DS/CFT Duality in Cosmology: Ramtin Amintaheri

dS/CFT Duality
in
Cosmology
Ramtin Amintaheri
Supervisor: Dr Toby Wiseman
Submitted in partial fulfillment of the requirements for the degree of Master of Science in Quantum Field
and Fundamental Forces of Imperial College London
September 2015
Contents
1 Introduction 1
1.1 motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2 Conformal Field Theory 4

2.1 Conformal Transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 Conformal Group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2.1 Generators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2.2 Special Conformal Transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2.3 Conformal Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2.4 Symmetries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3 Conformal Fields in Two Dimension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3.1 Complexification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3.2 The Witt Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.3.3 Global Conformal Transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.4 Conformal Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.5 Correlation Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.5.1 Two-Point Correlation Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3 Anti de Sitter Space 11

3.1 Constant Curvature Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.1.1 Einstein Static Universe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.2 Geometry of Anti de Sitter Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.3 Coordinate Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.3.1 Global Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.3.2 Poincare Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.4 Matter Field in AdS Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4 AdS-CFT Duality 17
4.1 Symmetries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
4.2 The Field - Operator Correspondence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
4.3 Green Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.3.1 Bulk-to-Bulk Green Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.3.2 Bulk-to-Boundary Green Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.4.1 Two- Point Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.5 Mass-Dimension Duality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.6 Radial Direction-Energy Duality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.7 Holographic Renormalisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.7.1 Domain Wall . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.8 Example of the Duality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.8.1 Supergravity and P-Branes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.8.2 String Theory and D-branes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
i
4.8.3 Maldacena’s Duality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.8.4 Limits of Validity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
5 De Sitter Space 29
5.1 Geometry of de Sitter Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
5.2 Coordinate Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
5.2.1 Global Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
5.2.2 Planar Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
5.2.3 Hyperbolic Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
5.3 Geodesics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
5.4 Matter Field in de Sitter Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
6 dS-CFT Duality 34
6.1 Symmetries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
6.1.1 Asymptotic Symmetries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
6.2 The Field - Operator Correspondence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
6.3 Mass-Dimension Duality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
6.4 Time Evolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
6.5 Green Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
6.7 Example of dS-CFT Duality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
6.7.1 Higher-Spin Gravity Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
6.7.2 EAdS - dS Correspondence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
6.7.3 GKPY Duality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
6.7.4 O(N) - Sp(N) Correspondence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
6.7.5 dS4 / Sp(N) CFT3 Duality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
7 Cosmology 45
7.1 Inflationary Epoch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
7.1.1 Metric Perturbation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
7.1.2 gauge transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
7.1.3 Statistics of Perturbation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
7.1.4 Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
7.1.5 Cosmological Power Spectra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
7.2 RG-Flow Universe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
7.3 Holography for Cosmology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
7.3.1 Domain-Wall / Cosmology Duality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
7.3.2 Hamiltonian Formalism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
7.3.3 CFT Two-Point Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
7.3.4 Weak versus Strong Gravitational Coupling . . . . . . . . . . . . . . . . . . . . . . . . 53
7.4 Holographic Phenomenology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
7.4.1 Prototype CFT Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
7.4.2 Energy-Momentum Tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
7.4.3 Cosmological Observables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
8 Conclusion 59
8.1 Futer Works . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
8.2 Acknowledgment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
A Anti de Sitter Equation of Motion a
B Anti-de Sitter Bulk-to-Boundary Propagator c
C Anti de Sitter Two-Point Function e
ii
D De Sitter Green Function g
D.1 Asymptotic behaviour . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . g
E De Sitter Two-Point Function i

E.1 Asymptotic Behaviour . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . j
F Holography for Cosmology k
iii
Abstract
In this dissertation, gauge/gravity correspondence in a de Sitter space, and their application in the field
of cosmology is studied. Basic material necessary to understand the topic including conformal field theory,
and geometrical properties of AdS and dS spaces are reviewed. The well-constructed counterpart theory of
AdS/CF T duality, and the renowned Maldacena’s conjecture are also explained. dS/CF T correspondence
has some special features, time is holographically emergent, and non-unitary conformal field lives in the
asymptotic future of the space time. In dS/CF T theory universe is an RG flow from the far future UV
point to the far past IR fixed point. We show that it is possible to formulate a holographic framework for
the inflationary universe, despite the fact that the direct map between cosmology and dual gauge field is not
understood. A holographic model is proposed that is able to provide predictions that are compatible with
recent observations. It is also argued that predictions of this model in the perturbative regime can describe
strongly-coupled quantum gravity state of the early universe, and offer some solutions for the initial condition
problem.
Chapter 1
Introduction
Gauge/gravity duality plays an important role in our understanding of modern physics as it connects two
supposedly different types of physical theories. On one side is the gauge theory which deals with quantum
fields, and on the other side is gravitational theories like general relativity. The duality states that any
theory of gravity in a given space-time can be completely described by a conformal field theory living on
the boundaries of the space-time. [28] Since field theory is in one dimension lower, it therefore functions as
a hologram encoding the information about higher-dimensional gravity formalism. [40] Any quantity in one
theory has a corresponding counterpart in the other, as if there is a dictionary translating one formalism to
the other. [39]
There are two known classes of dualities: the first one is the well-understood AdS/CF T correspondence
which deals with spaces of negative curvature, and the second is the newer version dS/CF T which is topic
of this report. Differernt aspects of AdS/CF T connection have been studied for a long time, and it is also
found to have realisations in different field of physics in particular quantum chronodynamics and condensed
matter. [54, 53] Nevertheless anti-de Sitter does not support expansion of space-time while all observational
evidence indicates that expanding universe has a positive cosmological constant. In addition at early times,
during inflationary epoc universe was a Sitter space, and in far future it will turn into a de Sitter again.
[17, 18] Therefore there is a need for formulation of the dS/CF T correspondence to apply gauge/gravity
duality to our universe.
Notion of principle of holography stemmed from G. t’Hooft, and later L. Suuskind works on the black
hole thermodynamics in 1993. [28, 29] Later in 1995 Polchinski showed that p-branes are the same objects
as the D-branes [27]: field theory describing dynamical endpoints of open strings on the brane corresponds
to the supergavity solutions in the space-time. Researches conducted following Polchinski’s work finally led
into discovery of the correspondence between gauge theories and gravity.
In 1997 Maldacena published a paper on his conjecture about AdS/CF T correspondence that is regarded
as the birth of gauge/gravity duality. [31] He could use type IIB string theory on a AdS5 × S 5 background
to describe the dual N = 4 super Yang Mills theory living on a four-dimensional D3 brane. His work excited
great interest among string theorists, and later was completed by S. Gubser, I. Klebanov, Polyakov, and
Witten [32, 33].Soon after a complete gauge/gravity dictionary was developed.
However all of the mentioned realisations was centred on space-times of negative cosmological constant,
and left this curious question unanswered that how we can apply the holography to the real universe with
positive cosmological constant. Despite considerable research there was no success in solving string theory in
de Sitter space as a conformal field on the boundary. [56] It was not until Strominger became interested in the
early works of Brown and Henneaux. They used asymptotic symmetries and central charges to demonstrate
relation between AdS3 and CF T2 without any reference to string theory [45]. In 2001 Strominger finally
published his paper and dS/CF T duality was brought to the physics’ world attention. Along the lines of
AdS/CF T correspondence, he could show that quantum gravity on dS3 is related to a Euclidean conformal
field of one lower dimension.
In the same year he wrote another paper about inflation and dS/CF T correspondence showing that
the main inspirations to find de Sitter analog of the duality stem from cosmology [75].His work was made
more precise by other scientists in particular K. Skenderis who could extend the holographic renormalisation
1
formalism to de Sitter space. He even could manage to offer a phenomenological prototype that provides
predictions consistent with latest observational data. [76, 77]
However in absence of string theory there was no example of the correspondence at the fundamental level,
and this problem somehow stymied the progress in the subject. Finally recently in 2011 Strominger could
explain that candidate theory in de Sitter space is higher spin gravity introduced in early 90’s by Vasiliev[62,
63]. By an analytical continuation of Giombi-Klebanov-Polyakov-Yin (GKPY) duality [69, 66, 71, 72] he
could establish that higher spin gravity in dS4 is dual to a fermionic Sp(N )-invariant CF T3 on the Euclidean
boundary space. [61]
1.1 motivation
From simple Newtonian models of mutual attractive forces through to Einstein’s geometrodynamical treat-
ment and more modern notions of quantum gravity and string theory, gravitational theories are becoming
more and more sophisticated. In general current gravity theories are complicated and not fully understood
in all details, many gravitational systems therefore can not be studied analytically. In addition numerical
approximations can not cover the whole physics, and this imposes a limitation on the scope of our knowledge.
Gauge gravity duality opened totally new avenues in fundamental physics as quantum fields has long been
studied, and well understood. It provided new tools to solve problems in different theoretical areas of gravity
through the corresponding field theories. [19]
Apart from this, AdS/CF T duality has significant physical interpretations: One of the fundamental forces
of the nature that is gravity is a phenomenon emerging from the underlying quantum fields. Furthermore
one of our microscopic dimensions is not fundamental and is just emergant. In other words the world can be
thought of as three-dimensional information. But dS/CFT correspondence even brought in more unprece-
dented results, ’time’ is emerging from more fundamental temporally-frozen quantum theories. Universe is a
fluid flowing from the far future infinity back to the far past in the space of renormalisation group theories.
[29, 76]
Inflation theory is undoubtedly one of the cornerstones of the contemporary cosmology. It has provided
the best solutions to the classical cosmological problems about horizon, flatness and monopole issues in the old
big-bang theory, and made correct predictions about formation of the large structures. Current observations
confirm that universe is spatially flat, homogeneous, and isometric with small nearly Gaussian perturbations
which has an almost scale invariant spectrum. Nevertheless it turns out that inflation is not a complete
theory yet as it suffers from some problems with cosmological constant, super-planck scales, and particularly
initial singularity. This theory does not remove the original singularity and can not provide any value for
initial conditions of the inflationary epoc. [80, 81]
Gauge-gravity correspondence, on the other hand, is a strong-weak duality which means when one theory
is strongly interacting the other one is weakly coupled. Accordingly the unknown area of strongly coupled
quantum gravity corresponds to a weakly interacting quantum field theory which is theoretically very well
understood. So holographic model can proceed beyond the standard inflation formalism, and potentially
can provide unrivalled predictions about initial values , and hopefully it will shed light on mysterious pre-
inflationary era of the universe. [39, 20]
Furthermore it is desirable that a theory of inflation does not suffer from ultra-violet divergences. As
a result considerable amount of work in the current research is focused on expressing the inflation in the
language of string theory. The holographic model explained here features a three-dimensional dual quantum
field and is super-renormalisable and hence UV complete. [76, 77]
1.2 Outline
Gauge/gravity correspondence draws upon various areas of theoretical physics. It is assumed that reader
has an introductory knowledge of quantum field theory, general relativity, string theory, super-symmetry,
black holes, and cosmology. Nevertheless, a brief review of conformal field theories, spaces of constant
curvature particularly de Sitter, and Anti-de Sitter, and inflationary cosmology will be provided in the
following chapters.
2
We start with conformal fields which are a generalisation of quantum fields with an additional scale
symmetry. It is aimed to provide a proper description of geometrical and physical properties of CFT by
studying the structure of conformal groups, algebra, and associated symmetries. At the end, effect of con-
formal symmetry on the general form of the correlation functions as the main physical observables will be
investigated.
In the third chapter we start studying study space forms in details because holographic framework applies
to the manifolds that asymptotically have a constant curvature. Geometrical properties of anti-de Sitter
which is a space with negative constant curvature are reviewed. Most commonly used coordinate systems
are introduced, and it is shown how they can describe topology and different symmetries of AdS. General
solutions to the equation of motion that are derived in the last section will be referred to frequently in this
paper.
Having reviewed conformal properties of the quantum fields on one hand, and geometrical and physical
aspects of anti-de Sitter, on the other hand,in chapter 4, we proceed to discuss about AdS/CF T correspon-
dence as the first and best understood realisation of gauge/gravity duality. Correspondence is motivated by
identification of isometry group of AdS with conformal group of the dual field. The one-to-one map between
operators of QFT, and the fields in the bulk is explained. Subsequently green functions of the theory are
derived which enable us to find an explicit expression for the correlation functions. In the last section, by
making a comparison between predictions of the Type IIB supergravity and N = 4, SY M superstring theory
on parallel D3 branes, the renowned Maldacena’s conjecture is explained in details.
In chapter 5 we start studying spaces of positive curvature by introducing geometrical properties of de
Sitter manifold. Common coordinate systems are introduced, and their applications are explained. Special
features of de Sitter space including existence of a horizon, problem of observables, and different asymptotic
boundaries are explained. In the last section dynamics of scalar matter field are studied, and equations of
motion in this space are solved.
Up to this point in the report, all the discussions about the gauge/gravity duality centred around the
spaces of negative curvature which based on observational data do not apply to the real universe. In chapter
6, therefore, we start studying the dS/CF T correspondence, and explore the similarities and differences
compared to the more familiar AdS/CF T counterpart. At first it is explained how asymptotic isometry
group of a de Sitter space is identified with conformal symmetry group on the boundary. This acts as the
main motivation to define the correspondence between conformal theory and de Sitter fields. Next different
entities on both sides of the correspondence are compared and a duality map is constructed. In particular
it is stressed that in dS/CF T time is holographically emergent, and thus notion of unitary time evolution is
no longer valid in this theory. In the last section we investigate the correspondence at a more fundamental
level by introducing higher spin formalism of gravity and defining the dual anticommuting Sp(N )-invariant
conformal field theory.
In the last chapter we begin exploring the dS/CF T in the context of cosmology as the main topic of
this report and the most important application of the duality. In the first section an introductory review of
perturbation theory during inflation is provided. Then the duality between time evolution of the universe,
and renormalisation group flow is explained. In the first step towards formulation of a holographic framework
for our universe, the correspondence between domain-wall spaces and cosmology is defined. Subsequently
using the well-established AdS/CF T formalism observables are mapped to the equivalent quantum field
correlators of the domain-wall, and finally by analytical continuation we find an expression for the QFT dual
the original cosmology. We finish this report by a detailed study of a proposed holographic prototype, and
demonstrate that how predictions of this quantum model complies with latest observational data.
Before proceeding to detailed scientific discussions it is worth clarifying some points about the notation.
Summation convention is used in this article, that is repeated indices are summed over. Also note
that repeated covariant (contra-variant) indices are also summed, and it should be interpreted as they are
contracted with a Kronecker delta function, i.e. xµ xµ ≡ δµν xµ xν , and xµ xµ ≡ δ µν xµ xν .
Space-time coordinates have Greek indices while for spatial dimension Latin indices are used.
We adopt the notation that coordinates on the boundary manifold xµ are written in bold compared to
the bulk space-time coordinates xµ .
3
Chapter 2
Conformal Field Theory
Conformal symmetry is a generalisation of Poincare invarience. In a simple language, when working with
conformally invariant fields we are interested in theories that are the same for observables at different scales.
Some field theories like four dimensional Yang-Mills theory are conformally invariant, in general all fields
living at fixed points of renormalisation group flow are conformal. Most quantum fields have a UV unstable
fixed point, flowing towards another trivial IR fixed point in the space of renormalisation groups. In this
paper we are interested in conformal fields because they correspond to the gravitational theories in spaces
that are asymptotically (anti) de Sitter.
In the first part of this chapter conformal transformation in an arbitrary dimension is defined, then the
mathematical prescription of conformal groups, generators of infinitesimal transformation, conformal algebra,
and finally associated symmetries is provided in the following sections. Subsequently we specialise to the
case of infinite dimensional field theory in a two-dimension space. Then the main properties of a conformal
field are reviewed, and finally correlation functions under conformal invariance restriction are computed.
2.1 Conformal Transformation

Conformal transformation is a map between two manifolds such that: [11]
′
gµν (x′ ) = Ω2 (x)gµν (x) (2.1)
′
Where gµν and gµν are original and transformed metrics. Scale factor Ω is a continuous, non-vanishing,
finite, real function. In this paper always two manifolds are the same which implies g ′ = g. Note in case of a
flat metric, given a unity scale factor Ω(x) = 1, conformal transformation turns into Poincar transformation
which is essentially Lorentz transformation and translation.
Geometrically a conformal transformation can be defined as a map which locally preserves the angle
between any two curves on the manifold (figure 2.1). [10]
Conformal transformation 2.1 consists of a Weyl rescaling gµν (x) → Ω2 (x)gµν (x), followed by a coordinate
transformation (diffeomorphism) x → x′ so that Ω2 (x)gµν (x) → gµν ′
(x′ ). It can be shown as: [10]
α β
′ ∂x′ ∂x′
gαβ (x′ ) = Ω2 (x)gµν (x) (2.2)
∂xµ ∂xν
A significant feature of conformal maps is that they preserve the casual structure of space-time. For an
arbitrary tangent vector uµ , it is clear that:
gµν uµ uν → gµν
′
uµ uν = gµν Ω2 (x)uµ uν (2.3)
So if a curve is originally time-like/null/space-like with respect to g, it remains the same with respect to
g ′ ; hence invariance of casual structure under conformal mapping. Also null geodesics in the original metric
are transformed into null geodesics. However it is not necessarily true of time-like/ space-like geodesics, e.g.
geodesics of a massive particle dose not correspond to its geodesic after conformal transformation. [3, 11]
4
Figure 2.1: Conformal transformation in 2-dimensional Euclidean surface. (reproduced from [10])
2.2 Conformal Group

Conformal group is the group of all invertible and finite conformal transformations, i.e. transformations
which preserve conformal structure of the manifold. Global conformal group like any other connected Lie
group can be built upon generators of infinitesimal transformation. In the following sub-section relevant
infinitesimal transformations will be computed.
2.2.1 Generators
Conformal fields considered in this dissertation live on a a flat pseudo-Riemannian manifold with metric
ηµν = diag(−1, ..., +1, ...). Given a general infinitesimal coordinate transformation to the first order:
x′µ = xµ + ǫµ (x) (2.4)
It turns out that ǫµ can be at the most quadratic in xµ for a conformal scaling of the form 2.1: [10]
ǫµ (x) = aµ + bµν xν + cµνλ xν xλ (2.5)
Where aµ , bµν , and cµνλ are arbitrary coefficients and cµνλ is symmetric in the last two indices. There are
three possibilities available. If ǫµ (x) is a scalar then ǫµ = aµ , and we have infinitesimal translation:
xµ → xµ + aµ (2.6)
with momentum
Pµ = −i∂µ (2.7)
as the corresponding generator. [5]
If ǫµ (x) is linear, then bµν can be divided into a symmetric part ηνµ , and an anti-symmetric part wνµ such
that bµν = αηνµ + ωνµ . Then symmetric part gives infinitesimal scaling x′µ = (1 + α) xµ with generator [5]
D = −ixµ ∂µ (2.8)
Finite scaling can be written as:

xµ → λxµ (2.9)
Where λ = 1 + α + O(α2 )
The anti-symmetric part describes infinitesimal rotation x′µ = (δνµ + ωνµ ) xν which has angular momentum
[5]
Lµν = i (xµ ∂ν − xν ∂ν) (2.10)
as the generator. The finite rotation is usually noted as:
xµ → Mνµ xν (2.11)
5
Figure 2.2: Special Conformal transformation in 2-dimensional Euclidean surface
and rotation matrix is expanded like Mνµ = δνµ + ω(x)µν + O(ω 2 )

Finally with ǫµ (x) being quadratic in xµ , one can reach to the restriction [10] cµνλ = ηµν cλ +ηµλ cν −ηνλ cµ
where cµ = d1 cνµν . This will give rise to infinitesimal form of so-called special conformal transformation
x′µ = xµ + 2 ηνλ xν cλ xµ + cµ x2 . The associated generator is: [10]

Kµ = −i 2xµ xν ∂ν − x2 ∂µ

(2.12)
This type of transformation will be studied in more details below.
2.2.2 Special Conformal Transformation

Special conformal transformation (SCT) is a type of spherical wave transformations (transformations which
leave a spherical wave invariant), and as mentioned above, one of the generators of conformal symmetry.
SCT’s could be obtained by a composition of space-time translation, and inversion. One can easily check
that an inversion xµ → xµ /x2 , followed by a translation → x′µ = xµ /x2 +cµ , and another inversion → x′µ /x′2
give rise to the following expression for finite SCT (figure 2.2) : [10]
xµ + x2 cµ
xµ → (2.13)
1 + 2xµ cν ηµν + c2 x2
Where expansion of denominator for small values of cµ results in infinitesimal version of transformation
discussed earlier. Note that for an arbitrary vector cµ , there is always a point xµ on the manifold such
that the denominator vanishes. It means that xµ is conformally mapped to infinity. In order to avoid this
problem, it is required to define conformal fields on a compactified space which includes points at infinity.
This will become more clear when CFT’s on compactification of space-time boundaries are discussed in the
following chapters. [10, 37]
2.2.3 Conformal Algebra

To summarise the results so far, we introduced the finite conformal transformations and the related generators
as follows:
1. Pµ generator of transaltion xµ → xµ + aµ
2. Mµν generator of the Lorentz transformation xµ → Λµν xµ
3. D generator of scaling xµ → λxµ

4. Kµ generator of special conformal transformation xµ → (xµ + x2 cµ )/(1 + 2xµ cν ηµν + c2 x2 )
6
Generators of infinitesimal conformal transformation constitute the conformal algebra which admits the
following commutation relations: [10]
[Mµν , Pρ ] = −i(ηµρ Pν − ηνρ Pµ ), [Mµν , Kρ ] = −i(ηµρ Kν − ηνρ Kµ ),

[Mµν , Mρσ ] = −i(ηµρ Mνσ − ηµσ Mνρ − ηνρ Mµσ + ηνσ Mµρ ), [Mµν , D] = 0, (2.14)
[D, Pµ ] = −iPµ , [D, Kµ ] = −iKµ , [Pµ , Kµ ] = 2iMµν − 2iηµν D
and all other commutations vanishes.
2.2.4 Symmetries
In order to find the dimension of the algebra, one needs to count the number of generators. Total number of
generators for Lµν , Pµ , Kµ , and D are:
1 1
d(d − 1) + d + d + 1 = (d + 2)(d + 1)
2 2
which equals the dimension of SO(p + 1, q + 1) symmetry group in a Mp,q manifold where p + q = d,.
The conformal group on a Mp,q manifold is SO(p + 1, q + 1).
If inversion xµ → xµ /x2 is included, then conformal group will be O(p + 1, q + 1) with connected subgroup
SO(p + 1, q + 1). Special cases of interest are d-dimensional Laplacian space Md−1,1 where conformal group
is SO(d, 2), as well as Euclidean d-dimensional manifold with a SO(d + 1, 1) conformal group.
Familiar generators of special orthogonal group can be defined via: [10]
1 1
Jµν = Mµν , Jµd = (Kµ − Pµ ), Jµ(d+1) = (Kµ + Pµ ), Jd(d+1) = D (2.15)
2 2
with µ, ν = 0, . . . d − 1. In matrix form it is clear that Jmn is antisymmetric with a Lorentz subgroup:
1 1
 
Mµν 2 (Kµ − Pµ ) 2 (Kµ + Pµ )
Jmn = − 21 (Kµ − Pµ ) 0 D 
− 21 (Kµ + Pµ ) D 0 (d+2)×(d+2)
The usual commutation relation is defined as: [11]
[Jµν , Jρσ ] = i (ηµσ Jνρ + ηνρ Jµσ − ηµρ Jνσ − ηνσ Jµρ ) (2.16)
2.3 Conformal Fields in Two Dimension

