Remarks on the

gauge/gravity duality

Juan Maldacena
 Field Theory = Gravity theory

Gauge Theories Quantum Gravity

QCD String theory
 Large N and string theory

• SU(N) theory when N  ∞ , with g2 N = fixed ‘t Hooft 74

• Planar diagrams dominate
• View them as the worldsheet of a string

+ + ==

• 1/N corrections  non-planar diagrams = strings

worldsheets with more complicated topologies
Look for a string theory in 4d  not
consistent

Polyakov
At least 5 dimensions
 D-branes in string theory
• Solitons in string theory. Excitations Polchinski 95
described by open strings
• Lowest energies  U(N) gauge fields
• Can also be viewed as black branes Horowitz
Strominger 91

A ij

i j
 A ij

i j

N=4 SU(N) Yang Mills String theory

theory in 3+1 dimensions on AdS5 x S5
³ ´4
g2N = R
ls JM 97
Conformal
symmetry

AdS/CFT

Duality:
³ ´4
g2N = R
ls
 Development of the dictionary
Gubser, Klebanov,
• Correlation functions Polyakov, Witten
• Wilson loops
• Various deformations: masses, marginal
deformations
• Many new examples, both conformal and
non-conformal
• Continues today…
 Lessons for gravity
• Use the field theory as a definition of string
theory or quantum gravity (finite N).
• Lessons for black holes.
• Emergent space time
Black holes

Entropy = area =
statistical entropy
in the field theory

Unitary evolution:
Quantum mechanics
and gravity are
compatible
 Counting supersymmetric black
holes
• More detailed black hole counting,
subleading corrections
• AdS3/CFT2 examples
• Connections to the topological string
• Connections between matrix models, N=1
quantum field theories and geometryGopakumar, Vafa,
Dijkgraaf, Vafa
 Holography
• Physics in some region described by a
theory on the boundary of the region with a
number of q-bits that grows like the area of
the region in Planck units.

Not clear how to extend the idea to other cases

- What are the degrees of freedom
- How do we define a ``boundary’’
 Emergent space time

Spacetime: like the fermi surface,

only defined in the classical limit
 Is there a dS/CFT ?

Witten
Asymptotic future
Strominger
Euclidean conformal JM
field theory
=? De-Sitter

Initial singularity

Wave function of Partition function of a

the universe Ψ[g] = Z[g] Euclidean field theory

No explicit example is known! Objections: - dS decays

How do we get “emergent” time ? - dS is thermal
Description of the string landscape,
and eternal inflation?

What are the future

boundaries?

Who measures that?

• What are the field theory duals of the AdS4
vacua in the landscape (preserving N=1
susy) ?
 How do describe the interior of a
black hole
• Crunching cosmology
• Intrinsically approximate description, up e-S
• There is no obvious “exact” sense in which the
interior exists.
• Probably relevant for describing cosmology

• There is probably no other region behind the

singularity.
• Can we have crunch-bang transitions ?
 Lessons for gauge theories
• We can view the geometry as the solution of the
theory in the large ‘t Hooft coupling limit.
• View these as toy model field theories that can be
explicitly solved. These theories capture many
interesting physical phenomena.
• Explicit examples of confining theories
• Thermal aspects are calculable. Transport
properties can be computed. e.g. RHIC
applications or as toy model for condensed matter
problems.
• Could be describing physics at higher energies
(Randall-Sundrum models).
• Examples of constructions of metastable vacua in
supersymmetric theories. Use to understand
supersymmetry breaking, supersymmetry breaking
mediation mechanisms.
• Many examples of conformal N=1 SUSY gauge
theories with a geometric description.
• Used to test other dualities.
 All coupling
• Connect the weak and strong coupling
regimes by computing exactly for all
coupling.
• Integrability in N=4 SYM.
• Cusp anomalous dimension known exactly
for all values of g2 N.
 Future
• More on the dictionary
• Toy model for analyzing field theory
dynamics.
• Further exact results in N=4.
 Big challenges
• Understand how to describe universes with
cosmological singularities
• Clear description of the interior of black holes

• String theory describing the large N limit of

ordinary Yang Mills theory, or other large N
theories whose duals have stringy curvature.
 Gravitational potential
in the extra dimension

Graviton is
localized in
the extra dimension

Extra dimension Massive spin 2

4 dimensions
particle in 4 dimensions
Jet Physics at strong coupling

Diego Hofman and J. M., to appear

 Jet physics at strong coupling
q
e -

γ
e+ q

J ¹ (p)j0i
 Why?
• New physics at the LHC could involve a strong or
moderately strongly coupled field theory
• We would like to understand the transition
between the weak coupling picture of the event
where one can qualitatively think in terms of
underlying partons and the gravity picture where
we do not see the partons in any obvious way.

