Humanly traversable wormholes

Juan Maldacena1 and Alexey Milekhin2

arXiv:2008.06618v2 [hep-th] 3 Nov 2020

1
Institute for Advanced Study, Princeton, NJ 08540, U.S.A.
2
Physics Department, Princeton University, Princeton, NJ 08544, U.S.A.

Abstract
We point out that there can be humanly traversable wormhole solutions in some
previously considered theories for physics beyond the Standard Model, namely the
Randall—Sundrum model.
Contents
1 Introduction 1

2 Review of traversable wormholes 2

2.1 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.2 Magnetic black holes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.3 The wormhole ansatz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.4 Example with free fermions . . . . . . . . . . . . . . . . . . . . . . . . . . 6

3 Traversable wormholes in the Randall-Sundrum II model 7

3.1 The emergence of a two dimensional CFT from the four dimensional one . 7
3.2 The Randall Sundrum II model that we consider . . . . . . . . . . . . . . 9
3.3 Requirements for a humanly traversable wormhole . . . . . . . . . . . . . 12
3.4 Some practical problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

4 Conclusions 14

A More details on the flow 14

B Rotation and the emission of radiation 16

1 Introduction
Traversable wormholes are a staple of the science fiction literature. In classical general
relativity, they are forbidden by the average null energy condition [1, 2, 3]. Interestingly,
they are allowed in the quantum theory, but with one catch, the time it takes to go
through the wormhole should be longer than the time it takes to travel between the two
mouths on the outsidea . Nevertheless they are interesting configurations that are allowed
by the laws of physics as we know them. In fact, based on the initial construction in
[5], four dimensional traversable wormholes were constructed in [6], see also [7, 8]. The
construction in [6] involved ingredients that are present in the Standard Model, but only
at short distances. Therefore only microscopic wormholes were found.
In this paper, we revisit the question and we engage in some “science fiction”. Namely,
we will introduce a dark sector with desirable properties for constructing macroscopic
traversable wormholes. This dark sector only interacts via gravity with the Standard
Model. Our main point is to emphasize the consistency of such wormholes with the laws
of physics as we presently know them, using only previous ideas for physics beyond the
Standard Model. In particular, we will use a version of the Randall-Sundrum model [9].
Our construction needs a dark sector consisting of a four dimensional conformal field
theory with a U(1) symmetry that is gauged by a four dimensional (dark) gauge field. A

a
This also implies that they can not be converted into time machines [4].

1
simple example would be a theory of many massless fermions coupled to a U (1) gauge
field [6]. Better wormholes are possible by using a Randall Sundrum II model [10] with a
U (1) gauge field. This model allows for large enough wormholes that could be traversed
humanely, i.e. surviving the tidal forces. Using them, one could travel in less than a
second between distant points in our galaxy. A second for the observer that goes through
the wormhole. It would be tens of thousands of years for somebody looking from the
outside.
This paper is organized as follows. In section two, we review the main idea for the
construction of traversable wormholes and discuss the example involving massless charged
fermions. In section three, we consider the Randall Sundrum II model and argue that we
can have wormholes large enough to allow for human travelers. In the discussion section we
mention some further practical issues that make such wormholes problematic in practice.

2 Review of traversable wormholes

2.1 Preliminaries
The theory we will be considering consists of gravity, plus a U (1) gauge field and a four
dimensional matter theory with a U (1) symmetry which we are gauging with the dynamical
gauge field

4 √ √
Z Z
1 1
S= d x gR − 2 d4 x gFµν F µν + Smatt [gµν , Aµ ] (2.1)
16πG4 4g4
where Fµν is the usual field strength, F = dA. Both Aµ and the matter theory are part of
a dark sector.
In [6], the matter theory, Smatt , was given by Nf free massless fermions charged under
the U (1). Here we consider a matter theory which is holographic, namely it has an AdS5
dual described by five dimensional gravity and a five dimensional U (1) gauge field which
we also call Aµ . We will later discuss it in more detail.
A crucial property for the construction is the following. In the presence of a constant
magnetic field, the four dimensional matter theory flows to a two dimensional theory with
a central charge that scales linearly with the total magnetic flux,

c2 ∝ BA = 2πq (2.2)

where B is the magnetic field, A is the transverse area and q the integer magnetic flux. We
can imagine dividing the transverse area into flux quanta and we have a certain central
charge per flux quantum. For the case with Nf free fermions, the central charge was
c2 = Nf q, due to q 2d massless modes [11, 12]. The transition between the four dimensional
description and the two dimensional one happens at a distance scale
1
lB ∝ √ (2.3)
B

2
2.2 Magnetic black holes
We start with the geometry of an extremal magnetically charged black hole
dr2 q
ds2 = −f dt2 + + r2 (dθ2 + sin2 θdφ2 ) , A = cos θdφ (2.4)
f 2
√
 re  2 πqlp p re
f = 1− , re ≡ , lp ≡ G4 , Me = (2.5)
r g4 G4
where q is the (integer) magnetic charge and Me the mass at extremality. re sets the
radius of curvature of the geometry in the near horizon region, and also the size of the two
sphere. At this extremal limit the geometry develops an infinite throat (as r → re ) where
the redshift factor, gtt , becomes very small but the size of the sphere remains constant, see
Figure 1(a).
When we put the matter theory on this background, it leads to a two dimensional CFT
at scales less than (2.3). Notice that the proper size of the magnetic field is
q
B= 2 (2.6)
2r
which for large q, q  1, is much larger than the inverse size of the sphere. This means that
we can ignore the curvature of the two sphere when we consider the transition between the
four dimensional theory and the two dimensional one. For that reason we can study this
transition in flat space. We will later discuss in detail how this occurs for a holographic
theory.

2.3 The wormhole ansatz

We will now consider two black holes with opposite charges at some distance d from each
other. Naively, they would attract and coalesce. However, let us first imagine they are
somehow held fixed so that we can ignore this issue for the moment. We will return to it
later.
If we had two separate extremal black holes each of them would have an infinite throat
like the one depicted in Figure 1(a). We now change the geometry of the throats by
connecting them to each other so that the geometry of the deep throat region is given by
dρ2
 
2 2 2 2 2 2 2
ds = re −(ρ + 1)dτ + 2 + (dθ + sin θdφ ) , − ρc ≤ ρ ≤ ρc (2.7)
(ρ + 1)
Notice that the size of the sphere is constant. This is a portion of the AdS2 × S 2 geometry,
which is a solution of the equations of gravity plus a magnetic field. Beyond ρ = ±ρc the
geometry is described by that of the extremal black hole (2.4). The final geometry looks
as in Figure (1)(b). It has a free parameter which is the total length of the wormhole. We
will define this quantity as the ratio of the time in (2.7) to the asymptotic time at infinity
in (2.4)
t
`= (2.8)
τ
3
Figure 1: (a) We see a sketch of the geometry of a single extremal black hole. It develops
a long throat with a locally AdS2 × S 2 geometry. (b) The wormhole geometry. We start
with two separate throats and we connect them. The geometry of the connected throats
is again AdS2 × S 2 . The two mouths of the wormhole are at some distance d. The red
brackets indicates the region of overlapping validity of the metrics (2.4) and (2.7). The
green lines denote the magnetic field lines. They form closed circles, entering one mouth,
traveling through the wormhole, exiting the other mouth and then going back to the first
mouth in the ambient space.

In terms of `, we can also find the relation between the radial coordinate in (2.7) and (2.4)

(r − re )
ρ=` , for ρ  1 , and r − re  re (2.9)
re2
The last two conditions ensures that we are far from the center of the wormhole but we
are deep inside the throat. For `  re , there is an overlapping region where we glue the
near horizon region of the black hole (2.4) to the wormhole region (2.7), see Figure (1)(b).
After this step we do not have a solution of the Einstein + Maxwell equations. In
order to obtain an actual solution we need to include the effects of matter. This matter
behaves as a two dimensional conformal field theory with the central charge proportional
to q, (2.2). This two dimensional CFT can be viewed as composed by somewhat separate
CFTs , each living along a magnetic field line. These magnetic field lines form circles.
They enter one mouth of the wormhole and they exit through the other, see Figure (1)(b).
A two dimensional theory on a circle gives rise to a negative Casimir energy. This gives
some additional negative energy which turns the wormhole into a solution for a special
value of the length parameter (2.8).
We will not repeat here the full argument for the solution, which can be found in detail
in [6]. This involves solving the Einstein equations with a source given by the quantum

4
stress tensor of the two dimensional field theory b .
Imposing the Einstein equations, we deform slightly the metrics (2.4), (2.7) and deter-
mine the parameter ` in (2.8). We will here only review a simple method to determine
the final parameters of the wormhole, which can be viewed as an energy minimization
argumentc .
It turns out that the gravitational energy cost for joining the AdS2 × S 2 region to
flat space for “length” ` is the same as the energy above extremality for a black hole
of temperature T = 1/(2π`) [6]. This can sound more plausible if we notice that the
rescaling in (2.8) is the same as the rescaling we would have between Rindler time in the
near horizon region of a near extremal black hole and the asymptotic time t. This energy
is twice the mass of a single near extremal black hole:

re3
Egravity = 2M = 2Me + (2.10)
G4 `2
The two dimensional conformal field theory develops a negative Casimir energy equal to
c2
ECasimir = − (2.11)
8`
This also includes a contribution due to the conformal anomaly in AdS2 . Here c2 is the
total central charge of the two dimensional CFT (2.3). In writing this expression we
assumed that d  ` d . Adding these two contribution and extremizing over ` we find