In this section we specialise to the case of conformal symmetry in two dimensions. It is of special importance
because the algebra is infinite-dimensional in this case. Consequently two-dimensional conformal fields have
an infinite number of conserved quantities, which means symmetry considerations alone suffice to make them
completely solvable. [10]
2.3.1 Complexification
Introduce complex variables:
z = x + iy, z̄ = x − iy (2.17)
which implies:
1 1
∂z = (∂x − ∂y ) , ∂z̄ = (∂x + ∂y ) (2.18)
2 2
7
Also metric will be:
ds2 = dx2 + dy 2 = dzdz̄ (2.19)
In two dimensions, the restriction that ǫ(x) be at the most quadratic in x no longer applies. In fact it
turns out that ǫ must be either an arbitrary function of z or of z̄. So the complexified version of infinitesimal
transformation 2.4 can be shown by a holomorphic function f (z) = z + ǫ(z) or f¯(z̄) = z̄ + ǭ(z̄). That is
infinitesimal two-dimensional transformation is an analytic coordinate transformation: [10]
z → f (z), z̄ → f¯(z̄) (2.20)
Which implies metric will transform like:
∂f ∂ f¯
ds2 → dzdz̄ (2.21)
∂z ∂ z̄
So scale factor in two dimensions is Ω = | ∂f 2
∂z | .
Minkowski Manifold
The complexification above was done on an Euclidean manifold, however one can generalise it to a Minkowski
space-time by defining the new coordinates u and v:
u = −t + v, v =t+v (2.22)
which changes the metric to:

ds2 = −dt2 + dx2 = dudv (2.23)
The infinitesimal transformation is therefore given by:
u → f (u), v → g(v) (2.24)
Euclidean and Minkowski versions are related to each other by a Wick rotation. So without loss of
generality we continue the discussion on a Euclidean manifold, and treat z and z̄ as two independent variables.
2.3.2 The Witt Algebra

In order to obtain infinitesimal transformation 2.4 in two dimensions, one needs to expand ǫ around zero in
Laurent series:
X
z′ = z + ǫ = z + an −z n+1 ,

n
X
′
ān −z̄ n+1

z̄ = z̄ + ǭ = z̄ +
n
The corresponding generators of infinitesimal transformation is thus given by: [10]
ln = −z n+1 ∂z , ¯ln = −z̄ n+1 ∂z̄ (2.25)

n+1

where ln is the generator of transformation with parameter ǫ = an −z . As usual to define the
algebra, it is required to work out the commutators of generators: [10]
[lm , ln ] = (m − n)lm+n ,
¯lm , ¯ln = (m − n)¯lm+n ,

(2.26)
lm , ¯ln = 0

This algebra is known as Witt algebra. Since n ∈ Z the algebra is infinite dimensional which is a unique
feature of conformal group in two dimensions. 1
1 In fact each commutation relation is an independent copy of Witt algebra because these two copies commute. This also
justifies our assumption about considering z and z̄ as two independent variables.
8
2.3.3 Global Conformal Transformation
The Witt algebra is generated by infinitesimal conformal transformation, in this section we proceed to consider
the global transformations in two dimensions. Note that generators of Witt algebra are singular at certain
points:
1. Consider 2.25 the generator ln = −z n+1 ∂z is non-singular at z = 0 only if n ≥ −1.
2. In order to study the behaviour near z → ∞, we perform change of variable w = 1/z, and investigate
w → 0. It can be seen that ln is non-singular at w = 0 only if n ≤ 1.
It can be concluded that the only globally defined generators of conformal transformation are l−1 , l0 and l1 .
Therefore conformal group in two-dimensions is locally described by Witt algebra, and has infinite dimensions,
in contrast, globally it is generated by the mentioned three generators and thus has six dimensions.
2.4 Conformal Fields

All relativistic quantum fields in the Standard Model of particle physics have Pioncare group ISO(3, 1)
symmetry. Conformal Field is a relativistic quantum field (in d-dimension) which is invariant under conformal
group SO(d, 2) not only Pioncare group ISO(d − 1, 1).
Physicists are interested in representations with operators that have well-defined scaling properties which
means operators that are eigenfunction of scaling operator D (with eigenvalue −i∆). ∆ is refereed to as
scaling dimension or conformal dimension. Under the scaling xµ → λxµ the operator is charged like:
O(x) → λ∆ O(λx) (2.27)
Back to commutation relations 2.2.3 it can be noticed that action of momentum Pµ increases the scaling
dimension while Kµ decreases the dimension. It turns out that there is a minimum value of conformal
dimension, and the corresponding operator which is annihilated upon action of Kµ at the origin x = 0 is
called primary operator. 2 Commutation relations of the conformal operator with generators of conformal
group are given by: [37]
[Pµ , O(x)] = i∂µ O(x), (2.28)
[Mµν , O(x)] = {i(xµ ∂ν − xν ∂µ ) + Aµν }O(x), (2.29)
µ
[D, O(x)] = i(−∆ + x ∂µ )O(x),
[Kµ , O(x)] = {i(x2 ∂µ − 2xµ xν ∂ν + 2xµ ∆) − 2xν Aµν }O(x) (2.30)
where Aµν is a matrix in the finite representation of the Lorentz group. It is worth mentioning that
because representations of conformal group are eigenfunctions of scaling operator, they are not necessarily
eigenfunctions of the Hamiltonian or mass operators. [10]
2.5 Correlation Functions

Correlation functions are the building blocks of the physical observables. High symmetries of conformal group
in comparison with Poincare , imposes strong restriction on the correlation functions. General form of the
most useful correlators that are two-point functions will be reviewed in this section.
2.5.1 Two-Point Correlation Function

Two-point correlation function of operators O1 and O2 with dimensions ∆1 and ∆2 is denoted as g(x1 , x2 ) =
hO1 (x1 )O2 (x2 )i.
Invariance under translation xµ → xµ + aµ means that the two-point function should depend on the
difference between the two points: [10]
g(x1 , x2 ) = g(x1 − x2 ) (2.31)
2 For a scalar field the minimum bound is ∆ ≥ (d − 2)/2 which describes the free scalar. [11]
9
Invariance under rescaling xµ → λxµ implies:
hO1 (x1 )O2 (x2 )i → hλ∆1 O1 (x1 )λ∆2 O2 (x2 )i

λ∆1 +∆2 g (λ(x1 − x2 )) ≡ g(x1 − x2 )
This amount to the following form:

d12
g(x1 − x2 ) = (2.32)
(x1 − x2 )∆1 +∆2
where d12 is a structure constant. Finally correlation function is invariant under special conformal trans-
formation which is composed of translation and inversion. Since restriction under translation was already
considered, we concentrate on inversion symmetry:
1 1 1 1
hO1 (x1 )O2 (x2 )i → h O1 (− )O2 (− )i
x2∆
1
1
x2∆
2
2 x1 x2
1 d12 d12
= ≡
x2∆ 1 2∆2
∆1 +∆2
x2 (x1 − x2 )∆1 +∆2

1 − x11 + 1
x2
Last equality holds if scaling wights of the two operators are the same ∆1 = ∆2 ≡ ∆. So ultimately
two-point correlator will be fixed by conformal symmetry as:
d12
hO(x1 )O(x2 )i = (2.33)
(x1 − x2 )2∆
10
Chapter 3
Anti de Sitter Space
Anti-de Sitter is a manifold with constant negative curvature analogous to Euclidean hyperbolic space.
Spaces of constant curvature play a significant role in gauge/duality correspondence as existence of such a
correspondence depends on cosmological constant. It turns out that boundary of compactified anti-de Sitter
is the compactified Minkowski space where the dual field theory lives. It is therefore important to be familiar
with interior and asymptotic regions of this space, in order to have a better understanding of correspondence
between gravity and conformal field theory.
This chapter starts with a review of spaces of constant curvature which includes anti-de Sitter and de
Sitter. Following this introduction geometrical properties of anti-de Sitter space are studied. Then we proceed
to introduce the most common coordinate systems and show how they provide a better understanding of
symmetries and topology. Final section focuses on the presence of matter field in this space-time, and general
solutions are derived. We will refer to these results frequently throughput this paper.
3.1 Constant Curvature Spaces

The spaces of constant curvature or space forms are the simplest examples of Riemannian manifolds where
the sectional curvature K has the same value at any point on the manifold. As a result Ricci Scalar R which
is defined as a multiple of the average of the sectional curvatures at a point is constant and will be given by
R = n(n − 1)K.[1]
In general for manifolds of more than three dimensions d > 3 neither Ricci scalar R or Ricci tensor Rµν is
sufficient to describe the curvature and the full Riemann tensor Rµνλσ is needed. The remaining components
of curvature tensor is represented by so-called Wyle Tensor Cµνλσ . [2]
However one can show that for a space form Riemann tensor can be locally defined by Ricci scalar alone
through the condition:
1
Rµνλσ = R (gµλ gνσ − gµσ gνλ ) (3.1)
12
This means that Wyle tensor vanishes and Cµνλσ = Rµν − 14 Rgµν = 0. A straight forward calculation
using contracted Bianchi identity verifies that Ricci scalar is constant throughout the manifold. [2, 3]
In order to get a physical interpretation of space-time one needs to calculate the Einstein tensor Gµν :
1 1
Gµν = Rµν − Rgµν = − Rgµν (3.2)
2 4
So spaces of constant curvature can be considered as the vacuum solutions to the Einstein equation with
cosmological constant Λ = 14 R. [3]
Space forms are important since they are locally maximally symmetric i.e.they have 12 n(n + 1) number
of local isometries.Also any maximally symmetric space has a constant curvature. These manifolds have
translation invariance and thus are homogeneous.[1]
11
3.1.1 Einstein Static Universe
In the following sections, in many occasions, manifolds of constant curvature will be mapped to the static
universe, so it is useful to have an overview of this cosmological space-time at this point.
Einstein static universe is a stationary cosmological model being temporally infinite but spatially finite. It
is completely homogeneous with spatial spherical symmetry and topology R × S 3 .[3]. The metric is described
by: [26]
dr2

2 2 2 2 2 2 2
ds = −dt + a √ + r dθ + r sin θdφ (3.3)
1 − r2
Note a is a constant independent of time. By change of variables r′ = Arcsin r metric can be re-written
as:
ds2 = −dt2 + dr′2 + sin2 r′ dθ2 + sin2 θdφ2

(3.4)
So one can regard Einstein static universe as an embedded cylinder:
x2 + y 2 + z 2 + w 2 = l 2 (3.5)
in a five-dimensional Minkowski manifold ds2 = −dt2 +dx2 +dy 2 +dz 2 +dw2 . Where l is radius of cylinder,
and has length dimension. In this paper for better visualisation, we usually suppress two coordinates θ and
φ, and represent the model as a cylinder in a three dimensional space.
Einstein universe is also a solution of Friedmann-Robertson-Walker (FRW) metric with a positive cur-
vature k = 1 if cosmological constant is at a critical value Λc = 3/a2 + 8πGρm where ρm is the matter
energy density. In this case the attractive force of the matter density is balanced by the repulsive force of
the cosmological constant. But it turns out to be an unstable solution thus not describing a physical state.
3.2 Geometry of Anti de Sitter Space

Given a (d + 1)-dimensional manifold Mp,q with metric:
ds2 = −dT i dT i + dX j dX j i + j = 0, . . . d (3.6)
Anti de Sitter space AdSd can be realised as a hyperboloid induced in one less temporal dimension
through:
−T i T i + X j X j = l2 i + j = 0, . . . d (3.7)
Parameter l referred to as anti de Sitter radius has dimensions of length being the only scale of the
geometry. In physics we are interested in the special case where q = 2 resulting into an embedded space-time
AdSp,1 . Anti de Sitter metric is nondegenerate and has Lorentzian signature, it has topology S 1 × Rd−1 . [1]
O(d − 1, 2) is the isometry group of anti de Sitter in d dimensions, and has SO(d − 1, 2) as the connected
subgroup.
In addition, anti de Sitter is the vacuum solution to the Einstein equation with negative cosmological
constant: [2]
1
Rµν − Rgµν + Λgµν = 0 Λ<0 (3.8)
2
Riemann tensor in AdSd manifold is given by: [3]
1
Rµνλσ = − (gµλ gνσ − gµσ gνλ ) (3.9)
l2
Ricci tensor is proportional to the metric, so anti de Sitter space is a an Einstein Manifold. [3]
d−1
Rµν = − gµν (3.10)
l2
Ricci scalar is given by: [3]
d(d − 1)
R=− (3.11)
l2
12
Figure 3.1: AdS3 space mapped into the 3-dimensional Einstein universe. Boundary θ = π/2 is an S 1 circle.
(reproduced from [37])
and finally cosmological constant will be: [3, 56]
(d − 1)(d − 2) d−2
Λ=− 2
=− R (3.12)
2l 2d
In a 4-dimensional space-time we have the relation R = 12/l2 = 4Λ which is the same result as in section
3.1.
3.3 Coordinate Systems

Due to geometrical symmetries, anti de Sitter space admits a number of coordinate systems. In this section
we review global coordinates and Poincare patch as the most commonly used systems. These coordinates
provide a better understanding about topology and symmetries of the AdSd , and we frequently refer to them
in this paper.
3.3.1 Global Coordinates

Global coordinates (τ, ρ, ω i ) covers the whole anti-de Sitter space, and obtained via parametrisation:
 0
 X = l cosh ρ cos τ,
X 1 = l cosh ρ sin τ, (3.13)
 i
X = l sinh ρ xi i = 2, ..., d
where xi is coordinate on a unit sphere S d−2 such that xi xi = 1. The metric takes the form: [3]
ds2 = l2 (− cosh2 ρ dτ 2 + dρ2 + sinh2 ρ dΩ2d−2 (3.14)
where ρ ≥ 0 and 0 ≤ τ < 2π. It can be seen that the metric has closed time-like curves in τ direction on
S 1 , so we need to unwrap this dimension such that −∞ < τ < ∞ and thus space-time is casual. This space
is called the universal covering of AdS, and has topology R4 . [3]
Isometry group of AdSd has the subgroup SO(d−1, 2) ⊃ SO(2)×SO(d−1) where SO(2) is time evolution
in τ direction, and SO(d − 1) corresponds to rotations in S d−2 . [37]
In order to investigate casual structure of AdS, one can demand tg θ = sinh ρ where 0 ≤ θ < π/2:
l2
ds2 = (−dτ 2 + dθ2 + sin2 θdΩ2(d−2) )
cos2 θ
13
Figure 3.2: Penrose diagram for AdS2 in Poincare patch. (reproduced from [37])
Now we can do a conformal compactification, and rescale the metric by cos2 θ/l2 , then metric takes the
form of Einstein static universe 3.4: [3]
ds2 = −dτ 2 + dθ2 + sin2 θdΩ2(d−2) (3.15)
However in since 0 ≤ θ < π/2, anti-de Sitter space is mapped to half of Einstein universe.
Note that constant τ hyper-surfaces has topology of a (d − 1)-hemisphere. Boundary of AdS corresponds
to θ = π/2 on this surface which is a S d−2 (figure 3.1). So boundary of con formally compactified AdSd i.e.
R × S d−2 is the conformally compactified Minkowski space Md−1,1 .
3.3.2 Poincare Coordinates

Poincare coordinates (t, u, xi ) cover half of the AdSd space, and are given by the parametrisation: [1]
 0
 X = (1/2u) 1 + u2 (l2 + xi xi − t2 ) ,

X 1 = (1/2u) 1 − u2 (l2 − xi xi + t2 ) , (3.16)

 i
X = luxi i = 2, ..., d
where u > 0, then the metric takes the form:

2
du
ds2 = l2 + u 2
(−dt 2
+ dx i
dx i
) i = 2, . . . d − 1 (3.17)
u2
Penrose diagram for AdS2 is illustrated in figure 3.2.
In this system isometry group of AdS has subgroups SO(d − 1, 2) ⊃ ISO(d − 2, 1) × SO(1, 1) where
ISO(d − 2, 1) is the Poincare transformation on (t, xi ) dimensions, and SO(1, 1) is scaling in the conformal
symmetry group SO(d − 2, 1).
By a change of coordinate z = 1/u the Poincare metric is commonly written as: [1]
l2
ds2 = dz 2 − dt2 + dxi dxi

2
i = 2, . . . d − 1 (3.18)
z
In this metric boundary is approached via z → 0, while AdS interior is the region z → ∞.
Poincare metric can also be brought into radial coordinates by a change of variable z = le−r/l :
ds2 = dr2 + e2r/l ηµν (dxµ dxν ) µ, ν = 1, . . . d − 1 (3.19)
In these coordinates boundary corresponds to r → ∞, and the deep interior is approached as r → −∞.
14
3.4 Matter Field in AdS Space
Having discussed geometrical properties of Anti de Sitter manifold so far, in this section we study presence
of matter field in this space. The simplest model will be a single scalar field on a fixed AdSd background.
The action in Poincare coordinates reads:
1 √ µν
Z
−g g ∂µ φ∂ν φ − m2 φ2 dd−1 xdz

S= (3.20)
2
Varying the action with respect to the field gives the Klein-Gorden equation:
( − m2 ) φ = 0 (3.21)
with usual definition of Laplacian:

1 √
φ = g µν φ,µ;ν = √ ∂µ ( gg µν ∂ν φ) (3.22)
g
Due to translation invariance on the x coordinates, we try separation of variables:
φ(z, x) = f (z)Φ(x) (3.23)
Following the steps specified in details in appendix A this leads to two independent equations for f (z)
and Φ(x): [38]
(∂x2 − k 2 )Φ(x) = 0 (3.24a)

d d 2 2 2 2

−z ∂z (z ∂z ) + m l + k z f (z) = 0 (3.24b)
where k is momentum, and ∂x2 indicates Laplacian on Minkowski x-plane. Solution to the Laplace equation
3.24a on the flat space are known to be the plane waves. Consequently the complete solution will be given
by superposition of plane waves and f (z):
dd−1 x
Z
φ(z, x) = f˜k eikx (3.25)
(2π)d−1
This indicates that the solution is actually the Fourier transform of f (z). By inverting the Fourier
transformation above, one realises that f˜k is in fact, the solution in momentum space. So the momentum-
space version of equation of motion 3.24b can be written as:
z 2 f˜k′′ − (d − 2)z f˜k′ − (m2 l2 + k 2 z 2 )f˜k = 0 (3.26)
By a change of coordinates this equation amount to the modified Bessel equation (see appendix A) with
the solution given by: [23]
f˜k (z) = ak (kz)(d−1)/2 Kδ (kz) + bk (kz)(d−1)/2 Iδ (kz) (3.27)

p
where δ = (d − 1)2 /4 + m2 l2 , and Iδ and Kδ are modified Bessel functions of first and second type
respectively. Imposing regularity in the interior and considering asymptotic behavior of Bessel functions
A.6, when approaching the boundary z → 0 the solution in the momentum space behaves like: [39]
f˜k ≈ φ̃− (k)z ∆− + φ̃+ (k)z ∆+ (3.28)
where ∆± are given by ∆± = (d − 1)/2 ± δ that is:

r
(d − 1) (d − 1)2
∆± = ± + m2 l 2 (3.29)
2 4
We come back to this relation later. Finally applying the inverse Fourier transformation, the solution to
the equation of motion for a scalar field in an anti-de Sitter background will be given by:
15
lim φ(z, x) = z ∆− φ− (x) + z ∆+ φ+ (x) (3.30)
z→0
The above asymptotic behavior of the field bulk when approaching the boundary will play a significant
role in gauge/gravity duality discussions in the following chapters.
16
Chapter 4
AdS-CFT Duality
Although the main focus of this report is on de Sitter spaces, understanding AdS/CF T correspondence is
in fact a prerequisite in studying the positive curvature counterpart i.e. dS/CF T duality. Formulation of
gauge/gravity duality in de Sitter space was developed along the lines of the anti-de Sitter counterpart, and
in many cases well-established results of AdS/CF T relation were used as a conjecture in dS/CF T version of
the duality. 1
In preceding chapters, on one hand a conformally invariant description of quantum fields, and on the
other hand gravitational properties of Anti-de Sitter geometry were studied. We are now ready to explain
the duality between these two fundamentally different theories. What is meant by duality is actually a one-to-
one correspondence between fields and operators on both sides, and consequently a map between correlation
functions in both theories.
Some literature start with an explicit example, usually the AdS5 × S 5 /N = 4, SY M correspondence due
to historical reasons, and then proceed to discuss the generalised prescription. In this dissertation, however,
a top-to-bottom formalism is selected: we first explain the abstract AdS/CF T duality conjectures, and then
will exhibit some examples of the correspondence.
In first section, we make a comparison between AdS space-time isometries and symmetries of conformal
QFT as the initial motivation behind the duality. In the following section a precise abstract formulation of
gauge theory / gravity identity is discussed. In particular the correspondence between quantum field operators
on the boundary and gravity fields in the bulk is defined. Computation of correlation functions as the main
operators of the theory is carried out in the next section. Subsequently a comprehensive mapping between
both sides of the duality will be exhibited in the following section. Next we proceed to give examples of
AdS/CF T conjecture, and start with the well-known Maldacena or AdS5 ×S 5 /N = 4, SY M correspondence.
It is discussed how Typ IIB supergravity theory on AdS5 ×S 5 geometry is identical to N = 4 super Yang-Mill
theory on parallel D3 branes. In addition three different types of correspondence including weak, semi-strong,
and strong forms will be reviewed.
4.1 Symmetries
Description of anti de Sitter space in radial coordinates 3.19 provides a suitable way to understand symme-
tries:
ds2 = dr2 e2r/l ηµν (dxµ dxν )
The boundary of AdS corresponds to the limit r → ∞ where the field theory lives. The metric is clearly
symmetric with respect to the transformations r → r + a, and ηµν → e−2a/l ηµν . Near the boundary the
second term dominates, so ds2 = ηµν (dxµ dxν ) describes the flat metric of the quantum fields. But the field
is invariant under the scaling of the metric η which means the field theory is conformal.
It is a necessary condition for AdS/CF T correspondence that unbroken symmetries on both sides of the
duality should be the same. It was discussed that isometry group of AdSd is SO(d − 1, 1) which is identified
1 In the last two chapters we will see explicit examples indicating how AdS/CF T prescription facilitates development of
dS/CF T formalism.
17
with the conformal symmetry group on a (d − 1)-dimensional flat space. This is the first step in establishing
a duality map between gravity in the bulk anti-de Sitter and conformal field theory on the boundary region.
4.2 The Field - Operator Correspondence

The duality between two theories, in some sense, means that there exists a one-to-one mapping between
their observables. Since conformal field theories do not have an S-matrix or asymptotic state, we therefore
consider operators, and related correlation functions as observables [10]. So it is expected that AdS/CF T
duality shows a connection between fields in anti-de Sitter space, and operators of the conformally invariant
field theory.
The first step is to define a boundary condition. Recalling the general form of solutions in 3.30, we
demand that limz→0 φ(z, x) = z ∆− φ− (x). This admits that near the boundary the bulk field factorises into
a product of a z-dependent part and a local operator φ− (x) ≡ φ0 (x). It is prescribed that boundary value
of the local bulk field φ0 (x) functions as the external source J(x) for the operators of CFT living on the
boundary of AdS space: [37, 39]
φ0 (x) = J(x) (4.1)
where x is to be understood as the coordinates on the boundary x ∈ ∂AdS. 2 Under this condition, it is
conjectured that partition function of the bulk classical gravity Zcl equals the generating functional of the
conformal field theory Z[J]: [37, 38]
R
Z
d−1
hei J(x)O(x)d x i = e−Scl [φ] Dφ

(4.2)
φ0 (x)=J(x)
Left hand
R
side describes CFT generating functional which is defined in a similar fashion to QFT via
d−1
Z[J] = he J(x)O(x)d x i,3 and right hand side is the classical gravity partition function Zcl = e−Scl [φ] Dφ.
R
It immediately follows that there is a duality between the correlation functions in the two theories: [41]
hO(x1 )...O(xn )iCF T ↔ hφ(x0 , x1 )...φ(x0 , xn )iAdS (4.3)
This is the main idea behind the gauge/gravity duality, meaning that any bulk fields φ is in a one-to-one
correspondence with a conformally invariant operator Oφ .
The n-point correlation function for CFT operators is therefore given by: [40]

−iδ −iδ
hO(x1 )...O(xn )i = ... Zcl (4.4)
δJ(x1 ) δJ(xn ) J=0
In this equation Zcl is classical gravity partition function.

