You are on page 1of 6

ARTICLES

PUBLISHED ONLINE: 23 MAY 2010 | DOI: 10.1038/NPHOTON.2010.120

Temporal cavity solitons in one-dimensional Kerr


media as bits in an all-optical buffer
François Leo1 *, Stéphane Coen2, Pascal Kockaert1, Simon-Pierre Gorza1, Philippe Emplit1
and Marc Haelterman1

Temporal cavity solitons are packets of light persisting in a continuously driven nonlinear resonator. They are robust
attracting states, readily excited through a phase-insensitive and wavelength-insensitive process. As such, they constitute
an ideal support for bits in an optical buffer that would seamlessly combine three critical telecommunication functions,
namely all-optical storage, all-optical reshaping and wavelength conversion. Here, with standard silica optical fibres, we
report the first experimental observation of temporal cavity solitons. The cavity solitons are 4 ps long and are used to
demonstrate storage of a data stream for more than a second. We also observe interactions of close cavity solitons,
revealing for our set-up a potential capacity of up to 45,000 bits at 25 Gbit s21. More fundamentally, cavity solitons are
localized dissipative structures. Therefore, given that silica exhibits a pure instantaneous Kerr nonlinearity, our experiment
constitutes one of the simplest examples of self-organization phenomena in nonlinear optics.

T
emporal cavity solitons (CSs) are pulses of light going round array processor with spatial CSs as independent pixels. Spatial CSs,
and round indefinitely in a driven nonlinear optical cavity1. however, place stringent demands on the transverse uniformity of
Being solitons, they do not spread. Chromatic dispersion is the cavity, posing serious technological challenges. Material impuri-
balanced by nonlinearity2. As ‘cavity’ solitons, they do not dissipate. ties act as trapping sites, restricting the positioning freedom of CSs,
Losses are balanced by an external continuous-wave (c.w.) coherent whereas any transverse gradients lead to drifts15,19,20. Temporal CSs,
driving3–5. This double balancing act completely sets the soliton par- albeit limited to being unidimensional, are intrinsically immune to
ameters, making them unique6. They are also robust attracting these issues.
states. Any suitable addressing pulse will write a persistent soliton Our experiments are based on a passive ring cavity made of stan-
in the cavity. Providing they are not too close to one another, they dard optical fibres and rely on the instantaneous pure Kerr non-
do not interact either1,7. Any incoming binary-coded data stream linearity of silica2. Remarkably, the Kerr cavity dynamics has been
can thus be captured and stored indefinitely as a sequence of tem- shown to be governed by a mean-field model21,22 that is analogous
poral CSs, realizing an all-optical buffer1,8. Because the stored data to reaction-diffusion equations encountered in chemistry and devel-
stream is sustained without an amplifier, it does not accumulate opmental biology6,23, and more generally in nonlinear dissipative
noise as it continues to circle the cavity. Rather, noise present chemical systems. CSs are therefore a class of dissipative structures
on the input data is dissipated by the cavity: the uniqueness of the (DSs) akin to those first introduced by Prigogine’s school24,25. In
CS state provides a robust all-optical reshaping mechanism. this sense, the Kerr cavity appears as the simplest optical system
Clearly, temporal CSs are natural candidates for bits in an optical that demonstrates the universality of self-organization phenomena
memory and in all-optical processing systems. However, they in nonlinear optics. For that reason, it has been dubbed the ‘hydro-
remain virtually unstudied experimentally. Mitschke and Schwache gen atom’ of nonlinear cavities3, which highlights the fundamental
reported indirect observations of temporal soliton ensembles, but character of our observations. Consequently, we must stress that
not of isolated CSs9. Frequency combs in Kerr microresonators our work not only constitutes the first experimental observation
may be a signature of temporal CSs10. In contrast, we report here of temporal CSs, but also of Kerr CSs.
the first direct experimental observation of these objects.
CSs have been intensely studied theoretically since the early Results
1990s, predominantly in the field of two-dimensional transverse CSs often arise under conditions of coexistence, for the same value
nonlinear optics3–5,7,8,11–14. The term ‘cavity soliton’ was introduced of parameters, of two stable stationary states, respectively a c.w. (or
in these studies for the subset of dissipative solitons6 found in coher- homogeneous) state and a modulated (or patterned) state13,14,19,26–28.
ently driven optical cavities3,5. Here, we extend this terminology to In such a case, it is possible for the intracavity field to be homo-
the temporal case. Note that this excludes solitons in lasers and geneous in one region and modulated in another. These type of
other incoherently driven cavities15,16 (see also ref. 6). Three-dimen- intermediate solutions are localized DSs, and a CS is a localized
sional CSs also exist and are known as cavity light bullets17,18. In the DS with a single peak. The passive Kerr cavity is well known to
spatial context, CSs are c.w. light beams of planar resonators and exhibit bistability of the homogeneous state29,30. Moreover, when
nonlinearity balances diffraction rather than chromatic dispersion. the Kerr medium (considered here to be self-focusing as in our
Those spatial CSs have been observed using semiconductor micro- experiment) exhibits anomalous dispersion, the upper branch of
cavities19. The experiment confirmed that two-dimensional spatial the bistable cycle is unstable in favour of modulated solutions21,31,32
CSs are independently addressable and can be moved in the trans- through a Turing bifurcation33,34. This branch of modulated sol-
verse plane of the driving beam, realizing the prospect of an optical utions is stable and coexists with the homogeneous lower-branch

1
Service OPERA-photonique, Université libre de Bruxelles (U.L.B.), 50 Avenue F. D. Roosevelt, CP 194/5, B-1050 Bruxelles, Belgium, 2 Department of Physics,
The University of Auckland, Private Bag 92019, Auckland, New Zealand. * e-mail: francois.leo@ulb.ac.be

NATURE PHOTONICS | VOL 4 | JULY 2010 | www.nature.com/naturephotonics 471

© 2010 Macmillan Publishers Limited. All rights reserved.