16re3
`= (2.12)
G4 c2
Inserting (2.12) into the expression for the energy we find the binding energy

c2 c2 G4
Ebind = Egravity + ECasimir − 2Me = − = − 28 3 (2.13)
16` 2 re

This means that forming the wormhole lowers the energy of the configuration, relative to
that of two separate extremal black holes. This energy is also important for the following
reason. To keep the wormhole open we need the negative Casimir energy. If we send

b
There are other contributions from the matter stress tensor which are SL(2) invariant in the throat
region. Such contributions can change slightly the black hole parameters. In particular, they include
contributions from the four dimensional conformal anomaly. The Casimir energy contribution from the
effective two dimensional theory is the leading contribution that violates the SL(2) symmetry and it is
the important one for the construction of the wormhole.
c
The justification of this argument is that the geometry we described above is almost a solution. We
get a relatively small action for the variable ` from gravity and also one from the matter fields. So we can
concentrate extremizing the effective action for the variable `.
d
The equations can also be solved when d ∼ `. It is very similar, see [6].

5
too much positive energy we will collapse the wormhole into separate non-extremal black
holes. The maximum energy we can send, before that happens, is of order (2.13).
We can interpret this bound on the energy from the information transfer perspective.
We can imagine sending signals through the wormhole using quanta with energy ∼ 1/`.
The constraint that the energy is smaller than |Ebin | in (2.13) implies that the total number
of quanta cannot exceed a number ∼ c2 . This means that we can not transfer more
information than can be carried by the CFT through outside the wormhole. A similar
bound exists for the Gao–Jafferis–Wall protocol [13]. However, in our case we can talk
about the information per unit time (namely of the order of ∼ c2 qubits per `), rather
than total transferable amount.
Let us note some geometric interpretations of `. ` sets the redshift factor between the
middle of the wormhole and the exterior. Namely, a massless particle with wavelength
1/re at the center of the wormhole would have an energy of order 1/` as seen from the
outside. We can call this the energy gap of the wormhole, the minimum energy excitation
of the configuration. The binding energy (2.13) is larger by a factor of c2 . In addition,
` sets the time it takes to travel through the wormholes as seen from the outside, which
is π`. In deriving (2.12) we assumed that the distance d between the two black holes is
smaller than `, d  `. Even when we do not make that assumption we find that the time
through the wormhole is always longer than through the outside, π` > d. On the other
hand the proper traversal time for an observer going through the wormhole is of order πre ,
which is much shorter than `. The ratio between these two times
`
=γ1 (2.14)
re
is the boost factor that a particle that starts non-relativistic outside would acquire at the
center of the wormhole. So, going through the wormhole is similar to accelerating to very
high speeds and then decelerating. The difference is that the acceleration and deceleration
is provided by gravity for free and one does not have to go through the ambient space.
These wormholes are the ultimate roller coaster.
We can now return to the problem of preventing the attraction of the two wormhole
mouths. Notice that from the outside these have, up to a small correction (2.13), the same
mass and charge as two oppositely charged extremal black holes. We can make they orbit
each other. It is possible to do this with an angular velocity Ω  1/` that is smaller than
the energy gap, so that we do not have a disruptive effect inside the wormhole. Of course,
such a configuration would slowly emit gravity and dark-U (1) waves which will eventually
cause them to coalesce. In Appendix B, we explain that this happens at time scales that
are parametrically longer than the traversal time, π`.

2.4 Example with free fermions

In this subsection we review the values of above parameters for the case of Nf free fermions.
In that case c2 = Nf q where q is the integer magnetic charge. This results in a wormhole

6
length and energy gap
` q
= 16π 2 , Ebin re = 2−8 g42 Nf2 (2.15)
re g4 Nf

This model is under control as long as g42 Nf < 1. Otherwise the gauge field becomes
strongly coupled. For reasons we will discuss in more detail later, due to tidal force reasons,
we need re > 107 m. If we further assume that the mass of the spaceship is about 103 kg,
we also need the condition |Ebin | > 103 kg. This then implies that

re |Ebin | ∝ Nf (g 2 Nf ) > 107 m × 103 kg → Nf > 1052 (2.16)

This value of Nf is too large if we want the UV cutoff of the local field theory to be
above the TeV scale, since we expect that Nf < Mpl2 /(Tev)2 ∼ 1032 due to Bekenstein-like
arguments. Since this suggests that a strongly coupled theory might be desirable, it is
natural to look for theories with an AdS5 gravity description.

3 Traversable wormholes in the Randall-Sundrum II

model
In this section we will study a theory which includes a dark sector based on the Randall-
Sundrum II model [10]. The feature that is important for us is that it includes an AdS5
region that extends towards the infrared. This can also be viewed as four dimensional
CFT coupled to gravity. This point of view is useful, but not essential, for describing the
construction. It is useful because we can just show that this AdS5 theory has the requisite
general properties described in the previous section. The main property that we will use
is the emergence of a two dimensional CFT in the presence of a magnetic field.
The five dimensional theory is described by the action

5 √ √
Z Z
1 1 6
S5 = d x g(R − 2Λ5 ) − 2 d5 x gFµν F µν , Λ≡− 2 (3.17)
16πG5 4g5 R5

which contains two dimensionless parameters R53 /G5 and R5 /g52 determining the (inverse)
couplings of gravity and the gauge field.

3.1 The emergence of a two dimensional CFT from the four

dimensional one
In this subsection argue that, in the presence of a magnetic field, the four dimensional
theory becomes effectively a two dimensional CFT.
When we start with free massless four dimensional charged fermions, the two dimen-
sional CFT comes from the lowest Landau level, which gives rise to massless fermions

7
propagating along the magnetic field lines. So in this case it is relatively easy to see. The
central charge is then c2 = Nf q.
We now start from a holographic theory, or an AdS5 space, and we study the effects of
adding a magnetic flux at the boundary of AdS5 . We will derive that, as we move away
from the boundary, the geometry transitions from AdS5 to AdS3 × R2 . The presence of
this AdS3 factor is interpreted as the two dimensional CFT of the previous discussion.
This necessary solution was discussed in [14], and we review it below.
We write the metric and gauge field ansatz consistent with the symmetries

ds2 = R52 e2λ (−dt2 + dx2 ) + e2σ (dy12 + dy22 ) + dρ2 ,

 
A = By1 dy2 (3.18)

Inserting this into Einstein’s equations we find

−4 −4σ
0 = −6 + 3lB e + λ02 + 4λ0 σ 0 + σ 02 , λ = λ(ρ) , σ = σ(ρ)
2 2
−4 −4σ 3g5 R5
0 = 4 − 4lB e − 2λ0 σ 0 − 2σ 02 − σ 00 , 4
lB ≡ (3.19)
4πG5 B 2
We now make the following observations.
For large ρ we have a solution λ = σ = ρ, where we neglect the term proportional to
lB . This is the asymptotically AdS5 form of the geometry. The term involving lB becomes
important when
e−σ ∼ lB (3.20)
This is can be viewed as the distance scale where we have the transition between the
√ four
dimensional behavior and the two dimensional one. This scale is proportional to 1/ B, as
we expect from general arguments, since B is the only dimensionful quantity in the four
dimensional sense.
At longer length scales σ is constant, e−σ = lB and the geometry becomes that of
AdS3 × R2 , with an AdS3 radius which is of order of the AdS5 radius. More precisely, in
this region e−σ = lB and R32 = R52 /3. The full numerical solution of (3.19) is indicated
in Figure 3 in appendix A. Note that in the IR, when e−σ = lB , the proper size of the
magnetic field q
in the transverse space goes to a constant independent of the original 4d
2 R3
flux, Be−2σ ∝ Rg 5 G55 , which is just the ratio of the five dimensional U (1) vs gravitational
couplings.
An important quantity is the two dimensional central charge [15, 16]
√ 3 s √
3R3 3R5 A 3/2 R53 R5 AB
c2 = = 2
= 2π × ×q , q= (3.21)
2G3 2G5 lB G5 g5 2π

Note that c4 ∝ R53 /G5 is the “central charge” of the four dimensional theory (not to
be confused with the two dimensional central charge on the left hand side of (3.21)). The
second factor is the effective five dimensional gauge coupling at the scale of the radius of
AdS5 . Both of them should be larger than one for the validity of the theory.

8
 If we were to put the four dimensional theory on R1,1 × S 2 with a large magnetic flux
on the S 2 , then we expect a slightly different solution. However, in the regime that the
magnetic flux is large enough, so that the length scale lB is much smaller than the radius
of the S 2 , then we expect that we can neglect the curvature of S 2 in studying the flow to
the AdS3 × S 2 geometry in the IR. Indeedpwe check that this is the case in Appendix A.
For our application, we have that lB /re ∼ R5 /re  1, so that this approximation works.
The situation is fairly similar when we put the four dimensional theory on AdS2 × S 2 .

3.2 The Randall Sundrum II model that we consider

We will now discuss the full model under consideration.
We imagine that we have a five dimensional gravity theory with a five dimensional
U (1) gauge field, as in (3.17). We then imagine that we have a so called “Planck brane”
at a finite distance into the UV direction of AdS5 . Its tension is fine tuned so as to make
the cosmological constant effectively zero, compared to the distance scales we consider. In
other words, we have the AdS5 metric

ds2 = e2ρ̃/R5 (−dt2 + d~x2 ) + dρ̃2 , ρ̃ < 0 (3.22)

Then the four dimensional metric is the metric at ρ = 0 where the Planck brane sits. We
imagine that the Standard Model fields live on the Planck brane. The four dimensional
Newton constant is
1 1 R5
= (3.23)
G4 2 G5
where we have set to zero a possible Einstein term we can add on the brane.
Similarly, the five dimensional gauge field leads to a four dimensional gauge field with
a coupling  
1 R5 LIR
= 2 log (3.24)
g42 g5 R5
where the logarithm can be interpreted as the running of the four dimensional coupling
constant due to the charged matter [17]. It is IR divergent, but this is not really a problem
for us. When we think about the near horizon region of the black hole geometry there is a
natural IR cutoff which is re . Moreover, the fact that the 4d theory flows to an essentially
two dimensional theory at scale lB provides an even shorter IR cutoff. So the log just gives
us an extra factor of order log(lB /R5 ).
We are interested in making a wormhole of a fixed size re as traversable as possible.
This will involve maximizing c2 , for a fixed re . This both makes ` shorter and the binding
energy stronger. From (3.21) we see that the last two factors are essentially the same as
the ones that appear in re in (2.5). So we want to maximize the first factor in c2 which
amounts to making R53 /G5 as large as possible. This in turn implies that we want to make
R5 as large as possible.

9
 According to [18], the experimental upper bound on R5 is roughly about 50µm =
5×10−5 m. So we pick this value for the estimates below. Note that (3.23) implies that
−1/3
G5  1 TeV.
The arguments we had above lead to a solution for this model if re  R5 , so that we
can think of the AdS5 region as a special four dimensional field theory. We are not using
a full blown AdS/CFT duality, we are only making the observation that the 5d problem
reduces effectively to a problem we can analyze using the 4d intuition. In fact, we are not
going to find the full 5d solution explicitly, we will only argue that it should exist based
on solving it by patches using effective field theory reasoning. As we explained in the
previous section the four dimensional solution depends only on the four dimensional gauge
coupling and c2 . These are given in (3.24) and (3.21). The five dimensional geometry
has a boundary that is determined by the four dimensional wormhole geometry. As it
extends into the fifth dimension, it locally has an AdS5 geometry in the UV region which
transitions into an AdS3 × S 2 geometry at a distance lB (3.19) which is
p
lB = [3 log (re /R5 )]1/4 re R5 (3.25)

where we used the four dimensional equations (2.5) that determined

√ the magnitude of the
magnetic field in the throat region. We also estimated lB ∝ re R5 inside the logarithm
in (3.24) with LIR ∼ lB , which yielded an extra 1/2.
This implies that lB is some intermediate scale between R5 and re , so that the flow to
AdS3 is well described by the discussion in section 3. Notice that this flow to a locally
AdS3 region occurs both in the wormhole region as well as the region outside where there
is some magnetic field. As the magnetic field is weaker outside we flow for longer and the
AdS3 region is reached at a distance scale which is larger in the four dimensional space.
Outside, in the region between the two mouths, the magnetic field also depends on the
angular coordinate on the S 2 that surrounds one of the mouths. This implies that the flow
distance also depends on this angular dimension and the five dimensional geometry will
not be a direct product.
As we go deeper into the AdS3 region, this space contains a time direction and the
spatial directions will be as drawn in Figure 2(b). We have a circle which follows the
magnetic field lines, where the AdS3 goes over to the AdS5 geometry. This circle is then
“filled in” by the extra dimension, see Figure 2(b). This AdS3 geometry leads to a classical
negative gravitational energy corresponds to the negative Casimir energy in the 2d CFT
interpretation. In this Figure the whole exterior region occupies a relatively short space,
a portion equal to d/(π`), indicated in purplee in Figure 2(b). In summary, the full
topology of the solution is the following. First, let us imagine fixing the two mouths of

e
The purple line in Figure 2(b) curves inwards because according to eq. (3.19), distance lB , where
AdS3 starts, is inversely proportional to the magnetic field. Since the magnetic field decreases as we go
away from the wormhole mouths into the flat space region, we flow to a lower scale and thus the metric
scale factor is smaller. The green line in Figure 2(b) also approaches the boundary at the wormhole ends
because the induced metric on this surface should be that of AdS2 .

10
Figure 2: (a) We display the topology of the full five dimensional space. It is essentially
filling the space “outside” the wormhole in the ambient space. We highlighted a special
slice that ends on a particular magnetic field line. (b) We display IR part of the geometry
of the highlighted slice. This geometry is a portion of global AdS3 , the time direction is
perpendicular to the page and we plot just the spatial direction. The Beyond the drawn
region, the geometry goes over to the AdS5 geometry as it approaches the Planck brane.
The part of the magnetic field line that is outside the wormhole was colored purple. For
each magnetic field line, or each point on the S 2 we have a slice like this.

the wormholes. Since gtt never vanishes, we can just forget about the time direction and
discuss only the topology of the spatial directions. The three spatial directions of the
Planck brane have a non-trivial topology that roughly looks like that of Figure 2(a). If we
compactify the point at spatial infinity, the topology of these three spatial dimensions is
S 1 × S 2 . After we add the extra dimension, the final topology of the spatial dimensions is
essentially D2 × S 2 , where we fill in the S 1 .
The fact that the S 1 is contractible in the full geometry is interesting. The topological
censorship theorems say that in classical general relativity we cannot have a non-trivial first
homotopy group, π1 . This is allowed when quantum effects are included. However, what
looks like a quantum Casimir energy in four dimensions is actually a classical effect in five
dimensions (the negative classical energy of AdS3 ). So in five dimensions the topological
censorship should work. It indeed works because the classical geometry in five dimensions
has a trivial π1 . More physically, in four dimensions, the null ray that goes through the
wormhole cannot be deformed continuously into one that stays outside the wormhole. This
is possible by moving the light ray into the fifth dimension.
We have argued that the solution exists, but one could imagine trying to find it nu-
merically in a more explicit form.
As a simple example where we can discuss the geometry more explicitly, let us take

11
the four dimensional geometry to be the Einstein static universe, R × S 3 , and we add
the two wormhole mouths at antipodal points on the S 3 . This is not a solution of the
four dimensional gravity theory we are considering. However, we can sketch the form
of the solution of the five dimensional equations with these four dimensional boundary
conditions. The advantage is that we have now an unbroken SO(3) rotation symmetry
and the geometry contains an S 2 factor (with varying radius) for each of the points of the
rest of the three dimensions. The four dimensional geometry has two pieces. One is the
wormhole interior with an AdS2 × S2 geometry. The exterior region is R × S 3 . Deep in the
bulk the metric is AdS3 in global coordinates. This continues up to a transition region,
with roughly the shape indicated by the green and purple lines in Figure 2. After this
region, the metric takes a locally AdS5 form with a local geometry determined by the flow
discussed in section (3.1). It is possible to describe the metric more explicitly and find the
solution numerically, but we will not do it here.

3.3 Requirements for a humanly traversable wormhole

Somebody traversing the wormhole will be subject to gravitational tidal forces, which
are similar to the ones felt by somebody falling into an extremal black hole. Since the
curvatures are of order 1/re2 , the tidal acceleration felt by the observer would be roughly
a ∼ (size)/re2 . We need to demand that this is smaller than the maximal sustainable
accelerations which are about 20g for short durations, where g = 9.8 m/s2
size
20g > a ∼ → re > 1.5×107 m ∼ .05s (3.26)
re2
where we assumed that the size is about 0.5 m. We see that the size of the wormhole
mouth should be very large, since we are very fragile. We have also written it as a time,
since this sets proper time it takes to traverse the wormhole, which is πre . It is useful to
express all the wormhole quantities in terms of re
Rr 4 re2 p π2 R52
c2 = 4π p 5 e , `= log(re /R5 ) , Ebin = −
G4 log(re /R5 ) π R5 16 G4 re log(re /R5 )
Taking the minimal size for re in (3.26) and the largest possible value for R5 , of 50 µm
[18], we see that the parameters are
`
re ∼ .05 s , c2 ∼ 7×1072 , ` ∼ 3×103 ly , γ= ∼ 2×1012 , Ebin ∼ −5×109 kg
re
(3.27)
where ` is given in light years. We see that the traversal time as seen from the outside is
rather long.
This time is basically unchanged as long as the distance, d, between the two mouths
is smaller d . `. For d of order ` the rotation frequency is much smaller than 1/`. So
if these wormholes were in an empty, zero temperature, flat space, they could exist for a
while and be traversable.

12
 One might be worried that a highly boosted observer going through a weakly curved
geometry will feel large tidal forces. This happens in general, but AdS2 is boost invari-
ant, so the traveller does not feel anything special. However, we also need to take into
account that there could be small non-boost invariant perturbations to the metricf . These
perturbations can be of order one near the mouth region, but as we go to the center of the
wormhole they are suppressed by a power of the ratio of the redshift factor at the center.
This works out to be of order 1/γ or smaller (2.14) (it is smaller for higher mass Kaluza
Klein modes). Here γ is also the boost factor (2.14). A boosted observer would see tidal
forces enhanced by a factor of γ 2 . However, the duration of the force is also decreased
by a factor of 1/γ, so that it integrates to a total momentum transfer of order one. This
implies that the net size of the effect of the perturbations is the same as it is in the mouth
region, where tidal effects are as in (3.26).

3.4 Some practical problems

When an object falls into the wormhole it develops a very large boost factor, γ, as seen
by an observer at the center (or bottom) of the wormhole (3.27). In particular, if a CMB
photon falls into the wormhole its energy will be enhanced by this factor. In addition,
the wormhole traveller that finds it inside will see it with the square of this factor. So
one would have to put the huge black hole inside a refrigerator in order to prevent this.
Therefore, it seems exceedingly difficult to prevent this problem. Of course, any particle
that falls into the wormhole would create a related problem. So the wormhole has to be
very clean before it can be traversed.
If particles that fall into the wormhole scatter and lose energy then they would accumu-
late inside, contributing some positive energy that would eventually make the wormhole
collapse into a black hole.
This solution is assuming that the dark sector is at a extremely low temperatures,
smaller than 1/`, or 10−26 eV . In our universe, even if we somehow arranged the cosmo-
logical evolution to produce such low temperatures, there would be processes that send
energy into the dark sector that heat it up to a much larger temperature. One simple
example of such a process is the soft radiation produced by particle collisions in a star.
This can have a 5th dimensional component, and the radiation thus produced is larger
than the purely four dimensional computation [19], which by itself is enough to heat it to
a much larger temperature.
Another problem seems to be producing the wormhole in the first place. It would be
interesting to understand whether they can be produced in the RS model. Since they
require topology change, this seems difficult.
We highlighted the wormholes that are good enough to send a person. However, we
could consider smaller wormholes, say of size re ∼ R5 which would correspond to a mass
M ∼ 1023 kg. In this case, we can separate them by a large distance d, say over the solar

f
We thank E. Martinec for prompting us to look at this issue.