Looking at the left hand side of equation 4.2, because quantum generating functional Z[J] includes
bubbles, the resulting correlation function will be disconnected. To have connected correlators we need to
define connected generating functional E[J] via Z[J] = e−iE[J] .
On the right hand side, we use saddle point approximation for classical gravity theory which means
in the limit where action Scl is large, the integral is dominated by the points which extremise the action
(saddle points) δScl /δφ = 0. So it is required to solve the equations of motion, and find the on-shell
solutions. Consequently
classical gravitational partition function to the leading order is approximated by:
Zcl ≈ e−Scl on−shell .
Under these assumptions, equation 4.2 is simplified to: [37, 41]

iE[J] = Son−shell [φ] (4.5)
φ0 (x)=J(x)
2 The motivation behind this will become clear when the AdS × S 5 /N = 4, SY M example is explained in the last section.
5
In this case boundary value of dilaton in the bulk corresponds to the coupling constant of super-Yang-Mill theory.
3 Similar to the case for QFT, taking J(x)O(x) as interaction Lagrangian, Z[J] will be sum of one-point vertices without
external legs. Note J(x) acts as position-dependent coupling. Z[J] includes sum of infinite number of separate pieces (bubbles)
which is simply Z[0].
18
Which means generator of connected correlation functions in the conformal theory is the on-shell classical
gravity action. Now quantum connected correlation functions can be obtained by taking functional derivatives
with respect to the on-shell gravitational action: [41]

δ δ
hO(x1 )...O(xn )iconnected = ... Son−shell (4.6)
δJ(x1 ) δJ(xn ) J=0
In fact, the same reasoning applies to other types of fields. It turns out that bulk spinor field ψa is related
to fermionic operator Oa on the boundary, bulk gauge field Aµ is dual to the boundary conserved current
Ji , and bulk metric gµν corresponds to the energy-momentum tensor Tij : [20, 21]
δSon−shell
φ(x0 , x) ↔ hO(x)i = (4.7)
δφ0 (x)
δSon−shell
ψa (x0 , x) ↔ hOa (x)i = (4.8)
δψa(0) (x)
δSon−shell
Aµ (x0 , x) ↔ hJi (x)i = (4.9)
δAi(0) (x)
δSon−shell
gµν (x0 , x) ↔ hTij (x)i = 2 (4.10)
δgij(0) (x)
where ψa(0) (x), Ai(0) (x) and gij(0) (x) are boundary values of the local spinor, gauge field and the metric.
4
4.3 Green Functions

In this section, we review derivation of green function as the response of the the bulk system in presence of a
point source with a specified boundary condition. The results will be used in the next section in formulation
of the correlation functions.
4.3.1 Bulk-to-Bulk Green Function

The bulk-to-bulk propagator is defined via:
i
+ m2 G(x; y) = − √ δ d+1 (x − y)

(4.11)
g
We adopt the notation that xi (in bold) indicates the transverse (d−1)-dimensional coordinates, and xµ is
d-dimensional bulk coordinates (z, xi ). Recall that Klein-Gorden equation 3.21 had two linear independent
solutions 3.30 which were multiples of the modified Bessel functions of first and second type:
f˜k (z) = (kz)(d−1)/2 {Kδ (kz), Iδ (kz)} (4.12)
Using these modes one can try to find the green function and make an ansatz of the form: [50]
dd−1 k
Z
G(x, y) = z (d−1)/2 h(k, z ′ ){θ(z − z ′ )Kδ (kz)Iδ (kz ′ ) + θ(z ′ − z)Kδ (kz ′ )Iδ (kz)}e−ik(x−y) (4.13)
(2π)d−1
One can check that this ansatz satisfy the boundary conditions at z = 0, and z → ∞, and it is continued
at z = z ′ . By matching this two region at z = z ′ we can find: h(k, z ′ ) = −z ′(d−1)/2 . The expression above
then can be integrated which gives the hypergeometric function: [39]
4 Any field f has an asymptotic Fefferman-Graham expansion of the form f (z, x) =

z 2m f(0) (x) + f(2) (x) + . . . + z 2n f(2n) (x) + ln z 2 h(2n) (x) + . . .

with different values of coefficients f(i) and h(j) , and
integers m and n. What we mean by local values of the field on the boundary is actually the leading term f(0) (x) in
Fefferman-Graham expansion. [41]
19
∆
C∆ ξ ∆ 1 d 2
G(x, y) = F , (∆ + 1), ∆ − + 1, ξ (4.14)
2∆ − d 2 2 2 2
the normalisation constant C∆ will be fixed later in 4.22, and
2zz ′
ξ= (4.15)
z 2 + z ′2 + (x − y)2
At short distance x − y → 0 bulk-to-bulk propagator behaves like: [39]

z1 z2
G(x, y) ≈ C∆ (4.16)
(z1 − z2 )2 + (x − y)2
4.3.2 Bulk-to-Boundary Green Function

When the point source is located on the boundary, we need to define another solution of Laplace equation 4.11
known as Bulk-to-Boundary Green Function K(x0 , x; y). For convenience we work in Poincare coordinate
√
3.18. On the boundary x0 ≡ z → 0, and since 1/ g ∼ z the right hand side of the equation vanishes and the
definition of green function is simplified to: [43]
− m2 K(z, x; y) = 0

(4.17)
As usual the response solution is given by the convolution with the source: [6]
Z
φ(z, x) = K(z, x; y)J(y)dd−1 y (4.18)
As discussed source is local part of the field on the boundary J(x) ≡ φ0 (x).
The methodology used is to solve Laplace equation 4.11, investigate if it is singular at some points as
singularity implies sources, and finally identify the kind of source. Pick the bulk point to be at infinity
z → ∞, then the transverse −dt2 + dxi dxi part of metric 3.18 vanishes because l2 /z 2 → 0, so the space
becomes a point and Laplace equation simplifies to : [43]
d
−z ∂z (z −d+2 ∂z ) + m2 l2 K(z) = 0

(4.19)
Try the ansatz K(z) = C∆ z ∆ which amounts to:
∆(d − 1 − ∆) + m2 l2 z ∆ = 0

(4.20)
p
The solution to the above equation clearly has two roots ∆± = (d − 1)/2 ± (d − 1)2 /4 + m2 l2 which is
the same relation as 3.29.
Due to asymptotic behavior near the boundary larger root ∆+ is chosen. The green function ought to be
z
invariant under conformal transformations that are isometries of AdS space. Doing an inversion z → z2 +x 2,
followed by a translation x → x − y, the boundary-bulk propagator becomes: [39]
∆ +
z
K(z, x; y) = C∆+ (4.21)
z 2 + (x − y)2
where normalisation constant is fixed as (refer to appendix B):
Γ(∆+ )
C ∆+ = d−1
(4.22)
π (d−1)/2Γ(∆+ − 2 )
20
Accordingly K vanishes on the boundary z → 0 except at one point, it diverges where (x − y) → 0
on the boundary. This suggests that propagator K is singular at x = y, and as shown in appendix B
Kz −∆− behaves like a delta function as z approaches the boundary. So the boundary condition for the
boundary-to-bulk propagator will be: [37]
lim K(z, x; y) = z ∆− δ (d−1) (x − y) (4.23)

z→0
So using equation 4.18, near the boundary the field behaves like:
lim φ(z, x) = z ∆− φ0 (x) (4.24)

z→0
That implies on the boundary, transverse dimension part of the field z ∆− factorises, leaving a local field
operator φ0 (x). [57]
Since the field φ0 (x) is dimensionless, the boundary Rfield φ(z, x) acting as the source J(x) has mass
dimension ∆− . Recall external source term in the action J(x)O(x)dd−1 x, counting dimensions yields that
the field in the bulk AdS is dual to an operator of dimension ∆+ on the boundary CFT.

Since all the observable can be expressed in terms of correlation functions, our knowledge of n-point correlators
allows us to work out all physical quantities describing the system. The correlation functions are also
important because they provide a way of testing the theoretical model.
It was shown earlier that how correlators could be computed by taking derivatives of supergravity action
with respect to the source field (equation 4.6). In this section we use green function results, and start with
two-point correlation functions which are the simplest example.
4.4.1 Two- Point Functions

In case of two-point function, only quadratic terms in the supergravity action are relevant since correlator is
given by: [43]
−iδ −iδ
hO(x1 )O(x2 )i = Ssugra (4.25)
δφ0 (x1 ) δφ0 (x2 ) φ0 =0
where source operator J(x) is the local part of the bulk field at the boundary J(x) ≡ φ0 (x). In addition
interaction term or any other higher terms can be ignored as source φ0 is put at zero at the end. So the
action considered here is that of a free scalar field in AdSd 3.20:
1 √ µν
Z

S=
2
After integration by part the on-shell supergravity action becomes: (refer to appendix C)
1 √ 1 √
Z Z
−gg µν φ ∂ν φ dd−1 xdz − −gφ( − m2 )φ dd−1 xdz

S= ∂µ (4.26)
2 2
It can be shown that (appendix C) on the boundary, on-shell action reduces to:
1
Z
lim S = − z 2−d φ ∂z φ dd−1 x (4.27)
z→0 2 ∂AdS
By finding the asymptotic behavior of ∂z φ, one finally finds:
1 ∆+ φ0 (x)φ0 (y) d−1 d−1
Z
S= d xd y (4.28)
2 C ∆+ |x − y|2∆+
Inserting into the equation defining two-point correlator 4.25, we find:
21
∆+ 1
hO(x1 )O(x2 )i = − (4.29)
2C∆+ |x − y|2∆+
Comparison of the above relation with correlation function of a conformally invariant operator 2.33 shows
that the dual operator has dimension ∆+ . This confirms the results obtained for this model in previous section
4.3.1
4.5 Mass-Dimension Duality

There is a relation between mass of the bulk scalar field in AdSD+1 , and scaling dimension of the dual
operator as shown in 3.29.
r
D D2
∆± = ± + m2 l 2
2 4
That is due to the fact that mass and scaling dimension are different representation of the second Casimir
operator for SO(D, 2) symmetry group. Since the roots of a stable solution need to be real, positivity of the
square root imposes the Breitenlohner - Freedman bound : [48]
D2
m2 l2 > − (4.30)
4
This means that unlike the flat space, in anti-de Sitter space certain negative mass square is allowed. For
values of mass beyond this limit, scaling dimensions are complex, and modes become oscillatory resulting in
an unstable system. [37]
There are different mass-dimension conventions for other types of the field . Note however that in a
supersymmetric field theory like N = 4 SY M , it is enough to consider scalars. Since when the chiral
primary field is known other fields could be obtained by application of the supersymmetric generators. [16]
The conventions in AdSD+1 can be summerised as: [52]
1
spinors: ∆ = (D + 2m/l) , (4.31)
2
1
spin-3/2: ∆ = (D + 2m/l) , (4.32)
2
1 p
vectors: ∆± = D ± (D − 2)2 + 4m2 l2 (4.33)
2
4.6 Radial Direction-Energy Duality

So far it was discussed that xi coordinates in the bulk are mapped to the same coordinates xi on the boundary
of anti-de Sitter, but not clear how the radial coordinate r is encoded in the gauge theory formalism. The
answer as usual stems from symmetries of the system. Consider AdSd space in Poincare coordinates 3.19:
ds2 = dr2 + e2r/l ηµν (dxµ dxν ) µ, ν = 1, . . . d − 1
The bulk metric is invariant under the transformation

′
r → r + r′ xi → e−r /l xi ≡ axi
Due to identification of xi coordinates it yields xi → axi on the boundary as well. In the momentum
space this amount to rescaling of the energy E → E/a. It can be therefore inferred that radial coordinate in
the bulk is dual to energy scale of CFT on the boundary: [13]
22
E ↔ r ∼ 1/z (4.34)
The boundary of anti-de Sitter r → ∞ or z → 0 corresponds to UV limits of energy scales E → ∞ on the

boundary, whereas deep interior of the space-time r → −∞ or z → ∞ is identified with IR limits E → −∞
in the energy spectrum of the conformal filed.
4.7 Holographic Renormalisation

In general a quantum theory depends on a set of coupling constants as free parameters. Following a trans-
formation in energy scales of the system if the theory could be defined by a finite number of parameters
then it will be renormalisable. Renormalisation group transformation consists of course graining and scaling
and allows investigation of the system at different energy scales. Changes of the parameters with respect
to scaling is described by β-function which induces renormalisation group flow in the space of all theories.
Fixed point correspond to conformally invariant fields, and are of great importance as they determine stable
macroscopic state of the system. [51]
Consider the AdS space in Poincare coordinates 3.18:
l2
ds2 = dz 2 + ηµν dxµ dxν

2
ν, µ = 1, . . . d
z
This metric is of particular use when studying scaling behaviour of the system in the asymptotic region.
When the field φ(z, x) approaches the boundary z → ǫz with ǫ < 1, the solution scales like φ0 (x) → ǫ∆− φ0 (x).
As already discussed value of the bulk field at a certain point corresponds to the coupling of the dual operator.
So it can be seen that change of the coupling with respect to the scale is geometrically identified with changing
the distance between the bulk field and the boundary. As the field approaches the asymptotic regions of the
AdS space, the coupling runs to the UV limits. While when the field moves towards the interior of the space,
coupling flows into the IR limits.
The metric diverges as z → 0 which is the result of the fact that volume of the interior is infinite and
suffers from infra red divergences. As explained earlier small scales in the bulk corresponds to large scales
on the boundary, so this will lead to ultra violet divergence for the field theory. Due to existence of duality,
it turns out that renormalisation of one theory amounts to renormalising the other. [49]
Steps taken to regularise the gravity is similar to quantum field, at first a cutoff scale is introduced,
divergence terms are subtracted and then the limit is taken to infinity. Note that cutoff in the bulk has the
geometrical interpretation of truncating the AdS at r = rΛ which is related to a UV cutoff Λ on the boundary
QFT.
To clarify the above statements, consider again the example of a massive scalar field in the AdSD+1
background, it was shown that near the boundary generic solution behaves like: 3.30
lim φ(z, x) = z (D−∆) φ− (x) + z ∆ φ+ (x)

z→0
where D(∆ − D) = −m2 l2 . Depending on the mass of the field there are three different regimes: [39]
1. If −D2 /4 < m2 l2 < 0 then the scaling dimension is in the range D/2 < ∆ < D. It means conformal
dimension could be ∆ or D − ∆, and either way it is positive, so near the boundary as z → 0, both
modes vanish, and classical action 3.20 is finite. In this case solutions are normal modes, and both of
them are valid. Depending on the boundary condition, if one mode is the source the other will be VEV
of the field in deep interior.
The coupling constant dies out near the UV limit, and gravity and field theory decouple at high energies.
This describes deformation by a relevant operator so the solution dose not change the boundary metric
drastically, and the space will be asymptotically anti-de Sitter.
2. In case m2 l2 = 0 then one of the dimensions vanishes. The solution will be constant, and corresponds
to a marginal dual operator. The dual quantum field is conformally invariant living at a fixed point in
23
the RG flow. However quantum correction may affect the dimension, making the operator relevant or
irrelevant.
3. Finally if m2 l2 > 0 then ∆ > D. In this case one of the solutions φ− blows up when approaching
the boundary space. The classical action will be asymptotically infinite, corresponding to a non-
normalisable mode. This mode will not be a valid solution to the equations of motion, but it acts as
the external source for the dual quantum operator. Scaling dimension ∆ − D will be negative, which
means magnitude of the coupling increases with scale, and the theory has an IR fixed point. The dual
operator in this case will cause irrelevant deformation and the asymptotic space-time will be totally
disfigured.
Changes of the coupling with respect to the scale is described by beta function meaning that variation of
the solution determines the beta function β ∼ ∂φ0 /∂ ln ǫ ∼ ∆− φ0 [42]
Also anomalous dimension is defined as the changes of wave-function renormalisation with respect to the
scale γφ ∼ ∂ǫ∆ /∂ ln ǫ ∼ ∆. [42]
and r = 1/z plays the role of the cutoff Λc .
It can be seen that AdS/CF T duality provides a gravitational description for renormalisation group flow
in the conformal field theory: [41]
r ↔ Λc η ↔ µ (scale) β ↔ ∆− φ 0 ∆ ↔ γφ (4.35)
Finding an exact expression seems non-trivial as regularisation and renormalisation schemes in the bulk
are not known, and it is not likely that it follows one of the recognised schemes like dimensional regularisation.
[41] One of the generally accepted schemes is Hamiltonian formalism which is the same as usual formulation
with the radial coordinate playing the role of time. It turns out that φ+ and φ− are canonically conjugate
pairs, and Hamiltonian equations corresponds to renormalisation group equations. [37] We will use this
scheme in the final chapter of this paper.
4.7.1 Domain Wall

In this section we briefly review an example of RG flows typically dealt with in the AdS/CF T context.
Consider a single bulk scalar field in a four-dimensional flat background minimally interacting with gravity.
The action takes the form:
1 √
Z
−g −R − g µν ∂µ φ ∂ν φ + 2κ2 V (φ) d4 x

S= 2 (4.36)
2κ
The arbitrary potential V (φ) has one or more critical points (figure 4.1). The general solution to the
equation of motion is known as domain wall metric, and can be written as: [39]
ds2 = dr2 + e2A(r) ηµν dxµ dxν φ = φ(r) (4.37)
The coordinate are separated into a three-dimensional xi coordinate with Poincare symmetry, and a radial
r coordinate. Clearly AdS4 metric in Poincare coordinates fit into the domain wall by requiring A(r) = r/l.
the boundary region corresponds to r → ∞ limit, and deep interior is approached as r → −∞. It turns
out that local maximum of the potential corresponds to the boundary of AdS, while local minimum is the
deep interior. So solutions to the domain wall space-time interpolate between these two critical points of the
potential.[39].
Substituting the metric above back to the action, one can reach to the following set of equations of motion:
[37, 42] ′′
φ + 4A′ φ′ = ∂V /∂φ
2 2 (4.38)
A′ = 16 (φ′ − 2V )
Finding the exact solution to the above second order non-linear set of differential equations is non-trivial,
but still we can understand main physical properties by linearising about the critical points of the potential
24
Figure 4.1: A typical potential for a domain wall solution.
V . Setting φ(r, x) = Φ(r, x) + ϕ(r), and A′ (r) = 1/l + a′ (r), and working with the potential to quadratic
order, one gets a similar equation of motion to that of AdS4 : [42]
4 m2
ϕ′′ + ϕ′ − 2 ϕ = 0
l l
with the solution:
ϕ(r) = Be(∆−4)r/l + Ce−∆r/l (4.39)
with the same mass-dimension relation ∆(∆ − 4) = −m2 l2 .
Based on what discussed earlier at the local maximum, the bulk field approaches the boundary and
causes a relevant deformation of the dual QFT at an unstable UV fixed point. On the other hand at the
local minimum, as the field entres the deep interior of the AdS, on quantum side RG flows towards a fixed IR
point. So the RG flows from the boundary to the interior of the dual domain wall space, and since potential
at UV turns out to be higher than IR the flow is irreversible. [39]
In the last chapter of this paper, it will be demonstrated that how identifying QFT dual for cosmology
can be eased using the domain wall solutions. The application of holographic renormalisation group theory
and holographic Hamiltonian formalism will also be explained in more details.
4.8 Example of the Duality

In this section Maldacena’s conjecture which was the first realisation of AdS/CF T correspondence is re-
viewed. As will be explained, the duality is established through a comparison between Type IIB supergravity
propagating in an AdS5 × S 5 space-time, and N = 4 super-Yang Mill theory on parallel D3 branes. In the
first section, we introduce P -branes and consider supergravity viewpoint, then in the next section, the same
situation is described using the gauge fields living on Dp branes. Subsequently the precise expression for the
duality is defined, and finally limits of validity and different forms of the conjecture are discussed.
4.8.1 Supergravity and P-Branes

Before continuing with description of AdS5 × S 5 /N = 4, SY M duality, it is important to discuss about p-
branes in more details. P-branes are classical solutions of 10-dimensional supergravity which have non-trivial
25
charge under a p + 1 form. The simplest example is R-R charged spherically symmetric black holes in 9 − p
dimensions which are objects localised in p spatial dimensions. [15]
It turns out that there is an inequality between black-hole charge N and its mass M : [37]
N
M≥ (4.40)
(2π)p gs lsp+1
where gs and ls are string interaction and string length receptively. We are particularly interested in
extremal p-branes whose mass is at ground state for a given R-R charge. The metric of an extremal black-
hole is given by: [15]
−1 1
ds2 = Hp 2 (r) −dt2 + dxi dxi + Hp2 (r) dr2 + r2 dΩ28−p

i = 1, .., p (4.41)

7−p 7−p
where Hp (r) = 1 + r+ /r , and it is related to vacuum expectation value of dilaton φ via eφ =
7−p
gs Hp (r)(3−p)/4 . Moreover parameter r+ is related to the charge through r+ = dp gs N ls7−p where dp is a
5
constant. Note that event horizon is located at r = 0.
There is a curvature singularity at r = 0 but investigation of supergravity metric above shows that 3-brane
is a special case as:
1. when p 6= 3 the horizon and curvature singularity coincide which is referred to as null singularity. The
supergravity approximation is only valid at some distance from r = 0.
2. when p = 3 then the factor of r in the denominator of H(r) cancels out with r2 coefficient in the S 5
metric. So r = 0 is regular and we can use supergravity to explore the solution near this point.
There is always a solution to supergravity as long as Hp (r) is a harmonic function in 9 − p transverse

directions. So it is possible to generalise to the case where there are more than one p-brane which amounts
to multi-centred solution: [15]
N
X rp7−p 7−p
H(r) = 1 + 7−p , ri+ = dp gs Ni ls7−p (4.42)
i=1 |r − ri |
which
P indicates N parallel extremal p-branes located at the positions ri , each with R-R charge Ni such
that i Ni = N .
As mentioned in p = 3 case it is possible to take the limit to r → 0 smoothly, and then the metric will
behave like:
r2 r2
ds2 = −dt2 + dxi dxi + +2 dr2 + r2 dΩ25

2 (4.43)
r+ r
By a change of coordinates z = l2 /r where l is AdS length defined through:

14
l ≡ r+ = 4πgs α′2 N (4.44)
The metric reads:

l2
ds2 =−dt2 + dxi dxi + dz 2 + l2 dΩ25

(4.45)
z2
One can recognise the first term as the anti de Sitter metric AdS5 in Poincar coordinates. It means that
effect of gravity changes the space-time into a product manifold AdS5 × S 5 near the horizon. We will come
back to this point when explaining the correspondence later.
Throughout discussion provided here we used the classical version of string theory. This classical approx-
imation is only valid where space time curvature, given by AdS length l in this case, is much bigger than
string length ls that is ls ≪ l. Alternatively using 4.44 this condition can e rewritten as: gs N ≫ 1. [15]
1

7−p
5 This constant is given by dp = 25−p π 2 (5−p) Γ 2
[37]. Note for p = 3 in case of D3 branes: d3 = 4π.
26
4.8.2 String Theory and D-branes
In string theory a Dp-brane is a p + 1 dimensional hyper-surface in the space-time where open strings can
end [13].By calculating R-R charges and open string tensions, and matching with supergravity solutions one
can conclude that D-branes and extremal p-branes are actually the same objects [27]. So we can look at the
configuration described in the previous subsection from a deferent viewpoint: consider N D3 branes on top
of each other in a 10 dimension space-time.
1. Excitations of empty space-time generate closed strings which give rise to gravity super-multiplet. In
the low energy limit the theory is described by type IIB supergravity in the bulk. [14]
2. Excitations of D3-branes causes open strings ending on the branes. They support an N = 4 vector
multiplet on the world-volume of the brane, and the low energy limit will be N = 4 super-Yang-Mills
theory. 6 [14]
3. Interaction between closed and open strings which produce Hawking radiation.7 [47]
So the action will be composed of:
S = Sbulk + Sbrane + Sint (4.46)
We are interested in the low energy behavior of the system i.e. scales lower than string length 1/ls .
Alternatively one can keep the energy fixed and instead take ls → 0.
In the bulk supergravity theory Sbulk , consider infinitesimal perturbation of the metric around the flat
Minkowski g = η + κh. The action reads: [13]
1 √
Z Z
Sint = 2 −gRd4 x = (∂h)2 + κ(∂h)2 h + κ2 (∂h)2 h2 + . . . (4.47)
2κ
since α′ = ls2 , in the ls → 0 limit, universal Regge slope vanishes and so does coupling κ ∼ gs α′2 → 0. So
gravity turns into a free theory at long distances. [14]
As to the last term Sint in 4.46, interaction Lagrangian is obtained through perturbation in powers of
α′ :
Lint ≈ α′ R + (α′ )2 R2 + . . . (4.48)
So contribution from interaction Lagrangian coupling the bulk to the branes in the α′ → 0 limit vanishes.
In conclusion, in the low energy limit there are two decoupled systems: gravity in the bulk, and N = 4 super-
Yang Mills gauge theory on the world-volume of the branes. Presence of Dp-branes breaks the space-time
symmetry into stability subgroup on the branes and R-symmetry which describes rotations on the 5-sphere
S5.
It is important to investigate limits of validity of perturbative string theory approach discussed here.
When N D-branes connected by open strings concise each open string loop contributes a product of Chan-
Paton factor N , and string coupling gs . In order for the purterbative expansion to be valid, one needs to
work with small overall coupling constant which is the regime N gs ≪ 1. [14]
4.8.3 Maldacena’s Duality

