INSTITUTE OF PHYSICS PUBLISHING CLASSICAL AND QUANTUM GRAVITY

Class. Quantum Grav. 22 (2005) 38333839 doi:10.1088/0264-9381/22/19/002

Superradiant scattering from a hydrodynamic vortex
T R Slatyer and C M Savage
Australian Centre for Quantum Atom Optics, Australian National University, ACT 0200,
Australia
E-mail: craig.savage@anu.edu.au
Received 1 July 2005, in nal form 25 July 2005
Published 8 September 2005
Online at stacks.iop.org/CQG/22/3833
Abstract
Sound waves scattered from a hydrodynamic vortex may be amplied. Such
superradiant scattering follows from the physical analogy between spinning
black holes and hydrodynamic vortices, as spinning black holes have an
ergoregion within which low frequency waves may be scattered with increased
amplitude. While black holes also have an event horizon, a uids analogous
sonic horizon requires the vortex to have a central drain, which may be
challenging to produce experimentally. We show that in the uid analogue,
a drain is not required in order for a vortex to scatter sound superradiantly.
Furthermore, this effect may occur even when the uid density drops to zero
at the vortex core, as is the case in a BoseEinstein condensate. We also
consider engineering the density prole, using repulsive light forces, to extend
the validity of the hydrodynamic approximation towards the vortex core.
PACS numbers: 03.75.Nt, 03.75.Kk, 04.62.+v
Hawkings prediction of radiation from black holes is one of the most surprising outcomes
of the quest to unify quantum mechanics and general relativity [1, 2]. It is problematic that
such an important theoretical prediction has no immediate prospect of direct experimental
investigation. Consequently, Unruhs observation of an analogy between wave scattering
from black holes, and sound wave scattering from sonic horizons in uids, has intrigued both
the astrophysical and condensed matter communities [37].
The generation and detection of the sonic analogue of Hawking radiation faces many
difculties, and a detailed experimental scheme has yet to be presented. Consequently, we
propose the initial investigation of the simpler, but closely related, superradiant amplication
of scattered waves by an ergoregion [8]. Quantum mechanically, Hawking radiation
and superradiance may be regarded as spontaneous and stimulated radiation counterparts.
However, stimulated radiation is easier to observe because of its much greater strength,
0264-9381/05/193833+07$30.00 2005 IOP Publishing Ltd Printed in the UK 3833
3834 T R Slatyer and C M Savage
and because of the control provided by the stimulating pulse. Furthermore, an analogue
ergoregion is realized by a vortex, which is experimentally simpler to achieve than a horizon.
Superradiance is related to the ZeldovichStarobinsky effect, which has been discussed in the
context of superuid helium by Volovik [10].
In this paper we are particularly concerned with issues that arise in assessing the feasibility
of superradiance experiments in dilute gas BoseEinstein condensates (BECs). We build on
previous work by Basak and Majumdar [11, 12] who showed that sonic superradiance occurs
for a draining vortex. However, they assumed that the uid density and the speed of sound
were constant. In a BEC with a non-draining vortex this is not true, because the density
drops to zero at the core of the vortex. Nevertheless, we nd that superradiance still occurs.
Furthermore, we extend the length scale on which the drop off occurs by assuming a repulsive
optical potential. This improves the applicability of the hydrodynamic approximation in
which the quantum pressure term is dropped. We provide quantitative details and numerical
simulations to argue that superradiance might be observed in dilute gas BECs [1318]. Note
that the superradiance we study is a stimulated effect, distinct from the collective spontaneous
effect in atomic physics known as Dicke superradiance [19].
According to general relativity, rotating black holes have an ergoregion: a region in
which no physical object can remain at rest with respect to an inertial observer at innity.
Low frequency waves, such as electromagnetic or gravitational waves, may be scattered from
the ergoregion with increased amplitude; that is superradiantly [9]. However, despite its
fundamental importance, this process is not expected to be directly observable for actual
astrophysical situations. In uids, regions where the uid speed exceeds the speed of sound
are analogue ergoregions [5]. A draining vortex with a central sink may also possess an
analogue event horizon, and hence is an analogue of a rotating black hole [20].
For a BEC a sink might be realized by outcoupling the condensate to form an atom laser
beam [21]. However, there are difculties with strong outcoupling [22, 23], and it will be
simpler to create a vortex without a sink, for example by stirring with laser beams [24] or
magnetic elds [25]. Then the simpler velocity function makes it feasible to obtain analytic
results for the variable density proles which are unavoidably associated with vortices in
BECs.
The wave equation for sound waves, that is linear perturbations of the velocity potential
of a barotropic, inviscid, irrotational uid, may be written [3, 5] as

