See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/239928060

The Randall-Sundrum Model

Article · January 2006

CITATIONS READS

8 1,664

1 author:

Maxime Gabella
Institute for Advanced Study
25 PUBLICATIONS   449 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

spectral networks and bps quivers View project

All content following this page was uploaded by Maxime Gabella on 26 January 2017.

The user has requested enhancement of the downloaded file.

 The Randall-Sundrum Model
Maxime Gabella

June 2006

IPPC, EPFL

Abstract
The Randall-Sundrum model was conceived in 1999 to address the
Higgs Hierarchy Problem in particle physics. It arose enormous inter-
est from theoreticians and phenomenologists ever since and revealed a
fruitful tool to explore the physics of extra dimensions. The aim of this
paper is to provide an introductory exposition of this model. After a
short survey of Kaluza-Klein theories, the setup of the RS model will
be exposed and its metric derived. We will explain how an exponential
hierarchy between the gravity scale and the weak scale can be natu-
rally generated, and how the standard 4D gravity emerges from this
model in the Newtonian limit. The Golberger-Wise mechanism will
be presented as a way to stabilize the radius of the extra dimension
without reintroducing a fine-tuning. Those topics will be presented in
an utterly pedagogical way. Here you will find what textbooks feel free
to disregard as too advanced but research papers consider as too basic
to even be mentioned.

1
Contents
1 Basics of Kaluza-Klein theories 3

2 Setup 5

3 Warped metric 5

4 Exponential hierarchy 9

5 Graviton modes 11

6 Graviton spectrum 17

7 Newtonian limit 18

8 Radius stabilization 21

A Einstein tensor 25

2
1 Basics of Kaluza-Klein theories
The existence of extra dimensions of space was first put forth in the middle
of the 1920’s by Theodor Kaluza and Oskar Klein [1] as a means of unifying
the electromagnetic and gravitational fields as components of a single higher-
dimensional field. As an illustration, consider the case of a five-dimensional
theory, with the extra dimension periodically identified:

x5 ∼ x5 + 2πR.

This procedure is called toroidal compactification [2]. The space obtained

is the product of the traditional four-dimensional Minkowski space with a
circle, noted M 4 ⊗ S 1 , which can be imagined as a 5D cylinder of radius R.
In such a theory, a massless scalar field φ(xµ , x5 ) would have a quantized
momentum in the periodic dimension:
n
p5 = ,
R
with n ∈ Z. We may then expand the field in Fourier series:
∞
X n 5
φ(xµ , x5 ) = φn (xµ )ei R x .
n=−∞

With this decomposition, the five-dimensional equation of motions (∂µ ∂ µ +

∂5 ∂ 5 )φ = 0 becomes

n2 n µ
∂µ ∂ µ φn (xµ ) = φ (x ).
R2
In this way, an infinite tower of fields with masses m2 = n2 /R2 is generated.
At energies small compared to R−1 , only the x5 -independent massless zero-
mode remains and the physics is effectively four-dimensional. At energies
above R−1 , the tower of Kaluza-Klein (KK) states comes into play.
An experimental bound on the size of the compactification radius R is
imposed by the fact that those KK states have not been detected at colliders
up to TeV energies. Their masses would thus have to be greater, n/R >TeV,
which implies a strong constraint on R:

R . 10−21 cm.

It is nearly hopeless to seek experimental confirmations of such minuscule

dimensions.

A way out of this restriction was suggested in 1998 by Arkani-Hamed,

Dimopoulos and Dvali (ADD) [3], based on an idea formulated in 1983 by
Rubakov and Shaposhnikov [4]. If the extra dimensions are accessible only

3
 +y

φ πR
0 πR 0
~ 2πR ~ ­πR

­y

L
y
0 πR

Figure 1: S 1 /Z2 orbifold.

to gravity and not to the SM fields, the bound on their size is fixed by
experimental tests of Newton’s law of gravitation, which has only been led
down to about a millimeter:

R . 1mm.

Such large extra dimensions could then perfectly exist and nevertheless have
escaped our vigilance so far!
In addition, this scenario provides a solution to one of the central prob-
lems of particle physics: the Hierarchy Problem. This problem arises in
quantum field theory because of the quadratically divergent corrections to
the Higgs field mass, which require an incredible fine-tuning in order to get
the expected mass of a few hundreds GeV. This problem can be equivalently
formulated in terms of the unnatural discrepancy between the strength of
gravity and those of the other three forces. In the ADD scenario, the weak-
ness of gravity compared to the other forces finds an explanation in the fact
that gravity gets diluted in the large volume of the extra dimensions. The
hierarchy between the four-dimensional Planck scale MP l ' 1019 GeV and
the scale of weak interactions MW 'TeV would in reality be only apparent.
However, this solution merely translates the Hierarchy Problem into the
problem of the discrepancy between the large size of the extra dimensions
R ' 1mm and their natural value R ' lP l ' 10−33 cm.

The model presented in [5] and [6] by Lisa Randall and Raman Sundrum
in 1999 provides a new explanation of the Hierarchy Problem.

4
2 Setup
The Randall-Sundrum model assumes the existence of one extra dimension
compactified on a circle whose upper and lower halves are identified (see fig.
1).
Formally, this means we work in S 1 /Z2 orbifold, where S 1 is the one-
dimensional sphere (i.e. the circle) and Z2 is the multiplicative group
{−1, 1}. This construction entails two fixed points, one at the origin y = 0
and one at the other extremity of the circle, at y = πR ≡ L. On each of
these boundaries stands a four-dimensional world like the one we live in. By
analogy with membranes enclosing a volume, these worlds with 3+1 dimen-
sions enclosing the 5D bulk have been called 3-branes. The picture is then
two 3-branes, at a distance L one from another, enclosing a 5D bulk (cf. fig.
2).
Taking into account the 5D cosmological constant Λ (which unlike the
effective 4D cosmological constant does not need to be vanishing or even
small) the fundamental action is the sum of the Hilbert-Einstein action SH
and a matter part SM :
Z +L
√
Z
S = SH + SM = d 4 x dy −g(M 3 R − Λ), (1)
−L

where M is the fundamental 5D mass scale, R the 5D Ricci scalar and g the
determinant of the metric, whose explicit form will be investigated in the
next section.

3 Warped metric
The first step is to find the metric for such a setup. Since we are looking
for solutions to the 5D Einstein equations that might fit the real world, we
require that the metric should preserve Poincaré invariance: the 4D universe
derived from this theory should appear flat and static. This leads to the
following Ansatz:

ds2 = e−2A(y) ηµν dxµ dxν + dy 2 , (2)

where ηµν = diag(−1, 1, 1, 1) is the 4D Minkowski metric. The prefactor

e−2A(y) , called the warp factor, is written as an exponential for convenience.
Its dependence on the extra dimension coordinate y causes this metric to
be non-factorisable, which means that, unlike the metrics appearing in the
usual Kaluza-Klein scenarios, it cannot be expressed as a product of the 4D
Minkowski metric and a manifold of extra dimensions. To determine the
function A(y), we have to calculate the 5D Einstein equations:
1
GM N = RM N − gM N R = κ2 TM N ,
2

5
 (1) (2)

5D bulk

3­brane 3­brane

Λ
4D
xν
xµ
y
0 L

Figure 2: Randall-Sundrum setup.

where the capital indices M and N take the values 0, 1, 2, 3 and 5, i.e.
M = (µ, 5) with µ the usual 4D Lorentz index, and so on. The 5D Newton
constant is defined as
1
κ2 ≡ ,
2M 3
and the energy-momentum tensor as
−2 δSM
TM N = √ , (3)
−g δg M N
√
so that a term in the action like −gV with V constant corresponds to
an energy-momentum tensor equal to V gM N . The Einstein tensors for the
metric parametrized by the Ansatz (2) are worked out in appendix A. The
55 component of the Einstein equation gives
−Λ
G55 = 6A02 = .
2M 3
Notice that a real solution for A only exists if the 5D cosmological con-
stant Λ is negative, which means that the space between the branes is anti-de
Sitter space, noted AdS5 . The case where A is purely imaginary corresponds
to an oscillating warp factor, which is not the concern of the RS model.
From that equation, we see that A02 is equal to a constant, which we call
2
k :
−Λ
A02 = ≡ k2 . (4)
12M 3

6
Integrating over y gets us the expression for A:

A(y) = ±ky.

As we want a solution that respects the orbifold symmetry, i.e. invariance

under the transformation y → −y, we choose

A(y) = k|y|.

Finally, the background metric in the Randall-Sundrum model is paramet-

rized by

ds2 = e−2k|y| ηµν dxµ dxν + dy 2 , (5)

with −L ≤ y ≤ L.
Let us look now at the µν component of the 5D Einstein equations.
Appendix A gives

Gµν = (6A02 − 3A00 )gµν .

From the solution we just found for A we see that the first derivative of A
is

A0 = sgn(y)k.

The term sgn(y) may be written as a combination of Heaviside functions as

sgn(y) = θ(y) − θ(−y),

so the second derivative is

A00 = 2kδ(y).

This delta function arose from the kink of A at the origin y = 0 (cf. fig. 3).
In the same way, the kink at y = L gives rise to another delta function, and
the complete expression for A00 is

A00 = 2k δ(y) − δ(y − L) .

Plugging those results into the expression of the Einstein tensor gives

Gµν = 6k 2 gµν − 6k δ(y) − δ(y − L) gµν .

The first term is equal to the µν components of the energy-momentum tensor

multiplied by the 5D Newton constant:
−Λ
κ2 Tµν = gµν = 6k 2 gµν .
2M 3

7
 A(y) A’(y) A’’(y)

L
y
0 L 0 L 0

Figure 3: The function A(y) and its first and second derivatives.

The second term however seems to have nothing to be matched to. The
resolution of this situation is to take into account the energy densities of
the branes themselves, called brane tensions. This is done by adding to the
action one term for each brane, corresponding to the brane tensions λ1 and
λ2 :
√ √
Z Z
S1 = − d4 x −g1 λ1 = − d4 xdy −gλ1 δ(y),
√ √
Z Z
S2 = − d4 x −g2 λ2 = − d4 xdy −gλ2 δ(y − L). (6)

The terms g1 and g2 stand for the determinants of the metrics induced on
the first brane and on the second brane respectively. The induced metrics
define distances along the branes:

ds2 = gµν
i
dxµ dxν
= gµν (x, yi )dxµ dxν ,

with i = 1, 2 and y1 = 0, y2 = L. Notice that with the metric given by (5),

g1 = gδ(y) and g2 = gδ(y − L) because g55 = 1.
In order to satisfy the Einstein equations we need to impose the relation

λ1 = −λ2 = 12kM 3 . (7)

Moreover, by the definition of k we have

λ21
Λ=− .
12M 3
Those two relations are consequences of the requirement that the 4D universe
be flat and static. The 4D brane sources are balanced by the 5D bulk
cosmological constant in order to get a vanishing effective 4D cosmological
constant.

8
4 Exponential hierarchy
Having presented the setup and found the metric of the Randall-Sundrum
model, we would like to investigate what the physical scales would be if,
in the spirit of [4], the matter fields were confined on the second brane at
y = L. Consider the Higgs scalar field with the action
√ h
Z

2 i
SHiggs = d4 x g2 g2µν Dµ H † Dν H − λ H † H − v 2
Z h
2 i
= d4 xe−4kL e2kL η µν Dµ H † Dν H − λ H † H − v 2 .

To get a canonically normalized action we redefine the Higgs field as H =

ekL H̃. The action becomes
Z h
2 i
SHiggs = d4 x η µν Dµ H̃ † Dν H̃ − λ H̃ † H̃ − (e−kL v)2 .

This is the action of a normal Higgs scalar, except for the vacuum expecta-
tion value (vev) which is exponentially suppressed:

vef f = e−kL v.

As the Higgs vev sets all the mass parameters in the Standard Model, this
means that all mass parameters are submitted to an exponential suppression
on the second brane. If the value of the bare Higgs mass is of the order of
the Planck scale, the physical Higgs mass could be warped down to the weak
scale, where we expect it to be. For this reason, the first brane at y = 0
is often called the “Planck” brane, whereas the second brane is called the
“TeV” brane. Since MW ' 10−16 MP l , the appropriate value for the size of
the extra dimension is given by

kL ' ln1016 ' 35.

We will see in section 8 how such a value can be obtained for the size L of
the extra dimension without reintroducing a fine-tuning.
To understand whether or not this exponential suppression is useful to
address the Hierarchy Problem, we must know how the effective scale of
gravity behaves with respect to the extra dimension. This information is
to be obtained from the way the 5D action S contains the 4D action S 4D .
Perturbating the action 1 around the background metric given by 5 produces
a term with the schematic form
Z Z +L q
3 4 −2k|y|
S 3 M d x dye −g (0) R4D (h(0)
µν )
−L
− −2kL Z
31 e
q
= M d x −g (0) R4D (h(0)
4
µν ).
k

9
 Planck M ~ MPl

hierarchy
veff = exp(­ky) v
hidden

TeV

y
0 L

Figure 4: The generation of an exponential hierarchy.

This term corresponds to the 4D action, so that we can read off the value
of the effective 4D Planck mass:

MP2 l = (1 − e−2kL )M 3 /k.

We see that it weakly depends on the size of the extra dimension L, provided
kL is moderately large.
Putting our two last results together, we see that the weak scale is ex-
ponentially suppressed along the extra dimension, while the gravity scale is
mostly independent of it (see fig.4).
In conclusion, in a theory where the values of all the bare parameters
(M, Λ, λ1 , v) are determined by the Planck scale, an exponential hierarchy
can be naturally generated between the weak and the gravity scales. Thus
the Randall-Sundrum model provides an original solution to the Hierarchy
Problem.

Remarkably, the effective Planck mass remains finite even if we take the
decompactification limit L → ∞. This case where there is only one brane is
known as the Randall-Sundrum II model (RS2). The fact that there could
be an infinite extra dimension and still a 4D gravity as we experience it
results from the localization of gravity around the brane at y = 0, which we
now turn our attention to.

10
5 Graviton modes
In order to understand how gravity works in the Randall-Sundrum model,
we first have to find explicit expressions for the gravitons, which correspond
to small fluctuations hM N (x, y) around the background metric given by
ds2 = e−2k|y| ηµν dxµ dxν + dy 2 .
That will be achieved by computing the solutions of the linearized Einstein
equation.