Reviewing the results from point of view of supergravity and from string theory perspective, in both cases
there are two decoupled systems in the low energy limit. In both cases one of the mentioned decoupled
systems is gravity in the bulk space-time. So we can ignore supergravity part in this description, and then
we are left with the same physical situation described by two different theories:
1. N = 4 super-Yang-Mills theories on (3+1) dimensions with SU (N ) gauge group, and
2. Type IIB supergravity on the AdS5 × S 5 .
6 When all N branes are coincident, all states go massless, and gauge group is enhanced from U (1)N to SU (N ). [13]
7 In
string theory picture one can explain Hawking radiation as a quantum process. Two open strings on the branes interact
and couple like a closed string. The resultant closed string is not bound to the brain and can propagate in the empty space time
as Hawking radiation. [47]
27
It is therefore reasonable to identify these theories and state Maldacena’s Conjecture: [31]
Type IIB supergravity on AdS5 × S 5 ↔ N = 4 super-Yang-Mills in (3 + 1) dimensions
Note that symmetries on both sides are identical. In the gravity view-point there is a SO(4, 2) × SO(5)
symmetry being isometry groups of AdS5 space and S 5 sphere. This is isomorphic to the superstring theory
SU (2, 2) × SU (4)R symmetry corresponding to symmetry group of N = 4 super Yang-Mills and R-symmetry
in transverse dimensions.
Gauge fields on the world-volume of branes are generated by open strings ending on them, and it turns
out that open string coupling is proportional to super-Yang Mills coupling such that: [13]
gY2 M = 4πgs (4.49)
We will need the identity above as well as 4.44 to analyse the limits and different forms of AdS5 /N =
4, SY M correspondence.
4.8.4 Limits of Validity

As already explained earlier, the perturbative field description is valid as long as the coupling is small. For
super-Yang-Mills theory this means: [39]
4
l
gY2 M N = 4πgs N = ≪1 (4.50)
ls
,and classical supergravity is applicable when AdS length is much larger than string length:
4
l
= 4πgs N = gY2 M N ≫ 1 (4.51)
ls
It can be seen that two theories are reliable in different regimes. This is the reason why AdS5 /N = 4, SY M
is taken as a strong-weak duality. This means when gauge fields are strongly coupled the gravity theory has
weak interactions and vice versa.
Depending on the range of validity, three forms of correspondence are proposed: [37]
1. Weak form: This version is valid when gs N is large which is the example described in this section. In
this case supergravity approximation is enough to explain the duality.
√ N is large but gs → 0 so that λ = gs N is fixed. It

2. Semi-strong form: This is the ’t Hooft limit where
means α′ correction are applicable as α′ /l2 = 1/ gs N , but gs corrections may not be.
3. Strong form: This is the ideal duality which is conjectured to be true for all values of α′ and N . However
it is difficult to prove as the full gravity theory is not well-understood.
As discussed above string theory serves as a tool to provide concrete evidence for the gauge-gravity
duality. For example consider deformation of the gauge theory by an external source which changes the
coupling constant. This will in turn change the string coupling via 4.49. But string coupling is related to
expectation value of dilaton on the boundary through gs = ehφi . So finally boundary condition for the dilaton
will change. So it is natural to consider boundary value of dilaton as the source for the conformal field theory
living there.
Following the steps explained about abstract AdS/CFT correspondence in earlier parts of this chapter,
one can define a map that connects fields on the supergravity theory to the gauge invariant operators of
SYM, and hence a correspondence between correlators on both sides of duality.
28
Chapter 5
De Sitter Space
De Sitter has some special features like existence of different horizons for observables, two asymptotic bound-
aries connected by null geodesics or the fact that purely spatial planes can reach to far past null infinity. In
addition to mathematical interests, studying geometrical properties of the de Sitter space is the first step in
understanding of gauge/gravity duality in a space of positive curvature.
In the first section a review of de Sitter geometry is provided, and some geometrical properties are
also displayed for reader’s convenience and further reference. In the following section, different common
coordinates for de Sitter space are derived, and their applications are explained. Penrose diagrams are used
to illustrate the casual structure of the space-time, and problem of de Sitter horizons for observers is discussed.
Geodesics of a test particle are explained in the next part, and finally dynamics of the scalar matter field in
dS space is studied, and general solutions to the equation of motion is derived.
5.1 Geometry of de Sitter Space

Consider a Minkowski manifold of (d + 1) dimensions Md,1 with metric:
ds2 = −(dX 0 )2 + dX i dX i i = 1, ..., d (5.1)
The d-dimensional de Sitter Space dSd can be defined as an embedded hyperboloid of one sheet realised
via: [1]
−(X 0 )2 + X i X i = l2 i = 1, ..., d (5.2)
Parameter l has dimensions of length, and known as de Sitter radius the unique scale of the manifold.1
De Sitter has topology R × S d−1 , it is non-degenerate, and has Lorentzian signature. [1]
Lorentz group O(d, 1) is the isometry group of de Sitter space. Excluding reflection, the connected
isometry group will be proper Lorentz group SO(d, 1). [3]
It turns out that de Sitter is a vacuum solution of Einstein equation with a positive curvature, and thus
positive cosmological constant: [3]
1
Rµν − Rgµν + Λgµν = 0 Λ>0 (5.3)
2
It describes a universe of uniform positive energy density ρ = T00 = Λ, and negative pressure P = Tii =
−Λ for i = 1, ..., d − 1. In this manifold the Riemannian tensor is given by: [3, 56]
1
Rµνλσ = (gµλ gνσ − gµσ gνλ ) (5.4)
l2
Ricci tensor is proportional to the metric: [3]
d−1
Rµν = gµν (5.5)
l2
1 Note replacing l2 with −l2 , gives a hyperboloid of two-sheet where each of them is a d-dimensional hyperbolic manifold. In
this case metric is Riemannian i.e. non-degenerate and positive-definite. This definition is not used in this report.
29
(a)
(b)
Figure 5.1: (a)Penrose diagram for de Sitter space. (b) De Sitter space mapped to the region − π2 < T < π
2
of Einstein universe.
hence de Sitter space is a kind of Einstein Manifold. Ricci scalar is given by: [3]
d(d − 1)
R= (5.6)
l2
and finally cosmological constant will be: [3, 56]
(d − 1)(d − 2) d−2
Λ= = R (5.7)
2l2 2d
Especially in a 4-dimensional space-time we recover R = 12/l2 = 4Λ which is the result obtained in
section 3.1.
5.2 Coordinate Systems

De Sitter space admits a number of coordinate systems which are obtained through different foliation of the
manifold. So spatial part of the metric could be closed, open, flat etc in different coordinates. We discuss the
most common systems that are global, planar, and hyperbolic coordinates, and explain why they are more
frequently used. Some of these coordinates cover the whole manifold but some only partially cover it.
5.2.1 Global Coordinates

A convenient choice (t, θi ) is obtained by closed spherical slicing using the usual parameterisation of a
hyperboloid as follows: 0
X = l sinh(t/l),
(5.8)
X i = l cosh(t/l) xi i = 1, ..., d
where xi is the coordinates of a unit S d−1 sphere such that xi xi = 1, as discussed earlier in section ??.
The induced metric will be:
ds2 = −dt2 + l2 cosh2 (t/l) dΩ2d−1 (5.9)
This metric covers the entire hyperboloid. the spatial section is compact, and describes a S d−1 sphere
with a time-dependant radius of l2 cosh2 (t/l). The spherical metric can be writen as:
dΩ2d−1 = dθ2 + sin2 θdΩd−2 (5.10)
with −π < θ < π. The point θ = π is usually referred to as north poles, in contrast to the south pole
corresponding to θ = −π
30
Figure 5.2: Regions O− corresponding to the casual past of the observer on the south pole (right), and O+
the casual future of that observer.
In order to construct the Penrose diagram for de Sitter manifold we need to define conformal time
coordinate T via T = 2Arctg et/l − π2 which means dT 2 = dt2 / l2 cosh2 (t/l) . This amounts to the following
metric in conformal coordinates:
l2
ds2 = −dT 2 + dΩ2d−1

2
(5.11)
cos T
where conformal time is within the range T ∈ − π2 , π2 . The Penrose diagram which contains all the

information about casual structure of de Sitter space is shown in figure 5.1a. In this figure, upper and lower
horizontal edges are future and past null infinities I + , I − , and left and right vertical edges are North and
South poles. Light rays are ±45◦ lines, time-like curves are more vertical, and space-like curves are more
horizontal. Each point on the diagram represents a S d−2 sphere.
Identifying r′ = T , de Sitter space can be mapped to the region − π2 < t < π2 in Einstein static universe,
figure 5.1b. It can be seen that de Sitter space is a finite slab in Einstein universe.
One of the features of the de Sitter manifold is that no observer have access to the whole space-time. It
is in contrast with say Minkowski manifold where eventually one can have the whole information of history
in his casual past. In figure 5.2 region O− illustrates the whole casual past for the observer sitting on the
south pole, whereas region O+ shows the casual future of this observer. The intersection O− ∪ O+ is reffered
to as southern diamond, and it is the only region that south pole observer has full access to (figure 5.1a).
The lower diamond is where he can never send any information to, and upper diamond is the region where
in future no messages reaches the observer from. The northern diamond is not accessible to him at all.
Also in de Sitter space, different observers have different horizons. The observer sitting on the south pole
will never see anything beyond the diagonal line of O− ; however, diagonal line of O+ is the horizon for the
observer on the north pole.
5.2.2 Planar Coordinates

Planar coordinates (t, xi ) are obtained by flat slicing of the dSd with spatial Euclidean planes of infinite
volume. Parameterisation goes as follows: [3]
 0
 X = l sinh(tl) + (1/2l)xi xi e−t/l ,
X 1 = l cosh(t/l) − (1/2l)xi xi e−t/l , (5.12)
 i i −t/l
X =xe i = 2, . . . , d
where xi xi = 1, and xi gives the coordinates of a S d−2 sphere. then the metric reads:
ds2 = −dt2 + e−2t/l dxi dxi i = 1, . . . d − 1 (5.13)
where asymptotic past I − corresponds to t → −∞. This coordinate only covers half of the dS space
which is the casual past of the observable at the South pole O− (figure 5.3). One of the features of the de
Sitter space is that spatial planes can reach to the asymptotic past.
It should be noted that time t is not a Killing vector in this coordinates. Space-time isometries are
rotations and translations in xi coordinates. [3]
31
Figure 5.3: Dashed lines represent are spaces of constant t in planar coordinates. (reproduced from [56])
Introducing conformal time τ = et/l , one reaches to the metric:

1
ds2 = −l2 dτ 2 + dxi dxi

2
i = 1, . . . d − 1 (5.14)
τ
where asymptotic past I − is approached when τ → 0.
In order to use the above planar metrics for the North pole observable one needs to take t → −t, and
τ → −τ . [58]
5.2.3 Hyperbolic Coordinates

In Hyperbolic coordinates (τ, ξ, θi ) de sitter space is foliated by open spaces of negative curvature that are
hyperbolic planes. Open slicing is achieved using the following parameterisation: [55]
 0
 X = l sinh(t/l) cosh ξ,
X 1 = l cosh(t/l), (5.15)
 i
X = l z i sinh(t/l) sinh ξ i = 2, . . . , d
where z i z i = 1, and z i gives the coordinates of a S d−2 sphere. The induced metric reads: [55]
ds2 = −dt2 + l2 sinh2 (t/l) dHd−1

2
(5.16)
2
here dHd−1 describes the metric for a hyperbolic space.
5.3 Geodesics
When working in de Sitter space, there is a measure of distance between two points P (x, y) that is more
convenient than the geodesics distance D(x, y). It is defined via: [1]
l2 P (x, y) = ηij xi xj (5.17)
One can write an explicit expression for P in any coordinate introduced in this chapter. In planar
coordinates it reads:
1 ′
P (t, x; t′ , y) = cosh(t − t′ ) − e−(t+t ) |x − y|2 (5.18)
2
In the casual diamond this invariant measure is related to the geodesics distance by: [3]
P = cos(D/l) (5.19)
If P = 1 then geodesics distance vanishes D(x − y) = 0 so two points x and y either coincide or are
separated by a null geodesics.
If P = −1 then x = −y so antipodal point of x lies on the light cone of y.
one can generalise this argument and summarise that for P > 1 geodesics separation is time-like, and for
P < 1 it is space-like. For P < −1 geodesic separation between x snd anti-podal point is timelike. [1]
32
5.4 Matter Field in de Sitter Space
Consider the simple case of a single scalar field φ with mass m on a fixed dSd background with flat slicing
5.14. The action reads:
1 √ µν
Z
−g g ∂µ φ∂ν φ − m2 φ2 dd−1 xdτ

S= (5.20)
2
Klein-Gorden equations of motion − m2 φ = 0 is obtained through variation of action where denotes

Laplacian with usual definition in 3.22. We follow similar steps as in anti-de Sitter case to find the solutions.
Separation of variables φ(τ, x) = ϕ(τ )Φ(x) amounts to two independent equations of motion in τ and x
coordinates: [55]
(∂x2 + k 2 )Φ(x) = 0 (5.21a)

d d 2 2 2 2

−τ ∂τ (τ ∂τ ) + m l + k τ ϕ(τ ) = 0 (5.21b)
with k 2 being modulus square of momentum, and ∂x2 indicating Laplacian on Euclidean manifold. Solution
to Euclidean Laplacian 5.21a are plane waves Φ(x) = eikx /(2π)d−1 , so the full solution to the equations of
motion is obtained through superposition:
dd−1 k
Z
φ(τ, x) = ϕ̃k eikx (5.22)
(2π)d−1
This shows that ϕ̃k is the solution in the momentum space. Writing the second equation of motion 5.21b
in Fourier space ends up with the Bessel equation of the form: [55]
(d − 1)2

2 ′′ ′ 2 2 2 2
(kτ ) ψ̃k + (kτ )ψ̃k − − m l + k τ ψ̃k = 0
4
where ψ̃k = τ (1−d)/2 ϕ̃k . The solution is a linear combination of first and second kind Bessel functions Jσ ,
and Yσ :
ψ̃k (kτ ) = ak Jσ (kτ ) + bk Yσ (kτ )
p
where in dS space σ = (d − 1)2 /4 − m2 l2 . Solution in terms of ϕ̃k can be written as: [58]
ϕ̃k (τ ) = ak (kτ )(d−1)/2 Jσ (kτ ) + bk (kτ )(d−1)/2 Yσ (kτ ) (5.23)
When approaching the origin, Bessel functions have the following behaviour:
τ σ σ
1 Γ(σ) 2
lim Jσ = lim Yσ = − (5.24)
τ →0 Γ(σ + 1) 2 τ →0 π τ
So near the boundary I− when τ → 0− , general form of the solution in momentum space can be written
as:
ϕ̃k (τ ) ≈ τ ∆− φ̃− (k) + τ ∆+ φ̃+ (k)
where mass-dimension relation in dSd is found to be:
r
(d − 1) (d − 1)2
∆± = ± − m2 l 2 (5.25)
2 4
In the real space, solution to the equations of motion when approaching past null infinity I − , in the
conformal planar coordinate, will be written as:
lim φ(τ, x) = τ ∆− φ− (x) + τ ∆+ φ+ (x) (5.26)

τ →0−
33
Chapter 6
dS-CFT Duality
So far in this report, all the discussions about the gauge/gravity duality focused on the spaces of negative
curvature; however, based on observational evidence, we know that universe does not have a negative cosmo-
logical constant. This is the main motivation to generalise the correspondence to a de Sitter space. A primary
difficulty with the dS/CF T duality is the absence of any example of a de Sitter solution in the context of
string theory; in spite of this, as will be explained, it is still possible to conjecture the duality along the lines
of well-constructed AdS/CF T correspondence. At the end we review a recent example proposing that the
actual gravitational theory is the higher spin formalism corresponding to a Sp(N )-invariant field theory.
In the first section asymptotic symmetries of de Sitter space are compared to those of conformal group
as the initial motivation for the duality. After defining the green functions in the bulk, correlation functions
which are the main observables of the theory are computed. An example of the duality is presented in the
last section, and correspondence between four-dimensional theory of higher spin gravity in de Sitter dS4 , and
conformal fields in three dimension CF T3 is explained.
6.1 Symmetries
It was discussed in section 2.2.4 that conformal group on a D-dimensional Euclidean space is Lorentz
group SO(D, 1) which is identified with the isometry group of the de Sitter space. This is basically the
main motivation to establish a one-to-one mapping between gravity in the bulk dS, and field theory on the
Euclidean boundary. In a theory of quantum field, gauge symmetries are diffeomorphism invariant, so it is
sufficient to find diffeomorphisms in de Sitter which can preserve the boundary conditions.
In the next chapter we consider the dS3 space related to the two-dimensional CF T , and at the end
generalise the results.
6.1.1 Asymptotic Symmetries

Recall metric for an dS3 5.13 in planar coordinates. We need to define boundary conditions that metric
should satisfy in order to be asymptotically de Sitter . The metric of an asymptotically past de Sitter space
will behave for t → −∞ as: [56] 1
gzz̄ = 12 e−2t + O(1)



gtt = −1 + O(e2t )

(6.1)

 gzz = O(1)
2t
gtz = O(e )

and all other metric components vanish. Asymptotic symmetries are diffeomorphisms that satisfy the
above boundary conditions. Changes in components of a metric δgµν under an active coordinate transforma-
tion ζV is given by the Lie derivative with respect to the vector field LV : [1]
δζ g = −LV g (6.2)
1 This asymptotic behavior can also obtained from analytic continuation of dS3 boundary conditions [45]
34
where Lie derivative of a metric is given by [1]
LV g = LV gµν dxµ ⊗ dxν + gµν (LV dxµ ) ⊗ dxν + gµν dxµ ⊗ (LV dxν ) (6.3)
So it is required to parametrise the diffeomorphism ζV in terms of the vector field. One can check that
the most general diffeomorphisms will be: [56]
1
ζ = V ∂z + V ′ ∂t + O(e2t ) + cc (6.4)
2
where prime denotes differentiation. Using definition 6.2 and 6.3, it is inferred that under this coordinate
transformation the metric changes like: [56]
2
δζ gzz = − l2 V ′′′
(6.5)
δζ gzz̄ = δζ gtz = δζ gtt = 0
It is not hard to see that these changes preserve the boundary conditions 6.1, and thus diffeomorphism
6.4 generates the asymptotic symmetry of dS3 . Now we need to identify this symmetry group generated
by ζ. In section 2.1 it was explained that a conformal transformation is composed of a Weyl scaling and
a diffeomorphism. Consider 21 (V ′ ∂t + V¯′ ∂t ) term in 6.4, it creates a time translation which is equal to
a Weyl scaling on the boundary complex plane. The other term i.e. V ∂z + V̄ ∂z̄ is simply a coordinate
transformation. This means the diffeomorphism ζ is a direct sum of a tangential term acting as a coordinate
transformation, and a normal term which functions as a Weyl scaling on the complex plane. Therefore in
conclusion, asymptotic symmetry group of de Sitter space is the conformal group on the boundary. [58]
In the special case where:
V = α + βz + γz 2 (6.6)
the V ′′′ term in 6.5 vanishes, so metric is invariant under the diffeomorphism generated by V that is
LV g = 0. In this case Killing vector V generates the isometries of dS3 . Note there are three complex free
parameters α, β and γ which give rise to SL(2, C) symmetry group. It is inferred that isometry group of dS3
is SL(2, C) which is a subgroup of asymptotic conformal group symmetry.
One can generalise the above argument to a d-dimensional de Sitter dSd , and conclude that related
isometry is SO(d, 1).
6.2 The Field - Operator Correspondence

There is an extension of the already discussed AdS/CF T theory to dS space. The conjecture is that corre-
lation functions of the fields in the de Sitter space are in a one-to-one correspondence with the corelators of
the Euclidean conformal field on the boundary.
hO(x1 )...O(xn )i ↔ hφ(t, x1 )...φ(t, xn )i (6.7)
This means bulk theory of gravity, and spatial conformal field theory are in fact two different representa-
tions of the same fundamental theory. Moreover it is conjectured that de Sitter space fields are in one -to-one
correspondence with single trace operators of the conformal theory.
When the field approaches the boundary of dS at either I + , or I − , it acts as the external source for
operators of the QFT. Partition function of the gravity theory is identified with generating functional of the
quantum field. Similar to AdS case, correlation functions are obtained by taking functional derivatives of the
bulk action with respect to the source.
6.3 Mass-Dimension Duality

In this part of the report we have a closer look at the mass-dimension duality relation 5.25 for a scalar field
in dSD+1 : r
D D2
∆± = ± − m2 l 2
2 4
35
1. In case 0 < m2 l2 < (D/2)4 , both conformal dimensions are real and positive 0 < ∆− < D/2 < ∆+ < d.
In this case both solutions vanish when approaching the boundary, so both are normalisable modes.
There is a freedom to choose dimension of the dual operator depending on the boundary conditions. If
boundary condition imposes the weight ∆+ (∆− ) for the operator then the source will be φ− (φ+ ) and
vacuum expectation value will be given by φ+ (φ+ ).
2. On the other hand if m2 l2 > (D/2)2 , then scaling dimensions are not real any more. Modes will be
oscillatory which means there is no stable scalar field on the de Sitter background. Complex dimensions
indicate that dual field theory is not unitary. Non-unitarity dose not rule out the gauge-gravity duality
as the field theory lives on a Euclidean space. In other words there is no reason that CFT must be
unitary. If it is, then according to AdS/CFT correspondence, the same field theory should be dual to
an AdS gravity which is puzzling.
6.4 Time Evolution

Recall a d-dimensional de Sitter space dSd is defined as a purely spatial submanifold of a bulk space-time of
(d, 1) dimensions. So in dS/CFT correspondence time itself is emerging from conformal field theory compared
to AdS/CFT duality where radial direction is emergent. So notion of time in de Sitter version of duality
challenges our quantum intuitions of unitary time evolution.
Consider the dSd metric in planar coordinates 5.13:
ds2 = −dt2 + e2t/l dxi dxi
The metric is invariant under following transformation:
t→t+τ xi → e−τ /l xi (6.8)
This transformation generates time evolution in first term, and scaling in the second term. This means
that time evolution in the bulk theory corresponds to the scaling transformation on the boundary QFT.
Physics at infinity is intriguing since infinite time in the gravity theory means infinite divergence in the
field theory. So introducing a cut-off in the energy scale of the quantum field stops time from extending
to infinity which is the starting point for holographic renormalisation formalism which discussed in earlier
chapters. From Wilsonian point of view this relation indicates that renormalisation group flow in CFT is
related to time-dependence of supergravity theory. Therefore infinite future in the de Sitter corresponds to
UV divergences of the dual CFT, while infinite past is dual to IR in the boundary field.
6.5 Green Function

Wightman green function as usual is defined via:
− m2 G(x − y) = δ (d) (x − y)

(6.9)
where is Laplacian in dSd . It is convenient to use translation symmetry of de Sitter to fix one point at
x = x0 . then we can solve the Laplace equation above away from this point, and at the end check that the
solution behaves like a delta function when these two points coincide. In this situation green function only
depends on a single parameter that is dS invariant length P already defined in section 5.3 so G(P ) ≡ G(x−y).
Laplace equation simplifies to: [58]
− m2 G(P ) = 0

(6.10)
Using the definition of Laplacian 3.22, one can compute it in dSd space as: [1]
x2
(1 − x2 )∂x2 − d x∂x

x = 2
(6.11)
l
where d is simply dimension. Then equation 6.10 simplifies to a one-variable ODE: [56]
(1 − P 2 )∂P2 G − d P ∂P G − m2 l2 G = 0 (6.12)
36
A change of variables z = 12 (P + 1) amounts to so-called hypergeometric equation: [23]
1
z(1 − z)G′′ + d( − z)G′ − m2 l2 G = 0 (6.13)
2
with hypergeometric function solution of the form: [56]

d P +1
G(P ) = C∆ F ∆+ , ∆− , , (6.14)
2 2
Γ(∆+ )Γ(∆− )
where the constant C∆ = (4π)d/2 Γ(d/2)
as shown in D.3, ∆± are given by:
1h p i
∆± = (d − 1) ± (d − 1)2 − 4m2 l2 (6.15)
2
and hypergeometric function is defined as: [25]
∞
X (a)n (b)n n
F (a, b, c, z) = z (6.16)
n=0
(c)n n!
here (x)n ≡ x(x + 1) . . . (x + n − 1) = Γ(x + n)/Γ(x). Now we need to examine the behavior of the
propagator G(x − y) when two points coincide. The hypergeometric function has a singularity at z = 1
[23], or equivalently P = 1 which yields D(x, y) = 0. So singularity occurs when two points coincide or are
separated by a null geodesics. Near the singularity hypergeometric function scales like (D/l)2−d [23] which
after action of Laplacian operator gives the right behavior of a delta function. Therefore the green function
G(x − y) satisfies all the requirements of Laplace equation 6.9.
Equation 6.12 is symmetric under the change P → −P , consequently G(−P ) is also another independent
solution:
d −P + 1
G(P ) = C∆ F ∆+ , ∆− , , (6.17)
2 2
Singularity of the above green function is at P = −1. This happens when first point x is located on the
light-cone of its antipode −x. 2 The green function for a de Sitter space therefore is a linear combination of
the above mentioned solutions G(P ) and G(−P ).