13
system and then the traversal time would be comparable to their distance. Of course, it is
very small and the tidal forces would be huge. But we could use them to send very secret
signals or qubits.

4 Conclusions
We have argued that the Randall Sundrum II model allows for traversable wormhole
solutions. In fact, it allows for solutions where the wormholes are big enough that a
person could traverse them and survive.
From the outside they resemble intermediate mass charged black holes. Their big size
comes from demanding that a human traveller can survive the tidal forces. They take a
very short proper time to traverse, but a long time as seen from the outside. The traveller
acquires a very large boost factor, γ, as it goes through the center of the wormhole.
We have argued this most clearly for the case that the wormhole exists in a cold and
flat ambient space, much colder than the present universe. We have not given any plausible
mechanism for their formation. We have only argued that they are configurations allowed
by the equations.
Acknowledgments
We are grateful to A. Almheiri, E. Martinec, F. Popov, L. Randall and N. Sukhov for
discussions.
J.M. is supported in part by U.S. Department of Energy grant DE-SC0009988, the
Simons Foundation grant 385600.
A.M. also would like to thank C. King for moral support.

A More details on the flow

In this Appendix we can include a numerical discussion of equations (3.19). Let us write
them again after setting lB = 1 by a rescaling of σ.

0 = −6 + 3e−4σ + λ02 + 4λ0 σ 0 + σ 02 (A.28)

0 = 4 − 4e−4σ − 2λ0 σ 0 − 2σ 02 − σ 00 (A.29)

These equations have only three integration constants. Two of them are trivial, one
corresponds to shifting λ by a constant and the other to shifting ρ by a constant. We
shifted σ to simplify the equation. And we are left with only one non-trivial integration
constant. We can solve for λ0 from the first equation (A.28) and replace it in the second.
This gives an √ordinary differential equation for σ. We solve it as follows. First we note that
σ = 0 , λ = 3ρ+constant is a solution. Then we expand in small fluctuations around
this solution, assuming σ is very small. This linear equation has two solutions which are
simple exponentials, one increasing and the other decreasing as a function of ρ. We want
to impose the boundary condition that we have AdS3 in the IR. This selects the increasing

14
solution and also fixes the non-trivial undetermined integration constant for the equations.
We then numerically solve the equations by choosing initial conditions at place where σ is
very small and given by the exponentially increasing solution. This leads to the numerical
solutions in Figure 3.
σ, λ
1.5

1.0

ρ + constant
0.5
constant + 3ρ
σ
ρ
-1.0 -0.5 0.5 1.0
λ
-0.5

-1.0

Figure 3: We show the numerical solution describing the flow. We have AdS√ 5 at large ρ,
0
where σ ∼ λ ∼ ρ+constant. For negative ρ this goes to σ ∼ 0 and λ = 3. Besides
plotting σ and λ we plotted the approximations that are valid when |ρ| is large.

We now consider the case that we start from the theory on a two sphere. In other
words we have R2 × S 2 at the boundary with q units of magnetic flux, or F = 2q sin θdθdϕ.
In this case the metric ansatz is the same as (3.18) replacing dy12 + dy22 → dθ2 + sin2 θdφ2 .
By conformal symmetry we set the radius of the two sphere to one. The new equations
are now
0 = −6 − e−2σ + 3ˆlB
−4 −4σ
e + λ02 + 4λ0 σ 0 + σ 02
2 2
0 = 4 + e−2σ − 4ˆlB
−4 −4σ
e − 2λ0 σ 0 − 2σ 02 − σ 00 , ˆl4 ≡ 3g R5 (A.30)
B
πG5 q 2
We see that the curvature of the sphere gives rise to the extra term involving e−2σ . Now
the quantity `ˆB is a distance in units of the radius of S 2 . The radius of the S 2 in the IR
is given by solving the second equation with σ 0 = 0 which gives
(ˆlB )3
e−σ = ˆlB + + ··· , for ˆlB  1 (A.31)
16
The second term is the correction due to the curvature of the sphere. The radius q of AdS3
has a similar small correction. We are indeed interested in the case that `ˆB = re ∝ Rre5 
l B

1, see equation (3.25). We can also numerically solve (A.30) and obtain answers similar
to those in Figure 3.

15
B Rotation and the emission of radiation
Here we briefly review the discussion in [6] for (temporarily) stabilizing the wormhole
mouths by rotation. When the mouths are far away, d  re we can use the Kepler formula
r
re
Ω= (B.32)
d3
We would like to be in a regime where this is much smaller than the energy gap 1/` of the
wormhole. This would ensure that the throat is unaffected by the radiation and Unruh
quanta [20]. For the parameters in (3.27) we get the condition d > .5 ly. If the two black
holes were orbiting the galaxy we also get that the frequency would be much smaller than
1/`.
We now consider the emission of four dimensional gravity waves as well as four dimen-
sional “dark” U (1) radiation. These rates are given by

2 re2 d4 Ω6
   
dE dE π2 2
= , = 2 q 2 d2 Ω4 = re2 d2 Ω4 (B.33)
dt grav 15 G4 dt U(1) g4 3 3

We see that the U (1) radiation dominates over the gravitational one, since for us dΩ <
`Ω  1.
The time, Tlifetime , that it takes the black holes to collapse on each other is
mvg2 3 d3
Tlifetime ∝ = (B.34)
2 (dE/dt)U(1) 4 re2

For the parameters in (3.27), and for the lowest d ∼ .5 ly, this gives a lifetime of about
1016 years.

References
[1] G. Galloway, K. Schleich, D. Witt, and E. Woolgar, “Topological censorship and
higher genus black holes,” Phys. Rev. D 60 (1999) 104039, arXiv:gr-qc/9902061.

[2] G. Galloway, K. Schleich, D. Witt, and E. Woolgar, “The AdS / CFT

correspondence conjecture and topological censorship,” Phys. Lett. B 505 (2001)
255–262, arXiv:hep-th/9912119.

[3] E. Witten, “Light Rays, Singularities, and All That,” arXiv:1901.03928 [hep-th].

[4] M. S. Morris, K. S. Thorne, and U. Yurtsever, “Wormholes, Time Machines, and the
Weak Energy Condition,” Phys. Rev. Lett. 61 (1988) 1446–1449.

[5] P. Gao, D. L. Jafferis, and A. C. Wall, “Traversable Wormholes via a Double Trace
Deformation,” JHEP 12 (2017) 151, arXiv:1608.05687 [hep-th].

16
 [6] J. Maldacena, A. Milekhin, and F. Popov, “Traversable wormholes in four
dimensions,” arXiv:1807.04726 [hep-th].

[7] Z. Fu, B. Grado-White, and D. Marolf, “Traversable Asymptotically Flat

Wormholes with Short Transit Times,” Class. Quant. Grav. 36 no. 24, (2019)
245018, arXiv:1908.03273 [hep-th].

[8] Z. Fu, B. Grado-White, and D. Marolf, “A perturbative perspective on

self-supporting wormholes,” Class. Quant. Grav. 36 no. 4, (2019) 045006,
arXiv:1807.07917 [hep-th]. [Erratum: [Link]. 36, 249501 (2019)].

[9] L. Randall and R. Sundrum, “Large mass hierarchy from a small extra dimension,”
Physical Review Letters 83 no. 17, (Oct, 1999) 3370–3373.
[Link]

[10] L. Randall and R. Sundrum, “An Alternative to compactification,” Phys. Rev. Lett.
83 (1999) 4690–4693, arXiv:hep-th/9906064 [hep-th].

[11] C. G. Callan, Jr., “Monopole Catalysis of Baryon Decay,” Nucl. Phys. B212 (1983)
391–400.

[12] V. A. Rubakov, “Adler-Bell-Jackiw Anomaly and Fermion Number Breaking in the

Presence of a Magnetic Monopole,” Nucl. Phys. B203 (1982) 311–348.

[13] J. Maldacena, D. Stanford, and Z. Yang, “Diving into traversable wormholes,”

Fortsch. Phys. 65 no. 5, (2017) 1700034, arXiv:1704.05333 [hep-th].

[14] E. D’Hoker and P. Kraus, “Magnetic Brane Solutions in AdS,” JHEP 10 (2009) 088,
arXiv:0908.3875 [hep-th].

[15] J. Brown and M. Henneaux, “Central Charges in the Canonical Realization of

Asymptotic Symmetries: An Example from Three-Dimensional Gravity,” Commun.
Math. Phys. 104 (1986) 207–226.

[16] A. Strominger, “Black hole entropy from near horizon microstates,” JHEP 02
(1998) 009, arXiv:hep-th/9712251.

[17] N. Kaloper, E. Silverstein, and L. Susskind, “Gauge symmetry and localized gravity
in M theory,” JHEP 05 (2001) 031, arXiv:hep-th/0006192.

[18] J. Lee, E. Adelberger, T. Cook, S. Fleischer, and B. Heckel, “New Test of the
Gravitational 1/r2 Law at Separations down to 52 µm,” Phys. Rev. Lett. 124 no. 10,
(2020) 101101, arXiv:2002.11761 [hep-ex].

[19] S. Weinberg, “Infrared photons and gravitons,” Physical Review 140 no. 2B, (1965)
B516.

17
[20] J. R. Letaw and J. D. Pfautsch, “The Quantized Scalar Field in Rotating
Coordinates,” Phys. Rev. D22 (1980) 1345.

18