Weak coupling Strong couling

?
 How do we describe the produced
state
• Not convenient to talk about partons
• Inclusive observable Basham, Brown, Ellis, Love 78
• Energy correlation functions.
h²(µ1 )i
θ1
h²(µ1)²(µ2)i
h²(µ1 )¢¢¢²(µn )i
θ2
 Integrated flux of energy at infinity
R
²(µ) = dtT 0i ni
R
h²(µ)i = h0jj y T j j0i

Three point function

Symmetries determine the three point function

up to two constants
h²(µ)i = a +b(cos2 µ ¡ 1)
3
weak coupling in QCD: (unpolarized e-beam)

²(µ) » 1 + cos2 µ θ is angle to beam

In any theory with a gravity dual:

²(µ) » 1

In N=4 SYM at weak and strong coupling:

²(µ) » 1

N=1 superconformal field theory, j = R-current:

3a ¡ c 2 1
²(µ) » 1 + (cos µ ¡ )
2 c 3
 Can we get a non-zero b from AdS ?

Bulk couplings between an on shell graviton

and two on shell gluons
R
S5¡ d » aF 2 +bR ¹ º±¾F ¹ º F ±¾

- Only two possible couplings in 5 dimensions

- a is fixed by the two point function of the currents
- The second coupling is a higher derivative
correction (which is present in bosonic string theory)
- If a ~ b  higher derivative correction as important
as the leading term)
 Two point function

Consider a state produced by a scalar operator.

Two point function is a somewhat complicated

function of the angle between the two detectors

At very weak coupling is goes like

h²(µ1)²(µ2)i » g2N µ12 ; µ12 ¿ 1
12

µ12
 Strong coupling computation

Since we integrate the stress

x tensor  wavefunction localized on
x an H3 subspace (SO(1,3) symmetry)
x
x
g(x,¢ )
R
h²(µ1)²(µ2)i = H3
ª ¤@¡2 ª f (µ1; x)f (µ2; x)
 General description
1/Δ

H3

²(µ; x) = f (µ; x)

In the gravity approximation the distribution of

energy depends just on three random variables:
a point on H3
 Small angle singularity
h²(µ1)²(µ2)i » g2N 4¡12¢
µ 12

The small angle singularity is governed by the

twist of the operators contributing to the light
cone OPE of the stress tensor
Twist = Δ - S
R R R
dx¡ T¡ ¡ y) dx¡ T¡
(~ ¡ (0) » y¡ 4+¿n dx¡ O¡n ¡ ¡

Balitsky Braun 88 Non-local operator of spin 3

 Single trace and double trace operators.

Double trace operators of the form

O¢ O¢
are the dominant contribution at strong coupling

At weak coupling single trace operators

have dimension one but at strong coupling
they get large dimension

¿3 = 2 + ¸ +¢¢¢ for ¸ ¿ 1
¿3 = 21=2¸ 1=4 ; for ¸ À 1
 Polchinski Strassler 02
Related to deep inelastic scattering

It is related to a particular moment of the deep

inelastic amplitude of gravitons.
R1
h²(¡ q)²(q)i » ¡ 1 dsA D I S (s;q2)
 Conclusions
• One can define inclusive, event shape variables for
conformal theories
• They can be computed at weak and strong coupling
• Small angle features governed by the anomalous
dimension of twist two operators
• Events are more spherically symmetric at strong coupling,
but not completely uniform.
• In a non-conformal theory one would need to face the
details of hadronization. Depending on these details there
could be small or large changes.
 Strings and Strong Interactions
Before 60s  proton, neutron  elementary
During 60s  many new strongly interacting particles
Many had higher spins s = 2, 3, 4 ….
All these particles  different oscillation modes of a string.

This model explained features of the spectrum

of mesons.

Rotating String model

m 2 ~ TJ max + const From E. Klempt hep-ex/0101031

Strong Interactions from Quantum ChromoDynamics
3 colors (charges)
Experiments at higher energies
revealed quarks and gluons They interact exchanging gluons

Electrodynamics Chromodynamics (QCD)

photon
g g
g g gluon
electron

Gauge group 3 x 3 matrices

Gluons carry
U(1) color charge, so
SU(3)
they interact among
themselves
Coupling constant decreases at high energy Gross, Politzer, Wilczek

g 0
at high energies QCD is easier to study at high energies

Hard to study at low energies

Indeed, at low energies we expect to see confinement

V = T L
q q
Flux tubes of color field = glue

At low energies we have something that looks like a string. There are
approximate phenomenological models in terms of strings.