(g

) = 0, (1)
where the coefcients g

are functions of the unperturbed uid density and velocity, and

g is their inverse determinant. This has the same form as the KleinGordon equation for a
massless scalar eld in curved spacetime with metric g

[26]. Provided the quantized uid

has suitable commutation relations, the analogy extends to the quantum eld [27].
All this suggests the possibility of using uid ows to mimic general relativistic spacetime
metrics. The analogy requires the length scales of the problem to be large enough that the
atomicity of the uid can be neglected. In BECs, furthermore, the hydrodynamic equation (1)
is an approximation neglecting the quantum pressure term [28]. This is valid when any
spatial variations in the BEC occur over length scales longer than the healing length of
the condensate [29]. In our case, the healing length must be smaller than both the sound
wavelength and the spatial scale of the vortex. This turns out to be an important constraint
on the conditions under which superradiance might occur in a BEC. For a BEC the velocity
potential represents perturbations of both the GrossPitaevskii mean eld and the solutions
to the Bogoliubovde Gennes equations for the quantized eld modes [14]. This duality is
important because perturbations of the GrossPitaevskii mean eld are subject to energy and
Superradiant scattering from a hydrodynamic vortex 3835
topological constraints that would make superradiance impossible. The non-condensed part
of the quantum eld is subject to less stringent constraints that allow the energy of the system
to decrease and the angular momentum of the system to change at the atomic level.
Let the unperturbed uids speed of sound be c, velocity be v and density be . In
cylindrical polar coordinates, a vortex centred at the origin, with v
r
= v
z
= 0, has wave
equation coefcients
g

=
1
c
_
_
_
_
1 0 v

/r 0
0 c
2
0 0
v

/r 0
_
v
2

c
2
__
r
2
0
0 0 0 c
2
_
_
_
_
, (2)
and g =
4
r
2
/c
2
. For irrotational ow, v

= /r, for some constant . The density and

speed of sound depend only on the radial coordinate r and approach asymptotic values

and
c

for r .
We consider cylindrical wave solutions to the wave equation (1) of the form(t, r, , z) =
(t, r) e
im
, with the angular wavenumber m. Assuming that the square of the speed of sound
is proportional to the density, as is the case for a BEC [28], the density may be eliminated
from the wave equation, which then becomes

t
2
2i
mv

1
r

r
_
rc
2

r
_
+
m
2
r
2
_
c
2
v
2

_
= 0. (3)
For single frequency waves of the form (t, r) = r
1/2
G(r

) e
it
, and dening a tortoise
coordinate r

by dr/dr

= c
2
, with r

r for r , where c = c/c

, we nd
d
2
G(r

)
dr
2
+
c
2
c
2

(
2
V
eff
)G(r

) = 0, (4)
V
eff
=
2mv

r
+
m
2
r
2
_
c
2
v
2

1
2r
_
c
2
2r

dc
2
dr
_
, (5)
where we have introduced an effective potential V
eff
to emphasize the similarity to the time-
independent Schr odinger equation.
In order to analyse superradiance, we consider the two limiting cases where the tortoise
coordinate r

approaches . For r large, r

r, and eliminating all terms except those of

highest order in r yields the asymptotic form
d
2
G(r

)
dr
2
+

2
c
2

G(r

) 0, (6)
with the general solution
G(r

) = Ae
i(/c

)r

+ B e
i(/c

)r

, (7)
where A and B are constant amplitudes of incoming and outgoing waves, respectively.
The analysis of the other limit, where r

, requires that the density prole of the

vortex be specied. In order to present analytical calculations, we use the density prole
(r) =
_
_
_