As discussed earlier, in the de Sitter space dSd near the boundary as t → −∞, the solution behaives like
φ ≈ e∆± t , and there is freedom in choosing either of these two boundary conditions. From this point forward,
we impose boundary condition:
lim φ(t, x) = e∆− t φ− (x) (6.18)
t→−∞
and φ− (x) functions as the source on the boundary.

All observables of the system can be constructed from a complete set of correlation functions. The simplest
case two-point correlation function is defined by taking the derivative of the bulk action two times:

δ δ
hO(x)O(y)i = Sgravity (6.19)
δφ− (x) δφ− (y) φ− =0
So only quadratic terms in the Lagrangian are required, and free theory formalism suffices for this purpose:
1 √ µν
Z
−g g ∂µ φ∂ν φ − m2 φ2 dt dd−1 x

S= (6.20)
2
2 This singularity is still physical because antipodal points in de Sitter space is always hidden beyond the horizon (section
5.2). So an observer cannot detect this singularity.
37
Following similar steps to AdS space explained in appendix E in details, the action in planar coordinates
can be written as:
↔ ↔
Z
′
′
lim S = − e(d−1)(t+t )/l φ(t, x)∂t G(t, x; t′ , y)∂t′ φ(t′ , y)dd−1 ydd−1 x (6.21)
t,t →−∞ I−
Near the past null boundary I− , the action becomes (refer to appendix E):
φ− (x)φ− (y) d−1 d−1

Z
S≈ d xd y (6.22)
|x − y|2∆+
Using the definition of two-point function in dSd it can be seen that tow-point function is the quadratic
coefficient in the action above: (appendix E)
∆+ 1
hO(x)O(y)i = (6.23)
2C∆+ |x − y|2∆+
which confirms the assumption that dual operator has conformal dimension ∆+ .
As discussed earlier in de Sitter space-time the bulk field can also asymptotically behave as φ(t, x) ≈
e∆+ t φ− (x). In this case correlation function will be:
∆− 1
hO(x)O(y)i = (6.24)
2C∆− |x − y|2∆−
which corresponds to a dual operator with scaling dimension ∆− . In fact this boundary condition can
only be imposed on the future infinity I + which is not within the casual past region O− . [56]
6.7 Example of dS-CFT Duality

Over a decade after introduction of dS/CFT duality by Strominger, there was no example of the correspon-
dence at the fundamental level.
It was not also possible to analytically connect the results of AdS/CF T correspondence to dS space due
to encountering complexification issues about the rank of the gauge group. Recalling the Maldacena’s duality
in AdS4 where it was shown that N 2 ∼ l2 /GN with l2 ∼ 1/Λ; one can attempt to analytically continue to dS4
by reversing the sign of cosmological constant Λ → −Λ while keeping the Newton’s constant unchanged. But
this yields N 2 → −N 2 meaning that rank of the gauge group is imaginary N → iN . This raises issues about
imaginary scaling weights, imaginary fluxes, ghost fields and negative energies that are not well-defined. [61]
In this section we explain the recent hypothesis that higher spin gravity in dS4 is actually dual to an anti-
commuting, Sp(N )-invariant conformal field in three dimensions. This conjecture rests on well-established
Giombi-Klebanov-Polyakov-Yin (GKPY) duality which relates HS gravity on AdS4 to O(N )-invariant CF T3 .
In GKPY duality conformal operators transform in the fundamental representation leading to the special
feature N ∼ 1/(ΛGN ), so one can continue to the dS4 space simply by N → −N which yields appearance of
symplectic group O(−N ) = Sp(N ). 3 [69]
The results of GKPY duality are used to derive HS dS4 /Sp(N ) CF T3 duality, and the methodology is
illustrated in figure 6.1. At first step we take an analytic continuation to move from higher spin theory in
AdS4 to EAdS4 . Then results of GKPY duality is used to map the HS gravity observables to correlators of
the dual O(N )-invariant conformal field. It will be discussed that by reversing the sign of the fundamental
index N one can switch to the fermionic Sp(N )-invariant field theory. At the end we are able to establish
the connection between HS dS4 and Sp(N ) CF T3 models.
38
Figure 6.1: Schematic diagram for methodology used to establish duality between HS dS4 and Sp(N ) CF T3
models.
6.7.1 Higher-Spin Gravity Theory

Higher spin gravity theory is a generalisation of general relativity which contains an infinite number of bosonic
fields with all non-negative integer spins. 4 HS theory comes in two types: Type A theory has a minimal
bosonic spectrum of gauge fields for each even spin whilst type B theory contains odd spin fields. In this
paper, we are interested in type A higher spin gravity in a background with constant curvature. [62]
To construct the algebra one needs to define star product of polynomials f and g:
Z
f (x, z) ∗ g(x, z) = eu.v f (x + u, z + u)g(x + v, z − v)d2 u d2 v (6.25)
which has the properties: [63]
xα ∗ xβ = xα xβ + ǫαβ , x̄α̇ ∗ x̄β̇ = x̄α̇ x̄β̇ + ǫα̇β̇ (6.26)
where ǫαβ is totally antisymmetric tensor with ǫ12 = 1. Then the Lie algebra of d-dimensional higher spin
theory hs(d) is defined with respect to the Lie bracket:
[f, g] = f ∗ g − g ∗ f (6.27)
with reality condition f † = −f , and projection condition Π f (x) ≡ f (ix) = −f . [63]

It is common to pack all of the gauge fields Aµ into a master field defined as:
Aµ (x, y, ȳ) = ǫµα1 α˙1 (x) yα1 ȳα˙1 + ωµα1 α2 (x) yα1 yα2 + ωµα̇1 α̇2 (x) ȳα̇1 ȳα̇2
X ∞
+ ... + Aαµ
1 ...αm α̇1 ...α̇n
(x) yα(1 . . . yαm ȳα̇1 . . . ȳα̇n) (6.28)
n,m=0
Imposing the constraint that only even spins are independent which gives rise to a tower of real fields of
even spins which describe all degrees of freedom for the theory. 5 Similarly one can package scalars φ, and
Weyle tensors φα1 ...αn into a scalar master field:
∞
X
Φ(x, y, ȳ) = φ(x) + φαα̇ (x) y α1 ȳ α˙1 + . . . + φα1 ...αm α̇1 ...α̇n (x) y α(1 . . . y αm ȳ α̇1 . . . ȳ α̇n) (6.29)
n,m=0
with constraint φαα̇ = ∂αα̇ φ, . . . , φα1 ...αm α̇1 ...α̇n = ∂α1 ...αr α̇1 ...α̇s φαr ...αm α̇s ...α̇n .
3 The bar in analytic continuation O(−N ) = Sp(N ) means Young tableux are transposed. This changes the symmetric
relations of bosons to anti-symmetric fermionic relations.

4 The spectrum obviously includes a scalar of spin zero.
5 In even spin reality condition reduces to the fact that component fields are real fields. [66]
39
6
In classical higher spin gravity the free action in d dimensions can be written as: [64]
1
Z
Sf ree ≈ (∂Φ)2 + ΛΦ2 dd x (6.30)
GN
where GN being the Newton’s constant, and Λ cosmological constant. The linearised equation of motion
is found to be:
∂ 2 Φ + ΛΦ = 0 (6.31)
It turns out that interaction couplings are inverse powers of cosmological constant, so interacting action
will be an expansion in ∂ 2 /Λ. For instance for cubic interactions between spins (2, s, s) the action reads: [64]
1 1
Z
S ≈ Sf ree + ∂ 2 hΦ2 + ΛhΦ2 + ∂ 2 h(∂Φ)2 + h(∂Φ)2 + . . . (6.32)
GN Λ
where h is the spin 2 metric fluctuation, and cubic interactions can be written as:
1 3
∂ 2 h ≈ ΛΦ2 + Φ∂ 2 Φ + (∂Φ)2 + [∂ Φ∂Φ + (∂ 2 Φ)2 ] + . . . (6.33)
Λ
So quantum corrections take the form of an overall factor of 1/(GN Λ) for the action. [65]
6.7.2 EAdS - dS Correspondence

In this section we establish the relation between Euclidean anti de Sitter EAdSd and de Sitter dSd spaces in
d-dimensions. Consider the Poincare patch coordinates in EAdS manifold:
2
lEAdS
ds2 = dz 2 + dxi dxi

z 2
This can be transformed into dS metric in the conformal planar coordinates 5.14:
2
ldS
ds2 = −dτ 2 + dxi dxi

τ 2
by a double wick rotation: z → −iτ , and lEAdS → −ildS . Note the last map is the familiar relation
between cosmological constants Λ → −Λ.
Having established the continuation between two types of manifold, we need to find the relation between
correlation functions. It was already discussed that in HS theory of gravity quantum corrections appear as
an overall factor of 1/(GN Λ) for the action. It turns out that this factor is proportional to the dimensionless
coefficient of two-point functions C2 ∼ −GN Λ. 7 Consider bulk-to-bulk propagator in anti de Sitter space
4.14. At short distances free anti de Sitter correlator behaves like 4.16: [66]
z1 z2
hφ(z1 , x1 )φ(z2 , x2 )i ≈ C2 (6.34)
(z1 − z2 )2+ (x1 − x2 )2
If only continuation z → −iτ is applied, we end up with a wrong overall minus sign. So in order to
continue EAdS to dS it is required to take C2 → −C2 . It can also be though of as taking Λ → −Λ while
keeping the Newton’s constant GN fixed.
This can be generalised to find the transformation rule for n-point functions of master fields between
EADSd , and dSd : [61]

hΦ(z1 , x1 ) . . . Φ(zn , xn )iEAdS = hΦ(−iτ1 , x1 ) . . . Φ(−iτn , xn )idS (6.35)
C2 →−C2
6 In fact action formalism is not known for higher spin theories. In order to derive the schematic form of the action it is
assumed in this paper, that equations of motion follow integrability conditions. [61]
7 Coefficient of two-point function in this section is denoted as C ≡ C .
2 ∆
40
6.7.3 GKPY Duality
Consider the O(N ) vector model consists of N -component conformal fields ϕa transforming in the fundamen-
tal representation. The theory has O(N ) symmetry with N being an even integer.
Free Theory
The free action in a 3-dimensional flat space can be written as:
1
Z
∂ i ϕ a ∂ i ϕ a + m2 ϕ a ϕ a d 3 x

S= (6.36)
2
The only class of singlet operators in this theory corresponds to the conserved currents of the form: [69]
(s)
Ji1 ...is = ϕa ∂(i1 . . . ∂is ) ϕa + . . . (6.37)
So that there is a different conserved current for each spin. GKPY duality conjectures that there is a one-
to-one mapping between these single trace operators, and the type A higher spin fields in AdS4 with Neumann
boundary condition. Accordingly connected generating functional of the boundary CF T3 in fundamental
representation of O(N ) are given by on-shell action of the HS gravity in the bulk AdS4 . There is therefore a
one-to-one correspondence between correlation functions on both sides.
Quantum field prescribes that correlation functions in the free theory are one-loop diagrams with the
operators running around the loop. Explicit calculations show that there is an overall factor of N , for
instance two-point correlators for conserved currents are given by: [66]
N
hJ (s) (x1 )J (s) (x2 )i ∼ (6.38)
|x1 − x2 |2
or three-point functions are:

N
hJ (s) (x1 )J (s) (x2 )J (s) (x3 )i ∼ (6.39)
x12 x13 x23
where xij ≡ |xi − xj |2 . One can compare the coefficient N in the above relations with that of a generic
conformally invariant correlator in section 2.5 to yield C2 ∼ 1/2N . Comparison with the relation obtained
on the gravity side C2 ∼ GN Λ yields the following identity in GKPY duality:
1
N∼ (6.40)
ΛGN
Interacting Theory
The action for an O(N ) invariant interacting theory is given by:
Z
1 a i a 1 2 a a λ
S= ∂i ϕ ∂ ϕ + m ϕ ϕ + (ϕ ϕ ) d3 x
a a 2
(6.41)
2 2 2N
where λ is quadric coupling. This theory is known to have a critical IR fixed point. [41] The usual practice
is to introduce an auxiliary field σ(x), then the action takes the form:
Z
1 a i a 1 2 a a a a N 2 3
S= ∂i ϕ ∂ ϕ + m ϕ ϕ + σϕ ϕ − σ d x (6.42)
2 2 2λ
The action is quadratic in ϕ, and it can be shown that the introduced auxiliary field removes all the
one-particle reducible diagrams in terms of σ(x). [69] 8
According to GKPY duality prescription the critical interacting O(N ) CFT in three dimensions corre-
sponds to the HS gravity theory in AdS4 with Dirichlet boundary conditions.
8 We will make use of this method in the next sections to find the analytic continuation between O(N ) and Sp(N ) models.
41
At the end it is essential to explain the difference between interacting fields in the fundamental and adjoint
representations. In the adjoint theory single-trace operators have the general form of tr (ϕ̂∂i1 ϕ̂ . . . ∂in ϕ̂) with
ϕ̂ being an N × N matrix charged under the adjoint. The number of these operators exponentially increases
with dimension. Higher spin theories do not contain sufficient number of fields to correspond to the dual
operators. In contrast to this, in the fundamental representation single trace operators take the generic form
of ϕa ∂i ϕa where ϕ is an N -component vector. The operators do not grow in number with respect to the
dimension, so it is possible to map them to the fields in an HS gravity theory. [69]
6.7.4 O(N) - Sp(N) Correspondence

In order to find the gauge theory dual to higher spin gravity, in de Sitter space, we need to define a new
quantum field. The conformal field theory of interest studied in this section is usually referred to as Sp(N)
vector model with N being an even integer. It consists of anticommuting scalar field transforming under
fundamental representation of Sp(N ), and thus can be represented as N-component vectors χa . It can be
shown that the only single particle state of the theory corresponds to single-trace operator Ωab χa ∂i χb where
Ω is the antisymmetric symplectic form.
Free Sp(N) vector model

It is more convenient to discuss the free conformal field theory at first step and then generalise the results
to the interacting case. The free Sp(N ) vector model can be described by a non-Dirac, two-derivative form
action, and in the free theory it reads: [67]
!
1 0 1N ×N
Z
a i b 2
D
S= Ωab ∂i χ ∂ χ + m χ.χ d x, Ωab = 2 2 (6.43)
8π −1 N × N 0
2 2 N ×N
and inner product is defined as χ.χ ≡ Ωab χa χb . As usual, one can work out the conserved current by
varying the Lagrangian:
Ji1 ...is = Ωab χa ∂(i1 . . . ∂is ) χb + . . . (6.44)
where s denotes all even spins. Using equation of motion ∂i ∂ i χ = 0 it is straight forward to check that
i1
∂ J(i1 . . . ∂is ) = 0. In the free theory current has conformal dimension ∆ = D − 2 + s. [66]
Similar to O(N ) vector model, correlation functions in the free theory are one loop diagrams with operator
circulating around the loop. However due to Fermi statistics there is an extra minus sign. Again analogous
to O(N ) model, there is an overall factor of N in the explicit expression of correlators (c.f. 6.38, and 6.39).
So by taking N → −N one can relate Sp(N ) invariant free correlators to O(N ) free functions: [61]
hJ (s1 ) . . . J (sn ) iSp(N ) = −hJ (s1 ) . . . J (sn ) iO(N ) (6.45)
Interacting Sp(N) vector model

Action for the Sp(N ) invariant interactive model can be written as: [68]
1
Z
Ωab ∂i χa ∂ i χb + m2 χ.χ + λ(χ.χ)2 dD x

S= (6.46)
8π
where λ is quadric coupling constant. The interacting theory has a critical IR fixed point. Similar to
O(N ) case one can introduce an auxiliary field σ(x) so that the action changes to:[68]
1
Z n √ o
S= Ωab ∂i χa ∂ i χb + m2 χ.χ + 2α λ(χ.χ) − α2 dD x (6.47)
8π
Consequently the Feynman diagrams reduces to closed loops of χ field linked by propagators of α. It can
be shown that analogous to O(N ) case there is a symmetry factor of N for each loop. [61] Since χ fields are
anti-commuting there is minus sign contribution form each loop. So the correlators in Sp(N ) vector model
are related to those of SO(N ) through the continuation N → −N . [67]
42
hJ (s1 ) . . . J (sn ) iSp(N ) = hJ (s1 ) . . . J (sn ) iO(N ) N →−N

(6.48)
It is worth mentioning that Sp(N ) scalar field is not unitary. First of all this does not affect the duality
as CFT lives on a Euclidean space. In addition Sp(N ) model turns out to be pseudo unitary. It means that
Hamiltonian is pseudo hermitian such that H † = CHC where C is a unitary operator and C 2 = 1. It can be
demonstrated that such a pseudo-hermitian Hamiltonian satisfy desirable properties of a unitary quantum
theory, especially that probabilities are conserved and eigenvalues are real. [70] 9
6.7.5 dS4 / Sp(N) CFT3 Duality

At this stage we have all the ingredients in place to express the first example of dS/CF T duality. It was
explained that GKPY duality prescribes that higher spin gravity in AdS4 with Dirichlet (Neumann) boundary
conditions is dual to interacting (free) O(N ) conformal field theory with identification N ∼ 1/ΛGN . EAdS
space can be analytically continued to dS via Λ → −Λ, and O(N ) vector model is related to Sp(N ) through
N → −N . So it can be conjectured that higher spin gravity in AdS4 with Dirichlet boundary conditions
corresponds to critical interacting Sp(N ) invariant conformal field in three dimensional Euclidean boundary,
and HS theory with Neumann conditions is dual to the free conformally invariant Sp(N ) model in three
dimensions. [61]
In order to show the statement explicitly, consider the scalar field φ(z, x) in the EAdS4 space. Near the
boundary zc → 0 where zc is cutoff scale. Using simple dimensional analysis and imposing O(4, 1) symmetry
constraints one can generalise the expression for the propagator 6.34 to n-point correlation functions as: [66]
z n∆+
(n) c
hφ(z1 , x1 ) . . . φ(zn , xn )iEAdS = AC2 (C2 )n (6.49)
x
(n) 10
where AC2 is a constant, and x is a function of distance between point which scales linearly. One can
always define an operator such that:
1
O= ∆
φ
C 2 zc +
As a result correlator for quantum operators is found to be: [61]
(n)
hO(x1 ) . . . O(xn )iEAdS = AC2 x−n∆+ (6.50)
So in this way any operator is in a one-to-one correspondence with a dual O(N )-invariant field which
is a statement of GKPY duality. In order to generalise the correspondence to dS space, we need to take
the analytical continuation. Using duality relation 6.40 correlation function for the bulk field in dS can be
written as: n∆+
(n) −iτc
hφ(τc , x1 ) . . . φ(τc , xn )idS = A−C2 (−C2 )n . (6.51)
x
Similarly operators of corresponding Sp(N )-invariant QFT can be defined as:

1
O= φ
−C2 (−iτc )∆+
So boundary conformal field correlation functions then will be give by:[61]

(n)
hO(x1 ) . . . O(xn )idS = A−C2 x−n∆+ (6.52)
9 In this quantum theory inner product is defined as hψ ′ |ψiC ≡ hψ ′ |C|ψi. [70]
10 This notation is chosen for convenience. As an example for a two-point function x2∆ = x2∆ 2∆ .
12 ≡ |x1 − x2 |
43
It can be seen that there is a one-to-one map between the correlators in the bulk HS gravity dS4 and
correlation functions on the three-dimensional conformal theory:
hO(x1 ) . . . O(xn )iCF T3 ↔ hφ(τc , x1 ) . . . φ(τc , xn )idS4 (6.53)
44
Chapter 7
Cosmology
Inflation was a period of exponential expansion in a de Sitter background during the early universe before
the hot big bang led into the radiation era. Inflation theory was successful in solving the old problems
about flatness, horizon, and monopoles, and could predict large structure formation based on inhomogeneous
quantum fluctuations. However inflation is not a complete theory and is unable to give any information about
the initial values of observables. In this section we study the application of dS/CF T duality in cosmology as
the main topic of this report, and try to show how one can define a holographic formalism for the universe
during the inflationary era. The ultimate goal is to introduce a quantum model that can provide quantitative
predictions that are compatible with latest observations.
In this chapter, at first a brief review of perturbation theory in cosmology is provided. At next stage, we
start applying dS/CF T duality to the cosmological universe by explaining the correspondence between bulk
time evolution and renormalisation group flow on the boundary, and address the issues regarding the number
of degrees of freedom. In the remainder of this chapter formulation of a holographic framework for cosmology
is review in details. Firstly we explain domain-wall cosmology correspondence, and employ the results of
familiar AdS/CF T connection to find the quantum field corresponding to the domain-wall. subsequently
analytical continuation N → −iN is used to map the quantities to the pseudo-QFT dual to the original
cosmology. Then it is shown how holographic formulae can be used to compute all inflationary observables
form predictions of the proposed holographic model. Finally a prototype system for phenomenological analysis
is introduced, and predictions of this model is verified against latest observational data.
7.1 Inflationary Epoch

Theory of inflation was originally introduced so as to provide solutions to some cosmological puzzles including
horizon, flatness and monopole problems. Later on it was realised that inflation can address another mysteri-
ous problem in cosmological particle physics which was generation of inhomogeneity. Quantum perturbations
on a homogeneous classical background are amplified during inflation and amount to formation of current
large structure of the universe. In this section we present an introductory review of selected topics about
primordial fluctuations generated by quantum fluctuations.
7.1.1 Metric Perturbation

In the context of general relativity, space-time is dynamical and will be perturbed as a result of fluctuations
in the matter. During inflationary period it is adequate to work with linear perturbation which involves small
inhomogeneities δgµν on the top of a homogeneous and isotropic background: 1 [18]
ds2 = {gµν (t) + δgµν (t, x)} dxµ dxν , |δgµν | ≪ |gµν | (7.1)
It is natural to decompose this metric perturbation into three modes: scalars φ, ψ, ν and χ, vectors ωi
and νi , and tensor γij . The vectors are transverse ∂i ω i = ∂i ν i = 0, and the tensor is transverse traceless
1 Linear approximation ceases to be valid at current time as inhomogeneities grew over time.
45
∂ i γij = δ ij γij = 0. There are therefore six constraints, and the perturbed metric has ten independent
components. So linear perturbation of FRW metric can be written as: [74]
ds2 = − (1 + 2φ) dt2 + 2a2 (t) (∂i ν + νi ) dtdxi + a2 (t) δij − 2ψδij + 2∂i ∂j χ + 2∂(i ωj) + γij dxi dxj (7.2)

It turns out that vector perturbations decay over time so they can be ignored. So we centre on the scalar
perturbations which affect structural formation, and tensor perturbation which generate gravitational waves.
Since these three modes can be considered independently, from now onwards we set vector modes to zero. 2
[74]
7.1.2 gauge transformation

′
A general perturbation in metric can be defined as the difference between the physical perturbed metric gµν
′ ′ ′
and the unperturbed background metric δgµν (p ) = gµν (p ) − gµν (p). But this comparison makes sense only
if one defines a gauge choice which is a one-to-one map between the point p′ in the perturbed space-time
and the point p in the background. A change in this map or choice of coordinates is referred to as gauge
transformation. For instance consider the linear gauge transformation:
xµ → xµ + ξ µ , ξ µ = (ξ 0 , ∂ i ξ + ξ⊥
i
) (7.3)
Then scalar fluctuations change like: [17]
φ → φ − ξ˙0


