news & views
are used in concert with staging to address
each of the problems individually11. However,
the reduced acceleration gradient per stage
and the coupling distance between stages
needed to inject the electron beam and fresh
laser pulses has the ultimate effect that a
TeV-level laser-wakefield accelerator would
still be over 100 m long.
For laser-wakefield accelerators to reach
very high energies in a very short distance, it
is clear that we need new approaches. Over
recent years, our ability to manipulate the
shape of high-power laser pulses in space and
time has increased. For example, in 2017 and
2018 Sainte-Marie et al.12 and Froula et al.13
developed techniques to create light pulses
that appear to travel at an arbitrary velocity in
vacuum. Both groups recognized that these
‘sliding’ or ‘flying’ focus techniques could
revolutionize laser-wakefield accelerators.
In 2019, Debus et al.3 published a
technique that overlaps two laser pulses
so that, in a scissor-like motion, their
interference generates an intensity peak that
travels at the speed of light in vacuum, phase
locking the acceleration. Caizergues et al.2
and Palastro et al.4 both submitted their
schemes to beat dephasing for publication
in 2019. These three new ideas address
dephasing, depletion and diffraction in
laser-wakefield accelerators and promise
100 GeV electron beams in accelerators
less than 1 m long using ultra-short,
high-energy laser pulses.
The solutions of Caizergues et al.2 and
Palastro et al.4 (shown in Fig. 1) both use
the ability of an axiparabola7 to provide an
extended region of high intensity in which
light arriving at each particular radius on
the axiparabola is focused at a different axial
position. Not only does this overcome the
issue of diffraction, but it also addresses
the issue of depletion as ‘fresh’ laser energy
NONLINEAR OPTICS
Disordering a superfluid of light
A phase transition between disordered and quasi-ordered states, known as the Berezinskii–Kosterlitz–Thouless
transition, has now been revealed in a two-dimensional photonic ‘fluid’. The interplay between phase singularities
and coherence may lead to new vortex-based optical applications.
Matthieu Bellec and Claire Michel
P
hotonic devices utilizing integrated
linear and nonlinear circuits, optical
metamaterials and fibre optics can be
considered as low-dimensional systems.
This arguably limits communication
Nature Photonics | VOL 14 | August 2020 | 466–474 | www.nature.com/naturephotonics
arrives at each axial position without having
to propagate through the whole accelerator.
The issue of dephasing is addressed by
applying a tailored, radially dependent pulse
front delay to the laser pulse. When the laser
is then focused by the axiparabola, this radial
pulse front delay translates to axial control
of the speed of the intensity peak. When
tuned correctly, this can set the speed of
the intensity peak inside the plasma to the
speed of light in vacuum, thereby locking the
accelerating structure to the electron beam
and eliminating dephasing. The works of
both Palastro et al.4 and Caizergues et al.2
impose this radial pulse front delay by
controlling the shape of the laser pulse in
both space and time before focusing, with
Palastro et al. proposing a stepped echelon
optic to implement this.
With the new schemes proposed by
Debus et al.3, Caizergues et al.2 and Palastro
et al.4, the incredibly high accelerating
fields that can be created in a high-density
plasma could, at least in theory, be properly
exploited for the first time. But to drive
a strong wake, it is still important for the
laser pulse duration to be shorter than the
period of the plasma wave. This means
that to produce very high-energy electron
beams, very short duration, very high-power
laser pulses will be required, and these will
need to be controlled at a level that has
yet to be achieved with high-energy laser
pulses. Other challenges will include the
development of suitable plasma sources, and
the fact that as the fresh laser energy arrives
from the side, it must pass through the edges
of the plasma and wakefield structure itself,
something that could significantly reduce the
intensity reached on-axis.
The three works2–4 present different
approaches to how the schemes might
scale: Palastro et al.4 suggest that a 100 GeV
and information processing capabilities.
Although efforts have been made to
increase dimensions, through for
example, optical modal multiplexing
electron beam will need a 500 J, 15 fs laser
pulse, Caizergues et al.2 suggest this could
be achieved with a 60 J, 15 fs laser, while the
scheme of Debus et al.3 indicates the same
could be achieved with as little as 30 J.
Of course, there must be some limit to
how low the laser energy can be: assuming
10% of the laser energy gets transferred to
the accelerated electron beam, a 100 J
laser could only ever accelerate 100 pC
of charge (less than 1 billion electrons)
up to 100 GeV.
Despite their differences, all three
schemes show that intricate control of the
shape of high-power laser pulses in space
and time could reap enormous benefits
for laser-wakefield accelerators. With the
theoretical concepts now proposed, the
race is on to demonstrate these schemes
in the lab. ❐
Rob J. Shalloo    ✉ and
Stuart P. D. Mangles    ✉
The John Adams Institute for Accelerator Science,
Imperial College London, London, UK.
✉e-mail: r.shalloo@imperial.ac.uk;
stuart.mangles@imperial.ac.uk
Published online: 27 July 2020
https://doi.org/10.1038/s41566-020-0670-5
References
1. Kneip, S. et al. Phys. Rev. Lett. 103, 035002 (2009).
2. Caizergues, C., Smartsev, S., Malka, V. & Thaury, C. Nat. Photon.
https://doi.org/10.1038/s41566-020-0657-2 (2020).
3. Debus, A. et al. Phys. Rev. X 9, 031044 (2019).
4. Palastro, J. P. et al. Phys. Rev. Lett. 124, 134802 (2020).
5. Leemans, W. P. et al. Nat. Phys. 2, 696–699 (2006).
6. Shalloo, R. J. et al. Phys. Rev. Accel. Beams 22, 041302 (2019).
7. Smartsev, S. et al. Opt. Lett. 44, 3414–3417 (2019).
8. Steinke, S. et al. Phys. Plasmas 23, 056705 (2016).
9. Guillaume, E. et al. Phys. Rev. Lett. 115, 155002 (2015).
10. Gonsalves, A. J. et al. Phys. Rev. Lett. 122, 084801 (2019).
11. Schroeder, C. B. et al. Phys. Rev. ST Accel. Beams 13, 101301 (2010).
12. Sainte-Marie, A., Gobert, O. & Quere, F. Optica 4,
1298–1304 (2017).
13. Froula, D. H. et al. Nat. Photon. 12, 262–265 (2018).
or 3D-nanoprinting techniques, there
are technological limitations.
However, low dimensionality is actually
not always a restriction. One-dimensional
(1D) or 2D photonic devices can lead to rich
471
news & views

Nonlinear crystal No vortices Vortex–anti-vortex pairs Isolated vortices

Out

2D fluid of light dynamics along the z axis

y In
z
T
x 0 T < Tc T = Tc T > Tc

Fig. 1 | The BKT transition in a fluid of light. Shining a laser on a nonlinear crystal allows a 2D ‘fluid of light’ to be created that propagates and evolves along
the z axis (left). Random initial speckle controls the effective temperature of the system (right, bottom row). As the photonic fluid propagates, vortices are
generated. Zoomed-in sketches of their phase distribution, usually measured at the output of the crystal, are presented (top row). At low temperature, vortices
are always paired. The phase coherence is affected only locally. This is a quasi-ordered state. Above the critical temperature Tc, free vortices proliferate.
The phase coherence is affected globally thus destroying the order.

and unexpected phenomena that may offer
applications in both classical and quantum
optics. For instance, topological photonics,
a rapidly growing research field, is steeped
in the 2D quantum Hall solid-state physics
effect. Even the simpler topological photonic
insulator 1D counterpart recently led to
intense investigations. Feasible applications
in robust optical transport, topological
lasing and quantum communication have
been proposed and realized1.
Likewise, the Berezinskii–Kosterlitz–
Thouless (BKT) transition2 is another
example showing that low-dimensional
condensed-matter physics yields rich
phenomena. This non-conventional phase
transition from disordered to quasi-ordered
states in 2D systems, mediated by
topological defects underlying the dynamics,
has been observed in quantum systems.
Nonetheless, this transition mainly involves
classical nonlinear wave dynamics.
Now, writing in Nature Photonics, Situ
and Fleischer report the first observation
of the BKT transition in optics, using a
continuous laser in a crystal exhibiting a
large optical nonlinear response, creating
a 2D ‘fluid of light’ that evolves along
the propagation axis3. In their work, the
associated topological defects are optical
vortices, defined as phase singularities in the
transverse optical wave field.
The science of structured light with
singularities has already led to strong
basic and applied scientific interests. Thus,
besides the fundamental importance of
this observation, the researchers’ results,
involving complex optical vortex dynamics,
naturally appeal to potential applications in
singularity-based photonic 2D devices4.
An important property of
low-dimensional physical systems (1D or
2D) is the absence of long-range order,
for example, crystalline, ferromagnetic or
superfluid states, due to inherent thermal
fluctuations at finite temperature. When
short-range interactions were additionally
considered, Berezinskii, and Kosterlitz
and Thouless independently predicted in
the 1970s that a phase transition between
ordered and disordered states may occur
in 2D systems2; it is usually said to be
an unconventional transition, being
continuous, mediated by topological defects
and without symmetry breaking. While 2D
crystals or magnets are rather complicated
to implement and study experimentally,
2D fluids of interacting particles, possibly
obeying Bose–Einstein statistics, are
much more accessible. Examples of such
systems range from liquid helium and
superconductors to cold atoms and, more
recently, quantum fluids of light. The order
is then characterized by a high degree of
phase coherence at long range, which is also
called the superfluid state.
The role of the vortices in the BKT
transition can be intuitively understood as
follows. Consider the lowest-energy state in
a quantum fluid that has constant amplitude
with flat phase. Adding a single vortex,
defined by a wavefront spinning around
a vanishing amplitude point, affects the
field in a global way. At high temperatures,
random motion of such vortices, which
proliferate as the temperature increases,
destroys large-scale phase coherence. But, if
two vortices spinning in opposite directions
are paired together, which may occur as the
temperature is decreased, the phase pattern
is altered only locally and their circulation
cancels, thus allowing long-range order to
appear. More precisely, the spatial phase
correlations decrease exponentially in the
disordered (normal) state and algebraically
(power law) in the quasi-ordered
(superfluid) state. At the BKT transition,
which occurs at a critical temperature
Tc, a universal jump in the superfluid
density is predicted. The first experimental

472 Nature Photonics | VOL 14 | August 2020 | 466–474 | www.nature.com/naturephotonics


measurements of this jump took place soon
after the BKT transition was predicted
in helium films5. In 2006, the transition
was observed in ultracold atoms and first
attempts to probe vortex dynamics were
reported6. More recently, quantum fluids of
light have been proposed as a rich alternative
playground to quantitatively study advanced
quantum hydrodynamics features7,
including superfluidity7,8 and the BKT
transition. Algebraic correlation functions
were reported with exciton-polaritons in
2D semiconductor microcavities9. Inspired
by a pioneering theoretical study10, Situ and
Fleischer characterized the optical BKT
transition in a nonlinear optical set-up,
allowing investigation of both the phase
coherence features of the fluid of light
and the complex vortices underlying the
dynamics of the transition3.
The experimental set-up consists of
a continuous laser beam propagating in
an optical crystal with a large nonlinear
photorefractive response (Fig. 1). As in
other nonlinear media, such as atomic
vapours11 or thermal liquids12, under the
paraxial approximation, the transverse
evolution of the optical beam is analogous
to a weakly interacting Bose–Einstein
condensate. The propagation axis plays
the role of time, while nonlinearities and
refractive index modulation act as effective
interactions and external potential. Such
nonlinear optical dynamics is usually
reformulated in terms of quantum
hydrodynamics where the laser intensity and
wavefront gradient correspond to the fluid
density and velocity, respectively. To deepen
the analogy, a uniform wavefront scenario
is typically analogous to a zero-temperature
Bose–Einstein condensate. Adding a
controlled amount of random-phase
components in the beam, achieved
experimentally by using a spatial light
modulator, allows the condensate’s effective
temperature to be increased. This imposes
density and phase fluctuations of the fluid of
light, eventually leading to the generation
of vortices, and enabling triggering of the
BKT transition. In principle, noise alone
should provoke the generation of the
vortices, but in practice this is not sufficient.
In the work of Situ and Fleischer a lattice is
photoinduced in the photorefractive crystal
to help trigger vortices.
For each parameter set, by varying the
effective temperature and the strength and
sign of the interaction, the output complex
Nature Photonics | VOL 14 | August 2020 | 466–474 | www.nature.com/naturephotonics
wavefunction (that is, amplitude and
phase of the optical field) is reconstructed
via a standard interferometric technique.
Two main quantities are then extracted to
characterize the BKT transition: (1) the
number of new vortices that have been
generated along the beam path along with
their pairing status (vortices are considered
as paired if their separation is smaller or
equal to three vortex cores); and (2) the
radial correlation function. The radial
correlation function’s cumulative integral
can be expressed, for all temperatures, T,
simply as Ω2κ(T), where Ω is the integrating
area and κ is an exponent to be fitted. In
theory, by looking at κ as a function of T,
the phase transition can be identified as
a jump from κ = 0.875 at just below the
critical temperature Tc, to κ = 0.5 just above
Tc, corresponding to the algebraic decay of
the spatial correlations discussed above in
the case of κ = 0.875, and exponential
decay for κ = 0.5.
The experimental results of Situ and
Fleischer show that the correlation functions
obey the exact scaling behaviour predicted
by the BKT theory. The values of κ are
directly linked to the (universal) jump
in the superfluid density. It is worth
mentioning that the expected step function
at Tc was not observed; instead a smooth
crossover was seen, an effect attributable to
the finite size of the experimental system,
which is an effect also observed in ultracold
atom experiments6.
Above the critical temperature,
the number of single free vortices
that proliferate drastically increases.
Counting them as a function of T allows
to quantitatively obtain the value of Tc.
Interestingly, the researchers observe that
Tc actually depends on two parameters.
Typical values of the normalized critical
temperature range from 0.6 to 0.95 in
accordance with previously reported results.
First, Tc is slightly shifted forward when
the nonlinear strength is larger, whatever
its sign. This is expected as the role of the
density fluctuations is smaller in this case.
Second, the researchers measure a linear
dependence of Tc with the lattice spacing.
A thermodynamics approach tells us that
smaller lattice spacings provide more
available sites for occupation (per a
fixed area), increasing the entropy and
indeed lowering the critical temperature
needed to overcome the vortex energy.
For a homogeneous system (that is, when
news & views
lattice spacing approaches zero), the
transition should occur at finite Tc. What
the researchers observe, however, is a
wave condensation phenomenon with no
evidence of vortex proliferation13. This
raises the important issue of the role of
the lattice in these experiments and the
interplay between continuous and discrete
dynamics. Situ and Fleischer found that the
experimental parameter ranges are limited
and that the vortices can’t be trigged by
only noise (temperature fluctuations) in
homogeneous fluids of light, thus requiring
an additional lattice.
The results reported by Situ and Fleischer
are of fundamental importance and
constitute a starting point for exploring
the full phase space of non-equilibrium
and vortex-coherence dynamics in fluids
of light. An interesting aspect to follow-up
on may be the role of the lattice arrangement
(with, for example, topological or disorder
features) on the transition. Fluids of
light — thanks to the various and
complementary existing photonic
platforms — offer a rich playground to
quantitatively study advanced quantum
hydrodynamics features. Although
restricted to 2D geometries, the work
of Situ and Fleischer also shows that
photonic systems generally exhibit complex
dynamics with new light transport
properties that may potentially lead to
important applications in photonics. ❐
Matthieu Bellec    ✉ and Claire Michel ✉
Institut de Physique de Nice, Université Côte d’Azur
& CNRS, Nice, France.
✉e-mail: matthieu.bellec@inphyni.cnrs.fr;
claire.michel@inphyni.cnrs.fr
Published online: 27 July 2020
https://doi.org/10.1038/s41566-020-0663-4
References
1. Ozawa, T. et al. Rev. Mod. Phys. 9, 015006 (2019).
2. Kosterlitz, J. M. & Thouless, D. J. J. Phys. Condens. Matter 5,
L124–L126 (1972).
3. Situ, G. & Fleischer, J. W. Nat. Photon. https://doi.org/10.1038/
s41566-020-0636-7 (2020).
4. Soskin, M. et al. J. Opt. 19, 010401 (2016).
5. Bishop, D. J. & Reppy, J. D. Phys. Rev. Lett. 40, 1727–1730
(1978).
6. Hadzibabic, Z. et al. Nature 441, 1118–1121 (2006).
7. Carusotto, I. & Ciuti, C. Rev. Mod. Phys. 85, 299–366 (2013).
8. Michel, C. et al. Nat. Commun. 9, 2108 (2018).
9. Roumpos, G. et al. Proc. Natl Acad. Sci. USA 109,
6467–6472 (2012).
10. Small, E. et al. Phys. Rev. A 83, 013806 (2011).
11. Fontaine, Q. et al. Phys. Rev. Lett. 121, 183604 (2018).
12. Vocke, D. et al. Optica 2, 484–490 (2015).
13. Sun, C. et al. Nat. Phys. 8, 470–474 (2012).
473