[(r r
0
)/]
2
2 + [(r r
0
)/]
2
, r > r
0
,
0, r r
0
.
(8)
This is similar to the charge l = 1 vortex density prole for a BEC [28], but with the scale
length given by the free parameter , rather than by the healing length . In the Thomas
Fermi limit of dominant atom interactions, the length scale for a charge l vortex, with angular
3836 T R Slatyer and C M Savage
momentum l h per particle, is l [30]. Although a vortex with l > 1 is unstable to decay
into l single charge vortices, it may be stabilized by a pinning optical potential [31]. Since
the density prole equation (8) was chosen primarily for analytical convenience, we do not
expect it to be exactly achievable experimentally, although something close to it should be.
For example, the optical dipole potential generated by far detuned light has been used to
engineer density proles in BECs [32]. Although our prole does not reproduce any specic
astrophysical metric, it does have an analogue ergoregion, and we shall show that it displays
the associated wave scattering physics. The other free parameter is the radius of the zero
density core r
0
, which might also be engineered by applied potentials. Since we have assumed
that the square of the speed of sound is proportional to the density, c
2
= /

, and the tortoise

coordinate is
r

=
_
1
c
2
dr = r
2
2
r r
0
. (9)
For r r
0
, r

2
2
/(r r
0
) and c
2
c
2

(r r
0
)
2
/(2
2
). Using these approximations
and retaining only terms of lowest order in r r
0
, we nd that asymptotically
r
2
d
2
G(r

)
dr
2
+
2
2

2
c
2

G(r

) = 0, (10)
m
_
r
2
0
. (11)
This is a Euler-type equation and has solutions of the form G(r

) = |r

, where
=
1
2
,
1
2
sign()
_
1 8
2

2
_
c
2

_
1/2
. (12)
The general solution to the asymptotic form of the wave equation is then
G(r

) = |r

|
1/2
(C|r

+ D|r

), (13)
where C and D are constants. According to equation (12), is either real or purely imaginary.
For imaginary this asymptotic solution is oscillatory and may be written in the form
G(r

) = |r

|
1/2
(C e
ln|r

|
+ De
ln|r

|
). (14)
By making |r

| the argument of a logarithm we have anticipated that r

will be made
dimensionless. This last form allows us to identify the two linearly independent solutions
as outgoing and ingoing waves, with amplitudes C and D, respectively. Note that in this
asymptotic limit r

< 0, and hence ln |r

| increases with decreasing r

. Note also that it is

the sign of the group velocity, not of the phase velocity, which determines the identication
of outgoing and ingoing waves [33], and the sign of the group velocity is independent of the
sign of .
Let the Wronskians (of the solution and its complex conjugate) corresponding to the
asymptotic solutions equations (7) and (14) be denoted by W() and W(). Then
W() =
2i
c

(|B|
2
|A|
2
). (15)
If is imaginary,
W() = 2 (|C|
2
|D|
2
). (16)
Following Vilenkin [33], consider a solution to the wave equation for G(r

) representing a
wave originating at r

= +and having the asymptotic forms, for imaginary ,

G(r

) =
_
e
i(/c

)r

+ Re
i(/c

)r

, r

,
T |r

|
1/2
e
ln|r

|
, r

.
(17)
Superradiant scattering from a hydrodynamic vortex 3837
Then by Abels theorem, the Wronskian of this solution is constant, since there is no rst
derivative term in the differential equation for G(r

), equation (4). Thus, setting A = 1 and

B = R in equation (15), and C = 0 and D = T in equation (16), and equating W() =
W(),
2i
c

(|R|
2
1) = 2 |T |
2
, (18)
and therefore,
|R|
2
= 1 sign()| ||T|
2
c

/. (19)
In particular, if is negative, and T = 0, then |R| > 1, and superradiance occurs, since then,
in the asymptotic region far from the vortex, the amplitude of the reected wave exceeds that
of the incident wave. This superradiance inequality < 0 expands to
< m
_
r
2
0
, (20)
which requires m > 0, so that the cylindrical waves propagate in the same direction as the
vortex ow.
Since the reected superradiant pulse carries energy away fromthe vortex, the vortex must
lose energy. This is possible if the ingoing pulse is interpreted as an anomalous negative energy
mode [34] of the fully quantized eld [14]. It is appropriate to interpret all the phonon modes
as non-condensed modes. This avoids regarding the incident pulse as a superposition with the
vortex background mean eld, and explains why the angular momentum is not constrained
to change in units of Plancks constant times the number of atoms, as it would be in a pure
condensate.
In contrast to the draining vortex examined by Basak and Majumdar [11, 12] and the
rotating black hole [33], here there are two independent necessary conditions for superradiance:
must be purely imaginary and must be negative. Combining these conditions gives a
frequency-independent condition for superradiance. For since 0, then m
_
r
2
0
,
and superradiance requires, 0 m
_
r
2
0
, and therefore
2
m
2

2
_
r
4
0
. If is to
be imaginary, we must have 1 < 8
2

2
_
c
2

, and since m > 0, the frequency-independent

necessary condition for superradiance is m > c

r
2
0
_
(2

2).
For a charge l vortex in a BEC, we can substitute expressions [28] for the vortex velocity
constant = l h/m
atom
, where m
atom
is the mass of an atom, and for c

in terms of the healing

length , c

= h/(

2m
atom
), to obtain > r
2
0
_
(4lm). This is always possible to full, at
least in principle, since the healing length is given by = (8na)
1/2
, where n is the atomic
number density and a is the s-wave scattering length, and both n and a are under experimental
control.
To determine the magnitude of the superradiant amplication, we numerically solved the
wave equation (3) using the XMDS package [35]. We used an initial Gaussian wavepacket of
the form (0, r) = A(r), with frequency
0
. Specically,
A(r) = e
(rr
init
)
2
/w
2
e
i
0
r/c

t
(0, r) = (i
0
2c

(r r
init
)/w
2
)A(r).
(21)
We found that superradiance occurred for a wide range of parameters, as long as the relevant
inequalities were fullled. Figure 1 shows a particularly strong and clean example of the
scattering of an m = 1 wavepacket. Comparing the Fourier components of the real parts of
the incident and reected wavepackets in the asymptotic region, we found that the dominant
Fourier power in the reected wavepacket is approximately doubled.
The parameters are given in the caption and are representative of conditions for a trapped
dilute gas BEC. They are made dimensionless, indicated by a tilde, by measuring distance and
3838 T R Slatyer and C M Savage
1
0
1
2
0
2
-600 -400 -200 0 200
1
0
1

r*
(a)
(b)
(c)

Figure 1. Wavepacket propagation from numerical solution of equation (3). The real part of the
wavepacket is plotted versus the dimensionless modied radial tortoise coordinate r

, dened in the
text, with = 800/801 so that r

= 800 at r = r
0
. The initial wavepacket, at t = 0, was given by
equation (21) with r
init
= 200 and w = 10. (a) t = 114: the wavepacket is propagating towards the
vortex. (b) t = 330: the wavepacket has just split into reected and transmitted parts. (c) t = 504:
the reected wavepacket. Dimensionless parameters: m = 1, = 2, c

= 1, = 20. The
spatial grid had 8192 points over r

[800, 400], and the time steps were 0.015 dimensionless

time units.
time in units of r
0
and
1
0
, respectively. For example, choosing r
0
= 1 mand
1
0
= 10
3
s,
the parameters of gure 1 correspond to c

= 10
3
ms
1
and = 20 m. For the case of a
Rb
87
BEC, m
atom
= 1.4 10
25
kg, and the asymptotic healing length 0.5, and = 2
corresponds to a vortex charge of about l = 3. Note that the real part of the wave is plotted
against the modied tortoise coordinate r

= r 2
2
/( r 1 + ). This magnies the scale
as r r
0
( r 1) , while limiting the lower bound of r

to 1 2
2
/. The ergoregion
boundary, or static limit, is at r 8 or r

92.
The local healing length determines the validity of the hydrodynamic approximation,
and is larger than the asymptotic value by the factor

/(r). Because of this the BEC

interpretation fails for gure 1(b). However, we have found a tradeoff between the strength of
the superradiance and the validity of the BEC hydrodynamic approximation. With increasing
m, and sound wavelength, reection occurs at larger values of r, and hence for smaller local
healing lengths, while the superradiance decreases in strength. We dene the reection point
to be at the maximum of the effective potential equation (5). Compared to the case of gure 1,
a transmitted wavelength comparable to the local healing length at the reection point,
accompanied by a 70% reduction in superradiance power, occurs for an initial wavelength
an order of magnitude longer, and a factor of 5 larger.
Numerical simulations of BECs under the proposed experimental conditions, and without
the hydrodynamic approximation, could explore the effect of its breakdown as the wavelength
Superradiant scattering from a hydrodynamic vortex 3839
of the transmitted wave decreases due to its inward propagation. This is related to the trans-
Planckian problem of Hawking radiation [36].
In summary, our analytical and numerical work has shown superradiant sound scattering
from hydrodynamic vortices. We have also argued for its experimental realization in dilute
gas BoseEinstein condensates. This might provide a useful step towards the ultimate goal of
observing the sonic analogue of Hawking radiation.
Acknowledgments
ACQAOis an Australian Research Council Centre of Excellence. This research was supported
by the Australian Partnership for Advanced Computing.
References
[1] Hawking S W 1974 Nature 248 30
[2] Hawking S W 1975 Commun. Math. Phys. 43 199
[3] Unruh W G 1981 Phys. Rev. Lett. 46 1351
[4] Zurek W H 1996 Phys. Rep. 276 177
[5] Visser M 2002 Articial Black Holes ed M Novello, M Visser and G Volovik (Singapore: World Scientic)
[6] Barcelo C, Liberati S and Visser M 2005 Preprint gr-qc/0505065
[7] Jacobson T 1999 Prog. Theor. Phys. Suppl. 136 1
[8] Bekenstein J D and Schiffer M 1998 Phys. Rev. D 58 064014
[9] Frolov V P and Novikov I D 1998 Black Hole Physics (Dordrecht: Kluwer)
[10] Volovik G E 2001 Phys. Rep. 351 195
[11] Basak S and Majumdar P 2003 Class. Quantum Grav. 20 2929
[12] Basak S and Majumdar P 2003 Class. Quantum Grav. 20 3907
[13] Giovanazzi S et al 2004 Phys. Rev. A 70 063602
[14] Leonhardt U, Kiss T and Ohberg P 2003 J. Opt. B: Quantum Semiclass. Opt. 5 S42
[15] Barcelo C, Liberati S and Visser M 2003 Int. J. Mod. Phys. A 18 3735
[16] Garay L J, Anglin J R, Cirac J I and Zoller P 2000 Phys. Rev. Lett. 85 4643
[17] Garay L J, Anglin J R, Cirac J I and Zoller P 2001 Phys. Rev. A 63 023611
[18] Barcelo C, Liberati S and Visser M 2000 Class. Quantum Grav. 18 1137
[19] Dicke R H 1954 Phys. Rev. 93 99
[20] Berti E, Cardoso V and Lemos J P S 2004 Phys. Rev. D 70 124006
[21] Bloch I, Hansch T W and Esslinger T 1999 Phys. Rev. Lett. 82 3008
[22] Robins N P et al 2004 Phys. Rev. A 69 051602(R)
[23] Robins N P et al 2005 Preprint cond-mat/0501747
[24] Madison K W, Chevy F, Wohlleben W and Dalibard J 2000 Phys. Rev. Lett. 84 806
[25] Hodby E, Hechenblaikner G, Hopkins S A, Marago O M and Foot C J 2001 Phys. Rev. Lett. 88 010405
[26] Birrell N D and Davies P C W 1982 Quantum Fields in Curved Spacetime (Cambridge: Cambridge University
Press)
[27] Unruh W G and Schutzhold R 2003 Phys. Rev. D 68 024008
[28] Pethick CJ and Smith H2002 BoseEinstein Condensation in Dilute Gases (Cambridge: Cambridge University
Press)
[29] Barcelo C, Liberati S and Visser M 2003 Phys. Rev. A 68 053613
[30] Svidzinsky A A and Fetter A L 1998 Phys. Rev. A 58 3168
[31] Simula T P, Virtanen S M M and Salomaa M M 2002 Phys. Rev. A 65 033614
[32] Stamper-Kurn D M et al 1998 Phys. Rev. Lett. 81 2194
[33] Vilenkin A 1978 Phys. Lett. B 78 301
[34] Feder D L, Svidzinsky A A, Fetter A L and Clark C W 2001 Phys. Rev. Lett. 86 564
[35] A description of the XMDS package may be found at: http://www.xmds.org
[36] Unruh W G 1995 Phys. Rev. D 51 2827