ν → ν + ξ 0 /a − aξ˙


(7.4)
 χ→χ−ξ

ψ → ψ + Hξ 0

However tensor perturbation is gauge invariant to linear order because there is no tensor in components
of ξ. 3
There is no unique gauge choice and this freedom in selection of coordinates will give rise to fictitious
gauge. In general, there two ways to avoid spurious gauges:
1. Impose a gauge constraint to remove ambiguity. A familiar example is comoving gauge which requires
that both comoving momentum δq and χ vanishes: 4 [18]
δq = 0, ξ=0 (7.5)
2. Work with gauge invariant quantities which do not change under a coordinate transformation. The
example we use in this chapter is curvature purterbation on hypersafes of uniform energy-density:
H
ζ = −ψ − δρ (7.6)
ρ̇
In addition to ζ there are other gauge invariant quantities obtained from metric perturbations: [74]
δρ
α=φ− (7.7a)
ρ̇
δρ
β = ν − χ̇ + 2 (7.7b)
a ρ̇
Imposing Hamiltonian and momentum constraints on the above invariant parameters yields: [74]
ζ̇ ζ ǫζ̇
α=− , β= + 2 (7.8)
H a2 H q
where ǫ = −Ḣ/H 2 . 5 This means that the dynamics depend only on two invariant free parameters that
are curvature perturbation ζ, and graviton γij .
2 This is the result of the translation invariance of the perturbed metric which means Fourier transformation of the modes do
not interact. Se it is possible to study them separately and set unwanted ones to zero. [74]
3 However γ
ij is not invariant in quadratic order. [17]
4 Comoving gauge requires that slicing is comoving i.e. comoving momentum perturbation equals shift scalar δq = χ, and
that threading is also comoving i.e. momentum perturbation vanishes δq = 0 which gives the definition 7.5 above. [18]
5 This is the same ǫ as the slow-roll parameter; however,note that slow-roll condition is not used here.
46
Helicity
In Fourier space, it is convenient to expand a tensor like γij in helicity basis:
(s)
X
γij (t, q) = γ (s) (t, q)ǫij (q) (7.9)
s=±2
where helicities satisfy orthogonality and normalisation conditions of: [17]

(s) (r)
X (s) (s)
ǫij (q) ǫij (−q) = 2δ sr , ǫij (q) ǫkl (−q) = 2Pijkl
s=±2
6
and Pijkl is traceless transverse projection operator.
7.1.3 Statistics of Perturbation

Consider a real scalar field like ζ. Due to homogeneity two-point correlation function should only depend on
the distance between two points: hζ(x)ζ(y)i = f (|x − y|). Taking the Fourier transform it can be written as:
Z
′
hζ(x)ζ(y)i = hζq ζq′ ie−i(qx+q y) d3 qd3 q ′
and symmetry constraints yield:[73]
hζq ζq′ i = (2π)3 ps (q)δ (3) (q + q′ ) (7.10)
where q ≡ |q|, and p(q) is known as 3D power spectrum and has dimensions of volume.
In an isotropic space we can do the angular Fourier transformation: [73]
d3 q
Z
hζ(x)ζ(x + r)i = ps (q)e−iq.r
(2π)3
Z ∞ 2
q sin qr
= p (q)
2 s
dq
2π qr
Z0 ∞
sin qr
= ∆2s (q) d log q
−∞ qr
where dimensionless power spectra ∆2 (q) representing power per logarithmic intervals in momentum is
defined as:
q3
∆2s (q) ≡ ps (q) (7.11)
2π 2
In order to evaluate scale-dependence of power spectra it is convenient to define scalar spectral index as:
[17]
d ln ∆2s
ns − 1 ≡ (7.12)
d ln q
Another useful measure of primordial scalar fluctuations is running of the spectral index :
dns
αs ≡ (7.13)
d ln q
For observational purposes, power spectrum is approximated by a power law form as: [17]
ns (q∗)−1+ 12 αs (q∗) ln(q/q∗)
q
∆2s (q) = ∆2s (q∗) (7.14)
q∗
6 It is defined via Pijkl = pik pjl + pil pjk − pijkl /2 where transverse projector is pij = δij − qi qj /q 2 . [17]

47
where q∗ is a chosen pilot scale.
The same statistical quantities can be defined for a tensor like graviton γ, but it should be noted that
a tensor has two polarisation modes γ (s) = γ +2 , γ −2 . power spectrum for tensor perturbations is defined as
the sum of power spectra of two modes ∆2T ≡ 2∆2γ .
q3
∆2T (q) ≡ pT (q) hγq γq′ i = (2π)3 pT (q)δ (3) (q + q′ ) (7.15)
2π 2
On historical grounds, tensorial spectral index is defined without the −1 term: [18]
d ln ∆2T
nT ≡ (7.16)
d ln q
and one can run the tensor spectral index in a similar fashion:
dnT
αT ≡ (7.17)
d ln q
Power spectra is fitted by a power law of the form:
nT (q∗)
q
∆2T (q) = ∆2T (q∗) (7.18)
q∗
and the ratio of the tensor to scalar power spectra is often shown as:
∆2T (q)
r≡ (7.19)
∆2s (q)
7.1.4 Dynamics
7
Action of a scalar field with potential V in a classical gravity background is given by:
1 √
Z
−g −R − g µν ∂µ Φ ∂ν Φ + 2κ2 V (Φ) d4 x

S= 2 (7.20)
2κ
We restrict to the case where Φ(t) is monotonic in time so that invert function t(Φ) is well-defined. Hubble
constant can therefore be expressed in terms of Φ such that H = −(1/2)W (Φ). Then known equations of
motion for inflation field on a classical background can be re-written as: [76, 17]
3
Φ̇ = ∂φ W, −2κ2 V = (∂φ W )2 − W 2 (7.21)
2
One can solve the Hamiltonian and momentum constraints 7.8 for free parameters ζ and γij , and insert
the results back into the action above. Then the action for perturbations at second order becomes: [18]
Z
1 1 1
S= 2 a3 ǫζ̇ 2 − aǫ (∂ζ)2 + a3 γ̇ij
2 2
− a∂k γij d4 x
κ 8 8
varying the action with respect to ζ and γij will result in the following equations of motion:

ǫ̇ 1 1 2 (s)
ζ̈ + 3H + ζ̇ + 2 q 2 ζ = 0, γ̈ (s) + 3H γ̇ (s) + q γ =0 (7.22)
ǫ a a2
Note that first equation implies that outside the Hubble horizon k/aH ≪ 1 on super-horizon scales scalar
mode ζ is frozen in time. This is a general property of matter fields when they leave the horizon. It will
remain constant until reentering the horizon after inflationary epoc. 8 [18]
7 This is the minimally coupled case. More details about coupling of a scalar to gravity will be given in section 7.4.2.
8 More precisely this is true if the process is adiabatic and no entropy is produced. [17]
48
Linear Equations of Motion
When working to the first order, linearised wave equations facilitates the required amount of calculations.
At first we need to define the conjugate momentum as:
δL δL 1
Π = κ2 = 2ǫa3 ζ̇, Πij = κ2 = a3 γ̇ij (7.23)
δ ζ̇ δ γ̇ij 4
where for future convenience a factor of κ2 is included in the usual definition. Then linear response
functions are defined so as to relate the conjugate momenta to the perturbation fields:
Π(t, q) = Ω(t, q)ζ(t, q), Π(s) (t, q) = E(t, q)γ (s) (t, q) (7.24)
as mentioned Π(s) denotes helicity component of the linear response function Πij . Substituting back into
the equations 7.22, linear equations of motion can be written as:
1 4 2 a 2
Ω̇ + Ω2 + 2aǫq 2 = 0, Ė + E + q =0 (7.25)
2a3 ǫ a3 4
In most cases like calculation of power spectra, linear response functions suffice for derivation of cosmo-
logical observables.
7.1.5 Cosmological Power Spectra

Having derived the classical equations of motion, standard prescription can be used to quantise the fields.
Scalar and tensor quantum fields are now operators, and follow the equal time commutation relation:
[ζ(t, q), κ−2 Π(t, q′ )] = i(2π)3 δ (3) (q + q′ ) (7.26a)
−2 ′ 3 (3) ′
[γij (t, q), κ Πkl (t, q )] = i(2π) δ (q + q )Πijkl (7.26b)
The latter commutation can be rewritten in the helicity basis 7.9 as:
i
[γ (r) (t, q), κ−2 Π(s) (t, q′ )] = (2π)3 δ (3) (q + q′ )δ rs (7.27)
2
with other commutations vanishing. Note that κ−2 Π and κ−2 Πij are the actual canonical momenta.
Operators can be expanded using creation and annihilation operators as follows:
ζ(t, q) = a(q)ζq (t) + a† (−q)ζq∗ (t), (7.28a)
(s) (s) †
γ (t, q) = b (q)γq (t) + b(s) (−q)γq∗ (t) (7.28b)
and they follow the usual commutation relations:
[a(q), a† (q′ )] = (2π)3 δ(q − q′ ), [b(r) (q), b(s)† (q′ )] = (2π)3 δ(q − q′ )δ rs (7.29)
Using the mode decomposition 7.28 and the explicit expression for momentum one can reach to the
Wronskian relations: [76]

2ǫa3 κ−2 ζq (t)ζ̇q∗ (t) − ζ̇q (t)ζq∗ (t) = i, (7.30a)
1 3 −2 i
γq (t)γ̇q∗ (t) − γ̇q (t)γq∗ (t) =

a κ (7.30b)
4 2
Now, we can use the definition of linear response functions E and Ω in order to find the amplitude of the
scalar and tensor perturbations. It follows that: [76]
2 κ2 2 κ2
|ζq (t)| = − , |γq (t)| = − (7.31)
2Im[Ω(t, q)] 4Im[E(t, q)]
Recalling the definition of power spectra, it is straightforward to define them in terms of constant late
time linear response functions:
49
κ2 q 3 κ2 q 3
∆2S (q) = − , ∆2T (q) = − (7.32)
4π 2 Im[Ω(0) (q)] 2π 2 Im[E(0) (q)]
The relation above is the first step towards the holographic treatment of inflationary cosmology. In the
next sections we try to find a similar relation between domain wall two-point correlation function and linear
responses.
7.2 RG-Flow Universe

Recent observations confirms that universe had a de Sitter geometry far in the past during inflation and will
become a de Sitter space again far in future. Value of cosmological constant raises by an order of about a
hundred. [82, 83] During this time the metric describing the space-time is:
ds2 = −dt2 + e2Ht dxi dxi (7.33)

p
where Hubble constant is H = Λ/3. As discussed earlier isometry under t → t + τ , and xi → e−Hτ xi
indicates that time evolution in the bulk corresponds to scaling of the conformal field on the boundary.
writing the latter in momentum space as k → eHτ k shows that temporal evolution is identified with energy
scale of the gauge theory, hence to renormalisation group flow. Conformal invariance indicates that the field
lives at a fixed point. This means that infra-red fixed point of the conformal field corresponds to the early
inflationary times, whereas ultra-violet point is related to the late cosmological constant-dominated times.
[56]
Between these two points, the universe is described by a flat FRW metric:
ds2 = −dt2 + a2 (t)dxi dxi (7.34)
The metric is no longer symmetric with respect to time evolution-scale transformation since Hubble
constant H(t) = ȧ(t)/a(t) is in general a function of time. Lack of symmetry in the bulk supposedly
corresponds to a boundary quantum field which is not conformal, and to a flow in the renormalisation group.
So RG flow starts at the UV point and flows to the end IR fixed point. [58]
The universe is a RG flow from UV to IR fixed points of the dual Euclidean field theory.
Note that renormalisation group flows opposite direction to the time evolution. Consistent increase in
the value of Hubble constant shows that RG flow is irreversible. [75]
7.3 Holography for Cosmology

As discussed in previous chapter the gravitational theory describing the bulk space-time in dS/CF T cor-
respondence is higher spin gravity, and it was shown that it corresponds to a fermionic Sp(N ) invariant
quantum field. Holographic renormalisation formalism for these theories is not yet very well understood. In
absence of such a prescription we use the correspondence between cosmology and domain wall spaces to move
to anti-de Sitter space, and then apply the known AdS/CF T methodology to find the dual QFT.
Theses steps are schematically described in figure 7.1. The first step is to switch from cosmology to
domain wall space time. Then the holographic renormalisation group flow machinery is employed to find
the quantum observables in terms of the corresponding domain wall variables. Finally we take an analytical
continuation to move to the QFT theory dual to the original cosmology. In this report, we refer to this gauge
theory as pseudo-QFT.
50
Figure 7.1: Schematic diagram illustrating methodology to find quantum field theory dual to the inflationary
cosmology (reproduced from [77])
7.3.1 Domain-Wall / Cosmology Duality

Consider a scalar field Φ with a minimally coupled potential V (Φ) on a flat four dimensional background,
the metric reads:
ds2 = σdz 2 + a2 (z)dxi dxi , Φ = Φ(z) (7.35)
where i = 1 . . . 3, and σ = ± is a constant which allows for two different physical interpretation of the
metric:
1. If σ = −1 then metric corresponds to the inflation field Φ on a flat FRW cosmology while z coordinates
interpreted as time t.
2. If σ = 1, the metric describes a domain wall space-time, and z will be the radial r coordinates.
Without loss of generality, the domain wall metric shown here is Euclidean, but one can do a Wick
rotation for one of xi coordinates to switch to Minkowski metric. Following similar procedures to section
7.1.4 one can derive common classical equations of motion 7.21 for this metric:
3
Φ̇ = ∂φ W, 2σκ2 V = (∂φ W )2 − W 2 (7.36)
2
as well as the equations of motion for scalar and tensor fields 7.22:

ǫ̇ σ σ
ζ̈ + 3H + ζ̇ − 2 q 2 ζ = 0, γ̈ (s) + 3H γ̇ (s) − 2 q 2 γ (s) = 0 (7.37)
ǫ a a
and relevant equations for linear responses 7.25:

1 4 2 σa 2
Ω̇ + Ω2 − 2σaǫq 2 = 0, Ė + E − q =0 (7.38)
2a3 ǫ a3 4
To recognise similar quantities in two pictures, we show variables in the domain wall with a tilde on the
top. The claim is that domain wall space is related to the cosmology space-time via the analytic continuation:
κ̃2 → −κ2 , q̃ → −iq (7.39)

The correspondence applies to both homogeneous background and linear perturbation around that:
1. First we consider background solutions. Continuation κ̃2 → −κ2 is due to the change in the sign of
potential in the background equation 7.36. The second map q̃ → −iq comes from reversing the sign of
q 2 in the equations 7.37. 9
9 The choice of minus sign in −iq results from the fact that behaviour of Bunch-Davies vacuum ζ, γ (s) ∼ e−iqτ at early times
τ → −∞, should correspond to solution of the domain-wall as it decays ζ, γ (s) ∼ eqτ . [76]
51
2. As to linear perturbations, it can be noticed that if E(z, q) and Ω(z, q) are solutions to the cosmological
equations 7.25; then the analytically continued responses Ẽ(z, −iq) and Ω̃(z, −iq) will solve the domain
wall version of the equation i.e. 7.38.
Recall the discussion about Maldacena’s duality where it was shown κ2 ∼ N 2 . So keeping the cosmological
constant Λ fixed, continuation κ̃2 → −κ2 yields:
Ñ → −iN, q̃ → −iq (7.40)
Choice of sign in −iN will be discussed later in section 7.4.3.
7.3.2 Hamiltonian Formalism

As explained earlier in this paper holographic version of Hamiltonian methodology is identical to the usual
Hamiltonian-Jacobi theory with radial coordinate replacing time. This method features a number of ad-
vantages: it admits an arbitrary potential for scalar fields, and also provides a universal formula for the
expectation value of energy-momentum: 10 [78]

2
hTnm i = − √ Πm (7.41)
−g n (3)
with Πm
n being radial canonical, and subscript three means components with conformal dimension three.
11
In an asymptotically AdS space radial momentum is given by: [76]

1 √
Πm
n = −g (Knm − Kδnm ) (7.42)
2κ̃2
where Kmn = 12 ∂z gmn is defined as extrinsic curvature on constant z hypersurfaces.
Finally by varying the Lagrangian, one can obtain Hamilton and momentum constraint equations 7.8.
7.3.3 CFT Two-Point Functions

In this section rules of holographic renormalisation already introduced in section 4.7 are employed to derive
an expression for boundary quantum observables in terms of bulk domain wall linear response functions.
This corresponds to the horizontal line on the top of schematic diagram 7.1. At first, holographic version of
Hamiltonian formalism is reviewed so as to derive a useful expression for the stress two-point correlator. finally
relation between domain wall response functions and two-point functions of the corresponding conformal field
is introduced.
In previous parts of this chapter the expression for cosmological power spectra in terms of response
functions was derived. So at the end, we would be able to define the relation between cosmological observables
and pseudo quantum fields.
The ultimate goal is to show how cosmological observables are related to quantum correlation functions in
order to find a dual description of the universe. In general correlation functions are computed by solving the
equations of motion for domain wall metric with asymptotic Dirichlet boundary conditions. In this section,
we focus on energy-momentum correltor, and decompose it into a traceless part and a trace part, encoded
by coefficients A and B respectively: [77]
hTmn (q̃)Trs (−q̃)i = A(q̃)Pmnrs + B(q̃)pmn prs (7.43)
10 In this section Latin letter indices stating from i, . . . are used for spatial coordinate in the bulk while those starting from
m, . . . used for quantum operators on the boundary.

11 This result is actually expected as in three dimensions energy-momentum has scaling dimension three. [76]
52
The aim is to find a relation between energy-momentum and domain-wall parameters, so we try to find
an expression for coefficients A(q̃) and B(q̃) above in terms of domain wall response function.
Using metric perturbation formalism already discussed, variation of expectation value of energy-momentum
is found to be (refer to appendix F for details):
1 (s)
δhT (s) (q̃)i = A(q̃)γ(0) (q̃), δhT (q̃)i = −4B(q̃)ψ(0) (q̃) (7.44)
2
On the other hand after expanding 7.41 in momentum space, variation of stress tensor to linear order
will be: (appendix F)
!
(s) 2 h −3 (s)
i 1 Ω̃(q̃) 2q̃
δhT (q̃)i = − 2 a Ẽ(q̃)γ , δhT (q̃)i = 2 − 2 ψ (7.45)
κ̃ (3) κ̃ a3 a H
and comparing two equations above, one can work out transverse traceless A(q̃), and trace B(q̃) pieces
as: [77]
1 −2
A(q̃) = −4κ−2 Ẽ(0) (q̃) B(q̃) = κ Ω̃(0) (q̃) (7.46)
4
where the zero subscripts refers to the term in the response function that has zero scaling dimension or
in other word Ẽ(0) and Ω̃(0) are independent of r. 12
To summerise the computations so far, at first in 7.32 we defined the cosmological power spectra in terms
of imaginary parts of cosmological linear response functions. Then in 7.46 we expressed energy-momentum
correlation function in terms of zero-scaling dimension pieces of domain-wall response functions. Now, all
the ingredients are in place. We only need to take the continuation 7.40 and find the imaginary parts of
the response functions ImΩ0 and ImE0 . The expression for dimensionless power spectrum ∆2s (q) and ∆2T (q)
7.32 will amounts to our principal result:
−q 3 −2q 3
∆2s (q) = , ∆2T (q) = (7.47)
16π 2 Im B(−iq) π 2 Im A(−iq)
This equation shows that power spectrum in the cosmology can be obtained straight away from correlation
functions in the dual quantum field. Note that the scalar power spectrum ∆2s (q) is dual to the trace piece of
energy-momentum two-point correlator while the tensor power spectrum ∆2T (q) corresponds to the traceless
piece.
7.3.4 Weak versus Strong Gravitational Coupling

In this section, It was explained that cosmological observables can be described in terms of correlation
functions of a dual quantum field. In particular it should be emphasised that from holographic point of
view, inhomogeneities are not originating from perturbations around a classical FRW background metric as
conventional inflation predicts, but form the fully quantum dynamics of the dual gauge field. This provides
a new insight into the theory of inflation; however the proposed holographic model can be used to offer a
broader scope beyond other current theories.
At early times before inflation universe was in a complicated strongly coupled quantum gravity phase with
no clear description in low energy effective fields. However gauge/gravity relation is a strong-weak duality so
that weakly coupled gravity era can be mapped to weakly interacting low energy fields in the dual quantum
field theory. [37, 56]
12 Ẽ and Ω̃(0) can also be thought of as the leading term in expansion of the response functions E and Ω as r → ∞.
(0)
53
It can be conjectured that early strongly-coupled quantum gravity phase smoothly transformed into the
weakly interacting inflationary epoc. So it is possible to investigate the early time behaviour of the universe
using the proposed holographic model in a well-understood weakly coupled regime of QFT, and provide
quantitative predictions about the unknown features of quantum gravity phase. holographic cosmology can
therefore offer a new perspective on the problem of initial conditions which is beyond the scope of current
standard inflation formalism. [76]
7.4 Holographic Phenomenology

The dual conformal field theory ideally should be derived from the quantum gravity theory in dS space;
however, the map between the cosmological gravity and the dual field is not theoretically very well understood.
Nevertheless we are still interested in a testable field theory that complies with recent observations. So in
this section, we leave the fundamental approach taken throughout this report, and set up a phenomenological
plan to find a prototype model. An intelligent choice is to use quantum field models which their dual gravity
theory is known. In what follows, at first step a prototype CFT model on the AdS boundary dual to domain-
wall gravity is defined. Then this model is used to compute all relevant QFT observables like two-point
correlation functions. At the end analytical continuation is performed to move to the conformal field theory
dual to the original cosmology, and thus to deduce such cosmological observables as power spectrum and
running of spectral index.
7.4.1 Prototype CFT Theory

Prototype theory is supposed to be a typical model emerging in AdS/CFT discussions. It turns out that
such a model needs to satisfy some restrictions: rank of gauge group N should be large, all the fields are
required to be massless, coupling constant must be dimensionful, and scaling dimension of Lagrangian terms
should not equal three. [59] The proposed model is a three-dimensional SU (N ′ ) Yang-Mills theory with the
following action: [76]
Z
1 1 I 1 1
S= F F Imn + Dm φJ Dm φJ + Dm χK Dm χK + ψ̄ L Dψ
/ L
gY2 M 2 mn 2 2

+ λM 1 M2 M 3 M4 Φ M1
Φ M2
Φ M3
Φ M4
+ µαβ M L1 L2
M L 1 L 2 Φ ψα ψβ d3 x (7.48)
where Fmn is field strength, φ denotes an ordinary scalar, but χ is a conformal scalar, Φ = (φ, χ) represents
a pair of these scalars, and ψ is a spinor. Note dimensionful coupling gY M has dimension one 13 , so the
theory is super-renormalisable, whereas quartic λ, and Yukawa µ couplings are dimensionless. Here capital
letters are used to denote gauge group indices, Greek letters show fermionic indices.
7.4.2 Energy-Momentum Tensor

At this point we need to review some background information about quantum field theory in the curved
space-time as energy-momentum tensor in a curved space is different than the one on a flat background.
Lagrangian for a massless scalar field in presence of gravity is given by: [7]
1 √ mn
−g g ∂m φ ∂n φ + ξRφ2

