundary. The Feynman rules are summarized in figure
2.9.
The propagator ∆µν (z, z ′ ) becomes logarithmically divergent if the arguments co-
incide. Therefore, we replace the logarithm of zero by its dimensionally regularized
version
Z 2 Z
′ d k k(z−z ′ ) ǫ d2−ǫ k
log z − z |z=z ′ = − e |z=z ′ = − lim µ , (2.3.3.15)
2k2 ǫ→0 2 (k2 + m2 )
where the momentum integral extends over a two dimensional plane. We introduced a
mass scale µ which is needed in order to keep the mass dimension fixed while changing
the momentum space dimension. In the last step we have introduced also an infrared
2. D-brane actions 82
(and replacing the bare gauge field by the renormalized one. Hoping that the renor-
malization program is sufficiently familiar we do not introduce sub- or super-scripts
indicating the difference between bare and renormalized couplings). The beta-function
d
of Aµ is obtained by applying µ dµ on the renormalized couplings and using the fact
that the bare couplings are independent of the cutoff. This leads to (now, Aµ denotes
the renormalized coupling)
d λν
βρA = µ Aρ = ∂ν Fρλ Ĝ−1 . (2.3.3.18)
dµ
In this case the vanishing of the beta function ensures conformal invariance. (We do
not encounter the subtleties which we met in section 2.1.3. Partially, this is the case
because we have written the action always in a manifestly gauge invariant form, i.e.
in terms of the gauge field strength. By performing partial integrations differently we
could have carried out the calculation in a slightly more complicated way, with the
same result.) The equation of motion for the gauge field is
βµA = 0. (2.3.3.19)
where p + 1 is the number of Neumann directions (i.e. for our discussion p + 1 = 10(26)
for the super (bosonic) string45 ). The factor in front of the integral in (2.3.3.20) has
not been fixed by our current discussion. We will explain how to fix it below. The
same applies to the dilaton dependence. (We discussed only the case of a constant
dilaton Φ.) Since we have rescaled Aµ by powers of α′ such that the α′ dependence
2. D-brane actions 83
∂µ F µν = 0. (2.3.3.21)
where the Φ dependence has been taken such that the result coincides with the small
α′ expansion of (2.3.3.20). Expanding (2.3.3.20) in powers of F and keeping only terms
R √
up to F 2 we find in addition to (2.3.3.22) a contribution dp+1 xe−Φ −G. From a
field theory perspective this is a tree level vacuum energy. So far, we did not properly
couple the open string excitations to gravity. We included the effects of bulk fields
on the equations of motion for open string excitations, but we did not encounter a
back reaction, i.e. that the field Aµ enters the equations of motion for the closed string
excitations. The reason is that the back reaction is an annulus effect. We will not
present a detailed annulus calculation but sketch the result. (In principle we have
done the necessary computations in the previous section.) Since the beta functions
depend on local features (short distance behaviors) one would guess that for the beta
function it may not matter whether the worldsheet is an annulus or a disc. However,
the annulus may degenerate as depicted in figure 2.11.
44
A different (maybe more elegant) way to deal with infrared divergences is discussed e.g. in [102],
see also the appendix of[228].
45
We view the result of our computation as a result for the bosonic modes of the superstring. Since
we did not specify the effective action for the closed bosonic string the factor in front of (2.3.3.20) is
irrelevant for the bosonic string (see also the discussion of the Fischler Susskind mechanism below).
2. D-brane actions 84
Figure 2.11: Fischler Susskind mechanism: The annulus degenerates into a punctured
disc as the inner circle shrinks to zero. This gives rise to a closed string counterterm
depending on the open string excitations.
This gives an additional short distance singularity. (The inner circle of the annulus
becomes short.) This singularity can be taken care of by adding counterterms to
the closed string action. The counterterms depend on open string modes. This in
turn leads to terms in the closed string beta functions which depend on the open
string modes. This process is known as the Fischler Susskind mechanism. The net
effect is that we add the open string effective action to the closed string effective
action (2.3.3.20) and obtain the equations of motion by varying the sum. This is
the expected back reaction. (In particular the Einstein equation now contains the
energy momentum tensor of the open string modes.) After taking the back reaction
into account, the coefficient in (2.3.3.20) does matter. In the previous section we
have computed the tension of the D-brane (2.3.2.29). This fixes the coefficient and the
dilaton dependence46 as given in (2.3.3.20). According to our discussion in the previous
section, the presence of a D-brane should also back-react on the RR background. We
could not see this in the present consideration since we did not take into account non
trivial RR backgrounds. (In fact, it is rather complicated to switch on non-trivial RR
backgrounds in the non-linear sigma model.) We will come back to the discussion of
RR contributions to the open string effective action below.
So far, we have studied the case of a single D-brane. How is this discussion modified
in the presence of multiple D-branes? We have focused on the case where we have only
Neumann boundary conditions. This means that multiple D-branes must sit on top
of each other, simply because there is no space dimension left in which they could be
separated. The effect of having more than a single brane is that the gauge field Aµ is
a U (N ) gauge field – it is a matrix. Calling the expression (2.3.3.2) an action does not
make much sense anymore since we would have a matrix valued action. Therefore, one
46
Note also that this dilaton dependence agrees with our general discussion in section 2.1.4. The
Euler number of the disc differs by one from the Euler number of the sphere.
2. D-brane actions 85
takes just the bulk part of the action (the first line in (2.3.3.2)) and computes instead
of the partition function the Wilson loop along the string boundary[142],
D R µ
E
W = tr P ei ∂M dtAµ Ẋ , (2.3.3.23)
where we have denoted the boundary of the worldsheet by ∂M and chosen some t
to parameterize this curve. The letter P stands for the path ordered product. The
expectation value is computed with respect to the bulk action only. Now, it is prob-
lematic to get an expression containing all orders in α′ . The leading α′ contribution
to the beta function results in the Yang-Mills equation
∇µ F µν = 0, (2.3.3.24)
where ∇ denotes a gauge covariant derivative. The effective action in leading ap-
proximation can be obtained as follows. We expand (2.3.3.20) to first order in F 2 .
We replace F 2 by trF 2 . In addition, we multiply the zeroth order term in F by the
number N of D-branes (the tension is N times the tension of a single D-brane). The
generalization of the Dirac-Born-Infeld action (2.3.3.20) to non-abelian gauge fields is
a subject of ongoing research, see e.g. [449, 75].
In the previous section we have discussed the case of having Neumann boundary
conditions in all directions. This means that the D-branes have been space filling
objects. In order to obtain results for D-branes extending along less dimensions we
will discuss T-duality for open strings, now.
At first, we focus on the case with trivial background fields. From section 2.1.5 we
recall that T-duality interchanges winding with momentum modes. For the open string
we have either winding or momentum modes in compact directions. A string with
DD boundary conditions along the compact dimension can have non-trivial winding
modes. Since the ends of the string are tied to the D-brane it cannot unwrap. On
the other hand, the DD string does not have quantized Kaluza Klein momenta. The
D-brane can absorb any momentum carried by the string in the compact direction.
For NN strings, opposite statements are true. If the string has Neumann boundary
conditions along the compact dimension, its ends can move freely in that direction –
it can continuously wrap and unwrap the compact dimension. On the other hand, the
string cannot transfer Kaluza-Klein momentum to the D-brane. The NN string has
non-trivial momentum modes. This consideration suggests that T-duality for open
strings interchanges Neumann with Dirichlet boundary conditions.
Let us substantiate these qualitative statements by studying the effect of T-duality
on the mode expansions. For the T-duality transformation we use the “recipe” (2.1.5.31).
2. D-brane actions 86
x9 ≡ x9 + 2πR. (2.3.3.25)
For the string with NN boundary conditions in the ninth dimension this implies that
the center of mass momentum is quantized
n
p9 = , (2.3.3.26)
R
with n being an integer. There are no integer winding numbers in the case of NN
boundary conditions. We rewrite the mode expansion (2.3.1.20) in a suggestive way
X 9 = XR9
+ XL9 , (2.3.3.27)
9 x9 n − i X 1 9 −inσ−
XR = + σ + α e , (2.3.3.28)
2 2R 2 n n
n6=0
x9 n + i X 1 9 −inσ+
XL9 = + σ + α e . (2.3.3.29)
2 2R 2 n n
n6=0
Applying the recipe (2.1.5.31), we obtain the mode expansion for the T-dual coordinate
T-DUALITY n X1
X9 −→ X̃ 9 = σ+ α9 e−inτ sin nσ. (2.3.3.30)
R n n
n6=0
1 α′
R′ = = . (2.3.3.31)
2R R
The interpretation is that the open string ends on a D-brane located at x9 = 047 .
The open string winds n times around a circle of radius R′ . It is rather obvious that
—starting from a DD string with mode expansion (2.3.3.30)— T-duality will take us
to an NN string with mode expansion (2.3.3.27), (the center of mass position depends
again on the way we distribute a constant between left and right movers). So, T-
duality inverts the compactification radius and interchanges Dirichlet with Neumann
boundary conditions. We leave it to the reader to verify that an investigation of the
worldsheet fermions and of ND directions is consistent with this picture.
In section 2.3.2 we have noticed that Dp-branes with p even (odd) interact via
the exchange of closed type IIA(B) strings. Our present observation that T-duality
along a compact direction interchanges Dirichlet with Neumann boundary conditions
implies that a Dp-brane with even p is mapped onto a Dq-brane with odd q, and vice
47
We could have obtained a different position by distributing the center of mass position of the NN
string x9 asymmetrically among left and right movers.
2. D-brane actions 87
versa (q = p ± 1). This goes along nicely with our earlier statement (section 2.1.5.4)
that T-duality along one circle interchanges type IIA with type IIB strings.
Finally, let us discuss T-duality for open strings in the presence of non-trivial
background fields. For the closed string we have done this in section 2.1.5.3. Because
the discussion of the closed string background fields is not affected by the open string,
we will focus on the special case where only the open string gauge field is non-trivial.
For simplicity we also restrict to one D-brane only (for multiple D-branes see e.g.
[143])48 . Let us first outline in words the procedure we are going to carry out. The
compactification has to be done in a Killing direction. (Shifts along the compact
direction are isometries.) We will take this dimension to be the ninth. The next step
is to gauge this isometry and to undo the gauge by forcing the corresponding gauge
field to be trivial. This will be done again by adding a Lagrange multiplier times the
field strength of the isometry gauge field. The Lagrange multiplier will become the
T-dual coordinate in the end. In particular the Lagrange multiplier lives on a circle
whose radius is inverse to the original compactification radius. This is derived from
the requirement that the radius of the isometry-gauge group (U (1)) agrees with the
compactification radius. We will not discuss the technical details of this derivation
(they are presented for example in the appendix of [15]). Instead, we will focus on
a detailed discussion of the boundary conditions. The boundary condition of the
isometry gauge fields is constrained by the boundary condition of the open string. This
will be implemented by a second Lagrange multiplier which lives only at the boundary
of the worldsheet. After integrating out the isometry gauge fields the integration
over this second Lagrange multiplier will give the boundary condition for the T-dual
coordinate (the “first” Lagrange multiplier)49 .
After having described the strategy, we will now present the details of the proce-
dure. Setting α′ = 21 the open string worldsheet action with a non-trivial U (1) gauge
field coupling to the boundary reads (for convenience we use a rescaled Aµ as compared
to (2.3.3.2) and choose Minkowskian worldsheet signature here)
Z Z
1 2 µ α µ µ µ
S= d z ∂α X ∂ X + dt ( Aµ ∂t X + Vµ ∂n X ) , (2.3.3.32)
2π M ∂M
where M denotes the worldsheet and ∂M its boundary (parameterized by t). With
∂n we denote the derivative into the direction normal to the boundary. We specify the
48
We will comment briefly on the case of multiple D-branes at the end of this section.
49
As we will see below integrating over the second Lagrange multiplier κ boils down to setting an
argument of a delta function to zero. This in turn implies boundary conditions on the Lagrange
multiplier λ.
2. D-brane actions 88
Boundary Condition δX 9 ∂n δX 9
Dirichlet fixed free . (2.3.3.33)
Neumann free fixed
This implies that for Dirichlet boundary conditions we set A9 = 0 whereas for Neu-
mann boundary conditions V9 = 0 is chosen. For the Neumann boundary conditions
(free varying ends) the variation of S gives the boundary condition (we denote the
normal vector with nα and the tangent vector with tα )
1
nα ∂α X 9 = − F 9ν ∂t xν , (2.3.3.34)
2
where Fµν is the field strength of the U (1) gauge field Aµ . For Dirichlet conditions we
obtain
V9 = 0. (2.3.3.35)
Since this should be in agreement with our assignment that the variation of the end
of the open string in the ninth direction is fixed (possibly related to the variations in
other directions), the function Vµ should be interpreted as a vector which is tangent to
the brane. Equation (2.3.3.35) then means that the D-brane is localized in the ninth
direction.
Since we have chosen the simplified case of trivial closed string backgrounds any
direction (in cartesian target space coordinates) is an isometry. Suppose that in ad-
dition the x9 derivative of the U (1) gauge field is pure gauge, i.e. zero modulo gauge
transformations. So, without loss of generality we restrict ourselves to the case that
the gauge background is X 9 independent. We also assume that the tangent vector Vµ
does not depend on X 9 . We specify the boundary condition on X 9 by the equation
bα ∂α X|∂M
9
= independent of X 9 , (2.3.3.36)
S = S̄ + S (9) , (2.3.3.37)
Z Z
1
S (9) = d2 z ∂α X 9 ∂ α X 9 + dt A9 ∂t X 9 + V9 ∂n X 9 ,(2.3.3.38)
2π M ∂M
50
A fixed boundary condition on a variation means that this variation depends on the boundary
values of variations of other fields (or is zero). In particular, if we do not vary the other directions we
can replace “fixed” by “zero” in (2.3.3.33).
2. D-brane actions 89
where S̄ stands for the X 9 independent part. The action is invariant under constant
shifts in X 9 . We transform this into a local symmetry by the replacement
∂α X 9 → Dα X 9 = ∂α X 9 + Ωα , (2.3.3.39)
where Ωα is the isometry gauge field. (We use this terminology in order to avoid
confusion with the open string excitation mode Aµ .) The isometry gauge field Ωα
transforms under local shifts in X 9 such that Dα X 9 is invariant. We introduce a bulk
Lagrange multiplier λ in order to constrain the Ω-field strength51
f = ǫαβ ∂α Ωβ (2.3.3.40)
where for later convenience we have performed partial integrations such that no deriva-
tive of Ωα appears in the action. The worldsheet vector tα denotes the tangent vector
to the boundary. The T-dual model will be obtained by integrating out Ωα . The
T-dual coordinate will be λ. Its boundary conditions are going to be fixed by the
integration over κ. Before going through the steps of this prescription, let us verify
that the gauged action is equivalent to the ungauged one. Integration over λ leads to
Ωα = ∂α ρ, (2.3.3.42)
bα ∂α ρ = 0. (2.3.3.43)
Because neither the background (nor bα ) depend on X 9 , the scalar ρ can be absorbed
completely into a redefinition of X 9 without spoiling the boundary condition (2.3.3.36).
(In addition ρ needs to live on a circle with radius equals the compactification radius.
51
In two dimensions we can hodge dualize the two form field strength to a scalar f .
52
As before we write the case of D and N boundary conditions into one formula. Recall that A9 = 0
and bα = tα for D boundary conditions, and V9 = 0 and bα = nα for N boundary conditions.
2. D-brane actions 90
This issue has been addressed in [15]. The discussion given there leads to the obser-
vation that λ lives on a circle with inverted radius.) Hence, the gauged and ungauged
models are equivalent.
In order to construct the T-dual model we first integrate out Ωα . Because the
action (2.3.3.41) does not contain any derivatives of Ωα (it is ultra local with respect
to the isometry gauge field), the functional integral over Ωα factorises into a bulk
integral and a boundary integral
Z Z Z
DΩM∪∂M (. . . ) = DΩM (. . . ) × DΩ∂M (. . . ) . (2.3.3.44)
Integrating out Ω in the bulk leads to the ungauged bulk action with X 9 replaced by
λ. This is exactly as in the closed string computation (up to a boundary term)
Z
(9) 1
S̃bulk = d2 z ∂α λ∂ α λ + 2ǫαβ ∂α λ∂β X 9 (2.3.3.45)
2π M
Z Z
1
= d2 z ∂α λ∂ α λ + dt 2λ∂t X 9 , (2.3.3.46)
2π M ∂M
Let us evaluate this delta function for the two cases: X 9 has Dirichlet boundary
conditions (bα = tα ) or Neumann conditions (bα = nα ). In the first case, the evaluation
of the delta function fixes κ in terms of λ and sets V9 = 0. This means that λ has free
varying ends, i.e. Neumann boundary conditions. Taking into account the boundary
term in (2.3.3.46) we obtain that the dual U (1) gauge field is determined by the
position of the original D-brane,
Recall that the original Dirichlet boundary condition may depend on the other direc-
tions, i.e. the rhs of (2.3.3.48) is some fixed function.
If X 9 satisfies Neumann conditions, the evaluation of the delta function leads to
κ = 0 and the Dirichlet boundary condition
1
λ|∂M = − A9 . (2.3.3.49)
2
In the T-dual string theory there is a D-brane located in x9 along the curve A9 (note
that A9 may depend on the coordinates different from x9 ). Note also that plugging
2. D-brane actions 91
the boundary condition (2.3.3.49) into (2.3.3.46) cancels the original A9 coupling to
the boundary.
To summarize, we have seen that T-duality interchanges Dirichlet with Neumann
boundary conditions. The position of the D-brane is interchanged with the U (1)
gauge field component in the T-dualized directions. Starting with Neumann boundary
conditions it is easy to see that gauge transformations do not change the sigma model
for the string, i.e. the field equations of the string excitations do not depend on gauge
transformations. Via T-duality this translates to changes of the position of a D-brane,
in particular constant shifts are moduli of the theory. From the above expressions it
is also clear that performing the T-duality twice will result in the original theory.
With these considerations we can go back to the effective action (2.3.3.20) and
generalize it to non space filling branes. This is done by simply replacing the Aµ
components where µ labels a Dirichlet direction by scalars. These scalars are the
collective coordinates of the lower dimensional D-brane. One can also parameterize
the worldvolume of the D-brane by an arbitrary set of parameters. In this case one
needs to replace bulk fields by the induced quantities. The effective D-brane action
for lower dimensional D-branes can be also computed in the sigma model approach
directly. This has been done in [315].
Finally, let us comment briefly on the case of multiple branes. We start with
Neumann boundary conditions. The gauge field A9 is now a matrix. Suppose that
this matrix is diagonal. In this case the above discussion is valid if we just replace A9
by a diagonal matrix everywhere. In the T-dual theory, the position of the D-brane is
a diagonal matrix. The interpretation is that each entry corresponds to the position
of a single D-brane. The matrix describes a set of D-branes. The more general case of
non-diagonal gauge fields is rather complicated. It is addressed e.g. in[143, 140, 141].
2.3.3.3 RR fields
So far, we have discussed D-brane effective actions only for trivial RR backgrounds.
The reason was mainly of technical origin. It is rather complicated to describe non-
trivial RR backgrounds in a sigma model approach. Later in section 4.3, we will use
such a description for a particular background. Now, we will not discuss the RR
background in a sigma model. Instead we will use our computation of section 2.3.2
and field theoretic arguments.
In section 2.3.2 we have seen that the Dp-brane carries RR charge with respect
to a p + 1 form RR gauge potential of type II theories. In section 2.3.3.1 we argued
that the interaction of D-branes via closed strings is obtained by adding the effective
D-brane action to the effective type II action (IIA for even p, and IIB for odd p).
2. D-brane actions 92
Combining these two observations, we infer that the effective D-brane action contains
an additional piece
√ Z
π 3−p
S1 = SDBI + 4π 2 α′ 2 dp+1 xA1,... ,p+1 , (2.3.3.50)
κ
where we assume that the D-brane worldvolume extends along the first p + 1 dimen-
sions. (In general, the D-brane can be parameterized by a set of p + 1 parameters. In
this case, the D-brane action is written in terms of induced fields.) We have abbrevi-
ated the action (2.3.3.20) with SDBI . Further, we used the result (2.3.2.31) to fix the
coefficient in front of the RR coupling.
The label in S1 has been introduced because now we will argue that there are
further couplings to RR fields. These occur if another a D-brane lies within the
worldvolume of the considered D-brane, or a D-brane intersects the considered D-
brane. In such a case there will be strings starting and ending on different D-branes.
They give rise to massless fields transforming in the fundamental representation of
the gauge group living on the considered D-brane. Under certain circumstances there
may be chiral fermions leading to potential gauge anomalies. Such anomalies can be
canceled by assigning anomalous gauge transformations to certain bulk RR fields and
adding an interaction term to the effective D-brane action. This procedure has been
carried out in detail in[218, 145]. Here, we just briefly give the result.
In cases that there is an anomaly, this anomaly can be canceled by adding a Chern-
Simons term to the D-brane action
S = S1 + SCS , (2.3.3.51)
with (for N coincident D-branes – for N > 1 also the DBI action needs to be modified
as discussed in the end of section 2.3.3.1)
Z q
iF
SCS = C ∧ tr e 2π Â (R). (2.3.3.52)
Bp
The way of writing the Chern-Simons term needs explanation. The integral is taken
over the worldvolume of the Dp-brane which is denoted by Bp . The integral is a
formal expression in differential forms. It is understood that only p + 1 forms out of
this expression are kept.
The first form C is an RR q + 1 form where q is the spatial dimension of the surface
in which the two D-branes (or sets of D-branes) overlap. The last term contains the
so called A-roof genus. This is a polynomial in the curvature two-form (for an explicit
definition see e.g. [218]). In addition to adding SCS to the D-brane action the RR form
C receives a contribution under gauge transformations. This comes about as follows.
2. D-brane actions 93
The definition of the RR field strength receives a correction (the correction is related
to a Chern-Simons form whose explicit form is not needed here)
H = dC + correction (2.3.3.53)
such that
q
iF
dH = 2πδ (Bp → M10 ) TrN e 2π Â (R), (2.3.3.54)
where the delta function means that this correction is supported on the worldvolume
of the D-brane, only. Even though the right hand side of (2.3.3.54) is gauge invariant,
C has to change under gauge transformations in order to ensure that H is invariant.
The construction is such that the change of SCS under gauge transformations cancels
the anomaly.
where we restrict to the case that θ ij are c-numbers. Because of the non commuting
coordinates one has to specify the ordering in say complex functions. For our purpose
the Weyl ordering is appropriate. The Weyl ordering is constructed as follows. The
starting point is the pair of the function and its Fourier transform in commutative
space (first with commuting coordinates)
Z
1
φ (x) = n dn k eikx φ̃ (k) . (2.3.3.56)
(2π) 2
The Weyl ordered functions are defined by replacing the commuting coordinates xi
with the non commuting ones x̂i in (2.3.3.56) (but keeping k as a commutative inte-
gration variable),
Z
1
φW (x̂) = n dn k eikx̂ φ̃ (k) . (2.3.3.57)
(2π) 2
2. D-brane actions 94
Multiplying the two exponentials on the rhs of (2.3.3.58) using the BCH formula and
afterwards dropping the hat on the coordinates leads to a natural way to deform the
algebra of ordinary functions (φ: commuting Rn → C) by replacing the ordinary
product by the Moyal product
∂ ∂
iθ ij
(φ ⋆ ψ) (x) = e ∂xi ∂y j φ (x) ψ (y)|x=y . (2.3.3.59)
A possible U (1) gauge background could be absorbed into Bij by use of Stoke’s the-
orem. However, we will restrict first to the case that Bij is constant, and add a U (1)
gauge field coupling to the boundary later. Note also that Bij has now mass dimension
two – the canonical dimension of a gauge field strength. We consider the case that
all coordinates X i have Neumann boundary conditions (coordinates with Dirichlet
boundary conditions do not play a role here and may be added as spectators). The
propagator for the X i can be easily obtained from the expressions (2.3.3.11). The
53
The index i instead of µ indicates here that we focus on space like target space dimensions. The
rescalings of fields have been done mainly in order to achieve agreement with the literature[418].
2. D-brane actions 95
θ ij z − z̄ ′
− log . (2.3.3.65)
2π z̄ − z ′
We take z = τ + iσ (hoping that this does not cause confusion due to the fact that
now τ and σ parameterize the upper half plane whereas they parameterized a strip
earlier (and will so in later sections)). Ordering with respect to real and imaginary
part, one obtains for (2.3.3.65)
!
θ ij (τ − τ ′ )2 + 2i (σ + σ ′ ) (τ − τ ′ )
− log . (2.3.3.66)
2π (τ − τ ′ )2 + (σ + σ ′ )2
With this propagator one can compute the following operator product
i i i ij i (τ )+iq i (0)
: eipi x (τ ) :: eiqi x (0) : = e− 2 θ p i qj ǫ(τ )
: eipi X iX
:, (2.3.3.69)
where the normal ordering means that self contractions within the exponentials are
subtracted. By use of Fourier transformation one can deduce the operator product for
generic functions
i ij ∂2
θ
: φ (X (τ )) :: ψ (X (0)) : = : e 2 ∂X i (τ )∂X j (0) φ (X (τ )) ψ (X (0)) : . (2.3.3.70)
In the limit of coincident arguments the operator product can be related to the Moyal
product
This expression suggests that we are likely to obtain noncommutative field theory if
we use the limiting procedure on the lhs of (2.3.3.71) as a way to regularize composite
operators. This regularization technique is known as point splitting. In composite
operators well defined (normal ordered) parts are taken at different points, and then
the limit to coinciding points is performed (after adding counterterms if needed).
In the following we are going to argue that we obtain an effective noncommutative
theory on the D-brane if we use the point splitting regularization instead of dimensional
(or Pauli-Villars) regularization. For a trivial worldsheet metric point splitting simply
means that we cut off short distances by keeping
τ − τ ′ > δ, (2.3.3.72)
and take δ to zero in the end. First, we add the following interaction term to (2.3.3.60)
Z
Sint = −i dτ Ai (X) ∂τ X i . (2.3.3.73)
δAi = ∂i λ. (2.3.3.74)
Now, we are going to observe that whether or not the partition function is invariant
depends on the regularization prescription. To this end, note that δZ contains a term
Z Z
δZ = −h dτ Ai (X) ∂τ X · dτ ′ ∂τ ′ λi + . . . .
i
(2.3.3.75)
where in the second step the connection between the operator product and the Moyal
product (2.3.3.71) has been used. Hence, when using the point-splitting regularization
(2.3.3.72), the string partition function is not invariant under ordinary gauge trans-
formations. However, the lack of invariance can be cured by replacing the gauge field
Ai with a “noncommutative” gauge field Âi with the deformed gauge transformation
Indeed, computing the effective action of the open string with the point-splitting
method, one finds the noncommutative version of the Dirac-Born-Infeld action (2.3.3.20).
We will not go through the details here, but refer the interested reader to[418] and
further references to be given in the end of this review.
The effective D-brane action was obtained by setting open string beta functions
to zero. Now, we have seen that the outcome can depend on the way we regularize
singularities: commutative Dirac-Born-Infeld e.g. for dimensional regularization and
noncommutative Dirac-Born-Infeld for point-splitting. From quantum field theory it
is known that beta functions which differ by the way of renormalization should be
identical up to redefinitions of the couplings. In our example the couplings are Ai in
the commutative case, and Âi in the noncommutative one. Therefore, there should
exist a field redefinition relating commutative gauge theory to noncommutative one.
Indeed, such a field redefinition has been found in[418], it is sometimes called the
Seiberg-Witten map.
The connection between D-branes and noncommutative field theory has many in-
teresting aspects, which we will, however not further discuss in this review.
