Photon Bubble Turbulence in Cold Atomic Gases

Jo
ao D. Rodrigues,1, Jose A. Rodrigues,2, 1 Antonio V.
Ferreira,1 Hugo Tercas,3 Robin Kaiser,4 and Jose T. Mendonca1

arXiv:1604.08114v1 [physics.atom-ph] 27 Apr 2016

Instituto de Plasmas e Fus

ao Nuclear, Instituto Superior Tecnico,
Universidade de Lisboa, 1049-001 Lisbon, Portugal
2
Departamento de Fsica, Faculdade de Ciencias e Tecnologia,
Universidade do Algarve, 8005-139 Faro, Portugal
3
Physics of Information and Quantum Technology Group, Instituto de Telecomunicaco
es, Lisbon, Portugal
4
Institut Non Lineaire de Nice, CNRS and Universite Nice Sophia Antipolis,
1361 route des Lucioles, 06560 Valbonne, France
Turbulent radiation flow is ubiquitous in many physical systems where light-matter interaction becomes
relevant. Photon bubbling, in particular, has been identified as the main source of turbulent radiation
transport in many astrophysical objects, such as stars and accretion disks. This mechanism takes place when
radiation trapping in optically dense media becomes unstable, leading to the energy dissipation from the
larger to the smaller bubbles. Here, we report on the observation of photon bubble turbulence in cold atomic
gases in the presence of multiple scattering of light. The instability is theoretically explained by a fluid
description for the atom density coupled to a diffusive transport equation for the photons, which is known
to be accurate in the multiple scattering regime investigated here. We determine the power spectrum of the
atom density fluctuations, which displays an unusual k4 scaling, and entails a complex underlying turbulent
dynamics resulting from the formation of dynamical bubble-like structures. We derive a power spectrum
from the theoretical photon bubble model which, to a high level of accuracy, explains the observations. The
experimental results reported here, along with the theoretical model we developed may shed light on the
analogue photon bubble instabilities in astrophysical scenarios.
PACS numbers: 37.10.Vz, 52.35.Ra, 47.27.-i

Due to multiple scattering in dense astrophysical systems such as massive stars, light propagation is often diffusive instead of ballistic. In such optically thick media,
photons exhibit high residence times and can therefore
be trapped [1]. One interesting consequence of such diffusive behaviour is the so-called photon bubbling [2], an
effect that is possible when the system dynamically responds to the propagation of light. Photon bubbling consists in the formation and growth of light bubbles that
eventually leave the medium or decay into smaller bubbles. This decay is at the origin of a specific form of
turbulence in optically thick media, the photon bubble
turbulence (PBT). Interestingly, the PBT spectrum significantly differs from the usual Kolmogorov (or similar)
decay cascades, which scale as k 5/3 . Physically, the
mechanism leading to turbulence is the saturation of the
photon-bubbling instability, and the corresponding decay spectrum strongly depends on the optical properties
of the media.
In this Letter, we reveal the experimental evidence of
a photon bubble instability in an optically thick, cloud of
cold atoms driven close to the atomic resonance, conditions under which diffusive photon transport takes place
[35]. We determine the spectrum of the density fluctuations, which we compare with an extension of a previously developed model of photons bubbles. The obtained
theoretical spectral dependence, which significantly differs from the Kolmogorov case, is shown to accurately

Corresponding author: joaodmrodrigues@tecnico.ulisboa.pt

describe all the features of the observed turbulence spectrum. We start by briefly introducing a fluid model for
the cold atoms in a laser trap, revealing the essential ingredients for the photon bubble instability. We describe
the experimental set-up and the conditions under which
such instability is observed. The turbulence spectrum is
measured and compared to the theory developed here. A
detailed analysis of the results is presented before stating
some final conclusions.
For conditions of high optical densities and near resonant driving, the laser cooling beams are deflected by
multiple scattering inside the medium, randomizing the
photon propagation direction thus turning photon transport into an isotropic diffusive process described by [6, 7]

I (DI) = a I,
t

(1)

where I is defined as the spectral energy density of the

electromagnetic radiation field W (, k), integrated over
all possible directions of propagation and a = Dka2 with
ka the inverse of the energy absorption length. The photon mean free path l = 1/nL , with n the atom density
and L the photon absorption cross-section, determines
the light diffusion coefficient as D = l2 / . The photon diffusion time, , is usually considered independent
from the atom density [8], yielding a dependence of the
form D n2 while providing and a strong coupling
between the atom dynamics and photon transport. On
the other side, the atom density n and the velocity field v
are determined by the usual continuity and Navier-Stokes

y (mm)

0 ms

3 ms

6 ms

9 ms

1
0
-1
-2
-2 -1 0

y (mm)

12 ms
x (mm)

-2 -1 0

15 ms
x (mm)

-2 -1 0

18 ms
x (mm)

-2 -1 0

21 ms
x (mm)

1
0
-1
-2
-2 -1 0

-2 -1 0

x (mm)

x (mm)

-2 -1 0

x (mm)

-2 -1 0

x (mm)

FIG. 1. (color online) Temporal evolution of the turbulent phase, after the photon bubble instability is fully developed. The
plots corresponds to the fluorescence signal (proportional to the atom density, integrated along the line-of-sight) fluctuations for each frame the average density profile, computed with as many as 500 frames, is subtracted. The blue and red colors then
represent regions of higher and lower densities than the local average density, respectively. The grey counters corresponds to
level lines of the average density, thus representing the shape of the average atom distribution. The inset plot depicts the real
atom density distribution (integrated along the line-of-sight of the camera) without the subtracted averaged, in false color code
(blue regions correspond to high atom densities).

equations
n
+ (nv) = 0,
t

and

v
F
P
+ (v )v =

v,
t
m
nm

(2)

(3)

respectively, where P is the thermodynamical pressure,

the damping rate from optical molasses and m the atom
mass. In the presence of multiple scattering of light a
radiation pressure force scaling as 1/r2 appears due to
exchange of photons between nearby atoms. This collective force F is determined by a Poisson-like equation,
F = Qn,

(4)

where Q = (R L )L I/c is related to an effective electrical charge for the atoms and L and R are the atomphoton absorption and re-emission cross-sections, respectively [3, 9]. This suggests the
p introduction of an effective plasma frequency, p = Qn0 /m [4, 5], and allows
for an analogy with an electrostatically self interacting
one-component trapped plasma, which has been explored
before in the context of exotic phenomena such as classical rotons [10] or instability of twisted excitations [11].
In typical experimental conditions
[12, 13] the effective

charge is of the order q = 0 Q 104 e, with e the electron charge, and the atom densities n0 1010 cm3 , resulting in plasma frequencies of the order p 100 200
Hz. Notice that the effective charge Q depends on the
laser intensity, which provides another coupling mechanism between the atom density n and photon intensity

I. Such coupling, determined by Eqs. (1), (2), (3) and

(4), eventually becomes unstable leading to the nucleation of photon bubbles, representing a local increase in
the photon number accompanied by a deplection in the
atom density [2]. Here we describe the experimental evidence of such an instability. We should note that the
photon bubble mechanism differs from the recently observed pattern formation instability in cold atoms [14].
In the former case, optical dipole forces, from far offresonant light, are coupled to the atomic density distribution, whereas here we rely on quasi-resonant radiation
pressure. Moreover, the pattern formation mechanism
results from coherent photon transport, whereas in the
photon bubble instability light is in the diffusive limit.
Our experiment consists of a magneto-optical trap
(MOT) [15], where 85 Rb atoms are collected from a dilute vapor at a background pressure of 108 Torr. Six
independent colling, and trapping, laser beams cross the
center of the trap with a beam waist of w 4 cm, power
per beam P 40 mW and wavelength 780 nm.
Since the beams are not retro-reflected we guarantee that
the observed instability does not originate from feedback
mechanisms [16]. The cooling laser operates on the D2
line of 85 Rb (F = 3 F 0 = 4), and is red-detuned by ,
which can be controlled by a double passage through an
acousto-optic modulator (AOM), with a frequency tuning
precision higher than the transition linewidth, /2 6
M Hz. A magnetic field gradient (B) with a zero field
at the center of the trap generates a spatially dependent
Zeeman split of the energy levels, allowing for magnetooptical trapping of the atoms. A finite probability exists