Conformally flat metric It is convenient to work with a conformally flat

metric, i.e. a metric proportional to flat space. To achieve this, we define a
new extra dimension variable z related to y through
dy 2 ≡ e−2k|y| dz 2 .
The integration of this equation produces a constant, which we set so as to
have the zero value of y corresponding to the zero value of z. The result is
k|z| = ek|y| − 1, (8)
and thus
1
e−2k|y| = .
(k|z| + 1)2
With this new coordinate, the metric is given by
1
ds2 = (ηµν dxµ dxν + dz 2 ).
(k|z| + 1)2
To underline the fact that it is conformally flat we rewrite it in the following
way:

ds2 = e−2A(z) ηM N dxM dxN ,

where we use the notation x5 = z. The function A(z) is given by

1
e−2A(z) = ,
(k|z| + 1)2
and so A(z) = ln(k|z| + 1). For later reference, we give its first and second
derivatives:
sgn(z)k
A0 (z) = , (9)
k|z| + 1
and
2k(δ(z) − δ(z − Lz )) k2
A00 (z) = − , (10)
k|z| + 1 (k|z| + 1)2
where we have used again that sgn0 (z) = 2δ(z).

11
Linearized Einstein equations To keep the calculations as concise as
possible we will not compute the Einstein tensor by brute force, but rather
use a formula about conformally related metrics (see [7] appendix D). Specif-
ically, if some metric gM N is a conformal transformation of another metric
g̃M N , for example

gM N = e−2A g̃M N ,

then the respective Einstein tensors are related by

˜ M A∇
˜ NA + ∇˜ M∇ ˜ NA

GM N (gM N ) = G̃M N (g̃M N ) + (n − 2) ∇
˜ RA − n − 3 ∇
˜ R∇ ˜ R A∇˜ R A) ,

−g̃M N (∇
2
where n is the number of spacetime dimensions. In the present case, the
perturbed metric has the form

gM N = e−2A (ηM N + hM N )

and n = 5, so the formula gives, taking into account the Christoffel symbols
contained inside the covariant derivatives:

GM N = G̃M N + 3 ∂M A∂N A + ∂M ∂N A − Γ̃R


M N ∂R A
−g̃M N (∂R ∂ A − Γ̃RS ∂ A − ∂R A∂ R A) .
R R S

(11)

To linear order, the Christoffel symbols are easily found:

1
Γ̃R R R R
M N = (∂M hN + ∂N hM − ∂ hM N ),
2
where we have used η M N to rise the indices. It is particularly convenient in
that kind of calculations to work with a gauge in which the fluctuations do
not have any extra dimension component and are transverse and traceless:

hM 5 = 0,
∂µ hµν = 0 and η µν hµν = hµµ = 0.

We verify that those 10 conditions restrict the number of degrees of freedom

of the symmetric 5 × 5 tensor hM N from 15 to 5, as appropriate for a spin-
two 5D particle (see [8] section 2.3 and [9] section 10.6). With this gauge
fixing, the second Christoffel symbol in equation (11) vanishes, whereas the
first one reduces to −∂ 5 hM N /2 given that it is contracted with ∂R A, whose
only non-vanishing component is ∂5 A. In addition, the expression of the
Einstein tensor for fluctuations around the flat metric (see [7] eq. (4.4.5))
shrinks to
1
G̃M N = − ∂R ∂ R hM N .
2

12
The µν component of the linearized Einstein tensor in this gauge is then
given by

1 3
Gµν = − ∂R ∂ R hµν + h0µν A0 − 3(ηµν + hµν )(A00 − A02 ). (12)
2 2

On the other hand, we have to compute the energy-momentum tensor

for the perturbed metric. Coming back to the expression of the action terms
for the brane tensions (6), we have to be careful about the fact that with
the conformally flat metric the determinants of the metrics induced on the
branes are now related to the determinant of the full metric by

g = gi g55 = gi e−2A(zi ) ,

with i = 1, 2 and z1 = 0, z2 = Lz . The corresponding actions read

√ √
Z Z
4
S1 = − d x −g1 λ1 = − d4 xdz −gλ1 eA(0) δ(z)
√ √
Z Z
4
S2 = − d x −g2 λ2 = − d4 xdz −gλ2 eA(Lz ) δ(z − Lz ).

The µν component of the energy-momentum tensor multiplied by the 5D

Newton constant is then

1
κ2 Tµν −Λ − λ1 eA δ(z) − λ2 eA δ(z − Lz ) gµν
 
=
2M 3
1
−Λe−2A − λ1 e−A δ(z) − λ2 e−A δ(z − Lz ) (ηµν + hµν ).
 
=
2M 3

Remembering the definition (4) of k as well as the relation (7) between the
brane-tensions and referring to the expressions (9) and (10) of the first and
second derivatives of A allows us to rewrite it as
 2 −2A
κ2 Tµν − 6k δ(z) − δ(z − Lz ) eA (ηµν + hµν )


= 6k e
= 6A02 − 3(A00 + A02 ) (ηµν + hµν )
 

= (3A02 − 3A00 )(ηµν + hµν ). (13)

When we put the two sides (12) and (13) of the µν component of the
linearized Einstein equation together, the terms proportional to ηµν repro-
duce the unperturbed Einstein equations, and we are left with the part due
to the perturbations:

− 12 ∂R ∂ R hµν + 23 A0 h0µν = 0.

13
Schrödinger-like equation An elegant way of solving this equation is to
rewrite it in the form of a Schrödinger equation.
As a start, in order to get rid of the first derivatives h0µν , we make the
following rescaling:

hµν → eαA hµν ,

with α a constant. A pencil and a small piece of scratch paper bring the
Einstein equations to
    
1 R 3 0 0 3 1 2 02 1 00
− ∂R ∂ hµν + − α A hµν + α − α A − αA hµν = 0.
2 2 2 2 2

For the choice α = 3/2, the coefficient of h0µν vanishes and we are left with
 
1 R 9 02 3 00
− ∂R ∂ hµν + A − A hµν = 0.
2 8 4

Performing a Kaluza-Klein decomposition,

∞
X
hµν (x, z) = hnµν (x)ψn (z),
n=0

with hnµν ≡ ∂ρ ∂ ρ hnµν = m2n hnµν , we get

−ψn00 (z) + 02 − 32 A00 (z) ψn (z) = m2n ψn (z).

9 
4 A (z) (14)

That looks just like a Schrödinger equation with potential

9 02 3
V (z) = A (z) − A00 (z)
4 2 !


9 k 2 3 2k δ(z) − δ(z − Lz ) k2
= − −
4 (k|z| + 1)2 2 k|z| + 1 (k|z| + 1)2


15 k2 3k δ(z) − δ(z − Lz )
= 2
− .
4 (k|z| + 1) k|z| + 1

The shape of this potential looks like a volcano (see fig. 5).

Boundary conditions To get the boundary conditions that the solutions

will have to obey, we integrate equation (14) over small domains around the
boundaries. For the boundary at z = 0 we get
Z 0+ Z 0+
dz(−ψn00 + V ψn ) = dzm2 ψn
0− 0−
−ψn0 (0+ ) + ψn0 (0− ) − 3kψn (0) = 0.

14
 V(z)

z
­L 0 L

Figure 5: Volcano potential.

The wave-function has to be an even function under the transformation

z → −z, and so its first derivative is an odd function: ψn0 (0− ) = −ψn0 (0+ ).
The boundary condition at the Planck brane is then
3k
ψn0 (0) = − ψn (0). (15)
2
Similarly, we get the boundary condition at the TeV brane:
3k
ψn0 (Lz ) = − ψn (Lz ). (16)
2(kLz + 1)

Zero-mode The zero-mode is the solution of the Schrödinger-like equation

with m0 = 0:
 
00 9 02 3 00
−ψ0 + A − A ψ0 = 0.
4 2

It is given by
3
ψ0 (z) = e− 2 A = (k|z| + 1)−3/2 ,

which satisfies the boundary conditions (15) and (16). We see that the
graviton zero-mode has a wave function that is peaked around the origin (cf.
fig. 5). As we are going to see in section 7, the gravitational interactions
are predominantly mediated by the graviton zero-mode. Gravity is thus
localized on the Planck brane, while on the TeV brane we feel only the
tail of the graviton wave-function. So in the RS model the reason of the

15
 ψ(0)(z)

z
0

Figure 6: Localization of the graviton zero-mode around the Planck brane.

weakness of gravity is that it is localized far away from where we live — in

contrast to the ADD scenario, which attributes it to the dilution of gravity
in the higher-dimensional volume.

Kaluza-Klein modes Between the boundaries, the massive Kaluza-Klein

modes have to satisfy the following equation:

k2
 
15
ψn00 + m2n − ψn = 0.
4 (k|z| + 1)2

This is a Bessel equation of order 2 (see [10] eq. 9.1.49.), and its solutions
are linear combinations of Bessel functions of first and second kinds:

ψn = (|z| + 1/k)1/2 an J2 mn (|z| + 1/k) + bn Y2 mn (|z| + 1/k) ,




(17)

with an and bn some coefficients.

To get an approximation of these wave-functions, we will use asymptotic
expressions of the Bessel functions (see [6]). We can rewrite the above
equation as

4k 2
 
1/2



ψn = Nn (|z| + 1/k) Y2 mn (|z| + 1/k) + J2 mn (|z| + 1/k) ,
πm2

where Nn is a normalization constant. The coefficient in front of J2 has

been determined using the boundary condition (15) and the asymptotic
expressions of the Bessel functions for small arguments (mn |z|  1):

4 1
Y2 mn (|z| + 1/k) ' − −
πm2n (|z| + 1/k)2 π

16
and

m2 (|z| + 1/k)2
J2 mn (|z| + 1/k) ' .
8
As k/mn  1, the term with J2 dominates in the expression of the wave
function. To evaluate the constant Nn , we use an approximation for large
values of mn |z|:
√
zJ2 (mn |z|) ' (2/πmn )1/2 cos(mn |z| − 5π/4),

which we plug in the normalization relation

Z +L
dz|ψ|2 = 1.
−L

We get
+L
32k 4 4
Z
2 32k
dzNn2 cos 2
(m n |z| − 5π/4) = N n 3 5 L = 1,
−L π 3 m5n π mn

which implies
r 5/2
π πmn
Nn = √ .
2 4k 2 L

Our approximation for the KK states wave-functions in the limit of large

mn |z| is then
√
ψn = cos(mn |z| − 5π/4)/ L. (18)

6 Graviton spectrum
The presence of two branes induces the quantization of the masses of the
KK states. To see it, let us look at the effect of the two boundary conditions
(15) and (16) on the general solutions (17). The derivative of these solutions
turns out to be (cf. [10] eq. 9.1.29)

ψn0 = mn (|z| + 1/k)1/2 an J1 mn (|z| + 1/k) + bn Y1 mn (|z| + 1/k)





3
− (|z| + 1/k)−1/2 an J2 mn (|z| + 1/k) + bn Y2 mn (|z| + 1/k) ,



2
so the boundary conditions become

an J1 (mn /k) + bn Y1 (mn /k) = 0,

an J1 mn (Lz + 1/k) + bn Y1 mn (Lz + 1/k) = 0.

17
This system has solutions only if its determinant vanishes, i.e. only if



J1 (mn /k)Y1 mn (Lz + 1/k) − J1 mn (Lz + 1/k) Y1 (mn /k) = 0.

Coming back to the coordinate y, which effectively represents to the distance

along the extra dimension (see equation (8)),

1 kL
Lz ' e  1,
k
we can write

J1 (mn /k)Y1 (mn ekL /k) − J1 (mn ekL /k)Y1 (mn /k) = 0.

In the approximation of small masses (mn /k  1), the Bessel functions of

first order behave like J1 (mn /k) ∼ mn /k and Y1 (mn /k) ∼ ln(mn /2k)mn /k
(see [10] eqs. 9.4.4 and 9.4.5.), so that we can assume that −Y1 (mn /k) 
J1 (mn /k). The requirement that the determinant vanishes reduces to
 
J1 mn ekL /k = 0.

The masses of the KK tower are thus given by

mn = ke−kL jn ,

where jn are the zeros of the Bessel function: J1 (jn ) = 0.

As the value of k is supposed to be of order of the Planck scale and
the factor exp(−kL) at the TeV brane has been fixed to solve the Hierarchy
Problem, the masses of the KK states are of order TeV. Furthermore, jn+1 −
jn ' π, so that the splitting of the masses is also of order TeV. This implies
the possibility to observe individual resonances of the first KK states at
colliders in the very near future [11]. Figure 6 shows an extrapolation of the
cross-section for the process e+ e− → µ+ µ− at a linear collider, with different
values of the ratio k/MP l . The resonances for the first and second KK modes
are well-defined and can be seen individually, in dramatic opposition to the
phenomenology of the ADD scenario, which predicts only collective effects,
given the smallness of the splitting between the KK modes.

7 Newtonian limit
We would like to verify that the gravitational interactions mediated by the
gravitons modes that we found are in agreement with Newton’s law. For that
purpose, we consider a minimal coupling of matter to gravity and look for the
values of the coupling constants. The action is composed of a gravity part

18
 k/MPl

Z0 1
0.7

h(1) 0.5

0.3

0.2

h(2) 0.1

Figure 7: Individual resonances of the Randall-Sundrum gravitons.

SG given by equations (1) and (6), and a part accounting for the interactions
between matter and gravity:

√
Z
S = SG + d4 xdy −gLM (Φ, gM N ),

where Φ stands for the fields residing on the branes.

For small graviton perturbations around the background metric

gM N = e−2A ηM N → gM
0
N =e
−2A
(ηM N + hM N ),

we expand the matter Lagrangian in Taylor series up to first order:

0 δLM i
LM (Φ, gM N ) = LM (Φ, gM N ) + hµν 0
+ O(h2 ).
δgµν 0 =g
gµν µν

Using the definition (3) of the energy-momentum tensor,

√
µν −2 δ −gLM i
T = √
−g δgµν 0 =g
gµν µν

δLM i
= −LM g µν − 2 ,
δgµν gµν
0 =g
µν

p
and the formula det(ηµν + hµν ) = 1 + h/2 + O(h2 ) with h = g µν hµν , we

19
can write
p 0 √ 0 2
g 0 LM (φ, gM N) = g(1 + h/2)LM (φ, gM N ) + O(h )

√ δLM i
= g LM (Φ, gM N ) + hµν 0
δgµν gµν0 =g
µν

h
+ LM (Φ, gM N ) + O(h2 )
2
 
√ 1
= g LM (Φ, gM N ) − hµν T µν + O(h2 ).
2
On the other side, when we expand SG up to the second order in the
perturbations, we get the following terms: one part independent of hµν that
vanishes because of the requirement of the vanishing of the effective cosmo-
logical constant; one linear part, which is the action leading to the linear
equations of motion, so that it vanishes on shell; and one quadratic part,
which corresponds the usual Pauli-Fierz Lagrangian LP F . Remembering the
solution we found for the KK modes after a rescaling by exp(3A/2) and a
KK decomposition drives us to
X
0 3
LM (Φ, gM N ) = LM (Φ, gM N ) + M LP F (hnµν (x))
n
3
A
Xe 2 ψ (n) (z)
− hnµν (x)T µν
n
2

In order to get a canonically normalized Pauli-Fierz Lagrangian, we proceed

to a field redefinition:
1
hnµν (x) → √ hnµν (x),
M3
and we finally obtain
X X e 23 A ψn (z)
LM (φ, g) = LM (Φ, η) + LP F (hnµν (x)) − √ hnµν (x)T µν ,
2 M 3
n n

from which we can read the expression of the gravity-matter coupling con-
stants:
3
e 2 A ψn (z)
an = √ .
2 M3
We can now compute the gravitational potential between two particles
with unit masses on the TeV brane at z = Lz , i.e. the static potential
generated by the exchange of the zero-mode and the massive KK states.
Like in the case of a Yukawa interaction (see [12] eq. (4.127)), it is given by
∞
X a2n e−mn r
V (r) = − .
4π r
n=0

20
 The contribution of the zero-mode ψ0 (z) = exp(−3A/2) to the gravita-
tional interaction is
1 1
V0 (r) = − 3
16πM r
GN
= − ,
r
with GN the Newton constant. This reproduces the 4D gravity.
With the help of the approximation (18) for the KK states wave-fun-
ctions, the non-relativistic gravitational potential mediated by the nth mas-
sive graviton on the TeV brane reads

k 3 L2 2 e−mn r
Vn (r) = − cos (m n Lz − 5π/4)
16πM 3 r
GN 3 2 2
= − k L cos (mn Lz − 5π/4)e−mn r .
r
These contributions to the gravitational potential are exponentially sup-
pressed, and thus may be neglected down to distances of order of the fermi,
r . 10−13 cm. The actual experimental tests of gravity having only probed
down to the millimeter scale, there is no perspective of detecting such small
corrections any time soon.
In conclusion, gravity in the RS model corresponds effectively to 4D
gravity as we experience it.

8 Radius stabilization
Until now we have treated the length, or equivalently the radius, of the
extra dimension as a parameter, and we felt free to set it to the appropriate
value to solve the Hierarchy Problem (see section 4). However, such a degree
of freedom would imply the existence in the effective theory of a massless
scalar field, corresponding to the fluctuations of the radius along the extra
dimension: the radion. This massless radion would cause a fifth force in
violation to the equivalence principle. Therefore, to preserve the viability
of the Randall-Sundrum model, the radion has to obtain a mass, i.e. to be
stabilized.
A way to do it is the Goldberger-Wise mechanism [13]. The idea is to
introduce a massive scalar field φ in the bulk with a potential V (φ) and add
some potentials V1 (φ) and V2 (φ) on the two branes at the boundaries. The
corresponding action reads

√ 
Z
1
S = d4 xdy −g M 3 R + ∂M φ∂ M φ − V (φ)
2

−V1 (φ)δ(y) − V2 (φ)δ(y − L) .

21
 The requirement of Poincaré invariance imposes to choose the metric
given by

ds2 = e−2A(y) ηµν dxµ dxν + dy 2 ,

and to restrict the dependence of the scalar field to the extra dimension:

φ(x, y) = φ(y).

To find φ(y), the scalar field and the Einstein equations should be solved
simultaneously. The scalar equation is
1 √ ∂Vtot
√ ∂M −gg M N ∂N φ = − ,
−g ∂φ
with Vtot = −V − V1 δ(y) − V2 δ(y − L). Only the 55 component gives a
non-vanishing result:
∂V ∂V1 ∂V2
φ00 − 4A0 φ0 = + δ(y) + δ(y − L).
∂φ ∂φ ∂φ
Referring to the expression of the Einstein tensors found in appendix A,
the 55 and µν components of the Einstein equations GM N = κ2 TM N are
κ2 02 κ2
A02 = φ − V (φ), (19)
12 6
and
κ2 02 κ2
2A02 − A00 = φ − (V + V1 δ(y) + V2 δ(y − L)).
6 3
We can simplify the second result by using the first one to eliminate φ0 :
κ2
A00 = (V1 δ(y) + V2 δ(y − L)).
3
To obtain the boundary conditions, we integrate those results on very
small domains around the positions y1 = 0 and y2 = L. The kinks of φ and
A at those positions will result in jumps in the derivatives:
iyi+ ∂Vi
φ0 = ,
yi− ∂φ
and
iyi+ κ2
A0 = Vi .
yi− 3
Together with the scalar and Einstein equations these equations form the
gravity-scalar system. It is quite hard to solve generally, so we will restrain
our study to a special case.

22
 Suppose that V has the special form
2
κ2 2

1 ∂W (φ)
V (φ) = − W (φ),
8 ∂φ 6

for some function W (φ), called “superpotential”. As equation (19) can be

written as
1 6
V (φ) = φ02 − 2 A02 ,
2 κ
we conclude that
1 ∂W κ2
φ0 = and A0 = W (φ).
2 ∂φ 6
We want the bulk potential to include a cosmological constant term
(independent of φ) and a mass term (quadratic in φ), so we choose for
example
6k
W = − uφ2 ,
κ2
with u a parameter. From that we have
1 ∂W
φ0 = = −uφ,
2 ∂φ
whose solution is easily found:

φ(y) = φP e−uy .

On the TeV brane we get

φT = φP e−uL ,

which can be inverted to

L = ln(φP /φT )/u.

The value of the radius is thus determined by the equation of motion. To

solve the Hierarchy Problem, we need kL ' 35, which only implies a modest
tuning of order O(50) on the input parameters. Thus the solution to the
Hierarchy Problem provided by the Randall-Sundrum model does not arise –
like in the case of large extra dimensions – at the cost of introducing another
fine-tuning.

23
Acknowledgments

One of us (MG) would like to express his gratitude to Antonios Papazoglou

for the patient guidance of his first hesitating steps in general relativity:
cheers!

References
[1] T. Kaluza, Sitzungsber. Preuss. Akad. Wiss. Berlin (Math. Phys.) K1,
966 (1921); O. Klein, Z. Phys. 37, 895 (1926).

[2] J. Polchinski, String Theory, Cambridge, Vol. I, chap. 8: Toroidal com-

pactification and T -duality (1995).

[3] N. Arkani-Hamed, S. Dimopoulos and G.R. Dvali, Phys. Rev. D59,

086004 (1999) hep-ph/9807344.

[4] V. A. Rubakov and M. E. Shaposhnikov, Do We Live inside a Domain

Wall?, Phys. Lett. B125, 136 (1983).

[5] L. Randall and R. Sundrum, A Large Mass Hierarchy from a Small

Extra Dimension, Phys. Rev. Lett. 83, 3370 (1999).

[6] L. Randall and R. Sundrum, An Alternative to Compactification, Phys.

Rev. Lett. 83, 4690 (1999).

[7] R. M. Wald, General Relativity, The University of Chicago Press, 1984.

[8] C. Csaki, TASI Lectures on Extra Dimensions and Branes (2002) hep-
ph/0404096.

[9] B. Zwiebach, A First Course in String Theory, Cambridge University

Press, 2005.

[10] M. Abramowitz and I. A. Stegun, Handbook of Mathematical Functions,

Dover Publications, 1972.

[11] H. Davoudiasl, J. L. Hewett and T. G. Rizzo, Phys. Rev. Lett. 84, 2080
(2000).

[12] M. E. Peskin and D. V. Schroeder, An Introduction to Quantum Field

Theory, Addison-Wesley Publishing Company, 1995.

[13] W. D. Goldberger and M. B. Wise, Modulus Stabilization with Bulk

Fields, Phys. Rev. Lett. 83, 4922 (1999) hep-ph/9907447.

24
A Einstein tensor
We want to calculate the Einstein tensor for the metric
ds2 = e−2A(y) ηµν dxµ dxν + dy 2
= gM N (y)dxM dxN ,
with
gM N (y) = e−2A(y) ηµν + δM
5 5
δN .
The inverse metric is
g M N (y) = e2A(y) η µν + δ5M δ5N .

Christoffel symbols
1
ΓPM N = g P R (∂M gN R + ∂N gRM − ∂R gM N ).
2
As gM N is a function of the extra dimension only, and this only in its µν
components, we have
∂L gM N = ∂5 gM N = ∂5 gµν .
That implies that only two types of Christoffel symbols are non-vanishing:
1 5R
Γ5µν = g (−∂R gµν )
2
1 55
= g (−∂5 gµν )
2
= A0 e−2A ηµν ,
and
1 νR
Γνµ5 = g (∂5 gRµ )
2
1 2A νρ
= e η (−2A0 e−2A ηρµ )
2
= −A0 δµν .

Ricci tensor
RM N = ∂P ΓPM N − ∂N ΓPM P + ΓPP Q ΓQ P Q
M N − ΓN Q ΓM P .

Rµν = ∂5 Γ5µν + Γσσ5 Γ5µν − Γσν5 Γ5µσ − Γ5νσ Γσµ5

= (A00 − 2A02 )e−2A ηµν − 4A02 e−2A ηµν
+A02 e−2A ηµν + A02 e−2A ηµν
= (A00 − 4A02 )gµν .

25
 Rµ5 = 0.

R55 = −∂5 Γσ5σ − Γσ5ρ Γρ5σ

= 4A00 − 4A02 .

Ricci scalar

R = g M N RM N
= g µν Rµν + g 55 R55
= 4(A00 − 4A02 ) + 4A00 − 4A02
= 8A00 − 20A02 .

Einstein tensor
1
Gµν = Rµν − gµν R
2
= (6A − 3A00 )gµν .
02

1
G55 = R55 − g55 R
2
= 6A02 .

26

View publication stats