Prepared for submission to JHEP

Holographic teleportation with conservation laws:

diffusion on traversable wormholes
arXiv:2206.03434v3 [hep-th] 14 Oct 2022

Byoungjoon Ahn,a Sang-Eon Bak,b Viktor Jahnke,a Keun-Young Kima

a
School of Physics and Chemistry, Gwangju Institute of Science and Technology, 123 Cheomdan-
gwagiro, Gwangju 61005, Korea
b
Department of Physics, Arizona State University, Tempe, AZ 85287, USA
E-mail: bjahn123@gist.ac.kr, sbak2@asu.edu, viktorjahnke@gist.ac.kr,
fortoe@gist.ac.kr

Abstract: We study the effects of conservation laws on wormholes that are made traversable
by a double trace deformation. After coupling the two asymptotic boundaries of an eternal
(d + 1) dimensional AdS-Schwarzschild black hole with U (1) conserved current operators,
we find that the corresponding quantum matter stress-energy tensor violates the average
null energy condition (ANEC) in the bulk, rendering the wormhole traversable. We discuss
how the wormhole opening depends on the charge diffusion constant, how this affects the
amount of information that can be sent through the wormhole, and possible implications
for many-body quantum teleportation protocols involving conserved current operators.
Contents

1 Introduction 1

2 Gravity set-up 5

3 Boundary conditions for scalar fields 6

4 Boundary conditions for vector fields 8

4.1 Dirichlet boundary conditions for vector fields 9
4.2 Neumann boundary conditions for vector fields 10

5 Opening the wormhole with conserved current operators 10

5.1 Modified bulk two-point function 12
5.2 1-loop stress tensor 13
5.3 Average null energy 14
5.4 Bound on information transfer 16

6 Discussion 18

A ANEC violation 20

B Vector field bulk-to-boundary propagators in the hydrodynamic limit 20

C Hydrodynamic approximation and the small-k limit 23

1 Introduction

In holography [1–3], maximally extended two-sided black holes are dual to a pair of con-
formal field theories entangled in a thermofield double state [4]. The wormhole connecting
the two sides of the geometry can be made traversable by introducing a non-local coupling
between the two asymptotic boundaries. This construction was first proposed by Gao, Jaf-
feris and Wall [5], and it allows us to send a message from one side of the geometry to the
other. From the point of view of the boundary theory, this can be viewed as a teleportation
protocol [5, 6].
In the simplest instance of quantum teleportation, two distant observers, often called
Alice and Bob, share a pair of maximally entangled pair of qubits. Alice has an additional
qubit |ψi that she wants to teleport to Bob. To do so, Alice performs a measurement on
the two particles in her possession in a particular basis (the Bell basis) and reports the
result to Bob through a classical communication channel. Finally, Bob performs a unitary
operation in his qubit to obtain the desired quantum state [7]. The essential ingredients of

–1–
this teleportation protocol are share entanglement, measurement, and classical communica-
tion, and it involves only three qubits. GJW traversable wormhole is related to a new type
of teleportation protocol, the so-called traversable wormhole (TW) teleportation protocol,
which involves two copies of a strongly interacting many-body system entangled in a ther-
mofield double (TFD) state. Let us call them the left (L) and the right (R) systems, and
consider the teleportation of a qubit from the L to the R system. In this setup, the quantum
information to be teleported is initially scrambled among the degrees of freedom of the L
system. Then, after a weak coupling between the L and R systems, the quantum informa-
tion reappears (unscrambles) in the R system after a time of the order of the scrambling
time. In this protocol, the TFD state plays the same role that the maximally entangled
pair plays in conventional teleportation protocols, while measurement and classical com-
munication are used to implement the coupling between the two systems. See Fig. 1. The
TW protocol can be implemented in rather general chaotic many-body quantum systems,
but the phenomenon of many-body quantum teleportation has distinct features in the case
of systems that have an emergent gravitational description. This property makes the TW
protocol a powerful experimental tool to gain insights into the inner-working mechanisms
of gauge-gravity duality [8–10].
Traversable wormholes violate the Average Null Energy Condition (ANEC), which
states that the integral of the stress energy tensor along complete achronal null geodesics
is always non-negative Z
Tµν k µ k ν dλ ≥ 0 , (1.1)

where k µ is a tangent vector and λ is an affine parameter. In classical theories, the con-
struction of traversable wormholes is prevented by the Null Energy Condition (NEC)
Tµν k µ k ν ≥ 0, which implies (1.1) and is valid in physically reasonable theories. Gao, Jaf-
feris and Wall (GJW) construction of a traversable wormholes overcomes this difficulty by
considering quantum mechanical effects. They work in the context of the semi-classical ap-
proximation, in which the gravitational field is treated classically, but the matter fields are
treated quantum mechanically. In this context, one writes Einstein’s equations as follows

Gµν = 8πGN hTµν i , (1.2)

where Gµν is the Einstein tensor, and hTµν i is the expectation value of the stress tensor in a
given quantum state. Initially, before introducing the deformation, the wormhole connecting
the two asymptotic boundaries is not traversable, which is consistent with the fact that the
two boundary theories are not interacting. One then introduces a non-local deformation of
the boundary theory
Z
Sbdry → Sbdry + dt dd−1 x h(t, x) OL (−t, x)OR (t, x) , (1.3)

which couples the two asymptotic boundaries. Here, Sbdry denotes the action for the two
copies of the boundary theory1 , and OL,R denotes a scalar operator that acts on the d-
1
In what follows, we will not need the explicit form of Sbdry , so we did not write it here.

–2–
dimensional left/right boundary theory. For certain choices of h(t, x), the deformation (1.3)
gives rise to a stress energy tensor whose expectation value violates ANEC,
Z
hTµν ik µ k ν dλ < 0 , (1.4)

rendering the wormhole traversable. The physical picture is that the deformation (1.3)
introduces negative energy in the bulk, whose backreaction is given in terms of a negative
energy shock wave that causes a time advance for the geodesics crossing it, as opposed to
the usual time delay caused by positive-energy shock waves. In this way, a signal originated
on the left boundary can cross the wormhole and reach the right boundary after interacting
with the negative energy, as shown in Fig. 1.
|ψi

many-body
time evolution
operator

oj Classical communication
V̂

many-body many-body
time evolution time evolution
operator operator

|ψi

signal
L TFD
TFD R

Figure 1. Left panel: Penrose diagram for the GJW traversable wormhole. The negative energy
shock wave (shown in blue) causes a time advance ∆U for geodesics crossing it. The geodesics
are actually continuous, but they appear to be discontinuous because the diagram is drawn using
discontinuous coordinates. Using the equations of motion one can see that the time advance is
R
proportional to the average null energy, i.e., ∆U ∼ hTV V idV . The signal (shown in red) originated
on the left boundary and propagating along the U = 0 horizon can cross to the other side of the
R
geometry if ∆U < 0, which happens when ANEC is violated, i.e, hTV V idV < 0. Right panel:
schematic representation of the traversable wormhole (TW) teleportation protocol proposed in [10].
The quantum information (shown in red) is introduced in L at early times, gets scrambled with
the other degrees of freedom, and reappears (unscrambles) in the R system at late times after a
weak coupling between the two systems. The coupling is implemented as follows. We measure some
operator OL in Rthe the L system, obtaining one of the possible values of oj . We then apply the
operator V̂ = ei hoj OR on the R system. For further details about this teleportation protocol, we
refer to [10].

The ANEC can be violated in the GJW setup because the geodesics crossing the
wormhole are not achronal – early and late times points along the horizon can be connected
by a timelike curve passing through the directly coupled boundaries.
GJW construction of a traversable wormhole involves a two-sided BTZ black hole,
but their construction can be extended to several other gravitational setups, including 2-

–3–
dimensional black holes [6, 11–13], rotating BTZ black holes [14], asymptotically flat black
holes [15], higher-dimensional hyperbolic black holes [16], and near-extremal traversable
wormholes [17]. Other interesting developments include [18–34]. For a recent review of
wormholes and their applications in gauge-gravity duality, see [35].
Despite the existence of a growing literature about wormholes that become traversable
by a double trace deformation, most works only considered deformations involving scalar
operators. It is then natural to question if such a construction is still possible for deforma-
tions involving other types of operators, such as vector and tensor operators. In particular,
conserved current operators have different scrambling properties as compared to (non-
conserved) scalar operators, and this is expected to affect the traversability properties of
the wormhole.
In this work, we are interested in deformations involving U (1) conserved current oper-
ators. In this case, the corresponding gauge fields in the bulk display a diffusive behavior
which is expected to affect the traversability properties of the wormhole. In general, the
diffusive behavior of bulk gauge fields leads to a power-law behavior of two-point functions
[36, 37] and out-of-time-order correlators2 at late times [39]. Since the wormhole opening
can be computed as an integral involving a product of two-point functions, we expect it
to display a power-law behavior at late times. Moreover, we expect the wormhole opening
and consequently the bound on information transfer to depend on the transport properties
of the black hole horizon. In particular, we would like to understand the effects of con-
servation laws on many-body TW protocols and obtain the expected behavior for systems
that admit a dual gravitational description. Is teleportation favored in this case? Or do
the different scrambling properties of conserved currents result in a less efficient teleporta-
tion protocol? The benchmark holographic behavior is well understood in the case where
the double trace deformation involves scalar field operators but has not yet been explored
for deformations involving vector and tensor operators, and this might be relevant in the
experimental realization of TW teleportation protocols, especially because of this possible
interplay between traversability and hydrodynamic behavior.

Summary of results
We consider a maximally extended AdS-Schwarzschild black hole in (d + 1) dimensions.
When the double trace deformation involves U (1) conserved current operators, we find that
the average null energy reads

∞ − d+1
h̃ d2 (d − 1) 3−d V0
Z  
1 2
dV hTV V i = − d−1 d+4 3 D 2 log + V 0 , (1.5)
V0 L 2 π 1 + V02 V0

where T is the black hole temperature and D = T Dc is related to the charge diffusion
constant Dc (5.16), and h̃ = h T d−2 is the dimensionless coupling constant. We considered
an instantaneous perturbation at time t0 , applying the hydrodynamic approximation. Note
2
For a review on out-of-time-order correlators, we refer to [38].

–4–
that there is an appropriate choice of the sign of the coupling constant h̃ that leads to the
ANEC violation.
Since the charge diffusion constant affects the average null energy, it is reasonable
to assume that it also affects the amount of information that can be transferred through
the wormhole. We show that the number of particles Nbits that we can send through the
wormhole without closing it is parametrically bounded as

rhd−1 3−d 3−d

Nbits . h̃ d−1
D 2 ∼ h̃ (T L)d−1 D 2 (1.6)
L
2
where rh = 4πTd L is the horizon radius, and L is the AdS length scale. Even though D
is just a fixed number in our setup, we write our results in terms of it to emphasize that
the wormhole opening depends on transport properties of the black hole horizon. In more
complicated setups, D might depend on additional parameters, and, in principle, its value
can be tuned to optimize traversability.

Organization of this work

This work is organized as follows. In Sec. 2, we introduce our gravity setup. In Sec. 3, we
review the boundary conditions for scalars fields associated with the GJW construction
of a traversable wormhole. In Sec. 4, we discuss boundary conditions for vector fields. In
Sec. 5, we consider a double trace deformation involving U (1) conserved current operators
and show that it leads to a violation of the average null energy condition. In Sec. 6, we
discuss our results. We relegate some technical details to the appendices A and B.

2 Gravity set-up

We consider a two-sided AdS-Schwarzschild black hole in (d+1) dimensions. In Schwarzschild

coordinates (z, t, xi ), the line element reads

L2 dz 2
 
2 2 i i
ds = 2 −f (z)dt + + dx dx , (2.1)
z f (z)
 d
z
f (z) = 1 − , (2.2)
zh

where L is the AdS length scale. We set L = 1 for simplicity. Here, (t, xi ) are the boundary
theory coordinates, with i running from 1 to d − 1, and z is the AdS radial coordinate. The
above solution has a horizon at z = zh and the boundary is located at z = 0. The black
hole’s inverse temperature is given by β ≡ 1/T = 4πz d .
h

To analyze the equations of motion near the horizon, it is convenient to define the
tortoise coordinate
 d !
dz 1 d+1 z
dr = − , or r(z) = −z 2 F1 1, , ; , (2.3)
f (z) d d zh

–5–
Figure 2. Kruskal diagram (Left) and Penrose diagram (Right) for a two-sided AdS-Schwarzschild
black hole.

in terms of which the boundary is located at r = 0 and the horizon at r = −∞. In terms
of the tortoise coordinate, the line element reads

1 
ds2 = 2 2 i i

f (z(r))(−dt + dr ) + dx dx . (2.4)
z(r)2

Finally, to describe the globally extended spacetime, we introduce Kruskal-Szekeres

coordinates (U, V, xi ) as

− 2π (t−r) 2π
(t+r)
U = −e β , V = +e β , (right exterior region)
− 2π (t−r) 2π
(t+r)
U = +e β , V = −e β , (left exterior region)
− 2π (t−r) 2π
(t+r)
U = +e β , V = +e β , (future interior region)
− 2π (t−r) 2π
(t+r)
U = −e β , V = −e β . (past interior region)

In terms of these coordinates, the line element takes the form

dxi dxi β 2 f (z)

ds2 = gU V (U V )dU dV + , gU V (U V ) = . (2.5)
z2 4π 2 U V z 2
where z is also a function of U V . In these coordinates the horizon is located at U = 0 or
at V = 0. The left and right boundaries are located at U V = −1 and the past and future
singularities at U V = 1. The Kruskal and Penrose diagrams for this geometry are shown
in Fig. 2.

3 Boundary conditions for scalar fields

In this section, we review the boundary conditions for scalar fields that are associated with
the GJW traversable wormhole. Let’s consider a minimally coupled scalar field with mass

–6–
m propagating in the background (2.1). Near the boundary, the field behaves as

φ(z, y) = α(y)z ∆− + β(y)z ∆+ , (3.1)

where y = (t, x) denotes a boundary point, and

d p
∆± = ± ν with ν = d2 /4 + m2 . (3.2)
2

For m2 > −d2 /4 + 1, the first term in (3.1) is non-normalizable and is associated to a
deformation of the boundary theory of the form
Z
Wα = dd y jα (y) Oα (y) with α(y) = jα (y) , (3.3)

where the single trace operator Oα (y) has scaling dimension ∆+ = d/2 + ν. The second
term in (3.1) is normalizable, and is related to the expectation value of Oα

β(y) = 2ν hOα (y)i . (3.4)

For −d2 /4 < m2 < −d2 /4 + 1, both terms in (3.1) are normalizable. In this regime, we are
free to impose boundary conditions on either α or β. Each choice corresponds to a different
boundary theory. In the boundary theory in which α(β) is fixed, the bulk field φ is dual
to an operator of dimension ∆+ (∆− ). We refer to these boundary theories as CFT∆+ and
CFT∆− . In particular, the so-called alternative boundary condition

β(y) = jβ (y) , (3.5)

is associated to deformations of the form

Z
Wβ = dd y jβ (y)Oβ (y) , (3.6)

in which the boundary operator Oβ (y) has scaling dimension ∆− = d/2 − ν. Note that,
since −d2 /4 < m2 < −d2 /4 + 1, one has d−2 d
2 < ∆− < 2 . In this case, the leading order
term is now associated to the expectation value of Oβ (y):

α(y) = −2ν hOβ (y)i . (3.7)

dd y β α.
R
The above equations allow us to symbolically write the deformation as Wβ ∼
A linear boundary condition relating the faster and the slower falloff parts

β = hα, (3.8)

corresponds to a double trace deformation of the form [40]

Z Z
Wβ ∼ d y h α ∼ dd y h Oβ2 ,
d 2
(3.9)

–7–
which is a relevant deformation, because ∆− < d/2. Starting from the CFT∆− , the defor-
mation (3.9) produces a renormalization group flow which is expected to end at the CFT∆+
in the infrared. We now explain how the linear boundary condition (3.8) can be used to
construct a traversable wormhole.
GJW construction of a traversable wormhole relies on the use of a relevant deformation,
because in this case the near boundary geometry is not modified by backreaction in an
uncontrolled way [5]. In this way, their construction can be shown to be embeddable in a
UV complete theory of gravity. To have a relevant deformation, they consider a massive
scalar field with mass in the range −d2 /4 < m2 < −d2 /4 + 1 with the alternative boundary
condition, in which the boundary operator has dimension ∆− = d/2 − ν. For the fields
propagating in the left and right exterior regions, we expect the following near boundary
behaviors

φ(z → 0L , y) = αL (y)z ∆− + βL (y)z ∆+ , (3.10)

∆− ∆+
φ(z → 0R , y) = αR (y)z + βR (y)z . (3.11)

In order to have a deformation coupling the two asymptotic boundaries, GJW impose the
following boundary conditions

βR = h αL , βL = h αR (3.12)

which correspond to a non-local double trace deformation of the form

Z Z
(L) (R)
Wβ ∼ d y h αL αR ∼ dd y h(y) Oβ (y) Oβ (y) .
d
(3.13)

The deformation (3.13) modifies the 1-loop expectation value of the scalar field stress-
energy tensor. For some choices of h(y), this leads to a violation of the ANEC that renders
the wormhole traversable.

4 Boundary conditions for vector fields

In this section, we briefly review possible boundary conditions for vector fields in AdS/CFT.
A general discussion about this topic was first presented in [41]. We consider a massless
vector field Aµ propagating on the background (2.1) with action

d+1 √ √
Z Z
1 µν
S=− d r −g Fµν F + dd y −γ nµ F µν Aν , (4.1)
4 M ∂M

where r = (z, t, x) ∈ M denotes a bulk point and y = (t, x) ∈ ∂M denotes a boundary

point. Here, γ denotes the determinant of the induced metric on the boundary, and nµ is
the outward pointing unit vector normal to ∂M. The equation of motion for Aµ reads

∇ν F µν = 0 . (4.2)

–8–
4.1 Dirichlet boundary conditions for vector fields
Near the boundary (z → 0), the gauge field behaves as

Aµ (z → 0) = aµ + bµ z d−2 + . . . (4.3)

In general, when d ≥ 4, the leading term aµ is non-normalizable and must be fixed. This
corresponds to imposing a Dirichlet boundary condition on Aµ . From the point of view of
the boundary theory, such boundary condition correspond to a deformation by a global
U (1) conserved current operator
Z
WA = dd y aµ J µ , (4.4)
∂M

with aµ acting as a source for J µ . Note that, since Aµ = aµ at the boundary, we can write
WA = ∂M dd y Aµ J µ . The expectation value of the conserved current is given by
R

hJµ i ∼ bµ . (4.5)

This allow us to write WA ∼ ∂M dd y aµ bµ . Similarly to the scalar field case, a double

R

trace deformation can be introduced by imposing a linear relation between aµ and bµ , i.e.,
aµ = Qµν bν , which leads to
Z Z
WA ∼ dd y Qµν bµ bν ∼ dd y Qµν Jµ Jν , (4.6)
∂M ∂M

which is an irrelevant deformation, with dimension 2d − 2, because the scaling dimension

of J µ is d − 1. Here Qµν is a set of parameters that we will specify later.
Just like in the scalar field case, one can construct a traversable wormhole by coupling
the two asymptotic boundaries of an asymptotically AdS two-sided black hole with con-
served current operators. This can be done as follows. First, we write the near boundary
behavior of the gauge field as follows
(R) (R)
Aµ (z → 0R ) = aµ + bµ z d−2 + . . . (4.7)
(L) (L)
Aµ (z → 0L ) = aµ + bµ z d−2 + . . . (4.8)

where the superscripts R and L denote the right anf left exterior regions, respectively.
Then, an irrelevant double trace deformation can be introduced by imposing the following
boundary condition
aµ(R) = Qµν b(L)
ν , aµ(L) = Qµν b(R)
ν , (4.9)

which leads to a deformation of the form

Z
WA ∼ dd y Qµν Jµ(L) Jν(R) . (4.10)
∂M

We show in the next section that for some choices of Qµν the deformation (4.10) leads to
a violation of the ANEC, rendering the wormhole traversable.

–9–
 In this work, we consider the deformation (4.10) because we are interested in the
hydrodynamic limit of the gauge fields, associated to long-distance physics. We can think
about our setup as the IR fixed point of a renormalization group (RG) flow. As explained
in the next section, a relevant deformation can be introduced if one considers gauge fields
satisfying Neumann boundary conditions. In this case, the initial UV theory flows under the
RG flow to an IR fixed point in which the gauge fields respect Dirichlet boundary conditions,
which corresponds to our case. By that reasoning, we expect that our construction has a
well-defined UV fixed point.

4.2 Neumann boundary conditions for vector fields

In the previous section, we adopted the most common boundary condition for vector fields
in AdS/CFT, which corresponds to imposing Dirichlet boundary conditions on Aν . As
pointed out in [41], for d = 3, 4 and 5, it is possible to consider an alternative boundary
condition for vector fields, in which one imposes Neumann boundary condition on the gauge
field by fixing
√
F ν = −γ nµ F µν , (4.11)

at the boundary. This boundary condition fixes the value bµ , but leaves aµ unconstrained.
For d = 3, this corresponds to a deformation of the boundary theory by an operator OF ν ,
with dimension one. One can show that this operator is a U (1) gauge field [41–43].
Imposing the condition bµ = Qµν aν corresponds to introducing a relevant double trace
deformation of the form [41]
Z
WF = dd y Qµν OF µ OF ν , (4.12)
∂M

which has dimension 2. Under the deformation (4.12), the boundary theory associated to
Neumann boundary conditions flows in the infrared to the boundary theory associated to
Dirichlet boundary conditions [41].
Once again, an wormhole connecting the two sides of an asymptotically AdS two-sided
black hole can be made traversable by introducing the following relevant deformation
Z
(L) (R)
WF = dd y Qµν OF µ OF ν , (4.13)
∂M

which is associated to boundary conditions of the form

bµ(R) = Qµν a(L) µ µν (R)

ν , b(L) = Q aν . (4.14)

5 Opening the wormhole with conserved current operators

In this section, we propose a generalization of the GJW traversable wormhole that involves
a double trace deformation constructed out of conserved currents
Z
δH(t) = dd−1 x Qµν (t, x)Jµ(L) (−t, x)Jν(R) (t, x) . (5.1)

– 10 –
For some choices of Qµν , the deformation (5.1) introduces negative null energy in the bulk,
which makes the wormhole traversable. For simplicity, we consider the coupling between
(R) (L)
the two conformal field theories as involving only Jt and Jt , i.e., we set Qµν as

Qµν (t1 , x1 ) = δtµ δtν h(t1 , x1 ) . (5.2)

Here, h(t1 , x1 ) controls the coupling between the two boundary theories. We choose a
perturbation that is instantaneous in time t0 and uniform in the transverse space
 
2π
h(t1 , x1 ) = h δ (t1 − t0 ) = h V0 δ(V1 − V0 ) , (5.3)
β

where h is a constant. By dimensional analysis, h has dimensions of [E]2−d . Later, we will

factor out this dimensional dependence and write this constant as h = h̃ T 2−d .
The traversability of the wormhole can be measured near the event horizon (U = 0)
by the average null energy Z
A= dV TV V , (5.4)

where TV V is the V V component of the Maxwell stress energy tensor. To study the effect
of the deformation on TV V , we compute the 1-loop Maxwell stress energy tensor using a
point splitting technique.
The Maxwell stress energy tensor is given by

1
Tµν = −Fαµ Fνα + gµν Fαβ F αβ . (5.5)
4
To simplify our calculation, we consider that only AV and AU are non-zero, and we con-
sider a metric of the form ds2 = gU V dU dV + gij dxi dxj . With these assumptions, the V V
component of the stress energy tensor reads

TV V = −FαV FVα = −FiV FVi = −g ij FiV FjV . (5.6)

Now, using FiV = ∂i AV − ∂V Ai = ∂i AV and the point splitting method, we write

hTV V i = − lim
0
g ij h∂i AV (r)∂j 0 AV (r0 )i (5.7)
r →r
= − lim
0
g ij ∂i ∂j 0 hAV (r)AV (r0 )i,
r →r

where hAV (r)AV (r0 )i is the bulk two-point function between two points r = (z, t, x), r0 =
(z 0 , t0 , x0 )3 .
3
Note that hTV V i can be written in terms of the gauge invariant quantity FiV = ∂i AV − ∂V Ai = ∂i AV ,
i.e, hTV V i = limr0 →r g ij hFiV (r)FjV (r0 )i. This makes it clear that the expression is gauge invariant, as
expected on physical grounds.

– 11 –
5.1 Modified bulk two-point function
To compute stress energy tensor involving the deformation δH in (5.1), we first compute
the modified bulk two-point function. Due to the deformation, the gauge field operators
R t
−i dt δH(t )
in (5.7) evolves in time with the operator U (t, t0 ) = T e t0 1 1
in the interaction
picture. Then, the modified bulk two-point function in (5.7) is given by
(H) (H)
GV V (r; r0 ) ≡ hAV (r)AV (r0 )i
(I) (I)
= hU (t, t0 )−1 AV (r)U (t, t0 )U (t0 , t0 )−1 AV (r0 )U (t0 , t0 )i , (5.8)

where we used the superscripts H and I to distinguish between the Heisenberg and in-
teraction picture, respectively. For notational simplicity, we will omit the superscript I in
what follows.
Working at first order in perturbation theory, we expand
Rt Z t
−i dt1 δH(t1 )
U (t, t0 ) = T e t0 =1−i dt1 δH(t1 ) , (5.9)
t0
(0) (1)
GV V (r; r0 ) = GV V (r; r0 ) + GV V (r; r0 ) , (5.10)

(0)
where GV V (r; r0 ) is the unperturbed two-point function of AV . The first correction to the
two-point function of AV is obtained as follows,
Z t0 Z t
(1)
GV V (r; r0 ) =i dt1 hAV (r)[δH(t1 ), AV (r )]i + i 0
dt1 h[δH(t1 ), AV (r)]AV (r0 )i (5.11)
t0 t0
Z t0
=i dt1 dd−1 x1 Qµν hAV (r)[Jµ(L) (−t1 , x1 )Jν(R) (t1 , x1 ), AV (r0 )]i + (t0 ↔ t)
t0
Z t0
=i dt1 dd−1 x1 Qµν hAV (r0 )Jµ(L) (−t1 , x1 )ih[Jν(R) (t1 , x1 ), AV (r)]i + (t0 ↔ t) .
t0

In the second line, we put (5.1) into the expression. In the third line, we assume that
the fields AV (r) and AV (r0 ) act on the right exterior region. Then, we can use large N
factorization and the fact that operator on the right exterior region commute with the
operators on the left boundary.
Let us introduce the Wightman and retarded bulk-to-boundary propagators for the
right exterior region as

w
Kµν (z, t, x; t1 , x1 ) ≡ −ihAµ (z, t, x)Jν(R) (t1 , x1 )i , (5.12)
ret
Kµν (z, t, x; t1 , x1 ) ≡ −ih[Aµ (z, t, x), Jν(R) (t1 , x1 )]i . (5.13)

for t > t1 . Then, the expression of the first order correction to the bulk two-point function

– 12 –
of AV can be rewritten as follows,
Z t0
(1) iβ
GV V (r; r0 ) =i dt1 dd−1 x1 Qµν KVwµ (z 0 , t0 , x0 ; −t1 − , x1 )KVretν (z, t, x; t1 , x1 ) + (t0 ↔ t)
t0 2
Z t0
iβ
=i dt1 dd−1 x1 h(t1 , x1 ) KVwt (z 0 , t0 , x0 ; −t1 − , x1 )KVrett (z, t, x; t1 , x1 ) + (t0 ↔ t) .
t0 2

(L)
In the first line, we used the KMS condition [44] which gives the relation between Jµ and
(R)
Jµ , and in the second line, we included the perturbation (5.2).
The bulk-to-boundary propagators (5.12) and (5.13) in the background geometry (2.1)
are not known in general, but they can be computed using a hydrodynamic approximation.
In this approximation, we focus on large times and large distances. This translates to solving
our equations of motion for low energy modes with large wavelengths. In this limit, the
bulk gauge field displays a diffusive behavior, controlled by a charge diffusion constant. For
µ = ν = t, the above propagators can be written as (see App. B or [39])

dd−1 k eik·(x−x1 )
Z
KVrett (V, x; t1 , x1 ) = 2i T Dc θ(V − e 2πT t1
) ∂V Dc k 2
, (5.14)
(2π)d−1
(V e−2πT t1 − 1) 2πT

dd−1 k eik·(x−x1 )
Z
1
KVwt (V, x; t1 , x1 ) = T Dc ∂V Dc k 2
, (5.15)
(2π)d−1 2 sin Dc k2
 
2T (1 − V e−2πT t1 ) 2πT

where Dc is the charge diffusion constant:

d 1
Dc = . (5.16)
d − 2 4πT
(1)
Using the above expressions, GV V is obtained as

V0 0
dd−1 k Dk2 eik·(x−x )
Z Z
(1) dV1
GV V = −2T 2
Dc2 h(V1 ) +(V ↔ V 0 ) ,
V0 2πT V1 (2π)d−1 2 sin(πDk ) (V /V1 − 1)Dk +1 (1 + V1 V 0 )Dk +1
(5.17)
Dc 2
where Dk = 2πT k , and V1 = e2πT t1 .

5.2 1-loop stress tensor

The V V component of the 1-loop expectation value can then be computed as
(1)
hTV V i = − lim
0
g ij ∂i ∂j 0 GV V (r; r0 ) , (5.18)
r →r

(0)
where we used the fact that GV V does not contribute to the opening of the wormhole.
Using (5.3) and performing the integral in V1 , the expectation value of the stress-energy
tensor be written as

h d2 Dc2 dd−1 k Dk2 k2

Z
hTV V i = . (5.19)
(2π)3 T (2π)d−1 2 sin(πDk ) (V /V0 − 1)Dk +1 (1 + V V0 )Dk +1

– 13 –
Writing k in spherical coordinates and substituting the integral of the angular part to
d−1
2π 2
Γ( d−1
, we find
2 )

d−1
∞ − Dc k2 −1
h d2 Dc3
Z  
2π 2 V 2πT
hTV V i ' d+9 d−1

dkk d+2 − 1 (1 + V0 V )
2 2 π d+4 T 2 Γ 2 0 V0
3−d
h̃ d2 (d2 − 1) V0 D 2
= d+9 i d+3 . (5.20)
2 2 π2 (V − V0 )(1 + V V0 ) h (V −V0 )(1+V0 V ) 2
log V0

In the first line of (5.20), we use a small k approximation in the integrand4 and only keep
the leading order terms in k 2 . Since we don’t have the full solution of the gauge field
propagator on this geometry, we introduce a regulator to properly study the effects of
diffusion. The low frequency modes contribute dominantly to the late-time behavior of the
stress energy tensor and the approximation gives qualitatively correct results.
We used that h = h̃ T 2−d and factor out the temperature dependence of the diffusion
constant as Dc = D/T . Fig. 3 shows the behavior of hTV V i as a function of V for several
values of V0 . Note that hTV V i is negative for h < 0. The stress energy tensor has a divergent
behavior near the insertion time V0 . However, as we will see, this is an integrable divergence
so that the integral of hTV V i from V0 to ∞ gives finite values.
hTV V i

V
Figure 3. Stress-energy tensor (5.20) versus V for V0 = 1, V0 = 2, and V0 = 3. Here, we set d = 4,
1
T = 2π and h = −1.

5.3 Average null energy

In this section, we compute the average null energy in the presence of double trace defor-
mation involving U (1) conserved current operators. By integrating (5.18) along complete
4
As we explain in the next subsection, another approach is to introduce a momentum cutoff kmax and
compute this integral numerically. This gives qualitatively similar results to the one obtained using a small
k approximation.

– 14 –
achronal null geodesics, the average null energy can be computed as
Z ∞
R∞ (1)
V0 dV hTV V i = − lim
0
dV g ij ∂i ∂j 0 GV V (V, x; V 0 , x0 ) (5.21)
r →r V0
∞ Z V
d2 Dc2 dd−1 k k2 Dk2
Z Z
dV1
= dV h(V1 ) .
(2π)3 T V0 V0 V1 (2π)d−1 2 sin(πDk ) (V /V1 − 1)Dk +1 (1 + V V1 )Dk +1

Using that h(V1 ) = hV0 δ(V1 − V0 ), and integrating in V1 , in the angular part, and in V , we
obtain
d−1 −1−2Dk
∞ ∞
h d2 Dc2 2π 2 dkk d 4Dk Dk2 Γ(−Dk )Γ(Dk + 1/2)
Z Z 
1
dV hTV V i = √ + V0 (5.22)
.
(2π)3 T Γ d−1 (2π)d−1


V0 2 0 2 π sin(πDk ) V0

To evaluate the remaining integral in k we should keep in mind that the high-k modes
ω
are inconsistent with the hydrodynamic approximation, which assumes 2πT  1 and
k
 1. When V > 1, the factor of 4 Dk (V + 1/V )−2D k suppresses these high-k modes,
2πT 0 0 0
and our calculation is consistent with the hydrodynamic approximation. However, when
V0 = 1, the factor of 4Dk (V0 + 1/V0 )−2Dk becomes 1, and no longer suppresses the high-
k modes. To make our results always consistent with the hydrodynamic approximation
we should introduce a momentum cutoff kmax to get rid of these high-k modes. How-
ever, if we still integrate in k from zero to infinity, but use a small-k approximation for
Dk
1 4√ −2Dk in the integrand suppresses
2 sin(πDk ) π Γ(−Dk )Γ(Dk + 1/2), the factor of (V0 + 1/V0 )
the contribution from the high-k modes, and we can get an analytic result which is qualita-
tively similar to the numerical results when kmax ∼ 2πT . Employing the above-mentioned
small-k approximation, we obtain

∞ d−1 −1 Z ∞ − Dk2

2hzh2 T Dc2
Z  
2π 2 1 d 1 πT
dV hTV V i = − d−1

+ V0 dkk + V0
V0 π Γ 2 (2π)d V0 0 V0
− d+1
h̃ d2 (d − 1) 3−d V0
 
1 2
= − d−1 d+4 3 D 2
2 log + V0 , (5.23)
L 2 π 1 + V0 V0

where in the second line we restored the dependence on the AdS length scale L, and factor
out the dimensional dependence of the coupling and diffusion constant as h = h̃ T 2−d and
Dc = D/T . Note that ANEC is violated and the wormhole becomes traversable for h̃ < 0.
Our corresponding numerical results are shown in Appendix C.
The relation between the average null energy and the wormhole opening ∆U reads (see
App. A) Z
4πGN
∆U = hTV V idV . (5.24)
(d − 1)
Combining the above result with (5.23), we find
− d+1
h̃GN d2
 
3−d V0 1 2
∆U = − d−1 d+2 2 D 2 2 log + V0 . (5.25)
L 2 π 1 + V0 V0

– 15 –
 R
In Fig. 4, we plot TV V dV versus V0 for several spacetime dimensions using (5.23). The
R
wormhole opening TV V dV decreases as a function of V0 , decreasing faster for bigger values
of d. This suggests that it is more difficult to open the wormhole in higher dimensional
setups, in accordance with previous results in the literature [16].
TV V dV
R

V0
R
Figure 4. TV V dV computed using (5.23) versus V0 for d = 4, d = 5 and d = 6. Here we set
1
T = 2π and h = −1. Qualitatively similar results are obtained if we used the numerical method
described in App. C.

It is interesting to note that the wormhole opening scales with the charge diffusion
3−d
constant as ∆U ∼ Dc 2 , which implies that the wormhole opening decreases as we increase
Dc , i.e., the diffusion properties of the gauge field make it harder to cross the wormhole.
In the GJW traversable wormhole, a signal is able to cross the wormhole because the
negative energy introduced by the double trace deformation produces a time advance in
the trajectory of the signal, which allows it to escape from the interior region. The same
thing happens when the double trace deformation involves conserved current operators,
however, in this case, due to the diffusion properties of the gauge field, the interaction
between the signal and the negative energy gets weaker because their collision is more
delocalized. As a consequence, the time advance ∆U is smaller. On the other hand, a
smaller diffusion constant could improve the conditions for traversability when d > 3.

5.4 Bound on information transfer

A traversable wormhole allows us to send a signal from one side of the geometry to the
other side. In this section, we derive parametric bounds on the amount of information
that can be transferred through the wormhole. The basic idea is that the backreaction of
each particle that we send through the wormhole has the effect of reducing |∆U |, i.e., it
closes the wormhole a bit. Therefore, if we try to send a signal with too many particles,
the corresponding backreaction will eventually reduce |∆U | to zero, closing the wormhole
completely.
Consider a message with Nbits particles, each one with momentum peach U , such that
the total momentum of the signal is ptotU = N peach . Using the uncertainty principle and
bits U
requiring that the signal ‘fits’ in the opening of the wormhole. The uncertainty principle

– 16 –
implies that
peach
U ∆Ueach & 1 . (5.26)

Let us assume that the double trace deformation open the wormhole by an amount ∆U .
Requiring that each particle’s wave function fits in the wormhole opening, we find that

∆Ueach ≤ |∆U |. (5.27)

Combining (5.26) and (5.27), we find

1
. ∆Ueach ≤ |∆U | . (5.28)
peach
U

ptot
Using (5.28), and the fact that Nbits = U
peach
, we find
U

Nbits . ptot
U |∆U | . (5.29)

Finally, we require that the backreaction of the signal is small. To impose this condition,
we need to model the stress energy tensor of the signal. For a very early perturbation, the
d−1
stress energy of the signal takes a simple form, which can be written as TU U ∼ ptot
U zh δ(U ).
d−1
The probe approximation requires GN ptot U zh  1, and it breaks down roughly at

zh−d+1
ptot
U . . (5.30)
GN

Now, using (5.30) and our result for |∆U |, we can derive a parametric bound on Nbits . The
wormhole opening ∆U given in (5.23) should be thought as valid for t0 & 1/T because it
was derived in the hydrodynamic limit. To obtain the maximal possible value of |∆U | that
1
is also roughly consistent with the hydrodynamic approximation, we set t0 = 2πT in (5.23),
obtaining the following scaling behavior

GN 3−d
∆U ∼ h̃ D 2 . (5.31)
Ld−1
Finally, plugging (5.30) and (5.31) into (5.29), and we obtain
 r d−1 3−d
h
Nbits . h̃ D 2 . (5.32)
L

d−1
where rh = 1/zh . The fact that the bound is proportional to rLh is consistent with
the analysis of [20] for higher dimensional traversable wormholes. The scaling behavior of
Nbits with the dimensionless diffusion constant D suggests that the diffusive properties of
the gauge field play an important role in determining the amount of information that can
be sent through the wormhole. We expect this scaling behavior to also take place in other
black hole solutions and in constructions in which the deformation involves other types of
conserved current operators. In our gravity setup the dimensionless diffusion constant is a
d
fixed number that depends on the dimensionality of the space-time, i.e., D = 4π(d−2) , but in

– 17 –
more complicated geometries D may also depend on additional parameters. For example,
in Einstein-Maxwell-Axion models [45] the diffusion constant D is known to depend on
m/T , where m is a momentum relaxation parameter. See for instance Eq. (2.5) of [46].
In particular, D becomes arbitrarily small as one increases the value of m, what strongly
suggests that the large-m limit favors traversability for d > 3.

6 Discussion

We constructed a traversable wormhole by coupling the two asymptotic boundaries of a

two-sided AdS-Schwarzschild black hole with U (1) conserved current operators. The non-
local coupling introduces a quantum correction to the expectation value of the stress-energy
tensor that violates ANEC, rendering the wormhole traversable.
The double trace deformation involving U (1) conserved current operators is dual to a
gauge field fluctuation in the bulk. We found that the diffusive properties of the gauge field
control the behavior of the wormhole opening ∆U , see Fig. 5. For a deformation introduced
at time t0 & 1/T , we found the following scaling behavior

1
∆U ∼ e−2πT t0 . (6.1)
(2πT t0 )(d+1)/2

The power law part is reminiscent of the power law behavior observed in two-point func-
tions [36, 37] and out-of-time-order correlators involving conserved current operators, e.g.
hO(t)Jt (0)O(t)Jt (0)i ∼ t−(d−1)/2 [39], and it is related to the diffusive behavior of the bulk
gauge field. The exponential part also takes place in the large time limit when the double
trace involves scalar operators. For scalar operators of dimension ∆, one finds [16, 20]
 2∆+1
V0
∆Uscalar ∼ , with V0 = e2πT t0 . (6.2)
1 + V02

In the limit t0  1/T , one obtains ∆Uscalar ∼ e−2πT (2∆+1)t0 . Comparing with (6.1), one
can see that the exponential decrease with t0 is present in both cases, but the power law
part only appears in the case involving conserved currents.

– 18 –
Figure 5. The boundary double trace deformation with U (1) conserved current operators corre-
sponds to gauge field fluctuations in the bulk. One can view the red lines as photons that are
delocalized with different longitudinal energies, and it indicates the diffusive behavior of the bulk
gauge fields. The right and left past (future) horizons intersect at the point P1 (P2 ). The size of the
wormhole opening ∆U can be obtained by integrating along an achronal null geodesic, which is
represented by the blue line.

We derived a parametric bound on the amount of information that can be transferred

through the wormhole. In particular, we showed that the bound on information transfer
−(d−3)/2
scales with the charge diffusion constant as Dc . This suggests that for d > 3, a small
diffusion constant would allow us to send more information through the wormhole, while
a large diffusion constant would make it harder to cross the wormhole. In particular, it
would be interesting to investigate this in the Einstein-Maxwell-Axion model proposed in
[45], because in this case the charge diffusion constant depends on a momentum relaxation
parameter and becomes arbitrarily small when this parameter is large [46]. This strongly
suggests that momentum relaxation favor traversability when d > 3. More generally, this
type of investigation might help us to find systems which favor traversability on the gravity
side, and many-body quantum teleportation on the boundary theory side.
Another interesting future direction would be to consider double trace deformations
involving other types of conserved currents, and study how the wormhole opening depends
on the corresponding transport coefficients.

Acknowledgments

We would like to thank Mitsuhiro Nishida, Kyung Kiu Kim, Hyun Seok Yang and Kyoung-
Bum Huh for helpful discussions. This work was supported by Basic Science Research
Program through the National Research Foundation of Korea(NRF) funded by the Min-
istry of Science, ICT & Future Planning (NRF- 2021R1A2C1006791), the GIST Research
Institute(GRI) and the AI-based GIST Research Scientist Project grant funded by the

– 19 –
GIST in 2022. V. Jahnke was supported by Basic Science Research Program through the
National Research Foundation of Korea(NRF) funded by the Ministry of Education(NRF-
2020R1I1A1A01073135). B. Ahn was also supported by Basic Science Research Program
through the National Research Foundation of Korea funded by the Ministry of Education
(NRF-2020R1A6A3A01095962, NRF-2022R1I1A1A01064342).

A ANEC violation

In this section, we review how the wormhole opening is related to the violation of ANEC.
We consider a background metric gµν that satisfies (1.2) with vanishing right hand side.
Assuming the double trace deformation introduces a non-zero expectation value for the
stress energy tensor hTµν i ∼ O(), we consider fluctuations gµν → gµν + hµν , with
hµν ∼ O(). Using Kruskal coordinates, we find that at U = 0,

+ V ∂V2 hij
 
1 ij ∂V hij
−Λ hV V + δ = 8πGN hTV V i . (A.1)
2 V zh2
Integrating both sides with respect to V we find
Z Z
8πGN hTV V idV = −Λ hV V dV . (A.2)

The wormhole opening can then be computed as

Z Z
1 4πGN
∆U = − hV V dV = hTV V idV . (A.3)
gU V (0) (d − 1)
2
where in the last equality we used that Λ = − d(d−1)
2L2
and gU V (0) = − 4Ld .

B Vector field bulk-to-boundary propagators in the hydrodynamic limit

In this section, we derive bulk-to-bulk propagators for a gauge field propagating in the
background (2.1). We start by writing Maxwell-Einstein equations with a source term
√
∂µ ( −gg µρ g νσ Fρσ ) = Jbulk
ν
, (B.1)

where the source satisfies the equation

µ
∂µ Jbulk = 0. (B.2)

The gauge field can then be written as

Z
Aµ (r) = dd+1 r Gµν (r − r0 )Jbulk
ν
(r0 ) , (B.3)

where r = (r, t, x) and Gµν (r − r0 ) solves (B.1) with a delta function as the source.

– 20 –
 Following [39], we use the gauge Ai = 0. By writing the propagators and the sources
in momentum space

dωdd−1 k
Z
Gµν (t, x) = Gµν (r, ω, k)e−iωt eik·x ,
(2π)d
dωdd−1 k ν
Z
ν
Jbulk (t, x) = Jbulk (r, ω, k)e−iωt eik·x , (B.4)
(2π)d

we can find a solution for Gµν (r, ω, k) in the hydrodynamic limit. To find Gtt , we require
r
Jbulk t
= 0, and Jbulk (r, ω, k) = δ(r − r0 ). Then (B.2) implies Jbulk x (r, ω, k) = ω δ(r − r 0 ). The
k
t component of (B.1) can then be written as [39]
" !#
1 (ω 2 − k 2 f )zhd−3
∂r2 Gtt + ∂r log ∂ r G tt + (ω 2
− k 2
f )G tt = δ(r − r0 ) .
zhd−3 (ω 2 − k 2 f ) k2
(B.5)
A solution to the above equation satisfying in-falling boundary conditions at the horizon
and Neumann boundary conditions on the boundary corresponds to the retarded bulk-to-
bulk propagator in momentum space:
Z
0
Gret
tt (r, r , ω, k) = −i dt dd−1 xeiωt−ik·x θ(t)h[At (r, t, x), At (r0 , 0, 0)]i . (B.6)

For small ω and small k, we can show that the solution in the near-horizon region reads
0
0 ω 2 e−iω(r−r )
Gret
tt (r, r , ω, k) = 2 . (B.7)
k iω − Dc k 2
d 1
where Dc = d−2 4πT is the charge diffusion constant.
The Wightman function i Gw 0
tt = hAt (r, ω, k)At (r , ω, k)i can be written as [47]

0 eω/2T 0
Gw
tt (r, r , ω, k) = 2i Im Gret
tt (r, r , ω, k) . (B.8)
eω/T − 1
In position space, the above bulk-to-bulk propagators can be written as

dd−1 k dω ret
Z
0 0 0 0 0
Gret
tt (r, t, x; r , t , x ) = −i d−1
Gtt (r, r0 , ω, k)e−iω(t−t ) eik·(x−x )
(2π) 2π
0 0
dd−1 k dω ω 2 e−iω(r+t−r −t ) ik·(x−x0 )
Z
= −i e
(2π)d−1 2π k 2 iω − Dc k 2
dd−1 k dω ω 2 (V /V 0 )−iω/2πT ik·(x−x0 )
Z
= −i e
(2π)d−1 2π k 2 iω − Dc k 2
2
dd−1 k 2 V −Dc k /2πT ik·(x−x0 )
Z  
= iDc2 k e θ(V − V 0 ) . (B.9)
(2π)d−1 V0

– 21 –
The Wightman function can be obtained as

 − Dc k2   Dc k2
 
0
dd−1 k k 2 eik·(x−x )
Z
0 0 0 V 2πT U 2πT
Gw 2
tt (U, V, x; U , V , x ) = Dc d−1
  0
− 0
 . (B.10)
(2π) 2 sin Dc k 2 V U
2T

The corresponding bulk-to-boundary propagators can be obtained from the above ex-
pressions by taking one of the bulk points to the boundary. Denoting the bulk point
as (U = 0, V, x) and the boundary point as (t1 , x1 ), we can write the retarded bulk-to-
boundary propagator as

dd−1 k 2 eik·(x−x1 )
Z
Kttret (V, x; t1 , x1 ) i Dc2 θ V − e2πT t1 ,


= k Dc k2
(B.11)
(2π)d−1
(V e−2πT t1 ) 2πT

and the Wightman bulk-to-boundary propagator as

dd−1 k k2 eik·(x−x1 )
Z
Kttw (V, x; t1 , x1 ) = Dc2 Dc k2
. (B.12)
(2π)d−1 2 sin Dc k2
 
2T (V e−2πT t1 ) 2πT

where we used that r(z → 0) = 0 and consequently V → e2πT t1 as z → 0.

In the calculation of ∆U , we need KVwt and KVrett , which can easily be obtained from
Kttw and Kttret . Since AV = A2πT
t +Ar 1
V , we can write hAV Jt i = 2πT V h(At + Ar )Jt i. This can be
simplified even more in the hydrodynamic limit, in which Ar = At . Using this last equality,
we find hAV Jt i = πT1V hAt Jt i. With the above considerations, we find

dd−1 k eik·(x−x1 )
Z
KVrett (V, x; t1 , x1 ) = 2iT Dc θ(V − e 2πT t1
) ∂V c k2
, (B.13)
(2π)d−1 (V /V ) D2πT
1
dd−1 k ik·(x−x )
Z
1 e 1
KVwt (V, x; t1 , x1 ) = T Dc ∂V d−1
  Dc k 2
. (B.14)
(2π) 2 sin D2T
ck
2
(V /V1 ) 2πT

where V1 = e−2πT t1 , and an overall factor of T was included for dimensional reasons. These
formulas were derived in the hydrodynamic limit, which means they are only well-defined
near the black hole horizon. Since we consider the expectation value of stress energy tensor
in the near horizon region, the large time separation modes (V /V1  1) give a dominant
contribution to it.
Also, the stress tensor hTV V i computed from the above propagators is independent on
the insertion time t0 . This happens because hTV V i is proportional to5 K ret (V, t0 )K w (V, −t0 )
and the t0 dependence cancels out. Consequently, the expression for the average null energy
R∞
V0 dV hTV V i derived using (B.13) and (B.14) is only valid at lates times. To fix this problem
and obtain a result for ANE that depends on the insertion time, we used the propagators
5
Here we omit the other coordinates for simplicity.

– 22 –
proposed by Swingle and Cheng in [39]:

dd−1 k eik·(x−x1 )
Z
KVrett (V, x; t1 , x1 ) = 2iT Dc θ(V − e 2πT t1
) ∂V Dc k2
, (B.15)
(2π)d−1
(V e−2πT t1 − 1) 2πT

dd−1 k eik·(x−x1 )
Z
1
KVwt (V, x; t1 , x1 ) = T Dc ∂V d−1
  Dc k 2
. (B.16)
(2π) 2 sin Dc k
2
2T (1 − V e−2πT t1 ) 2πT

According to [39], the above propagators have qualitatively the same behavior as (B.13)
and (B.14) in the late time limit and have the desired analytic properties connecting K ret
and K w . They allow us to study the dependence of ANE for any value of the insertion
time, which is more physical, since there is nothing special about t0 = 0.

C Hydrodynamic approximation and the small-k limit

In this section, we perform a numerical analysis to understand the validity of the small
k approximation we employ in Sec. 5, which is taken to get an analytic result (5.23). We
will introduce a momentum cutoff in our numerical integration to make it consistent to the
hydrodynamic approximation. After calculating the integral numerically, we will compare
this result to our analytic result.
To evaluate the ANE, we need to take integral in k from zero to infinity. When we
evaluate the integral, the high-k modes should be eliminated because we are considering
the hydrodynamic approximation. We are interested in the following integration

d−1 −1−2Dk
∞ ∞
hd2 Dc2 2π 2 dkk d−2 k 2 Dk2 4Dk
Z Z 
1
dV hTV V i = √ Γ(−Dk )Γ(Dk + 1/2) + V0 ,
(2π)3 T Γ d−1 (2π)d−1 2 sin(πDk ) π


V0 2 0 V0
(C.1)

Dc 2
where Dk = 2πT k . In the integrand, the oscillating and divergent behavior in a large k
regime arises due to the following factor.

Γ(−Dk )Γ(Dk + 1/2)

k d Dk2 . (C.2)
2 sin(πDk )

The fast oscillation in the high-k modes are suppressed due to the following factor

4Dk (V0 + 1/V0 )−2Dk . (C.3)

This term goes to zero in the large k regime. Using this fact, we can approximate (C.2)

– 23 –
without the highly oscillating behavior as
r
d Γ(−Dk )Γ(Dk+ 1/2) 9 d 3 d+3
k Dk2 '− √ k − k (C.4)
2 sin(πDk ) 4 π 2

− √9 k d for small k
4 π
' q . (C.5)
− 3 k d+3 for large k
2

We have checked the consistency between the right hand side and left hand side terms of
the first line in (C.4) by computing the integral (C.1) numerically. However, the suppression
of high-k modes does not work when V0 = 1, since the factor (C.3) becomes constant in
this case. Due to this fact, we have a still divergent behavior after taking the integral in k
from zero to infinity. This divergence comes from the fact that the high-k modes affect our
integral, which is not consistent to the hydrodynamic approximation. To make our result
consistent with the hydrodynamic approximation, we need to introduce a momentum cutoff
kmax in our calculation. Our numerical calculation was done by as follows
d−1 −1−2Dk
Z ∞
h d2 Dc2 2π 2
Z kmax
Dk d
 √ 3
 1
dV hTV V i = d+2 d+3 dk4 k 9 + 2 6πk + V0 .
2 π T Γ d−1


V0 2 0 V0
(C.6)

The numerical plot for various momentum cutoffs are shown in Fig. 6. By comparing the
numerical results to our analytic result in (5.23), we found that they have qualitatively
similar behavior. Also, the numerical result and analytic result coincides when kmax has
a value of order of 2πT , and the numerical results differ from our analytic results when
kmax is greater than 2πT . It means that eliminating the high-k modes yields a qualitatively
similar result with small-k approximation, which we used to obtain analytic result for the
ANE.

– 24 –
 TV V dV
R

V0
R
Figure 6. Numerical plots of TV V dV versus V0 for various momentum cutoffs. The numerical
plot without momentum cutoff (black curve) has a divergence at V0 = 0. This divergence can be
eliminated by introducing a momentum cutoff (blue curve). Here, we compare the numerical plots
with our analytic result (orange curve) obtained using a small-k approximation (5.23). Here, we set
1
T = 2π , h = −1 and d = 3.

References

[1] J. M. Maldacena, The Large N limit of superconformal field theories and supergravity,
Adv.Theor.Math.Phys. 2 (1998) 231–252, [hep-th/9711200].
[2] E. Witten, Anti-de Sitter space and holography, Adv. Theor. Math. Phys. 2 (1998) 253–291,
[hep-th/9802150].
[3] S. S. Gubser, I. R. Klebanov and A. M. Polyakov, Gauge theory correlators from non-critical
string theory, Phys. Lett. B428 (1998) 105–114, [hep-th/9802109].
[4] J. Maldacena, Eternal black holes in anti-de sitter, Journal of High Energy Physics 2003
(apr, 2003) 021–021.
[5] P. Gao, D. L. Jafferis and A. C. Wall, Traversable Wormholes via a Double Trace
Deformation, JHEP 12 (2017) 151, [1608.05687].
[6] J. Maldacena, D. Stanford and Z. Yang, Diving into traversable wormholes, Fortsch. Phys.
65 (2017) 1700034, [1704.05333].
[7] C. H. Bennett, G. Brassard, C. Crépeau, R. Jozsa, A. Peres and W. K. Wootters, Teleporting
an unknown quantum state via dual classical and einstein-podolsky-rosen channels, Phys.
Rev. Lett. 70 (Mar, 1993) 1895–1899.
[8] A. R. Brown, H. Gharibyan, S. Leichenauer, H. W. Lin, S. Nezami, G. Salton et al.,
Quantum gravity in the lab: Teleportation by size and traversable wormholes, 1911.06314.
[9] S. Nezami, H. W. Lin, A. R. Brown, H. Gharibyan, S. Leichenauer, G. Salton et al., Quantum
Gravity in the Lab: Teleportation by Size and Traversable Wormholes, Part II, 2102.01064.
[10] T. Schuster, B. Kobrin, P. Gao, I. Cong, E. T. Khabiboulline, N. M. Linke et al., Many-Body
Quantum Teleportation via Operator Spreading in the Traversable Wormhole Protocol, Phys.
Rev. X 12 (2022) 031013, [2102.00010].

– 25 –
[11] D. Bak, C. Kim and S.-H. Yi, Bulk view of teleportation and traversable wormholes, JHEP
08 (2018) 140, [1805.12349].
[12] D. Bak, C. Kim and S.-H. Yi, Transparentizing Black Holes to Eternal Traversable
Wormholes, JHEP 03 (2019) 155, [1901.07679].
[13] D. Bak, C. Kim and S.-H. Yi, Experimental Probes of Traversable Wormholes, JHEP 12
(2019) 005, [1907.13465].
[14] E. Caceres, A. S. Misobuchi and M.-L. Xiao, Rotating traversable wormholes in AdS, JHEP
12 (2018) 005, [1807.07239].
[15] Z. Fu, B. Grado-White and D. Marolf, Traversable Asymptotically Flat Wormholes with
Short Transit Times, Class. Quant. Grav. 36 (2019) 245018, [1908.03273].
[16] B. Ahn, Y. Ahn, S.-E. Bak, V. Jahnke and K.-Y. Kim, Holographic teleportation in higher
dimensions, JHEP 07 (2021) 219, [2011.13807].
[17] S. Fallows and S. F. Ross, Making near-extremal wormholes traversable, JHEP 12 (2020)
044, [2008.07946].
[18] A. Almheiri, A. Mousatov and M. Shyani, Escaping the Interiors of Pure Boundary-State
Black Holes, 1803.04434.
[19] J. Couch, S. Eccles, P. Nguyen, B. Swingle and S. Xu, Speed of quantum information
spreading in chaotic systems, Phys. Rev. B 102 (2020) 045114, [1908.06993].
[20] B. Freivogel, D. A. Galante, D. Nikolakopoulou and A. Rotundo, Traversable wormholes in
ads and bounds on information transfer, Journal of High Energy Physics 2020 (Jan, 2020) .
[21] H. Geng, Non-local entanglement and fast scrambling in de-Sitter holography, Annals Phys.
426 (2021) 168402, [2005.00021].
[22] T. Nosaka and T. Numasawa, Chaos exponents of SYK traversable wormholes, JHEP 02
(2021) 150, [2009.10759].
[23] A. Levine, A. Shahbazi-Moghaddam and R. M. Soni, Seeing the entanglement wedge, JHEP
06 (2021) 134, [2009.11305].
[24] D. Marolf and S. McBride, Simple Perturbatively Traversable Wormholes from Bulk
Fermions, JHEP 11 (2019) 037, [1908.03998].
[25] A. Al Balushi, Z. Wang and D. Marolf, Traversability of Multi-Boundary Wormholes, JHEP
04 (2021) 083, [2012.04635].
[26] R. Emparan, B. Grado-White, D. Marolf and M. Tomasevic, Multi-mouth Traversable
Wormholes, JHEP 05 (2021) 032, [2012.07821].
[27] N. Bao, A. Chatwin-Davies, J. Pollack and G. N. Remmen, Traversable Wormholes as
Quantum Channels: Exploring CFT Entanglement Structure and Channel Capacity in
Holography, JHEP 11 (2018) 071, [1808.05963].
[28] S. Hirano, Y. Lei and S. van Leuven, Information Transfer and Black Hole Evaporation via
Traversable BTZ Wormholes, JHEP 09 (2019) 070, [1906.10715].
[29] A. M. Garcı́a-Garcı́a, T. Nosaka, D. Rosa and J. J. M. Verbaarschot, Quantum chaos
transition in a two-site Sachdev-Ye-Kitaev model dual to an eternal traversable wormhole,
Phys. Rev. D 100 (2019) 026002, [1901.06031].
[30] T. Numasawa, Four coupled SYK models and nearly AdS2 gravities: phase transitions in

– 26 –
 traversable wormholes and in bra-ket wormholes, Class. Quant. Grav. 39 (2022) 084001,
[2011.12962].
[31] J. Maldacena, A. Milekhin and F. Popov, Traversable wormholes in four dimensions,
1807.04726.
[32] Z. Fu, B. Grado-White and D. Marolf, A perturbative perspective on self-supporting
wormholes, Classical and Quantum Gravity 36 (jan, 2019) 045006, [1807.07917].
[33] A. Anand and P. K. Tripathy, Self-supporting wormholes with a massive vector field, Physical
Review D 102 (dec, 2020) , [2008.10920].
[34] A. Anand, Self-supporting wormholes in four dimensions with scalar field, 2204.08178.
[35] A. Kundu, Wormholes and holography: an introduction, The European Physical Journal C
82 (may, 2022) , [2110.14958].
[36] P. Kovtun and L. G. Yaffe, Hydrodynamic fluctuations, long-time tails, and supersymmetry,
Physical Review D 68 (jul, 2003) .
[37] S. Caron-Huot and O. Saremi, Hydrodynamic Long-Time tails From Anti de Sitter Space,
JHEP 11 (2010) 013, [0909.4525].
[38] V. Jahnke, Recent developments in the holographic description of quantum chaos,
1811.06949.
[39] G. Cheng and B. Swingle, Scrambling with conservation laws, JHEP 11 (2021) 174,
[2103.07624].
[40] E. Witten, Multitrace operators, boundary conditions, and AdS / CFT correspondence,
hep-th/0112258.
[41] D. Marolf and S. F. Ross, Boundary Conditions and New Dualities: Vector Fields in
AdS/CFT, JHEP 11 (2006) 085, [hep-th/0606113].
[42] E. Witten, SL(2,Z) action on three-dimensional conformal field theories with Abelian
symmetry, in From Fields to Strings: Circumnavigating Theoretical Physics: A Conference in
Tribute to Ian Kogan, pp. 1173–1200, 7, 2003. hep-th/0307041.
[43] R. G. Leigh and A. C. Petkou, SL(2,Z) action on three-dimensional CFTs and holography,
JHEP 12 (2003) 020, [hep-th/0309177].
[44] R. Haag, N. M. Hugenholtz and M. Winnink, On the Equilibrium states in quantum
statistical mechanics, Commun. Math. Phys. 5 (1967) 215–236.
[45] T. Andrade and B. Withers, A simple holographic model of momentum relaxation, Journal of
High Energy Physics 2014 (may, 2014) .
[46] K.-B. Huh, H.-S. Jeong, K.-Y. Kim and Y.-W. Sun, Upper bound of the charge diffusion
constant in holography, Journal of High Energy Physics 2022 (jul, 2022) .
[47] D. T. Son and A. O. Starinets, Minkowski space correlators in AdS / CFT correspondence:
Recipe and applications, JHEP 0209 (2002) 042, [hep-th/0205051].

– 27 –