ARTICLES NATURE PHOTONICS DOI: 10.1038/NPHOTON.2010.120

Addressing EDFA 1,535 nm, 4-ps pulses


beam
AOM Pulsed laser

Driving BPF EDFA 1,551 nm


WDM beam
c.w. laser

Polarization Photo- 5 Gsa s−1


controller Fibre coupler Output port Fibre coupler diode oscilloscope
BPF
90/10 95/5

Fibre Fibre coupler


90 m WDM
cavity 50/50
290 m
Optical
Piezo-electric
isolator WDM
fibre stretcher Photodiode
Optical spectrum
Controller analyser

Figure 1 | Experimental set-up. A passive fibre cavity is externally driven by a c.w. laser beam. CSs are excited with 4-ps addressing pulses at a different
wavelength. Only a single addressing pulse, selected by an acousto-optic modulator (AOM), is needed to write a CS. The driving and addressing beams are
combined through a wavelength division multiplexer (WDM) after amplification with erbium-doped fibre amplifiers (EDFA). Part of the output beam is used
to actively stabilize the cavity length, while the rest is directed towards an oscilloscope and an optical spectrum analyser. Bandpass filters (BPF) improve the
signal-to-noise ratio of the measurements.

solutions in a suitable range of driving powers. This allows for the Once the driving beam is turned on, we choose the servo-con-
existence of CSs in the Kerr cavity. Although CSs can arise in con- troller setpoint corresponding to the lower level within the bistable
ditions more general than described here35, this is the regime in regime and then launch an addressing pulse into the cavity. The
which we ran our experiment. addressing pulse comes from a mode-locked fibre laser that emits
Our experimental set-up is depicted in Fig. 1. The passive cavity 4-ps pulses with a repetition rate of 10 MHz at 1,535 nm. After
is entirely made up of about 380 m of standard single-mode silica amplification, it is combined with the driving beam through a
optical fibre laid in a ring configuration and closed on itself by a WDM coupler placed before the cavity input port. Inside the
90/10 fibre coupler. The round-trip time is 1.85 ms. The coupler cavity, an addressing pulse has a peak power of about 20 W and
is arranged so that 90% of the intracavity power is recirculated, interacts with the driving beam through cross-phase modulation2.
which leads to a large effective nonlinearity. The cavity incorporates In essence, the part of the c.w. intracavity field travelling with the
an optical fibre isolator to prevent the build-up of stimulated addressing pulse accumulates an additional phase shift and is
Brillouin scattering radiation, which would otherwise deplete the forced to locally switch to the upper bistability branch. This is suffi-
driving beam2,32, a piezo-electric fibre stretcher for active stabiliz- cient to subsequently reshape the local field into a CS. We must
ation of the cavity length30,36, and a wavelength-division multiplex- emphasize that this writing process is very robust: it is relatively
ing (WDM) coupler to extract the addressing pulses, as will be insensitive to the exact shape of the addressing pulse, it works for
explained below. Overall, the internal cavity losses are only 18%. any combination of wavelengths, it is phase insensitive37, and a
Combined with the 10% of power exiting the cavity through the CS can be written with a single addressing pulse, all of which is
coupler, the total round-trip losses are 26%, corresponding to a very beneficial for practical applications. However, as the addressing
relatively high finesse of 24 and to 22-kHz-wide resonances. In pulse travels at a different velocity to the driving beam, it has the
such a cavity, numerical simulations predict 4-ps-long CSs (see potential to excite several CSs in a row. The WDM coupler incor-
Supplementary Information). porated into the cavity prevents this by removing the addressing
To drive the cavity coherently, that is, to ensure that the driving pulses before they can excite a second CS.
beam adds up coherently over successive round trips, the spectral An acousto-optic modulator (AOM) driven by a pattern genera-
width of the c.w. driving beam must be narrower than that of the tor is placed along the path of the addressing beam. It allows us
cavity resonances. For that reason, it is generated by a distributed- either to select a single pulse or to encode a binary pattern onto
feedback fibre laser exhibiting an ultranarrow 1-kHz linewidth. This the addressing beam. Figure 2a first illustrates the steady-state
laser operates in the C-band at a wavelength of 1,551 nm, and its cavity output after a single addressing pulse has been launched
power is boosted up to 225 mW by an erbium-doped fibre amplifier into the cavity. The sequence shown has been taken about 1 s
(EDFA) before the driving beam is launched into the cavity. after the addressing pulse has left the cavity, but essentially identical
The intracavity field is measured through the fourth port of the traces are observed at all other times. As can be seen, in this con-
90/10 fibre coupler. The output beam is divided into three parts. A dition, the cavity output is constituted of a sequence of nearly iden-
first part is sent towards a digital oscilloscope to observe the CSs tical pulses with a period equal to the 1.85-ms cavity round-trip time.
directly in the time domain. Here, to measure the long-term evol- This observation is consistent with the presence inside the cavity of
ution of the CSs and to test their stability, we used an oscilloscope a single pulse that circulates repeatedly. We must emphasize that
capable of acquiring 10 million data points at a high sampling despite total round-trip losses of 26%, the pulse that we observe per-
rate (5 Gsample s21) in a single shot, corresponding to observation sists in the cavity without decaying, only sustained by the c.w.
times of at least 2 ms or, equivalently, 1,000 round trips. In parallel, driving beam. The intracavity pulse therefore realizes the balance
a second part of the beam is sent to an optical spectrum analyser between losses and coherent driving, which is a first signature of a
while the rest is directed to the servo controller that drives the CS. To further analyse the nature of the intracavity pulse, we have
piezo-electric fibre stretcher and that actively stabilizes the measured its temporal duration with an intensity autocorrelator.
cavity length. The recorded trace is plotted in Fig. 2b (blue). It is compared