to excite the open transition F = 3 F 0 = 3, which, in

turn, is responsible for a ground state change via a spontaneous Raman transition. An additional repump beam,
operating on the hyperfine transition F = 2 F 0 = 3
repopulates the F = 3 hyperfine ground state, thus allowing the cooling process to continue. The repump detuning is set by searching for the maximum fluorescence
signal. We thus obtain a cold cloud with N 1010 atoms
at T 100 K. A CCD camera working at approximately 350 fps collects the fluorescence signal from the
cold atoms. In this way we dynamically measure the
atomic distribution of the trap, integrated along the lineof-sight of the camera.
The photon bubble instability described here is observed when the cooling beams are tuned very close to
the atomic resonance, = 0.5 , which we keep constant during the entire experiment time. In such conditions, significant photon trapping occurs inside the cold
gas and the diffusive approximation in Eq. (1) becomes
the appropriate description of the photon density evolution which, in turn, influences the atom dynamics according to Eq. (2), (3) and (4). We have also set the magnetic
field gradient to B 9 G/cm, although this parameters
has no strong influence on the observed atomic dynamics, in contrast with the laser detuning , which strongly
determines the system behaviour. In Fig. (1) we plot the
temporal evolution of a photon bubble instability, where
complex atomic motion takes place. Due to density inhomogeneities, localized photon trapping occurs, which
grows from the onset of the instability forming bubblelike structures. The growth of these structures is accompanied by a local decrease in the atom density, due to
radiation pressure, which eventually leads to the decay
of the bubble due to photon losses and the subsequent
restablishment of the local atom density. Such processes
occur at a time scale much slower than the typical plasma
frequency, as expected from the photon bubble model [2].
To sustain this claim we determined the frequency power
spectrum of the atom density fluctuations by averaging
the squared amplitude of the Fourier Transforms of the
time signals from each pixel of the CCD camera - see
Fig. (2). We start by noting that for an high laser detuning, = 1.5 , we observe isolated frequency peaks.
In fact, it has been reported before [12] that for such
region of experimental parameters the system undergoes
supercritical Hopf bifurcations resulting in self-sustained
oscillations in the collective atom dynamics. This scenario is consistent with the red line spectrum in Fig. (2).
When tuning the beams closer to the atomic resonance
however, the systems undergoes a significant transformation and an unforeseen regime of complex atom motion
takes place - = 0.5 , black line in Fig. (2). In
this new regime, the frequency power spectrum exhibits
broad band components at low frequencies, as expected
from dissipative turbulent systems [17]. This indicates a
weak turbulent-like evolution of the atom density, char-

P (f ) (arb. units)

3
10 0

10 -1

50

100

150

Frequency (s1 )
FIG. 2. (color online) The red line represents the power spectrum of the atom dynamics (in the time domain), for a region
of parameters ( = 1.5 ) where supercritical Hopf bifurcations are observed, resulting in self-sustained oscillations and
corresponding to isolated frequency peaks. In the darker line
( = 0.5 ) a new dynamical regime is depicted where a
characteristic turbulence power spectrum is observed. Broad
band components are present at low frequencies, as expected
in scenarios of dissipative turbulence - see Discussion.

acterized by fairly well defined dynamical structures, as

we are about to demonstrate.
Let us investigate the power spectrum of the density fluctuations, but now in the space domain. We
take as many as 500 fluorescence images of the atom
cloud, in the experimental conditions described before,
yielding the spatial atom distribution integrated along
the line-of-sight, which we shall denote by n
i (), with
= (x, y) and i = 1, 2, ..., 500. From here we compute the mean atom distribution by averaging over the
500 samples, n
0 () = h
ni ()i, and subtract to each
frame to determine the fluctuating density distributions

ni () = n
i () n
0 (). To determine the spectral dependence of the density fluctuations, we Fourier transform
each real space distribution
ni () to obtain
ni (k). Finally, we can average the squared amplitude of each twodimensional momentum distribution, P (k) = h|
ni (k)|2 i.
Note that, assuming isotropic density fluctuations allows
us to average the power spectrum in the angular dependence, yielding the desired power spectrum P (k) - see
Fig. (3). We shall at this point inquiry about the correctness of taking an integrated distribution and compute the Fourier Transform. The question that naturally
arises is if such spectrum does correspond to the real density fluctuations spectrum. The same problem arises, for
instance, when investigating turbulence in the inter stellar medium (ISM) [1, 18] from earth-based observations.
As it turns out, we can safely take the Fourier Transform
of the integrated signal and obtain the correct scaling of
the full three dimensional spectral dependence as long as
the depth (along the line-of-sight) of the probed medium
is at least of the order of the largest transverse scale we
are interested in [19, 20].
We shall now discuss the consequences and physical
meaning of such a particular k dependence in the density fluctuations spectrum. We start by noting that the

P (k) (arb. units)

4
10 0

spectral equations. In spherical coordinates r = (r, , )

the solutions are given by

10 -1

F (r) = j` (qr)Y`m (, ) ,

10 -2
10 0

10 1

Wavenumber k (mm1 )
FIG. 3. (color online) Experimental turbulence power spectrum (grey dots). An unusual k4 scaling at high wavenumbers is observed, indicating a fast decay of the density autocorrelation function and revealing the presence of well defined
dynamical spatial structures - see Discussion. The error bars
correspond to standard deviations when averaging over the
angular dependence of the power spectrum. The red line depicts a numerical fit to the theoretical power spectrum derived
in Eq. (11).

k 4 scaling of the turbulent spectral components corresponds to a much faster decay than the usual Kolmogorov spectrum, k 5/3 , and other similar turbulence
scaling laws. This indicates the existence of structures
with well defined scale lengths, and not self-similar or
fractal-like structures at different scales. This is easily
understood in terms of the auto-correlation function, defined as C(r) = hn(r)n(r + r)i. Note that, the autocorrelation function C(r) and the power spectrum P (k)
are conjugate quantities, forming a Fourier pair. For a
scaling of the form P (k) 1/(a2 +k 2 )2 , such as we experimentally observe, where a is a regularization at k 0
and related with the inverse of a typical scale, the autocorrelation function takes the form C(r) exp(ar)
with a fast decay at a 1/a scale, thus again evidencing
the existence of well defined spatial structures with 1/a
a typical size. This phenomenological analysis will be
made more precise in the subsequent discussion.
Let us now investigate to which extent the photon bubble model can account for such particular observations regarding the density fluctuations and the turbulence spectrum. In this sense we must go beyond the linear regime
considered in [2]. We begin by assuming slowly growing
photon density perturbations described by
n
(I,
) F (r) eit .

(5)

Here the mode frequency is much smaller than the effective plasma frequency defined before. Note that nonlinear mode saturation can be described by an Helmholtz
equation [2] of the form
2 F (r) = q 2 F (r) .

(6)

The real quantity q determines the scale of the bubble

solutions. We should note that Eq. (6) could be derived
directly from Eq. (1), (2), (3) and (4), yielding the same

(7)

where j` (r) are the spherical Bessel functions and Y`m the
spherical harmonics. We note that such functions may
not accurately describe realist photon bubble solutions,
as they are perturbed by background fluctuation, such as
sound waves, or by photon losses. The erosion associated
with lossy processes can be described by a radial absorption length , with is related with the damping term a
introduced earlier in Eq. (1). More realistic solutions are
then given by
F (r) = j` (qr)Y`m (, ) er .

(8)

The observed turbulent behavior is described by such

solutions, with random angular orientations. The corresponding (radial only) Fourier spectrum, average over
randomly oriented spherical harmonics, is given by
Z
P` (k) = 4
j` (qr)er eikr r2 dr,
(9)
0

where the 4 factor appears from the integration on the

angular variables. For the first two modes ` = 0 and
` = 1, we have
P0 (k) =

8 ( ik)
,
[q 2 + ( ik)2 ]2

P1 (k) =

and

8q
,
[q 2 + ( ik)2 ]2

(10)

(11)

respectively. Notice that, for k  q, the first mode, corresponding to spherical solutions, decays as k 3 while
the second mode, corresponding to dipolar-shaped bubbles, decays as k 4 . This suggest that we are indeed in
the presence of dipolar bubbles. To further clarify the
discussion we performed a numerical fit of the absolute
value of the expression in Eq. (11) to the experimentally measured power spectrum - see Fig. (3). Here, we
can verify the excellent agreement between the theoretical power spectrum and the experiment, over the entire range of accessible wavenumbers. Furthermore, from
the numerical fit we can extract the parameters of the
model. Particularly, we obtain q = 3.8 0.1 mm1 and
= 2.2 0.1 mm1 . Note that the power spectrum attains its maximum for k q and thus l0 = 2/q 1.6
mm corresponds to the typical bubble length scale. As
expected . q, which allows the bubbles to develop and
grow.
In conclusion, we demonstrated here that optically
thick samples of cold atoms driven close to the atomic
resonance can develop photon bubble instabilities due to
the strong coupling between the atom density and photon transport, which becomes diffusive due to multiple

5
scattering inside the medium, under appropriate experimental conditions. Such dynamical structures are at the
origin of a distinct low frequency turbulence with a spectrum scaling as k 4 , revealing the unsuspected dipolar
shape of the bubbles, according to the theory developed
here. Photon bubbles have been considered in astrophysical contexts, namely in the athmosphere of neutron stars
[21], black hole accretion disks [22, 23] and around young
massive stars [24]. Despite previous proposals of laboratory based simulation of such instabilities, namely with
high power lasers [25], this is, for the best of our knowledge, the first time photon bubbles resulting from radiation pressure effects are experimentally observed. Radiation pressure is also expected to be at the origin of
turbulence in the interstellar medium, with effects in star
formation or cosmic-ray propagation [1]. We thus pave
the way to investigate complex astrophysical situations
with cold atoms experiments.
JR acknowledges the Doctoral Programme in Physics
and Mathematics of Information (DP-PMI) and the financial support of FCT - Fundac
ao da Ciencia e Tecnologia through the grant number SFRH/BD/52323/2013.
HT thanks the support from Fundac
ao para a Ciencia e
a Tecnologia (Portugal), namely through the programme
PTDC/POPH. RK thanks Guillaume Labeyrie for many
discussions and experimental insights in large MOTs.

[1] B. G. Elmegreen and J. Scalo, Annual Review of Astronomy and Astrophysics 42, 211 (2004).
[2] J. T. Mendonca and R. Kaiser, Phys. Rev. Lett. 108,
033001 (2012).
[3] T. Walker, D. Sesko, and C. Wieman, Phys. Rev. Lett.
64, 408 (1990).
[4] J. Mendonca, R. Kaiser, H. Tercas, and J. Loureiro,
Phys. Rev. A 78, 013408 (2008).
[5] J. T. Mendonca and H. Tercas, Physics of Ultra-Cold
Matter (Spring Series on Atomic, Optical and Plasma
Physics Vol. 70, Berlin, 2012).

[6] M. C. W. van Rossum and T. M. Nieuwenhuizen, Rev.

Mod. Phys. 71, 313 (1999).
[7] A. Ishimaru, Wave Propagation and Scattering in Random Media (Academic, New York, 1978).
[8] G. Labeyrie, E. Vaujour, C. A. M
uller, D. Delande,
C. Miniatura, D. Wilkowski, and R. Kaiser, Phys. Rev.
Lett. 91, 223904 (2003).
[9] L. Pruvost, I. Serre, H. T. Duong, and J. Jortner, Phys.
Rev. A 61, 053408 (2000).
[10] H. Tercas, J. T. Mendonca, and V. Guerra, Phys. Rev.
A 86, 053630 (2012).
[11] J. D. Rodrigues, H. Tercas, and J. T. Mendonca, EPL
(Europhysics Letters) 113, 13001 (2016).
[12] G. Labeyrie, F. Michaud, and R. Kaiser, Phys. Rev.
Lett. 96, 023003 (2006).
[13] J. D. Rodrigues, J. A. Rodrigues, O. L. Moreira,
H. Tercas, and J. T. Mendonca, Phys. Rev. A 93, 023404
(2016).
[14] L. G., T. E., G. P. M., O. G.-L., F. W. J., R. G. R. M.,
A. A. S., K. R., and A. T., Nat Photon 8, 321 (2014).
[15] E. L. Raab, M. Prentiss, A. Cable, S. Chu, and D. E.
Pritchard, Phys. Rev. Lett. 59, 2631 (1987).
[16] D. Wilkowski, J. Ringot, D. Hennequin, and J. C. Garreau, Phys. Rev. Lett. 85, 1839 (2000).
[17] J. P. Eckmann, Rev. Mod. Phys. 53, 643 (1981).
[18] D. Falceta-Goncalves, G. Kowal, E. Falgarone, and
A. C.-L. Chian, Nonlinear Processes in Geophysics 21,
587 (2014).
[19] M.-A. Miville-Deschenes, F. Levrier, and E. Falgarone,
The Astrophysical Journal 593, 831 (2003).
[20] R. Romain, A. Jallageas, P. Verkerk, and D. Hennequin,
ArXiv e-prints (2016), arXiv:1602.06181 [physics.atomph].
[21] J. Arons, Astrophys. J. 388, 561 (1992).
[22] C. F. Gammie, Monthly Notices of the Royal Astronomical Society 297 (1998).
[23] M. C. Begelman, The Astrophysical Journal 643, 1065
(2006).
[24] N. J. Turner, E. Quataert, and H. W. Yorke, The Astrophysical Journal 662, 1052 (2007).
[25] S. J. Moon, S. C. Wilks, R. I. Klein, B. A. Remington, D. D. Ryutov, A. J. Mackinnon, P. K. Patel, and
A. Spitkovsky, Astrophysics and Space Science 298, 293.