How do strings emerge from QCD

Can we have an effective low energy theory in terms of strings ?
 Large N and strings
Gluon: color and anti-color
Take N colors instead of 3, SU(N)
t’ Hooft ‘74

Large N limit

g2N = effective interaction strength

when colors are correlated

Open strings  mesons

Closed strings  glueballs
General Idea

- Solve first the N=∞ theory.

- Then do an expansion in 1/N.

- Set 1/N =1/3 in that expansion.

 The N=∞ case

- It is supposed to be a string theory

- Try to guess the correct string theory

- Two problems are encountered.

1. Simplest action = Area

Not consistent in D=4 ( D=26 ? ) Lovelace

generate Polyakov
At least one more dimension (thickness)

2. Strings theories always contain a state with m=0, spin =2:

a Graviton.

But:
- In QCD there are no massless particles. Scherk-Schwarz
- This particle has the interactions of gravity Yoneya

For this reason strings are commonly used to study

quantum gravity. Forget about QCD and use strings as a theory of
quantum gravity. Superstring theory, unification, etc.

But what kind of string theory should describe QCD ?

We need to find the appropriate 5 dimensional geometry

It should solve the equations of string theory

They are a kind of extension of Einstein’s equations

Very difficult so solve

Consider a simpler case first. A case with more symmetry.

We consider a version of QCD with more symmetries.

 Most supersymmetric QCD

Supersymmetry
Ramond
Bosons Fermions Wess, Zumino

Gluon Gluino
Many supersymmetries
B1 F1
B2 F2

Maximum 4 supersymmetries, N = 4 Super Yang Mills

Susy might be present in the real world but spontaneously broken at low energies.
So it is interesting in its own right to understand supersymmetric theories.

We study this case because it is simpler.

Similar in spirit to QCD

Difference: most SUSY QCD is scale invariant

Classical electromagnetism is scale invariant
V = 1/r
QCD is scale invariant classically but not quantum mechanically, g(E)

Most susy QCD is scale invariant even quantum mechanically

Symmetry group

Lorentz + translations + scale transformations + other

These symmetries constrain the shape of the five dimensional space.

ds2 = R2 w2 (z) ( dx23+1 + dz2 )

redshift factor = warp factor ~ gravitational potential

Demanding that the metric is symmetric under scale transformations

x  λ x , we find that w(z) = 1/z
 ds2 = R2 (dx23+1 + dz2)
z2

R4

AdS5
Boundary

z
z=0 z = infinity

This metric is called anti-de-sitter space. It has constant negative

curvature, with a radius of curvature given by R.
w(z) Gravitational potential

z
(The gravitational potential does not have a minimum  can have massless excitations
Scale invariant theory  no scale to set the mass )
 Anti de Sitter space

Solution of Einstein’s equations with negative cosmological constant.

( De Sitter  solution with positive cosmological constant, accelerated expanding

universe )
Two dimensional
negatively curved space
Spatial section of AdS = Hyperbolic space
R = radius of curvature

Time

Light rays Massive particles

The space has a boundary.

It is infinitely far in spatial distance

A light ray can go to the boundary and back in finite time, as seen from
an observer in the interior. The time it takes is proportional to R.

The Field theory is defined on the boundary of AdS.

 Building up the Dictionary
Graviton stress tensor

Τµν
Gubser, Klebanov,
Τµν Polyakov - Witten

Τµν

< Τµν(x) Τµν(y) Τµν(z) >Field theory = Probability amplitude that gravitons
go between given points on the boundary
Other operators
Other fields (particles) propagating in AdS.
Mass of the particle scaling dimension of the operator

∆ = 2 + 4 + (mR) 2
 Most supersymmetric QCD

We expected to have string theory on AdS.

5
Supersymmetry D=10 superstring theory on AdS5 x (something)

S5
Type IIB superstrings on AdS 5x S 5 (J. Schwarz)

5-form field strength F = generalized magnetic field quantized

∫S5
F =N
 Veneziano
String Theory Scherk
Schwarz
Green
Free strings …..

1
String Tension = T = 2 , ls = string length
ls

Relativistic, so T = (mass)/(unit length)

Excitations along a stretched string travel at the speed of light

Closed strings

Can oscillate Normal modes Quantized energy levels

Mass of the object = total energy

ls
M=0 states include a graviton (a spin 2 particle)
First massive state has M 2 ~ T
String Interactions
Splitting and joining

String theory Feynman diagram

g

Simplest case: Flat 10 dimensions and supersymmetric

Precise rules, finite results, constrained mathematical structure

At low energies, energies smaller than the mass of the first massive string state

ls
Gravity theory
R

Radius of curvature >> string length  gravity is a good approximation

( Incorporates gauge interactions  Unification )

 Particle theory = gravity theory

Most supersymmetry QCD String theory on (J.M.)

theory = AdS5 x S 5

N colors N = magnetic flux through S5

Radius of curvature RS 5 = RAdS5 = g ( 2

YM N )
1/ 4
ls

Duality:
g2 N is small  perturbation theory is easy – gravity is bad

g2 N is large  gravity is good – perturbation theory is hard

Strings made with gluons become fundamental strings.

Where Do the Extra Dimensions Come From?
3+1  AdS5  radial dimension

Strings live here

Interior Gluons live here

z Boundary
 What about the S5 ?

• Related to the 6 scalars

• S5  other manifolds = Most susy QCD  less susy
QCD.
• Large number of examples

Klebanov, Witten,
Gauntlett, Martelli, Sparks,
Hannany, Franco, Benvenutti,
Tachikawa, Yau …..
 Quark anti quark potential

string q V = potential = proper length of the string

in AdS
Boundary
g2N
V ≈−
L
q

g2N
Weak coupling result: V ≈−
L
 Confining Theories
Add masses to scalars and fermions  pure Yang Mills at low energies
 confining theory. There are many concrete examples.
At strong coupling  gravity solution is a good description.

w(z) Gravitational potential

boundary or warp factor

z0 z
w(z0) > 0
String at z0 has finite tension from the point of view of the boundary
theory.
Graviton in the interior  massive spin=2 particle in the boundary theory
= glueball.
 Checking the conjecture
• It is hard because either one side is strongly
coupled or the other.
• Supersymmetry allows many checks. Quantities
that do not depend on the coupling.
• More recently, ``integrability’’ allowed to check
the conjecture for quantities that have a non-trivial
dependence on the coupling, g2N.
• One can vividly see how the gluons that live in
four dimensions link up to produce strings that
move in ten dimensions. …

Minahan, Zarembo, Beisert, Staudacher, Arutyunov, Frolov, Hernandez, Lopez, Eden

The relation connects a quantum field theory to gravity.

What can we learn about gravity

from the field theory ?
• Useful for understanding quantum aspects
of black holes
 Black holes
Gravitational collapse leads to black holes

Classically nothing can escape once it crosses the event horizon

Quantum mechanics implies that black holes emit thermal

radiation. (Hawking)
1 1 M 
T≈ ≈ T ≈ 10 −8 K  sun 
rs GN M  M 
Black holes evaporate 3
 M 
Evaporation time τ = τ universe  12 
 10 Kg 
Temperature is related to entropy
Area of the horizon (Hawking-Bekenstein)
dM = T dS S= 2
4 L Planck

What is the statistical interpretation of this entropy?

 Black holes in AdS
Thermal configurations in AdS.
Entropy:

SGRAVITY = Area of the horizon =

SFIELD THEORY =
Log[ Number of states]

Evolution: Unitary

Solve the information paradox raised by S. Hawking

 Confining Theories and Black Holes

Low temperatures High temperatures

Deconfinement=
Confinement black hole (black brane)
Gravitational
potential

Horizon

Extra dimension
Extra dimension
 Black holes in the Laboratory

QCD  5d string theory

z=0 z=z0 ,
High energy collision produces a black hole =
droplet of deconfined phase ~
quark gluon plasma .
Black hole Very low shear viscosity
similar to what is observed at RHIC:
“ the most perfect fluid”
z=0 z=z0 Kovtun, Son, Starinets, Policastro
Very rough model, we do not yet know the precise string theory
 Emergent space time

Spacetime: like the fermi surface,

only defined in the classical limit
Lin, Lunin, J.M.
 A theory of some universe
• Suppose that we lived in anti-de-sitter space
• Then the ultimate description of the
universe would be in terms of a 2+1
dimensional field theory living on the
sphere at infinity. (With around 10120 fields
to give a universe of the size of ours)
• Out universe is close to de-Sitter. Could we
have a similar description in that case ?
 Conclusions
- Gravity and particle physics are “unified”
Usual: Quantum gravity  particle physics.
New: Particle physics quantum gravity.

- Black holes and confinement are related

- Emergent space-time. Started from a theory
without gravity  got a theory in higher
dimensions with gravity.
- Tool to do computations in gauge theories.
- Tool to do computations in gravity.
 Future
Field theory:
Theories closer to the theory of strong interactions

Solve large N QCD

Gravity:
Quantum gravity in other spacetimes

 Understand cosmological singularities