2. Orientifold fixed planes 98
µ
X µ = XR σ − + XLµ σ + , (2.4.1.1)
with
µ 1 µ 1 µ − i X 1 µ −inσ−
XR = x + p σ + α e , (2.4.1.2)
2 2 2 n n
n6=0
1 µ 1 µ + i X 1 µ −inσ+
XLµ = x + p σ + α̃ e . (2.4.1.3)
2 2 2 n n
n6=0
µ
X −
ψ− = dµn e−inσ , (2.4.1.4)
n∈Z
µ
X +
ψ+ = d˜µn e−inσ , (2.4.1.5)
n∈Z
2. Unoriented closed strings 99
µ
X −
ψ− = bµr e−irσ , (2.4.1.6)
r∈Z+ 12
µ
X +
ψ+ = b̃µr e−irσ (2.4.1.7)
r∈Z+ 12
in the NS sectors.
We define an operator Ω which changes the orientation of the worldsheet. For the
closed string the action of Ω is
Ω : σ ↔ −σ. (2.4.1.8)
For left handed fermionic modes, we introduce an additional sign such that the product
of a left with a right handed fermionic mode is Ω invariant (recall that fermionic modes
from the left moving sector anti-commute with fermionic modes from the right moving
sector). In formulæ, this means
From this we see that Ω is a symmetry in type IIB theory – the only closed superstring
which is left-right symmetric. (Note that the GSO projection operator (2.1.2.59) in
the R sector contains an even number of d0 ’s. Hence, the sign in the transformation
+ +
(2.4.1.9) cancels out and e.g. PGSO is interchanged with P̃GSO .)
Let us study the action of Ω on the massless sector of type IIB excitations. We
take the vacuum to be invariant under worldsheet parity reversal. The massless NSNS
sector states are (in light cone gauge)
The action of Ω on this state interchanges the indices i and j. Thus the states surviv-
ing an Ω projection are symmetric in i, j – these are the graviton Gij and the dilaton
Φ. Since Ω relates the NSR with the RNS sector only invariant superpositions are
kept. Thus we obtain only one gravitino (56 components) and one dilatino (8 compo-
nents). Half of the target space supersymmetry is broken by the Ω projection. The
massless states in the RR sector are obtained from the tensor product of the left with
the right moving R vacuum. The R vacua are target space spinor components and
Ω interchanges the left with the right moving vacuum. Because spinor components
anti-commute the antisymmetrized tensor product survives the Ω projection. This is
′ . We obtain
the 28 dimensional SO(8) representation – the antisymmetric tensor Bij
2. Unoriented closed strings 100
the field content of the heterotic string without the internal fermions λA
+ . As we stated
before, a theory with such a massless spectrum suffers from gravitational anomalies. In
the heterotic theory we actually needed 32 worldsheet fermions λA + whose quantization
provided exactly the gauge multiplets needed to obtain an anomaly free massless spec-
trum. Later we will see that one needs to add D-9-branes to the unoriented type IIB
theory, for consistency. Before going into that let us study for a while the unoriented
closed stringtheory – even though it is not consistent yet.
The theory of unoriented type IIB strings contains orientifold-nine-planes – or short
O-9-planes. An O-plane is a set of target space points which is fixed under an element
of a discrete group which contains Ω (the element must contain Ω). Because Ω alone
does not act on the target space geometry the full target space is fixed under Ω. The
fixed set of points is space filling – it is an O-9-plane.
We have seen that when we compactify the type IIB string on a circle and perform
a T-duality we obtain type IIA theory compactified on a circle with inverted radius.
Let us study what happens to the O-9-planes in this process. Formally, we have the
expression (X 9 stands for the bosonic string coordinate)
T-DUALITY −1
ΩX 9 Ω−1 −→ T ΩT −1 T X 9 T −1 T ΩT −1 . (2.4.1.11)
T Ω T −1
XL9 + XR
9
−→ XL9 − XR
9 9
−→ XR 9
− XL9 −→ −XR − XL9 (2.4.1.12)
Thus we see that T ΩT −1 reflects the dimension in which T acts, and also interchanges
left with right movers (the second statement can be easily verified by drawing the
diagram (2.4.1.12) for the left or right moving piece alone). Thus, for T-duality in X 9
direction we can write
T ΩT −1 = R9 Ω, (2.4.1.13)
R9 : X 9 → −X 9 . (2.4.1.14)
The action on the worldsheet fermions can be studied likewise. Now we go to the
decompactification limit on the type IIA side. Instead of an O-9-plane we have an
O-8-plane, because now only points with X 9 = 0 are fixed under the action of ΩR9 .
Repeating this argumentations for more than one T-dualized circle we conclude that
2. O-plane interactions 101
1 + 21 =
2
Figure 2.12: The superposition of two strings with opposite orientation can be viewed
as a crosscap. The crosscap is a circle with diagonally opposite points being identified.
we have O-p-planes with even (odd) p in type IIA (B) theory. For an O-p-plane with
even p, Ω comes combined with a Z2 operator reflecting an odd number of dimensions.
+ −
In particular, this combination interchanges e.g. PGSO with P̃GSO , i.e. it is indeed a
symmetry of type IIA strings. The closed string is unoriented only when it is located
on an O-plane. A string off the O-plane is oriented. Its counterpart with the opposite
orientation is the R9 image of the string.
In figure 2.13 we have drawn a process in which two O-planes interact via the exchange
of closed strings. We restrict to the special case that the orientifold group element ΩR
squares to one. (Combining orbifold compactifications with orientifolds, one can have
the more general situation that the orientifold group elements square to a nontrivial
orbifold group element. This has to be the same for the two O-planes. Then a twisted
sector closed string is exchanged.)
In the following we are going to compute this process. As in the D-brane compu-
tation, we take the O-planes to be parallel. The range for the worldsheet coordinates
is
Oi Oj
Figure 2.13: O-plane Oi and O-plane Oj talking to each other by exchanging a closed
string.
Figure 2.14: The triangulated version of fig. 2.13 on the left. By manipulations
preserving the topology this can be mapped onto the triangulated version of a Klein
bottle on the right.
Like in section 2.3.2 we want to perform the computation in the scheme where the
role of τ and σ are interchanged – i.e. in the worldsheet dual channel. In this dual
channel, a virtual pair of closed strings pops out of the vacuum – one of the strings
changes its orientation before they rejoin. Therefore, this is called the loop channel.
Before performing the transformation to the loop channel, we need to describe the tree
channel process fig. 2.13 such that it is periodic in time. The method of doing so differs
slightly from the D-brane/D-brane interaction. It is best explained by looking at the
triangulated version of picture 2.13 and its double cover which is a torus. We draw this
in figure 2.14. In the left picture, the shaded region is the triangulated version of figure
2.13. The half-circles indicate the identifications of the crosscaps. The white region
shows our intention to obtain a description which is periodic in τ , with a period 4πl.
Now, one cuts the shaded part along the dotted line (with the indicated orientation),
and flips the upper rectangle once around its right vertical edge and afterwards shifts
it down in the vertical direction. We obtain a process which is indeed periodic in τ ,
and now τ ∈ [0, 4πl). (This periodicity appears due to the crosscap identifications
2. O-plane interactions 103
indicated by the left half circles. The crosscap identifications for the other O-plane
ensure that one can glue the upper rectangle to the lower one after the flip and the
shift.) It is difficult to describe this process as a tree channel closed string exchange.
Instead we can interchange the roles of σ and τ . Then the interpretation is that a pair
of closed strings of length 4πl pops out of the vacuum, one of the closed strings changes
its orientation before they annihilate after a worldsheet time π. This is a vacuum loop
amplitude which has the topology of the Klein bottle. The parameter range is
As in the D-brane case (section 2.3.2) we want to rescale the dualized worldsheet
coordinates such that their ranges are the canonical ones, which (now for the closed
string) are
H → 2l (2t)2 H. (2.4.2.5)
Analogous to the annulus discussion in section 2.3.2 we require that the action of the
rescaling on the time evolution operator is
detail here. We take the parameter range (2.4.2.3). We are interested in the behavior
of worldsheet fermions under shifts in τ by 4πt. Periodic behavior corresponds to tree
level RR exchange whereas anti-periodicity translates to NSNS exchange. Fig. 2.14
tells us how to continue in τ beyond 2πt
µ µ
ψ± (τ + 4πt, σ) = (−)F +F̃ ψ± (τ + 2πt, 2π − σ) , (2.4.2.8)
where the (−)F +F̃ reflects the boundary condition on worldsheet fermions under 2πt
shifts in τ . However, in the Klein bottle amplitude only states with (−)F = (−)F̃
contribute because of an Ω insertion in the trace over states. Therefore, the additional
factor in (2.4.2.8) is not relevant. Now, the 2πt shift in τ can be replaced by acting
with the trace insertion (−)F Ω. The (−)F insertion just cancels a sign included in
the definition of the trace over fermions for the right movers. By the same token we
have to insert a (−)F̃ for the left movers. Thus we obtain
µ µ µ
ψ± (τ + 4πt, σ) = (−)F̃ Ωψ± (τ, 2π − σ) Ω−1 = ψ± (τ, σ) , (2.4.2.9)
where in the last step we used our definition for Ω (2.4.1.9) and the 2π periodicity in
σ. Thus the (−)F insertion in the loop channel filters out the RR tree level exchange,
indeed. Strictly speaking the above consideration is correct only when the fermions
point in directions longitudinal to the O-plane (where the Z2 reflection R acts as the
identity). For the other directions there are two signs canceling each other and leading
to the same result. At first, there is an additional minus sign in (2.4.2.8) because the
half-circles in fig. 2.14 now contain also the (non-trivial) action of the Z2 reflection R.
This sign is canceled when we replace Ω by ΩR in (2.4.2.9).
We want to filter out the contribution due to RR exchange in the tree channel.
Then, the loop channel vacuum amplitude is given by the following expression
Z !
dt (−)F −2πα′ tH
Str ΩR e =
2t 2
Z
dt
tr ZERO ΩRe−2πα tH0 trBOSONS ΩRe−2πt(HB +H̃B )
′
2t MODES
!
(−)F −2πt(HNS +H̃NS )
tr NSNS ΩR e . (2.4.2.10)
FERMIONS 2
Here, we split the Hamiltonian into right and left moving parts H + H̃ and these in
turn into
α′ H = α′ H0 + HB + HN S (2.4.2.11)
2. O-plane interactions 105
with54
p2
H0 = (2.4.2.12)
4 !
8 ∞
X X 1
HB = αi−n αin − , (2.4.2.13)
24
i=1 n=1
8 ∞
X X 1
rbi−r bir − ,
HN S = (2.4.2.14)
48
i=1
1
r= 2
and the corresponding expressions for the left moving sector. The ΩR insertion
projects out contributions of states with zero mode momenta perpendicular to the
O-planes, since those states are mapped onto states with the negative momentum
in the perpendicular direction by the ΩR insertion. The result for the zero mode
contribution reads
Z
dp+1 k −2πα′ t k2 p+1 p+1
′ 2 − 2
tr = 2Vp+1 2 = 2· 2 2 V
ZERO
p+1 e p+1 8α π t . (2.4.2.15)
MODES (2π)
Also here there is an additional factor of two, due to the possible orientations of the
closed string. (The trace is taken over oriented strings with an Ω insertion. Another
point of view would be that one needs to add to the picture on the rhs of figure 2.14 a
picture with reversed orientations on the horizontal edges.) For the traces over excited
states we note that the insertion ΩR in the trace means that only states contribute
which are eigenstates of ΩR. This means that the left moving excitations have to be
identical to the right moving ones. Thus, it is straightforward to modify the calculation
presented in section 2.3.2 by just changing the power of the arguments in the functions
2.3.2.20 by two (since the identical left and right moving contributions add). We obtain
1
trBOSONS ΩRe−2πt(HB +H̃B ) = 8 −2πt (2.4.2.16)
f1 (e )
for the trace over the bosonic excitations, and
!
(−)F −2πt(HNS +H̃NS ) 1
tr NSNS ΩR e = − f48 e−2πt . (2.4.2.17)
FERMIONS 2 2
Thus we obtain
Z !
dt (−)F −2πtH
Str ΩR e =
2t 2
Z
1 p+1 dt p+1 f 8 e−2πt
− 4
− 2 2 Vp+1 8α′ π 2 t 2
. (2.4.2.18)
2 2t f18 (e−2πt )
54
From our treatment in section 2.1.2.2 we would get a factor of α′ 8/2 = 2 instead of α′ 4/2 = 1 in
the oscillator contributions. Recall, however, that we have changed the σ range from [0, π) to [0, 2π),
meanwhile. We have distributed the zero mode contribution symmetrically on H and H̃. Taking into
account the effect of rescaling, this gives the factor of 1/4.
2. O-plane interactions 106
We see that the result has almost the same structure as the one we obtained for the
D-brane/D-brane interaction in table 2.7. (Recall that now, we separated out the RR
sector exchange.) The differences are that we do not have the exponential dependence
on the distance and that we do have an additional factor of 2p+1 . The explanation for
the missing exponent is very simple. Since the orientifold planes are all located at a
fixed point of the Z2 action R, they cannot be separated in target space. (However,
we could for example compactify the dimensions transverse to the brane. In that case
winding modes would play the role of the distance.)
Before we can deduce the ratio of the O-plane RR to the D-brane RR charge, we
need to discuss a subtlety appearing because we have modded out reflections in the
transverse directions. This has the effect that each transverse direction is “half as
long” as in the D-brane computation. The implications of this effect are best seen
in a field theory consideration. The field theory result gives a “Coulomb potential”
which is of the structure charge-squared times density. (The density appears as the
inverse of a second order differential operator.) The charge is obtained as an integral
R
over the transverse space (analogous to Q = d3 xj 0 in electro-dynamics). In the
O-plane case this gives a factor of a half per transverse direction as compared to the
D-brane/D-brane interaction. On the other hand the density is multiplied by a factor
of two per transverse direction. Hence, the overall net-effect of this transformations is
an additional factor of 2p−9 which we need to put by hand into the O-plane/O-plane
result, before we can compare it with the D-brane/D-brane calculation.55 Taking this
into account, the ratio of the D-p-brane RR charge µp to the O-p-plane RR charge µ′p
We cannot fix the sign by the present calculation since the charges enter quadratically
the expressions we derived so far. Computing also the contributions without the (−)F
insertions to the Klein bottle, one obtains the square of the O-plane tension. Here, we
infer the result by supersymmetry instead. Since the ΩR projection leaves half of the
55
If we performed a detailed field theory calculation we would find this factor due to the different
target spaces (as argued in the text). Later, we will compactify the transverse dimensions. Then this
factor will appear “automatically” due to a Poisson resummation of the sum over the winding modes.
This must be the case since in the compactified theory D-branes and O-Planes will have the same
transverse space.
2. O-plane interactions 107
Di Oj
Figure 2.15: D-brane Di and O-plane Oj talking to each other via the exchange of
closed strings.
supersymmetries unbroken, the total one loop amplitude should vanish. This tells us
that the ratio of the D-brane tension Tp to the O-plane tension Tp′ is
where at the present stage of the calculation the sign is not known. In order to fix the
signs in (2.4.2.20) and (2.4.2.21) we need to study the interaction between D-branes
and O-planes. We will do so in the next subsection.
So far, we have seen that D-branes as well as O-planes interact via the exchange of
closed type II strings. This suggests that also D-branes interact with O-planes. Such
a process is depicted in figure 2.15. We consider the case of parallel D-branes and O-
planes. This implies that the D-brane is located in directions where the Z2 reflection
acts with a sign and extended along the other directions.
Again, the range for the worldsheet coordinates is
Di
Di Di
Figure 2.16: The triangulated version of fig. 2.15 on the left. By manipulations pre-
serving the topology this is mapped onto the triangulated version of a Möbius strip
on the right.
of the open strings changes its orientation before they finally annihilate. The topology
of this diagram is called Möbius strip (or Möbius band).
The range for the worldsheet coordinates after interchanging the role of time and
space is
H → 4l (2t)2 H. (2.4.2.26)
Finally the time evolution operator should take its canonical form
!
e−πH → e−2πtH . (2.4.2.27)
where the factor of (−)F appears due to the anti-periodic boundary condition of world-
sheet fermions under shifts of 2πt in τ . Let us study the case where we insert in the loop
channel trace just Ω (possibly combined with some target space Z2 action which we
will discuss below). Taking into account the sign when a trace is taken over worldsheet
fermions we obtain
µ µ
ψ± (τ + 4πt, σ) = Ωψ± (τ, π − σ) Ω−1
µ
= Ωψ∓ (τ, σ − π) Ω−1 , (2.4.2.30)
where in the second step we have used the mode expansions (2.3.1.22)–(2.3.1.29). In
the open string sector we define the action of Ω as taking σ → π − σ. This finally
results in
µ µ
ψ± (τ + 4πt, σ) = ψ∓ (τ, 2π − σ)
µ
= ψ± (τ, σ − 2π) , (2.4.2.31)
where once again the mode expansion has been used. We deduce that open string R
sector contributions correspond to closed string RR exchange. (This can also easily
be seen in the mode expansions (2.3.1.22)–(2.3.1.29)). The above consideration is
correct only in NN directions (in directions in which the D-brane extends). For DD
directions there are a couple of signs which cancel each other such that one gets the
same result. Since the Z2 reflection R acts with a sign in those directions, the first
line in (2.4.2.30) receives an additional minus sign. Looking at the mode expansion
(2.3.1.22)–(2.3.1.29) in DD directions we observe that this sign is canceled when going
to the second line in (2.4.2.30). Because now we need to replace Ω by ΩR, the first
line in (2.4.2.31) contains an additional minus sign which again is canceled by using
the DD mode expansion when going to the second line in (2.4.2.31).
In the above consideration we have only specified how Ω acts on the oscillators,
and not how it acts on the vacuum. (In the closed string we tacitly took the vacuum as
being invariant under Ω leaving in the NSNS sector the metric invariant and projecting
out the B-field.) The computations of the D-brane/D-brane and the O-plane/O-plane
interactions provided the absolute values of the corresponding RR charges. The result
for the D-brane/O-plane calculation will give the product of the O-plane times the
D-brane charge. This should be compatible with our previous result. As we will see
in a moment this leaves the two choices that the ΩR eigenvalue of the open string
R-vacuum is ±1. We will choose the minus sign. This corresponds to a D-brane. The
action on the NS sector can be inferred by supersymmetry, i.e. it should be such that
the complete one loop Möbius strip amplitude vanishes. The result is that the massless
states have eigenvalue minus one. (This holds as well for Neumann directions as for
2. O-plane interactions 110
Dirichlet directions, since a sign due to the different mode expansions cancels a sign
due to the non-trivial action of R in Dirichlet directions.)
Now, we have collected all the necessary information needed to write down the
loop channel amplitude which gives the tree channel RR exchange (recall that a (−)F
insertion leads to a vanishing R sector trace)
Z
dt 1 ′
− trR ΩR e−2πα H =
2t 2
Z
dt ′
− tr ZERO ΩRe−2πtα H0 trBOSONS ΩRe−2πtHB
2t MODES
1 −2πtHR
tr R ΩR e . (2.4.2.32)
FERMIONS 2
The expressions for the Hamiltonians can be directly taken from (2.3.2.14)–(2.3.2.18)
with the difference that we put y = 0 in (2.3.2.15) (because of the ΩR insertion in
the trace only D-branes at distance zero from the O-plane contribute). With this
difference the trace over the zero modes gives (see (2.3.2.19)
Z
dp+1 k −2πtα′ k2 p+1
2 ′ − 2
tr ZERO = 2Vp+1 e = 2V p+1 8π α t , (2.4.2.33)
MODES (2π)p+1
The next step is to split the product over integers in the definition of f1 (2.3.2.20) into
a product over even times a product over odd numbers. This gives finally
1
trBOSONS ΩRe−2πtHB = . (2.4.2.36)
f18 (e−2πt ) f38 (e−2πt )
Manipulations analogous to the bosonic trace give the result (recall that we have
chosen the ΩR eigenvalue of the R vacuum to be minus one)
ΩR −2πtHR
−tr R e = f28 e−2πt f48 e−2πt , (2.4.2.38)
FERMIONS 2
2. O-plane interactions 111
where the 16-fold degeneracy of the R vacuum has been taken into account. We are
interested in the contributions due to tree channel RR exchange and have computed
now everything we need to obtain the result. However, in order to specify the action
of ΩR on the NS vacuum one needs to compute the tree channel NSNS exchange. The
requirement that this cancels the RR interaction determines the action of ΩR on the
open string NS vacuum. We leave this as an exercise. The result is that the massless
vector is odd under ΩR. In the computation of the open string NS sector trace it is
useful to apply the identity (2.3.2.21) on the f functions with the shifted arguments
and afterwards to proceed as we did above, i.e. to split the product in the definitions
of the f ’s into a product over even and over odd numbers.
So far, we obtained the result
Z Z
dt 1 −2πα′ H dt p+1 8 −2pit f48 e−2πt
2 ′ − 2 f2 e
− ΩRtrR e = Vp+1 8π α t . (2.4.2.39)
2l 2 2t f18 (e−2πt ) f38 (e−2πt )
Expressing t in terms of l via (2.4.2.28) and using the properties (2.3.2.22) yields finally
the tree channel infrared asymptotics
Z
1 − p+1 p−9
2 Vp+1 dl 4π 2 α′ 2
2p−4 l 2 . (2.4.2.40)
2 l→∞
This expression has to be compared with the second line (RR contribution) of table
2.7 and (2.4.2.19), where (2.4.2.19) has to be multiplied with 2p−9 as discussed earlier.
(For the Möbius strip we do not need to put such a factor since there is a cancellation
between the O-plane charge and the density.) In (2.4.2.40) we have pulled out a factor
of two. If we take the D brane distance y to zero in (2.7) we can write down the
cumulative infrared asymptotics for a system consisting out of one D-brane and one
O-plane (situated at the origin in the transverse space)
Z
− p+1 p−9 2
−Vp+1 4π 2 α′ 2
dll 2 1 − 2p−4 . (2.4.2.41)
l→∞
2
In field theory one obtains a result proportional to µp + µ′p , (recall that µp is the
D-brane charge and µ′p the O-plane charge). Thus we obtain finally the ratio between
D-brane and O-plane RR charges
The Möbius strip computation fixed also the sign of this ratio. However, if we assigned
an ΩR eigenvalue of +1 to the open string R vacuum we would obtain an additional
minus sign in (2.4.2.42). There is an ambiguity here. In the next section we will use
our results to construct consistent string theories containing D-branes and O-planes.
In this construction this ambiguity cancels out. (In some sense it will turn out that our
2. Compactifying the transverse dimensions 112
present choice is the “natural” one.) The ratio of the D-brane tension to the O-plane
tension can be inferred by supersymmetry
where jp is the external U (1) current indicating the presence of the D-brane (O-plane).
Integrating this equation over a compact transverse space gives zero for the left hand
side and the D-brane (O-plane) charge on the right hand side. Therefore, the RR
charge on the rhs has to vanish. To overcome this problem one may want to add
D-branes and O-planes such that the net RR charge is zero. Since one needs more
than one D-brane in order to achieve a vanishing net RR charge, one has to specify
how ΩR acts on a set of multiple D-branes. For example it could (and actually will)
happen that ΩR (anti)symmetrises strings starting and ending on different D-branes.
Technically, we define a (projective) representation of the Z2 (generated by ΩR) on the
Chan-Paton labels carried by open string in case of multiple D-branes. The generating
element of this representation is denoted by γΩR . The ΩR action on an open string is
−1
ΩR : |ψ, iji → (γΩR )ii′ ΩR (ψ) , j ′ i′ γΩR j′j
. (2.4.3.2)
Here, ψ denotes the oscillator content of the string on which ΩR acts in the same way
as discussed previously. In addition, the order of the Chan-Paton labels is altered due
to the orientation reversal. Acting twice with ΩR should leave the state invariant.
This leads to the condition
T
γΩR = ±γΩR , (2.4.3.3)
i.e. γΩR is either symmetric or antisymmetric. By a choice of basis this gives the two
possibilities
!
0 iI
γΩR = I or γΩR = . (2.4.3.4)
−iI 0
Let N be the number of D-branes (ΩR images are counted). Then I denotes an N × N
N N
identity matrix for symmetric γΩR and an 2 × 2 identity matrix for antisymmetric
γΩR .
2. Compactifying the transverse dimensions 113
The trace in the open string amplitudes (annulus and cylinder) includes also a
trace over the Chan-Paton labels. For the annulus this is simply
N
X N
X
hij| iji = δii δjj = N 2 . (2.4.3.5)
i,j=1 i,j=1
In the Möbius strip amplitude the trace over the Chan-Paton labels yields the addi-
tional factor
N
X
−1 T
hij| ΩR |iji = tr γΩR γΩR = ±N, (2.4.3.6)
i,j=1
In the following we are going to investigate the case where the compact space is a torus.
The next issue we need to discuss are zero mode contributions due to windings in
the compact transverse dimensions. For open strings windings can appear in Dirichlet
directions. Since ΩR leaves the winding number of a state invariant these contribute to
the annulus, the Klein bottle and the Möbius strip. Including the sum over the winding
numbers into the corresponding traces leads to additional factors. The transverse space
is a 9 − p–torus:
T 9−p = S 1
· · × S }1 .
| × ·{z (2.4.3.7)
9−p factors
For simplicity we take the radii of these S 1 s to be identical and denote them by r. It
is useful to introduce a dimensionless parameter
r2
ρ= (2.4.3.8)
α′
for the size of the compact space.
With this ingredients the sum over the winding modes gives the following factors
dt
(under the 2t integral):
∞
!9−p
X
−2πtρw 2
e for the annulus, (2.4.3.9)
w=−∞
∞
!9−p
X
−πtρw 2
e for the Klein bottle, (2.4.3.10)
w=−∞
∞
!9−p
X
−2πtρw 2
e for the Möbius strip. (2.4.3.11)
w=−∞
2. Compactifying the transverse dimensions 114
In the annulus we have restricted ourselves to the special case that all D-branes are
situated at the same point. This configuration gives the correct leading infrared contri-
bution to tree channel amplitude. One can also include distances among the D-branes
into the computation. In that case the trace over the Chan-Paton labels cannot be
directly taken as in (2.4.3.5) because the zero mode contribution depends on the Chan-
Paton label. Taking the infrared limit in the tree channel removes this dependence
on the Chan-Paton labels and gives the same result as our slightly simplified compu-
tation.56 Now, we express t in terms of l using (2.3.2.8), (2.4.2.7) and (2.4.2.28) and
apply the Poisson resummation formula
∞ ∞
X (n−b)2 √ X 2 +2πisb
e−π a = a e−πas . (2.4.3.12)
n=−∞ s=−∞
We obtain
∞
!9−p
X 2
9−p p−9 − πlw
l 2 ρ 2 e ρ for the annulus, (2.4.3.13)
w=−∞
∞
!9−p
X 2
9−p p−9
9−p − 4πlw
l 2 ρ 2 2 e ρ for the Klein bottle, (2.4.3.14)
w=−∞
∞
!9−p
X 2
9−p p−9
9−p − 4πlw
l 2 ρ 2 2 e ρ for the Möbius strip. (2.4.3.15)
w=−∞
In the IR limit l → ∞ the sums become a factor of one. The common ρ dependent
factor is a dimensionless quantity representing the volume of the transverse space
(sometimes denoted by v9−p ). In the Klein bottle as well as in the Möbius strip
there is an additional factor of 29−p . In the Möbius strip this is simply the number
of O-planes. (The number of R-fixed points is two per S 1 .) Since the Klein bottle
amplitude is proportional to the square of the total O-plane charge we would expect
another factor of 29−p here. However, this is “canceled” by the correction factor we
put earlier in by hand. As promised in footnote 55 this factor appeared automatically
after we compactified the transverse dimensions. This is good because now we would
miss the argument for putting it in by hand.