L= (7.49)
2
where g is determinant of metric, and R is Ricci scalar. Note the last term ξRφ2 is the coupling between
scalar field and gravity, and on dimensional grounds it is the only possible option. The dimensionless coupling
constant ξ determines the strength of interaction. If it vanishes then there is minimal coupling i.e. gravity
and scalar are independent.
13 For simplicity we normalised all the field so as to gY M appears in front of the action.
54
Figure 7.2: Feynman diagram for two-point function of energy momentum tensor hTrs (q̃)Tmn (−q̃)i at one-loop
order. (reproduced from [76])
For an arbitrary field, the energy momentum tensor can be obtained through variation of the action with
respect to the metric: [2]
2 δS
Tmn = √ (7.50)
−g δg mn
Thus for a minimally coupled scalar energy-momentum is found to be: [8]
1
Tmn = ∂m φ ∂n φ − gmn g rs ∂r φ ∂s φ (7.51)
2
Another value of interest for coupling constant is ξ = (d − 2)/4(d − 1) which amounts to conformal
coupling. In this case metric is invariant under a conformal transformation 2.1. Energy-momentum tensor
for such a massless scalar χ will be given by: [7]
1
Tmn = (1 − 2ξ)∂m χ ∂n χ + (2ξ − )gmn g rs ∂r χ ∂s χ − 2ξ(∂m ∂n χ)χ
2
2 1 d−1
+ ξgmn χχ − ξ Rmn − Rgmn + 2 ξRgmn χ2 (7.52)
d 2 d
Furthermore energy-momentum of a spinor is found to be:
1
Tmn = ψ̄γ(m ∂n) ψ − (∂(m ψ̄)γn) ψ (7.53)
2
and for the gauge field will be:
r 1
Tmn = 2Fm Frn − gmn F rs Frs (7.54)
2
Replacing for covariant derivative, and adding the Yukawa and quartic interaction terms the total energy-
momentum tensor of the non-abelian model reads: [76]

1 Ir I 3 1
Tmn = 2 tr Fm Frn + Dm φJ Dn φJ + Dm χK Dn χK − (Dm Dn χK )χK
gY M 4 4

1 1 1 1
Rmn − Rgmn + Rgmn (χK )2 + ψ̄ L γ(m Dn) ψ L − (D(m ψ̄ L )γn) ψ L

−
8 2 6 2

1 1 1
− gmn F Irs Frs
I
+ g rs Dr φj Ds φj + g rs Dr χK Ds χK + χK χK
2 4 12

+ λM1 M2 M3 M4 ΦM1 ΦM2 ΦM3 ΦM4 + µαβ M L 1 L2 Φ M L1 L2
ψ α ψ β (7.55)
7.4.3 Cosmological Observables

In the following sections the aim is to calculate such cosmological observables as power spectra and running of
the spectral index, and make a comparison with current observations. Even though a different phenomenology
is used, compatible results would confirm validity of holographic model introduced here.
55
Figure 7.3: Feynman diagrams for two-point function of energy momentum tensor hTmn (q̃)Trs (−q̃)i at two-
loop order. (reproduced from [76])
In order to do so, first we need to find the coefficients A(q̃) and B(q̃) i.e. traceless and trace components
of energy-momentum 2-point function 7.46. Then the result is analytically continued to the pseudo-QFT,
and finally power spectra 7.47, and other observables could be readily computed. Also along the way, we
can check correctness of our assumptions regarding large N limit, and small effective interaction gef f .
1-Loop Order
Summing over contribution from different fields in 7.48, 2-point function for energy-momentum could be
obtained. In addition power counting can be used as a calculation shortcut. Because Tmn has dimension
three, there should be a q̃ 3 contribution. 14 [76] Also due to conformal structure, and large N counting the
general format of coefficients will be given by:[77]
A(q̃) = a(0) Ñ 2 q̃ 3 , B(q̃) = b(0) Ñ 2 q̃ 3 (7.56)
where the coefficient Ñ 2 stems from tracing over gauge indices. 15 Coefficients a(0) and b(0) are a function
of dimensionless effective coupling gef f = gY2 M Ñ /q̃, and they also depend on the field content: [76]
a(0) = (NA + Nφ + Nχ + 2Nψ ) /256, b(0) = (NA + Nφ ) /256 (7.57)

where NA indicates number of gauge fields A in the system etc. One-loop diagram for hTmn (q̃)Trs (−q̃)i
is shown in figure 7.2. Applying the continuation Ñ = −iN , and q̃ = −iq, the cosmological power spectrum
will be:16 [76]
1 2
∆2s (q) = 2 2 (0)
, ∆2T (q) = 2 2 (0) (7.58)
16π N b π N a
This clearly shows that power spectra is scale independent to the leading order which is consistent with
observational data.
The tensor to scalar amplitude ratio is found to be:
∆2s b(0)
r= 2 = 32 (0) (7.59)
∆T a
Maximum value of r will be constrained by observational data which in turn clarifies the field content
7.56. For instance if r is decreased, the number of gauge fields and scalars need to decrease or fermions and
conformal scalars must increase in number.
Using equation 7.58 above, one can verify our assumption about the large value of N . According to
WAMP data [82], scalar power spectrum is ∆2S (q∗ ) ≈ 10−9 which means N ≈ 104 .
2-Loop Order
In order to find correction to one-loop order calculations, one needs to evaluate the energy-momentum tensor
correlator at two-loop order. Feynman diagram for hTmn (q̃)Trs (−q̃)i at two-loop order is shown in figure
14 Note only comoving momentum q̃ with dimension one counts because Yang-Mills coupling g
Y M does not contribute at one
loop order.
15 Recall that there was a factor of κ̃2 in the expression for power spectrum 7.32 which corresponds to the factor of Ñ 2 here
on the CFT side. Also recall the relation κ̃2 ∼ Ñ 2 .

16 Dimensionless ’t Hooft coupling g
ef f remains invariant under this continuation hence justifying minus sign in Ñ = −iN .
56
Figure 7.4: Error bar plot for WAMP data from figure 4 in [82] which shows 68% and 98% CL constraints
at the pilot scale q ∗ = 0.002M pc−1 . The straight line is QFT prototype predictions.(reproduced from [77])
7.3. Scale-free theories which are super-renormalisable have infra red divergences. [12] Power counting shows
that each of the diagrams has a quadratic IR divergence, and after regularisation it is evaluated as q̃ 2 ln q̃/q̃ ∗ .
There is also an overall factor of Ñ 3 gY2 M . So in total contribution of each diagram to the stress correlation
function will be found to be Ñ 3 gY2 M q̃ 2 ln (q̃/q̃ ∗ ) = Ñ 2 q̃ 3 gef
2 ∗
f ln (q̃/q̃ ). Therefore at this order, the traceless
and trace coefficients can be written as: [76]
h i
A(q̃) = a(0) Ñ 2 q̃ 3 1 + a(1) gef 2
f ln (q̃/q̃ ∗
) , (7.60a)
h i
B(q̃) = b(0) Ñ 2 q̃ 3 1 + b(1) gef 2
f ln (q̃/q̃ )
∗
(7.60b)
where coefficients a(1) and b(1) depend on the field content and, Yukawa and quartic coupling constants.
Substituting this result back into the expression for hTmn (q̃)Trs (−q̃)i in 7.47 we obtain: [77]
h i
∆2S (q) = ∆2S (q ∗ ) 1 − b(1) gef
2
f ln(q̃/q̃ ∗
) , (7.61a)
h i
∆2T (q) = ∆2T (q ∗ ) 1 − a(1) gef
2
f ln(q̃/q̃ )
∗
(7.61b)
where ∆2s (q ∗ ) ≡ 1/(16π 2 N 2 b(0) ) and ∆2T (q ∗ ) ≡ 1/(16π 2 N 2 a(0) ).

In order to make a comparison, we expand 7.14 and 7.18 to first order in (nS − 1) and nT which yields:
∆2s (q) = ∆2S (q ∗ ) [1 + (nS (q) − 1) ln(q̃/q̃ ∗ )] , (7.62a)

∆2T (q) = ∆2T (q ∗ ) [1 ∗
+ nT (q) ln(q̃/q̃ )] (7.62b)
Comparing 7.61 and 7.62 indicates:
ns (q) = 1 − b(1) gef

2 4
f + +O(gef f ) nT = −a(1) gef
2 4
f + +O(gef f ) (7.63)
It can be seen that at two-loop order there is a slight deviation from the original scale-invariance of
spectral index. To investigate scale dependence, one can run the spectral index, wich yields:
dnS 4
αS = = −(nS − 1) + O(gef f) (7.64)
d ln q
However slow-roll approximation predicts that αS if of second order in (nS − 1). To make a better
judgment, error bar plot for WAMP data showing spectral tilt nS versus αS is illustrated in figure 7.4. It is
clear that predictions of the model studied lies close to the allowed region which means that the results are
in agreement with observational data. [82]
Depending on the sign of a(1) (b(1) ) red or blue power spectrum is favored. Power spectrum for positive
and negative values of the coefficient b(1) is plotted in figure 7.5. It can be seen that for large momenta,
57
Figure 7.5: Relative power spectrum in momentum scale. The lower graph illustrates negative value of b(1) ,
and upper one corresponds to positive values.(reproduced from [79])
spectrum is scale independent, and positive value of b(1) corresponds to a slightly blue index while negative
value amounts to a slightly red tilt.
Infra red divergences causes a pick at smaller values of momentum at ln |q/q̃| = 1. Nevertheless note that
the perturbative treatment used in analysing the energy-momentum is no longer valid for small momenta.
So the results around the IR peak could be unreliable. [79]
We also need to check the value of effective coupling gef f to verify validity of perturbation method used
here. According to WAMP data at q = 0.002M pc−1 spectral index reads (nS − 1) ≈ O(10−2 ) which gives
the small value of gef f ≈ O(10−2 ). [82]
Once values of N , gY M , and the field content are determined, one can compute any other cosmological
observables using this model. As discussed above, in conclusion, the phenomenological approach is able to
satisfy constraints given by current observations.
58
Chapter 8
Conclusion
In this report, gauge/gravity correspondence for de Sitter space was studied and different properties were
investigated. In the first chapters basic materials necessary to understand the duality was reviewed which
included conformal field theory, and geometry of constant curvature spaces specially anti-de Sitter and de
Sitter. The formulation of dS/CF T duality has been guided by the anti-de Sitter counterpart correspondence.
As a result, at first a review of AdS/CF T duality was provided, and different aspects of the correspondence
including concrete example of AdS5 × S 5 /N = 4, SY M relation were studied.
However since in the dS/CF T correspondence time direction in the gravitational theory is holographi-
cally reconstructed, this duality has some unique features. It was shown that corresponding conformal field
lives in the asymptotic future of the de Siter, and there exist a renormalisation group which flows back to
the asymptotic past in opposite direction to the time. It was also explained that the fundamental theory
describing the gravity in this space is higher spin theory which corresponds to a fermionic Sp(N ) invariant
conformal field.
The application of dS/CF T duality to cosmology as the main topic of this dissertation was studied
in details, it was explained that how holographic renormalisation group could be formulated for universe
during the inflationary period. The exact correspondence between cosmology and the gauge theory is not
yet well understood. Using the correspondence between cosmology and domain wall spaces, nevertheless it
was shown that there is a methodology which makes this connection through the well established anti-de
Sitter and conformal field AdS4 /CF T3 duality. The main point was to relate the cosmological observables
to correlation functions of the quantum field on the other side.
A proposed prototype quantum field model was reviewed in details and values of cosmological observables
in particular power spectra was computed using this model. It was explained that predictions of the prototype
quantum field is consistent with latest observational data available. We also argued that holographic model
is able to make predictions where current inflation theories do not apply.
The gauge-gravity connection is a kind of weak-strong duality which means when the gravity is strongly
coupled the corresponding quantum field has weak interactions. It was discussed that the dual QFT is
potentially able to provide predictions on the strongly-interacting era which is beyond the scope of current
standard inflationary theories. This means that the holographic model has the potential to solve major
problems in the contemporary cosmology about the initial conditions of the early universe.
8.1 Futer Works

In this section some suggestions for possible future works on applications of the gauge/gravity duality in cos-
mology are provided. They range from computational simulations to analytical studies and more fundamental
ideas at early conceptual stages.
More recent results of the Planck space observatory have been released providing a more detailed image
of cosmic microwave background. It would be interesting to investigate the effect of new information on
the predictions of the prototype quantum model, and to examine whether predictions of the model are still
compatible with resent observations. In addition over the coming years, large amount of observational data
will impose more constraints on the values of cosmological observables. If new results confirm the predictions
59
of the dual QFT model, it could be regarded as the first evidence for the holographic description of our
universe.
In principal holographic formalism is applicable to any cosmology with asymptotic de Sitter behaviour
not only the inflationary epoc. In particular cosmological holography explained in this paper could be applied
to the far late cosmological constant dominated universe. At first step we need to provide a clear definition
of cosmological observables for the late universe. Then we need to try to derive an RG flow formulation for
that era along the lines of what was accomplished for the inflationary period.
Before the inflationary epoc, universe was in an unknown intractable state where gravity was strongly
coupled when current standard theories of inflation can not provide any explanation. As discussed earlier,
dS/CF T prescription is able to map this system of strongly interacting gravity to a theory of quantum
field with weak coupling. The proposed holographic model of inflation can provide a complete description
including the early behavior of the system. It would be desirable to find an explicit definition of the dual
quantum model. Such a model will be able to give quantitative predictions about pre-inflationary state and
offer some solutions for the problem of initial conditions. √
It was discussed that loop corrections of the correlators in cosmology correspond to 1/ N corrections
to the quantum system. In the limit where N is large, the analytic continuation is still valid, and this
model is able to predict all cosmological observables. The next step is to move beyond the perturbative
limits and find a more general definition of the holographic model. The gravitational theory supporting the
dS/CF T correspondence is the higher spin formalism of gravity. A longer-term goal would be to find a
holographic framework in the context of correspondence between HS gravity and Sp(N )-invariant fermionic
theory of quantum field. In absence a clear relation between the mentioned theories one can try an analogous
methodology to what proposed in this report. Using analytic continuation it is possible to map de Sitter to
a Euclidean AdS space, and then use the known GKPY duality to provide a definition of the corresponding
QFT with a familiar O(N ) gauge group. All the computations can be carried out in this regular quantum
theory, and finally results are analytically mapped to the pseudo-QFT dual to the higher spin gravity.
8.2 Acknowledgment
I would like to acknowledge Dr Toby Wiseman supervisor of this project, for his instructive comments and
discussions over the course of the project.
I would also like to acknowledge Prof. Carlo Contaldi, and Dr Paul McFadden for their contributions to
this project.
60
Appendix A
Anti de Sitter Equation of Motion
The equation of motion for a scalar field in AdSd is given by:
( − m2 ) φ = 0
where Laplacian can be computed as:

1 √
φ = g µν φ,µ;ν = √ ∂µ ( gg µν ∂ν φ)
g
zd
d 2
ld z 2 2

l z
= d ∂z ∂z + d 2 ∂x
l z d l2 z l
2
z
= 2 z d−2 ∂z z 2−d ∂z + ∂x2

(A.1a)
l
z2
= 2 ∂z2 − (d − 2)z −1 ∂z + ∂x2

(A.1b)
l
Separating the variables via φ(z, x) = f (z)Φ(x), and using relation A.1a, equation of motion will be
written as:
z2
− 2 z d ∂z z 2−d f ′ Φ + f ∂ 2 Φ + m2 f Φ = 0

l
where ∂ 2 denotes Laplacian in Minkowski space-time. Dividing both sides by f Φ we reach to two inde-
pendent equations:
z d−2 m2 l 2 ∂2Φ
− ∂z z 2−d f ′ + 2 = = −k 2
f z Φ
where k µ turns out to be momentum with modulus square k 2 . Treating both of these equations indepen-
dently, one reaches to the set of equations of motion in 3.24:
(∂x2 − k 2 )Φ(x) = 0 (A.2)

d 2−d 2 2 2 2

−z ∂z (z ∂z ) + m l + k z f (z) = 0 (A.3)
and the complete solution is given by the superposition of the form φ(z, x) = Φ̃k (x)f˜k (z)dd−1 k. The
R
equation of motion for Φ is the known Klein-Gorden equation with plane wave solutions Φ(x) = eikx /(2π)d−1 .
So superposition of the results gives rise to 3.25:
dd−1 x
Z
φ(z, x) = f˜k (z) eikx
(2π)d
In addition, equation of motion for f can be rewritten as:
z 2 f˜k′′ − (d − 2)z f˜k′ − m2 l2 + k 2 z 2 f˜k = 0

a
change of variable g̃k = z (1−d)/2 f˜k gives:
(d − 1)2

z 2 g̃k′′ + zg̃k′ − 2 2 2 2
+ k z + m l g̃k = 0
4
Writing the above equation in terms of kz rather than z, we end up with modified Bessel equation: [38]
(d − 1)2

2 ′′ ′ 2 2 2 2
(kz) g̃k + (kz)g̃k − + m l + k z g̃k = 0 (A.4)
4
with the solution:

g̃k = ak Kδ (kz) + bk Iδ (kz)
p
2 2 2
where δ = (d − 1) /4 + m l , and Iδ and Kδ are modified Bessel functions of first and second type
respectively. Rewriting in terms of f˜k , the solution to the wave equation will be found as:
f˜k (z) = ak (kz)(d−1)/2 Kδ (kz) + bk (kz)(d−1)/2 Iδ (kz)
The solution has to be regular throughout the interior, and particularly when approaching the deep
interior z → ∞. Modified Bessel equations in the limit of infinity behave like: [23]
√ √
lim Iδ (z) = ez / z, lim Kδ (z) = e−z / z (A.5)
z→∞ z→∞
So first Bessel function Iδ diverges unless bk = 0, which reduces the solution to:
f˜k (z) = ak (kz)(d−1)/2 Kδ (kz)
On the other side, when approaching the origin, modified Bessel functions behave like:
z δ δ
1 Γ(δ) 2
lim Iδ (z) = , lim Kδ (z) = (A.6)
z→0 Γ(δ + 1) 2 z→0 2 z
So near the boundary solution z → 0 will be:

" δ δ #
˜ (d−1)/2 Γ(δ) 2 Γ(−δ) kz
fk (z) ≈ ak (kz) +
2 kz 2 2
Defining φ̃− ≡ ak 2δ−1 Γ(δ)k ∆− , and φ̃+ ≡ ak 2−(δ+1) Γ(−δ)k ∆+ one reaches to the asymptotic solution in
the momentum space 3.28:
f˜k ≈ φ̃− (k)z ∆− + φ̃+ (k)z ∆+
q
2
with usual scalar mass-dimension relation ∆± = 2 ± δ = 2 ± (d−1)
(d−1) (d−1)
4 + m2 l2 .
Finally applying the reverse Fourier transformation we find the familiar scalar solution in the anti-de
Sitter space, 3.30:
lim φ(z, x) = z ∆− φ− (x) + z ∆+ φ+ (x) (A.7)

z→0
b
Appendix B
Anti-de Sitter Bulk-to-Boundary

Propagator
Bulk-to-Boundary Propagator in AdSd is given by 4.21

∆ +
z
K(z, x; y) = C∆+ (B.1)
z 2 + (x − y)2
In this section we show that near the boundary z −∆− K behaves like a delta function, and fix the nor-
malisation constant C∆+ . So it is required to integral over x and expect to reach to a constant value. Due
to translation invariance we can set y = 0.
1
Z Z
K(z, x)dd−1 x = C∆+ z ∆+ dd−1 x
(z + x2 )∆+
2
Z ∞
rd−2
= C∆+ z ∆+ Ωd−2 dr
0 (z + r2 )∆+
2
changing the variable t = r/z:
∞
td−2
Z
d−1−∆+
= C ∆+ z Ωd−2 dt
0 (1 + t2 )∆+
Again change of variable x = t2 :

Z ∞ (d−3)/2
1 x
= C∆+ z d−1−∆+ Ωd−2 dx
2 0 (1 + x)∆+
1 d−1 d−1
= C∆+ z d−1−∆+ Ωd−2 B( , ∆+ − )
2 2 2
(d−1)/2 Γ d−1 Γ ∆ − d−1

1 d−1−∆+ 2π 2 + 2
= C ∆+ z
Γ d−1

2 2
Γ(∆+ )
d−1
where Ωd−2 = 2π (d−1)/2 /Γ

2 , and beta function is defined as: ??
∞
tx−1 B(x)B(y)
Z
B(x, y) = dt = (B.2)
0 tx+y B(x + y)
So normalisation constant C∆+ is fixed to:
Γ(∆+ )
C ∆+ = d−1
(B.3)
π (d−1)/2 Γ(∆+ − 2 )
c
such that: Z
K(z, x)dd−1 x = z −∆− (B.4)
This shows that K does not behave like a delta function but z −∆− K does:
lim z −∆− K(z, x; y) = δ (d−1) (x − y) (B.5)

z→0
d
Appendix C
Anti de Sitter Two-Point Function
It was shown in section 4.4.1 that action of interest for a two-point correlator in the AdSd is given by:
1 √ µν
Z

S=
2
For the first term, integration by part gives:
√ √ √
−gg µν ∂µ φ∂ν φ = ∂µ −gg µν φ ∂ν φ − φ −g φ

with usual definition of Laplacian in 3.22. Imposing the on-shell condition ( − m2 )φ = 0, the action
reduces to:
1 √ 1 √
Z Z
S= −g∂µ (g µν φ ∂ν φ) dd−1 xdz − −gφ( − m2 )φ dd−1 xdz
2 2
1 √
Z
= −g∂µ (g µν φ ∂ν φ) dd−1 xdz
2
Using Stoke’s theorem this can be integrated over the boundary: [43]
1 √
Z
S= −gφ (n.∇φ) dd−1 x (C.1)
2 ∂AdS
Without loss of generality one can choose the normal vector n parallel to z axis. Working in Poincare
√
coordinate −g = z −d and g zz = z 2 , when approaching the boundary the action behaves like:
1 √
Z
lim S = − −g g zz φ ∂z φ dd−1 x
z→0 2 ∂AdS
1
Z
=− z 2−d φ ∂z φ dd−1 x (C.2)
2 ∂AdS
Now, recall that field is given by the convolution of the bulk-to-boundary green function with the source
4.18:
Z
∂z φ(z, x) = ∂z K(z, x; y)φ0 (y)dd−1 y
" #
1 z ∆+
Z
= ∂z ∆
φ0 (y)dd−1 y
C ∆+ (z 2 + |x − y|2 ) +
Z " ∆+ +1 #
∆+ z + z ∆+ −1 |x − y|2 − 2z ∆+ +1
= ∆ +1
φ0 (y)dd−1 y
C ∆+ (z 2 + |x − y|2 ) +
When approaching the boundary this will behave like:
∆+ z ∆+ +1
Z
lim ∂z φ(z, x) = dd−1 y (C.3)
z→0 C ∆+ |x − y|2∆+
e
In addition, recall that near the boundary the field behaves like φ ≈ z ∆− φ0 . Inserting these results into
C.2 powers of z cancel out, and asymptotic action is simplified to:
1 ∆+ φ0 (x)φ0 (y) d−1 d−1

Z
S= d xd y (C.4)
2 C ∆+ |x − y|2∆+
Finally two-point correlation function can be computed by taking the derivative with respect to the source:

−iδ −iδ
hO(x)O(y)i = S
δφ0 (x) δφ0 (y) φ0 =0
φ0 (x)δ (d−1) (x − x′ ) φ0 (x)δ (d−1) (x − y′ )

δ ∆+
= − ′ ′ 2∆
+ ′ ′ 2∆
S
δφ0 (y) 2C∆+ |x − y | + |x − y | +
φ0 =0
(d−1) ′ (d−1) ′ (d−1)
(x − y′ )δ (d−1) (y − x′ )

∆+ δ (x − x )δ (y − y ) + δ
=−
2C∆+ |x′ − y′ |2∆+
which gives the expression 4.29 in the text:
∆+ 1
hO(x1 )O(x2 )i = − (C.5)
2C∆+ |x − y|2∆+
f
Appendix D
De Sitter Green Function
As explained in section 6.5 green function in the de Sitter space is the solution of Laplace equation:
− m2 G(x − y) = δ (d) (x − y)

which leads into the hypergeometric equation 6.13, and the solution will be a multiple of hypergeometric
function F :
d P +1
G(P ) = C∆ F ∆+ , ∆− , ;
2 2
here we need to fix the constant C∆ . The exact behavior of the hypergeometric function F (a, b, c; z) near
the singularity z = 1 is given by: [23]
1− d2
Γ d2 Γ d2 − 1

D2

d
F ∆ + , ∆− , ; z ≈ (D.1)
2 4 Γ (∆+ ) Γ (∆− )
where D is geodesics separation defined in section 5.3. A short distance singularity in de Sitter behaves
like: [55]
Γ d2 − 1

D2−d (D.2)
2(d − 2)π d/2
Value of coefficient C∆ can be obtained by equating D.1 and D.1 so that:
Γ d2

Γ (∆+ ) Γ (∆− )
C∆ = 41−d/2 ×
Γ d2 Γ d2 − 1 2(d − 2)π d/2

Γ (∆+ ) Γ (∆− )
= (D.3)
(4π)d/2 Γ d2

So the green function in de Sitter space is given by:

Γ (∆+ ) Γ (∆− ) d P +1
G(P ) = F ∆+ , ∆− , ; (D.4)
(4π)d/2 Γ d2 2 2
D.1 Asymptotic behaviour

Near the I − boundary, invariant distance between two points P diverges because recalling 5.18:
1 ′
lim P (t, x; t′ , y) = − e−(t+t ) |x − y| (D.5)
t,t′ →−∞ 2
g
So the last argument in D.4 above, diverges. There are transformation formulas for hypergeometric
function which relate the functions with different values in z, the last argument: [23]
Γ d2 Γ (∆− − ∆+ )

d −∆+ d−1 1
F ∆ + , ∆− , ; z = (−z) F ∆ + , ∆ + − , ∆ + − ∆ − + d − 2; + (∆+ ↔ ∆− )
Γ (∆− ) Γ d2 − ∆+

2 2 z
(D.6)
Also note:
F (a, b, c; 0) = 1
So asymptotic behaviour of green function near the I − boundary, will be:
d ′
(d + 1)∆+ Γ

′ 2Γ (∆− − ∆+ ) e∆+ (t+t )
lim G(t, x; t , y) = 2∆
+ (∆+ ↔ ∆− ) (D.7)
Γ (∆− ) Γ d2 − ∆+

t,t′ →−∞ |x − y| +
h
Appendix E
De Sitter Two-Point Function
As explained earlier, the green function in de Sitter space dSd is defined as 6.9:
− m2 G(x − y) = δ (d) (x − y)

Therefore the field can be expressed:

√
Z
φ(t, x) = −gφ(t′ , y)( − m2 )G(t, x; t′ , y)dt dd−1 y (E.1)
Integrating the first term twice by parts yields:

φ G = ∂µ (g µν φ ∂ν G) − g µν ∂µ φ ∂ν G
= ∂µ (g µν φ ∂ν G) + ∂µ (g µν ∂ν φ G) + φ G
So E.1 becomes:
√
Z
−g ∂µ (g µν φ ∂ν G + g µν ∂ν φ G) + ( − m2 )φ G dt dd−1 y

φ(t, x) =
√

↔
Z
= −g∂µ g G∂µ φ dt dd−1 y
µν
where in the second line on-shell constraint ( − m2 )φ = 0 is imposed. Using Stoke’s theorem this can
be rewritten as:
√
Z ↔
φ(t, x) = −gG(t, x; t′ , y)n.∇φ(t′ , y)dd−1 y
∂dS
Without loss of generality, we can take the normal vector n to be in parallel with time axis. Using planar
√
coordinates 5.13 have −g = e−(d−1)t/l , and gtt = −1. So near the boundary field goes like:
↔
Z
lim φ(t, x) = − e−(d−1)t/l G(t, x; t′ , y)∂t′ φ(t′ , y)dd−1 y (E.2)
t→−∞ I−
On the other hand, it was discussed in section 6.6 that required gravity action is given by:
1 √ µν
Z
−g g ∂µ φ∂ν φ − m2 φ2 dt dd−1 x

S=
2
Following the same steps as before, by integration by part, and applying the Stoke’s theorem on the past
null infinity I− , the on-shell action behaves like:
↔
Z
lim S = e−(d−1)t/l φ(t, x)∂t φ(t, x) dtdd−1 x (E.3)
t→−∞ I−
Now, inserting expression for the field E.2 into the above integral gives the following relation for the
gravity on-shell action:
i
↔ ↔
Z
′
′
lim S=− e−(d−1)(t+t )/l φ(t, x)∂t G(t, x; t′ , y)∂t′ φ(t′ , y) dd−1 ydd−1 x (E.4)
t,t →−∞ I−
which is the same equation as 6.21 in the text.
E.1 Asymptotic Behaviour

Referring to D.7, near the past boundary I − , green function behaves like:
′ ′
i ′ i e∆+ (t+t ) e∆− (t+t )
lim G(t, x ; t , y ) = C∆+ 2∆+
+ C ∆− 2∆−
(E.5)
t,t′ →−∞ |x − y| |x − y|
where C∆± are constants fixed in the appendix D. Inserting asymptotic green function above E.5 and
the field on the boundary 6.18 in the action E.4, we find:
φ− (x)φ− (y) d−1 d−1

Z
S≈ d xd y (E.6)
|x − y|2∆+
Taking the derivative of action with respect the source, two-point correlation function will be given by:

δ δ
hO(x)O(y)i = S
δφ− (x) δφ− (y) φ− =0
∆+ φ− (x)δ (d−1) (x − x′ ) φ− (x)δ (d−1) (x − y′ )

δ
= ′ ′ 2∆
+ ′ ′ 2∆
S
δφ− (y) 2C∆ |x − y | + |x − y | +
φ− =0
(d−1) ′ (d−1) ′ (d−1)
(x − y′ )δ (d−1) (y − x′ )

∆+ δ (x − x )δ (y − y ) + δ
=
2C∆ |x′ − y′ |2∆+
This gives the relation 6.24 in the main text:
∆+ 1
hO(x)O(y)i = (E.7)
2C∆ |x − y|2∆+
j
Appendix F
Holography for Cosmology
In this appendix it will be demonstrated how energy-momentum tensor two-point functions can be expressed
in terms of domain wall linear response functions. Hamiltonian formalism already introduced in section 4.7
is used to perform the computations.
Before dong so consider quantum field theory on the boundary which is a flat metric in absence of any
sources. Ward identity indicates that energy-momentum 2-point function must be transverse:
q̃ m hTmn (q̃)Trs (−q̃)i = 0
Considering symmetry constraints, there are only two ways to construct transverse tensors which implies
tow-point stress correlator can be decomposed as follows: (equation 7.43)
hTmn (q̃)Trs (−q̃)i = A(q̃)Pmnrs + B(q̃)pmn prs (F.1)

As explained in the text, it is desirable to express transverse traceless A(q̃), and trace B(q̃) coefficients
of energy-momentum tensor in terms of domain-wall response functions Ẽ(q̃) and Ω̃(q̃).
Note that in general variation of a two-point function like hTmn (q̃)Trs (−q̃)i is given by variation of the
one-point function with respect to the source, in this case g(0)ab . In the flat boundary metric, perturbation
7.1 can be written as gmn = δmn + δgmn , so :
1
Z
m mr
δhTn i = δ δhTrn i = − δ mr hTrn Tst i δg(0)
st 3
d x
2
Taking the Fourier transformation, in momentum space:
1
δhTnm (q̃)i = − δ mr hTrn (q̃)Tst (−q̃)i δg(0)
st
(q̃) (F.2)
2
Using the the stress two-point function decomposition F.1, and metric perturbation formalism 7.2 we
reach to the equation 7.44:
1 (s)
δhT (s) (q̃)i = A(q̃)γ(0) (q̃), δhT (q̃)i = −4B(q̃)ψ(0) (q̃) (F.3)
2
On the other side expanding canonical momentum 7.41 to linear order gives:
1
δhTnm i = (δK δnm − δKnm )(3)
κ̃2
and imposing Hamiltonian and momentum constraints 7.8:

1 1 m
δhTnm i = − 2 2ψ̇δnm + q̃ 2 χ̇ pm
n + γ̇ (F.4)
κ̃ 2 n (3)
k
Then using the gauge φ = ν = 0, one finds that: [76]
2ψ̇ = ρ̇ δρ, (F.5a)

2 2 2
2q̃ ψ 2q̃ ψ κ̃ Ω̃ζ
q̃ 2 χ̇ = 2
− ǫζ̇ = 2 −
a H a! H a3
2q̃ 2 κ̃2 Ω̃ κ̃2 Ω̃H
= 2
+ 3
ψ− 3 δρ (F.5b)
a H 2a a ρ̇
1 m Ẽ
γ̇ = 2 3 γnm (F.5c)
2 n a
But terms proportional to δρ contribute to the dual operator hOi, not to the transverse or traceless parts.
So these terms can be omitted and we end up with equation 7.45:
3 !
(s) 2 a (s) 1 Ω̃(q̃) 2q̃
δhT (q̃)i = − 2 γ , δhT (q̃)i = 2 − 2 ψ (F.6)
κ̃ Ẽ(q̃) (3)
κ̃ a3 a H
and finally following comparison with F.3, coefficients can be written in terms of response functions:
(equation 7.46)
κ̃−2
A(q̃) = −4κ̃−2 Ẽ(0) (q̃) B(q̃) = − Ω̃(0) (q̃) (F.7)
4
l
Bibliography
[1] M. Nakahara, Geometry, Topology and Physics, Taylor and Francis (2003).
[2] J. Hartle, Gravity: An Introduction to Einstein’s General Relativity, Pearson Education, (2002).
[3] S. Hawking, and G. Ellis, The Large Scale Structure of Space-Time, Cambridge University Press, (1975).
[4] E. Schrdinger, .Expanding Universe, Cambridge University Press, (1956).
[5] J. Sakurai, and J. Napolitano, Modern Quantum Mechanics, Pearson, (2011).
[6] M. Peskin, and D. Schroeder, An Introduction to Quantum Field Theory, Advanced Book Program,
(1995).
[7] N. Birrell, P. Davies, Quantum Fields in Curved Space, Cambridge University Press, (1984).
[8] T. Hack, Cosmological Applications of Algebraic Quantum Field Theory in Curved Spacetimes, Springer
International Publishing, (2015).
[9] Lee Smolin, Three Roads to Quantum Gravity, Perseus Book Groups, (2001).
[10] P. Di Francesco,and P. Mathieu Conformal Field Theory, Springer, (1999).
[11] R. Blumenhagen, E. Plauschinn Introduction to Conformal Field Theory, With Applications to String
Theory, Springer, (2009).
[12] V. Rivasseau From Perturbative to Constructive Renormalization, Princeton University Press (1991).
[13] K. Becker, M. Becker, and J. Schwarz String Theory and M-Theory: A Modern Introduction, Cambridge
University Press, (2006).
[14] J. Polchinski String Theory, Cambridge University Press, (1998).
[15] C. Johnson D-Branes, Cambridge University Press, (2003).
[16] J. Wess, and J. Bagger, Supersymmetry and Supergravity, Princeton University Press, (1992).
[17] S. Weinberg Cosmology, Oxford University Press, (2008).
[18] V. Mukhanov, Physical Foundations of Cosmology, Cambridge University Press, (2005).
[19] M. Ammon, and J. Erdmenger, Gauge/Gravity Duality Foundations and Applications, Cambridge Uni-
versity Press, (2015).
[20] K. Patrick, Gauge/Gravity Duality, Sdwestdeutscher Verlag fr Hochschulschriften, (2012).
[21] N. Nastase, Introduction to the ADS/CFT Correspondence, Cambridge University Press, (2015).
[22] M. Natsuume, The Ads/Cft Duality User Guide, Springer Verlag, (2015).
[23] A. Milton, and S. Irene Handbook of Mathematical Functions with Formulas, Graphs, and Mathematical
Tables, Dover, (1965).
m
[24] F. Olver, D. Lozier, R. Boisvert, and C. Clark, NIST Handbook of Mathematical Functions, Cambridge
University Press, (2010).
[25] E. Hille, Ordinary Differential Equations in the Complex Domain, Dover, (1976).
[26] A. Einstein Cosmological Considerations in the General Theory of Relativity, Sitzung der Physikalisch-
Mathematischen Klasse 8, 142, (1917).
[27] J. Polchinski Dirichlet-branes and ramond-ramond charges, Phys.Rev.Lett. 75, 4724, (1995), arXiv:hep-
th/9510017v3.
[28] G. t Hooft Dimensional reduction in quantum gravity, (1995), arXiv:gr-qc/9310026v2.
[29] L. Susskind The world as a hologram, J. Math. Phys. 36, 6377, (1995), arXiv:hep-th/9409089v2.
[30] H. Robertsoin Reletivistic Cosmology, Rev. Mod. Phys. 5, 62, (1933).
[31] J. Maldacena The Large N limit of superconformal field theories and supergravity, Adv. Theor. Math.
Phys. 2, 231, (1998), hep-th/9711200.
[32] E. Witten, Anti-de Sitter space and holography, Advances in Thl. and Math. Phys. 2, 253291arXiv:hep-
th/9802150.
[33] S. Gubser, I.R. Klebanov,and A.M. Polyakov, Gauge Theory Correlators from Non-Critical String The-
ory, arXiv:hep-th/9802109v2.
[34] A. Achucarro, and P. Townsend, A Chern-Simons Action For Three- Dimensional Anti-De Sitter Su-
pergravity Theories, Phys. Lett. B 180, 89 (1986).
[35] A. Achucarro, and P. Townsend, Quantization of Chern Gauge Group, Math. Phys. 137, 29 (1991).
[36] J. Polchinski, Dualities of Fields and Strings, (2014), arXiv:1412.5704v3 [hep-th]
[37] O. Aharony, S. Gubser, J. Maldacena, H. Ooguri, and Y. Oz, Large N field theories, string theory and
gravity, Phys. Rept. 323, 183, (2000).
[38] J. Maldacena, TASI 2003 Lectures on AdS/CFT, arXiv:hep-th/0309246v5.
[39] E. DHokera, and D. Freedmanb, Supersymmetric Gauge Theories and the AdS/CFT Correspondence,
arXiv:hep-th/0201253v2.
[40] J. Maldacena, The gauge/gravity duality, arXiv:1106.6073v2.
[41] K. Skenderis, Lecture notes on holographic renormalization Class.Quant.Grav. 19 5849 (2002).
[42] J. Boer, The Holographic Renormalization Group, Fortsch. Phys. 49, 339-358 (2001), arXiv:hep-
th/0101026v1
[43] E. Witten, Anti De Sitter Space And Holography, Adv.Theor.Math.Phys. 2 253 (1998), arXiv:hep-
th/9802150v2.
[44] S. Giddings, The boundary S-matrix and the AdS to CFT dictionary, Phys.Rev.Lett. 83 2707 (1999),
arXiv:hep-th/9903048v2
[45] J. Brown and M. Henneaux, Central Charges in the Canonical Realization of Asymptotic Symmetries:
An Example from Three-Dimensional Gravity, Commun. Math. Phys. 104, 207 (1986).
[46] J. Polchinski, Introduction to Gauge/Gravity Duality, arXiv:1010.6134v1 [hep-th] (2010).
[47] E, Witten, Bound States Of Strings And p-Branes, Nucl. Phys. B460 335-350, (1996) arXiv:hep-
th/9510135v2.
n
[48] P. Breitenlohner and D. Freedman, Stability in Gauged Extended Supegravity,Annals Phys. 144, 249,
(1982).
[49] L. Susskind, and E, Witten, The holographic bound in anti-de sitter space, (1998) arXiv:hep-
th/9805114v1.
[50] C. Burgess and C. Lutken, Propagators and Effective Potentials in Anti-de Sitter Space, Nucl. Phys.
B272, 661 (1986).
[51] K. Wilson, The renormalization group: critical phenomena and the Kondo problem, Rev. Mod. Phys.
47, 773, (1975).
[52] D. Freedman, S. Gubser, K. Pilch, and N. Warner, Renormalization group flows from holography super-
symmetry and a c theorem, Adv. Theor. Math. Phys. 3, 363-417 (1999) arXiv:hep-th/9904017v3.
[53] S. Hartnoll, Lectures on holographic methods for condensed matter physics, Class. Quant. Grav. 26,
224002 (2009) arXiv:0903.3246v3 [hep-th].
[54] S. Brodsky, and G. Teramond AdS/CFT and QCD, (2009) arXiv:hep-th/0702205v1.

[55] Y. Kim, Ch. Young Oh, N. Park, Classical Geometry of De Sitter Spacetime : An Introductory Review,
arXiv:hep-th/0212326v1.
[56] A. Strominger, The dS/CFT Correspondence, J. High Energy Phys., 0110, 034, (2001).
[57] L. Dyson, J. LindFpointesay, and L. Susskind, Is There Really a de Sitter/CFT Duality, J. High Energy
Phys. 0208, 045, (2002).
[58] D. Anninos, de Siter Musings, Int. J. Mod. Phys. A, 27, 1230013, (2012).
[59] I. Kanitscheider, K. Skenderis, and M. Taylor, Precision Holography for non-Conformal Branes, J. High
Energy Phys. 09, 094, (2008), http://iopscience.iop.org/1126-6708/2008/09/094
[60] D. Harlow, and D. Stanford, Operator Dictionaries and Wave Functions in AdS/CFT and dS/CFT,
arXiv:1104.2621v2 [hep-th].
[61] D. Anninos, Th. Hartman, and A. Strominger, Higher Spin Realization of the dS/CFT Correspondence,
arXiv:1108.5735v1 [hep-th].
[62] M. Vasiliev, Consistent equation for interacting gauge fields of all spins in (3+1)-dimensions, Phys. Lett.
B243, 378-382 (1990).
[63] M. Vasiliev, Higher spin gauge theories: Star product and AdS space, arXiv:hep-th/9910096v2.
[64] N. Boulanger, S. Leclercq and P. Sundell, On The Uniqueness of Minimal Coupling in Higher-Spin
Gauge Theory, J. High Energy Phys. 0808, 056, (2008), arXiv:0805.2764v2 [hep-th]
[65] F. Kristiansson and P. Rajan, Scalar field corrections to AdS(4) gravity from higher spin gauge theory,
J. High Energy Phys. 0304, 009, (2003), arXiv:hep-th/0303202v1
[66] S. Giombi and X. Yin, Higher Spin Gauge Theory and Holography: The Three-Point Functions, J. High
Energy Phys. 1009, 115, (2010), arXiv:0912.3462v4
[67] A. LeClair, and M. Neubert, Semi-Lorentz invariance, unitarity, and critical exponents of symplectic
fermion models, J. High Energy Phys. 0710, 027, (2007), arXiv:0705.4657 [hep-th]
[68] D. Robinson, E. Kapit, A. LeClair Lorentz Symmetric Quantum Field Theory for Symplectic Fermions,
J.Math.Phys.112301,50, (2009), arXiv:0903.2399v2 [hep-th]
[69] I. Klebanov and A. Polyakov, AdS dual of the critical O(N) vector model, Phys. Lett. B 550, 213, (2002),
o
[70] A. Mostafazadeh, PseudoHermiticity versus PT symmetry, J. Math. Phys. 43, 205, (2002), arXiv:math-
ph/0107001v3
[71] A. Mikhailov, Notes on higher spin symmetries, (2002), arXiv:hep-th/0201019v1
[72] E. Sezgin and P. Sundell, Massless higher spins and holography,Nucl. Phys. B 644, 303, (2002),
[73] T. Wiseman, Cosmology Lecture Notes, Cosmology and particle physics MSc course, (2014).
[74] V. Mukhanov, H. Feldmanc, and R. Brandenbergera, Theory of Cosmological Perturbations, Phys. Rep.
, 215, 213, (1992).
[75] A. Strominger, Inflation and the dS/CFT Correspondence, J. High Energy Phys. 0111, 049, (2001).
[76] P. McFadden, and K. Skenderis, The Holographic Universe, J. Phys. Conf. Ser. 222, 012007 (2010),
arXiv:1001.2007v1 [hep-th]
[77] P. McFadden, and K. Skenderis, Holography for Cosmology, Phys .Rev. D81, 021, (2010),
arXiv:0907.5542v2 [hep-th]
[78] I. Papadimitriou, K. Skenderis, Correlation functions in holographic RG flows, JCAP, 10, 075, (2004),
[79] R. Easther, R. Flauger, P. McFadden, and K. Skenderis, Constraining holographic inflation with WMAP,
JCAP, 2011, 1109-030, (2011), arXiv:1104.2040v2 [astro-ph.CO]
[80] N. Turok, A critical review of inflation, Class. Quant. Grav. 19, 3449-67, (2002), arXiv:0907.5542v2
[hep-th]
[81] R. Brandenberger, Inflationary cosmology: progress and problems, Class. Quant. Grav. 19, 3449-67,
(2002), arXiv:hep-ph/9910410v1
[82] E. Komatsu, J. Dunkley, M. Nolta, C. Bennett, B. Gold, G. Hinshaw, N. Jarosik, D. Larson, M. Limon,
L. Page, D. Spergel, M. Halpern, R. Hill, A. Kogut, S. Meyer, G. Tucker, J. Weiland, E. Wollack,
and E. Wright, Five-yearWilkinson Microwave Anisotropy Probe (WMAP) observations: cosmological,
Astrophys. J. Suppl. 180, 330-376, (2009), arXiv:0803.0547v2 [astro-ph]
[83] P. Ade et al., BICEP2 and Planck Collaborations, Phys. Rev. Lett. 114, 101301, (2015),
arXiv:1502.00612v2 [astro-ph.CO]
linktext

DS/CFT Duality in Cosmology: Ramtin Amintaheri

Uploaded by

Copyright:

Available Formats

DS/CFT Duality in Cosmology: Ramtin Amintaheri

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

DS/CFT Duality in Cosmology: Ramtin Amintaheri

Uploaded by

Copyright:

Available Formats

dS/CFT Duality

Supervisor: Dr Toby Wiseman

2 Conformal Field Theory 4

3 Anti de Sitter Space 11

A Anti de Sitter Equation of Motion a

B Anti-de Sitter Bulk-to-Boundary Propagator c

C Anti de Sitter Two-Point Function e

E De Sitter Two-Point Function i

F Holography for Cosmology k

Conformal Field Theory

2.1 Conformal Transformation

2.2 Conformal Group

x′µ = xµ + ǫµ (x) (2.4)

ǫµ (x) = aµ + bµν xν + cµνλ xν xλ (2.5)

Finite scaling can be written as:

and rotation matrix is expanded like Mνµ = δνµ + ω(x)µν + O(ω 2 )

This type of transformation will be studied in more details below.

2.2.2 Special Conformal Transformation

2.2.3 Conformal Algebra

3. D generator of scaling xµ → λxµ

[Mµν , Pρ ] = −i(ηµρ Pν − ηνρ Pµ ), [Mµν , Kρ ] = −i(ηµρ Kν − ηνρ Kµ ),

and all other commutations vanishes.

The conformal group on a Mp,q manifold is SO(p + 1, q + 1).

The usual commutation relation is defined as: [11]

2.3 Conformal Fields in Two Dimension

z → f (z), z̄ → f¯(z̄) (2.20)

Which implies metric will transform like:

which changes the metric to:

u → f (u), v → g(v) (2.24)

2.3.2 The Witt Algebra

The corresponding generators of infinitesimal transformation is thus given by: [10]

ln = −z n+1 ∂z , ¯ln = −z̄ n+1 ∂z̄ (2.25)

justifies our assumption about considering z and z̄ as two independent variables.

2.4 Conformal Fields

2.5 Correlation Functions

2.5.1 Two-Point Correlation Function

hO1 (x1 )O2 (x2 )i → hλ∆1 O1 (x1 )λ∆2 O2 (x2 )i

This amount to the following form:

Anti de Sitter Space

3.1 Constant Curvature Spaces

3.2 Geometry of Anti de Sitter Space

ds2 = −dT i dT i + dX j dX j i + j = 0, . . . d (3.6)

and finally cosmological constant will be: [3, 56]

3.3 Coordinate Systems

3.3.1 Global Coordinates

ds2 = l2 (− cosh2 ρ dτ 2 + dρ2 + sinh2 ρ dΩ2d−2 (3.14)

ds2 = −dτ 2 + dθ2 + sin2 θdΩ2(d−2) (3.15)

3.3.2 Poincare Coordinates

X 1 = (1/2u) 1 − u2 (l2 − xi xi + t2 ) , (3.16)

where u > 0, then the metric takes the form:

ds2 = dr2 + e2r/l ηµν (dxµ dxν ) µ, ν = 1, . . . d − 1 (3.19)

with usual definition of Laplacian:

Due to translation invariance on the x coordinates, we try separation of variables:

φ(z, x) = f (z)Φ(x) (3.23)

(∂x2 − k 2 )Φ(x) = 0 (3.24a)

z 2 f˜k′′ − (d − 2)z f˜k′ − (m2 l2 + k 2 z 2 )f˜k = 0 (3.26)

f˜k (z) = ak (kz)(d−1)/2 Kδ (kz) + bk (kz)(d−1)/2 Iδ (kz) (3.27)

f˜k ≈ φ̃− (k)z ∆− + φ̃+ (k)z ∆+ (3.28)

where ∆± are given by ∆± = (d − 1)/2 ± δ that is: