This content has been downloaded from IOPscience. Please scroll down to see the full text.

Download details:
IP Address: 148.88.244.159
This content was downloaded on 16/07/2014 at 04:18
Please note that terms and conditions apply.
Bogoliubov theory of the Hawking effect in BoseEinstein condensates
View the table of contents for this issue, or go to the journal homepage for more
2003 J. Opt. B: Quantum Semiclass. Opt. 5 S42
(http://iopscience.iop.org/1464-4266/5/2/357)
Home Search Collections Journals About Contact us My IOPscience
INSTITUTE OF PHYSICS PUBLISHING JOURNAL OF OPTICS B: QUANTUM AND SEMICLASSICAL OPTICS
J. Opt. B: Quantum Semiclass. Opt. 5 (2003) S42S49 PII: S1464-4266(03)55826-8
Bogoliubov theory of the Hawking effect
in BoseEinstein condensates
U Leonhardt
1
, T Kiss
1,2,3
and P

Ohberg
1,4
1
School of Physics and Astronomy, University of St Andrews, North Haugh,
St Andrews KY16 9SS, UK
2
ResearchInstitute for Solid State Physics and Optics, H-1525 Budapest, POBox 49, Hungary
3
Institute of Physics, University of P ecs, Ifj us ag u. 6 H-7624 P ecs, Hungary
4
Department of Physics, University of Strathclyde, Glasgow G4 0NG, UK
Received 4 November 2002
Published 2 April 2003
Online at stacks.iop.org/JOptB/5/S42
Abstract
Articial black holes may demonstrate some of the elusive quantum
properties of the event horizon, in particular Hawking radiation. One
promising candidate is a sonic hole in a BoseEinstein condensate. We
clarify why Hawking radiation emerges from the condensate and how this
condensed-matter analogue reects some of the intriguing aspects of
quantum black holes.
Keywords: Articial black holes, BoseEinstein condensates
1. Introduction
Picture a BoseEinstein condensate owing through a nozzle
where the condensate exceeds the speed of sound. Suppose
that the nozzle is designed such that the trans-sonic ow does
not become turbulent. One could build such a nozzle, the
equivalent of the Laval nozzle [1], out of light, using the dipole
force between light and atoms to conne the condensate in
an appropriate potential. Consider the fate of sound waves
propagating against the current of the trans-sonic condensate.
In the subsonic region sound waves may advance against the
ow, whereas in the supersonic zone they are simply swept
away. No sound can escape the point where the ow turns
supersonicthe sonic horizon (gure 1). The trans-sonic uid
acts as the acoustic equivalent of the black hole [2, 3].
An articial black hole [4] of this kind could be
employed to demonstrate some elusive quantum properties
of the event horizon in the laboratory, in particular Hawking
radiation[5, 6]. Hawking[7, 8] predictedthat the event horizon
emits quanta as if the horizon had a temperature given by the
gradient of the gravitational potential. To be more precise, the
horizon should spontaneously emit quantum pairs where one
particle of each pair falls into the hole and the other escapes
into space, constituting the radiation of the horizon. Both the
spectral distribution and the quantum state of the emerging
radiation are thermal. For solar-mass or larger black holes the
Hawking temperature is in the order of 10
7
K or below, which
makes the effect next to impossible to observe in astronomy.
In the case of sonic holes the Hawking temperature is given by
u
Figure 1. Schematic diagram of a sonic horizon. A uid is forced to
move through a constriction where the ow speed u becomes
supersonic (dashed line). The constriction may be formed by the
walls of a tube or, if the uid is an alkali BoseEinstein condensate,
by a suitable trapping potential. The picture shows a Laval
nozzle [1] where a supersonic uid is hydrodynamically stable.
the velocity gradient at the sonic horizon [2, 3],
k
B
T =
h
2
, (1)
and the emitted quanta are phonons. A velocity gradient
of 10
3
Hz would correspond to about 1.2 nK temperature.
In order to observe such subtle quantum effects one should
employ the best and coldest superuids availableBose
Einstein condensates of dilute gases [9, 10].
Some detailed schemes for sonic black holes in Bose
Einstein condensates have been investigated theoretically [11
15]. The ultimate design depends on experimental details and
on the state of the art in manipulating condensates, a rapidly
evolving eld. In this paper we analyse the general aspects
of the Hawking effect in BoseEinstein condensates. In the
rst part of the paper we collect and combine the ingredients
of the effect, results that are scattered in the literature. In the
second part we show how the Hawking effect arises naturally
1464-4266/03/020042+08$30.00 2003 IOP Publishing Ltd Printed in the UK S42
Bogoliubov theory of the Hawking effect in BoseEinstein condensates
within the Bogoliubov theory of the elementary excitations
in BoseEinstein condensates [16, 17]. For the rst time, to
our knowledge, we connect the quantum physics of the event
horizon to the behaviour of a realistic quantum uid.
2. Sound in uids
Consider the propagation of sound in uids moving with ow
speed u. Suppose that the ow varies little over the scale
of an acoustic wavelength. In this regime we can describe
sound propagation in geometrical acoustics (the equivalent of
geometrical optics [18] or of the semiclassical approximation
in quantum mechanics [19]). Sound rays follow Hamiltons
equations,
dr
dt
=

k
,
dk
dt
=

r
, (2)
where the dispersion relation between the frequency and
the wavevector k denes the effective Hamiltonian (r, k).
Assume that in each uid cell

2
= c
2
k
2
, (3)
where c denotes the speed of sound and

refers to the
frequencyinlocallycomoving frames. Inthe laboratoryframe,

is Doppler shifted,

= u k. (4)
In order to see why waves in uids are related to waves in
general relativity, we write the dispersion relation (3) in a
relativistic form. We introduce the spacetime wavevector
k

= (, k) (5)
and the matrix
g

=
2

1 u
u c
2
1I + u u

. (6)
The prefactor is an arbitrary non-vanishing function of the
coordinates called the conformal factor. In this notation the
dispersion relation appears in the relativistic form
g

= 0, (7)
adopting Einsteins summation convention. Therefore, sound
waves experience the moving uid as an effective spacetime
geometry with the metric g

, the inverse matrix of g

, given
by
g

=
2

c
2
u
2
u
u 1I

. (8)
The analogy between sound waves in uids and waves in
general relativity [2, 3, 20] turns out to be exact for an
irrotational uid [21] with arbitrary density prole
0
, ow u
and speed of sound c, where
0
, u and c may vary in space and
time. The velocity potential and the density perturbations

s
of sound obey the linearized equation of continuity and the
linearized Bernoulli equation [22]

s
+ (u
s
+
0
) = 0, (9)
(
t
+ u ) + c
2

0
= 0. (10)
As a consequence, the velocity potential of sound obeys the
equation [3]

0
c
2
(
t
+ u ) +

0
c
2
[u
t
(c
2
u
2
)] = 0, (11)
which can be written as the relativistic wave equation [2, 3]
D

=
1

gg

= 0 (12)
with the conformal factor , in d spatial dimensions, chosen
as [3]
=

0
c
3

1/(d1)
. (13)
The assumptions made inorder toderive the wave equation(11)
are that the uid [22] is irrotational (1) and isentropic (2).
BoseEinstein condensates naturally satisfy condition (1).
Condition (2) characterizes the hydrodynamic regime of
condensates [9, 10]. Here the local pressure depends only
on the density and on the temperature of the uid and the
quantum pressure is negligible [9, 10]. Condition (2) turns out
to be violated close to the sonic horizon.
3. Sonic horizon
Consider the propagation of sound waves in the vicinity of
the sonic horizon. Focus on the physics in the direction z
of the ow at the horizon in a quasi-one-dimensional model.
Assume that the speed of sound in the uid is constant. The
wave equation (11) reads explicitly

t
(
t
+ u
z
) +
z
[u
t
(c
2
u
2
)
z
] = 0. (14)
We obtain the general solution
=
0
(

t ),

dz
c u
. (15)
The

t refer to null coordinates in the frame comoving

with the uid
5
. In these coordinates sound waves propagate
exactly as in homogeneous space. In the laboratory frame
sound waves are accelerated or slowed down by their carrier,
the moving uid. An interesting behaviour occurs near the
horizon, say at z = 0, where for small z
u = c + z. (16)
The constant describes the velocity gradient of the
condensate at the sonic horizon. We see that

+
=
ln(z/z

. (17)
Wavepackets localized just before the horizon at z 0 take
an exponentially large time to advance against the current. On
the other side of the horizon, z 0, such waves drift equally
slowly in the direction of the ow. The horizon at z = 0 marks
a clear watershed, cutting space into two disconnected regions.
In terms of the coordinate (17) these regions are characterized
5
In a uid extending in more than one spatial dimension the acoustic space
time geometry is curved, in general. There is no longer a universal comoving
frame, because in this frame the metric is at.
S43
U Leonhardt et al
by the sign of z

. For stationary sound waves with frequency

we get from the general solution (15)
= Re{
A
z
i/
e
it
}. (18)
The phase of the wave, (/) ln(z/z

), diverges logarithmi-
cally at the horizon where, in turn, the wavenumber k develops
a pole,
k =

z

ln
z
z

=

z
, (19)
and the wavelength of sound shrinks beyond all scales,
=
2
k
=
0
z. (20)
However, when reaches the scale of the healing length
of the condensate [9, 10] (also called the correlation length)
the hydrodynamic description of sound in BoseEinstein
condensates is no longer valid [9, 10]. The acoustic theory
at the horizon predicts its own demise. Similarly, waves near
the event horizon of a gravitational black hole are compressed
beyond all scales. New physics beyond the Planck scale may
affect the event horizon [2326].
4. Bogoliubov dispersion
For BoseEinstein condensates the equivalent of any
hypothetical trans-Planckian physics is well known
Bogoliubovs theory of elementary excitations. In the
dispersion relation (3) we replace the right-hand side by
Bogoliubovs famous result [9, 10, 27, 28]
6

2
= c
2
k
2

1 +
k
2
k
2
c

. (21)
The parameter k
c
is the acoustic Compton wavenumber
k
c
=
mc
h
=
1

2
, (22)
with m being the atomic mass, also expressed in terms of the
healing length [9, 10] (the correlation length). Typically,
is in the order of 10
6
m and c reaches a few 10
3
m s
1
in BoseEinstein condensates (without exploiting Feshbach
resonances). We calculate the group velocity
v =

k
= u + v

, (23)
v

k
= c
2
k

1 +
2k
2
k
2
c

. (24)
Equation (23) shows that the group velocity obeys the Galilean
addition theorem of velocities. Equation (24) expresses the
group velocity in the uid frame, v

, in terms of the frequency

and the wavenumber. The acoustic Compton wavenumber,
k
c
, sets the scale beyond which v

deviates signicantly
from c. For large wavelengths, sound is communicated by
atomic collision, and the product of the condensates density
and the atomic collision strength gives mc
2
[9, 10]. For
wavelengths comparable with or shorter than the healing
6
Corley and Jacobson studied the effect of a Bogoliubov-type dispersion on
scalar waves in [29, 30].
length, the interaction-free Schr odinger dynamics of the
atoms dominates the transport of excitations. Perturbations
of the free wavefunction travel with innite velocity. So
the acoustic Compton wavenumber, k
c
, characterizes the
crossover between the speed of sound and the innite speed of
perturbations of free matter waves.
Close to the sonic horizon, the wavenumber (19) increases
dramatically, and, inturn, the effective speedof soundv

grows.
The horizon, dened as the place where the uid exceeds the
speedof sound, seems todissolve like a mirage. Nature appears
to prevent the existence of an event horizon. However, we
show in section 6 that the horizon still exists, but at a less
well dened location and for a particular class of elementary
excitations only. As long as k
2
is much smaller than k
2
c
we get
the acoustic relation
k

u c
. (25)
For the other extreme, where k
2
is much larger than k
2
c
, one
nds [29, 30]
k 2k
c

u
2
/c
2
1 +
u
c
2
u
2
. (26)
Consider the turning points z
0
, the points where the group
velocity of sound (23) vanishes. If the acoustic dispersion
relation (3) were universally valid the horizon would be the
turning point. Therefore |z| |z
0
| does indicate the spatial
scale of the trans-acoustic range around the horizon, which
denes the spatial delocalization of the horizon. To proceed
we recall that elementary excitations are small perturbations of
the condensate. Their energies h ought to be much smaller
than the mean-eld energy of the condensate, which is in the
order of mc
2
(with c being the speed of sound). Therefore,
=
h
mc
2
, || 1. (27)
We expand the solution z
0
of v = 0 as a power series in
1/3
and nd, to leading order, three turning points in the complex
plane given by [31]
z
0
=
c

3
2
3

2/3
. (28)
Far away from the horizon we may characterize the four
fundamental solutions of the dispersion relation (21) combined
with the Doppler shift (4) by their asymptotics (25) or (26).
However, close to a turning point geometrical acoustics alone
does not provide a good description of wave propagation
any longer. The turning points may cause scattering. The
connections between elementary excitations across the horizon
must be examined with care. Extending z to the complex plane
represents anelegant wayof analysingthis connection. We nd
in the appendix that the acoustic relation (19) remains valid on
either the upper or the lower half of the complex plane.
5. Bogoliubov modes
Elementary excitations are perturbations of the condensate,
ripples on the macroscopic wavefunction
0
of the condensed
atoms. The excitations constitute the non-condensedpart of the
S44
Bogoliubov theory of the Hawking effect in BoseEinstein condensates
atomic gas. Todescribe elementaryexcitations, the total many-
bodyeldoperator

of the atoms is split intotwocomponents,
the condensate with the mean-eld wavefunction
0
and the
non-condensed part. The mean-eld wavefunction comprises
the density prole
0
and the ow u, as

0
=

0
e
iS0
, u =
h
m
S
0
. (29)
The atomic eld operator

is split into the condensate and the
non-condensed part according to the relation

=
0
+ e
iS0

. (30)
The non-condensed part consists of Bogoliubov modes u

and
v

[9, 10, 16, 17],

(u

+ v

). (31)
The u

and v

are subject to the BogoliubovdeGennes

equations [9, 10, 16, 17]. If one requires that the Bogoliubov
modes satisfy the orthonormality relations

(u

) dz =

, (32)

(u

) dz = 0, (33)
then the a

and a

obey the commutation relations of Bose

annihilation and creation operators, as a consequence of the
fundamental commutator of atoms with Bose statistics
[

(z),

(z

)] = (z z

). (34)
Each pair of u

and v

characterizes the spatial shape and the

evolution of an excitation wave, and the Fock space of the a

and a

spans the state space of the excitation quasiparticles.

To see how the Bogoliubov modes are related to sound
waves, we write down the macroscopic wavefunction of the
condensate combined with one of the excitations,
=
0
+ e
iS0
(u

+ v

). (35)
We represent as
=

e
iS
, =
0
+
s
, S = S
0
+ s, (36)
where
s
denotes the local density of the sound wave and s is
proportional to the velocity potential
=
h
m
s. (37)
Assuming that
s
and s are small perturbations we get
u

+ v

s
2
0
+ is

. (38)
Assuming further that u

and v

are stationary waves with

frequency , we obtain from our solution (18) of the
hydrodynamic sound-wave equation and from the linearized
Bernoulli equation (10) the Bogoliubov modes
u

2z
+
mc
h

z
i/
e
it
,
v

2z

mc
h

z
i/
e
it
.
(39)
These asymptotic expressions are valid as long as the elemen-
tary excitations are sound waves with wavenumber (19). We
show in the appendix that this is the case sufciently far away
from the turning points and on either the upper or the lower
half plane. Here the term /(2z) in the expressions (39) is
always small compared with mc/ h. For real and positive z we
have
(z)
i/
= e
(2/)
z
i/
. (40)
The sign refers to the two ways in which we may circumvent
the trans-acoustic region, on the upper (+) or on the lower
() half plane. Modes with the acoustic asymptotics (39)
throughout the upper half plane are suppressed on the left-
hand side of the horizon and for positive frequencies and
enhanced for negative . Modes with the asymptotics (39) on
the lower half plane show the opposite behaviour.
6. Negative energy
Bogoliubov modes are normalized according to the scalar
products (32) and (33). Let us calculate the norm of the
modes (39) of the sonic hole. The energy parameter (27)
is small and so is the extension of the trans-acoustic region
around the horizon, measured roughly by the location of
the turning points (28). Consequently, we can neglect the
trans-acoustic contribution to the normalization integral (32).
We approximate the Bogoliubov modes by their asymptotic
expressions (39), utilize the relation (40) and get

(u

) dz
| A

|
2

+
0

2k
c
z
z
i(

)/
dz
= | A

|
2
(1 e
(2/)
)4k
c
(

). (41)
If we choose modes with the asymptotics (39) on the upper
half plane the norm is positive and the u

and v

may serve as
proper Bogoliubov modes. We nd the normalized amplitude
A

= [(1 e
2/
)4k
c
]
1/2
. (42)
Remarkably, the Bogoliubov norm is also positive for
modes with negative frequencies. The BogoliubovdeGennes
equations have a well known symmetry [16, 17]: if the (u

, v

)
are solutions then the complex-conjugated and interchanged
modes, (v

, u

), are solutions as well. Yet the norm of the

conjugate modes is the negative norm of the original (u

, v

).
In contrast, sonic black holes generate negative-frequency
modes with positive norm, which is the unusual feature that
gives rise to the acoustic analogue of Hawking radiation [6].
We also see how the mentioned symmetry of the Bogoliubov
deGennes equations [16, 17] appears in our case. If we chose
the u

and v

with the asymptotics (39) on the lower half plane

we would get negative normalization integrals (41) for both
positive and negative frequencies.
The negative frequencies of the positive-normBogoliubov
modes give rise to negative energies in the Hamiltonian of the
elementary excitations,

H =

h( a

+
a
+
a

)N d. (43)
S45
U Leonhardt et al
Here and later the subscripts refer to positive and negative
frequencies, respectively, and N denotes the density of modes.
We see that there is no natural ground state of the elementary
excitations. In practice, of course, the spectrum is limited by
the requirement (27) that the energies of the excitations ought
to be much smaller in magnitude than the condensates mean-
eld energy. We note that the Hamiltonian (43) is invariant
under the Bogoliubov transformations
a

= a

cosh a

sinh ,
u

= u

cosh v

sinh ,
v

= v

cosh u

sinh
(44)
with an arbitrary real parameter . In the case where we choose
tanh = e
/
(45)
we get a new set of modes, equation (39), with
A

(z)(4k
c
)
1/2
. (46)
The step function indicates that the primed modes appear on
either the left- or on the right-hand side of the trans-acoustic
region. Therefore, despite the trans-Planckian problem, a
sonic horizon exists, but at a less well dened location, within
|z| |z
0
|, and the horizon applies to a particular set of modes
only.
7. Hawking effect
The Bogoliubov transformations (44) relate one set of
quasiparticles to another one, both representing perfectly
valid energy eigenvalues, yet their quasiparticle vacua differ
(the states |0 or |0

that are annihilated by a

and
a

, respectively). This ambiguity has direct physical

consequences, because the cloud of non-condensed atoms [28]
depends on the vacuum state of the elementary excitations,
0|

|0 =

(|v
+
|
2
+ |v

|
2
)N d, (47)
0

|0

(|v

+
|
2
+ |v

|
2
)N d = 0|

|0. (48)
In general relativity the notion of the vacuum is observer
dependent. For example, the vacuum of empty space
in Minkowski coordinates appears as a thermal eld to
accelerated observers [5, 6, 32]. In the case of the black
hole, the gravitational collapse has created a state of quantum
elds that an inward-falling observer perceives as vacuum,
yet an external observer sees as thermal radiation, Hawking
radiation [58]. In our case, the equivalent of the gravitational
collapse, the formation of the sonic horizon, chooses the
quasiparticle vacuum, if the trans-sonic velocity prole has
initially been created from a condensate without a horizon.
Such a process must be sufciently smooth to keep the
condensate intact.
To analyse the quasiparticle vacuum, we use the
Heisenberg picture of quantum mechanics where observables
evolve while the quantum state is invariant. We describe
the initial (and nal) vacuum state with respect to one set of
continuous modes given before the formation of the horizon.
In the Heisenberg picture these modes evolve. We sort the
initial modes into left- and right-moving modes that, close
to complex , are analytic on the upper or on the lower
half plane, respectively, because here exp(ikz) converges for
positive k on the upper and for negative k on the lower half
plane. The upstream modes we are interested in stem from
right-moving modes. The formation of the sonic horizon, a
smooth process, cannot fundamentally alter the analyticity of
the vacuum modes. In particular, the process can never create
non-analytic modes of the type expressed in equation (46).
Consequently, the initial quasiparticle vacuum assumes the
analytic modes of equations (39) and (42).
Given this vacuum state, we determine the quantum
depletion of the condensate. We write the density of the non-
condensed atoms in terms of the primed Bogoliubov modes.
As we have seen, these modes describe the set of elementary
excitations that exhibit the sonic horizon. We nd that
0|

|0 = (|z|),
(z) =

[(|u

+
|
2
+ |v

+
|
2
) n() + |v

+
|
2
]N d.
(49)
Here n() denotes the average number of non-condensed
atoms per excitation mode,
n() =
1
e
2/
1
=
1
e
h/k
B
T
1
. (50)
The non-condensed atoms are Planck distributed with the
Hawkingtemperature (1). Therefore, as soonas the condensate
ows through the nozzle, breaking the speed of sound,
a thermal cloud of atoms is formed. This effect is the
signature of Hawking radiation for sonic holes in Bose
Einstein condensates. The thermal cloud due to the Hawking
effect should be observable when the initial temperature of the
atoms is below the Hawking temperature. On the other hand,
one could also regard the Hawking effect in the condensate as
the quantum depletion (47) of atoms at zero temperature with
respect to the analytic modes (42) that transcend the horizon.
This feature reects the ambiguity of the vacuum in general
relativity. Using techniques for measuring the population of
Bogoliubov modes [33, 34], one could perhaps demonstrate
the ambiguity of the vacuum in the laboratory.
Finally we note that within our model sonic black holes
are stable, provided of course that the trans-sonic ow is not
plagued by hydrodynamic instabilities. In reality, elementary
excitations interact with each other, giving rise to what is
known as LandauBeliaev damping [3537]. Since a sonic
horizon does not have a ground state, this damping mechanism
will lead to the gradual evaporation of the condensate.
Therefore, LandauBeliaev damping [3537] plays the role of
black-hole evaporation. It is tempting to turn matters around
and to approach cosmological problems from the perspective
of condensed-matter physics [4, 3841].
Acknowledgments
We thank M V Berry, I A Brown, L J Garay, T A Jacobson,
R Parentani and G E Volovik for discussions. Our work
was supported by the ESF programme Cosmology in the
S46
Bogoliubov theory of the Hawking effect in BoseEinstein condensates
laboratory, the Leverhulme Trust, the National Science
Foundationof Hungary(contract noF032346), the Marie Curie
Programme of the European Commission, the Royal Society
of Edinburgh and the Engineering and Physical Sciences
Research Council.
Appendix
Inthis appendixwe examine the asymptotics of the Bogoliubov
modes on the complex plane. We use the semiclassical
approximation [31, 42]
u

= U

exp

k dz it

,
v

= V

exp

k dz it

.
(51)
The wavenumber k should obey Bogoliubovs dispersion
relation (21) including the Doppler shift (4). We obtain
Figure A.1. Wavenumbers k of elementary excitations around a sonic black-hole horizon, analytically continued onto the complex plane.
The gure shows three roots of the dispersion relation [ h
2
k
2
/(2m) + mc
2
]
2
h
2
[ (c + z)k]
2
= m
2
c
4
for = 0.1 (mc
2
/ h) and
= 0.5 (mc
2
/ h), with the branch cuts of k chosen according to gure A.2. The top row displays the wavenumber of a sound wave that
propagates against the current. The picture indicates the characteristic /(z) asymptotics away from the branch points. The two lower
rows display two trans-acoustic branches of k. The fourth root of the dispersion relation is not shown, because it corresponds to the trivial
case of sound waves that propagate with the ow.
four fundamental solutions of this fourth-order equation.
Figure A.1 shows the three branches of k that are relevant in our
analysis. The amplitudes U

and V

obey the relation [31, 42]

z
(U
2

V
2

)v = 0, (52)
where v denotes the group velocity (23) of the elementary
excitation. Equation (52) formulates the conservation law of
the quasiparticle ux for stationary states [42] if z is real, where
U
2

V
2

gives |U

|
2
|V

|
2
up to a constant phase factor. The
relation (52) can be extended to the complex z plane and to
complex frequencies [31]. Equation (52) implies that the
amplitudes U

and V

diverge close to a turning point z

0
where
v vanishes. Consequently, the semiclassical approximation
breaks down at the turning point [19]. The turning point
causes signicant scattering, i.e. the conversion of one mode
with a given k into two modes, one with wavenumber k and
the other one with a different wavenumber that satises the
dispersion relation as well. At the turning point the two
branches coincide. To prove this, we regard for a moment
S47
U Leonhardt et al
A,C B,C B,A
C,A A,B C,B
C,B
C,A
A,B
Figure A.2. Stokes lines of elementary excitations at a sonic black
hole (dotted lines), given the choice of branch cuts made in
gure A.1. The pairs of letters indicate which branches of the
superposition (52) are connected by the lines. The rst letter of each
pair identies the exponentially dominant branch. We construct a
Bogoliubov mode that is acoustic (component A) on the upper half
plane by putting the coefcient c
C
to zero on the C, A Stokes line
originating from the left turning point.
z as a function of k at constant frequencies =
0
, dened
implicitly by equations (4), (16) and (21). We get
v =

k
= u +

k
(
0
uk) = k
z
k
. (53)
We see that the function z(k) reaches extrema at the point z
0
where v vanishes, i.e. at the turning point. Close to the z
0
we
nd after some algebra [31], by expanding z into a power series
in
1/3
,
z z
0

3(k k
0
)
2
8
, k
0
=
3

4, (54)
to leading order. Consequently, each turning point connects
two branches of the wavenumber k. In general the mode
conversion occurs near specic lines in the complex plane,
called Stokes lines in the mathematical literature [43]. Stokes
lines, originating from the turning point z
0
, are dened as
the lines where the differences between the phases

k dz of
the two connected k branches are purely imaginary. Here
one of the waves is exponentially small compared with the
other. We obtain from equation (54) that the difference
between the two branches is proportional to the square root
of z z
0
. Consequently, the phase difference is proportional
to (zz
0
)
3/2
, giving rise to three Stokes lines fromeach turning
point z
0
, as in the traditional case of Schr odinger waves in one
dimension [19, 44]. Figure A.2 shows the Stokes lines for the
three turning points close to the horizon and for the branch cuts
of k chosen in gure A.1.
At a Stokes line the phase difference between the two
connected branches is purely imaginary. One of the waves
exponentially exceeds the other and, within the semiclassical
approximation, the smaller wave is totally overshadowed by
the larger one, if the larger wave is present. In general,
the Bogoliubov modes consist of a superposition of the four
fundamental solutions that correspond to the four branches of
the dispersion relation
u

= c
A
u
A
+ c
B
u
B
+ c
C
u
C
+ c
D
u
D
,
v

= c
A
v
A
+ c
B
v
B
+ c
C
v
C
+ c
D
v
D
.
(55)
The u
A
and v
A
refer to the k branch where k obeys the
asymptotics (19), i.e. where the wavenumber satises the
dispersion relation (3) of sounds in moving uids, taking into
account the Doppler detuning (4), and where k corresponds to
an upstream wave. We call such Bogoliubov modes acoustic
modes. When crossing a Stokes line, the exponentially
suppressed solution may gain an additional component that is
proportional to the coefcient of the exponentially enhanced
solution [44]. If we wish to construct Bogoliubov modes
where only the exponentially smaller component exists in
the vicinity of a Stokes line we must put the coefcient of
the larger one to zero. In gure A.2 the pairs of letters
indicate which branches are connected by the lines, and the
rst letter identies the exponentially dominant branch. The
picture shows that with the choice of branch cuts made we
can construct a Bogoliubov mode that is acoustic on the upper
half plane. Trans-acoustic physics is conned to the lower
half plane. On the other hand, if we chose other branch
cuts of k we may get Bogoliubov modes that are acoustic
on the lower half plane and trans-acoustic on the upper one.
Therefore, according to equation (40), the choice of the k
branch determines whether a Bogoliubov mode is larger or
smaller beyond the sonic horizon at the real axis, for z < 0.
Branch cuts of k are fairly arbitrary. Given the Bogoliubov
mode on the right-hand side of the horizon, we cannot predict
within the semiclassical approximation the amplitude of the
mode on the left side. Therefore, the two sides are causally
disconnected. Within the semiclassical approximation, the
horizon is a genuine horizon, despite the acoustic analogue
of the notorious trans-Planckian problem [23, 24].
References
[1] Courant R and Friedrichs K O 1967 Supersonic Flow and
Shock Waves (New York: Wiley)
[2] Unruh W G 1981 Phys. Rev. Lett. 46 1351
[3] Visser M 1998 Class. Quantum Gravity 15 1767
[4] Novello M, Visser M and Volovik G E (ed) 2002 Articial
Black Holes (Singapore: World Scientic)
[5] Birrell N D and Davies P C W 1982 Quantum Fields in Curved
Space (Cambridge: Cambridge University Press)
[6] Brout R, Massar S, Parentani R and Spindel Ph 1995 Phys.
Rep. 260 329
[7] Hawking S M 1974 Nature 248 30
[8] Hawking S M 1975 Commun. Math. Phys. 42 199
[9] Dalfovo F, Giorgini S, Pitaevskii L P and Stringari S 1999
Rev. Mod. Phys. 71 463
[10] Pitaevskii L P and Stringari S 2003 Bose Einstein
Condensation (Oxford: Clarendon)
[11] Garay L J, Anglin J R, Cirac J I and Zoller P 2000 Phys. Rev.
Lett. 85 4643
[12] Garay L J, Anglin J R, Cirac J I and Zoller P 2001
Phys. Rev. A 63 023611
[13] Barcelo C, Liberati S and Visser M 2001 Class. Quantum
Gravity 18 1137
[14] Barcelo C, Liberati S and Visser M 2001 Preprint
gr-qc/0110036
S48
Bogoliubov theory of the Hawking effect in BoseEinstein condensates
[15] Sakagami M and Ohashi A 2002 Prog. Theor. Phys. 107 1267
[16] Fetter A L 1972 Ann. Phys., NY 70 67
[17] Fetter A L 1996 Phys. Rev. A 53 4245
[18] Born M and Wolf E 1999 Principles of Optics (Cambridge:
Cambridge University Press)
[19] Landau L D and Lifshitz E M 1977 Quantum Mechanics
(Oxford: Pergamon)
[20] Fischer U R and Visser M 2002 Phys. Rev. Lett. 88 110201
[21] For a wave equation of sound in uids with vorticity see
Bergliaffa S E P, Hibberd K, Stone M and Visser M 2001
Preprint cond-mat/0106255
[22] Landau L D and Lifshitz E M 1982 Fluid Mechanics (Oxford:
Pergamon)
[23] tHooft G 1985 Nucl. Phys. B 256 727
[24] Jacobson T 1991 Phys. Rev. D 44 1731
[25] Unruh W G 1995 Phys. Rev. D 51 2827
[26] Brout R, Massar S, Parentani R and Spindel Ph 1995
Phys. Rev. D 52 4559
[27] Bogoliubov N N 1947 J. Phys. (Moscow) 11 23
[28] Landau L D and Lifshitz E M 1980 Statistical Physics. Part 2,
Theory of the Condensed State (Oxford: Pergamon)
[29] Corley S 1998 Phys. Rev. D 57 6280
[30] Corley S and Jacobson T 1999 Phys. Rev. D 59 124011
[31] Leonhardt U, Kiss T and

Ohberg P 2003 Phys. Rev. A at press
(Leonhardt U, Kiss T and

Ohberg P 2002 Preprint
cond-mat/0211462)
[32] Unruh W G 1976 Phys. Rev. D 14 870
[33] Vogels J M, Xu K, Raman C, Abo-Shaeer J R and
Ketterle W 2002 Phys. Rev. Lett. 88 060402
[34] Ozeri R, Steinhauer J, Katz N and Davidson N 2002 Phys. Rev.
Lett. 88 220401
[35] Pitaevskii L P and Stringari S 1997 Phys. Lett. A 235 398
[36] Giorgini S 1998 Phys. Rev. A 57 2949
[37] Katz N, Steinhauer J, Ozeri R and Davidson N 2002 Phys. Rev.
Lett. 89 220401
[38] Chapline G, Hohlfeld E, Laughlin R B and Santiago D I 2001
Phil. Mag. B 81 235
[39] Volovik G E 2001 Phys. Rep. 351 195
[40] Volovik G E 2003 The Universe in a Helium Droplet (Oxford:
Clarendon)
[41] Zhang S-C and Hu J 2001 Science 294 823
[42] Csord as A, Graham R and Sz epfalusy P 1998 Phys. Rev. A 57
4669
[43] Ablowitz M J and Fokas A S 1997 Complex Variables
(Cambridge: Cambridge University Press)
[44] Furry W H 1947 Phys. Rev. 71 360
S49