Together with our previous results table 2.7, (2.4.2.19) and (2.4.2.40) we obtain
for the infrared limit of the total (tree level RR channel) amplitude
Z
1 p+1
′ 2 − 2 p−9
− 4α π Vp+1 ρ 2 dl N 2 + 322 ∓ 2N 32 , (2.4.3.16)
2 l→∞
where the ∓ sign is correlated with the ± sign in (2.4.3.3). We observe that the contri-
butions of the D-brane/D-brane, O-plane/O-plane and D-brane/O-plane interaction
56
In the limit l → ∞ the distance dependent exponential function in table 2.7 becomes one.
2. Compactifying the transverse dimensions 115
(32 ∓ N )2 . (2.4.3.17)
For consistency the total RR charge has to vanish. Thus, we are lead to the conditions
T
γΩR = γΩR (2.4.3.18)
and
N = 32. (2.4.3.19)
Note that the condition (2.4.3.18) is related to our choice that the ΩR eigenvalue of
the R vacuum (in the open string sector) is minus one. Since (2.4.3.18) implies that
we can choose a basis such that
γΩR = I, (2.4.3.20)
our choice of the ΩR eigenvalue of the R vacuum seems natural. The case p = 9
can be obtained from our previous considerations in two ways. The simplest is to set
p = 9 in (2.4.3.16). Requiring this expression to vanish yields equations (2.4.3.18)
and (2.4.3.17) independent of p.57 An alternative way is to perform T-dualities with
respect to all compact directions and to take the decompactification limit in the T-dual
model. Both methods lead to the same results. The massless closed string spectrum
has been discussed in section 2.4.1. In the NSNS sector one finds the metric Gij and
′ . The Ω invariant
the dilaton Φ. The RR sector provides the antisymmetric tensor Bij
combinations of the NSR with the RNS sector massless states yield the space time
fermions needed to fill N = 1 supermultiplets.
It remains to study the open string sector. The massless NS sector states are
i mn
ψ− 1 |k, mni λ . (2.4.3.21)
2
i T
mn
−ψ− 1 |k, mni λ . (2.4.3.22)
2
λ = −λT . (2.4.3.23)
The states (2.4.3.21) are vectors and thus should be interpreted as gauge fields of
a certain gauge group. In order to identify the gauge group, we require that the
57
Recall that a p-dependence cancels in the product of O-plane charges with the number of O-planes.
2. Compactifying the transverse dimensions 116
Table 2.8: Consistent string theories in ten dimensions containing open strings.
state (2.4.3.21) transforms in the adjoint and that gauge transformations preserve
the condition (2.4.3.23). Thus the commutator of a generator of the gauge group
with a 32 × 32 antisymmetric matrix (λ) should be a 32 × 32 antisymmetric matrix.
This consideration determines the gauge group to be SO(32). The R sector provides
fermions in the adjoint of SO(32). NS and R sector together yield an N = 1 SO(32)
vector multiplet. The list of consistent closed string theories in ten dimensions (table
2.1) is supplemented by the ten dimensional theory containing (unoriented) closed
strings and open strings in table 2.8.
−1
R: |ψ, iji → (γR )ii′ R (ψ) , i′ j ′ γR j ′j
. (2.4.3.24)
(9) (5)
Also, the γR can be split into two blocks: γR and γR . The requirement that perform-
ing twice the same Z2 action should leave the state invariant leads to the conditions
that every gamma-block containing an Ω in the subscript must be either symmetric
or anti-symmetric, whereas the gamma-blocks without an Ω in the subscript must
square to the identity58 . At this point, we need to discuss a subtlety of the five-nine
sector, i.e. strings with ND boundary conditions along the compact dimensions. In
that sector the Fock space state (without the Chan-Paton label) has an Ω2 and an
(ΩR)2 eigenvalue of minus one. Unfortunately, we did not develop the techniques
needed to show this, in this review. An argument employing an isomorphism between
the algebra of vertex operators and Fock space states can be found in[204]. Since Ω2
and (ΩR)2 should leave states invariant, this minus sign needs to be canceled by an
(9)
appropriate action on the Chan-Paton labels. For example a symmetric γΩ implies
(5) (5) (9)
an anti-symmetric γΩ , and a symmetric γΩR implies an anti-symmetric γΩR .
Let us now study the amplitudes in the loop channel. For strings starting and end-
ing on D-9-branes there is a tower of Kaluza-Klein momentum modes but no winding
modes. The D-9-branes are space filling and thus must lie on top of each other. Open
58
We fix a possible phase to one.
2. Compactifying the transverse dimensions 118
strings with both ends on a D-5-brane can have winding modes and can be separated
in the compact directions. (In our previous computation we did not consider this sep-
aration, because it is not relevant in the tree channel infrared limit. In order to see
explicitly that the dependence on the distance among D-5-branes drops out, we will
take it into account, here. We call the component of the position of the ith brane cai .)
Further, all amplitudes obtain an additional insertion
1+R
, (2.4.3.25)
2
ensuring that we trace only over states which are invariant under the orbifold group.
For terms containing an R insertion the D-5-brane must be located at R fixed points,
since otherwise the states in the (55) sector are not eigenstates under R. For the same
reason the winding or Kaluza-Klein momentum modes have to vanish in the presence
of an R insertion. Further, there will be additional signs for oscillators pointing into
a compact dimension. This modifies the oscillator contribution to the trace in a
straightforward way. (We leave the details as an exercise.) Taking into account all
these effects and the discussion in section 2.3.2 one finds for the annulus amplitude A
!4
Z 8 e−πt
∞
V6 dt f
2 ′ −3 4
(9) 2
X − 2πtn
2
A = − 8π α tr γ1 e ρ
4 t4 f18 (e−πt ) n=−∞
√
9 ∞ a −ca 2
X (5) (5) Y X ( 2πw ρα′ +c i j)
+ γ1 γ1 e−t 2πα′
ii jj
i,j∈5 a=6 w=−∞
16 !
f24 e−πt f44 e−πt X (5)
(9)
−2 4 −πt 4 −πt tr γR,I tr γR
f1 (e ) f3 (e )
I=1
4 −πt
4 −πt ( 16
)#
f3 e f4 e
(9) 2
X (5) 2
+4 4 −πt 4 −πt tr γR + tr γR,I , (2.4.3.26)
f1 (e ) f2 (e )
I=1
where we have formally assigned a gamma with subscript 1 to the action of the identity
element of the orbifold group on the Chan-Paton labels. The sum over i, j ∈ 5 means
that we sum over all Chan-Paton labels belonging to an open string end on a D-5-brane.
The index I = 1, . . . , 16 labels the fixed 5-planes, and a corresponding subscript at a
γ (5) indicates that the D-5-brane is located on the Ith fixed plane.
Next, we want to compute the Klein bottle amplitude K. It contains the insertions
Ω and ΩR. In principle, we have to take the trace over untwisted and twisted sector
states (with the (−)F insertion). Because half of the RR sector states have the opposite
(−)F eigenvalue than the other half, RR sector states do not contribute to the trace
with a (−)F insertion. The same applies to RR and NSNS twisted sector states.
Eigenstates of Ω have zero winding numbers whereas for eigenstates of ΩR the Kaluza-
2. Compactifying the transverse dimensions 119
Finally, for the Möbius strip amplitude M we need to trace over R sector states
with an Ω + ΩR insertion. Eigenstates correspond to open strings starting and ending
on the same brane. According to our earlier assignments, the ΩR eigenvalue of the R
vacuum corresponding to a string ending on a D-5-brane is minus one, and so is the
Ω eigenvalue of the R vacuum corresponding to a string ending on D-9-branes. To
determine the remaining eigenvalues one has to act with R on the Ramond vacuum.
R can be viewed as a 180◦ rotation and the R vacua as target space spinors. Hence,
half of the Ramond vacua have R eigenvalue minus one and the other half plus one.
For this reason, only D-9-branes contribute to the term with the Ω insertion whereas
only D-5-branes give a non-vanishing result for the trace containing an ΩR insertion.
The result for the Möbius strip is
Z 8
−2πt f 8 e−2πt
V6 dt 2 ′ −3 f2 e
4
M = 8π α
4 t4 f18 (e−2πt ) f38 (e−2πt )
!4
X ∞ 2
(9) −1 (9) T − 2πtn
tr γΩ γΩ e ρ
n=−∞
∞
!4
T
(5) −1
X
(5) −2πtρw 2
+tr γΩR γΩR e . (2.4.3.28)
w=−∞
With the next steps necessary to compute the total RR charge of the system we
1 1
are familiar by now. We replace t = 2l in the annulus, t = 4l in the Klein bottle
1
and t = 8lin the Möbius strip. In order to be able to read off the infrared (large l)
asymptotics we use formulæ (2.3.2.22) and (2.4.3.12). The final result is
Z
V6 3
A + K + M −→ − dl 4π 2 α′
4
l→∞
(9) 2 (9) −1
ρ2 trγ1 − 32tr γΩ (γΩ )T + 322
1 (5) 2
(5) −1 T 2
+ 2 trγ1 − 32tr γΩR (γΩ ) + 32
ρ
16
#
1 X (9)
(5) 2
+ trγR + 4trγR,I . (2.4.3.29)
4
I=1
2. Compactifying the transverse dimensions 120
The remaining representation matrices can be found by imposing that the gammas
should form a projective59 representation of the orientifold group (Z2 × Z2 ). We
simply choose
(5) (5) (5)
γR = γΩR γΩ (2.4.3.32)
(9) (9) (9)
γR = γΩR γΩ . (2.4.3.33)
where the rank of the identity matrix is 32. The antisymmetric form is
!
(5) (9) 0 iI
γΩ = γΩR = , (2.4.3.35)
−iI 0
with I being a 16 × 16 identity matrix, here. Note that our choice is consistent with
(·)
the requirement that γR squares to the identity. So far, we did not take into account
(·)
that the last term in (2.4.3.29) has to vanish. With γR being traceless this is ensured.
We have now all the ingredients needed to determine the open string spectrum. Let
us first study strings starting and ending on the D-9-branes, or in short the (99) sector.
We keep states which are invariant under each element of the orientifold group. (The
D-9-branes are fixed under each element of the orientifold group.) In the NS sector
we find massless vectors with the Chan-Paton matrix
!
(99) A S
λvector = , (2.4.3.36)
−S A
where A denotes a real antisymmetric and otherwise arbitrary 16 × 16 matrix and S
stands for a real 16 × 16 symmetric matrix. For the scalars in the NS sector one finds
!
(99) A1 A2
λscalars = , (2.4.3.37)
A2 −A1
59
“Projective” means up to phase factors, which drop out since the gamma acts in combination
with its inverse on the Chan-Paton label.
2. Compactifying the transverse dimensions 121
where the Ai are 16 × 16 antisymmetric matrices. Let us ignore the D-5-branes for
a moment and determine the gauge group and its action on the scalars in the (99)
sector. Since the vectors are in the adjoint of the gauge group, the gauge group should
be 162 dimensional. U (16) is a good candidate. Further, we know that the vector
should transform in the adjoint under global gauge transformations under which it
should not change the form specified by (2.4.3.36). Thus, we define an element of the
gauge group as
!
Ag Sg
g (9) = exp , (2.4.3.38)
−Sg Ag
where Sg (Ag ) are real anti-(symmetric) matrices with infinitesimal entries. A gauge
transformation acts on the Chan-Paton matrix as
" ! #
−1 Ag S g
λ(99) → g (9) λ(99) g(9) = ,λ . (2.4.3.39)
−Sg Ag
We observe that the vectors transform in the adjoint and the form of the Chan-Paton
matrix is preserved. Note also that g(9) is unitary and has 162 parameters. It is a
U (16) element. The U (1) subgroup corresponds to Ag = 0 and Sg proportional to the
identity. From our assignment that the Chan-Paton matrix transforms in the adjoint
of U (16) it is also clear that the Chan-Paton label i and j transform in the 16 and
16 of U (16) Thus, the scalars can be decomposed into the antisymmetric 120 + 120.
One may also explicitly check that the form of the Chan-Paton matrix for the scalars
is not altered by a gauge transformation. We leave the discussion of the fermions in
the R sector as an exercise. The result is that that half of them carry the Chan-Paton
matrix (2.4.3.36) and the other half the matrix (2.4.3.39). Altogether the (99) sector
provides a vector multiplet in the adjoint of U (16) and a hypermultiplet in the 120
+ 120.
Now, we include the D-5-branes. Here, we have to distinguish between the case
that a D-5-brane is situated at a fixed plane or not. In the first case the (55) strings
have to respect the ΩR and R symmetry, whereas in the second case these orientifold
group elements just fix the fields on the image brane. Suppose we have 2mI D-5-branes
at the Ith fixed plane. (The number of the D-5-branes per fixed plane must be even,
since otherwise they cannot form a representation of the orientifold group.) The NS
sector leads again to a massless vector and massless scalars with almost the same Chan-
Paton matrices as in the (99) sector ((2.4.3.36) and (2.4.3.39)). The only difference is
that the antisymmetric and symmetric matrices are now mI × mI instead of 16 × 16.
Hence, we obtain a vector multiplet in the adjoint of U (mI ) and a hypermultiplet in
the m mI
2 (mI -1) + 2 (mI -1).
I
2. Compactifying the transverse dimensions 122
Let 2nj D-5-branes be situated away from the fixed plane (but on top of each other).
For the (55) sector belonging to those D-5-branes we impose invariance under Ω, only.
(5)
The solution for γΩ is given in (2.4.3.35). Together with the minus eigenvalue on the
massless NS sector Fock space state, this leads to the result that the vector in the (55)
sector is an element of the U Sp (2nj ) Lie algebra in the adjoint representation. Taking
into account (part of) the R sector this is promoted to a U Sp (2nj ) vector multiplet.
The scalars together with the remaining R sector states form a hypermultiplet in the
antisymmetric nj (2nj -1) representation.
It remains to study the (95) sector. (Here, one has to take into account that
along the compact directions NS sector fermions are integer modded whereas R sector
fermions are half integer modded. This is quite similar to the twisted sector closed
string. In particular, the (95) NS sector ground state is already massless. Hence,
the NS sector does not give rise to massless vectors. We do not impose Ω or ΩR
invariance on (95) strings since they are mapped onto (59) strings by the worldsheet
parity inversion. If the considered D-5-branes are situated at one of the fixed planes
we impose R invariance. In this case, one finds in the NS sector two scalars with the
Chan-Paton matrix
!
(95) X1 X2
λ = , (2.4.3.40)
−X2 X1
where the Xi are general mi × 16 matrices. Together with the R sector this leads to
a hypermultiplet in the (16 , mI ) of U (16) × U (mI ), (the hypermultiplet is neutral
under the gauge group living on D-5-branes not situated at the Ith fixed plane). For
D-5-branes which are not a fixed plane the (95) sector provides a hypermultiplet in
the (16, 2nJ ) of U (16) × U Sp (2nj ).
Altogether we find the gauge group is
16
Y Y
U (16) × U (mI ) × U Sp (2nj ) , (2.4.3.41)
I=1 J
where j labels the D-5-brane packs away from fixed planes. In addition the total
number of D-5-branes has to be 32 (images are counted), i.e.
16
X X
2mI + 2 2nj = 32. (2.4.3.42)
I=1 j
where we have split the anti-symmetric representation of U Sp (2nj ) into its irreducible
parts and an index I, j refers to the gauge group on the D-5-brane pack at a fixed plane
and off a fixed plane, respectively. It can be checked that the effective six dimensional
theory is free of anomalies. The models belonging to different distributions of the
D-5-branes on and off fixed planes can be continuously transformed into each other.
In the field theory description this corresponds to the Higgs mechanism.
We have seen that adding to the orbifold compactification of section 2.2 O-planes
and D-branes gives a very interesting picture. Apart from the closed string twisted
sector states located at the orbifold fixed planes we obtain various fields from open
strings ending on D-branes. These D-branes can be moved within the compact direc-
tions while keeping the geometry fixed. The techniques described in this section can be
also applied to phenomenologically more interesting setups leading to four dimensional
theories. A description of such models is beyond the scope of the present review.
Chapter 3
Non-Perturbative description of
branes
3.1 Preliminaries
In the previous sections we gave a perturbative description of various extended ob-
jects: the fundamental string, orbifold planes, D-branes and Orientifold planes. The
string plays an outstanding role in the sense that field theories on the worldvolumes
of the other extended objects are effective string theories. The quantization of the
fundamental string is performed in a trivial target space (i.e. the target space metric
is the Minkowski metric and all other string excitations are constant or zero). Further,
the worldsheet topology is specified to the spherical (for closed strings) or disc (for
open strings) topology (after Wick rotating to Euclidean worldsheet signature). Our
treatment leads to a perturbative expansion in the genus of the worldsheet (see section
2.1.4). The perturbative expansion is governed by the string coupling
gs = ehΦi , (3.1.0.1)
which needs to be small. Perturbative closed string theory has an effective field theory
description which contains supergravity. How does one obtain insight into regions
where gs is large? Clearly, the perturbation theory breaks down in this case, and
indeed this region is rather difficult to study. There are, however, a few results one
can obtain also for strong couplings. Let us recall how non perturbative effects in Yang-
Mills theory can be studied. Apart from the trivial vacuum, (Euclidean) Yang-Mills
theory contains several other stable vacua, viz. the instantons. Studying fluctuations
around an instanton vacuum, one finds an additional weight factor in the path integral
which comes from the background value of the action and is of the form
n
−
e g2 ,
124
3. Universal Branes 125
where n is the instanton number and g is the Yang-Mills gauge coupling. As long as g is
small, the fluctuations around an instanton vacuum are heavily suppressed. However,
as soon as g becomes large, the suppression factor becomes large. Thus, knowing about
the instanton solutions in Yang-Mills theory gives a handle on non perturbative effects.
But how can one know, that one does not have to include strong coupling effects into
the theory before deriving the instanton solutions? The answer is that instantons are
stable, they are characterized by a topological number which cannot be changed in
a continuous way when taking g from small to large. Therefore, instantons can give
information about strongly coupled Yang-Mills theory even though they are found as
solutions to the perturbative formulation of Yang-Mills theories. States (vacua) with
such a feature are called BPS states.12
Therefore, our aim will be to find BPS states in string theory. In the low energy
limit, the various superstring theories are described by supergravities. Insights into
non-perturbative effects in string theory can be gained by finding the BPS states of
perturbative string theory. As a guiding principle, we will look for solutions to the
effective equations of motion that preserve part of the supersymmetry (i.e. are invariant
under a subset of the supersymmetry transformations). Roughly speaking, it is then
the number of preserved supersymmetries which cannot be changed continuously when
taking the string coupling from weak to strong. We will see that such solutions can
be viewed as branes. The number of branes takes the role of the instanton number
in the Yang-Mills example discussed above. We will be very brief in our analysis and
essentially only summarize some of the important results. The classical review on
branes as supergravity solutions is[152] and we will give more references in the end of
this review.
In the present section we will truncate all closed string effective field theories to (the
supersymmetric extension of) (3.2.0.1). This is consistent because we will restrict on
backgrounds where the discarded part of the action vanishes and the corresponding
equations of motion are satisfied trivially. By adding appropriate terms including
1
There are also stable non BPS states (for reviews see e.g. [424, 319, 195]). We will not discuss
these.
2
BPS stands for the names Bogomolny, Prasad and Sommerfield, and refers to the papers[76, 379].
3. Universal Branes 126
where Gµν is the string frame metric and gµν is the Einstein frame metric. The action
(3.2.0.1) takes the form
Z
1 10 √ 1 2 1 −Φ 2
SE,univ = 2 d x −g R − (∂Φ) − e H . (3.2.0.3)
2κ 2 12
We observe that (3.2.0.3) starts with the familiar Einstein Hilbert term (therefore the
name “Einstein frame”). Further, the kinetic term of the dilaton has the “correct”
sign now, and the coupling of the B field is Φ dependent. In the supersymmetric
extension, a gravitino and a dilatino are added. We do not give the supersymmetric
action explicitly. For us, it will suffice to know the supersymmetry transformations of
the gravitino and the dilatino. These are
1 −Φ
δψµ = Dµ ǫ + e 2 Γµ νρκ − 9δµν Γρκ Hνρκ ǫ, (3.2.0.4)
96
1 µ 1 Φ
δλ = − √ Γ ∂µ Φǫ + √ e− 2 Γµνρ Hµνρ ǫ, (3.2.0.5)
2 2 24 2
where ψµ denotes the gravitino and λ the dilatino. The Gamma matrices with curved
indices are obtained from ordinary Gamma matrices (16 × 16 matrices satisfying the
usual Clifford algebra in ten dimensional Minkowski space) by transforming the flat
index with a vielbein to a curved one. A Gamma with multiple indices denotes the
anti-symmetrized product of Gamma matrices. The spinor ǫ is the supersymmetry
transformation parameter.
Sometimes it is useful to formulate the theory in a slightly different way. To
this end, one adds to the action (3.2.0.3) a Lagrange multiplier term providing the
constraint of a fulfilled Bianchi identity. Calling the Lagrange multipliers Aµ1 ...µ6 ,
such a term looks like
Z
d10 x ǫµ1 ...µ10 Aµ1 ...µ6 ∂µ7 Hµ8 µ9 µ10 . (3.2.0.6)
The Aµ1 ...µ6 equation of motion yields the Bianchi identity of the B field strength
Hµνρ . However, one can alternatively solve the B field equation of motion with the
result
with
This means that we can trade the antisymmetric tensor B for a six form potential
A. Choosing an appropriate normalization for the Lagrange multiplier terms (3.2.0.6),
the effective action (3.2.0.3) in terms of the six form potential A reads
Z
1 √ 1 1 Φ 2
S̃E,univ = 2 d10 x −g R − (∂Φ)2 − e K . (3.2.0.9)
2κ 2 2 · 7!
Also in this form the action can be supersymmetrized. In terms of the six form
potential A the gravitino and dilatino supersymmetry transformations read
1 Φ
δψµ = Dµ ǫ + e 2 3Γµ ν1 ...ν7 − 7δµν1 Γν2 ...ν7 Kν1 ...ν7 ǫ, (3.2.0.10)
2 · 8!
1 µ 1 Φ
δλ = − √ Γ ∂µ Φǫ − √ e 2 Γµ1 ...µ7 Kµ1 ...µ7 ǫ. (3.2.0.11)
2 2 2 · 2 2 · 7!
In the following two subsections, we will present two solutions preserving half of
the supersymmetry.
only. Here, we have taken the second step before the first one in the sense that
first we should have thought about what kind of isometries we would like to obtain
and only afterwards we should have written down a general ansatz respecting the
isometries. Therefore, let us perform the first step now and discuss the isometries of
the ansatz. Clearly, there is an SO (1, 1) isometry acting on the xi . This means, that
up to Lorentz boosts, xi spans the worldvolume of a straight static string. There is no
further dependence on xi in the ansatz because we do not wish to distinguish a point
on the worldvolume of the string. The other isometry acts as an SO (8) on the y a .
This is the natural extension of the SO (3) isometry associated with non-rotating four
dimensional black holes. It is SO (8) now because the space transverse to the string is
eight dimensional (whereas in 4d black holes the space transverse to the hole is three
dimensional). The r dependence respects the SO (8) isometry. Distinguishing between
different values of r means specifying the position of the string, i.e. r measures the
radial distance from the string.
In order not to spoil the above symmetries, we choose for the remaining fields the
ansatz
where C is also a function of r only. All other components of B are zero. Viewed as a
two form, B is proportional to the invariant volume form of the string worldvolume.
The factor eC may depend on r.
The ansatz for the target space spinors is that they all vanish. As mentioned earlier
we are interested in situations where the solution preserves part of the supersymmetry
because this ensures that we can continuously take the string coupling from weak to
strong. In particular, the unbroken supersymmetry is parameterized by spinors ǫ for
which the gravitino and dilatino values of zero do not change under supersymmetry
transformations, i.e. those ǫ for which the rhs of (3.2.0.4) and (3.2.0.5) vanish. In order
to find such solutions for our ansatz it is convenient to represent the ten dimensional
Gamma matrices A, B = 0, . . . , 9,
ΓA = (γi ⊗ I, γ3 ⊗ Σa ) , (3.2.1.5)
3
The corresponding algebras are {γi , γj } = 2ηij and {Σa , Σb } = 2δab .
3. The fundamental string 129
γ3 = γ0 γ1 (3.2.1.6)
squares to the 2 × 2 identity matrix. Further, we have to take into account that the
ten dimensional N = 1 supersymmetry parameter ǫ is subject to the constraint
Γ11 ǫ = ǫ. (3.2.1.7)
Under certain conditions to be specified below the variations of the gravitino and the
dilatino vanish for
3Φ
ǫ=e 8 ε0 ⊗ η0 , (3.2.1.8)
where ε0 and η0 are SO (1, 1) and SO (8) constant spinors, respectively, which satisfy
the lower dimensional chirality conditions
9
!
Y
(1 − γ3 ) ε0 = 0 , 1 − Σa η0 = 0. (3.2.1.9)
a=2
This breaks the supersymmetry to half the amount of the perturbative (trivial) vac-
uum. (The condition (3.2.1.7) could be also satisfied by choosing simultaneously the
opposite chiralities in the two equations (3.2.1.9).)
We already mentioned that only under certain conditions we can find unbroken
supersymmetries at all. Requiring that asymptotically (r → ∞) we obtain the pertur-
bative vacuum, these conditions read
3
A = (Φ − Φ0 ) , (3.2.1.10)
4
1
B = − (Φ − Φ0 ) , (3.2.1.11)
4
3
C = 2Φ − Φ0 (3.2.1.12)
2
where Φ0 is the asymptotic value of Φ. Hence supersymmetry leaves only one function
out of our ansatz undetermined. This function can be taken to be the dilaton whose
equation of motion boils down to
i.e. the “flat” Laplacian of the transverse space (spanned by the y a ) acting on e−2Φ
has to vanish. As in the case of four dimensional black holes, we solve this equation
everywhere but at the origin r = 0, where there are additional contributions due to
a source string. (We do not add the source string explicitly here, but will infer its
3. The fundamental string 130
properties (tension and charge) in an indirect way below. For the explicit inclusion of
the source term see e.g.[152].) The solution to (3.2.1.13) reads
k
e−2Φ = e−2Φ0 1 + 6 , (3.2.1.14)
r
where k is an integration constant which will be related to the string tension below.
Plugging this back into (3.2.1.10) – (3.2.1.12) and in the ansatz gives the final solution.
Next, we would like to deduce the tension of the string source from our solution.
This is done by studying the Newtonian limit of general relativity. In particular, by
comparing the Einstein equation with the geodesic equation of a point particle (which
has constant mass in the Einstein frame) one finds that the Newton potential of the
string-source is encoded in the subleading term in a large r expansion of g00 . Therefore,
we first observe that for large r
3k
g00 = −1 + + ... . (3.2.1.15)
4r 6
The relation between g00 and the Newton potential of a string is explicitly such that4
1 3 δ (r)
7
∂r r 7 ∂r g00 = − κ2 TE (3.2.1.16)
r 2 Ω7 r 7
holds, with the understanding that terms denoted by . . . in (3.2.1.15) are neglected.
The string tension is denoted by TE . Further, the unit volume of a seven–sphere Ω7
enters the expression. Hence, we obtain
3k
TE = Ω7 . (3.2.1.17)
κ2
We put the index E at the tension in order to indicate that it is measured in the Ein-
stein frame. What we are actually interested in, is the tension in the string frame. This
is readily obtained by noticing that transforming back to the string frame (asymptot-
ically) implies
7Φ0
κ2 → e2Φ0 κ2 , Ω7 → e 2 Ω7 . (3.2.1.18)
1
Recalling that our “elementary particle” is a string of tension 2πα′ and requiring that
any string like object must consist out of an integer number N of elementary strings
we finally determine the integration constant k to be6
κ2 3Φ
− 20
k=N e . (3.2.1.20)
6πα′ Ω7
It remains to compute the U (1) charge carried by the vacuum. This is basically
done by integrating the B equation of motion over the transverse space. The result is
Z
1
µ= √ e−Φ ⋆ H, (3.2.1.21)
2κ S 7
where the integration is over an asymptotic seven-sphere enclosing the string source.
The U (1) charge is denoted by µ. Expressing the result in terms of the string tension
one obtains
√ N
µ= 2κ . (3.2.1.22)
2πα′
This equality is related to partially unbroken supersymmetry. If the configuration
was not stable the tension of the bound state would be larger than the sum of the
elementary tensions. Hence, the rhs of (3.2.1.22) is larger for general (non BPS) states.
The BPS state saturates a general inequality. Since the BPS state is stable, there can
be no state with less tension and the same charge since otherwise the BPS state would
decay into such a state. The lower bound on the tension set by the BPS state is called
the Bogomolnyi bound.
Φ = Φ (r) . (3.2.2.4)
All fermionic fields are again set to zero. There is an unbroken supersymmetry if
we can find a spinor such that the gravitino and dilatino transformations (3.2.0.10)
and (3.2.0.11) vanish. It turns out that half of the supersymmetry is preserved if the
following relations hold
1
A = − (Φ − Φ0 ) , (3.2.2.5)
4
3
B = (Φ − Φ0 ) , (3.2.2.6)
4
3
C = −2Φ + Φ0 , (3.2.2.7)
2
where Φ0 denotes again the asymptotic r → ∞ value of the dilaton. (In addition
to partially unbroken supersymmetry we have once again imposed that for large r
the vacuum should approach the perturbative vacuum.) Under these conditions the
equations of motion boil down to
We solve this equation everywhere but at r = 0 where we allow for additional contri-
butions due to source terms. One finds
k̃
e2(Φ−Φ0 ) = 1 + . (3.2.2.9)
r2
The integration constant k̃ can be fixed by exploiting the Dirac quantization condition.
To this end we compute the charge carried by the vacuum
Z √
1 Φ 2Ω3 k̃ Φ0
µ̃ = √ e ⋆K = e2 , (3.2.2.10)
2κ S 3 κ
3. The NS five brane 133
where the integral is over an asymptotic three–sphere surrounding the five brane and
Ω3 denotes the volume of a unit three–sphere. Now, the Dirac quantization condition
reads
µ̃µ = 2π Ñ N, (3.2.2.11)
where µ is the charge of N elementary strings (3.2.1.22). The number of five branes
(number of magnetic charges) is Ñ . This fixes the integration constant,
π Ñ − Φ0
k̃ = e 2 , (3.2.2.12)
TS Ω3
1
where TS is the elementary string tension (TS = 2πα′ in the string frame). By com-
puting the gravitational potential in the Newtonian limit, one finds the tension of the
five brane in the Einstein frame (with TS and κ also in Einstein frame units)
π Ñ − Φ0
T̃E = e 2 . (3.2.2.13)
TS κ2
2πα′ π Ñ
T̃ = e−2Φ0 (3.2.2.14)
κ2
in the string frame. We observe that the five brane tension behaves as 1/gs2 . In the
perturbative region the NS five brane is very heavy whereas it becomes lighter when
the string coupling increases.
The NS five brane is an extended object for which we did not give a perturbative
description. Indeed, such a description is not known. One could try to quantize strings
in the NS five brane background. This is possible only in certain spatial regions.
Firstly, for large r the background becomes flat, and we know how to quantize strings
there. But also in the background at r → 0 (the near horizon limit) one can find a
quantized string theory. In that limit the string frame metric reads
and the dilaton is linear in ρ = log r. With dΩ23 we denote the metric of a unit three–
sphere. The NSNS field strength H is a constant times the volume element dΩ3 . The
geometry factorises into a 5 + 1 dimensional Minkowski space times the direction on
which the dilaton depends linearly times an S 3 . Since S 3 is an SU (2) group manifold,
string theory can be quantized in such a background. For more details see[97] or the
review[96].
3. Type II branes 134
The dilaton is also taken to be a function of r. Let us first exclude the case p = 3 from
the discussion. For the p + 1 form gauge field we choose
where again the boundary condition that for r → ∞ the background should be trivial
has been imposed. C0 denotes the asymptotic value of C which is related to the
asymptotic dilaton value
p−3
C0 = Φ0 . (3.3.0.9)
4
The equations of motion reduce to
We solve this by
kp
e−C = e−C0 + . (3.3.0.11)
r 7−p
The RR charge of the vacuum is
Z
1 3−p 7−p
µp = √ e 2 Φ ⋆ Fp+2 = √ Ω8−p kp , (3.3.0.12)
2κ S 8−p 2κ
where the integration is over an asymptotic (8 − p)–sphere surrounding the p brane,
and Ω8−p is the volume of the unit sphere. Now we try whether we can identify the
type II p-branes with the D-branes discussed in section 2.3. This trial is motivated by
the observation that the D-branes considered in section 2.3 are also extended objects
carrying RR charge. In section 2.3 we computed the charge of a single D-brane to be
(see (2.3.2.31) and (2.3.2.29))
√ 3−p
µsingle
p
brane
= 2π 4π 2 α′ 2
. (3.3.0.13)
Assuming that the vacuum considered in the present section is composed out of an
integer number Np of single D-branes, we identify (TS denotes the frame dependent
tension of a single fundamental string)
3−p √
2κ 4π 2 /TS 2 π
kp = Np , (3.3.0.14)
(7 − p) Ω8−p
such that
µp = Np µsingle
p
brane
. (3.3.0.15)
is satisfied. After we have fixed the integration constant kp the tension of the brane
solution is determined. Because the vacuum considered here and the D-branes consid-
ered in section 2.3 are BPS objects the tension is related to the charge and we expect
3. Type II branes 136
that our tension should be in agreement with (2.3.2.29). Let us nevertheless compute
it explicitly. To this end, we write down the asymptotic expansion of g00
7 − p kp p−3 Φ0
g00 = −1 + e 4 + ... . (3.3.0.17)
8 r 7−p
The tension is given via (see also (3.2.1.16))
1 8−p
7−p 2 δ (r)
∂r r ∂r g00 =− κ Tp,E , (3.3.0.18)
r 8−p 4 Ω8−p r 8−p
where Tp,E denotes the tension in Einstein frame units and it is understood that terms
denoted by dots in (3.3.0.17) are neglected. This yields
7−p p−3
Tp,E = 2
Ω8−p kp e 4 Φ0 . (3.3.0.19)
2κ
p+1
Since Tp,E has mass dimension p + 1, it receives a factor of e− 4 Φ under the trans-
formation to the string frame. Taking this and (3.3.0.14) into account, we find for the
string frame tension
√ 3−p 1
Tp = Np e−Φ0 κ π 4π 2 α′ 2 , (3.3.0.20)
κ
in agreement with (2.3.2.29). Thus, we found that the p-brane vacuum can be viewed as
consisting out of an integer number of “elementary” (or magnetic) D-branes considered
in section 2.3.
So far, we have derived this result only in the case p 6= 3. In the case p = 3, the
condition (3.3.0.8) is changed into
Φ = Φ0 , C0 = 0. (3.3.0.21)
The selfduality condition can be imposed by replacing F5 from our ansatz with F5 +⋆F5 ,
F5 → F5 + ⋆F5 . (3.3.0.22)
such that
√
µ3 = 2πN3 . (3.3.0.26)
Applications
In this chapter, we are going to present some applications of the branes discussed so
far. In the following, we will show that branes are a useful tool in supporting duality
conjectures involving an interchange between strong and weak couplings. As a first
example we consider dualities among different string theories. Thereafter, field theory
dualities will be translated into manipulations within certain brane setups. Next, we
want to present the AdS/CFT correspondence – a duality between closed and open
string theory, or in first approximation between gravity and gauge theory. Finally, we
argue that branes allow constructions in which the string scale is about a TeV. Such
setups have the prospect of being discovered in the near future. There are many more
applications of branes in theoretical physics. Some of them we will list in chapter 6
containing suggestions for further reading.
138
4. String dualities 139
11 D SUGRA
Heterotic
E8 × E8 Type IIA
Type I
lated to the length of the interval such that the string coupling is small when the
interval is short. The E8 × E8 fields live as twisted sector fields at the ends of the
interval (the orbifold nine planes). If we take the string coupling of the E8 × E8 het-
erotic string to be strong, 11D supergravity on an interval provides the more suitable
description. The connection between the E8 × E8 and the SO (32) string was already
discussed in section 2.1.5.4. It does not relate strong with weak coupling but small
with large compactification radii in nine dimensions. The heterotic SO (32) string is
connected to type I strings by a strong/weak coupling duality. Now, let us go back
to the top of the star and go down clockwise. Type IIA supergravity can be obtained
by compactifying 11 dimensional supergravity on a circle. The radius of the circle
determines the vev of the dilaton. For small string coupling the circle is small, and
for strong coupling it is large. The connection between type IIA and type IIB strings
is seen by compactifying further down to nine dimensions and inverting the radius, as
argued in section 2.1.5.4. Type I theory is obtained by gauging worldsheet parity of
type IIB strings and adding the D-branes needed to ensure RR-charge conservation
(in a sense these can be viewed as twisted sector states).
Because the branes we have discussed are stable under deformations in the moduli
space, they should be mapped in a one-to-one way onto each other by string dualities.
Since eleven dimensional supergravity did not appear until now in our discussion (it
does not correspond to an effective weakly coupled string theory), we have to list
the relevant BPS branes of 11 dimensional supergravity. 11 dimensional supergravity
contains a three form gauge potential which can be Hodge dualized to a six form gauge
potential. Analogously to the solutions found in the previous chapter, one finds thus
a membrane (2 brane) and a five brane.
4. String dualities 140
Let us now walk once around the star in figure 4.1 in a clockwise direction and follow
the branes along this journey. Upon compactifying one of the eleven dimensions the
momentum into this direction becomes quantized. The off diagonal metric components
containing one 11 label become a Kaluza Klein gauge field – a one form potential, which
can be Hodge dualized (with respect to the non compact directions) to a seven form
potential. The associated BPS states are zero and six branes. These become the D0
and the D6 branes in the type IIA picture. For the branes which exist already in
the uncompactified theory, there are two options within the compactification. The
compact dimension can be transverse or longitudinal. Hence, the membrane will be
either described by a fundamental string or by a D2 brane in weakly coupled type
IIA theory, and the five brane yields the D4 brane and the NS five brane of type IIA
theory.
Compactifying further down to nine dimensions and taking the decompactification
limit after a T-duality transformation, type IIA theory goes over into type IIB theory.
The D-branes gain or lose one spatial direction due to the T-duality, and hence we
obtain all the D-branes of type IIB theory. Type IIB theory possesses a symmetry
which is not depicted in figure 4.1. This is an SL (2, Z) symmetry which we do not
want to discuss in detail. For later use we state that the SL (2, Z) symmetry contains
a transformation called S duality. S duality interchanges strong with weak coupling,
the D1 brane with a fundamental string and the D5 brane with the NS five brane.
The D3 brane stays a D3 brane under S-duality.1
Type I strings are obtained by projecting out worldsheet parity in type IIB strings.
This removes the fundamental string, the NS five brane, and the D3 brane from the
spectrum of BPS states. The remaining states are the D1 and the D5 brane. Under
the strong/weak coupling duality mapping of type I theory to the SO (32) heterotic
theory, these become the fundamental string and the NS five brane of the heterotic
string. The BPS spectrum is not affected when going over to the E8 × E8 heterotic
string via T-duality. The E8 × E8 theory is supposed to be dual to 11 dimensional
supergravity on S 1 /Z2 . Therefore, let us discuss which of the branes of 11 dimensional
supergravity survive the Z2 projection. First of all, the zero and the six branes are
projected out since the Kaluza-Klein gauge field is odd under changing the sign of the
eleventh coordinate. In order to deduce the Z2 action on the three form potential C,
we note that the action of 11 dimensional supergravity contains a Chern Simons term
Z
C ∧ dC ∧ dC.
This term is symmetric under the Z2 if C receives an additional sign, i.e. a C component
1
We do not include the D7 brane and its counterpart, the D instanton (related by Hodge duality),
into the discussion.
4. Dualities in Field Theory 141
NS5 Nf D6 NS5′
Nc D4
Figure 4.2: Brane setup for supersymmetric QCD. It has to be looked at in combination
with table 4.1.
0 1 2 3 4 5 6 7 8 9
NS5 – – – – – – · · · ·
NS5′ – – – – · · · · – –
D4 – – – – · · – · · ·
D6 – – – – · · · – – –
Table 4.1: Brane setup for supersymmetric QCD. The numbers in the first line label
the dimensions. Hyphens denote longitudinal and dots transverse dimensions.
branes. Throughout this section we will neglect the back reaction of branes on the
target space geometry, i.e. we take a limit where gravity decouples.
Our first task is to translate N = 1 SU (Nc ) supersymmetric gauge theory with
Nf chiral multiplets in the fundamental representation into a brane setup. A setup
yielding the desired theory is drawn in figure 4.2. Since it is difficult to draw pictures
in ten dimensions we supplement the figure by table 4.1 where hyphens stand for
longitudinal and dots for transverse dimensions. The D4-brane stretches in the sixth
direction between the two NS5 branes. Hence, its extension along the sixth dimension
is given by the finite distance of the NS5 and the NS5′ brane. If this distance is
shorter than the experimental resolution, the theory on the D4-branes is effectively
3+1 dimensional. The positions of the NS5, NS5′ and the D4 in the seventh dimension
must coincide (simply for geometrical reasons). We take Nc of such D4-branes in
order to obtain SU (Nc ) gauge theory. The position of the D4-branes in the transverse
directions is fixed by the condition that it stretches between the NS5 and NS5′ brane.
The scalar fields in the adjoint of the gauge group correspond to collective coordinates
for those positions. They are projected out by the boundary condition. Therefore,
4. Dualities in Field Theory 143
0 1 2 3 4 5 6 7 8 9
D5 – – – – – – · · · ·
D5′ – – – – · · · · – –
F1 – · · · · · – · · ·
D3 – · · · · · · – – –
Table 4.2: Dual brane setup for supersymmetric QCD. This can easily be checked to
be consistent.
the theory on the D4-branes can admit at most N = 1 supersymmetry (viewed from
a 3 + 1 dimensional perspective). We will argue in a moment that there is partially
unbroken supersymmetry in the above setup. Before, let us comment on the role of
the D6-branes. A string starting on a D6 and ending on a D4-brane transforms in
the fundamental representation of SU (Nc ). If we take Nf D6-branes we obtain the
Nf desired multiplets in the fundamental representation of the gauge group. The
SU (Nf ) gauge theory becomes the flavour symmetry of our supersymmetric QCD.
(Indeed, the effective four dimensional coupling is obtained by integrating over the
extra dimensions. It is inversely proportional to the volume of the extra dimensions.
Since the D6-brane worldvolume contains non-compact extra dimensions, the SU (Nf )
dynamics decouples and we are left with a global symmetry.)
After we successfully constructed a brane setup for the gauge theory we are in-
terested in, we should check whether this brane setup is consistent. One could, for
example, couple it to gravity and look for explicit solutions describing such a setup.
This is, however, rather complicated. What we will do instead, is to take a setup from
which we know that it is consistent and connect it to the above setup through a chain
of string dualities. A setup of which we know that it is consistent is given in table
4.2. Here, a fundamental string (F1) stretches between two D5-branes (D5 and D5′ ).
This is consistent by the definition of a D-brane. Above, we have argued that at most
N = 1 supersymmetry survives on the (interval compactified) D4-brane. An argument
that this is exactly the case can be given by counting the ND directions of the setup in
table 4.2 (see also section 2.3.1.3). The D5 and the D3-brane provide eight ND direc-
tions (all but the zeroth and sixth dimension). An open string starting on a D5′ -brane
and ending on a D3-brane has four ND directions (1237). Finally, the string stretched
between the D5 and D5′ -brane has mixed (ND) boundary conditions along the four
dimensions: 4589. Hence, the number of ND directions is always an integer multiple
of four, indicating that the spectrum possesses some supersymmetry. (There are also
more precise methods to investigate the number of preserved supersymmetries. One
can study the conditions for vanishing gravitino and dilatino variations in the rigid
4. Dualities in Field Theory 144
Nf D6 NS5 NS5′
Nf − Nc D4
Nc D4
Figure 4.3: Brane setup of figure 4.2 after pushing the D6-branes past the NS5 brane.
limit, see e.g. [158] for such an analysis within the present context.) It remains to see
that the setup in table 4.2 is connected to the one which we are interested in (table
4.1). Table 4.2 contains branes of type IIB theory. Therefore, we can apply an S-
duality (shortly described in the previous section) on this system. This takes the D5
and D5′ -brane to the NS5 and NS5′ brane of table 4.1. The fundamental string (F1)
turns into a D1 string and the D3-brane remains invariant under S-duality. Performing
a T duality along the first, second and third dimension (replacing type IIB by IIA)
yields the configuration of table 4.1.
In the following we will describe a path in the moduli space of the setup in figure
4.2 taking us to the dual theory found by Seiberg. We will do so by essentially
interchanging the position of the NS5 with the NS5′ brane. This involves however
some subtleties which we will mention but not elaborate on. For more details we ask
the interested reader to consult[158] or literature to be given in chapter 6. Our first
step is to move the D6-branes to the left of the NS5 brane. When the D6-branes cross
the NS5 brane, Nf additional D4-branes stretching between the D6-branes and the
NS5 brane are created[238]. After the D6-branes have been moved to the left of the
NS5 brane, there is a point in moduli space where there are no D4-branes stretching
between the NS5 and the NS5′ brane. This can be achieved by connecting the Nc
D4-branes stretching between NS5′ and NS5 branes with Nc out of the Nf D4-branes
which stretch between D6-branes and NS5′ branes. The result of performing this first
step in moduli space is drawn in figure 4.3.
Now, the boundary conditions are such that we can displace the NS5 brane in the
seventh dimension. After doing so, it can be moved to the right of the NS5′ brane
4. AdS/CFT correspondence 145
Nf D6 NS5′ NS5
Nf D4
Nf − Nc D4
along the sixth dimension.2 As soon as the NS5 brane is situated to the right of the
NS5′ brane, we realign it in the seventh dimension with the positions of the NS5′ and
the D4-branes. There are now (Nf − Nc ) D4-branes starting at the D6-branes passing
through the NS5′ branes and ending on the NS5 brane. These we break on the NS5′
brane. The picture drawn in figure 4.4 emerges.
Finally, we need to read off the perturbative formulation of the field theory cor-
responding to figure 4.4. The gauge group of the theory living on the D4-branes
stretching between the NS5 and the NS5′ brane is SU (Nf − Nc ). There are Nf chi-
ral multiplets in the fundamental of the gauge group coming from strings stretching
between the Nf D4-branes on the left and the Nf − Nc D4-branes in the middle. The
D4-branes to the left can move (fluctuate) in the eighth and ninth dimension. This
gives rise to Nf2 chiral multiplets which are singlets under the gauge group.
In this section we have seen that branes can be useful tools in deriving (or at least
illustrating) quite nontrivial connections between gauge field theories. Our purpose
was to provide the rough ideas on how this works within an example. The reader who
found this interesting is strongly advised to check the literature (chapter 6) for more
details and subtleties.
4πgs N α′ 2
e−C = 1 + , (4.3.1.1)
r4
where gs denotes the string coupling, and N = N3 is the number of D3 branes. Recall
also that the metric is (3.3.0.3) (use (3.3.0.6) and (3.3.0.7) and i, j = 0, . . . , 4, and we
parameterize the transverse space by polar coordinates, i.e. dΩ25 is the metric of a unit
five sphere)
C C
ds2 = e 2 ηij dxi dxj + e− 2 dr 2 + r 2 dΩ25 . (4.3.1.2)
some probe brane at a finite distance is constant. Performing the limit (4.3.1.3) the
metric (4.3.1.2) becomes5
2 ′ U2 i j
p dU 2 p 2
ds = α √ ηij dx dx + 4πgs N 2 + 4πgs N dΩ5 . (4.3.1.4)
4πgs N U
This describes an AdS5 × S 5 geometry. Before taking a short detour on the description
of AdS5 spaces as hypersurfaces of a six dimensional space let us check the validity
region of (4.3.1.4) by focusing on the S 5 part. The radius of the S 5 is
p
R2 = α′ 4πgs N . (4.3.1.5)
In order to avoid high curvature (where higher derivative corrections become impor-
tant, or even the effective gravity description may break down) one should take this
radius to be large, i.e.
gs N ≫ 1. (4.3.1.6)
In addition we should keep the string coupling small which implies that the number
of D3 branes we look at has to be large. Now, we recall that the field theory living on
the D3 branes is N = 4 supersymmetric SU (N ) gauge theory. The gauge coupling
is gY2 M = 2πgs (see (2.3.3.22)). So, at first sight it seems that the gauge coupling
is small whenever the string coupling is small. However, we should also impose the
condition (4.3.1.6), in particular the large N limit. For large N SU (N ) gauge theories
‘t Hooft developed a perturbative expansion in the parameter (the ‘t Hooft coupling)
gY2 M N [435]. The condition (4.3.1.6) implies that the ‘t Hooft coupling is large when-
ever the effective gravity description is reliable. We will argue below that gravity (or
closed type IIB strings) in the space (4.3.1.4) is dual to the gauge theory living on
the D3 branes. Because of (4.3.1.6) this is a strong/weak coupling duality. One of
the first things one should check before publishing a conjecture on dual pairs is that
the global symmetries of the dual descriptions should match. (Global symmetries are
observable.) Therefore, let us take a short detour and describe the AdS5 space as a
hypersurface in a six dimensional space. This will enable us to see the isometries of
AdS5 in much the same way as one sees the SO (6) isometry of S 5 when viewing it as
a hypersurface in six dimensional space.
The space in which we will find an AdSp+2 space as a hypersurface is a 2 + p + 1
dimensional space with the metric
p+1
X
2 2
ds = −dX−1 − dX02 + dXα2 . (4.3.1.7)
α=1
5
One can obtain this metric directly when dropping the requirement of an asymptotically trivial
background in the search for BPS branes in section 3.3.
4. The conjecture 148
Analogously to a sphere, the AdSp+2 space is defined as the set of points satisfying
the condition
p+1
X
2
−X−1 − X02 + Xα2 = −R2 , (4.3.1.8)
α=1
where R is called the radius of the AdSp+2 space. We solve this equation by
U2 2 dU
2
ds2 = ηij dx i
dx j
+ R . (4.3.1.12)
R2 U2
Comparison with (4.3.1.4) shows us that the limit (4.3.1.3) took us to an AdS5 × S 5
space where the radii of the AdS5 and the S 5 coincide and are given by (4.3.1.5).
After this detour we can easily read of the isometries of (4.3.1.4). The isometry is
SO (4, 2) × SO (6). These isometries show up in the field theory on the D3 branes as
follows. The SO (6) (= SU (4)) is the R symmetry of N = 4 supersymmetric Yang-
Mills theory. The beta function of the gauge theory vanishes exactly, i.e. the gauge
theory is a conformal field theory. The SO (4, 2) part of the isometry corresponds to
the conformal group which is a symmetry in the gauge theory. Taking into account the
preserved supersymmetries6 , one observes that the isometry group SO (4, 2) × SO (6)
can be extended to the superconformal group acting in the field theory. Thus the
global symmetries of the two descriptions match. In the asymptotic region U → ∞
the AdS part of the metric (4.3.1.4) becomes (up to a conformal factor) the 3+1
dimensional Minkowski space. This is the boundary of the AdS space. The SO (4, 2)
isometry acts as the group of conformal transformations on the Minkowski space.
In this sense, one can identify the boundary of the AdS space with the location of
the D3 branes, although one should not think of the two descriptions simultaneously,
because whenever the parameters are such that the gravity description is reliable, the
perturbative description of the gauge theory breaks down and vice versa.
Moreover, one can identify the SL (2, Z) duality of the type IIB string with the
Montonen Olive duality[346, 463, 360, 457] of N = 4 super Yang Mills theory.
6
These are enhanced in the near horizon limit.
4. Wilson loop computation 149
Thus, we have seen some evidence that the AdS/CFT correspondence conjecture
holds. More checks have been performed, but we will not discuss those here. In the
following section we want to illustrate the duality by computing Wilson loops in gauge
theory using type IIB superstrings. Before doing so, let us summarize the AdS/CFT
correspondence (duality) conjecture.
• The AdS5 and the S 5 have the same radius whose value (in units of α′ ) is
related to the ‘t Hooft coupling of the gauge theory via equation (4.3.1.5), and
gY2 M = 2πgs .
• The type IIB string theory is in its perturbative regime if gs is small, and higher
curvature effects are not dangerous as long as (4.3.1.6) holds. In this region, the
gauge theory is in the large N limit and strongly coupled.
In a somewhat weaker statement, one should replace “type IIB superstrings” by “type
IIB supergravity” and “open strings ending on D3 branes” by N = 4 SU (N ) gauge
theory. We will take the duality conjecture as stated in the items.
A Wilson loop is the (normalized) partition function of gauge theory in the presence
of an external quark anti-quark pair. A perturbative description of this situation in
a D3 brane setup for static quarks is drawn in figure 4.5. In order to employ the
AdS/CFT duality conjecture, we need to translate figure 4.5 to type IIB strings living
on AdS5 × S 5 . The prescription is that the open strings in figure 4.5 translate into
a background string of type IIB theory on AdS5 × S 5 . In the previous section we
have argued that the position of the N D3 branes is translated to the boundary of the
AdS5 space. (We should point out again that the emphasis is on “translated” since
the gauge theory description breaks down whenever the AdS prescription is reliable.)
Therefore, the background string should fulfill the boundary condition that its ends on
the AdS5 boundary are separated by a distance L. Classically, the background string
is then uniquely determined by the requirement of minimal worldsheet area. As we
will see in a moment, the picture drawn in figure 4.6 arises. The fact that the string
with minimal area goes down from the boundary into the AdS space and up again to
satisfy the boundary condition is a result of the non trivial metric. The corresponding
4. Wilson loop computation 150
N
D3 Branes
single
D3 Brane
Figure 4.5: The perturbative Wilson loop setup. The quark anti-quark pair cor-
responds to the ends of open strings on the N D3 branes. The open strings have
opposite orientation. The quark anti-quark pair is chosen to be static. The dynamics
of the quarks decouples as long as the single D3 brane is very far away from the N D3
branes. The distance between the quark and the anti-quark is L.
Boundary
of AdS5
Background
String
Figure 4.6: The non perturbative Wilson loop setup. The quark anti-quark pair
corresponds to a background string ending on the AdS5 boundary.
4. Wilson loop computation 151
calculation can be carried out explicitly. For the sake of a minor simplification, we
R2
redefine the coordinate U = α′ u
such that the metric (4.3.1.4) reads
2 2 2 i j du2 2
ds = R u ηij dx dx + 2 + dΩ5 . (4.3.2.1)
u
The worldsheet area of the background string is
Z
1 p
S= ′
dτ dσ − det g (4.3.2.2)
2πα
where gαβ is the induced metric (2.1.1.2). As an ansatz for the background string we
take
X0 = τ , X1 = σ , X 4 = U (σ) , (4.3.2.3)
and the rest of the string positions is constant in σ and τ . The indices are assigned in
the order in which coordinates appear in (4.3.2.1), and X 4 = U (the capital U denotes
the string position in the space with the metric (4.3.2.1) and should not be confused
with the capital U in (4.3.1.3)). The first two equations in (4.3.2.3) represent the static
gauge and the sigma dependence of U allows for the string to describe the curve of
figure 4.6. This is the simplest consistent ansatz for the given boundary conditions.
The induced metric is
!
ds2ind 2 2 2 (∂σ U )2
= −U dτ + U + dσ 2 . (4.3.2.4)
R2 U2
where the plus-minus sign appears due to the two branches of the curve in figure 4.6
(σ is a horizontal coordinate and U a vertical one in this figure). At the boundary
(U → ∞) the difference between the two values of σ should be L. Some straightforward
manipulations with the integral in (4.3.2.8) yield
L 1 3 1
= B , , (4.3.2.9)
2 4U0 4 2
where Z 1
Γ (α) Γ (β)
B (α, β) = dxxα−1 (1 − x)β−1 =
0 Γ (α + β)
denotes Euler’s Beta function. Using the identities
π 1 √
xΓ (x) = Γ (x + 1) , Γ (x) Γ (1 − x) = , Γ = π
sin πx 2
one finds for the integration constant
3
(2π) 2
U0 = 2 . (4.3.2.10)
Γ 41 L
This shows our earlier statement that the background string is uniquely determined
by the boundary condition. The Wilson loop W [C] is the partition function for the
background string. For the classical approximation we find
with
Z q
R2
E= dσ (∂σ U )2 + U 4 . (4.3.2.12)
2π
Plugging in the classical solution (4.3.2.7) (and taking into account a factor of two due
to the two branches) yields
Z ∞
R2 U2
E= dU p . (4.3.2.13)
π U0 U 4 − U04
Now we split this integral into two pieces (the motivation for this will become clear
below)
E = Ec + Es , (4.3.2.14)
4. Wilson loop computation 153
with
Z ∞
R2 U 4 + U04
Es = dU p , (4.3.2.15)
π U0 U 2 U 4 − U04
and
Z ∞
R2 U4
Ec = − dU p 0 . (4.3.2.16)
π U0 U 2 U 4 − U04
Let us first discuss the integral Es . This integral is divergent due to the upper inte-
gration bound and we regularize it by a cutoff Umax . The asymptotic expansion for
large Umax is
Z Umax
R2 U0 U0 y4 + 1
Es = p dy
π 1 y2 y4 − 1
"p # Umax
R2 U0 y 4 − 1 U0 R2
= = Umax + . . . , (4.3.2.17)
π y π
1
where the dots stand for terms going to zero as Umax is taken to infinity. Thus, Es
corresponds to the self energy of the two strings in figure 4.5. It does not depend on the
distance L and diverges as the length of the string is taken to infinity. Here, we observe
one interesting feature of the AdS/CFT correspondence. From the AdS perspective
Umax is a large distance, i.e. an IR cutoff. On the field theory side this appears as
a cutoff for high energies, i.e. a UV cutoff. This interchange between infrared and
ultraviolet cutoffs is a general characteristics of the correspondence[434]. Let us give a
technical remark in connection with the integral Es . Plugging in our classical solution
(4.3.2.7) into the induced metric (4.3.2.4), we find the classical value of the induced
metric (for later use we call this R2 hαβ )
U6 2
ds2class 2 α
≡ R hαβ dσ dσ = R β 2 2 2
−U dτ + 4 dσ . (4.3.2.18)
U0
U 4 + U04
R(2) = 2 . (4.3.2.19)
U4
With this information it is easy to verify that the structure of the self energy integral
is
Z
R2 √
Es = d2 σ −hR(2) . (4.3.2.20)
4πT
(This does not contradict the Gauss–Bonnet theorem because the worldsheet of the
background string is not compact.)
4. Wilson loop computation 154
Now, let us come to the second contribution in (4.3.2.14). This will turn out to be
the more interesting one. Its computation is quite similar to the computation of U0 in
terms of L. Therefore, let us just give the result
q
4π 2 2gY2 M N
Ec = − 4 , (4.3.2.21)
Γ 14 L
where (4.3.1.5) and gY2 M = 2πgs has been used. This is the part of the quark anti-quark
potential which arises due to gluon exchange among the two quarks. It is a Coulomb
potential. Since L is the only scale appearing in the setup and N = 4 supersymmetric
Yang–Mills theory has a conformal symmetry, there can be only a Coulomb potential.
Anything else would need another scale to produce an energy, but this cannot appear
due to conformal invariance.
In that respect models with less or none supersymmetry are more interesting be-
cause one can observe confinement in those models. The corresponding literature is
listed in chapter 6. The case we are considering here, is the one where the AdS/CFT
correspondence is perhaps best understood. We will study a question which is inter-
esting from a more theoretical perspective, namely whether there are corrections to
the result (4.3.2.21).
where δX denote bosonic fluctuations and δθ fermionic ones (the fermionic background
of the string is trivial). Before going into the details of the computation, let us describe
the expansion we are going to perform. From (4.3.2.2) and (4.3.2.1) we see that the
square root of the ‘t Hooft coupling appears as an overall constant in front of the metric.
(This is also true for terms containing fermions.) Therefore, the expansion parameter ~
(or α′ in section 2.1.3) is identified with the inverse square root of the ‘t Hooft coupling.
The expression (4.3.2.22) can be computed as a power series in this parameter. In
particular, the next to leading order correction to (4.3.2.21) will not depend on the ‘t
Hooft coupling. It is this correction we will discuss in some detail in the following. In
order to be able to use (4.3.2.22) for explicit calculations we need to know the type
IIB string action in an AdS5 × S 5 background. Fortunately, this has been constructed
in the literature[340]. These authors gave a type IIB action in the Green Schwarz
formalism, which is appropriate in the presence of non-trivial RR backgrounds. The
construction is similar to the one discussed in section 2.1.1.3. One uses target space
supersymmetry and kappa symmetry as a guide. The technicalities are rather involved
and we will not discuss them here. Because we will restrict ourselves to terms second
order in fluctuations we only need a truncated version of the action of type IIB strings
on AdS5 × S 5 . The complete action does not contain terms with an odd number of
target space fermions, in particular no terms linear in target space fermions. Since the
fermionic background is trivial, contributions quadratic in fluctuations can either have
two fermionic fluctuations and no bosonic fluctuation or only bosonic fluctuations.
The part of the action which is quadratic in the fluctuations consists of a sum of terms
with only bosonic fluctuations and terms with only fermionic fluctuations.
Let us discuss the bosonic fluctuations first. The type IIB action is a kappa
symmetric extension of (4.3.2.2). For the bosonic fluctuations only the contribution
(4.3.2.2) is relevant (for the lowest non trivial contribution). As in section 2.1.3 we pa-
rameterize the fluctuations by tangent vectors to geodesics connecting the background
with the actual value, i.e. we perform a normal coordinate expansion. The quantum
fields are
ξ a = Eµa ξ µ , (4.3.2.23)
where Eµa are the vielbein components obtained by taking the square root of the
diagonal metric components. A number as a label on a ξ will stand for a flat index
(a), unless stated explicitly otherwise. The computation of the term second order in
the ξ a is a bit lengthy but straightforward with the information given in section 2.1.3.
(The only difference to section 2.1.3 is that we expand now a Nambu-Goto action
instead of a Polyakov action.) Before giving the result it is useful to perform a local
4. Wilson loop computation 156
x4 = u
Background
String
ξ µ=4
ξ µ=1
x1 = σ
with
p
U2 U 4 − U04
cos α = 02 , sin α = . (4.3.2.25)
U U2
Note that the determinant of the matrix appearing in (4.3.2.24) is one. The fields ξ k
and ξ ⊥ describe fluctuations parallel and perpendicular to the worldsheet, respectively.
This is illustrated in figure 4.7. Fluctuations drawn into figure 4.7 carry Einstein in-
dices which we indicated explicitly. The angle γ is given by the slope of the background
string
tan γ = ∂σ U. (4.3.2.26)
The combination
vanishes for
ξ µ=4
tan γ = , (4.3.2.28)
ξ µ=1
7 L
We employ the symmetry of the background string and take 0 ≤ σ ≤ 2
and the plus sign in
(4.3.2.7).
4. Wilson loop computation 157
i.e. if ξ µ=1 + ξ µ=4 is tangent to the background string. Thus, the combination in
(4.3.2.27) is normal to the background string. Transforming the indices to flat ones
(with the vielbein) and ortho-normalization yields ξ k and ξ ⊥ with the given interpre-
tation. When writing down the Lagrangian second order in the fluctuations, we can
set R = 1 since we know the general R dependence (viz. none) from the argument
given above. The Lagrangian for the bosonic fluctuations comes out to be
(2) (2) (2)
Lbosons = LAdS5 + LS 5 (4.3.2.29)
with
(2) 1√ X 2 2
LAdS5 = −h ξ a ∆ξ a − 2 ξ 2 − 2 ξ3
2
a=2,3,⊥
2
+ R − 4 ξ⊥
(2)
, (4.3.2.30)
9
(2) 1√ X ′ ′
LS 5 = −h ξ a ∆ξ a , (4.3.2.31)
2 ′ a =5
where total derivative terms have been dropped (the fluctuations should satisfy Dirich-
let boundary conditions in order not to change the classical boundary conditions).
Further, ∆ denotes the two dimensional Laplacian with respect to the metric hαβ
(4.3.2.18) and R(2) is the corresponding scalar curvature (4.3.2.19). We observe that
the longitudinal fluctuations ξ 0 and ξ k drop out of the action. Hence, we can fix the
worldsheet diffeomorphisms via
ξ 0 = ξ k = 0. (4.3.2.32)
First, we need to explain some of the notation. The index â = 0, . . . , 9 labels the
tangent space coordinates of AdS5 × S 5 . Later, we will use an index a = 0, . . . , 4 for
4. Wilson loop computation 158
the tangent space of AdS5 and a′ = 5, . . . , 9 for the tangent space of S 5 . The vielbein
with a worldsheet index is
where X µ is the position of the background string. The indices I, J = 1, 2 label the
two target space supersymmetries. The derivative Dα is defined as
I IJ 1 µ ab ab i IJ
(Dα θ) = δ ∂α + (∂α X ) ωµ γ µ a a
− ǫ (∂α X ) Eµ γ θ J
4 2
i
≡ ∇α θ I − ǫIJ (∂α X µ ) Eµa γ a θ J . (4.3.2.35)
2
Here, ωµab denotes the target space spin connection and we have used the fact that our
′
background is trivial in S 5 directions. The gamma matrices γ̂ â = γ a , iγ a satisfy
SO (4, 1) and SO (5) Clifford algebras, respectively. The θ I are sixteen component
′
spinors each. They are conveniently labeled by a double spinor index θ αα where α is
a spinor index in the tangent space of the AdS5 , and α′ a spinor index in the tangent
′
space of S 5 . The γ a and γ a are four times four matrices tensored with four times
four identity matrices. (In the following we will suppress target space spinor indices
in order to avoid confusion with worldsheet indices which are also labeled by small
Greeks.)
We do not intend to give a derivation of (4.3.2.33) but let us have a brief look
at its structure before proceeding. Expression (4.3.2.35) is a tensor (density) with
indices αβ contracted either with hαβ or ǫαβ . The terms with hαβ can be thought
of as arising from the replacement (2.1.1.36)8 whereas the ǫαβ contracted terms come
from the Wess Zumino term (2.1.1.38) needed for kappa symmetry. The details differ
from the discussion in section 2.1.1.3 due to the different target space geometry and
the RR four form flux.
For our background, the Lagrangian (4.3.2.33) can be written in a compact way
! !
√ αβ
1 2
2iEµa (∂α X µ ) γ a P− ∇β 1−B θ1
LF = − −h θ̄ , θ̄ αβ , (4.3.2.36)
−1 − B 2iEµa (∂α X µ ) γ a P+ ∇β θ2
γ − ψ 1 = 0 , γ + ψ 2 = 0. (4.3.2.41)
We assume that the integration over spinors not satisfying (4.3.2.41) cancels the volume
of kappa transformations appearing as a normalization factor in the functional integral.
This may be problematic, and we will comment on this issue later. Spinors satisfying
the kappa fixing condition are then governed by the Lagrangian
! !
√ iγ +∇
+ 2 ψ 1
LF = − −h ψ̄ 1 , ψ̄ 2 , (4.3.2.42)
−2 iγ − ∇− ψ2
1 U2
∇± = eτ0 ∇τ ± eσ1 ∇σ = ∇τ ± 03 ∇σ , (4.3.2.43)
U U
where eτ0 and eσ1 are two dimensional (inverse) vielbein components obtained from the
square roots of the diagonal elements of hαβ . Note also that the covariant derivative
simplifies when acting on spinors satisfying (4.3.2.41). Defining partial tangent space
derivatives analogously to (4.3.2.43) one finds
ω± I
∇± ψ I = ∂± ± ψ , (4.3.2.44)
2
where
are tangent space components of the two dimensional spin connection ωα01 computed
from the zweibeinen defined in (4.3.2.43). Let us further define the matrices
! !
0 0 0 −γ 0
ρ+ = , ρ− = . (4.3.2.46)
γ0 0 0 0
4. Wilson loop computation 160
These are the same matrices as in (2.1.1.17) with i replaced by γ 0 /2. Finally, we
rewrite (4.3.2.42) in a suggestive way as follows
! !
√ −2 iγ − ∇− ψ1
LF = − −h ψ̄2 , −ψ̄1
−iγ + ∇+ −2 ψ2
! !
√ −2 2iγ 0 ∇− ψ1
= − h ψ̄2 , −ψ̄1
−2iγ 0 ∇+ −2 ψ2
!
√ ψ1
= 2 −h ψ̄2 , −ψ̄1 (iρm ∇m + 1) , (4.3.2.47)
ψ2
where in the second line (4.3.2.41) has been used and a repeated index m stands
for the sum over the labels + and −. Comparison with the expressions in section
2.1.1.2 shows that the part of the action containing target space spinor fluctuations
‘metamorphosed’ into an action for worldsheet spinors after imposing the kappa fixing
condition (4.3.2.41). The difference is that the derivative contains the spin connection
due to the non trivial worldsheet metric hαβ (4.3.2.18), and the mass terms appearing
due to the constant non vanishing curvature of the AdS space.
Now we have collected all the information needed to express the second order fluctu-
ation contribution to (4.3.2.22) in terms of determinants of two dimensional differential
√
operators. (For Dirac operators one uses the formal identity det A = det A2 .)
Integration over the fluctuations leads to determinants of operators which can be
read off from (4.3.2.30), (4.3.2.31) and (4.3.2.47). The corrected expression for the
Wilson loop reads
−T Eclass det −∆F − 14 R(2) + 1
W [C] = e 1 5 . (4.3.2.48)
det (−∆ + 2) det 2 −∆ + 4 − R(2) det 2 (−∆)
The exponential is the classical contribution with Eclass given by (4.3.2.14), (4.3.2.20)
and (4.3.2.21). Note also that the operator appearing in the numerator of (4.3.2.48) is a
four times four matrix. The Laplacian acting on worldsheet fermions ∆F is η mn ∇n ∇m .
Unfortunately it is not known how to evaluate the determinants in (4.3.2.48) exactly.
What is known exactly are the divergent contributions. These are given in (2.1.3.33).
They are of the form
Z
√
Ediv ∼ d2 σ −hR(2) . (4.3.2.49)
Comparing with (4.3.2.20), we find that this divergence renormalizes the self-energy
which is infinite anyway. A correction to the Coulomb charge of the quarks will be
finite. Unfortunately, we cannot give it in a more explicit way (as a number).
In addition, there is also a conceptual puzzle with the divergent contribution.
Although for our problem it is not relevant, it should not be there. The argument
4. Strings at a TeV 161
that something might have gone wrong goes as follows. The string action is equivalent
to a Polyakov type action, at least at a classical level (see section (2.1.1.1)). The
Polyakov action is conformally invariant, and in a consistent string background the
conformal invariance should not be broken by quantum effects. Therefore, divergences
which introduce a cutoff (or renormalization group scale) cannot occur. Indeed, it
was argued in[149] that a treatment analogous to ours but with a Polyakov instead
of the Nambu-Goto action leads to a finite result. This treatment is a bit more
complicated since the worldsheet metric appears as an independent field which also
fluctuates. The advantage is however that subtle contributions due to the occurring
integral measures are well understood. Such contributions are typically of the structure
(4.3.2.49)[16, 193, 328]. (Note, however, that if the worldsheet metric is identified with
the induced metric, the term in (4.3.2.49) is not really distinguishable from (∂X)2
terms (see e.g.[155])). In our derivation, we have mentioned already two places where
nontrivial measure contributions could arise. This could happen when we cancel the
integration over the longitudinal fluctuations against the volume of the worldsheet
diffeomorphisms and in the kappa fixing procedure. Unfortunately, the Nambu-Goto
case is less understood than the Polyakov formulation. (For a recent attempt to fix
the functional measures in the bosonic part see[347].) Fortunately, the result of the
better understood calculation in the Polyakov approach is identical to the one given
here (up to the irrelevant divergence)[149].
With these open questions we close our discussion on the AdS/CFT correspon-
dence. The reader who wants to know more will find some references in chapter 6.
where we drop details which are not relevant for the present estimate on scales. In
(4.4.0.1) lh is the heterotic string scale (set by α′ ), gh the heterotic string coupling
(fixed by the dilaton vev), and V is the volume of the compact space. The quantities
in which four dimensional physics is usually described are the four dimensional Planck
mass Mp and the gauge coupling gY M . These are related to the input data (gh , lh and
V ) as follows
V 1 V
Mp2 = , = . (4.4.0.2)
gh2 lh8 gY2 M gh2 lh6
Expressing gh in the first equation in terms of the second equation and further defining
the string mass scale as Mh = 1/lh the above equation can be rewritten as
√
V
Mh = gY M Mp , gh = gY M 3 . (4.4.0.3)
lh
Now we assume that a gY M ∼ 0.2 is a realistic value. (This is the gauge coupling of
the minimal supersymmetric standard model at the GUT scale.) Plugging gY M = 0.2
into the first equation in (4.4.0.3) we find that the heterotic string scale is
i.e. of the order of the Planck scale. The second equation in (4.4.0.3) implies that the
compact space is also of the Planck size if we want to stay within the region where the
string coupling is small.
Now let us investigate how the above estimates on scales are altered in a the-
ory containing branes. Phenomenologically interesting models arise also as orientifold
compactifications of type II theories. As we have seen in section 2.4, these contain typ-
ically D-branes on which the gauge interactions are localized whereas the gravitational
sector corresponds to closed string excitations which propagate in all dimensions. As-
suming that the gauge sector (and charged matter) is confined to live on Dp-branes
the effective action for the orientifold compactification will be of the form
Z Z
1 1
10
Sori = d x 2 8 R − dp+1 x 2
p−3 trF , (4.4.0.5)
gII lII gII lII
where lII and gII are the string scale and coupling of the underlying type II theory,
respectively. Assuming further that our orientifold construction is such that the com-
pact space has dimensions which are transverse to all relevant D-branes we denote by
V⊥ the volume of the compact space transverse to the branes and by Vk the volume of
the compact space longitudinal to the branes (such that the overall compact volume
is V = V⊥ Vk ). With this notation the four dimensional action reads
Z Z
4 Vk V⊥ Vk
Sori = d x 2 8 R − d4 x 2
p−3 trF , (4.4.0.6)
gII lII gII lII
4. Strings at a TeV 163
from which we obtain the four dimensional Planck length lp and gauge coupling gY M
1 Vk V⊥ 1 Vk
= 2 8 , = . (4.4.0.7)
2
lp gII lII gY2 M p−3
gII lII
Hence, the four dimensional Planck mass (Mp = 1/lp ) and the string coupling gII are
−2
v⊥ lII
Mp2 = , gII = gY2 M vk , (4.4.0.8)
vk (gY M )4
where
3−p p−9
vk = Vk lII , v⊥ = V⊥ lII (4.4.0.9)
are dimensionless numbers describing the size of the compact space in string scale
units. The relations (4.4.0.8) allow to take the string length lII larger than the four
dimensional Planck length lp . This can be achieved by taking v⊥ large. The size of the
parallel volume is taken to be of the “string size”, i.e. vk ∼ 1. If the parallel volume is
smaller than the string size, we T-dualize with respect to the smaller dimension. This
dimension will then contribute to the perpendicular volume since the string changes
boundary conditions. Hence, the vk < 1 case is T dual to the considered case of large
v⊥ . On the other hand if vk > 1, the second equation in (4.4.0.8) tells us that in this
case the string coupling becomes strong, and our description breaks down. (Moreover,
it is problematic for gauge interactions to be compactified on large volumes because
the corrections to the four dimensional gauge interactions are usually ruled out by
experimental accuracy.)
Let us analyse in some detail what happens if we choose a TeV for the string scale.
This is about the lowest value which is just in agreement with experiments. (For a
lower value massive string excitations should have shown up in collider experiments.)
With Mp ∼ 1016 TeV we find
1
v⊥ ∼ 1028 → V⊥ = 1028 . (4.4.0.10)
(TeV)9−p
The Planck length is about 10−33 cm and hence in our units one TeV corresponds to
1/ 10−18 m . Thus we obtain
For the case p = 8 (one extra large dimension) we obtain that the perpendicular
dimension is compactified on a circle of the size
Such a value is certainly excluded by observations. (In the next subsection we will
compute corrections to Newton’s law due to Kaluza-Klein massive gravitons and see
that the size of the compact space should be less than a mm.) For p = 7 we obtain
(distributing the perpendicular volume equally on the two (extra large) dimensions)
p = 7 → R⊥ ∼ 0.1mm. (4.4.0.13)
This value is just at the edge of being experimentally excluded. The situation improves
the more extra large dimensions there are. For example in the case p = 3 (and again
6 ))
a uniform distribution of the perpendicular volume on the six dimensions (V⊥ = R⊥
we obtain
p = 3 → R⊥ ∼ 10−10 m, (4.4.0.14)
exp
which is in good agreement with the experimental value (R⊥ = 0...0.1mm).
We have seen that D branes allow the construction of models where the string
scale is as low as a TeV. (Note also, that in the above discussion we can perform
T-dualities along the string sized parallel dimensions. This changes p but leaves the
large extra dimensions unchanged. Actually, it might be preferable to have p = 3 in
order to avoid Kaluza–Klein gauge bosons of a TeV mass.) This gives the exciting
perspective that string theory might be at the horizon of experimental discovery. In
near future collider experiments, massive string modes would be visible. In addition,
the extra large dimensions could be also discovered soon. This can happen either by
the production of Kaluza-Klein gravitons in particle collisions or by short distance
Cavendish like experiments. However, it might as well be the case that models with
less “near future discovery potential” are realized in nature.
Apart from the prospect of being observed soon, strings at a TeV scale are interest-
ing for another reason. If the string scale is at a TeV, we would call this a fundamental
scale. Thus the hierarchy problem would be rephrased. With the fundamental scale at
a TeV we should wonder why the (four dimensional) Planck scale is so much higher, or
why gravitational interactions are so much weaker than the other known interactions.
This hierarchy is now attributed to the size of the extra large dimensions. Super-
symmetry may not be necessary to explain the hierarchy between the Planck scale
and the weak scale. Therefore, in the above models supersymmetry could be broken
already by the compactification. In such models the question of stability is typically
a problematic issue.
The above considerations are also interesting if one does not insist on a direct
connection to string theory. If one just starts ‘by hand’ with a higher dimensional
setup containing branes, one would also obtain the first equation in (4.4.0.8). In this
4. Corrections to Newton’s law 165
case, one calls lII the higher dimensional Planck length, which in turn can be chosen
to be 1/TeV.
y ≡ y + 2πR. (4.4.1.15)
The n + 4 dimensional Newton potential Vn+4 of a point particle with mass µ located
at the origin is given by the equation
where ∆n+3 is the three dimensional flat Laplacian and Ωn+2 is the volume of a unit
n + 2 sphere. Any solution to (4.4.1.16) should be periodic under (4.4.1.15). This can
be ensured by expanding the potential in terms of eigenfunctions ψ k (y) of a Laplace
operator. The eigenvalue equation is
where k is an n dimensional vector with integer entries. We expand the higher di-
mensional Newton potential into a series of the eigenfunctions with r = |x| dependent
coefficients
X
Vn+4 = φk (r) ψk (y) . (4.4.1.19)
k
4. Corrections to Newton’s law 166
Plugging this ansatz into equation (4.4.1.16) determines the Fourier coefficients9
where the four dimensional and the higher dimensional Newton constant are related
via
Ωn Gn+4
G4 = . (4.4.1.22)
2 (2πR)n
For k = 0 we obtain the usual four dimensional Newton potential. The other terms
are additive Yukawa potentials. They arise due to the exchange of massive Kaluza
Klein gravitons.
Experimentalists usually parameterize deviations from Newton’s law via the ex-
pression [323]
G4 µ r
V4 (r) = − 1 + αe− λ . (4.4.1.23)
r
In the paper[323] the experimental values are discussed. These maybe outdated by
now but for us only the order of magnitude is important (and the fact that so far no
deviation from Newton’s law has been observed). Depending on the size of α an upper
bound on λ varying from the µm range to the cm range has been measured. This tells
us that a scenario with two extra large dimensions is almost excluded whereas setups
with more than two extra large dimensions are in agreement with the experimental
tests of Newton’s law.
9
Here, one uses the completeness relation satisfied by the ψk .
Chapter 5
In the last section of the previous chapter we have argued that branes allow for scenar-
ios with large extra dimensions transverse to the brane. This is because those extra
large dimensions can be tested only via gravitational interactions which are (due to
their weakness) measured only at scales down to about 0.1 mm. We obtained such
models via investigations of string theory. One could, however, just postulate the exis-
tence of branes (on which charged interactions are located). In this last chapter we will
take this latter point of view and not worry whether the setups we are going to discuss
have a stringy origin. Because in the presence of branes we can attribute the hierarchy
between the Planck and the weak scale to the size of the transverse dimensions, we
do not need supersymmetry in such setups. Without supersymmetry, quantum effects
usually create vacuum energies. A non vanishing vacuum energy on a brane will back
react on the geometry of the space in which the brane lives. Taking into account such
back reactions leads to so called warped compactifications. This means that the higher
dimensional geometry is sensitive to the position of a brane. The most prominent ex-
ample of such warped compactifications are the Randall Sundrum models which we
will discuss next.
167
5. The RS1 model with two branes 168
the form
where Sb1 and Sb2 denote the actions on the branes. For the bulk action we take five
dimensional gravity with a bulk cosmological constant,
Z Z π
√
Sbulk = d4 x dφ −G 2M 3 R − Λ , (5.1.1.2)
−π
where M denotes the five dimensional Planck mass and GM N is the five dimensional
metric. The branes are located in φ and we identify the brane coordinates with the
remaining 5d coordinates xµ , µ = 0, . . . , 3. Then the induced metrics on the branes
are simply
b1 b2
gµν = Gµν |φ=0 , gµν = Gµν |φ=π . (5.1.1.3)
We assume that fields being localized on the brane are in the trivial vacuum and
take into account only nonzero vacuum energies on the branes. Calling those vacuum
energies T1 and T2 , the brane actions read
Z p p
Sb1 + Sb2 = − d4 x T1 −g b1 + T2 −g b2 , (5.1.1.4)
where the first (second) term on the lhs is attributed to the first (second) term on
the rhs. Instead of working out the solutions to the system on an interval S 1 /Z2 it
is technically easier to construct a solution in a non compact space, such that the
solution is periodic in
φ ≡ φ + 2π, (5.1.1.5)
φ → −φ. (5.1.1.6)
A vacuum with this property yields then automatically a compact interval in φ. (The
equivalent1 and more complicated alternative is to define the theory on an interval from
the very beginning and take into account surface terms when deriving the equations
of motion as well as Gibbons Hawking[200] boundary terms (for a discussion in the
context of brane worlds see also[132, 133]).) With these remarks the Einstein equations
of motion read (capital indices run over all dimensions M, N = 0, . . . , 4)
√ 1
−G RM N − GM N =
2
" 2
#
1 √ X p
bi µ ν
− Λ −GGM N + Ti −g bi gµν δM δN δ (φ − φi ) , (5.1.1.7)
4M 3
i=1
1
I acknowledge discussions with Radoslaw Matyszkiewicz on this topic.
5. The RS1 model with two branes 169
|φ|
φ
−2π −π 0 π
with φ1 = 0 and φ2 = π. The delta functions appearing on the rhs of (5.1.1.7) are
defined on a real line. The most general metric ansatz possessing a four dimensional
Poincaré transformation as isometry is
We could rescale φ such that the rc dependence drops out, but that would change
the periodicity condition (5.1.1.5). Plugging this ansatz into the equations of motion
(5.1.1.7) yields (a prime denotes differentiation with respect to φ)
6σ ′ 2 Λ
= − , (5.1.1.9)
rc2 4M 3
3σ ′′ T1 T2
= δ (φ) + δ (φ − π) . (5.1.1.10)
rc2 4M 3 rc 4M 3 rc
This ensures that the solution is even under φ → −φ. In order to incorporate
(5.1.1.5), we define the modulus function on the real line by the periodic continu-
ation of (5.1.1.12). The resulting function is drawn in figure 5.1. Away from the
points at φ = 0 and integer multiples of π, the second derivative of σ vanishes and
5. The RS1 model with two branes 170
(5.1.1.10) is fulfilled in those regions. In order to take into account the delta function
sources in (5.1.1.10), one integrates this equation over an infinitesimal neighborhood
around the location of the brane sources. This gives rise to the constraints
Λ
T1 = −T2 = 24M 3 k , with k2 = − (5.1.1.13)
24M 3
on the parameters of the model. These constraints can be thought of as fine tuning
conditions for a vanishing effective cosmological constant in four dimensions. We will
come back to this point in section 5.2.3. Our final solution is
In principle we should also allow the four-four component of the metric rc2 to fluctuate.
Since rc is an integration constant, such fluctuations will be seen as massless scalars
in the effective four dimensional theory. This is a common problem known as moduli
stabilization problem. We will assume here that some unknown mechanism gives a
mass to the fluctuations of G44 and take it to be frozen at the classical value rc2 . The
Kaluza-Klein gauge fields Gµ4 are projected out by the Z2 . Plugging (5.1.1.15) into the
action and integrating over φ yields the effective action for four dimensional gravity
Z
√
Sef f = Mp2 d4 x −gR(4) (g) , (5.1.1.16)
where R(4) (g) denotes the four dimensional scalar curvature computed from gµν =
ηµν + hµν and the four dimensional Planck mass Mp is given by
Z π
2 3 M3
Mp = M rc dφe−2krc |φ| = 1 − e−2krc π . (5.1.1.17)
−π k
This tells us that choosing five dimensional scales of the order of the Planck scale gives
the correct order of magnitude for the four dimensional Planck scale.
5. The RS1 model with two branes 171
Now, let us consider matter living on the branes. On the first brane located at
φ = 0, the induced metric is just the Minkowski metric and Lagrangians for matter
living on that brane will just have their usual form. On the other hand, matter living
on the second brane (located at φ = π) feels the φ dependence of the bulk metric. Let
us focus on a Higgs field being located at the second brane. Its action will be of the
form
Z 2
b2 4 −4krc π 2krc π µν † 2 2
SHiggs = d xe e η Dµ H Dν H − λ |H| − vo , (5.1.1.18)
where the overall exponential factor originates from the determinant of the induced
metric. Rescaling the Higgs field H such that the kinetic term in (5.1.1.18) takes its
canonical form induces the rescaling
This means that a symmetry breaking scale which is written as v0 into the model
effectively is multiplied by a factor of e−krc π . Repeating the above argument for any
massive field, one finds that any mass receives such a factor
when going to an effective description in which kinetic terms are canonically normal-
ized. Choosing krc ≈ 10 (which is roughly a number of order one), one can achieve
that the exponential in (5.1.1.20) takes Planck sized input masses to effective masses
of the order of a TeV. Hence, in the above model we can obtain the TeV scale from
the Planck scale without introducing large numbers, provided we live on the second
brane.
In the previous section, we have already mentioned that the internal metric component
G44 gives rise to a massless field in an effective description. This means that its vev
rc is very sensitive against any perturbation and rather unstable. For the discussion
of the hierarchy problem it is important that the distance of the branes rc is of the
order of the Planck length. Therefore, it is desirable to stabilize this distance, i.e.
to give a mass to G44 in the effective description. In the present section we briefly
present a proposal of Goldberger and Wise how a stabilization might be achieved via
an additional scalar living in the bulk. We will neglect the back reaction of the scalar
field on the geometry. This means that we just consider a scalar field in the RS1
background constructed in the previous section. The action consists out of three parts
where Sbulk defines the five dimensional dynamics of the field and Sb1 and Sb2 its
coupling to the respective branes. We choose
Z Z π
1 √
Sbulk = d4 x dφ −G GM N ∂M Φ∂N Φ − m2 Φ2 , (5.1.1.22)
2 −π
where Φ is the scalar field and GM N is given in (5.1.1.14). The coupling to the branes
is taken to be
Z p 2
Sb1 = − d4 x −gb1 λ1 Φ2 − v12 , (5.1.1.23)
Z p 2
Sb2 = − d4 x −gb2 λ2 Φ2 − v22 , (5.1.1.24)
where vi and λi are dimensionfull parameters whose values will be discussed below.
With the ansatz that Φ does not depend on the xµ for µ = 0, . . . , 3 the equation of
motion for the scalar is
where the integration constants A and B will be fixed below. Plugging this solution
back into the Lagrangian yields an rc dependent constant, i.e. a potential for the
distance of the two branes,
V (rc ) = k (ν + 2) A2 e2νkrc π − 1 + k (ν − 2) B 2 1 − e−2νkrc π
2 2
+λ1 Φ (0)2 − v12 + λ2 e−4krc π Φ (π)2 − v22 . (5.1.1.27)
Because of the dependence of Φ on the modulus function (see figure 5.1) the second
derivative in the first term in (5.1.1.25) will lead to delta functions whose argument
vanishes at the position of the branes. Matching this with the delta function source
terms in (5.1.1.25) yields equations for the integration constants A and B. Instead
of writing down and solving those equations explicitly we suppose that λ1 and λ2 are
large enough for the approximation
m2
ǫ= ≪ 1. (5.1.1.31)
4k
In evaluating the potential V (rc ) (5.1.1.27), we neglect terms of order ǫ2 but do not
treat ǫkrc as a small number. This yields
2 ǫ
V (rc ) = kǫv12 + 4ke−4krc π v2 − v1 e−ǫkrc π 1+
4
−(4+ǫ)krc π −ǫkrc π
−kǫv1 e 2v2 − v1 e . (5.1.1.32)
In figure 5.2, we have drawn the potential in a neighborhood of the minimum (using
Maple). (What is actually drawn is V − kǫv12 .) With the appropriate choice for the
scales, the minimum of the potential is clearly visible. One should note, however, the
exponentially suppressed height of the right wall of the potential. If we had chosen
a larger scale for the drawing, figure 5.2 would just show a runaway potential which
rapidly reaches its asymptotic value. This might be a drawback of the stabilization
mechanism.
The expression for the stable distance between the branes (5.1.1.33) shows that
no extreme fine tuning is needed in order to obtain the wanted value of about ten
for krc . It remains to investigate whether the various approximations (including the
neglection of the back reaction) are sensible. This investigation has been carried out
in[215] by estimating the size of next to leading order corrections. The result is that
the approximations are fine.
To close this section, we should mention that the described stabilization method
is often called “Goldberger Wise mechanism” in the literature. We preferred to use
the term “proposal” because we are not certain that this mechanism is the commonly
established method for solving the problem of moduli stabilization. We decided to
present a brief description of the method because it is one of the most prominent lines
of thought in the context of the Randall Sundrum model. In general, the problem of
moduli stabilization is not very well understood.
5. The RS2 model with one brane 174
6e–09
4e–09
2e–09
–2e–09
–4e–09
–6e–09
y → −y. (5.1.2.34)
Further, we remove Sb2 from the action (5.1.1.1). Since the extra dimension is not
compact, we can perform rescalings of y in order to remove the rc dependence of the
ansatz (5.1.1.8). Without loss of generality we take rc = 1. Thus, in the single brane
case, the solution for the metric is
With a non compact extra dimension, one may worry that gravity is five dimensional
now. However, taking the rc → ∞ limit of (5.1.1.17), one finds that the effective
four dimensional Planck mass is finite. This means that the graviton zero mode is
normalizable and yields a four dimensional Newton law. Apart from the zero modes,
there will be also massive gravitons who lead to corrections of Newton’s law. In the
following subsection we will investigate these corrections.
where V denotes a fluctuation. In the presence of a point particle with mass µ at the
origin, the non relativistic limit of the linearized equation for V reads
h i
∆3 + e−2k|y| ∂y2 + 4kδ (y) − 4k2 V (x, y) = Gµ δ(3) (x) δ (y) , (5.1.2.37)
where G is the five dimensional Newton constant. The fact that V is indeed the Newton
potential can be confirmed by studying the geodesic equation of a point particle probe
and comparing it with the Newton equation of motion. The equation (5.1.2.37) is the
warped geometry analogon of equation (4.4.1.16). (The normalization of the higher
dimensional Newton constant is not really important here.) It is useful to redefine the
coordinate y according to
Analogous to section 4.4.1 we plan to expand the solution V into a series of eigenfunc-
tions, i.e. in the case at hand we are looking for solutions of the differential equation
15k2
∂z2 − 2
2 + 3kδ (z) ψ (m, z) = −m ψ (m, z) , (5.1.2.41)
4 (k |z| + 1)
5. The RS2 model with one brane 176
where we expect a continuous eigenvalue m now, since the “internal space” is not
compact. Let us discuss first the zero mode, i.e. the solution to (5.1.2.41) with m2 = 0.
The zero mode is found to be2
N0
ψ0 (z) ≡ ψ (0, z) = 3 , (5.1.2.42)
(k |z| + 1) 2
where N0 is an integration constant to be fixed later. Note that
Now, we take m > 0. For z > 0 the general solution to the above equation can be
written as a superposition of Bessel functions
r
1 1 1
ψ (m, z) = |z| + c1 J2 m |z| + + c2 Y2 m |z| + , (5.1.2.44)
k k k
where Jν denotes the Bessel functions of the first kind whereas Yν stands for the
Bessel functions of the second kind and c1,2 are constants to be fixed below. Because
the solution (5.1.2.44) is written as a function of |z|, the second derivative with respect
to z in (5.1.2.41) will yield a term containing a δ (z) (and other terms). One can fix
the ratio c1 /c2 by matching the factor in front of this delta function with the factor in
front of the delta function in (5.1.2.44). We will do this in an approximate way. The
most severe corrections to Newton’s law are to be expected from gravitons with small
m (because they carry interactions over longer distances). In matching the coefficients
of the delta functions, only a neighborhood around z = 0 matters. Therefore, we
replace the Bessel functions by their asymptotics for small arguments, which are
2
1 m2 |z| + k1
J2 m |z| + ∼ , (5.1.2.45)
k 8
1 4 1
Y2 m |z| + ∼ − 2 − . (5.1.2.46)
k πm2 |z| + 1 π
k
Plugging the asymptotic approximation into (5.1.2.44) and then into (5.1.2.41) one
finds that the overall coefficient in front of the delta function vanishes if
c1 4k2
= . (5.1.2.47)
c2 πm2
Hence, our general solution (5.1.2.44) reads
r
1 1 4k2 1
ψ (m, z) = Nm |z| + Y2 m |z| + + J2 m |z| + , (5.1.2.48)
k k πm2 k
2
In forthcoming expressions we will always imply that m > 0 when writing ψ (m, z). The zero
mode will be denoted by ψ0 (z) from now on.
5. The RS2 model with one brane 177
where we replaced c2 = Nm because this remaining integration constant will turn out
to depend on the eigenvalue m.
Recall that the extra dimension y (or z) is not compact. Thus the eigenvalue m is
continuous. Therefore, we normalize
Z
dz ψ (m, z) ψ m′ , z = δ m − m′ , (5.1.2.49)
The orthonormalization condition (5.1.2.49) fixes Nm . It turns out that the compu-
tation simplifies essentially in the approximation where the arguments of the Bessel
functions are large, since the corresponding asymptotics yields plane waves. Explicitly,
for large mz the Bessel functions are approximated by
r
√ 2 5π
zJ2 (mz) ∼ cos mz − , (5.1.2.52)
πm 4
r
√ 2 5π
zY2 (mz) ∼ sin mz − . (5.1.2.53)
πm 4
Because we are mainly concerned about large (> µm ≫ 1/Mp ) distance modifications
2
of Newton’s law we focus on the contribution of the “light” modes ( m
k2
≪ 1). (Recall
that k is of the order of the Planck mass.) Then (5.1.2.49) yields for the normalization
constant (for m > 0)
5
πm 2
Nm = . (5.1.2.54)
(4k2 )
Now, we expand V̄ (x, z) into eigenfunctions ψ0 (z) and ψ (m, z) with x dependent
coefficients ϕm (x)
Z ∞
V̄ (x, z) = ϕ0 (x) ψ0 (z) + dm ϕm (x) ψ (m, z) . (5.1.2.56)
0
5. The RS2 model with one brane 178
By plugging the ansatz (5.1.2.56) into (5.1.2.40), we find that for m ≥ 0 and r = |x|
Gµ −mr
ϕm (x) = − e am , (5.1.2.57)
r
with the constants am taken such that
Z
a0 ψ0 (z) + dm am ψ (m, z) = δ (z) . (5.1.2.58)
G4 = Gk, (5.1.2.60)
where once again we took into account only modes with m/k ≪ 1 such that we could
use the asymptotics (5.1.2.45) and (5.1.2.46) in order to evaluate ψ (m, 0). Finally,
performing the integral in (5.1.2.61) leads to
µ 1
V (x, 0) = −G4 1+ 2 2 . (5.1.2.62)
r r k
For k being of the order of the Planck mass (5.1.2.62) is in very good agreement
with the experimental values. This may look a bit surprising. Even though the extra
dimension is not compact, we obtain a four dimensional Newton potential for observers
who live on the brane at y = 0. This non trivial result finds its explanation in the
exponentially warped geometry. It is this geometry which is responsible for the fact
that the amplitude of the zero mode has its maximum at the brane and vanishes rapidly
for finite z. On the other hand, the massive modes reach their maximal amplitudes
asymptotically far away from the brane. Therefore, they have very little influence on
the gravitational interactions on the brane, although the masses of the extra gravitons
can be arbitrarily small.
In the following subsection we are going to rederive (5.1.2.62) in a different way.
5. The RS2 model with one brane 179
In section 4.3 we have described a duality between a field theory living on the boundary
of an AdS5 space and a theory living in the bulk of an AdS5 space. This correspondence
is sometimes called the holographic principle since it allows to reproduce bulk data
from boundary data (and vice versa). Now we are going to apply this principle to the
RS2 setup. Before doing so, we will establish that the RS2 setup has something to do
with an AdS5 space (namely it is a slice of an AdS5 space). To this end, we first write
down the RS2 metric (5.1.2.35) in terms of the coordinate z defined in (5.1.2.38). This
results in
1
ds2RS2 = µ ν 2
2 ηµν dx dx + dz . (5.1.2.63)
(k |z| + 1)
For symmetry reasons the coordinate z can be restricted to the half interval between
zero and infinity. The singularity at z = 0 is caused by the brane.
Now, let us recall from section 4.3 that the AdS5 metric is (see (4.3.1.12))
U2 2 dU
2
ds2AdS = ηµν dx µ
dx ν
+ R . (5.1.2.64)
R2 U2
Changing the coordinates according to (−R < z < ∞)
R2
U= (5.1.2.65)
z+R
yields an AdS metric of the form
1
ds2Ads = ηµν dx
µ
dxν + dz 2 . (5.1.2.66)
z 2
1+ R
Comparing (5.1.2.66) with (5.1.2.63), we observe that the RS2 geometry describes a
slice of an AdS5 space. The radius of the AdS5 space is 1/k, and the space is cut off at
z = 0. Since the boundary of the AdS5 space is situated at U → ∞, the cutoff at z = 0
means that we lost the region between U = R and the boundary. Hence, the position of
the brane in the RS2 setup can be viewed as an infrared cutoff for gravity on an AdS5
space. This suggests that we may apply the AdS/CFT conjecture on the RS2 scenario.
(Note however, that we do not have any supersymmetry now. Without supersymmetry
the AdS/CFT conjecture has passed less consistency checks. Nevertheless, let us
assume that the conjecture is correct also without supersymmetry.) The field theory
dual of the RS2 setup is thus a conformal field theory with a UV cutoff3 given by
k. (The cutoff actually breaks the conformal invariance. The conformal anomaly
3
Recall from section 4.3.2.1 that an IR cutoff in the bulk theory corresponds to a UV cutoff in the
dual field theory.
5. The RS2 model with one brane 180
induces a coupling of the field theory to gravity.) In particular, we plan to employ the
AdS/CFT duality conjecture for the computation of corrections to Newton’s law. As
a preparation let us sketch how Newton’s potential is related to the gravity propagator
in four dimensions. If we did not have an extra dimension, the gravity propagator in
momentum space is (up to a polarization tensor) 1/ Mp2 p2 . The Newton potential
can be obtained from this propagator by formally setting the p0 component to zero
and Fourier transforming4 with respect to the spatial momentum components. The
result in position space is then 1/(Mp2 r). Therefore, we will use the AdS/CFT duality
conjecture to compute the corrected graviton propagator and deduce the corrected
Newton potential via the above description.
The dual picture for the RS2 setup is that we have four dimensional gravity plus
the CFT dual of gravity on AdS5 with a UV cutoff k. Corrections to four dimensional
gravity are caused by the interaction of gravity with the CFT. The effective graviton
propagator is obtained by integrating over the CFT degrees of freedom. The one loop
corrected graviton propagator will be schematically of the form
1 1
1 + hTCF T (p) TCF T (−p)i 2 2 , (5.1.2.67)
Mp2 p2 Mp p
where TCF T stands for the energy momentum tensor of the CFT dual. (The coupling
of gravity to the CFT fields is given by the energy momentum tensor.) For any four
dimensional CFT, the two point function of the energy momentum tensor is fixed to
be of the form
where we imposed that the UV cutoff is k. We will not derive this result here, but
just give two comments. First, notice that (5.1.2.68) is the four dimensional analogon
of (2.1.3.51). The number c quantifies the conformal anomaly. The second remark
is, that the reader may get some impression on how the expression (5.1.2.68) arises
by computing it explicitly for pure gauge theory. We are interested in the order of
magnitude of c. This has been computed in[241] to be
M53 Mp2
c≈ = , (5.1.2.69)
k3 k2
where M5 denotes the five dimensional Planck mass and the radius of the AdS space
dual to the CFT is 1/k. In the second equality of (5.1.2.69) we used the relation
4
We use the following prescription for performing the Fourier transformation. Transforming the
equation ∆3 f (x) = δ (3) (x) one finds that 1/p2 transforms into 1/r. Later we will also have to compute
Fourier transforms with additional powers of p in the numerator or denominator. An additional power
of p in the numerator transforms into ∂r whereas powers of p in the denominator can be generated by
∂p which in turn transforms into r.
5. The RS2 model with one brane 181
between the five and four dimensional Planck mass (M5 and Mp ) which, in the RS2
setup, is obtained by taking rc → ∞ in (5.1.1.17).
The corrected Newton potential is obtained by setting formally p0 to zero in
(5.1.2.67) and performing a three dimensional Fourier transformation to the posi-
tion space. Thus, the effect of integrating over the CFT fields results in the following
replacement of the Coulomb (Newton) potential
1 1 1
− →− 1+ 2 2 . (5.1.2.70)
r r k r
This result agrees with the expression (5.1.2.62) computed in the previous section.
Thus, we have learned that integrating over the CFT fields yields the same corrections
to four dimensional gravity as taking into account the massive “Kaluza-Klein” gravi-
tons. Employing the AdS/CFT correspondence, the computational effort decreases
substantially. We will make use of this fact when we combine the RS1 with the RS2
scenario in the next subsection.
In the previous two subsections we have seen that the RS2 setup has the exciting
feature of giving rise to effectively four dimensional gravitational interactions even
though the extra dimension is not compact. On the other hand, we observed before
that the RS1 model is capable to explain the hierarchy between the Planck scale and
the weak scale without introducing large numbers. How can we combine these two
models? We should introduce a brane with the observers at y = πrc into the RS2
setup. However, this brane should not cause a change of the RS2 metric (5.1.2.35).
The observers on the additional brane (at y = πrc ) can achieve this by performing a
fine tuning such that the vacuum energy on their brane vanishes. In the following we
will call the brane at y = πrc the SM (Standard Model) brane. The SM brane can
be viewed as a probe in the RS2 background. The hierarchy can now be explained in
the same way as it is explained in the RS1 setup. What we should worry about are
the gravitational interactions as viewed by an observer on the SM brane. In principle,
these can be computed along the lines of section 5.1.2.1. The situation is, however,
slightly more complicated since the approximation has to be refined. In particular,
replacing the Bessel functions by their plane wave asymptotics in the computation
of Nm is too rough an estimate. Now, this would imply that the observer on the
SM brane sees the Bessel functions as plane waves. As argued in[332] this is not the
case, in particular for the light continuum modes. The authors of[332] refined the
5. The RS2 model with one brane 182
for the Newton potential observed on the SM brane. Here, Mw is of the order of a TeV
if we take rc such that the hierarchy problem is solved. Instead of going through the
tedious refinement of the approximations performed in section 5.1.2.1, we employ the
AdS/CFT correspondence to motivate (5.1.2.71). The introduction of the SM brane
modifies the RS2 dual such that it consists out of four dimensional gravity, the CFT
dual of the RS2 AdS5 slice and the Standard Model of the probe brane. Note that
yc = rc π is translated to U0 −Uc = T eV in the course of the coordinate transformations
(5.1.2.38) and (5.1.2.65), where U0 denotes the position of the brane at the origin and
Uc the position of the SM brane. This means that SM fields and CFT fields interact via
fields with masses of the order of a T eV .5 Integrating out those fields yields effective
coupling terms between SM fields and CFT fields. (This is analogous to generating
the Fermi interaction via integrating out the W and Z bosons.) The structure of the
possible interaction terms is restricted by symmetries to[37]
1 µν
T TµνCF T . (5.1.2.72)
Mw4 SM
Note the similarity to the coupling of the SM fields to gravitons. Apart from charged
interactions, the SM fields interact via gravitons and via CFT fields. This suggests
that for an observer on the SM brane the effective graviton propagator is
1 1 1
1 + hTCF T (p) TCF T (−p)i + 8 hTCF T (p) TCF T (−p)i, (5.1.2.73)
Mp2 p2 Mp2 p2 Mw
where the first two terms are the same as in (5.1.2.67), and the last term means
that the observer will interpret the interaction (5.1.2.72) as gravitational interaction.
In computing the contribution due to the last term we use (5.1.2.68). Applying the
recipe of the previous section, we obtain out of the propagator the modified Newton
potential (5.1.2.71). This potential is still in agreement with the observational bounds
on deviations from Newton’s law. Hence, adding a probe brane at y = πrc in the
RS2 setup one obtains a model which explains the hierarchy and possesses effectively
four dimensional gravitational interactions, even though there is a non compact extra
dimension. However, we should remark that we discussed the setup only classically and
showing its stability against quantum corrections may be a problematic issue. This
5
One may view the field theory dual as a stack of D-branes on which the CFT lives and the SM
probe brane separated by a distance 1/T eV from the CFT branes. The interaction between the CFT
and the SM can be thought of as arising due to open strings stretching between the corresponding
branes.
5. Inclusion of a bulk scalar 183
where y is the coordinate labeling the extra dimension, and the sum over i stands for a
sum over the branes. The index bi at the integral means that y is fixed to the position
(yi ) of the brane bi . The function V (Φ) is a bulk potential for the scalar and fi (Φ) is
the coupling function of the scalar to the brane bi .
For later use let us also generalize the ansatz (5.1.1.8) to
where ḡµν denotes the metric of a four dimensional maximally symmetric space, i.e.
diag
(−1, 1, 1, 1) for M4
√ √ √
ḡµν = diag −1, e2 Λ̄t , e2 Λ̄t , e2 Λ̄t for dS4 . (5.2.1.3)
√ √ √
3 3 3
diag −e2 −Λ̄x , e2 −Λ̄x , e2 −Λ̄x , 1 for AdS 4
6
For simplicity we set the five dimensional Planck mass to one. It can be introduced if needed by
a simple analysis of the mass dimensions.
5. A solution generating technique 184
The constant Λ̄ is related to the constant curvature of the de Sitter (dS4 ) and the anti
de Sitter (AdS4 ) slices.
Let us first discuss the simplest case with Λ̄ = 0. As usual we consider fields which
depend only on y and denote a derivative with respect to y by a prime. The equations
of motion for Λ̄ = 0 are
8 ′′ 32 ′ ′ ∂V X ∂fi
Φ + AΦ − − δ(y − yi ) = 0, (5.2.1.4)
3 3 ∂Φ ∂Φ
i
2 2 ′ 2 V
6 A′ − Φ + = 0, (5.2.1.5)
3 2
4 ′ 2 1 X
3A′′ + Φ + fi δ(y − yi ) = 0. (5.2.1.6)
3 2
i
First, we analyze this system of equations in absence of the branes. We start with the
ansatz
A′ = W (Φ) . (5.2.1.7)
The jump conditions (5.2.1.10) and (5.2.1.11) are still of the same form. We observe
that a constant shift in A enters the equations of motion. Hence, there is no fine
5. Consistency conditions 186
tuning to be expected if we do not fix the value of Λ̄ in the ansatz. For completeness,
we note that the equations of motion can be reduced to a set of first order equations
like in the Λ̄ = 0 case. The corresponding first order equations are
27 1 ∂W (Φ) 2
V = − 12W (Φ)2 , (5.2.1.17)
4 γ(r)2 ∂Φ
A′ = γ(r) W (Φ) , (5.2.1.18)
9 1 ∂W (Φ)
Φ′ = − , (5.2.1.19)
4 γ(r) ∂Φ
s
Λ̄
γ(r) = 1+ e−2A . (5.2.1.20)
W (Φ)2
To find a solution to this system of first order equations looks more complicated than
in the Λ̄ = 0 case. The equation (5.2.1.17) now couples to the rest of the equations
due to the γ dependent factor.
where
is the metric on the four dimensional slices. It is taken to be independent of y, and the
background metric ḡµν is defined in (5.2.1.3). If we do not consider other fluctuations
than hµν , the action for four dimensional gravity will be of the general form
Z p
2
S4 = Mp d4 x −g̃ R̃(4) − λ , (5.2.2.3)
5. Consistency conditions 187
where R̃(4) is the four dimensional scalar curvature computed from g̃µν . The cosmo-
logical constant λ is fixed by the condition that g̃µν = ḡµν should be a stationary point
of (5.2.2.3). This yields
λ = 6Λ̄. (5.2.2.4)
We should also recall that the effective four dimensional Planck mass is given by
Z
Mp = dy e2A(y) ,
2
(5.2.2.5)
where A takes its classical value. The vacuum value of the Lagrange density in (5.2.2.3)
can be easily computed to be
L̄4 = Mp2 R̄(4) − λ = 6Λ̄Mp2 , (5.2.2.6)
where gµν is the metric induced on the brane (see (5.1.1.3)). The classical value of R
can be easily computed by taking the trace of (5.2.2.7) with the result
4 5 4X
R= (∂Φ)2 + V (Φ) + fi δ(y − yi ) . (5.2.2.9)
3 3 3
i
Plugging this into (5.2.1.1), we obtain the classical value for the four dimensional
Lagrangian
Z !
2 1 X
L̄4 = dy e4A V (Φ) + fi δ(y − yi ) , (5.2.2.10)
3 3
i
5. The cosmological constant problem 188
where it is understood that A and Φ satisfy the equations of motion. Comparing with
(5.2.2.6) and using (5.2.2.8), we obtain finally the consistency condition
Z
1
− dy e4A T00 + T55 = 6Λ̄Mp2 . (5.2.2.11)
3
We should emphasize again that (5.2.2.11) is just a consequence of the equations of
motion. For Λ̄ = 0, (5.2.2.11) implies that the vacuum energy density of the solution
has to vanish.
Before closing this subsection we want to describe an alternative way to obtain the
same (or equivalent) consistency conditions. First, we note that
!
′ n − 4 ′ 2 nV 1 X
A′ enA = enA Φ − + (n − 1) Λ̄e−2A− fi δ(y − yi ) . (5.2.2.12)
9 12 6
i
This can be easily checked with the equations (5.2.1.15) and (5.2.1.16). With the
expression (5.2.2.8) we rewrite (5.2.2.12) in the following way
′ nA ′ nA 1 0 n 1
5 −2A
Ae =e T + − T5 + (n − 1) Λ̄e . (5.2.2.13)
6 0 12 6
Assuming that for a consistent solution the integral over the total derivative on the
lhs of (5.2.2.13) vanishes we find
Z n Z
1 nA 0
− dy e T0 + − 1 T5 = 2 (n − 1) Λ̄ dy e(n−2)A .
5
(5.2.2.14)
3 2
We observe that for n = 4 this condition is identical to the previously derived condition
(5.2.2.11). In the next subsection we will discuss solutions with singularities. For those
solutions, one could argue that the preposition of condition (5.2.2.14) is not necessarily
satisfied. If there are singularities, an integral over a total derivative may differ from
zero, and one may not worry about (5.2.2.14) in such a case. For n = 4, we have shown
that (5.2.2.14) encodes the statement that the effective four dimensional cosmological
constant is compatible with the curvature of the four dimensional slices. This should
be the case also in the presence of singularities. We leave it as an exercise to verify
that the Randall Sundrum models satisfy all the consistency conditions.
Taking into account the leading order contribution of quantum field theory, one obtains
5.2.3.1 An example
The model we are going to discuss is a special case of (5.2.1.1) with a single brane at
y = 0 as well as V (Φ) ≡ 0 and f0 (Φ) = T ebΦ . Hence, the action reads
Z Z
√ 4 2 √
5
S = d x −G R − (∂Φ) − d4 x −gT ebΦ |y=0 , (5.2.3.3)
3
7
Recent observations seem to hint at a small but non zero constant. For the discussion of the fine
tuning problem this value is too small to be relevant.
5. The cosmological constant problem 190
where b and T are constants. In what follows we will focus on the case b 6= ± 34 . The
case b = ± 34 is similar and discussed in[268, 267], [36], [186, 187]. We take the ansatz
(5.2.1.2) with Λ̄ = 0. From equation (5.2.1.5) one finds that
1
A′ = ± Φ′ . (5.2.3.4)
3
We choose
(
1 ′
A′ = 3Φ , y<0
. (5.2.3.5)
− 13 Φ′ , y>0
The reader may verify that taking the same sign on both sides of the brane does not
lead to a consistent solution. The only other choice is to interchange the signs in
(5.2.3.5). This can be undone by redefining y → −y and hence the ansatz (5.2.3.5)
is general (for b 6= ± 34 ). The rest of the equations of motion is easily solved with the
result
(
3 4
4 log 3 y + c1 + d1 , y<0
Φ(y) = , (5.2.3.6)
− 43 log 43 y + c2 + d2 , y>0
where ci and di are integration constants. The condition that Φ should be continuous
at y = 0 fixes d2 in terms of the other integration constants. The jump conditions
(5.2.1.10) and (5.2.1.11) determine c1 and c2 in terms of d1 according to
2 3b 1 3b
= − − T ebd1 |c1 | 4 , (5.2.3.7)
c2 8 2
2 3b 1 3b
= − + T ebd1 |c1 | 4 . (5.2.3.8)
c1 8 2
Together with possible constant shifts in A, two integration constants are not fixed by
the equations of motion.
The next step is to ensure that an observer will experience four dimensional gravi-
tational interactions (plus possible small corrections). This is the case only if the four
dimensional Planck mass is finite. The expression for the four dimensional Planck
mass is given in (5.2.2.5). If the parameters (T and b) of the model are such that
there is no singularity at some y > 0 (y < 0) the integration region in (5.2.2.5) extends
to (minus) infinity. In one or both of these cases the four dimensional Planck mass
diverges, and an effective four dimensional theory decouples from gravity. This is not
what we are interested in since with decoupled gravity the problem of the cosmological
constant does not occur. Therefore, we have to choose our parameters such that there
are singularities at which we can cut off the integration over y. Explicitly this imposes
5. The cosmological constant problem 191
the conditions
1 3b
T − > 0,
2 8
1 3b
T − − < 0. (5.2.3.9)
2 8
These conditions are easy to satisfy without fine tuning of the parameters. So far,
it looks as if we have achieved to find a solution with vanishing four dimensional
curvature without the necessity of a severe fine tuning of input parameters.
It remains to check whether the consistency condition (5.2.2.11) is satisfied. Since
we have taken the ansatz with Λ̄ = 0, the condition states that the vacuum energy
density of our solution should vanish. The vacuum energy density is most easily
computed from (5.2.2.10). To be specific, we fix the integration constant in A via
A = 31 Φ for y < 0. Taking further into account that our background is static, we find
for the vacuum energy density
1 2 8 4
E = − T e4A+bΦ |y=0 = − e 3 d1 6= 0. (5.2.3.10)
3 3 4 − 3b
We see that the consistency condition is not satisfied. Since the condition of vanishing
vacuum energy density E = 0 is derived from the equation of motion, (5.2.3.10) im-
plies that the equations of motion are not solved. Indeed, with the parameter choice
(5.2.3.9), the second derivatives of Φ and A contain delta functions which are not can-
celed by source terms in the equations of motion. We have to cure this inconsistency
by adding additional source terms to the setup, i.e. to extend the single brane scenario
to a three brane scenario. From our considerations in section 5.2.1, we know already
that this will lead to fine tuning conditions on the input parameters. For illustrative
purposes, let us demonstrate the appearance of the fine tuning explicitly. We modify
our action (5.2.3.3) by two additional source terms, i.e.
S → S + S+ + S− , (5.2.3.11)
with
Z
S± = − d4 xT± eb± Φ |y=y± . (5.2.3.12)
The quantities b± and T± are now input parameters of the model. The value of y±
gives the locations of the singularities,
3 3
y− = − c1 , y+ = − c2 . (5.2.3.13)
4 4
5. The cosmological constant problem 192
The additional source terms give rise to four more jump conditions to be satisfied by
the solution. These jump conditions are
8 ′
Φ (y± + 0) − Φ′ (y± − 0) = b± T± eb± Φ(y± ) , (5.2.3.14)
3
1
∓ Φ′ (y± + 0) − Φ′ (y± − 0) = − T± eb± Φ(y) . (5.2.3.15)
2
Before solving these additional jump conditions we need to give a prescription how to
continue our solution beyond the singularities. There are several possible descriptions.
For example, one may continue in such a way that the setup becomes periodic in y.
The simplest choice is to effectively cut off the space at the singularities (at y = y± ) by
freezing the fields to the singularity values for y 6∈ [y− , y+ ] such that the first derivatives
vanish beyond the singularities. (The final conclusion is not affected by the particular
way of continuing the solution beyond the singularities.) With our prescription the
conditions (5.2.3.14) and (5.2.3.15) are solved by
4
b± = ± (5.2.3.16)
3
and
4 4
T− e− 3 d1 = T+ e 3 d2 = −2. (5.2.3.17)
One should recall that d2 is already fixed by the jump conditions at y = 0. We observe
that the input parameters need to be fine tuned.
The contribution of the branes at y = y± to the vacuum energy density is
1
E+ + E− = − T+ e4A+b+ Φ |y=y+ + T− e4A+b− Φ |y=y−
3
2 4 d1 8
= e3 , (5.2.3.18)
3 4 − 3b
where we have employed the jump conditions and fixed an integration constant in
A by the choice A = 13 Φ for y < 0. Hence, the contribution (5.2.3.10) is exactly
canceled by the additional branes and the model is consistent now. However, we failed
to construct a brane setup yielding a vanishing effective four dimensional cosmological
constant without fine tuning of the parameters. If the fine tuning is not satisfied, there
exist Λ̄ 6= 0 solutions[267]. (The situation is slightly different in the b = ± 34 case where
the possible value of Λ̄ is fixed by the bulk potential V , which needs to be fine tuned
to zero for Λ̄ = 0 to be a solution[187]. In addition there is a fine tuning due to the
necessity of additional branes for b = ± 43 , too.)
In the next subsection we will show that our failure to find a Λ̄ = 0 solution without
fine tuning is not caused by an unfortunate choice of the model we started with but
rather a generic feature of brane models with a bulk scalar.
5. The cosmological constant problem 193
5.2.3.2 A no go theorem
The prepositions for the no go theorem for a “brany” solution to the cosmological
constant problem are:
• The model does not contain singularities apart from the one corresponding to
the single brane source.
In a first step, we are going to show that these prepositions imply that the five
dimensional space must be asymptotically (for large |y|) an AdS space. Suppose the
warp factor asymptotically shows a power like behavior,
eA ∼ |y|−α . (5.2.3.19)
The four dimensional Planck mass is computed in (5.2.2.5). With a single brane and no
further singularities the integration is taken over y ∈ (−∞, ∞). A necessary condition
for
Z ∞
dy e2A < ∞ (5.2.3.20)
−∞
is
1
α> . (5.2.3.21)
2
On the other hand, equation (5.2.1.6) tells us that in the bulk (in particular asymp-
totically)
A = −k |y| , (5.2.3.23)
with k being a positive constant. In the following we will show that in this case there is
a fine tuning similar to the fine tuning of the RS2 model. Before going into the details,
let us sketch the outline of the proof. The asymptotic behavior (5.2.3.23) suffices to
5. The cosmological constant problem 194
reproduce the superpotential for all y. Plugging this into the matching conditions
(5.2.1.12) and (5.2.1.13) will show that the input parameters of the model need to be
fine tuned. Let us now present the details of the slightly tedious construction of W
from its asymptotics.
From equation (5.2.1.6) we learn that Φ must be asymptotically constant. We de-
note the asymptotic values of Φ by Φ±
c corresponding to the limits y → ±∞. Equation
(5.2.1.8) implies that
∂W
= 0. (5.2.3.24)
∂Φ |Φ=Φ±
c
∂2W ∂2W
>0 , < 0. (5.2.3.27)
∂Φ2 |Φ=Φ+
c ∂Φ2 |Φ=Φ−
c
This implies that the asymptotic values of V must be negative. Further note that
the asymptotic values of W are fixed in a unique way with the additional conditions
(5.2.3.25). So far, we know the asymptotic value of W in terms of the input parameters
and the asymptotics of the first derivative of W (5.2.3.24).
In order to compute the higher derivatives of W , it is useful to express the nth
derivative of V in terms of W via (5.2.1.9). The corresponding expression is
n
∂nV X n − 1 ∂ k W 27 ∂ n−k+2 W ∂ n−k W
= 2 − 12 . (5.2.3.29)
∂Φn k − 1 ∂Φk 4 ∂Φn−k+2 ∂Φn−k
k=1
5. The cosmological constant problem 195
This formula is most easily proven in the following way. First, apply the Leibniz rule
(F and G are arbitrary functions of Φ)
n
∂ n (F G) X n ∂ k F ∂ n−k G
= (5.2.3.30)
∂Φn k ∂Φk ∂Φn−k
k=0
∂W 2
on ∂Φ2
= 2W ∂W
∂Φ in order to show that
n
∂ n W 2 X n − 1 ∂ n−k W ∂ k W
= 2 . (5.2.3.31)
∂Φn k − 1 ∂Φn−k ∂Φk
k=1
In a second step use (5.2.3.31) with W replaced by its first derivative and redefine the
summation index k → n + 1 − k.
In the following we will employ (5.2.3.29) to compute the asymptotics of all deriva-
tives of W . Since there are no singularities between the brane and the asymptotic
region, this will enable us to expand W in a Taylor series yielding its value arbitrarily
close to the brane.
The second derivative of W needs some separate discussion. With the result
(5.2.3.24) we obtain the relation
" 2 #
27 ∂ 2 W ∂2W ∂2V
− 24W − = 0. (5.2.3.32)
2 ∂Φ2 ∂Φ2 ∂Φ2 ±
|Φ=Φc
This equation has real solutions for the asymptotics of the second derivative of W
provided that
∂2V 8 ±
> V Φ c . (5.2.3.33)
∂Φ2 |Φ=Φ±
c 9
Taking into account that the asymptotic value of W is fixed uniquely by (5.2.3.28)
and (5.2.3.25), and that the sign of the asymptotic value of the second derivative of
W is determined by (5.2.3.27), one finds that (5.2.3.32) can be solved in a unique way.
Note, that (5.2.3.25) and (5.2.3.27) imply
∂2V
> 0. (5.2.3.34)
∂Φ2 |Φ=Φ±
c
The computation of the higher derivatives of W in the large |y| region is somewhat
simpler. First, we notice that asymptotically on the rhs of (5.2.3.29) the nth derivative
of W is the highest occurring derivative (see (5.2.3.24)). Terms containing the nth
derivative correspond to k = 2, n. The expression (5.2.3.29) evaluated at Φ±
c takes the
form (n > 2)
n
∂nV ∂ W 27 ∂ 2 W
= n − 24W + . . . , (5.2.3.35)
∂Φn |Φ=Φ±
c ∂Φn 2 ∂Φ2 |Φ=Φ±
c
5. The cosmological constant problem 196
where the dots stand for terms containing lower derivatives of W . The relation
(5.2.3.35) allows to determine all derivatives of W provided that the coefficient at
the nth derivative of W differs from zero. This is ensured by equations (5.2.3.32)
and (5.2.3.34). Indeed, requiring the coefficient in front of the nth derivative of W to
vanish yields
2
∂ V 32 1
=− 1− , (5.2.3.36)
V ∂Φ2 |Φ=Φ± c
9n n
which is not compatible with (5.2.3.34) and (5.2.3.28). We conclude that in the Taylor
expansion
P
∞ 1 ∂nW n
n=0 n! ∂Φn |Φ=Φ− (Φ − Φ−
c ) , y <0
c
W (Φ) = P∞ 1 ∂nW n (5.2.3.37)
n=0 n! ∂Φ |Φ=Φ+
n (Φ − Φ+
c ) , y >0
c
all the coefficients are fixed uniquely by the model parameters. Then the jump condi-
tion (5.2.1.12) will fix the value of Φ at y = 0, whereas (5.2.1.13) imposes generically
a fine tuning of the model parameters.
It may look somewhat disappointing to close a review with the proof of a no go
theorem. However, often no go theorems help to find a way leading to the desired
aim. This way should then start with a model not satisfying the prepositions of the
no go theorem. Indeed, there have been proposals for not fine tuned solutions with
Λ̄ = 0. These proposals are based on the idea of introducing more integration constants
without increasing the number of jump conditions. We provide the corresponding
references in section 6. Here, we should remark that (so far) there is no commonly
accepted solution to the cosmological constant problem within brane world setups.
The explanation of the observed value of the cosmological constant remains a great
challenge. Whether branes will be helpful in a solution of this problem has to be seen
in the future.
Chapter 6
Throughout the text I have already given some references. However, this I did only
when I felt that a direct hint on results obtained in the literature would be useful for
the reader at that particular point. Of course, these notes are based on many more
publications than already given in the text. In the present chapter, I will provide
all the references I used and give suggestions for further reading. However, there are
many more contributions to this field. I apologize to all those authors whose work
could have been listed but is not.
6.1 Chapter 2
6.1.1 Books
In [222, 223], [87], [331], [371], [269, 270] I list the textbooks on string theory of which
I am aware. In section 2.1 I used mainly[222] but also[331]. For the discussion of
orbifold planes in section 2.2 I borrowed some results presented in[371]. D-branes and
orientifolds are also covered in[371]. Since string theory is a conformal field theory the
book[191] may be also a useful reference. The subject of Calabi-Yau compactifications
entered the text rather as a side remark. Apart from the discussions presented in the
above mentioned textbooks on strings, the book[251] is perhaps a useful reference for
people who are interested in Calabi-Yau spaces. Let me also mention three books on
supersymmetry. Often the conventions of the standard reference[41] are used in the
literature. Ref.[456] contains (at least in its second edition) a discussion of supersym-
metry in two dimensions. Finally, [399] is not really a textbook but a collection of
papers dealing with supergravity in various dimensions. For each dimension there is
a summary of the possible supermultiplets.
197
6. Review articles 198
[296], [12, 11], [448], [207], [298]. T-duality has been also discussed for backgrounds
with non Abelian isometries e.g. in [126], [11, 13], [214], [297], [305], [160], [113, 114,
112], [17], [450], [9], [343], [242]. The T-duality relation between type IIA and type IIB
strings can be found in [136], [122], [57], [289]. The connection between compactified
E8 × E8 and SO (32) strings is presented in [206].
The techniques for orbifold compactifications of string theory have been devel-
oped in [137, 138]. More papers on orbifolds are (including explicit constructions of
phenomenological interest): [451], [237], [42], [286], [256], [257], [253], [174], [173],
[255], [318], [175]. T-duality for orbifold compactifications is for example discussed in
[312, 313].
The importance of D-branes was realized in[369], where the connection to BPS
solutions of supergravity was discovered. In the text I have given conditions imposed
by the requirement unbroken supersymmetry on the number of ND directions. More
generally, D-branes can intersect at certain angles[59]. The computation of D-brane
interactions is presented in [372], [92, 93]. References concerning the beta function
approach are given together with the other references for the beta function approach
to effective field theories, above. D-brane actions are also e.g. discussed in[400]. The
interchange of Dirichlet with Neumann boundary conditions via T-duality has been
pointed out in [122], [243], [217]. For general backgrounds, T-duality for open strings
with respect to abelian isometries is presented in[15], [143]. T-duality with respect to
non-Abelian isometries has been performed in [184], [77]. (The boundary Lagrange
multiplier has been introduced in [184, 185].) A different method of performing T-
duality transformations in general backgrounds has been proposed in [306]. The Wess-
Zumino term in the D-brane action has been derived (in steps) in [145], [320], [61, 62],
[218]. Our discussion of open strings and non commutative geometry follows closely
(the introductory section of) [418]. Constant B-fields and non commutative geometry
have been connected earlier in e.g. [147], [106], [404]. The connection between non
commutativity and the renormalization scheme is further investigated e.g. in [24, 23].
A more abstract conformal field theory approach to D-branes can for example be found
in [194], [388]. There are many more aspects of D-branes for which I would like to ask
the reader to consult one of the given reviews and the references therein.
Orientifolds were introduced in[396]. For early papers on orientifold constructions
see also [378], [216], [244], [63, 64]. The cancellation of divergences in string diagrams
of type I SO (32) strings is observed in[221]. The model of section 2.4.3.2 has been
first constructed in [63, 64]. The presentation in the text follows [204]. Indeed, it
has been the paper [204] which triggered an enormous amount of research devoted to
orientifolds. This research resulted in a lot of papers out if which I list only “a few”:
[120, 121], [203], [60], [68], [272], [469], [179], [178], [357], [5], [71, 70], [115], [377],
6. Chapter 3 200
[183], [254], [31], [25], [8], [26], [28], [27], [383], [302], [300], [301], [72], [73], [182], [382],
[74], [7, 6], [117, 116].
6.2 Chapter 3
As far as I know there are no books devoted to solutions of ten dimensional super-
gravity.
6.3 Chapter 4
The presented applications of branes are not a subject of a book. A discussion of
string dualities can be found in[371].
considered in[65], [129], [317], [19], [33]. The duality of N = 4 super Yang-Mills theory
is presented in [358], [239].
The standard review article on the AdS/CFT correspondence is[2]. Two more
introductory notes are listed in [366], [299]. Lecture notes dealing with Wilson loops
in the context of the AdS/CFT correspondence are e.g.[430].
Settings where the string scale is the TeV scale are reviewed in[32], [29].
in[43]. The discussion of the stringy corrections to the Wilson loop follows[180, 181].
A similar approach (in the conformal gauge) and more examples are discussed in[149].
This paper also addresses the problem of the divergence and gives a numerical estimate
of the correction. String fluctuations as a source for corrections to the Wilson loop are
also discussed in[226], [351], [466], [294], [262], [322]. Corrections to the field theory
calculation are derived in[163], [467], [164],[367]. An attempt to apply the techniques
for computing corrections to the Wilson loop on the M5 brane case is reported in[177].
That branes allow constructions with the string scale at a TeV has been pointed
out in[30]. (Relating the hierarchy problem to the size of extra dimensions has been
proposed before in a field theory context[35].) The argument that in compactifications
of the perturbative heterotic string the size of the compact space is of the order of the
Planck size is given in[281]. Our discussion of corrections to Newton’s law follows[287].
6.4 Chapter 5
Since there are no books on the subject of brane world setups I start directly with a
list of review articles.
Section 5.2.1 is closely related to[128]. The consistency condition that the effec-
tive cosmological constant should be compatible with the metric on the brane is also
mentioned in[128]. The derivation and form of the consistency condition in section
5.2.2 is presented in[187]. The alternative method of integrating a total derivative is
developed in[161]. The connection between the two conditions has been pointed out
in[187]. The complete set of consistency conditions (as it appears in the text) is given
in[199]. (Different consistency conditions are discussed in[279].)
That the cosmological constant problem is rephrased within a brane world setup
is discussed in[395]. The example of section 5.2.3.1 (and a closely related example)
appear in[268], [36]. That the effective cosmological constant does not vanish in this
models is observed (simultaneously) in [465], [186]. To reach consistency by adding
branes and consequently fine tuning input parameters is proposed in [186]. (Problems
with singularities in warped compactifications are considered e.g. also in[231], [291].)
The proof of the no go theorem is taken from[110].
There are too many papers on warped brane world scenarios to be listed. Therefore,
the following list is restricted to papers dealing with the cosmological constant problem
(and most likely this list is also incomplete): [101], [123], [280], [124], [247], [227], [45],
[308, 307], [105], [67], [321], [271], [278], [104], [55], [85], [100], [103], [134], [86], [240].
Papers containing proposals on avoiding the fine tuning problem of the cosmological
constant by going beyond the prepositions of the no go theorem (section 5.2.3.2)
are [288], [292, 293], [109, 229]. A different proposal for addressing the cosmological
constant problem in brane world scenarios is put forward in[453], [401, 402, 403].
Warped compactifications in the context of string theory are e.g. discussed in[330,
329], [53], [150], [69], [84].
Observational bounds on extra dimension scenarios are e.g. presented in[323], [342],
[341].
Acknowledgments
The topics of these notes represent the experience with string theory that the author
gained over the past years. During this period I collaborated with and received in-
valuable help from various people. First, I thank Hans-Jörg Otto and Harald Dorn
who patiently taught me the basics of string theory, the beta function approach to
the effective field theory, and non critical strings. I enjoyed very much collaborating
with Klaus Behrndt on[50, 51, 52]. These collaborations have been very efficient even
at times when we communicated only via email. I would like to thank the Jerusalem
group: Shmuel Elitzur, Amit Giveon and Eliezer Rabinovici for many insights into the
subjects of T-duality, marginal deformations of conformal field theories, and strongly
6. Research papers 204
[2] O. Aharony, S. S. Gubser, J. Maldacena, H. Ooguri, and Y. Oz. Large N field the-
ories, string theory and gravity. Phys. Rept., 323:183–386, 2000. hep-th/9905111.
[4] K. Akama. An early proposal of ’brane world’. Lect. Notes Phys., 176:267–271,
1982. hep-th/0001113.
205
BIBLIOGRAPHY 206
[15] E. Alvarez, J. L. F. Barbón, and J. Borlaf. T-duality for open strings. Nucl.
Phys., B479:218–242, 1996. hep-th/9603089.
[17] O. Alvarez and C. Liu. Target space duality between simple compact Lie groups
and Lie algebras under the Hamiltonian formalism: 1. Remnants of duality at
the classical level. Commun. Math. Phys., 179:185–214, 1996. hep-th/9503226.
[29] I. Antoniadis. String and D-brane physics at low energy. 2001. hep-th/0102202.
[32] I. Antoniadis and A. Sagnotti. Mass scales, supersymmetry breaking and open
strings. Class. Quant. Grav., 17:939–950, 2000. hep-th/9911205.
[35] N. Arkani-Hamed, S. Dimopoulos, and G. Dvali. The hierarchy problem and new
dimensions at a millimeter. Phys. Lett., B429:263–272, 1998. hep-ph/9803315.
[42] J. A. Bagger, Jr. Callan, C. G., and J. A. Harvey. Cosmic strings as orbifolds.
Nucl. Phys., B278:550, 1986.
[44] T. Banks and E. Martinec. The renormalization group and string field theory.
Nucl. Phys., B294:733, 1987.
[46] K. Behrndt. Classical and quantum aspects of 4 dimensional black holes. Nucl.
Phys. Proc. Suppl., 56B:11–22, 1997. hep-th/9701053.
[47] K. Behrndt and M. Cvetič. Anti-de Sitter vacua of gauged supergravities with
8 supercharges. Phys. Rev., D61:101901, 2000. hep-th/0001159.
[48] K. Behrndt and H. Dorn. Boundary effects relevant for the string interpretation
of sigma models. Phys. Lett., B266:59–64, 1991.
[49] K. Behrndt and H. Dorn. Weyl anomaly and Curci-Paffuti theorem for sigma
models on manifolds with boundary. Int. J. Mod. Phys., A7:1375–1390, 1992.
[53] K. Behrndt and S. Gukov. Domain walls and superpotentials from M theory on
Calabi-Yau three-folds. Nucl. Phys., B580:225–242, 2000. hep-th/0001082.
[57] E. Bergshoeff, C. Hull, and T. Ortin. Duality in the type II superstring effective
action. Nucl. Phys., B451:547–578, 1995. hep-th/9504081.
[61] M. Bershadsky, C. Vafa, and V. Sadov. D-Branes and Topological Field Theories.
Nucl. Phys., B463:420–434, 1996. hep-th/9511222.
[63] M. Bianchi and A. Sagnotti. On the systematics of open string theories. Phys.
Lett., B247:517–524, 1990.
[64] M. Bianchi and A. Sagnotti. Twist symmetry and open string Wilson lines.
Nucl. Phys., B361:519–538, 1991.
[65] A. Bilal. Duality in N=2 SUSY SU(2) Yang-Mills Theory: A pedagogical intro-
duction to the work of seiberg and witten. 1995. hep-th/9601007.
[68] J. D. Blum and A. Zaffaroni. An orientifold from F theory. Phys. Lett., B387:71–
74, 1996. hep-th/9607019.
BIBLIOGRAPHY 210
[69] R. Blumenhagen and A. Font. Dilaton tadpoles, warped geometries and large
extra dimensions for non-supersymmetric strings. Nucl. Phys., B599:241–254,
2001. hep-th/0011269.
[73] R. Blumenhagen, B. Körs, and D. Lüst. Type I strings with F- and B-flux.
JHEP, 02:030, 2001. hep-th/0012156.
[74] R. Blumenhagen, B. Körs, D. Lüst, and T. Ott. The standard model from stable
intersecting brane world orbifolds. 2001. hep-th/0107138.
[75] J. Bogaerts, A. Sevrin, J. Troost, W. Troost, and S. van der Loo. D-branes and
constant electromagnetic backgrounds. 2000. hep-th/0101018.
[77] J. Borlaf and Y. Lozano. Aspects of T-duality in open strings. Nucl. Phys.,
B480:239–264, 1996. hep-th/9607051.
[85] P. Brax and A. C. Davis. Cosmological evolution on self-tuned branes and the
cosmological constant. JHEP, 05:007, 2001. hep-th/0104023.
[86] P. Brax and A. C. Davis. On brane cosmology and naked singularities. Phys.
Lett., B513:156–162, 2001. hep-th/0105269.
[87] L. Brink and M. Henneaux. Principles of string theory. NEW YORK, USA:
PLENUM (1988) 297p.
[88] J. H. Brodie. Duality in supersymmetric SU (Nc ) gauge theory with two adjoint
chiral superfields. Nucl. Phys., B478:123–140, 1996. hep-th/9605232.
[105] H. Collins and B. Holdom. The cosmological constant and warped extra dimen-
sions. Phys. Rev., D63:084020, 2001. hep-th/0009127.
[107] P. Creminelli, A. Nicolis, and R. Rattazzi. Holography and the electroweak phase
transition. 2001. arXiv:hep-th/0107141.
[109] C. Csaki, J. Erlich, and C. Grojean. Gravitational Lorentz violations and adjust-
ment of the cosmological constant in asymmetrically warped spacetimes. Nucl.
Phys., B604:312–342, 2001. hep-th/0012143.
BIBLIOGRAPHY 213
[111] G. Curci and G. Paffuti. Consistency between the string background field
equation of motion and the vanishing of the conformal anomaly. Nucl. Phys.,
B286:399, 1987.
[113] T. Curtright and C. Zachos. Currents, charges, and canonical structure of pseu-
dodual chiral models. Phys. Rev., D49:5408–5421, 1994. hep-th/9401006.
[115] M. Cvetič, M. Plumacher, and J. Wang. Three family type IIB orientifold string
vacua with non-Abelian wilson lines. JHEP, 04:004, 2000. hep-th/9911021.
[122] J. Dai, R. G. Leigh, and J. Polchinski. New connections between string theories.
Mod. Phys. Lett., A4:2073–2083, 1989.
[123] S. P. de Alwis. Brane world scenarios and the cosmological constant. Nucl.
Phys., B597:263–278, 2001. hep-th/0002174.
[125] J. de Boer, K. Hori, H. Ooguri, Y. Oz, and Z. Yin. Mirror symmetry in three-
dimensional gauge theories, SL(2, Z) and D-brane moduli spaces. Nucl. Phys.,
B493:148–176, 1997. hep-th/9612131.
[128] O. DeWolfe, D. Z. Freedman, S. S. Gubser, and A. Karch. Modeling the fifth di-
mension with scalars and gravity. Phys. Rev., D62:046008, 2000. hep-th/9909134.
[131] P. Di Vecchia and A. Liccardo. D-branes in string theory. II. 1999. hep-
th/9912275.
[132] R. Dick. Which action for brane worlds? Phys. Lett., B491:333–338, 2000.
hep-th/0007063.
[135] K. R. Dienes. String Theory and the Path to Unification: A Review of Recent
Developments. Phys. Rept., 287:447–525, 1997. hep-th/9602045.
[136] M. Dine, P. Huet, and N. Seiberg. Large and small radius in string theory. Nucl.
Phys., B322:301, 1989.
[140] H. Dorn. Nonabelian gauge field dynamics on matrix D-branes. Nucl. Phys.,
B494:105–118, 1997. hep-th/9612120.
[141] H. Dorn. The mass term in non-Abelian gauge field dynamics on matrix D-
branes and T-duality in the sigma-model approach. JHEP, 04:013, 1998. hep-
th/9804065.
[142] H. Dorn and H. J. Otto. Open bosonic strings in general background fields. Z.
Phys., C32:599, 1986.
[143] H. Dorn and H. J. Otto. On T-duality for open strings in general abelian and non-
abelian gauge field backgrounds. Phys. Lett., B381:81–88, 1996. hep-th/9603186.
[144] H. Dorn and H. J. Otto. On Wilson loops and Q anti-Q potentials from the
AdS/CFT relation at T ≥ 0. 1998. hep-th/9812109.
[147] M. R. Douglas and C. Hull. D-branes and the noncommutative torus. JHEP,
02:008, 1998. hep-th/9711165.
[148] N. Drukker, D. J. Gross, and H. Ooguri. Wilson loops and minimal surfaces.
Phys. Rev., D60:125006, 1999. hep-th/9904191.
[150] E. Dudas and J. Mourad. Brane solutions in strings with broken supersymmetry
and dilaton tadpoles. Phys. Lett., B486:172–178, 2000. hep-th/0004165.
[151] M. J. Duff. TASI lectures on branes, black holes and anti-de Sitter space. 1999.
hep-th/9912164.
[152] M. J. Duff, Ramzi R. Khuri, and J. X. Lu. String solitons. Phys. Rept., 259:213–
326, 1995. hep-th/9412184.
[153] M. J. Duff and James T. Liu. Complementarity of the Maldacena and Randall-
Sundrum pictures. Phys. Rev. Lett., 85:2052–2055, 2000. hep-th/0003237.
[156] S. Elitzur, A. Forge, A. Giveon, and E. Rabinovici. More results in N=1 super-
symmetric gauge theories. Phys. Lett., B353:79–83, 1995. hep-th/9504080.
[157] S. Elitzur, A. Forge, A. Giveon, and E. Rabinovici. Effective Potentials and Vac-
uum Structure in N=1 Supersymmetric Gauge Theories. Nucl. Phys., B459:160–
206, 1996. hep-th/9509130.
[162] U. Ellwanger and J. Fuchs. String field equations and the dual S matrix from
the exact renormalization group. Nucl. Phys., B312:95, 1989.
[170] W. Fischler and L. Susskind. Dilaton tadpoles, string condensates and scale
invariance. Phys. Lett., B171:383, 1986.
[171] W. Fischler and L. Susskind. Dilaton tadpoles, string condensates and scale
invariance. 2. Phys. Lett., B173:262, 1986.
[172] A. Font, L. E. Ibáñez, D. Lüst, and F. Quevedo. Strong - weak coupling duality
and nonperturbative effects in string theory. Phys. Lett., B249:35–43, 1990.
[178] S. Förste and D. Ghoshal. Strings from orientifolds. Nucl. Phys., B527:95–120,
1998. hep-th/9711039.
[180] S. Förste, D. Ghoshal, and S. Theisen. Stringy corrections to the Wilson loop
in N = 4 super Yang-Mills theory. JHEP, 08:013, 1999. hep-th/9903042.
[181] S. Förste, D. Ghoshal, and S. Theisen. Wilson loop via AdS/CFT duality. 1999.
hep-th/0003068.
BIBLIOGRAPHY 218
[185] S. Förste, A. A. Kehagias, and S. Schwager. Non Abelian T-duality for open
strings. Nucl. Phys. Proc. Suppl., 56B:36–41, 1997. hep-th/9610062.
[188] S. Förste and J. Louis. Duality in string theory. Nucl. Phys. Proc. Suppl.,
61A:3–22, 1998. hep-th/9612192.
[190] E. S. Fradkin and A. A. Tseytlin. Quantum string theory effective action. Nucl.
Phys., B261:1–27, 1985.
[194] J. Fuchs and C. Schweigert. Branes: From free fields to general backgrounds.
Nucl. Phys., B530:99–136, 1998. hep-th/9712257.
[197] T. Gherghetta and Y. Oz. Supergravity, non-conformal field theories and brane-
worlds. 2001. arXiv:hep-th/0106255.
[198] T. Gherghetta and A. Pomarol. Bulk fields and supersymmetry in a slice of AdS.
Nucl. Phys., B586:141–162, 2000. arXiv:hep-ph/0003129.
[199] G. Gibbons, R. Kallosh, and A. Linde. Brane world sum rules. JHEP, 01:022,
2001. hep-th/0011225.
[201] S. B. Giddings and E. Katz. Effective theories and black hole production in
warped compactifications. 2000. hep-th/0009176.
[202] P. Gilkey. Invaraince theory, the heat equation and the Atiyah-Singer index
theorem. CRC Press, second edition, 1995.
[205] P. Ginsparg. Applied conformal field theory. Lectures given at Les Houches
Summer School in Theoretical Physics, Les Houches, France, Jun 28 - Aug 5,
1988.
[207] A. Giveon. Target space duality and stringy black holes. Mod. Phys. Lett.,
A6:2843–2854, 1991.
[209] A. Giveon and E. Kiritsis. Axial vector duality as a gauge symmetry and topology
change in string theory. Nucl. Phys., B411:487–508, 1994. hep-th/9303016.
[210] A. Giveon and D. Kutasov. Brane dynamics and gauge theory. Rev. Mod. Phys.,
71:983–1084, 1999. hep-th/9802067.
BIBLIOGRAPHY 220
[211] A. Giveon, M. Porrati, and E. Rabinovici. Target space duality in string theory.
Phys. Rept., 244:77–202, 1994. hep-th/9401139.
[215] W. D. Goldberger and M. B. Wise. Modulus stabilization with bulk fields. Phys.
Rev. Lett., 83:4922–4925, 1999. hep-ph/9907447.
[216] J. Govaerts. Quantum consistency of open string theories. Phys. Lett., B220:77,
1989.
[217] M. B. Green. Space-time duality and Dirichlet string theory. Phys. Lett.,
B266:325–336, 1991.
[218] M. B. Green, J. A. Harvey, and G. Moore. I-brane inflow and anomalous cou-
plings on D-branes. Class. Quant. Grav., 14:47–52, 1997. hep-th/9605033.
[225] B. R. Greene, A. Shapere, C. Vafa, and S. Yau. Stringy cosmic strings and
noncompact Calabi-Yau manifolds. Nucl. Phys., B337:1, 1990.
BIBLIOGRAPHY 221
[226] J. Greensite and P. Olesen. Worldsheet fluctuations and the heavy quark poten-
tial in the AdS/CFT approach. JHEP, 04:001, 1999. hep-th/9901057.
[228] M. T. Grisaru, D. I. Kazakov, and D. Zanon. Five loop divergences for the N=2
supersymmetric nonlinear sigma model. Nucl. Phys., B287:189, 1987.
[230] D. J. Gross and H. Ooguri. Aspects of large N gauge theory dynamics as seen
by string theory. Phys. Rev., D58:106002, 1998. hep-th/9805129.
[231] S. S. Gubser. Curvature singularities: The good, the bad, and the naked. 2000.
hep-th/0002160.
[232] S. S. Gubser. AdS/CFT and gravity. Phys. Rev., D63:084017, 2001. hep-
th/9912001.
[235] M. Haack, B. Körs, and D. Lüst. Recent developments in string theory: From
perturbative dualities to M-theory. 1998. hep-th/9904033.
[236] M. Haack, J. Louis, and H. Singh. Massive type IIA theory on K3. JHEP,
04:040, 2001. hep-th/0102110.
[237] S. Hamidi and C. Vafa. Interactions on orbifolds. Nucl. Phys., B279:465, 1987.
[238] A. Hanany and E. Witten. Type IIB superstrings, BPS monopoles, and three-
dimensional gauge dynamics. Nucl. Phys., B492:152–190, 1997. hep-th/9611230.
[241] M. Henningson and K. Skenderis. The holographic Weyl anomaly. JHEP, 07:023,
1998. hep-th/9806087.
[242] S. Hewson and M. Perry. Exact non-abelian duality. Phys. Lett., B391:316–323,
1997. hep-th/9603015.
[243] P. Hořava. Background duality of open string models. Phys. Lett., B231:251,
1989.
[244] P. Hořava. Strings on world sheet orbifolds. Nucl. Phys., B327:461, 1989.
[246] P. Hořava and E. Witten. Heterotic and type I string dynamics from eleven
dimensions. Nucl. Phys., B460:506–524, 1996. hep-th/9510209.
[248] G. T. Horowitz and A. Strominger. Black strings and p-branes. Nucl. Phys.,
B360:197–209, 1991.
[254] L. E. Ibáñez, F. Marchesano, and R. Rabadan. Getting just the standard model
at intersecting branes. 2001. hep-th/0105155.
[256] L. E. Ibáñez, H. P. Nilles, and F. Quevedo. Orbifolds and Wilson lines. Phys.
Lett., B187:25–32, 1987.
[257] L. E. Ibáñez, H. P. Nilles, and F. Quevedo. Reducing the rank of the gauge group
in oribifold compactifications of the heterotic string. Phys. Lett., B192:332, 1987.
[259] K. Intriligator and P. Pouliot. Exact superpotentials, quantum vacua and duality
in supersymmetric SP (Nc ) gauge theories. Phys. Lett., B353:471–476, 1995. hep-
th/9505006.
[262] R. A. Janik. String fluctuations, AdS/CFT and the soft pomeron intercept.
Phys. Lett., B500:118–124, 2001. hep-th/0010069.
[270] M. Kaku. Introduction to superstrings and M-theory. New York, USA: Springer
(1999) 587 p.
[274] R. Kallosh and A. Linde. Supersymmetry and the brane world. JHEP, 02:005,
2000. hep-th/0001071.
[275] R. Kallosh and J. Rahmfeld. The GS string action on AdS5 × S 5 . Phys. Lett.,
B443:143–146, 1998. hep-th/9808038.
[278] S. Kalyana Rama. Brane world scenario with m-form field: Stabilisation of
radion modulus and self tuning solutions. Phys. Lett., B495:176–182, 2000. hep-
th/0010121.
[280] P. Kanti, K. Olive, and M. Pospelov. Static solutions for brane models with a
bulk scalar field. Phys. Lett., B481:386–396, 2000. hep-ph/0002229.
[281] V. S. Kaplunovsky. Mass scales of the string unification. Phys. Rev. Lett.,
55:1036, 1985.
[282] A. Karch. Field theory dynamics from branes in string theory. 1998. hep-
th/9812072.
[284] A. Karch and L. Randall. Locally localized gravity. JHEP, 05:008, 2001.
arXiv:hep-th/0011156.
BIBLIOGRAPHY 225
[285] A. Karch and L. Randall. Open and closed string interpretation of SUSY CFT’s
on branes with boundaries. JHEP, 06:063, 2001. arXiv:hep-th/0105132.
[287] A. Kehagias and K. Sfetsos. Deviations from the 1/r 2 newton law due to extra
dimensions. Phys. Lett., B472:39–44, 2000. hep-ph/9905417.
[289] A. A. Kehagias. Type IIA/IIB string duality for targets with abelian isometries.
Phys. Lett., B377:241–244, 1996. hep-th/9602059.
[291] H. D. Kim. A criterion for admissible singularities in brane world. Phys. Rev.,
D63:124001, 2001. hep-th/0012091.
[292] J. E. Kim, B. Kyae, and H. M. Lee. A model for self-tuning the cosmological
constant. Phys. Rev. Lett., 86:4223–4226, 2001. hep-th/0011118.
[296] E. Kiritsis, C. Kounnas, and D. Lüst. A large class of new gravitational and
axionic backgrounds for four-dimensional superstrings. Int. J. Mod. Phys.,
A9:1361–1394, 1994. hep-th/9308124.
[297] E. Kiritsis and N. A. Obers. A New duality symmetry in string theory. Phys.
Lett., B334:67–71, 1994. hep-th/9406082.
[298] E. B. Kiritsis. Duality in gauged WZW models. Mod. Phys. Lett., A6:2871–2880,
1991.
[301] M. Klein and R. Rabadan. Orientifolds with discrete torsion. JHEP, 07:040,
2000. hep-th/0002103.
[305] C. Klimčik and P. Severa. Dual nonAbelian duality and the Drinfeld double.
Phys. Lett., B351:455–462, 1995. hep-th/9502122.
[306] C. Klimčik and P. Severa. Poisson-Lie T-duality: Open Strings and D-branes.
Phys. Lett., B376:82–89, 1996. hep-th/9512124.
[308] A. Krause. A small cosmological constant, grand unification and warped geom-
etry. 2000. hep-th/0006226.
[310] D. Kutasov, A. Schwimmer, and N. Seiberg. Chiral rings, singularity theory and
electric-magnetic duality. Nucl. Phys., B459:455–496, 1996. hep-th/9510222.
[312] J. Lauer, J. Mas, and H. P. Nilles. Duality and the role of nonperturbative effects
on the world sheet. Phys. Lett., B226:251, 1989.
[314] J. Lee. Zero norm states and enlarged gauge symmetries of closed bosonic string
in background massive fields. Nucl. Phys., B336:222–244, 1990.
[315] R. G. Leigh. Dirac-Born-Infeld action from Dirichlet sigma model. Mod. Phys.
Lett., A4:2767, 1989.
[317] W. Lerche. Introduction to Seiberg-Witten theory and its stringy origin. Nucl.
Phys. Proc. Suppl., 55B:83–117, 1997. hep-th/9611190.
[319] A. Lerda and R. Russo. Stable non-BPS states in string theory: A pedagogical
review. Int. J. Mod. Phys., A15:771–820, 2000. hep-th/9905006.
[320] M. Li. Dirichlet Boundary State in Linear Dilaton Background. Phys. Rev.,
D54:1644–1646, 1996. hep-th/9512042.
[321] T. Li. Time-like extra dimension and cosmological constant in brane models.
Phys. Lett., B503:163–172, 2001. hep-th/0009132.
[328] H. Luckock. Quantum geometry of strings with boundaries. Ann. Phys., 194:113,
1989.
BIBLIOGRAPHY 228
[331] D. Lüst and S. Theisen. Lectures on string theory. Berlin, Germany: Springer
(1989) 346 p. (Lecture notes in physics, 346).
[332] J. Lykken and L. Randall. The shape of gravity. JHEP, 06:014, 2000. hep-
th/9908076.
[335] J. Maldacena. The large N limit of superconformal field theories and supergrav-
ity. Adv. Theor. Math. Phys., 2:231–252, 1998. hep-th/9711200.
[336] J. Maldacena. Wilson loops in large N field theories. Phys. Rev. Lett., 80:4859–
4862, 1998. hep-th/9803002.
[338] J. M. Maldacena. Black holes and D-branes. Nucl. Phys. Proc. Suppl., 61A:111–
123, 1998. hep-th/9705078.
[344] T. Mohaupt. Black holes in supergravity and string theory. Class. Quant. Grav.,
17:3429–3482, 2000. hep-th/0004098.
[345] T. Mohaupt. Black hole entropy, special geometry and strings. Fortsch. Phys.,
49:3–161, 2001. hep-th/0007195.
[346] C. Montonen and D. Olive. Magnetic monopoles as gauge particles? Phys. Lett.,
B72:117, 1977.
[347] W. Mück. Ideas on the semi-classical path integral over embedded manifolds.
Fortsch. Phys., 49:607–615, 2001.
[351] S. Naik. Improved heavy quark potential at finite temperature from anti-de
Sitter supergravity. Phys. Lett., B464:73–76, 1999. hep-th/9904147.
[352] K. S. Narain. New heterotic string theories in uncompactified dimensions < 10.
Phys. Lett., B169:41, 1986.
[356] N. A. Obers and B. Pioline. U-duality and M-theory. Phys. Rept., 318:113–225,
1999. arXiv:hep-th/9809039.
[357] D. O’Driscoll. General abelian orientifold models and one loop amplitudes. 1998.
hep-th/9801114.
[358] D. I. Olive. Exact electromagnetic duality. Nucl. Phys. Proc. Suppl., 45A:88–102,
1996. hep-th/9508089.
BIBLIOGRAPHY 230
[359] H. Ooguri and Z. Yin. TASI lectures on perturbative string theories. 1996.
hep-th/9612254.
[360] H. Osborn. Topological charges for N=4 supersymmetric gauge theories and
monopoles of spin 1. Phys. Lett., B83:321, 1979.
[361] J. Pawelczyk and S. Theisen. AdS5 × S 5 black hole metric at O α3 . JHEP,
09:010, 1998. hep-th/9808126.
[362] A. W. Peet. The Bekenstein formula and string theory (N-brane theory). Class.
Quant. Grav., 15:3291–3338, 1998. hep-th/9712253.
[364] I. Pesando. A kappa gauge fixed type IIB superstring action on AdS5 × S 5 .
JHEP, 11:002, 1998. hep-th/9808020.
[371] J. Polchinski. String Theory, volume 1 & 2. Cambridge University Press, 1998.
[372] J. Polchinski and Y. Cai. Consistency of open superstring theories. Nucl. Phys.,
B296:91, 1988.
[374] J. Polchinski and E. Witten. Evidence for Heterotic - Type I String Duality.
Nucl. Phys., B460:525–540, 1996. hep-th/9510169.
BIBLIOGRAPHY 231
[377] G. Pradisi. Type I vacua from diagonal Z3 -orbifolds. Nucl. Phys., B575:134–150,
2000. hep-th/9912218.
[378] G. Pradisi and A. Sagnotti. Open string orbifolds. Phys. Lett., B216:59, 1989.
[379] M. K. Prasad and C. M. Sommerfield. An exact classical solution for the ’t Hooft
monopole and the Julia-Zee dyon. Phys. Rev. Lett., 35:760–762, 1975.
[382] R. Rabadan. Branes at angles, torons, stability and supersymmetry. 2001. hep-
th/0107036.
[383] R. Rabadan and A. M. Uranga. Type IIB orientifolds without untwisted tad-
poles, and non-BPS D-branes. JHEP, 01:029, 2001. hep-th/0009135.
[386] L. J. Randall and R. Sundrum. A large mass hierarchy from a small extra
dimension. Phys. Rev. Lett., 83:3370–3373, 1999. hep-ph/9905221.
[389] S. Rey. The confining phase of superstrings and axionic strings. Phys. Rev.,
D43:526–538, 1991.
BIBLIOGRAPHY 232
[391] S. Rey and J. Yee. Macroscopic strings as heavy quarks in large N gauge theory
and anti-de Sitter supergravity. 1998. hep-th/9803001.
[392] M. Roček and E. Verlinde. Duality, quotients, and currents. Nucl. Phys.,
B373:630–646, 1992. hep-th/9110053.
[396] A. Sagnotti. Open strings and their symmetry groups. Talk presented at
the Cargese Summer Institute on Non- Perturbative Methods in Field Theory,
Cargese, France, Jul 16-30, 1987.
[398] N. Sakai and I. Senda. Vacuum energies of string compactified on torus. Prog.
Theor. Phys., 75:692, 1986.
[401] C. Schmidhuber. AdS5 and the 4d cosmological constant. Nucl. Phys., B580:140–
146, 2000. hep-th/9912156.
[409] J. H. Schwarz and A. Sen. Duality symmetries of 4-D heterotic strings. Phys.
Lett., B312:105–114, 1993. hep-th/9305185.
[410] J. H. Schwarz and A. Sen. Duality symmetric actions. Nucl. Phys., B411:35–63,
1994. hep-th/9304154.
[412] N. Seiberg. Exact results on the space of vacua of four-dimensional SUSY gauge
theories. Phys. Rev., D49:6857–6863, 1994. hep-th/9402044.
[414] N. Seiberg. The power of duality: Exact results in 4-d SUSY field theory. 1995.
hep-th/9506077.
[415] N. Seiberg and E. Witten. Spin structures in string theory. Nucl. Phys.,
B276:272, 1986.
[417] N. Seiberg and E. Witten. Monopoles, duality and chiral symmetry breaking in
N=2 supersymmetric QCD. Nucl. Phys., B431:484–550, 1994. hep-th/9408099.
[418] N. Seiberg and E. Witten. String theory and noncommutative geometry. JHEP,
09:032, 1999. hep-th/9908142.
BIBLIOGRAPHY 234
[419] A. Sen. Equations of motion for the heterotic string theory from the conformal
invariance of the sigma model. Phys. Rev. Lett., 55:1846, 1985.
[420] A. Sen. Electric magnetic duality in string theory. Nucl. Phys., B404:109–126,
1993. hep-th/9207053.
[421] A. Sen. Dyon - monopole bound states, selfdual harmonic forms on the multi -
monopole moduli space, and SL(2, Z) invariance in string theory. Phys. Lett.,
B329:217–221, 1994. hep-th/9402032.
[422] A. Sen. Strong - weak coupling duality in four-dimensional string theory. Int.
J. Mod. Phys., A9:3707–3750, 1994. hep-th/9402002.
[424] A. Sen. Non-BPS D-branes in string theory. Class. Quant. Grav., 17:1251–1256,
2000.
[425] A. Shapere and F. Wilczek. Selfdual models with theta terms. Nucl. Phys.,
B320:669, 1989.
[428] K. Skenderis. Black holes and branes in string theory. Lect. Notes Phys., 541:325–
364, 2000. hep-th/9901050.
[430] J. Sonnenschein. What does the string / gauge correspondence teach us about
Wilson loops? 1999. hep-th/0003032.
[434] L. Susskind and E. Witten. The holographic bound in anti-de Sitter space. 1998.
hep-th/9805114.
[435] G. ’t Hooft. A planar diagram theory for strong interactions. Nucl. Phys.,
B72:461, 1974.
[436] R. Tatar. Dualities in 4D theories with product gauge groups from brane con-
figurations. Phys. Lett., B419:99–106, 1998. hep-th/9704198.
[437] IV Taylor, W. Lectures on D-branes, gauge theory and M(atrices). 1997. hep-
th/9801182.
[445] A. A. Tseytlin. Sigma models and renormalization of string loops. Lectures given
at 1989 Trieste Spring School on Superstrings, Trieste, Italy, Apr 3-14, 1989.
[446] A. A. Tseytlin. Sigma model Weyl invariance conditions and string equations of
motion. Nucl. Phys., B294:383, 1987.
[448] A. A. Tseytlin. Duality and dilaton. Mod. Phys. Lett., A6:1721–1732, 1991.
[451] C. Vafa. Modular invariance and discrete torsion on orbifolds. Nucl. Phys.,
B273:592, 1986.
[453] E. Verlinde and H. Verlinde. RG-flow, gravity and the cosmological constant.
JHEP, 05:034, 2000. hep-th/9912018.
[455] S. Weinberg. The cosmological constant problem. Rev. Mod. Phys., 61:1–23,
1989.
[457] E. Witten. Dyons of charge e theta / 2 pi. Phys. Lett., B86:283–287, 1979.
[458] E. Witten. String theory dynamics in various dimensions. Nucl. Phys., B443:85–
126, 1995. hep-th/9503124.
[459] E. Witten. Solutions of four-dimensional field theories via M-theory. Nucl. Phys.,
B500:3–42, 1997. hep-th/9703166.
[460] E. Witten. Anti-de Sitter space and holography. Adv. Theor. Math. Phys.,
2:253–291, 1998. hep-th/9802150.
[461] E. Witten. Anti-de Sitter space, thermal phase transition, and confinement in
gauge theories. Adv. Theor. Math. Phys., 2:505–532, 1998. hep-th/9803131.
[462] E. Witten. The cosmological constant from the viewpoint of string theory. 2000.
hep-ph/0002297.
[464] D. Youm. Black holes and solitons in string theory. Phys. Rept., 316:1–232,
1999. hep-th/9710046.
[465] D. Youm. Bulk fields in dilatonic and self-tuning flat domain walls. Nucl. Phys.,
B589:315–336, 2000. hep-th/0002147.
[467] K. Zarembo. String breaking from ladder diagrams in SYM theory. JHEP,
03:042, 2001. hep-th/0103058.