472 NATURE PHOTONICS | VOL 4 | JULY 2010 | www.nature.com/naturephotonics

© 2010 Macmillan Publishers Limited. All rights reserved.


NATURE PHOTONICS DOI: 10.1038/NPHOTON.2010.120 ARTICLES
a 1.85 µs
Output power (a.u.)

Output power (a.u.)


U L B
Round-trip time 1 0 1 0 1 01 1 00 00010

Time (10 µs div−1)


Time (5 µs div−1)
b 1.5
2

power (W)
Intracavity Figure 3 | The acronym of our institution stored all-optically with CSs. ULB
4 ps (Université Libre de Bruxelles) is encoded as a 15-bit data stream in our
Autocorrelation (a.u.)

passive fibre cavity. Each letter is represented with five bits by its ordinal
1
position in the alphabet (U ¼ 21, L ¼ 12, B ¼ 2).
0
−50 0 50
Time (ps) thousand kilometres of fibre. This clearly illustrates its stability
6 ps
and robustness. For comparison, in our fibre, chromatic dispersion
0.5 should double the temporal duration of 4-ps pulses in only about
230 m (ref. 2). Nonlinearity must therefore effectively balance chro-
matic dispersion for the intracavity pulse to propagate undistorted
for so long, a fact that highlights the solitonic character of the intra-
0 cavity pulse. All together, these observations prove that we have
−20 −10 0 10 20 successfully excited a temporal CS inside our cavity.
Time (ps)
An ideal temporal CS should keep its amplitude rigorously con-
c stant over successive round trips. In contrast, the measurement
0 reported in Fig. 2a clearly reveals the presence of pulse-to-pulse fluc-
tuations. We attribute these to several factors. First, as seen in the
bottom of Fig. 2a, the background noise of the photodiode placed in
Power density (dB)

−10 front of the oscilloscope could well account for the totality of the
difference between adjacent pulses. This noise is mainly due to ampli-
fied spontaneous emission (ASE) in the driving laser amplifier, and we
−20
have checked that a larger level of ASE correlates with much stronger
observed pulse-to-pulse variability. The relatively poor signal-to-noise
−30 ratio is largely due to the limited bandwidth of our photodiode and
oscilloscope, as explained in the Methods. With a larger bandwidth
detector, the detected CSs would appear with a larger amplitude on
−40 the same background, so would display comparatively less variability.
−0.2 −0.1 0 0.1 0.2 Second, relaxation oscillations of the driving laser lead to fluctuations
Frequency (THz)
of the driving power and perturb the CSs. We have actually checked
Figure 2 | Observation of an isolated CS. a, Oscilloscope recording (linear that the long-term fluctuations of the CS we observe in Fig. 2a
scale). The delay between subsequent pulses matches the 1.85-ms cavity largely correlate with the fluctuations of the driving power. Longer-
round-trip time and confirms the presence inside the cavity of a single term fluctuations not visible in Fig. 2a are also due to limitations
persistent pulse circulating repeatedly, only sustained by the c.w. driving of the cavity-length stabilization system. The CSs are nevertheless
beam. b, Experimental (blue) intensity autocorrelation trace of the intracavity sufficiently robust to withstand all these perturbations.
pulse. It is compatible with a 4-ps-long CS, as revealed by its comparison To further assess that what we are observing are genuine temporal
with a numerically simulated trace (red) that corresponds to the temporal CSs, we have verified that they are independently addressable and that
intensity profile of the CS shown in the inset. c, Experimental (blue) and several CSs can coexist in the cavity. To this end, we first used our
simulated (red) optical spectrum. The simulated spectrum corresponds to AOM to encode a binary-coded data stream onto the addressing
the CS shown in the inset in b. beam and in this way store some data in the cavity in the form of a
sequence of CSs. Note that the input data stream is only sent once
with the result of a numerical simulation obtained for the same par- to the cavity input port. Figure 3 shows the cavity output measured
ameters as the experiment (red) and that corresponds to the tem- with the oscilloscope after having written a 15-bit sequence encoding
poral intensity profile of the temporal CS shown in the inset. the acronym of our institution, ULB. Each letter is represented by its
Clearly, the agreement is remarkable. The dips in the c.w. back- ordinal position in the alphabet (U ¼ 21, L ¼ 12, B ¼ 2) coded on
ground on both sides of the main peak—which are a well-known five bits. The presence of a CS encodes a binary 1, and its absence
feature of CSs38–40—are particularly well reproduced. This confirms in the corresponding time slot encodes a 0. As can be seen, the
that the recorded autocorrelation trace is compatible with a 4-ps- stored data stream is repeated at the output without degradation,
long temporal CS. In Fig. 2c, we observe a similarly excellent agree- over and over again. As with a single CS, we have demonstrated a
ment between the experimental and simulated spectra. The wings of storage time exceeding one second, only limited by environmental
the spectrum are the signature of the presence of a short pulse inside perturbations that eventually bring the piezo-electric stretcher to
the cavity, and the central peak corresponds to the c.w. background the end of its course. Clearly, this truly demonstrates that our fibre
on which it is superimposed. Finally, we must stress that during the cavity operates as an all-optical memory.
second that separates the addressing pulse from the observations The low 10-MHz repetition rate of our pulsed addressing laser
reported in Fig. 2a, the intracavity pulse has propagated along prevented us from writing more than 18 bits per cavity round
about 540,000 cavity round-trips or, equivalently, several hundred trip. To investigate how CSs interact and the maximum storage

NATURE PHOTONICS | VOL 4 | JULY 2010 | www.nature.com/naturephotonics 473

© 2010 Macmillan Publishers Limited. All rights reserved.


ARTICLES NATURE PHOTONICS DOI: 10.1038/NPHOTON.2010.120

a b
0 0

−10 −10

−20 −20
CSs: 39 ps delay
Power density (dB)

Power density (dB)


−30 −30

−40 −40

−50 −50

−60 −60

−70 −70

−80 −80

−90 Addressing pulses: 24.5 ps delay −90 Addressing pulses: 30 ps delay

−150 −100 −50 0 50 100 150 −150 −100 −50 0 50 100 150
Frequency (GHz) Frequency (GHz)

c d
0 0

−10 −10

−20 −20
CSs: 41.5 ps delay CSs: 63.5 ps delay
Power density (dB)

−30 Power density (dB) −30

−40 −40

−50 −50

−60 −60

−70 −70

−80 −80

−90 Addressing pulses: 40 ps delay −90 Addressing pulses: 61 ps delay

−150 −100 −50 0 50 100 150 −150 −100 −50 0 50 100 150
Frequency (GHz) Frequency (GHz)

Figure 4 | Experimental observation of the interactions between close CSs. The green curves are the spectra of the addressing pulses (shifted vertically for
clarity) and the blue curves are the cavity output spectra 1 s later. Periodic spectral fringes marked with sets of vertical dotted lines indicate the presence of
two close pulses and are used to measure the temporal separation between those pulses as displayed on each graph. a, With an initial separation of 25 ps,
no fringes are ever observed on the cavity output spectrum. Only a single CS is excited in this condition. b–d, For larger separations between the addressing
pulses, the CSs repel each other.

density of our optical memory, we added an unbalanced interferom- sensitive function of the initial conditions. Either only a single CS
eter with a variable delay line on the path of the addressing beam. In is ever generated, or the two CSs initially excited interact attractively
this way, we created pairs of addressing pulses the temporal separ- and quickly merge into a single CS. In contrast, with addressing
ation of which could be continuously varied between 20 and pulses separated by 30 ps (Fig. 4b), the two CSs strongly repel:
60 ps. We then checked what happens to the two CSs that are sub- their separation increases to 39 ps, or by about 30%, as they freely
sequently excited in the cavity and whether they attract or repulse propagate for about 1 s in the cavity. Larger separations of the
each other. Because of the limited bandwidth of the oscilloscope, addressing pulses lead to a much weaker repulsion, as observed in
direct time-domain observations were not possible. Instead, we Fig. 4c,d. The relatively weaker repulsion observed in Fig. 4c in com-
relied on the presence of spectral interference fringes on the parison with Fig. 4d may signal the existence of a bound state of CSs.
cavity output spectrum to detect the presence of stable pairs of So far, however, we have not been able to conclusively identify such
CSs and to measure their temporal separation. Note that those a bound state in our experiment, probably due to fluctuations of the
measurements were performed with multiple CS pairs circulating driving power. All our observations are nevertheless compatible
simultaneously in the cavity (see Methods). with earlier studies, in particular those performed with spatial
Figure 4a–d shows the results for four different temporal separ- CSs38–40, and are typical of interacting localized DSs41,42. Note that
ations of the two addressing pulses. In each case, we measured the an all-optical buffer may be affected by the repulsive interactions
spectrum of the addressing pulses before launching them into the between CSs. However, above a separation of 40 ps, the repulsion
cavity (green), then, about 1 s later, the cavity output spectrum clearly becomes very weak, particularly considering that our obser-
(blue). Fringes on the spectra (if any) indicate the presence of two vations are performed after the CSs have propagated and interacted
close pulses for which the temporal separation is deduced by over several hundred thousand cavity round trips. Such weak repul-
Fourier analysis and displayed on the graph (see Methods for sion is actually very challenging to model. We believe that a shallow
more details on the finer fringe structure). With addressing pulses modulation superimposed on the driving beam will easily trap the
separated by 25 ps or less (Fig. 4a), the output spectrum never exhi- CSs into specific time slots while at the same time providing a retim-
bits any fringes. Simulations reveal that what happens here is a ing functionality8. It would allow our fibre cavity to store about

474 NATURE PHOTONICS | VOL 4 | JULY 2010 | www.nature.com/naturephotonics

© 2010 Macmillan Publishers Limited. All rights reserved.


NATURE PHOTONICS DOI: 10.1038/NPHOTON.2010.120 ARTICLES
45,000 bits at 25 Gbit s21. This places our cavity memory right in driving and addressing beams as well as to extract the writing beam from the cavity
the operating range of realistic telecommunications systems43,44. and from the path to the active stabilization system are three-port coarse WDM
couplers. They separate/mix two channels centred at 1,530 and 1,550 nm with
Storage of data streams with higher bit rates would be made possible 13-nm-wide passing bands at 0.5 dB. The cavity length stabilization system is based
by using a fibre with different characteristics (such as a lower chro- on a programmable microcontroller driven by a 3-kHz bandwidth photodiode. The
TM
matic dispersion coefficient). driving beam is generated by a Koheras AdjustiK laser, which is an erbium-doped
distributed feedback fibre laser. It presents a linewidth of ,1 kHz over 120 ms. After
Discussion amplification, a narrow tunable bandpass filter (1 nm bandwidth at 0.5 dB) centred
on the driving laser wavelength is used to reject most of the ASE noise of the EDFA
Our study provides clear experimental evidences of Kerr temporal before the driving beam is launched into the cavity. This filter helps to improve the
CSs in a c.w. coherently driven all-fibre passive optical cavity. signal-to-noise ratio when observing the CSs on the oscilloscope. The oscilloscope
Both an isolated CS as well as pairs of CSs and their interactions used had a large 5 Gsample s21 sampling rate but was limited to a bandwidth of
have been observed. The solitonic nature of the 4-ps pulses that cir- 1 GHz. The photodiode used in front of the oscilloscope also had a 1-GHz
culate in our cavity was established by observing that they realize the bandwidth. As a result, the oscilloscope response to the short 4-ps-long CSs was very
broad (1 ns) and severely flattened. As the CSs are superimposed on a c.w.
double balance of chromatic dispersion and losses with, respectively, background (the residual c.w. driving beam), the detection contrast of these pulses
nonlinearity and external driving. In our experiment, we have on the oscilloscope was very poor. For that reason, we used a sequence of two
observed these pulses propagating for several seconds, correspond- identical narrow bandpass filters in front of the oscilloscope. These filters are dense
ing to several hundred thousand kilometres, limited only by our WDM fibre couplers centred at 1,552 nm, slightly above the wavelength of the c.w.
cavity-length stabilization system. Also, we have shown that we driving beam, with a spectral width of 0.9 nm at 0.5 dB. This effectively completely
removes the c.w. background of the CSs and greatly improves the signal-to-
can store a binary-coded data stream in our cavity, realizing an noise ratio.
all-optical memory. In our configuration, the cavity could poten-
tially store up to 45,000 bits at 25 Gbit s21. Autocorrelation and spectrum of a single CS (Fig. 2b,c). When a single CS is
In addition to implementing an all-optical memory, our passive present in the cavity as in Fig. 2a, the energy contained in the 4-ps-long CS is about
20,000 times weaker than the energy contained in the c.w. background filling the
cavity system seamlessly provides the function of an all-optical cavity. This makes an autocorrelation or a spectral measurement of the intracavity
reshaper and wavelength converter. Reshaping is obtained thanks pulse very difficult to perform. Therefore, for the results presented in Fig. 2b and c,
to the robust attractive nature of the CS state, whereas wavelength we actually improved the pulse-to-background energy ratio by filling the cavity with
conversion naturally follows from the use of a phase-insensitive several thousands CSs before taking the measurements. This was obtained by
keeping the AOM selecting the addressing pulses open for about 100 ms (therefore
and wavelength-insensitive all-optical interaction to excite the
letting 1,000 addressing pulses pass through) every 10 ms. The 10-ms re-writing
CSs. Indeed, the CSs are generated with the same wavelength as time interval was chosen to match the scanning period of the autocorrelator. In these
the driving beam, which is different from that of the addressing conditions, even if the CSs disappear because the piezo-electric fibre stretcher
beam. This is potentially very useful for applications in all-optical reaches the end of its course, the system is reset to its original state for the next scan
routers. Simultaneous re-timing of the incoming data stream of the measuring apparatus. The autocorrelation was averaged over 512 scans. Note
that the number of solitons filling the cavity in those conditions was the only fit
would also be obtained by superimposing a shallow modulation parameter used to match the experimental and simulated results superimposed in
on the c.w. driving beam1,8. Finally, we must stress that our Fig. 2b and c. Additionally, we should point out that, for these measurements, the
system is not critically sensitive to parameters such as the wave- optical signal was not obtained through the fourth port of the 90/10 cavity input
length of the lasers, the fibre characteristics or the fibre length. coupler. Instead, we used the output of a 1% tap coupler that was incorporated into
Our cavity has actually been built with standard off-the-shelf com- the cavity in place of the internal WDM coupler. In this way, we obtained a signal
that was directly proportional to the intracavity intensity and for which the
ponents and fibres, which highlights the strong potential of our background was not modified by interference with the reflected driving beam. This
all-optical memory scheme for practical applications. change in the set-up led to slightly increased total cavity round-trip losses of 27%
For applications as an all-optical memory, the writing time is of (instead of the original 26%). Finally, the signal-to-noise ratio of these
course very important. Although we have not discussed it here measurements was further improved in two other ways. First, the driving power was
specifically, we have observed that the CSs become established in increased from 154 mW in Fig. 2a to 224 mW in Fig. 2b,c. Numerical simulations
indicate that this only weakly changes the temporal width of the CS (from 4.4 to
about 100 cavity round trips after the addressing pulses are launched 4.1 ps). Second, an EDFA was used in front of the autocorrelator to boost the average
(see Supplementary Information). The writing time can be reduced signal power to 100 mW.
by using a shorter and/or lower finesse cavity, to the detriment of
the requirement of a higher driving power. The driving power Measurements of the spectra of pairs of CSs (Fig. 4). Again, to avoid an over-
whelming c.w. background on the spectral measurements, we filled the cavity with a
could then be lowered by using either highly nonlinear fibres, or large number of pairs of CSs before measuring the cavity output spectrum. Here,
highly nonlinear serpentine waveguides45 or microresonators10 inte- however, to correctly study the interactions between close CSs, it is important to
grated on a chip. Chip integration would also provide a simple way make sure that the pairs of CSs are not written too close to one another. As the pairs
to make the cavity less sensitive to environmental perturbations. are written and interleaved over several round trips, and because the repetition time
Addressing pulses about 50% more powerful than what is strictly of the addressing laser is not a simple fraction of the cavity round-trip time, this is
not completely trivial to assess. To this end, we refined our measurements of the two
necessary to excite a CS in the cavity can also be used to selectively timescales involved. On the one hand, the repetition time of the addressing laser was
erase an existing CS in a phase-insensitive manner (see measured by observing the harmonics of the repetition rate of the laser with a 1-GHz
Supplementary Information). This process in fact implements an photodiode connected to a radio-frequency spectrum analyser and was found to be
all-optical XOR-logic gate between an incoming data stream and T ¼ 100.1 ns. On the other hand, the cavity round-trip time was deduced from a
binary data stored in the cavity as CSs. long oscilloscope recording of a single CS propagating in the cavity as in Fig. 2a and
was found to be tR ¼ 1,854.3 ns. With those figures, we found that we could keep the
To conclude, we have reported the first experimental observation AOM open for 250 ms at a time, or write about 2,500 pairs of CSs over 135 cavity
of both temporal and Kerr CSs. Given the simplicity and the para- round trips, while guaranteeing that no two adjacent pairs were closer than 0.7 ns,
digmatic character of the Kerr cavity model, combined here with the which is quite comfortable. The spectral measurements presented in Fig. 4 were
unidimensionality inherent to fibre optics, our experiment can be therefore obtained with the AOM open for 250 ms every 1 s, and the spectra were
acquired some time during the 1-s interval between writing sequences. The driving
considered as the most fundamental example of self-organization
power was set at 200 mW (this corresponds to the same normalized parameters as
phenomena in nonlinear optics. for Fig. 2b,c once the difference in internal cavity round-trip losses is factored in; see
Supplementary Information). Additionally, we should point out that, instead of a
Methods standard OSA with a resolution of 0.1 nm (or 10 GHz) as used for Fig. 2c, we had
Experimental set-up. The cavity is composed of standard telecommunications access here to an interferometric OSA with an ultrahigh resolution of 100 MHz.
single-mode silica optical fibre. At 1,550 nm, this fibre presents a second-order This explains that even with the largest 60 ps delay (Fig. 4d) we have studied, the
dispersion coefficient b2 ¼ 220 ps2 km21 and a nonlinearity coefficient of spectral fringes are nicely resolved. The resolving power of the OSA is actually good
g ≈ 1.8 W21 km21. The intracavity isolator preventing stimulated Brillouin enough to resolve the spectral fringes resulting from the temporal separation
scattering has a 60-dB isolation factor. The three WDM couplers used to mix the between the adjacent pairs of CSs filling the cavity. This is what causes the fine fringe

NATURE PHOTONICS | VOL 4 | JULY 2010 | www.nature.com/naturephotonics 475

© 2010 Macmillan Publishers Limited. All rights reserved.


ARTICLES NATURE PHOTONICS DOI: 10.1038/NPHOTON.2010.120

structure visible on the blue curves in Fig. 4b–d. These fringes confirm that adjacent 29. Fraile-Peláez, F. J., Capmany, J. & Muriel, M. A. Transmission bistability in a
pairs are not closer than about 0.7 ns. In our operating conditions, the sweep time of double-coupler fiber ring resonator. Opt. Lett. 16, 907–909 (1991).
the high-resolution OSA is, however, relatively large, at 0.4 s. This leads to some 30. Coen, S. et al. Experimental investigation of the dynamics of a stabilized
uncertainty and averaging in the delays we measure as the two CSs within each pair nonlinear fiber ring resonator. J. Opt. Soc. Am. B 15, 2283–2293 (1998).
keep moving with respect to one another while their spectrum is being measured. 31. Haelterman, M., Trillo, S. & Wabnitz, S. Dissipative modulation instability in a
It also prevents us from observing how the temporal separation between two CSs nonlinear dispersive ring cavity. Opt. Commun. 91, 401–407 (1992).
evolves dynamically with time. 32. Coen, S. & Haelterman, M. Continuous-wave ultrahigh-repetition-rate pulse-
train generation through modulational instability in a passive fiber cavity. Opt.
Received 20 January 2010; accepted 6 April 2010; Lett. 26, 39–41 (2001).
published online 23 May 2010 33. Turing, A. M. The chemical basis of morphogenesis. Phil. Trans. R. Soc. B 237,
37–72 (1952). Reprinted in Bull. Math. Biology 52, 153–197 (1990).
References 34. Coen, S. & Haelterman, M. Competition between modulational instability and
1. Wabnitz, S. Suppression of interactions in a phase-locked soliton optical switching in optical bistability. Opt. Lett. 24, 80–82 (1999).
memory. Opt. Lett. 18, 601–603 (1993). 35. Rozanov, N. N. Dissipative optical solitons in the absence of bistability and
2. Agrawal, G. P. Nonlinear Fiber Optics, Optics and Photonics Series 4th edn modulation instability. Optics & Spectroscopy 96, 569–574 (2004).
(Academic Press, San Diego, 2006). 36. Jackson, D. A., Priest, R., Dandridge, A. & Tveten, A. B. Elimination of drift in a
3. Firth, W. J. & Weiss, C. O. Cavity and feedback solitons. Opt. Phot. News 13(2), single-mode optical fiber interferometer using a piezoelectrically stretched coiled
54–58 (2002). fiber. Appl. Opt. 19, 2926–2929 (1980).
4. Lugiato, L. A. Introduction to the feature section on cavity solitons: an overview. 37. Barbay, S. et al. Incoherent and coherent writing and erasure of cavity solitons in
IEEE J. Quantum Electron. 39, 193–196 (2003). an optically pumped semiconductor amplifier. Opt. Lett. 31, 1504–1506 (2006).
5. Ackemann, T. & Firth, W. J. Dissipative solitons in pattern-forming nonlinear 38. Schäpers, B., Feldmann, M., Ackemann, T. & Lange, W. Interaction of localized
optical systems: cavity solitons and feedback solitons, in Dissipative Solitons structures in an optical pattern-forming system. Phys. Rev. Lett. 85,
Vol. 661, Lecture Notes in Physics 55–100 (Springer, 2005). 748–751 (2000).
6. Akhmediev, N. N. & Ankiewicz, A. (eds). Dissipative solitons: from optics to 39. Ramazza, P. L. et al. Tailoring the profile and interactions of optical localized
biology and medicine. in Lecture Notes in Physics Vol. 751 (Springer, 2008). structures. Phys. Rev. E 65, 066204 (2002).
7. Brambilla, M., Lugiato, L. A. & Stefani, M. Interaction and control of optical 40. Tlidi, M., Vladimirov, A. G. & Mandel, P. Interaction and stability of periodic
localized structures. Europhys. Lett. 34, 109–114 (1996). and localized structures in optical bistable systems. IEEE J. Quantum Electron.
8. McDonald, G. S. & Firth, W. J. Spatial solitary-wave optical memory. J. Opt. Soc. 39, 216–226 (2003).
Am. B 7, 1328–1335 (1990). 41. Aranson, I. S., Gorshkov, K. A., Lomov, A. S. & Rabinovich, M. I. Stable particle-
9. Mitschke, F. & Schwache, A. Soliton ensembles in a nonlinear resonator. J. Opt. like solutions of multidimensional nonlinear fields. Physica D 43,
B: Quantum Semiclass. Opt. 10, 779–788 (1998). 435–453 (1990).
10. Del’Haye, P. et al. Optical frequency comb generation from a monolithic 42. Bödeker, H. U., Liehr, A. W., Frank, T. D., Friedrich, R. & Purwins, H.-G.
microresonator. Nature 450, 1214–1217 (2007). Measuring the interaction law of dissipative solitons. New J. Phys. 6, 62 (2004).
11. McLaughlin, D. W., Moloney, J. V. & Newell, A. C. Solitary waves as fixed points 43. Moores, J. D. et al. 20-GHz optical storage loop/laser using amplitude
of infinite-dimensional maps in an optical bistable ring cavity. Phys. Rev. Lett. modulation, filtering, and artificial fast saturable absorption. IEEE Photon.
51, 75–78 (1983). Technol. Lett. 7, 1096–1098 (1995).
12. Rosanov, N. N. & Khodova, G. V. Diffractive autosolitons in nonlinear 44. Boyd, R. W., Gauthier, D. J. & Gaeta, A. L. Applications of slow light in
interferometers. J. Opt. Soc. Am. B 7, 1057–1065 (1990). telecommunications. Opt. Phot. News 17(4), 18–23 (2006).
13. Tlidi, M., Mandel, P. & Lefever, R. Localized structures and localized patterns in 45. Madden, S. J. et al. Long, low loss etched As2S3 chalcogenide waveguides for all-
optical bistability. Phys. Rev. Lett. 73, 640–643 (1994). optical signal regeneration. Opt. Express 15, 14414–14421 (2007).
14. Firth, W. J. & Scroggie, A. J. Optical bullet holes: Robust controllable localized
states of a nonlinear cavity. Phys. Rev. Lett. 76, 1623–1626 (1996). Acknowledgements
15. Tanguy, Y., Ackemann, T., Firth, W. J. & Jäger, R. Realization of a The authors are grateful to A. Mussot and M. Taki from Laboratoire de Physique des Lasers,
semiconductor-based cavity soliton laser. Phys. Rev. Lett. 100, 013907 (2008). Atomes et Molécules (PhLAM), Université de Lille 1 (Lille, France) and to T. Sylvestre and
16. Bakonyi, Z., Michaelis, D., Peschel, U., Onishchukov, G. & Lederer, F. Dissipative J.-M. Merolla, Université de Franche-Comté (Besançon, France) for lending us some
solitons and their critical slowing down near a supercritical bifurcation. J. Opt. experimental parts, as well as to M. Tlidi, Université Libre de Bruxelles (Brussels, Belgium),
Soc. Am. B 19, 487–491 (2002). for fruitful discussions. This work was supported by the Belgian Science Policy Office
17. Brambilla, M., Maggipinto, T., Patera, G. & Columbo, L. Cavity light bullets: (BELSPO) Interuniversity Attraction Pole (IAP) programme under grant no. IAP-6/10.
three-dimensional localized structures in a nonlinear optical resonator. Phys. F.L. acknowledges the support of the Fonds pour la formation à la Recherche dans
Rev. Lett. 93, 203901 (2004). l’Industrie et dans l’Agriculture (FRIA) (Belgium). The participation of S.C. to this project
18. Jenkins, S. D., Prati, F., Lugiato, L. A., Columbo, L. & Brambilla, M. Cavity light was made possible thanks to a Research & Study Leave granted by The University of
bullets in a dispersive Kerr medium. Phys. Rev. A 80, 033832 (2009). Auckland and to a visiting fellowship from the Fonds National de la Recherche Scientifique
19. Barland, S. et al. Cavity solitons as pixels in semiconductor microcavities. Nature (FNRS) (Belgium). S.C. and The University of Auckland also provided the high-sampling
419, 699–702 (2002). rate oscilloscope necessary for this project. The work of S.C. is supported by a New
20. Pedaci, F. et al. All-optical delay line using semiconductor cavity solitons. Appl. Economy Research Fund (NERF) grant from The Foundation for Research, Science and
Phys. Lett. 92, 011101 (2008). Technology of the New Zealand government.
21. Lugiato, L. A. & Lefever, R. Spatial dissipative structures in passive optical
systems. Phys. Rev. Lett. 58, 2209–2211 (1987).
22. Haelterman, M., Trillo, S. & Wabnitz, S. Additive-modulation-instability ring Author contributions
F.L. performed the experiments, starting with the set-up built by S.C. for other studies, and
laser in the normal dispersion regime of a fiber. Opt. Lett. 17, 745–747 (1992).
analysed the results. S.C. helped analyse the results, supervised the experiments, and wrote
23. Murray, J. D. Mathematical biology. in Biomathematics Texts Vol. 19, 2nd edn
the paper. Overall, F.L. and S.C. contributed equally to this work. P.K. provided day-to-day
(Springer-Verlag, 1993).
support both in the laboratory and on theoretical aspects. S.-P.G. helped with the
24. Glansdorff, P. & Prigogine, I. Thermodynamic Theory of Structure, Stability and
autocorrelation measurement and in the analysis of the spectral fringes of pairs of CSs.
Fluctuations (Wiley, 1971).
Ph.E. is the group leader and obtained funding for this work. M.H. supervised the
25. Nicolis, G. & Prigogine, I. Self-Organization in Nonequilibrium Systems: From
overall project.
Dissipative Structures to Order through Fluctuations (Wiley, New York, 1977).
26. Scroggie, A. J. et al. Pattern formation in a passive Kerr cavity. Chaos, Solitons &
Fractals 4, 1323–1354 (1994). Additional information
27. Firth, W. J. & Lord, A. Two-dimensional solitons in a Kerr cavity. J. Mod. Opt. The authors declare no competing financial interests. Supplementary information
43, 1071–1077 (1996). accompanies this paper at www.nature.com/naturephotonics. Reprints and permission
28. Firth, W. J. et al. Dynamical properties of two-dimensional Kerr cavity solitons. information is available online at http://npg.nature.com/reprintsandpermissions/.
J. Opt. Soc. Am. B 19, 747–752 (2002). Correspondence and requests for materials should be addressed to F.L.

476 NATURE PHOTONICS | VOL 4 | JULY 2010 | www.nature.com/naturephotonics

© 2010 Macmillan Publishers Limited. All rights reserved.

You might also like